MINIREVIEW Neuronal growth-inhibitory factor (metallothionein-3): reactivity and structure of metal–thiolate clusters* Peter Faller 1,2 1 CNRS, LCC (Laboratoire de Chimie de Coordination), Toulouse, France 2 Universite ´ de Toulouse, UPS, INPT; LCC; Toulouse, France Introduction Metallothionin-3 (MT3) was originally dubbed neuro- nal growth-inhibitory factor (GIF) [1] because of the discovery that it is a factor in brain extract with the ability to inhibit neuronal outgrowth. Moreover, MT3 or GIF was reported to be down-regulated in extract from Alzheimer’s disease (AD) brain. Later on it became clear that GIF belongs to the metallothionein (MT) family based on its high cysteine and metal con- tents. In mammals, the MT family consists of four dif- ferent subfamilies designated MT1 to MT4 [2–4]. Mammalian MTs are composed of a single polypep- tide chain of 61–68 residues. They are characterized by a conserved array of 20 cysteines and the absence of His and aromatic amino acids. MT3 contains 68 amino acids with 70% sequence identity to the MT1 and MT2 (MT1 ⁄ 2) isoforms. The MT3 sequence contains two inserts: an acidic hexapeptide in the C- terminal region and a Thr in position 5. Moreover, a conserved Cys-Pro-Cys-Pro motif between positions 6 and 9 is unique to MT3 [1,4,5]. All mammalian MTs can bind a variety of different mono-, di- and trivalent metal ions via their cysteine residues. Most relevant under normal conditions in biology is the binding of Zn 2+ and Cu + . However, MTs can also bind other metals (Cd 2+ ,Hg 2+ ,Ag + , Pt 2+ ,Pb 2+ and Bi 3+ ) when they are administered to animals. Classically, the metal ions studied in detail [mainly Zn(II), Cu(I), Cd(II), Hg(II) and Ag(I)] are Keywords bioinorganic chemistry; copper; growth inhibitory factor; metallothionein; metal– thiolate clusters; protein structure; zinc Correspondence P. Faller, CNRS, LCC 205, route de Narbonne, 31077 Toulouse, France Fax: +33 5 61 55 30 03 Tel: +33 5 61 33 31 62 E-mail: peter.faller@lcc-toulouse.fr *This article is dedicated to Prof. M. Vasak on the occasion of his retirement (Received 3 December 2009, revised 4 May 2010, accepted 17 May 2010) doi:10.1111/j.1742-4658.2010.07717.x Metallothionein-3, also called neuronal growth-inhibitory factor, is one of the four members of the mammalian metallothionein family, which in turn belongs to the metallothionein, a class of ubiquitously occurring low- molecular-weight cysteine- and metal-rich proteins containing metal– thiolate clusters. Mammalian metallothioneins contain two metal–thiolate clusters of the type M(II) 3 -Cys 9 and M(II) 4 -Cys 11 [or Cu(I) 4 -CysS 6-9 ]. Although metallothionein-3 shares these metal clusters with the well- characterized metallothionein-1 and metallothionein-2, it shows distinct bio- logical, structural and chemical properties. This short review focuses on the recent developments regarding the chemistry of the metal clusters in metal- lothionein-3, in comparison to those in metallothionein-1 and metallothion- ein-2, and discusses the possible biological and functional implications. Abbreviations Ab, b-amyloid; AD, Alzheimer’s disease; apo-T, apo-thionein; DTNB, 5,5¢-dithiobis(2-nitrobenzoic acid)); GIF, neuronal growth-inhibitory factor; M(II), divalent metal ion; MT, metallothionein. FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works 2921 bound in mammalian MTs collectively in two metal– thiolate clusters located in independent domains (a b-domain for the N-terminal cluster and an a-domain for the C-terminal cluster). In the absence of metals, apo-thionein (apo-T) is predominantly unstruc- tured and only upon metal coordination is a defined 3D structure formed. The cluster structures for the divalent metals have been resolved for MT1 ⁄ 2 and they were found to be highly unusual compared with other proteins (Fig. 1). The N-terminal b-domain (amino acids 1-30) binds three divalent metal ions by nine deprotonated cysteines in a hexane-like cluster [M(II) 3 - Cys 9 ], including three bridging and six terminal cyste- ines. The C-terminal a-domain (amino acids 31-61) binds four divalent metal ions by 11 deprotonated cysteines in an adamantane-like cluster [M(II) 4 -Cys 11 ], including five bridging and six terminal cysteines. All seven divalent metals are bound in tetrahedral coordi- nation (Fig. 1) [4]. The precise cluster structure of the Cu(I) X -CysS Y moieties in mammalian MTs has not yet been determined, and the only structure available is from the Cu(I) 8 -Cys 10 cluster in yeast MT (Fig. 2) [6]. It is unlikely, however, that mammalian MTs contain such a cluster. Spectroscopic experiments on mamma- lian MT1 ⁄ 2 indicate that they contain Cu(I) 4 -CysS 6-9 , Cu(I) 6 -CysS 9-11 or Cu(I) 3 ⁄ 4 -Zn x CysS x clusters and hence have a distinctly different cluster organization. In all cases, Cu(I) shows a preference (at least partial) for binding to the N-terminal b-domain [4,7]. These metal–thiolate cluster structures are unusual for metal-binding proteins and are responsible for the typical reactivity properties of MTs. First, the metal binding is thermodynamically relatively stable, with the following order Cu(I)>Cd(II)>Zn(II). The cluster structure raises the possibility of a cooperative binding of the metal ions. Cooperative binding has been reported but is not always found, being dependent on the type of metal, the type of cluster and the pH [2,5,8]. In contrast to Zn(II)-MT and Cd(II)-MT, apo- T is is oxidatively unstable under aerobic conditions as a result of the formation of disulfide bridges [2]. Sec- ond, in contrast to the high thermodynamic stability, the binding is kinetically labile, allowing rapid intra- molecular and intermolecular metal transfer. This is a direct consequence of the relatively high structural dynamics and flexibility typical of all MTs [5]. Third, the deprotonated cysteines bound to the metal ions (i.e. thiolates) are good nucleophiles, conferring a high reactivity with radical species (HO·, O 2 · ) , NO·) as well as with alkylating and oxidizing agents. Such reactions result in oxidation or derivatization of the cysteines with subsequent metal release [4]. Although MT3 belongs to the MT family and hence shares the unusual properties of their metal–thiolate clusters, there are important differences between MT3 and the well-characterized MT1 ⁄ 2 [4]. Such chemical and structural differences are probably important for the biological roles of MT3, such as its growth-inhibitory activity, the non-inducibility of its gene by diverse metal ions and other compounds known to elicit the formation of MT1 ⁄ 2, its predominant localization in the central nervous system with an accumulation in zinc-enriched neurons, and its possibility of being excreted into the extracellular space [4,9,10]. Note, that the latter may not be restricted to MT3 as evidence is accumulating that MT1 ⁄ 2 also occurs extracellularly [4]. This is summa- rized in Table 1 and discussed in more detail below. Metal content of MT3 Initially, MT3 was isolated from human brain with a metal content of four Cu(I) and three Zn(II) per MT3 Fig. 1. Scheme of the two metal–thiolate clusters containing Zn ⁄ Cd in mammalian MT1 ⁄ 2. Left: four-metal cluster [M(II) 4 -Cys 11 ] localized in the C-terminal a-domain. Right: three-metal cluster [M(II) 3 -Cys 9 ] localized in the N-terminal b-domain. The scheme is based on structures obtained by X-ray and NMR for MT1 ⁄ 2. MT3 contains the same type of four-metal cluster. M, divalent metal, S, cysteine thiolate. Adapted from a previous publication [4]. Fig. 2. Scheme of the Cu(I) 8 (Cys) 10 cluster from yeast CuMT. Cu, copper; S, cysteine thiolate. Adapted from a previous publication [6]. Metal–thiolate clusters in metallothionein-3 P. Faller 2922 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works Table 1. Comparison between MT3 and MT1 ⁄ 2. NOS, nitric oxide synthase; ROS, reactive oxygen species. Feature Similarities Differences MT1 ⁄ 2 and MT3 MT3 MT1 ⁄ 2 Putative biological function Involved in Zn and Cu metabolism Synthesis not inducible Synthesis induced by Zn, Cu, Cd Growth-inhibitory activity ROS and NOS scavengers (antioxidant) Protects cultured neurons against Ab toxicity [42] Does not protect [42] Localization Can occur extracellularly [14] Predominantly in the brain, neurons and ⁄ or astrocytes? [3] Ubiquitous; in the brain mostly in astrocytes [3] Primary structure 70% sequence identity 68 amino acids 61 amino acids 20 cysteines, same arrangements Acidic 6-amino-acid insert in the C-terminal domain Absence of 6-amino-acid and Thr insert Thr insert at position 4 Cys(6)-Pro-Cys-Pro(9) Cys(5)-Ser-Cys-Ala(8) Metal content Binds without metal exposure Zn and perhaps Cu. No heterometallic Zn ⁄ Cu clusters Isolated as a mixture of Cu and Zn [1,11], but perhaps in vivo predominantly Zn [13] Isolated predominantly as Zn only Zn binding 7 Zn(II) bound to 20 thiolates Additional specific (eighth) Zn-binding site [30,31] Similar overall apparent K d of Zn Cu 4 Zn 4 : 4.2 · 10 )12 M[32] Zn7: 1.6 · 10 )11 M [17] Zn7: 3.2 · 10 )12 M [17] Individual K d at pH 7.4 Zn ⁄ Cd binding stronger in 4-metal cluster than in 3-metal cluster [18] Zn 3 and Zn 4 cluster: non-cooperative? [21,22,30] Not known Additional eighth site: 1 · 10 )4 M 7th Zn: 2.0 · 10 )8 M 6th Zn: 1.1 · 10 )10 M 5th Zn: 3.5 · 10 )11 M First 4 Zn: 1.6 · 10 )12 M Cd binding 7 Cd(II) bound to 20 thiolates Additional specific eighth binding site [30] No specific additional sites [30] Cd binding similar to Zn Cd 7 form less compact than Zn7 [26] Similar compactness [26] Similar overall apparent K d of Cd Cd 7 : 5.0 · 10 )15 M [17] Cd 7 : 1.4 · 10 )15 M [17] More non-cooperative Cd binding [17] More cooperative Cd binding (at pH 7.4) [17] Zn ⁄ Cd–thiolate clusters Cd ⁄ Zn 3 -CysS 9 in b-domain and Cd ⁄ Zn 4 -CysS 11 in a-domain Same connectivities in Cd form in Cd 4 cluster [5,23,24] Dynamics Cd 3 -CysS 9 more dynamic than Cd 4 -CysS 11 [5] Cd 3 -CysS 9 very dynamic (precluded structure determination by NMR) [24,27,39] Cd 3 -CysS 9 less dynamic, NMR structure available [5] Reaction with NO Cysteine oxidation and Zn release Zn release is faster [38] Zn release is slower [38] Reaction with ROS Cysteine-oxidation and Zn-release rates relatively similar [38] Reaction with Pt compounds Reaction with Cys, Pt bound to Cys Cisplatin and transplatin react faster [36] Cisplatin and transplatin react slower [36] Cu(I) binding Cooperative formation of Cu 4 - CysS x [33,43] Further forms: Cu 8 MT (two Cu 4 -CysS x clusters in each domain) Cu 12 MT (Cu 6 -CysS 9 and Cu 6 -CysS 11 )[33] Cu 4 -CysS 8-9 [32,33] Cu 4 -CysS 6-7 [44] Established that first Cu 4 -CysS 8-9 is in the N-terminal domain [33] Not established, but evidence provided Cluster stable in air [12,32] Cluster not stable in air [44] Redox-labile site in Zn 4 cluster of Cu 4 Zn 4 MT-3 [32] formation of Cu 4 Zn 3 MT-3 with disulfide Not studied K d of Cu(I) K d estimated to be 1 · 10 )19 M Stronger Cu(I) affinity? [35] Weaker Cu(I) affinity? [35] Cu(II) binding to Zn 7 -MT MT binds Cu(II) after reduction to Cu(I) through cysteine oxidation Formation of Cu(I) 4 Zn 4 MT-3 with two disulfide bridges [37] Zn form not studied P. Faller Metal–thiolate clusters in metallothionein-3 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works 2923 [1]. A mixture of Zn(II) and Cu(I) has also been found in MT3 isolated from bovine and equine brain [11,12] as well as from mice with disrupted Mt1 and Mt2 genes [13]. In all cases only homometallic clusters have been reported and the predominant Zn ⁄ Cu form might thus be Cu(I) 4 Zn(II) 4 -MT3[4]. The physiological accu- mulation of Cu in MT3 has been questioned as it was speculated that in vivo MT3 binds Zn almost exclu- sively [13]. This draws support from the demonstration that during the purification of Zn(II)-MT3, metal exchange reactions do occur. It could allow picking up the more firmly binding Cu + during purification, but, of