Evidence for H-bonding interactions to the μ-η²:η²-peroxide of oxy-tyrosinase that activate its coupled binuclear copper active site

Abstract

The factors that control the diverse reactivity of the μ-η²:η²-peroxo dicopper(II) oxy-intermediates in the coupled binuclear copper proteins remain elusive. Here, spectroscopic and computational methods reveal H-bonding interactions between active-site waters and the μ-η²:η²-peroxide of oxy-tyrosinase, and define their effects on the Cu(II)₂O₂ electronic structure and O₂ activation.

Graphical Abstract

The broad family of coupled binuclear copper (CBC) metalloproteins includes hemocyanins (Hc), catechol oxidases (CaOx), tyrosinases (Ty), and the o-aminophenol monooxygenase (NspF), all of which contain two copper centers, each coordinated by three histidine residues, and held in close proximity to each other (3.5–4.0 Å).¹ In their reduced bicuprous deoxy state, these active sites bind dioxygen reversibly and with great affinity (K_d < 16–40 μM) to form a μ-η²:η²-peroxo dicopper(II) oxy-intermediate.^1,2 Crystallographic and X-ray spectroscopic investigations into the structure of these oxy-intermediates in different CBC members reveal a broadly conserved μ-η²:η²-peroxo dicopper(II) active site.^3–6 Similarly, all oxy-CBC proteins exhibit equivalent spectroscopic features, including (i) two distinct electronic absorption bands (ε_345nm ~14,000–20,000 M⁻¹cm⁻¹, ε_600nm ~1,000 M⁻¹cm⁻¹) corresponding to two peroxide→Cu(II) charge-transfer (CT) transitions, (ii) a set of characteristic resonance Raman (rR) features corresponding to vibrational modes of the [Cu(II)₂O₂]²⁺ cofactor (Fig. 1), and (iii) the absence of an electron paramagnetic resonance (EPR) signal due to the strong antiferromagnetic coupling between the two Cu(II) centers facilitated by the μ-η²:η²-peroxo superexchange pathway.^1,7,8 Despite these apparent similarities in the geometric and electronic structure of their [Cu(II)₂O₂]²⁺ active sites, oxy-CBC proteins exhibit distinct reactivities that facilitate diverse physiological functions. The extracellular O₂-carrier protein Hc contains an extended terminal domain that prevents phenolic substrates from accessing its Cu(II)₂O₂ active site, thus lacking any enzymatic reactivity.⁹ Unlike Hc, CaOx catalyze the oxidation of catechols to ortho-quinones, and are found predominantly as part of natural product biosynthesis pathways and defense mechanisms in plants. The ubiquitous Ty is bifunctional as it possesses both catechol oxidation and monophenol monooxygenation reactivities, and catalyzes the conversion of L-tyrosine to L-DOPAquinone, which constitutes the rate-limiting step in melanin biosynthesis in a variety of organisms from soil bacteria to humans. Finally, in addition to catechol oxidation and monophenol monooxygenation, NspF also catalyzes the monooxygenation of o-aminophenols to o-nitrosophenols in bacterial siderophore biosynthesis.

Fig. 1.

Resonance Raman (351nm laser excitation) spectral regions for the key [Cu(II)₂O₂]²⁺ vibrational modes of oxy-Ty from S.Glaucescens (0.75mM, 0.1M CHES, pH/D 9.0) in H₂O (red) and D₂O (blue) corrected for background contributions, with gaussian fits (in regions with multiple peaks, fits without the baseline components are also shown), and each region scaled for clarity (reference normalized peak intensity ratios are ~16:1:2:6 for Cu-Cu : Cu-O fundamental : O-O : Cu-O overtone), (A) the Cu-Cu (indicated with solid line) and Cu-N stretches (dashed lines), the negative feature (asterisk) at the solvent ice peak region results from background correction; the insert expands the region of the two higher-energy isotope-sensitive Cu-N stretches, (B) the Cu-O (B_3u) fundamental stretch (the black gaussian indicates a non-isotope sensitive peak present in both solvents), (C) the O-O stretch, and (D) the Cu-O (B_3u) overtone stretch. For entire rR spectra before and after background corrections and for band-shape discussion see Fig. S5; for additional experimental details see SI methods 1.1.5–1.1.6.

