Artificial diiron proteins: From structure to function

Abstract

De novo protein design provides an attractive approach for the construction of models to probe the features required for the function of complex metalloproteins. These minimal models contain the essential elements believed necessary for activity of the protein. In this article, we summarize the design, structure determination, and functional properties of a family of artificial diiron proteins. © 2005 Wiley Periodicals, Inc. Biopolymers (Pept Sci), 2005

INTRODUCTION

It is quite remarkable that metalloproteins use only a handful of metal ion cofactors to help catalyze a myriad of diverse reactions in the living cell. The diiron/dimanganese cluster is a particularly versatile cofactor found in functionally diverse enzymes. They are able to catalyze both hydrolytic and oxygen-dependent reactions. Similar diiron sites are capable of catalyzing the ferroxidase (ferritins),1–3 hydroxylase (methane monooxygenase4–7 and toluene oxygenases8–10), catalase (Mn catalase),11, 12 alkane desaturation [acyl carrier Δ⁹ desaturase (Δ⁹ ACP desaturase)]13, 14 reactions, or are capable of acting as a radical generator in ribonucleotide reductase.15–17 They have a common dimetal site consisting of two metal ions, bridged by a combination of oxo, hydroxo, or carboxylate donors, and are housed within a very simple four-helix bundle motif (Figure 1). How do their structures tune the chemical properties of a common dimetal cofactor to obtain such a wide diversity of functions? This question has classically been addressed through the study of variants of natural proteins as well as from the study of small complexes.18–31

Figure 1. The four-helix bundle motif that houses the diiron site (top) and the arrangement of the ligands and cofactor in diiron proteins (bottom). (A) Bacterioferritin (PDB: 1BCF), (B) ribonucleotide reductase 2 (PDB: 1RIB), (C) hydroxylase component of methane monoxygenase (PDB: 1FYZ), and (D) acyl carrier protein Δ desaturase (PDB: 1AFR) as derived from X-ray crystallography. Figures were generated with molscript.93

The current state of the art in the development of small synthetic models of diiron-oxo proteins has recently been reviewed by Tshuva and Lippard.32 Small molecule models have been devised to reproduce the structures and spectroscopic properties of the various redox states of the cofactor, including a high-valent diferryl intermediate.33, 34 These models share similarities with the intermediates in the reaction cycle of the natural proteins, thus allowing a better interpretation of the chemistry supported by the natural systems. A current important goal is to design more complex models to allow the evaluation of environmental variables, such as the role of second shell ligands, and to provide sites for binding substrates as well as the dimetal cofactor. This can be achieved through different strategies, like the use of sterically hindered ligands, dendrimer structures, or peptide based models.

Peptide models stand at the crossroads of small molecule models and large proteins. Their structures are simple, and hence more easily understood than their natural counterparts. They simultaneously have sufficient size and chemical diversity to allow the construction of functional binding and catalytic sites. The development of peptide-based models takes advantage of recent progress in de novo protein design. Recent advances in computational protein design35–40 have allowed for the construction of rigid scaffolds able to accommodate substantial changes in their sequence while retaining their native three-dimensional structures. As the principles and methods for de novo protein design have matured, it has recently become possible to design structurally defined models for metalloproteins. Previous work from our lab focused on the design of heme-binding proteins containing between one and four heme groups.41, 42 These proteins provided excellent frameworks for determining how the protein matrix and heme–heme interactions affect electrochemical midpoint potentials of their bound cofactors.

More recently, we have turned our attention to the diiron class of proteins. Our goal is to elucidate how the electrostatic environment, polarity, and solvent accessibility of the metal-binding site influence the properties of diiron proteins, including the electrochemical midpoint potential and reactivity. The ultimate goal is to construct synthetic protein-based catalysts with enhanced stability and novel functions.

STRUCTURAL ANALYSIS OF DIIRON PROTEINS

Our initial design of diiron peptides evolved from a retro-structural analysis of the six different carboxylate-bridged diiron proteins (ferritin, bacterioferritin, rubrerythrin, ribonucleotide reducatse R2 subunit, Δ⁹ ACP desaturase, and the catalytic subunit of methane monooxygenase), which all contain two tandem helical Glu–Xxx–Xxx–His motifs, where Glu and His are liganding residues.43 Proteins of this class have low sequence identity (less than 5%), and they span a variety of different tertiary structural families. Nevertheless, their active sites are housed within a very simple pseudo-222-symmetric four-helix bundle.

Figure 2 illustrates the structure of the catalytic subunit of Δ⁹ ACP desaturase. In contrast to the complex nature of the overall structure, the active site is bounded by a structurally simple antiparallel four-helix bundle. The arrangement of the helices is particularly simple, and the overall bundle can be described to within about an Å by a D₂-symmetric model. The idealized helix contains hydrophobic residues (Hb) that encapsulate and structurally stabilize the diiron site. Each helix contains a Glu side-chain ligand. The position labeled in Figure 2 as “Xxx” is a His ligand in two of the helices. In the other two helices, this position helps define the environment and access surrounding the diiron center.

