Electrochemical and Structural Study of the Buried Tryptophan in Azurin: Effects of Hydration and Polarity on the Redox Potential of W48

Abstract

Tryptophan serves as an important redox-active amino acid in mediating electron transfer and mitigating oxidative damage in proteins. We previously showed a difference in electrochemical potentials for two tryptophan residues in azurin with distinct hydrogen-bonding environments. Here, we test whether reducing the side chain bulk at position Phe110 to Leu, Ser, or Ala impacts the electrochemical potentials (E°) for tryptophan at position 48. X-ray diffraction confirmed the influx of crystallographically resolved water molecules for both the F110A and F110L tyrosine free azurin mutants. The local environments of W48 in all azurin mutants were further evaluated by UV resonance Raman (UVRR) spectroscopy to probe the impact of mutations on hydrogen bonding and polarity. A correlation between the frequency of the ω17 mode—considered a vibrational marker for hydrogen bonding—and E° is proposed. However, the trend is opposite to the expectation from a previous study on small molecules. Density functional theory calculations suggest that the ω17 mode reflects hydrogen bonding as well as local polarity. Further, the UVRR data reveal different intensity/frequency shifts of the ω9/ω10 vibrational modes that characterize the local H-bonding environments of tryptophan. The cumulative data support that the presence of water increases E° and reveal properties of the protein microenvironment surrounding tryptophan.

Introduction

Tyrosine and tryptophan play critical roles in redox reactions by mediating long-range electron transfer (ET), often coupled to proton transfer.^1,2 For example, tyrosine radicals, formed from proton-coupled electron transfer (PCET) reactions, are linked to function in several protein systems, including cytochrome c oxidase, photosystem II, ribonucleotide reductase (RNR), cyclooxygenase, and fatty acid α-oxygenase.^3,4 Tryptophan, one of the least abundant amino acids, has higher prevalence in oxidoreductases, where it has been implicated in redox “wires” that may help avert undesirable oxidative damage.⁵ Tryptophan also has functional roles in ET, such as in DNA photolyase, cryptochromes, mauG, and an engineered peroxidase, where strings of tryptophan residues mediate long-range ET and/or PCET.⁶⁻⁸

Complex protein systems, such as those involved in bioenergetics, enzyme catalysis, and redox homeostasis, often exploit aromatic amino acids for long-range and reversible ET. The local environment is expected to tune the reduction potentials of the amino acid radicals that modulate the properties of ET and directionality of electron flow.⁹ How the local, heterogeneous environment of a folded protein induces this control is not well understood. This challenge of direct electrochemical detection and characterization of specific amino acid radicals in complex systems is due, in part, to unwanted contributions from other redox-active species. To overcome these limitations, previous thermodynamic and kinetic studies have focused on amino acid mimics to assess the impact of solvent, intramolecular hydrogen bonding, and proton/deuterium transfer on the rate and driving force of ET.¹⁰⁻¹⁶ Biomimetics, including the designed (poly)peptides with single redox-active amino acids, have expanded the scope of the small molecule studies to simulate the anisotropic nature of the biological realm. These systems have begun to resolve how sequestration of redox centers from solvent influences the thermodynamic properties of amino acids.^17,18

There is a strong interest in understanding how water may influence the PCET reactions facilitated by tyrosine and tryptophan.¹⁹⁻²² There have been limited examples of tuning the local environment in biomimetic systems through strategic mutagenesis to assess how subtle changes in local polarity, hydrogen bonding, and hydration might influence the thermodynamic properties of these amino acid side chains embedded within a protein matrix. Recently, we demonstrated that varying the hydration in the local protein environments surrounding the tryptophan in the model protein azurin (Az) may influence the Trp^•/TrpH electrochemical potentials.²³ The P. aeruginosa azurin is a 128 amino acid, predominantly β-barrel protein that contains a single natural tryptophan (W48) in the core of the barrel. W48 in the tyrosine-free variant of azurin (AzW48) has been described as rigid and residing in a nonpolar environment (Figure 1A) characterized with blue-shifted fluorescence.²⁴ By comparison, a modified azurin variant (AzW108) was previously constructed in which W48 and the two native tyrosine residues were mutated to redox-inert Phe and a single tryptophan was inserted into a more polar environment at position 108.²⁵ A structural model, generated from molecular dynamics (MD) simulations, predicted an interaction between the indole nitrogen of W108 and a water molecule at the protein/solvent interface, consistent with its red-shifted fluorescence.²³ The electrochemical potential of tryptophan in AzW108 was found to be significantly increased by +100 mV, relative to AzW48, across a broad pH range, from 3 to 9.²³ The increased polarity near the tryptophan residue of AzW108 relative to AzW48 raises the question on the role of hydrogen bonding on the electrochemical potentials of tryptophan.

Figure 1.

Designed azurin variants. X-ray structure of wild-type azurin (PDB: 1E67) in panel A. In panel B, structures of modified azurins to investigate the effect of the local environment around W48 from the present study are shown; F110S is from PDB: 1ILU.

In the current work, we expand upon these previous studies to investigate newly constructed azurin variants, centered on W48, in which the tryptophan is exposed to variable degrees of solvation, introduced by cavity altering mutations (Figure 1). Previously, an X-ray structure of an azurin mutant showed the presence of water molecules in the cavity surrounding W48 when Phe-110 was mutated to Ser (AzF110S).²⁶ Additional mutants to F110, described here, show similar penetration of water in the W48 cavity caused by smaller aliphatic residues. The presence of this water impacts the electrochemical potential of W48 and suggests a correlation between degree of polarity/hydrogen bonding and the redox potential.

