Heme orientation modulates histidine dissociation and ligand binding kinetics in the hexacoordinated human neuroglobin
Neuroglobin (Ngb) is a globin present in the brain and retina of mammals. This hexacoordinated hemoprotein binds small diatomic molecules, albeit with lower affinity compared with other globins. Another distinctive feature of most mammalian Ngb is their ability to form an internal disulfide bridge that increases ligand affinity. As often seen for prosthetic heme b containing proteins, human Ngb exhibits heme heterogeneity with two alternative heme orientations within the heme pocket. To date, no details are available on the impact of heme orientation on the binding properties of human Ngb and its interplay with the cysteine oxidation state. In this work, we used 1H NMR spectroscopy to probe the cyanide binding properties of different Ngb species in solution, including wild-type Ngb and the single (C120S) and triple (C46G/C55S/C120S) mutants. We demonstrate that in the disulfide-containing wild-type protein cyanide ligation is fivefold faster for one of the two heme orientations (the A isomer) compared with the other isomer, which is attributed to the lower stability of the distal His64–iron bond and reduced steric hindrance at the bottom of the cavity for heme sliding in the A conformer. We also attribute the slower cyanide reactivity in the absence of a disulfide bridge to the tighter histidine–iron bond. More generally, enhanced internal mobility in the CD loop bearing the disulfide bridge hinders access of the ligand to heme iron by stabilizing the histidine–iron bond. The functional impact of heme disorder and cysteine oxidation state on the properties of the Ngb ligand is discussed.
KeywordsNMR Neuroglobin Molecular dynamics Heme orientation Disulfide bridge
Similarly to other globins, small ligands (O2, NO, CO) can bind deoxyNgb (Fe2+), whereas the metNgb (Fe3+) form is able to bind small ligands such as CN−, NO, and N3 − . However, the low affinity for O2 precludes a role as an O2 carrier [3, 4, 5, 13]. The function of Ngb has not been fully resolved, and different roles for Ngb have been proposed in the past [2, 3, 4, 5, 13, 14, 15, 16, 17]. For example, its potential interaction with the NO radical led to a neuroprotective role in the case of oxidative stress in cells being proposed [18, 19, 20]. Furthermore the expression of Ngb increases during hypoxia, which may reveal the action of Ngb as a radical scavenger [2, 3, 4, 5].
Despite the low (25 %) sequence homology between Ngb and other members of the globin family, Ngb shares the highly conserved amino acid positions characteristic of the globin fold [21, 22, 23]. Not surprisingly, the crystal structure of Ngb [22, 23] revealed strong structural similarities between Ngb and other globins. Similarly to a few other globins such as cytoglobin and nonsymbiotic plant hemoglobins, Ngb possesses two key histidine residues in the iron coordination sphere, including the distal His64 (E7) and the proximal His96 (F8) that close the coordination sphere of the heme iron atom, resulting in a hexacoordinated iron even in the absence of an external ligand. This situation contrasts with that in most globins, for which only the proximal histidine is coordinated to the iron and the distal pocket is directly available to host an external diatomic ligand without the requirement for conformational change. It has been proposed that in Ngb the distal histidine has first to be dissociated to make the distal pocket accessible to the incoming ligand [24, 25]. A recent experimental study on mouse Ngb  and a theoretical study on human Ngb  also suggested that His64–iron bond breaking is accompanied by heme sliding instead of histidine side chain rotation.
In the case of globins, protoheme IX is not directly linked to the peptidic backbone. Owing to the lack of true twofold symmetry, heme b often adopts two stable conformations, A and B, in the heme pocket; they differ by 180° rotation around the α–γ meso axis (see the vertical dashed line in Fig. 1a). For example, the M1 methyl (and V2 vinyl) group in the A form occupies a spatial position similar to that of the vinyl V4 (and M3 methyl, respectively) group in the B form, and vice versa. The ratio of A and B populations is highly sensitive to the composition of the heme pocket and thus depends on protein sequences and species. For example, the crystal structures of mouse Ngb (Protein Data Bank entries 1Q1F and 2VRY) contain two heme molecules refined as 70 and 30 % occupancy and labeled as A and B, respectively . The two stable orientations of heme are also reflected in the 1H NMR spectra of mouse and human Ngb, with a major (70 %) B conformation and a minor (30 %) A conformation [28, 29, 30]. As a source of confusion , different conventions were used for the NMR and crystallographic studies. In this article, we use the same nomenclature as previous NMR studies of Ngb. Following this naming convention, the groups contacting the Phe42 side chain are M8/M1 in the B conformer and M5/V4 in the A conformer.
