Environmental Health PerspectivesVol. 54, pp. 135-145, 1984

NMR Analysis of the Structure and MetalSequestering Properties of Metallothioneinsby Catherine T. Hunt,* Yvan Boulanger,t Stephen W. Fesikland Ian M. Armitage*

113Cd-NMR studies have been used to elucidate the structure of the metal-binding sitesin mammalian and invertebrate (Scylla serrata) metaliothioneins (MTs). Chemical shiftdata have shown that all Cd ions are tetrahedrally coordinated to four cysteine thiolateligands with single cysteinyl sulfurs bridging adjacent metals. Homonuclear decouplingexperiments have shown that the 7 g-atoms of metal bound per mole of mammalianprotein are located in a three- and a four-metal cluster while the 6 g-atoms of metal in theinvertebrate MT are located in two three-metal clusters. The different metal bindingaffinities ofthe two mammalian clusters have been determined by 113Cd-NMR. The three-metal cluster prefers Cu > Zn > Cd whereas exactly the reverse order applies in the four-metal cluster. Proteolytic cleavage of the protein produced a 32-residue fragment whichcontained the four-metal cluster and demonstrated the presence oftwo separate domainsin the protein. 500 MHz 'H-NMR has been employed to elucidate the arrangement ofthese metal clusters in the tertiary structure of the protein. The 1H resonances wereassigned from their scalar and dipolar connectivities obtained from extensive one andtwo-dimensional NMR experiments. A specific application of 2D correlation spectros-copy (COSY) to the assignment of the 1H resonances in crab MT-1 is discussed. Amolecular model, representing the three-dimensional solution structure of this protein,has been constructed based on an analysis of all thepe data. Detailed structural featuresof this model are discussed, with particular emphasis on their relationship to the functionand evolution of the protein.

IntroductionMetallothioneins (MTs) are a class of low mo-

lecular weight (-6100 dalton), cysteine-rich,metal-binding proteins found ubiquitously in na-ture (1-4). The protein is known to bind variousmetal ions such as cadmium, zinc, copper andmercury (5), and its biosynthesis is closely regu-lated by the level of exposure of an organism tosalts ofthese metals (6 - 8). For these reasons it isquite widely accepted that MTs function as detox-ifying agents by sequestering toxic metals (9), butit has also been suggested that MTs function inthe regulation and/or metabolism of essentialheavy metals (10,11).

*The Department of Molecular Biophysics and Biochemis-try, Yale University, New Haven, CT 06510.

tPresent affiliation: Institut de Genie Biomedical, Univer-site de Montreal, C.P. 6128, Succ. A, Montreal, Quebec, Can-ada H3C 3J7.

tPresent affiliation: Abbott Laboratories, D-47F, AbbottPark, North Chicago, IL 60064

Shortly after the discovery (12) and sequencingof metallothionein from horse kidney cortex(13,14), it was found that 20 of the 61 amino acidresidues of mammalian MTs are cysteines, all ofwhich participate in the ligation of 7 g-atoms ofmetal (5). Subsequent studies revealed that aremarkable homology exists in the amino acidsequences of all mammalian MTs (2,13-18). Infact, the positions of all the 20 cysteine residuesare invarient along with those of 18 other resi-dues. This suggests that specific metal-thiolateinteractions are important for the structural andfunctional viability of the protein. This will bediscussed in light of the fact that MTs from non-mammalian sources possess shorter amino acidchains with a smaller number of cysteines (Fig. 1)and bind fewer metals. MT generally exists in twoisoprotein forms, MT-1 and MT-2, wlhich differ inamino acid composition and total charge (19).Structural characterization of this protein hasbeen slow because of its elusive properties such

5 10 15 20 25 30 35 40 45 50 55 60HUMAN MT-II N-Ac-M-D-P-N-C-S-C-A-A-G-0-S-C-T-C-A-G-A-C-K-C-K-E-C-K-C-T-S-C-K-K-S-C-C-S- C-C-P-V-G-C-A-K-C-A-Q-G-C- I -C-K-G-A-S-D-K-C-C-S-C-A-OH

5CRAB MT-I H2N-P-G-P-C-C-N-D-K-C-V-C-K-E-G-G-C-K-E-G-C-Q-C-T-S-C-R-C-S-P-C-E-K-C-S-S- G-C-K-C-A-N-K-E-E-C-S-K-T-C-S-K-A-C-S-C-C-P-T-OH

Neurospora MT H2N-G-D-C-G-C-S-G-A-S-S-C-N-C-G-S-G-C-S-C-S-N-C-G-S-K-OH

FIGURE 1. Amino acid sequence of metallothionein from different sources.

as, the large number of repetitive residues (Cys,Lys, Ser) and the lack of both aromatic residuesand histidine (19). Previously reported UV andCD data (5,20-23) are indicative of metal-sulfurligation but assertions of tetrahedral coordina-tion (23) based on these data must be viewed withcaution because of the severe overlap of themetal-thiolate charge transfer bands. The metalions commonly found in MTs-Cu+ (24), Zn2+ andCd2+-are all diamagnetic and therefore unsuit-able for ESR studies.In view ofthese factors, 113Cd-NMR has become

a key in the study of the structure of the metal-binding sites in metallothionein. The tremendoussensitivity of 113Cd chemical shifts to subtle differ-ences in metal environments (25) has made itpossible to observe a separate 113Cd resonance foreach of the metal ions in the protein, in spite oftheir similar coordination sites (26). Analysis of113Cd-NMR data from rabbit (27), human (28), calf(29) and crab (30,31) MTs provided the first directevidence for the arrangement of these metals intwo separate polynuclear clusters. For mamma-lian MTs, the 7 g-atoms of metal are present in athree- and a four-metal cluster, while in an inver-tebrate MT the 6 g-atoms of metal are present intwo three-metal clusters (27,32). These and fur-ther studies have shown that the two clusters inmammalian MTs exhibit significant differencesin their affinities for different metal ions and thatthey function quite independently of one another(28-30,33).

