Quantum model of catalysis based on a mobile proton revealed by subatomic x-ray and neutron diffraction studies of h-aldose reductase

Abstract

We present results of combined studies of the enzyme human aldose reductase (h-AR, 36 kDa) using single-crystal x-ray data (0.66 Å, 100K; 0.80 Å, 15K; 1.75 Å, 293K), neutron Laue data (2.2 Å, 293K), and quantum mechanical modeling. These complementary techniques unveil the internal organization and mobility of the hydrogen bond network that defines the properties of the catalytic engine, explaining how this promiscuous enzyme overcomes the simultaneous requirements of efficiency and promiscuity offering a general mechanistic view for this class of enzymes.

Enzymatic catalysis frequently involves proton translocation (PT). PT can occur on very short time scales (nanoseconds) and over long distances (tens of Angstroms), often involving multiple proton relay sites and proton wires (Grotthus-like mechanisms) (1). The nature of PT processes combined with the weak x-ray scattering signal from hydrogen atoms, which excludes their observation at resolutions <1.2 Å, make the structural characterization of PT processes very difficult. Recent advances in synchrotron and neutron sources, detectors, cryocooling, and software, coupled with the ability to obtain high-quality crystals, have led to an improved level of detail in protein structures, as exemplified in our recent structure of human aldose reductase (h-AR) determined at subatomic resolution (2).

h-AR (Enzyme Commission 1.1.1.21) is a NADPH-dependent enzyme that reduces a wide range of substrates, such as aldehydes, aldoses, and corticosteroids. Because it reduces d-glucose into d-sorbitol, it is believed to cause severe degenerative complications of diabetes (3). The catalytic reaction involves a hydride transfer from carbon C4 of NADPH, which becomes NADP⁺, and a proton donation from the enzyme (4). To visualize h-AR interactions with the NADPH cofactor, substrates, and inhibitors, and to determine the details of the catalytic mechanism, several x-ray analysis (5–7), site-directed mutagenesis (8–11), and modeling studies (12) have been performed. The extraordinary quality of h-AR crystals allows structural studies at a level of detail not available before for an enzyme of this size.

h-AR folds as a (β/α)₈ triosephosphate isomerase barrel with the nicotinamide ring of the NADP⁺ buried at the bottom of a deep cleft (5) (Fig. 1A). The catalytic carbon C4 of the nicotinamide ring of NADP⁺ is accessible through this cleft, thus defining the catalytic zone of the active site. Residues Asp-43, Tyr-48, Lys-77, His-110, Asn-160, and Gln-183, highly conserved in all enzymes of the aldo-ketoreductase superfamily, form a tight network of hydrogen bonds linked to the nicotinamide ring (Fig. 1B).

Fig. 1.

X-ray results. (A) Ribbon drawing of hydrogenated (S1) (red) and deuterated (S2) (blue) structures (rmsd = 0.7 Å), superposed with x-ray 2F_o−F_c density map (S2) structure, 1.5 rms magenta contours for the inhibitor IDD594. (B) Active-site conformation showing the residues around NADP⁺ and inhibitor IDD594 for the deuterated structure (S2). (C) Closeup of Lys-77 and Asp-43 in deuterated h-AR superposed with the two difference maps [deuterated (S2), 2 rms blue contours; and hydrogenated (S1), 2 rms red contours]. These maps show the hydrogen atom in the H-bond Asp-43–Lys-77 in two alternative conformations. This shared partial protonation is indicated also by the bond lengths (corresponding to structure S2) in the carboxylate of Asp-43. The H-atom is placed on the Lys-77 side.

