pH-dependent reaction triggering in PmHMGR crystals for time-resolved crystallography

Abstract

Serial crystallography and time-resolved data collection can readily be employed to investigate the catalytic mechanism of Pseudomonas mevalonii 3-hydroxy-3-methylglutaryl (HMG)-coenzyme-A (CoA) reductase (PmHMGR) by changing the environmental conditions in the crystal and so manipulating the reaction rate. This enzyme uses a complex mechanism to convert mevalonate to HMG-CoA using the co-substrate CoA and cofactor NAD⁺. The multi-step reaction mechanism involves an exchange of bound NAD⁺ and large conformational changes by a 50-residue subdomain. The enzymatic reaction can be run in both forward and reverse directions in solution and is catalytically active in the crystal for multiple reaction steps. Initially, the enzyme was found to be inactive in the crystal starting with bound mevalonate, CoA, and NAD⁺. To observe the reaction from this direction, we examined the effects of crystallization buffer constituents and pH on enzyme turnover, discovering a strong inhibition in the crystallization buffer and a controllable increase in enzyme turnover as a function of pH. The inhibition is dependent on ionic concentration of the crystallization precipitant ammonium sulfate but independent of its ionic composition. Crystallographic studies show that the observed inhibition only affects the oxidation of mevalonate but not the subsequent reactions of the intermediate mevaldehyde. Calculations of the pKa values for the enzyme active site residues suggest that the effect of pH on turnover is due to the changing protonation state of His381. We have now exploited the changes in ionic inhibition in combination with the pH-dependent increase in turnover as a novel approach for triggering the PmHMGR reaction in crystals and capturing information about its intermediate states along the reaction pathway.

Significance

The uniquely complex mechanism of HMGR is a promising application of time-resolved X-ray crystallography that can overcome the inherent limitations of static crystal structures in elucidating the details of complex enzyme mechanisms. The observation of an inactive ternary complex and the pH dependence of the rate of the reaction enable a new triggering method suitable for time-resolved crystallographic data collection in PmHMGR crystals. Control of the conditions in the crystal environment also allows for the direct observation of the formation of the thiohemiacetal intermediate previously postulated. Determination of the time frame of the biochemical reaction in the crystal provides essential information for estimating the required rate of data collection for future time-resolved experiments.

Introduction

Class II 3-hydroxy-3-methylglutaryl (HMG)-coenzyme A (CoA) reductase from Pseudomonas mevalonii (PmHMGR) catalyzes the oxidative acylation of mevalonate to form HMG-CoA using two NAD⁺ cofactors and the co-substrate CoA. (1,2,3). HMG-CoA is further converted into acetoacetate, which is involved as a carbon source in a large number of biochemical pathways (4). In other prokaryotic and eukaryotic organisms, the reaction catalyzed by HMGR runs predominantly in the opposite direction, reducing HMG-CoA to mevalonate used in isoprenoid biosynthesis (5,6). The mevalonate pathway is critical for the survival of pathogenic gram-positive bacteria, such as Staphylococcus aureus, Streptococcus pneumonia, and Enterococcus faecalis; virulence of these pathogens without the HMG-CoA reductase (mevalonate synthase [mvaS]) gene is substantially reduced in mice, suggesting that bacterial HMGR could be a potential target for novel antibiotics (5,7,8). PmHMGR has a highly conserved active site fully comparable to its homologs in three low G + C gram-positive bacteria that have shown antibiotic resistance, methicillin-resistant S. aureus, drug-resistant S. pneumoniae, and vancomycin-resistant E. faecalis (5,9,10,11)^, and so makes an ideal model system for mechanism-based antibiotic development. Time-resolved experiments to examine the details of the HMGR reaction in these bacterial species have become possible with the production of cubic I4₁32 crystals of PmHMGR. These crystals have wide channels open to the active site for rapid diffusion of added molecules, the ability to accommodate large conformational changes necessary to build the active site around the substrate and cofactors, and allow the reaction to run in the crystalline environment.

The overall reaction catalyzed by PmHMGR is extraordinarily complex (12) and requires the formation of two intermediates, mevaldehyde and mevaldyl-CoA, to form HMG-CoA (Fig. 1). In the two hydride transfer steps, the cofactor accepts electrons from the substrate (mevalonate) and intermediate (mevaldyl-CoA), respectively, to generate HMG-CoA and two equivalents of NADH. As will be discussed in more detail below, this unusual catalysis of two different reactions in a single active site requires large-scale reorientations of HMGR domains to allow cofactor exchange.

Figure 1.

Reactions catalyzed by PmHMGR.

Biochemical studies indirectly demonstrated the involvement of mevaldehyde as an intermediate in the HMGR reaction pathway in yeast and P. mevalonii, demonstrating that HMGR can convert mevaldehyde to mevalonate using NADH (1,12,13). PmHMGR can also catalyze the oxidative acylation of mevaldehyde (12). Additional studies identified a second intermediate in the HMGR reaction in yeast where the enzyme converts mevaldehyde into a thiohemiacetal (mevaldyl-CoA) (14). Although the cofactor CoA is chemically involved as a substrate that forms a thioester bond with mevalonate to form HMG-CoA, its binding also plays a structural role in the PmHMGR mechanism by increasing the rate of conversion of mevaldehyde to mevalonate 1.5- to 2.5-fold (1,13).

A number of available crystal structures provide additional insights into the mechanism. PmHMGR crystals have a cubic symmetry (I4₁32) with equal unit cell dimensions of 226 Å and two molecules per asymmetric unit (15). Each monomer has a domain region that binds to cofactors CoA (large domain) and NAD⁺ (small domain) with the substrate mevalonate binding at the homodimer interface containing the catalytic residues Glu83 and Lys267 (15,16,17). Movement of the 50-residue C-terminal region of the enzyme, the flap domain, completes the active site, covering the ligand-bound region upon substrate and cofactor binding and positioning the catalytic histidine (His381) (17). Quantum mechanics/molecular mechanics, site directed mutagenesis, and kinetic studies have shown that the stabilization of residues Leu407, Glu399, and Glu396 in the closed flap transition state can facilitate the hydride transfer step (18,19). For the reaction, Glu83 and Lys267 in the active site act as a proton donor and oxyanion hole, respectively, and are crucial for enzyme activity (16,20,21,22), whereas His381 on the flap domain acts as a general base and allows the protonation of CoA-S⁻ after the breakdown of the thiohemiacetal bond in mevaldyl-CoA. NAD(H) is exchanged without release of the intermediates during the reaction, which requires a mechanism-driven movement of the flap domain whose details are not yet fully described (17,23,24). Structural studies with PmHMGR, E. faecalis HMGR and the S. pneumoniae and Delftia acidovorans enzymes suggest that interactions with the CoA cofactor and small domain residues in concert with the positioning and re-orienting of the flap domain over the cofactor and substrate-binding site can regulate NAD⁺/NADH exchange for subsequent steps of the enzymatic reaction (25,26,27).

Additional static structures of ligand-bound structural states provide partial snapshots of the reaction pathway, illustrating conformational changes induced by binding before full construction of the active site or flap domain closure (16). Nonproductive complexes with HMG-CoA and mevalonate (PDB: 1QAX) gave more clues to potential steps in the reaction (17), but a structure with what was thought at the time to be an inhibitor, dithio-HMG-CoA, bound to the enzyme with NADH, revealed that dithio-HMG-CoA was actually a slow substrate and the enzyme was active in the crystal. The dithiohemiacetal intermediate that mimics mevaldyl-CoA had been captured in this crystallographic structure (PDB: 4I4B) (16). The closed flap and the substrate contacts with the enzyme identified several residues of interest, suggesting possible mechanistic steps, but with large gaps between where the mechanism was still undefined.

