Interactions of Bismuth Complexes with Metallothionein(II)*

Bismuth complexes are widely used as anti-ulcer drugs and can significantly reduce the side effects of platinum anti-cancer drugs. Bismuth is known to induce the synthesis of metallothionein (MT) in the kidney, but there are few chemical studies on the interactions of bismuth complexes with metallothionein. Here we show that Bi3+ binds strongly to metallothionein with a stoichiometry bismuth:MT = 7:1 (Bi7MT) and can readily displace Zn2+ and Cd2+. Bismuth is still bound to the protein even in strongly acidic solutions (pH 1). Reactions of bismuth citrate with MT are faster than those of [Bi(EDTA)]−, and both exhibit biphasic kinetics.1H NMR data show that Zn2+ is displaced faster than Cd2+, and that both Zn2+ and Cd2+ in the β-domain (three metal cluster) of MT are displaced by Bi3+ much faster than from the α-domain (four metal cluster). The extended x-ray absorption fine structure spectrum of Bi7MT is very similar to that for the glutathione and N-acetyl-l-cysteine complexes [Bi(GS)3] and [Bi(NAC)3] with an inner coordination sphere of three sulfur atoms and average Bi–S distances of 2.55 Å. Some sites appear to contain additional short Bi–O bonds of 2.2 Å and longer Bi–S bonds of 3.1 Å. The Bi3+ sites in Bi7MT are therefore highly distorted in comparison with those of Zn2+ and Cd2+.

Metallothionein (MT) 1 is an intriguing, low molecular mass (ϳ7 kDa), cysteine-and metal-rich protein. It was first isolated from equine renal cortex 40 years ago (1) and contains 61 amino acids, of which 20 are cysteine residues. Since then, similar proteins have been isolated from the kidney, liver, and intestines of a variety of animal species (2), fungi (3,4), plants (5), and metal-resistant bacteria (6 -8). The two major isoforms of mammalian MT (MT(I) and MT(II)) differ only in minor sequence changes and overall charge. Recently, the discovery of a growth inhibitory factor (GIF) from human brain tissue and nerve and its characterization as a metallothionein (MT-III) has stimulated new interests in studying this small protein (9,10).
The functions of metallothionein are still not fully understood. It appears to play a fundamental role in the metabolism of copper and zinc ions under various physiological conditions (11,12), including its ability to donate metal ions to apo-Zn 2ϩ enzymes (13,14). Metallothionein may also be important for sequestering toxic Cd 2ϩ ions, and probably also Hg 2ϩ (15), Au ϩ (16,17), and Pt 2ϩ (18,19), thereby preventing reactions with other cellular targets in mammals and other higher organisms (20). Metallothionein also appears to play a role in radical scavenging, stress response, and the pharmacology of metallodrugs and alkylating agents (12,21,22).
The best characterized mammalian metallothioneins contain a single polypeptide chain with seven bound metal ions (either Zn 2ϩ or Cd 2ϩ ). The x-ray crystal structure of rat liver Zn 2 Cd 5 -MT(II) (23) and NMR solution structures of rabbit liver Cd 7 -MT(II) (24), rat liver Cd 7 -MT(II) (25), and human liver Cd 7 -MT (26) show that metallothionein contains two structurally independent ␣ (C-terminal) and ␤ (N-terminal) domains, which are linked in the protein via two amino acids. The seven metal ions are present in clusters of four and three metals bound to bridging and terminal cysteine thiolate ligands, with metal-tothiolate ratios of M 4 S 11 and M 3 S 9 for the ␣and ␤-domains, respectively (23). When both Zn 2ϩ and Cd 2ϩ are present, Cd 2ϩ binds preferentially to the ␣-domain, whereas Zn 2ϩ is found preferentially in the ␤-domain (23,28). A Zn 2 Cd three-metal cluster (␤ domain) in MT has the same structure as a Cd 3 cluster (23). The ␣-domain binds Cd 2ϩ ions cooperatively (29). All 20-cysteine residues participate in metal binding, and each of the seven Zn 2ϩ or Cd 2ϩ ions is tetrahedrally coordinated to four cysteine thiolate sulfur atoms (30,31).