Various aspects of the h-AR catalytic mechanism have been studied, and several models have been proposed, with the identity of the residue donating the proton proving controversial. Based on previous low-resolution results of site-directed mutagenesis studies of key active-site residues (Tyr-48, His-110, and Lys-77), Bohren et al. (11) concluded that Tyr-48 is the most likely proton donor, and that His-110 plays a role in orienting the substrate. They also suggested that Lys-77 plays an important role in catalysis. After several papers with alternative propositions (9–14), the Tyr-48-mediated pathway, for which the proton is transferred from Tyr-48 (which has a lowered pK_a), while His-110 is singly protonated at Nε₂, gained significance (14). This hypothesis was confirmed by the subatomic-resolution (0.66-Å) structure of the complex of h-AR with the brominated inhibitor IDD594 (Fig. 1A) (2), which revealed most of the hydrogen atoms for the well ordered catalytic region, including the single protonation state of His-110. However, although this h-AR structure provided details rarely observed before (e.g., a direct Br-O electrostatic contact and unexpected h-AR main chain conformations), information was still required for a full explanation of the catalytic mechanism about the protonation state of other key catalytic residues, such as Asp-43, Lys-77, and Tyr-48.

Results

Experimental Work.

To better visualize the active-site residue protonation states, we produced fully deuterated samples (15) and collected both x-ray and neutron diffraction data at room temperature. Neutron diffraction is particularly efficient at locating Deuterium (D) atoms, because the scattering length of D is similar to that of C, N, and O atoms. Three different structures of the complex of either the hydrogenated or fully deuterated h-AR with the cofactor NADP⁺ and the inhibitor IDD594 were analyzed in this work [see Table 1 and supporting information (SI) Text for data collection and refinement statistics], as follows: structure 1 (S1), the hydrogenated enzyme (x-ray, 0.66 Å, 100 K) (2); S2, the fully deuterated enzyme (x-ray, 0.80 Å, 15 K, this work); and S3, the fully deuterated enzyme (neutron, 2.2 Å and x-ray, 1.75 Å, both at 293K, this work). Determination of partially occupied H+/D atoms was done by a combination of information from electron density maps, neutron scattering maps, and bond length analysis.

Table 1.

Summary of data collection and refinement statistics for the three different structures analyzed in this work

	H-AR100K x-rays (S1)	Full D-AR15K x-rays (S2)	Full D-AR293K neutrons (S3)	Full D-AR293K x-rays (S3)
Diffraction data
Resolution range, Å	20.0–0.66	50.0–0.8	50.0–2.2	50–1.75
R(I)merge, %	2.9	2.3	22.8	2.9
Completeness, %	89.1	99.8	73.5	98.4
I/σ(I) (all)	14.5	28.1	5.1	26.8
Refinement
Reflections (work/test)	485,662/25,585	306,026/16,125	11,884/992	31,524/2,952
Rcryst/Rfree with H (D), %	9.36/10.32	9.6/11.3	25.65/29.08	12.88/16.53
Rcryst/Rfree without H (D), %	10.56/10.94	11.6/12.5	34.67/34.76	14.20/17.30
PDB ID code	1US0	2QXW	2R24	2R24

Combined Structure Analysis at the Catalytic Region.

Analysis of the deuterated structure (S2) showed no overall significant difference with the structure (S1) (Fig. 1). The Cα rmsd was 0.7 Å overall and 0.15 Å in the well ordered active site, implying that the information from both structures can be combined. The electron-density maps for structure (S2) showed clear density for active site amino acid, cofactor, and inhibitor atoms (Fig. 1A) (see SI Text for statistics).

The active site of the 0.66-Å structure (S1) of hydrogenated h-AR (2) was analyzed for partially occupied hydrogen atoms of key catalytic residues. This revealed a significant difference in the C-O bond lengths of Asp-43 carboxylate [1.243(3) vs. 1.271(3) Å], indicating a partial protonation of the carboxylate O2 oxygen in its H-bond with Lys-77. However, this observation was not confirmed by the difference electron density maps, which showed a peak corresponding to a hydrogen atom at the Lys side but not at the Asp side (Fig. 1C).