A schematic depiction for the conversion of mevalonate to HMG-CoA catalyzed by PmHMGR can be seen in Fig. 2 (22). After binding of mevalonate, CoA, and one equivalent of NAD⁺ (step 1), the flap domain closes to form the reactive quaternary complex (step 2) that transfers a hydride to form NADH and mevaldehyde (step 3), which is trapped as a thiohemiacetal by CoA (step 4). Opening of the flap domain allows cofactor exchange (step 5), after which the flap domain closes again (step 6) to allow the second hydride transfer to produce HMG-CoA and NADH (step 7), which are released after opening of the flap domain (step 8). Note that the time frame of cofactor exchange is not certain. Although it is expected to occur after the first hydride transfer, it could happen before or after mevaldyl-CoA formation. The proposed mechanism here only suggests one of the two possibilities.

Figure 2.

Graphic representation of the steps of a postulated HMG-CoA reductase catalytic reaction.

Despite this wealth of structural and biochemical information about PmHMGR, a number of questions about the mechanism remain. Mevaldehyde has not been observed directly in crystal structures or in various assays. Neither C¹⁴ labeled HMG-CoA nor semicarbazide trapping in yeast HMGR gave an indication of mevaldehyde, suggesting that either it was not present as an intermediate or transitioned into mevalonate too fast to be detected (6). Similar experiments run with C¹⁴ labeled HMG-CoA and mevalonate in PmHMGR also did not detect any C¹⁴ in isolated pools of mevaldehyde that would indicate its presence as an intermediate in the reaction (12). The observed activity in the crystal led to an investigation of the possibility of time-resolved crystallography for PmHMGR. Attempts to crystallographically study the reaction of mevalonate, CoA, and NAD⁺ in PmHMGR did not lead to the observation of any product or intermediate despite having the previous evidence of turnover in the PmHMGR crystal containing dithio-HMG-CoA and NADH (16). As outlined in more detail below, some conformational changes associated with initiation of the reaction occur but no turnover is observed.

To investigate this apparent contradiction, we investigated the impact of the crystallization conditions such as pH, concentration, and ionic composition on enzymatic activity and various reaction steps. We will show that the systematic investigation yields novel ways to control the reaction, which can be used to trigger catalysis for time-resolved crystallography studies.

Materials and methods

Reagents

NAD⁺ and CoA for all diffraction, spectroscopic, and kinetic experiments were purchased from Sigma-Aldrich. Mevalonate was prepared from mevalonolactone (Sigma-Aldrich) by incubation with 1 M NaOH until the pH is stable at pH 11–12 and stored as an aqueous solution at pH 7 with addition of 1 M HCl (28,29). Tris, potassium phosphate, and glycine used to prepare the optimized pH-rate profile buffer were obtained from Sigma-Aldrich. Ammonium sulfate and 2,2′-[(2-amino-2-oxoethyl)imino]diacetic acid (ADA) used to prepare the crystallization solution were obtained from Sigma-Aldrich. Glycerol used in the crystallization buffer was obtained from Thermo Fischer Scientific. Various salts used to study enzyme activity in different ionic environments (ammonium acetate, ammonium chloride, sodium sulfate, and lithium sulfate) were obtained from Thermo Fischer Scientific. Sodium chloride, also used to study enzyme turnover in the absence of ammonium sulfate, was obtained from Thermo Fischer Scientific. PEG-400 used in the pH-jump buffer was obtained from Sigma-Aldrich.

Crystallization of PmHMGR bound to mevalonate, CoA, and NAD⁺

To observe the interaction between ligands and the enzyme in the initial crystallization environment, we solved the structure of PmHMGR bound to mevalonate, CoA, and NAD⁺. PmHMGR was purified as described previously (1,20). The enzyme was then concentrated to 20 mg/mL and crystallized using the sitting drop vapor phase exchange method with 1.2 M ammonium sulfate, 100 mM ADA, and 10% glycerol at pH 6.7. A 0.25 × 0.1 × 0.1-mm crystal grown in the same buffer was crushed to prepare a seeding solution: 0.9 μL of the seeding solution that was diluted 10³- to 10⁴-fold and was added to 10 μL of reservoir solution, 10 μL of 20 mg/mL purified PmHMGR, and 2.5 μL of nanopure water (16). To cryoprotect these crystals, the concentration of glycerol was gradually increased to 32% by slowly introducing a solution with 1.2 M ammonium sulfate, 100 mM ADA, and 32% glycerol into the crystal sitting drop and then storing it for 16–20 h. The cryoprotectant solution was slowly exchanged into a 23.4-μL sitting drop by adding incremental volumes from 3 to 20 μL with intervals of 5 min between additions. When adding volumes of or greater than 10 μL of cryoprotectant solution, an equal volume was also removed from the sitting drop. Crystals of the size 0.4 × 0.4 × 0.2 mm were selected for structure determination. The ligands were slowly introduced into the crystal sitting drop in 100 μL of 1 mM CoA, 1 mM NAD⁺, and 5 mM mevalonate solution by gradually replacing the sitting drop solution in the same manner as the cryoprotectant solution. The crystals were soaked in the substrate solution for a minimum of 4 h.

Crystallization of PmHMGR bound to mevaldehyde and CoA

To determine if the crystallization environment is inhibiting mevalonate oxidation or the subsequent thioester bond formation from mevaldehyde, the structure of PmHMGR bound to mevaldehyde and CoA was determined. If solution conditions do not interfere, this combination of substrates should spontaneously form a thioester bond by proximity in the active site. Mevaldehyde was prepared using a mevaldic acid precursor (the hemi-[N,N′-dibenzylethylenediammonium] salt of mevaldic acid) purchased from Sigma-Aldrich (29). PmHMGR was purified using the method described in the previous section (1). PmHMGR crystals of size 0.4 × 0.2 × 0.2 mm in 1.2 M (NH₄)₂SO_4, 100 mM ADA, and 32% glycerol at pH 6.7 were transferred by capillary into another sitting drop well with the same solution. Mevaldehyde and CoA were gradually introduced into the solution to obtain a final ligand concentration of 4 mM for both after exchanging out the starting solution. To achieve a 4 mM concentration of mevaldehyde and CoA in crystals, solutions containing 4 mM of each ligand were separately added to the sitting drop solution, with an equal volume removed before addition. The ligand solutions were introduced in incremental volumes ranging from 0.5–10 μL. The soaking time was 30 min between additions. The exchange step with 10 μL of 4 mM mevaldehyde and CoA solutions was carried out three or four times. The crystals were soaked in the final ligand solution for 14 h followed by cryoprotection by gradually introducing glycerol into the ligand solution to a final concentration of 30%. The equilibration time in glycerol was 5 h. The crystals were then frozen in liquid nitrogen.

Crystallographic data collection and analysis

Diffraction data for the mevalonate-, CoA-, and NAD⁺-bound crystals were obtained at the Advanced Photon Source, Argonne National Laboratory (23-ID-D) using a wavelength of 1.013 Å. Crystallographic data from the crystal extended to 2.0 Å using the CC_1/2 cutoff of 0.3 (30). The diffraction images were obtained on a Dectris Pilatus3 6M detector. Diffraction data with the mevaldehyde- and CoA-bound crystal were obtained at the Advanced Photon Source, Argonne National Laboratory (19-BM-D). Crystallographic data from the crystal extended to 1.65 Å with an I/σ cutoff of 4.5. Details of the data collection and refinement statistics are shown in Table 1.

Table 1.

