Heme Sensor Proteins*

Heme is a prosthetic group best known for roles in oxygen transport, oxidative catalysis, and respiratory electron transport. Recent years have seen the roles of heme extended to sensors of gases such as O2 and NO and cell redox state, and as mediators of cellular responses to changes in intracellular levels of these gases. The importance of heme is further evident from identification of proteins that bind heme reversibly, using it as a signal, e.g. to regulate gene expression in circadian rhythm pathways and control heme synthesis itself. In this minireview, we explore the current knowledge of the diverse roles of heme sensor proteins.


Heme-responsive Proteins: Defining Features
Heme homeostasis is crucial. Free heme at Ͼ1 M is cytotoxic, mainly by producing reactive oxygen species. Iron intake accounts for only a small proportion of mammalian requirements, so iron recycling (particularly from heme) is critical (1). Tight regulation of heme synthesis/breakdown is needed. Heme regulates its own fate and controls several other biological processes. Heme-responsive proteins elicit cellular responses by binding/debinding heme or via changes in heme ligation (e.g. by gases) or redox state. They can be divided into two classes: nuclear receptor (NR) 2 hemoproteins and heme sensors with no NR function. Each heme-responsive protein has a heme regulatory motif (HRM), usually containing a CP motif (2). The Cys residue of the CP motif is an axial heme ligand. No other residues are conserved across the protein class. The wider relevance of the CP motif is unclear because other hemoproteins (e.g. prostaglandin E 2 synthase, chloroperoxidase, and some cytochromes P450) also have a CP motif. The properties of members of the two main classes are described below.

NR Hemoproteins
Several heme-binding proteins occur in the NR superfamily, which is the largest transcription factor superfamily (summarized in Table 1). NRs bind specific DNA motifs in response to small molecule signaling. They generally share conserved domain architecture, with a DNA-binding region containing two zinc fingers, a ligand-binding domain, and activation domains (3). Usually, ligand binding induces conformational changes that dissociate partner proteins, allowing DNA binding and/or changes to dimerization status or cellular localization that enable the protein to elicit a response. A subfamily of NRs binds heme at their ligand-binding domain and so act as heme sensors, as discussed below.

Circadian Rhythm Heme Sensors
Circadian rhythms are regulated by feedback loops at transcriptional/translational levels. Heme is critical in this process (4). In vertebrates, the expression of several day/night cycle genes, as well as Rev-erb␣, is controlled by binding a heterodimer of Clock (or NPAS2 (neuronal PAS domain protein 2)) and Bmal1 to their 5Ј-UTR, thus switching on transcription. Expression of Bmal1 and NPAS2/Clock is repressed in turn by binding of Rev-erb␣ to a homodimeric partner (5). Rev-erb␣ and homologs (Rev-erb␤ and E75, with the latter involved in insect ecdysone signaling) are heme-regulated NRs (6,7). When heme-bound, partner interactions (with NR corepressors) are favored, and target gene transcription is repressed. Heme-free Rev-erb␣/␤ does not bind partners, and transcription proceeds (7). The resting state heme of human Rev-erb␤ is ferric and six-coordinate, with axial ligation from His-568 and Cys-384 in a CP motif (8). In ferrous Rev-erb␤, a major fivecoordinate high-spin form with a His axial ligand and a minor low-spin six-coordinate form are seen (8). Both the phase and period of circadian cycles are thought to be controlled by NO/CO levels (9). NO/CO displaces the Cys ligand in ferrous Rev-erb␤ to form His-Fe 2ϩ -NO/CO species (8). The ability of NO-bound Rev-erb␤ to recruit corepressors is reduced, decreasing transcriptional repression (10). A protein thiol/disulfide redox switch likely regulates Rev-erb␤ heme binding. In the reduced (dithiol) protein state, Rev-erb␤ retains the Cys-384 ligand with a 5-fold lower K d for ferric heme than the oxidized form, where Cys-384 and Cys-374 form a disulfide, and an unknown neutral ligand replaces Cys-384. The K d of reduced Rev-erb␤ for ferric (and ferrous) heme is ϳ20 nM, similar to the intracellular heme concentration and consistent with its heme sensor role (11).
Covalent heme binding to protein occurs in insect E75 isoforms. The inability to dissociate heme suggests a gas sensor/ redox role, and target gene transcription is induced by binding NO and CO to E75 heme (12). The transcription factor (TF) NPAS2 is also heme-regulated. It has two PAS (Per-Arnt-Sim) domains, common in signaling proteins, each of which can bind one heme and sense CO, leading to dissociation of its DNAbinding complex with Bmal1, leaving nonproductive Bmal1 homodimers (13). Spectroscopic analysis of its PAS-A domain suggests bis-His-ligated heme, with CO replacing His in the ferrous state to enable signal transduction (14). The pathways of the circadian rhythm proteins are summarized in Fig. 1.
The photosynthetic bacterium Rhodobacter sphaeroides utilizes PpsR, a heme and redox sensor protein that binds heme at a HRM as part of its light/dark cycle (15). PpsR has two redoxactive cysteines and, under oxidizing conditions, binds promoters to block tetrapyrrole gene transcription, preventing heme and bacteriochlorophyll synthesis. Cys and His residues were proposed as PpsR axial ligands (15). Heme-bound PpsR does not bind DNA to inhibit transcription of light-harvesting II peptides and bacteriochlorophyll synthesis genes, but it can bind and repress heme synthesis genes, preventing heme accumulation (15). Under oxidizing conditions, the heme Cys ligand instead forms a disulfide to stimulate binding to and repression of target genes by PpsR apoprotein. PpsR is regulated by binding the blue light sensor flavoprotein AppA, an antirepressor that is reduced as O 2 tension decreases. AppA may also bind heme in a domain indispensable for normal gene regulation (16). PpsR has a CI rather than a CP motif. This novel HRM is not in its PAS domain, as in most other heme-sensing TFs, but is in the C-terminal helix-turn-helix, with a proposed His sixth axial ligand coming from one of the two PAS domains (15).