In order to further extend our knowledge of thesecondary and tertiary structure of the protein,Chou-Fasman calculations and 500 MHz 'H-NMRstudies have been employed (34). Scalar and dipo-lar connectivities between the protons of severaldifferent residues were obtained from extensiveone- and two-dimensional NMR experiments.Based on an analysis of these and previouslysummarized physicochemical data (34), a molecu-lar model representing the three-dimensional so-lution structure of this protein has been con-structed. Detailed structural features of thismodel will be discussed with particular emphasison the existence of two independent domains andtheir relationship to the function and evolution ofthis protein.

'3Cd-NMR of InvertebrateMetallothioneinMetallothionein from mud crab (Scylla serrata)

hepatopancreas binds 6 g-atoms ofmetal per moleof protein and contains 58 (MT-1) or 57 (MT-2)amino acids, 18 of which are cysteines (35). Thisinvertebrate MT, like the mammalian MTs, canbe induced by the administration of cadmium(36). Native crab MT isolated from l3Cd-injectedcrabs is homogeneous in its metal composition(36,37) and is therefore particularly amenable tostudy by 113Cd-NMR (31). Figure 2A shows the'H-decoupled 113Cd-NMR spectrum of native crab113Cd-MT-1. The chemical shifts of the 113Cd reso-nances (between 620 and 660 ppm) are consistentwith coordination to four cysteinyl sulfurs(38,39). The six separate 113Cd resonances (3 and 4overlap slightly) correspond to the individualmetal binding sites. The multiplet structure ofeach of the resonances is due to 113Cd-113Cd scalarcoupling and has provided evidence that all six113Cd2+ ions are located in polynuclear metal clus-ters (25-27,32). The structure of these clusterswas determined by homonuclear decoupling ex-periments (30), two ofwhich are shown in Figures2B and 2C. In Figure 2B, selective irradiation ofresonance 1 resulted in the collapse of resonances2 and 6, while irradiating the overlapping reso-nances 3 and 4 in Figure 2C collapsed resonance 5to a singlet. It was therefore concluded that themetals are located in two three-metal clusters,one containing the 113Cd2+ ions giving rise toresonances 1, 2 and 6 and the other containingthe metals corresponding to resonances 3, 4 and 5.The magnitude of the observed 113Cd-113Cd spincoupling (19-43 Hz) is suggestive of a two-bondinteraction (31), presumably via a bridging thio-late ligand. From these data, it is possible topropose a structure for the three-metal cluster inwhich each cluster forms a six-membered ringconsisting of three sulfur and three cadmiumatoms (Fig. 3, cluster B). Each Cd2+ is tetrahe-drally coordinated to four cysteines, and thusnine cysteines are involved in the formation ofeach cluster, Cd3(Cys)9. This structure for the twothree-metal clusters in crab MT is therefore con-sistent with the participation of all 18 cysteines

136 HUNTETAL.

NMR ANALYSIS OF METALLOTHIONEINS

Cd

Cd Cd2 6

3Cd

Cd Cd4 5

Cluster A4-Metal Cluster

1 23,4 6

A

B

C A I

660 650 640pp m

630 620

FIGURE 2. ll3Cd-NMR spectra at 44.4 MHz of native crab MT-1 (-8 mM; Cd:Zn = 60:1) in 0.01 M ¶Tis, 0.1 M NaCl, pH9.0: (A) proton decoupled spectrum (23,000 transients);(B,C) same as A but with homonuclear decoupling pulsesapplied at the frequencies indicated by an arrow and theaffected resonances designated by X (31).

in metal binding. In the other isoprotein of crabmetallothionein, MT-2, the 113Cd chemical shiftsare slightly different as a result of the differentamino acid composition but the results of thehomonuclear decoupling experiments again dem-onstrate the presence oftwo three-metal clusters.

113Cd-NMR of MammalianMetallothioneinsMammalian MTs bind 7 g-atoms of metal per

mole of protein and contain 61 amino acids, 20

S7NS.\6 / \C5

Cd Cd 7 /-Cd iC S

Cd \

/ s\

Cluster B3-Metal Cluster

/ I Xs

Id S4Cd "SCd -s 14/~~/ S

FIGURE 3. Proposed structures of the four-metal and three-metal clusters of mammalian metallothioneins based on"13Cd-NMR data. The Cd numbering corresponds to thechemical shifts of the ll3Cd-NMR multiplets (Fig. 4) (27).

being sequence invarient cysteine residues (13-19). The positions of 18 other residues, mostlylysine and serine, are also invarient. This sug-gests that their positions are important to thephysiological function of the protein. The individ-ual metal coordination sites of rabbit (27), human(28) and calf (29) liver metallothioneins havebeen elucidated by 113Cd-NMR experiments simi-lar to those used in the study of native 113Cd-induced crab MT. While native crab 113Cd-MT wasfound to be homogeneous in metal content, thiswas not the case for the 113Cd-induced mamma-lian MTs studied. The cadmium-induced rabbitMTs contained, in addition to Cd2+, a substantial

137

HUNT ETAL.

amount of Zn2+ which was not randomly distrib-uted among the multiple metal binding sites (32).This heterogeneity in metal composition resultedin a 113Cd-NMR spectrum of native rabbit113Cd,Zn-MT which was complicated due to thesensitivity of the 113Cd chemical shifts to thepresence of Zn2+ at neighboring sites in the clus-ters.

