Mapping the zinc ligands of S100A2 by site-directed mutagenesis.

S100 family proteins are characterized by short individual N and C termini and a conserved central part, harboring two Ca(2+)-binding EF-hands, one of them highly conserved among EF-hand family proteins and the other characteristic for S100 proteins. In addition to Ca(2+), several members of the S100 protein family, including S100A2, bind Zn(2+). Two regions in the amino acid sequences of S100 proteins, namely the helices of the N-terminal EF-hand motif and the very C-terminal loop are believed to be involved in Zn(2+)-binding due to the presence of histidine and/or cysteine residues. Human S100A2 contains four cysteine residues, each of them located at positions that may be important for Zn(2+) binding. We have now constructed and purified 10 cysteine-deficient mutants of human S100A2 by site-directed mutagenesis and investigated the contribution of the individual cysteine residues to Zn(2+) binding. Here we show that Cys(1(3)) (the number in parentheses indicating the position in the sequence of S100A2) is the crucial determinant for Zn(2+) binding in association with conformational changes as determined by internal tyrosine fluorescence. Solid phase Zn(2+) binding assays also revealed that the C-terminal residues Cys(3(87)) and Cys(4(94)) mediated a second type of Zn(2+) binding, not associated with detectable conformational changes in the molecule. Cys(2(22)), by contrast, which is located within the first EF hand motif affected neither Ca(2+) nor Zn(2+) binding, and a Cys "null" mutant was entirely incapable of ligating Zn(2+). These results provide new information about the mechanism and the site(s) of zinc binding in S100A2.

The S100 family of small cation-binding proteins currently comprises 16 closely related members and two proteins containing the S100 sequence as part of a fusion protein. These and 13 of the S100 genes are clustered in the epidermal differentiation complex located on human chromosome 1 (1). The Ca 2ϩ -binding sites are well defined as two EF-hands. The Cterminal EF-hand follows the canonical consensus of the common EF hand motif, while the N-terminal one is S100-specific (2,3). Although Zn 2ϩ binding has been demonstrated unequivocally for several S100 proteins, including S100A2 (4), the region(s) involved in this interaction have not yet been identified (5)(6)(7)(8)(9). In particular, molecular characterization has been hampered by the absence of a conserved cysteine-or histidine-containing Zinc-finger-like motif (10). Recently, Brodersen and colleagues reported the crystal structures of the human S100A7 (psoriasin) dimer, both in the Zn 2ϩ -loaded and the free form and determined two zinc binding sites per dimer (11). From their data, these authors further delineated two regions, one located in the first, variable EF-hand of one monomer and the second residing in the C terminus of the other monomer, which could function as potential ligands for zinc ions. Ca 2ϩ and Zn 2ϩ binding for wild type S100A2 and for mutants in which the EF-hands were substituted for parvalbumin EFhands have been studied previously (4). However, the detailed mechanism of Zn 2ϩ binding remained elusive.
Human S100A2 protein has attracted particular interest based on its potential function as a tumor suppressor-related gene (12). This hypothesis is based on the observations that S100A2 expression is down-regulated in breast cancer (13,14) and that the S100A2 promoter can be activated by p53 (15). In addition, localization studies have shown that a substantial fraction of the S100A2 protein is present in the nucleus in interphase cells (16 -18).
Human S100A2 harbors four cysteines, located at strategically interesting sites, that may be involved in Zn 2ϩ -binding. One cysteine residue, Cys 1(3) (the number in parentheses indicating the position in the sequence of S100A2) is located at the very N terminus of the molecule. A cysteine at the same position has been proposed as a potential Zn 2ϩ ligand in calcyclin (S100A6) by Kordowska et al. (9). The second cysteine, Cys 2 (22) resides within the first EF-hand. The first EF-hand of S100 proteins has only low affinity for Ca 2ϩ , is 2 amino acid residues longer than canonical EF-hands and, in the present case, contains a cysteine flanked by a highly conserved histidine residue. Moreover, the corresponding site in S100A7 seems to be involved in Zn 2ϩ binding (11). The third and fourth cysteine in human S100A2, Cys 3(87) and Cys 4(94) , are situated in the C terminus adjacent to the penultimate helix-loop-helix motif. In this region, histidine and cysteine residues are found in most human S100 proteins. Based on theoretical models (19), NMR structures of the apo-form of related S100 family proteins, and sequence alignments (10), these residues have been hypothesized to be ligands for Zn 2ϩ .
