Abstract
Introduction
Intrinsic Ligand Selectivity of Heme
Heme, by itself, shows a wide range of reactivity with diatomic gases, anions, and bases. The anionic ligands, azide, cyanide, nitrite, thiocyanate, and halides only bind to hemin (Fe(III)-protoporphyrin IX), which has a net +1 charge at the heme core. On the other hand, the Lewis bases with nonbonded electron pairs, including the diatomic gases NO, CO, and O2, tend to coordinate primarily with heme (Fe(II)-protoporporphyrin IX) (120). The prominent interaction in this case is not charge but backbonding between the metal d orbitals (93) and the two ligand π antibonding (π*) orbitals. CO has two empty degenerate π* orbitals (LUMO), NO has one electron in these two π* orbitals (HOMO), and O2 has both π* orbitals occupied (as triplet O2) (61, 72). Model hemes differentially bind NO, CO, and O2, which are separated by 1 electron in their outer shell, 10–12 electrons, respectively. The binding affinities are in the order NO >> CO >> O2. In the case of O2, there is little back-bonding capability with filled π* orbitals, and as a result, O2 dissociation rate constants are normally several orders of magnitude greater than those for CO and NO (Table 1).
Unless otherwise specified, the experimental temperature was 20°C–25°C.
Calculated based on the koff and kon for 6c-NO complex determined in this study at 24°C.
Determined by O2 binding isotherm under high pressure (132).
N/A, not applicable; N/D, not determined.
The free radical nature of NO causes it to have the highest reactivity with iron and, therefore, the highest bimolecular association rate constants, which normally represent the rate of ligand movement into the protein active site (33, 83, 104). NO is also able to react with both the ferric and ferrous forms of all heme proteins (110). The Fe(II)-NO complex exhibits geometric flexibility (24). Normally, a linear conformation (sp hybridization) is observed for the ferric complex, where NO has triple bond character, and a bent conformation (sp2 hybridization) is observed for the ferrous complex, where the bound NO has a double bond character. However in both cases, the molecular orbital pattern is on the borderline between linear and bent geometries according to the general rule developed by Feltham and Enemark for {MNO} n , where n is the sum of the metal d-electrons and the NO π* electrons (31). The M-N-O conformation is predicted to be linear for n≤6, but bent for n>6. For the Fe(III)NO and Fe(II)NO, n equals 6 and 7, respectively. Thus any change in redox state of the iron or steric hindrance and electrostatic interactions with the bound NO may easily change the ligand conformation. For the same transition metal series, NO ligated to Mn(II)PPIX, n=6, should be linear, whereas NO coordinated to Co(II)PPIX, n=8, should be bent. Because it is a 3-electron donor, NO is a strong field ligand, and both the six-coordinate and five-coordinate NO heme complexes are low-spin due to large energy splitting between the t2g and eg d orbitals (72). In contrast to FeNO, FeCO always assumes an almost linear geometry, Fe-C-O ∼180°, and O2 always shows a bent conformation, Fe-O-O ∼110°–120° (92, 93, 103, 125).
Effect of Protein on Heme Ligand Selectivity
Although the intrinsic geometries of model heme-gas complexes remain relatively unchanged in proteins, the relative affinities for NO, CO, and O2 binding can be altered dramatically when the heme is associated with protein. The chemical nature and geometry of the axial ligand provided by the protein on the proximal side of the heme group can alter the absolute reactivity of the iron atom over several orders of magnitude (47, 84). Steric constraints and electrostatic interaction with the bound ligand on the distal side of the heme group can dramatically influence ligand selectivity (84, 90). Different stereochemical strategies appear to have evolved to optimize both absolute ligand affinity and selectivity between NO, CO, and O2 and to meet specific physiological needs for O2 storage and transport [e.g., myoglobin (Mb), hemoglobin (Hb) (109)], O2 sensing (e.g., FixL, EcDos, HemAT) (39), NO storage/transport (nitrophorins) (126), NO sensing (e.g., sGC, cyt c′, and other sensors of NO) (15, 22, 26), and CO sensing (e.g., CooA and NPAS in circadian rhythm control) (8, 29, 95).
For example, the affinity of CO for model heme with a proximal imidazole base is roughly 4000 times greater than that for O2, whereas in mammalian Mbs and Hbs the ratio KCO/KO2 is only 25–200, indicating a selective increase in affinity for O2 in the globins (84) (Table 1). In these proteins, the highly conserved distal E7 histidine donates a strong hydrogen bond to preferentially stabilize bound O2 by a factor of roughly 1000 (84, 90). Direct steric hindrance from protein residues on the bound ligands, although greatly affecting absolute affinity, causes little (≤2–3-fold) additional discrimination between NO, CO, and O2.
Nitrophorins are NO storage hemeproteins present in the saliva of the blood-sucking insect Rhodnius proxlixus and achieve complete selectivity for NO by keeping the iron atom oxidized and incapable of binding O2 and CO (126). When the uncharged NO binds to the hemin iron atom, it displaces water molecules from the distal portion of the heme pocket and induces movement of two flexible loops to close the active site by a mechanism referred to as a hydrophobic trap. These conformational changes due to NO binding are in sharp contrast to the lack of structural alterations when anionic ligands, such as cyanide, bind to nitrophorin (127).
Hemeproteins involved in sensing diatomic gases are typical sensor/effector two-component systems (39). For some of these proteins, functional selectivity does not occur directly during the ligand binding step but instead occurs during coupling with the effector or enzymatic domain. For example, the O2 sensor FixL from nitrogen fixing Rhizobia has an N-terminal heme domain and a histidine kinase C-terminal domain (74). In this protein, the partial negative charge on bound O2 attracts a key basic residue, Arg220, to move toward the center of the porphyrin (10). This movement of Arg220 is believed to be the conformational switch that leads to signal transduction and activation of the histidine kinase in FixL. Although NO and CO also bind well to ferrous FixL, the resultant neutral FeNO and FeCO complexes fail to induce the required Arg220 conformational change and the kinase domain remains inactive (1, 10).