Another unique characteristic of human Ngb is the existence in the CD–D region of two cysteine residues, Cys46 (CD7) and Cys55 (D5). The first experimental evidence for the existence in human Ngb of an internal disulfide bridge involving these cysteine residues was reported in [32, 33]. Remarkably, an increase in O2 affinity [32, 33] and CO rebinding rate  was observed upon formation of the bridge, suggesting that cysteines in human Ngb may act as a strong regulator of O2 binding and may be used as redox sensors. Unfortunately, no high-resolution structure is yet available for human wild-type (WT) Ngb with the internal disulfide bond. However, the crystal structures of mouse Ngb and the cysteine-depleted human Ngb, both lacking the crucial cysteines, as well as molecular dynamics (MD) simulations [27, 35] suggest the requirement for a local rearrangement in the CD–D region located between the two α-helices C and D to accommodate the formation of the internal disulfide bridge. The cysteine redox-dependent ligand binding properties of Ngb were then proposed to be associated with the large conformational change of the CD loop required to bring the thiol groups in close proximity [27, 35]. Our understanding of the conformational change and of the precise role of the cysteine redox state in human Ngb function is still limited, emphasizing the need for better characterization of Ngb with oxidized (WTox) or reduced (WTred) cysteines. Mammalian Ngb sequences are highly conserved (more than 90 % sequence identity). However, in rodent Ngb, a conserved glycine is systematically found at the position equivalent to Cys46 in the human Ngb sequence and represents the sole substitution in the otherwise strictly conserved CD loop. Hence, despite bearing the conserved cysteines at positions 55 and 120 (in the human sequence), rodent Ngb does not have the ability to create an internal disulfide bridge, which may reflect possible functional differences between rodents and other mammals.
Similarly to many hemoproteins, the Ngb protein is capable of existing in several stable states characterized by the oxidation state of the heme iron, the heme orientation, and the nature of the sixth ligand of the iron (distal histidine or external small ligand). The possibility to form an internal disulfide bridge is unique to nonrodent mammalian Ngb compared with other globins and further increases the structural diversity. A growing body of evidence has been gathered in the last decade demonstrating the profound influence of the conformation of Ngb on its ligand binding properties [32, 33, 34], but a systematic study was still lacking. In this work, we explored the spectroscopic properties of the various human Ngb forms by 1H NMR spectroscopy. The excellent spectral dispersion due to the paramagnetic high-spin ferric ion (Fe3+) and its ability to directly monitor various species in solution that cannot be purified to homogeneity, such as the slowly interconverting heme orientations, establish NMR spectroscopy as a convenient tool to assess the influence of the disulfide bridge and/or the heme orientation on the binding kinetics and thermodynamics of Ngb with the small-ligand cyanide anion. The Fe3+–CN− complex is believed to be isostructural and isoelectronic to the physiologically relevant Fe2+–CO complex and its very slow binding permits time-resolved NMR investigations, which are impossible in the case of the (fast) binding of CO or O2. Furthermore, we provide new MD trajectories highlighting the role of dynamics in the ligation mechanism. Beyond the specific study of Ngb, the understanding of the influence of heme orientation on protein reactivity will extend our knowledge of hemoproteins.
Materials and methods
Cloning, expression and purification of recombinant Ngb
Recombinant human WT Ngb was cloned into the expression vector pET15b and overexpressed in Escherichia coli BL21 (DE3). The sequence of Ngb encompassed the six His-tag at the N-terminal extremity followed by Leu-Val-Pro-Arg-Gly-Ser. The overexpression and the preparation of a crude Ngb extract was performed as described by Dewilde et al. . The proteins were purified on a nickel nitrilotriacetic acid resin column using an equilibrium buffer [50 mM tris(hydroxymethyl)aminomethane (Tris)–HCl pH 8.0, 8 mM imidazole] and an eluent buffer (50 mM Tris–HCl pH 8.0, 150 mM imidazole). The His-tag of the different samples was cleaved using a thrombin kit (Sigma) and concentrated in 20 mM Tris–HCl pH 7.4. Ngb samples were stored at 193 K prior to use. For Ngb mutants, the QuikChange site-directed mutagenesis method was used to generate the C120S single mutant and the C46G/C55S/C120S triple mutant starting from human Ngb as a template.
