The Structure of Formaldehyde-Inhibited Xanthine Oxidase Determined by 35 GHz ²H ENDOR Spectroscopy

Abstract

The formaldehyde-inhibited Mo(V) state of xanthine oxidase (I) has been studied for four decades, yet it has not proven possible to distinguish unequivocally among the several structures proposed for this form. The uniquely large isotropic hyperfine coupling for ¹³C from CH₂O led to the intriguing suggestion of a direct Mo-C bond for the active site of I. This suggestion was supported by the recent crystal structures of glycol- and glycerol-inhibited forms of aldehyde oxidoreductase (AOR), a member of the xanthine oxidase family. ^1,2H-ENDOR spectra of I(C^1,2H₂O) in H₂O/D₂O buffer now unambiguously reveal that the active-site structure for I contains a CH₂O adduct of Mo(V) in the form of a four-membered ring with S and O linking the C to Mo, and rule out a direct Mo-C bond. DFT computations are consistent with this conclusion. We interpret the large ¹³C coupling as resulting from a 'transannular hyperfine interaction'.

Xanthine oxidase is a molybdo-enzyme that catalyses the oxidative hydroxylation of a variety of heterocyclic and aldehyde substrates, including the physiological substrates hypoxanthine and xanthine.¹ Of the large number of EPR-active Mo(V) intermediates exhibited by this enzyme,² one of the most intriguing is the CH₂O inhibited Mo(V) form (I) first described by Bray and coworkers.³ Its most remarkable feature, revealed by Howes et al., is the presence of a carbon derived from CH₂O that EPR and ENDOR spectroscopies show to have a uniquely large ¹³C isotropic hyperfine coupling, a_iso (¹³C) ~ 43.0 MHz.^4,5 This value contrasts with the five-fold smaller coupling for the Mo-O- ¹³C of the “2-hydroxy-6-methyl-purine (HMP) very rapid” form (a_iso = 7.9 MHz).⁶ Although I has been studied for 40 years, it has not proven possible to distinguish unequivocally among the viable candidates for its structure (Scheme 1).^5,7–13

Scheme 1.

The large isotropic ¹³C hyperfine coupling for I led Howes et al. to the intriguing suggestion that it contains a CHO fragment with a direct Mo-C bond, Scheme 1, C.^4,5 This suggestion recently received support from the crystal structures of glycol and glycerol inhibited forms of aldehyde oxidoreductase (AOR), a member of the xanthine oxidase family.⁷ These structures exhibit Mo-C bond distances of 2.36 Å and 2.72 Å, respectively, the former in particular being suggestive of direct Mo-C bonding interactions. In contrast, analysis of the smaller ¹³C coupling for a substrate-derived species bound to Mo(V) of the “very rapid” state of xanthine oxidase indicated there was no direct Mo-C bond,⁶ and this was confirmed by recent crystal structures for that intermediate.^14,15

We here report that ^1,2H-ENDOR spectra of I(C^1,2H₂O) prepared¹⁶ in H₂O/D₂O buffer rule out all models proposed for the active-site structure of I (Scheme 1) except for model A, a four-membered cyclic adduct of CH₂O with S and O linking the C to Mo. DFT calculations are consistent with the ENDOR finding that A represents the structure for the active site of I, and not the direct Mo-C bond of C.

The X-band EPR spectrum of I(^12,13C¹H₂O) in H₂O shows a doublet splitting from the ¹³C-nucleus, with each line further split into a doublet from a single proton that derives from CH₂O.⁴ The 35 GHz echo-detected EPR spectrum of I(^12,13C¹H₂O) in H₂O show the ¹³C doublet from ¹³C¹H₂O, but the ¹H splitting is not observed, Fig S1.^17,18 The EPR spectrum of I(^12,13C¹H₂O) further shows hyperfine splitting from ^95,97Mo (natural abundance: 15.9% ⁹⁵Mo; 9.6% ⁹⁷Mo). The g₁ and g₃ splittings are observable in both X and Q band spectra; those associated with g₂ are only resolved at Q-band (Fig S1). Simulations of the EPR spectra gave principal g-values and A for the ¹³C of CH₂O that agree with those of Howes et al.: g = [1.988, 1.974, 1.948], A(¹³C) = [51.5, 40, 40] MHz, a_iso = 43.8 MHz.

Fig 1 presents 35 GHz Davies ¹H and ^95,97Mo (upper), and Mims ²H (lower) ENDOR spectra collected for I(¹²C^1,2H₂O) in H₂O/D₂O near g₂, the magnetic field corresponding to the maximum EPR intensity. The Davies spectra of I(C¹H₂O) in H₂O and D₂O are essentially the same (A), both exhibiting a ¹H doublet centered at the ¹H-nuclear Larmor frequency and split by A ≈ 13 MHz. This doublet is absent in the spectrum of I(C²H₂O) in H₂O (B) and thus is associated with a proton, ¹H_A, derived from CH₂O. The ¹H doublet in Fig 1 rides on ^95,97Mo-ENDOR signals and is more clearly seen upon subtraction of the spectrum of I(C²H²O, B) from that of I(C¹H₂O, A), spectrum A–B. The ¹H_A doublet is replaced by a corresponding ²H_A signal when C²H₂O is the substrate. Davies ²H-ENDOR spectra of I(C²H₂O) in H₂O exhibit a ²H_A doublet without resolved quadrupole splitting (not shown). The hyperfine interaction, A(²H_A) corresponds to a smaller value of A(¹H_A) than that seen directly in the ¹H-ENDOR spectrum of I(C¹H₂O) in H₂O/D₂O because of a substantial isotope effect on the hyperfine couplings.^19,20

Figure 1.