Bismuth is known to induce the synthesis of renal metallothionein (32), and it has been shown that pretreatment with bismuth complexes can prevent the toxic side effects of the anti-cancer drug cisplatin without compromising its anti-tumor activity (33)(34)(35)(36). The protection probably involves platinum binding to Bi-induced metallothionein. However, there are few chemical studies of the reaction of Bi 3ϩ with metallothionein (37,38). Such interactions could also play a crucial role in the pharmacology of widely used bismuth anti-ulcer drugs including colloidal bismuth subcitrate (De-Nol ® ) and ranitidine bismuth citrate (Pylorid ® and Tritec ® ) (39 -41). Here we report investigations of reactions of EDTA and citrate complexes of Bi 3ϩ with metallothionein studied by UV, NMR, and x-ray absorption spectroscopy (XAS).  (43). To study reactions with Bi 3ϩ complexes, 0.5 ml of Zn 7 MT(II) solution was placed into a 1-cm cuvette and sealed with parafilm. After temperature equilibration for ϳ10 min in the cuvette, 40 mol eq of Bi(III) complexes ([Bi(cit)] Ϫ or [Bi(EDTA)] Ϫ ) were added and the course of the reaction was monitored by UV spectrophotometry using a computer controlled Perkin-Elmer Lambda 16 spectrometer equipped with a PTP-1 temperature programmer. The absorbance recorded after 2 or 3 days was assumed to represent the equilibrium situation. The kinetic data were analyzed by a nonlinear least squares fitting based on an exponential function using the program Kaleidagraph (Synergy Software). Two kinetic steps were resolved which obeyed first-order kinetics.

Materials
Stoichiometry of Bismuth Metallothionein-Appropriate volumes of a 3.75 mM [Bi(EDTA)] Ϫ solution were added to 1.0-ml aliquots of 15 M Zn 7 MT(II) (in 20 mM Tris-HCl, 10 mM NaCl, pH 7.4) to produce different molar ratios of [Bi(EDTA)] Ϫ to protein and the samples were left for about 1-2 days at 298 K to equilibrate. The absorbance at 350 nm (Bi-S ligand-to-metal-charge-transfer band) was recorded, and the stoichiometry of Bi-MT was obtained from the titration curve. The total sulfur content (due to cysteine and methionine), bismuth, cadmium, and zinc contents were also measured using the ICP-AES (Thermo Jarrell Ash, IRIS) at 180.731 nm (sulfur), 223.061 nm (bismuth), 213.856 nm (zinc) and 226.502 nm (cadmium). The Bi-MT sample for ICP-AES was prepared as follows. After addition of 40 mol eq of [Bi(EDTA)] Ϫ to Zn 2 Cd 5 MT(II) solution (pH 7.6, 50 mM phosphate buffer) and equilibration for 48 h at 298 K, excess of Bi 3ϩ , and displaced Zn 2ϩ and Cd 2ϩ were removed by ultrafiltration (Centricon 3, Amicon). The final solution was diluted with 2% HNO 3 , and the metal and sulfur contents were measured without any digestion of the sample (44).
1 H NMR Studies of Reactions of Bi 3ϩ with Zn 2 Cd 5 MT(II) and Zn 7 MT(II)-1 H NMR spectra of 0.5 mM rabbit liver Zn 2 Cd 5 MT(II) or Zn 7 MT(II) (50 mM phosphate buffer, pH* 7.6) were recorded on a Bruker DMX500 spectrometer operating at 500.13 MHz at 298 K. Typical pulsing conditions for 1 H were: 7-s pulse width, 16,384 data points, 3-s recycle delay, 16 -64 transients. A two-dimensional TOCSY spectrum (mixing time 65 ms) was acquired using 2048 data points in f2 dimension, acquisition time 0.13 s, 48 scans, and 320 increments in the f1 dimension, in a total time of 14 h. All solutions were purged with nitrogen for at least 5 min. After recording an initial proton NMR spectrum, a 40-fold excess over protein of either [Bi(EDTA)] Ϫ or ranitidine bismuth citrate was added to the Zn 2 Cd 5 MT(II) or Zn 7 MT(II) solution, and an NMR spectrum was recorded immediately (with ϳ3 min of mixing). Further spectra were recorded at frequent intervals over the next 9 h. Protein solutions were then allowed to react overnight, and small molecules were removed by ultrafiltration (Centricon 3) and washing four times with 50 mM deuterated phosphate buffer, each about 1.5 h, and then a two-dimensional TOCSY spectrum (mixing time 65 ms) was recorded. The peaks for [Zn(EDTA)] 2Ϫ and [Cd-(EDTA)] 2Ϫ were integrated relative to the internal reference dioxane.
