Co(II) Substitution Enhances the Esterase Activity of a de Novo Designed Zn(II) Carbonic Anhydrase

Abstract

Carbonic Anhydrases (CAs) have been a target for de Novo protein designers due to the simplicity of the active site and rapid rate of the reaction. The first reported mimic contained a Zn(II) bound to three histidine imidazole nitrogens and an exogenous water molecule, hence closely mimicking the native enzymes’ first coordination sphere. Co(II) has served as an alternative metal to interrogate CAs due to its d⁷ electronic configuration for more detailed solution characterization. We present here the Co(II) substituted [Co(II)(H₂O/OH⁻)]_N(TRIL2WL23H)₃ⁿ⁺ that behaves similarly to native Co(II) substituted human-CAs. Like the Zn(II) analogue, the cobalt-derivative at slightly basic pH is incapable of hydrolyzing p-nitrophenylacetate (pNPA); however, as the pH is increased a significant activity develops, which at pH values above 10 eventually yields a catalytic efficiency that exceeds that of the [Zn(II)(OH⁻)]_N(TRIL2WL23H)₃⁺ peptide complex. X-ray absorption analysis is consistent with an octahedral species at pH 7.5 that converts to a 5-coordinate species by pH 11. UV-vis spectroscopy can monitor this transition, giving a pK_a for the conversion of 10.3. We assign this conversion to the formation of a 5-coordinate Co(II)(N_imid)₃(OH)(H₂O) species. The pH dependent kinetic analysis indicates the maximal rate (k_cat), and thus the catalytic efficiency (k_cat/K_m), follow the same pH profile as the spectroscopic conversion to the pentacoordinate species. This correlation suggests that the chemically irreversible ester hydrolysis corresponds to the rate determining process.

The Co(II) analogue of the designed peptide Zn(II)(TRIL23H)₃ is made as a spectroscopic probe for the carbonic anhydrase activity exhibited by this de novo designed system. The coordination geometry of Co(II) and Zn(II) are vastly different at high and low pH. Nonetheless, the Co(II) analogue at high pH exceeds the catalytic efficiency of the parent Zn(II) protein.

Graphical Abstract

Introduction

The design of artificial metalloproteins has been a productive area of research.^[1] Within this rubric de Novo protein design has been shown to be a powerful strategy for mimicking biologically relevant metal sites, recreating their function and understanding the relationship of the protein structure to the bound metal ion’s activity.^[2] The de Novo design approach allows an investigator to employ a novel sequence in order to define a desired protein structure and then probe whether a metal binding site can be incorporated within this minimal unit. One may then assess whether a well-folded protein scaffold is obtained and whether it can carry out a desired function (e.g., templating a specific fold, carrying out electron transfer reactions or completing either redox or non-redox chemical transformations).^{[2f, 3]} Although the question of how the specific location of a metal site along the primary sequence impacts the complex activity has been addressed,^[4] the important question about how similar, but different, metal ions affect reactivity has not been probed.

Studies in our research group focus on the metalation of three stranded coiled coils (3SCCs). Previously, we reported the structure and hydrolytic activity of [Zn(II)(H₂O/OH⁻)]_N(TRIL2WL23H)₃ⁿ⁺ (see below in Table 3) toward p-nitrophenyl acetate (pNPA, k_cat/K_m = 23.3 M⁻¹ s ⁻¹ at pH 9.5) and CO₂ (k_cat/K_m = 1.8 × 10⁵ M⁻¹ s ⁻¹ at pH 9.5).^{[3e, 4–5]} At the time, this model was the fastest de Novo designed metalloenzyme relative to a native enzyme under physiological conditions, only 1500-fold less active for pNPA hydrolysis than the fastest carbonic anhydrase (CA). While other designed systems have since been reported that exceed these values, TRIL2WL23H peptide still holds the record with respect to the native CO₂ hydration chemistry (being within a factor of several hundred fold).^[6] X-ray crystallographic analysis showed that the Zn(II)His₃(H₂O) first coordination sphere of ZnCA was recapitulated in this 3SCC, despite the fact that the native enzyme contains significant β-sheet structure whereas the TRI proteins are fully α-helical. These studies demonstrated that the Zn(II) first coordination sphere alone was enough to confer significant hydrolytic activity and retain the catalytic efficiency (k_cat/K_M).^[4]

Table 3.

pH-Dependent Kinetic Parameters for pNPA hydrolysis by Co(II) and Zn(II)-Bound TRIL2WL23H Peptides.^[a]

Peptide complex	pH[b]	Kcat (s−1) (x 10–3)	KM (mM)	Kcat/KM (M−1 s−1)
[Co(II)(H2O/OH−)]N(TRIL2WL23H)3n+	8.0	2.4 ± 0.5	1.9 ± 0.5	1.3 ± 0.1
	8.75	6.9 ± 0.5	1.5 ± 0.3	4.6 ± 1.7
	9.5	28.6 ± 0.3	2.0 ± 0.3	14.4 ± 1.0
	10.0	49.5 ± 0.8	2.0 ± 0.5	24.5 ± 1.6
	10.5	73.4 ± 0.2	2.6 ± 0.1	28.7 ± 2.0
	10.7	95.4 ± 0.3	2.4 ± 0.2	38.8 ± 1.5
[Zn(II)(H2O/OH−)]N(TRIL2WL23H)3n+ [c]	7.5	1.1 ± 0.2	2.7 ± 0.8	0.41 ± 0.03
	8.0	2.9 ± 0.5	2.7 ± 0.6	1.07 ± 0.06
	8.5	6.0 ± 0.7	1.8 ± 0.3	3.3 ± 0.2
	9.0	16 ± 1	1.8 ± 0.2	8.9 ± 0.4
	9.25	22 ± 1	1.7 ± 0.2	12.9 ± 0.4
	9.5	33 ± 2	2.1 ± 0.2	15.5 ± 0.4

^[a]Error bars result from fitting all individual initial rates measured (three per concentration of substrate, without averaging) to the Michaelis–Menten equation in Prism 8 (GraphPad Software). 5 μM active Co(II)-bound peptide complex.