Iron, Heme, and Oxidative Stress Response
Iron regulatory proteins (IRPs) control mammalian cell iron levels. Under low iron conditions, IRPs bind iron-responsive elements (IREs; 28-nucleotide stem-loop regions of mRNA in 3Ј-or 5Ј-UTRs of genes important in iron homeostasis) and prevent mRNA processing/translation. IRP binding upstream of the gene prevents its translation, whereas binding at the 3Ј-end stabilizes mRNA by preventing nuclease attack (17). Mitochondrial aconitase, ALAS2, and ferroportin I genes all have IREs.
Two IRPs are known and are differentially regulated by cell iron concentration. IRP1 binds a [4Fe-4S] cluster at high iron concentrations. This is lost under low iron conditions, allowing binding to IREs. Under high iron conditions, IRP2 is ubiquitinated in the C-terminal region and degraded (18,19). IRP2 binds a Cys-ligated heme in a HRM, and heme binding underpins its iron-sensing mechanism (19). Heme may bind only to a truncated IRP2 that is specifically proteolyzed as part of the regulatory mechanism (20). However, another proposal is that C-terminal cysteines bind iron directly and that heme has no role in IRP2 degradation (21).
Irr (iron response regulator) in ␣-proteobacteria functions similarly to IRP2. Heme-free Irr binds iron control elements 5Ј to the regulated gene. In Bradyrhizobium japonicum, Irr has a HRM that binds ferric heme, likely Cys-ligated. Heme-bound Irr is unstable and rapidly degraded, triggering transcription of several heme biosynthesis genes. A second His-ligated ferrous heme was also postulated in Irr (22,23).
Saccharomyces cerevisiae Hap1 activates transcription of numerous genes in response to oxidative stress. It has seven HRMs, six in its heme domain and one in the HRM7 domain. Heme-free Hap1 binds several proteins to form a large complex. Heme-bound Hap1 binds DNA and activates transcription. The single HRM7 CP motif is needed for heme binding. Deletions of other CP motifs do not abolish Hap1 heme activation (24,25).

