Re-evaluation of the binding of ATP to metallothionein.

In a recent paper Jiang et al. (Jiang, L. J., Maret, W. & Vallee, B. L. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 9146-9149) reported that metallothionein interacts with adenosine triphosphate (ATP) to form a 1:1 complex with a dissociation constant of K(d) = 176 +/- 33 microM at pH 7.4. In an effort to characterize further this interaction using nuclear magnetic resonance spectroscopy, titration calorimetry, gel-filtration chromatography, affinity chromatography, and ultrafiltration, we were unable to find any evidence for the binding of ATP to metallothionein.

In an effort to characterize further this interaction using nuclear magnetic resonance spectroscopy, titration calorimetry, gel-filtration chromatography, affinity chromatography, and ultrafiltration, we were unable to find any evidence for the binding of ATP to metallothionein.
Metallothioneins (MTs) 1 are a class of small (Ͻ7 kDa) proteins with a high cysteine content (ϳ30%) and the highest known metal content after ferritins, which bind both essential (Cu 1ϩ and Zn 2ϩ ) and nonessential (Cd 2ϩ and Hg 2ϩ ) metals. MTs are ubiquitous proteins, found in animals, higher plants, eukaryotic organisms, and some prokaryotes. Metal coordination in metallothionein has a high thermodynamic but low kinetic stability. This means that metal binding is very tight, but very facile metal exchange occurs with other proteins. For this reason, MTs are thought to function biologically as intracellular distributors and mediators of the metals they bind (2,3).
Despite the fact that MTs have been investigated for over 40 years, no clear physiological role can be unambiguously assigned to this protein. A recent series of articles by Vallee and co-workers (4 -7) on the redox state dependence of the amount of zinc bound to MT has attracted much attention (7). They have shown that an oxidoreductive mechanism modulates the affinity of zinc for the cysteine thiolate ligands, and key players are GSH, GSSG, and other oxidizing agents (4). Glutathione has been shown to bind specifically to metallothionein (8,9), and with the exception of metals, this ligand has been the only compound shown to bind to MTs. Phosphate ions have been reported to bind to dimeric metallothionein and a special synthetic form of MT termed Cd 13 -(P i ) 2 -MT (10). A phosphate ion was also found in the crystal structure between the two do-mains of metallothionein (11,12).
Recently, Jiang et al. (1) reported the stoichiometric binding of ATP to metallothionein. They used the Hummel-Dreyer method (13) to extract the ATP dissociation constant, and they found that the release of zinc from MT and the behavior of MT on a gel-filtration column was modified in the presence of ATP. In our experimental effort to characterize the structural details of the binding of ATP to metallothionein and the identity of the binding site, we could not find any evidence of an interaction between ATP and MT. In these experiments, several independent methods were used that included NMR spectroscopy, titration calorimetry, gel-filtration chromatography, affinity chromatography, and ultrafiltration.

EXPERIMENTAL PROCEDURES
Materials-Recombinant mouse [Cd 7 ]-MT1 was expressed from a pET3d vector (Novagen) in the Escherichia coli BLR(DE3) strain. 5 ml of LB starter cultures were inoculated with 5 l of glycerol stock and grown for 4 -6 h before two of these were used to inoculate 200-ml LB cultures. These were pelleted after 3 h, resuspended in brain heart infusion media, and used for inoculation of fresh circle grow medium with 1.5% (v/v) preculture. All cultures were grown at 37°C and had 150 mg/liter ampicillin. Expression of MT1 was induced by 1 mM isopropyl-␤-D-thiogalactopyranoside at an A 600 of 1.2, followed by a 30-min growth before addition of 0.4 mM CdSO 4 . Cells were harvested 12 h post-induction, pelleted, washed with 20 mM Tris-HCl, 0.25 M sucrose, pH 8.0, and resuspended in the same buffer with 0.04% ␤-mercaptoethanol. They were lysed with a French pressure cell. After centrifugation the supernatant was loaded onto a POROS HQ 20 anion exchange column (2 ϫ 25 cm) run with an Ä kta fast performance liquid chromatography system (Amersham Pharmacia Biotech) and equilibrated with 20 mM Tris-HCl, pH 8.6. MT1 was eluted by washing the column with 3.5 column volumes of the same buffer. Fractions were monitored at 280, 254, and 220 nm, and their cadmium content was assessed by atomic absorption spectroscopy (Varian SpectrAA-100). Pooled fractions that contained MT1 were concentrated by ultrafiltration (Amicon, YM3 membrane) in the presence of 1.2 mM dithiothreitol before loading onto a Sephadex G-75 gel filtration column (2.6 ϫ 80 cm) equilibrated with 20 mM Tris-HCl, pH 8.6. The final peak was pooled and concentrated for further studies.