Methods

Azurin Mutants and Peptide

The genes encoding Pseudomonas aeruginosa azurin variants AzF110S (W48/Y72F/Y108F/F110S), AzF110L (W48/Y72F/Y108F/F110L), and AzF110A (W48/Y72F/Y108F/F110A) were prepared by QuikChange site-directed mutagenesis. Other variants include AzW108 (W48F/Y72F/Y108W) and AzW48 (Y72F/Y108F). Azurin was expressed in E. coli BL21 (DE3) cells as previously described.²³ The protein was purified to homogeneity from a semipurified periplasmic fraction using a HiTrap SP column attached to an AKTA FPLC. Typical yields were 10–30 mg pure protein/L of cell culture. Protein purity was characterized by SDS-PAGE and size exclusion chromatography (Figure S1). Using this method, the protein contained a mixture of Cu²⁺ and Zn²⁺. Because our previous square wave voltammetry (SWV) results indicated no significant difference in the reduction potentials of W48 in Cu²⁺- or Zn²⁺-containing forms of azurin, no further steps were taken to reconstitute the protein. The gene encoding α3W32 was synthesized by GenScript (Piscataway, NJ) and subcloned into a pET-28a(TEV) expression vector. The bacterial expression and purification of α3W32 was performed as described previously.²⁷ The N-terminal his₆ tag was removed by treatment with TEV and further purified by Ni-NTA prior to use. The 12-amino acid peptide, β-W2 (RWVEVNGOKIFQ), was synthesized previously.²⁸

Circular Dichroism (CD) Spectroscopy

The impact of the mutations on the azurin secondary structure was assessed using CD measurements (Jasco CD spectrometer J-815). The CD spectra of azurin mutants were measured at room temperature in a 10 mM APB (acetate/phosphate/borate) buffer at pH 7.0 and pH 9.0. The protein concentration was 25 μM to ensure the voltage remained at or lower than 600 V within the spectral range of 190–260 nm.

Differential Scanning Calorimetry (DSC)

To assess the impact of the mutations on the stability of the protein, azurin variants were studied via DSC. DSC experiments were conducted on a TA Instruments Nano-DSC microcalorimeter. These protein samples were 50–100 μM with 25 mM phosphate buffer (pH 7.0). The DSC experiments were carried out in heat-only mode, from 30–90 °C at a rate of 1 °C min^–1. The pressure was maintained at 3 atm. These experiments were performed in duplicate or triplicate.

Immobilization and SWV of Azurin

Pyrolytic graphite electrodes (2 mm diameter) were polished on a SiC pad, sonicated briefly in water, rinsed with ethanol, and dried under a stream of argon gas. Electrodes were prepared following layer-by-layer (LbL) methods to immobilize the protein on the surface: 30 μL drops of poly(diallyldimethylammonium) (PDDA, 2 mg mL^–1; 50 mM NaCl), polystyrenesulfonate (PSS, 2 mg mL^–1; 50 mM NaCl), and protein were applied for 15 min each cycle. The following architecture resulted on the electrode surface: PDDA/PSS/(Protein/PSS)₂/Protein. The final layer was rinsed with water, and the electrodes were capped and stored at 4 °C until use. The presence of the protein on the electrode was analyzed using a CH Instruments 600A potentiostat with Pt counter and saturated calomel (saturated KCl) reference electrodes. SWV scans were performed with the following parameters: potential scan range 0.0 to 1.0 V (vs SCE), step increments of 0.004 V, amplitude of 0.025 V, frequency of 200 Hz, quiet time of 2 s, and a sensitivity of 1 × 10^–5 A. Origin Pro software was used to analyze electrochemical data, and the data were background subtracted, subjected to 15-point Savitzky–Golay smoothing, and corrected for NHE.

The electrochemical potential of the azurin Cu(II/I) couple was determined by cyclic voltammetry (CV). Azurin was immobilized on alkanethiol (1:1 6-mercapto-1-hexanol:1-hexanethiol) self-assembled monolayer-modified gold electrodes. Solutions of the 1:1 alkanethiol mixture (5 μM) were applied to an electrochemically cleaned gold electrode surface for 1 h. The electrodes were rinsed, and the azurin protein sample (2 mg/mL) was applied to the surface overnight at 4 °C. CV data were collected in 25 mM potassium phosphate buffer (pH 7.0).

Electron Paramagnetic Resonance (EPR) Spectroscopy

EPR spectra on copper-containing azurin in fused quartz sample tubes were recorded at 77 K on a Bruker Instruments (Billerica, MA) EMX 10/12 X-band EPR spectrometer, equipped with a ER4116DM cavity and a Wilmad (Buena, NJ) liquid nitrogen dewar. Azurin samples were prepared at 0.3–1 mM final concentration in 20 mM potassium phosphate buffer, pH 7.5. The EPR spectrometer conditions were as follows: microwave frequency, 9.63 GHz; microwave power, 2 mW; modulation frequency, 100 kHz; modulation amplitude, 5 G; time constant, 81.92 ms; conversion time, 81.92 ms; resolution, 1024.

