Heme-based Globin-coupled Oxygen Sensors: Linking Oxygen Binding to Functional Regulation of Diguanylate Cyclase, Histidine Kinase, and Methyl-accepting Chemotaxis*

An emerging class of novel heme-based oxygen sensors containing a globin fold binds and senses environmental O2 via a heme iron complex. Structure-function relationships of oxygen sensors containing a heme-bound globin fold are different from those containing heme-bound PAS and GAF folds. It is thus worth reconsidering from an evolutionary perspective how heme-bound proteins with a globin fold similar to that of hemoglobin and myoglobin could act as O2 sensors. Here, we summarize the molecular mechanisms of heme-based oxygen sensors containing a globin fold in an effort to shed light on the O2-sensing properties and O2-stimulated catalytic enhancement observed for these proteins.


What Is a Heme-based Oxygen Sensor?
Hb, the prototypical O 2 -binding heme protein, is the initial binding site for external O 2 molecules in the body and, as such, is important for numerous important subsequent physiological functions that use oxygen (1,2). Mb, which has a similar protein structure to Hb, acts as an O 2 storage site (1,2). Hb and Mb additionally perform other functions, such as detoxification of NO via dioxygenation to form nitrate (3,4). The histidine (E7) imidazole located at the heme distal side in the heme Fe(II)-O 2 complex of Hb and Mb plays a significant role in preferentially stabilizing bound O 2 via hydrogen bonding and the resistance of globin-O 2 complexes to autoxidation (5,6). An O 2 -binding site is also present in other heme proteins, including hemebased monooxygenases such as cytochrome P450 and heme oxygenase. As such, many O 2 -binding heme proteins contain the heme iron complex as their active site/center.
In an extension of this concept, an emerging class of hemebased oxygen sensors uses the heme iron complex as the O 2 -sensing site for intramolecular signal transduction. In general, the heme-based oxygen sensors are composed of an N-terminal heme-bound oxygen-sensing/binding domain containing globin, PAS (Per-Arnt-Sim), or GAF (cGMP-specific and cGMP-stimulated phosphodiesterases, adenylate cyclases, and Escherichia coli FhlA) folds and a C-terminal functional/catalytic domain ( Fig. 1) (7)(8)(9)(10)(11)(12)(13). O 2 (the first signal) binds to/dissociates from the heme iron complex and induces a structural change (the second signal) in the heme-bound oxygen-sensing domain of the protein. These induced structural changes in the protein are then transduced to the functional domain, switching on (or off) the function of the C-terminal catalytic domain. The C-terminal domain may have various catalytic functions, such as diguanylate cyclase (DGC) 5 activity, phosphodiesterase (PDE) activity toward cyclic diGMP (c-diGMP), and histidine kinase (HK) activity. A similar heme-based structural change underlies the signal transduction function of methyl-accepting chemotaxis proteins (MCPs) (Figs. 1 and 2 and Table 1).
In this minireview, we illustrate the concept of heme-based oxygen sensors by 1) exploring the structure-function relationships of globin-coupled oxygen sensors (GCSs), which contain a heme-bound globin fold (14 -16) or heme pocket that constitutes the O 2 -sensing/binding site, and 2) comparing the heme Fe(II)-O 2 complex equilibrium dissociation constant (O 2 affinity) and autoxidation rate constant (stability of the heme Fe(II)-O 2 complex) of these GCSs with those of other hemebased oxygen sensors containing heme-bound PAS or GAF domains ( Table 1).

The Globin-coupled Oxygen Sensor
Globins have been identified in 1185 of 2275 bacterial genomes and 32 of 140 archaeal genomes (16) and further categorized into myoglobin-like globins with a 3/3 ␣-helix fold, sensor globins with a 3/3 ␣-helix fold, and truncated globins with a 2/2 ␣-helix fold. The sensor globins constitute GCSs, protoglobins, and sensor single-domain globins (14 -18). It must be noted that not all sensor globins have enzyme function. Moreover, numerous sensor globins involved in signaling have unknown functions that remain to be established, such as GCS from Geobacter sulfurreducens (GsGCS) (Fig. 1) (19). GCSs are defined as chimeric proteins composed of heme-bound globin and functional domains ( Fig. 1) (14 -18). The globin domain of GCS lacks the entire D-helix and part of the E-helices of Mbs and Hbs, which have a common genetic ancestry with sensor globins. The resulting shortened globin fold is predicted to have special characteristics that should be beneficial for heme-based oxygen sensors that use the shortened globin fold for oxygen sensing (14 -16, 18).
