Organization and Assembly of Metal-Thiolate Clusters in Epithelium-specific Metallothionein-4*

Mammalian metallothionein-4 (MT-4) was found to be specifically expressed in stratified squamous epithelia where it plays an essential but poorly defined role in regulating zinc or copper metabolism. Here we report on the organization, stability, and the pathway of metal-thiolate cluster assembly in MT-4 reconstituted with Cd2+ and Co2+ ions. Both the 113Cd NMR studies of 113Cd7MT-4 and the spectroscopic characterization of Co7MT-4 showed that, similar to the classical MT-1 and MT-2 proteins, metal ions are organized in two independent Cd4Cys11 and Cd3Cys9 clusters with each metal ion tetrahedrally coordinated by terminal and bridging cysteine ligands. Moreover, we have demonstrated that the cluster formation in Cd7MT-4 is cooperative and sequential, with the Cd4Cys11 cluster being formed first, and that a distinct single-metal nucleation intermediate Cd1MT-4 is required in the cluster formation process. Conversely, the absorption and circular dichroism features of metal-thiolate clusters in Cd7MT-4 indicate that marked differences in the cluster geometry exist when compared with those in Cd7MT-1/2. The biological implication of our studies as to the role of MT-4 in zinc metabolism of stratified epithelia is discussed.

maintenance of intracellular redox balance (7)(8)(9)(10). Differential expression of mammalian MT isoforms is tightly regulated during development and in pathological situations (9,11). The extensively studied mammalian MT-1 and MT-2 show ubiquitous expression regulated at the transcriptional level (12). Their biosynthesis is inducible by a variety of compounds and stress conditions, such as metals, glucocorticoids, cytokines, and reactive oxygen species (9). MT-3 and MT-4 are relatively unresponsive to these inducers. MT-3 is primarily confined to the central nervous system, where it represents a major component of the intracellular Zn 2ϩ pool in zinc-enriched neurons (13). Lower expression levels of MT-3 have also been reported in pancreas, kidney, reproductive tissues, and maternal deciduum. This protein exhibits growth inhibitory activity in neuronal cultures and was found down-regulated in Alzheimer disease (reviewed in Ref. 14).
The expression of the last identified mammalian metallothionein isoform, MT-4, has been found restricted to cornified, stratified, squamous epithelium, a tissue providing a protective surface on skin, footpath, tail, tongue, the upper part of the alimentary tract, and the vagina of rodents (15). Besides these tissues, the developmentally regulated MT-4 expression in maternal deciduum together with the expression of entire MT gene locus have been reported in mouse (16). Gene expression profiling by microarray hybridization of wild-type and nude mice back skin identified that the transcription factor Whn regulates the MT-4 expression together with other proteins involved in the metabolism of keratin (17). However, the question of whether MT-4 is involved in copper or zinc metabolism in epithelia is still debated. Thus, in the recent studies on the metal binding abilities of MT-4 using heterologously expressed MT-4 in Escherichia coli with zinc, cadmium, and copper in combination with the in silico protein sequence analyses, the copper binding nature of MT-4 has been suggested (18). On the other hand, based on the highly regulated and specific expression pattern of MT-4 in stratified squamous epithelium, the documented switch in expression of MT-1 in the basal layer to MT-4 expression in the next layer during differentiation of this tissue and the zinc content of the isolated protein, a special role of MT-4 in the regulation of zinc-dependent keratinocyte differentiation or in the regulation of proteolytic processing of keratins has also been suggested (15). Moreover, MT-4 expression protects cell culture against cadmium toxicity, and its expression is induced in the upper stomach of mice exposed to high zinc levels in drinking water. These observations suggest that MT-4, similar to MT-1 and MT-2, may be involved in the regulation of zinc levels and cadmium detoxification (15). However, the metal binding specificity of MT-4 is difficult to assess, as to date, only limited information regarding the metal binding properties of MT-4 containing divalent metal ions is available.