Protein Crystallization and Structure Determination

Purified azurin was concentrated to 10 mg/mL with an Amicon 5 kDa centrifugal filter (Millipore Sigma); this process simultaneously exchanged the buffer to 50 mM Tris, pH 8.0. Azurin crystals were obtained by hanging drop vapor diffusion using ViewDrop 96-well seals (SPT Labtech) or 24-well VDX plates (Hampton Research). The initial crystals were obtained from crystallization screening conditions of JCSG Plus and PACT premier (Molecular Dimensions), and Wizard (Emerald BioSystems), using Oryx8 (Douglas Instrument). Azurin was crystallized in 22–25% (w/v) PEG 3350, 0.1 M Hepes, pH 7.0, with various salts: 0.2 M ammonium chloride, 0.1 M ammonium sulfate, or 0.1–0.2 M sodium formate. Crystals appeared at room temperature within 2–3 days. X-ray diffraction data were collected at Beamlines 19-BM (Structural Biology Center) and 23ID (National Institute of General Medical Sciences and National Cancer Institute Structural Biology Facility) at the Advanced Photon Source, Argonne National Laboratory, Lemont, IL. Initial phasing was determined by molecular replacement using Phaser²⁹ based on PDB ID 4KOC. Residue tracing was performed with ARP/wARP³⁰ and Coot.³¹ The model was refined with Phenix³² and Refmac.³³

Ultraviolet Resonance Raman (UVRR) Spectroscopy

Ultraviolet resonance Raman spectra were acquired with 228 nm excitation from the fourth harmonic of a 1 kHz Ti:sapphire laser.^34,35 The excitation beam was focused on a vertical quartz capillary (100 μm inner diameter, 160 μm outer diameter). The sample was flowed through this capillary at a flow rate of 0.16 mL/min to ensure fresh sample for each excitation pulse. After a single pass through the capillary, the sample was discarded. The incident power on the sample was 0.6 to 2.6 mW for the different trials; power-dependence experiments indicated the UVRR signal from azurin was linear up to the highest power of 2.8 mW. The scattered light was collected and filtered through a prism prefilter (slit opening of 110 μm) to reject Rayleigh scattering, followed by dispersion in a 0.5 m spectrograph equipped with a 3600 gr/mm holographic grating. The scattered light was detected by a CCD camera (Princeton Instruments). The Raman shift was calibrated with acetonitrile; the accuracy based on Gaussian fits to the acetonitrile bands is ±0.5 cm^–1, and day-to-day variation in the CCD pixels for a given peak led to a repeatability of ±1 cm^–1. Gaussian fits to the ω17 and overlapping ω9/ω10 regions were averaged for multiple trials.

Density Functional Theory (DFT) Calculations

All calculations were performed by the ORCA 5.0.3 quantum chemistry package.^36,37 The hybrid functional B3LYP with the Pople basis set 6-311++G(2df,2dp) with dispersion correction (Becke Johnson damping scheme) was used to optimize the 3-ethyl-indole (3EI) systems (3EI with or without explicit water and with or without various implicit solvents) to their equilibrium geometry.^38,39 B3LYP/6-311++G(2df,2dp) was also used to calculate the vibrational frequencies of the 3EI systems. There were no virtual frequencies, confirming the systems were at a local minimum of the potential energy surface. For systems with implicit solvent, the conductor-like polarizable continuum model (CPCM), as implemented in ORCA 5.0.3, was used to incorporate several different implicit solvents (water, methanol, and cyclohexane)⁴⁰ for the optimization of the system geometry and the calculation of the vibrational frequencies.

Results and Discussion

Characterization of the Local Environment of Azurin Mutants

The absorption and fluorescence spectra of the purified azurin variants were collected at pH 7.0 (Figure 2A–D, Figure S2). Note that these azurin mutants contain only one redox-active side chain (W48) as the two native tyrosine residues have been mutated to Phe. In all cases, the UV absorption signature exhibits fine features for tryptophan, similar to the wild-type azurin absorption spectrum. The absorption and fluorescence spectra can be compared to those of skatole solutions in solvents with varying polarity (Figure 2E–H). Skatole dissolved in a low dielectric solvent (e.g., cyclohexane, chx) shows fine features in the UV–vis absorbance spectrum, similar to the azurin mutants. Furthermore, fluorescence peaks, which are known to be a marker for the local polarity, support the nonpolar character of the W48 environment.⁴¹⁻⁴³ For comparison, the skatole emission shifts from 370 nm in water to 307 nm in cyclohexane (Figure 2E–H, Table S1). The addition of hexamethylphosphoramide (HMPA), a proton acceptor, to the skatole solution in cyclohexane shifted this peak to approximately 325 nm (Figure 2I), supporting the ability of fluorescence to report on the polarity and/or hydrogen bonding of the local environment. Across the azurin variants, the peak emission wavelengths were different (Figure 2A–D). AzW48, which has the single tryptophan in the native position of the wild-type protein, has an emission peak of 308 nm, which is the most blue-shifted wavelength that has been found in a natural protein.²⁴ This result is consistent with its previously characterized X-ray structure,⁴⁴ in which the tryptophan is buried in a rigid and solvent-free β barrel (Figure 1A). Conversely, AzW108 exhibits a red-shifted emission peak (λ_em) at 335 nm (Figure 2D). From the MD structure, the indole ring of W108 is moderately solvent accessible (SASA = 7–9%) and is predicted, to form a hydrogen bond interaction with a water network that leads to the surface of the protein.²³ This MD model provides structural support for the increase in polarity surrounding W108, consistent with the λ_em of 335 nm.

Figure 2.

UV–vis absorption (continuous line) and fluorescence (dashed line) spectra of (A) AzW48, (B) AzF110S, (C) AzF110A, (D) AzW108, and (E–I) skatole solutions. The protein solutions were 20 μM in 25 mM phosphate, pH 7.0 for absorption. For fluorescence, the protein samples were diluted to 5 μM. As isolated from bacterial cultures, these azurin samples contain a mixture of Cu²⁺ and Zn²⁺ (see Figure S4). The excitation wavelength for the azurin fluorescence spectra was 290 nm. In panels E–H, the skatole concentrations were 200 (absorption) or 10 (fluorescence) μM in the solvents: aqueous buffer, ethanol (EtOH), dioxane (diox), and cyclohexane (chx). In panel I, 0.08% (v/v) of the proton acceptor, HMPA, was added to cyclohexane.

