A Heme-based Redox Sensor in the Methanogenic Archaeon Methanosarcina acetivorans*

Background: Multidomain sensory kinases are involved in numerous receptive processes in all kingdoms of life. Results: The multidomain sensory kinase MA4561 covalently binds a redox-active heme cofactor, which triggers kinase activity. Conclusion: Covalently bound heme is utilized to detect redox changes. Significance: Learning how Archaea perceive environmental stimuli will enhance our understanding of the evolution of prokaryotic signal transduction. Based on a bioinformatics study, the protein MA4561 from the methanogenic archaeon Methanosarcina acetivorans was originally predicted to be a multidomain phytochrome-like photosensory kinase possibly binding open-chain tetrapyrroles. Although we were able to show that recombinantly produced and purified protein does not bind any known phytochrome chromophores, UV-visible spectroscopy revealed the presence of a heme tetrapyrrole cofactor. In contrast to many other known cytoplasmic heme-containing proteins, the heme was covalently attached via one vinyl side chain to cysteine 656 in the second GAF domain. This GAF domain by itself is sufficient for covalent attachment. Resonance Raman and magnetic circular dichroism data support a model of a six-coordinate heme species with additional features of a five-coordination structure. The heme cofactor is redox-active and able to coordinate various ligands like imidazole, dimethyl sulfide, and carbon monoxide depending on the redox state. Interestingly, the redox state of the heme cofactor has a substantial influence on autophosphorylation activity. Although reduced protein does not autophosphorylate, oxidized protein gives a strong autophosphorylation signal independent from bound external ligands. Based on its genomic localization, MA4561 is most likely a sensor kinase of a two-component system effecting regulation of the Mts system, a set of three homologous corrinoid/methyltransferase fusion protein isoforms involved in methyl sulfide metabolism. Consistent with this prediction, an M. acetivorans mutant devoid of MA4561 constitutively synthesized MtsF. On the basis of our results, we postulate a heme-based redox/dimethyl sulfide sensory function of MA4561 and propose to designate it MsmS (methyl sulfide methyltransferase-associated sensor).

Based on a bioinformatics study, the protein MA4561 from the methanogenic archaeon Methanosarcina acetivorans was originally predicted to be a multidomain phytochrome-like photosensory kinase possibly binding open-chain tetrapyrroles. Although we were able to show that recombinantly produced and purified protein does not bind any known phytochrome chromophores, UV-visible spectroscopy revealed the presence of a heme tetrapyrrole cofactor. In contrast to many other known cytoplasmic heme-containing proteins, the heme was covalently attached via one vinyl side chain to cysteine 656 in the second GAF domain. This GAF domain by itself is sufficient for covalent attachment. Resonance Raman and magnetic circular dichroism data support a model of a six-coordinate heme species with additional features of a five-coordination structure. The heme cofactor is redox-active and able to coordinate various ligands like imidazole, dimethyl sulfide, and carbon monoxide depending on the redox state. Interestingly, the redox state of the heme cofactor has a substantial influence on autophosphorylation activity. Although reduced protein does not autophosphorylate, oxidized protein gives a strong autophosphorylation signal independent from bound external ligands. Based on its genomic localization, MA4561 is most likely a sensor kinase of a two-component system effecting regulation of the Mts system, a set of three homologous corrinoid/methyltransferase fusion protein isoforms involved in methyl sulfide metabolism. Consistent with this prediction, an M. acetivorans mutant devoid of MA4561 constitutively synthesized MtsF. On the basis of our results, we postulate a heme-based redox/dimethyl sul-fide sensory function of MA4561 and propose to designate it MsmS (methyl sulfide methyltransferase-associated sensor).
Heme-based sensors are a diverse group of sensor proteins that are able to sense small gaseous ligands like oxygen (O 2 ), carbon monoxide (CO), and NO and changes of the cellular redox state. So far, several heme-binding sensor domains have been characterized (for a review, see Ref. 1). Among them, a variety of globin-coupled sensors with a globin fold as the heme-binding site have been identified (2). Other examples are the heme NO/oxygen-binding proteins (3), which have similarity to eukaryotic soluble guanylate cyclases (4); the SCHIC (sensor containing heme instead of cobalamin) domain in AppA (5); and the CO-sensing transcriptional activator CooA from Rhodospirillum rubrum (6). Additionally, heme-containing PAS 3 domains (7), like those in FixL (8), EcDos (9), and AxPDEA1 (10), and GAF domains, like those in DevS (DosS) and DosT (11,12), have been reported. In virtually all cases, the proteins contain non-covalently bound b-type heme. Only two examples of covalently bound heme in sensor proteins have been described in the literature. DcrA from Desulfovibrio vulgaris contains a c-type heme in a periplasmic cytochrome c-like fold (13)(14)(15), and the protein GSU0303 from Geobacter sulfurreducens was shown to contain a periplasmic heme-binding PAS domain. In contrast to the sensory input, which is determined by the chemical nature of heme, the physiological output is as variable as the heme-binding domains themselves. HemATs function as regulators of aerotaxis (2), whereas other globincoupled sensors regulate concentrations of second messengers (16). EcDos and AxPDEA1 are phosphodiesterases (10,17), and CooA is a direct transcriptional activator (6), whereas AppA is an antirepressor for a transcriptional regulator (18). FixL, DevS, and DosT are sensor histidine kinases of two-component systems (8,19). Recently, a globin-coupled histidine kinase also was identified (20). The only known heme-based sensor from Archaea is the aerotaxis transducer Hs-HemAT from Halobacterium salinarum (2).
Within the archaeal kingdom, one-component systems are more widely distributed than two-component systems (21), and indeed genes encoding two-component systems are only found in some euryarchaeal genomes and neither in the Crenarchaeota nor in the Nanoarchaeota (22). Therefore, two-component systems most likely evolved in bacteria and radiated into Archaea via horizontal gene transfer (21,23). Typical bacterial two-component systems consist of a sensor histidine kinase and a corresponding response regulator (24). The sensor kinases are often multidomain proteins with a conserved kinase output domain fused to a variety of sensor domains that detect certain signals. The response regulators have a conserved input domain and non-conserved output domains (25). Upon signal recognition by the sensor kinase, a conserved histidine residue is autophosphorylated, and the phosphoryl group subsequently is transferred to a conserved aspartate residue in the response regulator input domain, resulting in activation of the output domain.
Based on sequence analyses and phylogenetic studies, the protein MA4561 from the methanogenic archaeon Methanosarcina acetivorans was predicted to be a phytochrome-like protein (26). Another study searching for two-component systems in Archaea showed that MA4561 contains a histidine kinase-like ATPase (H_ATPase) domain but seems to lack the H-box containing the conserved histidine residue that functions as the phosphoacceptor during autophosphorylation (22).
