Search tips
Search criteria 


Logo of actad2this articlesearchopen accesssubscribesubmitActa Crystallographica Section D: Biological CrystallographyActa Crystallographica Section D: Biological Crystallography
Acta Crystallogr D Struct Biol. 2016 July 1; 72(Pt 7): 892–903.
Published online 2016 June 28. doi:  10.1107/S2059798316009049
PMCID: PMC4932920

Visualizing the Bohr effect in hemoglobin: neutron structure of equine cyanomethemoglobin in the R state and comparison with human deoxyhemoglobin in the T state


Neutron crystallography provides direct visual evidence of the atomic positions of deuterium-exchanged H atoms, enabling the accurate determination of the protonation/deuteration state of hydrated biomolecules. Comparison of two neutron structures of hemoglobins, human deoxyhemoglobin (T state) and equine cyanomethemoglobin (R state), offers a direct observation of histidine residues that are likely to contribute to the Bohr effect. Previous studies have shown that the T-state N-terminal and C-terminal salt bridges appear to have a partial instead of a primary overall contribution. Four conserved histidine residues [αHis72(EF1), αHis103(G10), αHis89(FG1), αHis112(G19) and βHis97(FG4)] can become protonated/deuterated from the R to the T state, while two histidine residues [αHis20(B1) and βHis117(G19)] can lose a proton/deuteron. αHis103(G10), located in the α11 dimer interface, appears to be a Bohr group that undergoes structural changes: in the R state it is singly protonated/deuterated and hydrogen-bonded through a water network to βAsn108(G10) and in the T state it is doubly protonated/deuterated with the network uncoupled. The very long-term H/D exchange of the amide protons identifies regions that are accessible to exchange as well as regions that are impermeable to exchange. The liganded relaxed state (R state) has comparable levels of exchange (17.1% non-exchanged) compared with the deoxy tense state (T state; 11.8% non-exchanged). Interestingly, the regions of non-exchanged protons shift from the tetramer interfaces in the T-state interface (α12 and α21) to the cores of the individual monomers and to the dimer interfaces (α11 and α22) in the R state. The comparison of regions of stability in the two states allows a visualization of the conservation of fold energy necessary for ligand binding and release.

Keywords: neutron crystallography, cyanomethemoglobin, alkaline Bohr effect, H/D exchange

1. Introduction  

Hemoglobin (Hb) is one of the best-characterized macromolecules and is regarded as an ideal model for studies of protein evolution, cooperativity and allostery. In vertebrates, two heterodimers (α1β12β2) comprise the oxygen-transporting functional tetramer contained at high concentration (~5 mM tetramer, normal mean corpuscular hemoglobin per mean corpuscular volume) inside the red blood cells (RBCs). This containment is crucial since the tightly coupled deoxy tetramer (K d [similar, equals] 0.01 nM) would dissociate into liganded relaxed (R-state) dimers (K d [similar, equals] 1 µM) at lower concentrations (Atha & Riggs, 1976  ). Binding of oxygen in the lung, driven by the increased partial pressure of oxygen, the tetramer transitions into the fully liganded R state. As RBCs move away from the lungs, the partial pressure of oxygen declines, leading to ligand dissociation, promoted by the stabilizing effect of the deoxy tense (T-state) tetramer. In the tissue capillaries, the RBCs also encounter an increase in the concentration of dissolved CO2 and a corresponding decrease in pH, where hemoglobin assists in the transport of CO2 and excess protons. Several heterotropic allosteric hemoglobin modulators, including H+, CO2 and 2,3-bisphosphoglycerate (2,3-BPG), regulate the reversible binding of O2. The release and uptake of protons by hemoglobin is known as the ‘Bohr effect’ and ionizable residues influenced by this effect are known as ‘Bohr groups’. Perutz and coworkers proposed that salt bridges present in the T state primarily influence the Bohr effect. The salt bridges involve the α N-terminus, Val1(NA1), bridged to the α C-terminus, Arg141(HC3), and the β C-terminus, His146(HC3), bridged to both αLys40(C6) and βAsp94(FG1) (globin fold positions are denoted in parentheses; Perutz et al., 1969  , 1980  ; Perutz, 1970  ). Multiple studies have implicated additional residues that contribute to the Bohr effect (Berenbrink, 2006  ; Lukin & Ho, 2004  ). From a structural point of view, the transition from the R state to the T state alters the local environments of potential Bohr groups. As expected, small changes in relative positions shift the pK a values of ionizable residues, altering the proton affinity of a particular Bohr group (Thurlkill et al., 2006  ). Despite the numerous theoretical and experimental studies on hemoglobin, the specific changes in the protonation state of potential Bohr groups in the R and T states remains uncertain, as the techniques that have previously been utilized do not provide direct observation of the protonation state.

The cooperative binding behavior of O2 to Hb has been investigated both kinetically and thermodynamically (Parkhurst, 1979  ; Ackers, 1980  ). In most models, Hb is proposed to exist in two states: a high O2 affinity R state and a low O2 affinity T state. The two states are revealed in the biphasic on-rates of O2 (33 and 7 µM −1 s−1 for the R and T states, respectively) and in the sigmoidal shape of the oxygen-dissociation curves (Parkhurst, 1979  ; Imai et al., 1970  ). Detailed analysis of oxygen-dissociation curves has yielded a number of mathematical models for the R-to-T transition. These include the Hill equation, where all monomers are treated equally, the Adair equation, where individual equilibrium constants are introduced, and the symmetry model (extended Adair equation), where the R and T states are included in the binding steps (Parkhurst, 1979  ). The introduction of induced-fit and free-energy penalties for subunit association provided the best overall understanding of the thermodynamics of oxygen delivery supported by the view of hemoglobin provided by structural analysis (Ackers, 1980  ; Perutz et al., 1980  ; Koshland et al., 1966  ). In Acker’s free-energy penalty model, the presence of ligand affects the stability of dimer interfaces (α21 and α12), with the greatest strain introduced when ligands bind on opposing dimers (Ackers, 1980  ). The first ligand-binding step is restricted owing to the tetramer stability of the T state, while the last ligand has high affinity owing to the transfer of stability to the dimer. The free-energy difference between the two states is ~6.3 kcal mol−1, where the T-state tetramer is more thermodynamically stable. Structurally, the oxygenation of Hb disrupts major dimer–dimer interactions and stabilization is shifted towards the individual monomers.

