¹³C ENDOR Spectroscopy-Guided MD Computations Reveals the Structure of the Enzyme-Substrate Complex of an Active, N-linked Glycosylated Lipoxygenase

Abstract

Lipoxygenase (LOX) enzymes produce important cell-signaling mediators yet attempts to capture and characterize LOX-substrate complexes by X-ray co-crystallography are commonly unsuccessful, requiring development of alternative structural methods. We previously reported the structure of the complex of soybean lipoxygenase, SLO, with substrate linoleic acid (LA), as visualized through the integration of ¹³C/¹H electron nuclear double resonance (ENDOR) spectroscopy and molecular dynamics (MD) computations. However, this required substitution of the catalytic mononuclear, nonheme iron by the structurally faithful, yet inactive Mn²⁺ ion as a spin-probe. Unlike canonical Fe-LOXs from plants and animals, LOXs from pathogenic fungi contain active mononuclear Mn²⁺ metallocentres. Here, we report the ground-state active-site structure of the native, fully glycosylated fungal LOX from rice blast pathogen M. oryzae, MoLOX complexed with LA, as obtained through the ¹³C/¹H ENDOR-guided MD approach. The catalytically important distance between the hydrogen donor, carbon-11 (C11), and the acceptor, Mn-bound oxygen, (donor-acceptor distance, DAD) for MoLOX-LA complex derived in this fashion is 3.4 ± 0.1 Å. The difference of the MoLOX-LA DAD from that of the SLO-LA complex, 3.1 ± 0.1 Å, is functionally important, yet is only 0.3 Å, despite the MoLOX complex having a Mn-C11 distance of 5.4 Å and a ‘carboxylate-out’ substrate binding orientation, whereas the SLO complex has a 4.9 Å Mn-C11 distance and a ‘carboxylate-in’ substrate orientation. The results provide structural insight into reactivity differences across the LOX family, give a foundation for guiding development of MoLOX inhibitors, and highlight the robustness of ENDOR-guided MD approach to describe LOX-substrate structures.

Graphical Abstract

Introduction

Lipoxygenases (LOXs) are a family of enzymes widely represented in plants, animals, fungi and select prokaryotes.^1–3 LOXs oxidize polyunsaturated fatty acids to form a diverse array of potent bioactive cell-signaling mediators. For example,¹ the six human LOX isozymes produce a variety of pro-inflammatory oxylipins linked to chronic inflammation, including atherosclerosis, diabetes, stroke, asthma and select cancers.⁴ Despite the clinical importance of these human LOXs, there is only one FDA approved anti-inflammatory drug, Zileuton, that targets human 5-LOX activity linked to asthma.^{5, 6} The dearth of LOX therapeutics is due in part to the lack of high-resolution structural information on the enzyme-substrate (ES) complexes, with only a few representative co-crystal structures available of LOXs with natural or mimetic substrates, or isoform-selective inhibitors.^7–10 This contrasts with cyclooxygenase (COX) enzymes, where the availability of high-resolution structures of protein-inhibitor complexes for both isoforms have aided the development and evolution of non-steroidal anti-inflammatory drugs (NSAIDs).⁶

In addition to its medical relevance, the initial C-H bond cleavage associated with the LOX reaction occurs by a non-classical, hydrogen tunneling process, and extensive experimental and theoretical studies have advanced our understanding of the critical link between protein structure, dynamics, and the origins of enzyme catalysis.¹¹ Thus, implementation of alternative structural methods, complementary to traditional x-ray structural determination, is needed for acquiring detailed information about active protein configurations that can provide a framework to facilitate the design of structure-based inhibitors for therapeutic intervention and to shed additional light onto the structural regulation of enzymatic hydrogen tunneling.

LOXs are also important in pathogenic organisms. Of the ten most pathogenic plant fungi,¹² the rice blast fungus, M. oryzae, is considered the most destructive: M. oryzae is responsible for the loss of nearly 1/3 of the world’s rice crops, causing the loss of enough food to feed up to 60 million people. The lipoxygenase from M. oryzae (MoLOX), the focus of this report, is synthesized at the onset of pathogenesis and is secreted by the fungus along with lipases.³ In vitro studies have shown that MoLOX oxidizes both polyunsaturated fatty acids and phospholipids, generating unique bis-allylic hydroperoxides (Scheme 1) that are thought to induce oxidative damage and necrosis of rice leaves.^{3, 13} Thus, it has been proposed as one of the prime agents causing the pathology of rice blast disease.⁴ Development of inhibitors for MoLOX could help to determine the mechanism(s) of MoLOX in pathogenesis and potentially provide a successful intervention for rice blast disease.

Scheme 1.

Mechanism of oxidation of linoleic acid, LA, by MoLOX, involving a rate-limiting C-H cleavage catalytic step by proton-coupled electron transfer (PCET) involving a hydrogen tunneling mechanism.²⁸ MoLOX can also generate 9S/R- and 13S-hydroperoxides through β-fragmentation.

As with most secreted enzymes, MoLOX is a glycoprotein decorated with N-linked glycans on its surface. One of the inherent challenges in studying glycoproteins is the difficulty in characterization of the native protein structure because the branched and inherently flexible carbohydrates inhibit crystallization.¹⁴ To overcome this challenge, many studies resort to the modification or (partial or full) removal of the glycan(s) to enable crystal packing for high-resolution structural analysis. A crystal structure has been reported for MoLOX with modified glycans and in the absence of a substrate at 2.04 Å resolution.¹⁵ This previous study used a MoLOX sample treated with endonuclease H (EndoH), an endoglycosidase which removes mannose-rich oligosaccharides linked to asparagine residues on the surface of proteins (Figure 1A).¹⁵ However, for many N-linked glyco-enzymes, removal of the carbohydrates is linked to altered catalytic proficiency.^16–19 As there is of yet no structure of the fully glycosylated form of MoLOX, the impact of the removal of the N-linked glycans on the ES complex structure and on its enzymatic function is not well known. Therefore, to facilitate structure-guided inhibitor design for MoLOX, it is imperative to resolve the structural details of the ES complex for the native glyco-enzyme.

Figure 1.

Structure¹⁵ of MoLOX (A) with the predicted site asparagine residues for N-linked glycosylation depicted in green spheres (N18 and N28 are missing from the X-ray structure). The parentheses represent the predicted potential for N-linked glycosylation. Also shown in spheres are the catalytic manganese center (purple) and the structurally conserved active site residues, L331 (red) and F526 (blue), that mediate substrate positioning for efficient hydrogen tunneling. (B) and (C) represent the first- and second-shell ligands to the mononuclear metal center in MoLOX and soybean lipoxygenase (SLO), a model iron LOX. The second-shell ligands are labeled, and color coded for reference. Note that the MoLOX numbering here is based on the processed enzyme that lacks the N-terminal signaling peptide sequence.