The presented studies were conducted with the aim of gaining insight into the metal binding properties and the structural features of MT-4, which are likely responsible for its function in zinc metabolism. In our studies, we have replaced the spectroscopically silent Zn 2ϩ ions in the mouse Zn 7 MT-4 structure by the Cd 2ϩ and Co 2ϩ ions and subjected these metal derivatives to detailed spectroscopic investigations. In the past, both Cd 2ϩ and Co 2ϩ ions have proven to be useful probes for zinc binding sites in MTs, affording a wealth of structural information on the native protein (6,19 -21). From the two-dimensional [ 113 Cd-113 Cd] COSY of 113 Cd 7 MT-4, evidence for two distinct metal-thiolate clusters (i.e. a 3-metal and a 4-metal cluster) in this protein was obtained. Further studies on Cd 7 MT-4 and Co 7 MT-4 by electronic absorption, CD, and MCD spectroscopy and by mass spectrometry of partially and fully metal-loaded MT-4 afforded to us information regarding the coordination geometry of the metal sites, the pathway of cluster assembly, and their stability. The results are discussed in comparison with information available on the well characterized mammalian MT-1/2 isoforms.

EXPERIMENTAL PROCEDURES
Materials-Media for protein expression were purchased from BD Biosciences. All other standard reagents were purchased of the highest purity available from common commercial sources.
Construction of the MT-4 Expression Vector-The pBluescript plasmid containing the coding sequence of mouse MT-4 was received from Prof. R. Palmiter, University of Washington. The expression plasmid encoding MT-4 was constructed from the cloning vector pET24d (Novagen) carrying a T7 promoter. The T7 tag sequence was eliminated upon NcoI/EcoRI (Fermentas) digestion, and the MT-4 coding sequence was subcloned into the same restriction sites. The MT-4 sequence and the expression vector were ligated using T4 DNA ligase (Fermentas, Lithuania) and transformed into E. coli strain DH5␣ and BL21pLysS. The DNA sequence of the insert was verified by DNA sequencing (BigDye, Amersham Biosciences).
Protein Expression and Purification-Mouse MT-4 was expressed in E. coli strain BL21pLysS and purified as described previously for MT-3 (22) with the following modifications. Kanamycin (30 mg/liter) was used as a selecting antibiotic during cultivation, a 0.5 mM concentration of isopropyl-␤-D-thiogalactopyranoside was used for induction, and a 0.6 mM concentration of Zn(CH 3 COO) 2 was added upon induction.
Preparation and Characterization of Cd 7 MT-4-The apo-form of MT-4 was generated by the method of Vašák (23), and fully Cd 2ϩ -or 113 Cd 2ϩ -loaded MT-4 was prepared by reconstitution (23). Metal-toprotein ratios were determined using a small aliquot of the sample. The metal concentration was determined by flame atomic absorption spectrometry (SpectrAA-110, Varian Inc.) and that of the protein via sulfhydryl group quantification (20 Cys/protein). The concentration of sulfhydryl groups was determined photometrically with 2,2Ј-dithiopyridine in 0.2 M sodium acetate/1 mM EDTA (pH 4) using ⑀ 343 ϭ 7600 M Ϫ1 cm Ϫ1 (24). The metal-to-protein ratio of 7.0 Ϯ 0.5 was obtained.
The absence of disulfides and the full metal loading were confirmed by ESI-MS measurements of the reconstituted metalloprotein.
Preparation and Characterization of Co(II) 7 MT-4-The solutions used in Co 2ϩ reconstitution were rendered oxygen-free by five freezepump-thaw cycles on a vacuum line and all sample preparation steps were performed in a nitrogen-purged glove box. The air-sensitive Co 7 MT-4 was obtained by the mixing of seven equivalents of CoCl 2 with apoMT-4 (0.5 mM) in 0.1 M HCl. The solution mixture was then adjusted to pH 8.0 with 1 M Tris base, and the final salt concentration of 17 mM NaCl was adjusted with 2 M NaCl. The metal-to-protein ratio was checked using a small aliquot of the sample as described above. In the glove box, samples were transferred into 1 or 0.1 cm cap-sealed cuvettes for spectroscopic measurements.