^[b]pH 8.0 measured in 50 mM HEPES; pH 8.75 – 9.5 measured in 50 mM CHES. pH 10.0 – 10.7 measured in 50 mM CAPS at 298K.

^[c]Previously reported data of Zn(II)-bound peptide complex included for comparison.^[4]

While kinetically well characterized, the main limitation of this system is the fact that Zn(II) is spectroscopically silent. This characteristic makes solution structural determination of the Zn(II)-bound peptide extremely difficult. For example, there is no convenient way to assess whether the catalytically determined pK_a for the enhanced activity at high pH correlates with the hydrolysis of a metal bound water to form the active Zn(II)His₃(OH) species. Metals such as Co(II) and Cd(II) have often been substituted for Zn(II) in many zinc proteins and can act as powerful spectroscopic probes in biological systems. Fortunately, there have been many studies showing that the Co(II) ion may be used as an alternative metal in CA.^[7] This metal substitution could allow for the determination of a spectroscopic pK_a for hydrolysis that may potentially be correlated to the kinetic pK_a observed for Zn(II). Although substitution with high-spin Co(II) is a common approach for zinc enzymes (given that significant activity is often retained and the site can be examined using UV-Vis or EPR spectroscopies),^{[7a, 8]} the two ions often display different first coordination sphere geometries in proteins.^{[7a, 9]}

For Co(II)-substituted CA, a pH-dependent spectroscopic change in the UV-Vis absorption spectrum is observed and correlates well with the kinetic pK_a for Zn(II)-catalyzed hydrolysis, representing a transition from metal-bound water to an active metal-bound hydroxide species.^[10] In order to assess whether the kinetically derived pK_a for Zn(II)(TRIL2WL23H)₃ was consistent with metal hydrolysis, we incorporated Co(II) into this designed scaffold. We show here that while Co(II)(TRIL2WL23H)₃ is not isostructural with the Zn(II) complex, p-nitrophenyl acetate esterase activity is enhanced, albeit at slightly more basic conditions.

The peptide sequences used in the present work are reported in Table 1.

Table 1.

Peptide sequences.

Peptide[a]	Sequence (N→C terminus)
TRIL2W	Ac-G WKALEEK LKALEEK LKALEKK LKALEEK G –NH2
TRIL2WL23H	Ac-G WKALEEK LKALEEK LKALEKK HKALEEK G –NH2

^[a]N- and C-termini are capped by acetyl (Ac) and NH₂ groups, respectively.

Experimental Section

Peptide Synthesis and Purification.

Peptides were synthesized using standard Fmoc chemistry solid-phase peptide synthesis on a rink amide methylbenzhydrylamine (MBHA) resin with a Biotage Initiator+ Alstra peptide synthesizer using standard protocols.^[11] The crude peptide was simultaneously cleaved and deprotected using a cleavage cocktail consisting of trifluoroacetic acid (TFA), ethanedithiol (EDT), anisole, and thioanisole (90:3:2:5). The peptide was precipitated with ice-cold diethyl ether, recovered by vacuum filtration, dissolved in 50% acetonitrile in doubly distilled water (ddH2O), and lyophilized to dryness. The lyophilized crude product was then dissolved in 10% acetic acid in ddH2O and loaded onto a 300 mm × 50 mm DeltaPAK preparative C18 column using a Waters 600 HPLC system with the following parameters: flow rate: 10 mL/min, solvent A: 0.1% TFA in water, solvent B: 0.1% TFA in (90:10) acetonitrile:water, and linear gradient: (70:30) A:B increasing to 20:80 over 35 min.^[12] Purity and identity of the peptide were confirmed by electrospray ionization-mass spectrometry (ESI-MS) (Agilent Q-TOF), and fractions containing pure peptide were combined and lyophilized to yield a pure, white powder. The concentration of peptide was determined based on the tryptophan absorbance at 280 nm using ε = 5500 cm⁻¹ M⁻¹.

Circular Dichroism (CD) and Ultraviolet–Visible (UV–Vis) Spectroscopies.

CD and UV–Vis spectra were recorded in quartz cuvettes at 25 °C on a Jasco J-1500 CD-Spectrometer and a Cary 100 Bio UV–Vis spectrometer, respectively.

The apparent binding constant at pH 7.8 (50 mM HEPES with 0.1 M Na₂SO₄) was determined by direct titration of CoSO₄ into a peptide solution. The titration was performed at room temperature in a 1 cm quartz cuvette. Co(II) binding at pH 10.8 (50 mM CAPS with 0.1 M Na₂SO₄) was evaluated in the same way, but in the absence of oxygen to prevent oxidation of the metal, oxidation of the peptide or Co(OH)₂ formation, which can occur at high pH. The Co(II) concentration (CoSO₄) was determined using ICP-MS (Perkin-Elmer Nexion 2000). Aliquots of cobalt were added, and the UV–visible absorption spectra were recorded after 10–15 min to ensure the binding reaction was complete. For all titrations, difference spectra were obtained by subtracting the background spectrum of the control peptide (TRIL2W) under identical conditions, to ensure that any changes in the measured absorption were due to metal interaction with the His₃ binding site. The titration at pH 7.8 was analyzed using non-linear least square fits (in Prism 8 (GraphPad Software)) of the absorbance at λ_max = 517 nm as a function of the concentration of titrant (CoSO₄) added. Titration at pH 10.8 could not be completed due to Co(OH)₂ precipitation above 1.2 equivalent of Co(II) per trimer. At pH 10.8, two species contribute to the Co(II) titration absorbance. The species which prevails at low pH (λ_max = 517 nm) and the species which prevails at higher pH (λ_max = 580 and 638 nm) are both present, hence a sigmoid pattern is obtained.