In this study, we investigated the function of MA4561 as well as different truncated variants thereof in vitro and in vivo. We show that MA4561 harbors a covalently bound heme cofactor in its second GAF domain. To our knowledge, this is the first report of a covalently attached heme cofactor in a cytoplasmic sensor protein. Autophosphorylation of the kinase domain is shown to be strongly dependent on the heme redox state. As deletion of MA4561 led to deregulated synthesis of a dimethyl sulfide-dependent methyltransferase, we argue for a redox/dimethyl sulfide signaling function for MA4561.

EXPERIMENTAL PROCEDURES
Construction of Expression Plasmids-The M. acetivorans C2A ma4561 coding region and shorter variants were PCRamplified from genomic DNA and cloned according to standard procedures. Details can be found in the supplemental Experimental Procedures and supplemental Tables S1 and S2.
Overexpression and Purification of His 6 -tagged MA4561 Variants-Recombinant full-length MA4561 and all variants fused to a C-terminal His 6 tag were produced using a T7/lac promoter-driven expression system (Novagen) in Escherichia coli strain BL21(DE3). Cells were grown in LB medium containing 100 g/ml ampicillin at 37°C to an A 578 of 0.5. Protein production was induced by the addition of 0.1 mM isopropyl ␤-thiogalactoside (full-length MA4561-His 6 and variants GAF2-His 6 , PAS2-His 6 , GAF1-His 6 , and PAS1-His 6 ) or 0.25 mM isopropyl ␤-thiogalactoside (variants sGAF2-His 6 and sGAF2C656A-His 6 ), and cells were further incubated for 18 -20 h at 17°C. The cell pellet was either frozen and stored at Ϫ20°C or directly resuspended in elution and washing (EW) buffer (50 mM Tris/HCl, pH 7.0, 300 mM NaCl, 10% glycerol, 0.05% Tween 20, 1 mM dithiothreitol, 0.25 mM 4-(2-aminoethyl)benzenesulfonyl fluoride; in the case of variants sGAF2-His 6 and sGAF2C656A-His 6 , no glycerol and Tween 20 were added) (6 ml of buffer/g of cell weight) and disrupted by sonication (Branson sonifier 250) or by acceleration disruption (Constant Systems Ltd. cell disruption system) at 40 k.p.s.i. (variants sGAF2-His 6 and sGAF2C656A-His 6 ). Cell debris was removed by centrifugation (90 min at 28,000 ϫ g at 4°C), and the supernatant was either directly loaded onto an Co 2ϩ -based metal affinity chromatography column (BD Talon), which had been preequilibrated with EW buffer, or incubated with 100 M hemin (in DMSO) for 1 h at 4°C prior to this step. Unwanted proteins were washed off with 10 column volumes of EW buffer containing 10 mM imidazole (Im). Elution of His 6 -tagged proteins was performed with EW buffer containing 150 mM Im. Alternatively, the column was washed extensively with a minimum of 20 column volumes of EW buffer (without Im) after loading the protein (incubated with hemin), and all residual proteins were washed of the column with a total of 3 column volumes of EW buffer containing 200 mM EDTA. Fractions containing the desired protein were combined and dialyzed against an appropriate volume of dialysis buffer (50 mM Tris/ HCl, pH 7.0, 100 mM NaCl; 50 mM sodium phosphate, pH 7.0, 100 mM NaCl; or 50 mM CAPS, pH 10.0, 100 mM NaCl) at 4°C overnight. Proteins were concentrated if necessary using Vivaspin concentrator devices (molecular weight cutoff, 10,000; Sartorius Stedim Biotech). If higher purity was desired, the protein was subjected to gel permeation chromatography on a Superdex 200 10/300GL column (GE Healthcare). Fractions containing the desired protein were combined and again concentrated to an appropriate concentration (for calculated molecular masses and extinction coefficients, see supplemental Table S3).
Overexpression and Purification of StrepII-tagged MA4561 Variants-Recombinant full-length MA4561 and variant sGAF2 fused to a C-terminal StrepII tag were produced using a tet promoter-driven expression system (IBA) in E. coli strain BL21(DE3). Cell growth and purification were performed as described for the His 6 -tagged protein with the exception that protein production was induced with 200 ng/ml anhydrotetracycline, and the buffer for the purification procedure was Buffer W (100 mM Tris/HCl, pH 8.0, 300 mM NaCl, 1 mM EDTA, 10% glycerol, 0.05% Tween 20, 0.25 mM 4-(2-aminoethyl)benzenesulfonyl fluoride; in the case of variant sGAF2, no glycerol and Tween 20 were added). For the affinity chromatography, a Strep-Tactin chromatography column (IBA) preequilibrated with Buffer W was used. Unwanted proteins were washed off with 10 column volumes of Buffer W. Elution of StrepII-tagged proteins was performed with Buffer E (Buffer W containing 2.5 mM desthiobiotin). Fractions containing the desired protein were combined, dialyzed, and concentrated as described for His 6 -tagged proteins (for calculated molecular masses and extinction coefficients, see supplemental Table S3).
Overexpression and Purification of MA4561 Full-length Protein from M. acetivorans-MA4561-His 6 was produced in M. acetivorans strain WWM73 (27) after liposome-mediated transformation (28) with ptet4561. Cells were grown under strictly anaerobic conditions at 37°C to an A 578 of 1.5 in high salt medium (29) containing 125 mM methanol as the sole energy source. 2 g/ml puromycin was added for selection of the pac gene encoding puromycin acetyltransferase (30), and 100 g/ml tetracycline was added for induction of gene expression controlled by the hybrid mcrB(tetO1) promoter (27). Cells were harvested by centrifugation, washed three times with anaerobic buffer (50 mM Pipes, pH 6.8, 400 mM NaCl, 13 mM KCl, 54 mM MgCl 2 , 2 mM CaCl 2 , 2.8 mM cysteine HCl, 0.4 mM Na 2 S, 4 M resazurin) and stored at Ϫ80°C until use. Purification of homologously produced MA4561-His 6 was performed as described for the His 6 -tagged protein produced in E. coli.
Heme Staining Procedure-Covalently bound heme in proteins can be detected with enhanced chemiluminescence (ECL) reagents due to peroxidase activity of the heme group (31). Proteins were separated by SDS-PAGE using sample buffer without ␤-mercaptoethanol and subsequently electroblotted onto a polyvinylidene fluoride (PVDF) (Roth) membrane. The membrane was covered with Luminata Forte Western blot horseradish peroxidase (HRP) substrate (Millipore) (about 1 ml for a 10 ϫ 7-cm-sized membrane) and incubated for 1 min. Signals were detected using a charge-coupled device imaging system (Multi Image II, Alpha Innotech).
Heme Titration Experiments-Variant sGAF2-His 6 purified without externally added heme was used to perform heme titration experiments in vitro. Increasing amounts of hemin (stock solution 4.6 mM in DMSO) were added to a 10 M concentration of the protein, incubated for 1 h at room temperature, and subjected to the heme staining procedure.