We previously utilized electron nuclear double resonance (ENDOR) spectroscopy to capture high-resolution, 3-D structural information about the elusive ES complex of the model plant lipoxygenase from soybean (SLO) with its substrate linoleic acid, LA.^{20, 21} Although ENDOR is most commonly used to characterize the coordination sphere of biological metal ions by interrogating ligands coordinated to the paramagnetic center,²² it can also provide high-precision structural information for atoms in the active site that are in close proximity to the metal ion but not covalently linked to it,^{23, 24} as is the case in the ES complex of SLO-LA.¹⁴ In the previous SLO study, to achieve the proper electronic properties for ENDOR analysis of the lipoxygenase the nonheme mononuclear iron was substituted by the non-native, high-spin (S = 5/2) Mn²⁺ ion as a spin-probe (Mn-SLO). ¹³C/¹H ENDOR spectroscopy then revealed the position and orientation of substrate LA relative to the enzymatic metal center, and further gave information about the distribution in positioning.^{20, 21} ENDOR analysis and Molecular Dynamics (MD) simulation combined not only provide the important acceptor (Mn-bound Oxygen) to donor (C11 of LA) distance (DAD) in the ground state, but also corroborate the previous theoretical understanding that, enzymes achieve reactive hydrogen tunneling geometries at the active site through protein thermal motions, starting from their ground state geometries.

Unlike the canonical Fe-LOXs found in prokaryotes, plants and animals, LOXs found in pathogenic fungi, such as MoLOX, are naturally equipped with a catalytically active Mn²⁺ cofactor (Mn-LOX).^{13, 25, 26} In the current study, we use EPR and ENDOR spectroscopies to guide MD simulations that yield the first active-site, ground state ES structure for the native, N-linked glycosylated form of the fungal Mn-LOX, MoLOX with bound LA substrate. We further compare that to the similarly-derived ES structure of a de-glycosylated form of MoLOX to test the effects of the surface glycans on the active-site structure. The results provide structural insights into the evolutionarily divergent subfamily of fungal LOXs that give a foundation for guiding development of MoLOX inhibitors.

Materials and Methods

Materials.

Dideuterated and ¹³C labeled linoleic acid substrates were synthesized previously.²⁷ All yeast/bacterial cells, media, salts, and buffers were purchased from Fisher Scientific, Sigma-Aldrich, or VWR at the highest grade possible.

MoLOX Expression and Purification.

WT MoLOX was expressed in and purified from P. pastoris X-33 cells as previously described.²⁸ G. graminis lipoxygenase, GgLOX (gene synthesized by Genscript), was expressed and purified in a similar manner as MoLOX. The final purification for all variants was carried out using a HiPrep 26/60 Sephacryl S-200 column on an AKTA FPLC system with 50 mM HEPES (pH 7.5), 150 mM NaCl, and 10% glycerol. The fractions that corresponded to the peak with lipoxygenase activity were concentrated to 100 μM, frozen in aliquots with N₂ (l), and stored at −80°C. Mn content was determined using ICP-MS. Site-directed mutants were generated with the Qiagen QuickChange kit.

EndoH Deglycosylation.

An endoglycosidase H (EndoH) gene from Streptomyces plicatus was synthesized by GenScript for recombinant expression in E. coli and subcloned in-line with his-tagged MBP construct (2CT-10 vector, Addgene plasmid #55209; gift from Scott Gradia). Recombinant MBP-EndoH fusion was expressed in and purified from E. coli BL21 (DE3) cells. In brief, EndoH expression was induced by 1 mM IPTG once OD₆₀₀ reached 0.6. Upon induction, the cultures were dropped to 18 °C and incubated overnight. Cells were lysed by sonication in lysis buffer (50 mM sodium phosphate, 100 mM NaCl, 8% glycerol, 2 mM MgSO₄, pH 7.5, supplemented with lysozyme, DNAse I, and AEBSF). The protein was purified using NiNTA chromatography, dialyzed in 50 mM Tris, 150 mM NaCl, pH 8 buffer and stored at −80°C until use.

EndoH MoLOX was prepared by treating WT MoLOX with EndoH in a 20:1 (MoLOX-EndoH) mass ratio. The reaction was carried out at 20°C overnight in 50 mM sodium acetate, pH 5.5. The optimal reaction conditions were determined using SDS-PAGE. In our hands, the reaction did not require α-mannosidase for complete digestion, as previously suggested.¹³ For large scale preparations, after the reaction, MoLOX was separated from MBP-EndoH by passing the reaction over a NiNTA column equilibrated with 50 mM Tris (pH 8), 500 mM NaCl and 10 mM imidazole and then further purification using SEC FPLC. The removal of the glycans was confirmed by SDS-PAGE (Figure S1).

Circular Dichroism (CD).

MoLOX secondary structure and thermal stability was assessed by CD spectroscopy using a Jasco model J-815 CD spectrometer with bandwidths of 2 nm using a Starna cuvettes (path length of 0.1 cm). Samples were prepared at 3 μM in 25 mM potassium phosphate (pH 7). Under these conditions, the PMT voltage (HT) remained ≤ 600V in the range of 190–260 nm. Measurements for stability were also carried out with wavelengths set to 222 nm and the temperature range of 25–90°C (2°C intervals) with a temperature ramp up rate of 0.6°C/min.

EPR/ENDOR Measurements.

35 GHz Pulsed EPR/ENDOR measurements were collected at ∼2 K on a spectrometer described previously, with SpinCore PulseBlaster ESR_PRO 400 MHz digital word generator and Agilent Technologies Acquiris DP235 500 MS/s digitizer using SpecMan4EPR software.^{29, 30} The pulsed EPR employed two-pulse electron spin echo sequence, π/2−τ−π−τ-echo. The Mims pulsed ENDOR sequence (π/2−τ−π/2−T_rf−π/2−τ−echo, with rf pulse T_rf) was used to probe the ¹³C hyperfine coupling of C10, C11 nuclei of labeled LA substrate. The Davies ENDOR sequence (π-T_rf-π/2-τ-π-τ-echo, T_rf is the rf pulse) was used to measure ¹H couplings of H₂O coordinated to Mn²⁺. 35 GHz CW EPR spectra were collected using a modified Varian E-110 spectrometer equipped with a helium immersion dewar (∼2 K).³¹ Simulation of EPR spectra was carried out using EasySpin.³² Simulation of ¹³C10, ¹³C11, ¹H, ENDOR spectra were carried out using a purpose-written (ad hoc) program in Mathcad Prime.^®.²¹

Orientation and length of Mn-C10/11 vectorss.

The orientation of Mn-C vector in the ZFS coordinate frame is defined through Euler angles (θ, ϕ), where “θ” is the angle away from the unique ZFS axis (z) and “ϕ” defines the rotation in the x-y plane away from the ZFS x-axis. The ZFS-splitting of the S = 5/2 Mn²⁺ ions introduces orientation selection in the ¹³C ENDOR spectra for a labelled LA site such that the ¹³C-ENDOR spectrum at each external field across the EPR envelope interrogates a well-defined set of orientations of the field relative to the Mn-¹³C vector, r, and analysis of this effect allows determination of the orientation of T relative to the ZFS frame. When the magnetic field is set to the low-field edge of the EPR spectrum, the signal arises only from the $m_s-5/2\leftrightarrow -3/2$ manifold (Figure 2A) and one observes a single-crystal-like ENDOR spectrum with the field along the ‘y’ ZFS axis, while at the high-field edge one obtains an ENDOR spectrum from the same manifold, with the field pointing along the ‘z’ ZFS axis. ENDOR at each intermediate field is the sum of contributions from a well-defined set of orientations associated with the $-5/2\leftrightarrow -3/2$ manifold, with the $m_s=-3/2\leftrightarrow -1/2$ manifold contributing detectably. The lengths of the Mn-C10/11 vectors, namely the Mn-C10/11 distances, were determined from the magnitudes of the ¹³C hyperfine couplings as described below.