In order to facilitate the assignment of the 113Cdspectrum, all of the Zn2+ ions in several rabbitMT preparations were replaced with 113Cd2+ byadding cadmium salts directly to the tissue ho-mogenate during extraction (27). The 113Cd-NMRspectrum of reconstituted rabbit 113Cd-MT-1 isshown in Figure 4A and is considerably simplerthan that ofthe native rabbit MT (27). The chemi-cal shifts of the 113Cd resonances (610-670 ppm)span a slightly larger range than those of crabMT (Figs. 4A and 4D) but are entirely consistentwith coordination to four cysteinyl sulfurs. Selec-tive homonuclear decoupling experiments on re-constituted rabbit 113Cd-MT identified the pres-ence of two separate polynuclear metal clusters.There is a three-metal cluster corresponding tothe 113Cd2+ ions giving rise to resonances 2, 3 and4 and a four-metal cluster corresponding to the113Cd2+ ions giving rise to resonances 1, 5, 6 and 7.Duplication in the resonances from the four-metal cluster (1', 5', 6' and 7') presumably stemsfrom the sensitivity of the 113Cd chemical shifts toheterogeneity in the primary structure of the MT(25). That is, each of the major isoproteins (MT-1and MT-2) from rabbit liver is a mixture of atleast two proteins.The 113Cd-NMR spectrum of reconstituted hu-

man 113Cd-MT-1, which is homogeneous in metalcomposition, is remarkably similar to that of rab-bit as seen by comparing spectra A and B inFigure 4. Complete analysis confirmed that hu-man Cd-MT contains the same two-cluster ar-rangement (28). There was no duplication of 113Cdresonances for human MT-2 because there is noheterogeneity in its primary sequence. HumanMT-1, however, shows duplication in resonance 4of the three-metal cluster and resonance 5 of thefour-metal cluster. This is consistent with the factthat human MT-1 is a mixture of at least threeproteins (40) with sequence heterogeneity at sixpositions (18).Even after reconstitution with 113Cd2+ in a

manner similar to that used for rabbit and humanMTs, calf liver MT was heterogeneous in metals,containing both Cu+ and Cd2+ in a ratio of ap-proximately 3 to 4 (29). However, unlike the hy-brid rabbit 113Cd,Zn-MT, the hybrid calf 113Cd,Cu-MT exhibited a remarkably simple spectrum as

670 650 630 610 590ppm

FIGURE 4. Comparison of the proton-decoupled 113Cd-NMRspectra at 44.4 MHz of 113Cd reconstituted mammalianmetallothioneins and ll3Cd-induced invertebrate metal-lothionein: (A) rabbit liver Cd-MT-1 (- 8 mM; Cd:Zn =40:1); (B) human liver Cd-MT-1 (-7 mM; Cd:Zn = 50:1);(C) calf liver Cd,Cu-MT-1 (-6 mM; Cd:Cu = 1.45:1); (D)Scylla serrata hepatopancreas Cd-MT-1 (-8 mM; Cd:Zn =60:1).

shown in Figure 4C. There were only four majorresonances, and these displayed chemical shiftsin positions already assigned to the four-metalcluster of human and rabbit MT. This was con-firmed by homonuclear decoupling experiments.The lack of 113Cd resonances corresponding to thethree-metal cluster (between 635 and 665 ppm)indicated that it was occupied by Cu+. The differ-ent metal affinities of the two clusters will bediscussed in a subsequent section.

In order to be consistent with these data and inorder to utilize all 20 cysteines in metal ligation,the structures shown in Figure 3 have been pro-posed for the two separate clusters in mammalianMTs. The three-metal cluster forms a cyclohex-ane-like six-membered ring requiring nine cys-teine thiolate ligands, Cd3(Cys)9, as seen in crabMT. The four-metal cluster forms a bicyclo[3:1:3]structure requiring the remaining 11 cysteinethiolate ligands, Cd4(Cys)11. This results in twotypes ofmetal coordination. Two metals (labeled 6and 7) are coordinated by three bridging sulfur

138

NMR ANALYSIS OF METALLOTHIONEINS

ligands and one nonbridging sulfur, and thesegive rise to the most shielded resonances in thespectrum. The other five metals are each ligandedto two bridging and two nonbridging thiolatesand give rise to more deshielded resonances (Figs.4A and 4B).

Metal-Binding AffinityThe 113Cd-NMR studies summarized above

have provided considerable insight into the rela-tive affinities of the two mammalian metal clus-ters for different metal ions. The 113Cd-NMRstudy of calf liver 113Cd,Cu-MT has shown thatthe 113Cd2+ ions are bound to the four-metal clus-ter, whereas the ESR-silent Cu+ ions are locatedin the three-metal cluster (29). Thus, the four-metal cluster exhibits a preference for Cd2+ overCu+, while the opposite is true for the three-metalcluster. Furthermore, since reconstituted calfliver 113Cd,Cu-MT was prepared by in vitro re-placement of Zn2+ by"13Cd2+ in native Zn,Cu-MT(29), one can conclude that the native proteincontained Zn2+ in the four-metal cluster and Cu+in the three-metal cluster. This observation es-tablishes the preference of the four-metal clusterfor Zn2+ over Cu+. When the protein was inducedin rabbits by the injection of a 113CdC12 solution,native 113Cd,Zn-MT was extracted from the liver,and an analysis of its 113Cd-NMR spectrum re-vealed that Zn2+was predominantly bound(>85%) to the three-metal cluster (25). 113Cd-NMR analysis of the metal distribution in thesethree metal-hybrid MTs therefore demonstratedthat the two metal clusters display opposite bind-ing affinities. For the four-metal cluster, the pref-erence is Cd > Zn > Cu, and for the three-metalcluster, Cu > Zn > Cd.The reverse selectivity of the two mammalian