The increase in pH of (TRIL2WL23H)₃ and 0.8 equiv. of CoSO₄ (to ensure complete Co(II) binding to the peptide) at room temperature was monitored by UV-Vis spectroscopy in a 1 cm quartz cuvette. The titration was initiated at pH 7.3 and small increments of concentrated NaOH were added up to pH 11.88. At least 10 minutes equilibration time was allowed between additions to ensure that the equilibrium had been reached. The titration was analyzed using non-linear least square fits (in Origin Pro Software) of the absorbance at λ_max = 638 nm as a function of the pH.

This solution was back-titrated to low pH with H₂SO₄ to assess whether the spectral changes were reversible. Starting the titration from the endpoint of the direct titration (pH 11.88) increments of concentrated H₂SO₄ were added up to pH 7.3. At least 10 minutes equilibration time was allowed between additions to ensure that equilibrium had been reached.

The same pH titration was performed under an inert atmosphere to ensure that the Co(II) initially bound to (TRIL2WL23H)₃ was not oxidized to Co(III). The titration was initiated at pH 7.0 and the pH was raised to pH 10.7 by addition of concentrated NaOH. As reported for the aerobic titration, a back-titration with concentrated H₂SO₄ was performed to assess the titration reversibility.

To test the oxidation state of cobalt bound to the peptide at lower pH further, 10 equivalents per trimer peptide of ascorbate were added to a sample of cobalt-peptide at pH 8.0 and UV-Visible spectra were recorded. The sample of metal-peptide complex was prepared as previously described for the apparent binding constant.

X-ray Absorption Spectroscopy.

Samples with final concentrations of 1.5 mM (TRIL2WL23H)₃, 1 mM CoSO₄, 50 mM buffer (HEPES for pH 8.0, CHES for pH 8.75–9.5, and CAPS for pH 10.0–10.8), and 50% glycerol were frozen and transported in liquid nitrogen. The samples were initially prepared at pH 7.5 and then pH-adjusted. In this way, formation and precipitation of Co(OH)₂ or Co(OH)₃ were prevented. After 15 minutes equilibration time the samples were frozen. All XAS measurements were performed at the beamline 9–3 at Stanford Synchrotron Radiation Lightsource (SSRL). A Si(220) double crystal monochromator was utilized, with a flat Rh-coated vertically collimating harmonic rejection monochromator. The X-ray energy was calibrated by collecting the absorption spectrum of a Co foil as a reference, placed between two ionization chambers situated behind the sample, at the same time as the fluorescence data were collected, with the first inflection point of the foil assigned as 7709 eV. The samples were maintained at a temperature of ~10 K during data acquisition, and the data were measured in a fluorescence mode with a 100-elements energy-resolving Ge detector; based on the visual inspection of each channel, 97 channels were used for data analysis. The data were collected using 0.25 eV steps in the XANES region (1 s integration time) and 0.05 Å⁻¹ increments for the extended X-ray absorption fine structure (EXAFS) region up to k = 13.5 Å⁻¹, with an integration time of 1–20 s (k³ weighted) in the EXAFS region for a total scan time of ~40 min. The maximum incident count rate for the channel with the highest counts was kept below 10,000 to avoid detector saturation. The EXAFS data were analyzed using the Athena/Artemis suite of programs with FEFF version 6.0.^[13] For the XANES data they were normalized for comparison using the MBACK program.^[14]

Esterase Activity Assays.

The esterase activities of Co(II)-bound peptides were determined spectrophotometrically with p-nitrophenyl acetate (pNPA, 200–3500 μM) substrate at 25 °C using a Cary 100 Bio UV–Vis spectrometer. Measurements were made at 348 nm (ε = 5000 cm⁻¹ M⁻¹).^[15] The procedure is similar to that which was previously described.^[4–5] The substrate solution was prepared by quickly diluting a 0.1 M pNPA acetone solution into ddH₂O to a concentration of 5 mM. The procedure for measuring esterase activity was as follows: in a 0.1 cm path length quartz cuvette, buffer (50 mM, HEPES for pH 8.0, CHES for pH 8.75–9.5, and CAPS for pH 10.0–10.8), ddH₂O, Na₂SO₄ (100 mM) and metal–peptide solution were mixed. pNPA was added, mixed, and the absorbance increase recorded every 0.5 s for 5–20 min. Metal-peptide solutions contained excess peptide in order to ensure all Co(II) was bound to the peptide (trimer:Co(II) 1:0.8). The initial rates of controls containing same amount of apo-peptide were subtracted from those of Co(II)-bound peptide samples. There was some background activity from the free peptide, due to unbound His residues, which was subtracted from the observed rates.^[4–5] Initial rates determined from linear fits of the first 2–10% of the reaction were plotted as a function of pNPA concentration and fitted to the Michaelis–Menten equation in Prism 8 (GraphPad Software). The concentration of enzyme is 5 μM and is accounted for in all reported values.

Results

CD and UV-Vis of Co(II)(TRIL2WL23H)₃

The TRI family of designed peptides has been shown previously to bind a variety of metals^{[3h, 3i, 4, 16]} without change to the association of three peptides to form an α-helical 3SCC. The same behavior is observed with this peptide when Co(II) is added as shown by the essentially invariant CD spectra over a broad range of pH values (Figure S1). These CD spectra are indicative of helical peptides arranged as a coiled coil.