X-ray Absorption Spectroscopy Data Collection and Sample Preparation-X-ray spectra were recorded at the bismuth L III edge on EXAFS station 7.1 at the Daresbury Laboratory Synchrotron Radiation Source. The operating energy was 2 GeV, and average current was approximately 200 mA. Powder samples were finely ground and mounted either as thin films or diluted appropriately with boron nitride. Solution samples were contained in sample cells with mylar windows and a 2-mm path length. In addition to the model compounds used for calibrating phase shifts, two model complexes [Bi(GS) 3 ] and [Bi(NAC) 3 ] were also studied by XAS for comparison. Data for the glutathione and N-acetyl-L-cysteine complexes were collected at ambient temperature in transmission mode using argon/helium-filled ionization chambers (up to three scans per sample). Data for the Bi 7 MT samples were collected at 100 K in fluorescence mode using a scintillation detector (five to eight scans). The monochromator was detuned to 70% to reject harmonics. X-ray damage was checked by comparison of the data sets during collection and by the absence of color changes or edge shifts; there was no evidence of radiation damage to any samples. Samples were prepared as follows. For [Bi(GS) 3 ] and [Bi(NAC) 3 ], 3 mol eq of GSH or NAC were added to 50 mM [Bi(cit)] Ϫ followed by adjustment of the pH to ϳ7.0. Powder samples were prepared by freeze-drying the solution samples. For Bi 7 MT, 5 mg of Zn 2 Cd 5 MT(II) was dissolved in ϳ1 ml of 20 mM Tris-HCl buffer, pH 7.6, deoxygenated with N 2 , and 40 mol eq of [Bi(cit)] Ϫ was added and equilibrated overnight. Some yellow participate appeared in this yellow solution and was separated by centrifugation, washed three times with water, and finally freeze-dried giving a slightly brown solid (hereafter called "brown MT precipitate"). The excess of [Bi(cit)] Ϫ was removed from the yellow solution by ultrafiltration three times using water as eluant (Centricon 3), and the final yellow solution was freeze-dried to give a yellow solid (hereafter called Bi 7 -MT).
X-ray Absorption Spectroscopy Data Analysis-Background subtraction was achieved using the motif-based SPLINE program (45,65), modified for use with EXCURV. Data analysis was accomplished using EXCURV98 (46) via the single scattering curved-wave method for EX-AFS calculations. Edge positions were calibrated against a Au foil at the L II edge, taking the first derivative maximum as 13,734 eV. Phase shifts were derived from ab initio calculations within EXCURV98 and were extensively checked against the crystallographically characterized  (49). Agreement between EXAFS and crystallographic parameters was better than Ϯ0.01 Å for sulfur (Ϯ0.02 Å for 3-4 Å), Ϯ0.02 Å for oxygen with sulfur present (Ϯ0.01 Å if no heavy scatterers present), and Ϯ10% for coordination numbers. The EXAFS data were weighted by k 3 to compensate for the diminishing amplitude at high k. The data range used for analysis varied marginally between samples.

