The crystal structure of yeast copper thionein: The solution of a long-lasting enigma

Abstract

We report here the crystal structure of yeast copper thionein (Cu-MT), determined at 1.44-Å resolution. The Cu-MT structure shows the largest known oligonuclear Cu(I) thiolate cluster in biology, consisting of six trigonally and two digonally coordinated Cu(I) ions. This is at variance with the results from previous spectroscopic determinations, which were performed on MT samples containing seven rather than eight metal ions. The protein backbone has a random coil structure with the loops enfolding the copper cluster, which is located in a cleft where it is bound to 10 cysteine residues. The protein structure is somewhat different from that of Ag₇-MT and similar, but not identical, to that of Cu₇-MT. Besides the different structure of the metal cluster, the main differences lie in the cysteine topology and in the conformation of some portions of the backbone. The present structure suggests that Cu-MT, in addition to its role as a safe depository for copper ions in the cell, may play an active role in the delivery of copper to metal-free chaperones.

Our understanding of the physiological chemistry of metal ions is still at the beginning, despite the fact that, in recent years, progress has been made on the understanding of how metal ions are transported and stored within the prokaryotic and eukaryotic cells (1, 2). Among the essential transition metal ions, copper is characterized by high toxicity related to its redox chemistry, which is potentially able to damage any molecule in a cell (3, 4). Efforts from different laboratories are successfully unveiling the fascinating chemistry and biochemistry involved in copper homeostasis, transport, and regulation in prokaryotic and eukaryotic organisms (5-13). One of the still-missing pieces of the picture is related to the role of metallothioneins in general and of copper thionein (Cu-MT) in particular. The physiological role of these molecules is still a matter of debate (14, 15), but a role as depository for copper transfer into apo-copper proteins (16, 17) and apo-copper chaperones (1, 11, 12) involved in copper trafficking has been proposed. Several studies have also demonstrated that Cu-MT in yeast and mammals is related to copper detoxification (refs. 3, 15-18, and refs. therein). In both roles, Cu-MT should be able to bind copper strongly but also quickly and, furthermore, it should need, at least in the first role, to quickly release it when required. In this respect, the 3D structure of Cu-MT and especially the details of the Cu(I) arrangement within the copper cluster will provide a starting model to validate the different functional hypotheses.

Cu-MT contains 53 amino acids, of which 12 are cysteines and six to eight Cu(I) ions (15, 18-21). Crystallization attempts are as old as the protein's discovery in 1975 (22) and have always failed. Crystallization failures are usually not recorded in the literature, but one of us (U.W.) has a long list of personal communications in this respect.

In the absence of x-ray data, solution structures of increasing accuracy were provided by NMR (23-27). However, no structure of the copper cluster could be inferred. A step forward was provided by the finding that four residues at the N terminus and 13 residues at the C terminus were disordered in the NMR structures. Two cysteines located in the disordered C-terminal part were therefore confirmed as not being involved in copper coordination (27), as already suggested by extended x-ray absorption fine structure on mutants lacking these two cysteines (28). But the knowledge of the protein frame did not solve the problem of how the copper ions are loaded, stored, and released by Cu-MT.

A model for the copper cluster was suggested by the Ag₇-MT derivative, where the scalar couplings between the silver nuclei and the cysteine α protons did provide a cluster topology (23-26). In this model, 10 cysteines were found to be involved in metal coordination (Fig. 1A; ref. 23). However, the transferability of the structural information from silver to copper should not be taken for granted (27). Indeed, the protein backbones derived from the nonphysiological Ag₇-MT derivative (Fig. 1 A) and from the physiological Cu₇-MT derivative obtained from NMR data (Fig. 1B; refs. 25 and 26) were similar but not identical beyond the experimental error. Particularly, the cysteine sulfur cage hosting the seven copper ions was different from the cage hosting the seven silver ions (26) (Fig. 1).

Fig. 1.

Least-squares superposition of the present Cu₈-MT structure (cyan tube) with Ag₇-MT NMR model (green tube) (A) (25) and Cu₇-MT NMR model (red tube) (B) (26). The Cu₈-MT copper atoms are represented as cyan spheres, whereas the silver and copper atoms of the Ag₇-MT and Cu₇-MT NMR models are represented as green and red spheres, respectively. The cysteine side chains are also displayed.