Acidified Butanone Extraction-Acidified butanone extractions were carried out as described before (32). In detail, an aqueous protein solution was titrated with 1 M HCl to give a pH of 1.5-2.0. An equivalent volume of ice-cold 2-butanone was added, and the solution was mixed gently. After cooling on ice, an upper organic layer and a lower aqueous layer were distinguishable. Covalently bound heme stayed in the aqueous layer, whereas non-covalently bound heme was extracted to the organic layer.
Western Blot Analysis-Detection of His 6 -tagged protein was performed using the HisTag Antibody HRP Conjugate kit (Novagen) according to the manufacturer's instructions. Extracts of M. acetivorans strains were electrophoretically separated, transferred to nitrocellulose membranes, and probed with polyclonal antiserum specifically cross-reacting with MtsF (but not with MtsD or MtsH) as described (62).
Optical Absorption Spectra-Optical absorption spectra were recorded on an 8453 UV-visible spectrophotometer (Agilent Technologies) at room temperature. For anaerobic conditions, rubber-sealed cuvettes were used. To obtain the Fe(II) complex, the protein was reduced by adding some crystals of sodium dithionite before sealing the cuvette. The solution was sparged with N 2 gas for at least 30 min to make it completely anaerobic. For preparation of the Fe(II)-CO complex, this solution then was flushed with CO gas for 30 min. To obtain complexes with Im or dimethyl sulfide (DMS), anaerobic stock solutions thereof were used to add specific concentrations to anaerobically prepared Fe(III) or Fe(II) complexes in rubbersealed cuvettes.
Pyridine Hemochrome Assays-Pyridine hemochrome spectra were obtained according to modified, published protocols (33). Specifically, 500 l of a 10 M protein solution were mixed with 62.5 l of pyridine and 62.5 l of 0.5 M NaOH. Then an excess of crystalline sodium dithionite was added to the solution and mixed, and a spectrum of the reduced pyridine hemochromogen was recorded.
Resonance Raman Spectroscopy-Resonance Raman (RR) spectra were acquired using a WITec ␣300AR Raman microscope (Ulm, Germany). The excitation wavelength used was 532 nm from a frequency-doubled neodymium-doped yttrium aluminum garnet laser (CrystaLaser, Reno, NV). The exciting laser radiation was coupled into a Zeiss microscope through a wavelength-specific single mode optical fiber with a diameter of 50 m. The incident laser beam was collimated via an achromatic lens and passed through a holographic band pass filter before being focused into the sample through a Nikon Plan Fluor (20ϫ/0.45 numerical aperture) dry objective. Raman back-scattered light was collected through a microscopic objective and passed through a holographic edge filter onto multimode fiber (50-m diameter) and into a 300-mm-focal length monochromator incorporating an 1800 g/mm grating blazed at 500 nm. Detection of Raman spectra was provided by a backilluminated deep depletion charge-coupled device camera operating at Ϫ60°C. The laser power at the sample point was adjusted to 3 milliwatts. Raman shifts were calibrated with cyclohexane. The accuracy of the peak positions of well defined Raman bands was Ϯ1 cm Ϫ1 . The protein concentrations for the RR experiments were 500 M in 50 mM sodium phosphate, pH 7.0, 100 mM NaCl.
Magnetic Circular Dichroism-Magnetic circular dichroism (MCD) spectra were measured on a Jasco J815 spectropolarimeter fitted with a Jasco MCD-1B magnet at a magnetic field strength of 1.41 T at 4°C using a 0.2-cm-path length quartz cuvette. Jasco software was used for data acquisition and manipulation as reported previously (34). Because the molar absorptivity of the pyridine hemochromogen of the heme in variant sGAF2-His 6 (without Im) could not be determined, concentrations of the sGAF2 samples were tentatively determined based on an estimated value of ⑀ ϭ 115 mM Ϫ1 cm Ϫ1 (the value for ferric cytochrome b 5 ) for the Soret absorption peak (415 nm) of the Im complex of ferric sGAF2 at pH 7.0. UVvisible absorption spectra were recorded before and after each MCD measurement to track sample integrity. sGAF2 protein samples were studied in 50 mM sodium phosphate buffer containing 100 mM NaCl (NaP i ) at pH 7.0 and in 50 mM CAPS buffer at pH 10.0. The Im and tetrahydrothiophene (THT) adducts were prepared by adding 2 and 14 mM Im and THT to the protein, respectively. The ferrous species of sGAF2 was pre-pared from ferric protein in the presence or absence of exogenous ligands (Im and THT) by flushing a rubber-sealed cuvette with nitrogen gas for about 10 -15 min followed by the addition of a small amount of solid sodium dithionite.
Kinase Assays-Autophosphorylation activity was assayed in a reaction mixture containing 50 mM sodium phosphate, pH 8, 5 mM MgCl 2 , 0.2 mM ATP, 20 M MA4561-StrepII full-length protein, and 0.16 nmol of [␥-32 P]ATP (0.0185 MBq) at room temperature. Assays were performed in rubber-sealed cuvettes. The reaction was terminated after different time points by the addition of SDS sample buffer. Samples were separated on a 10% SDS-polyacrylamide gel and electroblotted onto a PVDF membrane. The membrane was exposed to an imaging plate (BAS-MS 2040, Fujifilm) for about 18 h, and the signals were recorded using a bioimager (FLA-3000, Fujifilm).
An sGAF2-StrepII solution at pH 7.0 containing the mediator dyes was injected into the electrochemical cell, and a potential of Ϫ465 mV was applied for 1 min to completely reduce residual oxygen in the cell. Then the potential was reset to ϩ440 mV to produce the Fe(III) complex of MA4561. The potential was reduced by 10 mV steps, and the reaction was equilibrated for 15 min. After equilibration, a spectrum was recorded, and the potential was again reduced. The redox reaction of sGAF2-StrepII was followed by the absorbance change in the Soret band at 428 nm.
To measure the potentials accurately for the reduction and oxidation, well designed mediator mixtures were added (concentration, 0.1 mM) to a 0.4 mM concentration of the protein.
Knock-out and Complementation-The gene encoding MA4561 was deleted in-frame from the M. acetivorans chromosome using pMP44del4561 and a markerless mutagenesis procedure (35). Plasmid pMP44del4561was constructed by cloning (via SacI and SpeI) of a 910-bp fragment (PCR-amplified using primers 19 and 20 (see supplemental Table S1) and genomic DNA of M. acetivorans as template) corresponding to the upstream region of ma4561 into pMP44 (35), which resulted in pMP44del4561up. Inserting into this plasmid (via SphI and SpeI) a 1003-bp fragment (PCR-amplified using primers 21 and 22; supplemental Table S1) corresponding to the downstream region of ma4561 gave pMP44del4561. The genotype of M. acetivorans D4561 (deleted for ma4561) was verified by DNA hybridization (Southern) using an internal fragment of ma4561 (PCR-amplified using primers 23 and 24; supplemental Table S1) as probe. M. acetivorans D4561 was complemented using ptet4561.

