The structure of carbonmonoxy neuroglobin reveals a heme-sliding mechanism for control of ligand affinity

Abstract

Neuroglobin (Ngb), a globular heme protein expressed in the brain of vertebrates, binds oxygen reversibly, with an affinity comparable to myoglobin (Mb). Despite low sequence identity, the overall 3D fold of Ngb and Mb is very similar. Unlike in Mb, in Ngb the sixth coordination position of the heme iron is occupied by the distal histidine, in the absence of an exogenous ligand. Endogenous ligation has been proposed as a unique mechanism for affinity regulation and ligand discrimination in heme proteins. This peculiarity might be related to the still-unknown physiological function of Ngb. Here, we present the x-ray structure of CO-bound ferrous murine Ngb at 1.7 Å and a comparison with the 1.5-Å structure of ferric bis-histidine Ngb. We have also used Fourier transform IR spectroscopy of WT and mutant CO-ligated Ngb to examine structural heterogeneity in the active site. Upon CO binding, the distal histidine retains (by and large) its position, whereas the heme group slides deeper into a preformed crevice, thereby reshaping the large cavity (≈290 ų) connecting the distal and proximal heme sides with the bulk. The heme relocation is accompanied by a significant decrease of structural disorder, especially of the EF loop, which may be the signal whereby Ngb communicates hypoxic conditions. This unexpected structural change unveils a heme-sliding mechanism of affinity control that may be of significance to understanding Ngb's role in the pathophysiology of the brain.

Neuroglobin (Ngb) is a recently discovered vertebrate heme protein expressed in the brain (1). The 3D structure of ferric Ngb (metNgb) from human and mouse has recently been published (2, 3). The protein displays all key determinants of the canonical 3-over-3 α-helical globin fold (4) despite the very low sequence identify with canonical hemoglobins and myoglobins (Mbs). However, in contrast to Mbs, the heme iron in metNgb is hexacoordinated by the distal and proximal histidines (His-64 and His-96) in the absence of an exogenous ligand (2, 3); spectroscopic data show hexacoordination also for ferrous deoxy Ngb (1, 5, 6). Endogenous ligation at the sixth coordination position has already been reported for nonsymbiotic plant Hbs and bacterial Hbs (7, 8). Interestingly, whereas sperm whale Mb (swMb) displays only small matrix cavities (9), metNgb contains a huge internal tunnel (≈290 ų) that connects the distal and proximal sides of the heme to the bulk (2, 3). Such a large cavity implies a substantial free energy cost estimated at several kcal/mol (10).

Hexacoordination has been proposed as a novel mechanism to regulate ligand affinity in heme proteins (5–8, 11, 12) because the covalent bond between the heme iron and the distal His has to be broken for an exogenous ligand to bind. In Ngb, the latter reaction has been observed in flash photolysis experiments, in which recombination of CO was found to occur in two steps. The first process (on microsecond time scales at [CO] = 1 atm) arises from the competition between CO and the endogenous His-64 imidazole for the vacant sixth coordination; in a second step (at ≈100 s^–1), the His-64 ligand dissociates thermally and is replaced by the more tightly bound CO (12).

The physiological role of Ngb is intriguing but as yet unclear. Ferrous Ngb binds O₂ reversibly, with a half-saturation pressure P₅₀ ≈2.0 torr (1 torr = 133 Pa) at 20°C (1), similar to that of swMb (P₅₀ ≈1.0 torr); nevertheless, Ngb has a much greater tendency toward autoxidation (1, 5). Its low expression level (in the micromolar range) suggests a physiological function different from simple O₂ storage and transport as in Mb (13). In recent studies, up-regulation of Ngb expression was observed under conditions of hypoxia or ischemia in vitro and in vivo (14, 15), which may foster neuronal survival after a stroke. Moreover, the expression of Ngb can also be induced by hemin (16). These results imply that more than one signal transduction pathway is involved in the regulation of Ngb expression. Interestingly, only metNgb was found to interact with the GDP-bound form of the α subunit of the heterotrimeric G protein, which led to the proposition that Ngb might be a stress-responsive sensor for signal transduction in the brain (17). Other possible functions, such as NO scavenging or NADH oxidase activity, have also been discussed (18–20). In summary, Ngb is apparently important for the pathophysiology of stroke (15); however, its precise physiological role is as yet a conundrum.

