Rational Design of a Histidine-Methionine Site Modeling the M-center of Copper Monooxygenases in a Small Metallochaperone Scaffold.

Abstract

Mononuclear copper monooxygenases peptidylglycine monooxygenase (PHM) and dopamine β-monooxygenase (DBM) catalyze the hydroxylation of high energy C-H bonds utilizing a pair of chemically distinct copper sites (CuH and CuM) separated by 11 Å. In earlier work, we constructed single-site PHM variants that were designed to allow study of the M- and H-centers independently in order to place their reactivity sequentially along the catalytic pathway. More recent crystallographic studies suggest that these single-site variants may not be truly representative of the individual active sites. In this work we describe an alternative approach that uses rational design to construct an artificial PHM model in a small metallochaperone scaffold. Using site-directed mutagenesis, we constructed variants that provide a His₂Met copper-binding ligand set that mimics the M-center of PHM. The results show that the model accurately reproduces the chemical and spectroscopic properties of the M-center, including details of the methionine coordination, and the properties of Cu(I) and Cu(II) states in the presence of endogenous ligands such as CO and azide. The rate of reduction of the Cu(II) form of the model by the chromophoric reductant N,N’-dimethyl phenylenediamine (DMPD) has been compared with that of the PHM M-center, and the reaction chemistry of the Cu(I) forms with molecular oxygen has also been explored, revealing an unusually low reactivity towards molecular oxygen. This latter finding emphasizes the importance of substrate triggering of oxygen reactivity, and implies that the His₂Met ligand set while necessary, is insufficient on its own to activate oxygen in these enzyme systems.

Graphical Abstract

Introduction

Copper active sites are found in an array of essential enzymes capable of catalyzing oxidative conversions of substrates as diverse as aromatic rings, catecholamines, peptide hormones, saturated alkanes, and polysaccharides.¹ The copper center is first activated via dioxygen binding, forming a Cu-dioxygen intermediate that then reacts further with substrate via chemistry that requires a finely tuned metal binding site, capable of redox cycling. The mononuclear monooxygenases are an important class of oxygen-activating cuproproteins which catalyze the insertion of a single oxygen atom from O₂ into the strong C-H bond of organic substrates. They include peptidylglycine α-hydroxylating monooxygenase (PHM),² dopamine β-monooxygenase (DβM),^{3, 4} its insect analog tyramine β-monooxygenase (TβM),^{5, 6} and the lytic polysaccharide monooxygenases (LPMO).⁷

Despite differing structures and substrate specificities, both PHM and DBM utilize a catalytic core made up of two nonequivalent mononuclear copper centers,^{4, 8–10} and react via an identical chemical mechanism.^11–13 Extensive spectroscopic,^14–17 kinetic,^{3, 13, 18, 19} and computational^20–22 work on PHM suggests that one site (CuM) controls catalysis, exhibiting an unusual His₂Met ligand set, while the other site (CuH) is responsible for electron storage and transport and exhibits a His₃ ligand set. In a proposed canonical mechanism, the Cu(I)M - substrate complex reacts with oxygen to form a Cu(II)-superoxo intermediate with sufficient electrophilicity to abstract a H atom from the nearby peptide substrate.^{18, 23, 24} An X-ray structure of “oxy-PHM”²⁵ has allowed a close analogue of the reactive superoxo species to be visualized as a four-coordinate, distorted tetrahedral entity containing an end-on superoxide ligand. Enzyme-bound cupric superoxo species have also been proposed as intermediates in LPMO catalysis where rapid re-oxidation of the mononuclear Cu(I) center by O₂ is suggestive of the formation of an unstable Cu(II)O₂^·− species,²⁶ but spectroscopic characterization of either intermediate has yet to be achieved.

The quest for spectroscopic signatures of mononuclear copper intermediates has stimulated the synthesis of a number of biomimetic model compounds whose chemistry with dioxygen has been probed at low temperature. As reviewed by Tolman and coworkers²⁷ a number of inorganic mononuclear cupric superoxo species have been characterized, and include both side-on (η²−),^{20, 28} and end-on (η¹−) complexes.^29–38. Early synthons were built from 3- or 4-coordinate all-N donor scaffolds, since studies on related Cu(II)-³⁹ and Cu(III)-peroxo⁴⁰ species had shown little or no stabilization via thioether coordination. More recent work on a N₃S(thioether)-Cu(I) complex which reacts with oxygen to form a superoxo intermediate⁴¹ suggests that the thioether moiety may increase the electrophilic character of the superoxo with respect to H-atom abstraction. While these studies have offered great insight into the chemistry and spectroscopy of cupric superoxo species they require the use of low temperature (−125°C) and aprotic solvents, and often exhibit a strong driving force towards formation of peroxo-bridged dinuclear complexes which lack electrophilic reactivity.

