Uno Ferro, a de novo designed protein, binds transition metals with high affinity and stabilizes semiquinone radical anion

Abstract

Metalloenzymes often utilize radicals in order to facilitate chemical reactions. Recently, DeGrado and coworkers have discovered that model proteins can efficiently stabilize semiquinone radical anion produced by oxidation of 3,5-di-tert-butylcatechol (DTBC) in the presence of two zinc ions. Here, we show that the number and the nature of metal ions have relatively minor effect on semiquinone stabilization in model proteins, with a single metal ion being sufficient for radical stabilization. The radical is stabilized by both metal ion, hydrophobic sequestration, and interactions with the hydrophilic residues in the protein interior resulting in a remarkable, nearly 500 mV change in the redox potential of the SQ^•−/catechol couple as compared to bulk aqueous solution. Moreover, we have created 4G-UFsc, a single metal ion-binding protein with pM affinity for zinc that is higher than any other reported model systems and is on par with many natural zinc-containing proteins. We expect that the robust and easy-to-modify DFsc/UFsc family of proteins will become a versatile tool for mechanistic model studies of metalloenzymes.

Graphical Abstract

Bound to transition metal ions, proteins of the DFsc family facilitate oxidation of 3,5-di-tert-butylcatechol by dioxygen into corresponding semiquinone radical. We have created 4G-UFsc, a single metal ion-binding protein with pM affinity for zinc that is higher than any other reported model systems and is on par with many natural zinc-containing proteins.

Enzymes facilitate an incredible variety of challenging chemical reactions with high efficiency and selectivity by concerted action of multiple functional groups in a precisely tuned local environment. Even highly reactive radicals can be successfully employed by enzymes to promote reactions that are absolutely necessary for life: conversion of ribonucleotides into deoxyribonucleotides, oxygen evolution in photosystem II, molybdenum cofactor biosynthesis, nitrogenase cluster M biosynthesis.^[1] In many cases, enzymes require metallocofactors to generate radicals, e.g. in amino acid oxidation and reductive cleavage of S-Adenosylmethionine, and metal ions are used to stabilize radicals.^[2] Due to the inherent complexity of enzymes and a transient nature of radicals, discerning the contribution of various factors that determine the overall reactivity of radical-promoting enzymes is often challenging.

Model proteins that recreate basic arrangements of functional groups found in enzymatic active sites, but have much lower overall complexity provide insight into the factors that determine protein structure and function.^[3] DeGrado and coworkers pioneered use of Due Ferri (DF) helical bundles designed from first principles to replicate coordination geometry, and second sphere interactions of non-heme diiron enzymes such as ribonucleotide reductase, ferritin, methane monooxygenase and dimanganese enzymes: catalase and arginase.^[4] DF family proteins bound to metal ions can facilitate multiple chemical transformations such as oxidation of hydroquinones into quinones^[5] and N-hydroxylation of anilines.^[6] Recently, model systems have been also utilized to investigate strategies used by proteins to stabilize radicals and prevent harmful side reactions.^[7]

In a major advance, DeGrado and coworkers showed that a zinc-containing protein [3His-G2DFsc-Zn₂] derived from DFsc,^[8] stabilizes a semiquinone radical anion formed by comproportionation of 3,5-di-tert-butylcatechol (DTBC) and the corresponding quinone (Q), Scheme 1.^[2a] 3His-G2DFsc-Zn₂ is an evolved mutant of DFsc, a single chain version of a DF bundle, created to lift the limitations imposed by the D₂ symmetry of the original design.^[9]

Scheme 1.

SQ^•− is an unstable intermediate between fully reduced catechol (DTBC) and fully oxidized quinone (Q) states.

DTBC, widely used to study catechol oxidation,^[10] is an excellent model substrate for establishing basic principles that govern mechanisms of radical enzymes. Interestingly, coordination of the substrate to a non-redox active zinc metal ion in [3His-G2DFsc-Zn₂] has a major impact on semiquinone production. In this work, we set out to explore the role of the metal in radical production in model systems to understand the factors that contribute to radical stabilization by proteins.