Biochemical characterizations in reconstituted systems using purified enzymes have been reported for only seven of the 420 previously reported putative bacterial GCSs (14 -16). Reports on the physicochemical characteristics of these proteins have focused on the structure of the heme environment, with only a few studies investigating biochemical reactions involving specific substrates and products.

MCP-GCS: HemAT
HemAT is likely the MCP-GCS protein that controls the movement direction of archaea and bacteria such as Halobacterium salinarum and Bacillus subtilis (HemAT-Bs) in an O 2 gradient (20,21). HemAT is the only MCP-GCS in which the molecular characteristics of the heme-bound domain have been well studied. Crystal structures of heme Fe(II) and heme Fe(III)-CN (cyano) complexes of the isolated heme-bound domain of HemAT-Bs have been solved (Fig. 3) (22). The crystal structures revealed that the tyrosine hydroxyl of Tyr-70(B10) in one subunit of the dimer of the heme Fe(II)-bound domain faces the protein surface. In contrast, the tyrosine hydroxyl in the same subunit interacts with CN in the heme Fe(III)-CN complex. Based on the structures of the heme Fe(II) and Fe(III)-CN complexes of HemAT-Bs and amino acid sequence alignment of GCSs, it is possible to speculate about the nature of the O 2 -binding site (Fig. 3) (14,15,23 (21). Moreover, based on the crystal structure, HemAT is proposed to utilize heterogeneity or negative cooperativity (or conversion from an asymmetrical to a more symmetrical form upon O 2 binding) to expand the dynamic rage for detecting O 2 and to transfer structural information to downstream HK (22,24). However, no concrete functional regulation by O 2 binding and/or change in redox status has been reported in reconstituted systems in vitro, possibly because no specific substrate (and thus, no product) of a HemAT-mediated reac- Similarly, DevS and DosT have two tandem GAF domains; in both cases, the heme iron complex is bound to the first GAF domain. Note that globin folds are not always localized at the N terminus but are inserted into diverse regions, as predicted from the amino acid sequences of sensor globins that are yet to be characterized (14 -18). AxPDEA1, A. xylinum PDEA1. Once autophosphorylated, AfGcHK transfers the phosphate group to the cognate response regulator. Although, the enzyme is depicted here as a monomer for the sake of simplicity, all HKs reported to date, including AfGcHK, are dimers (or tetramers). It has been suggested that autophosphorylation is exerted "crosswise" such that the kinase in one subunit phosphorylates the His residue of HK in the other subunit of the dimer (or tetramer). aa, amino acids; Rec, receiver domain. This figure is adapted from Ref. 44. tion has been identified. The C-terminal domain interacts with the HK CheA, which is a component of the CheA/CheY twocomponent signal transduction system in bacterial aerotaxis (20,21,31).
Azotobacter vinelandii GReg (AvGReg)-Both the full-length and isolated heme-bound domains of GCS from A. vinelandii were cloned, overexpressed in E. coli, and characterized (35). Spectral and ligand-binding kinetic studies were conducted for recombinant proteins. The heme Fe(III) complex of the fulllength enzyme was in the bis-histidine hexacoordinate form. The heme Fe(II) complex of the full-length enzyme was also hexacoordinated with bis-histidine, whereas that of the isolated heme-bound domain was pentacoordinated. O 2 affinities for both proteins were high. The heme Fe(II)-O 2 complex of the isolated heme-bound domain remained stable over several hours. Furthermore, examination of ligand (such as O 2 and NO)-binding kinetic properties suggested that AvGReg plays a role in O 2 -mediated NO detoxification via dioxygenation of NO to form nitrate (3,4,35). A characterization of the pheno-types of Salmonella typhimurium cells overexpressing AvGReg revealed that biofilm formation and swimming motility were influenced by AvGReg (35,36). Although the DGC activity of AvGReg has not been characterized in vitro, these latter results suggest that AvGReg is involved in c-diGMP synthesis in vivo.