In addition to these two previously characterized azurin variants, we constructed variants of AzW48 in which the cross-stranded Phe110 was mutated to Ser (AzF110S), Ala (AzF110A), or Leu (AzF110L). The λ_em values for tryptophan in the AzF110S and AzF110A variants are both 322 nm (Figure 2B,C), a value that lies in between the λ_em values for AzW48 (308 nm) and AzW108 (335 nm). The fluorescence spectrum of AzF110L was quenched, despite a scan of the excitation wavelength, absence of Cu²⁺, and attempts with various buffer solutions. These fluorescence data suggest that altering the bulk at the cross-strand Phe to smaller side chains results in red-shifted fluorescence maxima indicative of enhanced polarity and/or hydrogen bonding in the cavity.

To evaluate the effect of the mutations on the entatic state of the copper center in azurin, we measured the electrochemical potential of the Cu(II/I) couple using cyclic voltammetry with the proteins immobilized on modified Au electrodes.^23,45 The electrochemical potential of the cupric center in AzF110A was only slightly reduced (ΔE° ≤ 30 mV) relative to the parent protein, AzW48 (Figure S3). AzF110S and AzW108 did not produce a strong electrochemical response, likely due to the lower copper content in these isolated protein samples. AzW48 is isolated from E. coli cultures with mixtures of Cu²⁺ and Zn²⁺, typically in a 3:2 ratio, based on inductively coupled plasma (ICP) data. While we did not collect ICP data for these mutants, the copper content of azurin can also be estimated based on the ratio of the absorbance of the ligand-to-metal charge transfer (LMCT) band of Cu²⁺ at 625 nm relative to the protein A₂₈₀, with an expected A₆₂₅/A₂₈₀ ratio of 0.88⁴⁶ corresponding to 100% copper occupancy. Our as-isolated AzW48 sample exhibited an A₆₂₅/A₂₈₀ of 0.52 (Figure S4), consistent with 60% copper loading measured from ICP data²³ and comparable to the typical range of 0.50–0.53.^47,48 For comparison, AzF110A had slightly elevated copper content (A₆₂₅/A₂₈₀ = 0.61), whereas AzF110S and AzW108 were isolated with low copper content, A₆₂₅/A₂₈₀ of 0.07–0.13 (Figure S4). While there was variability in copper content, the peak position and shape of the LMCT band as well as the characteristic azurin-derived Cu²⁺ EPR signals⁴⁹ are nearly identical among this suite of azurin mutants (Figure S5). Thus, despite introducing altered polarity in the central cavity of azurin, the Cu²⁺ coordination site of the protein remains intact.

Structures of Azurin

Circular dichroism (CD) spectroscopy was performed to confirm the azurin β sheet secondary structure (Figures S6, S2C). Azurin variants exhibited the typical CD trace for a predominantly β sheet secondary structure, with only subtle shifts in the CD spectrum among the AzF110X variants.

Cryogenic X-ray structures of the various AzF110X variants are presented in Figure 3. The AzW48 and AzF110S structures have been previously determined (Figure 3A,B, respectively).^26,44 Compared to AzW48, the substitution of the phenylalanine to serine, F110S, introduced water molecules into the cavity originally occupied by Phe (Figure 3B).²⁶ In addition, we present high-resolution structural models obtained for the AzF110L (1.25 Å, Figure 3C) and AzF110A (1.15 Å, Figure 3D) variants. The global structures of all proteins are virtually superimposable with RMSD values of 0.38–0.43 Å with respect to the parent AzW48. The structures of AzF110L and AzF110A (Figure 3C,D) also show infiltration of water molecules (blue spheres) into the azurin β barrel cavity, with one (occupancy: 60%) and three (occupancy: 100, 60, and 90%) waters, respectively. The published structure of AzF110S shows two-to-four water molecules in the β barrel cavity, depending on the chain. The average B-factors of the Trp side chain correlate with the number of waters in the cavity, from 5.8 Ų in AzW48 to 10.4 Ų for AzF110S (Table S3). This tracks global RMSD values. Furthermore, the reduction of the side chain bulk at Phe110 to smaller aliphatic residues (Ala or Leu) is consistent with the influx of water into the hydrophobic cavity as well as the observed red-shifted fluorescence maximum for AzF110A.

Figure 3.

X-ray structures of the environment surrounding W48. The structures in panels A through D were determined at 2.14 (PDB: 1E67), 2.3 (PDB: 1ILU), 1.25 (PDB: 8F5L), and 1.15 (PDB: 8F5K) Å resolution, respectively. Crystallographically resolved water molecules inside the azurin cavity are represented by blue spheres. Relevant distances (dashed black lines) are in Å. See Table S2 for crystal data statistics. The figure was generated in PyMoL.⁵⁰

Mutational Effects on Azurin Stability

To rule out artificial shifts in λ_em from protein instability/unfolding due to the influx of water in the hydrophobic cavity, we examined the protein folding thermodynamics using differential scanning calorimetry (DSC). Representative DSC thermograms are presented in Figure S7. The DSC derived melting temperature (T_m) and enthalpy of folding (ΔH°) values for AzW48 (Table 1) are consistent with previous literature values.⁵¹ AzF110A and AzF110S exhibit lowered T_m values compared to AzW48 caused by the infiltration of water in the cavity. Their folding enthalpy values, related to the overall protein stability, were modestly reduced from AzW48 (ΔΔH° = 72–88 kJ/mol) and within the range for other azurin mutants that maintain copper-mediated ET reactivity.⁵² These results are also consistent with the trends found from mutational studies on the model protein lysozyme in which reducing the side chain bulk of interior aliphatic residues led to decreases in ΔH°.⁵³ In contrast, an extreme case is the azurin variant AzF110L, which despite a moderate T_m value of 57–59 °C, the folding enthalpy is greatly reduced from 450 ± 15 kJ/mol in AzW48 to 58 ± 2 kJ/mol in AzF110L. Upon thawing AzF110L samples stored at −80 °C for a few months, the protein absorption and CD spectra indicated loss of structure (Figure S2), while other stored azurin variants retained spectra and folding thermodynamics comparable to fresh samples. These data further support that the AzF110L structure is less stable than the other azurin variants. We propose that this destabilization could be attributed to the penetration of high-energy or “frustrated” water molecules in the interior of the protein.^54,55 Because of the lowered structural stability of the AzF110L variant, it was not studied further for its electrochemical or spectroscopic properties.