Although the overall structures (S1 and S2) were seen to be similar, analysis of the active site of structure (S2) provided new fine detail. The Nζ atom of Lys-77 makes H-bonds with the O atom of Cys-44 (as proton donor) and the Oδ2 atom of Asp-43 (as proton acceptor) (Fig. 1B). The S2 structure shows also significant differences in the Asp-43 C-O bond lengths [1.245(4) vs. 1.279(4) Å, Fig. 1C, very similar to those observed for structure S1] and indicates a partial protonation of the carboxylate O2 oxygen in its H-bond with Lys-77.

To locate the H-atom in this hydrogen bond, the difference maps for structures S1 and S2 were further compared (Fig. 1C). They suggest indeed that this H-atom is present in two conformations, because a peak is observed near Lys-77 in structure S1 and near Asp-43 in structure S2 (Fig. 2A), and in both structures, the C-O bond lengths indicate partial protonation of Asp-43. The neutron map confirms the neutral state of Lys-77 (Fig. 2 B and C). Furthermore, the room temperature x-ray 2F_o−F_c map shows a diminished density of Cε in Lys-77 (SI Text and SI Fig. 5), which indicates some mobility of the side chain in the neutral state.

Fig. 2.

Comparison of x-ray and neutron results. (A) X-ray model of fully deuterated h-AR-IDD594 complex [0.80-Å data collected at 15K, S2; treatment with HKL2000 (16); refinement with SHELX (17)] superposed with electron density difference map (F_o−F_c, 2 rms blue contours, phases calculated from model without deuteriums). The map suggests partial deuteration of the Asp-43. The model shows the neutral state for the Asp-43-Lys-77 pair. (B) Model from joint x-ray/neutron refinement of fully deuterated h-AR-IDD594 complex (S3) (both data collected at room temperature) superposed with neutron scattering-density map calculated with phases from the model with all deuteriums (2F_o−F_c, 2 rms red contours, 1 rms gold contours, S3). Note that the deuterium atom D (marked in magenta) was included in the model but is only weakly present in the map, confirming the partial protonation of the Asp-43-Lys-77 pair. (C) Closeup of B centered on the Lys-77 head. The map (2 rms red contours) shows strong density for only the two deuteriums marked in light gray.

These interpretations about both related motilities of the proton and the lysine residue were included in a new model of the catalytic mechanism, as described below.

Quantum mechanics/molecular mechanics (QM/MM) modeling of the catalytic mechanism was used to help explain and analyze the different protonation states of the Asp-43–Lys-77 pair, the geometry of the Asp-Lys-Tyr triad, and the proton-transfer energetics after the initial hydride transfer step. Hybrid QM/MM, using density functional theory (DFT) calculations for the bond-breaking and forming processes, provided a satisfactory but perhaps not unique explanation of the observed facts.

MM/MD calculations were in general agreement with the crystal structures. In the catalytic region, this agreement strongly depends on the charge state of the Asp-43–Lys-77 pair. The PT process itself was studied by using QM/MM simulations. Productive trajectories (with complete PT from Tyr to substrate or inhibitor head; see Fig. 3 and SI Movies 1 and 2) were obtained for the neutral/inhibitor and neutral/substrate models in all five cases simulated for each model. This analysis using the QM/MM neutral pair/substrate model results in the following distances: Tyr-48(O)/IDD594(O) = 3.0 Å; Tyr(O)/Lys(N) = 3.1 Å, and Lys(N)/Asp(O) = 3.2 Å, in excellent agreement with experiment. On the other hand, QM/MM simulations using a charged pair/substrate model results in the recoil of the Tyr head, hindering its ability to donate the proton to the substrate [Tyr-48(O)/IDD594(O) = 4.0 Å; Tyr(O)/Lys(N) = 3.2 Å, and Lys(N)/Asp(O) = 3.0 Å]. Thus, the charged form appears to be a kinetic trap that must be avoided. This is achieved through the destabilization of the charged Lys by the positive nicotinamide head, causing the back donation of the proton from Lys(+) to Asp.