Preparation of CdMT-4 Samples for Spectroscopic Titration-All solutions used in the Cd 2ϩ titration experiments were rendered oxygenfree by three freeze-pump-thaw cycles on a vacuum line and all samples prepared in a nitrogen-purged glove box. Individual apoprotein samples (9 M) in 0.1 M HCl were titrated with increasing Cd 2ϩ equivalents (1-8 equivalents). Subsequently, the pH value was adjusted with metal-free 1 M Tris base to 7.2. The final protein sample was contained in 10 mM Tris/HCl, 20 mM NaCl, pH 7.2. In the glove box, samples were transferred into a 1-cm cap-sealed cuvette for spectroscopic measurements. Determination of Apparent Cd 2ϩ Binding Constant by Photometric pH Titration-The apparent binding constants of Cd 7 MT-4 at pH 7.0 was determined as previously described by Kägi and co-workers (25,26) using the adapted expression of Wang et al. (27). Briefly, the release of Cd 2ϩ from Cd 7 MT-4 was achieved by lowering of the pH value by the addition of increasing amounts of oxygen-free 1 M HCl. Prior to measurements, a stock solution of Cd 7 MT-4 was diluted to a final concentration of 5 M in 10 mM Tris/HCl, 20 mM NaCl, pH 8.0, and rendered oxygen-free by three freeze-pump-thaw cycles on a vacuum line. The pH titration was performed on independent samples prepared in a nitrogen-purged glove box. Metal release was followed by recording absorption spectra between 320 and 210 nm in a sealed 1-cm cuvette. The pH was determined immediately after spectra recording using a microelectrode. The degree (D) of Cd 2ϩ release is illustrated by plotting the pH values against the percentage of metal dissociation described by D ϭ (A 250i Ϫ A 250,pHϭ1.5 )/(A 250,pHϭ8 Ϫ A 250,pHϭ1.5 ), where A 250i is the absorbance at 250 nm of the Cd 2ϩ -thiolate complex at different pH values and A 250,pHϭ8 and A 250,pHϭ1.5 represent the corresponding absorbance of Cd 7 MT-4 and the apoprotein, respectively. In the calculation of the apparent Cd 2ϩ binding constant, the pK a values of the cysteine side chains were assumed to be equal to those reported for rabbit MT-1/2 isoforms (pK a ϭ 8.9) (25).

Preparation of CdMT-4 Samples for ESI-MS Titration-
Spectroscopic Measurements-UV-visible absorption spectra were recorded on a Cary 3 spectrophotometer (Varian). CD and MCD measurements were performed using a Jasco (Model J-810) spectropolarimeter equipped with a 1.5-tesla electromagnet for room temperature MCD measurements. A 1-cm quartz cuvette was used for Cd x MT-4 measurements and a 1-or 0.1-cm quartz cuvette for Co 7 MT-4 measurements. The CD spectra are expressed as molar ellipticity [⍜] in units of degrees dmol Ϫ1 cm 2 , and the MCD spectra are expressed as [⍜ M ] in units of degrees dmol Ϫ1 cm 2 tesla Ϫ1 .
110.9-MHz one-dimensional 113 Cd NMR spectra of 113 Cd 7 MT-4 (1.7 mM) in 20 mM Tris/HCl and 40 mM NaCl, were recorded on a Bruker DRX-500 spectrometer at 293 K using an inverse-gated broad band proton decoupling, a 34,483-Hz spectral width, a 2.2-s acquisition time, and a 4.2-s pulse repetition rate (averaging 6000 free induction decays/ spectrum). The two-dimensional [ 113 Cd-113 Cd] COSY spectrum of 113 Cd 7 MT-4 (7 mM) was acquired using the standard COSY sequence in phase-sensitive mode with proton decoupling during acquisition. The evolution period t 1 was varied in 207 increments from 0 to 10.5 ms covering a spectral width in the indirect ( 113 Cd) dimension of 9.8 kHz. A total of 128 transients were accumulated for each value of t 1 . Chemical shifts are reported in parts/million with respect to 113 Cd resonance of the external standard 0.1 M Cd(ClO 4 ) 2 in 2 H 2 O. The NMR samples contained 20% 2 H 2 O to provide the field frequency lock and were measured in 5-mm NMR tubes.

RESULTS AND DISCUSSION
Expression and Purification of MT-4-As described under "Experimental Procedures," MT-4 was expressed and purified as the zinc protein. The presence of a single mass peak of 6275.5 Da (calculated mass 6276.5 Da) in the ESI-MS spectrum of metal-free MT-4 established the correctness of the recombinant protein. The cadmium-containing MT-4 was generated by the method of metal reconstitution as described above. The 113 Cd-reconstituted 113 Cd 7 MT-4 revealed a molecular mass of 7052.1 Da, which is in agreement with the calculated mass for MT-4 containing seven 113 Cd 2ϩ ions. The analytical gel filtration experiments revealed monomeric protein with an apparent molecular mass of ϳ20 kDa (data not shown). The increased apparent molecular masses have also been reported for MT-1/2 and attributed to the non-globular shape of these molecules (22).