RESULTS
The Putative Sensor Kinase MA4561-In silico analysis of MA4561 revealed the presence of two alternating N-terminal PAS and GAF domains fused to a C-terminal H_ATPase domain (Figs. 1 and 2A). A previous bioinformatics characterization of both GAF domains predicted the protein to be a phytochrome-like photoreceptor mainly because of the presence of a conserved cysteine residue in the first GAF domain (26). We overproduced and purified different variants of MA4561 ( Fig.  1) to identify possible cofactors and characterized their biochemical properties.
Heterologously Purified MA4561 Contains a Heme Cofactor in the Second GAF Domain-Purified full-length MA4561-His 6 contained heme as indicated by a typical heme absorption spectrum with a Soret band at 415 nm and a visible band at 532 nm with a shoulder at about 565 nm (Fig. 2B). The protein was not fully saturated with heme as the Soret band was very low in comparison with the protein absorption at 280 nm. Addition of hemin to the clarified lysate as described under "Experimental Procedures" led to a significant increase in Soret band absorption (Fig. 2C). Purified variant GAF2-His 6 (aa 1-751) showed the same absorption properties as the full-length protein where the heme content also increased by addition of external hemin (data not shown). Variants PAS2-His 6 (aa 1-580), GAF1-His 6 (aa 1-457), and PAS1-His 6 (aa 1-290) were also purified but did not show any heme absorption properties even when hemin was added to the clarified lysate (Fig. 2, B and C; only variant PAS2 is shown). Variant sGAF2-His 6 (aa 580 -804) was purified with very little heme bound but showed an intense Soret band at 414 nm and a visible band at 533 nm with a shoulder at about 564 nm when external hemin was added (Fig. 2, B and C).
The protein-heme complex of purified full-length MA4561-His 6 could not be further oxidized by the addition of potassium ferricyanide, indicating the existence of an Fe(III) complex (data not shown). The spectrum of an Fe(II) complex was obtained by reduction with dithionite and had a Soret band at 429 nm and ␣and ␤-bands at 562 and 528 nm, respectively. Upon exposure of this Fe(II) complex to CO gas, Fe(II)-CO complex formation was observed with a Soret band at 424 nm and visible bands at 532 and 560 nm (Fig. 3A). The Fe(II) complex was easily oxidized by aeration with oxygen. As the autoxidation rate in the presence of dithionite is very high, the Fe(II) complex (with excess dithionite) was subjected to gel permeation chromatography on a NAP5 column (GE Healthcare) under aerobic conditions to remove free dithionite. Under these conditions, the Fe(II) complex either binds the oxygen from air to form an Fe(II)-O 2 complex (hemoglobin served as a control; data not shown) or is oxidized to form an Fe(III) complex. No stable Fe(II)-O 2 complex was detected under this condition. Furthermore, aeration of the Fe(II)-CO complex with O 2 did not alter the spectrum (data not shown). The spectra of the Fe(III) and the Fe(II) complexes were stable over a pH range from 6.0 to 10.0 (data not shown). All obtained heme complexes (Fe(II), Fe(II)-CO, and Fe(III)) of the heme binding variants (i.e. full-length MA4561-His 6 , GAF2-His 6 , and sGAF2-His 6 ) have optical absorption properties indicative of six-coordinate complexes with only slight differences (Fig. 3).
To test whether any other cofactors bind to other domains of MA4561, we used variant PAS2-His 6 , which only lacks the heme-binding second GAF domain and the H_ATPase domain, to perform binding assays with the putative phytochrome chromophores biliverdin IX␣ and phycocyanobilin as well as blue light receptor chromophores FAD and FMN but did not obtain any binding (data not shown). Thus, we ruled out the former prediction of MA4561 being a phytochrome-like protein.
Homologously Produced MA4561 Also Binds Heme-To exclude artifacts resulting from heterologous protein production in E. coli and to test whether heme is the native cofactor to MA4561, full-length MA4561-His 6 was produced in M. acetivorans. The purification was identical to that of the protein derived from E. coli except that no external hemin was added to the lysate. The His 6 -tagged protein was detectable in Western blots (Fig. 4A) and showed the same absorption properties as that purified from E. coli (Fig. 4B).
Imidazole Influences the Coordination Structure of the Heme Cofactor-As mentioned above, all His 6 -tagged variants showed optical absorption properties indicative of six-coordinate complexes. In many heme proteins, a histidine residue serves at least as one amino acid to coordinate the heme central iron. Standard purification protocols for native His 6 -tagged proteins use Im to elute the tagged proteins from the metal affinity chromatography column. Because Im resembles the side chain of histidine residues and has the potential to be coor-

The Heme-based Sensor Kinase MA4561
dinated by heme, EDTA was used to elute the His 6 -tagged proteins from the column (as described under "Experimental Procedures") to rule out the possibility of any effect on the coordination structure of the heme cofactor by Im.
Interestingly, the optical absorption spectra of EDTA-purified sGAF2-His 6 showed significant differences. The Soret band had a maximum at 408 nm with a lower overall absorption that was significantly broadened. In the visible region, very weak absorption peaks were obtained with maxima at 504 and 630 nm (Fig. 5A). As already observed for the Im-purified variant, addition of potassium ferricyanide as an oxidizing agent did not alter the spectrum, indicating that the heme is in an Fe(III) state (data not shown). Upon reduction with dithionite to form the Fe(II) complex, the Soret band shifted to 432 nm, and a peak occurred in the visible region with a maximum at 556 nm and shoulders on both sides at around 528 and 574 nm, respectively. Additionally, a small peak at 623 nm was obtained (Fig. 5A). Aeration with CO led to an Fe(II)-CO complex with the same absorption properties as for the Im-purified variant (compare Figs. 5A and 3B, dashed lines). Upon addition of Im to the Fe(III) complex as well as to the Fe(II) complex, optical absorption spectra changed drastically. The obtained complexes exactly matched the "Fe(III)" and "Fe(II)" complexes when purification was performed with Im (compare Figs. 5B and 3B), indicating formation of Fe(II)-Im and Fe(III)-Im complexes. Both complexes could be converted into each other by reducing the Fe(III)-Im complex with dithionite or oxidizing the Fe(II)-Im complex with potassium ferricyanide or O 2 . Addition of CO to the Fe(II)-Im complex produced the above described Fe(II)-CO complex (e.g. Figs. 3B and 5A). To completely rule out effects of the tag and purification system used, we constructed, produced, and purified variants MA4561-Stre-pII (full-length) and sGAF2-StrepII (aa 580 -804). All optical absorption properties for full-length MA4561-StrepII (data not shown) and sGAF2-StrepII (Fig. 5, C and D) were identical to those of the His 6 -tagged variants when purified without Im.