X-Ray data collection and refinement statistics

	Mevalonate, CoA, and NAD+	Mevaldehyde and CoA (Thiohemiacetal)	Mevaldyl-CoA intermediate (5 min at pH 9)
Space group	I4(1)32	I4(1)32	I4(1)32
PDB ID	8GDN	8SZ6	8VLQ
Unit cell dimensions
a, b, c (Å)	226.0	225.9	226.0
α = β = γ (deg)	90	90	90
Resolution (Å)	48.19 –1.99 (2.02 – 1.99)a	28.70 – 1.65 (1.67 – 1.65)a	44.32 – 2.07 (2.14 – 2.07)a
No. of unique reflections	66,787	116,412	59,611
Mean I/Σi	13.6 (0.6)	32.0 (4.0)	20.4 (2.7)
Completeness (%)	82.9 (1.7)	99.8 (100.0)	99.0 (93.8)
Data redundancy	16.4	1.1	9.5
Wilson B-factor (Å2)	26.3	32.0	23.5
Refinement
Rwork (%)	18.0	15.5	21.5
Rfree (%)	22.3	17.6	24.8
RMSD bond lengths (Å)	0.012	0.016	0.007
RMSD bond angles (deg)	1.09	1.40	0.82
No. of protein residues	799	754	796
No. of water molecules	478	777	429
Average B. factor, protein (Å2)	31.8	23.5	33.9
Average B factor, ligand (Å2)	36.6	32.9	41.8
Average B factor, water (Å2)	37.0	36.9	40.7
Residues in Ramachandran plot regions (%)
Most favored	96.5	97.5	95.7
Additional allowed	3.5	2.5	4.2
Outliers	0.00	0.00	0.13

RMSD, root-mean-square deviation.

^aValues in parentheses represent the highest-resolution bin.

All collected datasets were indexed, integrated, and scaled with HKL2000 using Denzo and Scalepack (31). Scalepack2MTZ was used to generate a reflection file in the CCP4 suite of programs (32). Phasing information for both the structures was obtained using the PhaserMR program in CCP4 (33). A high-resolution native structure (PDB: 4I64) was used as a search model for the structure of the enzyme bound to mevaldehyde and CoA and for the enzyme bound to mevalonate, CoA, and NAD⁺. Structure and parameter files for the ligands were obtained using the eLBOW program or a small-molecule library from the Phenix suite of programs (34,35). LigandFit from Phenix was used to fit new ligand molecules in the generated 2Fo-Fc density (36,37). The structure was then optimized by a rigid body refinement using refmac5 in CCP4 followed by a simulated annealing (Cartesian) refinement using the phenix.refine function (38,39,40). The structure was then processed through several refinements in phenix.refine after using Molprobility and real-space correlation to correct errors after each refinement until convergence of Rwork and Rfree values was achieved (41,42). Data collection and refinement statistics for these structures are summarized in Table 1. PDB entries are 8GDN for the mevalonate, CoA, and NAD⁺ structure and 8SZ6 for the mevaldehyde and CoA structure. The program composite omit_map in Phenix was used to generate omit maps for the ligands at the enzyme’s active site (43). TLS refinement was used in the final iterations of the refinement process (40). An alignment of the thiohemiacetal produced in the mevaldehyde and CoA structure was done with the dithiohemiacetal and NAD⁺-bound PmHMGR structure (PDB: 4I4B) using the align function on Pymol; a superposition of all the protein atoms between the two structures was done before comparing the conformation of the bound ligand.

pH-jump method to initiate turnover in PmHMGR crystals

The application of a pH-dependent method to trigger mevalonate oxidation in PmHMGR crystals was tested by obtaining UV-visible (UV-vis) spectra of pre- and post-reaction triggered crystals. The absorbance at 340 nm for NADH was used to monitor the reaction in the crystal. PmHMGR crystals were grown in sitting drops using the same seeding protocol used in obtaining crystals for diffraction measurements (16). Crystals were stabilized in cryoprotectant as for the crystallographic data collection before moving between tested buffer solutions. Crystals with dimensions 0.4 × 0.2 × 0.2 mm were first cryoprotected in a solution of 1.2 M ammonium sulfate, 100 mM ADA (pH 6.7), and 32% glycerol as outlined above. These crystals were slowly introduced into the pH-jump buffer, a solution of 1.2 M ammonium acetate, 100 mM ADA, and 32% PEG-400 pH 6.7 by gradually replacing the sitting drop solution and equilibrated for 8 h. This buffer environment was chosen since it had the lowest turnover around crystallization pH (9.6 ± 5.4 min⁻¹) and a mild increase in rate at pH 9 (32.1 ± 11.7 min⁻¹) that suggested it would allow the reaction to progress in the order of minutes. The addition of the pH-jump buffer was done in incremental volumes ranging from 3 to 20 μL in the same manner as the addition of the cryoprotectant solution. While adding this solution and for subsequent additions, volumes of approximately 3 μL are first removed from the crystal drop before being replaced with the pH-jump buffer solution. The ligands mevalonate, CoA, and NAD⁺ were gradually introduced into the sitting drop solution at concentrations of 5, 1, and 1 mM, respectively, in the pH-jump buffer by gradually introducing and replacing the current solution in the sitting drop with the ligand-soaking solution.

At pH 6.7 the enzyme is inactive and the ligands can be slowly added to crystals that then remain in the soaking solution for 4 h. The soaking time was optimized to ensure sufficient binding of the ligands as observed with their high occupancy in the reported crystal structure of the mevalonate-, CoA-, and NAD⁺-bound complex. The ligand-bound crystals were subsequently placed in another well with the pH-jump buffer of 1.2 M ammonium acetate, 100 mM ADA, and 30% PEG-400 adjusted to pH 9. The transfer of these crystals to a higher pH initiates enzyme activity. These crystals were soaked for different time periods ranging from 80 s to 10.5 min in the pH 9 environment. The crystals were then scooped out and frozen in liquid nitrogen to freeze-trap the post-reaction state captured in the pH 9 soaked crystal. A crystal with dimensions 0.25 × 0.1 × 0.1 mm was used as a control to obtain the basic unreacted absorbance spectra after the pH-jump buffer and ligands had been introduced into the crystal at pH 6.7. For crystals of this size, an optimized soaking time of 2 h was used to introduce the ligands mevalonate, CoA, and NAD⁺.

The same process was subsequently applied to crystals measuring 0.1 × 0.05 × 0.05 μm for time points ranging from 65 s to 30 min after a pH jump. Using smaller crystals resulted in a quicker soaking time for pH changes, which was aimed at achieving more synchronized initiation of the reaction and a faster transition of the enzyme to its active state.

The UV-vis spectra of the frozen crystal samples were collected at the Stanford Synchrotron Radiation Lightsource (Beamline 9-2) using an in situ microspectrophotometer (44). The fully automated microspectrophotometer incorporates a Hamamatsu deuterium and halogen light source, reflective Newport Schwardchild objectives, and an Ocean Optics QE65000 Spectrum Analyzer. The crystals were mounted on the BL9-2 goniometer and maintained at 100 K in a nitrogen gas stream. To eliminate any accumulated ice on the samples, cryo-annealing was used, where the cryogenic nitrogen gas stream was turned off for a predetermined amount of time (30 s) to allow the ice present on the crystal surface to melt and vitrify.

To confirm that the expected reaction occurs in the crystal on changes with the spectroscopic readout, the mevalonate-CoA-NAD⁺-soaked crystal at an intermediate time point of 5 min post pH jump was chosen for examination. We obtained a 2.07-Å diffraction dataset at 23-ID-D using a 1.0332-Å wavelength on a Dectris Pilatus3 6M detector. The reflections obtained on this dataset were cut at an I/σ of 2.6. We employed the same molecular replacement and refinement protocols as before, starting with the initial model (PDB: 4I64), to solve the structure. After several refinement cycles without ligands, we introduced mevaldyl-CoA and NADH into our model. Refinement continued with phenix.refine until R-factors converged. Electron density observations for substrate and intermediate led to the addition of CoA, mevaldehyde, and NAD⁺ models to our crystal structure. We conducted B-factor refinements with phenix.refine, adjusting occupancies for mevaldehyde/CoA and mevaldyl-CoA from 0–1.0, ensuring their sum remained 1. Occupancy values eliminating Fo-Fc density for both ligand sets at the active site were identified. We then incorporated both ligand sets with these occupancies into the active site and performed further occupancy refinement cycles with TLS parameters using phenix.refine. This process generated our final structure model with the most accurate proportions of ligands at the active site.