CooA is a CO sensor/transcription factor found in Rhodospirillum rubrum and represents another case where activation of the nonheme domain is ligand specific. This protein has a heme regulatory N-terminal domain and a C-terminal DNA-binding domain (95). Reduction of the heme iron center causes a proximal ligand shift from a Cys thiolate (Fe+3) to a His imidazole (Fe+2) (9). CO only binds to the reduced heme and thus greatly increases the reduction potential of the iron atom. The concomitant proximal ligand shift causes a large movement of the C-helix and leads to a conformational change in the DNA binding domain. Thus, signaling in CooA involves both a redox and a conformational switch (7, 8). O2 does appear to bind CooA, but very quickly oxidizes the iron atom and thus cannot sustain the conformation required for affecting DNA binding (7, 96). NO binding to CooA results in a 5c-NO-heme complex with a very different protein conformation that is not competent for activating the conformational switch (94).
EcDos is an O2 sensor found in Escherichia coli and has an N-terminal PAS heme domain and a phosphodiesterase C-terminal domain (39). Reduction of the ferric form of this enzyme causes binding of a Met side chain to form a hexacoordinate heme complex, which in turn causes a scissor-like motion in the quaternary structure of the EcDos homodimer (58). This conformational change leads to the association of the phosphodiesterase domains to form an active enzyme (58). O2 and CO bind to ferrous EcDos with a similar affinity (KD ∼10 μM) and both displace the bound Met side chain, inactivating the enzyme. The association rate constants for all three diatomic gases (O2, CO, and NO) are very low, ≤0.003 μM −1s−1 (40). Thus, signal transduction, at least for O2 and CO, is determined by the natural abundance of individual gaseous ligand, which in the case of O2 is high enough to inactivate the enzyme under aerobic conditions.
sGC: A Highly Selective NO Receptor
The signaling mechanisms described for FixL, CooA, and EcDos and the NO storage and delivery mechanisms of nitrophorin all have structural support from crystallographic data. No crystal structure has been determined for sGC, and as a result, mechanistic information regarding signal transduction in this enzyme relies on spectroscopic and kinetic studies of wild-type and mutant proteins. In contrast to the gas sensors described previously, sGC is a heterodimer with the heme center present in the β subunit, and the cyclase active sites present at the interface between the α and β subunits (91). As isolated, ferrous sGC appears to bind NO with a KD of ∼4 pM, which is similar to that for NO binding to mammalian Mbs (30, 93). However, the affinity of sGC for CO is 100 million-fold lower, and the observed KD for CO binding is on the order of 100–300 μM, which is roughly 10,000-fold larger than KD for CO binding for mammalian Mb (17, 112) (Table 1). More importantly, there is no evidence for the formation of an oxygenated ferrous complex under atmospheric pressure in air or pure O2. Autooxidation is remarkably slow, and purified sGC does not show any heme optical spectral shift for several weeks at 4°C, even after substantial loss of cyclase activity has occurred. Thus, gas selectivity for sGC appears to be determined almost exclusively at the ligand binding step.
Total exclusion of O2 is a special property of sGC and is observed for only a few gas sensors, including cyt c′. Four possible mechanisms have been proposed to explain the absence of O2 binding by ferrous sGC: 1. Lack of a hydrogen-bond donor, such as histidine or tyrosine, on the distal side of the heme to preferentially stabilize bound O2. 2. A negative electrostatic field adjacent to bound ligands that selectively destabilizes the polar Fe-O2 complex. 3. Steric restriction of ligand access to the iron atom by distal heme pocket amino acid side chains that directly inhibit bond formation and sterically hinder the bound ligand. 4. Inhibition of in-plane movement of the Fe atom due to an unfavorable Fe-proximal base orientation or other steric restraints that prevent movement of the proximal amino acid toward the heme group.
Preferential Stabilization of Bound O2 by Hydrogen Bonding and Destabilization by Negative Electrostatic Fields (Mechanisms 1 and 2)
The distal histidine located at the seventh position along the E-helix in Mb has been shown to stabilize bound O2 over a 100-fold by hydrogen bonding in Mb and Hb (84, 90, 93). Replacing His(E7) with apolar amino acids causes marked, 100-fold decrease in O2 affinity of mutant Mbs and Hbs (70, 80, 98, 111). In many bacterial and some invertebrate Hbs, a tyrosine at the B10 helical position serves as a strong H-bonding donor to preferentially stabilize bound O2, and in some of these cases, a glutamine at the E7 position further stabilizes the Fe(II)O2 complex (16, 30, 38, 45, 76, 89). In these proteins, replacement of TyrB10 by phenylalanine or GlnE7 by leucine markedly increases the O2-dissociation rate constant indicating loss of these favorable interactions (30, 38, 73, 77).
Marletta and his coworkers have suggested that a lack of H-bonding donors in the distal heme pocket of sGC may account for O2 exclusion based on their investigations of several other gas-sensing heme proteins that show significant sequence homology with the sGC heme-binding domain (13 –15, 87). This hypothesis was proposed based on the structural information from a heme sensor isolated from the anaerobe, Thermoanaerobacter tengcongensis (Tt), which belongs to the H-NOX family of proteins. The crystal structure revealed a direct H-bond between Tyr140 and bound O2 and secondary H-bonds from Trp9 and Asn74 to Tyr140 (13 –15, 87) (Fig. 1, top). Crystallographic data for the same protein, also called Tt sensor of NO (SONO), obtained by Raman and his colleagues (81) showed different side chain orientations of Trp9 and Asn74, which do not appear to donate H-bonds to Tyr140 (Fig. 1, bottom). However, the bond length between the O atom of Tyr140 and the second O2 atom of the O2 is the same in both structures. Bound O2 is assigned an orientation in the Tt SONO structure almost opposite to that in the Tt H-NOX structure (Fig. 1). When Tyr140 is replaced with Leu in Tt H-NOX, O2 affinity decreases more than one order of magnitude, and the double Y140L/W9F mutant shows little or no O2 binding (13). While involvement of Trp9 and Asn74 in regulating O2 binding is tenuous, it is clear the Tyr140 stabilizes bound O2 in this sensor.