Bach1
Heme homeostasis is partly controlled at the level of transcription of heme-metabolizing genes. Genes such as the oxidative stress-responsive ho-1 and erythroid-specific ALAS-E (ALAS2) genes contain Maf recognition elements that bind Maf TFs, which switch from gene repressor to activator depending on their partner in a heterodimeric complex. Maf partners include the transcriptional repressor leucine zipper proteins Bach1 and Bach2 (26). Bach1 has six CP motifs and binds heme (27). No single CP motif is essential, but four motifs in the C-terminal leucine zipper region are important for heme binding. Removing all CP motifs prevents heme binding (27). Heme binding to Bach1 does not cause Maf dissociation but displaces the heterodimer from the Maf recognition element-binding site, allowing binding of a different partner to activate transcription. Heme binding targets Bach1 for ubiquitination and degradation (28).

Non-NR Heme Sensors
The proteins discussed below typically do not bind DNA/ RNA but instead have heme sensor functions.

TABLE 1 Summary of key proteins with known or proposed heme sensor or heme-based gas sensor functions
The functions of the proteins listed are detailed in text. For IRP2, regulation by iron (rather than heme) was proposed (21). HRM-mediated heme binding also remains controversial for HO-2 (33). The role of heme in regulating ALAS1 and ALAS2 import into mitochondria is also disputed (36,37). Heme inhibits potassium efflux from BK channels. However, how heme binds and whether CO stimulates K ϩ efflux in a heme-dependent or heme-independent manner has still to be established. A cytochrome c-type CXXCH motif is present rather than the typical HRM (45)(46)(47)(48). Further studies are also required to establish the physiological relevance of the apparent heme-mediated inhibition of the arginine transferase activity of R-transferase and of putative heme binding to stanniocalcin glycoprotein hormones (52)(53)(54). For DGCR8, heme may have a structural role, although a heme redox state dependence on RNA binding was proposed (50,51).

HRM Function
Ref.

Heme Homeostasis
Heme synthesis/breakdown is directly regulated by heme. Heme oxygenases HO-1 and HO-2 oxidatively degrade heme to biliverdin using three molecules of O 2 and yielding one molecule of CO and ferrous iron and three H 2 O molecules (29). CO is a signaling molecule in several pathways, and HO is the main CO source in mammals. The HOs are 55% identical, but HO-2 has three HRMs with CP motifs, whereas HO-1 has none (30). Incubation of HO-2 with heme yielded a protein with three hemes, two binding at HRMs and one at the active site of HO-2. However, no change in activity or heme binding occurred when CP motif cysteines were mutated to Ala (31), and His-45 ligates the substrate heme for oxidation (30). In normoxia, a disulfide forms between Cys-265 and Cys-282 of two HRMs, with a small increase in active site heme affinity (32). Disulfide reduction to dithiol may lower active site heme affinity and increase the cellular heme pool. CP motifs may thus be nonessential for HO-2 heme binding and instead have signaling or molecular interaction functions (31)(32)(33).
The ALASs catalyze pyridoxal phosphate-dependent condensation of glycine and succinyl-CoA, forming ␦-aminolevulinic acid in the first heme synthesis step. There are two mammalian forms: ALAS2 (needed for hemoglobin synthesis) and ubiquitously expressed ALAS1 (maintains basal heme levels) (34). Large precursor ALAS proteins (pre-ALAS) are transported to mitochondria and cleaved to form mature enzymes. Several mechanisms control ALAS function, e.g. binding of IRP2 (under iron-deficient and/or hypoxic conditions) to an IRE in the 5Ј-UTR of ALAS2 mRNA inhibits its translation (35). ALAS1 and ALAS2 have three HRMs with CP motifs, two (HRM1 and HRM2) in the presequence and one (HRM3) in each mature enzyme. Hemin inhibits mouse ALAS2 transport into mouse mitochondria, but mutating both presequence CP motif cysteines to serine eliminated the hemin inhibition (36). However, studies with ALAS-transfected quail fibroblasts showed that rat ALAS2 mitochondrial import was not affected by heme. However, ALAS1 import was inhibited by heme, and studies on Cys-to-Ser mutant CP variants showed that CP motifs in HRM1 and HRM3 are important in this heme-mediated inhibition (37).