One-and Two-dimensional 113 Cd NMR Characterization of 113 Cd 7  Cd NMR proved to be a powerful tool in the investigation of the nature of metal binding sites and their organization in Cd-MTs (19,20). The 113 Cd NMR experiments have been performed at 110.9 MHz and at 293 K. The one-dimensional 113 Cd NMR spectrum of 113 Cd 7 MT-4 shows seven major 113 Cd signals at 667, 666, 661, 659, 640, 630, and 599 parts/million, corresponding to seven distinct metal binding sites. The 113 Cd resonances have been numbered from 1 to 7 in the order of decreasing chemical shifts ( Fig. 2A). The chemical shift positions of the seven major resonances of 113 Cd 7 MT-4, which are very similar to those reported for other mammalian 113 Cd 7 MTs (20), and the presence of 113 Cd-113 Cd spin splitting indicate that the seven Cd 2ϩ sites are organized in a cluster structure(s) in which both bridging and terminal thiolate ligands participate in metal binding.
Besides these major resonances, an additional four low intensity 113 Cd resonances marked with an asterisk are also discerned ( Fig. 2A). We found that, in different NMR samples, their intensities varied between 15 and 30% compared with those of the major 113 Cd signals and that the overall 113 Cd NMR profile was unaffected by a temperature increase from 293 to 323 K. The latter indicates that these resonances do not originate from a different cluster conformation. In addition, analytical gel filtration experiments performed after each NMR run also showed, besides the major chromatographic peak of a monomeric species, a peak of dimers (15-30%) that was formed at millimolar protein concentrations required in the NMR studies (data not shown). Because no dissociation of these dimers occurred after their rechromatography, this suggests that the dimers are linked through disulfide bond(s).
Two-dimensional [ 113 Cd-113 Cd] COSY has been used to elucidate metal cluster organization in the previously studied MT isoforms (19,28). Therefore, to investigate the topology of the Cd-thiolate clusters in Cd 7 MT-4, the two-dimensional [ 113 Cd-113 Cd] COSY spectrum has been recorded at 293 K. In the spectrum analysis, only 113 Cd connectivities among the major 113 Cd signals of the monomer were considered. The two-dimensional plot of 113 Cd 7 MT-4 COSY spectrum (Fig.  2B) on a "noise-free" level clearly shows the presence of five strong cross-peaks and an additional three weaker and partially resolved crosspeaks. Based on 113 Cd connectivities, the topology of both metal clusters was determined (Fig. 2C). The selective association of the strong five cross-peaks with resonances 2, 5, 6, and 7 and the three weaker with resonances 1, 3, and 4 partitioned the seven 113 Cd signals into two independent linkage groups. Thus, the seven metal binding sites in the protein are organized in two independent clusters, i.e. a 4-metal cluster (resonances 2, 5, 6, 7) and a 3-metal cluster (resonances 1, 3, 4). It may be noted that the observed 113 Cd connectivities among the four low intensity 113 Cd resonances of the dimers marked with the asterisk are in line with a slightly altered 4-metal cluster in this species. Considering that all of the cysteine residues in monomeric Cd 7 MT-4 are involved in metal binding (see also next paragraph), the NMR data suggest the presence of the Cd 4 Cys 11 and Cd 3 Cys 9 clusters in which five and three cysteine thiolates act as bridging ligands, respectively. The identical cluster topologies have also been found in the mammalian Cd 7 MT-1/2 (29,30). Moreover, in view of the conserved array of 20 cysteine residues in all mammalian MTs, the Cd 4 Cys 11 cluster of MT-4 would be located in the C-terminal ␣-domain and the Cd 3 Cys 9 cluster in the N-terminal ␤-domain.