Kinetics of Reactions of Bismuth Complexes with Zn 7 MT(II)-Reaction of bismuth complexes with rabbit liver
Zn 7 MT(II) produced a new UV absorbance band centered at 350 nm. This was used to monitor the progress of reactions of bismuth complexes with metallothionein under pseudo firstorder conditions (40-fold molar excess of Bi 3ϩ over MT). Fig. 1 shows absorption spectra recorded for the reaction of excess [Bi(EDTA)] Ϫ with Zn 7 MT(II) at different times at 298 K, and the time course for the absorbance changes is shown in Fig. 2. The overall reaction was relatively slow, requiring over 10 h for completion. The reaction of [Bi(cit)] Ϫ with Zn 7 MT(II) gave rise to similar spectral changes, although it was much faster. About one-third of the total absorbance change occurred within the first 5 min. However, the total increase in absorbance for both reactions was similar after 1 or 2 days, indicating that equilibrium probably involved the formation of a similar final Bi-MT product.
Kinetic data were analyzed using the nonlinear least squares best fits based on an exponential function. The reaction of [Bi(EDTA)] Ϫ and [Bi(cit)] Ϫ appeared to be biphasic; the rate constants are listed in Table I Stoichiometry of Binding of Bi 3ϩ to Metallothionein-The extent of bismuth binding to Zn 7 MT was investigated in two ways: by determination of the change in absorption at 350 nm and by determination of the bismuth:sulfur ratio in the product via ICP-AES. Fig. 3 shows the change in the extinction coefficient at 350 nm (⌬⑀) with variation of the [Bi(EDTA)] Ϫ :Zn 7 MT molar ratio. The value of ⌬⑀ increases linearly, and plateaus at a ratio of ϳ7.0, reaching a final value of ϳ15,000 M Ϫ1 cm Ϫ1 . By ICP-AES, the stoichiometry of purified Bi-MT (see "Experimental Procedures") was determined to be bismuth:sulfur ϭ 1.0:3.0 (Ϯ0.1), and the mol ratio of bismuth:zinc Ͼ 170:1. Since there are 21 sulfur atoms in MT (20 Cys ϩ 1 Met), the mol ratio of bismuth to MT is 7:1.
Effect of pH on Bi 7 Fig. 4. The intense resonances at ϳ3.1 ppm can be assigned largely to ␤ protons of Cys and Lys residues, and the two singlets at 2.10 and 2.16 ppm to the N-acetyl-CH 3 and ⑀CH 3 , respectively, of the terminal N-acetylmethionine residue (51). The resonances at 0.8 -1.7 ppm can be assigned to methyl groups of Ala, Ile, and Lys residues on the basis of their chem-ical shifts and coupling constants (51 3, which suggested that all Zn 2ϩ and Cd 2ϩ had been displaced from the protein by Bi 3ϩ . Gradual changes were also observed for the metallothionein resonances during the course of the reaction. There was an overall broadening of the resonances, the peak at 1.16 ppm, which can be assigned to the ␤ protons of Ala 23 (CH 3 ) (24) disappeared, the intense peak at ϳ3.10 ppm decreased in intensity, and broadening of the two singlets at 2.10 and 2.16 ppm and peaks from 0.8 -1.7 ppm was notable.
It can be seen clearly from Fig. 5 that Zn 2ϩ was displaced from Zn 2 Cd 5 MT(II) by Bi 3ϩ (added as [Bi(EDTA)] Ϫ ) very rapidly (within 3 min), while Cd 2ϩ was displaced relatively slowly (Ͼ4 h). By the time the first spectrum was recorded, the [Cd-(EDTA)] 2Ϫ peak at 2.76 ppm had reached one-fifth of its final intensity (attained after overnight equilibration). After a further 5 min the peak had doubled in intensity. The rate constants for Cd 2ϩ displacement were determined using nonlinear least squares best fits (Fig. 5). This was a biphasic process with pseudo first-order rate constants of 5.8 ϫ 10 Ϫ3 s Ϫ1 and 1.0 ϫ 10 Ϫ4 s Ϫ1 (half-lives of 2 min and 1.9 h, respectively), whereas the rate of Zn 2ϩ displacement was too fast to determine accurately by NMR spectroscopy.