Cellular Homeostasis
Phosphorylation of eIF2 at its ␣ subunit is a cellular stress response mechanism. eIF2 aids recruitment of Met initiator tRNA to the 40 S ribosome. The eIF2␣-bound GTP is hydrolyzed to GDP prior to translational initiation. Phosphorylation of eIF2 inhibits GDP release, preventing eIF2 from binding a new GTP and reactivating (38). Different eIF2␣ kinases respond to distinct stress signals, e.g. low amino acid levels, UV stress, or viral infection. The heme-regulated inhibitor (HRI; eIF2␣ kinase 1) phosphorylates eIF2␣ during heme deprivation or oxidative stress (39). In heme-replete cells, HRI binds heme with Cys/His coordination and is inactive (40). At low heme levels, the heme dissociates, and HRI autophosphorylates and then phosphorylates eIF2 (41). NO may also regulate hemebound HRI, potentially forming a five-coordinate NO-bound ferrous heme to activate eIF2␣ phosphorylation (42). BK (big potassium or large conductance Ca 2ϩ -activated K ϩ ) channels are K ϩ -selective and regulated by voltage and calcium. All known BK channels have a CXXCH motif (typical of c-type cytochromes that bind heme covalently) and bind heme to inhibit K ϩ efflux under hypoxic conditions independent of Ca 2ϩ concentration (43). CO from HO-2 activates BK channels, putatively by binding ferrous heme iron to allow K ϩ efflux (44). Studies on a peptide containing the CXXCH motif (which binds heme by His coordination) led to a conclusion that heme is the CO-binding site (45). However, mutating the CXXCH motif His did not diminish the effect of CO on channel permeability (despite heme loss), possibly because CO might also interact with an unidentified metal cluster (46,47). It was also proposed that BK channels do not bind heme but that CXXCH motif cysteines form a thiol/disulfide redox switch, similar to that proposed for Rev-erb␤ (48). The significance of the BK channel CXXCH motif thus remains to be established. DGCR8 (DiGeorge critical region 8) protein is essential for cleavage of long primary microRNAs (miRNAs) into short pre- In other organs, the NPAS2 paralog Clock fulfills this function. The PER and CRY proteins can also self-regulate, and their binding to the NPAS2-Bmal1 heterodimer causes dissociation from DNA and switches off their transcription (C). In addition, binding of heme and CO to NPAS2 causes dissociation of the heterodimer, switching off PER/CRY transcription (D). NPAS2-Bmal1 expression is also regulated in a heme-dependent manner (shown for NPAS2 here). In a heme-bound state, Rev-erb␣ recruits heterodimeric NR corepressor (NCoR) partners to repress transcription. When heme-free, it cannot repress transcription (E).
cursor miRNA hairpins ready for further processing to form mature miRNA. Primary miRNA cleavage is catalyzed by the RNase Drosha, which forms the "microprocessor" complex with its DGCR8 partner. DGCR8 binds heme with a PC (rather than CP) motif (49). Spectroscopy suggested that DGCR8 has two Cys axial ligands, a unique hemoprotein coordination state (50). The role of DGCR8 is not resolved, but RNA may bind ferric (not ferrous) hemoprotein (51). Heme may enable DGCR8 homodimer formation, with a Cys from each polypeptide as axial ligands (50).
There are two other examples. 1) R-transferase is an Arg-tRNA-protein transferase that links an Arg to the N terminus of a protein, targeting its breakdown by the N-end rule pathway. It has two HRMs with CP motifs. Heme binding to mouse/yeast R-transferase inhibits arginylation and induces formation of a disulfide (Cys-71 and Cys-72 in mice) in one HRM (52). 2) Stanniocalcins are glycoprotein hormones with roles in vertebrate calcium uptake. STC1 was suggested to contain a HRM and bind heme from studies with a 10-amino acid peptide of the postulated heme-binding region (53). Heme binding to intact STC1 was not shown. Heme binding was proposed to occur at a CS (not CP) motif. However, it was also reported that this Cys forms a disulfide and does not bind heme (53,54) and that STC1 may operate a thiol/disulfide redox switch, as described above (54). HRM-containing heme sensors are summarized in Table 1.