metal clusters may be partly rationalized interms of geometric considerations. Although Cd2+and Zn2+ both prefer tetrahedral coordinationwith sulfur ligands, it is interesting to note thatthe Zn-S bond length (2.34 A) is shorter than theCd-S bond length (2.52 A) as a result of thesmaller ionic radius of Zn2 + (0.69 A) relative toCd2 + (1.03 A) (41). Cd2+ binding to the four-metalcluster may be favored over Zn2+ because bindingof the latter would result in a reduction of thevolume of the four-metal cluster leading to in-creased steric hindrance between the residues inthis more complex structural unit. As a compari-son, it has been calculated that the volume of theidealized Cd4 tetrahedron is -30% larger thanthe corresponding Zn4 tetrahedron in the[M4(SC6H5)102- compounds which form adaman-

tane-like structures and are therefore closely re-lated to the proposed four-metal cluster of MT(42). Cu+ is known to adopt trigonal coordinationwith sulfur and therefore its mode of binding maydiffer from that of Cd2+ or Zn2+ (41). The cysteinepositions in the three-metal cluster may be morefavorably disposed to trigonal (or other) coordina-tion than those associated with the four-metalcluster. This is consistent with the different bind-ing stoichiometries and chromatographic proper-ties which have been reported for Cu-MT relativeto Cd,Zn-MTs (43,44). It is, however, possible thatthe Cu+ coordination in MT is tetrahedral, asreported from an EXAFS study of the relatedyeast Cu-MT (45).

In the case ofhomogeneous rabbit "3Cd-MT, themetal ions are more strongly bound to the four-metal cluster. Evidence for this preference is theselective release of Cd2+ from the three-metalcluster during isolation (32), or titration withethylenediaminetetraacetate ion (EDTA), as seenby 113Cd-NMR (Y. Boulanger and I. M. Armitage,unpublished observations). The simultaneous dis-appearance of the 113Cd2+ ions from this clusterduring EDTA titration is also indicative of cooper-ative dissociation. Although metal exchangewithout protein aggregation is readily accom-plished in the tissue homogenate during isolationby addition of the appropriate metal salts (32),exchange in an aqueous solution of the isolatedprotein in our hands, even under reducing condi-tions, is not trivial. In an effort to further charac-terize the mechanism of exchange, experimentsare underway to determine the appropriate com-bination(s) of pH, ionic strength, reducing agentsand counter ions which will mimic the conditionsin the homogenate.

Two Clusters, Two DomainsPreliminary evidence for the independence of

the two mammalian metal clusters is that there isno observable effect on the 113Cd resonances of thefour-metal cluster when the three-metal clusteris depleted of Cd2+ by EDTA (Y. Boulanger and I.M. Armitage, unpublished observations). Evenmore striking is the analysis of the 113Cd-NMRspectrum of a polypeptide fragment, designatedal, which was obtained by proteolytic cleavage ofrat liver MT-1 by subtilisin (46). This fragment'samino acid composition corresponded to the car-boxyl-terminal portion of the metallothioneinchain (residues 30-61). It can be seen in Figure 5that there is an almost direct correspondencebetween the chemical shifts of the four "3Cd reso-nances of the a,-fragment and those attributed to

139

HUNT ET AL.

2 3_

2 3

MAMMALIAN

Addition of Exon

INVERTEBRATE _ VERTEBRATE

PRIMITIVE ANCESTRAL TYPE

670 650 630 610ppm

FIGURE 5. Proton-decoupled l13Cd-NMR spectra (44.4 MHz):(A) Cd-al-MT fragment from subtilisin digestion of ratliver MT-1 (4.2 mM); (B) 1'3Cd reconstituted liver MT-2(7.1 mM). The acquisition parameters were: temperature10°C; spectral width, 10 kHz; pulse angle, 2.5 psec (40°);recycle tiime, 0.7 sec; and number of transients 15,000(MT-2) and 220,000 (a,-MT) (33).

the four-metal cluster of intact human MT-2 (33)This provided the first unequivocal evidence forthe existence of two separate domains in thestructure of metallothionein.

Gene Structure and EvolutionA recent hypothesis states that exons of euka-

ryotic genes code for the structural and functionaldomains of proteins (47). While in many casesthis holds true (47), there are certain instanceswhere this proposal falls short (48,49). For sev-eral proteins, introns in the genetic sequence mapfor proteolysis cleavage points on the surface ofthe protein (49).With this in mind, it is interesting to examine

the primary structure of the mouse MT-1 (50,51)and human MT-2 (40) genes. Both these mamma-lian MT genes have two introns which divide thechain into three exons corresponding to aminoacid residues 1-9, 10-31 and 32-61, respectively.The three exons code for 2, 7 and 11 cysteines,respectively. Although our studies are not con-

2 3

Evolution of Exon 3

Accumulation of Mutations in 2'2 2o

,Duplication of Exon 2

I MT GENE (Exon 2)NEUROSPORA _

ffi. ISIMT GENEPRIMITIVE EUCARYOTE -MT GE

FIGURE 6. Speculation on the phylogeny of the metallothio-nein gene.

sistent with the 20 cysteines being distributedover three separate domains, it is apparent thatexon 3 codes for the domain containing the four-metal cluster. It is also interesting to note thatthe cleavage point for subtilisin is two residuesremoved from the intron 2/exon 3 junction whichis consistent with the views of Fletterick et al.(49) on the dispositions of proteolysis sites.