Table 1. Thermodynamics of Azurin Folding Stability by DSC.

	Tm (°C)	ΔH° (kJ/mol)
AzW48	78.1 ± 0.5	450 ± 15
AzF110S	61.3 ± 0.3	378 ± 15
AzF110A	61.2 ± 0.1	362 ± 26
AzF110L	59.4 ± 0.2	58 ± 2
AzW108	70.1 ± 0.3	392 ± 31

Electrochemical Analysis of Azurin Variants

The impact of these cavity altering mutations, with infiltrated waters, on the tryptophan electrochemical potentials was measured by SWV using LbL immobilization, as described for AzW48 and AzW108.²³ It was previously shown that AzW108, which is located proximal to the surface loops and forms hydrogen bonds with solvent, exhibits a ∼100 mV increase in the net electrochemical potential (E_net) relative to W48. Here, we expected to see a similar trend with increases in E_net for AzF110A and AzF110S, based on the red-shifted fluorescence maxima and the X-ray resolved waters in the azurin cavity. Note that the previous electrochemical characterization of AzW48 and AzW108 was performed with protein samples in which the metal was reconstituted with Zn²⁺. Efforts were made to reconstitute AzF110S and AzF110A to purely Zn²⁺ forms. However, DSC thermograms of reconstituted AzF110S showed a significantly reduced enthalpy of folding (Table S4). The Zn²⁺-reconstituted AzF110S sample also exhibited an altered UV–visible absorption spectrum (Figure S8), consistent with protein unfolding. The reason for the destabilization upon incorporation of Zn²⁺ is not clear, though we hypothesize that AzF110S did not refold properly after the dialysis procedure with cyanide. This behavior contrasts that of AzW48, which retains its well-defined absorption features and folding enthalpy following metal reconstitution. Note that we observed no significant difference in protein stability between Cu- and Zn-forms of AzW48 (Table S4), though the apo-AzW48 is seen to exhibit a greatly reduced T_m (Δ = −22.2 °C) and enthalpy (ΔΔH° = −160 kJ/mol) relative to those of the as-isolated form. The latter is consistent with recent MD simulations that predicted a destabilized protein structure for the metal-deficient azurin.⁵⁶ Importantly, the peak potentials of Zn²⁺-substituted and as-isolated AzW48 samples are nearly identical, though the as-isolated sample produced slightly reduced currents.²³ Thus, the electrochemical potentials for the new suite of Az variants are reported for the as-isolated samples.

LbL-SWV was performed at pH 7.0 for AzF110S and AzF110A. The peak potentials (E_net) from the background-subtracted square wave voltammograms for Trp in these azurin variants are listed in Table S5. AzF110A did not produce a clear electrochemical response for the tryptophan at pH 7.0. At these elevated potentials of ∼1 V, large background currents can arise from water reactions at the electrode surface and may overlap with or obscure the Faradaic responses from desired Trp redox centers. To reinforce the electrochemical trends, SWV was also performed at pH 9.0, at which the Trp potentials will decrease due to Nernstian behavior. The E_net values changed by ∼62 mV/pH (Figure 4, Table S5), in accordance with the expected 59 mV/pH change associated with proton release upon Trp oxidation. Voltammograms were collected with a constant square wave frequency, f, of 200 Hz, at which we previously found the intensities of the forward and reverse voltammogram to be equal and the peak potentials for the forward and reverse voltammograms matched.²³ Under these conditions, the electrochemical responses are considered reversible and E_net will approximate E°′.¹⁷ With our setup, the noise and background signal dwarfed any Faradaic current originating from the azurin tryptophan redox center at f ≥ 400 Hz.

Figure 4.

Relationship between electrochemical potential and peak emission from fluorescence. Data are listed in Table 2 and Table S5. Top values are pH 7.0 (orange dashed) and pH 9.0 (black dashed); the dashed lines are linear fits to the data. The values for α4W were obtained from ref (57) and not included in the linear fit for pH 7.0.

The E_net values for AzF110S and AzF110A (pH 9.0) were found to lie between the potentials reported for AzW48 and AzW108. The data are surprisingly well correlated, with a linear relationship between the potential and peak emission from fluorescence with the latter reporting on the combination of local polarity and hydrogen bonding for tryptophan side chains embedded within a protein (Figure 4). We also found that a previously characterized, single-tryptophan-containing four-helix protein, α4W,⁵⁷ also tracks well with the electrochemical potential–emission maximum correlation with our azurin variants, despite having a very different protein scaffolds.

A 12 amino acid peptide (RWVEVNGOKIFQ) that folds into a well-defined β-hairpin was also studied. In the predicted structure (Figure S9), tryptophan (β-W2) forms a cross-strand, staggered π stacking interaction with a phenylalanine (F11). This peptide was designed to mimic the staggered π stacking between W48 and F110 in azurin (Figure 1A). The electrochemical potential of the β-W2 was found to be 989 ± 9 mV at pH 7.0;²⁸ this result is comparable to the potential of the tryptophan model, N-acetyl-l-tryptophan ethyl ester, NATE, E_net = 983 mV, suggesting that the solvation surrounding the tryptophan impacts the potential. Similarly, a series of tyrosine-containing β-hairpins exhibited electrochemical potentials comparable to that of the side chain alone. However, thermodynamic properties of tyrosine in these peptides, such as pK_a, were influenced by varying cross-strand interactions involving hydrogen bond acceptors (e.g., His) or cation−π interactions with arginine.^58,59 These small peptide studies further underscore the importance of shielding amino acids from solvent to modulate electrochemical potentials.