Absorbance spectra, within the wavelength range of 250–900 nm, were collected using a 40-μm (full width at half maximum) light spotlight focused onto the sample. A dark spectrum was first obtained to obtain a measure of readout noise on the internal QE65000 CCD. Additionally, a reference spectrum was obtained, which measures maximum intensity from the light source in the absence of the sample. To monitor artifacts due to ice, the nylon loop, internal reflections from the frozen mother liquor, and other effects, each sample was rotated by the beamline goniometer as a series of spectra were measured (45). The phi-rotation mode was utilized to determine the best orientation to obtain UV-vis spectra and to reduce artifacts due to ice, cracks, nylon loop, and internal reflections from the frozen mother liquor while measuring the absorbance. To sample areas around the entire crystal, a rotation range of 0°–360° was used with a 5°–10° step size. The sample orientation that produced the highest-quality spectra was selected and used to obtain the final UV-vis spectra. All the spectra obtained were referenced to the lowest absorbance value in each of the spectral measurements at λ > 600 nm. Sample, dark, and reference spectra were used to calculate the sample absorbance using Eq. 1, where A is the calculated sample absorbance at wavelength $\lambda$ (AU), S is the sample transmitted intensity at wavelength $\lambda$ (counts), D is the background intensity at wavelength $\lambda$ (counts), and R is the reference incident intensity at wavelength $\lambda$ (counts).

$$A={\log }_{10}\left[\frac{R-D}{S-D}\right]$$ (1)

pH-rate profile measurements

The pH-rate profiles were generated by measuring turnover of the enzyme during the conversion of mevalonate to HMG-CoA in buffers with pH values ranging from 4 to 11 with mevalonate, CoA, and NAD⁺. The turnover is measured spectroscopically by following the rate of NADH formation by monitoring absorption at 340 nm during the PmHMGR reaction (Fig. 1). All measurements were conducted at ambient temperature at 1 atm.

Initial experiments compared the crystallization buffer to a buffer standard for turnover measurements. The optimized pH-rate profile buffer consisted of 100 mM Tris, 100 mM potassium phosphate, and 100 mM glycine and was designed to have a buffered pH range from 5.8–10.6 with a combination of potassium phosphate (pH range 5.8–8.0), Tris (pH 7.0–9.0), and glycine (pH 8.8–10.6). In addition, the effect of ammonium sulfate in the crystallization buffer was tested by measuring the pH-rate profile in additional ionic environments. These environments were composed of ammonium and sulfate salts with different counter ions. We also measured the pH-rate profile in the presence of sodium chloride to test the effect of high salt concentration in the absence of both ammonium and sulfate ions. To measure the turnover of the enzyme, the conversion of mevalonate to HMG-CoA was run in buffers with pH values ranging from 4 to 11 with mevalonate, CoA, and NAD⁺. The buffers used were 1) the crystallization buffer (1.2 M ammonium sulfate, 100 mM ADA, and 32% glycerol), 2) ammonium chloride buffer (1.2 M ammonium chloride, 100 mM ADA, and 32% glycerol), 3) ammonium acetate buffer (1.2 M ammonium acetate, 100 mM ADA, and 32% PEG-400), 4) sodium sulfate buffer (1.2 M sodium sulfate, 100 mM ADA, and 32% glycerol), and 5) lithium sulfate buffer (1.2 M lithium sulfate, 100 mM ADA, 32% glycerol).

The substrates mevalonate, CoA, and NAD⁺ were added to a 96-well plate with an assay volume of 100 μL. Then 5 μL of PmHMGR was added at a concentration of 0.1 mg/mL (20). Concentration of the substrates mevalonate, CoA, and cofactor NAD⁺ were 4, 0.51, and 2 mM respectively. These ligands were kept at the saturated concentrations referred to in previous assays of mevalonate oxidation (1). They were mixed with the protein in their respective pH solution and the change in absorbance at 340 nm was recorded for 10 min using a BioTek Synergy H1 Hybrid reader. The rate of NADH production was calculated from the absorbance values according to Eq. 2:

$${rate}_{NADH}=\frac{\Delta A340}{\left(T_f-T_i\right){\epsilon }_{NADH}l}$$ (2)

Here, ΔA340 is the maximum change in absorbance observed in the linear region of the progress curve in the activity assay for our enzyme. T_i and T_f are the initial and final time points, respectively, for the absorption measurement. Using the Beer-Lambert law, ${A340}_T={\epsilon }_{NADH}l{conc}_T$ where ${A340}_T$ and ${conc}_T$ are defined as the absorbance and concentration of NADH at time point T. Here, ${\epsilon }_{NADH}$ is the extinction coefficient of NADH (6220 M ⁻¹cm⁻¹) and l is the path length of the microplate reader well, in this case 0.3 cm (46). $\Delta A340$ = ${A340}_{Tf}-{A340}_{Ti}$ where ${A340}_{Ti}$ and ${A340}_{Tf}$ are the initial and maximum measured absorbance values observed over the period of first-order linear increase. The turnover of the enzyme was then measured using the following equation:

$$\text{Turnover}=\frac{\left({rate}_{NADH}\times {MW}_{enzyme}\right)}{\left(mg.ofenymeused\times Vol.ofwell\times {10}^{-3}\right)}$$ (3)

The molecular weight of the obligate dimer of PmHMGR is 91,181 g/mol (2,15). Seven replicate measurements and one control without the enzyme were obtained at each of the pH values that were tested in both the crystallization and enzymatic assay buffer conditions. An average of the seven replicates was used to determine the final turnover and standard deviations calculated at each pH.The turnover of the enzyme at various pH values was then plotted for different buffer conditions. Similar experiments were done for the variations in ionic composition.

Concentration-rate profile measurements

The turnover rate of PmHMGR was measured in the presence of varying concentrations of ammonium sulfate ranging from 0.02–1.8 M to measure the effect of the ion concentration of the crystallization precipitant salt ions NH₄⁺ and SO₄²⁻ on enzymatic activity. This range covers the concentration of ammonium sulfate used in the crystallization buffer (1.2 M). The turnover of the enzyme was measured as outlined in the pH-rate profile measurements but with varying concentrations of ammonium sulfate while keeping the concentrations of ADA and glycerol constant at the values used in the crystallization buffer (100 mM and 32% (v/v), respectively). Seven replicate measurements and one control without the enzyme were obtained at each of the concentrations of ammonium sulfate that were tested. An average of the seven replicates was used to determine the final turnover at each ammonium sulfate concentration.

pKa calculations of active site residues and ligands

To investigate the effect of the protonation state of the catalytic residues and ligands on the enzymatic reaction in the crystal environment, the pKa of each charged residue and ligand of the PmHMGR structure bound to CoA, mevalonate, and NAD⁺ was calculated using PROPKA 3.4 (47).

Results

Effect of the PmHMGR crystal environment on mevalonate oxidation from crystallographic studies

To uncover the steps in the PmHMGR reaction mechanism that have not been previously observed via structural studies, we were interested in using time-resolved techniques with the PmHMGR crystals, running the reaction with mevalonate, CoA, and NAD⁺ to observe the oxidative acylation of mevalonate (Figs. 1 and 2, step 2–3). Previous studies with dithio-HMG-CoA and NADH (Fig. 2, step 5–6) have shown that the reaction can proceed in the crystal under the same buffer conditions in which they were grown, resulting in the formation of a thiohemiacetal intermediate, accompanied by major conformational changes of the flap domain (16). We recently obtained a 2.0-Å structure soaked with the co-substrates mevalonate, CoA, and cofactor NAD⁺ in those same conditions in what appears to be an inactive state. Only one catalytic site of the homodimer in the asymmetric unit is fully occupied; the other has only density for mevalonate as this second active site is partially blocked by crystal contacts. Electron density was observed in the mevalonate-, CoA-, and NAD⁺-bound regions (Fig. 3 A) but not in the region between mevalonate and CoA that would denote the bond formation for the thiohemiacetal. This indicates that the reaction does not progress toward the formation of mevaldyl-CoA in the quaternary complex (Fig. 2, step 3–4), which was unexpected since this intermediate was shown to form when the reaction was run in the reverse direction. It should be noted that the difference between mevalonate and mevaldehyde is the changed oxidation state of a terminal hydroxyl or aldehyde group (Fig. 2, step 1–3). Thus, the substrate and the first intermediate of Fig. 1 are not distinguishable at our current resolution and it is possible that the structure reflects a single turnover step resulting in mevaldehyde and NADH.