In previous work we reported the preparation and properties of PHM variants that were designed to allow study of the M- and H-centers independently.¹⁷ PHM H242A bound a single Cu at the H-center while H107AH108A bound a single Cu at the M-center. More recent crystallographic studies⁴² have indicated conformational mobility in these copper depleted mutants, suggesting that the single-site variants may not be truly representative of the individual active sites. Notwithstanding, the ability to interrogate the individual chemistry of the two non-equivalent copper sites is an important goal in fully understanding the mechanism of PHM and its congeners, leading us to seek alternative strategies. One approach to modeling the reactivity of enzyme active sites has made use of small peptide or protein scaffolds inside which metal binding sites can be assembled.^43–46 This has the advantage that the protein polypeptide can provide protection from solvent and can furnish any necessary H-bonding interactions from second sphere interactions while still maintaining aqueous solubility. We have applied this approach to the PHM system via use of the CusF scaffold a small mononuclear Cu(I) metallochaperone expressed in the E. coli copper export system, CusCBAF.⁴⁷ The metal binding site of native CusF is situated in a hydrophobic pocket and ligated by two sulfurs (from methionines M47, M49), the Nε of histidine 36 (H36) and a cation-π interaction with the aromatic ring of W44.^48–51 Using site-directed mutagenesis, we constructed the M49H CusF variant, along with its M47H analogue in the M8IM59IW44A triple mutant background. These engineered scaffolds provide a His₂Met copper-binding ligand set that mimics the M-center of PHM where the single Met residue in the protein is coordinated to copper. Specifically the system holds promise for obtaining a deeper understanding of a number of attributes of PHM M-site chemistry that remain enigmatic. For example, how does substrate hydroxylation remain fully coupled to oxygen consumption even with extremely slow variants (implying that substrate binding must trigger oxygen reactivity)?^{52, 53} What is the role of the Met ligand, and why is it essential for catalysis?^{6, 52, 54, 55} Can a cupric superoxo species be isolated, and what structural features are required for its stability?

Here we report results that show that the CusF-model accurately reproduces the chemical and spectroscopic properties of the M-center. We probe details of the methionine coordination and the properties of Cu(I) and Cu(II) states in the presence of exogenous ligands such as CO and azide. We also explore the reaction chemistry of Cu(I) forms with molecular oxygen and the Cu(II) forms with chromophoric reductants. The data reveal a close structural and electronic similarity to the PHM M-center, but an unusually low reactivity towards molecular oxygen which emphasizes the important role played by the PHM substrate complex in oxygen activation.

Figure 1.

Structural depiction of the PHM M-site model constructed in the CusF scaffold. The graphic shows an energy minimized representation of the CusF M49H variant in a CusF M8IM59IW44A triple mutant background (right) compared with the PHM M-site (left). The energy minimized structure was obtained by mutating the WT structure (PDB: 2VB2) in Chimera. The new structure was then submitted for energy minimization to the YASARA Energy Minimization Server (Krieger et al., Proteins: Structure, Function, and Bioinformatics 77, 114–122) and the final structure rendered in PyMOL (The PyMOL Molecular Graphics System, Version 1.8 Schrödinger, LLC).

Materials and Experimental Details

Construction of CusF mutants.

W44A M49H and W44A M47H mutation were introduced into CusF_6–88 M8IM59I trx-his6-tev sequence (both non-coordinating Met residues mutated to Ile) using overlap extension polymerase chain reaction (OE-PCR). Sense and antisense oligonucleotide primers encoding ~20 bases downstream and upstream of the mutation were used for site-directed mutagenesis and paired with primers upstream and downstream of two restriction enzyme sites, HindIII and Nco1. PCR products were purified on agarose gels. Final PCR products were extracted with phenol and chloroform, digested using restriction enzymes (NEB), separated on agarose gels, purified via a Qiagen PCR kit, and ligated into the pETDuet-1 expression vector. All variants were confirmed by DNA sequencing. The resulting plasmid pETDuet-1-CusF was then transformed into E. coli BL21-(λDE3) cells.

Expression and Purification of CusF.

A freshly streaked plate of BL21(DE3) cells containing pET-Duet1-CusF with the appropriate mutation was used to inoculate 10 ml of LB media containing 100 mg/ml ampicillin. After overnight incubation at 37°C with shaking at 250 rpm, the starter culture was used to inoculate a 1L flask of LB culture medium with 100 μg/mL ampicillin and 0.2% D-glucose. The culture was grown at 37°C until an OD₆₀₀ of 0.6 – 0.8 was reached at which point it was induced with 500 μM isopropyl β-D-1-thiogalactopyranoside. Growth was continued at 17 °C for 16–20 hrs after which cells were harvested by centrifugation. Pellets were re-suspended in column buffer (20 mM Tris, 150 mM NaCl, 10 mM imidazole, 5% glycerol). For purification, MgCl₂, Dnase and protease inhibitor were added and cells were lysed using an EmulsiFlex-C3. The lysate was centrifuged at 12,000 rpm for 50 minutes to remove cell debris. The supernatant was filtered using a 0.45 μm syringe filter and then applied to Ni-NTA column, rinsed with buffer, and eluted with 250 mM imidazole. The His₆-Trx tag was cleaved using tobacco etch virus (TEV) protease during an overnight incubation at 20°C by the addition of 1 part TeV to 100 parts protein and 5 mM β-mercaptoethanol. The protein was dialyzed overnight, then reapplied to the Ni-NTA column. The flow-through represented >95% pure, apo CusF protein with the His-tag cleaved. The final product was analyzed via SDS/PAGE (8–25% gradient stained with Coomassie brilliant blue R-250) which showed a single band at the appropriate molecular weight (~10 kDa). Proteins were then dialyzed and stored in 50 mM sodium phosphate pH 8.0. Protein concentration was quantified by bicinchoninic acid assay (BCA) and if needed concentrated using a 3 kDa molecular weight cutoff concentrator (Amicon).

Sample Preparation.