Figure 2. Acyl carrier protein Δ desaturase structure (PDB: 1AFR). The architecture of the diiron binding site can be described as a pseudo-symmetric four-helix bundle (middle). A single helix used to generate the complete bundle contains hydrophobic residues (Hb) and a liganding glutamate residue (Glu); Xxx represents a His ligand in 2 of the helices, while in the other 2 helices Xxx helps to define the environment and access surrounding the dimetal center.

Looking in more detail at the active site in many members of this class of proteins, a highly symmetric arrangement of liganding side chains appears. The sites generally have four Glu residues (Figure 3) that project toward the center of the bundle: two bridge both metal ions, while the other two carboxylates interact with a single metal ion in a monodentate or bidentate chelating interaction. The two His residues at position i + 3 relative to the two bridging Glu side chains form additional monodentate ligands. The absence of a His ligand in the other two helices leaves two adjacent coordination sites available for interaction with exogenous ligands.

Figure 3. Structure of the dimetal site in DF1, an idealized diiron protein. The primary Glu and His ligands are shown along with the Zn²⁺ ions.

The helical bundles of many diiron proteins conform to a coiled-coil geometry throughout their structures, and all approximate this motif in the region immediately surrounding the metal-binding site. If the heptad nomenclature typically applied to coiled coils (the residues in the heptad are designed as a, b, c, … , g) is adopted, the four liganding Glu side chains can be seen to project from “a” positions, and the two His side chains from “d” positions. By convention, these positions project toward the center of the bundle (Figures 3 and 4).

Figure 4. Helical-wheel diagram of the active site of a diiron protein. Residue positions are labeled according to the heptad repeat usually applied to coiled coils. The 4 Glu side chains (a position) and the 2 His side chains (d position) point toward the center of the four-helix bundle. The second-shell hydrogen bonds are shown in red. The positions that help define the active-site access and environment are shown in blue.

The interfaces between helices are defined by “b” and “e” residues (top and bottom) or “g” and “c” residues (left and right). The residues along the b/e interface in diiron proteins appear to serve a primarily structural role, and they often form a tightly packed “Alacoil” motif.44 In contrast, the residues at one c/g interface hydrogen bond to the liganding His residue, forming second-shell interactions. The opposite c/g interface defines the entry to the active site. Finally, the residues at the “a” and “d” positions above and below the liganding residues form the top and bottoms of the active site. These side chains are often hydrophobic, and their association provides part of the driving force for the association of the bundle. Residues in these positions are also important for function; for example, in proteins with ferroxidase activity, a Tyr residue at an “a” position often hydrogen bonds to the carboxylate oxygen of one of the liganding Glu residues.45, 46 Also, a Tyr at an analogous position forms a free radical in R2.47

The first two subsets of the DueFerri family (DF), DF1 and DF2, contain two noncovalently associated helix–loop–helix motifs, which bind the dimetal cofactor near the center of the structure (Figure 4).43, 48 Our design evolved with the construction of heterotetramers, DFtet, consisting of four disconnected helices.49, 50 These helices can be combinatorially assembled to create a large number of helical bundles, which can be screened for activity.51 Finally, using a computational approach, we explored the feasibility of constructing a monomeric four-helix bundle, encompassing the dimetal cofactor, DFsc.52

DESIGN AND CHARACTERIZATION OF THE DUEFERRI CLASS OF PROTEINS

Based on the analysis above, we have prepared a series of models for the diiron protein. The first was DF1, which was a dimer of helix–loop–helix motifs. We also have designed a single-chain version of the protein (DFsc), as well as versions in which the four helices are distinct chains that come together by noncovalent self-assembly (DFtet) (Figure 5). Each class has different advantages for functional measurements. For example, the symmetric nature of the dimeric derivatives simplifies interpretation of data, whereas analogues of the four-chain DFtet constructs can be mixed and matched in indexed arrays to allow easy generation of combinatorial diversity.

Figure 5. DF protein constructs. Models are shown of dimeric (DF1), four-chain (DFtet) constructs, and single-chain (DFsc) diiron proteins. DF1 was first characterized and then used to design a system that would allow for the generation of mutants combinatorially, DFtet. To facilitate kinetic and spectroscopic characterization, a monomeric DF protein was designed, DFsc.

