Flexibility of the metal-binding region in apo-cupredoxins

Abstract

Protein-mediated electron transfer is an essential event in many biochemical processes. Efficient electron transfer requires the reorganization energy of the redox event to be minimized, which is ensured by the presence of rigid donor and acceptor sites. Electron transfer copper sites are present in the ubiquitous cupredoxin fold, able to bind one or two copper ions. The low reorganization energy in these metal centers has been accounted for by assuming that the protein scaffold creates an entatic/rack-induced state, which gives rise to a rigid environment by means of a preformed metal chelating site. However, this notion is incompatible with the need for an exposed metal-binding site and protein–protein interactions enabling metallochaperone-mediated assembly of the copper site. Here we report an NMR study that reveals a high degree of structural heterogeneity in the metal-binding region of the nonmetallated Cu_A-binding cupredoxin domain, arising from microsecond to second dynamics that are quenched upon metal binding. We also report similar dynamic features in apo-azurin, a paradigmatic blue copper protein, suggesting a general behavior. These findings reveal that the entatic/rack-induced state, governing the features of the metal center in the copper-loaded protein, does not require a preformed metal-binding site. Instead, metal binding is a major contributor to the rigidity of electron transfer copper centers. These results reconcile the seemingly contradictory requirements of a rigid, occluded center for electron transfer, and an accessible, dynamic site required for in vivo copper uptake.

Long-range electron transfer (ET) in proteins is a key chemical event in many essential biological processes such as cellular respiration and photosynthesis (1–3). According to Marcus’ semiclassical theory, the outstanding efficiency of these processes is based on the maximization of the superexchange coupling between donor and acceptor and the minimization of the reorganization energy, given the low driving forces found in most biological systems (4–6). Thus, with driving forces as low as 0.1 eV and distances larger than 10 Å, efficient long-range electron transfer is only possible if the nuclear reorganization energy of the reactants is below 1 eV (4, 7).

Transition metal ions such as copper and iron are ubiquitous in ET chains because their redox potentials and electronic structures can be tuned by the protein environment to match the requirements of different biological redox events. Cycling between the coordination geometries preferred by these metals in each redox state would in principle lead to high reorganization energies (7). Electron transfer iron centers avoid this by using rigid cofactors such as iron-sulfur clusters and heme groups (8–10). Instead, copper ions in ET centers are only bound to protein residues, as observed in type 1 (blue) copper and Cu_A centers; and thus the protein fold around the metal ion is expected to impart rigidity to these otherwise flexible metal centers (10–12). Particularly, outer-sphere coordination involving hydrogen bonding networks has been proposed as responsible for the low reorganization energies observed in copper centers in proteins (13, 14). The strain introduced by the protein fold is then predicted to be responsible for the unusual functional and spectroscopic features observed in type 1 copper and Cu_A centers (15). This scenario, known as the entatic/rack-induced state hypothesis, was originally formulated for protein-bound cofactors by Lumry and Eyring in 1954 (16) and then extended to electron transfer metalloproteins by Malmström and Williams in the late 1960s (17, 18). This concept has been matter of debate along many years (7, 15, 19–26). In particular, Ryde et al. have reported in vacuum quantum calculations suggesting the absence of any strain in the geometry adopted by the copper in the protein framework (21, 22).

The entatic/rack-induced state hypothesis applied to ET copper proteins was strongly supported by several X-ray structures of type 1 copper proteins revealing that the conformation of the copper ligands in the metal-depleted (apo) form was identical to that found in the holoproteins (27–31). These results prompted the idea that the protein fold determines the coordination geometry of the metal center by creating a preformed chelating site with very little flexibility, thus minimizing conformational changes that would normally be exhibited during Cu(II)/Cu(I) redox cycling (19). This idea is reinforced by the fact that all ET copper centers are bound to a rigid Greek-key β-barrel domain, known as the cupredoxin fold (31–34). Instead, iron-sulfur and heme centers are found in different folding motifs, with diverse intrinsic mobilities (35–37).