Bordetella pertussis GReg (BpeGReg)-BpeGReg was cloned and overexpressed in E. coli and was the first GCS that was unequivocally proven to possess DGC activity (36). This study showed that the catalytic activity of BpeGReg is significantly stimulated by O 2 binding to the heme Fe(II) complex. Binding of NO and CO to the heme Fe(II) complex also enhances the catalytic activity of this enzyme. The study provided an interesting homology model based on the crystal structures of HemAT-Bs and PleD, a DGC from Caulobacter crescentus, and proposed that c-diGMP binds to inhibitory sites, causing feedback inhibition of enzyme activity. More importantly, the middle domain between the heme-bound globin and DGC domains was suggested to be required for proper folding of the DGC domain, but not the heme-bound domain. Relationships between O 2 binding, active dimer formation, and autophosphorylation were also implicated (36). The involvement of BpeGReg in biofilm formation and bacterial motility was demonstrated by protein overexpression in S. typhimurium or gene knock-out in B. pertussis.
HemDGC-HemDGC from Desulfotalea psychrophila was identified and characterized (37). It was shown that only the heme Fe(II)-O 2 complex of HemDGC exhibits DGC activity, whereas heme Fe(II), heme Fe(II)-NO, heme Fe(II)-CO, and heme Fe(III) complexes do not. A discussion of the protein structure of the heme distal side emphasized the role of Tyr-55(B10) at or near the O 2 -binding site ( Fig. 3 and Table 1).
YddV (EcDosC)-In the first report of YddV, a heme-based oxygen sensor DGC from E. coli (also designated EcDosC) (38), it was shown that incubation of YddV in a solution containing GTP (the substrate of YddV) and EcDOS (with PDE activity toward c-diGMP) produced linear diGMP (pGpG) as the final product, providing indirect evidence of DGC activity. A separate study from our group (32) Table 1). The heme Fe(III) and heme Fe(II)-CO complexes were also active, whereas the heme Fe(II) and heme Fe(II)-NO complexes were inactive. Analyses of the physicochemical properties of mutant proteins suggested that Tyr-43(B10) is adjacent to the O 2 -binding site on the heme distal side (Fig. 3) (32, 39). The hydroxyl of Tyr-43(B10) appears to preferentially stabilize O 2 bound to the heme Fe(II) complex via hydrogen bond donation, as observed for other GCSs. The electron of O 2 bound to the heme Fe(II) atom is partially removed with a negative charge and interacts strongly with proton donors or positive electrostatic fields induced by Tyr-43. However, the residues that are situated at or near CO-or NO-binding sites in the heme distal side possibly differ from Tyr-43, as suggested from ligand-binding kinetics and resonance Raman spectra of the Tyr-43 mutant proteins. This structural difference in the ligand-bound heme complex in YddV appears to reflect the variations in catalytic regulation by CO and NO. The crystal structure of HemAT suggests that the partial negative charge of Tyr-70(B10) on bound O 2 can "pull" hydrogen-bonding donors toward it, whereas CO and NO would not be able to cause such a movement of Tyr-70(B10) (22), similar to the case of the distal Arg side chain for FixL (40,41). Moreover, it has been speculated that the environmental structure of the heme Fe(II) complex in YddV is significantly different from that of the heme Fe(III) complex, as observed for FixL (40,41) and EcDOS (42,43), which are heme-based oxygen sensors with a PAS fold, resulting in heme redox-dependent catalytic regulation. Biofilm formation by E. coli is significantly affected by overexpression of YddV, suggesting again that YddV is involved in biofilm formation in vivo.

HK-GCS: Anaeromyxobacter sp. Strain Fw109-5 Globin-coupled HK (AfGcHK)
AfGcHK, isolated from the soil bacteria Anaeromyxobacter sp. strain Fw109-5, is the first reported GCS with HK activity. The HK activity of AfGcHK is markedly stimulated by the binding of an O 2 molecule to the heme Fe(II) complex (44). As noted above, evolutionary considerations led to the discovery of GCS enzymes involved in MCP and c-diGMP homeostasis (synthesis), but GCS enzymes with other activities were not known (14 -16). On the basis of bacterial genomic sequences, it was predicted that soil bacteria encode a protein containing an amino acid sequence corresponding to the GCS globin fold. In addition, it was thought that the same protein has an HK domain similar to that of other heme-based oxygen sensor HKs involved in two-component systems, such as FixL, DevS, and DosT ( Fig. 1, left).