Dimethyl Sulfide Binds to the Heme Cofactor-Because MA4561 is encoded on the chromosome in the vicinity of a system involved in metabolism of methyl sulfides (as discussed later), we tested whether DMS can bind to the heme cofactor of MA4561. DMS is a typical thioether heme ligand resembling the side chain of methionine residues. Therefore, a coordination by the heme iron is feasible. Addition of DMS led to either Fe(III)-DMS or Fe(II)-DMS complexes, which could be converted into each other by reduction and oxidation as already observed for the Im complexes (Fig. 5G). The Fe(III)-DMS and Fe(II)-DMS complexes were nearly indistinguishable from the Fe(III)-Im and Fe(II)-Im complexes with only slightly shifted Soret bands. The Fe(III)-DMS Soret maximum was shifted to a longer wavelength by about 2-3 nm in comparison with the

Different pH Values Influence the Coordination Structure of the Heme
Cofactor-Variant sGAF2-StrepII was used to test the pH stability of the optical absorption spectra. The Fe(III) complex showed no significant changes in the spectrum in buffers with pH 5.4, 7.0, and 10.0. The very broad Soret peak at 429 nm and the weakly resolved visible region with a maximum at 557 nm with shoulders on each side was detectable at pH 7.0 and 5.4. In contrast, the Soret peak at pH 10.0 was sharper and shifted to 425 nm, and in the visible region, two clearly resolved peaks were detectable with the ␤-band at 527 nm and the ␣-band at 557 nm. In all cases, an additional peak at around 625 nm was observed ( Fig. 5H; only pH 7.0 and 10.0 are shown).
Heme Is Covalently Bound to MA4561 at a 1:1 Stoichiometry-Pyridine hemochromogen assays were initially performed to determine the protein to heme ratio. Interestingly, the ␣-band of full-length MA4561 (His 6 -and StrepII-tagged variants) as well as variant GAF2-His 6 and variant sGAF2 (His 6 -and StrepII-tagged variants) had a maximum at 553 nm ( Fig. 5F; only variant sGAF2-StrepII is shown). Typically, the pyridine hemochromogen of heme b shows an ␣-band maximum at 556 nm, whereas the ␣-band maximum of heme c pyridine hemochromogen is 550 nm (36 -38). The shift from 556 nm in heme b to 550 nm in heme c is due to saturation of both heme vinyl groups when covalently attached to two cysteine residues in a CXXCH motif of the protein.
It was reported that in CXXAH and AXXCH variants of a c-type cytochrome the ␣-band maximum is 553 nm due to the saturation of only one heme vinyl group, whereas the other stays unsaturated (38 -40). Because the pyridine hemochromogen of MA4561 displays exactly this absorbance, we assume that the heme group in MA4561 is also bound covalently to the protein via one thioether linkage to a cysteine residue.
To further strengthen the assumption that heme is covalently bound to MA4561, the protein was subjected to a specific heme staining procedure. Variant sGAF2-His 6 (with Im) purified with hemin in the clarified lysate contains covalently bound heme (Fig. 6A, lane 1). In contrast, this variant purified without hemin contains only a little heme (Fig. 6A, lane 2). The covalent attachment of the heme to the protein was additionally verified with other methods, such as MALDI-TOF-MS and denaturing HPLC (data not shown), as well as acidified butanone extraction (Fig. 6B). In addition, the covalently bound heme was verified for the full-length MA4561-StrepII protein (Fig. 6C, lane  3).
As variant sGAF2-His 6 contained very little bound heme when directly purified from E. coli, it was used in titration experiments with hemin. 10 M protein was titrated with increasing amounts of hemin and stained for protein as well as for covalent heme (Fig. 6A). The heme signal increased with increasing amounts of heme up to a hemin concentration of 5 M (Fig. 6A, lane 5). Upon the addition of 7.5 M hemin, an additional signal in the heme stain became visible that is due to the accumulation of free heme (Fig. 6A, lanes 6 -9, marked with an asterisk). The accumulation of free heme at substoichiometric concentrations might be due to the formation of a -oxobridged dimer or heme-DMSO complex formation of the free heme at higher concentrations in the solution (41).
Based on a calculation from the Soret to protein absorption maxima ratio, about 10% of the protein was already saturated with heme before the titration experiment (Fig. 2, B and C). Therefore, a stoichiometry of about one heme molecule per protein monomer can be derived from these calculations. When the same titration experiment was performed under anaerobic conditions and titrated with reduced hemin, identical results were obtained (data not shown). Additionally, titration experiments were performed using denaturing HPLC to determine the amount of heme-bound protein (data not shown). These experiments also confirmed a stoichiometry of one heme molecule per protein monomer.
Heme Is Covalently Bound to Cysteine 656-The pyridine hemochrome spectra shown in Fig. 5F pointed toward a cys-

The Heme-based Sensor Kinase MA4561
teine residue as the site for covalent attachment. Sequence analysis of the heme-binding second GAF domain revealed that it contains only one cysteine residue at position 656. Therefore, the variant sGAF2C656A-His 6 was constructed to examine the role of the cysteine 656 residue in heme binding. This variant was no longer able to covalently bind heme as shown by specific heme staining (Fig. 6C, lane 2) and acidified butanone extraction (Fig. 6B). Also, the pyridine hemochrome spectrum showed an ␣-band maximum at 556 nm, indicating that both heme vinyl groups are unsaturated (Fig. 5F). Notably, UV-visible spectra were identical to those of variant sGAF2 with the exception that no band at 625 nm was observed (Fig. 5E), and Im complexes were formed with the same absorption properties as in wild type sGAF2 (data not shown). Based on these data, the heme cofactor is covalently attached to the protein via cysteine 656 (Cys 656 ). Exchange of this residue to alanine did not dramatically influence the optical absorption properties of the complex (Fig. 5E). Also, RR spectra indicated that the heme coordination is the same in the wild type and the C656A variant (Fig. 7). Therefore, non-covalently bound heme most likely occupies the same position within MA4561 as covalently bound heme.
Axial Ligation of the Heme Cofactor (RR Spectroscopy)-RR spectroscopy was performed to further analyze the coordination structure of the heme cofactor. Fig. 7 depicts the visible excited (532 nm) RR spectra of the Fe(III) (A) and Fe(II) complex (B) of the sGAF2-His 6 (without Im) (a) and sGAF2C656A-His 6 (without Im) (b) variants in the high wave number region. They reflect the vibrations of the heme in these variants. It is well established that the Raman spectra of heme proteins in the high wave number region comprise porphyrin in-plane modes, which are sensitive to oxidation, spin, and coordination states of the heme iron (42)(43)(44)(45).