DF1

We first designed DF1, beginning with a D₂-symmetrical array of helices, each containing a single Glu residue. The positions of the helices and low-energy rotamers for the Glu side chains were chosen to allow ideal ligation of the diiron site. Next we introduced the two His side chains, which reduced the symmetry from D₂ to C₂. The asymmetric addition of the His side chains on only two of the four helices leaves a space free for binding of O₂ and substrates. Second-shell interactions are also important for maintaining the structures of metal-binding sites and for tuning their properties. They were carefully engineered in DF1: the nonbridging Glu carboxylates form a hydrogen bond with a Tyr side chain of a neighboring helix; similarly, an Asp residue forms a hydrogen bond with the imidazole Nϵ of a His ligand. We refer to the residues that form key active-site and second-shell interactions as the “keystone residues.” Together with the other primary interactions, they help to uniquely define the backbone of the protein, and set the stage for the design of the remainder of the sequence.

The side chains at the remaining solvent-inaccessible positions were chosen to be apolar to provide a strong driving force for folding. They were designed to pack efficiently in the interior of the protein.53 The remaining interfacial positions were selected to favor a unique topology through the formation of specific electrostatic interactions.54, 55 Helix-stabilizing Glu and Lys residues were selected for many of these positions. The design of DF1 also included an intermonomer salt bridge between Lys38 of one unit and Asp35′ of a neighboring helix. A similar interaction is present in the active sites of natural diiron proteins, such as methane monooxygenase, where the lysine residue is replaced by an arginine to form an Arg–Asp–His cluster (Figure 6).56

Figure 6. Second- and third-shell ligands in the crystal structures of DF1 (PDB: 1EC5) and methane monooxygenase (MMO) (PDB: 1MTY). The hydrogen-bonded network consisting of Lys/Asp/His in DF1 vs. Arg/Asp/His, as observed in MMO, is shown.

Finally, an idealized γ–α_L-β interhelical loop was included between the two pairs of helices. This is one of the most frequently occurring turns found in the structures of helical hairpins of natural proteins.57 Figure 7 illustrates the amino acid sequence and structure of DF1.

Figure 7. Crystal structure of DF1 (PDB:1EC5) and amino acid sequence of DF1, DF2, and DF2t.

DF1 was prepared in good yield by standard solid-phase methodology. It adopts a folded, helical conformation in aqueous solution irrespective of its ligation state. The CD spectrum for both the apo and Co(II)-reconstituted protein is consistent with its proposed helical structure, and equilibrium analytical ultracentrifugation indicated that the protein was dimeric. DF1 was able to bind Zn(II), Co(II), and Fe(II). The spectrum of the di-Co(II) complex shows absorption maxima, whose position and intensities are in reasonable agreement with the values reported in the literature for the Co(II) derivative of bacterioferritin.58 This finding supports the notion that the coordination of the cobalt at the dinuclear site is in approximately the designed geometry.43

We have solved the crystal structure of the di-Zn(II) form of the protein (PDB: 1EC5),43 as well as the NMR structure of the apo form (PDB: 1NVO).59 The structures conform extremely well with the intended designed models. Each Zn(II) atom is five-coordinate, with an open site trans to the two His side chains (Figure 3). Each metal ion has 3 glutamate ligands and 1 histidine ligand; Glu 36 and 36′ interact with both zinc ions in a bidentate bridging interaction, while Glu 10 and 10′ interact with individual ions in a chelating manner. His residues 39 and 39′ complete the liganding environment about the dimetal site, by their Nδ atoms. The Zn(II)–Zn(II) distance is 3.9 Å, close to the distance observed between metal ions in the parent natural proteins.1, 60, 45, 61 The pentacoordinate ligand arrangement leaves a pair of vacant sites along one face of the dimetal center. The vacant sites are oriented trans to the His ligands (Figure 3), and are well oriented for interaction with exogenous ligands.

All of the second-shell interactions were observed to form the desired hydrogen bonds. Tyr17 donates a second-shell hydrogen bond to the nonbridging Glu10′ of the other monomer in each of the dimeric units (the same interaction exists between the symmetrically related pair, Tyr17′ and Glu10). Also, consistent with the design, Asp35′ forms a hydrogen bond with the ligating His39 from a neighboring helix in the dimer. The intermonomer salt bridge between Lys38 of 1 helix and Asp35′ of the neighboring helix has also been observed in the crystal structure of DF1 (Figure 6). In summary, the crystal structure of DF1 is in excellent agreement with the designed model and demonstrates all specific H-bonded and metal–ligand interactions included in the model.

The NMR structure of the apo form of DF1 showed that the protein is folded, with the four Glu and two His side chains buried in the center of the protein, almost entirely preorganized for binding metal ions.59 As in native proteins, the apolar side chains were located in the core of the dimeric structure and generally have unique conformations. These studies established for the first time the ability to design and structurally characterize not only cofactor-binding sites, but also intricate networks of hydrogen-bonded second-shell interactions essential for tuning the properties of the cofactor.