Figure 3.

Electron density maps and models at the active site of the PmHMGR crystals. (A and B) Three sigma Fo-Fc omit map density (green) generated for ligands in the PmHMGR structures. (A) Omit map from the 2.0-Å X-ray structure for HMGR crystals soaked with mevalonate, CoA, and NAD⁺ showing clearly separated substrate and cofactors. (B) Omit map from the 1.65-Å crystal structure of mevaldehyde and CoA PmHMGR active site showing bond density between mevaldehyde and CoA in the PmHMGR active site. (C) Thiohemiacetal conformation in the structure obtained with mevaldehyde and CoA (green backbone) in comparison with the dithiohemiacetal conformation obtained in structure with dithio-HMG-CoA and NAD⁺ (purple backbone; PDB: 4I4B). Sulfur atoms are colored in yellow. (D) Key interactions between the small and flap domain in the NAD-binding site in the quaternary complex structure (green). (E) Comparisons of the ligand-binding region for the quaternary complex (green) and the mevaldehyde and CoA complex (teal).

To investigate the possibility that the crystal environment inhibits mevalonate oxidation (Fig. 2, step 2–3), we eliminated that step and crystallized the enzyme with the ligands representing the first reaction intermediate, mevaldehyde and CoASH, to test for mevaldyl-CoA formation (Fig. 2, step 3–4) under the crystallization conditions. The formation of a thioester bond between mevaldehyde and CoASH would then indicate that the crystallization conditions inhibit the first hydride transfer from mevalonate to NAD⁺ in the quaternary complex (Fig. 2, step 2–3). The refined F_o-F_c omit map density in the crystal structure shows clear density in regions where we expect CoA and mevaldehyde to bind and the emergence of an electron density in the C–S bond region between mevaldehyde and CoA (Fig. 3 B). A comparison with the previous dithiohemiacetal intermediate structure (PDB: 4I4B), produced from soaking experiments with a productive combination of a slow substrate dithio-HMG-CoA and NADH, demonstrates a close match of these structures (Fig. 3 C).

The similarity in thioester bond configuration after alignment indicates that the 5-carbon of mevaldehyde portion is sp³ hybridized as expected for mevaldyl-CoA. This provides clear evidence for the formation of a thiohemiacetal intermediate from CoA and mevaldehyde in the absence of NAD⁺ (Fig. 2, step 3–4). The differences seen in Fig. 3 C near the second sulfur atom in the structure of the dithiohemiacetal bound to PmHMGR (PDB: 4I4B) can be attributed to the differences in van der Waals radius of sulfur in comparison to oxygen. Since a thiohemiacetal can be created from mevaldehyde and CoA in the HMGR active site, these results indicate that the crystal environment inhibits mevalonate oxidation (step 2–3 in Fig. 2).

Although our focus in the quaternary complex and the mevaldehyde-CoA crystal structures is the demonstration that the forward reaction for PmHMGR is blocked at the point of hydride transfer, other mechanistic details can be derived from a comparison of these structures. The role of NAD⁺ in the closure of the flap is clearer, emphasizing how the active site construction is completed by delivering His381.

Other movements in the enzyme orchestrated by the contacts created on closure of the flap domain are observed in the quaternary complex. Movements in the adjacent parts of the small (Gly620–Ala725) and large domain (Pro824–Ser841) regions associated with cofactor binding are observed. Flap domain residues from Arg379–Leu422 are only observed in the NAD⁺-bound structure. We also see subsequent regions of the flap domain forming stable contacts with the diphosphate linkages of NAD⁺. In the active site then His381 forms contacts with Asn688 in the small domain, which then forms a stable linkage with the ribose region of NAD⁺ (Fig. 3 D).

In the absence of the NAD⁺ cofactor the conformation of the mevaldehyde- and CoA-bound structures appears to resemble the apo enzyme, indicating that the binding of CoA and mevalonate/mevaldehyde do not bring significant conformational changes to the enzyme. This is due to the easy and open access for CoA to bind via solvent channels and the small size of mevaldehyde/mevalonate. It is the binding of the cofactor that brings about significant changes in conformation resulting in the closure of the flap domain and contraction of the small and large domains. Minor changes are also observed between the apoenzyme and the thiohemiacetal-bound structure generated using mevaldehyde and CoA where regions of the flap domain from Gly376 to Gln378 are stabilized via hydrophobic interactions with the pantothenic acid region of mevaldyl-CoA.

Although both the structures are otherwise largely the same, they offer insight into the effect of substrate and cofactor on conformational changes that enable turnover. The closure of the flap domain observed in the cofactor-bound structure is shown to not be critical for thiohemiacetal formation as can be seen in the mevaldehyde- and CoA-bound structure. This indicates that either a mobile flap domain can still facilitate the deprotonation of CoASH for the formation of a thiohemiacetal or that even in the absence of a flap domain, the bond formation between mevaldehyde and CoA can take place via a purely chemical pathway accelerated by proximity. However, the closure of the flap domain in the presence of NAD⁺ does indicate that the flap closure and opening could have evolved to enable the exchange of the NAD⁺ cofactor post reduction that would allow a new cofactor to bind for the formation of HMG-CoA from mevaldyl-CoA oxidation. The changes in the flap domain that occur after the protonation of His381 could perhaps subsequently lead to changes that facilitate the release of the cofactor. Time-resolved studies that can follow changes that take place during and after thiohemiacetal formation would allow us to understand the changes that facilitate the re-opening of the flap domain.

The stereochemistry of the newly formed thiohemiacetal, mevaldyl-CoA, does not differ significantly from the mevalonate-, CoA-, and NAD⁺-bound structure. What is shown to be the conformation of R-mevalonate is also conserved in the placement of the thiohemiacetal. A drastic difference is observed in the placement of the thiol in CoA, which now forms a strong bond density with the C5 atom of mevaldehyde. These changes could occur after the deprotonation of CoASH by His381 after proton abstraction leading to subsequent changes in the placement of Ser85 and His381 (Fig. 3 E). Although the flap domain is no longer observed, we do observe changes in Ser85 placement indicating a loss of interaction between His381, Ser85, and the thiol atom of CoA. The changing coordination state of His381 with respect to CoASH might not be enough to result in an opening of the flap domain; however, it could trigger an onset of changes in the flap domain that could lead to flap domain opening and subsequent cofactor release. The placement of catalytic residues Glu83 and Lys267 that are present adjacent to the mevalonate portion of the thiohemiacetal is not affected, indicating that most changes that lead to the interaction of the CoA with mevaldehyde and subsequent bond formation are at the CoA end of the intermediate.

Investigating inhibition of mevalonate oxidation in crystallization conditions

To identify the factors of the crystallization environment that affect enzymatic turnover, we studied the individual crystallization buffer components with the goal of developing methods to initiate mevalonate oxidation in PmHMGR crystals for time-resolved X-ray crystallography. In addition to the buffer salts, we hypothesized that the crystallization pH might affect the turnover of the enzyme based on previously characterized pH-rate profiles for mevalonate oxidation in PmHMGR (12). Using these optimized conditions, we measured the pH-rate profile for mevalonate oxidation in the crystallization solution at saturated substrate concentrations. Upon measuring the rate of NADH production in the crystallization buffer in comparison to the enzymatic assay buffer, we found that, at the crystallization pH of ∼7, the turnover of the enzyme decreases by 92.7%, from 271.0 ± 22.6 min⁻¹ in the enzymatic assay buffer to 19.8 ± 3.3 min⁻¹ in the crystallization buffer (Fig. 4). The pH-rate profile also indicates that the maximum turnover of the enzyme decreases by 86. 4% in the crystallization buffer in comparison to the enzymatic assay buffer. The maximum turnover is observed at pH 9 in both environments (Fig. 4). Around pH 11, disappearance of enzyme activity is observed. These turnover measurements in the crystallization environment further demonstrate the pH dependence of the enzyme turnover with an increase in enzymatic activity between pH 5 and 9.