In view ofthe proposal that duplication, modifi-cation and shuffling of exons are mechanisms ofevolution (49), it seems reasonable to speculate onthe development of the MT genes (see Fig. 6). Forinstance, the MT gene of the mold, Neurosporacrassa, codes for a protein containing seven cys-teines, suggesting that the primordial gene formetallothionein could have been a single exon,similar to exon 2. Perhaps, because of selectivepressure to handle a greater volume of essentialand nonessential metal ions, this exon was dupli-cated and with an accumulation of mutations,there evolved a primitive ancestral type gene thatcontained two exons. Through further evolution,this ancestral gene diverged on the one hand toan invertebrate form which contains 18 cysteinesand on the other hand to a vertebrate from whichcontains 20 cysteines. It is tempting to speculatethat the polypeptide sequence of the invertebrateMT with its two three-metal domains correspondsto the polypeptide coded for by exons 2 and 3 ofthe mammalian gene. Further speculation, how-

.

140

NMR ANALYSIS OF METALLOTHIONEINS

ever, must await the sequencing of an inverte-brate MT gene. At present, our data on the inver-tebrate do not reflect any differences in the metalaffinities ofthe two three-metal clusters. It is onlywith the addition of two more cysteines, whichenables the binding of a fourth metal ion, that themetal clusters exhibit different metal affinities. Itis therefore plausible to speculate that the evolu-tion of two polynuclear clusters with differentphysicochemical properties occurred in order toconfer on higher organisms a more sophisticatedmechanism for metal homeostasis and/or detoxifi-cation.

1H-NMR StudiesSubstantial data on the three-dimensional

structure of metallothioneins have been obtainedfrom the 'H-NMR studies of metallothioneinsfrom different sources (calf, human, rat, rabbit,crab and Neurospora crassa) (34, S. W. Fesik andI. M. Armitage, unpublished observations). Forexample, the downfield region (6.5-9.5 ppm) ofthe 'H-NMR spectra of intact 113Cd-reconstitutedrabbit MT and of the rat all-fragment in 95% H20shows 11-13 and 6 well-resolved 'H resonances,respectively (34). These resonances are due toslowly exchanging amide protons. The number ofresonances is consistent with the results of theChou-Fasman calculation (52) which predict 11and 6 ,8-bends for the intact MT and the fragment,respectively. Each 1-bend is formed by hydrogenbonding of the NH group of residue n to the COgroup of residue (n + 3), thus forming a hairpinturn. Additional three-dimensional structuraldata has been obtained by 'H-'H nuclear Over-hauser enhancement (NOE) difference spectrawhich enables the spatial proximity of severalresidues to be established (34).The complete assignment of the 'H-NMR reso-

nances of metallothionein is not possible solely onthe basis of one-dimensional NMR data even at500 MHz. Several assignments have thereforebeen made by using two-dimensional NMR tech-niques [2D J-resolved (53), spin-echo correlatedspectroscopy (54) and 2D correlated spectroscopy(COSY) (55,56,)] (S. W. Fesik and I. M. Armitage,unpublished observations). Figure 7 shows anexcellent example of the application of one such2D technique to the elucidation of the 'H-NMRspectrum of crab MT-1. The upper part of thefigure shows the one-dimensional spectrum in-which most of the resonances cannot be assigneddue to the large number ofrepetitive residues andoverlapping resonances. The lower portion of Fig-ure 7 shows a contour plot of a 2D COSY experi-

all (ppm)

FIGURE~~~7. Cotu3lto w-imninlcreaeNMRspctu of crbl3dM lOM n2 b

CH2) 1116as~~~~~~~~~fl

Ala(H

Thr(CH135

S~(ppm)FIGURE 7. Contour plot of a two-dimensional correlated 'H-NMR spectrum of crab "'Cd-MT-1 (10mM) in 'H2O ob-tained at 500 MHz. A (90-t1-9O-acquire)n pulse sequencewas used in which the value of t1 was incremented by 300,usec, and 64 scans were acquired for each tl. The data setconsisted of 512 free induction decays (FID), each con-taining 2K data points. To process the data, the FIDs werezero-filled, multiplied by a sine-bell window function, andFourier-transformed.

ment (55,56) in which a number of unequivocalassignments are indicated. The cross peaks in thecontour plot correspond to the frequencies of theprotons that are scalar coupled. For example, thescalar connectivities between the methyl protonsand the a-CH (Ala) and ,B-CH (Val, Thr) protonsare illustrated by a dashed line in the figureconnecting the cross peak to the two coupledproton frequencies located on the diagonal. TheVal resonances were assigned from their charac-teristic chemical shifts and the fact that bothmethyl resonances are scalar coupled to the same1-CH proton. The Thr CH3 protons are spin-coup-led to a 1-CH which is scalar connected to an a-CH proton; whereas, the Ala CH3 is spin-coupleddirectly to an a-proton. On this basis, the Thr andAla proton resonances were distinguished. Theregion of the contour plot within the solid linedisplays an area of off-diagonal peaks correspond-

141

HUNT ET AL.

FIGURE 8. Photograph of our proposed three-dimensional model for mammalian MT. The numbers correspond to the residuenumbers in the primary structure, and the arrow indicates the subtilisin cleavage point.

ing to the 1-CH2 protons of the 18 Cys residuesspin-coupled to their corresponding a-CH protons.By spreading the data into two dimensions, theLys resonances in this region appear above thediagonal and the different Cys cross peaks areeasily resolvable, allowing all of the a-CH Cysprotons to be identified. The analysis of the datafrom the application of these different techniquesto MTs from different sources and of differentstructures is in progress and should allow theassignment of the vast majority of the 'H reso-nances. These assignments will then be used forfurther conformational analysis.