The Raman bands of the ferric form of wild type sGAF2-His 6 at 1377, 1506, 1573, and 1645 cm Ϫ1 (Fig. 7A, a) are assigned to 4 , 3 , 2 , and 10 , respectively, and their wave numbers indicate that the heme adopts a six-coordinate low spin (6c-LS) state. Similar results were observed for the sGAF2C656A-His 6 variant (Fig. 7A, b), implying that the mutation of Cys 656 to alanine has no significant effect on the heme coordination structure and the spin state. The 4 , 3 , 2 , and 10 bands of the Fe(II) complex of wild type sGAF2-His 6 (Fig. 7B, a) are observed at 1360/1369, 1498, 1592, and 1631 cm Ϫ1 , respectively, suggesting a 6c-LS type of heme. Besides, 4 is located at 1360 and 1369 cm Ϫ1 , suggesting the presence of two conformations in which the heme adopts a 6c-LS state. The mutation of Cys 656 perturbs the populations of the 1360 and 1369 cm Ϫ1 fractions. The population of the 1360 cm Ϫ1 species is much higher than that of 1369 cm Ϫ1 (Fig. 7B, b). Also, the intensity of the Raman bands at 1555 and 1607 cm Ϫ1 is significantly enhanced than that in the wild type protein, indicating that the interactions of the heme vinyl groups with the surroundings are altered upon mutation of cysteine residue 656.
Axial ligation of the heme cofactor (MCD)-MCD spectra of ferric and ferrous states of sGAF2-His 6 (without Im) are shown in Fig. 8 where analogous states of horse heart myoglobin (Mb), HRP, and cytochrome b 5 (cyt b 5 ) are overlaid for comparison. At pH 7.0, ferric sGAF2 exhibits an MCD spectrum (Fig. 8A, solid line) that is similar to that of ferric HRP (dashed line) and ferric Mb (dotted line), five-coordinate and water-ligated sixcoordinate heme proteins, respectively, in the visible region (450 -700 nm). However, a closer comparison of the trough intensities at ϳ640 nm indicates that the spectrum of ferric sGAF2 is more similar to that of HRP than Mb. In the Soret region (350 -450 nm), an asymmetric MCD spectrum having a prominent peak at ϳ404 nm (intensity, ϳ7.5 M Ϫ1 cm Ϫ1 T Ϫ1 ) and two troughs at ϳ370 and ϳ430 nm (Ͻ5 M Ϫ1 cm Ϫ1 T Ϫ1 ) does not resemble either a weak negative feature (a trough at ϳ420 nm; ϳ4 M Ϫ1 cm Ϫ1 T Ϫ1 ) of HRP or a broadly symmetric derivative-shaped spectral feature (Ϯϳ15 M Ϫ1 cm Ϫ1 T Ϫ1 ) of Mb. The MCD spectrum of ferric sGAF2 at pH 10.0 (gray line) is clearly different from that at pH 7.0 in that one of the visible region troughs (ϳ550 nm) is shifted to ϳ575 nm, and a Soret region feature becomes derivative-shaped with considerably enhanced intensity (ϩ34/Ϫ40 M Ϫ1 cm Ϫ1 T Ϫ1 ). The resultant spectral features are indicative of an increased 6c-LS species.
In the ferrous state, sGAF2 (Fig. 8B, solid line) has MCD spectral features that are broadly similar to those of ferrous Mb (which contains a five-coordinate heme; Fig. 8B, dotted line) including quite intense Soret peaks (ϳ125 M Ϫ1 cm Ϫ1 T Ϫ1 ) except for a more enhanced and derivative-like shape in the visible region around 555 nm for sGAF2 than for Mb. At pH 10.0, this derivative-shaped spectral line becomes much more intense. Concomitantly, the Soret peak intensity diminishes by ϳ50% (Fig. 8B, gray line). These MCD spectral changes are indicative of an increase in the fraction of a ferrous 6c-LS heme such as bisamine (or bis-His)-ligated ferrous heme (see below). Fig. 8C compares ferric Im (solid line) and ferric THT (a cyclic thioether) adducts (dashed line) of sGAF2 with ferric cyt b 5 (dotted line). These three heme adducts display similar MCD spectral patterns, namely relatively intense derivative-shaped Soret spectra and mainly positive (450 -560 nm) and negative (560 -700 nm) features with a prominent trough at 570 -580 nm. The MCD spectral similarity for the ferric Im complex of sGAF2 and ferric cyt b 5 (which contains a bis-His-ligated heme) suggests identical ligation modes for the two heme proteins. Because bisamine (or bis-His)-ligated and bisthioether-ligated ferric hemes have been shown to have similar MCD spectra (46), mono-Im (or His)/monothioether-bound ferric heme also is expected to share this similarity. In Fig. 8D, MCD spectra of ferrous Im and ferrous THT adducts are compared with the spectrum of ferrous cyt b 5 . As seen in their ferric state (Fig. 8C), the MCD spectral patterns of these three ferrous complexes show common patterns. In the visible region, extremely intense derivative-shaped features around ϳ555 nm are notable. In the Soret region, the MCD spectra of the THT-bound sGAF2 and cyt b 5 are similar to each other in pattern (except for some intensity difference), whereas the Im complex of sGAF2 shows a somewhat different pattern with a noticeably intense trough at ϳ425 nm. A possible cause of this difference could be incomplete formation of the Im complex; however, an increase of Im concentration (2-50 mM) does not change the outcome.
Model Structure of sGAF2 Variant-To structurally support the spectroscopic data, the sGAF2 domain (aa 608 -768) was modeled using the PHYRE2 server (47). Based on six template sequences, which were all GAF domains or GAF domain-like structures, a model of the second GAF domain of MA4561 was generated (Fig. 9). The modeled structure was also used to predict putative ligand-binding sites using 3DLigandSite (48). Best hits were obtained for heme molecules. One of the heme vinyl groups is in direct vicinity to the cysteine 656 residue, which is the covalent attachment site for the heme. Putative axial ligands like histidine residues or methionine residues are located in the vicinity of the heme iron and strengthen the conclusion of the MCD data.
Autophosphorylation of MA4561 Is Dependent on the Heme Redox State but Not on External Ligands-As MA4561 contains an H_ATPase domain at its C terminus, we tested whether the protein is autophosphorylated using radioactively labeled ATP. Kinase assays of all heme complexes of full-length MA4561-StrepII were performed (Fig. 10). Although the Fe(III) complex showed increasing autophosphorylation over time, the Fe(II) and Fe(II)-CO complexes were not autophosphorylated (Fig.  10). Binding of DMS had no influence on the kinase activity with the Fe(III)-DMS complex showing autophosphorylation activity and the Fe(II)-DMS complex not showing autophosphorylation activity (Fig. 10). The same was true for Im complexes (data not shown).
Redox Potential of MA4561-To gain insight into the oxidation state of MA4561 within M. acetivorans, the redox potential of recombinant variant sGAF2-StrepII was determined using redox titration experiments. These measurements obtained midpoint potentials of Ϫ95 Ϯ 10 and Ϫ75 Ϯ 10 mV for the oxidation and reduction reactions, respectively (Fig. 11). Fulllength MA4561-His 6 (with Im) showed comparable midpoint potentials of Ϫ91 Ϯ 10 and Ϫ85 Ϯ 10 mV for the oxidation and reduction reactions, respectively (data not shown).