Gas Sensor Proteins
Many gas-binding hemoproteins are known. Oxygen transporters (e.g. hemoglobin) and O 2 reductases (e.g. cytochrome c oxidase and P450 cytochromes) are "poisoned" by CO and NO binding to ferrous and ferric/ferrous hemes, respectively. NOS enzymes make NO as a cellular effector (55). CO is produced endogenously by HOs. Both gases have regulatory roles, many mediated by binding hemoproteins. Signaling is often achieved by structural perturbation following heme iron ligation and transmission of effects by altered activity (e.g. DNA binding or phosphorylation) in an adjacent domain. Thus, free energy changes upon gas binding/debinding are transformed into conformational changes to trigger alterations in activity and signaling induction. General features of the burgeoning class of heme gas sensors are given below (56).

CooA
The purple proteobacterium Rhodospirillum rubrum uses CO as an energy source under anaerobic dark growth conditions. CooA regulates the pathway, binding CO to activate expression of genes for oxidation of CO to CO 2 and reduction of protons to H 2 (57). CooA is a homodimeric heme-binding cAMP receptor protein (CRP) transcriptional regulator family member that binds CO cooperatively (58). Inactive ferric CooA has a Cys-75 thiolate heme ligand, which switches to His-77 in ferrous CooA. In the active CO-bound form, His-77 coordination is retained (59). The reduced CooA dimer structure confirmed His-77 as the axial ligand, revealing the N-terminal proline (Pro-2) of the opposite monomer as the heme sixth ligand in both units (60). CO displaces Pro-2 to activate CooA (supplemental Fig. S1). CooA resembles Escherichia coli CRP in structure, but reorientation of CooA DNA-binding domains is needed to produce a transcriptionally active state (61). CO binding may accompany a shift in heme position to a hydrophobic cavity and movement of the C-helix of CooA toward the opposite heme to restructure the CO-binding pocket. This switch reorganizes a hinge between the C-and D-helices to enable DNA-binding domains to interact with DNA (61). Carboxydothermus hydrogenoformans CooA structures support this model, showing a semi-apo state of the CooA dimer, with heme-bound monomer in a CO-bound form revealing a heme and C-helix displacement (62). An imidazole-"sensing" CooA mutant enabled distal imidazole ligation to both Cys-75 (Fe 3ϩ ) and His-77 (Fe 2ϩ ) proximally coordinated forms. Only the Fe 2ϩ -imidazole adduct was transcriptionally active, consistent with the importance of the heme Fe 2ϩ -dependent ligand switch and heme repositioning in activating CooA (63).  (68). The five-coordinate (deoxy) FixL is "active" and directs induction of genes encoding high O 2 affinity terminal oxidases for microaerobic respiration (the likely main role of FixL in non-N 2 -fixing bacteria/archaea) and for nitrogenase subunits (S. meliloti) or denitrification enzymes (B. japonicum) (67). Activation relates to His kinase activity, and the O 2 -bound form is inactive. The ␥-phosphate transfer from ATP (and O 2 sensing) occurs in the FixL-FixJ dimer complex (69). Conformational change in the five-coordinate state enables autophosphorylation of a conserved His in each FixL that precedes phosphate transfer to a target Asp on each FixJ to activate transcription (Fig. 2) (67). FixL heme is ferrous, but oxidation to the ferric form either does not affect FixJ phosphorylation rate (B. japonicum) or inhibits it by 100-fold (S. meliloti) (70,71). This likely relates to conformational differences that impact on phosphate transfer efficiency. The high-spin ferrous heme iron in deoxy-FixL is active, but O 2 binding shifts it to low spin, alters its position relative to the heme plane, and induces conformational changes, including reorientation of a conserved arginine (Arg-220