Given the major implications of CBC proteins in both human diseases¹⁰ and critical applications in agriculture (fruit browning¹¹) and biotechnology (protein bioconjugation¹²), obtaining a mechanistic understanding of their diverse reactivities has been a major goal with broad scientific interest. However, putative differences in the electronic/geometric structures of their Cu(II)₂O₂ active sites as well as the molecular-level factors that control their reactivity are not fully understood despite extensive research efforts.^8,13–15 Second-sphere interactions involving active-site residues and/or solvent molecules with the Cu(II)₂O₂ core could have important mechanistic implications, but have been challenging to define experimentally, in part, due to the fact that most of the structural insights for catalytic oxy-CBC enzymes have derived from crystallographic studies on their latent pro-forms (PDB:6JUA-D)¹⁶ or binary complexes with their caddie proteins (PDB:1WX2, 1WX4–5)⁴, where in all cases the Ty active sites are occupied by additional residues, paralleling oxy-Hc (Fig. S4). Here, we have been able to directly probe and define previously unknown interactions between active-site water molecules and the μ-η²:η²-peroxide of the [Cu(II)₂O₂]²⁺ active-site in the monomeric and fully-catalytic oxy-Ty from Streptomyces glaucescens by employing rR spectroscopy in tandem with solvent isotopic (H₂O/D₂O) perturbations, and correlating these experimental results to molecular dynamics (MD) simulations coupled with quantum-mechanics/molecular-mechanics (QM/MM) calculations. Comparison of the active sites of different CBC enzymes reveals variations in the key second-sphere residues that facilitate the water-peroxide H-bonding interactions in oxy-Ty, and correlate with the monooxygenation reactivity within this protein family.

By generating the oxy-Ty intermediate at high concentration and high purity in both H₂O and D₂O solvents (SI methods 1.1.5) we were able to obtain high-quality spectroscopic data. Upon laser excitation of the intense peroxide→Cu(II) CT transition at 351nm, the rR spectrum of oxy-Ty in H₂O exhibits the characteristic resonance-enhanced vibrational modes of the μ-η²:η²-peroxo dicopper(II) core, previously defined for oxy-CBC proteins.^17,18 These include (i) the intense Cu-Cu (A_g) stretch at ∼280 cm⁻¹, (ii) the O-O (A_g) stretch at ∼740 cm⁻¹, (iii) the weak Cu-O (B_3u) stretch, which is non-totally symmetric, formally forbidden in rR, at ∼540 cm⁻¹ along with its intense symmetry-allowed first-overtone at ∼1075 cm⁻¹, and (iv) four of the six Cu-N(His) modes in the 200–350 cm⁻¹ region (Fig. 1). Upon solvent deuteration, the Cu-O (B_3u) overtone frequency increases by 6±2 cm⁻¹, with its corresponding fundamental mode increasing by 4±2 cm⁻¹ (Fig. 1D and 1B, respectively), the two higher-energy Cu-N(His) stretches decrease by 2±1 cm⁻¹ and 4±2 cm⁻¹, respectively (Fig. 1A insert), while the remaining modes, including the Cu-Cu and O-O stretches, are not perturbed (Fig. 1A and 1C). Paralleling the two higher-energy Cu-N(His) stretches in oxy-Ty, the H₂O→D₂O perturbations of the rR spectrum of oxy-Hc also exhibit a 1–2 cm⁻¹ decrease in these two modes, which was attributed to the mass increase of the deuterated imidazole rings of the histidine ligands.¹⁹ However, unlike oxy-Ty, no solvent isotopic (H₂O/D₂O) perturbations were observed for the Cu-O (B_3u) fundamental or overtone modes in the rR spectrum of oxy-Hc.¹⁹ Since solvent accessibility is among the most prominent active-site differences between oxy-Hc and the monomeric active oxy-Ty,^1,20 the presence of direct interactions between the [Cu(II)₂O₂]²⁺ core and water molecules is likely responsible for the observed H₂O→D₂O upshift of the Cu-O (B_3u) frequency. Such rR H₂O→D₂O upshifts of metal-oxo stretches have been observed in binuclear iron metalloenzymes, including oxy-hemerythrin and met-ribonucleotide reductase.^21,22 In both cases, the frequency increase of the symmetric Fe-O-Fe stretch upon solvent deuteration was attributed to H-bonding interactions between the oxo-bridge and the respective metal-coordinated hydoperoxo or water-derived hydroxyl ligands. However, these correlations between the presence of H-bonding interactions and the H₂O→D₂O frequency increase of metal-oxo stretches remain empirical, lacking an analysis of their molecular origin.