MA4561 Regulates the Synthesis of MtsF, a Corrinoid/Methyltransferase Fusion Protein-Because we suspected MA4561 to play a role in signal transduction for regulating the vicinally encoded MtsH and potentially MtsD and MtsF, three highly similar corrinoid/methyltransferase fusion proteins involved in DMS metabolism (62), a markerless in-frame deletion of the ma4561 coding region was constructed. The strain (D4561) grew indistinguishably from the wild type under all conditions tested (growth on methanol, trimethylamine, and acetate; data not shown). To test the influence of MA4561 on the synthesis of the corrinoid/methyltransferase fusion proteins, Western blot analyses were performed with the only available antiserum specific for the MtsF protein. MtsF was not detectable in Western blots when wild type cells were grown on methanol (Fig. 12, lane 1) (62). Addition of 5 mM DMS to the growth medium resulted in a strong MtsF signal (Fig. 12, lane 2). Interestingly, in strain D4561 carrying the "empty" complementation vector pMR08, this DMS-dependent regulation of MtsF was abol- The second GAF domain (aa 608 -768) of MA4561 was modeled using the PHYRE2 server (47), and the modeled structure was used to predict putative ligands binding to this structure using 3DLigandSite (48). The heme cofactor with highest priority is shown in the middle as sticks with the central iron atom as a sphere, and the protein structure is shown as a ribbon (interesting aa are shown as sticks). The cysteine residue (Cys 656 ) that is the covalent heme-binding site is located close to one of the two heme vinyl groups. Two histidine residues (His 702 and His 704 ) and a methionine residue (Met 653 ) are located at the proximal heme site. Two tyrosine residues (Tyr 632 and Tyr 665 ), a methionine residue (Met 645 ), and a histidine residue (His 646 ) are located at the distal heme site. ished, resulting in constitutive synthesis of MtsF (Fig. 12, lanes  3 and 4). When D4561 was complemented with the ma4561 allele encoding the MA4561-His 6 variant (which was also used for in vitro analyses; see above), the wild type situation was restored (Fig. 12, lanes 5 and 6).

DISCUSSION
MA4561 Is Not a Phytochrome-The multidomain protein MA4561 from M. acetivorans was previously predicted to be a phytochrome-like photoreceptor with a histidine kinase-like output domain (22,26). However, MA4561 is unable to bind any known chromophores of phytochrome photoreceptors (i.e. open-chain linear tetrapyrroles). Based on this result and because it is unknown whether the methanogenic archaeon M. acetivorans is able to produce open-chain tetrapyrroles as their classical biosynthesis requires molecular oxygen, we rule out the possibility that this protein is a phytochrome. In addition, the domain organization is different from that observed in bona fide phytochromes, which consist of a PAS-GAF-PHY (phytochrome) domain structure. Although recently so-called cyanobacteriochromes have been described that lack the PHY domain and are still photochromic (49), MA4561 lacks all spectral features of a photoreceptor.
MA4561 Harbors an Intrinsic Heme Lyase Activity-Nonetheless, we were able to demonstrate that MA4561 binds a tetrapyrrole cofactor. Heme was shown to be covalently attached via a thioether linkage to cysteine 656 in the second GAF domain. Our spectroscopic data point toward a linkage between the sulfhydryl group and one of the heme vinyl groups similar to the linkage observed in cytochrome c. Formation of c-type cytochromes in vivo depends on the c-type cytochrome maturation system (50,51). However, in vitro formation of c-type cytochrome is also possible in the absence of the maturation machinery (36). In this case, a correct attachment is only observed when both heme and apocytochrome are in a reduced state; all other combinations lead to incorrect attachment of the heme. Covalent attachment of heme to MA4561 occurs regardless of whether the protein is produced in the native organism M. acetivorans or heterologously in E. coli (with the exception that not all the protein is saturated with heme). The cofactor also binds covalently in vitro as it can be titrated afterwards, directly in the cell lysate and even to the purified protein, independently of the redox state of the titrated hemin and the protein itself. Our data suggest that the ability of MA4561 to covalently bind the heme is an intrinsic property of the protein itself. MA4561 binds heme and shows autophosphorylation activity dependent on the heme redox state, indicating a correctly bound cofactor.
The axial coordination of the heme iron in the protein is not significantly affected by a mutation of the cysteine residue to an alanine residue as shown by UV-visible (Fig. 5E) and RR spectra (Fig. 7). Although the heme in the Fe(II) complex seems to be found in a mixture of two slightly different six-coordinate states (see below), one of those is more dominant in the cysteine to alanine mutation. The in vivo relevance of this covalent attachment is thus unclear and requires further characterization. Differences in biophysical properties other than spectral properties might be influenced by the covalent binding.
MA4561 Contains Heme in a Large, Accessible Binding Pocket-Extensive UV-visible spectral analyses of differentially tagged heme binding variants of MA4561 unraveled an influence of the purification system used on the heme cofactor. When Im was used for the purification of the His 6 -tagged variants, the UV-visible spectral properties differed from those obtained with StrepII-tagged variants and Im-free His 6 -tagged variants. The tag itself had no influence on the optical absorption properties as the spectra of the StrepII-tagged variants were identical to those of Im-free His 6 -tagged variants. With addition of Im to the Im-free variants (independent from the tag used), the spectra obtained perfectly matched the spectra of the Im-purified protein. As a consequence, we suggest a heme pocket architecture that is accessible even to larger molecules like Im. This assumption is strengthened by the fact that also DMS (or alternatively the even bulkier THT) can coordinate the heme.
Coordination Structure of the Heme Cofactor Is Heterogeneous-For more insight into the coordination structure of the heme, we performed extensive UV-visible spectral analyses combined with RR and MCD measurements. The UV-visible spectra of the Fe(III) complex as purified from E. coli without Im with a His 6 tag as well as with a StrepII tag showed a Soret band at about 407 nm and weak bands in the visible region of about 505 and 630 nm at pH 7.0. These absorption properties are indicative of a six-coordinate high spin heme as found in HemAT (52). The absorption properties clearly differ from the 6c-LS heme found in the direct oxygen sensor EcDos from E. coli (53,54), recombinant barley hemoglobin (55), CO sensor CooA from R. rubrum (6,56), and globin-coupled histidine kinase AfGcHK from Anaeromyxobacter sp. Fw109-5 (20) and from the five-coordinate high spin heme found in YddV (57) and BjFixL from Bradyrhizobium japonicum (58).
In the RR measurements, a 6c-LS heme was observed, whereas MCD data suggest a high spin state and a five-coordinate structure rather than a six-coordinate structure. Resonance conditions used in the RR experiments most likely would not allow detection of a five-coordinate high spin heme species. Therefore, a clear assignment of the coordination structure cannot be made, suggesting a mixture of different coordination states.
When the pH of the buffer was changed to pH 10.0, a mixture of a high spin and a low spin component was also detectable in MCD spectra, whereas the differences in UV-visible spectra were minor. This low spin component appears to be arising from a bis-His-type coordination structure as judged from the similarity to the Fe(III)-Im complex (as discussed below).  1 and 2), D4561 carrying pMR08 (mutant containing empty vector) (lanes 3 and 4), and D4561 carrying ptet4561 (complemented mutant) (lanes 5 and 6) grown on methanol in the absence (lanes 1, 3, and 5) or presence of 5 mM DMS (lanes 2, 4, and 6) were separated by SDS-PAGE, transferred onto a nitrocellulose membrane, and probed with antiserum specific for MtsF.