Fig. 3.

QM/MD results. (A–C) Three steps of the QM/MD calculations before (A), during (B), and after (C) PT. The calculations suggest that the presence of the charged nicotinamide head has a marked influence in the behavior of the reaction centre region. The initial hydride transfer step provokes a driving force for PT, because of both the charge on the substrate and the effect on the Lys component of the salt bridge, which donates back a proton to Asp to reduce its positive charge, thus triggering the movement of Tyr and the consequent PT. Note the shortening of the distance between Lys 77 and Tyr-48 between frames A and B. (D) Map corresponding to PT step (B) computed using quantum chemical methods. The image shows the important residues in the pathway, especially Asp-43, Lys-77, and Tyr-48, as well as the substrate. Note the density (red arrow) around the oxygen atom of Tyr-48 extending over the Lys-77 nitrogen, in good agreement with the experiment.

Before every productive PT event, there is a shortening of the Tyr-Lys distance (Fig. 3 SI Text, and SI Appendix). This event is largely reversible in the neutral/inhibitor model. The close values of the energies of the contributing protonation forms in the neutral/substrate model result in oscillatory events with the proton bouncing back and forth from the inhibitor carboxylate to the Tyr. This process seems induced by the presence of the Tyr− after the Tyr proton is transported to the inhibitor. The reaction pathway deduced from this analysis shows rapid proton transport after the hydride donation by the nicotinamide. As shown in Fig. 4A, the pathway follows steps I to VI in numerical order. The previously suggested pathway (14) is obtained by joining steps I, III, VII, VIII, and VI. The energy diagram (Fig. 4B) shows our calculations for all these states, including those corresponding to the previous pathway (VII and VIII), which are in good agreement with the published ones (14), as shown in the SI Fig. 7. Step IV shows a similar density distribution to the experimental one across the Lys-Tyr region (Fig. 3D). Transits IV to V and V to VI are barrierless (Fig. 4B, energy diagram, blue lines) in the free energy hypersurface, resulting from a concerted mechanism (18) (for further details, see SI Text and SI Appendix). Furthermore, step VI* (IDD594 model), showing a similar density along the Lys-Tyr pair than step IV, has the closest geometry to the experimental one from the modeled ones, with Tyr-48(O)/IDD594(O) = 3.2 Å [Experiment (Exp.): 2.92 Å]; Tyr(O)/Lys(N) = 3.4 Å (Exp.: 3.05 Å), and Lys(N)/Asp(O) = 3.2 Å (Exp.: 2.72 Å). By comparison, the distances in step VII, which is part of a previouslysuggested reaction pathway (14), were Tyr-48(O)/IDD594(O) = 4.6 Å; Tyr(O)/Lys(N) = 3.0 Å, and Lys(N)/Asp(O) = 3.1 Å.

Fig. 4.

Proposed reaction pathway. (A) I–VI show the proposed reaction pathway. VII and VIII show the proton donation for a “charged” model, in which both Lys-77 and Asp-43 are charged. VI* shows the complex with the inhibitor IDD 594. (B) Energies calculated for each reaction step. The proposed pathway I–VI is shown in black (before hydride donation) and blue (after hydride donation) lines. The “charged model” pathway is included (red lines) after hydride donation (VII–VIII–VI). Note that for the proposed model, the proton donation is barrierless (blue lines), whereas for the “charged model,” there is an energy barrier (red lines).

In summary, after the hydride donation step, the positive charge in the nicotinamide head stabilizes the neutral bridge between Asp-43 and Lys-77, allowing the relaxation of the Tyr geometry toward the substrate head and the subsequent proton donation. After this last step, Tyr-48 recoils to a locked geometry where the Asp-43–Lys-77 salt bridge is formed again and completes a tight four-charge sandwich involving the Asp(−), Lys(+), Tyr(−), and nicotinamide(+) head (step VI).