Model for MammalianMetallothioneinBased on the structural and physicochemical

data available, a molecular model of mammalianmetallothionein has been constructed and isshown in Figure 8 (34). The protein is constructedwith two domains, one containing the three- and

the other the four-metal cluster, corresponding tothe amino-terminal and carboxyl-terminal por-tions of the protein, respectively. The metal clus-ters consist of tetrahedral metal centers boundand bridged by cysteine sulfurs, as evidenced by113Cd-NMR. The 11 1-bends predicted by theChou-Fasman calculations are also included inthis model (52).Whereas it is not possible at this time to deter-

mine a unique distribution of the cysteine resi-dues among the individual metal binding sites,the results from the NMR studies of the proteoly-tically cleaved a,,-fragment allowed us to dividethe chain into two domains containing: the three-metal cluster (residues 1-29) and the four-metalcluster (residues 30-61). The domain ofthe three-metal cluster contains four Cys-X-Cys groups(specifically residues 5-7; 13-15; 19-21 and 24-26) and one isolated cysteine (residue 29) (see Fig.1). In deploying these cysteine residues in thecluster, we have assumed that each Cys-X-Cysgroup binds to the same metal, because of the

142

NMR ANALYSIS OF METALLOTHIONEINS 143

short distance between the cysteines. The five I-bends predicted by the Chou-Fasman program forthis region can also be easily built into the model.The domain of the four-metal cluster containsthree cysteine groupings: a Cys-Cys-X-Cys-Cysgroup in positions 33-37, a Cys-X-Cys-Cys groupin positions 57-60 and four cysteines in positions41-50. In the proposed structure of the four-metalcluster we have assumed that adjacent cysteinesbind the same metal. Therefore, the cysteine resi-dues in positions 33-37 and 41-50 are ligands ofCdl and Cd5, respectively, and the cysteine resi-dues in positions 57-60 are ligands of Cd6 andCd7. The cysteines at residues 33, 37, 41 and 50are bridging ligands and are therefore furtherlinked to a second metal ion. The six 3-bends,which were predicted for this domain by theChou-Fasman calculation, are readily incorpora-ted into our model. It is interesting to note that inthis model the polypeptide chain is wrapped moretightly around the four-metal cluster. Furthertertiary structure is provided by electrostatic in-teractions and hydrogen bonding involving theside chains ofvarious residues (Ser, Thr) with, forexample, the amide carboxyl of the peptide chain(0H ... 0 = C) (57). Some of these interactionshave been built into our model (34).

Conclusion113Cd-NMR studies have provided detailed in-

formation on the structural and spacial relation-ship of the individual metal sites in MTs. Espe-cially significant are the findings that crab MTshares with mammalian MTs the property ofbinding its metals in two separate metal-thiolateclusters, but differs in that both are type-B, three-metal clusters, rather than a type B, three- and atype A, four-metal cluster. Despite the evolution-ary distance between the invertebrate (crab) andthe vertebrates, a high degree of sequence homol-ogy exists between MTs from these sources (35).In this light, it is possible that the type A, four-metal cluster is a relatively recent evolutionarydevelopment occurring sometime after the diver-gence of invertebrates and mammals from theirprimordial ancestor. The selective pressure forthis is unclear because of the uncertainty whichstill exists regarding the physiological function ofthe protein. However, the results of the 113Cd-NMR experiments have demonstrated the oppo-site selectivity of the two mammalian metal clus-ters for various metal ions. This would enable thefour-metal cluster to sequester the toxic metalion, Cd2", while leaving the three-metal clusteravailable for the regulation and/or metabolism of

Zn2+ and Cu+. In other words, this developmentmight have occurred in order to impart to verte-brates the ability to handle both essential andnonessential metals with selectivity.1H-NMR studies have extended our knowledge

of the tertiary structure of the protein and willcertainly contribute more substantially in thefuture. Our present model for the three-dimen-sional solution structure of mammalian metal-lothionein is consistent with all the physicochem-ical data presently available. It should be noted,however, that this is not a unique solution. Thatis, a final refinement of our model will be neces-sary when the X-ray crystal structure, currentlyin progress in the laboratory of C. D. Stout, be-comes available. Nevertheless, we are confidentthat the crystal structure will confirm many ofthe features of our model and we are continuingin our efforts to elucidate the biological func-tion(s) of this protein.The financial assistance of National Institutes of Health

Grant AM 18778, National Science Foundation grant CHE-7916210, and a North Atlantic Treaty Organization Postdoc-toral Fellowship (Y.B.) is gratefully acknowledged.

REFERENCES

1. Nordberg, M., and Kojima, Y. Report from the First Inter-national Meeting on metallothionein and other low molec-ular weight metal-binding proteins. In: Metallothionein(J. H. R. Kagi and M. Nordberg, Eds.), Birkhauser, Basel,1979, pp. 41-116.

2. Kojima, Y., and Kagi, J. H. R. Metallothionein. TrendsBiochem. Sci. 3: 90-93 (1978).

3. Prinz, R., and Weser, U. A naturally occurring Cu-thio-nein in Saccharomyces cerevisiae. Hoppe-Seyler's Z. Phy-siol. Chem. 356: 767-776 (1975).

4. Lerch, K. Copper metallothionein, a copper-binding pro-tein from Neurospora crassa. Nature 284: 368-370 (1980).

5. Kagi, J. H. R., and Vallee, B. L. Metallothionein: a cad-mium and zinc-containing protein from equine renal cor-tex. J. Biol. Chem. 236: 2435-2442 (1961).

6. Vallee, B. L. Metallothionein: Historical review and per-spectives. In: Metallothionein, (J. H. R. Kagi and M.Nordberg, Eds.), Birkhauser, Basel, 1979, pp. 19-40.