DF2 and DF2t

As originally designed, the DF1 protein had low water solubility (10 μM), and it was necessary to perform early experiments in the presence of DMSO. Examination of the structure suggested that it contained too many hydrophobic residues on its surface, which were converted to polar residues in DF2. Also, the dimetal site in DF1 was relatively inaccessible; it was blocked by a pair of hydrophobic leucine residues (L13 and L13′).62 Therefore, in order to increase accessibility to the dimetal site in DF2, residue L13 was changed to a smaller Ala residue. The DF2 protein was expressed in bacteria, and it proved to have much better water solubility (0.5 mM). Although the protein is not as stable as DF1, it nevertheless retained sufficient stability in aqueous solution to be folded, even in the absence of metal ions.48 The NMR and crystal structures of the diZn(II) form of DF2 have been solved and are essentially the same as DF1.63

Although the DF2 protein was significantly more soluble than the corresponding DF1 protein, DF2 had a tendency to aggregate, and was expressed at relatively modest levels. A careful examination of the structure showed that the interhelical turn adopted a strained conformation, which might account for these problems. Although the analysis and design of turns that connect the strands in antiparallel β-hairpins has reached an advanced state, much less is known concerning turns between antiparallel helices in helical hairpins. We therefore conducted an analysis of the structures and sequence preferences of interhelical hairpins, in a large database of crystallographically characterized proteins.63 Interestingly, a very limited number of interhelical turns were found to occur with high frequency. Furthermore, these turns occurred only within well-defined structural contexts. This analysis confirmed that the turn introduced into DF2 did not occur within an optimal structural context. Therefore, a longer loop with a γ–β–γ–β conformation was inserted between the two helices in the protein, and a sequence was chosen to stabilize its conformation. This protein, DF2t expressed in significantly better yield, and did not show the same tendency to aggregate in solution as DF2. The structure DF2t was solved by X-ray crystallography (PDB: 1MFT)63 as well as NMR (PDB: 1U7M)63; and it was shown to confirm exactly to the design (Figure 8).

Figure 8. Structure of DF2 (red backbone) vs. DF2t (green backbone). The right panel shows detailed interactions included to stabilize the turn conformation of DF2t including the N- and C-capping interactions involving Asn27 and Lys31, respectively.

Four-Chain Derivatives of Due Ferri, DFtet analogues

The ultimate goal of our work with diiron proteins is to understand how the natural proteins direct the reactivity of the diiron cofactor, to give rise to so many different enzymes with properties ranging from ferroxidases, to mono-oxygenases, to desaturases. Even with our very simple DF2t protein, there appeared to be many possible mutations that we wanted to examine. Ideally, we wanted to examine six residues that define the active-site access channel, plus the residues responsible for the primary and second-shell ligands. Clearly, we needed a system for the combinatorial generation of pure mutants in sufficient quantities for biochemical measurements. Towards this end, we engineered a four-chain heterotetramer. First, an A₂B₂ heterotetramer was generated.49 By combining n variants of A and m of B, such a system would generate n · m variants (e.g., 10 variants each of A and B would provide 100 combinations). We next designed an A_aA_bB₂ heterotetramer, in which the two A chains were distinct.50 Ten variants of each of these three chains can now generate 1000 different peptides.

The A₂B₂ heterotetramer (designated DFtet A₂B₂) was designed to be an antiparallel coiled coil (Figure 9). The helices of DFtet were extended relative to those of DF1 (33 residues in DFtet vs. 24 residues in DF1) to increase the stability of the DFtet system by increasing the size of the hydrophobic core. Next, side chains were placed onto the A and B helices, resulting in an A₂B₂ heterotetramer with C₂ symmetry. The residues within a 12-residue region of the binding site were essentially identical to the corresponding residues of DF2.

Figure 9. Helical-wheel diagram of DFtet A₂B₂ (left), and computer model and sequence of the DFtet A_aA_bB₂ protein (right). The model of DFtet A_aA_bB₂ shows the electrostatic interactions that stabilize the A_aA_b interface. Also, below are the sequences of each subunit, with the changes (from DFtet A) in A_a and A_b underlined.

Residues at the remaining a and d positions were modeled as leucine because this side chain effectively filled the interior volume of the bundle. The nature of the residues at the remaining “e”, “g”, “b,” and “c” positions were chosen to specifically stabilize only one of the possible topologies for an antiparallel A₂B₂ heterotetramer. For this work, a simplified energy function, which considers only the charge of the residues, was used. The target function was selected for the sequence with the largest energy gap between the desired structure and an alternatively folded structure, rather than with the aim to simply minimize the energy of the desired conformation. This represented the first example in which “negative design” against alternatively folded structures had been explicitly coded and experimentally tested. The resulting proteins were found to assemble into a helical tetramer with considerable thermodynamic stability. It bound Co(II), Zn(II), and Fe(II) in the expected stoichiometry, and an analogue of DFtet has been crystallized, and structure determination is in progress.