The addition of Co(II)SO₄ to a neutral solution (pH 7.8) of (TRIL2WL23H)₃ yields the visible spectrum shown in Figure 1. The λ_max (517 nm) and ε (43 ± 1 M⁻¹ s⁻¹) for this broad transition are similar, but distinct, from that of Co(II)SO₄ added to (TRIL2W)₃ at pH 7.8 (Figure S2). Based on this titration curve, the Co(II) binding affinity at pH 7.8 can be determined using a simple model of Co(II) + (TRIL2WL23H)₃ ⇆ Co(TRIL2WL23H)₃. A similar titration completed at pH 10.8 is more complex (Figure S3). Although Co(II) binds to the protein, as was seen at low pH; the equilibria also include the binding of the metal as a hydrolyzed form (vide infra) with competition of the corresponding cobalt oxide which precipitates under these conditions. Thus, an accurate assessment of high pH complexation cannot be accomplished.

Figure 1.

UV-Vis titration of CoSO₄ into a solution of 0.44 mM (TRIL2WL23H)₃ at pH 7.8 in 50 mM HEPES buffer, 0.1 M Na₂SO₄. The data are plotted as molar extinction coefficient (ε, determined using the peptide trimer concentration) vs wavelength. The spectrum at 1:1 Co to peptide trimer is shown in blue. The inset to the figure displays the titration curve plotted as absorbance at 517 nm (λ_max) vs equivalents of Co(II) per peptide trimer. The concentration of CoSO₄ is corrected for dilution.

The pH titration of Co(TRIL2WL23H)₃ indicates that Co(II) binding occurs at pH 5.5, with a spectrum identical to that described above for pH 7.8, and with only a slight broadening of the peak at 517 nm up to pH 9.50 (Figure 2). As the pH is increased further, the spectral features begin to include additional peaks at ~580 nm and ~630 nm. These peaks continue to grow until pH 10.8, after which Co(II) begins to precipitate. The final spectrum resembles that reported for tetrahedral Co(II) in basic solution ([Co(OH)₄]²⁻ or [Co(OH)₃(H₂O)]⁻), although the absorption energies are slightly shifted and the ratios of the extinction coefficients between the 580 nm and 630 nm absorptions are quite different.^[17] It should be noted that the pH titration is reversible if H₂SO₄ is used to lower the pH of the solution (Figure S4). A fit to these data using an equilibrium Co(H₂O)(TRIL2WL23H)₃ ⇆ Co(OH)(TRIL2WL23H)₃ + H⁺ yields a relatively high pK_a (pK_a = 10.25 ± 0.11). The spectra in Figure 2 were collected without precaution to remove dioxygen. Identical spectra are obtained if the titration is performed in an inert atmosphere box to exclude dioxygen, indicating that the Co(II) is stable under aerobic conditions (Figure S5). Furthermore, the low pH form was subjected to reduction using ascorbate at pH 8, again to test whether the low pH form was Co(III). The recorded spectra demonstrate that there is no change in absorbance at 520 nm after the addition of 10 equivalents of ascorbate (Figure S6).

Figure 2.

Top: UV-Vis spectra representing the pH titration of 0.3 mM (TRIL2WL23H)₃ + 0.24 mM CoSO₄. The data are plotted as molar extinction coefficient (ε, determined using the peptide trimer concentration) vs wavelength. pH values are color-coded. Bottom: Titration curve plotted as molar extinction coefficient at 638 nm (λ_max) vs pH fitted to determine the spectroscopic pK_a. The concentration of CoSO₄ is corrected for dilution.

X-ray Absorption Spectra

X-ray absorption near-edge structure (XANES) spectroscopy was used to characterize the coordination geometries and the oxidation states of the Co ion in the complexes over the pH range 8.0 to 10.8 (Figures 3 and S7). The pre-edge features of the Co(TRIL2WL23H)₃ were around 7710 eV, while the main absorption peak is ~7720 eV for all pH values. The energy of both the rising edge and the 1s-3d transitions are in good agreement with those of authentic Co(II) models and distinctly lower than Co(III), confirming the oxidation state assignment. The intensity change in the pre-edge feature in the Co K-edge XANES spectra associated with 1s→3d pre-edge peak for pH 8 is significantly smaller than those at higher pH 10.8 is consistent with conversion from a more centrosymmetric octahedral geometry to a lower less symmetric coordination environment. This trend is evident with the increasing in intensity of the pre-edge peak (Figure S7).

Figure 3.

XANES of Co(TRIL2WL23H)₃ as a function of pH. The spectrum at pH 8.0 is shown in orange, at pH 9.5 in blue, at pH 10.5 in red and at pH 10.8 in black.

The EXAFS (extended X-ray absorption fine structure) spectroscopy data are shown in Figure 4. As with the XANES, it is apparent that the pH 8 data is distinct from the other three samples. In all 4 cases, the EXAFS is dominated by a strong nearest-neighbor shell, with a significantly larger amplitude at pH 8. For pH 8, the data show the outer-shell scattering characteristic of imidazole scattering, while at high pH, the outer-shell peaks are weaker and broader. While this might, in principle, arise from loss of histidine ligands at higher pH, given the totality of the data, we believe that it is better explained as arising from increased disorder in the higher pH samples, consistent with a range of Co-N_His distances and a mixture of 5 and 6-coordinate sites.

Figure 4.

k³ -weighted EXAFS (top) and Fourier transforms (FTs) (bottom) of the EXAFS data for [Co(II)(H₂O/OH⁻]_N(TRIL2WL23H)₃ⁿ⁺.

