Structure of mammalian metallothionein.

All mammalian metallothioneins characterized contain a single polypeptide chain of 61 amino acid residues, among them 20 cysteines providing the ligands for seven metal-binding sites. Native metallothioneins are usually heterogeneous in metal composition, with Zn, Cd, and Cu occurring in varying proportions. However, forms containing only a single metal species, i.e., Zn, Cd, Ni, Co, Hg, Pb, Bi, have now been prepared by in vitro reconstitution from the metal-free apoprotein. By spectroscopic analysis of such derivatives it was established that all cysteine residues participate in metal binding, that each metal ion is bound to four thiolate ligands, and that the symmetry of each complex is close to that of a tetrahedron. To satisfy the requirements of the overall Me7(Cys-)20 stoichiometry, the complexes must be combined to form metal-thiolate cluster structures. Experimental proof for the occurrence of such clusters comes from the demonstration of metal-metal interactions by spectroscopic and magnetic means. Thus, in Co(II)7-metallothionein, the Co(II)-specific ESR signals are effectively suppressed by antiferromagnetic coupling of juxtaposed paramagnetic metal ions. By monitoring changes in ESR signal size occurring on stepwise incorporation of Co(II) into the protein, it is possible to follow the building up of the clusters. This process is biphasic. Up to binding of four equivalents of Co(II), the ESR amplitude increases in proportion to the metal content, indicating generation of magnetically noninteracting high-spin complexes. However, upon addition of the remaining three equivalents of Co(II), these features are progressively suppressed, signaling the formation of clusters. The same mode of cluster formation has also been documented for Cd and Hg. The actual spatial organization of the clusters and the polypeptide chain remains to be established. An attractive possibility is the arrangement of the tetrahedral metal-thiolates in adamantane-like structures surrounded by properly folded segments of the chain providing the ligands. 1H-NMR data and infrared absorption measurements are consistent with a tightly folded structure rich in beta-type conformation.

All mammalian metallothioneins characterized contain a single polypeptide chain of 61 amino acid residues, among them 20 cysteines providing the ligands for seven metalbinding sites. Native metallothioneins are usually heterogeneous in metal composition, with Zn, Cd, and Cu occurring in varying proportions. However, forms containing only a single metal species, i.e., Zn, Cd, Ni, Co, Hg, Pb, Bi, have now been prepared by in vitro reconstitution from the metal-free apoprotein. By spectroscopic analysis of such derivatives it was established that all cysteine residues participate in metal binding, that each metal ion is bound to four thiolate ligands, and that the symmetry of each complex is close to that of a tetrahedron.
To satisfy the requirements of the overall Me7(Cys-)20 stoichiometry, the complexes must be combined to form metal-thiolate cluster structures. Experimental proof for the occurrence of such clusters comes from the demonstration of metal-metal interactions by spectroscopic and magnetic means. Thus, in Co(II)7-metallothionein, the Co(II)-specific ESR signals are effectively suppressed by antiferromagnetic coupling of juxtaposed paramagnetic metal ions. By monitoring changes in ESR signal size occurring on stepwise incorporation of Co(II) into the protein, it is possible to follow the building up of the clusters. This process is biphasic. Up to binding of four equivalents of Co(II), the ESR amplitude increases in proportion to the metal content, indicating generation of magnetically noninteracting high-spin complexes. However, upon addition of the remaining three equivalents of Co(II), these features are progressively suppressed, signaling the formation of clusters. The same mode of cluster formation has also been documented for Cd and Hg.
The actual spatial organization of the clusters and the polypeptide chain remains to be established. An attractive possibility is the arrangement of the tetrahedral metal-thiolates in adamantane-like structures surrounded by properly folded segments of the chain providing the ligands. 1H-NMR data and infrared absorption measurements are consistent with a tightly folded structure rich in 3-type conformation.