The notion of a preformed rigid binding site is conflicting with the finding that copper centers are loaded by specific metallochaperones (38, 39). The concept of metallochaperones is relatively new and it has replaced the idea that metal uptake depends entirely on the chelating properties of the apoprotein. Several copper chaperones have been identified and characterized so far, requiring specific protein–protein interactions that can be metal-mediated in some cases (40). In the latter situation, the metal ion is simultaneously bound to ligands from both the metallochaperone and the target protein during the transfer step (40). In any case, the assumption of a highly rigid metal site, which should also be hidden from the bulk solvent to prevent unwanted side reactions, does not favor metal delivery by another protein. Thus, the demands for efficient electron transfer and in vivo metallation are hard to reconcile given the current knowledge in the field.

We have recently elucidated the chaperone-mediated mechanism of copper uptake by the Cu_A-binding cupredoxin domain in Thermus thermophilus ba₃ cytochrome c oxidase (41). In that work, we observed that the NMR spectra of the apo and metallated forms of this protein show significant differences, suggesting that there might be nonnegligible structural perturbations upon metal binding, consistent with the finding that copper uptake is mediated by a metallochaperone (Fig. S1). Here we report a detailed NMR study showing the solution structure and dynamic features of this cupredoxin domain in the nonmetallated form. We have found that not only is the metal-binding site disordered in the apoprotein but it also exhibits significant dynamics in the microsecond to second timescale. Moreover, to prove that our results can be extended to other cupredoxins, we report an NMR-based study of Pseudomonas aeruginosa apo-azurin, which has been considered as the paradigm of blue copper proteins and long-range ET (42–44). Also in this case, we show there is significant dynamics in the metal-binding region only in the absence of the copper ion. Thus, our findings indicate that the rigid copper center needed for efficient long-range ET is not preformed in the nonmetallated cupredoxin. Instead, metal binding does significantly contribute to the rigidity of these centers and the metal-binding region in the apoprotein is flexible enough to allow in vivo metallation.

Results

The Cu_A-Binding Cupredoxin Domain.

Subunit II of T. thermophilus ba₃ oxidase consists of a soluble Cu_A-binding domain anchored to the membrane through an N-terminal transmembrane helix (32, 45). This domain adopts a cupredoxin fold similar to the one found in type 1 copper proteins, with a longer loop that allows binding of an additional metal ion (12). The two copper ions of the Cu_A site are bridged by two cysteine ligands (Cys149 and Cys153), giving rise to a rigid Cu₂S₂ core. Each copper ion is also coordinated to a terminal histidine residue (His114 and His157) and a weak axial ligand: a methionine sulfur (Met160) and a backbone carbonyl (Gln151), respectively. The soluble fragment herein studied comprises only the soluble cupredoxin domain of subunit II, where the transmembrane helix is not part of the purified protein. This fragment retains the structure observed in the whole oxidase (32, 45) and it is competent for ET with its biological redox partner (46, 47). Given that residues 1–42 are absent in our protein construct, residue numbering starts at 43 in the results presented here, following that employed in the X-ray structure of oxidized holoCu_A [Protein Data Bank (PDB) ID 2CUA] (32).*

Resonance Assignment and Solution Structure of apoCu_A.

¹H, ¹³C, and ¹⁵N resonances were assigned for most residues in the nonmetallated form of the soluble domain of T. thermophilus ba₃ oxidase subunit II (apoCu_A, hereafter) (Table S1). Resonances corresponding to eight non-proline residues were missing in the ¹H, ¹⁵N-heteronuclear single quantum coherence (HSQC) spectrum of apoCu_A, whereas 19 residues presented duplicated NH cross-peaks (see below). Five of the residues with missing resonances (Cys149-Gln151, His157, and Gln158) are located at the loop bearing five out of the six copper ligands (ligand loop, hereafter) whereas the remaining correspond to His114 (the N-terminal histidine coordinated to one of the copper ions), its adjacent residue Gly115, and the N-terminal Met43. These eight residues are thus involved in an exchange process whose frequency lies right on an intermediate timescale that broadens signals beyond detection. Among residues with duplicated resonances in the ¹H, ¹⁵N-HSQC spectrum, five are located in the N-terminal region (Val44, Ile45, Ala47, Lys49, and Leu50) and duplication here arises from the cis-trans isomerization of Pro46 in a slow regime, as reported for the reduced holoprotein [BioMagResBank (BMRB) entry 5819; ref. 48]. Most of the remaining residues with duplicated correlations are located around the metal-binding site (Ala85, Ala87, Tyr90, Val112, Ile113, His117, Gly120, Asn124, Gly154, Gly156, Met160-Thr163). Comparison of the assignments obtained for apoCu_A with those available for the reduced holoprotein revealed that most chemical shift perturbations also map close to the metal-binding site (Fig. 1 and Fig. S1).