Figure 2:

(A) MoLOX EPR spectra: absorption-display EPR envelopes of individual contributions from five electronic transitions as differentiated by colored lines. (B) 35 GHz ESE-EPR spectra of MoLOX with and without LA. The simulations of the spectra are shown as red dashed lines. Experimental conditions: microwave (MW) frequency, 34.9 GHz; MW pulse length (π/2), 40 ns; τ, 500 ns; repetition rate,100 Hz; temperature: 2 K. For simulation parameters, see Table 1.

Docking and Molecular Dynamics.

The initial structure of the WT MoLOX enzyme was obtained from the PDB (code 5FNO¹⁵). The H++ webserver^{33, 34} was used to determine the protonation and to add hydrogen atoms under pH=9.0. Afterwards, the protonation states of the metal site residues were manually corrected if necessary due to the H++ webserver does not consider the metal ions or water molecules when determining the protonation states. The AutoDockVina program³⁵ was used to dock the linoleic acid (LA) into the protein binding pocket. A cubic box centered at the active site (with center has the coordinates [8.50 Å, 47.65 Å, 124.90 Å]) with length of 20 Å was used for the search space during the ligand docking. The MoLOX model was represented by the AMBER12 forcefield^36–39 with extensions for the LA ligand and Mn metal site. The partial charge for the LA⁴⁰ and force constant of the Mn-ligand⁴¹ are taken from the literature. More details about the metal site parameters can be referred to our previous research.²⁰

In the next step, hydrogen atoms were added and optimized the system. A periodic rectangular box of size 96 Å x 86 Å x 76 Å was used to solvate the protein-ligand system. The TIP3P water model was used to model the solvent molecules.⁴² 48 Na⁺ ions and 46 Cl⁻ ions were added to neutralize the system and provide a salt solution of NaCl with concentration of ~0.13 M in the production runs. The following simulation procedure was performed for the protein-ligand system: (1) In the first equilibration step, with the protein-ligand system fixed, the solvent molecules and Na⁺ and Cl⁻ ions were minimized for 10000 steps and equilibrated for 100 ps at 300 K with a 2 fs timestep in the NVT ensemble. (2) Then the system was heated to 300 K in four steps 50K🡢100K🡢200K🡢300K. In each temperature steps, the system was minimized for 10000 steps and then propagated for 100 ps with a 2 fs timestep under 1 atm in the NPT ensemble. (3) Afterwards, two replicas were performed, for each has the system minimized for 20000 steps and then equilibrated for 20 ns at 300 K and 1 atm with a 2 fs timestep in the NPT ensemble, and finally has the production run for 20 ns at 300 K and 1 atm with a 2 fs timestep in the NPT ensemble.

During Steps 2–3, the ENDOR derived distances were applied to restrain the distance between Mn²⁺ ion and C10/C11 atoms of substrate, with the force constant as 200 kcal/mol/Ų, a value which was utilized and optimized previously for Mn-SLO,²⁰ and comparable to other force constants used for SLO simulations.⁴³ As was previously acknowledged for MD simulations with SLO,⁴³ restraints were necessary to prevent the dissociation of the substrate from the MoLOX active site during the timescale of the MD simulations.⁴⁴ While the stiffness of the MD restraints may influence the absolute DADs determined from the MD simulations, there is close correspondence of the DADs in the WT SLO-LA complex calculated from the restrained, classical MD simulations ²⁰ and the ‘unrestrained’ quantum mechanical/molecular mechanical (QM/MM) simulations.⁴⁴ Given the use of the same force constant for both sets of MD simulations, we expect a similar influence on the SLO-LA and MoLOX-LA complexes. Note that there was no restraint imposed on the carboxylate of substrate in the active site pocket. Snapshots were saved every 40 ps during the production MD runs, yielding a total of 1000 frames for the final analyses. All the MD calculations were performed by using the NAMD-2.13b1 program.⁴⁵

Results

Mn²⁺ EPR Spectroscopy of Mn-LOXs:

We start by describing the Mn²⁺-EPR spectrum of native MoLOX in the light of protocols established earlier for the manganese-substituted soybean lipoxygenase (Mn-SLO).²⁰ The EPR spectrum of the $\text{S}=5/2$ Mn²⁺ ion of MoLOX (Figure 2A), like that of Mn-SLO (Figure S2), is the sum of the ‘envelopes’ of the five orientation-dependent $\Delta m_s=\pm 1$ transitions between adjacent pairs of the six electron-spin $m_s$ substates, $-5/2\le m_s\le +5/2$. Each envelope has a well-defined shape that is spread over a range of fields defined by its $m_s$ values and the magnitudes of the zero-field splitting (ZFS) parameter, $D$, and a rhombicity parameter, $\text{λ}$, with each envelope weighted by the thermal population of the contributing $m_s$ levels. The overall breadth of the EPR spectrum is proportional to the magnitude of $D$, which increases with deviations of the Mn²⁺-coordination sphere from spherical symmetry. The shape of the spectrum is determined by the rhombicity parameter, $0\le \text{λ}\le 1/3$, which reflects deviations from axial symmetry around the principal ZFS axis $(\text{λ}=0)$.

The value of $D$ of MoLOX is three-fold larger than that of Mn-SLO (Figure 2B and S2, see Table 1) which is manifested as three-fold larger magnetic field span of the EPR spectrum of MoLOX than that of Mn-SLO. The Mn-coordination sphere of MoLOX is a distorted octahedron that doesn’t overlap exactly with the Fe-SLO structure (Figure 1B,C).¹⁵ This subtle difference in the position and orientation of Mn-coordinating ligands in the two Mn enzymes causes the appreciable differences in ZFS as evidenced in the EPR spectra. Another native, functional Mn-LOX from the fungal family Gaeumannomyces graminis, GgLOX also shows a broad EPR spectrum like that of MoLOX and has similar ZFS parameter (Figure S2, Table 1).²⁶

Table 1.

ZFS Parameters of Mn²⁺ in Mn-LOX

Mn-LOX	substrate	D (MHz)a	$\text{λ}$	$f$
MoLOX	−LA	+3500	0.14	0.6
	+LA	+3000	0.17	0.3
GgLOXb	−LA	+2100 to +3000	0.13–0.23	N.R.
Mn-SLO c	−LA/+LA	+1200	0.17	0.3

^a)D is the axial ZFS parameter, $\text{λ}=E/D$ is the rhombicity. In the simulations, $D$ and $E$ each has a width that is the fraction, $f$, times the parameter (See SI).

^b)Native manganese lipoxygenase from fungus, G. graminis; values reported from ref ²⁶ N.R., not reported.

^c)values reported from ref ²⁰

Unlike Mn-SLO, the EPR spectrum of MoLOX is notably changed by the addition of the substrate, LA, (Figure 2B and S1) with well-defined sharpening of the contribution from the different electron spin manifolds, corresponding to a two-fold decrease in the distribution $f$ in the ZFS parameters, along with a small reduction in the ZFS splitting, $D$ (Table 1). The distribution $f$ in ZFS parameters reflects the degree of flexibility to the Mn coordination sphere. Although LA is not directly ligated to Mn²⁺and rather is in the first coordination sphere as we will learn from ENDOR analysis (see below), the sharpened Mn²⁺ EPR spectrum upon substrate binding indicates improved ordering of the Mn coordination geometry, while the small reduction of ZFS splitting suggests a subtle change in the ligand arrangement.