Cd,Zn-metallothionein II (Cd,Zn-MT2) and ATP were purchased from Sigma, and NHS-activated Hi-Trap affinity columns were obtained from Amersham Pharmacia Biotech.
NMR Spectroscopy-The 31 P NMR experiments were carried out on a Varian Unity INOVA 500 MHz NMR spectrometer and all other NMR experiments on a Varian Unity INOVA 800 MHz NMR spectrometer. Watergate solvent suppression (14) was used for the experiments involving proton detection. All measurements were carried out at 25°C, and the pH was kept constant at 7.4. The proton spectra were referenced using the chemical shift of water, which is 4.76 ppm at 25°C (15). For the titrations of metallothionein with ATP, the concentration of mouse-MT1 was kept at 0.29 mM, and depending on the sensitivity of the experiment various ranges of ATP concentrations were used (between 0.1 and 6.0 mM). NOESY spectra (16) with mixing times between 100 and 400 ms were acquired on samples containing 0.6 mM mouse-MT1 and concentrations of ATP varying between 0.2 and 20 mM in a search for transfer NOE peaks. Longitudinal (T 1 ) 31 P NMR relaxation times were acquired with the inversion recovery method (17), whereas the proton T 1 relaxation times were obtained with a diffusion-edited inversion recovery experiment (18) in order to suppress the signals of the protein.
Titration Calorimetry-All calorimetric experiments were carried out using the OMEGA titration calorimeter from MicroCal, Inc. (Northampton, MA) (19), at 25°C and pH 7.4. The buffers used for the titration were degassed in vacuo for at least 30 min. After an equilibration time of 900 s, a 2-ml metallothionein solution was titrated with 40 equal volumes of 6.25 l of 10 mM ATP with a time interval of 210 s between each addition of ATP. The solution was stirred at 400 rpm during the titration. Changes in temperature in the cell containing the MT sample were recorded as a function of the volume of added ligand solution. Since the reference cell of the calorimeter acts only as a thermal reference for the sample cell, it was filled with water containing 0.01% azide.
Gel-filtration Chromatography-A Sephadex G-75 column (1.5 ϫ 80 cm) was used to evaluate possible changes in the shape or size of MT upon ATP binding. It was equilibrated with a 50 mM Hepes⅐NaOH, 10 mM NaCl buffer, pH 7.4, or the same buffer including 0.5 mM ATP. The flow rate was set to 0.3 ml/min, and all experiments were carried out at 23°C. The applied sample size was 0.5 ml and the eluate monitored at 250 nm.
To investigate the binding of ATP to MT using the Hummel-Dreyer method (13,20), a Sephadex G-25 column (1 ϫ 30 cm) was equilibrated with 50 mM Hepes⅐NaOH buffer, pH 7.4, and various concentrations of NaCl (0 or 100 mM) and ATP (50, 100, or 200 M), whereby the concentrations of both ATP and NaCl were kept constant during each chromatographic separation. A UV detector was used at 250 nm to screen the eluate. At this wavelength, both MT and ATP show an absorbance, and the steady concentration of ATP in the buffer gives rise to an elevated base line. MT is dissolved in the elution buffer containing the same concentration of ATP (50, 100, or 200 M) and applied to this column. In the case of binding, the concentration of free ligand is reduced by an amount equivalent to the hypothetical protein-ligand complex formed. As this complex emerges at the end of the column, the total amount of UV-detectable ligand in the eluate rises above the equilibrium value. Correspondingly, at some point after the protein peak the concentration of ligand in the eluate is decreased below the base-line level to form a trough.