The Bohr effect is the linkage of decreasing oxygen affinity to increasing CO2 concentration, mediated by the formation of carbaminohemoglobin and the acidity of the dissolved CO2, where proton binding influences the release of oxygen. Approximately two CO2 molecules and two H+ ions are added to the Hb tetramer during the R-to-T transition (O’Donnell et al., 1979  ; Perutz et al., 1980  ). In order to identify a potential Bohr group, the protonation state of histidine (His) needs to be established in both the T and the R state (Fig. 1  ). These His residues are titratable at neutral pH. Comprehensive studies using NMR, H/D exchange and site-directed mutagenesis of recombinant Hb have determined pK a differences for 24 of the 38 possible Bohr-group His residues (Sun et al., 1997  ; Fang et al., 1999  ; Ohe & Kajita, 1980  ). Recent pK a calculations based on high-resolution crystallographic models have provided a detailed prediction of the protonation states of the Bohr residues (Park et al., 2006  ).

Figure 1
Overall structure of EqCNmetHb: a ribbons model of EqCNmetHb with α subunits in wheat, β subunits in gray, heme groups in yellow and His residues in black.

This study directly visualizes the positions of exchangeable H atoms in equine cyanomethemoglobin (EqCNmetHb; Fig. 1  ) using neutron crystallography and deuterium exchange. EqCNmetHb is a stable model for the liganded hemoglobin R state crystallized at physiological pH. Equine hemoglobin has 88% sequence identity to human hemoglobin and is considered to be isomorphous with human oxyhemoglobin, with most histidine positions conserved (Ladner et al., 1977  ). A 2.0 Å resolution neutron data set was collected at the Protein Crystallography Station using the pulsed spallation neutron source of the Los Alamos Neutron Science Center (Langan et al., 2004  , 2008  ). Comparison of this R-state model with the previously reported neutron crystal structure of human deoxyhemoglobin (PDB entry 3kmf; HuDeoxyHb) provides a unique opportunity to compare and contrast the extent of protonation/deuteration (H+/D+) of the R and T states (Kovalevsky, Chatake et al., 2010  ). While obviously more appropriate, crystals of human R-state Hb are not amenable to neutron diffraction analysis owing to smaller crystal size and an excessively long unit-cell dimension (space group P41212, unit-cell parameters a = 53.7, c = 193.5 Å). The use of equine Hb as the model R-state hemoglobin (space group C2, unit-cell parameters a = 108.8, b = 63.0, c = 54.6Å, β = 110.7°) has long been established, and for long-term ligand stability the use of ferricyanide was essential (Perutz, 1968  ).

Variations in histidine H+/D+ states are coupled to changes in the local environment involving changes in solvent accessibility, changes in hydrogen-bond networks and disruption of proton shuttles. Comparison of the two neutron structures provides proper orientation and discerns changes in the histidine rings while also providing a qualitative assessment of the presence or absence of H+/D+ at the δ and [sm epsilon] positions. These studies indicate that six histidine residues show significant changes in their H+/D+ states, with some gaining protons, supporting the Bohr effect, while others lose protons, in opposition. The backbone amides of the taut state (deoxy T state) exhibit a comparable level of non-exchangable H atoms to that seen in the relaxed state (liganded R state). Interestingly, the regions of non-exchangeability change location, providing a direct observation of the preservation of fold energy during ligand binding. These differences in backbone exchange indicate an increase in flexibility of the R-state α12 interface, with a corresponding decrease in the flexibility of the α11 interface and the heme pocket correlated to ligand binding.

2. Methods  

2.1. Crystallization and data collection and reduction  

EqCNmetHb crystallization was carried out using the batch method at pH 7.2 in a nine-well depression plate (Hampton Research) as described previously (Mueser et al., 2000  ; Kovalevsky, Fisher et al., 2010  ). This method produced several large crystals; the crystal used in this study had a volume of ~10 mm3 (Kovalevsky, Fisher et al., 2010  ). The EqCNmetHb crystal underwent H/D vapor exchange in a quartz capillary for one month before data collection (Kovalevsky, Fisher et al., 2010  ). A neutron time-of-flight Laue diffraction experiment was carried out at the Protein Crystallography Station (PCS) located at Los Alamos Neutron Scattering Center (LANSCE). The data-collection and neutron data-reduction details have been described previously (Kovalevsky, Fisher et al., 2010  ). After the neutron data set had been collected, the crystal was subjected to X-ray data collection and the X-ray data were initially integrated and scaled using d*TREK during data collection (Pflugrath, 1999  ) as described previously (Kovalevsky, Fisher et al., 2010  ). During joint refinement, it became apparent that the previously reported X-ray data were not optimally processed. Therefore, reintegration and rescaling of the X-ray diffraction data was performed using iMosflm and SCALA incorporated into the CCP4 program suite (Evans, 2006  ; Winn et al., 2011  ; Leslie, 2006  ). The reprocessed data have a slightly higher overall R merge than the previously reported data. However, the crystal mosaicity was significantly lower (0.42 versus 1.36°), with a higher signal-to-noise ratio (12.9 versus 10.9 overall; 3.9 versus 2.1 in the highest resolution shell). The reprocessed data also produced noticeably better quality maps in both X-ray and joint refinement.