ENDOR of Mn²⁺ (S = 5/2) in MoLOX:

For a single molecular orientation of a paramagnetic center of spin $S=\mathbf{5}/\mathbf{2}$, the first-order ENDOR spectrum for an $I=1/2$ nucleus (¹³C, ¹H) obtained by monitoring an EPR transition between adjacent substates, $m_s\iff m_s+1$ is a doublet with a signal from each of the substates. We here write the two frequencies in terms of the nuclear Larmor frequency, ${\text{ν}}_{\text{N}}$, the substate spin projections, and $A$, the orientation-dependent electron-nuclear hyperfine coupling interaction.

$$\begin{gathered} {\text{ν}}_{ms}=\left|{\text{ν}}_N-m_{s}A\right|\hspace{1em}\hspace{1em}\hspace{1em}{m}_s=\text{−5/2, −3/2, − 1/2, 1/2, 3/2} \\ {\text{ν}}_{ms+1}=\left|{\text{ν}}_N-\left(m_s+1\right)A\right| \end{gathered}$$ (1)

A ¹³C nucleus of a non-coordinated, isotope-enriched LA substrate bound in the vicinity of the active-site Mn²⁺ experiences a through-space electron-nuclear hyperfine dipolar interaction with the electron spin of the Mn²⁺ center. Although the Mn²⁺ spin is partially delocalized over its coordinating ligands, we have shown²⁰ that for the electron-nuclear interaction with the remote ¹³C of bound LA, this interaction can be taken as a point dipole coupling with the Mn²⁺ ion, ²⁰ with the dipolar hyperfine tensor $T\left(r\right)$ defined as,

$$T\left(r\right)=\left[-T(r),-T(r),2T(r)\right]$$ (2)

$$T(r)=\frac{{\rho }_{eff}{g}_e{\beta }_e{g}_C{\beta }_C}{r^3}\stackrel{{\rho }_{eff}\approx 1}{\to }\frac{19.9}{r{(\AA )}^3}MHz$$

where, because of the large values of r that occur, the effective Mn spin density can be taken as, ${\text{ρ}}_{\text{eff}}\approx 1$ρ_eff ≈ 1, and thus $T\left(r\right)$ is simply determined by the length of the Mn²⁺ → ¹³C vector $\text{(}r\text{)}$, while the unique axis points along this vector.²⁰

As described in the Methods section, analysis of the full 2D field-frequency pattern of multiple ENDOR spectra collected across the entire EPR envelope of MoLOX-LA, as exemplified by the experimental 2D patterns for ¹³C10 and ¹³C11 of LA bound to MoLOX (Figure 3), enables determination of the dipolar interaction tensor, T, for each of the two nuclei ¹³C (C10, C11), thereby yielding the Mn-C distances $r$ (eq 2) and the spatial orientation of the two Mn-C vectors relative to the ZFS coordinate frame for the Mn²⁺ ion.

Figure 3:

35 GHz 2D field-frequency pattern ¹³C Mims ENDOR for MoLOX WT with ¹³C10-LA and ¹³C11-LA. Simulations highlight peak positions, not shapes (in red and blue ENDOR contributions from two electron-spin transitions as mentioned; simulation parameters are listed in Table 2). Experimental conditions: MW frequency, ∼ 34.8 GHz; MW pulse length (π/2), 50 ns; τ, 1500 ns; RF pulse length, 20 μs; repetition rate,100 Hz; temperature, 2 K.

The analysis of the 2D ENDOR patterns of ¹³C10 and ¹³C11 of LA bound to MoLOX (Figure 3), began by comparing these patterns to the corresponding patterns previously seen for LA bound to Mn-SLO.²⁰ In SLO, both C10 and C11 of LA lie in the x-z ZFS plane (Euler angle ϕ = 0), and so at the low-field edge of the EPR spectrum, where the magnetic field pointed along the y-axis, the magnetic field is perpendicular to their Mn-C vectors, and the ENDOR spectra show a sharp doublet whose frequencies precisely give $T\left(r\right)$, which yields the Mn-C distances directly (eq 2); the full analysis of the 2D pattern then determined the remaining Euler angle, θ. Comparison of the 2D ¹³C ENDOR patterns for MoLOX in Figure 3 with 2D patterns calculated for trial Mn-C orientations chosen from a grid of (θ, ϕ) values showed that the Mn-C vectors of C10 and C11 for the MoLOX-LA complex both lie in the ZFS (x, y) plane (θ = 90°). However, unlike Mn-SLO, ϕ ≠ 0, and thus the Mn-C vectors of MoLOX do not point along the x-axis. As a result, the frequencies of the peaks in ENDOR spectra collected at the low-field edge of the MoLOX EPR spectrum, with the magnetic field along the y axis, do not simply yield $T\left(r\right)$, as they do for SLO.

To precisely determine $T\left(r\right)$ for MoLOX, we therefore used the fact that as the field is varied across the EPR envelope, the maximum frequencies of the ¹³C ENDOR doublet observed at frequencies greater than the ¹³C Larmor frequency (Figure 3) are associated with the orientation in which the external field lies parallel to the Mn-C vector, where the hyperfine coupling precisely equals $2T\left(r\right)$ and the observed ENDOR frequencies are,

$${\text{ν}}_{\text{m}}={\text{ν}}_{\text{N}}-\left(2T\right)m_s\hspace{1em}{m}_s=-5/2,-3/2$$ (3)

The measured frequency of the $m_s=-5/2$ peak (eq 1, 3) yield a most-probable value for 5T(r). Then, consideration of the small uncertainties in frequencies of 5T(r) (±0.01 MHz), arising from peak breadth and spectrum noise, leads in turn to the result, T(r) = 0.11±0.002 MHz for ¹³C10 and 0.13±0.002 MHz for ¹³C11 (Figure 4). The use of eq 2 then yields the corresponding Mn-C distances and uncertainties: r = 5.65±0.04 Å for C10 and 5.35±0.04 Å for C11. Both C10/C11 carbons of substrate LA are significantly farther from the metal center in MoLOX than in Mn-SLO, where Mn-C10, r = 4.8±0.04 Å and Mn-C11, r = 4.9±0.04 Å. Below, the Mn-C10/C11 distances are used as restraints on MD calculations that yield the Mo-LOX ground-state active-site structure.

Figure 4:

35 GHz ¹³C Mims ENDOR of WT MoLOX with ¹³C10-LA and ¹³C11-LA. Simulations of the ENDOR contributions from two electron-spin transitions are displayed as red and blue lines, as mentioned in the text. Conditions are that same as in Figure 3.