The first-shell data can be modeled with a single Co-N shell, with a single Co-O shell, or with a mixture of Co-N and Co-O ligation. Since the accessible resolution of the data are not sufficient to permit independent refinement of the Co-O and Co-N shells, we adopted a procedure that minimized the number of variable parameters. Based on the known stoichiometry, we included 3 Co-N and 2 or 3 Co-O in the fit, and refined the average Co-nearest neighbor distance, the difference in Co-O and Co-N distances (ΔR), and a single Debye-Waller factor which was assumed to be the same for both the Co-N and Co-O shells. ΔE₀ was fixed at 7722 eV and S₀² at 0.9 based on fits to model compounds. The best fits for each sample is summarized in Table 2. It is important to note that identical average nearest-neighbor distances were obtained for ΔR ranging from −0.1 to +0.1 Å; all that changes is that the fit is slightly worse and the Debye-Waller factor is slightly different. We are therefore confident that this is a reliable average nearest-neighbor distance even if the individual Co-O and Co-N distances cannot be resolved. Since the pH 8.0 data shows outer-shell scattering typical of imidazole ligation, it was fit using a combination of a rigid Co-Imidazole 5-membered ring and Co-O scattering, with similar constraints. The outer-shell scattering was calculated using FEFF 6.0 ^[13] and the 14 most important multiple-scattering paths were included in the fit. The initial Debye-Waller factors for all of the imidazole paths were taken from Bunker^[18] and scaled by adding either a fitting a single static disorder parameter that was added to each path’s Debye-Waller factor, or by fitting three shell-dependent (R~2 Å, R~3.2 Å or R~4.2 Å) static disorder parameters. In no case did the average nearest-neighbor distance change, although the fits improved slightly as the number of fitted parameters increased. A representative fit is shown in the Figure S9.

Table 2.

Best EXAFS Fitting Parameters.

pH	Avg. Co-N/O [Å]	|ΔR| [Å]	σ2 x103[Å2]
8.0	2.14	0.12	8.6
9.5	2.11	0.12	5.4
10.5	2.13	0.01	10.2
10.8	2.11	0.02	9.2

pNPA Hydrolysis Kinetics

The hydrolysis of pNPA by Co(TRIL2WL23H)₃ was investigated over the pH range 8 to 10.8 using initial rate studies in order to extract Michaelis-Menten parameters. The initial rate profiles as a function of pNPA concentration are shown in Figure 5 and the catalytic parameters derived from these studies are provided in Table 3.

Figure 5.

pH dependence of the initial rate for pNPA hydrolysis by Co(II)(TRIL2WL23H)₃. Error bars result from fitting all individual initial rates measured (three per concentration of substrate, without averaging) to the Michaelis–Menten equation in Prism 8 (GraphPad Software). 5 μM active Co(II)-bound peptide complex. pH 8.0 measured in 50 mM HEPES; pH 8.75 – 9.5 measured in 50 mM CHES. pH 10.0 – 10.7 measured in 50 mM CAPS at 298K.

Clearly, there is a strong pH dependence to the hydrolytic activity. The k_cat values increase by a factor of 50 on going from pH 8 to 10.7. Similarly, there is a 30-fold increase in the catalytic efficiency, which is given by k_cat/K_M, over the same pH range. In contrast, the substrate binding parameter, K_M, is only modestly changed over the same range of hydrogen ion concentrations. These pH dependence of the three kinetic parameters (k_cat, K_M and k_cat/K_M) are provided graphically in Figure 6. The data are compared to the values predicted by using the pK_a determined spectroscopically. The pK_a determined from a fit of the kinetic parameters is shown in Figure S10.

Figure 6.

pH dependence of the kinetic parameters for pNPA hydrolysis by [Co(II)(H₂O/OH⁻](TRIL2WL23H)₃ⁿ⁺. pH dependence of k_cat (A), K_M (B) and k_cat/K_M (C). Error bars result from fitting all individual initial rates measured (three per concentration of substrate, without averaging) to the Michaelis-Menten equation in Prism 8 (GraphPad Software). The curve in red is the predicted change in the respective parameter based on the Henderson-Hasselbach equation using the spectroscopically determined pK_a (pK_a = 10.25).

Discussion

The replacement of the native metal in a metalloenzyme often has been used to glean important insight about the active site of such a system. Proteins containing Zn(II) are particularly attractive targets for such substitution because spectroscopic methods to evaluate the structure of zinc enzymes are limited. Related divalent ions containing unpaired d electrons may then serve as surrogates that allow more detailed interrogation of the active site ligands, geometry, and coordination number. Furthermore, properties such as hydrolysis may be quantified and intermediates in the catalytic process may be identified more easily. However, such substitutions are not always benign, leading to coordination environments that are altered from the native enzyme and catalytic activity that is significantly different.^[19] These distinct characteristics can arise even when the metal ion used is broadly similar to the native ion, including properties such as charge, size and allowed coordination geometries. One example is that when Zn(II) in CA is substituted by another transition metal ion such as Co(II), a decrease in the catalytic activity is registered (halved activity).^{[19a, 20]} Different metal coordination geometries arising through metal replacement in the non-native CA likely play an essential role in the different catalytic efficiencies by altering substrate/product binding or Lewis acidity in the enzyme.^{[19a, 21]}

Protein design, and in particular de Novo Design, allows for the interrogation of metal active sites in minimal functional units. One may ask whether a metal structure may be recapitulated in a dramatically different protein fold (e.g., an α-helical bundle vs β-sheet) and if so, whether the minimal unit containing the desired metal is competent to afford catalysis at a reasonable rate. Several years ago Zn(TRIL2WL23H)₃ was shown to both reproduce the active site structure of CA and to exhibit highly efficient hydratase (CO₂ +H₂O ⇆ HCO₃⁻ + H⁺) and esterase (RCO₂R’ ⇆RCO₂⁻ + R’OH) activities.^[4] While an X-ray structure of a mercurated analogue of Zn(TRIL2WL23H)₃ was reported (PDB 3PBJ), details about the high pH, catalytically active form remained elusive. For this reason, we have resorted to preparing and characterizing the Co(II) analogue of Zn(TRIL2WL23H)₃ in order to understand this designed enzyme more completely.

Addition of CoSO₄ to a neutral, buffered aqueous solution leads to the formation of Co(TRIL2WL23H)₃, where this nomenclature is intended solely to represent Co bound to a 3SCC without defining the hydration/hydrolytic state of the metal nor its oxidation state. The TRI peptides have been shown to form 3SCCs in the absence of bound metals and the CD spectra (Figure S1) over a broad range of pH values demonstrate that Co(II) binding does not negatively impact the folding of the system.