7. Chang, C. C., Lauwerys, R., Bernard, A., Roels, H., Bu-chet, J. P., and Garvey, J. S. Metallothionein in cadmium-exposed workers. Environ. Res. 23: 422-428 (1980).

8. Squibb, K. S., Cousins, R. J., and Feldman, S. L. Control ofzinc-thionein synthesis in rat liver. Biochem. J. 164: 223-228 (1977).

9. Winge, D. R., Premakumar, R., and Rajagopalan, K. V.Metal-induced formation of metallothionein in rat liver.Arch. Biochem. Biophys. 170: 242-252 (1975).

10. Cherian, M. G., and Goyer, R. A. Minireview: metallothio-neins and their role in the metabolism and toxicity ofmetals. Life Sci. 23: 1-10 (1978).

11. Johnson, D. R., and Foulkes, E. C. On the proposed role ofmetallothionein in the transport of cadmium. Environ.Res. 21: 360-365 (1980).

12. Margoshes, M., and Vallee, B. L. A cadmium protein fromequine kidney cortex. J. Am. Chem. Soc. 79: 4813-4814(1957).

144 HUNT ET AL.

13. Kojima, Y., Berger, C., Vallee, B. L., and Kagi, J. H. R.Amino acid sequence of equine renal metallothionein-lB.Proc. Natl. Acad. Sci. (U.S.) 73: 3413-3417 (1976).

14. Kojima, Y., Berger, C., and Kagi, J. H. R. The amino acidsequence of equine metallothioneins. In: Metallothionein(J. H. R. Kagi and M. Nordberg, Eds.), Birkhauser, Basel,1979, pp. 153-161.

15. Huang, I. -Y., Yoshida, A., Tsunoo, H., and Nakajima, H.Mouse liver metallothioneins: complete amino acid se-quence of metallothionein-I. J. Biol. Chem. 252: 8217-8221 (1977).

16. Huang, I. -Y., Kimura, M., Hata, A., Tsunoo, H., andYoshida, A. Complete amino acid sequence of mouse livermetallothionein-II. J. Biochem. 89: 1839-1845 (1981).

17. Kimura, M., Otaki, N., and Imano, M. Rabbit liver metal-lothionein. Tentative amino acid sequence of metallothio-nein. In: Metallothionein (J. H. R. Kagi and M. Nordberg,Eds.), Birkhauser, Basel, 1979, pp. 163-168.

18. Kissling, M. M., and Kagi, J. H. R. Amino acid sequence ofhuman hepatic metallothioneins. In: Metallothionein(J. H. R. Kagi and M. Nordberg, Eds.), Birkhauser, Basel,1979, pp. 145-151.

19. Kissling, M. M., and Kagi, J. H. R. Primary structure ofhuman hepatic metallothionein. FEBS Letters 82: 247-250 (1977).

20. Rupp, H., and Weser, U. Circular dichroism of metal-lothioneins. A structural approach. Biochim. Biophys.Acta 533: 209-226 (1978).

21. Buhler, R. H. O., and Kagi, J. H. R. Spectroscopic proper-ties of zinc-metallothionein. In: Metallothionein (J. H. R.Kagi and M. Nordberg, Eds.), Birkhauser, Basel 1979, pp.211-220.

22. Law, A. Y. C., and Stillman, M. J. The effect of pH onCd2+ binding to rat liver metallothionein. Biochem.Biophys. Res. Commun. 94: 138-143 (1980).

23. Vasak, M., Kagi, J. H. R., and Hill, H. A. 0. Zinc (II),cadmium (II), and mercury (II) thiolate transitions inmetallothionein. Biochemistry 20: 2852-2856 (1981).

24. Rupp, H., and Weser, U. Conversion of metallothioneininto Cu-thionein, the possible low molecular weight formof neonatal hepatic mitochrondrocuprein. FEBS Letters44: 293-297 (1974).

25. Armitage, I. M., and Otvos, J. D. Principles and applica-tions of l13Cd NMR to biological systems. In: BiologicalMagnetic Resonance, Vol. 4 (L. R. Berliner and J. Reuben,Eds.), Plenum Press, 1982, pp. 79-144.

26. Otvos, J. D., and Armitage, I. M. ll3Cd NMR of metal-lothionein: direct evidence for the existence ofpolynuclearmetal binding sites. J. Am. Chem. Soc. 101: 7734-7736(1979).

27. Otvos, J. D., and Armitage, I. M. Structure of the metalclusters in rabbit liver metallothionein. Proc. Natl. Acad.Sci. (U.S.) 77: 7094-7098 (1980).

28. Boulanger, Y., and Armitage, I. M. ll3Cd NMR study ofthe metal cluster structure of human liver metallothio-nein. J. Inorg. Biochem. 17: 147-153 (1982).

29. Briggs, R. W., and Armitage, I. M. Evidence for site-selective metal binding in calf liver metallothionein. J.Biol. Chem. 257: 1259-1262 (1982).

30. Armitage, I. M., Otvos, J. D., Briggs, R. W., andBoulanger, Y. Structure elucidation of the metal-bindingsites in metallothionein by ll3Cd NMR. Fed. Proc. 41:2974-2980 (1982).

31. Otvos, J. D., Olafson, R. W., and Armitage, I. M. Structureof an invertebrate metallothionein from Scylla serrata. J.Biol. Chem. 257: 2427-2431 (1982).

32. Otvos, J. D., and Armitage, I. M. Elucidation of metal-lothionein structure by ll3Cd NMR. In: Biochemical

Structure Determination by NMR (B. D. Sykes, J. Glick-son, and A. A. Bothner-By, Eds.), Marcel Dekker, NewYork, 1981, pp. 65-96.