The values for T(r) and θ = 90° for C10 and C11 are well determined, as is ϕ ≈ 15° for C10. However, uncertainties in the optimum values of ϕ for C11 led us to carry out an additional joint optimization, as follows. Taking the Mn-C distances given above for ¹³C10 and ¹³C11, along with a C10-C11 bond distance of 1.55 Å, the law of cosines gives the angle between the Mn-C vectors, δϕ ≈ 15–16°. Then, as both vectors were found to lie in the ZFS xy plane (θ = 90°), the well determined value ϕ ≈ 15° for C10, then leads to ϕ ≈ 30° for C11. The resulting simulations of the 2D ENDOR patterns for the two carbons are shown in Figure 3. As the simulations show, the ¹³C ENDOR spectra of C10, C11 (Figures 3 and 4) are dominated by responses from −5/2 to −3/2 electronic transition (in red). However, the incorporation of the weak ENDOR contribution from −3/2 to −1/2 transition enables the simulations to match additional peaks in the ENDOR pattern.

Interestingly, the ¹³C ENDOR of Mn-SLO also revealed the presence of a minority, inactive ‘b’ state with Mn-C11 distance, 5.7 Å, much longer than the Mn-C11 distance of 4.8 Å in the active ‘a’ state. If the substrate-bound MoLOX exhibited such a minority conformer, with an even longer Mn-C11 distance than 5.35 Å as found here, then the resulting dipolar interaction would be so small as to give unresolved ENDOR features that are buried within the ‘distant’ ENDOR signal centered at the ¹³C Larmor frequency, which arises from natural abundance ¹³C in the enzyme, and they could not be detected.

¹H₂O ENDOR of MoLOX-LA:

To orient the C10-C11 fragment of LA relative to the Mn-OH₂ linkage, and thereby generate the information necessary to estimate the metal-bound oxygen-to-carbon distance central to catalytic H atom abstraction, we carried out ¹H Davies ENDOR measurements on the exchangeable protons of the metal-bound H₂O (Figure S3), isolating the ¹H signals from bound H₂O signals from those of the constitutive protons of coordinated histidine by subtracting spectra collected with enzyme in D₂O buffer from those of enzyme in H₂O buffer. Simulation of the resulting ¹H 2D ENDOR pattern (Figure 5) shows that the Mn-¹H vector lies along the ZFS x-axis (θ, ϕ) = (90°, 0°). As a result, at the low (H = 9.5 kG) and high (H = 16.5 kG) field edges of the EPR spectrum, ¹H ENDOR yields single crystal-like spectra; the low-field ¹H doublet splitting is T = 4.2 MHz, which corresponds to a Mn-H distance of 2.7 Å for a through-space Mn-H dipolar interaction. There is no detectable response from a second water proton. Taking the Mn-O distance of 2.23 Å observed in the X-ray structure, the Mn-H distance given by the ENDOR measurement, and the expected O-H distance of ~1 Å, this leads to an Mn-O-H angle of ~ 109°. Simple geometric considerations then determine a range of possible orientations of the Mn-O vector relative to the zero-field splitting axes (allowed positions of O relative to Mn), and in turn determine the range of possible dihedral angles between the Mn-O-H and Mn-O-C11 planes. These geometrical considerations in turn show that the O↔C11 distance can be as short as ~ 3.1 Å.

Figure 5:

35 GHz 2D field-frequency pattern ¹H Davies ENDOR for WT MoLOX with LA and simulations (in red and blue) ENDOR contributions from two electron-spin transitions as mentioned. Conditions: MW frequency, ∼ 34.8 GHz; MW pulse length (π/2), 50 ns; τ, 500 ns; RF pulse length,15 μs; repetition rate,100 Hz; temperature, 2 K. Simulation parameters are listed in Table 2.

Studies of EndoH Modified MoLOX:

The published X-ray structure of MoLOX (Figure 1A)¹⁵ was collected from a modified form of the protein in which the enzyme was pre-treated with endonuclease H (EndoH), an endoglycosidase which removes mannose-rich oligosaccharides linked to asparagines on the surface of proteins, leaving an N-GlcNAc attached. Wild-type (WT) MoLOX has seven predicted sites for N-linked glycans.²⁸ Removal of the N-glycans by EndoH reduces the molecular mass of the protein from 110–130 kDa in the native state to ~70kDa, as shown in our SDS-PAGE analysis (Figure S1) and in agreement with the literature.¹³ Such a change in molecular mass raises the question as to what degree could removing the N-glycans from the MoLOX surface influence the active site structure and enzyme function. To validate the use of the EndoH-modified X-ray structure of MoLOX in the absence of fully glycosylated one for MD simulations, we also examined the ground state structure of the substrate-bound de-glycosylated sample using ENDOR spectroscopy.

¹³C ENDOR of EndoH-Treated MoLOX:

The EPR spectrum of EndoH-treated MoLOX is identical to WT both in the presence and absence of substrate LA, which suggests that the local Mn²⁺ coordination environment is unaffected by the surface modifications (Figure S4). The field dependence of the ¹³C Mims ENDOR spectra of ¹³C10, ¹³C11 labeled LA of EndoH-treated MoLOX-LA matches nicely with that of WT (Figure S4). The ENDOR peak positions and linewidth are essentially unchanged, with the spectra completely overlaying those of WT (Figure 6 and S5). The unchanging peak frequencies show that the orientation and positioning of the substrate doesn’t change upon clipping of glycans, while the unaffected linewidths indicate that the precision with which the active site positions the substrate also does not change. In short, the ENDOR measurements indicate the ground state active site structure of the ES complex is unchanged by the clipping of glycans by EndoH, which validates the use of the EndoH-modified X-ray structure of MoLOX for MD simulations.

Figure 6:

Overlay of 35 GHz ¹³C Mims ENDOR spectra of native (black) and EndoH (red) MoLOX with ¹³C10-LA and ¹³C11-LA. Conditions are the same as in Figure 3. The complete field dependent 2D ENDOR spectra are shown in Figure S5.

Kinetic Studies of WT and EndoH MoLOX:

The ENDOR-derived invariant positioning of LA in WT and EndoH-treated MoLOX is consistent with the analysis of its steady-state kinetic parameters (Table 3). Primary deuterium isotope effects $\left({}^{\text{D}}{k}_{\text{cat}}\right)$ and their temperature dependence, $\Delta E_{\text{a}}$, (Table 3) report on the proficiency of substrate positioning and dynamics at the active site for enzymatic C-H reactions.⁴⁶ The $\Delta E_{\text{a}}$ of the EndoH-treated MoLOX variant (−0.9 ± 1.5 kcal/mol) was, within uncertainty, the same as that of the native, fully glycosylated enzyme, with both being close to ~0, supporting a hydrogen tunneling mechanism. The unchanged $\Delta E_{\text{a}}$ following EndoH processing demonstrates that the loss of surface glycans does not alter the properties of tunneling associated with catalysis. Taken together, the kinetic and ENDOR observations indicate the substrate positioning in the active site is not significantly perturbed when the size of the N-glycans are reduced from the surface of MoLOX by EndoH.

Table 3.

Kinetics of WT and EndoH-treated MoLOX

	WT	EndoH
$k_{\text{cat}}\left({\text{s}}^{-1}\right)$ a	2.41 ± 0.17	2.38 ± 0.15
$k_{\text{cat}}/K_{\text{M}}\left({\mu \text{M}}^{-1}{\text{s}}^{-1}\right)$ a	0.17 ± 0.03	0.12 ± 0.02
$E_{\text{a}}\left(\text{H}\right)\left(\text{kcal/mol}\right)$	9.5 ± 0.5	8.5 ± 1
${}^{\text{D}}{k}_{\text{cat}}$ a,b	70 ± 2	70 ± 6
$\Delta E_{\text{a}}\left(\text{kcal/mol}\right)$ c	−1 ± 0.6	−0.9 ± 1.5

^*In borate buffer, pH 9.0 @ 20°C.