Figure 4.

pH-rate profile showing the trendline for turnover of HMG-CoA reductase in an optimized pH-rate profile buffer (Tris, potassium phosphate, and glycine, blue) and crystallization buffer (ammonium sulfate, ADA, and glycerol, green). Error bars represent the standard deviation from seven replicates at each pH.

Measuring the effect of ions from the crystallization buffer on enzymatic activity

The observation of a significant reduction in turnover in the crystallization environment led us to speculate that a cause for the drop in turnover in the crystallization buffer could be the significantly higher concentration of ions from the precipitant salt ammonium sulfate in the crystallization environment. Previous studies of ribonucleases provide precedence for reduction of turnover due to binding of sulfate ions from the common crystallization salt ammonium sulfate (48,49,50,51). To determine if a similar effect is present in the case of PmHMGR, we measured the turnover of the enzyme at different concentrations of ammonium sulfate ranging from 0.02 to 1.8 M in the presence of the other crystallization buffer constituents at a constant concentration (100 mM ADA and 32% glycerol). As shown in Fig. 5, there is a steady decline in turnover with increasing concentration of ammonium sulfate between 0.2 and 1.8 M with the activity decreasing by ∼97% from the lowest concentration to the highest concentration. This indicates a correlation between the concentration of the precipitant ions, NH₄⁺ and SO₄²⁻, in the crystallization environment and observed turnover.

Figure 5.

Turnover of PmHMGR at different concentrations of ammonium sulfate ranging from 0.02–1.8 M with 100 mM ADA and 32% glycerol (pH 6.7). Error bars represent the standard deviation from seven replicates at each pH.

At this point, it was unclear if this inhibitive effect was the outcome of a specific interaction between NH₄⁺ and SO₄²⁻ ions with the enzyme. To investigate this question, we determined the pH-rate profile of PmHMGR in the presence of several ammonium and sulfate salts. The concentration of the salts was kept the same as that of the crystallization precipitant (1.2 M). Substituting one of the ions in the crystal precipitant salt in these measurements, we determined if there is a specific effect of NH₄⁺ and SO₄²⁻ on enzyme turnover as a function of pH. The pH-rate profiles with several ammonium salts are shown in Fig. 6, indicating a similar degree of inhibition in the absence of sulfate ions with ammonium chloride and ammonium acetate across pH 4–11 with maximum activity at pH 9. The reduction in turnover is slightly lower in the presence of ammonium chloride (80.1%) in comparison to that observed with ammonium sulfate (86.4%) in the crystallization solution. In contrast, there is a stronger reduction in the presence of ammonium acetate (98.0%) across the entire pH range of pH 4–11, with maximum activity at pH 9. The turnover observed in the presence of ammonium salts indicated that PmHMGR is inhibited in the presence of a high concentration of NH₄⁺ ions even in the absence of SO₄²⁻.

Figure 6.

pH-rate profile showing the turnover of HMG-CoA reductase in an ammonium chloride, ADA, and glycerol buffer (brown); ammonium acetate, ADA, and PEG-400 buffer (blue); sodium sulfate, ADA, and glycerol buffer (yellow); lithium sulfate, ADA, and glycerol buffer (purple); crystallization buffer (red); and optimized pH-rate profile buffer (green). Error bars represent the standard deviation calculated from seven replicates at each pH.

Next, we investigated whether the reduced activity from the high salt concentration in the crystallization environment is entirely dependent on the presence of ammonium ions or depends on the presence of sulfate ions. The pH-rate profile was measured in the presence of lithium sulfate and sodium sulfate with the concentrations of ADA and glycerol kept constant. These experiments also indicated a significant drop in turnover across pH 4–11. The reduction in activity at the pH of maximum turnover (lithium sulfate, 89.3%; sodium sulfate, 82.5%) was equivalent to that measured in the presence of ammonium salts. This indicates that the reduction of turnover in the crystallization buffer is a result of a high concentration of both ammonium and sulfate ions each of which could have a separate effect on the enzyme (Fig. 4).

The observed turnover in the presence of both ammonium and sulfate salts suggests the possibility of the ion-dependent reduction in activity in the crystallization environment being dependent on the concentration rather than the chemical nature of the ions. To test the effect of a high ionic concentration in the absence of the crystallization salt ions NH₄⁺ and SO₄²⁻, we measured the turnover of PmHMGR at pH 7 and pH 9 in the presence of 1.2 M sodium chloride. The reduction in turnover in the presence of sodium chloride and was found to be very close to that observed in the crystallization buffer at pH 7 and 9 (Table 2). The significant reduction in turnover in the absence of both NH₄⁺ and SO₄²⁻ ions indicates that the ionic inhibition in the crystallization environment is not dependent on ionic composition but is largely a result of the high ionic concentration.

Table 2.

Turnover of PmHMGR at pH 7 and pH 9 in various buffers testing the inhibitive effect of sodium chloride in comparison to the crystallization and enzymatic assay buffer

Buffer conditions (pH 7)	Turnover pH 7 (min−1)	Turnover pH 9 (min−1)
100 mM Tris, 100 mM potassium phosphate, 100 mM glycine (enzymatic assay buffer)	270.9 ± 20.9	1367.94 ± 108.85
1.2 Ammonium sulfate, 100 mM ADA, and 32% glycerol (crystallization buffer)	19.80 ± 3.31 (92.6%)	190.11 ± 27.42 (86.1%)
1.2 M sodium chloride, 100 mM ADA, and 32% glycerol	36.92 ± 9.55 (86.4%)	124.35 ± 14.29 (90.9%)

Percentage drop in turnover in the presence of ammonium sulfate and sodium chloride are also included with turnover values.

Calculated pKa values of catalytic residues in the crystal structure

To gain further insight into the nature of the effect of the crystallization conditions, we assessed the protonation state of the catalytic residues in the PmHMGR structure. Identifying their protonation state would indicate whether these residues are suited to facilitate the proposed steps in the reaction mechanism at the crystallization pH. We calculated the pKa values of the catalytic residues Glu83, Lys267, and His381 involved in the enzymatic reaction from the crystal structure of the quaternary complex bound to mevalonate, NAD⁺, and CoA (Table 3). These calculations indicate that the pKa values of Glu83, Lys267, and His381 are 1.11, 11.27, and 7.45, respectively. These values indicate a high percentage of Glu83 and Lys267 that are charged and can act as proton acceptor and oxyanion hole, respectively, at the crystallization pH of 6.7. Conversely, a significantly lower population of His381 in the crystal would be in the neutral form (∼18%) at pH 6.7. The percentage of deprotonated His381 is expected to increase with pH, whereas the percentage of deprotonated CoASH with a calculated pKa of 12.65 would continue to remain relatively low under pH 12. Calculations made using PROPKA3 have been reported to have a root-mean-square error of 1.0 pKa for buried histidines and 0.79 pKa for aspartate and glutamate residues. In our case, where the supporting evidence for the changing protonation state of His381 is concerned, the range of ±1.0 pKa unit would still support the hypothesis that it is the increased concentration of deprotonated histidine above the calculated pKa (7.45) that results in increased activity (47).

Table 3.