Based on the NMR evidence of disordered N- and C-terminal residues, which could prevent the formation of crystals, a minimal construct was recently produced lacking the first 4 and the last 13 residues of yeast MT. An extensive NMR analysis showed that, whatever their arrangement, the seven copper ions are arranged identically within this minimal frame as they are in the native Cu₇-MT (27). After several attempts, small but suitable crystals for x-ray analysis could indeed be obtained from this construct, and diffraction data of good quality were measured at a synchrotron radiation source.

Materials and Methods

All chemicals were of analytical-grade quality or better. The protected amino acids were obtained from Nova Biochem, and the resin for peptide synthesis was from Rapp-Polymere (Tübingen, Germany). All other chemicals were purchased from Merck (Darmstadt, Germany).

Peptide Synthesis. The truncated form of yeast copper thionein was synthesized on an Eppendorf ECOSYN P solid-phase peptide synthesizer using a 9-fluorenylmethoxycarbonyl (Fmoc)-Asn(Trt)-PHB R resin (29). All amino acids were incorporated with the α amino functions protected by the Fmoc group. Side-chain functions were protected as tert-butyl esters (aspartic acid), tert-butyl ethers (serine), and the trityl derivatives (cysteine, histidine, and asparagine). Coupling was performed by using a 4-fold excess of protected amino acids and the coupling reagent 2-(1H-benzotriazole-1-yl)1,1,3,3-tetramethyluronium tetrafluoroborate (TBTU) plus 1.2 ml of diisopropylethylamine [DIEA; 12.5% solution in dimethylformamide (DMF)] over the resin loading. Before coupling the protected amino acids, the Fmoc groups were removed from the last amino acid of the growing fragment by using 25% piperidine in DMF. After cleavage of the N-terminal Fmoc group, the peptide was removed from the resin under simultaneous cleavage of the amino acid side-chain protecting groups incubating in a mixture of trifluoroacetic acid (TFA; 12 ml), ethanedithiol (0.6 ml), thioanisole (0.3 ml), anisole (0.3 ml), water (0.3 ml), and triisopropylsilane (0.1 ml) for 3 h. The mixture was filtered, washed with TFA, and the combined filtrates precipitated with anhydrous ether. The crude product was further purified by HPLC on a Nucleosil 100 C18 (7 μm) 250 × 10-mm column (Macherey & Nagel, Düren, Germany) by using a gradient from 10% to 90% solution B (solution A, 0.07% TFA/H₂O; solution B, 0.059% TFA in 80% CH₃CN) in 32 min. The elution was monitored at 214 nm. The polypeptide was assayed for purity by analytical HPLC, amino acid analysis, and electrospray ionization (ESI)-MS (Fig. 2).

Fig. 2.

Averaged (50 scans) positive-ion mode ESI mass spectrum acquired from the synthetic truncated form of yeast MT. Average molecular mass of 3,842.10 Da could be detected by the formation of the triple ([M + 3H]³⁺, m/z 1,281.6), and quadruple-charged molecular ions ([M + 4H]⁴⁺, m/z 961.5). The dried peptide was dissolved in 50% MeOH, 1% formic acid in water, and analyzed by using syringe pump infusion (5 μl per min flow rate) on an Esquire3000plus ion trap mass spectrometer (Bruker Daltonics) equipped with a standard ESI source. (Inset) The HPLC elution profile. The HPLC was run on a Nucleosil 100 C18 (7-μm) 250 × 10 mm column (Macherey & Nagel) by using a gradient from 10% to 90% solution B (solution A, 0.07% TFA/H₂O; solution B, 0.059% TFA in 80% CH₃CN) in 32 min. The elution was monitored at 214 nm. Besides the elution peak of the pure product, no other peaks are seen.