Ngb samples were prepared in 20 mM Tris–HCl buffer at pH 7.4 at 100 μM protein concentration and 90/10 % H2O/D2O. NMR experiments were performed at 298 K using a Bruker AVANCE II 600 MHz spectrometer equipped with a TCI cryoprobe. Data processing and analysis were performed using NMRpipe  and CCPNMR  software programs, respectively. The CN− binding experiment was initiated by the addition of 500 μM KCN to the protein solution and followed by a series of 1D 1H spectra. Solvent peak suppression was achieved using a water on-resonance 1-s presaturation. The spectral width was set to 80 ppm. One 1D spectrum was typically collected every 20 min (1,024 transients) to follow spectral changes. The KCN binding experiment was performed on the WT, C120S, and triple-mutant proteins. For the WT and C120S proteins, the internal disulfide bridge was either oxidized or reduced prior to KCN addition. The reduction of the disulfide bridge was obtained by adding 2 mM dithiothreitol (DTT) to the protein solution and the excess DTT was removed by a three-step dilution/concentration method using Vivaspin centrifugal concentrators (Sartorius Stedim, M r = 10,000, 3,000g). The dilution step was achieved using the 20 mM Tris–HCl (pH 7.4) buffer.
MD simulations were performed with the Gromacs software package [39, 40, 41] using the OPLS all-atom force field . The starting coordinates employed for the simulations were taken from the experimental X-ray structure of the cysteine-depleted human Ngb at 1.95-Å resolution (Protein Data Bank entry 1OJ6, B chain) . The charges and parameters for the prosthetic group were previously described by Bocahut et al. . The protein was solvated in a 78-Å cubic box, using periodic boundary conditions, with explicit single-point charge  water molecules; six Na+ ions were added to neutralize the system, which contained a total of around 48,000 atoms. The dynamics was performed at 1 atm and 300 K, maintained with the barostat and thermostat of Berendsen et al. . Long-range electrostatic interactions were treated using the particle mesh Ewald method , with a grid spacing of 0.12 nm and a nonbond pair list cutoff of 9.0 Å with updating of the pair list every five steps. We chose a time step of 2 fs by constraining bond lengths involving hydrogen atoms with the LINCS algorithm . The solvent was first relaxed by an energy minimization followed by a 100-ps equilibration step under restraint, and then heated slowly until a temperature of 300 K was reached for the system; 12-ns production runs were eventually performed, from which the last 10 ns was kept for analysis. In particular, the MD trajectory was investigated using principal component analysis [46, 47, 48, 49] on the first eight normal modes of the protein in order to retrieve the most significant fluctuations occurring along the collective modes of motion of Ngb.
NMR characterization of the oxidized WT Ngb (WTox)
NMR characterization of the DTT-reduced WT Ngb (WTred)
To characterize the properties of Ngb with reduced cysteines, we incubated the Ngb sample with DTT and extensively washed the protein solution to remove oxidized or reduced DTT prior to NMR measurement. The reduction of the disulfide bridge was also checked by mass spectrometry. We observed a 2.4-Da increase in mass upon reduction by DTT, in agreement with the formation a single disulfide bridge in Ngb before DTT treatment and with the two additional hydrogens in reduced cysteines. This analysis was also consistent with a highly predominant monomeric form of the protein before and after DTT treatment. The 1H NMR spectrum of Ngb subjected to this treatment is shown in Fig. 2, spectrum B and differed significantly from the 1H spectrum before DTT treatment (Fig. 2, spectrum A). Nevertheless, the 1H resonances could be easily assigned, including the M8B and M5A methyls, which shifted to 36.7 and 35.4 ppm, respectively, while retaining a constant population ratio of approximately 2:1.
Structural effects of disulfide bridge formation in WT Ngb
Most well-resolved heme resonances could be assigned in both WTred and WTox (see Fig. 2) using 2D nuclear Overhauser effect spectroscopy and total correlation spectroscopy experiments (data not shown) and on the basis of the assignment of the mouse  and human  Ngb. Heme resonances shifted slightly upon cysteine reduction, indicating that the loss of the disulfide bridge perturbs the paramagnetic contact and pseudocontact shifts that dominate heme chemical shifts. In the absence of experimental structures for the human WT Ngb, the chemical shift analysis suggests a significant structural rearrangement in the heme vicinity upon cysteine oxidation. Nevertheless, the very similar overall pattern of the heme 1H spectrum in the presence or absence of disulfide bridge is consistent with only a limited structural change for the heme. As an illustration, the conserved order and relative distance for heme methyls indicated mostly unaffected nodal plane orientation for the iron ligands His96 and His64 [47, 48].