Soluble Guanylate Cyclase
cGMP is made from GTP by guanylate cyclases (GCs) in response to Ca 2ϩ levels (73). cGMP activates kinases, ion channels, and phosphodiesterases (PDEs) (73). Mammalian soluble GCs (sGCs) are NO-binding hemoproteins (55). The sGCs are heterodimers of ␣/␤ subunits found in most tissues. The ␣/␤ subunits are homologous, and two isoforms of each are known (73). The most common sGC combination is ␣ 1 /␤ 1 , but ␣ 2 /␤ 1 is highly expressed in some tissues (e.g. brain). The ␤ 2 subunit has not been well studied but may be the only sGC form functional as a homodimer (74). The ␣/␤ subunits are different sizes (e.g. 690/619 amino acids for rat ␣ 1 /␤ 1 ) with four component domains (73). The minimal heme-binding region is in the first ϳ200 residues of ␤ 1 . The domain is a H-NOX (heme-nitric oxide/oxygen binding) family member (73). This is followed by PAS and coiled-coil domains, both likely involved in sGC dimerization. The C-terminal domains are catalytic subunits that dimerize to enable cGMP formation, although the ␣ 1 N-terminal region may also be important in dimerization (75,76). Bovine sGC binds ϳ1 heme/heterodimer, and studies of N-terminally truncated human ␣ 1 indicated that the heme binding and NO sensitivity of the ␣ 1 /␤ 1 subunit complex were unaffected (77). The bovine sGC ␤ 1 H105F mutant produced a non-NO-responsive heme-free heterodimer, revealing His-105 as a heme ligand. A His residue is absent in the corresponding part of the ␣ 1 domain, but this domain may still have heme affinity (73,78,79). Insights into sGC heme domain structure come from prokaryotic H-NOX proteins (Fig. 3) (73). The Thermoanaerobacter tengcongensis H-NOX structure revealed His-102 as a heme ligand (80). Structural/spectroscopic data for sGC ␣ 1 /␤ 1 and H-NOX proteins indicate a distorted heme occupying different conformations (73). Binding of chemical activators of sGC ␣ 1 /␤ 1 leads to a more planar heme, but it is unclear whether this induces or results from enzyme activation (81). The sGC ␤ 1 iron-His bond breaks rapidly on binding NO, FIGURE 2. FixL mechanism. The upper panel shows the FixL mechanism in the presence of oxygen, whereas the lower panel shows the mechanism under anaerobic conditions. FixL is shown as a red square, with the heme iron represented as a dot in the center with lines to indicate the tetrapyrrole ring. The iron is shown in purple for the high-spin ferrous state (for deoxy-FixL) and in red for the low-spin ferrous state (when bound to oxygen). In the absence of oxygen (lower panel), the His-ligated FixL dimer associates with a FixJ dimer (blue circles) to form a complex that can bind ATP. The FixL dimer can also bind ATP in the absence of FixJ but with much lower affinity. FixJ preferentially exists as a heterodimeric complex with FixL when it is non-phosphorylated. Once ATP is bound, a conserved FixL histidine residue is phosphorylated, releasing ADP. The phosphate is then transferred to a conserved aspartate residue on FixJ and stabilizes the FixJ dimer. FixJ undergoes a conformational change, and the FixL-FixJ complex dissociates. Free phosphorylated FixJ then activates transcription of target genes. Binding of O 2 (or to a lesser extent, CO or NO) causes a conformational change in FixL and switches the heme iron spin state from low spin to high spin (upper panel). CO and NO bind FixL more tightly than O 2 but are less effective inhibitors of kinase activity. In the O 2 -bound state, association with FixJ and binding of ATP are unaffected, but FixL cannot be phosphorylated, and therefore FixJ cannot be switched on as a transcriptional activator. Oxygen dissociation reactivates the system (67)(68)(69)(70)(71). and structures of a H103G mutant and the Fe 2ϩ -CO complex of WT Shewanella oneidensis H-NOX show that large changes in both heme and H-NOX conformation accompany His ligand cleavage. Precisely how ligand-dependent events regulate activities of the associated His kinase (in S. oneidensis H-NOX) or GC domains is unclear (73,82). However, sGC Fe 2ϩ -NO and (less effectively) Fe 2ϩ -CO complexes stimulate cGMP synthesis. O 2 is disfavored as a ligand to ferrous iron despite its higher cellular concentration. A lack of H-bond-donating residues (as found in oxygen-binding globins) resulting in faster O 2 dissociation rates in sGC/H-NOX is a potential explanation for weak O 2 binding (73,83). Steric constraints are likely also important (84). This discrimination allows sGC to respond efficiently to Ca 2ϩ -stimulated NO production. Mg 2ϩ -GTP or Mg 2ϩ /cGMP/ PP i accelerates NO-bound ferrous sGC ␤ 1 iron-His bond cleavage, whereas ATP inhibits formation of a five-coordinate Fe 2ϩ -NO complex (85). Elevated NO levels also stimulate sGC. A second NO binds to the proximal (opposite) side of the heme, transiently forming bis-NO sGC. The distal NO then dissociates, leaving a more active sGC with NO bound at the heme proximal face (Fig. 3)