The sequence of DFtet A₂B₂ was further modified to create DFtet A_aA_bB₂. In this design, one helix–helix interface was redesigned to provide high specificity for the desired three-component assembly. In particular, Glu, Lys, and Arg residues were introduced such that they would be able to interact favorably in the desired 3-component hetero-oligomer (Figure 9). CD spectroscopy, size exclusion chromatography, and analytical ultracentrifugation showed that the desired tetramer formed when the three components were combined in the proper stoichiometry (one A_a and A_b per two moles of B). As will be described below, it has been possible to evolve this system into a phenol oxidase.

DFsc

While the DFtet family of peptides provides excellent systems to screen for potential activities, we nevertheless wished to obtain a single-chain version of the protein that would have an unambiguous three-dimensional structure. Then, mutations identified with the DFtet systems could be cassetted into the single-chain protein to facilitate detailed kinetic and catalytic characterization. We therefore designed a 114-residue protein, beginning with the helical backbone of the DFtet peptides52 (Figure 10). The identities of 26 residues were predetermined, including the primary and secondary ligands in the active site, residues involved in active-site accessibility, and the turn sequences (which were based on database analyses). The remaining 88 amino acids were determined using a side-chain repacking algorithm written by Saven and coworkers termed scads (statistical computationally assisted designed strategy).64 This algorithm is based upon a recently developed statistical theory of protein sequences. Rather than sampling sequences, the theory directly provides the site-specific amino acid probabilities, which are then used to guide sequence design. The resulting sequence (DFsc) expresses well in E. coli and is highly soluble (at least 2.5 mM). DFsc stoichiometrically binds a variety of divalent metal ions, including Zn(II), Co(II), Fe(II), and Mn(II), with μM affinities. ¹⁵N heteronuclear single quantum correlation (HSQC) NMR spectra of both the apo and Zn(II) proteins reveal narrow linewidths and excellent dispersion in the amide and aliphatic regions, and the NMR structure of di-Zn (II)-DFsc confirms the overall fold of the protein.

Figure 10. Computational model and sequence of DFsc. The primary ligands, second-shell Tyr ligands and metal ion are shown. The second-shell Asp ligands are not shown for clarity. The alignment of DF1 and DFsc sequence is shown. Helices 1 and 2 of DFsc are aligned with helix 1 of DF1; helices 3 and 4 of DFsc are aligned with helix 2 of DF1.

METAL INTERACTIONS

The metal-binding affinity and specificity of the DF family of proteins has been most extensively studied for DF2, DF2t, and DFsc, because of the excellent water solubility of these proteins. The stoichiometry and affinity of binding metal ions and protons to these proteins has been evaluated by measuring the thermodynamic stability of the protein as a function of pH and metal ion concentration. In the absence of metal ions and between pH 4 and 6, approximately two protons are taken up when these proteins fold.48, 52 This finding has also been observed in natural diiron proteins such as apo bacterioferritin58 and apo ribonucleotide reducatase,65 and is consistent with the expectation that folding would be most favorable when the sum of the formal charges on the metal-binding site is zero. As the pH is increased above 6, the apo-protein becomes increasingly unstable, and higher order aggregates are formed near pH 7.0 in DF2, but not in the more precisely engineered DF2t or DFsc.

The DF proteins bind to a variety of metal ions, which specifically stabilize the protein at pH 7.0. The binding affinities for first-row divalent transition metal ions showed a reasonable correlation with the Irving–Williams series for Mn²⁺, Co²⁺, Fe²⁺, and Zn²⁺. However, the affinities are substantially weaker for Cu²⁺and Ni²⁺ than expected from this relationship. Thus, the protein displays a significant degree of geometric specificity, showing the highest affinity for metal ions that are classically known to prefer octahedral or tetrahedral coordination geometries.

DF Structures

The structures of DF1, DF2t, and DF2 have been extensively studied by X-ray crystallography in complex with d10 metal ions Cd(II) and Zn(II). We have also extensively studied the structure of the DF1 protein complexed with Mn(II), as a mimic of the dimanganese catalases.11, 66–68 We also have solved structures of DF2 and DF2t in complex with Fe(II/III), although these studies have yet to be reported. Finally, we have reported quantum calculations of the ligand field and flexibility of DF1 in complex with Zn(II) and Mn(II).69, 70

A major motivation for these studies is to determine how the protein adjusts to changes in the charge of the cofactor, its coordination number, and the binding of exogenous ligands. Flexibility in the coordination mode of active-site ligands in natural diiron proteins have been structurally characterized in methane monooxygenase and the R2 subunit of ribonucleotide reductase.71–73 Three of the four liganding carboxylates in ribonucleotide reductase undergo geometric shifts (termed carboxylate shifts74), whereas one of the four is mobile in methane monooxygenase. This flexibility is not surprising as carboxylate ligands are able to adopt several coordination modes. Carboxylate-coordinating changes have been postulated to serve as an important catalytic redox mechanism as they allows for facile changes in the metal coordination spheres.74 Furthermore, different coordination modes may be important for stabilizing different metal ion oxidation states. Indeed, the active-site metal ions cycle between +2, +3, and/or +4 oxidation states during the catalytic hydroxylation of hydrocarbons by methane monooxygenase and during the generation of the tyrosyl radical in the R2 subunit of ribonucleotide reductase.75 The binding of exogenous ligands to the active site of the R2 subunit of ribonucleotide reductase has also been shown to result in a carboxylate shift.76