Marletta and coworkers have suggested that replacing apolar residues to tyrosines in the heme active sites of NO-binding sensors that normally exclude O2 can convert them into H-NOX domains that bind O2 (13). To test this hypothesis, they replaced the apolar isoleucine located in the heme-binding pocket of sGC to tyrosine. They reported that the I145Y mutation in the β subunit heme domain of sGC [β1(1–385)] allows O2 binding, but their estimated KD of ∼70,000 μM demonstrates that O2 is still excluded from binding to the iron atom under physiological conditions, even when a tyrosine side chain is near the bound ligand (13). Insertion of a tyrosine into the active site of Legionella pneumophila H-NOX (L2 H-NOX) gave more definitive results indicating reversible O2 binding as judged by both kinetics and visible spectra (13).
Based on these mutagenesis results, Marletta and his colleagues concluded that (i) a distal pocket tyrosine is necessary and sufficient for O2 binding in the H-NOX family and (ii) lack of a distal pocket H-bond donor in sGC eliminates O2 as a functional ligand, making the sGC H-NOX domain selective for NO. However, we have recently shown that the full-length I145Y mutant of α1/β1-sGC does not show O2 binding even in buffer saturated with pure O2 (68) (Fig. 2). In addition, the SONO protein from Clostridium botulinum (CB SONO or CB H-NOX), which contains a conserved distal pocket Tyr, has an apparent KD for NO binding of ∼10−15 M and still does not bind O2 (81). Thus, the presence of an active site-tyrosine does not assure O2 binding.
Similarly, a completely apolar heme site for gas ligands does not necessarily preclude O2 binding. Replacing the distal histidine in Mb with Val, Leu, Ile, and Phe creates a completely apolar active site and causes a selective ∼100-fold increase in the KD for O2 binding from ∼1 to 50–300 μM (98, 111). However, O2 binding is easily observed in air or 100% O2 and facilitates rapid autooxidation (111). Thus, loss of hydrogen bonding by itself cannot account for the complete exclusion of O2 binding by sGC.
Denninger et al. (27) have argued that negative electrostatic fields in the vicinity of the bound ligand could preclude O2 binding by repulsion of the polar Feδ(+)-O2 δ(−) complex. However, sequence homology with the bacterial NO sensors and their structures indicate that there are no electron donors or negative charges in the active sites of these heme proteins, except for the possibility that the nonbonded electrons of the tyrosine Oζ atom in Tt H-NOX could be pointing toward the bound ligand and not the hydroxyl H atom. Nevertheless, even when negative fields occur near bound O2 in V68T Mb (108) and wild-type Cerebratulus lacteus miniglobin (88), O2 binding is still observed.
Direct Steric Inhibition of Ligand Binding (Mechanism 3): Lessons Learned from Alcaligenes xylosoxidans Cytochrome c′
The third explanation for O2 exclusion by sGC is that access to the iron atom is sterically hindered by large distal amino acid side chains. In globins, the direct steric restriction of iron-ligand bond formation appears to uniformly decrease the affinities of all three diatomic ligands (85). This effect is best seen for the Val68 to Ile mutation in Mb, where the Cδ-methyl group lies directly over the iron atom and decreases the affinity of Mb for all three gases by ∼5-fold. Similar nonselective effects are seen in Mb and Hb when Leu at the B10 helical position is increased in size to Trp and causes a ∼25-fold decrease in all three ligand affinities (12, 86, 93).
Severe steric restriction of ligand binding does occur in cyt c′ from A. xylosoxidans, which has biochemical, biophysical, and ligand binding properties that are very similar to those of sGC (59), and for which there is a high resolution crystal structure. There are six similarities between sGC and cytochrome c′ that are relevant to ligand binding (2
–5, 37, 49, 67, 71): 1. Both proteins have very hydrophobic distal heme pockets (Figs. 3 and 4) (59, 81). 2. Both proteins bind CO and NO but not O2 (Table 1) (49, 113). 3. As isolated, the heme centers of both proteins are 5c high-spin ferrous iron (3, 112). 4. The KDs for CO binding are much greater than those for other reduced, high-spin hemeproteins with a proximal histidine ligand, and the association rate constants for CO binding are very small (50, 112, 114). 5. The ferric form of each hemeprotein shows poor cyanide and azide binding in comparison with Mb or model hemes (50, 75, 115). 6. NO binding first forms a six-coordinate NO complex (6c-NO) complex and then transforms to a five-coordinate NO complex (5c-NO complex). The conversion from 6c-NO complex to 5c-NO complex is dependent on [NO] for both proteins, indicating a two-step scheme, both of which involve NO binding (2, 131).
Steric constraints in the distal pocket appear to be the key factor repressing O2 binding to cyt c′, and extensive ligand binding (49) and resonance Raman (rR) and Fourier transformed infrared spectroscopy studies (3, 4, 37) and quantum mechanics/molecular mechanics computations (67) support this hypothesis. Comparison of the crystallographic data of the CO-free and CO-bound forms of cyt c′ from A. xylosoxidans (AXCP) suggest that the side chain of Leu16 blocks access to the iron atom and must be pushed away from the center of the heme ring to accommodate bound CO (6). This movement also causes extensive rearrangements of other distal residues, including Pro55 (Fig. 3). Antonyuk et al. (6) have shown that this steric hindrance decreases CO affinity 4 to 5 orders of magnitude compared to that for heme models or Mb, and the value of the KD for CO binding to cyt c′ is very similar to that for sGC (67) (Table 1).