To elucidate the mechanism of exogenous ligand binding to hexacoordinate globins and to eventually explain the role of the huge internal cavity seen in metNgb (2, 3), we have determined the 3D structure of murine CO-ligated Ngb (NgbCO) by x-ray crystallography. To examine structural heterogeneity in the distal pocket of NgbCO in detail, we have also measured Fourier transform IR (FTIR) spectra of heme-bound CO in WT murine Ngb, active-site mutants His-64–Leu, Lys-67–Leu, and double mutant His-64–Leu–Lys-67–Leu. The crystallographic data show substantial structural changes upon binding of CO to the ferrous heme iron, involving a sliding motion of the heme and a topological reorganization of the large internal cavity and its connectivity with the bulk. We conclude that heme sliding and distortion of the internal cavity are crucial for the control of ligand affinity in Ngb.

Materials and Methods

Crystallization. Murine Ngb was expressed in Escherichia coli and purified as described (12). Crystals of metNgb double mutant C55S–C120S were grown by using the hanging-drop vapordiffusion method (3). To prepare the CO derivative, crystals were transferred into vials (containing mother liquor with 25% glycerol added as cryoprotectant) that were sealed with rubber septa and deaerated by purging with N₂. The heme iron was reduced by adding sodium dithionite to a final concentration of 5 mM. Afterward, N₂ in the vials was replaced by CO, and the CO-ligated crystals were frozen and stored in liquid N₂ until data collection.

Structure Determination. X-ray diffraction data were collected at the Elettra Synchrotron source (Trieste, Italy) by using an x-ray wavelength λ = 1.2 Å. At 100 K, the NgbCO crystals diffracted up to 1.7-Å resolution on a MarCCD detector (MAR Research, Hamburg). The crystals were isomorphous with the metNgb crystals measured previously (3). They belong to space group R32, with cell parameters a = b = 83.37 Å and c = 110.98 Å. Data were indexed and scaled by using the programs denzo and scalepack (21); the CCP4 program suite (22) was used for subsequent data manipulation and structure refinement.

The structure of NgbCO was determined by Fourier difference methods by using the metNgb structure at 1.5 Å (3). After rigid body refinement of the initial model, electron density maps were computed and displayed with quanta (Accelerys, San Diego). New features appearing in the structure, including the heme-bound CO and 150 water molecules, were fitted manually. The model was subjected to multiple rounds of refinement by using refmac (23). The final model contains most of the protein sequence (amino acids 3–148) and 116 water molecules (Table 1). The geometry is consistent with the resolution, with 95.3% of the residues within the most favored region of the Ramachandran plot, 3.9% in allowed positions, and only one residue in generously allowed regions [computed by using procheck (23)]. The C and N termini of the protein did not appear in the electron density maps. Consequently, the initial and terminal residues in the polypeptide chain show B factors higher than the rest of the protein.

Table 1. Summary of NgbCO structure analysis.

Diffraction data
Wavelength, Å	1.2
Space group	R32
a = b, Å	88.37
c, Å	110.98
Resolution range, Å	20.0-1.70 (1.73-1.70)
Unique reflections	17,586 (880)
Completeness, %	95.9 (97.5)
Redundancy	5.2 (3.2)
Rmerge(overall)*	0.061 (0.226)
〈l〉/〈σ〉†	20.8 (2.94)
Refinement
Rcryst	0.22
Rfree	0.26
No. of protein atoms	1,223
No. of water molecules	116
rms deviation from ideality
Bonds, Å	0.016
Angles, °	2.51
Average B value, Å
All atoms	29.2
Main chain	25.6
Heme	30.8
Side chains and water	32.1

Numbers in parentheses refer to the highest resolution shell.