course, yields no proof that MT3 functions solely as a Zn protein(II) [13]. Clearly, further investigations are needed to clarify how much, and under what condi- tions, Cu(I) is bound to MT3. The binding of Cu to MT3 has to be considered to occur not only intracellu- larly because it is now known that MT3 also occurs extracellularly [14] and evidence is accumulating that Cu can be released into the synaptic cleft [15]. More- over, there is evidence that MT3 can bind Cu(I) when Cu homeostasis breaks down, such as in AD [16]. Binding of Zn and Cd to MT3 MT3 binds predominantly seven Zn(II) or Cd(II) ions with overall apparent dissociation constants, at pH 7.4, of 1.6 · 10 )11 m and 5.0 · 10 )15 m, respectively [17]. The three-metal cluster was less stable than the four-metal cluster for Zn(II) and Cd(II) [18]. Initially, information about the presence of two separate metal– thiolate clusters came from spectroscopic studies and comparison with other MTs of Zn(II)- and Cd(II)- MT3, as well as their individual domains [12,19–22]. Precise structural data were obtained by NMR show- ing a Cd(II) 4 -CysS 11 cluster in the C-terminal domain with Cd(II)-Cys connectivities identical to those found in the structure of human MT2 [23,24]. Such informa- tion is still lacking for the N-terminal cluster, but molecular dynamics simulation proposed a Cd(II) 3 - CysS 9 cluster structure essentially identical to that of MT2 [9,25]. Thus, apart from determining the Cd(II) 3 - CysS 9 structure in MT3, the confirmation that Cd replaces Zn isostructurally (as shown for MT2) is still required. This seems important because the isostructur- al replacement has been challenged for MT3 (but not MT2) based on the observed difference in signal intensity of the charge states of Zn(II)-MT3 and Cd(II)-MT3 in ESI-MS, suggesting that with Cd(II) the N-terminal domain [the M(II) 3 -CysS 9 cluster] has a less compact structure with Cd than with Zn [26]. No comparable difference was seen in MT2, indicating that the more open structure in MT3 with Cd(II) is not just the result of the larger ionic radius of Cd(II) over Zn(II) and that the difference in amino acid sequence plays a role. One of the most important aspects is the greater dynamics of the Cd(II) 3 -CysS 9 cluster of MT3. It was observed that the resonances of Cd(II) 3 -CysS 9 in the 113 Cd(II) NMR were much less intense than those of Cd(II) 4 -CysS 11 . Increasing the temperature did sharpen them but their intensity was not enhanced [27]. Recently, Wang et al. [24] confirmed the low intensity of the resonances of Cd(II) 3 -CysS 9 , although their dif- ference from those of Cd(II) 4 -CysS 11 was smaller. Moreover, NMR measurements of mouse MT3 and human MT confirmed a high dynamical structure caused by rapid internal motion (mostly of the first 12 amino acids) [24] and this was viewed as conforma- tional exchange broadening. This dynamic structure, only observed in the Cd(II) 3 -CysS 9 of MT3, is intrigu- ing because it correlates with the growth-inhibitory activity [9,28] as well with the higher chemical reactiv- ity of the metal cluster (e.g. with NO·, see later). Moreover, it is also the reason why the determination of the spatial structure of the N-terminal domain con- taining the Cd(II) 3 -CysS 9 cluster was thus precluded. For a more detailed discussion of the dynamics of the protein structure and its implication for the biological function, see the minireview by Huang et al. in this minireview series [29]. What is the reason for the higher dynamics of the MT3 structure? Initially it was proposed that slow exchange occurs between alternative configurations involving CysS-Cd(II) bond breaking ⁄ formation in the conversion, which may include a cis–trans isomeriza- tion of the Cys-Pro amide bonds in the Cys 6 -Pro-Cys- Pro(6–9) motif [17]. Only one configuration is detected by 113 Cd(II)-NMR, whereas the other is broadened beyond detection by fast exchange processes. However, another possibility has been proposed by Palumaa et al. [30]. They found, by ESI-MS, that Zn(II) and Cd(II) do not bind cooperatively to MT3 at pH 7.3. Upon adding seven equivalents of Cd(II) or Zn(II) to MT3, distributions of metal loading were detected, ranging from five to nine for Cd(II) and from six to eight for Zn(II) (note that identical experiments with MT1A showed more homogeneous M(II)-binding of seven metals per MT [30]). This would mean that Cd(II) 7 -MT3, and, to a lesser extent, also Zn(II) 7 - MT3, are heterogeneous in their metal content, and that metal-exchange reactions between different metal-loaded forms could be the reason for the higher dynamics detected in NMR. However, this conclusion has its limits, first because the MS analysis is not quantitative and, second, because more recent MS Metal–thiolate clusters in metallothionein-3 P. Faller 2924 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works measurements did not confirm binding of more than eight equivalents of Zn(II) [31]. Importantly, the bind- ing of the eighth metal ion was confirmed by other methods and studied in more detail. [31] An additional equivalent of Zn(II) can bind to Zn(II) 7 -MT3 [but not to Zn(II) 7 -MT2] with an apparent K d of 0.1 mm. Simi- larly, Cd binds to Cd(II) 7 -MT3 but with an even stron- ger affinity (exact K d not reported). Binding of either additional ion disturbs the CD signatures of the thio- late clusters, indicating interference with the cluster structure. Moreover, a decrease in the Stokes radius was observed, suggesting a mutual approach of the two domains. Also, the additional binding of M(II) induced slow, but appreciable, non-covalent dimeriza- tion of MT3 [40% for Zn(II) and 80% for Cd(II)]. Subsequent analysis of 113 Cd(II) NMR revealed com- plete loss of the Cd 3 -CysS 9 resonances. The reported data indicate that the b-domain provides the binding site for the eighth equivalent of Zn(II) [and Cd(II)]. Moreover, the occurrence of a weaker eighth binding