Unless otherwise stated, samples for spectroscopy and kinetics were equilibrated into combination buffer comprising 50 mM each of formate, MES, and HEPES, adjusted to the appropriate pH using HCl or NaOH as required.

Oxidized Samples.

Protein was reconstituted with CuSO₄ at a ratio of 2.5:1 via syringe pump (25 μl/hr). The mixture was then allowed to incubate on ice for 1 hr with stirring. Excess Cu(II) was removed by overnight dialysis in 50 mM sodium phosphate (NaP) pH 8. Holo protein was either used immediately in experiments or flash-frozen and stored in liquid nitrogen for future use. Metal-to-protein concentrations were verified by ICP-OES and BCA assay. For pH dependence, buffer exchange was achieved by a rapid 4-fold dilution of the 4-fold concentrated protein into combination buffer at the desired pH. A small protein loss was observed, but the copper to protein ratios remained at 1:1 as confirmed by ICP.

Reduced Samples.

Cu(II) reconstituted protein in the buffer of choice was reduced anaerobically by addition of a 2-fold excess of ascorbate buffered at the same pH. Protein was shown to be fully reduced by the complete loss of the Cu(II) EPR signal. For protein used in re-oxidation studies, the ascorbate was removed by two cycles of desalting on spin columns (Zeba).

Stopped-Flow Spectrophotometry.

Pre-weighed DMPD, syringes, argon-purged deionized water and buffer were made anaerobic by overnight storage in the anaerobic chamber. The 50 mM DMPD stock solution was prepared by dissolving pre-weighed salt in 1 ml of deionized water. This DMPD stock solution was diluted to 2 mM using deoxygenated deionized water. Stopped-flow experiments were conducted under anaerobic conditions at room temperature on a SX20 Applied Photophysics stopped-flow instrument enclosed in a Vacuum Atmospheres anaerobic chamber with oxygen levels of ≤ 1 ppm.

Analysis of stopped-flow data.

DMPD undergoes one-electron oxidation to the radical cation. The DMPD radical is a chromophoric agent which absorbs strongly at 515 nm (A₅₁₅) and 550 nm (A₅₅₀) with molar absorptivity at 515 nm of 5200 M⁻¹cm⁻¹. The formation of the DMPD radical strongly correlates with the reduction of the copper centers making possible the measurement of copper reduction rates by monitoring the variation of absorbance at 515 nm (A₅₁₅) versus time. The concentration of the reduced copper was determined from the equation

$$Concentrationofthereducedcopper=\frac{\mathrm{\Delta }{A}_{515}}{{\in }_{DMPD}}$$ (1)

Where ΔA₅₁₅ is the difference in absorption at 515 nm at time t, and ∈_DMPD = 5200 M⁻¹cm⁻¹. The reduced copper concentration was calculated for every time point and plotted against time using SigmaPlot 12.0. The data were fitted to a double exponential rate equation

$$A_t=A_0+A_1\left(1-e^{-k_{1}t}\right)+A_2\left(1-e^{-k_{2}t}\right)+ct$$ (2)

Where A₁ and A₂ are the absorbance as a function of time for each of the two exponential time courses with rate constants k₁, k₂ respectively. A₀ is the initial absorbance at time zero. In all experiments it was found that DMPD oxidation continued in a slow, almost linear phase with coefficient c for some time after the absorbance change indicated complete reduction of the copper center.

Spectroscopic Data Collection and Processing

Azide Titrations.

2 M sodium azide (Sigma) was added via syringe pump to CusF M49H Cu(II) in combination buffer pH 5.5 at a rate of 25 μl/hr to a final azide concentration of 100 mM. Samples were stirred continuously and spectra were collected every 20 seconds on a Cary 50 UV/Vis. The data were fit to various models using the program DYNAFIT.⁵⁶

X-ray Absorption Spectroscopy.

Samples were mixed with 20% (vol/vol) ethylene glycol and measured as frozen glasses at 10 K. Cu K edge (8.9 keV) extended X-ray absorption fine structure (EXAFS) and X-ray absorption near edge structure (XANES) were collected at the Stanford Synchrotron Radiation Lightsource on beamlines 9–3 and 7–3 using a Si 220 monochromator with a φ= 90° crystal set and a Rh-coated mirror located upstream of the monochromator using a 13 keV energy cut-off to reject harmonics. Kα fluorescence was collected using a 100-element (beamline 9–3) or 30-element (beamline 7–3) Canberra Ge array detector. A Z-1 metal oxide filter and Soller slit assembly was placed in front of the detector to attenuate the elastic scatter peak. A buffer blank was subtracted from the raw data to produce a flat pre-edge and eliminate residual Ni Kβ fluorescence of the metal oxide filter. Energy calibration was achieved by placing a Cu metal foil between the second and third ionization chamber. Data averaging, background subtraction, and normalization were performed using EXAFSPAK.⁵⁷ The experimental energy threshold (k=0) was chosen as 8985 eV. Spectral simulation was carried out by least-squares curve fitting, using full curved wave calculations as formulated by the program EXCURVE 9.2 as previously described.^{15, 58, 59}

Fourier-Transform Infrared Spectroscopy.