Di-Zn(II) and Di-Cd(II) Structures

We have crystallographically characterized the structures of di-Cd(II)-DF2,63 di-Zn(II)-DF1 (PDB: 1EC5),43 and di-Zn(II)-DF2t (PDB: 1MFT).63 Multiple dimers were observed in the unit cells of the DF1 and DF2 structures, providing a measure of the degree of variability of the protein active site. The metal sites in all of theses structures are approximately 5-coordinate: two Glu residues bridge the two metal ions, two Glu interact with individual metal ions in an approximately bidentate (chelating) interaction, and the two His residues interact with individual metal ions. There is also a vacant site trans to the His ligands (Figure 3). The inter-Zn(II) distance ranges from 3.9 to 4.0 Å, while the inter-Cd(II) distance is approximately 3.7 Å. The greatest degree of variability lies in the bridging Glu residues, whose carboxylate groups frequently lie in the same plane as the two metal ions, but in some structures are rotated out of this plane. Two orientations of the bridging carboxylates as observed in the two dimers of the crystal structure of DF1 are shown in Figure 11. The chelating carboxylates also show a high degree of variability in the angle of the carboxylate relative to the bound metal ion. By contrast, the orientation of the His residues remains essentially invariable in all the structures.

Figure 11. Superposition of the structure of the two independent dimers in di-Zn(II)-DF1.The two orientations of the bridging and chelating carboxylates observed in the two dimers of the crystal structure are shown. The side chains of 1 dimer is in gray while the other side chains of dimer 2 are green (carbon), red (oxygen), and blue (nitrogen).

The structure of the di-Zn(II) site and occurrence of carboxylate shifts in DF1 has been investigated by first principles (Car-Parrinello)69 and hybrid quantum mechanical/molecular modeling (QM/MM) molecular dynamic simulations.70 These studies indicated that a carboxylate shift from chelated to monodentate occurs during the simulations at least on one of the two bridging glutamates. The calculations also showed that the second-shell interactions contribute significantly to the structural stability of the active site, and that the bulk solvent water molecules play a critical role in fine-tuning the dynamics of the system.

Di-Mn(II) Structures

The metal ion-binding sites of Mn(II) catalases are highly similar to the active sites in diiron proteins. We therefore have examined the structural properties of the di-Mn(II) forms of DF1 and two variants of this protein in which the bulk of the residue at the “d” position immediately adjacent to the metal ion site was systematically varied. In the original DF1, this residue is Leu, and it was converted to Ala (DF1L13A, PDB: 1JMO)62 and Gly (DF1L13G, PDB: 1LT1)77 in successive mutants. Thus, these variants allowed a systematic exploration of the role of the environment on the metal-binding site. The di-Mn(II) site in native DF1 is highly similar to that of the di-Zn(II)-DF1, although the limited resolution of the crystals precluded an in-depth analysis of the geometry.

The crystal structure of the di-Mn(II) form of L13A-DF1 was solved to a resolution of 1.7 Å.62 The asymmetric unit of L13A di-Mn(II)-DF1 contains three dimers, whose overall tertiary structures are nearly identical to that of di-Zn(II)-DF1 (RMSD of about 0.7 Å for the backbone atoms), with the greatest differences being located at the helix termini and the loops. Furthermore, the designed second-shell interactions, involving hydrogen bonds to the liganding side chains, were also observed. As intended in the design, the dimetal site lies at the bottom of a deep, water-filled pocket. In addition to the active site Mn(II) ions and protein side chains, a DMSO molecule lies in a depression formed between the bridging ligands in all three dimers of the asymmetric unit (the crystallization buffer contained 10% DMSO). In addition to DMSO, a number of ordered water molecules fill the channel formed by the leucine replacements in DF1.

The crystal structure of L13G, solved to a resolution of 1.9 Å,77 has an even larger and more polar opening in its active site. It crystallized with four dimers in the asymmetric unit. In three of these dimers, a water molecule bridges between the metal ions (Figure 12), in a manner analogous to the di-Mn(II) form of the R2 subunit of ribonucleotide reductase.60 In the fourth dimer, two solvent molecules are bound trans to the His ligands. This finding was particularly interesting, because the two structures resemble putative intermediates in the reaction mechanism of manganese catalases.78 The major difference in the two sites was an increase of 0.6 Å in the metal–metal distance in the di-aqua vs. the bridged structure.