2.2. Structure determination and refinement  

The room-temperature (RT) X-ray structure was refined using REFMAC5 from the CCP4 program suite (Murshudov et al., 2011  ); the initial starting model for rigid-body refinement was PDB entry 1g0b (Mueser et al., 2000  ; Kovalevsky, Fisher et al., 2010  ) and the statistics are highlighted in Table 1  . This RT X-ray structure was later used as the starting model for the joint refinement process. The joint X-ray/neutron refinement was executed in nCNS, starting with rigid-body refinement followed by several cycles of positional, atomic displacement parameter (i.e. B-factor) and D-occupancy refinement (Adams et al., 2009  ). Between each refinement cycle, D atoms and D2O molecule positions and orientations were manually adjusted using Coot (Emsley & Cowtan, 2004  ). In Coot, the orientation and the H+/D+ state of each ionizable residue were inspected with nuclear |F o| − |F c| and 2|F o| − |F c| Fourier X-ray density maps with σA-type weights. In order to provide evidence of the protonation/deuteration state of specific side chains (His, Lys, Glu, Asp, Tyr, Thr and Ser), D atoms were deleted from the model, the structure was refined and the omitted sections were inspected again by calculating |F o| − |F c| and 2|F o| − |F c| maps. Positive density peaks at the location of the omitted atoms permitted direct determination of the deuteration states of specific side chains. For refinement purposes, disordered side chains (lack of both electron and neutron density) were assigned based upon their pK a at pH 7.4. D2O molecules were assigned and built into the positive density peaks during X-ray refinement and orientated based upon hydrogen bonding and the nuclear |F o| − |F c| and 2|F o| − |F c| density maps. The figures and solvent-accessibility percentages were generated using PyMOL. The solvent-accessibility percentage was estimated using the ‘get_area’ option in PyMOL (dot density = 4, dot_solvent = on). Side-chain exposure was calculated by differencing values with (histidine) and without (alanine) the imidazole ring.

Table 1
Neutron crystallographic data-collection and refinement statistics

2.3. Hydrogen/deuterium-exchange (HDX) analysis  

The scattering lengths of D and H are 6.67 × 10−15 and −3.74 × 10−15 m, respectively, which results in the scattering length of H equalling −0.56 of the D scattering length. During the individual occupancy refinement in nCNS the minimum value associated with the occupancy of an atom is set to −0.56 (Adams et al., 2009  ). At this point in the refinement the atomic displacement parameters are constrained. Because of the chemical nature of a peptide bond, the backbone amide proton has the ability to exchange with D2O. In the case of an amide proton that is stabilized by a strong hydrogen-bonding network, the proton will be shielded from exchange with D2O. For the individual occupancy refinement the occupancy of the backbone amide protons can refine to between −0.56 and 1.0, with −0.56 if the amide proton is H and 1.0 if it is D. The refined backbone amide proton occupancies were analyzed for EqCNmetHb and HuDeoxyHb, with occupancies ranging from −0.56 to 0 assigned as a non-exchanged proton, occupancies of 0–0.6 assigned as partially exchanged H and occupancies of 0.6–1.0 assigned as fully exchanged D. These results from the individual occupancy refinement allow the areas of stability and dynamics in the two Hb neutron structure compared here to be visualized.

3. Results and discussion  

3.1. Neutron structure determination of EqCNMetHb  

EqCNmetHb crystals are inherently more stable and easier to handle than equine carbonmonoxy-Hb crystals (PDB entry 1g0b), allowing large crystals (>5 mm3) to be obtained (Mueser et al., 2000  ; Kovalevsky, Fisher et al., 2010  ). The neutron time-of-flight (TOF) data-collection and refinement statistics are outlined in Table 1  . The EqCNmetHb neutron structure was solved to 2.0 Å resolution using joint X-ray/neutron refinement, a resolution sufficient to accurately locate D atoms in macromolecules (Adams et al., 2009  ; Blakeley et al., 2008  ). The resultant real-space final 2|F o| − |F c| nuclear density maps allowed the assignment of most labile H atoms, with the exceptions being a few highly disordered, solvent-exposed side chains that are also absent in the electron-density maps. While the diffraction resolution is not high enough to accurately determine partial occupancies, they can establish the presence or absence of D (positive nuclear density peaks) or the presence of a non-exchanged H (negative nuclear density peaks). All Asp and Glu residues were interpreted as deprotonated. The majority of hydroxyl-containing side chains (Thr, Ser and Tyr) showed positive peaks for D atoms and were interpreted as exchanged with D2O. The exchange rate of tyrosine hydroxyl groups is relatively rapid and decreases as a function of strong hydrogen-bond networks associated with the side chain (Takeda et al., 2009  ). For a few residues, αTyr24(B5), αThr38(C3), αSer81(F2), αSer131(H14), βSer44(CD3) and βSer70(E14), the hydroxyls did not show positive deuteration peaks, suggesting the absence of exchange. However, these groups are located on the surface and lack any strong hydrogen-bond networks; therefore, exchange with D is likely but perhaps may not be visible owing to a high freedom of motion.

The equine hemoglobin tetramer has 38 His residues: 20 in the α chains and 18 in the β chains. The C2 symmetry of the EqCNmetHb crystals aligns the molecular dyad with a crystallographic axis, reducing the asymmetric unit to an αβ dimer and the number of His residues to analyze to ten for the α chain [residues 20(B1), 45(CD3), 50(CD8), 58(E7 distal), 72(EF1), 87(F8 proximal), 89(FG1), 103(G10), 112(G19) and 122(H5)] and nine for the β chain [residues 63(E7 distal), 69(E13), 76(E20), 77(EF1), 92(F8 proximal), 97(FG4 switch), 117(G19), 143(H21) and 146(HC3)]. The His positions in the equine (Eq) α chains are identical to those in HuDeoxyHb (Hu), whereas those in the β chains differ at four locations: Eq Gln2/Hu His2(NA2), Eq His69/Hu Gly69(E13), Eq His76/Hu Ala76(E20) and Eq Arg116/Hu His116(G18) (Fig. 2  ). Histidines have a slightly acidic pK a (~6.8) on average with environmental variability, with buried side chains becoming more acidic. EqCNmetHb crystals were grown at pH 7.2 (pD 7.6), where a solvent-exposed His is expected to have N[sm epsilon]2 fully occupied with H+/D+ and Nδ1 partially occupied with H+/D+ (~2.5:1 His0:His+). In the EqCNmetHb neutron structure the His residues were initially doubly occupied and the occupancy values were refined (ranging from 0 to 1) to estimate His0:His+ ratios with the final model containing the predominant form. Neutral histidine (His0) can exist in two tautomeric states with H+/D+ either at the Nδ1 or N[sm epsilon]2 position, with N[sm epsilon]2 being the more likely (Fig. 2  ). Two conserved His residues [αHis20(B1) and βHis117(G19)] were determined to be doubly protonated (His+), with the occupancies of both D atoms refining to a value of 1. The H+/D+ occupancies of these two residues are clearly influenced by neighboring conserved acidic groups: αHis20(B1) is hydrogen-bonded to αGlu23(B4) and βHis117(G19) is hydrogen-bonded to βGlu26(B8). Interestingly, two His residues that are unique to equine Hb, βHis69(E13) and βHis76(E20) (Gly and Ala in human Hb, respectively), are the only others that appear to be doubly protonated in the R state and are also influenced by neighboring groups: βHis69(E13) is hydrogen-bonded to βGlu20(B2) (Val in human) and to βGlu73(E16) (Asp in human), and βHis76(E20) is buried in a crystal contact. Four His residues appear to be disordered [αHis50(CD8), αHis89(FG1), βHis143(H21) and βHis146(HC3)] and were assigned to be neutral (His0) with N[sm epsilon]2 occupied by H+/D+. All other His residues in the EqCNmetHb neutron structure were determined to be singly H+/D+-occupied (His0). The proximal His residues, αHis87(F8) and βHis92(F8), have N[sm epsilon]2 coordinated to the Fe ion and Nδ1 exchanged with D. Of the remaining nine His residues, all have N[sm epsilon]2 occupied with H+/D+ as expected, with the exception of βHis97(FG4 switch), which has Nδ1 occupied with H+/D+ (Fig. 2  ).