33. Boulanger, Y., Armitage, I. M., Miklossy, K.-A., andWinge, D. R. 1'3Cd NMR study of a metallothionein frag-ment. J. Biol. Chem. 257: 13717-13719 (1982).

34. Boulanger, Y., Goodman, C. M., Forte, C. P., Fesik, S. W.,and Armitage, I. M. Model for mammalian metallothio-nein structure. Proc. Natl. Acad. Sci. (U.S.) 80: 1501-1505(1983).

35. Lerch, K., Ammer, D., and Olafson, R. W. Crab metal-lothionein: primary structure of metallothioneins 1 and 2.J. Biol. Chem. 257: 2420-2426 (1982).

36. Olafson, R. W., Kearns, A., and Sim, R. G. Heavy metalinduction of metallothionein synthesis in the hepatopan-creas of the crab Scylla serrata. Comp. Biochem. Physiol.62B: 417-424 (1979).

37. Olafson, R. W., Sim, R. G., and Boto, K. G. Isolation andchemical characterization ofthe heavy metal-binding pro-tein metallothionein from marine invertebrates. Comp.Biochem. Physiol. 62B: 407-416 (1979).

38. Murphy, P. DuBois, Stevens, W. C., Cheung, T. T. P.,Lacelle, S., Gerstein, B. C., and Kurtz, D. M., Jr. High-resolution ll3Cd NMR of solids. Correlation of spectrawith the,molecular structure of a decanuclear cadmium(II) complex. J. Am. Chem. Soc. 103: 4400-4405 (1981).

39. Haberkorn, R. A., Que, L., Jr., Gillum, W. O., Holm, R. H.,Liu, C. S., and Lord, R. C. Cadmium-113 Fourier trans-form nuclear magnetic resonance and Raman spectro-scopic studies ofcadmium (II)-sulfur complexes, including[Cd10(SCH2CH2OH)16]4+. Inorg. Chem. 15: 2408 (1976).

40. Karin, M., and Richards, R. I. Human metallothioneingenes-primary structure of the metallothionein-HI geneand a related processed gene. Nature (London) 299: 797-802 (1982).

41. TIssell, J. A., and Vaughan, D. J. Relationships betweenvalence orbital bonding energies and crystal structures incompounds of copper, silver, gold, zinc, cadmium andmercury. Inorg. Chem. 20: 3333-3340 (1981).

42. Hagen, K. S., Stephan, D. W., and Holm, R. H. Metal ionexchange reactions in cage molecules: the systems[M4_ M'.(SC6H5)10]2 - [M, M' = Fe(II), Co(II), Zn(II),Cd(II)] with adamantane-like stereochemistry and thestructure of [Fe4(SC6H5)10]2-. Inorg. Chem. 21: 3928-3936 (1982).

43. Suzuki, K. T., and Maitani, T. Metal-dependent propertiesof metallothionein: replacement in vitro of zinc in zinc-thionein with copper. Biochem. J. 199: 289-295 (1981).

44. Winge, D. R., Geller, B. L., and Garvey, J. Isolation ofcopper thionein from rat liver. Arch. Biochem. Biophys.208: 160-166(1981).

45. Bordas, J., Koch, M. H. J., Hartmann, H.-J., and Weser,U. Tetrahedral copper-sulphur coordination in yeast Cu-thionein. FEBS. Letters 140: 19-21 (1982).

46. Winge, D. R. and Miklossy, K.-A. Domain nature of metal-lothionein. J. Biol. Chem. 257; 3471-3476 (1982).

47. Gilbert, W. Why genes in pieces? Nature 271: 501 (1978).48. Craik, C. S., Rutter, W. J., and Fletterick, R. Splice

junctions: association with variation in protein structure.Science 220: 1125-1129 (1983).

49. Craik, C. S., Sprang, S., Fletterick, R., and Rutter, W. J.Intron-exon splice junctions map at protein surfaces. Na-ture (London) 299: 180-182 (1982).

50. Durnam, D. M., Perrin, F., Gannon, F., and Palmiter,R. D. Isolation and characterization of the mouse metal-lothionein-I gene. Proc. Natl. Acad. Sci. (U.S.) 77: 6511-6515 (1980).

51. Glanville, N., Durnam, D. M., and Palmiter, R. D. Struc-

NMR ANALYSIS OF METALLOTHIONEINS 145

ture of mouse metallothionein-I gene and its mRNA.Nature 292: 267-269 (1981).

52. Fasman, G. D. Prediction of protein conformation fromthe primary structure. Ann. N.Y. Acad., 348: 147-159(1980).

53. Aue, W. P., Karhan, J., and Ernst, R. R. Homonuclearbroad band decoupling and two-dimensional J-resolvedNMR spectroscopy. J. Chem. Phys. 64: 4226-4227 (1976).

54. Nagayama, K., Wuthrich, K., and Ernst, R. R. Two-di-mensional spin echo correlated spectroscopy (SECSY) for1H NMR studies of biological macromolecules. Biochem.

Phys. Res. Commun. 90: 305-311 (1979).55. Aue, W. P., Bartholdi, E., and Ernst, R. R. Two-dimen-

sional spectroscopy. Application to NMR. J. Chem. Phys.64: 2229-2246 (1976).

56. Wagner, G., Kumar, A., and Wuthrich, K. Systematicapplication of two-dimensional 1H NMR techniques forstudies of proteins. Eur. J. Biochem. 114: 375-384 (1981).

57. Vasak, M., Galdes, A., Hill, H. A. O., Kagi, J. H. R.,Bremner, I., and Young, B. W. Investigation of the struc-ture of metallothioneins by proton nuclear magnetic reso-nance spectroscopy. Biochemistry 19: 416-425 (1980).