Affinity Chromatography-Prepacked NHS-activated 1-ml Hi-Trap columns from Amersham Pharmacia Biotech were used for all affinity chromatography experiments. The gel is based on highly cross-linked agarose beads with 6-atom spacer arms attached to the matrix by epichlorhydrine and activated by N-hydroxysuccinimide, which binds to primary amine groups. The substitution level of this gel is Ϸ10 M NHS groups/ml gel. A coupling buffer of 0.2 M NaHCO 3 , 0.5 M NaCl, pH 8.3, was used, and unreacted NHS groups were deactivated by repeated washing with 0.5 M ethanolamine, 0.5 M NaCl, pH 8.3, and 0.1 M acetate, 0.5 M NaCl, pH 5, solutions. Sample volumes of 50 l were put onto the column and operated at a flow rate of 0.44 ml/min.
Ultrafiltration-The retention of ATP due to binding to MT was investigated by ultrafiltration using 2-ml Centricon concentrators (molecular mass cut-off 3 kDa) and centrifugation at 6000 ϫ g. Inside the micro-concentrator was a solution containing 0.7 or 0.4 mM metallothionein and 2 mM ATP in a 20 mM Tris-HCl buffer, pH 7.4. The relative concentrations of ATP in the retentate and in the eluate were determined by 31 P NMR spectroscopy with an internal standard of 1 mM PO 4 3Ϫ and relaxation delays long enough to allow complete restoration of equilibrium magnetization after each scan.

RESULTS AND DISCUSSION
NMR Spectroscopy-Mouse metallothionein-1 was titrated with ATP, and the 1 H NMR chemical shifts of both mouse-MT1 and ATP and the 31 P NMR chemical shifts of ATP were monitored. Binding phenomena whether or not they involve structural changes should be easily detected by this method and manifested in changes in either or both the respective chemical shifts and line widths. However, no changes in any of the chemical shifts or line widths were observed (Fig. 1). More dramatically perhaps are the superimposable NOESY spectra of mouse MT1 and mouse MT1 ϩ ATP as shown in Fig. 2 for the fingerprint region. The indifference of even the labeled side chain lysine protons, which according to Jiang et al. (1) should be the closest to the binding, is clearly manifested. It is worth mentioning that care needs to be exercised when dissolving ATP in the 20 mM Tris buffer at pH 7.4 since the acidity of ATP disodium salt is high enough to change the pH in this buffer even with minute amounts of ATP. The chemical shifts of the 31 P resonances are particularly susceptible to pH changes, and thus careful matching of the pH of the ATP solution used for the titration to the pH of the MT sample is necessary.
Another sensitive NMR parameter to detect binding of a small ligand to a larger protein is by observing the line widths or more accurately the relaxation times of the ligand. The longitudinal (T 1 ) relaxation times of 1 H and 31 P resonances of ATP in Fig. 1 do not indicate a change in the motional behavior of ATP, which can only be explained if ATP is not bound to or interacting with metallothionein to any significant extent.
The binding of a small ligand to a protein can be effectively studied by transfer NOE experiments (18,21,22) in the case of fast exchange between the bound and free ligand. These transfer NOE peaks result from the change in rotational correlation time and thus relaxation times of a small ligand upon the binding to a relatively large protein. The free ligand has small and positive NOEs, but when it is bound to a macromolecule, the effective correlation time is increased which results in large and negative NOEs. We acquired NOESY spectra on samples with ATP:mouse-MT1 ratios of 2:1, 5:1, 10:1, and 20:1 but could not detect any transfer NOE peaks. In fact, the small and positive NOE peaks that were also found for free ATP did not change their appearance at any ATP:MT ratio.