These workers have recently shown that L16A and L16G mutations in cyt c′ decrease KD(NO) and KD(CO) to remarkably small values, ∼70×10−15 and ∼3×10−12 M, respectively. Most remarkably, these cyt c′ variants bind O2 with high affinity, KD(O2) ≈50 nM (Table 1) (6, 36). Such high ligand affinities indicate that the intrinsic reactivity of the heme iron in cyt c′ is very high as previously suggested by theoretical calculations (6, 67). Thus, the severe restriction of O2 binding in wt cyt c′ is due exclusively to the iso-butyl side chain of Leu16, which also reduces the stability of 6c-CO and -NO complexes.
In contrast to CO binding, the reaction of NO with reduced cyt c′ is complex and involves formation of an initial unstable hexacoordinate complex, which shows a very poor affinity for NO (i.e., KD≈0.1–1 μM, Table 1) (2, 36, 37). Then, the proximal imidazole bond breaks, a second NO binds to the iron from the proximal pocket, displacing the original ligand on the distal side to form a stable 5c-NO complex. In this equilibrium 5c complex, the bound NO is on the proximal side of the heme. The preference for NO binding on the proximal side appears to be a consequence of the highly congested distal pocket, due primarily to Leu16 and stabilization of bound NO by Arg124 on the proximal side of the porphyrin ring (36, 59). The binding of a second NO in the proximal pocket to create a transient NO-heme-NO species provides the basis for the [NO] dependence of conversion from 6c- to 5c-NO complex (2, 67).
Proximal Coordination Geometry and Inhibition of Ligand Binding (Mechanism 4)
The fourth proposal for O2 exclusion is that steric constraints on the proximal side of the heme inhibit in-plane movement of the iron atom, weaken the pentacoordinate Fe-proximal His bond, and greatly inhibit formation of hexacoordinate complexes with all three ligands. However, NO is capable of forming a strong pentacoordinate complex and breaking the sterically-restricted Fe-proximal His bond, whereas CO and O2 cannot. Consequently, a high affinity 5c NO-heme complex can be formed by displacing the proximal histidine. This idea was first proposed by Traylor, Sharma, and colleagues in the 1980–1990s (106, 107, 119), and is supported by the multi-step NO binding scheme proposed for cyt c′ and by multiple electron paramagnetic resonance (EPR) and resonance Raman (rR) studies, which show that at equilibrium, the sGC-NO complex is pentacoordinate (25, 28, 64), as represented in Figure 8.
The stretching frequency of the Fe-His bond in the ligand-free reduced sGC is 204 cm−1, which is at the low end of all known measurements, where ν(Fe-His) ranges from 200 to 300 cm−1 (129). In many other proteins, the presence of an H-bond acceptor adjacent to the proximal histidine Nδ-H atoms usually increases the Fe-imidazole bond strength by ∼20 cm−1 due to enhanced basicity of the coordinating Nɛ atom (89). In sGC, weak π-donation by the histidine Nɛ atom causes weak Fe dπ donation to O2 π*. There is weak linear correlation between ν(Fe-His) and ν(Fe-O) for model heme compounds (30), but rR data for the Fe-His stretching of VC H-NOX and Tt H-NOX are opposite to the observed O2 binding capability of these two heme sensors (48). Moreover, a low value of ν(Fe-His) does not necessarily predict O2 exclusion. Many hemeproteins have a weak Fe-His bond but can easily bind O2. For example, some deoxygenated model Fe(II)PPIX complexes have ν(Fe-His)=201–205 cm−1 but still show moderate O2 affinities (23, 129). FixL(νFe-Nɛ=208 cm−1) (118), T-state α-chain Hb (νFe-Nɛ=207 cm−1) (79), ovine prostaglandin H synthase, or oPGHS-1, (νFe-Nɛ=206 and 225 cm−1) (62, 105) also have Fe-His bond strengths similar to that of sGC but bind O2 under physiological conditions. Ferrous o-PGHS-1 reacts readily with O2 and autooxidizes rapidly to form ferric prostaglandin H synthase type 1 (PGHS-1) (Tsai et al., unpublished results). Tomita and his colleagues have concluded that there is no correlation between Fe-His bond strength as measured by ν(Fe-His) and O2 binding parameters (118).
Even if the Fe-His bond is strong in the pentacoordinate reduced state, there may be large steric restraints to the in-plane movement of the iron that is required for O2 and CO binding. As result, both the Fe-His and Fe-ligand bonds can be weakened markedly in hexacoordinate low-spin complexes, and there is strong evidence that this occurs in sGC (118). The sGC-CO complex exhibits the lowest Fe-CO stretching (472 cm−1) that has been observed for heme proteins containing a neutral proximal histidine (25), and the weak Fe-CO bond correlates with the very large CO dissociation rate constant (koff) observed for this protein (Table 1) (13, 68, 112). The remarkably large koff(CO) value cannot be the result of an apolar pocket because hydrogen bonding to bound CO is weak and can only stabilize the bound ligand by a factor of 2–5 based on mutagenesis studies with Mb (60, 90). In contrast, the proximal constraints associated with the R to T transition in human Hb cause koff(CO) to increase 20-fold from 0.01 to 0.2 s−1 (124), and even larger effects have been observed for model hemes with proximal constraints (119). Thus, proximal constraints do play a very significant effect in regulating ligand binding affinity.
Lessons from Nostoc H-NOX, a Structurally Homologous Sensor to sGC
The recent crystal structure of Ns H-NOX has provided very useful insights into ligand selectivity by gas sensors (63). Ns H-NOX, (also called NP-SONO) has the highest sequence identity (31%) and structural homology with the sGC heme domain compared to the other SONOs from C. botulinum and T. tengcongensis (81). The proposed model for the heme pocket for sGC is similar to that reported for Ns H-NOX. Six hydrophobic amino acid residues line the distal heme pocket and are highly conserved even in their side chain orientations (V5/V5, I9/L9, F70/F70, W74/F74, M143/I145, and L147/I149, respectively, for Ns H-NOX/sGC), and the only polar amino acid (T77/C78) also appears to adopt the same side chain orientation relative to the heme porphyrin by molecular modeling (Fig. 4).