As a model protein for our studies, we selected the original version of DFsc.^[8] DFsc binds two metal ions, is extensively structurally characterized (Figure 1), very stable and can be easily expressed in E. coli in high yield. Notably DFsc is folded in both the presence and the absence of metal ions. Excitingly, oxidation of DTBC by dioxygen in the presence of zinc-bound DFsc produces spectroscopically detected semiquinone radical anion with yield comparable to that of [3His-G2DFsc-Zn₂] (Figure S1).

Figure 1. DFsc and ligands in primary coordination sphere.

Metal binding site of DFsc showing the side chains that coordinate zinc ions (purple spheres). (A) NMR structure of DFsc generated from PDB 2HZ8.^[8b] (B) Ligands that coordinate to metal ions; Tyr51 and Tyr18 are also shown as sticks.

The DF family of proteins, moreover, are specifically designed to bind metal ions. Hence, the first question we set out to explore was whether the metal ion is in fact essential for radical stabilization. To evaluate the importance of the metal for radical stabilization, we tested DFsc in its apo- or one metal bound forms. Apo DFsc promoted essentially no radical production under typical reaction conditions (Figure S2A). DFsc promoted formation of the semiquinone radical anion with approximately 30% less yield (based on the absorbance at 760 nm) in the presence of one equivalent of zinc (Figure S3A) compared to the fully metallated protein (Figure S4A).

To further evaluate the contribution of metal to radical stabilization, we converted DFsc that can bind up to two metal ions into a single metal binding protein by mutating the bridging Glu104 to a histidine (Uno Ferro single chain, UFsc). Additionally, we created 4G-UFsc by introducing A10G, A14G, A43G and A47G mutations to widen the channel that leads to the metal center (Figure S5). Increase in the channel size has been demonstrated before to improve substrate access^[11] and resulted in higher rate of substrate oxidation.^[5-6] Titration of UFsc and 4G-UFsc with Co²⁺ as well as ITC data demonstrated that the proteins bind only one metal ion (Figure 2) and are folded even in the apo form (Figure S6). Importantly, the proteins bind zinc with very high affinity; the dissociation constants measured by ITC competition experiment with triethylenetetramine (TETA) for the corresponding zinc complexes are 76 ± 13 pM for UFsc and 30.2 ± 4.4 pM for 4G-UFsc (Figures S7, G-H and S8, G-H). These values are about 3 orders of magnitude lower than those obtained for DFsc (Figure S9, G-H, Table 1), These affinity levels are quite extraordinary for de novo designed proteins and are at least two orders of magnitude tighter than the best example reported to date.^[12] This is even more remarkable, considering that UFsc proteins do not rely on cysteines for zinc binding. Interestingly, metal binding is entropically driven in the case of DFsc and UFsc, but is driven by both negative change in enthalpy and positive change in entropy in the case of 4G-UFsc. 4G-UFsc especially displays the highest affinity for zinc of all of the reported designed proteins to date. Furthermore, both UFsc and 4G-UFsc promote semiquinone radical formation in the presence of one equivalent of Zn²⁺ (Figures S10 and S11), in the latter case, the radical yield is almost the same as for DFsc with two equivalents of zinc bound. These results demonstrate that only one metal ion fulfills the radical stabilization role in the designed protein.

Figure 2. Metal binding characterization.

Co²⁺ titration monitored by visible absorption spectroscopy for UFsc (A) and 4G-UFsc (B).UFsc (150 μM) binds a maximum concentration of 140 μM Co²⁺ as monitored by a change in absorbance at 561 nm; 4G-UFsc (150 μM) binds a maximum concentration of 180 μM Co²⁺ as monitored by a change in absorbance at 550 nm. Insets display absorbance spectra with increasing [Co²⁺]. Dashed spectrum corresponds to a Co-protein mixture, where Co²⁺ saturates UFsc and 4G-UFsc at 140 μM and 180 μM, respectively. The thermogram (C) and the integrated binding isotherm (D) of 4G-UFsc-zinc binding obtained from the competition titration with triethylenetetramine (150 μM zinc with 500 μM TETA was titrated with 1.2 mM 4G-UFsc in 25 mM HEPES, 100 mM NaCl, pH 7.6 at 25°C). The dissociation constant of 30 ± 4 pM was derived from the non-linear least square fit of the isotherm using ‘one set of sites’ model.