In Fig. 2 Inset, the HPLC elution profile is presented, revealing one single homogenous peak. For the ESI-MS measurement, the dried peptide was dissolved in 100 μl of 50% MeOH, 1% formic acid in water (vol/vol) and analyzed by using syringe pump infusion (5 μl/min flow rate). The ESI-MS experiment was carried out on an Esquire3000plus ion trap mass spectrometer (Bruker Daltonics, Bremen, Germany) equipped with a standard ESI source. Peptide ions were analyzed in the positive-ion mode. Dry gas (6 liters per min) temperature was set to 245°C, and electrospray voltage was set to -3,400 V. The main ESI-MS peak represents the 3-fold charged species ([M + 3H]³⁺) and appears at a molecular mass of 1,281.6 Da. The corresponding 4-fold charged species ([M + 4H]⁴⁺) was detected at a molecular mass of 961.5 Da. An average molecular mass was allocated at 3,842.1 Da (expected in theory, 3,843.22; monoisotopic, 3,840.34).

Reconstitution of the Apo-Peptide. For reconstitution, the freezedried metal free peptide was dissolved to a peptide concentration of 15 mg/ml in 400 mM potassium phosphate buffer, pH 7.4, containing 4% (vol/vol) 2-mercaptoethanol, and a 7-fold molar surplus of copper sulfate was added. Excessive 2-mercaptoethanol was expected to maintain copper in its monovalent state. After 30 min of incubation at 20°C, the holopeptide was purified and desalted on a water-equilibrated Sephadex G-25 column. Fluorescent fractions were pooled and lyophilized. The freezedried holopeptide was dissolved in deionized water containing 0.1% mercaptoethanol to a concentration of 1.2 mg/ml and stored at -80°C until utilization.

Crystallization Conditions. Crystallization trials on the truncated Cu-MT were performed by the sitting-drop method by using the protein solution described above. Potassium dihydrogen phosphate and lithium bromide were added to the drop to yield final concentrations of 25 and 1 mM, respectively. The drop was allowed to equilibrate by vapor diffusion against 1 ml of 3.9 M ammonium sulfate solution in a 24-well cell-culture plate at 20°C. Cubic crystals started to grow after ≈14 h.

The crystallization trials have been continuously monitored to know the time of appearance of the crystals. The crystals have been prepared directly on the experimental site, and the diffraction experiment has been performed on the crystal as soon as it appeared in the drop and reached a reasonable size (≈17 h). This procedure was needed to avoid excessive phosphate saturation, which would lead to the formation of salt microcrystals on the protein crystal itself. Two diffraction experiments at 100 K have been performed at European Molecular Biology Laboratory BW7A beamline at Deutsches Elektronen Synchrotron (Hamburg, Germany) equipped with a MAR charge-coupled detector 165-mm detector (MAR Research, Hamburg, Germany): a single-wavelength anomalous dispersion measurement carried out on the copper edge wavelength (1.377 Å) and a remote energy measurement carried out at 0.919-Å wavelength.

The first crystal diffracted up to 1.7-Å resolution and the second one up to 1.4-Å resolution; both crystals belonged to the cubic space group P4₃32 (a = 62.17 Å), with one molecule in the asymmetric unit and a solvent content of ≈50%. The data were collected with the rotation method by using 0.5° steps. The two datasets were processed by using the program mosflm (30) and scaled by using the program scala (31) with the TAILS and SECONDARY corrections on (the latter restrained with a TIE SURFACE command) to achieve an empirical absorption correction. Table 1 reports the data collection statistics for both datasets.

Table 1. Data collection and reduction statistics.

λ, Å	0.919	1.377
Spacegroup	P4(3)32	P4(3)32
Cell dimensions	a = 62.16 Å	a = 62.21 Å
Resolution, Å	27.8–1.44	31.1–1.65
Total reflections	173,047 (25,070)*	170,252 (22,414)
Unique reflections	7,922 (1117)	5,475 (764)
Overall completeness, %	100 (100)	100 (100)
Anomalous completeness, %	—	98.6 (99.7)
Rsym, %	7.0 (35.8)	8.1 (33.7)
Ranom, %	—	9.5 (13.6)
Multiplicity	21.8 (22.4)	31.1 (29.3)
I/σ(I)	9.3 (2.1)	8.4 (2.1)
B factor from Wilson plot, Å2	9.15	7.21
Phases FOM before solvent flattening	—	0.25
Phases FOM after solvent flattening	—	0.78

FOM, figure of merit.