Direct titration of apo-(TRIL2WL23H)₃ with CoSO₄ can be followed by monitoring the UV-vis spectra shown in Figure 1. Based on these data one 3SCC binds a single equivalent of Co(II) and curve fitting gives a stability constant for the reaction Co(II) + apo-(TRIL2WL23H)₃ ⇆ Co(TRIL2WL23H)₃ at pH 7.8 of K_Co = 35 ± 4 μM. This binding affinity is sufficient to prepare stoichiometric Co(TRIL2WL23H)₃ under both spectroscopic and catalytic conditions. Similar Co to peptide titrations were performed at pH 10.8; however, quantitative analysis of these data were complicated by metal hydrolysis and precipitation at this high basicity. Clearly, at high pH, there is a kinetic limitation to the binding of the metal to the protein. None-the-less, if a sample of Co(TRIL2WL23H)₃ is prepared at pH 7.5 and the solution pH then raised to 10.8, the Co peptide remains stable for long periods. The pH dependent UV-vis spectra of the cobalt peptide are provided in Figure 2. Up to pH 9.5 the broad, weak feature of the low pH form of Co(TRIL2WL23H)₃ is retained; however, under more basic conditions, new red-shifted features with slightly greater extinction coefficients appear at 590 and 640 nm. These features continue to grow in through pH 10.8. Under more basic conditions, the Co(TRIL2WL23H)₃ complex begins to lose Co, presumably as the insoluble oxide.

Previous work with CA has shown that Zn(II) binds to three histidine residues and a water molecule (Zn(His)₃(OH₂)) at low pH and converts to a tris(histidine)hydroxy form (Zn(His)₃(OH)) with a pK_a~ 6.8 that is the active hydration/esterase catalyst.^{[3e, 4–5]} The Zn(II) ion may be replaced by Co(II) to generate a less active CoCA. Both Zn and Co forms have been characterized by X-ray crystallography at different pH values and with active site mutations.^[6] The visible spectra of CoCA crystals^[22] are similar to the spectra of Co(TRIL2WL23H)₃ at low and high pH (Figure 7). CoCA shows a broad maximum at 540 nm at neutral pH. As the pH is raised, two new features at 622 and 642 nm grow in. For Co(TRIL2WL23H)₃, a broad feature at pH 9 and below is observed at 520 nm with a similarly low extinction coefficient. As the pH is raised to 10.8, two new absorption features appear at 590 and 640 nm. This strong similarity in absorption spectra suggests that related structures are formed in the native and designed proteins.

Figure 7.

Comparison of pH dependent visible spectra for crystalline CoCA^[22] (Top) and Co(TRIL2WL23H)₃ (Bottom). Spectra for CoCA correspond to dark blue, pH 7.0; light blue, pH 8.0; magenta, pH 9.0; red, pH 10.0. Spectra for Co(TRIL2WL23H)₃ correspond to orange, pH 8.0; magenta, pH 8.75; blue, pH 9.5; green, pH 10.0; red, 10.5; black, 10.7.

As with ZnCA, the CoCA form has been assigned as a Co(hydroxy) species at higher pH. The pK_a is elevated compared to the Zn enzyme (Zn, 6.8–7.3; Co, 8.4). The calculated pK_a of 10.25 for Co(TRIL2WL23H)₃ is also elevated from the value of 9.2 for Zn(TRIL2WL23H)₃. As is the case for ZnCA, Zn(TRIL2WL23H)₃ and CoCA, this pK_a is assigned to the conversion of a metal bound water to the hydroxy species. One should notice that the CA systems consistently exhibit lower pK_as than the metalated (TRIL2WL23H)₃, likely due to a nearby threonine that forms a hydrogen bond to the coordinated water/hydroxide. The designed proteins do not contain this residue, so there is a 2.1 ± 0.3 unit pK_a shift to more basic values for Zn or Co(TRIL2WL23H)₃. In both the CA and (TRIL2WL23H)₃ systems, the Co enzymes are always more basic than the Zn congeners, reflecting the stronger Lewis acidity of Zn(II).

While there is clarity that native CAs have a clean conversion from a Zn(II)His₃(H₂O) form to a catalytically active 4-coordinate Zn(II)His₃(OH) environment, there has been considerable controversy about the correlation of the low pH spectrum to the active site geometry and oxidation state of the CoCA derivative. In many systems, Co(II) has the propensity to adopt higher coordination numbers when substituted for Zn(II).^[9] Early studies had suggested that the low pH form was a Co(II) compound,^[8b] likely pseudotetrahedral in nature;^{[21b, 22]} however, a five-coordinate geometry has been observed in the presence of sulfate (bound at pH 6.0)^[21b] and in the presence of citrate (pH 8.5, CoHis₃(H₂O)₂).^[22] An octahedral cobalt center was reported at pH 6.0 in the presence of citrate.^[22] That center was proposed to have oxidized to Co(III) on the basis of similar Co(III)-CA octahedral geometry.^[23] It should be noted that the conditions required to observe this oxidation (i.e. excess of H₂O₂) are far from our conditions. Based on an X-ray structure,^[6] a 540 nm band has been assigned to a Co(III) hexacoordinate species. These authors assign the spectra of the high pH form to a tetrahedral Co(II) species. Given the variety of Co coordination environments suggested in CA, we wanted to be especially careful in assigning oxidation state and structure to the Co(TRIL2WL23H)₃ peptides.

We have shown previously^[5] that Zn(TRIL2WL23H)₃ at neutral pH forms a 4-coordinate Zn(II)His₃(H₂O) complex (an alternate Zn(II)His₃Cl compound has also been structurally characterized). The higher pH form was characterized by X-ray absorption spectroscopy and is best fit as a 4-coordinate ZnN₃O species that we have assigned as Zn(II)His₃(OH). Such an assignment is consistent with the high activity of this compound by analogy to the native ZnCA enzyme. To assess the properties of the Co(TRIL2WL23H)₃ peptide, we have undertaken several different approaches.