Figure 12. Electron density map of the dinuclear metal binding site of di-Mn-DF1–L13G. (A) 2Fo-Fc electron-density maps (contour levels are 1.5σ) in the AB dimer showing the water molecule bridging the Mn²⁺ ions (B) 2Fo-Fc electron-density maps (contour levels are 1.5σ) in the GH dimer showing the dihydroxy complex of di-Mn-DF1–L13G. These figures were generated with Bobscript.94

A rigid body shift of the helices accompanied this change in the metal ligation. While the C-terminal helices (2 and 2′) of the four crystallographic dimers are virtually invariant among the structures, the two copies of helix 1 undergo a shift in opposite directions (approximately 0.7 Å) along the Z-axis, away from the metal-binding site in the dihydroxy structure (Figure 12). Indeed, this shift increases the length of the metal–metal distance. This sliding-helix mechanism may also occur in natural metalloproteins, in order to accommodate changes in their coordination environment. However, such motions have not been observed in the parent natural metalloproteins, probably because it may be more difficult to be observed in a large complex molecule.

Although the DF1 and DF2 variants bind Mn(II), they fail to catalyze the Mn(II) catalase reaction to an appreciable rate, because they are kinetically and/or thermodynamically very stable in the Mn(II) state, whereas the manganese catalases are able to facilely transition between the Mn(II) and Mn(III) oxidation states by alternately oxidizing and reducing HOOH to O₂ and water. Examination of the structures of DF1 and its variants, relative to the manganese catalases,11, 68 suggests a structural reason for this disparity. The manganese catalases lack one of the two bridging residues found in the DF class of proteins and natural diiron proteins. Instead, the protein recruits a second bridging solvent molecule, which can assist in shuttling protons in response to changes in the oxidation state of the metal ion and might additionally specifically stabilize the Mn(III) oxidation state. Furthermore, the DF proteins lack general acid/base groups in the vicinity of the active site that might assist in protonation and deprotonation of the peroxide moiety. Thus, it should be of considerable interest to determine how the addition of such interacting groups may affect activity.

Interactions with Fe(II/III)

A number of variants of DF1 bind to Fe(II) and catalyze the oxidation to ferric ions under aerobic conditions (the ferroxidase reaction). These variants have relatively open active sites, and include DF1 L13A, L13G, as well as DF2 and DF2t (which have Ala at the position analogous to L13 in DF1). The final product of the reaction appears to be an oxo-bridged diferric species (Scheme 1), as is observed for a variety of diiron proteins.79–82 This assignment is based on the presence of strong ligand-to-metal charge transfer bands in the visible spectrum of the product, and preliminary crystallographic and spectroscopic analyses are consistent with this assignment.

An analysis of a series of analogues of DFtet showed that the rate of formation of oxo-bridged species correlates with the bulk of the residues lining the active-site channel, as will be discussed below in more detail. Also, given the fact that that there is sufficient solvent access to the dimetal center, some forms of the diferric proteins are also able to bind exogenous ligands such as phenol and acetate. Interestingly, some variants of DFtet and DFsc with active sites of intermediate accessibility form a long-lived intermediate, tentatively assigned to the diferric peroxo species (Scheme 1). This finding is of considerable interest, because this is a transient intermediate that has either been directly observed or proposed to exist in the reaction mechanism of all natural diiron enzymes.83–87 Generally, this species can be studied in small molecule models only at low temperature and in the absence of aqueous solvents. The availability of model proteins that form a related intermediate that is stable at room temperature for hours should provide attractive possibilities for examining the chemistry of this species.

INTRODUCING A CATALYTIC SITE INTO DFTET

Recently, we have designed a phenol oxidase, by introducing a substrate-binding site adjacent to the metal ion in DFtet.51 The intended reaction mechanism (Scheme 2) involved the use of O₂ to oxidize the diferrous protein to a diferric species. The diferric protein then reacts with 4-amino-phenol (4AP), producing benzoquinone monoimine. Then the reduced diferrous form is oxidized by O₂, thereby initiating another catalytic cycle. The released quinone mono-imine product is quenched in solution and spectroscopically detected using a reaction [Eq. (4) in Scheme 2], whose kinetics are rapid relative to the protein-catalyzed reactions.88

To accomplish this reaction, it was necessary to carve out a site adjacent to the dimetal site, which could accommodate the substrate. The original DF1 protein had a pair of Leu residues at positions 9 and 13 (and 9′ and 13′ of neighboring helices), which entirely blocked access to the metal ion site (Figure 13). As described above, by decreasing the bulk of this residue with Ala or Gly variants we were able to induce the formation of a cavity just above the dimetal site (Figure 13). However, to allow tighter binding of 4AP it appeared that it would be necessary to simultaneously mutate Leu9 to a smaller residue for rapid and efficient substrate access.