Historically, it was in human kidney that the Russian geochemist D. P. Malyuga first serendipitously discovered the natural accumulation of cadmium in vertebrate tissues (1). Based on this observation, Margoshes and Vallee then proceeded to isolate from equine kidney the first cadmium-containing protein (2). This material, subsequently named metallothionein (3), was characterized by the following features (4): molecular weight 6000-7000; high metal content; characteristic amino acid composition (high cysteine content, no aromatic amino acids); unique distribution of cysteinyl residues in the amino acid  (5), the definition was adopted that any protein resembling equine renal metallothionein in several of these criteria can be classified as a metallothionein. Proteins satisfying this liberal definition have now been found to occur not only in many animal species but also in cexrtain eukaryotic microorganisms (6)(7)(8), in some plants (9,10), and even in prokaryotes (11,12). Table 1 lists the species and tissues in which the natural occurrence of metallothionein has been demonstrated and for which data on metal composition are available. The variability in the proportions of the different group-2B metal ions and copper has been a unique and unorthodox feature of these proteins since the earliest reports (2,3). Clearly, there is no evidence either for a stoichiometric partitioning among the metals or for the existence of metal-specific sites. In fact, the occupation of the binding sites by a given metal seems to be determined not so much by the chemical affinity of the sites as by the supply of the metal and by biological variables, such as the tissue of origin (13,14), age, stage of development (15), and the activity of certain still poorly understood homeostatic mechanisms (16). While zinc is the more abundant component in nearly all mammals examined, copper tends to be more prevalent in fish. Thus, the patent variability in metal composition alone should guard against the assignment of a function of metallothionein with respect to a single metal. Most likely, the very broad metal-binding specificity provides for a multiplicity of biological roles that metallothionein may have acquired during evolution (Table   1).
Many metallothioneins are related to one another in their primary structure as documented by amino acid sequence analysis. Figure 1 summarizes most of the structural information available of what we propose to designate as class I metallothioneins. In addition to all mammalian metallothioneins thus far characterized, this class includes proteins from a fish (Pleuronectes platessa), a marine crab (Scylla serrata) and from a mold (Neurospora crassa). The average 42% identity between the aligned sequences of the arthropode and mammalian forms and the average 32% identity between Neurospora metallothionein and the corresponding N-terminal segment of all other metallothioneins included in the comparison leave no doubt about a common origin of the genes coding for them. With the limited number of sequences available it is, of course, premature to attempt the construction of a phylogenetic tree. However, from the manifest differences between the sequences of metallothioneins of phyla whose separation in evolution can be dated, it is feasible to obtain an estimate of the average rate of evolution of this class of proteins. The time course of the emergence of structural changes expressed in percent difference of the aligned segments of pairs of sequences is shown in Figure 2. After correcting for back mutations (45) and by assuming that separation of vertebrates and invertebrates occurred some 600 million years ago, one calculates from the comparison of the crab metallothioneins with all mammalian forms an average rate of 7.2 x 10-10 amino acid substitutions per year corresponding Pig Liver Limpet Soft tissue + + + + + + (33) aAll data refer to metallothionein obtained from organisms not subjected to experimental pretreatment with metals. KAGI ET AL. 30 C K K S C C S C C C K K S C C S C C C K K S C C S C C C K K S C C S C C  (6). The numeration refers to the sequence of mammalian metallothioneins. The residues enclosed within parentheses require further identification. Dots between adjacent residues denote deletions introduced for optimal alignment. Residues 58 and 59 of the human and equine metallothioneins were reassigned following sequence reexamination (M. Kimura   to a unit evolutionary period of about 14 million years. A similar figure is obtained by comparing the Neurospora protein with all other metallothioneins. They range between those calculated previously for the globins and for cytochrome c, suggesting that class I metallothioneins are rather slowly evolving proteins ( Table 2). The sequencing studies have also established that mammalian metallothioneins display very Fibrinopeptide A 42.9 x 10-10 aCalculated from differences between mammalian and anthropode metallothioneins by the procedure of King and Jukes (46). The figures included for comparison were taken from the same reference. substantial genetic polymorphism. There are two isometallothioneins in rodents (47) and there are at least four in horse (13) and in rabbit (48). An even larger multiplicity of forms exists in human tissues where not less than 11 genes or pseudogenes coding for metallothionein have been reported (49). The structural divergence of the isoproteins ranges from 1 to 15 amino acid substitutions, indicating that they must have arisen at various times during evolution. While isoproteins differing in electric charge can be isolated by electrophoretic techniques or by ionexchange chromatography (5,50), many isoforms are not resolved by these methods. A remarkably good resolution of many isoproteins can be obtained, however, by reversed-phase HPLC   (51). (48,51). As shown in Figure 3, it is now possible to separate in a single step five of the isometallothioneins present in the supernatant of human liver.