We first addressed the oxidation state of the metal in Co(TRIL2WL23H)₃ at various pH values. On their own, the UV-visible spectra do not distinguish the two possibilities of cobalt oxidation state as both Co(II) and Co(III) are predicted to have broad, low intensity visible absorption features. However, we can use these spectra to assess the behavior of the Co center under different conditions of redox poise and pH. Titration of (TRIL2WL23H)₃ with CoSO₄ under aerobic and anaerobic conditions yields identical spectra as shown in Figures 1 and S5, indicating that ambient dioxygen is unable to oxidize the Co(II) to Co(III) at low pH. Furthermore, if 10 equivalents of ascorbate are added to Co(TRIL2WL23H)₃, the UV-vis spectrum is invariant (Figure S6). These neutral pH experiments strongly support the assignment of Co(II) to the metal center. We have also examined the stability of Co(II)(TRIL2WL23H)₃ as a function of pH. As shown in Figure S5, the UV-Vis spectra of solutions prepared and maintained in the presence or absence of dioxygen over the pH range of 8.0 to 10.8 are identical. Furthermore, titrating the high pH form of Co(II)(TRIL2WL23H)₃ with acid and with and without dioxygen reveals identical behavior. For these reasons we assign the metal as Co(II) across the studied range of pH values.

This oxidation state assignment is confirmed using the energies of both the pre-edge and the rising edge features. The intensity change in the pre-edge feature (Figure S7) is consistent with the conversion from a centrosymmetric pseudo-octahedral geometry to a lower symmetry site as would be seen from a decrease in coordination.

Next, we examined UV-Vis spectral features of Co(TRIL2WL23H)₃ to characterize the ligand type and number. The molar absorptivity values of visible d-d bands for Co(II) compounds have been shown to be related to the coordination number of these complexes. In general, ε values lower than 50 M⁻¹ cm⁻¹ are associated with six-coordinated species, while values higher than 300 M⁻¹ cm⁻¹ are associated with four-coordinate compounds. Values of ε between these two extremes suggest a five-coordinate structure. In addition, six-coordinate systems appear as a single broad band whereas 5-coordinate compounds typically exhibit more complex spectra due to the lower symmetry.^[24] An octahedrally coordinated Co(II) is suggested at pH 8.0 for Co(II)(TRIL2WL23H)₃ according to the absorptivity (ε = 43 ± 1 M⁻¹ s⁻¹) of the single broad band at 517 nm. Given the tris(histidine) environment offered by the peptide scaffold, we suggest that the structure is likely a facial Co(II)(His)₃(H₂O)₃ environment. By raising the pH, a more complex d-d spectrum develops with absorption maxima at 590 and 640 nm having estimated ε ~85 M⁻¹ cm⁻¹ (the measured epsilons are 60–65 M⁻¹ cm⁻¹ at pH 10.7, which when corrected for the pK_a to yield 100% of the high pH form, give these values). Thus, the Co(II) center appears to have adopted a pentacoordinate structure given the complexity of the spectra and the intermediate magnitude of extinction coefficients.

At this point, our predicted structure is Co(II)(His)₃(H₂O)(OH). A more direct determination of the Co structure can be obtained from EXAFS spectra. The correlation between coordination number and Debye-Waller factor makes it difficult to determine the coordination number based solely on the EXAFS amplitude. However, the average nearest neighbor distance for pH 8 is in good agreement with the average found for the Co(II)(imid)₃O₃ structures in the Cambridge Crystallographic Database (2.13 ± 0.01 Å) and the decrease of ~0.03 Å similarly matches that seen for crystallographically characterized Co(II)(imid)₃O₂ structures. The one anomaly in these data is the sample at pH 9.5. While the titration curve at room temperature suggests that the samples at pH 10.5 and 10.8 should be nearly 100% in the high pH form and the sample at pH 8 should be 100% in the low pH form, consistent with the X-ray absorption data, the sample at pH 9.5 is not expected to be more than 15% in the high pH form. We attribute this to a temperature dependent change in the sample pH on freezing. For samples well below or well above the pKa (i.e., pH = 8, 10.5, and 10.8) small changes in pH on freezing will not causes any significant change in composition. However, for the pH 9.5 sample, with a nominal pH close to the pKa, even a small temperature dependent change in pH will cause a large change in composition.

The UV–visible absorption and XAS data together reveal how Co(II) binds to His₃ site in this 3SCC. We can conclude that under neutral to slightly basic conditions (e.g., pH 8.0) Co(II) binds to the His₃ site with three additional H₂O/OH⁻ molecules coordinated, forming an CoN₃O₃ structure. With an increase in pH, the coordination changes to CoN₃O₂, a configuration more conducive for nucleophilic attack to lead to the ester hydrolysis. It should be emphasized here that in neither case (low or high pH conditions) does the Co(II) recapitulate the Zn(II)(TRIL2WL23H)₃ structures that are known to be pseudotetrahedral (ZnN₃O) under the full range of pH conditions.

With the structure and oxidation states of the Co(II)(TRIL2WL23H)₃ complex now well-defined, we explored the catalytic behavior of this designed construct. The enzymatic activity of TRIL2WL23H in presence of Zn(II) was found to increase on moving from neutral to basic conditions.^[4] Therefore, we anticipated that similar behavior might be observed for Co(II)(TRIL2WL23H)₃. We chose to examine the esterase activity of our cobalt protein using p-nitrophenylacetate due to the simplicity measuring rate constants for this reaction. We collected initial rate data for the production of p-nitrophenol and fit these data using a standard Michaelis-Menten analysis.