However, a recent spectroscopic and kinetic study revealed interesting differences in gas binding to Ns H-NOX versus sGC (122). Ns H-NOX sequentially forms two different but relatively stable 6c-NO complex intermediates with Soret absorption peaks at 418 and 414 nm, respectively, after mixing the unliganded, reduced 5c complex with NO. Formation of the second 6c-NO complex also shows dependence on [NO], indicating a complex multi-step mechanism. Only partial conversion to 5c-NO complex occurs even with excess NO (122). The restricted formation of the 5c-NO complex in Ns H-NOX, and the substantial conformational change upon heme oxidation implies that Ns H-NOX may be a redox sensor rather than an NO-sensing protein. In addition, Ns H-NOX binds O2 very weakly and autooxidizes slowly (122), whereas sGC is totally inert to O2.
The major structural difference between the heme pockets of Ns H-NOX and sGC appears to be the replacement of a Trp74 in H-NOX by a Phe74 near the heme iron in sGC (63) (Fig. 5A). In the CO- and NO-bound Ns H-NOX structures, the indole ring of Trp74 is in van der Waal's contact with the gaseous ligand, and the indole nitrogen may serve as a H-bond donor to the nitrogen atom of bound NO (the oxygen atom of NO does not appear to be properly aligned for H-bonding even though it is closer in distance) (Fig. 5B, C). The bulky Trp74 side chain appears to cause an ∼15° tilt from a perfect linear Fe-CO geometry. The same steric effect likely causes a Fe-N-O angle of ∼150° rather than a 120° bent conformation predicted by the rule of Feltham and Enemark. The equivalent residue for sGC is Phe74, which is less bulky and has no H-bonding donor capability.
The presence of a Trp74 H-bond donor may provide a stabilization mechanism for the observed 6c-NO complexes in Ns H-NOX, and the H-bonding would favor the bent form of NO and a less negative trans effect. Such stabilization of the 6c-NO-sGC complex by Trp74 may account for the lower NO-dependent activation observed for the α-βF74W sGC mutant (63). Careful kinetic measurements for NO and CO binding to wild-type Ns H-NOX, its W74F mutant, and to the αβF74W sGC should provide a clearer picture of the role of residue74 in ligand selectivity.
The Sliding Scale Rule for NO, CO, and O2 Binding Affinity to 5c Heme Proteins with a Neutral Proximal Imidazole Base
To understand the unusually high affinity for NO but complete exclusion of O2 by sGC, we quantitatively examined the relationship between ligand binding parameters and the chemical nature of the gas ligand for >75 different single and multiple variants of globins, heme proteins, and model heme compounds (120). Plots of the logarithm of KD versus ligand species for the series NO, CO, and O2 yield sets of parallel lines for heme proteins containing a neutral proximal histidine side chain and samples of these data are shown in Figure 6. The absolute KD values are spread over a vertical range of 8–9 orders of magnitude. The highest affinities and lowest KD values are observed for L16A cyt c′, in which distal steric hindrance was alleviated by an alanine replacement (36) and H61L leghemoglobin (Lba) mutant, which has a completely apolar and unhindered distal pocket (41). The lowest affinities and highest KD values were observed for wild-type cyt c′, wild-type sGC, and sGC with site-specific replacement of isoleucine by a tyrosine on residue 145 (I145Y sGC) (Fig. 6). The KD values for chelated model heme, unhindered Mb mutant (H64V), and Ns H-NOX lie in between the two limits. The slopes of the lines for log(KD) versus ligand type, from NO to CO to O2, show that, in general, there is a ∼3- to 4-order separation between the KD values of NO and CO and between those for CO and O2 (Fig. 6). This parallel relationship of log KD for the three gases occurs over a dynamic range of 8–9 orders of magnitude for the individual equilibrium dissociation constants. This linear relationship describes a sliding scale rule for ligand discrimination that applies to all heme proteins containing a neutral proximal imidazole ligand and an apolar distal pocket (121).
The most remarkable result in Figure 6 is the effect of the L16A mutation on ligand binding to cyt c′. Relief of distal steric hindrance by replacement of Leu16 with either alanine or glycine in cyt c′ moved the KD values of all three gas ligands from the upper (weakest affinity) to the lower (highest affinity) limits of the log(KD) versus ligand plot. This result demonstrates dramatically the critical role of steric hindrance by the Leu16 side chain in regulating ligand access to and dissociation from the heme iron in this bacterial heme sensor. Once steric restriction is removed in cyt c′, the heme iron shows an affinity for ligands that is as high as that of the unhindered leghemoglobin (H61L lb). The cause of the ultra-high affinity of the cyt c′ mutant is the staggered geometry of the proximal histidine ring with respect to the pyrrole N atoms (6, 36, 43), which is compared to the eclipsed geometry found in Mb in Figure 7b. Soybean Lb shows a similar staggered proximal geometry (41, 56, 57), and a variety of mutagenesis and kinetic studies have shown that this conformation accounts for the high reactivity of Lb with all ligands compared to Mb (41, 56, 57). In wt cyt c′, the iron atom is still highly reactive, but the binding of all ligands is dramatically inhibited by the sterically restricted distal active site, in which the iso-butyl side chain of Leu16 is held rigidly over the axial position of the heme iron atom (Fig. 3).
Deviations from the Sliding Scale Rule: sGC and Ultra-High O2 Affinity Globins
The major outlier from the sliding scale rule shown in Figure 6 is the sGC. Despite a KD(CO) similar to that for cyt c′ and I145Y sGC, the reported overall equilibrium dissociation constant for NO binding to sGC is ∼5 orders lower than that predicted by a line parallel to the data for all the other heme proteins. This deviation from the sliding scale rule led us to reassess the literature value for the NO binding equilibrium constant. The reported value for KD(NO) value for sGC was calculated as the ratio between the koff for the final equilibrium 5c-NO complex and the kon for the initial 6c-NO complex formed on mixing unliganded ferrous sGC with NO anaerobically (52, 131). Thus, the KD(NO) value does not represent the equilibrium dissociation constant for the initial 6c-NO complex (Fig. 6, sGC1st step) but instead is an estimate of the KD value for the overall reaction to form the final 5c complex (see Fig. 6, sGCeq).