Fig. 1.

NMR-based solution structure of T. thermophilus apoCu_A. (A) Cartoon representation of the X-ray structure of oxidized holoCu_A (from PDB ID 2CUA; ref. 32). (B) Ribbon representation of the best 20 out of 80 structures calculated by CYANA for apoCu_A in solution (this work, PDB ID 2LLN). Regions with significant chemical shift perturbations (Eq. S1) between apo- and holoCu_A and structural disorder in the calculated structure of the apoprotein are highlighted in blue (ligand loop, residues 149–161), and orange (residues 86–90).

The calculated three-dimensional structure of apoCu_A is well defined by 2,147 (including 879 long-range) NOE-based distance restraints and 215 experimental torsion angle restraints (Fig. S2 and Table S2). The atomic rmsd values for the 20 models in the refined structure are 1.0 ± 0.2 and 1.4 ± 0.1 Å for the backbone and all heavy atoms, respectively. Considering only residues from the rigid β-barrel domain (residues 53–148 and 162–168), the rmsd values for backbone and all heavy atoms are 0.49 ± 0.07 and 1.01 ± 0.08 Å, respectively.

The apoprotein in solution adopts a β-barrel fold where most regions are well-defined except the N terminus and loops 86–90 and 149–161 (ligand loop) (Fig. 1 and Fig. S3). The disorder observed at the ligand loop might be due to the lack of experimental constraints (Fig. S2) and/or true mobility of this region in the absence of the metal ions. Indeed, the low number of constraints in this region parallels the missing correlations in the ¹H, ¹⁵N-HSQC, suggestive of some degree of flexibility (see below). It is then clear that the polypeptide presents a similar Greek-key β-barrel fold both in the apo- and copper-loaded forms, and that the structural perturbations between both forms of the protein are confined to the metal-binding region, with the ligand loop showing the highest degree of disorder.

Dynamics of apoCu_A and Reduced holoCu_A.

Subnanosecond timescale.

The dynamics of apo- and reduced holoCu_A in the pico-to-nanosecond timescale was studied by measuring ¹⁵N R₁ and R₂ and ¹H-¹⁵N heteronuclear NOE. Average correlation times of 8.3 ± 0.1 ns (apoCu_A) and 8.2 ± 0.1 ns (holoCu_A) were estimated and used for relaxation data analysis with the model-free approach (49–51). Order parameter values (S²) for apo- and holoCu_A are remarkably high along most of the sequence, averaging 0.9 ± 0.1 in both proteins and confirming the high level of rigidity of the β-barrel (Fig. S4). As expected, the order parameter values for both protein forms drop below 0.6 at the N terminus, due to the high mobility usually observed in this region. S² values lower than 0.6 are also observed for residues 154–156 in apoCu_A. However, a slight decrease in S² values is also found in this region for holoCu_A. Overall, these data do not reveal significantly different dynamic features in the pico-to-nanosecond timescale between the apo and holo forms of the Cu_A-binding domain.

Microsecond to millisecond timescale.

Protein motions in the micro-to-millisecond timescale were probed by constant-time Carr–Purcell–Meiboom–Gill (CPMG) relaxation dispersion experiments (52, 53). Fig. 2A shows the ΔR₂ profiles obtained for apoCu_A and reduced holoCu_A, where ΔR₂ is the difference between R₂ values measured at CPMG frequencies of 33 and 966 Hz, validated by fitting of the dispersion curves to Eq. S2 (Fig. 2, Inset and Fig. S5). The ΔR₂ profile for apoCu_A shows two regions experiencing significant exchange in comparison to the holoprotein. Both regions map close to the metal-binding site, as shown in Fig. 3A. Thus, residues in the metal-binding region exhibit significant micro-to-millisecond dynamics in the apoprotein, and this flexibility is lost upon binding of the metal ions.