^b${}^D{k}_{\text{cat}}=k_{\text{cat}}\left(\text{H}\right)/k_{\text{cat}}\left(\text{D}\right)$.

^c$\Delta E_a=E_a\left(\text{D}\right)–E_a\left(\text{H}\right)$.

MD Simulated Model of the MoLOX-LA ES Complex:

The final model of the LA structure in MoLOX was generated by MD simulations that used the X-ray crystal structure of EndoH-MoLOX to obtain the starting coordinates and by then imposing the C11−Mn and C10−Mn ENDOR-derived distances as restraints. Forty ns of MD trajectories, treated as two independent 20 ns trajectories, were carried out for the ES complex of EndoH-MoLOX and LA, using the experimental ENDOR-derived Mn-C11 and Mn-C10 distances as restraints for the MD simulations (see Methods). The final snapshot of the second 20 ns trajectory is presented in Figure 7A. In this model, LA is positioned within the U-shaped substrate channel in the MoLOX active site proximal to the catalytic Mn center (shown in yellow sticks and purple sphere). The carboxylate moiety (yellow/red sticks) of LA is located near the surface and is predicted to hydrogen bond with Arg525 and Arg528. This “carboxylate out” binding configuration is consistent with that predicted from a published docked model (Figure S6),¹⁵ and also is consistent with the X-ray co-crystal structure of the coral 8R-LOX with arachidonic acid.⁸ The ensemble average distance from the C11 hydrogen donor to the Mn-bound oxygen acceptor (i.e., DAD; Figure 7B) obtained from these MD simulations is 3.4 ± 0.1 Å (Figure 7C). This DAD is in close agreement with the sum of the van der Waals radii for carbon and oxygen, ca. 3.2 Å.

Figure 7.

MD model of LA bound to active site of MoLOX. (A) MD snapshot of the MoLOX complex with LA (final snapshot). For comparison, overlays of the in silico docked model and snapshots of the two MD trajectories are shown in Figure S6. The substrate is represented as yellow sticks. Potentially important MoLOX sidechains involved in substrate positioning are labeled and represented as green sticks. (B) Representation of the DA distance between carbon-11 (black) of substrate and Mn-bound oxygen (from water). (C) C11↔O distance along the two 20 ns MD trajectories for LA in native MoLOX (cf. Figure S7). The ensemble-average DAD is 3.4 ± 0.1 Å, with the error representing the standard deviation of the DADs determined along the 40 ns MD simulations.

Kinetic Studies of MoLOX Active Site Mutants:

To validate the MD model in Figure 7A, we prepared a series of enzyme variants in which five aliphatic residues predicted to line the substrate-binding site (i.e., within 4 Å of LA): L331, F332, L337, L522, and F526, were individually mutated to alanine and tested the kinetic impact (Table 4). Similar strategies have been employed previously to map the important residues involved in substrate binding and positioning in other lipoxygenases.^47–49 For example, alanine substitution of conserved Leu residues in SLO (Table S1) and human 12-LOX have been reported to decrease the rate constant $k_{\text{cat}}$ by 100- to 1000-fold.^{49, 50} The alanine substitutions at the homologous residues in MoLOX, L331 and F526, which sit within 4 Å of C11 of LA, cause an ~20-fold decrease in $k_{\text{cat}}$, which confirms the expectation that these residues aid in positioning C11 of LA close to the metal-bound oxygen in MoLOX.

Table 4.

Kinetic parameters of MoLOX active site mutant^a

MoLOX	$k_{\text{cat}}\left({\text{s}}^{-1}\right)$	${\text{K}}_{\text{M}}\left(\mu \text{M}\right)$	$k_{\text{cat}}/{\text{K}}_{\text{M}}\left({\text{μM s}}^{-1}\right)$	Tm (°C)
WT	2.1 ± 0.2	14 ± 2	0.15 ± 0.03	63.4 ± 0.6
L331A	0.12 ± 0.02	10 ± 2	0.012 ± 0.003	62.1 ± 0.5
F332A	0.76 ± 0.03	16 ± 2	0.048 ± 0.006	62.5 ± 0.2
L337A	1.9 ± 0.1	9 ± 3	0.21 ± 0.07	62.6 ± 0.1
L522A	0.24 ± 0.04	17 ± 8	0.014 ± 0.007	60.4 ± 0.3
F526A	0.11 ± 0.01	9 ± 3	0.012 ± 0.004	60.8 ± 0.7

^aReactions were conducted with LA at 20°C in 0.1 M borate, pH 9 buffer. The kinetic and melting temperature $\left({\text{T}}_{\text{m}}\right)$ parameters are averages of triplicate measurements collected from one biological sample on three independent days. The error bars represent standard error of the mean (s.e.m).

In the MoLOX-LA model, F332 is within 4 Å of C15 and C16 of LA and adjacent to L331 (see above). Substitution of Phe-for-Ala results in a modest 3-fold decrease in $k_{\text{cat}}$ (Table 4). It is important to note that F332 is located in a loop that undergoes significant movement during the 40 ns MD trajectory. The Cα of F332 in the final snapshot of the MD simulation (Figure 7) is displaced by 3 Å relative to the Cα position in the X-ray structure (Figure S6). To probe if these dynamics have any relevance to catalysis, we also measured the activation energy ($E_{\text{a}}$) for F332A. In the context of hydrogen tunneling, changes in the $E_{\text{a}}$ value are consistent with altered isotope-independent conformational (thermal) motions of the ES complex.^{11, 51} The E_a of F332A is reduced (7.2 ± 0.7 kcal/mol) relative to WT MoLOX, $E_{\text{a}}=9.5\text{ kcal/mol}$ (Table S1 and Figure S8), supporting a functional role for the dynamic nature of this loop. The kinetic behavior of F332A MoLOX is reminiscent of the SLO I553X mutants (X = L or G) that showed little-to-modest changes in $k_{\text{cat}}$, but was accompanied by altered $E_{\text{a}}$ values that are associated with disruption of active site sidechain packing and a network of protein motions.^52–54

Finally, we explored alanine mutations at L337 and L522 in MoLOX (Table 4). Kinetic analysis of L337A and L552A show distinct catalytic impacts in MoLOX compared to their homologous mutants, I553A and V750A,⁵⁴ in SLO (Table S1). Mutation to alanine of L522, which is positioned behind and contacts both L331 and F526, produces a 9-fold decrease in $k_{\text{cat}}$. This demonstrates the importance of proper hydrophobic packing in the MoLOX active site. In contrast, the homologous V750A mutation in SLO slightly reduces $k_{\text{cat}}$ (1.4-fold) relative to WT. Finally, the L337A MoLOX variant has a nearly identical $k_{\text{cat}}$ and unchanged $E_{\text{a}}$ relative to WT (Figure S8) whereas the homologous I552A-SLO mutant results in modest 4-fold decrease in $k_{\text{cat}}$ and appreciable change in $E_{\text{a}}$ (Table S1). Note that the second-order rate constants, $k_{\text{cat}}/{\text{K}}_{\text{m}}$, for this suite of MoLOX mutants follow similar trends to $k_{\text{cat}}$ (Table 4). Overall, the cumulative kinetic properties validate the LA-EndoH-MoLOX binding model in Figure 7A while the differences from the kinetic properties of SLO illustrate the influence of subtle differences in substrate positioning by aliphatic, active-site residues.