Ultraviolet Resonance Raman (UVRR) Spectroscopic Characterization of Tryptophan Microenvironments

We next turned to UVRR spectroscopy to provide more detailed insights into the nature of the local environment and characterize the hydrogen bonding and polarity surrounding the tryptophan in azurin. UVRR spectra were collected with 228 nm excitation; this wavelength is resonant with the B_b transition of tryptophan, and thus, the vibrational features of this aromatic amino acid are enhanced with minimal signal from other molecules or solvent. Because the azurin variants are tyrosine deficient, the UVRR data primarily reflect tryptophan. A UVRR spectrum of a Trp-null protein (W48F/Y72F/Y108F) was also collected to subtract contributions from Phe so that the resulting double difference UVRR spectra of azurin only showcase Trp-derived vibrational bands, with no contribution of buffer or Phe. The UVRR spectra collected at 228 nm for AzW48, AzW108, AzF110S/A, α3W32, and β-W2 are presented in Figure S10; the vibrational features of the closed shell tryptophan residues of AzW48 and AzW108 have been assigned previously.³⁵ The majority of the vibrational bands for these azurin variants exhibited similar frequencies, indicating that the cavity mutations at position 110 do not significantly affect the protein structure. To further validate the similarity in the local environment, the intensity ratios of the Fermi doublet (RFD), calculated as I_∼1360/I₁₃₃₅, are in the range of 1.81–1.85 for the AzF110X variants (Table 2). Values in excess of 1.7 are consistent with tryptophan side chains in nonpolar environments, whereas smaller RFDs of 1.1 are a marker of polar environments.⁴² Thus, the RFDs confirm that the proteins are well folded and that the Trp side chain in the AzF110X azurin variants are buried in a nonpolar environment.

Table 2. Summary of UVRR Data for Azurin Tryptophan Variants and Relationship to λ_em. Number of trials for UVRR experiments is indicated in parentheses and if appropriate, average values are reported.

	Ratio FD	ω9 (cm–1)	ω10 (cm–1)	ω17 (cm–1)	λem (nm)
AzW48 (n = 3)	1.81	1247	1232	873	308
AzF110S (n = 2)	1.83	1248	1233	876	322
AzF110A (n = 3)	1.85	1248	1233	875	322
AzW108 (n = 1)	1.55	1260	1238	876	335
α3W32 (n = 3)	2.39	1253	1236	878	325
β-W2 (n = 1)	1.13	1256	1236	877	355

There are notable frequency shifts in the ω17 mode around 870–880 cm^–1 (Table 2, Figure 5). This vibrational mode corresponds to a planar ring deformation with displacement of the N–H moiety (Figure S11), and its frequency has been historically considered a spectral marker for hydrogen bonding.⁶⁰ The frequencies of the ω17 mode for AzW48 and AzW108 were measured to be 873 and 876 cm^–1, respectively. The frequencies of the ω17 mode for the AzF110X azurin variants lie in between AzW48 and AzW108 (Table 2).

Figure 5.

Expanded view of the ω17 (left) and ω9/ω10 (right) regions of the UVRR double difference spectra for representative experiments of azurin mutants as well as peptides β-W2 and α3W32 (labeled α3W). Solid curves are experimental data, dotted curves are the results from Gaussian decompositions, and dashed curves are the sum of the Gaussians.

Previous off-resonance Raman studies of small molecules indicated that increased ω17 frequency correlates with weaker hydrogen bonding.^60,61 This interpretation implies that AzW48 is H-bonded, presumably to water, and AzW108 is not. However, this interpretation is inconsistent with several experimental results. The X-ray structure of AzW48 shows no polar side chains or water in the vicinity of this tryptophan. Furthermore, the fluorescence maximum of W48 is 308 nm, consistent with a nonpolar, non-H-bonding environment. In the case of AzW108, the fluorescence maximum of AzW108 (λ_em = 335 nm) is red-shifted relative to AzW48, indicating an increased polarity and/or hydrogen bonding microenvironment. This result is supported by our previous 100 ns MD simulations, in which cluster analysis of the entire trajectory generated a model predicting a hydrogen bond between the N–H of W108 and a water molecule proximal to the surface.²³ The ω17 frequency for W108 is also comparable to that observed for the solvent-exposed Trp in the β-W2 peptide, 877 cm^–1. Azurin F110 mutations to Ala and Ser resulted in X-ray structures with water-filled cavities and moderately red-shifted λ_em values (322 nm).