In this study, we map the amino acids that can act as ligands for the zinc ion in S100A2 by site-directed mutagenesis and correlate the effects of the absence or presence of a single zinc-ligating amino acid with the binding characteristics of the wild type molecule. Serial substitutions of each of the four cysteines to serine residues did not alter the Ca 2ϩ -binding properties of human S100A2 but induced significant changes in the Zn 2ϩ binding characteristics of the protein. In the wild type protein, Zn 2ϩ binding induced conformational changes in the protein. In addition, the affinity for Zn 2ϩ were raised 5-fold in the presence of Ca 2ϩ . In the absence of a conserved sequence motif that would provide all of the four necessary ligand amino acids for the Zn 2ϩ ion, we hypothesize an involvement of the entire intact dimer to contribute to the Zn 2ϩ binding activity of S100A2.

MATERIALS AND METHODS
Electrophoresis and Western Blotting-Analytical SDS gel electrophoresis on 10 -26% gradient polyacrylamide minislab gels and Western blotting onto nitrocellulose (Hybond; Amersham Pharmacia Biotech) were performed essentially as described elsewhere (20). Transferred proteins were detected by incubation with a monoclonal anti-S100A2 antibody (SH-L1; Sigma) and visualized using horseradish peroxidase-coupled secondary antibodies and the ECL chemiluminescence system (Amersham Pharmacia Biotech).
Purification of Recombinant S100A2 Mutants-Recombinant proteins were expressed in the Escherichia coli host strain BL-21. Purification was carried out according to a modified protocol by Becker et al. (22). Bacteria were grown in LB medium containing 100 g/ml ampicillin and induced at logarithmic growth by the addition of 0.5 mM isopropyl-␤-D-thiogalactopyranoside. After growth for an additional 3 h, bacteria were harvested by centrifugation in a Sorvall GS-3 rotor at 2000 ϫ g for 15 min at 4°C. The bacterial pellet was resuspended in buffer C1 (20 mM Imodazole, pH 7.5, 20 mM NaCl, 0.5 mM EDTA, 0.5 mM dithioerythritol, 0.5 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, pepstatin) and opened with a French pressure cell (Spectronic Unicam, Cambridge, UK). After removal of cell debris by centrifugation in a Sorvall SS-34 rotor at 20,000 ϫ g for 30 min at 4°C, the supernatant was dialyzed against buffer C1. Proteins were applied onto a Whatman DE-52 anion exchange column (60 ml) and eluted with a linear gradient from 20 mM to 1 M NaCl. The fractions containing recombinant S100A2 were pooled, dialyzed against buffer C1 containing 2 mM Ca 2ϩ , and applied onto a 10-ml phenyl-Sepharose column (Amersham Pharmacia Biotech). After elution of the recombinant protein with 4 mM EGTA, the protein was Ͼ90% pure. For further purification, the protein was applied to an Amersham Pharmacia Biotech S100-HR gel filtration column (120 ml) in buffer C1.
Mass Spectrometry-Mass spectra of wild type S100A2 or mutant proteins were generated on a PE SCIEX API 365 LC/MS/MS electrospray ionization mass spectrometer (Perkin-Elmer). The protein samples were dialyzed against water and diluted to a final concentration of 50 g/ml in running medium containing 50% acetonitrile and 0.1% acetic acid or 3 mM ammonium acetate. 20 l of the protein solution were injected into the electrospray ionization mass spectrometer (ESI-MS) 1 in the presence of 300 M CaCl 2 , 50 M ZnCl 2 or without metal ions.
Analytical Ultracentrifugation-Sedimentation assays were carried out on a Beckmann Optima XL-1 (Beckman Coulter, Fullerton, CA) centrifuge at 56,000 rpm and 20°C using 0.3 mg/ml recombinant S100A2 in buffer C1 containing 2 mM Ca 2ϩ , 100 M Zn 2ϩ , or 2 mM EDTA. The molecular mass of the complexes in solution (i.e. their oligomeric state) was analyzed as described earlier (23,24).