As predicted, the autophosphorylation of AfGcHK at His-183 and the AfGcHK-mediated phosphorylation of Asp-52 and Asp-169 in the response regulator (via a phosphotransfer reaction) were significantly stimulated by the binding of O 2 to the Fe(II) heme complex in the N-terminal sensor domain, whereas the heme Fe(II) complex alone showed no catalytic activity. CO binding to the Fe(II) heme complex also enhanced catalysis. These gas-induced catalytic enhancements and the redox-dependent catalytic regulation of AfGcHK are similar to those observed for YddV (32).
Tyr-45(B10) appears to be at or near the O 2 -binding site in AfGcHK based on an amino acid sequence alignment of GCSs, the crystal structure of the heme Fe(III)-CN complex of HemAT (Fig. 3), and physicochemical properties of mutant proteins. Again, the hydroxyl of Tyr-45(B10) interacts with O 2 in the heme Fe(II)-O 2 complex and stabilizes the complex via hydrogen bond donation, as observed for YddV (32).

GCSs with Unknown Function
The x-ray crystal structures and physicochemical properties of the isolated heme-bound domain of GsGCS have been reported (Fig. 3) (19). The heme Fe(III) complex of this protein displays bis-histidyl hexacoordination, whereas the heme Fe(II) complex is an admixture of penta-and hexacoordinated complexes. Interestingly, distal heme coordination of the heme Fe(III) complex in GsGCS is provided by a His residue unexpectedly located at the E11 topological side, distinct from that at the E7 site of Hb, Mb, and other GCSs. Resonance Raman spectral and ligand (O 2 , CO, and NO)-binding kinetic studies were additionally conducted. Although no functional properties were examined, GsGCS is the only GCS reported to date that has a C-terminal transmembrane signal transduction domain.

Specific Characteristics of GCSs
Globin sensor domains that lack heme binding are known, although their functions have not been determined (45,46). However, given the accumulating information about the O 2 -mediated catalytic regulation and physicochemical properties of GCSs, it is worth summarizing specific characteristics of GCSs in comparison with those of non-GCS heme-based oxygen sensors. Note that the characteristics of non-GCS enzymes are not described here in detail; instead, see Table 1  GCS Active Sites Contain Distal Tyr and Thr Residues-On the basis of amino acid sequence alignments of GCSs and crystal structures of heme Fe(II) and Fe(III)-CN complexes of HemAT-Bs, it was suggested that the side chain phenolate of the Tyr residue (B10) is located in the heme distal side and is at or near the O 2 -binding site for most GCSs (Fig. 3). Although the crystal structure of the heme Fe(II)-O 2 complex of GCSs has not been determined, O 2 -binding kinetics and resonance Raman studies of heme Fe(II)-O 2 complexes of wild-type and mutant HemDGC, YddV, and AfGcHK proteins suggest that the Tyr residue could be located at or near the O 2 -binding site ( Table 1) (Table 1). Interestingly, although the protein folding of the GAF domain is similar to that of the PAS domain, the amino acid in the distal pocket adjacent to the ligand-binding site in the heme Fe(II)-O 2 complex is different between the GAF and PAS folds.
Thus, the O 2 -binding sites of heme-based oxygen sensors are different from those of Hb and Mb, where O 2 in the heme Fe(II)-O 2 complex interacts with a His residue (E7) in the distal side. For the oxygen sensors, it is probably important for the O 2 association/dissociation process to respond smoothly to O 2 concentration in the environment without being influenced by environmental pH or ionic strength. This is likely why Tyr, Thr, or Arg, but not the His imidazole, is used for the oxygen-binding site of these heme-based oxygen sensors.
Note that the His (E7) side chain preferentially stabilizes O 2 bound to the heme Fe(II) complex in all vertebrate Hbs and Mbs. However, distal Gln (E7) and Tyr (B10) side chains stabilize bound O 2 in many invertebrate Hbs, particularly those with extremely high O 2 affinities (73). In contrast, the hydroxyl group of Tyr (B10) can destabilize bound O 2 by the non-bond electrons or partial negative charge on the oxygen atom of the side chain in Cerebratulus lacteus Hb (99).