site in Cd(II) 7 -MT could be involved in metal- exchange reactions and hence contribute to the flexibil- ity of the Cd(II) 3 -CysS 9 cluster. This is supported by the almost complete loss of resonance intensity upon binding of an additional Cd(II). Disulfide bond formation might be another way to produce heterogeneity and ⁄ or increased structural dynamics in the b-domain, leading to partial disulfide bonds and ⁄ or to disulfide exchange reactions, respec- tively. No evidence for disulfide bond formation in Zn(II) 7 -MT3 has been reported. Nevertheless, it might be worthwhile to re-investigate this point because oxidation of cysteine was noted in the Zn(II) 4 -CysS 11 cluster of freshly prepared Cu(I) 4 Zn(II) 4 -MT3 [32]. Definitive studies will be important to explore the isostuctural replacement of Zn(II) with Cd(II), by monitoring the peptide structure and the flexibility of Zn(II) and Cd(II) by NMR spectroscopy. Another interesting (and to my knowledge not yet reported) experiment would be assessment of the growth-inhibi- tory activity of the Cd-containing form of MT3. If affirmed, this would support true isostructural replace- ment, as this MT3-specific activity is believed to be a structure-dependent feature. A problem with such a measurement could be the toxicity of Cd(II). Neverthe- less it might work as Cd(II) is tightly bound to MT-3 and Cd(II)-MT1 ⁄ 2 could be used as a control, the issue of Cd toxicity may be overcome. So far the struc- ture determination of the b-domain of Cd7-MT-3 by NMR was hampered by the high structural dynamics and exchange reactions. The recently gained insights discussed above (dimerization, disulfide formation, additional Zn(II) ⁄ Cd(II) binding and Mg ⁄ Ca effects) might be used to slow down or increase the time regime of exchange reactions as such that they are more favorable for NMR studies [30–32]. It seems clear now that Zn 7 -MT3 and Cd 7 -MT3 can bind specifically an additional eighth equivalent of Zn(II) or Cd(II). Binding of more Zn(II) ⁄ Cd(II) is very likely to be non-specific. Whether the binding of seven equivalents of Zn(II) ⁄ Cd(II) in MT3 is more heteroge- neous than the binding of seven equivalents of Zn(II) ⁄ Cd(II) to MT1 ⁄ 2 is still not known. There are indications that binding of Cd is less cooperative in MT3 compared with MT2 [17]. For Zn this is less clear. Binding of Cu(I) to MT-3 The spectroscopic characterization of MT3 isolated from bovine and equine brain showed that Cu is bound in the oxidation state I in a four Cu(I)–thiolate cluster, Cu(I) 4 -CysS X [12,20]. Furthermore, reconstitu- tion experiments with human apo-T3 and its separate domains reproduced closely the features of the isolated native MT3 forms (bovine, equine) [32,33]. Moreover, titration experiments of Cu(I) to apo-T3 (or to apo- a- and apo-b-domains) showed cooperative formation of the Cu–thiolate cluster involving eight or nine cyste- ines [i.e. Cu(I) 4 -CysS 8-9 ] [21,22,33]. In apo-T3 the first cluster formed, Cu(I) 4 -CysS 8-9 , was localized in the N-terminal domain [33]. The structure of this Cu(I) 4 - CysS 8-9 cluster is not known, but extended X-ray absorption spectroscopy data yield a Cu-Cu distance of 2.67 A ˚ and a Cu-S distance of 2.26 A ˚ . The latter points to mainly trigonal coordination of Cu(I) [the correlation would predict one or two digonal bound Cu(I)]. This clearly suggests that Cu(I) and Zn(II)⁄ Cd(II) bind preferentially to the b- and a-domains, respectively, and that no heterometallic clusters con- taining both Cu(I) and Zn(II) ⁄ Cd(II) are formed. Addition of Cu(I) beyond four equivalents results in the cooperative formation of a Cu(I) 4 -CysS 8-9 cluster in the C-terminal domain. After completion of the two Cu(I) 4 -CysS 8-9 clusters, further addition of Cu(I) results in the formation of Cu(I) 6 -CysS 9 clusters and Cu(I) 6 -CysS 11 in the N-terminal and C-terminal domains, respectively [33]. The dissociation constant of Cu(I) to MT3 has not been measured, but by analogy with other MTs it can be estimated to be around 10 )19 m [34]. Judged from the higher reactivity of Cu(I)-MT1 with 5,5¢-dithio- bis(2-nitrobenzoic acid) (DTNB) compared with Cu(I)- MT3, it has been proposed that the affinity of MT3 for Cu(I) is higher than that of MT1 ⁄ 2 [35]. One of the remarkable features of Cu(I) 4 -CysS 8-9 clusters in the isolated Cu(I) 4 -Zn 3-4 MT-3 is its stability in air. P. Faller Metal–thiolate clusters in metallothionein-3 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works 2925 The spectroscopic features of isolated Cu(I) 4 -Zn 3-4 MT- 3 did not change over time in air [12]. The Cu(I) 4 clus- ter in freshly reconstituted Cu(I) 4 -Zn 4 MT-3 also seemed to be stable, although oxidation occurred in the Zn cluster [32]. Indeed, exposure of Cu(I) 4 - Zn 4 MT3 to air resulted in the slow formation of a disulfide linkage by cysteine oxidation of the a-domain and a concominant release of one Zn ion, but the spec- troscopic features of the Cu(I) 4 cluster did not change. This might indicate that in the isolated Cu(I) 4 -Zn 4 MT3 this oxidation had already occurred (for further discus- sion see below). The molecular basis of the air stability of the Cu(I) 4 cluster is not known. Cu(I)–thiolates are normally oxidized by molecular oxygen, resulting in the formation of disulfide bonds (Eqn 1). Thiyl radi- cals have been observed as an intermediate. The mech- anism might be oxidation of Cu(I) by oxygen (Eqn 2). The formed superoxide might oxidize a further equiva- lent of Cu(I) (Eqn 3). A two-electron reaction of Cu(I) with oxygen yielding H 2 O 2 is probably favored (Eqn 4) as the reduction of O 2 to O 2 •) is thermodynamically unfavored, but the two-electron oxidation of O 2 to H 2 O 2 is favored. 2CuðIÞþ2S À þ O 2 þ 2H þ ! 2CuðIÞþS À S þ H 2 O 2 ð1Þ CuðIÞþO 2 ! CuðIIÞþO À 2 ð2Þ CuðIÞþO 2 À þ 2H þ ! CuðIIÞþH 2 O 2 ð3Þ ðEqn1þ Eqn3Þ2CuðIÞþO 2 þ2H þ ! 