^*R_merge = Σ_hkl Σ_j|l_j(hkl) — 〈l(hkl)〉|Σ_hkl Σ_j〈l(hkl)〉, with l_j(hkl) representing the intensity of measurement j and 〈l(hkl)〉 the mean of measurements for the reflection hkl.

^†〈l〉/〈σ〉 = ratio between the mean intensity and the mean error of the intensity.

Careful visual inspection of the electron density maps of NgbCO indicates that a minor fraction (<10%) is still in the unligated state. This finding is only evident from a clear positive density observed at 3 σ, centered at the position occupied by the iron in the unligated structure. Corresponding features from the lighter atoms remain undetected in the electron density map because of their lower electron densities.

FTIR Spectroscopy. For the FTIR experiments, lyophilized protein was dissolved in 75%:25% (vol/vol) glycerol/potassium phosphate buffer to a final concentration of ≈10 mM. Solutions were equilibrated with 1 bar of CO and reduced by adding a 2-fold molar excess of sodium dithionite. The solution was centrifuged before loading into the sample cell to remove any undissolved protein. Difference spectra were calculated from FTIR spectra taken at 3 K before and after photolysis with visible light at 532 nm. Data collection procedures have been described in detail (12).

Results

General Features of the Structure. The overall tertiary fold of murine Ngb is only moderately affected by iron reduction and CO binding (Fig. 1), which is in line with observations on swMb and other monomeric globins (24, 25). A slight movement of the entire NgbCO molecule within the unit cell was readily adjusted with a few rounds of rigid body model refinement. However, shifts of individual tertiary elements, such as helix F and turns CD, EF, and FG as well as single amino acid side-chain rearrangements, are clearly evident.

Fig. 1.

Stereo diagrams of the structures of NgbCO (red) and unligated metNgb (magenta) in ribbon representation. (A) Overall structures, the distal histidine (His-64), the proximal histidine (His-96), Phe-106, the heme group, and loops EF and FG are highlighted. Green regions indicate the positions of the two point mutations, Cys-55–Ser and Cys-120–Ser. (B) Close-up view of the active site, also including Tyr-44 and Phe-106.

To assess similarities between the structures of NgbCO on one hand and unligated Ngb and swMb on the other, we carried out structural superpositions by automatic C_α alignments. To quantify structural differences between each pair, we calculated rms displacements (rmsds) of the C_α atoms from each structure; they are plotted in Fig. 2A as a function of residue number. The structural changes in swMb and Ngb caused by ligand binding are small; the overall rmsds averaged over all residues are 0.49 and 0.54 Å, respectively. The structures of NgbCO and sperm whale CO-ligated Mb (swMbCO) are slightly closer together than the unliganded forms (rmsd 1.85 vs. 2.0 Å), mainly because of a better superposition of the E and EF segments. In detail, however, a comparison between Ngb and Mb reveals larger differences. In Fig. 2B, we have plotted rmsds of the C_α atoms of metNgb and NgbCO as obtained from isotropic B factor refinement of each individual structure (B = 8 π² rmsd²). They represent (static and dynamic) structural variability of the atoms around their mean positions. For the majority of residues, the rmsds are significantly lower in the CO-ligated derivative (Fig. 2B, solid line), implying a better defined structure. The overall rmsds, averaged over all atoms, are 0.72 and 0.61 Å for metNgb and NgbCO, respectively.

Fig. 2.

Plots of rmsds versus primary sequence position. (A) rmsds between C atoms of pairs of structures after optimal alignment of swMbCO and deoxy swMb (gray dashed line), NgbCO and metNgb (gray solid line), NgbCO and swMbCO (black dashed line), and metNgb and met swMb (black solid line). (B) rmsds representing disorder within a single structure estimated from isotropic B factor refinement of the structures of NgbCO (solid line) and metNgb (dashed line). Bars and capital letters indicate the helical segments in murine Ngb.