^*Numbers in parentheses refer to the high-resolution shell (1.52–1.44 for the 0.919-Å dataset and 1.73–1.65 for the 1.377-Å dataset)

Structure Determination. The analysis of the anomalous Patterson performed with the program solve (32), using the 31-fold redundant dataset collected at the copper edge (1.377 Å), provided the positions of seven copper atoms. The preliminary phases obtained [figure of merit (FOM) = 0.25] were then improved with the density modification technique up to a FOM of 0.78 by using a solvent content of 50% with the program resolve (32). The chain-tracing routine of resolve was not able to trace any residues in the electron density map, whereas it was possible to trace 20 of 36 residues without side chains by using arp/warp (33); at this point, the electron density clearly showed the position of an additional eighth copper ion, which was further confirmed by the presence of eight large peaks in the anomalous Fourier difference map (Fig. 3A).

Fig. 3.

Different aspects of copper binding in Cu₈-MT. (A) Close-up of the electron density in the metal-binding region showing the 10 Cys residues and the eight coppers bound to them. 1σ contoured 2F_o-F_c map is shown for the Cys residues and the coppers (cyan) and 17σ contoured anomalous difference Fourier map for the coppers (red). (B) Bonding scheme of the eight Cu(I) ions (copper spheres) to the 10 cysteine sulfur atoms (yellow) reporting the numbering of the residues. The brown bonds connect Cu(I) ions with interatomic distances <3.0 Å. (C) Close-up of the Cys-7-Cu44 region of the Cu₈-MT x-ray structure, superimposed by least-squares fitting to the Cu₇-MT NMR structure. The different orientation of the Cys 7 (copper bound) and of the corresponding Cys 11 (free) side chains is shown.

The use of the dataset at higher resolution allowed for the extension of the phases and the tracing of 34 residues of 36. The two remaining residues were then added and all of the side chains correctly placed manually by using the software xtalview (34).

The refinement was then carried out by using refmac5 (35, 36) on this latter dataset. Between the refinement cycles, the model was subjected to manual rebuilding by using xtalview (34). Water molecules have been added by using the standard procedure within arp/warp (33). The stereochemical quality of the refined model was assessed by using the program procheck (37). The Ramachandran plot was of good quality, with no residues in disallowed regions.

It is worth mentioning that previous attempts to solve the structure by using a three-wavelength multiwavelength anomalous dispersion dataset collected at the European Synchrotron Radiation Facility ID-29 beamline (Grenoble, France) and one single-wavelength anomalous dispersion dataset taken at ELETTRA (Trieste, Italy) were unsuccessful. In all cases, spots coming from the diffraction of salt microcrystals grown on the surface of the protein crystal were present in the diffraction pattern, possibly spoiling by coincidence the experimental intensities. Furthermore, all these datasets had a maximum redundancy not higher than 15, whereas the successful dataset had a redundancy of 31, a finer slicing per frame (0.5° instead of 1°), and the absence of spurious spots coming from the salt microcrystal diffraction.

It is also interesting to notice that reducing the redundancy of the successful dataset to 15 causes the failure of the structure solution. In this case, redundancy, thus accuracy, of the data appears to be the crucial factor to solve the structure despite the presence of a very large anomalous signal. Table 2 reports all of the refinement statistics.

Table 2. Refinement statistics.

Resolution, Å	27–1.44
Total reflections used	7,873
Reflections in working set	7,239
Reflections in test set (8%)	634
Rcryst/Rfree, %	14.5/17.1
Protein atoms	256
Ligand atoms	8
Water molecules	64
rmsd bonds, Å	0.024
rmsd angles, °	2.1
Average B factor (including metals), Å2	12.9

rmsd, rms deviation.