DosS and DosT
DosS and DosT (with response regulator DosR) regulate the transition of Mycobacterium tuberculosis from a replicating form to a "persistent" latent state that is unresponsive to antibiotics (87). Understanding how M. tuberculosis switches in and out of latency is crucial for development of effective drugs. Conditions associated with M. tuberculosis latency are decreased [O 2 ] and/or increased [NO] (88). The Dos gene regulon is induced under such conditions, controlled by the twocomponent regulators dosS/dosR and dosT/dosR (89). Dos regulon induction is essential for M. tuberculosis survival/recovery on exposure to hypoxia, NO, and CO, conditions faced on host infection and immune response activation (87). DosS/ DosT are autokinases that phosphorylate a conserved His and then transfer phosphate to DosR Asp-45 to up-regulate target genes (90). DosS and DosT have a similar architecture, with two N-terminal GAF (cGMP-specific phosphodiesterases, adenylyl cyclases and FhlA) domains. Heme binds only to GAF-A. The topology of GAF is similar to that of PAS domains, but the Mycobacterium smegmatis DosS GAF-B structure shows alterations (relative to other GAF domains) that suggest that cyclic nucleotide binding is unlikely and that it may instead interact with GAF-A to help structure its heme-binding pocket (91). His-149 is the M. tuberculosis DosS GAF-A heme ligand, but heme still binds DosS in a non-ligated form, unless hemin is omitted from growth medium in DosS-expressing cells (92,93). The M. tuberculosis DosT structure confirms the invariant His- The heme tetrapyrrole ring is represented as a square with the ferrous (Fe 2ϩ ) iron bound. In the resting state, the heme has His coordination with no sixth ligand (A). This form has only very low activity (red circle). Binding of CO as the sixth ligand yields a form with a 4-fold increase in activity (B; dotted green circle). At stoichiometric NO levels, distal NO binding first produces a six-coordinate species (C). The proximal His then dissociates, yielding a five-coordinate distally NO-bound form with low GC activity (D; solid green circle) (73). Alternatively, in more rapid reactions in the presence of excess NO, a second NO molecule binds at the proximal side of the heme, displacing the His (E). The distal NO then dissociates, leaving a high GC activity five-coordinate proximally NO-bound form (F; double green circle) (86). The low (D) and high (F) activity five-coordinate NO-bound forms can be distinguished by EPR spectroscopy. Preincubation of sGC substrate (Mg 2ϩ -GTP) or products (Mg 2ϩ /cGMP/PP i ) with sGC at stoichiometric NO concentrations may also lead to the high activity form (curved arrow), although ATP competes with GTP and can instead lead to formation of a low activity form (likely species D) (73). The low activity NO-bound form (D) may be less stable and more prone to deactivation in a process that may not involve dissociation of the distal NO ligand (73,86). A distinct "desensitization" of sGC to repeated exposure to NO may result from nitrosation of a sGC protein thiol (73). The sGC allosteric stimulator YC-1 (5-(1-(phenylmethyl)-1H-indazol-3-yl)-2-furanmethanol) may convert the low activity form (D) to the high activity state (F) and also substantially stimulates the sGC activity of the CO-bound form (B). YC-1 was also reported to decrease the rate of sGC deactivation, despite enhancing the NO dissociation rate (73). Other such allosteric stimulators of sGC are also known (e.g. BAY 41-2272 (3-(4-amino-5-cyclopropylpyrimidine-2-yl)-1-(2-fluorobenzyl)-1H-pyrazolo [3,4-b]pyridine)) (73,85). DosT O 2 complex. In DosS, Tyr-167 provides a H-bond to the sixth ligand H 2 O molecule (92). DosS ferrous heme iron is rapidly autoxidized to the met (ferric) state, whereas DosT forms a stable Fe 2ϩ -O 2 complex. Autokinase activity is increased in ferrous DosS and in deoxy-DosT, suggesting that DosS is a redox sensor and that DosT is a hypoxia sensor, with activities modulated by O 2 , NO, and CO in vitro (94). However, the in vivo situation is more complex. Differential expression of dosT (constitutive in anaerobic dormancy and in aerobic growth) and dosS (induced by hypoxia, CO, and NO) suggests that DosT induces the Dos regulon in response to hypoxia, and then DosS (induced by the action of DosT) continues to induce the regulon once DosT is inactivated (87). Ascorbate (a cytochrome c reductant) and menaquinone induce the Dos regulon under aerobic conditions, suggesting links to the M. tuberculosis respiratory chain (87).