The pH-dependent kinetic parameters for pNPA hydrolysis by Co(II)(TRIL2WL23H)₃ and Zn(II)(TRIL2WL23H)₃ peptides are compared in Table 3. The Zn data are from reference [4]. The first point to emphasize is that in both cases we see a progressive increase in esterase activity (measured as either k_cat or k_cat/K_M) as the pH is increased. At the same time, both Zn and Co enzymes have essentially identical K_M values (~ 2 ± 0.5 mM) that are effectively pH invariant. This latter point indicates that there is very similar access of the substrate to the active site in either metal derivative and that the K_M is not the basis for the different catalytic parameters between Co and Zn scaffolds. As shown by the plots of pH vs Michaelis constants given in Figure 6, the K_M values are also not responsible for the differential between low and high pH activities.

These data demonstrate that the Zn and Co enzymes follow a similar pH dependence on k_cat and k_cat /K_M, but are displaced by roughly one pH unit from one another. The solid red line in each graph is the predicted progression of K_M, k_cat and k_cat/K_M based on the 10.25 pK_a that was determined spectrophotometrically for [Co(II)(H2O)₃](TRIL2WL23H)₃ⁿ⁺. These same data may also be fit directly to determine the kinetic pK_a for the process in Prism 8 (GraphPad Software) as was previously done with the Zn(II)-peptide complex for which a determination of a spectroscopic pK_a was impossible. The analysis was performed with respect to k_cat, K_M and k_cat/K_M vs pH (Table 4 and Figure S10). This analysis yielded a pK_a = 10.32, which is in close agreement with the spectroscopic value for [Co(II)(H₂O)₃](TRIL2WL23H)₃ⁿ⁺. One should also note that there is an elevation of the pK_a of the Co system as compared to the Zn designed enzyme, which is analogous but slightly smaller in magnitude that reported for the ZnCA vs CoCA. Because of the high pH instability of the [Co(II)(H₂O/OH⁻)](TRIL2WL23H)₃ⁿ⁺ we were unable to directly measure the largest kinetic constants; however, since the kinetic data closely follow the pK_a dependence of hydrolysis we can estimate the maximal activities of [Co(II)( H₂O/OH⁻)](TRIL2WL23H)₃ⁿ⁺. These are the kinetic parameters reported in Table 4.

Table 4.

Kinetic pK_a, Maximal Efficiency, and Maximal Rate Values for pNPA hydrolysis by Metal(II)-Bound TRIL2WL23H Peptides^a

Peptide complex	pKa a	Kcat/KM (max) (M−1 s−1) b	kcat(max) (s−1) (x 10−3) c
[Co(II)(H2O/OH−)]N(TRIL2WL23H)3n+	10.32 ± 0.01	52.6 ± 1.9	~ 129.3
[Zn(II)(H2O/OH−)]N(TRIL2WL23H)3n+ d	9.2 ± 0.1	25 ± 2	~ 55

^[a]Determined by fitting individual k_cat/K_M values versus pH.

^[b]Maximal catalytic efficiency from the fitting of k_cat/K_M values versus pH (assuming that 100% active enzyme complex is present).

^[c]Estimated maximal rate determined as described in the text with eq. 1 for Co(II).

^[d]Previously reported data included for comparison.^[4]

Based on catalytic efficiencies (k_cat/K_M) the [Co(II)(H₂O/OH⁻)](TRIL2WL23H)₃ⁿ⁺ system is actually twice as efficient as the Zn designed enzyme. Both systems also retain an approximately factor of 2 increase in maximal rate. That both the catalytic efficiency (which provides information about the first chemically irreversible step) and the maximal velocity (which provides information about the rate limitation for the complete reaction) follow the same pH dependence supports the conclusion that the chemically critical conversion is rate limiting and that substrate access or product release do not influence significantly the catalytic process.

An added bonus of these kinetic studies is that they provide insight for the oxidation state assignment for Co(TRIL2WL23H)₃ at neutral pH. The catalytic activity of [Co(II)(H₂O)₃](TRIL2WL23H)₃ⁿ⁺ at pH 8.0, while significantly lower than the maximal activity at high pH, is still substantial (2400 s⁻¹). This is another piece of evidence supporting the Co(II) oxidation state at neutral/slightly basic pH. When Co(III) is bound to CA, it renders the enzyme completely inactive, whereas Co(II) in CA retains most of the activity (for both pNPA hydrolysis and CO₂ hydration).^[25]

Conclusions

In recent years a number of highly active de Novo designed metalloenzymes have been reported. The goal of these systems is to build catalytically active metal sites often into non-natural protein folds to test how important the native protein structure is for catalysis and to assess whether possibly more simplistic and stable systems are capable of carrying out desired chemical reactions. In this report we accomplish two significant advances in this field. First, we provide the first designed example where a native activity (in this case Zn esterase) mimicking carbonic anhydrase can be achieved without the native metal. Second, the (TRIL2WL23H)₃ system not only can recapitulate the active site structure and activity of native ZnCA, it can also be substituted with Co(II) to generate an enzyme with even higher catalytic efficiency than the original Zn(TRIL2WL23H)₃. This behavior is opposite to native CA enzymes where the substitution of Zn(II) with Co(II) halved the catalytic efficiency.

While an X-ray structure of a Zn(TRIL2WL23H)₃ analogue at neutral pH is available, there has been little other structural characterization of this system, especially at high pH. In this report, we can now use Co(II) as a spectroscopic probe to interrogate catalytic intermediates and high pH active sites. With the knowledge gained from these studies, we can immediately conclude that Co(II) does not structurally reflect the Zn(II) enzyme either at neutral or high pH conditions and yet matches or exceeds the catalytic ability of the initial system. Therefore, even in a relatively non-complex scaffold which exhibits closely similar catalytic activities, the interchange of closely related cations can lead to major structural alteration of the metalloenzyme active site.