NMR characterization of C120S and C46G/C55S/C120S Ngb mutants
Ngb contains three cysteines, and Cys120 is suspected to form intermolecular bridges. To probe such a possibility, the C120S mutant was produced. The 1H NMR spectra of this mutant in oxidized and reduced states (Fig. 2, spectra C and D) were essentially indistinguishable from those of their WT counterparts. This observation confirmed that a cross-linked Ngb monomer, if it exists, does not significantly contribute to the NMR spectra of WT Ngb, consistent with the findings of mass spectrometry. In addition, this demonstrated the intact heme environment upon mutation in line with the location of Cys120 in the remote GH loop, 17 Å from the heme. To understand how chemical modifications in the CD loop may affect protein function, we also produced the cysteine-free triple mutant (C46G/C55S/C120S). The crystal structure of this mutant is available , and despite a recent report on WT Ngb crystallography , it is still the sole experimental structure available for human Ngb. As visible in Fig. 2, spectrum E, the 1H NMR spectrum of triple-mutant Ngb was globally very similar to the spectra of other Ngb sequences, indicating a conserved global fold in triple-mutant Ngb. Notably, the closest similarity was observed with WTred and reduced C120S Ngb, in line with the absence of a disulfide bridge in these species. Nevertheless, resolved heme resonances in triple-mutant Ngb slightly shifted with respect to those in WTred, indicating significant but limited reorganization in the heme vicinity upon mutations in the CD loop. As a conclusion, chemical modifications in the CD–D region bearing Cys46 and Cys55, such as oxidation of cysteines or amino acid substitution, consistently modify the heme environment. In contrast, the heme environment and structure are not sensitive to amino acid substitution in the GH loop bearing Cys120.
Kinetics of cyanide binding to WTox
Association rate constants for binding of cyanide to neuroglobin (Ngb)
k obs (M−1 s−1)
WTox (A isomer)
WTox (B isomer)
WTred (A isomer)
0.26 (K d = 153 μM)
WTred (B isomer)
0.12 (K d = 316 μM)
C120Sox (A isomer)
C120Sox (B isomer)
C120Sred (A isomer)
C120Sred (B isomer)
TM (A isomer)
0.12 (K d = 2 mM)
TM (B isomer)
0.14 (K d = 5 mM)
Decreased binding rates upon reduction of cysteines
To obtain more insight into the influence of the disulfide bridge on ligand binding kinetics in human Ngb, we also monitored the kinetics of cyanide binding to a cysteine-reduced Ngb (WTred) sample in the same conditions as for WTox (Fig. 4b). During this long cyanide binding experiment, the absence of reducing agent in solution such as DTT might result in air reoxidation of the disulfide bridge that may interfere with the binding process. The rate of air reoxidation was estimated to be about 60 h from a similar experiment where cyanide was omitted and by following the increasing peak intensity of the WTox form. We concluded that the kinetics represented in Fig. 4b unambiguously reported on cyanide binding to WTred.
As illustrated in Fig. 4b, both the A isomer and the B isomer showed very slow binding in WTred, with time constants of 127 and 265 min, respectively. In the absence of a disulfide bridge, the A isomer also bound cyanide at a faster rate than the B isomer. When compared with WTox, the on-rate values decreased by a factor of 5 and 2.5 for the A isomer and the B isomer, respectively, upon cysteine reduction, suggesting that the disruption of the disulfide bridge globally decelerates cyanide ligation. At equilibrium, and despite the excess of CN− anion in solution, neither of the isomers were saturated in cyanide, and about 42 % of the A isomer and 24 % of the B isomer remained unbound. Thus, reduced cysteines in Ngb are associated with decreased affinity for cyanide.