Purified CusF was concentrated to approximately 1 mM and transferred to an airtight conical vial in an anaerobic COY chamber. Samples were purged with a gentle stream of CO for 5 min and allowed to incubate for 10 min. Protein solutions (~ 1 mM) were loaded anaerobically into an IR cell (50 μm path length). Samples were equilibrated inside the IR sample chamber at room temperature for 15 min to purge water vapor and CO₂ prior to data collection. FTIR data were recorded on a Bruker Tensor 27 FTIR spectrophotometer continuously purged with CO₂-free dry air as previously described.⁶⁰ One thousand scans were collected for both protein sample and buffer blank from 2250 to 1900 cm⁻¹ at a resolution of 2 cm⁻¹. Spectral analysis including subtraction of the buffer-blank was performed using GRAMS AI spectroscopy software (Thermo).

Electron Paramagnetic Resonance.

Electron paramagnetic resonance spectra were measured on a Bruker Elexsys e500 spectrometer with the following experimental conditions: frequency 9.63 GHz, T = 100 K, microwave power 20 mW, gain 10 dB, modulation amplitude 10 G and sweep time 84 s. To determine the relative concentrations of paramagnetic copper in a given sample, a series of standard solutions containing 150–600 μM Cu(II)-EDTA in 50 mM Hepes buffer, pH 7.5 was measured to create a calibration curve. The concentrations of paramagnetic copper were determined by double integration compared to the standard curve. Spectral analysis was performed using GRAMS AI spectroscopy software (Thermo). EPR spectra were simulated using EASYSPIN.⁶¹

Results and Discussion

The purpose of this study was to develop an aqueous model system which could mimic the M-center of the mononuclear monooxgenases, leading ultimately to a deeper understanding of the underlying reaction chemistry of this active site. A number of methodologies exist for engineering single-site constructs all of which offer both promise and potential pitfalls. For example our laboratory has already advanced the study of the H242A (M-site empty) and the H107AH108A (H-site empty) single-site mutants which in solution have ‘native-like’ properties. However, other mutants (H107A, H108A) that loose copper from the H center in crystallo, exhibit “non-native” conformational mobility⁴² raising the possibility that single-site copper loss may lead to non-native conformations. Constructs that express only the M-site sub-domain may be a possible route to M-site reactivity given that limited proteolysis using LysC cleaves the inter sub-domain linker at K219.⁶² However single sub-domain constructs could be structurally compromised on account of their hydrophobic inter-domain interface and have not yet been reported. As an alternative, the present paper describes the rational design of an M-site model containing the His₂Met ligand set engineered into the CusF protein scaffold which we evaluate alongside other approaches for its ability to provide fundamental information of the chemical properties of the His₂Met ligand set.

For the reduced monooxygenases, parameters to mimic include coordination by the His₂Met ligand set provided by residues H242, H244, M314 for PHM, the presence of an unusually weak Cu(I)-S(Met) bond, and the ability to bind and/or activate exogenous diatomics (CO,O₂). A different coordination has routinely been observed for the oxidized forms of PHM¹⁵ and DBM⁶³ where the Met ligand is displaced and water and/or other exogenous ligands occupy vacant equatorial positions of a tetragonal coordinate structure. For example, the oxidized M site is known to bind azide, nitrite, and peroxide in crystallo,⁶⁴ and the latter has been postulated as an intermediate in the reaction pathway. Because the model does not yet have the capability to bind a substrate, our initial studies focus on validating the structural, spectroscopic and reactivity markers of the CuM active site.

Reduced Forms.

CusF M47H and M49H variants each bind one equivalent of cuprous ion to form 1:1 Cu(I) derivatives. The coordination chemistry of these Cu(I) forms was studied by X-ray absorption spectroscopy (XAS). Figure 2 shows the Fourier transform and extended X-ray absorption fine structure (EXAFS) of ascorbate-reduced CusF M49H in MES buffer pH 5.5, while data for the ascorbate-reduced M47H variant is depicted in Supporting Information Figure S1. Both reduced His₂Met variants produce spectra similar to that previously reported for the reduced M-center of PHM, with the best fit to the experimental data modeled by two N(His) at 1.96 Å and one S(Met) ligand with Cu-S = 2.25 – 2.28 Å, respectively (Table 1). This compares with values for Cu-N(His) and Cu-S(Met) of 1.98 Å and 2.18 Å respectively for the isolated M-center reported for the H107AH108A PHM variant,¹⁷ and 1.92 Å and 2.24 Å reported for the WT PHM.¹⁵ As noted earlier, recent crystallographic studies on PHM variants lacking copper in the H-center show altered structural features at the M-center,⁴² and in one isoform of the H108A variant show a closed conformation with a single copper coordinated by residues from H and M. Thus, while recognizing that the WT has both copper centers occupied, the Cu-S(Met) bond length of WT is a better indicator of similarities, and clearly shows a closer correspondence between model and enzyme. Absorption edge data (Figure 2, bottom) show edge features in the region 8983 eV as expected for 3-coordination, but with some differences in geometric or electronic structure.

Figure 2.

(a) Fourier transform and EXAFS (inset) for the ascorbate-reduced Cu(I)-M49H PHM M-site model. Black traces represent experimental data while red traces are simulations using EXCURVE 9.2. The parameters used in the fits are listed in Table 1. (b) Cu K absorption edges for the model compared with those of the WT enzyme (M-site and H-site metallated) and its H107AH108A variant (M-site only metallated: blue trace, M49H model, red trace WT, green trace M-site PHM variant.

Table 1.