2CuðIIÞþH 2 O 2 ð4Þ CuðIIÞÀS À ! CuðIÞÀS  ð5Þ 2S  ! S À S ð6Þ Then, Cu(II) can oxidize thiolate to thiyl (Eqn 5) and two thiyls form a disulfide bridge (Eqn 6). [It is also possible for S • to react with a neighboring thiolate to yield a disulfide radical anion (S-S •) ), which can be further oxidized to disulfide]. In the framework of this mechanism several possibilities can be envisaged to explain the air-stability of Cu(I) 4 -Zn 4 MT. The first possibility is no accessibility of oxygen to the cluster. This is unlikely because MT3 is more dynamic and hence more exposed to the solvent. The second possi- bility is steric hindrance to form disulfide bridges, and the third is that the redox potential of Cu(I) [or the Cu(I)–thiolate moiety] is high enough that it is not oxi- dized by molecular oxygen. In this context it is note- worthy that absorption data indicate that a different number of cysteines is bound to Cu(I) (i.e. eight or nine in MT3 and six or seven in MT1 ⁄ 2). This indi- cates that the cluster structure is different. More cyste- ines involved means either fewer bridging cysteines or a higher coordination number of MT3 compared with MT1 ⁄ 2, features that might be responsible for the sta- bility in air. Reactivity of the metal–thiolate clusters in MT3 First, some general considerations about the reactivity of metal–thiolate clusters, normally concerning all metallothioneins, are given. In the case of Zn (and Cd) the metal is bound by thiolates (i.e. deprotonated thiol groups of cysteine). Thiolate is a soft ligand and therefore shows a preference for soft metals. More- over the structures of metallothioneins are not rigid and hence there is little selectivity concerning the size of the ion. Thus, the affinity of the different metal ions in MTs is governed primarily by the thermody- namic stability of the thiolate–metal bond. Therefore, soft metals such as Cu(I), Cd(II), Hg(II), Pb(II) and Pt(II) bind more strongly than Zn(II), leading to Zn(II) release. The cysteine side chains are also very reactive. At physiological pH, free cysteine is predominantly pro- tonated. The availability of Zn(II) leads to their depro- tonation at physiological pH, yielding Zn–thiolate complexes. This renders thiolates more nucleophilic than thiols. However, their binding to Zn(II) generally also inhibits the formation of disulfides under aerobic conditions. The latter occurs with uncoordinated thio- lates. Thus, Zn(II) binding elicits the formation of thiolates, which are more reactive than thiols, but less reactive than uncoordinated thiolates. This can be con- sidered as the sulfur reactivity of MT on a biological time scale is controled by the Zn-binding state. Thiols would react too slowly, whereas free thiolates would react too fast and be difficult to control. Metal-centered reactions of MT3 In the framework of considering thiolates as simple metal ligands, metal-exchange reactions such as Cd(II) or Hg(II) with Zn(II)-MT have been studied with MT1 ⁄ 2. However, as MT3 synthesis is not inducible by exposure to metal ions, a role in detoxification is less likely, which is probably the reason why metal- exchange reactions with Cd(II), Hg(II), Pb(II), etc., have not been studied so far. Moreover, the interac- tion of Zn(II) 7 MT3 with biologically relevant Cu(I) has not been reported. By contrast, the reaction of MT3 with cis-amminedichloridoplatinum(II) (cisplatin) and trans-amminedichloridoplatinum(II) (transplatin) has been studied. These reactions are of interest because MTs play an important role in the acquired Metal–thiolate clusters in metallothionein-3 P. Faller 2926 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works resistance of platinum-based anticancer drugs. Here, MT3 is important because its gene is overexpressed in a number of cancer tissues ([36] and references therein). It was shown that cisplatin and transplatin react with the cysteines in Zn(II) 7 MT3, causing the stoichiometric release of Zn(II). The reactions were much faster than with MT2. Transplatin reacted more quickly, but retained the two ligands. By contrast, the slow-reacting cisplatin had all ligands replaced with thiolates. Cisplatin binds preferentially to the b-domain (but binding with transplatin has not been determined). A special case of metal-exchange reaction is Cu(II), because it involves, in addition to its exchange with Zn, also the reduction to Cu(I) by the oxidation of cysteine. Meloni et al. [37] showed that Zn(II) 7 MT3 scavenges free Cu 2+ ions through reduction to Cu(I) and binding to the protein. In this reaction, thiolate ligands are oxidized to disulfides concomitant with Zn 2+ release. The binding of the first four Cu 2+ is cooperative, forming a Cu(I) 4 –thiolate cluster in the N-terminal domain of Cu(I) 4 ,Zn(II) 4 MT3 together with two disulfide bonds. Because four zinc ions remain bound, it seems likely that the four-metal Zn 4 – thiolate cluster in the a-domain stayed intact. As a consequence the two disulfides would be localized in the b-domain with the Cu(I) 4 –thiolate cluster. The formed Cu(I) 4 –thiolate cluster has spectroscopic prop- erties similar to the isolated and the Cu(I)-reconsti- tuted Cu(I) 4 Zn(II) 4 MT3 described above, including stability in air. The reaction of Zn(II) 7 MT3 with Cu(II) has been proposed to be the underlying mecha- nism for the protective effect of MT3 against b-amy- loid (Ab) neurotoxicity linked to AD. It was shown that a metal swap between Zn(II) 7 MT3 and soluble and aggregated Cu(II)–Ab abolishes the production of reactive oxygen species and the related cellular toxicity [16]. Thus, MT3 might have a role in protecting Ab from aberrant Cu binding. Thiolate-centered reactions (NO • , reactive oxygen species, DTNB) Metals in MTs are relatively deeply buried in the pro- tein. The modification of the surface-accessible sulfur of cysteine ligands is thought to be the key that unlocks the metals from the protein [38]. Thus, in gen- eral, reactions of the cysteine thiolates result in a con- comitant release of Zn(II). A