Titration Calorimetry-The binding isotherm corresponding to a plot of integrated heats of titrating a sample of 0.34 mM mouse MT1 or 0.2 mM Cd,Zn-MT2 in 20 mM Tris-Hcl, pH 7.4, with solutions of 10 mM ATP in the same buffer and at the same pH did not show any release of heat upon mixing, which is consistent with a zero change in enthalpy. Due to restricted mobility and flexibility of both the protein and the watersoluble ligand, a binding interaction would lead to an entropy change that would be negative for this reaction, and therefore any binding must be accompanied by a larger negative enthalpy change to give a negative free energy ⌬G ϭ ⌬H Ϫ T⌬S.
Gel-filtration Chromatography-Jiang et al. (1) reported that the apparent molecular size of metallothionein as measured by a gel-filtration column was reduced in the presence of ATP. We used a Sephadex G-75 column (1.5 ϫ 80 cm) to investigate this behavior and found no changes in the retention times of Cd 7 -MT1 in the chromatograms (Fig. 3). The trough at the position of ATP in the chromatogram of mouse-MT-1 on a column equilibrated with a buffer containing 0.5 mM ATP (Fig. 3D) stems from the fact that the sample applied to that column contained just 0.36 mM ATP.
A reduction in the apparent molecular size of MT would very likely be related to a change in the tertiary structure of MT as mentioned by Jiang et al. (1). The dumbbell-shaped metallothionein molecules are known to elute at an apparent molecular mass about twice the actual molecular mass of about 7 kDa. Elution at a lower molecular weight indicates a more compact structure, which could arise from enhanced interdomain interactions that would result in a more compact metallothionein structure and a loss of the dumbbell shape. Another example of a related behavior in MTs was described by Palumaa et al. (10) when MT2 was treated with excess cadmium in the presence of phosphate ions. This resulted in a form of MT that contained 13 cadmium atoms and 2 phosphate groups per monomer which was termed Cd 13 -(P i ) 2 -MT. What is even more concerning in the report by Jiang et al. (1) is that phosphate competes with ATP for the binding to MT, since phosphate has already been shown to bind just to dimerized MT, to MT crystals, and to the above-mentioned Cd 13 -form of MT. Is it possible, therefore, that Jiang et al. (1) may have somehow produced a form of MT resembling Cd 13 -(ATP) x -MT? In this regard it is perhaps noteworthy that our attempts to produce such a species according to the procedure used to produce Cd 13 -(P i ) 2 -MT were unsuccessful.
Another way to analyze the binding of a ligand to a protein by gel-filtration chromatography is the Hummel-Dreyer method. For this procedure, the column is equilibrated with a buffer containing the ligand. The protein is dissolved in the buffer containing the same concentration of ligand as the elution buffer. The ligand that needs to be detectable by the monitor gives rise to a constant but elevated base line. Binding of the ligand to the protein leads to a reduced concentration of the ligand in the buffer where the sample is applied onto the column. The deficit of ligand travels with the speed of the ligand through the column Re-evaluation of the Binding of ATP to MT and gives a negative peak or trough at the position where free ligand would elute. By repeating the Hummel-Dreyer method as reported by Jiang et al. (1), we too found a trough at the position of ATP if MT was applied onto the column. However, we also found troughs for every other protein tested in this way (lysozyme, bovine serum albumin, ␣-chymotrypsinogen, RNase, and insulin). Critical evaluation of this experiment revealed that by increasing the ionic strength, the trough disappeared. The Sephadex gel (G-25) used for this experiment is known to bind positively charged and aromatic molecules to some extent (23)(24)(25), and the partial exclusion of negatively charged ATP from this column material has been reported (24). It is also known that increasing the ionic strength reduces or eliminates the repulsion of ATP on Sephadex columns (24). We therefore postulate that by loading a protein onto the column, the exclusion of ATP is decreased by either slight binding of the protein to the column or increasing the local ionic strength and thereby decreasing the effective charge of ATP. As soon as the protein moves down the column, ATP molecules can occupy more space due to the now partially shielded charges, leading to a higher local concentration of ATP in the moving protein zone and thereby reducing the concentration of ATP behind the protein, leading to the observed trough. In other words, the free energy ⌬G of ATP on the column is increased in the presence of MT, leading to an accumulation of ATP around MT that could be misinterpreted as binding to the protein. Further work to study this effect is in progress.