Fits obtained to the ascorbate-reduced Cu K EXAFS of the M49H and M47H PHM models at pH 5.5 by curve-fitting using the program EXCURVE 9.2.

	Fa	Nob	R (Å)c	DW (Å2)	Nob	R (Å)c	DW (Å2)	Nob	R(Å)c	DW (Å2)	E0
Sample		Cu-N(His)d			Cu-S/N/O			Cu-COd
Cu(I)-M49H	0.51	2	1.94	0.017	1S	2.28	0.023				0.18
Cu(I)-M49HCO	0.58	2	1.99	0.007	1S	2.33	0.015	1 C C≡O angle	1.84 1.02 174°	0.006	−1.1
Cu(I)-M47He	0.88	2	1.97	0.009	1S	2.25	0.013				0.47
Cu(I)-M47HCOe	0.87	2	2.00	0.0009	1S	2.31	0.012	1 C C≡0 angle	1.82 1.05 175°	0.009	−1.6
Cu(II)-M49H	0.27	2	1.99	0.016	2N/O	1.99	0.016				−4.6
Cu(II)-M49H-azide	0.44	2	1.96	0.014	2N/O	1.99	0.014				−3.1

^aF is a least-squares fitting parameter defined as $F^2=\frac{1}{N}\sum _{i=1}^N{k}^6{(Data-Model)}^2$

^bCoordination numbers are generally considered accurate to ± 25%

^cIn any one fit, the statistical error in bond-lengths is ±0.005 Å. However, when errors due to imperfect background subtraction, phase-shift calculations, and noise in the data are compounded, the actual error is probably closer to ±0.02 Å.

^dFits included both single and multiple scattering contributions from the imidazole ring and/or the linear CO unit.

^eM47H samples were measured in 50 mM sodium phosphate buffer pH 8.

The value of the Debye-Waller (DW) factor (2σ², Table 1) for the Cu-S(Met) interaction is also of interest. In WT PHM and DBM the Cu-S DW is unusually large and variable, simulating in the range 0.012 – 0.025 Ų. The large DW for the enzyme has been suggested to arise from either two conformations at the M-center involving Met-on and Met-off forms (only one of which is active), or alternatively, scaffold-based protein dynamics that create a specific protein architecture and a fluxional Cu-S bond which might couple with other specialized vibrational modes involved in H-tunneling.^{14, 60, 65} While these suggestions are elegant, the finding that the model also exhibits a large DW for Cu-S(met) at temperatures close to 10 K, establishes that this is a property of the ligand set, rather than a property induced by the PHM scaffold. Sufficient data exist to compare Cu-S(Met) DW values across a series of 3-coordinate CusF and CusB His_xMet_y ligand sets. In this series Cu-S(Met) DW factors range from 0.007 Ų for the Met₃ site of CusB,⁵⁸ to 0.011 Ų for the HisMet₂ site of WT CusF⁵⁰ to 0.023 Ų for the His₂Met site of the CusF M49H model (this work). A plausible explanation for this trend may be put forward in which the Cu-S(Met) bond is inherently fluxional: when present together with two other Met residues in the Met₃ ligand set, the dynamics may favor the reversible dissociation of any one Cu-S, but would leave the other two tightly coordinated with a low average DW. For the HisMet₂ case the stronger donor His ligand would remain coordinated while one out of the two Met ligands would be in flux leading to a larger DW, while in the His₂Met case the Met ligand would always be fluxional. This argument also explains why Met residues often line potential transport pathways in importers and exporters such as CTR1^66–68, CusA⁶⁹ and CopA^{70, 71} since a single Cu-Met interaction provides a balance between selective binding and kinetic lability.

Carbon Monoxide Binding.

Carbon monoxide (CO) binding to reduced copper centers generates Cu(I)carbonyl complexes that with few exceptions exhibit 4-coordinate tetrahedral coordination. Since CO is a dioxygen analogue, CO binding is often used as a surrogate for O₂ reactivity, and with this in mind, CO has been shown to bind both the reduced M-center of PHM^{60, 72} and the reduced form of DBM,⁷³ forming carbonyls with stretching frequencies of 2093 cm⁻¹ and 2089 cm⁻¹ respectively. The ability of reduced CusF M49H and M47H to also form CO adducts would be indicative of analogous chemistry, and was tested using Fourier transform infrared spectroscopy (FTIR). Figure 3 shows that both the reduced M47H and M49H model systems produce a v(CO) peak indicating the formation of a carbonyl. The stretching frequency of an exogenous π-acceptor ligand such as CO coordinated to a metal is dependent on the degree of back-bonding from filled metal d-orbitals into the empty π*- antibonding orbitals of the CO ligand which in turn is influenced by the donor strength of the endogenous protein ligands. Based on a previously established library of Cu(I)carbonyl stretching frequencies, Cu(I) carbonyls are known to have a stretching frequency range of 2012 – 2045 cm⁻¹, where the upper range is typical of rare 3-coordinate His₂CO synthons⁷⁴ and the latter the lowest frequency observed in the His₃CO ligand set of arthropodal hemocyanins (see reference⁷⁵). The trend in decreasing ν(CO) with increasing donor strength is exemplified by comparison of the WT PHM M-site carbonyl (2092 cm⁻¹) with its M314H variant (2075 cm⁻¹) where replacement of thioether with histidine results in a 17 cm⁻¹ downshift.⁶⁰

Figure 3.