Fig. 2.

Protein dynamics and compactness of apoCu_A (black squares) and reduced holoCu_A (gray triangles). (A) ΔR₂ profiles showing dynamics in the micro-to-millisecond timescale. Fits of experimental data to Eq. S2 for some apoCu_A residues are shown as examples in the inset of the figure (Gln91, circles; Met160, triangles; and Gly162, squares, showing significant relaxation dispersion; and Ala101, diamonds, located in a region showing no exchange). (B) λ_NOE profiles showing proton density around the backbone amides.

Fig. 3.

Mapping of the dynamic features in apoCu_A. Residues experiencing dynamics in the absence of the copper ions (i.e., in apoCu_A) are highlighted in the structure of holoCu_A (PDB ID 2CUA, ref. 32, used to evidence the proximity of these residues to the metal ions). (A) Dynamics in the micro-to-millisecond timescale. Residues with ΔR₂ < 1.3 s^-1, ΔR₂ > 1.3 s^-1, and missing residues are shown in blue, orange, and gray, respectively. (B) Slow dynamics. Residues with zero (missing), one or two (duplicated) correlations in the ¹H, ¹⁵N-HSQC spectrum of apoCu_A are shown in red, gray, and yellow, respectively.

Protein compactness.

Protein compactness was studied by a ¹H, ¹⁵N heterogeneity-band-selective optimized-flip-angle short-transient-heteronuclear multiple quantum coherence (HET-SOFAST-HMQC) experiment with irradiation of aliphatic protons. The λ_NOE parameter calculated from this experiment reports on the proton density around the probed amide, with low λ_NOE values (< 0.3) indicating well-structured compact segments and higher values indicating a less compact environment (which might be due in turn to enhanced dynamics) (54). Fig. 2B shows the λ_NOE profiles obtained for apo- and reduced holoCu_A. Both protein forms present a similar global compactness along most of the sequence, except for a significant difference observed at the ligand loop, where the higher λ_NOE values obtained for the apoprotein suggest a less compact structure in the absence of the copper ions.

Millisecond to second timescale.

The absence (for eight non-proline residues) and duplication (for 19 residues) of NH cross-peaks in the ¹H, ¹⁵N-HSQC spectrum of apoCu_A reveals chemical exchange in the intermediate and slow regimes on the chemical shift timescale, respectively. As detailed above, most of these residues correspond to copper ligands and/or map to the surroundings of the metal-binding site. We note that only one correlation is observed for all these residues in the ¹H, ¹⁵N-HSQC spectrum of reduced holoCu_A (BMRB entry 5819). Thus, slow dynamic features on the copper ligands and surrounding residues are only observed in the absence of the copper ions (Fig. 3B).

Dynamics of apo-azurin.

In order to extend these results to other cupredoxins, we have studied the solution dynamics of P. aeruginosa apo-azurin. The blue copper protein azurin has been extensively studied as a paradigmatic protein for electron transfer (42–44), and Cu(I)-azurin has been the subject of several NMR studies (55–58). Instead, apo-azurin has not been characterized by NMR.

Backbone ¹H and ¹⁵N resonances were assigned for most residues of P. aeruginosa apo-azurin. Resonances corresponding to 17 non-proline residues were missing in the ¹H, ¹⁵N-HSQC spectrum, indicating they are involved in an intermediate exchange process on the chemical shift timescale. These residues are Ala1-Cys3, Gly9-Asp11, Asn18, His35-Leu39, Gly45, His46, Gly88-Glu91, most of them mapping to the surroundings of the metal-binding site (Fig. 4B). This behavior contrasts with that of Cu(I)-azurin, whose ¹H, ¹⁵N-HSQC spectrum shows one cross-peak for every non-proline residue in the protein (except the N-terminal Ala1) (56), indicating that this slow dynamic process takes place only in the nonmetallated form, in agreement with the findings on apoCu_A. Moreover, most chemical shift perturbations resulting from metal removal in azurin cluster close to the copper center (Fig. 4A and Fig. S6), again resembling the situation met in the Cu_A-binding domain.