Affinity Chromatography-The immobilization of proteins on a solid support matrix and the use of this material to bind protein-specific ligands is known as affinity chromatography which has been described extensively in the literature (20). Alternatively, the ligand can be bound to the support matrix and used to fish out specific binding proteins. Mouse-[Cd 7 ]-MT1, Cd,Zn-MT2, or ATP was immobilized to NHS-activated Hi-Trap affinity columns via their primary NH 2 groups through 6-atom-spacer arms. Samples of 50 l, 50 M ATP were applied onto the MT affinity columns and eluted with a flow rate of 0.275 ml/min. The elution volume in the absence of binding was obtained with solutions of MT and other proteins, like lysozyme, bovine serum albumin, or ␣-chymotrypsinogen. Another column was prepared with hexokinase bound to serve as a reference. The elution times of various ligands on the protein binding columns are shown in Fig. 4, along with the elution times of various proteins on the ATP column. Although the binding of ATP to hexokinase and glutathione to metallothionein was clearly demonstrated by this technique (Fig. 4), no indication of an interaction between either MT isoform and ATP was found (Fig. 4, B-D). The proposed sites on MT responsible for ATP binding are lysine NH 2 groups, and these groups are also responsible for the binding of MT to the affinity columns. However, MT contains 7 (MT2) or 6 (MT1) lysine residues, and just one is needed for the binding to the NHSactivated columns. Therefore, after binding to the affinity columns there should still be enough lysine residues available. Alternatively, the fact that the ATP affinity column did not bind MT could theoretically indicate an involvement of this group in the binding of MT. However, Jiang et al. (1) suggested the binding to occur via the NH 2 groups of the lysine residues on MT and the competition with phosphate ions which would indicate that the binding would occur via the negatively charged phosphate groups of ATP.
Ultrafiltration-A very simple method to analyze the binding behavior of proteins to small ligands is ultrafiltration. The protein and ligand are mixed together, and the separation of bound and free ligand is achieved by filtration devices with a molecular weight cut-off between the ligand and the protein. Therefore, if the concentration of ligand before the filtration is x and y is the portion of x bound to the protein, then the concentration of ligand in the filtrate is x Ϫ y. We used Centricon concentrators with a cut-off of 3 kDa. Ultrafiltration in the absence of any protein showed that ATP (2 mM) was retarded by the membrane to a small extent (ϳ5%). This served as our blank and in the presence of either 0.7 mM [Cd 7 ]-MT1 or 0.4 mM Cd,Zn-MT2 no increase in this amount was observed. One can calculate that binding with a K d ϭ 176 M, as reported (1), would give a decrease of ATP concentration in the filtrate of 0.62 mM (mouse-MT1) or 0.36 mM (Cd,Zn-MT2).
The fact that each and every one of these very different experiments to probe molecular interactions did not give any indication of an interaction between MT and ATP should perhaps not be too surprising considering that MT does not carry any ATP binding consensus sequence (26), and simple electrostatic interaction between a lysine side chain and the phosphate groups on ATP is not sufficient to give a K d ϭ 176 M. Jiang et al. (1) also reported an increased metal release and metal transfer between metallothionein and sorbitol dehydrogenase upon the addition of ATP to MT. We can only hypothesize that either the transfer or release of metal from metallothionein is enhanced by the presence of weak electrostatic interactions with ATP or that by dissolving ATP at the concentration (1 mM) used, the pH of the sample was lowered, which is well known to result in increased metal release in MTs (27)(28)(29).
In conclusion, with the use of five independent methods to probe molecular interaction, we could not find any indication of the binding of ATP to metallothionein or a structural change of MT in the presence of ATP.