Electronic Absorption, CD, and MCD Characterization of Co 7 MT-4 and Cd 7 MT-4-Because no direct information regarding the geometry of the metal binding sites and the clusters can be obtained from [ 113 Cd-113 Cd] COSY experiments, further spectroscopic investigations of the Co 7 MT-4 and Cd 7 MT-4 derivatives were conducted. To unambiguously establish the coordination geometry and ligation of metal binding sites in MT-4, the Co 7 MT-4 derivative was characterized by electronic absorption and room temperature MCD spectroscopy (Fig. 3). The visible region of the corresponding absorption spectrum shows a d-d profile characterized by two broad bands with maxima at 744 and 693 nm and a poorly resolved shoulder at ϳ610 -620 nm. The former two bands have their respective molar absorbances of 2700 and 2950 M Ϫ1 cm Ϫ1 , i.e. ⑀ ϭ 386 M Ϫ1 cm Ϫ1 and 421 M Ϫ1 cm Ϫ1 /Co 2ϩ -bound to protein, respectively. The position and the intensity of the d-d features are typical of tetrahedral tetrathiolate coordination reported for inorganic model complexes and other protein cobalt derivatives with cysteine thiolate ligands. The resolved d-d pattern can be assigned to the spin-allowed 3 [ 4 A 2 3 4 T 1 (P)] transition, as already described for rabbit liver MT-1 (31). Evidence for tetrahedral tetrathiolate Co 2ϩ coordination is provided by the MCD spectrum of Co 7 MT-4. The low energy region of the MCD spectrum shows a strong negative band at 743 nm with a pro-nounced shoulder at around 690 nm. In addition, a weak positive band at 628 nm and a weak negative band at 573 nm were observed. A similar overall pattern has been theoretically predicted and observed in a number of inorganic tetrahedral and pseudotetrahedral model compounds and in other Co 2ϩ protein derivatives possessing these coordination geometries (Ref. 31 and references therein). Splitting of the 3 transition in the absorption spectrum into three components with an energy separation of ϳ2000 and 1000 cm Ϫ1 is larger than that expected from spin-orbit coupling alone. This and the molar extinction coefficient of the most intense band at 693 nm (421 M Ϫ1 cm Ϫ1 /Co 2ϩ ) is in agreement with a pseudotetrahedral symmetry of the sites (⑀ Ͼ 250 M Ϫ1 cm Ϫ1 ) (32). Evidence for cobalt-sulfur coordination was obtained from the high energy part of the absorption spectrum, where a strong absorption band at 325 nm (⑀ ϭ 20200 M Ϫ1 cm Ϫ 1) and a shoulder ϳ390 nm was observed. The position of the CysS-Co 2ϩ LMCT transition at 325 nm and its intensity (⑀ ϭ 1010 M Ϫ1 cm Ϫ1 /CysS-Co 2ϩ bond) are consistent with 20 Cys residues being involved in metal binding (33).
To learn more about the cluster geometry in Cd 7 MT-4, its electronic absorption, CD, and MCD spectra were examined (Fig. 4). The electronic absorption spectrum of Cd 7 MT-4 (Fig. 4A) shows a characteristic shoulder at ϳ250 nm, a feature commonly found in mammalian Cd 7 MTs. By analogy, we assign the underlying metal-induced bands to CysS-Cd 2ϩ LMCT transitions (34). It should be noted that metal-free MT-4 (apoMT-4) does not show appreciable absorption above 230 nm due to the absence of aromatic amino acids and histidine. The molar extinction coefficient at 250 nm of ϳ100,000 M Ϫ1 cm Ϫ1 reveals a value of 5000 M Ϫ1 cm Ϫ1 /CysS-Cd 2ϩ bond. The calculated value is closely similar to that reported for the CysS-Cd 2ϩ bond in a number of Cd 2ϩsubstituted metalloproteins (35), confirming that all 20 cysteines in MT-4 are involved in metal binding. The corresponding MCD spectrum of Cd 7 MT-4 (Fig. 4C) shows a biphasic profile with bands at (Ϫ)258 and (ϩ)236 nm with the inflection point at 247 nm reported also for other mammalian Cd 7 MTs (36,37). These features have been assigned to a positive A term originating from Cd 2ϩ binding sites possessing a Td-type symmetry (21). However, the CD profile of Cd 7 MT-4 ( Fig. 4B) with extrema at (ϩ)266, (Ϫ)245, and (ϩ)228 nm substantially differ from those reported for different members of Cd 7 MT-1/2 subfamilies (21,36 -38). Thus, in the structurally well characterized rabbit Cd 7 MT-2 and mouse Cd 7 MT-1, characteristic CD bands at (ϩ)261, (Ϫ)241, and (ϩ)228 nm have been reported. Moreover, it has been demonstrated that the oppositely signed low energy CD bands represent an envelope of biphasic CD bands originating from the excitonically coupled transition dipole moments of the bridging thiolate ligands within the cluster structure. This was further confirmed by the identity of the positions of the first low energy CysS-Cd 2ϩ LMCT transition at 249 nm (obtained from a Gaussian analysis of the corresponding difference absorption profile) with the crossover point in the corresponding CD spectra at 250 nm (36,39). Because the CD spectra are highly sensitive to structural changes, the observed red shift of both low energy CD bands with a crossover point at 258 nm suggest marked alterations of the cluster geometry in Cd 7 MT-4. This conclusion is supported by detailed analysis of the absorption and CD profiles presented in the next paragraph.