Fig. 4.

NMR results on apo-azurin. Residues showing significant chemical shift perturbations or enhanced dynamic features are mapped on the tridimensional structure of Cu(II)-azurin (PDB ID 4AZU; ref. 33). (A) Chemical shift perturbations between apo-azurin and Cu(I)-azurin backbone amide resonances. Residues with chemical shift perturbations (CSP) < 0.2, 0.2 < CSP < 0.5, CSP > 0.5 are shown in blue, yellow, and red, respectively. Residues with missing NH correlations are shown in gray. (B) Microsecond to second dynamics of apo-azurin. Residues with ΔR₂ < 2 s^-1, ΔR₂ > 2 s^-1 are shown in blue and orange, respectively. Residues with missing correlations in the ¹H, ¹⁵N-HSQC spectrum are shown in red. In both representations, prolines are shown in gray. (C) ΔR₂ profiles showing dynamics in the micro-to-millisecond timescale. Fits of experimental data to Eq. S2 for some apo-azurin residues are shown as examples in the inset of the figure (Ile87, circles; Met44, squares, showing significant relaxation dispersion; and Thr61, triangles, located in a region showing no exchange).

Protein dynamics in the micro-to-millisecond timescale were evaluated by relaxation dispersion experiments in apo-azurin. As observed in Fig. 4C, the ΔR₂ profile, validated by fitting of the dispersion curves to Eq. S2 (Fig. S7), shows three defined regions in the micro-to-millisecond timescale. Residues in these regions are next to residues with missing correlations in the ¹H, ¹⁵N-HSQC of apo-azurin, thus evidencing a common exchange process. Comparison of these data with those reported by Orekhov and coworkers for Cu(I)-azurin (55, 57) reveals an increased mobility at the metal-binding region in the absence of the copper ion.

Discussion

In this work, we present an NMR study on apoCu_A, revealing that the metal-binding region (mainly the ligand loop) shows a disordered and less compact structure in the apoprotein compared to the holo form (Figs. 1 and 2B) and that this disorder, in turn, is due to a dynamic behavior of this region in the microsecond to second timescale (Figs. 2A and 3). The blue copper protein azurin displays similar dynamic features in its apo form which also map to the metal-binding region (Fig. 4). The absence of such dynamic features both in holoCu_A and Cu(I)-azurin allows us to conclude that this region is significantly flexible in the apoproteins and that only metal binding quenches the dynamics of these loops in the cupredoxin fold. Given the different loop length and structure in Cu_A and azurin, these results strongly suggest that this behavior is a general feature of cupredoxins.

Dynamics of Apo-Cupredoxins and the Entatic/Rack-Induced State.

The entatic/rack-induced state in ET copper proteins refers to the organized protein structure around the metal-binding site that ensures a low reorganization energy for efficient electron transfer (7). There have been numerous reports and reviews supporting the entatic/rack-induced state hypothesis (7, 13, 19, 20, 26, 59). On the other hand, theoretical and experimental studies have questioned the role of the protein matrix in defining the geometry of the copper center (21, 22, 24, 25). X-ray studies on the apo form of several cupredoxins, showing no changes in the ligands’ conformations upon removal of the copper ion, were considered as experimental evidence of the entatic/rack-induced hypothesis for ET copper proteins (27–31). In addition, those results put forward the idea that the protein matrix creates a preformed rigid chelating site in the absence of the metal ions, thus entirely determining the geometry of the copper center (19).

The present NMR-based results provide compelling evidence of structural disorder, flexibility, and conformational exchange in the metal-binding region of the apo form of the Cu_A-binding cupredoxin domain. Flexibility and slow chemical exchange were also observed here for apo-azurin, supporting the extension of these findings to other cupredoxins and contrasting the mentioned X-ray studies. This discrepancy may be attributed to the fact that most of the reported structures of apo-cupredoxins were obtained by crystallization of the holoprotein and subsequent metal removal from the crystal by soaking with a chelating agent (28–30). Instead, direct crystallization of apo-azurin gave rise to some (minor) structural heterogeneity selectively on the copper ligands (27). However, analysis of the structure of apo-azurin does not reveal a significant increase in the B factors in the metal-binding region nor the lack of electron density, suggesting that the absence of dynamics in these cases is due to crystal packing.