High O 2 Affinity-For the GCSs reported to date, the heme Fe(II)-O 2 complex is the active form, whereas the heme Fe(II) complex is the inactive form. The equilibrium constants for the dissociation of O 2 from the heme Fe(II) complexes of GCSs are low (between 0.077 and 14 M) as shown in Table 1; thus, the O 2 -binding affinity for the heme Fe(II) complex is very high. These equilibrium dissociation constant values are in contrast to those of heme-based oxygen sensors containing the hemebound PAS fold, such as FixL and EcDOS, where the corresponding values are 140 -340 M. However, the value for DevS (0.58 M) is in the same range as that for GCSs. This suggests that GCSs, as well as DevS/DosT, act under anaerobic or semi-anaerobic (micro-aerobic) conditions and need to sense changes in the concentration of trace amounts of environmen-tal O 2 , whereas FixL and EcDOS operate under aerobic conditions and sense a decrease in O 2 concentrations from normal levels. Because the O 2 -binding affinity of YddV (14 M) is higher than that of other GCSs, it may be that E. coli YddV must work under semi-anaerobic, but not strictly anaerobic, conditions in the large intestine. If it is assumed that only YddV and EcDOS are involved in homeostasis of c-diGMP in E. coli in specific organs (Fig. 2), the two enzymes would function distinctly and synergistically, depending on the O 2 concentration, to regulate the c-diGMP level in response to physiological stimuli.
It should be noted that Mb and Hb function by taking up and releasing O 2 for transport and delivery to respiring mitochondria. PAS sensors discern the presence of O 2 and activate genes related to aerobic metabolism, and GCS sensors respond to hypoxia and activate the associated genes. Therefore, the O 2 affinities for these heme proteins differ to accommodate their functional purposes.

Specific Characteristics of Heme-based Oxygen Sensors, Including Non-GCSs
The increasing availability of information about the biochemical and physicochemical characteristics of heme-based oxygen sensors, including non-GCSs, has made it worthwhile to compare the specific properties of these sensors with those of other heme proteins, such as Hb and Mb ( Table 1).

The Heme Fe(II)-O 2 Complex Is Stable with a Low Autoxidation Rate Constant-
The autoxidation rate of heme proteins is controlled by several factors (5,6). The hydrogen bond provided by the neutral imidazole side chain of the distal His residue (E7) plays the most crucial role in the inhibition of Mb heme iron autoxidation. This interaction prevents both dissociation and protonation of bound oxygen. The pH dependence of the rate observed at or above 7.0 may be attributable to protonation of the heme Fe(II)-O 2 complex. In addition, the superoxide (the product of autoxidation of the heme Fe(II)-O 2 complex) dissociation pathway predominantly determines the rate. Moreover, the accessibility of the distal pocket to solvent water molecules increases protonation of the heme Fe(II)-O 2 complex, resulting in an increased rate.
The importance of the heme Fe(II)-O 2 complex in catalytic enhancement or suppression of oxygen sensors is evidenced by the stability of the heme Fe(II)-O 2 complex. In fact, the stability of the heme Fe(II)-O 2 complex of GCSs is such that the spectrum of the heme Fe(II)-O 2 complex does not change for hours to days at room temperature. This stability is reflected in the very low GCS autoxidation rate constants, which are Ͻ0.025 min Ϫ1 (Table 1). It is reasonable to infer that the heme Fe(II)-O 2 complexes of other oxygen sensors, such as FixL and EcDOS (which contain the heme-bound PAS domain) and DevS (which contains the heme-bound GAF domain), are similarly stable, given that the heme Fe(II)-O 2 complex is the key form of their functions. This stability of the heme-based oxygen sensor is in contrast to that of other heme proteins, such as cytochrome P450, where the O 2 molecule is activated in the heme active site; thus, the heme Fe(II)-O 2 complex is not as stable as that in heme-based oxygen sensors. The hydrogen bond from Tyr, Arg, or Thr plays important roles in stabilizing the heme Fe(II)-O 2 complex in heme-based oxygen sensors, similar to that from the distal His residue (E7) in Hb and Mb (Table 1).