Determination of the KD for the initial 6c-NO sGC complex, which should be predictable by the sliding scale rule analysis in Figure 6, requires accurate measurement of the NO association (kon) and dissociation (koff) rate constants for the formation of this transient intermediate. Using rapid multi-mixing methods, we determined kon and koff to be 4.8×108 M−1s−1 and 27 s−1 at 24°C for the initial 6c NO-heme-His complex of full length sGC (121). Thus, the KD(NO) for the initial 6c-NO complex is 5.4×10−8 M, which is similar to the KD values for the initial and more kinetically stable 6c NO complexes of cyt c′ and I145Y mutant of sGC (37, 68, 122) and the value predicted by the sliding scale rule with the KD for CO binding (Fig. 6, red circle).
This correlation between the predicted and observed KD values for the 6c-NO complex of sGC supports the use of the sliding scale to estimate the KD value for O2 binding to sGC. As shown in Figure 6 (red dashed line and red circle with grey edge), the predicted KD for O2 binding to sGC is ∼1 M, way above the O2 concentration of air- or even O2-saturated buffer. Similar high KD(O2) values are predicted for I145Y sGC and cyt c′ (∼10 M, Fig. 6), and thus the sliding scale rule and the measured values of KD(NO) and KD(CO) for 6c complexes empirically predict the exclusion of O2 binding to these proteins under aerobic conditions.
As a test of the validity of the sliding scale rule, we looked for O2 binding to Ns H-NOX at high O2 pressures. Ns H-NOX was chosen because of its lower KD for CO binding (Fig. 6) and because it does slowly autooxidize at a rate of 0.05 h−1, indicating some reactivity with O2 in air (122). Linear extrapolation of the line connecting the log KD values for NO and CO binding to Ns H-NOX predicts a KD for O2 binding of ∼15 mM (blue circle with grey edge in Fig. 6). We then examined O2 binding to Ns H-NOX experimentally using a high-pressure cell and obtained an isotherm with a KD≈13 mM (Fig. 6, blue circle) (121), which matched almost exactly the prediction from the sliding scale rule. Thus, the sliding scale rule applies to the 6c complexes of all three of the NO sensors and predicts empirically that none of them can bind O2 under physiological conditions.
There are other dramatic deviations from the sliding scale rule due to selective lowering of the KD(O2) values for globins as result of selective stabilization of the polar Fe-O2 complex (121). This selective discrimination in favor of O2 binding has been the focus of extensive study over the past 30 years and is due to electrostatic stabilization of bound O2 by hydrogen-bonding with distal amino acids, most noticeably distal E7 His and Gln and B10 Tyr side chains in the globin family of heme proteins (84, 93). This selective stabilization is absent in mutants, where His(E7), Gln(E7), and Tyr(B10) are replaced with apolar amino acids, (i.e., the H64V Mb mutant in Fig. 6) (121).
Sliding Scale Rules for Ligand Association (kon) and Dissociation Rate Constants (koff)
We carried out similar correlation graphical analyses of the koff and kon, rate constants for NO, CO, and O2 binding to the same large library of heme proteins. As shown in Figure 7a (left panel), a set of parallel V-shaped lines are observed for plots of log(kon) versus NO, CO, and O2, with absolute values spanning 5–6 orders of magnitude and kon(CO) always showing the lowest value (120). There are two reasons why CO shows the lowest kon relative to NO and O2. First, the in-plane movement of the heme iron is required to vacate the electron in the dz2 orbital to accommodate two electrons from the nonbonding orbitals of CO; and second, the spin-forbidden nature of iron-carbonyl bond formation markedly increases the activation barrier (33, 116). In contrast, bond formation can occur without a spin-state change of the iron when NO and O2 are the ligands, because both contain unpaired electrons in the HOMO (46). The bimolecular association rate constant for NO binding is limited almost exclusively by the speed of ligand movement to the iron atom, whereas CO binding can be limited by rates of iron movement into the heme plane and bond formation, which can be slow relative to diffusion of the ligand into the protein. O2 binding can be limited by both processes, but in many cases, kon(O2) is similar to kon(NO).
Soybean H61L Lb sets the upper limit of the V kon relationship and wild-type cyt c′ sets the lower limit, representing totally free and completely sterically-restricted access to the iron atom, respectively. Steric constraints from the protein structure clearly plays a dominant role in inhibiting ligand binding to cyt c′, regardless of which ligand is examined. Replacement of Leu16 with Ala causes a 2–5 order of magnitude increase in kon for the binding of all three gas ligands to cyt c′ (Fig. 7a, left panel). Interestingly, both cyt c′ and Lb mutants show similar shallow kon dependences on ligand type, with kon(CO) being large and much closer to kon(NO) and kon(O2). This behavior is indicative of a highly reactive iron atom, high rates of bond formation, and diffusion limitation of ligand association. Even though their KD values are almost identical, the kon values for L16A cyt c′ are uniformly ∼2 orders of magnitude smaller than those for ligand binding to H61LlLb. Thus, movement into the active site of L16A cyt c′ is still restricted by the densely packed amino acid side chains surrounding the active site (Fig. 3).
Graphical analyses of koff values versus ligand type are also shown in Figure 7a (right panel). The parameter values can be represented as a set of parallel reversed L patterns. The general trend is that koff(O2) is much larger than either koff(NO) or koff(CO), and in most cases, koff(NO) is 10- to 100-fold lower than koff(CO). The koff values indicate the stability of the ligand-bound complex, and the lack of backbonding capability of O2 relative to NO and CO is the basis for its faster dissociation from the heme iron. CO can use both π* orbitals for backbonding, which greatly strengthens the Fe-CO bond. NO only has one π* orbital and one extra electron available compared to CO, and its backbonding strength is not much weaker than CO (24). Relief of one antibonding interaction by losing one backbonding interaction in the bonding plane is offset by a stronger backbonding in the perpendicular plane (y-z plane) with a lower energy π*. Thus, even though the bent form of Fe(II)-NO seems to be favored over the linear form according to the Feltham and Enemark rule (31), the energy difference between these two forms is not large and is subject to fine tuning by the protein environment. This extra flexibility and the radical nature of NO probably enhance the stability of the NO complex, matching that of the CO complex. The decreased value of koff(NO) compared to koff(CO) is due primarily to high rate of NO-Fe bond reformation within the protein after thermal dissociation compared to the rate of ligand escape from the protein (82, 83).