The chemically most significant information derived from comparative sequence analysis pertains to the mode in which the polypeptide chains provide for metal binding. The essential conditioning factor for metal complexation are the 20 cysteine residues whose position is completely preserved in all mammalian sequences known thus far (Fig. 4). The most unique feature is the repetitive occurrence of the cys-x-cys-tripeptide sequence, where x usually stands for a residue other than cysteine. Occurring seven times along the chain, these sequences have been suggested to serve as primary chelation sites for the seven group-2B metal ions usually bound to this protein and are thought to be essential for the formation of the metal-thiolate clusters typical of this protein (vide infra). That the metal ions are coordinated to the sulfur group of the cysteine residues was inferred early from the observation that native metallothionein contains no free sulfhydryl groups (21). All 20 cysteine residues are deprotonated, and hence they must occur as thiolate ligands bound to the seven bivalent metal ions, yielding the rather unusual stoichiometry of about three thiolate ligands per metal. This mode of metal binding entails the generation of an extra negative charge at six of the seven metal centers, thereby bringing about the overall negative charge typical of all metallothioneins on record (5,52).
The earliest direct evidence for binding of the metal to cysteine sulfur was obtained from the electronic absorption spectrum of Cd(II)-containing forms of metallothionein which is characterized by a broad absorption shoulder at 250 nm superimposed upon protein absorption (21,53). The same feature occurs in Cd(II)-complexes of 2mercaptoethanol (53) and of synthetic peptides containing the cys-x-cys sequence (54). It disappeared completely upon removal of the metal at low pH, and hence it was attributed to a ligandmetal electron transfer transition (21,53). This assignment is also supported by a recent comparative spectroscopic study of a series of monosubstituted derivatives of metallothionein containing a full complement of Zn(II), Cd(II), Hg(II), Pb(II), and Bi(III) (55). The striking features of these spectra (Fig. 5) are the strong metal-induced absorption envelopes and the large red shift which increases on going from the lighter to the heavier metals.
According to the semiempirical theory of  The values for the metals were deduced from absorption spectra for tetrahedral halide complexes (61,62).
The values of the optical electronegativities are determined by the chemical nature and the coordination geometry of the complexes and can be obtained from the spectra of model complexes. That this theory holds for metallothionein is documented in Table 3, where the predicted positions of these bands calculated for tetrahedral model complexes of Zn(II), Cd(ll), Hg(ll), and Pb(ll) with thiolate ligands are compared with the actual location of the first electron transfer transitions in the absorption spectra of the corresponding monosubstituted metallothioneins. The excellent agreement between the two sets of data strongly suggests that in metallothionein the metal ions are bound in tetrahedral geometry and that each of the seven metal ions is surrounded by four cysteine residues. Identical inferences have also been drawn from spectroscopic studies of CO(II)7metallothionein (57).
Since the protein contains 20 cysteine residues and usually binds seven bivalent metal ions, it follows from these spectroscopic data that the options for the accommodation of the metal ions are severely restricted. Clearly, there are not enough cysteines to provide every metal ion with four separate thiolate ligands, and hence there must be a sharing of some of these ligands by adjacent metal ions giving rise to the formation of metal-thiolate cluster complexes. The observed stoichiometry dictates that the 20 cysteine residues present in the molecule must be partitioned into eight doubly coordinated bridging thiolate ligands and twelve singly coordinated terminal thiolate ligands.
The first direct evidence for the existence of metal-thiolate clusters in metallothionein came from the "13Cd-NMR studies of Otvos and Armitage (59), who deduced from homonuclear decoupling measurements that the protein contains two separate metal-thiolate clusters, one with three and one with four metal ions, respectively. Similar conclusions are also obtained from spectroscopic and magnetic studies of Co(II)-substituted metallothionein (60). Co(II) has chemical properties similar to those of Zn(II). Therefore, this metal can replace zinc in many metalloproteins including metallothionein. Due to its transition-metal nature, the introduction of Co(II) into metallothionein produces an emerald-green derivative whose absorption spectrum offers proof that the metal is bound tetrahedrally to four thiolate ligands (Fig. 6).