Results and Discussion

Despite the very modest size of the protein, solving this structure turned out to be particularly challenging. Molecular replacement techniques based on the protein structure obtained by NMR in solution failed, probably due to the unfavorable known protein/unknown metal ratio, the contribution of the latter representing ≈35% of the total scattering. Direct methods could not be applied for phasing because the resolution, although good, was not at the true atomic level. On the other hand, the large anomalous signal coming from the copper cluster (≈20% of the diffracted intensities) could not be straightforwardly used to solve the structure. Metal clusters present in proteins have been used on several occasions to obtain phases exploiting either the isomorphous or the anomalous effect or both (38-41). It is interesting to note that the large anomalous effect comes from the global scattering of the cluster as a whole: in this sense, the cluster can be used as a group scatterer, and it can yield phases up to a resolution corresponding to ≈60-70% of the average diameter of the cluster itself, thus ≈5-8 Å. This happens because, at low resolution, all atoms in the cluster scatter in phase and act as a super-heavy atom. The difficulty in extending phases to higher and more useful resolutions comes from the difficulty in gathering information on the atomicity of the single components of the cluster. In other words, there is an intrinsic problem in finding the correct orientation of the cluster itself. The correct orientation, thus the correct positions of the metals in the cluster, can be achieved only with very accurately measured data. As the present structure has shown, a very high-quality dataset is still required to solve the structure exploiting the anomalous diffraction effect of a large metal cluster.

The exclusive existence of Cu(I) bound to cysteinyl thiolate sulfur atoms in yeast Cu-MT was proven by using x-ray photoelectron spectrometry (42, 43), whereas the exact number of Cu(I) ions per Cu-MT molecule has been a matter of debate for quite a long time. In any case, there is much convincing evidence that M₇-MT is a stable species, and our reconstitution strategy in the present work was based on a 7:1 Cu(I)/protein molar ratio. To our surprise, eight rather than seven copper atoms are found in the x-ray structure, coordinated to all 10 cysteines (Fig. 1 and Fig. 3 A and B) and arranged in a Cu₈-thiolate cluster. This is an example of an octanuclear copper core present in a protein, and it is remarkable how this is realized in a protein fragment of such low molecular mass. Of the eight coppers, two (Cu38 and Cu44) are digonally coordinated to cysteines, whereas all of the others are trigonally coordinated. In two different preceding studies, digonal and trigonal coordination was anticipated on the basis of extended x-ray absorption fine-structure measurements, although the coexistence of both in this single Cu-thiolate cluster had not yet been seen (28, 44). As shown in Fig. 3B, the eight Cu(I) ions can be described as grouped in two sets of four, both having a flattened tetrahedron structure. One of the tetrahedra (Cu39, Cu40, Cu41, Cu44) has all of the Cu atoms at very short distances, ranging from 2.59(6) to 2.91(6) Å, whereas the other (Cu37, Cu38, Cu42, Cu43) displays slightly longer distances [2.78(6)-3.12(6) Å]. The Cu-Cu distances below 2.8 Å (sum of Cu VdW radii) imply some metal-metal bonding character. Cu39 of set 1 makes a close approach to set 2, being at 2.73 and 2.78 Å from Cu42 and Cu37, respectively, thereby connecting the two polyhedra (Fig. 3B).

Like Cu(I)-thiolates in general, yeast Cu-MT also exhibits a characteristic orange luminescence emission with a maximum at 609 nm (45-47). The noteworthy high quantum yield of native yeast Cu-MT was determined at 0.0058 and is due to the effective shielding of the Cu-thiolate cluster from solvent interaction (45). Both the previously published NMR structure (26) and the now-available x-ray structure could show that the cysteinyl sulfur atoms responsible for Cu(I) binding are allocated in the inside of the polypeptide. At the same time, all Cu(I) ions are shielded by the hydrophobic thiolate sulfurs, being inaccessible for solvent molecules. As a consequence, solvent quenching of the luminescence is diminished, which becomes apparent in the relatively high quantum yield.

Comparison with the Ag₇-MT and the Cu₇-MT structure (Fig. 1) shows that the protein backbone and the metal-cysteine topology are similar but not identical. Least-squares fitting of the backbone atoms of the present molecule with Ag₇-MT and Cu₇-MT resulted in rms deviations (rmsds) of 1.56 and 1.12 Å, respectively. A closer inspection shows that for Ag₇-MT, part of the backbone and several side chains have different conformations, leading to a different arrangement of some of the cysteine side chains. The Cu₇ cluster inferred from the Ag₇-MT shows a largely different clustering of the seven atoms with little superposition (the rmsd varies from 1.03 to 1.05 Å, depending on which extra Cu atom is taken away, Cu39 or Cu43). On the other hand, in the solution structure of Cu₇-MT, all cysteines but one (see below) have a side-chain arrangement similar to the present x-ray structure.