The koff values for the 6c-sGC complexes define the upper limits in these plots, and remarkably the koff values for L16A cyt c′ define the lower limits (Fig. 7a, right). The koff(O2) for sGC is expected to be large based on its predicted KD, ∼1M (Fig. 6). We estimate that the koff(O2) for sGC is 107–108 s−1 based on the product of kon and the projected KD value (Fig. 7a, right). Again, the most remarkable result is the dramatic ∼5 order of magnitude decrease in koff for ligand dissociation from cyt c′ due to the L16A mutation. It is clear that distal steric hindrance greatly destabilizes bound ligands in the w.t. cyt c′ complex causing dramatic increases in the rates of thermal breakage of all the iron-ligand bonds. When Leu16 is removed, even O2 binds with a high affinity (KD of 49 nM, Table 1) and a low dissociation rate constant (36). The reversed L sliding scale rule applies to this cyt c′ mutant, which shows lines parallel to the heme, Lb, and Mb model systems. Assuming the same parallel relationship, the value of koff(O2) for wild-type cyt c′ was estimated to be ∼5000 s−1, which is similar to what would be obtained from the predicted values of kon(O2) and KD(O2) from the sliding scale rule analyses of the data in Figures 6 and 7a, left panel.
The Basis of Ligand Discrimination in NO Sensors
The graphical sliding scale analyses demonstrate that ligand discrimination is built into the chemistry of 5c heme complexes with a neutral imidazole side chain. There are intrinsically ∼3 to 4 orders of magnitude separation between the affinities of the NO/CO and CO/O2 pairs. Except for preferential electrostatic stabilization of bound O2, adding protein around the heme either uniformly reduces or uniformly enhances the affinities of the protein for all three gaseous ligands by changing steric hindrance on the distal side or altering the Fe-imidazole geometry on the proximal side of the porphyrin ring. In the case of sGC, adding a polar distal tyrosine by the I145Y mutation had little effect on the relative or absolute affinities for CO and O2, suggesting that electrostatic interactions play a minor or no role in regulating ligand binding in this sensor (Fig. 6).
The large KD values for the binding of all three gaseous ligands to wild-type sGC and cyt c′ appear to be due to severe proximal constraints on iron reactivity and marked steric hindrance of bound ligands on the distal side of the heme plane, respectively. The observed kon value for the initial step of NO binding to w.t. sGC is in the range of 100 to 400 μM −1s−1, indicating little or no resistance to movement into the distal pocket or access to the iron atom from the distal side of the heme (Table 1) (123, 131). Similar large values for kon(NO) are observed for the unhindered model heme, H64V Mb, and H61L Lb (Table 1, Fig. 7a, left panel). The I145Y sGC mutation does decrease kon for NO binding to 15 μM −1s−1, a value similar to that for NO binding to wt Mb (Table 1), presumably because the large Tyr side chain does begin to restrict movement to the iron atom. In contrast, the observed bimolecular rate constant for NO binding to cyt c′ is only 0.04 μM −1s−1, which is a 10,000-fold smaller than that for NO binding to sGC. Replacement of Leu16 with Ala in cyt c′ causes an increase in kon(NO) to 3.0 μM −1s−1. Thus, it is clear that cyt c′ excludes O2 binding by severe distal steric hindrance.
The structural mechanism for exclusion of O2 and the poor affinity of CO for sGC are not as well established, but all the observed data suggest that proximal constraints dramatically lower the reactivity of the heme iron, as originally proposed by Traylor and Sharma over 30 years ago (119). Indeed, Shiro and coworkers (65) have recently emphasized that the heme proximal strain, which limits iron movement toward the porphyrin plane, is the crucial effect that leads to the ultra-weak affinity of sGC for O2. The exceptionally positive midpoint potentials (E M) of +187 mV and +234 mV for sGC from humans and Manduca sexta support the idea of restriction of in-plane movement of the iron atom during oxidation, which in turn favors the reduced pentacoordinate, out of plane geometry (34, 65). Thus, cyt c′ and sGC have evolved structurally distinct mechanisms to achieve O2 exclusion, that is, severe distal steric hindrance versus proximal strain, respectively.
A Multi-Step Mechanism Evolved to Achieve High NO Affinity When O2 Binding Is Excluded
Most heme protein sensors need to respond to low nanomolar levels of NO under aerobic conditions. O2 exclusion is required to prevent both autooxidation and more importantly NO dioxygenation to nitrate, which would rapidly remove the signaling molecule without eliciting the desired physiological effect (i.e., smooth muscle relaxation in the case of sGC). As described in the last section, cyt c′ achieves O2 exclusion by physically inhibiting ligand binding at the distal axial position, whereas sGC appears to limit the binding of O2 by inhibiting in-plane movement of the iron atom. Both mechanisms severely inhibit formation of 6c-ligand complexes and markedly increase the KD values for all ligands as shown in Figure 6. In order to increase the sensitivity to low levels of NO, these proteins evolved mechanisms to form high-affinity 5c-NO complexes with the ligand bound on the proximal side of the heme group.