variety of reagents reacting with thiolate have been studied in MTs [5], but such reactions are limited for MT3. The reaction with NO • has attracted particular attention [38,39]. Most interest in the reaction with NO • comes from the possible role of MTs as NO • scavengers and in the conversion of a NO • signal to a Zn(II) signal. The reaction with NO • providing S-nitrosothiols is sug- gested to be a transnitrosation (i.e. translocation of NO + from the S-nitrosothiols to the cysteine of MT), which then releases NO ) during the formation of disul- fides [38]. This means that NO ) is not stable in the MTs and a storage function of MTs for NO· is less likely. By comparing the reaction of NO and S-nitros- othiols in MT3 and MT1 ⁄ 2, Chen et al. [38] found that MT3 was much more reactive, whereas the activi- ties with reactive oxygen species (H 2 O 2 , OCl ) ,O 2 •) ) were comparable. In line with this, MT3 was also more potent in protecting rat embryonic cortical neurons against S-nitrosothiols. The b-domain showed greatest reactivity in Zn(II) release and cell protection. The higher reactivity of the b-domain in MT3 versus NO has been confirmed by NMR studies on Cd(II) 7 MT3 [39]. Moreover, the NMR data suggest a non-selective release of the metals from the b-domain first, followed by a partial release of two Cd(II) ions from the a-domain, without a significant change in the poly- peptide structure. Further addition of NO resulted in a complete loss of protein structure. DTNB has often been used to probe the nucleophilic reactivity of thiolates in MTs. DTNB contains an intra- molecular disulfide bond and, upon nucleophilic attack from MT, disulfide exchange occurs resulting in an in- termolecular disulfide bond between MT and 5-thio-2- nitrobenzoic acid. It has been shown that Cd(II) 7 MT3 reacts faster with DTNB than MT1 ⁄ 2 does [35,40]. For Zn(II) 7 MT3, similar kinetics have been reported for human MT3 and rat MT1, but this was monitored in the presence of EDTA and hence is more likely to reflect the reactivity of the unstructured apo-T than that of the metal-loaded form [19]. In general, it can be suggested that MT3 is more reactive than MT1 ⁄ 2, which is probably related to its more flexible b-domain and hence to a better access of compounds to the metal–thiolate cluster. This is also in line with the general (but not exclusive) observation that the difference in reactivity is more pronounced for larger molecules (DTNB, providing S-nitrosothiols) than for small molecules (H 2 O 2 , OCl ) ,O 2 •) ). Conclusions and Perspectives MT3 is clearly a member of the MT family as it shares with them several biological and chemical properties; however, there are also very distinct chemical features that might be directly relevant to the particular biolog- ical properties of MT3. Therefore, comparison of the properties of MT3 with those of the well-studied P. Faller Metal–thiolate clusters in metallothionein-3 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works 2927 MT1 ⁄ 2 forms could yield information on the structural and chemical features responsible for the biological peculiarities of MT3, such as its growth-inhibitory activity. The fact that synthesis of MT3, in contrast to that of MT1 ⁄ 2, is not inducible by metal ions (Zn, Cu, Cd, etc.) suggests that it has no essential role in sequester- ing toxic or an overload of essential metals. This is in line with a lower binding affinity to Cd(II) for MT3 compared with MT1 ⁄ 2. One of the most striking features is the greater struc- tural dynamics and flexibility of MT3 and in particular of the N-terminal b-domain. This seems to be connected with the lower metal-binding affinity and with the higher reactivity towards nucleophilic reagents (NO • ,Pt compounds, DTNB) and the ease of Zn release. One could speculate that higher structural dynamics should also result in a faster metal transfer from and to MT3 and hence is in line with an involvement in the traffick- ing of Zn(II) in particular in zinc-containing neurons [41]. However, metal-exchange rates have not yet been measured experimentally. Importantly, MT3 shows increased Zn(II) release compared with MT1 ⁄ 2 upon reacting with NO • and other compounds, supporting a function in Zn meta- bolism. It was suggested that MT3 could be involved in turning a NO • signal into a Zn signal [38]. Although the question of whether Cu(I) is physio- logically bound to MT3 is not as yet resolved (see ear- lier), it is clear that MT3 has the capacity of binding Cu(I) under conditions of Cu-homeostasis breakdown. There are several unanswered questions concerning Cu(I)MT3. First the exact cluster structures of the dif- ferent forms are not known [i.e. Cu(I) 4 ,Zn(II) 4 MT3 (with or without disulfides), Cu(I) 4 ,Cu(I) 4 MT3 and Cu(I) 6 ,Cu(I) 6 MT3]. The determination of a 3D struc- ture seems crucial for understanding the differences between Cu–MT3 and Cu–MT1 ⁄ 2 (structure also not known) and the presence and role of the disulfide bonds. This would also give insight into the stability of the Cu(I) 4 -CysS X cluster in the b-domain of MT3 towards oxidation by molecular oxygen. The data seem to be contradictory (see above) as the Cu(I) 4 cluster is stable in air, but freshly reconstituted Cu(I) 4 ,Zn(II) 4 MT3 shows an oxidation reaction forming disulfide bonds in the Zn-loaded a-cluster. One could ask why this reaction seems not to occur in the Zn(II) 7 -MT-3, as in Cu(I) 4 -Zn(II) 4 -MT-3 the Cu(I) 4 -cluster seems not to be involved? Moreover, in the reaction of Cu(II) with Zn(II) 7 MT3, a Cu(I) 4 cluster and two disulfide bridges are formed in the b-domain, while the remain- ing Zn cluster seems stable. The spectroscopic features of this Cu(I) 4 cluster are very similar to those of the reconstituted forms. One partial explanation would be that the spectroscopic features of the Cu(I) 4 cluster are not affected by the presence of disulfide bridges and this cluster is only stable in air in the presence