The reaction of [Bi(EDTA)] Ϫ with Zn 7 MT(II) was also investigated by 1 H NMR spectroscopy. About three-seventh of the Zn 2ϩ was displaced within 3 min of mixing, and the rate could not be determined by NMR. The remaining Zn 2ϩ was displaced by Bi 3ϩ in a biphasic process. Using nonlinear least squares best fits, rate constants of 7.2 ϫ 10 Ϫ3 s Ϫ1 and ϳ5.9 ϫ 10 Ϫ5 s Ϫ1 (half-lives of 1.6 min and 3.3 h, respectively) were determined.
Studies of the reaction of the anti-ulcer compound ranitidine bismuth citrate with metallothionein using UV and NMR spectroscopy were hampered by the strong UV absorption from ranitidine and intense 1 H NMR signals of ranitidine and citrate. However, yellow solutions were obtained when ranitidine bismuth citrate was added to metallothionein under conditions similar to those used for [Bi(EDTA)] Ϫ and [Bi(cit)] Ϫ . The product obtained after ultrafiltration to remove excess ranitidine bismuth citrate and other small molecules was almost identical to Bi 7 MT obtained from the reaction of [Bi(EDTA)] Ϫ with metallothionein as judged by 1 H NMR spectroscopy (data not shown).
The two-dimensional 1 H TOCSY NMR spectrum (mixing time 65 ms) of purified Bi 7 MT is shown in Fig. 6 together with that for Zn 7 MT for comparison. Notable is the absence of several cross peaks assignable to Cys ␤ CH 2 protons (26,52,53) in the region 2.7-3.3 ppm for Bi 7 MT (boxed in Fig. 6).
X   (Table II and Fig. 7) gave rise to coordination numbers of 3.0 Ϯ 0.3, with sulfur as the only coordinating atom, with an average Bi-S bond length of 2.56 Å.
The EXAFS spectra of the two Bi-MT samples (Bi 7 -MT and brown MT precipitate) were quite distinct from each other. By comparing the EXAFS of Bi 7 MT (Fig. 7ii) with those of [Bi(GS) 3 ] (Fig. 7A, i) 3 ]. The Fourier transform for Bi 7 -MT is dominated by a single shell best simulated with 3.0 Ϯ 0.1 sulfur at 2.55 Ϯ 0.01 Å, a similar distance and coordination number as those determined for the model complexes. Fits were attempted with other sulfur coordination numbers, but on refinement the coordination number always returned to 3. The Fourier transform suggests that additional weakly scattering shells are present, including a small amount of oxygen (0.5 Ϯ 0.3) at 2.18 Ϯ 0.02 Å. The coordination number for this shell is poorly defined, but it is certainly less than unity. There is also reasonable evidence that the small peak in the Fourier transform at ϳ3 Å is due to additional sulfur scattering (1.9 sulfurs) at 3.09 Ϯ 0.1 Å. The refined coordination number of 1.9 for this additional shell has a large uncertainty (Ϯ2) and is highly correlated with its own Debye-Waller factor. Therefore, this parameter must be treated with caution. This type of split sulfur ligation has a number of precedents, including the model compound [Bi(SC 6 F 5 ) 3 {SϭC(NHMe) 2 } 3 ] used in this study (3 sulfurs at 2.72 Å, 3 sulfurs at 2.95 Å). X-ray fluorescence measurements on Bi 7 MT gave bismuth: zinc Ͼ 17:1 (the very low level of zinc limited the measurement), in agreement with the ICP-AES measurement.