Hough et al. (43) have provided a detailed multi-step mechanism for high affinity NO binding to cyt c′, based on ultra-fast kinetic measurements by Negriere's group (55) and their own mutagenesis and crystallographic studies. The first step involves bimolecular NO binding to form an initial 6c NO-heme-His complex, which is followed by a second catalytic, bimolecular NO binding process to form a highly transient and an unstable NO-heme-NO complex with the proximal His displaced. A final rapid, first order dissociation of the distal NO generates the final equilibrium 5c heme-NO complex (43). The second and third steps are driven by steric hindrance in the distal pocket of cyt c′, which forces the NO to the proximal side of the protein. CO is greatly inhibited and O2 is effectively excluded because 5c complexes with these ligands are extremely weak.
Animal sGCs evolved a similar multi-step binding mechanism to achieve high affinity NO binding, even though O2 exclusion appears to be achieved by proximal constraints of in-plane iron movement and not distal steric hindrance (Fig. 8). Again, the initial reaction of NO with ferrous sGC (with a 432 nm Soret peak, species A in Fig. 8) generates a weak 6c Fe(II)NO complex (420 nm Soret, species B in Fig. 8). A distal 5c Fe(II)NO complex (399 nm Soret) forms at low NO:heme stoichiometries (≤1) (species D* in Fig. 8), presumably due to the strong proximal constraints that greatly weaken the Fe-His bond when NO is bound (123). At higher NO:heme ratios, a proximal 5c NO-heme complex is formed in a second bimolecular process, which depends on [NO] (species D in Fig. 8) (123, 131). As in the case of NO binding to cyt c′, a bis-NO complex is proposed as a transient intermediate at excess NO (species C in Fig. 8). Although this very transient species has not been verified by direct spectroscopic identification, its formation is strongly supported by the observed linear [NO]-dependence of the rate of 2nd step to form the final equilibrium 5c complex. In addition, our recent EPR freeze-trap study using 15NO and 14NO isotope ligands in sequential stopped-flow experiments confirms that heme iron coordinates the second NO molecule (123, 132). As in the case of cyt c′, the net result of the additional NO binding step is the generation of a high affinity 5c proximal heme-NO complex.
An alternative proposal is that the second NO is bound covalently to a cysteine thiol or at another nonheme site during the activation process (19, 32, 100 –102). However, nitrosation of cysteine does not occur in the absence of O2, and even in the presence of O2, the association and dissociation reactions are too slow to contribute as an intermediate to the rapid formation of the final equilibrium 5c-NO complex with only one ligand bound (53, 69).
Complex mechanisms also appear to have evolved to allow Ns H-NOX and CB SONO, or CB H-NOX to have femptomolar KD values for NO but only very weak or no O2 binding under physiological conditions. In the case of Ns H-NOX, a multiple-step NO binding mechanism does occur, but the second [NO]-dependent step is a conversion from one type of 6c-NO complex (418 nm peak) to another (414 nm peak). The conversion from the 2nd 6c-NO complex to the final equilibrium 5c-NO complex is extremely slow, first order, and incomplete (122). The equilibrium 5c-NO complex formed for CB H-NOX is air stable, indicating that O2 cannot readily displace the bound NO, as is also the case for sGC (81).
Physiological Importance of O2 Exclusion and High NO Affinity
A summary of the mechanisms that have evolved to achieve O2 discrimination and high NO affinity by a variety of heme proteins is given in Table 2 and based on the graphical analyses and kinetic parameters discussed in the previous sections of this review. While the biological functions of some of the other heme proteins are not certain, sGC behaves as a bona fide NO sensor. sGC requires high selectivity for NO over CO and complete exclusion of O2 binding to prevent NO dioxygenation. The measured cellular level of NO in human embryonic kidney 293 T cells (HEK) cells, endothelial cells, and neuronal cells varies from picomolar to micromolar (11, 21, 128). Thus under normal conditions, the interaction between NO and sGC could occur at both high and sub-stoichiometric ratios.
Most “Yes” or “No” are on the relative scale to indicate the major and minor factors.
ND means three-dimensional structural information unavailable.
The second [NO]-dependent binding is conversion from 6c to 5c-NO complex.
“?” indicates possible but not proved function.
Indicates the imidazole ring orientation relative to the pyrrole nitrogens (Fig. 7b).
The second [NO]-dependent binding is the conversion from one 6c to another 6c-NO complex.
A mixture of orientations due to absence of protein structure to achieve a rigid ring orientation.
Our recent study indicates that stoichiometric NO is still able to activate sGC to a level similar to that observed at excess NO, indicating that both 5c-NO complexes (D and D* in Fig. 8) may be capable of cyclase activation or that D* slowly isomerizes to D, even at low NO levels (123). The multi-step NO reaction with sGC leads to Fe-His bond breaking, formation of a 5c-NO complex with an apparent KD(NO) of ∼10−12 M, and generation of a protein conformation that is coupled to cyclase activation. Reversible CO binding to form a simple 6c CO-heme-His complex (A↔E in Fig. 8) does not appear to strongly activate sGC. The critical proximal strain that results in formation of the active 5c NO-heme complex also evolved to drastically reduce the affinity of sGC for O2.
Exclusion of O2 binding inhibits oxidation of the reduced heme iron to its inactive ferric form with simultaneous production of superoxide and, more importantly, prevents rapid dioxygenation of NO by bound O2 to form biologically inactive nitrate. All heme proteins that reversibly bind O2 also rapidly dioxygenate NO (35, 82). Thus, exclusion of O2 by sGC is required for sensing low levels NO under aerobic conditions. This need for stringent ligand selectivity must be considered in searching for the biological functions for the other heme sensors, including cyt c′, Ns H-NOX, CB H-NOX, etc., where NO sensing may need to occur at markedly different levels of O2, particularly in the photosynthetic microorganisms.
Footnotes
Acknowledgments
We thank Elena Bogatenkova for expression and purification of αβI145YsGC for ligand binding studies. This work was supported by U.S. Public Health Service Grants HL088128 (E.M.), HL095820 (A.-L.T.), GM35649 (J.S.O.), HL47020 (J.S.O.), American Heart Association, South Central Affiliate Grant-in-Aid 09GRNT2060182] (E.M.), and Robert A. Welch Foundation Grant C0612 (J.S.O.).