Table 1.

Dissociation constants (in nM) determined by ITC for metal complexes of DF family proteins.

	Ni2+	Co2+	Mn2+	Zn2+
DFsc	Kd1 2,090 ± 230	56 ± 3	197 ± 37	109 ± 34
	Kd2 10,400 ± 1,400	133 ± 10	228 ± 37	213 ± 14
UFsc	112 ± 28	269 ± 116	173 ± 56	0.076 ± 0.013[a]
4G-UFsc	220 ± 37	67 ± 4	9,250 ± 520	0.030 ± 0.004 [a]

^[a]Dissociation constants determined by ITC competition experiments with triethylenetetramine (TETA).

Natural proteins commonly utilize redox active metal ions, thus, we probed the role of the nature of metal ion in semiquinone stabilization. We have quantitatively measured metal binding affinities for DFsc, UFsc and 4G-UFsc with Ni²⁺, Co²⁺, Mn²⁺ and Zn²⁺. The stoichiometry of binding (two for DFsc and one each for UFsc and 4G-UFsc) remains constant regardless of the nature of the metal ion. The binding affinities for these metals while significantly lower than those for zinc (Table 1) but are still very high (sub micromolar in nearly all cases).

Next, we measured SQ^•− formation for DFsc, UFsc and 4G-UFsc in the presence of different amounts (one and two equivalents) of various metal ions (Figures S3, S4, S10, S11). In all cases, formation of the semiquinone radical anion was observed. Addition of one equivalent of the metal results in radical stabilization and addition of the second metal ion only leads to minor (if any) increase in the yield of the radical as evidenced by an appearance of colored species with λ_max ~ 760 nm. Production of the radical species that is stable in aqueous solution for at least two weeks was confirmed in all cases by EPR spectra (Figure S12). The yield of the radical is dependent on the nature of the metal ion: DFsc bound to one equivalent of Ni²⁺ shows the highest stabilization (36.8 % higher than DFsc-Zn₂), addition of the second equivalent of Ni had no effect on radical production.

The coordination environment of two metal-binding sites in DFsc is symmetric but both UFsc and 4G-UFsc bind only one metal ion and we set out to establish the coordination geometry in these proteins. DeGrado, Solomon and coworkers demonstrated that Y51 (Figure 1) coordinates to Fe³⁺ through the side chain phenolate when iron-loaded DFsc protein reacts with dioxygen, showing a transient absorbance band with λ_max = 520 nm.^[13] When the same experiment was done with UFsc, no band at 520 nm was observed suggesting that there is no iron in the vicinity of Y51 and the metal ion is bound to the site composed of His77, Glu44, and Glu74 (there could be more ligands coordinating to the metal ions, Figure 3).

Figure 3. Kinetics of interaction between O₂ and Fe²⁺-loaded proteins DFsc and UFsc.

(A) UV-Vis spectra of the reaction after 2 s, containing protein (50 μM) loaded with Fe²⁺ (150 μM) in oxygenated buffer (25 mM HEPES, 100 mM NaCl, pH 7.6). (B) Absorbance at 520 nm as a log function of time for DFsc and UFsc proteins.

To quantitatively characterize the degree of stabilization of the semiquinone radical anion, we determined the redox potential of the SQ^•−/DTBC couple when bound to DFsc in the presence of nickel. The radical, generated by incubation of DTBC with the protein, was titrated by dithionite in the presence of indigo tetrasulfonate (ITS, redox potential –46 mV vs NHE^[14]) as a standard (Figure S13). The Ni₂-DFsc-bound SQ^•−/DTBC couple shows redox potential of about −100 mV, nearly 0.5 V lower than that of the free DTBC (380 mV^[15])! Interestingly, the observed redox potential was even lower than that of DTBC bound to [3His-G2DFsc-Zn₂] (−21 mV).^[2a]

The results presented above, show that while the yield of the semiquinone radical anion changes somewhat based on the metal ion and how much of it is used, the nature of the metal ion is not nearly as important as the environment provided by the protein. In agreement with previous studies by DeGrado and coworkers, we find that substrate interactions with the protein are critical to radical formation.^[2a][4][5] No semiquinone radical production is observed when DTBC is slowly oxidized to form the corresponding quinone (Q) in the absence of protein (Figure S14). Moreover, hydrophobic sequestration of the substrate alone leads to very little amounts of SQ^•−: incubation of DTBC in oxygenated buffer in the presence of n-dodecylphosphocholine (DPC) micelles or bovine serum albumin (BSA) produces the corresponding quinone and very little SQ^•− (Figure S15).