The dissociation constants K d,A and K d,B for dissociation of cyanide from the A isomer and the B isomer, respectively, were calculated to 153 and 316 μM, respectively, on the basis of the concentrations of the free and bound species at equilibrium. The ratio K d,B/K d,A was then estimated to be 2.1, which is very similar to the k obs A /k obs B ratio of on-rate values (Table 1), calculated to be 2.2. This analysis demonstrated the very similar off-rate values for both complexes, estimated to be 3.9 × 10−5 s−1. The increased affinity of the A conformer versus the B conformer might then be explained by a faster on rate in the A isomer, suggesting easier access to heme iron or enhanced reactivity of the heme iron in this conformation.
Cyanide binding to triple-mutant Ngb and C120S Ngb
Cyanide binding experiments were also conducted on the C120S mutant in the cysteine oxidized and reduced states. The time constants are reported in Table 1 and were almost identical to the values measured on the WT protein. As a consequence, the replacement of a sulfur atom by an oxygen atom at position 120 in the Ngb sequence has a negligible effect on the kinetics and thermodynamics of cyanide binding to Ngb, whether the internal disulfide bridge is formed or not. This observation parallels the similar heme environment in WT and C120S forms.
The analysis of 1H NMR spectra of the triple mutant suggested detectable conformational changes upon C46G/C55S double mutation near the heme pocket. To assess the impact of the mutation on the functional properties of the protein, we also monitored the reaction of potassium cyanide binding to triple-mutant Ngb (see Fig. 4c). In contrast to WT Ngb, quite similar binding time constants were obtained for the A isomer and the B isomer (223 and 262 min, respectively). The similar binding time constants for the A conformer and the B conformer in the triple mutant suggest that heme orientation had less impact on cyanide binding in this mutant. More importantly, and in line with the slower on rate, the affinity of triple-mutant Ngb for CN− was significantly reduced as judged from the incomplete binding at equilibrium (90 and 80 % of free B isomer and free A isomer, respectively, at the end of the reaction). K d and k off were evaluated to be 2.0 mM and 2.5 × 10−4 s−1, respectively, for the A isomer and 5.0 mM and 6.9 × 10−4 s−1, respectively, for the B isomer. Since the properties of cyanide binding to WT Ngb and the C120S mutant were highly similar, this analysis pointed to the major impact of C46G/C55S mutations in the CD loop on the cyanide binding kinetics and thermodynamics of Ngb.
Internal flexibility in the CD loop probed by MD
In this work, we assigned and analyzed most of the well-resolved heme NMR resonances of human WT Ngb in four stable states characterized by distinct heme orientations and cysteine oxidation states. To date no high-resolution NMR or crystal structure is available for human WT Ngb. The only human Ngb derived structure was obtained on the triple-cysteine-depleted mutant , and we also analyzed this triple mutant under the same sample conditions as the WT protein. The resonance assignment was greatly facilitated by the very similar heme 1H chemical shift pattern in all spectra, indicating that the heme environment is essentially conserved and that the WT protein and the triple mutant globally adopt similar structures. The population ratio of the A isomer and the B isomer, which is a sensitive probe of the local heme environment, was comparable in all protein states (WTox/WTred/triple mutant) and further reinforces the impression of a conserved heme environment.
The Cys46–Cys55 distance predicted from the structure of triple-mutant Ngb is not compatible with easy S–S bond creation, and the CD–D region requires a conformational transition to accommodate the new bond. This structural change may in turn be transmitted to the heme crevice as a slight rearrangement, such as heme reorientation or rotation of the g tensor. Indeed, subtle conformational differences in the heme vicinity could be detected on the basis of heme 1H chemical shift variations upon cysteine oxidization/reduction or depletion. Such a conclusion is also supported by recent EPR  and MD studies, and the Phe28 (B10) residue was proposed to mediate the communication between the CD–D region and the heme . There is now accumulating experimental and computational evidence that the dynamics of the CD–D region is profoundly enhanced upon disruption of the internal disulfide bond in WT Ngb [27, 52], suggesting that the cysteine redox state of the protein may influence protein function by modulating the flexibility and the conformation of the CD–D region. In this work we also found that the mutation of cysteine residues in the CD–D region resulted in heme displacement compared with WTred, in agreement with a previous Fourier transform IR study , further confirming the strong connection between the CD loop and heme conformation. Unfortunately, the absence of a high-resolution structure of WT Ngb excludes a more detailed description of the transition. We also demonstrated using MD simulations that the CD loop becomes even more flexible upon mutation of the cysteine residues to glycine or serine, which is likely due to the enhanced flexibility of glycine introduced at position 46. Of interest, mouse and human Ngb share high sequence identity (94 %), with an identical sequence in the CD loop except for position 46, which is occupied by a glycine in the mouse sequence and a cysteine in the human sequence. It is then expected that enhanced mobility in the CD loop might contribute to possible distinct functional properties for mouse and human Ngb.