CD Spectra-The ␣-helical content of the protein and its change during Ca 2ϩ binding were determined by measuring far-UV circular dichroism in a spectropolarimeter model 62D (Aviv Instruments, NJ). Spectra of 100 M recombinant S100A2 in buffer C1 in the absence or presence of Ca 2ϩ and/or Zn 2ϩ were taken from 205 to 250 nm. Ellipticity at 222 nm is negative proportional to the ␣-helical content of the protein when normalized to the residue concentration ( MRW ϭ 0 ϫ 115 g/mol/(1 mm ϫ 1 mg/ml). Data were processed as described elsewhere (4). Photometric and Fluorometric Assays-Internal tyrosine fluorescence emission spectra were obtained on a Hitachi F2000 spectrofluorimeter (Hitachi Instruments, CA) at room temperature and ext ϭ 280 nm and em ϭ 260 -360 nm. Emission scans were performed with 150 M recombinant S100A2 in buffer C1 in the presence or absence of varying Ca 2ϩ -, Zn 2ϩ -, and EDTA concentrations. The maximum change in fluorescence at 315 nm was plotted as a function of the cation concentration against the fluorescence emission at 315 nm. Before the titrations, recombinant S100A2 and mutant proteins were reduced by an overnight incubation with 100 mM dithioerythritol in buffer C1 at room temperature followed by gel filtration on a PD10 column (Amersham Pharmacia Biotech) with degassed buffer C1 lacking dithioerythritol.
The amount of free sulfhydryl was determined by measurement of the absorbance at 412 nm after a 15-min incubation with 1 mM 5,5Јdithiobis-2-nitrobenzoic acid (DTNB). During the reaction with free sulfhydryl residues, an equivalent amount of 5-nitro-2-benzoate anions is released, which can be quantified by monitoring the absorption at 412 nm. The extinction coefficient of the reaction product is 13,600 OD/M ϫ cm. The reactivity of the cysteine residues in dependence of metal binding was determined by measuring the change of the absorption at 412 nm with the time after the addition of 2 mM Ca 2ϩ , 40 M Zn 2ϩ , or 4 mM EDTA and 100 M DTNB. All photometric measurements were carried out on a Hitachi U3000 Spectrophotometer (Hitachi Instruments, CA).
UV absorption spectra were scanned from 240 to 310 nm. The absorbance at 278 nm was used to calculate the sample concentration with ⑀ 10 g/liter ϭ 3.03.
Solid Phase 65 Zn Binding Assay-2-10 g of the recombinant human S100A2 wild type or mutant proteins diluted in 20 l of buffer C1 was spotted onto nitrocellulose membrane (PALL, Linz, Austria), employing a vacuum dot blot apparatus (Bio-Rad). After 20 min of incubation, the blot was washed twice in buffer C1 while still in the apparatus and subsequently transferred to a Petri dish and equilibrated in metal binding buffer (20 mM HEPES, pH 7.4, 100 mM NaCl). The membrane was then cut into strips (11 ϫ 0.7 cm) containing 12 spots corresponding to the different mutant proteins (or protein concentrations) and incubated for 20 min in 2 ml of metal binding buffer containing varying ZnCl 2 concentrations traced with 65 ZnCl 2 . After incubation, the membrane was washed twice in metal binding buffer containing no metal ions and exposed to an x-ray film (Eastman Kodak Co.) overnight.
Cell Fractionation Experiments-Cells (porcine LLC-PK1 or human A431) were grown to 90% confluence in 10-cm tissue culture Petri dishes (Falcon), washed in HBS (10 mM Hepes, pH 7.4, 130 mM NaCl, 15 mM KCl, 2 mM MgCl 2 ), and harvested in 2 ml of HBS using a tissue culture scraper. For harvesting the cells and all following steps, the HBS contained either 1.5 mM CaCl 2 , 5 mM EGTA, or 50 M ZnCl 2 , respectively. The cell suspension was homogenized with a glass-glass tissue grinder and fractionated by ultracentrifugation for 30 min at 100,000 ϫ g (Sorvall, RC M150 GX) at 4°C. The pellets were resuspended in 2 ml of HBS, containing the corresponding metal ions. For human A431 cells, the extract was mixed with 5 g of recombinant S100A2 protein (c end ϭ 2.5 g/ml). The microsomal fraction was incubated for 5 min on ice and ultracentrifuged for 30 min at 4°C and 100,000 ϫ g. The pellet was washed by a second resuspension and centrifugation step in the corresponding buffer. All pellets and supernatants were precipitated following the method of Wessel and Flugge (25) and resuspended in SDS sample buffer, and an equivalent of 2 cm 2 of cells was applied to an SDS-polyacrylamide gel.