Determination of Apparent Binding Constant for Cd 2ϩ Bound to MT-4-To investigate the stability of metal-thiolate clusters in MT-4, the apparent Cd 2ϩ binding constants were determined through pH titration (Fig. 5). This spectroscopic method is based on the competition between protons and metals for cysteine thiolates (25,27). In very recent studies, this method has been used to compare the stability of cadmiumthiolate clusters in MT-1, -3, and -4. However, in these studies, no apparent stability constants for the MT-4 clusters were calculated (40).
The decreasing intensity of the CysS-Cd 2ϩ LMCT band at 250 nm on progressive acidification reflects metal release from the protein.
Cd 7 MT-4 shows a two-step titration profile with the pH midpoint values of 4.1 and 2.9, indicating different stabilities of its two metal-thiolate clusters. Based on the absorbance decrease at the plateau between the two titration steps, the number of thiolate ligands involved in metal binding in each cluster can be derived. Because a total of 20 cysteines are involved in metal binding, the decrease of the CysS-Cd 2ϩ LMCT absorption of ϳ0.45 corresponding to nine cysteines can be assigned to the less stable 3-metal cluster (␤-domain) and that of ϳ0.55 to the 4-metal cluster containing 11 cysteines (␣-domain). The pH midpoint values for each titration step have been used for independent calculation of the apparent Cd 2ϩ binding constant (K app,pHϭ7 ) of both clusters. The obtained values of K app,pHϭ7 of 6.7⅐10 13    Cd 7 MT-2 (27). However, compared with Cd 7 MT-2, the 3-metal cluster in Cd 7 MT-4 shows a slightly lower metal binding affinity with the metal release occurring already below pH 6 and over a wider pH range. This behavior may suggest the presence of metal binding sites with slightly different binding constants within this cluster.
Pathway of Cluster Assembly in Cd 7 MT-4-To gain an insight into the pathway of cluster assembly, the filling up process of apoMT-4 with Cd 2ϩ has been followed by spectroscopic (electronic absorption, CD, and MCD) and spectrometric (nano-ESI-QTOF-MS) techniques. In the absorption spectra, the stepwise addition of Cd 2ϩ equivalents to apoMT-4 (at pH 7.2) resulted in an incremental increase of the absorption profile (Fig. 6A). This is better documented in the difference absorption spectra of metal-bound protein from which the absorption spectrum of apoMT-4 was subtracted (Fig. 6A, inset). The difference absorption envelope of the Cd-MT-4 complexes increases monotonically until the protein saturation with seven Cd 2ϩ equivalents is reached. Further addition of Cd 2ϩ equivalents was without effect, confirming that, in Cd 7 MT-4, all 20 cysteines are involved in metal binding. In the corresponding CD spectra, dramatic changes as a function of the metal binding site occupation occurred (Fig. 6B). The addition of the first Cd 2ϩ equivalent introduced a broad positive CD envelope with an extremum at ϳ240 nm. Further metal addition led to a progressive development of a multiphasic CD profile characterized by two positive ellipticity bands and an interspaced negative ellipticity band. Both the positive CD band at 228 nm and the negative at 245 nm gradually increased in intensity up to seven Cd 2ϩ equivalents added. However, the low energy positive CD band reached maximum intensity already with the first four Cd 2ϩ equivalents. The binding of the remaining three Cd 2ϩ equivalents resulted in a progressive red shift of this CD band from 259 to 266 nm. A plot of the changes in intensity at 258 nm as a function of Cd 2ϩ equivalents bound reveals a clear break point between four and five Cd 2ϩ equivalents (Fig. 6B, inset). These results, together with the results presented below, indicate the progressive formation of the more stable 4-metal cluster, which is followed by the formation of the 3-metal cluster. In the corresponding MCD spectra of CdMT-4 complexes, the characteristic positive A-term signal of Cd 2ϩ binding sites possessing a T d -type of symmetry is preserved throughout the cluster formation process (Fig. 6C).