To fully define these H-bonding interactions between the [Cu(II)₂O₂]²⁺ core and water molecule(s), as well as to elucidate the origin of the Cu-O (B_3u) H₂O→D₂O upshift, our spectroscopic results were correlated with a combination of computational methods. First, from the crystal structure of oxy-Ty from Streptomyces castaneoglobisporous (PDB:1WX2)ⁱ we generated a fully solvated protein model of the monomeric oxy-Ty (Fig. 2A, SI methods 1.2.1–1.2.3) and performed 0.5 μs MD simulations (SI video 1, and SI methods 1.2.4). Our MD simulations clearly demonstrate three distinct positions in the active site of oxy-Ty that are occupied by water molecules (Fig. S7B). One of these waters (W3) has been previously observed crystallographically and was proposed to act as the acceptor of the monophenolic proton during the monooxygenation reaction.²³ The other two water molecules (W1, W2), which have not been previously observed, reside above the Cu₂O₂ plane and each of them forms an H-bond with one of the O atoms of the μ-η²:η²-peroxide.ⁱⁱ H-bonding to the μ-η²:η²-peroxide represents a new type of interaction, not previously defined in other CBC enzymes or synthetic [Cu(II)₂O₂]²⁺ model complexes.

Fig. 2.

(A) The full-length QM/MM-optimized structure for the experimentally-calibrated oxy-Ty active site from Streptomyces castaneoglobisporous (oxy_48, blue), (B) its active site, and (C) its alignment with the met-CaOx from Ipomoea batatas (PDB:1BT1, red). Key second-sphere residues and the μ-η²:η²-peroxide ligand and waters (W1–3) in oxy-Ty are shown as solid sticks, the water (W3) and μ-hydroxide ligand in CaOx as red spheres, the first-sphere histidines as transparent sticks, and the copper centers as brown spheres. H-bonds are indicated as dashed yellow lines.

We next generated 64 distinct QM/MM-optimized oxy-Ty structures with different numbers (2–8) and conformations of active-site waters (intital structures were obtained from MD simulations or generated manually; SI methods 1.2.5–1.2.10, Table S2). The optimal number of active-site waters in oxy-Ty was determined by systematically evaluating the thermodynamic stability of oxy-Ty at different hydration levels (SI methods 1.2.11–1.2.12; Table S1). Our calculations indicate that the most thermodynamically favorable QM/MM-optimized oxy-Ty structure (oxy_48) has four water molecules occupying its active site, including W1–3, with W1 and W2 H-bonding to the μ-η²:η²-peroxide (Fig. 2B), in agreement with our MD simulations.

To evaluate the contributions of these H-bonding interactions to the H₂O→D₂O frequency changes in the vibrational modes in oxy-Ty, we calculated the vibrational frequencies for oxy_48 and correlated them to our rR data (SI methods 1.2.13–1.2.15). Our calculations show that oxy_48 closely reproduces the H₂O→D₂O frequency perturbations (Δν[Cu-Cu]=−1 cm⁻¹, Δν[Cu-O (B_3u)]=+10 cm⁻¹, Δν[O-O]=+1 cm⁻¹), observed by rR (Δν[Cu-Cu]< ±2 cm⁻¹, Δν[Cu-O (B_3u)]=+4±2 cm⁻¹, Δν[O-O]<±2 cm⁻¹; Fig. 1). As a control, a less thermodynamically stable QM/MM-optimized structure (oxy_59 in Fig. S9) that lacks W1 and W2, does not reproduce our rR results, as it does not exhibit any isotopic perturbation of its Cu-O (B_3u) mode (Table S1).