The availability of the Cu₈-MT x-ray structure poses the question of whether Cu₈-MT should now be considered the formula of yeast copper thionein in vivo. In this respect, we should recall that six to eight coppers per thionein molecule have been found in the past, and that the last two copper ions were found to be more labile than the others (48). It might well be that crystals of Cu₈-MT have formed from a solution of Cu₇-MT, perhaps leaving some Cu₆-MT molecules in solution. In this case, there may be three different stoichiometries of Cu(I) binding to MT, which are eight, seven, or even six Cu(I) per MT molecule, and all these forms may have physiological significance.

In light of the present Cu₈-MT structure and its comparison with the best-available NMR structure of the protein part of Cu₇-MT (26), we propose here a model for the Cu₇-MT cluster in solution. Fig. 3C shows a closeup of the superposition of the x-ray and NMR structures in the region of Cys-7 (Cys-11 NMR). The conformation of the two Cys side chains is different, with the thiolate sulfurs pointing in opposite directions. In the present structure, Cys-7 is bound to Cu44 at 2.15 Å, whereas the conformation of the corresponding Cys-11 in solution is not compatible with copper binding. This evidence, together with the fact that Cu44 is positioned at one of the ends of the cluster that it is exposed to the solvent, and that it is coordinated by only two cysteines (Cys-7 and Cys-26), suggests that Cu44 can be the copper ion missing in the solution structure. In this respect, it is interesting to note that the two digonally coordinated Cu(I) ions (Cu38 and Cu44) are located at the opposite ends of the cluster, suggesting for the Cys-7-Cys-26 and Cys-3-Cys-16 pairs bound to them the role of uptake and release of these Cu(I) ions by side-chain movements similar to that described above.

The safe and finely tuned transport of reactive Cu(I) in aqueous biological systems was one of the most intriguing possible physiological functions of yeast Cu-MT (3, 15-18). On the basis of its crystal structure reported here, this role is supported by the finding that the Cu(I) ions are sequestered into the core of the polypeptide, thereby being protected from the solvent molecules. They are tightly bound to cysteinyl thiolate sulfurs and stabilized in their monovalent oxidation states. In this way, they are prevented from uncontrolled redox reactions that might be hazardous for many cellular components. This copper-scavenging role, together with the observation that copper-exposed yeast cells are capable of releasing entire Cu-MT, argue for a copper detoxification function of yeast Cu-MT (49).

Cu-MT is also considered to act as depository for copper designated for the effective transfer into the apo forms of copper proteins (1, 11, 12, 16, 17). Such function is beautifully backed by the finding of two digonally coordinated copper ions that may well correspond to the two more labile bound coppers observed in a competitive replacement experiment (48). Fulfilling this task of a copper storage and delivery agent, metallothionein might well interact with copper chaperones, possibly using the same mechanisms as those discussed for the transfer of copper from copper chaperones to their target proteins (1).

It is interesting to note that the full-length Cu-MT possesses two additional cysteine residues in the C-terminal tail that are missing in the truncated form used for crystallization. These two cysteine residues, which were shown not to participate in Cu(I) binding (27, 28), might also be involved in copper uptake and release and in the interaction with copper chaperones.

In the context of biological Cu(I) transport, the recent observation of cerebral Cu-MT being possibly involved in neurodegenerative diseases (50) is a promising endeavor for future studies. Unlike the established prevalence of two labile Cu(I)-thiolates of digonal coordination in yeast Cu-MT, no such structural features have been clearly assigned in mammalian Cu-MT. From circular dichroism and fluorescence emission, there may be similar, if not identical, Cu(I)-thiolate coordination in either Cu-MT (51).

Taking these considerations together, it could be speculated that the physiological function of yeast Cu-MT and of Cu-MT in general might be a safety overflow basin that stores copper within the respective cells and releases it under copper scarceness conditions. Under copper concentrations exceeding the cells' requirements, metallothioneins are able to bind excess copper and can be exported into the medium.