Fourier transform infrared spectra of the CO complexes of the PHM M-site models. (a) Cu(I)M49H-CO, (b) Cu(I)M47H-CO. Peaks were fit to either one or two Gaussian functions with frequencies and line widths as reported in the text.

Both CusF His₂Met models bind CO within the His₂Met ligand set range, with a v(CO) of 2086 and 2089 cm⁻¹ for CusF M47H and M49H, respectively. The peak shape of both the M47H and M49H carbonyls was characterized with Gaussian peak fitting. M49H carbonyl fits to a single peak, with a width at half height (FWHH) of 11.8 cm⁻¹. The M47H carbonyl when fit to a single peak gave a larger FWHH value of 15.5 cm⁻¹, but could also be fit by two Gaussians with FWHH values (11 cm⁻¹) that matched M49H-CO within the resolution of the spectrometer. These data suggest that the M49H carbonyl is a single species, while the M47H carbonyl may form two conformers.

The coordination chemistry and Cu-O bond angle of the His₂Met carbonyls was further investigated using XAS. Figure 4(a) shows the Fourier transform and EXAFS for the M49H carbonyl with absorption edge comparisons in Figure 4(b) while Figure S2 (Supporting Information) shows the EXAFS and absorption edges of the M47H Cu(I)-carbonyl. Metrical parameters used in the simulations are given in Table 1. The experimental data can be modeled by Cu-C bond distances of 1.82 – 1.84 Å, with a Cu-C-O angle close to linearity and a Cu-S(Met) distance of 2.32 – 2.34 Å. The Cu(I)-His₂Met carbonyls are expected to be four coordinate with addition of the CO ligand, and the ~0.07 Å increase in Cu-S bond length is consistent with the increase in coordination number. Both CusF His₂Met carbonyls compare well with metrical parameters (Cu-C = 1.80 Å with a ∠Cu-C-O = 179°) previously reported for the isolated M-center of the H107AH108A PHM variant although the 2.26 Å Cu-S(Met) distance is significantly shorter.¹⁷ The carbonyl parameters are also consistent with those reported for WT PHM, although the latter are extracted from data on samples that have both Cu sites occupied. While both CusF His₂Met carbonyls appear structurally identical on the basis of XAS analysis, the M49H carbonyl is electronically more similar to the enzymes with a stretching frequency only 3 wavenumbers less than PHM, and identical to DBM.

Figure 4.

(a) Fourier transform and EXAFS (inset) for the Cu(I) M49H-CO complex. Black traces represent experimental data while red traces are simulations using EXCURVE 9.2. The parameters used in the fits are listed in Table 1 and include metrical details of the coordinated CO ligand determined from multiple scattering analysis. (b) Cu K absorption edges for the model CO complex compared with those of the WT enzyme (M-site and H-site metallated) and its H107AH108A variant (M-site only metallated): blue trace, M49H model, red trace WT, green trace M-site PHM variant.

The M49H carbonyl was further investigated, for structural and electronic changes as a function of pH, using XAS and FTIR, respectively. Over the pH range 4 – 10 the overall coordination chemistry of the Cu(I) site was unchanged with the best fit to experimental data modeled by 2 Cu-N(His) bonds at 1.99(1) Å, a single S(Met) bond at 2.33(1) Å, Cu-C at 1.82(2) Å and a linear Cu-C-O angle. All M49H carbonyls produced FTIR spectra with a peak centered at 2089 cm-1 and width at FWHH of 11 cm⁻¹ indicating the carbonyl complex is pH independent (Supporting Information Figs S3, S4 and Table S1).

As previously discussed in both WT PHM and the CusF His₂Met models the S(Met) DW factor is large and variable, the latter due in part to large errors associated with simulating a very weak signal. The CusF M49H carbonyl provides an additional opportunity to define the origin the large DW factor observed in the His₂Met ligand set. The M49H carbonyl produces a single peak in the FTIR establishing a single 4-coordinate species, with the EXAFS analysis predicting a contribution from 1 Cu-S with a correspondingly large DW between 0.012 and 0.023 Ų. Further, this single species remains structurally identical across the full pH range. This in turn provides strong evidence for a single Met-on configuration, since an equilibrium between Met-on and Met-off forms of the carbonyl, must lead to two distinct peaks (or at least peak broadening) in the FTIR. Therefore we conclude that the large DW factors are indeed the result of fluxional Cu-S(Met) dynamics as suggested above for the WT protein.

Whereas the CusF M49H model carbonyl stretching frequency matches that of the PHM M-site carbonyl in the absence of substrate, binding of substrates to the enzyme causes an additional drop in the ν(CO) in a substrate-dependent fashion with acetyl-YVG dropping the frequency to 2062 cm-1.⁶⁰ The 30 cm⁻¹ red-shift implies a large increase in the electron donating power of the CuM ligand set induced by substrate binding and signals electronic activation of Cu(I)-bound diatomics, suggesting a plausible mechanism for substrate triggering of catalysis. Notwithstanding, the structural origin of the substrate-induced ν(CO) downshift is unclear since EXAFS data show no changes on substrate binding. The crystal structure of PHM reveals that M-site ligand H242 forms a strong H-bond (2.8 Å) between its Nε imino group and the amide oxygen of the side chain of Q272. In the presence of substrate an additional slightly longer interaction forms between H242 Nε and the substrate carboxy terminus. Taken together these interactions suggest a protein-mediated deprotonation of the imidazole ring of H242 which is enhanced on substrate binding and could provide CO/O₂ activation. The M49H model provides some insight into this hypothesis. WT CusF H36 forms a weak 3.3 Å H-bond with its own main chain amide O, and although no structure exists for CusF M49H, inspection suggests no strong H-bonding ligand within H-bonding distance of the H49 distal N in this variant. We can therefore conclude that the 2089 cm⁻¹ frequency represents a non H- bonded conformer which in turn would imply that in solution the 2092 cm⁻¹ band in PHM is also a non H-bonded form. This raises the possibility that substrate binding induces the strongly H-bonded form observed in the crystal structures of PHM, leading to the observed drop in ν(CO).