Discussion

ENDOR analysis of Mn-SLO vs MoLOX:

We previously used EPR/ENDOR spectroscopy to identify the enzyme-substrate (ES) ground state structure in metal substituted (Mn²⁺-for-Fe²⁺) soybean lipoxygenase, SLO.^{20, 21} Here we report the corresponding EPR/ENDOR study of the ES complex of a fungal Mn-lipoxygenase, MoLOX, which thus did not require metal substitution. Measurement of the ¹³C hyperfine coupling of C10 and C11 nuclei of LA, yielded the position and orientation of LA at the active site. Incorporation of this information as restraints on MD simulations that used the EndoH-MoLOX crystal structure as starting coordinates has yielded the first high-resolution image of the active site of the MoLOX-LA co-structure, the first for a fungal lipoxygenase in its native, fully N-glycosylated state, Figure 7A.

To begin, we compare the structural information derived from EPR/ENDOR results for ES complex of MoLOX-LA in the light of canonical Mn-SLO-LA. The EPR spectrum of Mn²⁺ at the enzymatic site provides information about the symmetry of the coordinating ligands. Mn(II)-MoLOX shows a broader EPR spectrum than for Mn(II)-SLO as the result of a larger ZFS splitting, which suggests a greater distortion of the Mn-coordination sphere from octahedral symmetry for MoLOX than Mn-SLO. This reflects the fact that the structures of the Mn-cofactor in the two enzymes, as revealed by X-ray diffraction, are not identical and do not overlap (Figure 1B, C), with small but distinct differences in orientation and position of metal ligands.

The most obvious difference between the ¹³C ENDOR-derived LA positioning in Mn-SLO and MoLOX is the significantly greater distance between Mn and the C11 target in MoLOX, 5.35±0.04 Å, than in Mn-SLO, 4.9±0.04 Å (Figure 8). However, despite this difference plus the differences in the orientation of LA (carboxylate-out for MoLOX, carboxylate-in for SLO) and in the orientation of the Mn-bound H₂O in MoLOX and SLO, simple geometrical analysis shows the distance between the aqua oxygen and target C11 in MoLOX can be as short as ~ 3.1 Å, the same as the estimated average ground-state distance of 3.1 Å in the active conformer of Mn-SLO. ENDOR analysis of Mn-SLO led to the identification of a minority LA-inactive conformer distinguished by Mn-LA distance Mn-C11 5.7±0.04 Å, whereas if such a minority conformer exists in MoLOX, the substrate LA would be too far away from the Mn to be detected.

Figure 8:

ENDOR-derived structure of active site of native MoLOX-LA (left) and SLO-LA active conformer (right) in the zero-field splitting (ZFS) coordinate-frame ($D_{\text{x}}$, $D_{\text{y}}$, $D_{\text{z}}$). C11-Mn (red dashed line) and C10-Mn (blue dashed line) distances are obtained experimentally from ENDOR measurements and the C11-O DAD (purple dashed line) is obtained from MD simulations using these restraints. The Mn-O distance (2.23 Å) is taken from the X-ray crystal structure. Dashed light-blue circles represent the possible orientations of the Mn-O vector (allowed positions of O relative to Mn), determined as described above. The SLO structure (right) was adapted with permission from ref 20. Copyright 2017 American Chemical Society.

Overall, although ENDOR-derived Mn-C11 distances differ substantially in the two enzymes, differences in the organization of aliphatic sidechains along the substrate binding channel, nonetheless allow the MoLOX ES complex to access comparable values for the catalytically important O↔C11 donor-acceptor distance (DAD), but with a small difference that is nonetheless functionally important (see below).

Importantly, the ENDOR derived Mn-C10, Mn-C11 distances, which provide two important input parameters for the MD simulation, remain unchanged after clipping glycans (EndoH-MoLOX), which indicates that the substrate LA positioning and orientation is not perturbed in the active site upon de-glycosylation. This observation, taken together with the kinetic impact of the volume-reducing alanine mutational screening of a series of aliphatic residues that line the putative substrate channel in MoLOX, (Figure 7A), validate the use of the docked model of LA into the available X-ray structure of the de-glycosylated enzyme, and the MD structure as a reliable ground-state 3-D model of the active site ES complex.

Comparison of the native MoLOX ground-state ES structure to that of Mn-SLO:

Analysis of the two 20 ns MD simulations of the MoLOX ES complex reveals the DAD between C11 of LA and the hydrogen acceptor on the Mn-bound oxygen to be 3.4 ± 0.1 Å. The 0.1 Å error represents the standard deviation of the DAD distances determined along the 40 ns MD simulations; it should be noted that this value is not the visual width of the distributions, as seen in Figure 9. This DAD value is the same as that observed in the X-ray crystal co-structure of coral 8R-LOX with arachidonic acid as substrate (3.4 Å for ‘C↔O’).⁸ Based on the comparison of the distribution of DADs in the ensemble of ground-state distances for LA in MoLOX and SLO, as determined by a combination of ENDOR distance restraints and multiple MD trajectories, the DAD in MoLOX is Δ = 0.3±0.14 Å longer than that in Mn-SLO, 3.1 ± 0.1 Å (Figure 9 and S7).

Figure 9.

Ground-state DA distance distributions from ENDOR-guided MD trajectories for MoLOX (blue) and SLO (red).

In the ENDOR-guided MD model of MoLOX (Figure 7A), the carboxylate group of LA is positioned near the surface at the substrate entrance in a “carboxylate out” orientation and with predicted hydrogen bonding interactions to two arginine residues, R525 and R528. Mutation of the Arg525 to alanine largely abolished catalytic activity for the reaction with the natural substrates (i.e., LA and α-linolenic acid).¹⁵ While R528 mutational studies have not been reported, it is conserved across 48 predicted fungal LOXs. Thus, R525 (and perhaps R528) of MoLOX appears to help tether the carboxylate group of C18 fatty acid substrates, supporting a “carboxylate out” substrate binding orientation.¹⁵ The ‘carboxylate out’ mode for MoLOX is also consistent with the enzyme’s ability to oxidize phosphatidylcholine substrates with comparable catalytic efficiency as LA.¹³ In contrast, LA is expected to occupy the ‘carboxylate in’ mode in SLO with the carboxylate group within H-bonding to R707 (Figure S9).²⁰ This orientation in SLO is preferred based on previous MD simulations,²⁰ though the substrate orientation in SLO has been of some debate.^55–57 Similarly, a QM/MM simulation of the SLO ES complex modeled the substrate in this ‘carboxylate in’ mode.⁴⁴ From this QM/MM model, the ground-state C11-O distance was predicted to be ~3.3 Å, in good agreement with the results from ENDOR-guided MD.²⁰ Because of the lack of a structure for the native, N-linked glycosylated form, ENDOR permits an experimental assessment that reducing the size of the glycan structures negligibly impacts substrate positioning in the fungal LOXs, including MoLOX.