Heme Displacement upon Ligand Binding. CO binding induces a major heme displacement within the frame of the globin fold (Fig. 1), which may define a new mechanism for ligand affinity modulation in hexacoordinated globins. The average rmsd of all heme atoms is 4.5 Å. The maximal rmsd within the porphyrin macrocycle is observed for the CHC atom (3.04 Å), whereas the heme iron undergoes a 2.0-Å displacement. Apparently, when CO binding releases the positional constraint imposed by the Fe–His-64 distal bond, the heme is free to tilt toward the distal side and slide deeper into its crevice, relaxing within the globin frame in a position closer to that found in swMb (24, 25). Similar movements of the heme group were noted in the x-ray structures of two Hbs from the sea cucumber Caudina arenicola, the low spin hemichrome of the monomeric Hb-C chain, and the cyanomet dimeric Hb-D chain (26, 27); however, in these globins, the ferrous heme iron is not hexacoordinate, unlike in Ngb.

In metNgb, the heme internal rim contacts are less extensive than in other globins, to the point of being permissive of heme orientational disorder within its binding pocket (3, 28). The NgbCO structure unveils the functional significance of this peculiar feature: the absence of close contacts on the CHC side (opposite to the propionates) is essential to allow the heme sliding movement upon ligand binding; thereby, the heme group forms a number of new van der Waals contacts (with Val-101, Phe-106, and Met-144). Very unexpectedly, the proximal branch of the large cavity observed in metNgb disappears (Fig. 3), and the huge tunnel acquires a partially unique topology. Comparison of the overall B factors (rmsds) of metNgb and NgbCO suggests that rupture of the Fe–His-64 bond and the subsequent docking of the heme in its new position represent a relaxation of Ngb to a more stable structure. In particular, we notice a decrease of the rmsds, which is substantial in the EF loop and just significant in the CD loop (Fig. 2B).

Fig. 3.

Cavities in NgbCO (blue) and unligated metNgb (yellow) as determined by using surfnet (46). The position of Ser-55 is highlighted in green.

Consequences of Heme Sliding. Repositioning of the heme group coupled to CO binding affects the conformation of some adjacent amino acids (Fig. 1). Surprisingly, the distal His-64 C_α moves by only 0.5 Å and its side chain N_ε atom by 0.7 Å. The moderate increase of its side-chain B factor is consistent with the rupture of the covalent bond to the heme iron. By contrast, the proximal His-96 remains bound to the heme iron. Its C_α atom shifts by 0.83 Å, its N_ε shifts by 1.5 Å, and the N_ε-Fe-NA angle changes from 84.6° to 88.6°, which is closer to ideal geometry. Consistent with the scenario of heme sliding into a preformed docking site, only Phe-106 is drastically repositioned (Fig. 1), with its C_α shifting by 1.26 Å and its side-chain apical carbon (C_ξ) shifting by 3.83 Å; this shift is accompanied by a rearrangement of the preceding FG loop. The His-96 displacement causes a shift of the F helix that is propagated to the EF loop. Moreover, a shift of Tyr-148 by 1.2 Å is made possible by the flexibility of Gly-147.

In addition to these protein rearrangements in the EF-F-FG region, the CD loop moves to a more “open” conformation upon ligand binding (Fig. 1 A). Modeling of exogenous ligand binding based on the structure of murine metNgb (3) led us to propose that the incoming ligand could access the binding site if the distal histidine were to swing either deeper inside or outside of the distal cavity. The unexpected repositioning of the entire heme group omits the necessity of the swinging motion; still, the minor movement observed for His-64 is in an outward direction, inducing the CD loop displacement via the contact His-64–Tyr-44. Thereby, the network of electrostatic interactions (3) involving Lys-67, one of the heme propionates, and the Tyr-44 hydroxyl group is weakened, as indicated by the increase of the relevant interatomic distances (from 2.7 to 4.57 Å for Lys-67/propionate, and 2.6 to 3.0 A for Tyr-44/propionate, Fig. 1B).