Because Co(ll) has unpaired electrons it can also serve as a paramagnetic probe of its molecular environment. Accordingly, Co(II) can sense the presence of other Co(II) ions when they occur in close proximity, i.e., in a cluster complex. Such Co(II)-Co(ll) interactions manifest themselves mainly by intensity changes in the electron spin resonance (ESR) properties. The ESR spectrum of Co(ll-metallothionein shown in Figure 7, (left), is typical of high-spin Co(ll) as expected for tetrahedral complexes. The same profile is also seen in partially saturated Co(II)-metallothionein, but its intensity monitored by the amplitude of the major signal at 1000 Gauss (g. unusual dependen'ce on metal-to-proteiln stoichiometry. On titration, of the metal-free p-rotein -(apometallothionein) with Co(II) (Fig.,.7, right), the signal -increases at first in proportion'to the paramagnetic metal ion added. However, this trend changes abruptly after incorporation of nearly four Co(II). On further addition of Co(II) there is a progressive loss of signal, yielding an ESR spectrum of the fully reconstituted derivative that has an even lower amplitude than that recorded after adding only 1 mole of Co(II). The explanation for this anomnalous titration behavior is spin canceling due to antiferromagnetic coupling of vicinal paramagnetic Co(II)-centers.
Hence, together with magnetic susceptibility measurements (60), these data establish unambiguously that in Co(II)-metallothionein the meta ions are lnked in one or more oligonuclear clusters. The remarkably sharp transition from the paramagnetic nonclustered to the diamagnetic clustered structure upon filling up of the metal-binding sites of metallothionein indicates that under the conditions employed magnetically noninteracting complexes are formed first and, hence, that they must be energetically more stable than the clustered complexes. The same noncooperative mode of binding in which spatially segregated metal complexes are formed before the clusters are completed was recently observed also in the formation of Cd7-metallothionein (54) and of Hg7-metallothionein (55). The cluster model fitting the observed spectroscopic features best is the adamantane-like metal-thiolate cage depicted in Figure 8. This decahedron first proposed by Dance on the basis of crystallographic studies of synthetic model compounds (63,64) is similar to the structures inferred from the 13Cd-NMR studies (59). Its attractive feature is that it fully preserves the tetrahedral coordination of each metal ion while allowing their interconnection through thiolate ligands. Its occurrence in metallothionein, albeit in a somewhat modified form, affords a ready explanation for the remarkable simplicity of the electronic spectra discussed above. The adamantane model also illustrates nicely the partitioning of the sulfur ligands into two classes, the bridging thiolates together with the metal forming the decahedron, and the' terminal thiolates located outside the' polyhedral cage. The model implies that the cluster is located in the interior of the prGtein surrounded by the segments of the polypeptide chain providing the ligands.
Although some estimates of the gross physical features of metallothionein have been made from hydrodynamic measurements (20) and although there is now good evidence that the two metal-  (65), little detailed documentation is as yet available on the spatial organization of this protein (66). Techniques sensitive to polypeptide conformation such as circular dichroism, infrared absorption spectroscopy, and 'H-NMR spectroscopy have, however, given some information on protein folding both in the metal-free and in the metal-containing forms.
Representative far-UV circular dichroism spectra of the two forms are shown in Figure 9. Both spectra display the strongest ellipticity band at about 200 nm, suggesting a large proportion of disordered structure (67). At longer wavelengths, there are, however, substantial differences which can in part arise from differences in polypeptide folding (vide infra) but which to an appreciable extent are also the result of the superposition of ellipticity contributions of the Zn(II)-thiolate chromophore upon the circular dichroism of the apoprotein. Accordingly, secondary structure analysis by computer fitting of the circular dichroism spectrum using parameters derived from crystallographically defined proteins (68) is feasible only for the metal-free form. Table 4 shows the result of such a study carried out on equine liver metallothionein-lA. It suggests that the apoprotein still contains about 45% secondary structure, most of it as 1-sheet and 13 amino acid sequence using the secondary structure prediction method of Chou and Fasman (69).