Our QM frequency calculations also reveal the underlying molecular mechanism for the Cu-O (B_3u) H₂O→D₂O upshift. Both W1 and W2 have two related vibrational modes, a twisting and a higher-energy rocking mode, which are present in the 450–650cm⁻¹ region in H₂O, above the Cu-O (B_3u) mode energy calculated at 425cm⁻¹,ⁱⁱⁱ while upon solvent deuteration, both of these water modes decrease in energy by ∼150–200 cm⁻¹ and are now below the Cu-O (B_3u) mode (SI methods 1.2.16). Consequently, in H₂O, the energy of the Cu-O (B_3u) mode is decreased due to its mixing with the nearby higher-energy water twisting mode (Fig. 3, left), while in D₂O, the Cu-O (B_3u) energy is now increased due to its mixing with the nearby lower-energy water rocking mode (Fig. 3, right). Thus, the combined effect of the Cu-O(B_3u) frequency decrease in H₂O and increase in D₂O via this mode-mixing mechanism results to the experimentally observed and computationally reproduced Cu-O (B_3u) H₂O→D₂O upshift.^iv This mode mixing mechanism establishes a molecular-level foundation for this solvent isotope-dependent rR v(metal-oxygen) increase, which has been routinely employed to identify H-bonds to metal/oxygen active sites in other enzymes,^21,22 and extends it beyond empirical identification and towards obtaining insights into electronic structure changes induced by H-bonding.

Fig. 3.

Vibrational coupling of the Cu-O(B_3u) stretch of the Cu(II)₂O₂ active site with two key water modes for W1 (twisting in H₂O and rocking in D₂O).

Previous structure/function investigations into the CBC family indicated that unlike plant CaOx, Ty lacks a bulky gatekeeper phenylalanine and contains a conserved Asn/Glu cluster (Fig. 2C), however the functional significance of these second-sphere correlations remains a subject of debate.^{13,13,15,20,24–27}. Inspection of our experimentally-calibrated structure of the oxy-Ty active site reveals that these key second-sphere residues are critical in allowing solvent access to the active site and stabilizing W2, and thus enable the H-bonding to the μ-η²:η²-peroxide (Fig. 2B). Although further experimental studies are required to establish the catalytic significance of these second-sphere residues, our QM/MM calculations on oxy-Ty reveal that H-bonding of W1 and W2 to the μ-η²:η²-peroxide weakens its O-O bond strength (Fig. S11). Indeed, H-bonds to the μ-η²:η²-peroxide are expected to stabilize its σ* lowest unoccupied molecular orbital, which would increase the back-bοnding donation from the Cu(II) d orbitals, resulting in weakening of the O-O bond.²⁸ This H-bonding perturbation of the O-O bond further correlates to previous reports where introduction of H-bond donor groups into the ligand scaffold of a μ-η²:η²-peroxo dicopper(II) complex promotes O-O bond cleavage and conversion to its bis-μ-oxo dicopper(III) isomer.²⁹ While such intramolecular H-bonding has been found to enhance both the H-atom abstraction reactivity of η¹-superoxo monocupric model complexes,^30,31 and the ammine oxidation reactivity of the η¹-hydroperoxo monocupric active site of an artificial metalloenzyme,³² the relevance of intra- or intermolecular H-bonding on the monophenol monooxygenation and catechol oxidation reactivities of μ-η²:η²-peroxo dicopper(II) complexes has not been investigated.

In conclusion, our spectroscopic and computational investigations define a previously elusive H-bonding network within the oxy-Ty active site, where water molecules (W1–3) form H-bonds with the μ-η²:η²-peroxide and conserved second-sphere residues (Fig. 2B). These key second-sphere residues vary within the CBC protein family, and likely affect the H-bonding interactions to the μ-η²:η²-peroxide in their respective oxy-intermediates. This study opens the way for further investigations into H-bonding interactions to the μ-η²:η²-peroxide of [Cu(II)₂O₂]²⁺ biological active sites and their model complexes, as well as determination of their mechanistic implications on O₂ activation, mono/diphenol substrate binding, and their subsequent reactivity.