Oxidized Forms.

CusF M49H binds one equivalent of cupric ion to form a Cu(II) derivative. We investigated the properties of this species using, XAS, EPR and exogenous ligand binding. Figure 5(a) shows the Fourier transform and EXAFS for oxidized CusF M49H. When simulated, the best fit to experimental data was obtained using two histidines and two water-derived ligands, with distances consistent with those reported for the oxidized CuM site of PHM¹⁵ (Table 1). In complete agreement with EXAFS analysis of the oxidized enzymes, CusF M49H shows no contribution from the S(Met) ligand, further supporting CusF M49H as a structural model for the M-site PHM.

Figure 5.

(a) Fourier transform and EXAFS (inset) for oxidized Cu(II)-M49H PHM M-site model. Black traces represent experimental data while red traces are simulations using EXCURVE 9.2. The parameters used in the fits are listed in Table 1. (b) X-band CW EPR spectra of the oxidized M49H PHM M-site model (top trace) and its azido derivative (bottom trace) determined at pH 5.5. Black traces are experimental data while red traces are simulations using EASYSPIN. The oxidized model was simulated by two components in a ratio of 1 : 0.3, with g and A values for component 1: gx=2.049, gy=2.076, gz=2.301, Az=481 MHz; and for component 2: gx=2.062, gy=2.084, gz=2.238, Az=472 MHz. The azido species also exhibited a two component fit in the ratio 1 : 0.25 with g and A values for component 1: gx=2.041, gy=2.068, gz=2.253, Az=476 MHz; and for component 2: gx=2.029, gy=2.107, gz=2.269, Az=390 MHz. Spectra were collected at a temperature of 100 K, microwave frequency 9.688 GHz, 100 Khz modulation, 10 G modulation amplitude, 20 mW microwave power and 1000 G sweep width with the field centered at 3100 G.

Oxidized CusF M49H exhibits an EPR spectrum that has contributions from two Cu(II) species which vary in a pH-dependent fashion (Figure 5(b)). Also of note, spin quantitation against a Cu(II)-EDTA standard shows that the amount of EPR-detectable copper is also pH dependent, decreasing to less than 50% above pH 5 (Table S2 and Figure S5 Supporting Information). Taken together the data suggest an equilibrium between monomers and hydroxy-bridged dimers formed by ionization of one or more coordinated solvent molecules. Attempts to confirm the presence of dimers using size exclusion chromatography were unsuccessful due to copper dissociation on the column.

Azide binding has been used as a reporter ligand for the oxidized copper centers in type 2 copper proteins since it gives rise to a LMCT band around 400 nm.^76–78 The Cu(II) form of CusF M49H also binds azide generating a species that elicits close to 90 percent EPR detectable Cu(II) consistent with formation of mononuclear azido adducts that resists dimerization. The azido complex gives rise to the expected LMCT band at 390 nm (Figure 6), which can be titrated to yield a formation curve best fit by a single azido species with K_D=3.4 mM. This chemistry is typical of type 2 Cu(II) sites in proteins, and closely resembles published PHM^{64, 79} and DBM⁷⁶ reactivity.

Figure 6.

Titration of the oxidized M49H PHM M-site model with azide at 390 nm. The right panel shows a fit to the experimental UV/vis data modeled by a single Cu(II)-azido adduct with K_D = 3.4 mM.

Oxygen Reactivity.

The oxygen reactivity of the reduced form of CusF M49H was tested in two series of experiments. The first series examined the rate of oxidation of ascorbate -reduced M49H with molecular oxygen using an oxygen electrode. All excess reductant was first removed by two cycles of desalting in spin columns. The ascorbate-free anaerobic Cu(I) complex was then added to an oxygen saturated solution of combination buffer pH 5.5, and the oxygen consumption monitored. No oxygen was consumed in this experiment, indicating a remarkable lack of reactivity. When the experiment was repeated with the carbonylated ascorbate-free protein, a similar lack of reactivity was observed. In the second series, we tested the ability of the M49H model to undergo a simple redox turnover reaction by adding sub-stoichiometric amounts of the oxidized protein to oxygenated buffer in the presence of excess ascorbate. Here the reduced Cu(I) form of the protein could form by ascorbate reduction and subsequently react with oxygen to regenerate the Cu(II) form in a catalytic redox cycle. Again no reactivity was observed. Adding the PHM substrate dansyl-Tyr-Val-Gly to this solution did not lead to observable oxygen consumption as expected on the basis that the model is not yet designed to bind a peptide substrate. Therefore the Cu(I)-His₂Met ligand set is unreactive to molecular oxygen in agreement with the substrate-triggering of activity exhibited by the ES complex of PHM.

Reduction by N,N’-dimethyl phenylenediamine (DMPD).