The brown MT precipitate sample gave much weaker EXAFS amplitudes compared with all the other samples. Fitting the data with a single shell of sulfurs resulted in large residuals in the Fit Index and the model was clearly incomplete in both the EXAFS and Fourier transform. Extensive analysis resulted in two different plausible models of equal quality. Both models suggest that the weaker EXAFS is due to destructive interference of two (or more) shells, rather than lower coordination numbers. Model 1 (Table II)   bismuth as a back-scatterer were unsuccessful. Although the models have equal fit indices, model 2 has a marginally lower R-factor (19.1% versus 20.3%). Additionally the difference in the Debye-Waller factors for the oxygen (0.004 Å 2 ) and sulfur (0.017 Å 2 ) in model 1 seems implausibly large. The refined coordination number for oxygen in model 1 is very highly correlated to a number of other parameters due the interference of the two scattering waves. Because of these potential problems with model 1 and the similarity of model 2 to known model systems, we suggest that model 2 is the more reliable. Brown MT precipitate may represent an intermediate (and less soluble) stage on the path to Bi 7 -MT. It is not clear if it is a fully or partially loaded structural intermediate. DISCUSSION Metallothionein appears to play an important role in human health and disease as well as in the mechanism of action of therapeutic agents. Its synthesis is induced in biological systems by a variety of metal ions including Zn 2ϩ , Cd 2ϩ , Cu ϩ , and Bi 3ϩ , and this induction may provide protection from toxicity by allowing sequestration of the metal (2). Metallothionein may also play a role in cellular resistance to Pt anti-cancer drugs (18,19). Administration of bismuth can induce the synthesis of metallothionein in kidney but not in tumor tissue. This reduces the renal toxicity of cisplatin without compromising its chemotherapeutic activity (33,36). Therefore it has been suggested that bismuth compounds are ideal for clinical application as adjuncts in chemotherapy with cisplatin. Bismuth compounds have been used in medicine for centuries, mainly for the treatment of peptic ulcers and gastric disorders. Therefore, MT may play an important role in the pharmacological activity of bismuth anti-ulcer drugs.
Metal ions have been reported to bind to MT with a variety of stoichoimetries, ranging from 7, 10, to 12 and even up to 20 metal ions per protein molecule and with different geometries (from tetrahedral, square-planar, to linear). These include monovalent metal ions such as Au ϩ (16) and Cu ϩ (54, 55), divalent metal ions such as Zn 2ϩ , Cd 2ϩ , Co 2ϩ , and Hg 2ϩ , trivalent metal ions such as In 3ϩ and Sb 3ϩ (38), and TcO 3ϩ (56,57). In the present studies we determined the stoichiometry of Bi 3ϩ binding to MT as 7:1 using both UV titration and ICP-AES to measure the metal and sulfur contents. This result is in agreement with previous brief reports (37, 38) on bismuth metallothionein. X-ray and NMR studies of Zn x Cd 7-x MTs have shown that the metals are distributed in two clusters as M 4 S 11 (␣) and M 3 S 9 (␤) with the metals coordinated tetrahedrally (30). Previously reported structural data are summarized in Table III. Surprisingly, our EXAFS data suggest that Bi 3ϩ (ionic radius 1.03 Å) coordinates strongly to only three cysteine sulfurs with an average Bi-S bond length of 2.55 Å. This bond length is almost identical to that found in x-ray structures of low M r Bi(III) thiolate complexes (Bi-S 2.5 to 2.6 Å) (49,58). The fit for the EXAFS data for Bi 7 MT suggests the presence of additional sulfur scattering at ϳ3.1 Å. To retain the metal clusters a much distorted tetrahedral geometry is required with three Bi-S bond distances of 2.55 Å and a longer Bi-S contact of ca. 3.1 Å. Although the 3.1 Å distance is well defined, the associated coordination number is not, and the required coordination number for distorted tetrahedral geometry is well within the error range. In the crystal structure of [Bi(SC 6 F 5 ) 3 ], Bi 3ϩ coordinates to three sulfurs with Bi-S bond lengths of 2.53-2.58 Å, and an additional long Bi-S bond of 3.32 Å (49). Molecular modeling of Hg 7 MT (59) suggests that the geometry of Hg 2ϩ is distorted away from tetrahedral due to extensive interactions with solvent water. The recent EXAFS studies on Hg 7 MT have shown that the nearest coordination number for Hg 2ϩ is 2, with Hg-S bond lengths of 2.33 Å and two less well defined long bonds of ϳ3.4 Å (60). It has been thought that the binding site cage of MT is too small to accommodate the volume of tetrahedral HgS 4 units. However, Cd 2ϩ , which has a similar ionic radius (0.92 Å) and the same charge, coordinates tetrahedrally to sulfurs of Cys residues. The coordination geometry is therefore largely dependent on the metal ion, e.g. Ag ϩ , digonal (CN, 2); Cu ϩ , trigonal (CN, 3) (61); and TcO 3ϩ , square pyramidal (CN, 5) (57).