Redox potential would also be expected to contribute to the stability of the heme Fe(II)-O 2 complex and autoxidation rate constant. However, currently available data indicate that this prediction is not always borne out for heme-based oxygen sensors. For example, redox potential values of Ϫ17 and 45 mV versus the standard hydrogen electrode for YddV (32) and EcDOS (68), respectively, do not strictly correspond with the associated stability and autoxidation rate constants (YddV, 0.0076 min Ϫ1 ; and EcDOS, 0.005 min Ϫ1 ) of these heme proteins.
NO and CO Exhibit Divergent Functional Regulation, but Selectivity among O 2 , NO, and CO Is Determined by a "Sliding Scale Rule"-Other diatomic ligands, such as NO and CO, bind to GCS as well, and it is likely that these are more important during early evolution than O 2 (14 -16). As summarized in Table 1, these gas molecules play divergent roles in the regulation of catalytic function. For example, addition of NO to the heme Fe(II) complex inhibits the catalytic activity of YddV and HemDGC, whereas the same treatment enhances BpeGReg activity. Similarly, addition of CO to the heme Fe(II) complex of YddV, BpeGReg, and AfGcHK stimulates the catalytic functions of these enzymes but has no effect on HemDGC. Liganddependent functional regulation is similar for non-GCS oxygen sensors. Specifically, FixL, EcDOS, DevS, and DosT accept NO and CO in addition to O 2 , although the catalytic activities of the ligand-bound complexes are variable (Table 1).
The sliding scale rule described by Tsai et al. (92,93) provides a thorough quantitative analysis of how O 2 , NO, and CO sensors selectively bind specific diatomic ligands based on combined graphical analysis of ligand-binding data for libraries of heme sensors, globins, and a model heme. According to this rule, linear logarithm plots of K d(O2) , K d(CO) , and K d(NO) values for heme proteins with a conserved neutral proximal histidine reveal marked ligand discrimination, with the relative affinities consistently in the order O 2 Ͻ Ͻ CO Ͻ Ͻ NO, indicating characteristic dependence of affinity on ligand type. It has been suggested that factors such as 1) preferential stabilization of bound O 2 via hydrogen bonding, 2) destabilization of the heme Fe(II)-O 2 complex by negative electrostatic fields, 3) direct steric inhibition of ligand binding, and 4) proximal coordination geometry are crucial in determining the affinities of diatomic ligands for heme proteins. These factors, in combination with the intrinsic chemical differences between ligands, lead to a remarkably wide range of ligand affinities and account for why gas sensor, storage, and transport heme proteins maintain the neutral form of the proximal imidazole.
The Heme Iron Complex Serves to Suppress Catalysis Because Heme-free Forms Have Sufficient Catalytic Activities-Why the heme iron complex is bound to heme-based oxygen sensors is an interesting question. Contrary to expectations, heme-free forms and heme-binding domain-truncated forms of YddV, AfGcHK, and EcDOS have sufficient catalytic activity (32,44,72,94). Instead of stimulating catalysis, the heme iron complex on the heme-based oxygen sensor serves to suppress catalysis, and oxygen association/dissociation releases this catalytic suppression.
Full-length Enzymes Are Dimers or Tetramers-The DGC-GCSs identified and characterized to date, such as HemAT-Bs (22), AvGReg (35), BpeGReg (36), HemDGC (37), and YddV (32,38,39), are dimers or tetramers. Such an arrangement suggests that each subunit catalyzes the cyclization reaction possibly in synergy with other subunits in the dimer/tetramer. The autophosphorylation of AfGcHK would be expected to occur in a crosswise manner.

Summary
GCSs appear to represent prototypical heme-based oxygen sensors with a basic similarity to Hb and Mb, with which they share a common genetic ancestry. O 2 binding to the heme iron complex in GCSs significantly enhances associated catalytic activity, including DGC and HK activities, and MCP function; thus, the heme Fe(II)-O 2 complex in GCSs is the physiologically relevant form, similar to the case for Hb and Mb. This characteristic contrasts with that of non-GCS oxygen sensors, where either O 2 dissociation from or association with the heme iron complex stimulates catalysis. The high O 2 affinity for the heme Fe(II) complex, the stability of the heme Fe(II)-O 2 complex, and the role of the Tyr residue (B10) of the distal site in the stability of the O 2 complex are essential contributors to the specific characteristics of GCSs. The sliding scale rule accounts for ligand discrimination of GCS, as well as other gas sensors (92,93).