Calculated pKa values for catalytic residues of PmHMGR with bound mevalonate, CoA, and NAD⁺

Residue	Glu 83	Lys 267	His 381	CoASH
Calculated pKa	1.11	11.27	7.45	12.65

Triggering of the reaction in the crystal using the pH-jump method

The dependence of enzymatic turnover in the crystal environment on the ion concentration and the crystallization pH suggests the intriguing possibility of triggering mevalonate oxidation by controlled changes in the PmHMGR crystal environment. Specifically, we hypothesized that the enzymatic reaction can be initiated in the crystal by transferring ligand-bound PmHMGR crystals from a pH where the enzyme is inactive to a higher-pH environment. This pH-jump approach is well established in spectroscopic (52) and crystallographic (53) studies of enzymatic reactions. We tested the viability of this approach by measuring changes in the UV-vis spectroscopic signature of NADH at 340 nm in crystals of mevalonate, CoA, and NAD⁺ bound to PmHMGR obtained in a 1.2 M ammonium acetate, 100 mM ADA, and 32% PEG-400 environment. This pH-jump buffer has a range of 4.75–9, allowing the transfer from the pH at crystallization of 6.7 to a higher-pH environment.

Insignificant absorbance at 340 nm is observed in the control crystal that is kept at pH 6.7 (blue in Fig. 7A), showing that the enzyme in the crystal is inactive at the crystallization pH. After the crystals are transferred to a pH 9 environment, the absorbance at 340 nm increases with time from 80 s to 8.5 min and then drops significantly after a soaking time longer than 10 min (Fig. 7 A). A notable exception in this shift is a drop in the absorbance at 5.5 min, which has an absorbance lower than that observed at 3.5 min. Such variations can be expected given that cofactor exchange is a key part of the reaction mechanism after the first hydride transfer and would result in an intermittent drop in absorbance at 340 nm. However, the UV-vis absorbance measurements in PmHMGR crystals before and after a pH change show the onset of NADH formation after a pH jump, indicating the expected initiation of enzyme activity. This confirms the possibility that a pH jump can trigger PmHMGR activity in the crystal for time-resolved studies.

Figure 7.

UV-vis spectra in ligand-bound (mevalonate, CoA, and NAD⁺) PmHMGR crystals of the size range (A) 400 × 200 × 200 μm and (B) 100 × 50 × 50 μm that have been introduced to a pH 9 environment for the selected time periods. The control measurement is from a 250 × 100 × 100-μm ligand-bound crystal at crystallization pH (6.7).

We also tested the feasibility of the pH jump in smaller PmHMGR crystals (100 × 50 × 50 μm) where the pH change can affect the crystal environment more rapidly, thereby achieving greater synchronization of reaction initiation. We still observe a continuous buildup of NADH across time (65 s to 6 min) followed by an eventual drop in the absorbance at 340 nm at 10 and 20 min (Fig. 7 B). We surprisingly also observe an increase in absorbance 30 min after reaction initiation (Fig. 7 B).

Last, we were able to obtain a single point 2.07-Å structure of the mevalonate-, CoA-, and NAD⁺-bound crystal 5 min after it had been introduced into a pH 9 environment. The aim of this test was to see if we could structurally observe product formation using the pH jump. Our crystal structure (Fig. 8.) shows a strong 2Fo-Fc density (1 σ) for the thiohemiacetal bond between the mevalonate/mevaldehyde and CoA-binding regions. Upon attempting a B-factor and subsequent occupancy refinement of the structure, we observed an occupancy of 0.3 and 0.5 for mevaldyl-CoA and CoA respectively, indicating a significant presence of the thiohemiacetal population in the mixture of states. Arriving at this set of occupancy values after successive refinements removed any Fo-Fc density present in the ligand-bound state. The flap domain appears to be closed in this structure with slight variations in its helical positions while forming contacts with NAD⁺ and CoA at the active site. Further analysis that is beyond the scope of this study is needed to assess the flap-ligand contacts that are affected during thiohemiacetal formation and the subsequent conformational changes that might follow.

Figure 8.

2Fo-Fc (1 σ) map obtained with mevalonate-, CoA-, and NAD⁺-soaked PmHMGR crystal after being introduced to a pH 9 environment for 5 min showing thiohemiacetal bond density.

Evidence from both spectroscopy and crystallography demonstrates that, under these new pH-jump conditions, the reaction can proceed at a controllably slow rate. This then should allow access to time-resolved steps that can fill in the gaps between the snapshots of the PmHMGR mechanism.

Discussion

The lack of turnover of the mevalonate oxidation in the crystal of the quaternary complex of mevalonate, NAD⁺, CoA, and PmHMGR under the initial crystallization buffer conditions has several important consequences that can provide additional insights into the complex mechanism of PmHMGR. First, using mevaldehyde as a substrate with CoA allows a direct observation of the production in a crystal of the previously postulated thiohemiacetal intermediate using natural substrates. A thiohemiacetal intermediate had been suggested previously based on the overall reaction mechanism (Fig. 1) (22) and earlier studies with the slow dithio-HMG-CoA substrate (16). The ability of the enzyme to proceed from the first intermediate in Fig. 1 to the second intermediate in the crystal has now been demonstrated by our crystallographic studies. Our conclusion is that the crystallization environment appears to inhibit the first hydride transfer from mevalonate to form mevaldehyde, although the mechanism for this is still unclear. However, this observation offers intriguing possibilities to control the reaction in the crystal. We have been able to spectroscopically observe the progression of this reaction step at higher pH, so now can apply this pH-jump method in time-resolved structural studies to understand the structural changes pertaining to mevalonate oxidation that prevent it from occurring at the crystallization pH. We have also demonstrated using a crystal structure at a single point after using the pH jump for reaction initiation that we can use it to observe in crystallo reaction intermediates for PmHMGR.

Detailed studies were made of the possible factors leading to the observed lack of activity in the crystal structure of what should be the productive quaternary complex, exploring the effect of the crystal buffer constituents and pH on the reactivity of PmHMGR. Initial observation of the pH-rate profile in the crystallization buffer shows a significant reduction of enzymatic activity across a range of pH 4–11 relative to the optimized buffer conditions for pH-rate profile measurements (Fig. 4). The ∼90% reduction in activity of the enzyme under these neutral pH conditions correlates with the lack of formation of the thiohemiacetal in the crystal. However, the pH-rate profile shows a trend of increasing turnover from pH 4 to 9 and subsequently drops from pH 10 to 11, providing access to a control mechanism for controlling in crystallo turnover.

The effect of the crystallization pH on turnover can be understood by considering the protonation states of the catalytic residues in PmHMGR in both soluble and crystallized states. Based on the calculated pKa of the catalytic residues in Table 3, Glu83 and Lys267 are expected to be fully deprotonated and protonated in the crystal, respectively. As shown by previous quantum mechanics/molecular mechanics calculations considering the reaction starting with HMG-CoA (18), these active-site residues in these protonation states can catalyze the reaction in the crystallization environment. However, the calculated pKa of His381, the residue essential for the deprotonation of CoASH, indicated that, at pH 6.7, the histidine sidechain is largely (85%) protonated in the crystal. Since a deprotonated form of the sidechain is needed to extract a proton from CoASH and subsequently form the thiohemiacetal intermediate with mevaldehyde (step 3–4, Fig. 2) it is less likely to occur at acidic pHs, potentially blocking mevaldyl-CoA formation. Increase in pH leads to an increased proportion of deprotonated His381 in the crystal environment, shifting the equilibrium toward thioester bond formation and the subsequent HMG-CoA formation as shown in Fig. 9. The changed turnover at alkaline pH may also be affected by the pKa of the CoA thiol, which is 9.83 in solution (12,54) and calculated to be at a much higher value in the crystal (Table 3).

Figure 9.

Proposed catalytic steps for PmHMGR.

Additionally, CoAS⁻ has been shown to result in a fivefold increase in the mevalonate oxidation reaction rate (12). Although the reasons for this have not been clearly established, it is possible that the presence of a highly nucleophilic thiol in proximity to mevalonate increases its electrophilicity during the hydride transfer step, thereby making it more prone to donate a hydride to NAD⁺. In contrast, CoASH in hamster HMG-CoA reductase has been shown to be an inhibitor (55).