We note that our results do not necessarily support nor refute the entatic/rack-induced state hypothesis because they apply to the nonmetallated protein and do not assess the strain the protein environment exerts on the Cu(I) and Cu(II) centers. Instead, our findings show that the entatic/rack-induced state, when operative on the metallated protein, does not require a preformed rigid chelating site. This conception is in line with the results reported by Wittung-Stafshede and coworkers, showing that a rigid, ET-functional blue copper center can be obtained from Cu(II) binding to unfolded azurin, suggesting that the constraints imposed by the protein to the geometry of the metal center do not require a preorganized binding site defined in the polypeptide (60–62).

Dynamics of Apo-Cupredoxins and in Vivo Copper Uptake.

The copper cargo in the Cu_A-binding cupredoxin domain of T. thermophilus oxidase is due to the action of a specific Cu(I)-metallochaperone (PCu_AC) (41). Similar copper-loading mechanisms have been proposed for the type 1 protein plastocyanin (63). Copper transfer mediated by chaperones generally takes milliseconds to seconds, overlapping with the fluxional timescales measured in this work for the metal-binding region of apoCu_A and apo-azurin (although no copper chaperones have been reported for azurin to date). This observation suggests that the observed flexibility of this region in the apo form of the protein allows the cupredoxin fold to sample different conformations, enabling copper transfer.

Our unique picture of a folded apo-cupredoxin consisting in a β-barrel with disordered flexible loops is fully consistent with the need of specific protein–protein interactions and defines an exposed, dynamic metal-binding site allowing for efficient metallochaperone-mediated copper binding in vivo. In turn, the fully conserved Greek-key β-barrel scaffold would be able to render a rigid and efficient ET site once the copper ions are bound. Our results thus allow us to reconcile the seemingly contradictory requirements of a rigid, occluded center for electron transfer, and an accessible, dynamic site for in vivo copper uptake. Nature has solved this problem by exploiting the cupredoxin fold, which is endowed at the same time with rigid and dynamic features that are finely tuned by the binding of the metal ions.

Materials and Methods

Protein Preparation.

Samples of azurin and apo- and holoCu_A proteins were uniformly labeled with ¹⁵N or ¹³C and ¹⁵N (Cambridge Isotope Laboratories, Inc.) and produced as described elsewhere (40, 47, 64). The apoCu_A protein was expressed and purified in the absence of copper, whereas azurin was purified as Cu(II)-azurin and the apoprotein was obtained by dialysis against 50 mM KCN, 200 mM Tris·HCl pH 8.5, followed by several dialysis steps without the complexing agent. Protein samples for NMR experiments were prepared in 100 mM phosphate buffer pH 7.0 for azurin and pH 6.0 for the Cu_A domain, adding 100 mM KCl in the case of the holoprotein or 2 mM DTT for the apoproteins in order to avoid oxidation of the cysteine ligands. Protein concentration was about 1 mM.

Nuclear Magnetic Resonance Spectroscopy.

NMR experiments were carried out on a 600 MHz Bruker Avance II Spectrometer equipped with a triple resonance inverse (TXI) probe head and on a 900 MHz Bruker Avance II Spectrometer equipped with a triple resonance inverse (TCI) cryoprobe. All experiments were carried out at 298 K using standard techniques, as described in SI Text. The obtained family of structures for apoCu_A is deposited at the Protein Data Bank under PDB ID 2LLN. Chemical shifts for all ¹H, ¹³C, and ¹⁵N nuclei assigned in this work are deposited at the Biological Magnetic Resonance Data Bank under accession numbers 18081 and 18254 for apoCu_A and apo-azurin, respectively. Resonance assignments for the reduced metallated proteins, holoCu_A and Cu(I)-azurin, were taken from the literature (48, 56) and transferred by us to the sample conditions used in this work.