Figure 2
Protonation states of histidine residues in deoxyHb versus CNmetHb with pK a values determined by previous NMR and H/D-exchange studies, and by molecular mechanics calculations based on high-resolution crystal structures. The protonation states of His ...

3.2. Comparison of R and T states and visualization of the Bohr effect  

The availability of two neutron structures (EqCNmetHb and HuDeoxyHb) allows direct visualization of the changes in the hydrogen-bonding networks associated with the potential Bohr groups. Essential experimental conditions impose caveats on this analysis as listed. (i) The use of equine Hb instead of human Hb as the R-state model was required to obtain crystals that were suitable for neutron diffraction. (ii) The duration of the studies, several months for deuterium exchange and data collection, necessitated the use of the stable ferric cyanomet derivative instead of a ferrous heme with CO or O2. (iii) Crystallization of the HuDeoxyHb crystal was completed at pH 6.3 (pD 6.7), which potentially exaggerates Bohr-group titration, while the R-state pH was completed at closer to the physiological conditions (pH 7.2, pD 7.6).

The pH/pD difference between crystal forms can account for changes in the protonation states of several His residues and is noted when relevant. Comparisons of His-residue protonation in the R-state and T-state models are tabulated in Fig. 2  along with the published pK a values estimated by NMR, H/D exchange and molecular mechanics calculations (Sun et al., 1997  ; Fang et al., 1999  ; Ohe & Kajita, 1980  ; Zheng et al., 2013  ). The pK a values and shifts determined by NMR and H/D exchange have discrepancies, and predictions of protonation do not necessarily correlate with what is seen in the neutron structures. For example, the pK a value determined by NMR and H/D exchange suggests that no histidines in the R state would be doubly protonated. However, two conserved His residues [αHis20(B1) and βHis117(G19)] clearly show protonation at both the Nδ1 and N[sm epsilon]2 positions. Also, the lower pH of the T-state crystal should have the primary Bohr groups, βHis146(HC3), both doubly protonated. Instead, the two crystallographically distinct C-termini, un­affected by crystal contacts, show variation: one is doubly protonated, while the other is singly protonated, suggesting a more modest involvement in the overall Bohr effect. The values of pK a are not accurately determinable from the deuterium occupancies of titratable groups. Instead, visualization of the changes in local environment that affect hydrogen-bonding networks correlates to the changes in pK a values, especially in residues with limited solvent accessibility (Supplementary Table S1). Interestingly, the largest overall change in solvent accessibility is noted for βHis146(HC3): from 44% exposed in the R state to 0% exposed in the T state.

The physiologically relevant alkaline Bohr effect is a decrease in oxygen affinity as the arterial Hb at pH ~7.4 encounters the pH of ~7.2 in the peripheral tissue, allosterically promoting oxygen release in the transition from the R state to the T state. During this R-to-T-state transition, the binding of an estimated two protons per Hb tetramer occurs (Perutz et al., 1980  ). Changes in His pK a owing to the structural transition will either support proton uptake (negative ΔpK a on transitioning from T to R) or oppose proton uptake (positive ΔpK a on transitioning from T to R). As Hb returns to higher O2 concentrations, the increased oxygen tension promotes proton release and the subsequent increase in oxygen affinity known as the Haldane effect. During the normal delivery of oxygen, Hb retains two ligands on average, adopting a mixed population of the T state with two ligands on the same dimer and the R state with two ligands on opposing dimers (Ackers, 1980  ).

Allosteric effectors of Hb, such as 2,3-BPG, enhance the stability of the T-state tetramer, promoting ligand release. In the HuDeoxyHb T state, the βHis146(HC3) C-terminus forms a salt bridge to βAsp94(FG1) during the R-to-T transition. NMR studies determined that βHis146(HC3) has a negative ΔpK a from T to R (−1.51 by NMR, −1.1 by H/D exchange), contributing to the alkaline Bohr effect (Fang et al., 1999  ; Sun et al., 1997  ; Ohe & Kajita, 1980  ). This observation is partly supported by the neutron structures. In the T-state model, HuDeoxyHb β1His146(HC3) was determined to be doubly protonated and β2His146(HC3) was determined to be singly protonated. EqCNmetHb (R state) has βHis146(HC3) solvent-exposed with a high freedom of motion and it is likely to be singly protonated (His0). While both HuDeoxyHb βHis146(HC3) residues are proposed to form ‘intramolecular salt bridges’ with βAsp94(FG1), encountering nearly identical hydrogen-bonding networks, no nuclear density was observed at the Nδ1 position of β2His146(HC3). A subtle difference between the two HuDeoxyHb βHis146(HC3) residues is the N(...)O hydrogen-bond distance to βAsp94(FG1), which is 2.8 Å in the doubly protonated β1 and 3.0 Å in the singly protonated β2. The shorter distance is likely to be promoted by the added electrostatic interaction of His+ with Asp in β1; however, both remain at distances normally assigned as hydrogen bonds (2.8–3.2 Å) and are not in the electrostatic range (2.4–2.6 Å). The neighboring His located at the end of the H helix, βHis143(H21), has been reported to have a pK a that becomes more acidic from T to R (+0.87 by NMR, +0.5 by H/D-exchange studies; Fang et al., 1999  ; Ohe & Kajita, 1980  ) and is predicted to be singly protonated in both the T and the R states. In the R state, βHis143(H21) appears to be singly protonated. In the T state, the dimers adopt states that are exactly the opposite of their βHis146(HC3) counterparts, with β1His143(H21) singly protonated and β2His143(H21) doubly protonated. All structural evidence suggests that these two neighboring residues work in concert to contribute to the overall Bohr effect.