Structural Changes in the Internal Cavities. The large (≈287 ų) internal cavity observed in metNgb (2, 3) is reshaped upon CO binding (Fig. 3). The branch at the proximal heme side is occupied by the heme and Phe-106, leaving only a small niche (8.2 ų) close to the proximal His-96. The distal cavity branch is deepened and forms a small open space (13.3 ų) lined by residue Ser-55 (a Cys in WT Ngb). Reshaping of the large tunnel causes a miniscule increase of its total volume to 306 ų. It still connects to the surface of the protein via a channel lined by Leu-70, Val-71, Ala-74, Leu-85, and Tyr-88, as in metNgb. The pronounced decrease of the rsmds in the EF loop (Fig. 2B), around which these residues are located, could, however, restrict its accessibility from the exterior. The cavity corresponding to the Xe4 site in swMb (9) is not modified appreciably, whereas the site which in metNgb is occupied by a water molecule just next to the distal His-64 has disappeared.

FTIR Spectroscopy. Fig. 4A shows FTIR absorbance difference spectra of the heme-bound CO in WT NgbCO at two different pH values. At pH 7.3, we observe essentially two CO stretching bands, A₁ at 1,938 cm^–1 and A₂ at 1,980 cm^–1. They are associated with two different heme pocket conformations, as reported from Raman studies (29). At pH 5.3, two additional bands become apparent, A₀ at 1,968 cm^–1 and A₃ at 1,923 cm^–1. Mutant Lys-67–Leu (Fig. 4B) also shows multiple stretching bands. At pH 7.3, three bands at 1,933, 1,947, and 1,960 cm^–1 can be identified, whereas at low pH, bands can also be seen at 1,920 and 1,975 cm^–1. A₁ and A₀ dominate the spectrum, with an additional minor contribution of A₃. Replacement of His-64 by Leu results in a more homogeneous environment of the bound CO, as indicated by only a single band at 1,972 cm^–1 (Fig. 4C); an identical spectrum is observed for the double mutant His-64–Leu–Lys-67–Leu.

Fig. 4.

FTIR absorbance difference spectra of the stretching bands of hemebound CO in WT Ngb at pH 7.3 (solid line) and pH 5.3 (dashed line) (A), mutant Lys-67–Leu at pH 7.3 (solid line) and pH 5.5 (dashed line) (B), and mutants His-64–Leu (pH 7.4) and His-64–Leu–Lys-67–Leu (pH 7.3) (C). Difference spectra were calculated from absorbance spectra taken before and after illumination with 532-nm light at 3 K.

The widths of the CO stretching bands of WT and mutant Ngb are substantially larger than their counterparts in swMbCO, which suggests pronounced structural heterogeneity for the heme pocket of NgbCO. At pH 7.3, the full width at half maximum of the A₁ band in NgbCO is ≈14 cm^–1, which is almost twice the value obtained for the same band in swMb (at pH 7 and 15 K) (30).

Discussion

Structural Heterogeneity at the Active Site. Multiple CO stretching bands in the FTIR spectra are clear evidence of different active-site conformations of WT NgbCO (Fig. 4). They are known to arise from differing electrostatic interactions between the CO molecule and its environment (30–35). The A₁ band at ≈1,940 cm^–1 is typical of a globin in which the His-64 imidazole is protonated at N_ε2; the positive partial charge near the CO oxygen decreases the bond order and causes a lower stretching frequency. By contrast, the higher stretching frequency of A₂ at 1,980 cm^–1 can be explained by a negative partial charge near the CO oxygen, which can arise from the tautomeric state in which the N_δ1 atom of the imidazole is protonated, leaving a lone pair at N_ε2. When lowering the pH, the absorbance of A₀ increases at the expense of A₁ and A₂. A similar increase in the A₀ population has been observed for swMbCO, where it has been assigned to protonation of the His-64 side chain with concomitant swinging out of the heme pocket to better solvate the charge. Indeed, the stretching frequency of A₀ near ≈1,970 cm^–1 is characteristic of mutants in which the His is replaced by a small, unpolar side chain (35). Moreover, a pK_a of ≈4.5 has been reported for His-64 in both swMbCO (35) and NgbCO (6). The low frequency of the A₃ band in Fig. 4A indicates an even stronger interaction of the CO with a positive charge than in A₁. A possible explanation for the appearance of A₃ at low pH may be that a fraction of the protonated His remains in the distal heme pocket.