The prior off-resonance Raman report that contradicts the current interpretation focused on the relationship between the ω17 mode with the N–H stretch (and hence, H-bonding) for crystalline and solid model indole compounds.⁶¹ To further investigate the expected correlation, we carried out density functional theory (DFT) calculations. We calculated the vibrational frequencies for 3-ethylindole (3EI) in vacuum and in three implicit solvents (cyclohexane, methanol, and water). Furthermore, systems with an additional explicit H₂O molecule were also investigated. The DFT results are presented in Figure 6. In the absence of explicit water, the frequencies of the ω17 mode and N–H stretch for 3EI are correlated, with the trend: vacuum > cyclohexane > methanol/water. This trend indicates that the frequency of the ω17 mode is highest in low dielectric solvent (cyclohexane) and lowest in high dielectric solvent (water). The presence of an explicit H-bond acceptor (NH···OH₂) consistently resulted in a decrease of the N–H stretch frequency, as expected, but had differential effects on the ω17 mode. In polar environments (methanol and water), the ω17 frequency also decreased, which is consistent with the N–H stretch and the previous off-resonance Raman study. However, in nonpolar environments (vacuum and cyclohexane), the ω17 frequency increased (vacuum) or stayed roughly the same (cyclohexane). In all calculations, the magnitude of shifts for the ω17 mode was modest, and within 3 cm^–1. This modest range contrasts with the previous experimental range⁶¹ of ∼13 cm^–1 for the crystals and ∼5 cm^–1 for the solvents, though the calculated range of N–H stretch frequencies (Δν ∼ 240 cm^–1) is similar to the range of experimental Δν of ∼220 cm^–1. Attempts were made to calculate the effect of an H-bond donor above the indole plane to mimic the effect of perturbation of the π cloud, and the results indicated that the ω17 frequency is complicated in terms of H-bond donors and acceptors. Thus, we cannot rule out the possibility that the presence of the F110 π cloud influences the observed peak position of the ω17 mode in AzW48. Overall, while the limited calculations presented here did not fully reproduce the experimental observations, the calculations indicate that the ω17 mode does not solely report on hydrogen bonding and that the local polarity may play an important role in the frequency as well.

Figure 6.

Frequencies calculated for ω17 mode and N–H stretch for closed-shell 3EI in different solvents and with explicit water (triangles) or without explicit water (circles). Inset shows the equilibrium structure of 3EI with one explicit water molecule.

Based on these results, we propose that the inclusion of water molecules near tryptophan in the nominally solvent excluded environment of azurin increases the ω17 frequency, consistent with the calculations, but contrary to the trend from the earlier model study.⁶¹ We can speculate on the origin of the discrepancy of the earlier work with the spectroscopic and structural data presented here. The earlier work focused on four crystals whose local polarity is not known. Furthermore, comparisons to dynamic solution structures may not be straightforward. The earlier work also studied skatole in the three solvents of cyclohexane, CS₂, and CS₂/dioxane mixtures. Cyclohexane and CS₂ are nonpolar solvents, while dioxane is more polar; the fluorescence maxima of skatole in cyclohexane and dioxane are 307 and 335 nm, respectively.⁴¹ Consistent with this prior report, a separate Raman study showed that the addition of an indole hydrogen bond acceptor, such as HMPA, in cyclohexane, downshifted the vibrational frequency of ω17, from 882 to 876 cm^–1.⁶² As shown in Figure 2, the addition of HMPA also shifted the fluorescence maximum of the tryptophan side chain in a similar fashion as that reported for the cyclohexane/dioxane mixtures. These prior reports of downshifted ω17 frequencies in the presence of H-bond acceptors may reflect the combination of an increased local polarity and enhanced H-bonding upon addition of dioxane or HMPA in cyclohexane; if evaluated in isolation, the transition to a more polar environment and addition of a H-bond acceptor in a nonpolar environment would have opposing effects on the ω17 frequency based on the present calculations. Thus, it is questionable to assign the previous trends in ω17 frequency solely to H-bonding.

This interpretation of an increase in ω17 frequency upon addition of 1–2 water molecules in a hydrophobic cavity is further substantiated by the pattern of the ω9 and ω10 modes. AzW48 and the AzF110X variants showed an intense ω9 peak at 1247 (AzW48) or 1248 (AzF110A and AzF110S) cm^–1. This mode is accompanied by a weak intensity for the ω10 peak centered at 1232 (AzW48) or 1233 (AzF110A and AzF110S) cm^–1. Based on previous small molecule UVRR studies, this pattern of the ω9/ω10 region is characteristic of a nonpolar environment and similar to that of skatole in cyclohexane (Figure S12). The slight, average 1 cm^–1 upshift in the ω9 and ω10 frequencies for both AzF110A and F110S compared to AzW48 could reflect a subtle change in the polarity of the cavity from the water molecule(s). In contrast to the AzW48 and the AzF110X mutants, AzW108, which is located more proximal to the protein surface, shows a different spectral pattern in the ω9/ω10 region. The frequencies of ω9 and ω10 are upshifted by 13 and 5 cm^–1 to 1260 and 1238 cm^–1, respectively. Additionally, the intensity pattern shifts such that the ω9 band is reduced while the ω10 peak is enhanced. This trend is similar to that of β-W2. These features are consistent with H-bonded W108 in a polar environment. While polar, the environment of AzW108 is distinct from the completely solvated β-W2 peptide. First, the UV absorption and fluorescence spectra of W108 are indicative of a partially shielded tryptophan (Figure 2). Second, there are subtle differences in the line shapes of the ω9/ω10 region (Figure S12), in which β-W2 matches l-Trp in water, and AzW108 matches skatole in dioxane. Finally, the ratio of the Fermi doublet (FD) of 1.55 for W108 (Table 2) lies in between the value for the hydrophobic environment of AzW48 (1.8) and the solvated β-W2 or l-Trp (1.1). Thus, the tryptophan in AzW108 can be described with a hydrogen-bonding interaction with water in a polar, yet mostly solvent excluded environment.

Overall, the UVRR results of RFD, ω17, and ω9/ω10 support hydrophobic environments for AzW48 and the AzF110X mutants, with evidence of H-bonding with the waters in the cavity for the AzF110X mutants. The presence of water in the F110X mutants has implications for the PCET reaction for W48. Upon UV excitation, W48 undergoes ET followed by PT to generate the tryptophan neutral radical that has a decay time of 7 h.^46,56 The proton acceptor has not yet been identified, and given the lack of evidence of water in the hydrophobic W48 cavity, it has been hypothesized that the proton acceptor is the nearby F110 residue.³⁵ However, protein fluctuations that permit transient interactions of W48 with water cannot be excluded.⁵⁶ For the F110X mutants, the presence of an aqueous H-bond suggests that a water molecule in the W48 cavity could serve as the proton acceptor. In contrast to the AzF110X mutants, the UVRR features of AzW108 and β-W2 support more prominent H-bonding and polar environments. These vibrational results correlate with the trends from fluorescence spectroscopy and support the interpretation that the electrochemical potentials of protein-embedded tryptophan side chains are influenced by H-bonding.