The remaining potential Bohr groups are mostly surface residues and the protonation states of surface residues are heavily influenced by the pH of the solvent. Both states crystallized in monoclinic space groups, HuDeoxyHb in P21 (with a tetramer in the asymmetric unit) and EqCNmetHb in C2 (with a dimer in the asymmetric unit) and, while related, the two exhibit different crystal contacts, potentially affecting their protonation states (Kovalevsky, Chatake et al., 2010  ). Adding to the complexity for comparison is the fact that the HuDeoxyHb crystal was grown at pH 6.3 (pD 6.7), increasing the ratio of protonated to deprotonated (~1:1) His residues on the surface (Chatake et al., 2007  ). To contribute to the alkaline Bohr effect, a histidine residue must be singly protonated in the R state (His0) and become doubly protonated in the T state (His+). The NMR studies identified three residues that support the alkaline Bohr effect: αHis89(FG1) (ΔpK a = −0.55 by NMR and −1.6 by H/D exchange), βHis97(FG4 switch) (ΔpK a = −0.26 by NMR and no change by H/D exchange) and βHis146(HC3) (ΔpK a = −1.51 by NMR and −1.1 by H/D exchange) (Fig. 2  ; Sun et al., 1997  ; Fang et al., 1999  ; Ohe & Kajita, 1980  ). In comparing the two neutron structures, three residues in the α chain and one in the β-chain have protonation changes that support the alkaline Bohr effect: αHis72(EF1), αHis103(G10) and αHis112(G19), and βHis97(FG4 switch), respectively. For the first α residue, the HuDeoxyHb αHis72(EF1) T state is doubly protonated, while αHis72(EF1) is only protonated at the N[sm epsilon]2 position in the EqCNmetHb R state. In both states, αHis72(EF1) is exposed to solvent and does not exhibit any hydrogen bonds; therefore, the differences seen can be attributed to differences in the pH of the crystallization buffers.

The second α residue, αHis103(G10), is located within the α11 interface (solvent accessibility = 0) near the hinge region on the F helix of hemoglobin and is the most intriguing of those that support the alkaline Bohr effect. In the NMR studies, the pK a shifts were not determined for αHis103(G10) owing to the incorporation of the α11 interface into the core (Sun et al., 1997  ; Fang et al., 1999  ). In the H/D-exchange experiment, the pK a of αHis103(G10) was determined to be 7.1 and was unaffected by the change of state, predicting partial occupancy in the T state (HA:A ratio of 1:0.7). Calculations estimate a strongly acidic pK a (3.2 in the T state and 3.9 in the R state) precluding any involvement in the Bohr effect (Zheng et al., 2013  ). Visually, both α1His103(G10) and α2His103(G10) in the HuDeoxyHb T state were determined to be doubly protonated with full occupancy (Figs. 3  a–3  d). The EqCNmetHb R-state α1His103(G10) is singly protonated (Figs. 3  e and 3  f). In the EqCNmetHb R state, the Nδ1 atom of αHis103(G10) acts as a hydrogen-bond acceptor to a water which is bridged by another water to the side-chain N atom of βAsn108(G10). In the HuDeoxyHb T state, the Nδ1 atom of αHis103(G10) acts as a hydrogen-bond donor to the nearest water and the coupling to βAsn108(G10) is lost. The hydrogen-bond network coupling αHis103(G10) to βAsn108(G10) bridges the G and H helices, involving βGln127(H4) and βGln131(H8). All four of these hydrogen-bond-linked residues are highly conserved and have known clinical variants. The βGln127(H4)→Arg (Hb Dieppe) and βGln131(H8)→Arg variants are reported to be highly unstable since a positive charge is being buried in the α11 dimer interface (Girodon et al., 1992  ; Beneitez Pastor et al., 2007  ). More intriguing are the αHis103(103)→Arg (Hb Contaldo) and βAsn108(G10)→Ile (Hb Schlierbach) variants, which form unstable and low O2 affinity Hbs, respectively (Sciarratta et al., 1984  ; Frischknecht et al., 1999  ). Both mutations would disrupt the R-state hydrogen-bond network and the destabilization promoting the adoption of the low-affinity T state.

Figure 3
αHis103(G10) in the α11 interface supports the alkaline Bohr effect: T-state HuDeoxyHb α11 interface (a, b), T-state HuDeoxyHb α22 interface (cd) and R-state EqCNmetHb α1 ...

For the third α residue, HuDeoxyHb T-state α1His112(G19) and α2His112(G19) are both doubly protonated, while in the EqCNmetHb R state the Nδ1 D atom is absent (Fig. 4  ). In both states, αHis112(G19) is hydrogen-bonded to αGlu27(B8); the N(...)O distances are 2.8 Å (EqCNmetHb) and 3.4 Å (HuDeoxyHb). In the case of HuDeoxyHb αHis112(G19), the protonation state can be attributed to the hydrogen-bond interactions with αGlu27(B8) and a weak hydrogen bond to αTyr24(B5). Interestingly, this interaction seems to affect the capability to observe the D atom on αTyr24(B5). With the loss of the hydrogen bond between αTyr24(B5) and αHis112(G19) in the R state, an increase in the freedom of motion of the tyrosyl phenol D atom of αTyr24(B5) is observed, as indicated by the refined negative occupancy. The only potential Bohr group in the β chain, βHis97(FG4 switch), appears to be doubly protonated in the HuDeoxyHb T state, and is located in close proximity to αThr41(C6) and αPro44(CD2) in the switch region of the hemoglobin tetramer (Figs. 5  a and 5  b). In the EqCNmetHb R state, βHis97(FG4) is singly protonated and is further away from the α12 interface, located between αThr38(C3) and αThr41(C6) but exposed to solvent (Fig. 5  c).