Other Gas Sensors
Another sensor hemoprotein is human cystathionine ␤-synthase (CBS). CBS is critical to sulfur metabolism, catalyzing pyridoxal phosphate-dependent condensation of homocysteine and serine to make cystathionine, which is hydrolyzed by cystathionine ␥-lyase form cysteine (95). CBS binds heme in an N-terminal domain with His-65/Cys-52 ligands. CO displaces Cys-52, forming a six-coordinate ferrous complex, whereas NO forms a five-coordinate complex and dissociates both heme axial ligands (96). CO inhibits CBS completely, although the mechanism of signal transduction and the physiological role are unclear (96). However, CBS and cystathionine ␥-lyase can produce H 2 S in non-classical reaction pathways, and CO-mediated inhibition of H 2 S release by CBS might improve bile output by stimulating HCO 3 Ϫ excretion through H 2 S-sensitive channels (97). Another key example is the E. coli direct oxygen sensor. E. coli Dos binds heme to the first of two PAS domains in the N-terminal region of the protein. E. coli Dos has both cAMP and cyclic di-GMP PDE activity, in the former case, producing 5Ј-AMP and preventing cAMP from interacting with its receptor CRP. It is a ferrous protein with His-77/Met-95 heme ligands. Met-95 is displaced by O 2 (and, for example, NO and CO), resulting in enhanced PDE activity mediated by the C-terminal region, which contains GGDEF and EAL subdomains (98) associated with cyclic di-GMP synthesis and PDE activities, respectively, and named after amino acid residues associated with activities of such domains (56). The GGDEF subdomain is likely inactive because it lacks key residues needed for diguanylate cyclase activity. Arg-97 H-bonds to heme-bound O 2 , stabilizing the oxyferrous form by lowering its autoxidation rate (99). E. coli Dos was also suggested to be a redox sensor because the ferric state (readily formed in vitro with H 2 O replacing the Met-95 ligand) has diminished PDE activity with cAMP (98). However, the reduced cytoplasm of E. coli and stabilizing Arg-97/O 2 interactions probably maintain the oxyferrous form in vivo. A further example is RcoM-2 from Burkholderia xenovorans, which undergoes a redox-dependent heme ligation switch (Cys to Met) and is implicated in the transcriptional response to CO (100).

Summary
The recognition of novel physiological roles for heme in, for example, gas sensing and transcriptional regulation has stimulated extensive research. Key roles for heme in regulating physiological functions such as circadian rhythms, ion channel activity, and miRNA biogenesis are now clear, as is heme function in gas sensing, e.g. in cGMP synthesis and signal transmission, and in regulating microbial respiration and denitrification. However, many challenges remain to clarify roles of CP "motifs" in heme binding in some cases (cf. forming disulfides, for instance) and to define mechanisms by which gas ligand binding induces structural changes that activate enzymes such as sGCs for cGMP synthesis and DosT for autophosphorylation and phosphate transfer to regulate transcription. A century after its structural characterization, the versatility of heme in biology continues to surprise, and its myriad functions are still not fully understood.