From the difference CD spectra in which the contribution of the apoprotein was subtracted a better understanding of the pathway of the 4-metal cluster formation could be obtained (Fig. 7). From the occurrence of a broad positive envelope in the first titration step and the development of two isodichroic points in the following three titration steps, it appears that initially a Cd 1 MT-4 species is formed followed by the cooperative formation of the 4-metal cluster. The presence of a nucleation intermediate Cd 1 MT-4 suggests that only a partial cooperativity exists in this cluster. This conclusion is supported by the corresponding ESI-MS titration experiments in which the co-existence of apoMT-4, Cd 1 MT-4, and Cd 4 MT-4 species is observed when less than four Cd 2ϩ equivalents were added (Fig. 8). This suggests that the formation of the nucleation intermediate Cd 1 MT-4 is required to create a template for the cooperative binding of the next three metal ions. It may be noted that the monometallic nucleation intermediate has also been observed in the formation of the Cu I 4 -thiolate cluster in copper chaperone Cox17, which apoform is mainly unstructured (41).
Evidence for a full cooperativity in the 3-metal cluster formation came from the development of isodichroic points at ϳ217 and 239 nm upon the binding of the last three Cd 2ϩ equivalents to Cd 4 MT-4 (see Fig. 6B, CD panel) and from the corresponding ESI-MS data (Fig. 8). In the ESI-MS spectra, besides the mass peak of Cd 7 MT-4, the mass peak of Cd 6 MT-4 was also discerned. The observed mass for Cd 6 MT-4 may originate from the formation of other intermediate species or from the required conditions for the ESI-MS analysis (pH ϭ 6). However, in view of a partial metal release from Cd 7 MT-4 already at pH 6 (see "Determination of Apparent Binding Constant for Cd 2ϩ Bound to MT-4"), we favor the latter effect.
Analysis of Metal-induced Absorption and CD Profiles of Cd 7 MT-4 and Cd 4 MT-4-As discussed above in MT-4, the identical cluster organization of tetrahedral metal binding sites exist when compared with  the MT-1/2 proteins. However, although the CD profile of Cd 7 MT-4 substantially differs from those of Cd 7 MT-1/2, that of the Cd 4 cluster formed during the structure assembly was found to be similar. To gain understanding regarding these differences, we have performed a Gaussian analysis of the difference electronic absorption spectra of Cd 7 MT-4 versus apoMT-4 and Cd 4 MT-4 versus apoMT-4 and compared the underlying absorption bands with the corresponding CD profiles (Fig.  9). It may be noted that the fit of the position of the first two low energy Gaussian bands was rather insensitive to the changes in the position of the only partially resolved high energy Gaussian band. The metal-induced absorption envelope of Cd 7 MT-4 ( Fig. 9B) was resolved into three Gaussian bands at 3.91, 4.21, and 4.72 m Ϫ1 , which correspond to absorption bands at 256, 238, and 212 nm, respectively. Similar analyses of the difference spectrum of Cd 4 MT-4 versus apoMT-4 (Fig. 9C) revealed underlying Gaussian bands at 4.04, 4.37, and 4.73 m Ϫ1 , i.e. at 248, 229, and 211 nm. The position of the latter bands closely resembles the resolved absorption bands at 250, 231, and 201 nm reported for Cd 7 MT-1/2. As shown in Fig. 9, the positions of the first low energy LMCT band in both CdMT-4 forms correspond to the crossover points of the corresponding biphasic CD profile, confirming their excitonic origin. In view of the identical cluster organization in mammalian MT-1, MT-2, and MT-4 reported here, we conclude that the red shift in both absorption and CD features of Cd 7 MT-4 originates from geometrical alterations of the cluster structure(s) and thus in changes of relative orientation of the transition dipole moments of bridging cysteine ligands.
In the well defined structure of rabbit MT-2, the cluster geometry of the 3-metal clusters can be best described as a distorted boat cyclohexane-like ring and that of the 4-metal clusters related to an adamantane-type with the metals and donor atoms arranged in two fused six-membered rings, which almost exclusively adopt distorted boat conformations (6,29). Thus, the identity of excitonic features of the 4-metal cluster in Cd 4 MT-4 with those of the 4-metal cluster in partially and fully cadmium-loaded rabbit MT-2 (36) suggests that their geometries ought to be comparable (6). The CD changes accompanying the cooperative filling of the 3-metal cluster in MT-4 clearly indicate that the formation of this cluster is responsible for alteration of cluster geometries (Fig. 6B, CD panel). However, at present, the obtained data do not afford information indicating whether the geometry of only one or both clusters is altered. In this context, it may be noted that the conformation-dependent excitonic interactions within the clusters, giving rise to strong, conformation-dependent absorption and CD spectra, was not manifested in the MCD spectra. MCD is usually considered to be less sensitive to conformational perturbations. Moreover, to our knowledge, the effect of excitonic splitting on the MCD spectra of metalthiolate clusters has so far not been reported. Taken together, there are substantial geometrical differences of cluster structure(s) in MT-4 compared with those in MT-1/2. Moreover, although a full cooperative formation of both clusters in Cd 7 MT-2 exists, only a partial cooperativity is present in the 4-metal cluster of Cd 7 MT-4.