Since tetrahedral binding of the seven metal ions to the 20 cysteine residues introduces not fewer than 42 additional crosslinks into the molecule, it is to be anticipated that metal complexation will impose new conformational constraints on the folding of the chain, bringing about some changes in secondary structure. This is, in fact, confirmed by a comparison of the metal-free and metal-containing forms of metallothionein by infrared absorption spectroscopy, a method which yields information on secondary structure folding. Over far-UV circular dichroism measurements it has the advantage of the conformationdependent spectral features not being obscured by contributions from the metal-thiolate chromophores (Fig. 10). In this technique, position and shape of the amide I vibration near 1650 cm-1 are monitored in a solution of the protein in 2H20 (70,71). For comparison, spectra of deuterated reference proteins are also included. The predominantly a-helices containing myoglobin shows its amide I absorption maximum at 1650 cm-1 andstructure-containing ribonuclease at 1639 cm-1.
The disordered polypeptide conformation in proteins displays a maximum at about 1645 cm-1 and a weak shoulder at 1670 cm-'. P-Structure produces both a maximum at 1634 cm-1 and a strong and a weak shoulder at 1660 and 1680 cm-', respectively. The maximum of apometallothionein is located at 1647 cm-1, confirming the presence of a large proportion of disordered structure. The shape of the absorption band of the holoprotein is similar but the displacement of the amide I band to 1643 cm-', the intensification of the shoulder at 1660 cm-', and the increase in total amplitude are suggestive of an appreciable increase in 13-structure content on metal binding (D. Gilg and J. H. R. Kagi, unpublished observation).  The stabilizing influence of metal binding on the secondary and tertiary structure is further underscored by the early observation that unlike the metal-free protein (65), the metal-containing form is highly resistant to proteolysis (73), and that some of the peptide hydrogens are shielded from exchanging with the solvent (74). This is now also supported by comparative 'H-NMR studies (76). The differences between the two forms are particularly striking in the low-field region (6.5-9.5 ppm), where only resonances of the secondary amide groups of the polypeptide chain and of the N-bound protons of the amino acid side chains occur (Fig. 11). While in the apoprotein in 'H20 these resonances are collected to relatively narrow bands, in the metal-containing form they are spread over a chemical shift span of nearly 3 ppm, indicating substantial hydrogen bonding.
The structural differences are also manifest in the high-field region (0-4 ppm) (75,77). However, because of the absence of aromatic residues in the molecule and, hence, of aromatic ring current effects, shifts arising from changes in the environ-Apometallothionein (pH < 1) in H20 -Metallothionein (pH 7) in H20-' 8 4 0 ppm FIGURE 11. 270-MHz 1H-NMR spectra of (bottom) equine metallothionein-lA and (top) apometallothionein. Ib obtain the spectra in H20, the solvent signal was suppressed by a gated pulse (75). ment of the protons are less pronounced and are difficult to interpret. These limitations inherent in one-dimensional 'H-NMR spectra of proteins are now being overcome, however, by the application of two-dimensional 'H-NMR techniques already applied successfully to a number of other systems (78). Studies on rabbit liver metallothionein currently conducted in collaboration with the Institute of Molecular Biology and Biophysics, ETH, Zurich, have led to the resolution of nearly all individual 'H spin systems of the protein and have permitted sequential assignment of about 80% of the signals (79). It is anticipated that this technique, in conjunction with 'H-113Cd heteronuclear coupling studies, will provide a realistic spatial model of the folded molecule.
In conclusion, the chemical and spectroscopic data presented reflect the progress made in recent years in the understanding ofthe structure ofthis highly unusual protein. Among the major findings are (1) the remarkable conservation of the primary structure manifesting itself in the nearly complete invariance of the positions of the metalbinding cysteine residues and (2) the detailed knowledge gained from physical studies on the structural organization of the metal-binding sites, i.e., the unambiguous demonstration that bivalent metal ions are bound in tetrahedral microsymmetry to four thiolate ligands and that these complexes are joined to form discrete metal-thiolate clusters. These oligonuclear complexes constitute the first example of clusters of group-2B metal ions in a biological system. Their unique occurrence in metallothionein implies that cluster formation with these bivalent metal ions is dependent on the very special distribution of cysteine residues characteristic for this protein.
In fact, it seems most likely that it is the stringency of the stereochemical requirements for cluster formation that is responsible for the remarkable preservation of the positions of the cysteine residues in class I metallothionein throughout evolution. Dedicated to Professor F. Leuthardt, Zurich, on the occasion of his 80th birthday.