These bound waters in the AzF110X variants also lead to altered conformational stability of the protein that will, in turn, influence the local protein motions. The latter is further suggested by the increase in the average B-factors of the tryptophan side chain in the X-ray structures of AzF110A and AzF110S, relative to AzW48 (Table S3). These protein motions likely play a role in tryptophan oxidation, such as permitting exchange of the cavity water(s) with bulk solvent upon protonation of the cavity water. Further, protein motions may influence the radical stability and electrochemical properties of W48. However, we expect the main contributor to the trends in the reduction potentials to arise from the local electrostatics. This is supported by the properties of the AzW108 congener, which exhibits the greatest shift in the Trp electrochemical potential, though its protein folding thermodynamics are more comparable to AzW48.

Comparison to a Protein Mimetic

This analysis extends beyond azurin. For example, a de novo, triple helix bundle was designed with a single tryptophan inserted in the interior of the polypeptide, α3W32. Previous electrochemical studies reported a very high reduction potential for this tryptophan.¹⁸ With the present LbL-SWV method, the E_net for α3W32 at pH 7.0 was determined to be 1134 ± 16 mV, which is in good agreement with the reported value of 1095 ± 4 mV acquired with protein thin films.¹⁸ This high potential was proposed to arise from the exclusion of the side chain from water.⁶³ Herein, the UVRR spectrum of α3W32 exhibits an RFD of 2.39, consistent with the interpretation of a buried tryptophan side chain in a nonpolar environment. However, the red-shifted λ_em of 325 nm for this tryptophan (Figure S13A) indicates a different environment than described above for AzW48. The different environment is further corroborated by the ω9/ω10 region of the UVRR spectrum for α3W32, which has a similar set of frequencies as seen for β-W2, but with a distinct intensity pattern. The ω10 band shows a growth of the 1236 cm^–1 peak with a loss of intensity of ω9 at ca. 1250 cm^–1 (Figure 5), which supports an H-bond in a nonpolar environment that agrees well with both the properties of the RFD and λ_em. This ω9/ω10 pattern is similar to that of skatole in cyclohexane with a low concentration of dioxane (Figure S13). Thus, while the elevated potential in α3W32 was originally considered to arise from solvent exclusion alone, the UVRR and fluorescence data support that this Trp, embedded in a nonpolar protein environment, participates in hydrogen bonding. Inspection of the static NMR structure of α3W32 shows that the tryptophan is mostly sequestered from direct interaction with solvent, with a low solvent-accessible surface area (2–6%).⁶³ This low solvent-accessibility raises the question of the nature of the proton acceptor. A previous kinetic study of the oxidation of the tyrosine in α3Y32 suggested that water is the primary proton acceptor.²⁰ Based on 1 μs MD simulations, Y32 is predicted to form transient interactions with water through dynamic fluctuations of the α3 scaffold.²⁰ These interactions may also be mediated by a glutamic acid residue (E13). Further, quantum calculations used to predict the redox properties of Y32 suggested that dynamic interactions with this charged residue and water may play a significant role in the observed increase in electrochemical potentials, relative to tyrosine solutions.^18,64 In comparison with this model for α3Y32, we previously predicted a direct interaction (2.7 Å) between E13 and W32 (Figure 13B).²³ Future mutational studies may help to shed light onto the role of these side chain interactions toward the thermodynamic and electrochemical properties of α3W32 and α3Y32.

Conclusion

This work presents an electrochemical and structural study of a series of azurin variants that have been shown to adopt varying degrees of water influx into this well-structured protein. Water occupancy in the azurin cavity created by substitution of smaller side chains at position 110 was confirmed by high-resolution X-ray structures. The azurin series provides insight into the depth and scope to which the local polarity and hydrogen bonding may influence the redox properties of tryptophan and, by extension, the driving force for Trp-mediated electron transfer. The description of water’s role in hydrogen bonding versus polarity in azurin was further assessed by UV resonance Raman spectroscopy. With a combination of the results from X-ray structures of azurin variants and steady-state fluorescence along with the differential behavior of the ω9/ω10 Raman modes, we offer a new interpretation for the ω17 mode. The frequency of ω17 is sensitive to both hydrogen bonding and local polarity, whereas the frequencies and relative intensities of the ω9/ω10 modes report on hydrogen-bonding strength in different environments. The assignment of the degree of polarity and H-bonding for each azurin variant aids in the understanding of the impact of water on the thermodynamics on Trp oxidation. In AzF110X variants, the effect of water ingress in a hydrophobic cavity on the electrochemical potentials of protein-encapsulated tryptophan residues is modest, varying by 40–75 mV, relative to AzW48. For AzW108, the potential is even more elevated relative to AzF110X; based on the UVRR data, W108 forms a hydrogen bond in a polar, yet semishielded environment. More drastic impacts on the potential can arise from the inclusion of hydrogen-bonding interactions that may involve charged side chains with tryptophan in low dielectric environments, such those reported by the de novo α3Y32 or α3W32 biomimetic proteins.^17,18,57 The cumulative results offer new structural intuition into how the redox properties of tryptophan in proteins can be modulated by hydrogen bonding.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jpcb.2c06677.

• CV, UV–vis, fluorescence, and CD characterization of azurin; UVRR spectra of azurin and skatole (PDF)

Author Contributions

^∇ K.T. and C.B.T. contributed equally.

The authors declare no competing financial interest.