The presence of additional short Bi-O bonds (2.18 Å) for at least some of the Bi 3ϩ ions in Bi 7 MT suggests that some Bi 3ϩ ions also have bound water, or more likely hydroxide or oxide, or possibly oxygen donors from amino acid side chains such as Ser, Thr, Asp, or Glu. Alkoxide donors from Ser or Thr are attractive to consider because Bi-OR (alkoxide) bonds are known to be short and strong in several bismuth citrate adducts (41). No Ser or Thr side chains are particularly close to the metal clusters in zinc and cadmium metallothioneins with known structures, although Ser 32 and Ser 35 are less than 7 Å away from the M 4 S 11 cluster in the ␣ domain, and Ser 2 , Ser 18 , and Ser 28 are less than 7 Å from the M 3 S 9 cluster in the ␤ domain (23)(24)(25). Alkoxide coordination to Bi 3ϩ often gives rise to lone-pair effects and distorted coordination spheres for Bi 3ϩ (41). At biological pH, Zn x Cd y MT(II) (x ϩ y ϭ 7) is negatively charged (Ϫ2), whereas Bi 7 MT(II) would have an overall positive charge (ϩ5). Thus the additional oxygen ligands may serve to neutralize the excess charge.
Previously we have shown (62) that Bi 3ϩ binds to the Cys residue of glutathione and induces large low field shifts of the 1 H NMR resonances of ␤ CH 2 protons (ϳ1.4 ppm). Free and bismuth-bound glutathione exchange at an intermediate rate on the NMR time scale at biological pH (ϳ1500 s Ϫ1 ). The NMR data were also consistent with sulfur-only binding, in agreement with the EXAFS data. The overall broadening of the 1 H NMR spectrum of Bi 7 MT and the disappearance of the twodimensional TOCSY cross-peaks for ␤ CH 2 of Cys residues (Fig.  6) suggest a facile exchange of Bi 3ϩ between different sites, even though binding is thermodynamically very strong. Other cross-peaks, notably those for Lys and Thr, residues that are wide spread throughout the protein, remain unchanged in comparison with Zn 7 MT, indicating that Bi 7 MT probably has a folded conformation although the overall three-dimensional structure may be different. Our gel filtration chromatography studies show that Bi 7 MT migrates in a similar manner to Zn 7 MT, suggesting that their molecular masses and shapes are similar, i.e. Bi 3ϩ has not induced MT polymerization. The disappearance of the ␤ CH 3 1 H NMR resonance of Ala 23 (1.16 ppm) suggests that the structure of Bi 7 MT in this region may be more flexible, and involved in dynamic exchange between different conformations. Previous molecular modeling studies have shown that Cys 26 in the ␤ domain is more solvent accessible and can be displaced from Zn 2ϩ binding by the thiolate sulfur of glutathione (63).
Bi-S bonds in MT(II) appear to be remarkably stable even down to pH values near 1.0, in contrast to Zn 2ϩ and Cd 2ϩ , which are 50% dissociated at pH 4.6 and 3.05, respectively (55), but similar in strength to Cu ϩ -MT (pH1 ⁄2 0.44) (55). The affinity of Bi 3ϩ for MT(II) is therefore higher than that of Zn 2ϩ and Cd 2ϩ .