Structural heterogeneity persists for mutant Lys-67–Leu, but the CO stretching spectrum is shifted to lower frequencies. In this mutant, the pK of His-64 is 5.6 (6), which implies a more hydrophilic distal pocket; solvent accessibility may be enhanced because of the lacking side chain of Lys-67. In WT Ngb without a bound exogenous ligand, this side chain is held in place by interactions with the heme propionate and Tyr-44 to sequester the distal heme pocket from the solvent. The appearance of a single IR stretching band in mutant His-64–Leu (Fig. 4C) provides clear evidence that His-64 is responsible for the spectroscopically detected structural heterogeneity in the active site of Ngb.

How are these heterogeneities in the spectroscopic data related to structure? The x-ray structure of murine metNgb shows two discrete heme conformations, with a population ratio of 70:30 (3). NMR studies have shown earlier that heme orientational disorder changes from ≈2:1 in metNgb to ≈1:1 upon binding of cyanide (28); on the other hand, heme heterogeneity was not resolved in the x-ray structure analysis of the CO-ligated species, possibly because of slightly lower resolution. Because the crystals were grown at pH 6, the A₁ and A₂ conformations (likely related to His-64 tautomers) should be predominant in the crystal, but they remain undetected at the present resolution. There is further heterogeneity within each of the discrete distal pocket conformers, as evidenced from the heterogeneous broadening of the CO stretching bands. The larger widths in comparison to swMbCO suggest that this heterogeneity is more pronounced in NgbCO.

Ligand Binding and Dissociation in metNgb. Whereas the ferrous heme iron is pentacoordinate in vertebrate Mbs and Hbs (13), in both metNgb and deoxy Ngb hexacoordination prevails (5, 6, 29, 36). Thus it was intriguing that metNgb binds O₂ reversibly and with a fairly high affinity (P₅₀ ≈2 torr), despite its hexacoordinate deoxy state.

Laser photolysis experiments have shown that ligand binding to pentacoordinate metNgb is extremely fast (5, 12), the bimolecular association of WT Ngb being close to the diffusion limit. Spectroscopic studies have shown that the association rate coefficient is finely tuned by pH. The pK_a of His-64 changes from 4.5 for the CO-ligated form to 6.0 for the pentacoordinate WT protein (6); the latter pK_a suggests that His-64 and hence the distal pocket is more solvent accessible. Therefore, gaseous ligands will likely experience little, if any, resistance by the protein to entering the distal pocket; there, they have a high probability of binding because of a very low enthalpy barrier at the heme iron (6, 12).

At normal tissue O₂ pressures, the majority of ligands that have dissociated from the heme iron will quickly rebind, whereas the heme is expected to reside deeply inside its crevice. In a small fraction of Ngb molecules, His-64 succeeds in competing for the sixth coordination, which requires that the heme shifts into the position observed in the met form. Only under conditions of severe hypoxia (or upon autoxidation of the heme iron) will the heme slide back to enable the His-64 imidazole side chain to block the sixth coordination. Thereafter, exogenous ligand binding has to wait for the distal His-64 to thermally dissociate from the heme iron (5, 12, 18, 36). For murine Ngb, the His-64 dissociation rate coefficient, k_off(His) ≈1 s^–1, at room temperature is much smaller than the one for O₂, k_off(O₂) ≈200 s^–1, but still much larger than k_off(NO) ≈0.0002 s^–1 and k_off(CO) ≈0.007 s^–1 [the latter was determined only for human Ngb (36)]. The propensity of the heme to slide deeper into the pocket destabilizes His-64 binding and thus increases k_off(His). Association and dissociation of exogenous ligands (and thus the affinity) are controlled by these conformational changes, especially the (transient) coordination of His-64 to the ferrous heme iron.