DMPD is a chromophoric reductant that has been shown to support substrate hydroxylation in PHM and DBM,⁸⁰ and has been used previously to investigate the rates of reduction of the copper centers in PHM.⁸¹ We investigated the rate of reduction of the Cu(II)-M49H model by DMPD using stopped-flow spectrometry. The stopped-flow trace and rate of reduction of the Cu(II)-M49H model is shown in Figure 7. The data were fit by non-linear regression to a rate equation which describes the process as a combination of fast and slow pseudo first order reactions and is identical to that used previously to analyze the DMPD reduction of the WT PHM.⁸¹

$$A_t=A_0+A_1\left(1-e^{-k_{1}t}\right)+A_2\left(1-e^{-k_{2}t}\right)+ct$$ (3)

Where A₁ and A₂ are the absorbance as a function of time for each of the two exponential time courses with rate constants k₁, k₂ respectively. A₀ is the initial absorbance at time zero. As before, it was found that DMPD reduction continued in a slow almost linear reaction after all the Cu(II) complex had been reduced. In a reaction of 20 μM Cu(II)-M49H with 1 mM DMPD in combination buffer at pH 5.5 under strictly anaerobic conditions k₁ and k₂ had values of 102 ± 2 s⁻¹ and 4.0 ± 0.1 s⁻¹ and appeared to be present in a ratio of 2 : 1 respectively. The data show that the Cu(II) complex is reduced rapidly but at least two species likely exist with differing rates of reduction. We note that the EPR spectroscopy described above also identifies two components whose composition and g/A values are pH dependent. Therefore, we suggest that the biphasic reduction kinetics are the result of differing rates of reduction for protonated and deprotonated forms of the Cu(II)-M49H complex. Although the ratios of major and minor components differ somewhat between the stopped-flow and EPR data, we note that they represent data collected at room temperature versus 100 K respectively which can alter the protonation equilibrium significantly.

Figure 7.

Stopped-flow measurement of the reduction of 20 μM Cu(II)-M49H by 1 mM DMPD in combination buffer at pH 5.5 and 23°C. The top panel is a plot of calculated Cu(I) concentration versus time using ε₅₁₅ = 5200 M⁻¹cm⁻¹. The data are simulated by a double exponential rate equation with fast and slow rate constants of 102 and 4 s⁻¹ respectively in an approximate ratio of 2:1.

Conclusions.

We have described a model system for the M-center of PHM and DBM which has been designed by mutagenic re-purposing of a metallochaperone copper binding site. Although a crystal structure is not yet available, energy minimization suggests a structure close to that of the crystallographic M-centers, while spectroscopic investigations confirm strong correlations between the structure and electronics of model versus enzyme. In particular, adduct formation between CO in the reduced forms, and azide in the oxidized forms, demonstrates similar reaction chemistry. Of significance, the data show clearly that structural attributes such as the weak Cu-S(Met) bond (XAS), and the binding of diatomics (ν(CO)) are properties of the His₂Met ligand set itself, rather than electronics induced by the enzyme protein scaffold.

We also investigated the oxygen reactivity of the reduced forms of the model finding that the Cu(I)-His₂Met site was unreactive to O₂. Consensus has arisen around the notion that 3-coordinate Cu(I) complexes with N₃ and/or N₂S ligand sets are reactive towards oxygen, and there is a substantial inorganic literature that supports this premise in aprotic solvents.⁸² In proteins, there is likewise a large body of work which demonstrates fast reactivity of O₂ with Cu(I)His₃ active sites albeit mainly in binuclear (hemocyanin, tyrosinase, cytochrome oxidase) or trinuclear (multi-copper oxidase) systems.¹ With the exception of LPMOs where the Cu(I) state appears to be extremely reactive towards dioxygen,^{26, 83} the data on type 2 mononuclear copper centers is extremely sparse, and in the case of PHM and DBM is complicated by the presence of two chemically distinct spatially separated metal sites. We believe that our model data provide some of the first conclusive evidence that mononuclear Cu(I) centers in aqueous environments may be less reactive towards O₂ than presumed, and require substrate activation. While this may seem surprising, it is nonetheless entirely consistent with the observed reactivity of the enzymes. Previous work has clearly established that oxygen reduction in PHM^{52, 53, 60} is 100 percent coupled to substrate hydroxylation even with slow substrates and mutants with low activity, implying that the M-center is unreactive until substrate is bound. While the origin of this activation is still unknown, it raises questions relating to what structural and/or chemical properties induce the oxygen reactivity that is observed during catalysis.

The development of this model system is in its early stages, but we anticipate further efforts to introduce reactivity that will model substrate binding and oxygen activation. Such efforts have been successful in other rationally designed enzyme mimics such as the three-helix bundles designed by Pecoraro and coworkers to model nitrite reductase⁴⁴ or the engineered myoglobin systems of Lu and coworkers that are successful mimics of nitric oxide reductase.^{84, 85} Modification of second sphere H-bonding effects to induce diatomic activation⁴⁶ or to alter binding constants of exogenous ligands⁸⁶ has also been achieved by Borovik and coworkers in rationally designed copper binding sites engineered into streptavidin scaffolds. These emerging systems all leverage the potential of biomimetic modeling in aqueous environments that can couple the effects of protonation and H-bonding with the promise of induced substrate reactivity. Our future studies on the M-site model will explore these and other areas of biomimetic chemistry.