35 GHz (Upper) Davies ¹H and ^95,97Mo (A, B), ENDOR spectra of I(C^1,2H₂O) in H₂ and D₂O buffer. (Lower) Mims ²H spectra of I(C²H₂O) in H₂O (green/blue), I(C¹H₂O) in H₂O and D₂O buffer. Horizontal bars indicate hyperfine splittings for ^1,2H_A,B non-exchangeable protons. Conditions: g = 1.97; T = 2 K. Davies: π-pulse = 80 ns, τ = 600 ns, repetition time = 50 ms, 34.84 GHz; Mims, ²H: π/2 pulse = 50 ns, τ = 800 ns, repetition time = 50 ms, 34.87 GHz.

To test for both exchangeable and non-exchangeable ²H signal(s) with smaller hyperfine couplings than ^1,2H_A, Mims ²H-ENDOR spectra were collected for I(C¹H₂O) in D₂O and I(C²H₂O) in H₂O at g₂ (Fig 1, lower). The nearly featureless ²H spectrum of I(C¹H₂O) in D₂O shows clearly that there are no exchangeable protons in the active site of I. However, the ²H spectrum of I(C²H₂O) in H₂O shows two doublets from deuterons derived from the C²H₂O (Fig 1). The larger coupling corresponds to ²H_A observed in the Davies ¹H-ENDOR spectra of I(C¹H₂O) in H₂O/D₂O, again with the isotope effect on the hyperfine coupling. The second doublet in the Mims spectrum for I(C²H₂O) also does not exchange in H₂O buffer, and thus corresponds to the second C–H deuteron introduced with the C²H₂O, denoted ²H_B.

To confirm that the Mims ²H spectrum is not significantly distorted by suppression `holes' associated with this technique²¹ and to determine the ²H hyperfine and quadrupolar tensors, we collected a complete 2-D field-frequency ²H-ENDOR pattern of spectra acquired at numerous magnetic fields across the EPR envelope and simulated this pattern with inclusion of the suppression effects (Fig S2).²² At magnetic fields near g₁, the spectra show the ²H_A and ²H_B doublets centered at ²H-Larmor frequency and split by A ~ 1.8 MHz and 0.4 MHz respectively. The ²H-ENDOR pattern does not change significantly as the field is moved towards g₂ and g₃, except in intensity, indicating that the hyperfine couplings of ²H_A and ²H_B are largely isotropic. The 2-D pattern is well simulated by the 1:1 summation of simulations of²H_A and ²H_B, with, A(²H_A) = [1.8, 1.8, 1.9] MHz; A(²H_B) = [0.44, 0.4, 0.39] MHz. Neither ²H exhibits quadrupolar splitting (I = 1) because of the large ENDOR line widths. As noted above for the g₂ spectrum, the tensor obtained from simulations of the ²H_A-ENDOR pattern only approximately matches that obtained by fitting a 2-D ¹H Davies ENDOR pattern for ¹H_A because of an isotope effect on the hyperfine couplings.^19,20

The observation that both aldehydic protons of CH₂O are non-exchangeably hyperfine-coupled to the Mo(V) of I rules out all candidates in Scheme 1 for the active-site structure of I except the cyclic CH₂O adduct, A.

DFT calculations are consistent with experiment in showing a large a_iso for the ¹³C. of Model A. This model yielded hyperfine tensors for ¹³C, ²H_A and ²H_B of CH₂O that all are in satisfactory agreement with the experimentally observed values: a_iso(¹³C) ~ 47.9 MHz, A(¹³C) = [53.6, 45.4, 44.6] MHz; A(²H_A) = [4.0, 3.3, 3.0] MHz, A(²H_B) = [0.38, −0.67, −0.72] MHz. Most importantly, the carbon-bound CHO complex, model C, already ruled out by the absence of H_B, is calculated to have almost three-fold and five-fold smaller ¹³C and ¹H_A hyperfine couplings, respectively: a^iso (¹³C) = ~ 16.1 MHz, A(¹³C) = [23.2, 13.4, 11.7] MHz; A(²H_A) = [0.51, −0.43, −0.66] MHz.

Why is the ¹³C coupling so large for the carbon of CH₂O that is not directly coordinated to Mo(V) in the four-membered ring of model A, and not for the carbon directly bonded to the metal ion in C? We interpret this as resulting from a `transannular hyperfine interaction'.^23,24 The carbon is in line with a lobe of the half-occupied Mo(d_xy) orbital, and this allows overlap between Mo(d_xy) and orbitals of carbon; the large a_iso for ¹³C corresponds to only ~ 1.2% spin density in a carbon 2s orbital. Analogous behavior was first observed for a phosphorous atom that formed part of four membered cyclic structures of Mo(V) and V(IV) with dialkyl/aryldithio-phosphinato ligands.^23,24 A weaker hyperfine coupling was observed when the 31P is bound to the metal-ion directly through a monodentate M-O-P linkage.²⁵ Such considerations may have relevance to a recent proposal of a direct Fe-C bond based on a large ¹³C coupling.²⁶

Finally, regarding the use of distances in the X-ray structure to infer a direct Mo-C bond, we note that the DFT-optimized cyclic structure A has a Mo-C distance of 2.76 Å, the same as found by x-ray diffraction for the glycerol-inhibited form of AOR.⁷