Our understanding of how heme disorder affects ligand reactivity can now be greatly improved. As seen before, the apparent association rate (k obs) depends on the efficiency of the diffusion of the ligand towards the pentacoordinated heme iron (k on,L) and on the His64–iron bond stability (K H). In principle, heme heterogeneity may affect both parameters. Nevertheless, since the heme controls the precise position of the heme iron atom, it is expected that heme primarily governs the histidine–iron interaction, which was initially proposed to explain the heme-orientation-dependent ligation rate for mouse Ngb . Convincingly, different k −H values were extracted from the fast and slow phases of CO ligand binding to Ngb, whereas similar k on,L values were fitted for the two phases , suggesting that heme orientation essentially conditions the iron–histidine bond stability. In mouse Ngb, the dissociation of His64 from iron has been associated with heme sliding towards the bottom of the heme cavity, while keeping the His64 side chain at the same position . It is then likely that the energy barrier corresponding to heme sliding is lowered in the case of the A conformer, thus accelerating the histidine–iron bond rupture. For example, the steric hindrance in the vicinity of the Phe106 side chain, which exhibits the largest conformational change upon heme sliding, may play a significant role in controlling the structural transition. Indeed, this region is less crowded in the A conformer than in the B conformer.
A marked deceleration in cyanide complex formation was observed upon reduction of the disulfide bridge, which in turn was associated with reduced affinity. A similar observation was also reported for O2 and CO ligands [32, 34]. Therefore, faster ligand binding, and increased affinity for the ligand, appears to be a general hallmark of Ngb containing an internal disulfide bridge, irrespective of the type of ligand. The histidine–iron bond was measured to be stabler in the absence of a disulfide bridge [32, 61], which may explain the greater difficulty for the ligand to access iron in this state. A theoretical work further supported the observation that the pentacoordinate form was stabilized by the existence of the disulfide bridge . Therefore, the stabler histidine–iron bond in the absence of an internal disulfide bridge might explain the slower reactivity. Notably, the ratio of A/B binding rates also significantly decreases from 5 to 2 upon cysteine reduction, which may indicate reequilibration of histidine dissociation rates between the two conformations in the reduced Ngb. This trend is further confirmed by the similar on-rate values for the two heme orientations in the triple mutant. An increase in backbone flexibility was found in the CD–D region upon cysteine reduction and was even greater upon C46G/C55S mutations (MD study), which suggests that elevated flexibility in this region may contribute to stabilize the His64–Fe3+ bond. This analysis further indicates that the triple-mutant Ngb poorly mimics the function of the WT protein. This may be attributed to the small structural changes caused by the mutations as revealed from the heme chemical shift analysis. This reinforces the need for an experimental structure of the native protein.
Although the precise function of Ngb in the cell is not yet fully understood, there is now a wealth of experimental data demonstrating the redox-dependent ligand binding properties of Ngb in vitro. This is illustrated here by the threefold to sevenfold reduction in cyanide ligation rate for Ngb upon cysteine reduction. The redox-dependent activity was proposed to explain the role of Ngb in protecting neuronal cells against oxidative stress induced under conditions such as hypoxia, ischemia, and stroke [13, 16, 17, 76]. However, Ngb displays large heterogeneity in binding properties owing to heme disorder and, for example, the B conformation with an internal disulfide bridge and the A conformation with reduced cysteines have similar ligand reactivity. Therefore, although from the macroscopic point of view Ngb experiences lower averaged reactivity in reducing conditions, the distributions of activity for the oxidized and reduced Ngb partially overlap owing to heme disorder, which sheds light on the requirement of a better description of the behavior of Ngb at the microscopic level. Alternatively, since heme orientation modulates Ngb activity, it may play a yet unrecognized role in regulating its function.
The authors thank F. Rusconi and V. Steinmetz (Université Paris-Sud UMR 8000) for mass spectrometry experiments.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
- 41.Lindahl E, Hess B, van der Spoel D (2001) J Mol Model 7:306–317Google Scholar