Based on our evidence and theoretical postulations surrounding the changing protonation state of His381 thus far, it appears that, at pH 6.7, due to the high population of protonated His381, CoA exists in a protonated state as well. Based on this speculation, we hypothesize that the protonated CoA interferes with the process of hydride transfer for mevalonate in the following manner: Since the protonation states of Glu83 and Lys267 allow the proton abstraction from mevalonate to take place, we still expect the mevalonate to lose one proton to Glu83. Although Lys267 is then hypothesized to be stabilizing the oxyanion hole, we postulate that the presence of CoASH near the deprotonated radical quickly results in a re-protonation of the mevalonate from CoASH, resulting in the formation of mevalonate and CoAS⁻ at the active site and thereby preventing the oxidative acylation of mevalonate from taking place. At pH 9, we expect CoA to be donating a proton to His381, where the population of this residue should be largely deprotonated. Therefore, we expect CoA to be largely in the CoAS⁻ form where it should not be able to re-protonate the mevalonate after the proton abstraction step, thereby allowing the mevalonate oxidation step to proceed.

Our experiments have also shown that the turnover rate of the soluble PmHMGR enzyme depends on the ionic concentration but not on the identity of the ions, as demonstrated by the similar degree of inhibition across both ammonium (ammonium chloride and ammonium acetate) and sulfate (sodium sulfate and lithium sulfate) salts. The observation suggests that the inhibition in the crystal is not due to specific interactions of the enzyme or its substrates/cofactors with the ions in the crystal environment. This hypothesis is further supported by the observation that the presence of sodium chloride at pH 7 and 9, where neither of the crystallization precipitant salt ions were present (Table 2), leads to the same degree of inhibition. In the presence of the different ammonium salts studied, all the pH-rate profiles showed maximum turnover at pH 9. The pH at which maximum turnover was recorded was found to change to pH 10 and 11 in the absence of ammonium with sodium and lithium sulfate, respectively. This difference between ammonium and sulfate salts indicates that, although the ionic composition of the crystallization salt does not significantly affect the degree of inhibition of turnover across pH, it may influence the pH-dependent variation in turnover.

Analysis of these results provides the basis for the design of a new reaction-triggering method for crystalline PmHMGR catalysis of mevalonate to HMG-CoA based on a pH jump (52,53). The implementation of this method uses crystals obtained at the previously described (16) pH of 6.7 and transferred into a buffer environment consisting of 1.2 M ammonium acetate, 100 mM ADA, pH 6.7, and 32% PEG-400 (cryoprotectant). Slow addition soaking can now be done with the ligands mevalonate, CoA, and NAD⁺ into an enzyme that is inactive at this pH. These ligand-bound crystals are then rapidly transferred to a higher-pH environment (pH 9) to initiate enzymatic activity with a relatively slow turnover. At different time points after the transfer, the crystals can be freeze-trapped and used for time-resolved studies. As confirmation for turnover in the crystal using this approach, we used UV-vis spectroscopy to monitor NADH formation at 340 nm. The absorbance increases for several minutes, reflecting the hydride transfer reaction, followed by a steep decline in absorbance that indicates the eventual release and diffusion of the reduced cofactor out of the enzyme crystal before it was frozen. The rate of change of absorbance in these crystals shows that the enzymatic reaction rate in this crystalline environment is on the order of minutes. This outcome is confirmed in the structure of a crystal frozen at a time point 5 min into the pH-jump protocol, where clear density for formation of a thiohemiacetal is revealed.

Although the trend across both the 0.4 × 0.2 × 0.2-μm and 0.1 × 0.05 × 0.05-μm crystals is largely toward an increasing absorbance for NADH at 340 nm, we observe an outlier for the crystal obtained 5.5 min after being introduced to a pH 9 environment where the absorbance, although present, goes below that of the previous time point at 3.5 min. Such an outlier can be expected in this measurement since the reaction also has a cofactor exchange step that occurs after the first hydride transfer. Although such a change would be hard to determine structurally if NADH were to be replaced with NAD at the active site with a closed flap domain, it would still result in a lower absorbance at 340 nm.

Although the largely increasing rate of change in NADH absorbance over time and presence of intermediate density in the post pH-jump structure indicate a slow unidirectional progression of the reaction in crystallo, one cannot discount the possibility of there being multiple turnover steps occurring at the same time. In case additional pH-jump structures do not point toward a synchronized progression of the reaction at other time points, the pH-jump method can be applied to obtain diffraction data within the time frame that allows us to observe changes for a single turnover.

This pH-jump method is a promising approach for triggering the mevalonate oxidation step in PmHMGR crystals to make it amenable for capturing reaction intermediates using freeze-trapped methods. Slowing the reaction mechanism by pH manipulations also could allow us to investigate earlier dynamic states that could reveal why the hydride transfer may be affected in the conversion of mevalonate to mevaldehyde. Alternatively, the pH jump could be used with new serial crystallography/time-resolved X-ray techniques to obtain dynamic structural information for various transitory states along the reaction pathway, thereby providing an understanding of the structural and biochemical features of PmHMGR that facilitate intermediate formation, cofactor exchange, and product release along the reaction pathway of Fig. 2.

With the advent of new approaches such as the use of on-chip crystallization, microfluidic channels, and liquid droppers for introducing chemical triggers in enzyme crystals, we believe that the pH-jump method can also be utilized to obtain time-resolved crystallographic information from single PmHMGR crystals, thereby reducing the variability inherent to freeze-trapping multiple samples (56,57,58,59,60,61). Exploiting the inhibitive properties of crystal environments arising from the ionic concentration of various precipitants used and the subsequent differences in enzyme activity with pH has the potential to be utilized to design pH-dependent reaction-triggering methods across various biological systems. By appropriate control of the pH, the reaction rate after triggering the enzyme reaction can be adjusted to match available data collection capabilities and the dynamics of the enzymatic systems involved.

Conclusions

The PmHMGR crystal environment that has been previously found to be suitable for observing the conversion of HMG-CoA to a thiohemiacetal was found to inhibit mevalonate oxidation (16). We demonstrated that overall inhibition of this reaction step is a result of a high ion concentration from the precipitant salt in the crystal environment. This ion-induced inhibition has been found to be independent of ion composition and can be observed in the presence of various salts. The enzyme has also shown pH-dependent changes in turnover across environments where its turnover is largely observed to increase from pH 5 to 9 and to subsequently decrease at pH 10 and 11.

The drastic reduction in activity from high precipitant salt concentrations coupled with pH-dependent changes in turnover has been used to develop a triggering pH-jump method to observe mevalonate oxidation in the PmHMGR crystal environment. Making use of an alternate environment, we introduced the ligands mevalonate, CoA, and NAD⁺ into an inactive enzyme. We have then been able to subsequently trigger activity by rapidly transferring these crystals to a higher pH in the same buffer conditions. The triggering of enzymatic activity has been observed using UV-vis spectroscopy measurements to detect the appearance of a unique absorbance peak for the reduced cofactor NADH and confirmed with the structure of a post pH-jump crystal. We plan to utilize this pH-dependent reaction triggering method to obtain dynamic time-resolved structural information along the PmHMGR reaction pathway and further study the molecular details of the complex mechanism of HMG-CoA reductase.

Author contributions

V.P., C.N.S., A.R.R., C.V.S., C.J.C., P.H., O.W., A.M., and A.E.C. contributed to the design of the research and the data analysis. V.P., C.N.S., C.J.C., and T.S. performed the crystallographic research and data processing. V.P., C.N.S., and A.R.R. performed the kinetics and pH profile experiments. V.P. and A.E.C. performed the UV spectroscopy with HMGR crystals. C.V.S., P.H., and O.W. supervised the research and were primary contributors, with V.P. in writing the manuscript.

Editor: Ronald Koder.