The remarkable difference in orientation of LA bound in MoLOX versus SLO is attributed to different orientation and structural alterations of the substrate channel shape,^{15, 58} which is influenced by sidechain from primarily hydrophobic residues whose mutant kinetic properties are highlighted in Tables 4 and S1. Awareness of this difference in turn would strongly impact efforts to design inhibitors of MoLOX.

Relevance of ground-state DADs to the efficiency of hydrogen tunneling in MoLOX:

Lipoxygenases, and in particular SLO, have led to the recognition of some of the most significant kinetic features in the quantum treatment of hydrogen transfer in enzyme catalysis. A key characteristic is the large, nearly temperature independent kinetic isotope effects for the C-H bond cleavage during the oxidation of fatty acids. The explanation of such behavior for hydrogen-tunneling reactions has advanced our understanding of enzyme catalysis, placing a focus on the importance of dynamic active-site compaction for effective wavefunction overlap. Based on the multidimensional analytical rate model for enzymatic H-tunneling, effective hydrogen wave function overlap in native enzymes is attained at the so-called tunneling-ready state (TRS), which exhibits compacted DADs of 2.7–2.8 Å.^{11, 44, 59–61} The ground-state C11-O, DAD distances .n both MoLOX (3.4 Å) and SLO (3.1 Å) are longer than the TRS distances that are required for productive wavefunction overlap and tunneling. This further supports the understanding that an enzyme-substrate (ES) complex transiently achieves the effective configuration(s) for tunneling through compaction across multiple tiers of C11↔O distances.^{11, 20, 46}

Due to the compact wavefunction for a hydrogen atom, small changes in this corresponding distance are expected to have profound effects on catalytic efficiency. For example, a double mutant (DM) of SLO, constructed from the volume-reducing mutation to alanine of two conserved leucine residues adjacent to the reactive carbon of LA, exhibited a 10⁵-fold decrease in $k_{\text{cat}}$.⁶² This arises, in part, from an elongated DAD.⁶³ For DM SLO, the ground state DAD is 3.8 Å based on ENDOR-guided MD,²⁰ with the comparable prediction of ~3.6 Å from QM/MM simulations.⁴⁴ Both values are significantly elongated relative to the 3.1 Å distance for WT SLO, thus requiring an even more energy-demanding compaction in achieving reactivity. In the current study, the ENDOR-guided MD data present a small, but statistically significant difference $\left(p<0.001\right)$ in the magnitude of the ground-state DADs of SLO and MoLOX, 3.1 Å vs 3.4Å (Figure 9). With the further elongated DAD in the MoLOX ES complex, the probability of achieving a productive TRS would be diminished compared to that for SLO. Such an expected decrease in catalytic rate agrees well with the trends in the experimental $k_{\text{cat}}$ values, with MoLOX exhibiting a 100-fold decreased catalytic rate constant relative to the SLO reaction.

The observation of a longer ground-state DAD in MoLOX compared to SLO, could also explain why $E_{\text{a}}$ for MoLOX is larger than that of SLO: the $E_{\text{a}}$ for the reaction of MoLOX is 10 kcal/mol, compared to a conspicuously low value of 2 kcal/mol for the WT SLO reaction.^{28, 50} The H-atom tunneling process is intrinsically temperature independent, and the temperature dependence of the observed catalytic rate constant reflects the contributions of isotope-independent, protein conformational motions, including those remote of the site of C-H cleavage, that create short, tunneling-ready distances and configurations.^{11, 54} We thus propose that the decreased catalytic efficiency of MoLOX, relative to SLO, originates from the differences in substrate positioning in the active site. As a consequence, MoLOX may require more extensive thermal sampling of the protein-substrate complex to reach compacted configurations with effective tunneling distances. Note that the magnitude of the $E_{\text{a}}$ will also reflect the stiffness of the relevant vibrational degrees of freedom along the reaction coordinate. Thus, elevated $E_{\text{a}}$ values can also arise from less stiff vibrational modes linked to the C-H bond cleavage event of catalysis. The argument is supported by the observed difference in the root-mean square fluctuations (RMSF) of the substrate carbon positions derived from the MD simulations of SLO and MoLOX (Figure S10). SLO substrate carbons were found to exhibit RMSF of ~ 1.5 Å for C1-C15, whereas in MoLOX the substrate LA undergoes significantly larger fluctuations in the MD trajectories, RMSF > 2.1 Å (Figure S10).

Conclusions

The pathogenic fungal lipoxygenase, MoLOX, contains a catalytically active, mononuclear Mn²⁺ metallocentre with coordination-sphere geometry distinct from that of the canonical iron LOXs, and that provides a natural electron-spin probe that is used here to examine the geometry of substrate binding by EPR/ENDOR spectroscopy. Combination of the spectroscopic results with MD simulations yields for the first time a high-resolution model of the ground-state active-site structure of MoLOX complexed with its natural substrate, LA. The distance between C11, which bears the hydrogen donor of LA, and the metal-bound oxygen (DAD) is ~3.4 Å, in excellent agreement with an X-ray derived co-structure of the animal 8R-LOX with AA. The ‘carboxylate-out’ orientation of substrate binding, found here for MoLOX and visualized crystallographically for 8R-LOX, contrasts with the ‘carboxylate-in’ binding in the model plant lipoxygenase, SLO. The structural information gleaned from this study, not least determination of the ‘carboxylate-out’ LA binding to MoLOX, is expected to guide future in silico screening efforts and rational inhibitor design of this enzyme, an emergent target for efforts to tame the devastating rice blast disease.

The structural and kinetic results presented herein implicate a relationship between the achievement in LOX enzymes of a thermally-equilibrated active site that is geometry-optimized for H-atom transfer, and the degree to which subsequent protein thermal fluctuations are needed to dynamically reach TRS configurations favorable for catalysis. Thus, the MoLOX DAD ~3.4 Å is slightly, but meaningfully, longer than the corresponding value for SLO (3.1 Å), while the activation energy for MoLOX H-atom transfer is significantly larger than the anomalously low activation energy seen for SLO. Indeed, representatives of mammalian, fungal, and prokaryotic LOXs have been reported with a range of reduced first-order reaction rate constants compared to that of SLO and enlarged $E_{\text{a}}$ values (8–13 kcal/mol) compared to SLO.^{64, 65} Future structural, kinetic, and theoretical studies on orthologues across the lipoxygenase family will further clarify the relationship of the activation energy barriers for hydrogen transfer reactions to the geometry of substrate binding in the ground state. The current work demonstrates the versatility of the ENDOR-guided MD approach to characterizing the ground-state configurations of the substrate in the active site of the diverse and biologically important family of LOXs, which often defy X-ray crystallographic efforts.

Table 2.

ENDOR hyperfine parameters for ¹³C-LA

nuclei	T(MHz)a	r(Å)b	(θ, ϕ)c
MoLOX
C10	0.11	5.65	90, 15
C11	0.13	5.35	90, 30
H2O	4.2	2.67	90, 0
Mn-SLO
C10	0.17	4.8	50, 0
C11	0.17	4.8	40, 0
H2O	4.0	2.71	70, 0

Estimated uncertainties:

^aT = ±2 kHz

^br = ±0.04 Å

^cϕ = ±5; ^cθ = ±5°