Discussion

A detailed structural analysis of a medium-size protein (36 kDa, h-AR in ternary complex with NADP+ and IDD594 inhibitor) is presented here. The x-ray (293K, 100K, and 15K) and neutron (293K) data describe the structure of h-AR corresponding to the state after hydride transfer but just before proton transfer. The experimental data were supplemented with quantum mechanics and molecular dynamics calculations for the entire reaction. These calculations are fully consistent with the state observed in the x-ray and neutron structures. The maps reveal details that cannot be extrapolated from single-method structural studies, including nonconventional protonation states of critical side chains and direct evidence of geometries and densities compatible with rapid proton relocation along an extended network of hydrogen bonds. Our data underscore the importance of Tyr-48 and substantiate this residue in proton transfer, as suggested (14). However, our data show also the role of the Asp-Lys pair operating on the Tyr donor as a crucial motif responsible for the catalytic properties of the enzyme. This proposition, which implies that Tyr-48 is negatively charged after the reaction (step VI), has been independently confirmed by microcalorimetric studies showing that a proton is accepted by Tyr-48 during ligand binding to the h-AR-NADP⁺ complex (corresponding to step VI) (19).

The ability of Tyr-48 to donate a proton may help enable adoption of multiple productive geometries and allow the processing of different substrates, a key feature of a promiscuous enzyme. Under this assumption, the catalytic center can be conceptually understood as having its catalytic zone placed one step away from the substrate, with Tyr-48 acting as a flexible final proton carrier. This mode of operation of the catalytic site is an extension of the interpretation of enzymes as transition-state (TS) binders, with the TS defined by the geometric arrangement of atoms of the enzyme and substrate in direct contact. The results presented here suggest this enzyme may overcome the difficulty of simultaneously satisfying the requirements of being an effective catalyst and a promiscuous one by using a distal proton donor (Asp-43–Lys-77 pair) acting through the flexible side chain of the final proton donor (Tyr-48), capable of accommodating different substrates.

Although this mechanistic suggestion should be further verified by other enzymology approaches, its application to other promiscuous enzymes is conceptually straightforward and might provide a cross-validation. The direct observation of the properties of the proton network, extending from Asp-43 to IDD594, was possible only by using state-of-the-art x-ray and neutron diffraction data. This work highlights the possibilities provided by the combination of these complementary techniques in structural biology.

Materials and Methods

Experimental.

The x-ray data (S2) were collected at ANL/APS/SBC sector 19ID at 15K. Refinement resulted in a model with 315 residues, NADP+, IDD594, two citrate molecules, and 555 water molecules. The neutron data (S3) were collected from a single crystal of volume 0.15 mm³ (unusually small for neutron diffraction) on the Laue diffractometer, LADI, at the Institut Laue Langevin (ILL) (15, 20, 21). The x-ray data (S3) were collected at Swiss Light Source [(SLS) Villigen, Switzerland], to 1.75-Å resolution and were used simultaneously with the neutron data in a joint refinement using PHENIX (22). The protonation states of amino acid residues were determined, and 301 water molecules, in which both deuterium and oxygen atoms were visualized, were assigned as full D₂O molecules with the characteristic “boomerang” shape density (23). Resulting R factor/R_free were 12.9%/16.6% for x-ray data and 25.6%/29.1% for neutron data. For comparison, the final R factor of a refinement with the neutron data only was 25.6% and R_free 31.4%. The joint refinement decreased the neutron R_free factors and improved significantly the neutron maps (see SI Text and SI Fig. 6). The main difference between the two cases was in the B factors, which reflected better the disordered regions in the joint refinement. On the other hand, the mean deviation of Cα atoms between the neutron-only and combined refinements was only 0.1 Å.

Modeling.

Molecular dynamics (MD) was performed by using both a conventional MM/MD methodology and a hybrid QM/MM approach, linking a quantum chemistry code to our own MD driver program Dynga (24). Details are given in SI Text and SI Appendix.