Figure 4
Alternate protonation states of αHis20(B1) and αHis112(G19). Located on the surface near the α11 interface, αHis20(B1) is doubly protonated in T-state HuDeoxyHb α1 (a), singly protonated in T-state HuDeoxyHb ...
Figure 5
βHis97(FG4) supports the alkaline Bohr effect: βHis97(FG4) is located in the switch region of the α12 interface. In T-state HuDeoxyHb (a, b), βHis97(FG4) is tightly positioned between αThr41(C6) and αPro44(CD2) ...

In addition to βHis143(H21) as discussed above, the NMR studies identified three additional residues that could potentially counter the alkaline Bohr effect: αHis20(B1) (ΔpK a = +0.06 by NMR and −0.6 by H/D exchange), αHis112(G19) (ΔpK a = +0.04 by NMR and no change by H/D exchange) and βHis77(EF1) (ΔpK a = +0.6 by NMR and no change by H/D exchange) (Fig. 2  ; Sun et al., 1997  ; Fang et al., 1999  ; Ohe & Kajita, 1980  ). When comparing the two neutron structures, two different His residues were identified, αHis20(B1) and βHis117(G19), which were determined to be doubly protonated in the R state and singly protonated in the T state. The solvent-exposed αHis20(B1) is weakly hydrogen-bonded to αGlu23(B4) (Fig. 4  ). In EqCNmetHb αHis20(B1) is protonated at both the Nδ1 and N[sm epsilon]2 positions, in which Nδ1 acts as a weak hydrogen-bond donor to αGlu23(B4) (N(...)O distance of 3.5 Å). The HuDeoxyHb αHis20(B1) is doubly protonated in α1 (N(...)O distance of 4.5 Å) but not in α2 (6.2 Å N(...)O distance), losing hydrogen bonds to αGlu23(B4) in both the α1 and α2 subunits. In EqCNmetHb, βHis117(G19) is clearly protonated at both the Nδ1 and N[sm epsilon]2 positions, as shown by the 2F oF c nuclear density map (Fig. 6  ). In HuDeoxyHb, both the β1His117(G19) and β2His117(G19) were determined to not be protonated and the β1His117(G19) protonation state was justified by the side chain being disordered. In EqCNmetHb βHis117(G19) N[sm epsilon]2 donates a hydrogen bond to βGlu26(B8) (N(...)O distance of 3.0 Å), while in the HuDeoxyHb neutron structure β2His117(G19) is rotated 180° and βGlu26(B8) is 7.0 Å away from the Nδ1 position of β2His117(G19). Also, in EqCNmetHb βHis117(G19) Nδ1 makes a weak electrostatic interaction with βGlu22(B4) (N(...)O distance of 3.4 Å) and in HuDeoxyHb βGlu22(B4) is more than 8.0 Å away (Fig. 6  ). It is important to note that βHis116(G18) in HuDeoxyHb is not conserved in equine Hb [βArg116(G18)]; both side chains contain a positive charge but allow distinctive interactions to occur between the two states.

Figure 6
βHis117(G19) counters the alkaline Bohr effect. Solvent-exposed and located near the end of the G helix, βHis117(G19) appears to be singly protonated with dual conformation in T-state HuDeoxyHb β1 (a), singly protonated in the ...

In summary, for most of the His residues in the R state His0 is prevalent as expected, whereas the T state adopts a mixture of single and double protonation states, as was observed for βHis143(H21) and βHis146(HC3), including residues αHis50(CD8), αHis58(E7 distal), αHis89(FG1), βHis2(NA2) and βHis63(E7 distal). The His residues that have no changes in apparent protonation state are αHis45(CD3), αHis87(F8 proximal), αHis122(H5), βHis77(EF1) and βHis92(F8 proximal). Here, we report several His residues that do contribute to the alkaline Bohr effect, including αHis103(G10), αHis112(G19) and βHis146(HC3), and two His residues that counteract the Bohr effect, αHis20(B1) and βHis117(G19). It is important to note that no other types of residues exhibited changes in protonation state that are related to the Bohr effect. Directly determining the atomic position of D atoms that are bonded to His residues contributes to the understanding of how H+ can act as a heterotopic modulator. Interestingly, this is the first time that αHis103(G10), an α11 interface contact, has been shown to have a definitive structural linkage to the Bohr effect where NMR, H/D exchange and calculations anticipate no differences.

3.3. Comparison of R and T states and the distal histidine  

Another unique difference between the R and T states (EqCNmetHb and HuDeoxyHb) is the change in distal His(E7) protonation states. The HuDeoxyHb neutron structure indicates that the protonation state of the distal His varies between the two dimers: α1β1 His(E7) residues are doubly protonated while α2β2 His(E7) residues are singly protonated. All distal His(E7) residues in HuDeoxyHb have a D2O molecule accepting a D atom connected to the N[sm epsilon]2 atom by a hydrogen bond and forming O—H(...)π contacts with the hemes. Curiously, in the α1β1 dimer the distal His residues [α1His58(E7) and β1His63(E7)] are doubly protonated while the neighboring Lys residues [α1Lys61(E10) and β1Lys66(E10)] were determined to be deprotonated (ND2) (Fig. 7  a). The opposite was observed in the α2β2 dimer: α2His58(E7) and β2His63(E7) are singly protonated while the neighboring Lys residues [α2Lys61(E10) and β2Lys66(E10)] were determined to be protonated (Nδ1) (Fig. 7  b).

Figure 7
The distal histidines, α1His58(E7) and β1His63(E7): alternate protonation states. In the α1β1 dimer of T-state HuDeoxyHb both distal histidines are doubly protonated (a) (α1 is shown, but β1 is similar), ...