To gain further insight into the interactions of the quinones with the designed protein, we undertook an in-depth structural characterization of substrate interaction with the apo protein by NMR. The overall fold of DFsc is not perturbed by DTBC binding (Figure S16). The ¹⁵N-¹H HSQC NMR spectrum of DFsc shows excellent signal dispersion (Figure S17), which allowed us to obtain structural information on the substrate binding to the protein. We have assigned all the signals of DFsc using [¹³C,¹⁵N]-labelled protein and standard methods as described in the Supporting Information. Titration of the protein with DTBC identified residues with chemical shifts most affected by substrate binding (Figures S18-S20) and established the corresponding K_d value (52 ±18 μM, Figure S21). NH signals of Ile17, Ile40, Lys42 shifted by more than 0.15 ppm in the presence of substrate, suggesting that DTBC comes in close contact with these residues.

Additionally, strong intermolecular NOE contacts between the 3-tBu group of catechol and Hα of Ala14, 3-tBu and Hα of Ile40 were obtained from 2D (¹³C,¹⁵N)-half-filtered NOESY and 3D (¹³C,¹⁵N)-filtered, ¹³C-edited NOESY-HSQC spectra. These constraints allowed us to build a structural model of DFsc-DTBC interaction based on the previously reported NMR structure of DFsc using Rosetta (see Supporting Information for experimental detail).^[16] In the lowest energy model, the substrate is located inside a hydrophobic pocket lined by Ala10, Ala14, Ala43, Ala47 with OH groups facing the negatively charged interior (Figure 4, Figure S22); the 3-tBu group is positioned between Ala14 and Ile40. Importance of the tBu groups in substrate-protein interaction was also indirectly confirmed by the lack of semiquinone radical produced during the oxidation of 4-tert-butylcatechol (4-TBC) in the presence of DFsc (Figure S14).

Figure 4. NMR model of DFsc-DTBC interaction generated using Rosetta.

(A) Surface representation of the lowest energy model with the substrate sitting inside the pocket lined by Ala10, Ala14, Ala43, Ala47 (light green). (B) Interactions between the 3-tBu group of the substrate and Hα of Ala14, Hα of Ile40. Residues shown in orange (Δδ > 1.5 ppm) and green (Δδ > 0.1 ppm) demonstrate significant chemical shift perturbation upon substrate binding.

In summary, we have shown that the number and the nature of metal ions have relatively minor effect on semiquinone radical anion formation in model proteins, with a single metal ion being sufficient for radical stabilization. These results have major implications in understanding the mechanism of metalloenzyme function. For example, catechol dioxygenase uses Fe²⁺ coordinated through a 2-His-1-carboxylate motif to facilitate electron transfer from the bound substrate to dioxygen and formation of semiquinone radical intermediate.^[17] It has been universally accepted that coordination of the catechol substrate to a redox active metal ion is absolutely required for radical production in metalloenzymes. Recently, DeGrado and coworkers showed that the metal ion does not have to be redox active and a model di-zinc protein provides excellent radical stabilization.^[2a] In this work, we show that a semiquinone radical can be stabilized by a protein containing various single metal ions, at the same time specific interactions of polar functional groups buried in the hydrophobic core of a model protein with the substrate are very important for semiquinone radical anion stabilization. Moreover, we have created 4G-UFsc, a single metal ion-binding protein with pM affinity for zinc that is higher than any other reported model systems and is on par with many natural zinc-containing proteins. We expect that robust and easy-to-modify DFsc and UFsc family of proteins will become a versatile tool for mechanistic model studies of oxidases.