CONCLUSION
Considering the results obtained in this investigation on Cd 2ϩ and Co 2ϩ binding to MT-4, it can be concluded that this protein possesses divalent metal binding properties similar to the ubiquitously expressed MT-1/2 isoforms. MT-4 is able to bind seven Cd 2ϩ through 20 cysteine thiolates into two separate Cd 4 S 11 and Cd 3 S 9 clusters located in two independent protein domains. All metal ions present in the clusters are tetrahedrally coordinated by both terminal and bridging thiolates. On the other hand, significant differences to mammalian MT-1/2 also exist. Thus, although the pathway of cluster formation shows a similar cooperative and sequential formation of both metal-thiolate clusters, a single metal intermediate is formed within the 4-metal cluster of MT-4 during its assembly. This feature is unique to this MT isoform. In addition, the Cd 2ϩ ions in the 3-metal cluster are bound with a slightly lower affinity and the filling up process of this cluster leads to substantial alterations of the cluster(s) geometry. The reason for these differences may originate from the differences between the primary structures of MT-4 and MT-1/2. Although within the MT-4 subfamily a high sequence homology conservation exists (93.4% identity between mouse MT-4 and human MT-4), a poor degree of conservation in non-cysteine amino acids is seen among MTs present in the same organism, which suggests stronger functional evolutionary constraints for this isoform. From the comparison of MT-4 and MT-1 sequences, a higher divergence for the ␤-domain (3-metal cluster) than for the ␣-domain (4-metal cluster) of MT-4 has been observed (18). In the ␣-domain of MT-4, characteristic Cys-Pro and Cys-Pro-Pro peptide sequences are found (Fig. 1), which are conserved within the MT-4 subfamily. Prolines are known to introduce structural constraints due to the turns imposed on the polypeptide chain by the peptidyl-prolyl bonds and the high energy barrier for their cis-trans isomerization. The influence of such structural constrains on the properties of the ␤-domain of MT-3 has been reported (37,38). In this protein, the conserved Cys-Pro-Cys-Pro motif introduced changes in the 3-metal cluster geometry and its dynamics, features found to be essential for the biological activity of MT-3 (37,38). The higher content of bulky amino acid and a Glu insert in position 5 in the ␤-domain of MT-4 together with the in silico protein sequence analyses led to the conclusion that this protein domain would be substantially more bulky than that in MT-1 (see Fig. 1). In contrast, similar analyses of the ␣-domain revealed that the change in the volumes would be rather small (18). These features of both protein domains in MT-4 are apparently responsible for its structure upon metal binding. Thus, although the geometry of the 4-metal cluster in Cd 4 MT-4 was found similar to that of MT-1/2, the formation of the 3-metal cluster plays a major role in generating the final protein structure. Hence, to account for our data, we hypothesize that the metal filling of the bulky ␤-domain may lead to a mutual interaction between the ␤and ␣-domains of MT-4, which in turn, would lead to a change in geometry, presumably in both clusters. In this case, the structural constraints imposed by the aforementioned prolines may play an important role. In the previous studies the in silico protein sequence analyses of MT-4 together with the results of heterologous E. coli expression of the protein with zinc, cadmium, or copper led to the suggestion that MT-4 differentiated toward copper thionein character, playing a role in copper metabolism in the stratified squamous epithelium (18). However, a role of MT-4 in zinc homeostasis has also been suggested (15). The latter and our data showing similar cadmium binding affinities of the metal-thiolate clusters in MT-4 and MT-1/2, overall similarities in the cluster topologies, the metal binding geometry, and the pathway of cluster assembly make the role of MT-4 in zinc metabolism in keratinocytes evenly likely. Taken together, particular structural features of MT-4 in relation to the metal homeostasis of the extremely specialized tissues in which MT-4 is expressed suggest that this protein may be involved in both zinc and copper metabolism.