Upon reduction of MA4561 with dithionite, an Fe(II) complex with a broad Soret band at about 432 nm and a visible band at about 557 nm with shoulders at about 528 and 574 nm is formed. An additional peak was observed at about 625 nm for the sGAF2 variants. These optical absorption properties cannot be clearly assigned to one coordination and spin state but rather look like a mixture of different states. With RR data, a mixture of two different 6c-LS states was observed, and the existence of two species was confirmed in the MCD measurements. At pH 10.0, one of the two states becomes predominant as judged from the MCD data and is similar to the bis-His-ligated Fe(II) form of cyt b 5 .
The addition of Im or DMS (or rather THT for the MCD measurements) led to typical 6c-LS hemes for the Fe(III)-(Im/ DMS/THT) as well as the Fe(II)-(Im/DMS/THT) forms (53,54). The Im complexes were hardly distinguishable from the DMS (THT) complexes with only minor differences in UVvisible as well as MCD spectra. Similarities of the Im complexes to cyt b 5 led to the conclusion that a His/Im or in the case of DMS (THT) a His/DMS (THT) complex is formed that would resemble a bis-His (cyt b 5 ) or hardly distinguishable His/Met (many c-type cytochromes) ligation structure.
The structural model of sGAF2 generated supports the spectral data with putative axial ligands in proximity of the heme cofactor. Specifically, two histidine residues (His 702 and His 704 ) are located at the proximal side of the heme pocket that might be involved in coordinating the heme iron (Fig. 9). Initial mutational analyses of these residues suggest His 702 as the proximal ligand. 4 The distal side is occupied by two tyrosine residues (Tyr 632 and Tyr 665 ), one methionine residue (Met 645 ), and one histidine residue (His 646 ) that might be directly or indirectly involved in coordinating the heme iron depending on the redox state. Because ligands like CO, DMS, and Im are able to bind to the heme in MA4561, we envision the distal side of the heme pocket large enough to allow these molecules to enter and coordinate the heme iron. The above mentioned amino acids might be involved in stabilizing different ligation states of the heme without being a direct ligand to the heme iron. Alternatively, they might be exchanged as a direct ligand upon binding of external ligands as most data suggest six-coordinate states also in the absence of external ligands. Interestingly, the cysteine residue that binds the heme covalently is located near one of the heme vinyl groups, giving the modeled structure more confidence.
MA4561 Is a Redox Sensor-The heme in MA4561 is easily oxidized and reduced by different oxidizing and reducing agents independently of the coordination of the external ligands Im and DMS. Autophosphorylation is strongly dependent on the heme redox state, and activity was observed for the Fe(III), the Fe(III)-DMS, and the Fe(III)-Im complexes. In contrast, all reduced complexes had no autophosphorylation activity (i.e. Fe(II), Fe(II)-DMS, Fe(II)-Im, and Fe(II)-CO). As M. acetivorans is an obligate anaerobe (59), it is very unlikely that oxygen is the physiological signal leading to the oxidation of the heme iron and subsequently to an activation of the kinase.
Although it was shown that the organism can cope with microaerophilic conditions (60), we consider a change in redox state of the cell as a more realistic signal (see below).
The positive and negative ends of the redox potentials to which M. acetivorans is exposed to may be defined by hydrogen at the negative end (with a redox potential of about Ϫ420 mV) and by the terminal electron acceptor of the methanogenic pathway, the heterodisulfide of coenzyme M and coenzyme B at the positive end (with a redox potential of about Ϫ140 mV) (61). The redox potential of the heme in MA4561 was determined to be about Ϫ80 to Ϫ90 mV. Under physiological conditions, the heme in MA4561 is therefore likely in the Fe(II) state, and consequently, the kinase activity is switched off. We envisage that MA4561 might be involved in redox sensing. Local accumulation of oxidized cofactors could lead to oxidation of the heme iron in MA4561 that in turn would switch on kinase activity. Alternatively, sulfur-containing metabolites like methyl sulfides could also be mediators for redox sensing either directly or indirectly (see below). Further analyses are warranted to substantiate this idea.
MA4561 Is Part of a System Regulating Methyl Sulfide Metabolism-MA4561 is encoded directly upstream of the transcriptional regulator MsrG (MA4560) (62). This regulator is the direct transcriptional activator of the protein MtsH (MA4558), which is one of three highly similar corrinoid/methyltransferase fusion protein isoforms required for methyl sulfide metabolism (62,63). These enzymes are involved in DMS formation when grown with CO and are essential for growth with DMS as the sole energy source (64). There are two other Mts isoforms present in M. acetivorans, MtsD (MA0859) and MtsF (MA4384), which are also regulated individually by vicinally encoded transcriptional regulators. MtsD transcription is activated by MsrF (MA0862), and MtsF transcription is activated by MsrC (MA4383). MsrF is encoded directly downstream of the putative sensor kinase MA0863, which shares 68% identity and 84% similarity with MA4561. The regulator proteins all belong to the Msr family of regulators specific to Archaea (65). It can therefore be speculated that MsrF and MsrG, both encoded in putative transcriptional units with the respective kinases MA0863 and MA4561, might form twocomponent systems with these sensor kinases.
Interestingly, regulatory cross-talk among MtsD, MtsF, and MtsH has been demonstrated previously (63). Abolishing DMS-dependent regulation of synthesis of MtsF by removing MA4561 and restoring this regulation by complementation with its His 6 -tagged variant demonstrates that (i) the tagged variant of the protein used for our in vitro studies is functional in vivo, (ii) MA4561 is directly or indirectly involved in regulation of MtsF (and presumably of the other Mts proteins as well), and (iii) MA4561 is a negative effector of MtsF synthesis. Based on these findings, we propose to name the protein MsmS for methyl sulfide methyltransferase-associated sensor. Removing MsmS leads to constitutive MtsF synthesis, which indicates constant MsrG (or MsrC)-dependent activation of mtsF expression.
Because autophosphorylation of MsmS is dependent on the heme redox state rather than on ligand binding, the mechanism underlying this deregulation cannot easily be explained by CO-or DMS-dependent phosphoryl transfer from the sensor kinase to the respective response regulator to activate it. Instead, binding of CO or DMS might keep MsmS in an unphosphorylated state as both gases are weak reductants. Thus, the presumed phosphoryl transfer could effect the inactivation of the response regulator. Although such a scenario is consistent with our observations, much of it is speculative. It is unknown whether the phenotype seen in D4561 is due to a direct interaction of DMS with MsmS, leading to a signal output other than kinase activity, or an indirect effect (e.g. accumulation of oxidized cofactors during growth on DMS). These questions will be addressed in the future.