In the R-state neutron structure, both of the distal His residues [αHis58(E7) and βHis63(E7)] are singly protonated (Fig. 7  c). In the presence of ligand, greater order in the region of the distal histidine appears, with a water molecule that is not present in the T state positioned between the distal His(E7) and Lys(E10). The deprotonated Nδ1 atom of the distal His(E7) accepts a hydrogen bond from this bridging water, while Lys(E10) donates a hydrogen bond (Fig. 7  c). The ordering of Lys(E10) stabilizes its coupling to the heme propionate group, an interaction that is substantially weaker in the T state. The protonation states of the distal histidine residues do not support their contribution to the Bohr effect. However, this is the first time that any evidence of a possible proton transfer linked to the distal histidine that occurs between the two states of hemoglobin has been reported.

3.4. Comparison of R and T states and backbone amide H/D exchange  

The partial double-bond character of the backbone amide bond allows H/D exchange (with the exception of proline) through acidic (protonation of N), basic (deprotonation of N) and imidic acid (protonation of O) mechanisms. This provides information pertaining to regions of flexibility that allow exchange and to regions of stability where exchange is prohibited (Engen, 2009  ). The two crystals used in the neutron diffraction experiments were mounted in quartz capillaries and exposed to D2O by vapor diffusion for a minimum of one month before data collection. The HuDeoxyHb mother liquor had a pH of 6.3 (pD 6.7), while the EqCNmetHb mother liquor had a pH of 7.2 (pD 7.6). The rate of exchange is affected by the pH/pD of the solution, with rates decreasing at lower pH (Connelly et al., 1993  ). The rates of exchange for pD 6.7 and pD 7.6 are not significantly different. The two crystals were crystallized and exchanged with deuterium using different methods. HuDeoxyHb underwent buffer exchange with D2O for several hours (~1 × 10−4 s−1) before the deoxygenation step, followed by crystallization, which required several days, while EqCNmetHb was exchanged with D2O after crystallization (Kovalevsky, Chatake et al., 2010  ; Kovalevsky, Fisher et al., 2010  ). This indicates that the highly reactive backbone amide protons (exchange rates lower than 1 × 10−4 s−1) cannot be the subject of this analysis. Instead, this comparison focuses on the protected backbone amide protons with partial exchange (rates of ~10−5–10−6 s−1) and with no exchange (rates of >3 × 10−6 s−1) (Englander, 1975  ). HuDeoxyHb has 56.8% of the amide protons fully exchanged (occupancy 0.6–1.0), 26.7% of the amide protons partially exchanged (occupancy 0.0–0.6) and 11.8% with no exchange (occupancy −0.56 to 0.0) (Fig. 8  a). EqCNmetHb has 56.1% fully exchanged, 22.6% partially exchanged and 17.1% with no exchange (Fig. 8  b).

Figure 8
Amide H/D exchange indicates that T-state HuDeoxyHb has greater flexibility, while R-state EqCNmetHb displays less exchange in the αβ dimer interface. Superimposed on the ribbon figures of T-state HuDeoxyHb (a) and R-state EqCNmetHb ( ...

Overall, the HuDeoxyHb T state has 0.7 and 4.1% more fully and partially exchanged main-chain amides, respectively, versus the EqCNmetHb R state. This may be due in part to the differences in D2O exposure prior to crystallization that would primarily affect the rapidly exchanging protons. While no controlled studies on the comparison of vapor versus liquid exchange exist, the extensive studies of xylose isomerase (XI) offer some insight. For the nine neutron structures of XI (PDB entries 3cwh, 4dvo, 3kcj, 3qza, 2gve, 4qdw, 4qdp, 3kcl and 3kco), the percentage non-exchanged varied from 9 to 36%, where two were liquid-exchanged (PDB entries 3chw and 2gve) and the rest were vapor-exchanged. As might be expected if differences exist, PDB entry 3chw does have the most exchange. However, PDB entry 2gve was subjected to several months of liquid exposure prior to microgravity crystallization as part of a NASA Shuttle mission (Katz et al., 2006  ). At 30.8%, PDB entry 2gve is in the group that are least exchanged. These XI results suggest that liquid and vapor exchange are comparable techniques, at least for comparing the non-exchanged regions. For the hemoglobin neutron structures, the T-state and R-state amide H/D exchanges are depicted relative to the sequence (Fig. 8  c). In the HuDeoxyHb T state, the primary areas of stability are localized in the α12 and α21 interfaces (Figs. 8  a and 8  c). The presence of ligand in the EqCNmetHb R state shifts stability to the cores of the individual monomers and extends the stability of the α11 interface (Figs. 8  b and 8  c). These results are in agreement with Acker’s free-energy penalty model of ligand binding (Ackers, 1980  ). Hemoglobin is a passive transporter where strain is not lost or gained but shifted from one region to another when switching between states. As ligands are lost, the interfaces forming the tetramer in the T state are stabilized. As ligands bind, the R-state monomer interactions are stabilized at the expense of the dimer–dimer interface.

Supplementary Material

PDB reference: neutron structure of equine cyanomethemoglobin, 5c6e

Supplementary Table S1.. DOI: 10.1107/S2059798316009049/mn5110sup1.pdf


The Protein Crystallography Station (PCS), located at Los Alamos Neutron Scattering Center (LANSCE), is funded by the Office of Biological and Environmental Research of the US Department of Energy. AYK, SZF and PL were supported by PCS. PL was partly supported by an NIH–NIGMS-funded consortium (1R01GM071939-01) between LANL and LBNL to develop computational tools for neutron protein crystallography. TCM, SS and BLH were supported by NSF (446218). SD was supported by The University of Toledo.