Structural Mechanism of Ligand Control. To be physiologically effective, globins have to adjust their affinities to be capable of binding and releasing O₂ efficiently in response to metabolic demands. Globins operate under a wide variety of conditions and play different metabolic roles in vertebrates and plants, including enzymatic reactions (37–40), and thus display a high degree of functional plasticity. Affinity tuning can be achieved through ligand stabilization or destabilization by the environment provided in the distal pocket, by gating ligand access to the binding site, as well as by creating or deleting docking sites and cavities within the protein moiety (41, 42). The discovery of Ngb and cytoglobin (refs. 20 and 43 and references therein) with their unusual internal hexacoordination has started a search for their specific functions. To test the numerous hypotheses put forward to unveil the in vivo function of Ngb, a thorough functional and structural characterization is mandatory.

The structure of NgbCO reveals how the unusual cytochrome-like iron-distal His coordination is made sufficiently labile to allow binding of exogenous ligands. A primary factor that counterbalances the presence of internal coordination (which a priori should be fairly stable) seems to be the docking of the heme in a new position, leading to an overall more stable protein conformation with drastically reduced rmsds in the EF loop (Fig. 2B). The coupling of heme relocation to reshaping of the large internal tunnel, accompanied by a substantial restriction of the connecting path to the bulk, might be relevant to more complex functions of Ngb. The conformational changes coupled to ligand binding may allow internal trapping of further, possibly different, exogenous ligands to enhance multiple collisions and possibly more complex chemistry. The opening of an additional cavity lined by Ser-55 (Cys in the WT protein) in close proximity to the large open space is worthy of a further remark. It was reported recently that formation of an internal disulphide bond (Cys-55–Cys-46) in human Ngb raises O₂ affinity by 10-fold (44). This effect, however, cannot be extended to murine Ngb, where Cys-46 is absent. Nevertheless, the conservation of Cys-55 and its involvement in cavities dynamics suggests a role in the as-yet-undetermined mechanism by which Ngb protects the nervous tissue from hypoxic damage.

A final point emerges from examining the NgbCO structure. Previously, we had proposed that the large internal tunnel connecting both sides of the heme might allow alternative ligand binding to the proximal side of the heme (3). However, in the CO adduct, the bond between the heme iron and the proximal His-96 is maintained (essentially unaltered); thus, the possibility that an exogenous ligand (notably NO) may bind to the proximal side seems ruled out.

Physiological Implications. Ngb, with its P₅₀ of ≈2 torr, is most likely reduced and ligated by O₂ under physiological conditions (normoxia). Under O₂ deficiency, Ngb will release its hemebound ligand and adopt the hexacoordinate deoxy structure; the same coordination change is expected upon autoxidation and release of O^–_2. The rmsd values in Fig. 2B indicate that exogenous ligand release will be accompanied by a pronounced increase in the mobility of the EF loop. The loosening of the structural constraints in the EF loop may be the signal encoding the information “low oxygen.” This signal may be transmitted further by interaction with GDP-bound G_αi and thereby liberating G_βγ, which protects the cell against neuronal death (17). Increased Ngb expression, as observed under conditions of hypoxia (14), will enhance this signaling pathway. Moreover, hypoxia is known to reduce the pH of a neuronal cell (45), which is significant because acidity decreases the rate of exogenous ligand binding to the pentacoordinate species, favoring hexacoordination (6), and thereby the coupled conformational change involving the EF loop.

Conclusions

The 3D structure of murine NgbCO presented here, together with the metNgb structure from the same species reported previously (3), allows us to assess the conformational changes that accompany exogenous ligand binding in Ngb. Based on this comparison, some conclusions can be drawn regarding the mechanism of ligand affinity regulation in hexacoordinate heme proteins. Given that the distal histidine efficiently blocks the binding site in the unligated state, we believe that the major heme displacement occurring upon exogenous ligand binding will assist rupture of the distal His–iron coordination bond, facilitating O₂ binding. Moreover, heme sliding is accompanied by a dramatic topological reorganization of the large internal cavity and a marked decrease of the overall rmsd; locally, the largest decrease of the rmsds is observed for the EF loop. This pronounced change in flexibility and conformation could represent the molecular basis of the signaling function of Ngb in the case of hypoxia.