The kinetics of reactions between bismuth complexes and MT(II) were elucidated in two ways: by observing the absorbance changes at 350 nm due to the formation of Bi-S bonds and the appearance of new 1 H NMR peaks resulting from the formation of either zinc or cadmium EDTA complexes after displacement of EDTA from Bi 3ϩ . 1 H NMR also allowed the kinetics of Zn 2ϩ and Cd 2ϩ displacement by Bi 3ϩ to be monitored separately. Biphasic processes were observed for both [Bi(cit)] Ϫ and [Bi(EDTA)] Ϫ . In the fast step, the displacement of Zn 2ϩ from Zn 7 MT by [Bi(cit)] Ϫ occurred four times faster than that by [Bi(EDTA)] Ϫ , while the second step was very similar for both of these bismuth complexes. This is probably due to differences  (48) in which the Bi 3ϩ ion has apparent vacant coordination sites (perhaps occupied by the 6 s 2 lone pair of electrons) and therefore may be able to attack Cys sulfur of MT more readily than [Bi(EDTA)] Ϫ in which the EDTA ligand is wrapped around Bi 3ϩ (42). By comparing the kinetics of reaction of [Bi(EDTA)] Ϫ with Zn 7 MT and with Zn 2 Cd 5 MT (with Zn 2ϩ in the ␤-domain), we have shown that metal ions in ␤-domain of the MT can be replaced preferentially, and cooperatively and rapidly (within minutes) by Bi 3ϩ , and there is little difference between the reactivity of Zn 3 and Zn 2 Cd ␤-domains. The first step maybe involve displacement of metal ions in the ␤-domain by Bi 3ϩ and one metal ion in the ␣ domain, as judged from NMR data, and the second step may involve displacement of the other metal ions in the ␣-domain. There have been several previous studies of metal and ligand displacement from Zn 7 MT and Cd 7 MT (18,19), and multiphase processes have always been observed. Biphasic kinetics have usually been explained on the basis of formation and breakdown of intermediates (18,19). However only a single kinetic step for metal displacement has been observed for reaction of EDTA with Cd 2ϩ in the ␤-domain of Cd 7 MT, and also one Cd 2ϩ in the ␣ domain can be extracted by EDTA more readily than the other Cd 2ϩ ions (64). Our NMR data indicated the rapid removal of a small amount of Ca 2ϩ from the protein (0.5 Ca 2ϩ per MTII) during reaction with [Bi(EDTA)] Ϫ . We assume that this is surface bound Ca 2ϩ which has little effect on the cluster structures. Its presence has been observed previously (50).
The high stability of bismuth metallothionein suggests that it could play a significant role in the mechanism of action of bismuth-based drugs both in bacteria and in man. It may be an important species for bismuth transport from the liver to the kidney and be involved in bismuth storage in the kidney. In addition, the neurotoxicity of bismuth drugs (encephalopathy) could be related to the binding of Bi 3ϩ to the brain-specific metallothionein(III), since the distribution of Bi 3ϩ in mouse brain is very similar to that of Cu ϩ and Ag ϩ (27). CONCLUSIONS We have established that Bi 3ϩ originating from bismuth citrate anti-ulcer compounds and from [Bi(EDTA)] Ϫ binds very strongly to MT(II) and readily displaces Zn 2ϩ and Cd 2ϩ . Despite its higher charge, Bi 3ϩ also forms a 7:1 complex (Bi 7 -MT), the same stoichiometry for binding as zinc and cadmium. EX-AFS data show that on average each Bi 3ϩ is coordinated strongly to only 3 cysteine sulfurs (at 2.55 Å) in contrast to Zn 2ϩ and Cd 2ϩ , which have clusters based on M(Cys) 4 centers. The presence of less well defined longer Bi-S contacts of 3.1 Å for some Bi 3ϩ ions suggests that Bi 7 -MT does contain clusters, although Bi-Bi contacts were not detectable. For at least some of the Bi 3ϩ ions in Bi 7 -MT, there is additional oxygen coordination as short Bi-O bonds of 2.2 Å. These could arise from bismuth-alkoxide linkages to Ser or Thr side chains or perhaps from oxide coordination. Remarkably, Bi 3ϩ is still bound to MT, even at low pH values (e.g. pH 1), again in contrast to Zn 2ϩ and Cd 2ϩ . Displacement of Zn 2ϩ and Cd 2ϩ from the ␤-domain (M 3 S 9 ) by Bi 3ϩ was much faster than from the ␣-domain. These differences between the structure and reactivity of bismuth metallothionein compared with zinc metallothionein are likely to have implications for its biological properties. Since bismuth citrate complexes are widely used as anti-ulcer drugs, and pretreatment with bismuth compounds can protect against some of the toxic side effects of the anti-cancer drug cisplatin, further studies of bismuth metallothionein are warranted.