of one or two disulfide bonds. This could explain the sensitiv- ity to oxidation of freshly reconstituted Cu(I) 4 , Zn(II) 4 MT3 and the stability of Cu(I) 4 , Zn(II) 4 MT3 [isolated, incubated or generated upon Cu(II) binding] in air. However, this does not explain why the disulfide bridge is formed in the a-domain (instead of the b-domain) upon oxidation of freshly reconstituted Cu(I) 4 ,Zn(II) 4 MT3. To shed more light on this issue it might be worthwhile investigating the number and localization of disulfide bridges in diverse preparations of MT3. With regard to the putative role of MT3 in Cu traf- ficking, it would be important to determine the binding constants. Very little is known about Cu(I) affinity in the MTs, and the values in the literature are mostly estimates. This is mainly because of the very low disso- ciation constants, with K d estimated to be about 10 )19 m, and the lack of suitable competing ligands with well-known binding constants. However, even rel- ative affinities could give important insights, such as comparison of MT3 with other MTs. The comparison reported in the literature, based on reactivity with DTNB, is indirect (see above) [35]. Relative Cu(I) affinities between MT3 and other MTs should be mea- sured by a competition assay using MS, as accom- plished previously for Zn(II) and Cd(II) [18], or by NMR analysis. The high affinity of Cu(I) to MT3 (or to MTs in general) means that Cu(I) release into solvent, as a result of thiolate bonding, is too slow to be biologi- cally relevant. Therefore, transfer of Cu(I) is impossi- ble via ‘free’ Cu(I). One way to transfer Cu(I) to another protein on a biologically relevant time scale is through the formation of a ternary complex [MT– Cu(I)–protein] (i.e. by a coordination bridge forming an interaction between MT and the acceptor protein). Other possibilities are that the transfer is assisted by cysteine oxidation ⁄ modification, by protonation or by protein breakdown. If the Cu(I) transfer did not taking place, MT would just be a sink for Cu(I). This could be sufficient for a redox-silencing role of MT3 for Cu. In this context it might be interesting to search for possible binding partners of Cu(I) 4 MT3. Since the discovery of MT3 [1] almost 20 years ago, it has been discovered that this member of the family has unusual biological and chemical properties, clearly distinct from the widely expressed MT1 ⁄ 2. This holds also for structure and reactivity of the metal–thiolate clusters, in particular for the cluster in the b-domain. Metal–thiolate clusters in metallothionein-3 P. Faller 2928 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works Several intriguing facets have been observed, such as high dynamics, formation of disulfides, high reactivity, stability of the Cu(I) 4 -cluster, etc. A better understand- ing of these features will help to shed light on the specific biological roles of MT3. Acknowledgement Gabriele Meloni (Caltech, USA) and Milan Vasak (Univ. Zu ¨ rich) are acknowledged for very helpful discussion. References 1 Uchida Y, Takio K, Titani K, Ihara Y & Tomonaga M (1991) The growth inhibitory factor that is deficient in the Alzheimer’s disease brain is a 68 amino acid metal- lothionein-like protein. Neuron 7, 337–347. 2 Rigby Duncan KE & Stillman MJ (2006) Metal-depen- dent protein folding: metallation of metallothionein. J Inorg Biochem 100, 2101–2107. 3 Hidalgo J, Aschner M, Zatta P & Vasak M (2001) Roles of the metallothionein family of proteins in the central nervous system. Brain Res Bull 55, 133–145. 4 Hidalgo J, Chung RS, Penkowa M & Vasak M (2009) Structure and function of vertebrate metallothioneins. Met Ions Life Sci 5, 279–317. 5 Romero-Isart N & Vasak M (2002) Advances in the structure and chemistry of metallothioneins. J Inorg Biochem 88, 388–396. 6 Calderone V, Dolderer B, Hartmann HJ, Echner H, Luchinat C, Del Bianco C, Mangani S & Weser U (2005) The crystal structure of yeast copper thionein: the solution of a long-lasting enigma. Proc Natl Acad Sci USA 102, 51–56. 7 Dolderer B, Echner H, Beck A, Hartmann HJ, Weser U, Luchinat C & Del Bianco C (2007) Coordination of three and four Cu(I) to the alpha- and beta-domain of vertebrate Zn-metallothionein-1, respectively, induces significant structural changes. Febs J 274, 2349–2362. 8 Krezel A & Maret W (2007) Dual nanomolar and picomolar Zn(II) binding properties of metallothionein. J Am Chem Soc 129, 10911–10921. 9 Ding ZC, Zheng Q, Cai B, Ni FY, Yu WH, Teng XC, Gao Y, Liu F, Chen D, Wang Y et al. (2008) Study on structure-property-reactivity-function relationship of human neuronal growth inhibitory factor (hGIF). J Inorg Biochem 102, 1965–1972. 10 Howells C, West AK & Chung RS (2010) Neuronal growth inhibitory factor (metallothionein-3): evaluation of the biological function of GIF in the injured and neurodegenerative brain. FEBS J 277, 2931–2939. 11 Pountney DL, Fundel SM, Faller P, Birchler NE, Hunziker P & Vasak M (1994) Isolation, primary struc- tures and metal binding properties of neuronal growth inhibitory factor (GIF) from bovine and equine brain. FEBS Lett 345, 193–197. 12 Bogumil R, Faller P, Pountney DL & Vasak M (1996) Evidence for Cu(I) clusters and Zn(II) clusters in neuro- nal growth-inhibitory factor isolated from bovine brain. Eur J Biochem 238, 698–705. 13 Masters BA, Quaife CJ, Erickson JC, Kelly EJ, Froelick GJ, Zambrowicz BP, Brinster RL & Palmiter RD (1994) Metallothionein III is expressed in neurons that sequester zinc in synaptic vesicles. J Neurosci 14, 5844–5857. 14 Lynes MA, Zaffuto K, Unfricht DW, Marusov G, Sam- son JS & Yin X (2006) The physiological roles of extra- cellular metallothionein. Exp Biol Med (Maywood) 231, 1548–1554. 15 Schlief ML, Craig AM & Gitlin JD (2005) NMDA Receptor Activation Mediates Copper Homeostasis in Hippocampal Neurons. J Neurosci 25, 239–246. 16 Meloni G, Sonois V, Delaine T, Guilloreau L, Gillet A, Teissie J, Faller P & Vasak M (2008) Metal swap between Zn7-metallothionein-3 and amyloid-beta-Cu protects against amyloid-beta toxicity. Nat Chem Biol 4, 366–372. 