Figure 13. Structure analogues of DF1 in which the residues that define the active-site access and environment are varied. Panel 1: wild-type DF1 with position Leu 9 and Leu 13. Panel 2: Leu 9 was changed to a smaller residue Ala. Panel 3: both Leu were changed to Ala residues. Panel 4: position 13 was changed to an even smaller residue Gly. The residue at position 13 is in purple, the residue at position 9 in white and the metal ions are shown in yellow.

The DFtet series of proteins was chosen to explore the introduction of catalytic activity, because of the ease of making a variety of mutants of the protein. The interior of the symmetrical DFtet was sculpted for creating a pocket capable of binding a substrate, 4-aminophenol (4AP). Modeling suggested that when this molecule is inserted in the active site of the original DFtet A₂B₂, and modeled such that its phenolic oxygen bridges the metal ions, it makes unfavorable contacts with Leu-15 and Ala-19 (corresponding to positions 9 and 13, respectively, of DF1) The steric bulk of these residues was therefore reduced in variants in which Ala-19 is changed to Gly and Leu-15 is changed to either Ala or Gly in both of the A chains.

A number of asymmetrical variants of DFtet A_aA_bB₂ and symmetrical variants of DFtet-A₂B₂ G₄-DFtet were prepared and evaluated. The variants of DFtet A₂B₂ are designated G4-DFtet (in which Leu-15 and Ala-19 of both A chains were substituted with Gly), L₂G₂-DFtet (in which Leu-15 was retained and Ala-19 was changed to Gly), A₂G₂-DFtet, and G₂A₂-DFtet (which have Gly or Ala at the indicated positions). First, all the variants were screened for their ability to react with Fe(II) and O₂ [Eq. (1), Scheme 2]. The variant with the fewest steric restrictions, namely G₄-DFtet, showed the most rapid rate of oxidation and formation of the oxo-species with no detectable intermediates. The same variant showed the greatest binding ability toward phenol, which binds to the diferric site.

Finally, we measured the two-electron oxidation of 4-aminophenol to the corresponding quinone monoimine, catalyzed by the proteins in atmospheric O₂. Again, G₄-DFtet showed a ≈1000-fold rate enhancement, relative to the background reaction when the initial rates of the reaction in the presence and absence of the protein were compared. The G₄-DFtet catalyzed this reaction for at least 100 turnovers. Changing either of the Gly residues at positions 19 or 15 to Ala gave a protein whose rate was decreased between 2.5- and 5-fold. All the results on DFtet variant clearly showed that the catalytic efficiency is sensitive to changes in size of a methyl group in the protein, illustrating the specificity of the design. The data above demonstrates the promise of a combinatorial approach for identifying potential candidates for defined applications. Having tested the potential of our system, the next step is to prepare a library of the variants, with the active-site cavity designed for a specific function.

THE THERMODYNAMIC COST OF FUNCTION

The introduction of an active-site cavity was expected to destabilize the folding of the proteins. We therefore conducted an analysis of the thermodynamics of folding. Three mutants were analyzed—DF1, DF1-L13A, and DF1-L13G—by chemical denaturation with Gdn · HCl.59 DF1 has exceptional stability, with a dissociation constant of approximately 0.001 femtomolar (fM); the corresponding values for DF1-L13A and DF1-L13G were 100 fM and 0.6 nM fM, respectively. The difference in stability associated with mutation of Leu-13 to Ala [ΔΔG°] was found to be 5.6 kcal/mol dimer (2.8 kcal/mol monomer). This value is within the range expected for mutating a single buried Leu to an Ala in a native protein.89, 90 The mutation of position 13 from Ala to Gly resulted in a further destabilization of 5.2 kcal/mol (2.6 kcal/mol monomer), reflecting contributions from both a decrease in the helix propensity of Ala relative to Gly [≈1kcal/mol monomer91, 92] and a decreased hydrophobic driving force [≈1.3 kcal/mol monomer90]. Even though a large thermodynamic price is paid by opening up the active-site access for function, the extreme stability of DF1 provides adequate stability to compensate for the mutation. Similar studies on functional variants of DFtet and DFsc have shown similar results (unpublished results). These studies illustrate the importance of using a very stable and highly optimized protein for functional design, which can withstand multiple changes to the sequence while retaining a folded form.

CONCLUSIONS

The DF family of proteins provides a highly stable and simple scaffold for examining how the sequence and structure of the protein affect the reactivity of the cofactor. Using this scaffold, it is possible to examine how the structure of the protein affects the specificity and affinity for metal ion binding, the electrochemical midpoint potential of the metal ions, and the chemical reactivity of the dimetal cofactor. Clearly, we are only beginning to see the full potential of minimal models for metalloproteins, and expect much more to be accomplished not only in the design of diiron proteins but also for a variety of other motifs.

Acknowledgements

The authors thank Herschel Wade for helpful discussion and help with figure preparation. This work was supported by the NIH (GM54616) and by the Italian Ministry of the University and Scientific Research (PRIN 20003037580).