  • Ackers, G. K. (1980). Biophys. J. 32, 331–346. [PubMed]
  • Adams, P. D., Mustyakimov, M., Afonine, P. V. & Langan, P. (2009). Acta Cryst. D65, 567–573. [PMC free article] [PubMed]
  • Atha, D. H. & Riggs, A. (1976). J. Biol. Chem. 251, 5537–5543. [PubMed]
  • Beneitez Pastor, D., Carrera, A., Duran-Suárez, J. R., Durán Pastor, M. A., Álvarez Vinent, F. & Melich, C. (2007). Hemoglobin, 31, 183–185. [PubMed]
  • Berenbrink, M. (2006). Respir. Physiol. Neurobiol. 154, 165–184. [PubMed]
  • Blakeley, M. P., Langan, P., Niimura, N. & Podjarny, A. (2008). Curr. Opin. Struct. Biol. 18, 593–600. [PMC free article] [PubMed]
  • Chatake, T., Shibayama, N., Park, S.-Y., Kurihara, K., Tamada, T., Tanaka, I., Niimura, N., Kuroki, R. & Morimoto, Y. (2007). J. Am. Chem. Soc. 129, 14840–14841. [PubMed]
  • Connelly, G. P., Bai, Y., Jeng, M.-F. & Englander, S. W. (1993). Proteins, 17, 87–92. [PubMed]
  • Emsley, P. & Cowtan, K. (2004). Acta Cryst. D60, 2126–2132. [PubMed]
  • Engen, J. R. (2009). Anal. Chem. 81, 7870–7875. [PMC free article] [PubMed]
  • Englander, S. W. (1975). Ann. NY Acad. Sci. 244, 10–27. [PubMed]
  • Evans, P. (2006). Acta Cryst. D62, 72–82. [PubMed]
  • Fang, T.-Y., Zou, M., Simplaceanu, V., Ho, N. T. & Ho, C. (1999). Biochemistry, 38, 13423–13432. [PubMed]
  • Frischknecht, H., Speich, R., Bloch, K. E., Fehr, J., Tuchschmid, P. & Jenni, R. (1999). Hemoglobin, 23, 83–87. [PubMed]
  • Girodon, E., Ghanem, N., Vidaud, M., Riou, J., Martin, J., Galactéros, F. & Goossens, M. (1992). Ann. Hematol. 65, 188–192. [PubMed]
  • Imai, K., Morimoto, H., Kotani, M., Watari, H., Hirata, W. & Kuroda, M. (1970). Biochim. Biophys. Acta, 200, 189–196. [PubMed]
  • Katz, A. K., Li, X., Carrell, H. L., Hanson, B. L., Langan, P., Coates, L., Schoenborn, B. P., Glusker, J. P. & Bunick, G. J. (2006). Proc. Natl Acad. Sci. USA, 103, 8342–8347. [PubMed]
  • Koshland, D. E. Jr, Némethy, G. & Filmer, D. (1966). Biochemistry, 5, 365–385. [PubMed]
  • Kovalevsky, A. Y., Chatake, T., Shibayama, N., Park, S.-Y., Ishikawa, T., Mustyakimov, M., Fisher, Z., Langan, P. & Morimoto, Y. (2010). J. Mol. Biol. 398, 276–291. [PubMed]
  • Kovalevsky, A. Y., Fisher, S. Z., Seaver, S., Mustyakimov, M., Sukumar, N., Langan, P., Mueser, T. C. & Hanson, B. L. (2010). Acta Cryst. F66, 474–477. [PMC free article] [PubMed]
  • Ladner, R. C., Heidner, E. J. & Perutz, M. F. (1977). J. Mol. Biol. 114, 385–414. [PubMed]
  • Langan, P., Fisher, Z., Kovalevsky, A., Mustyakimov, M., Sutcliffe Valone, A., Unkefer, C., Waltman, M. J., Coates, L., Adams, P. D., Afonine, P. V., Bennett, B., Dealwis, C. & Schoenborn, B. P. (2008). J. Synchrotron Rad. 15, 215–218.
  • Langan, P., Greene, G. & Schoenborn, B. P. (2004). J. Appl. Cryst. 37, 24–31.
  • Leslie, A. G. W. (2006). Acta Cryst. D62, 48–57. [PubMed]
  • Lukin, J. A. & Ho, C. (2004). Chem. Rev. 104, 1219–1230. [PubMed]
  • Mueser, T. C., Rogers, P. H. & Arnone, A. (2000). Biochemistry, 39, 15353–15364. [PubMed]
  • Murshudov, G. N., Skubák, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). Acta Cryst. D67, 355–367. [PMC free article] [PubMed]
  • O’Donnell, S., Mandaro, R., Schuster, T. M. & Arnone, A. (1979). J. Biol. Chem. 254, 12204–12208. [PubMed]
  • Ohe, M. & Kajita, A. (1980). Biochemistry, 19, 4443–4450. [PubMed]
  • Park, S.-Y., Yokoyama, T., Shibayama, N., Shiro, Y. & Tame, J. R. H. (2006). J. Mol. Biol. 360, 690–701. [PubMed]
  • Parkhurst, L. J. (1979). Annu. Rev. Phys. Chem. 30, 503–546.
  • Perutz, M. F. (1968). J. Cryst. Growth, 2, 54–56.
  • Perutz, M. F. (1970). Nature (London), 228, 726–739. [PubMed]
  • Perutz, M. F., Kilmartin, J. V., Nishikura, K., Fogg, J. H., Butler, P. J. G. & Rollema, H. S. (1980). J. Mol. Biol. 138, 649–668. [PubMed]
  • Perutz, M. F., Muirhead, H., Mazzarella, L., Crowther, R. A., Greer, J. & Kilmartin, J. V. (1969). Nature (London), 222, 1240–1243. [PubMed]
  • Pflugrath, J. W. (1999). Acta Cryst. D55, 1718–1725. [PubMed]
  • Sciarratta, G. V., Ivaldi, G., Molaro, G. L., Sansone, G., Salkie, M. L., Wilson, J. B., Reese, A. L. & Huisman, T. H. J. (1984). Hemoglobin, 8, 169–181. [PubMed]
  • Sun, D. P., Zou, M., Ho, N. T. & Ho, C. (1997). Biochemistry, 36, 6663–6673. [PubMed]
  • Takeda, M., Jee, J., Ono, A. M., Terauchi, T. & Kainosho, M. (2009). J. Am. Chem. Soc. 131, 18556–18562. [PubMed]
  • Thurlkill, R. L., Grimsley, G. R., Scholtz, J. M. & Pace, C. N. (2006). Protein Sci. 15, 1214–1218. [PubMed]
  • Winn, M. D. et al. (2011). Acta Cryst. D67, 235–242. [PMC free article] [PubMed]
  • Zheng, G., Schaefer, M. & Karplus, M. (2013). Biochemistry, 52, 8539–8555. [PubMed]

Articles from Acta Crystallographica. Section D, Structural Biology are provided here courtesy of International Union of Crystallography