17 Hasler DW, Jensen LT, Zerbe O, Winge DR & Vasak M (2000) Effect of the two conserved prolines of human growth inhibitory factor (metallothionein-3) on its biological activity and structure fluctuation: comparison with a mutant protein. Biochemistry 39, 14567–14575. 18 Palumaa P, Tammiste I, Kruusel K, Kangur L, Jornvall H & Sillard R (2005) Metal binding of metallothionein- 3 versus metallothionein-2: lower affinity and higher plasticity. Biochim Biophys Acta 1747 , 205–211. 19 Sewell AK, Jensen LT, Erickson JC, Palmiter RD & Winge DR (1995) Bioactivity of metallothionein-3 correlates with its novel beta domain sequence rather than metal binding properties. Biochemistry 34, 4740– 4747. 20 Bogumil R, Faller P, Binz PA, Vasak M, Charnock JM & Garner CD (1998) Structural characterization of Cu(I) and Zn(II) sites in neuronal-growth-inhibitory factor by extended X-ray absorption fine structure (EXAFS). Eur J Biochem 255, 172–177. 21 Faller P & Vasak M (1997) Distinct metal-thiolate clus- ters in the N-terminal domain of neuronal growth inhibitory factor. Biochemistry 36, 13341–13348. 22 Hasler DW, Faller P & Vasak M (1998) Metal-thiolate clusters in the C-terminal domain of human neuronal growth inhibitory factor (GIF). Biochemistry 37, 14966– 14973. 23 Oz G, Zangger K & Armitage IM (2001) Three-dimen- sional structure and dynamics of a brain specific growth P. Faller Metal–thiolate clusters in metallothionein-3 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works 2929 inhibitory factor: metallothionein-3. Biochemistry 40, 11433–11441. 24 Wang H, Zhang Q, Cai B, Li H, Sze KH, Huang ZX, Wu HM & Sun H (2006) Solution structure and dynamics of human metallothionein-3 (MT-3). FEBS Lett 580, 795–800. 25 Ni FY, Cai B, Ding ZC, Zheng F, Zhang MJ, Wu HM, Sun HZ & Huang ZX (2007) Structural prediction of the beta-domain of metallothionein-3 by molecular dynamics simulation. Proteins 68, 255–266. 26 Palumaa P, Njunkova O, Pokras L, Eriste E, Jornvall H & Sillard R (2002) Evidence for non-isostructural replacement of Zn(2+) with Cd(2+) in the beta- domain of brain-specific metallothionein-3. FEBS Lett 527, 76–80. 27 Faller P, Hasler DW, Zerbe O, Klauser S, Winge DR & Vasak M (1999) Evidence for a dynamic structure of human neuronal growth inhibitory factor and for major rearrangements of its metal-thiolate clusters. Biochemis- try 38, 10158–10167. 28 Romero-Isart N, Jensen LT, Zerbe O, Winge DR & Vasak M (2002) Engineering of metallothionein-3 neuroinhibitory activity into the inactive isoform metallothionein-1. J Biol Chem 277, 37023–37028. 29 Ding ZC, Ni FY & Huang ZX (2010) Neuronal growth inhibitory factor (metallothionein-3): structure-function relationships. FEBS J 277, 2912–2920. 30 Palumaa P, Eriste E, Njunkova O, Pokras L, Jornvall H & Sillard R (2002) Brain-specific metallothionein-3 has higher metal-binding capacity than ubiquitous metallothioneins and binds metals noncooperatively. Biochemistry 41, 6158–6163. 31 Meloni G, Polanski T, Braun O & Vasak M (2009) Effects of Zn(2+), Ca(2+), and Mg(2+) on the struc- ture of Zn(7)metallothionein-3: evidence for an addi- tional zinc binding site. Biochemistry 48, 5700–5707. 32 Roschitzki B & Vasak M (2003) Redox labile site in a Zn4 cluster of Cu4,Zn4-metallothionein-3. Biochemistry 42, 9822–9828. 33 Roschitzki B & Vasak M (2002) A distinct Cu(4)-thio- late cluster of human metallothionein-3 is located in the N-terminal domain. J Biol Inorg Chem 7, 611–616. 34 Hamer DH (1986) Metallothionein. Annu Rev Biochem 55, 913–951. 35 Toriumi S, Saito T, Hosokawa T, Takahashi Y, Numata T & Kurasaki M (2005) Metal binding ability of metallothionein-3 expressed in Escherichia coli. Basic Clin Pharmacol Toxicol 96, 295–301. 36 Karotki AV & Vasak M (2009) Reaction of human me- tallothionein-3 with cisplatin and transplatin. J Biol Inorg Chem 14, 1129–1138. 37 Meloni G, Faller P & Vasak M (2007) Redox silencing of copper in metal-linked neurodegenerative disorders: reaction of Zn7metallothionein-3 with Cu2+ ions. J Biol Chem 282, 16068–16078. 38 Chen Y, Irie Y, Keung WM & Maret W (2002) S-nitr- osothiols react preferentially with zinc thiolate clusters of metallothionein III through transnitrosation. Biochemistry 41, 8360–8367. 39 Wang H, Li H, Cai B, Huang ZX & Sun H (2008) The effect of nitric oxide on metal release from metallothi- onein-3: gradual unfolding of the protein. J Biol Inorg Chem 13, 411–419. 40 Ding ZC, Teng XC, Cai B, Wang H, Zheng Q, Wang Y, Zhou GM, Zhang MJ, Wu HM, Sun HZ et al. (2006) Mutation at Glu23 eliminates the neuron growth inhibitory activity of human metallothionein-3. Biochem Biophys Res Commun 349, 674–682. 41 Knipp M, Meloni G, Roschitzki B & Vasak M (2005) Zn7metallothionein-3 and the synaptic vesicle cycle: interaction of metallothionein-3 with the small GTPase Rab3A. Biochemistry 44, 3159–3165. 42 Irie Y & Keung WM (2001) Metallothionein-III antago- nizes the neurotoxic and neurotrophic effects of amyloid beta peptides. Biochem Biophys Res Commun 282, 416– 420. 43 Jensen LT, Peltier JM & Winge DR (1998) Identification of a four copper folding intermediate in mammalian copper metallothionein by electrospray ionization mass spectrometry. J Biol Inorg Chem 3, 627–631. 44 Pountney DL, Schauwecker I, Zarn J & Vasak M (1994) Formation of mammalian Cu8-metallothionein in vitro: evidence for the existence of two Cu(I)4-thio- late clusters. Biochemistry 33, 9699–9705. Metal–thiolate clusters in metallothionein-3 P. Faller 2930 FEBS Journal 277 (2010) 2921–2930 Journal compilation ª 2010 FEBS. No claim to original French government works . MINIREVIEW Neuronal growth-inhibitory factor (metallothionein-3): reactivity and structure of metal–thiolate clusters* Peter Faller 1,2 1. of cysteine). Thiolate is a soft ligand and therefore shows a preference for soft metals. More- over the structures of metallothioneins are not rigid and