Biophysical Characterization of Wild Type S100A2
Recombinant S100A2 was purified to homogeneity ( Fig. 1A) as described under "Materials and Methods." Similar to other members of this family S100A2 formed dimers both in the presence and absence of metal ions, as shown by analytical ultracentrifugation (Fig. 1B). The conformational changes taking place during Ca 2ϩ -binding resulted in a significant change of the tyrosine fluorescence, presumably due to the exposition of a hydrophobic protein interface (Fig. 1C). The half-maximal saturation for Ca 2ϩ in the presence and absence of Zn 2ϩ was in the range of 350 M. The addition of Zn 2ϩ to the wild type S100A2 molecule in the presence or absence of Ca 2ϩ led to an increase in tyrosine fluorescence greater than that induced by the addition of Ca 2ϩ alone. (see below). Upon exposure to low concentrations of Zn 2ϩ , the recombinant protein oligomerizes (Fig. 1B) and starts to precipitate in the upper micromolar range. CD spectra also detected conformational changes in the ␣-helical content during Ca 2ϩ binding in the absence of Zn 2ϩ , whereas no change was observed during Zn 2ϩ binding in the absence of Ca 2ϩ . Remarkably, the change induced in CD spectra by Ca 2ϩ is reversed upon further addition of Zn 2ϩ (Fig. 1D).
Recombinant S100A2 can bind Zn 2ϩ above its physiological range causing the protein to precipitate. This is assumed to be due to the presence of four cysteines in the protein, which are all capable of chelating Zn 2ϩ as shown in the DTNB test. Upon the addition of 40 M Zn 2ϩ , the reactivity of all four sulfhydryl residues of S100A2 to DTNB was dramatically decreased (see also Fig. 6). To overcome this problem and to identify the cysteine(s) involved in Zn 2ϩ binding, we constructed a battery of cysteine-deficient mutants. The cysteines are tentatively named Cys 1(3) , Cys 2 (22) , Cys 3(86) , and Cys 4(94) , the numbers 1-4 referring to the order in the amino acid sequence and the numbers in parentheses indicating the positions in the amino acid sequence. The mutants were numbered as ⌬n, where the number n refers to the cysteine(s) substituted for serine. The mutations and tentative names are summarized in Table I. The actual molecular mass as detected by ESI-MS and the theoretical mass calculated with the program GPMAW is shown in Table II. The deviation of 1-2 daltons between calculated and detected values can be caused by rounding the decimal places in the computer program, inaccuracies in the measurement, or loss of protons due to formation of disulfide bonds. monoclonal anti-S100A2 recognizing both the monomer (M) and the dimer (D). B, pure S100A2 in solution is a dimer as determined by analytical ultracentrifugation. The dimeric state is stable in buffer C1 without metal ions (1), in buffer C1 containing 1 mM Ca 2ϩ (2), and in buffer C1 containing 2 mM EGTA (3). The addition of 100 M Zn 2ϩ leads to oligomerization and precipitation during the experimental procedure (4). C, the half-maximal saturation concentration for Ca 2ϩ was determined in titration experiments using freshly reduced protein and found to be c(Ca 2ϩ ) 50 ϭ 300 Ϯ 10 M Ca 2ϩ . The shape of the resulting curve identifies positive cooperativity of the two Ca 2ϩ -binding sites. D, UV CD spectra revealed an increase in the ␣-helical content of the protein in the presence of 1 mM Ca 2ϩ (dotted line) compared with control conditions in the absence of metal ions (solid line). Upon the addition of 50 M Zn 2ϩ in the presence of 1 mM Ca 2ϩ , the spectrum is reversed to the level determined under metal-free conditions (dashed line).
FIG. 1. Biophysical characterization of recombinant human wild type S100A2. A, the elution profile from the final Sephacryl S100-HR gel filtration purification step revealed one major peak for S100A2. Insets, C, Coomassie-stained gel; W, Western blot probed with Biophysical Characterization of Cysteine-deficient S100A2 Mutants Ca 2ϩ binding and associated conformational changes are unaffected by the substitution of cysteines. Almost identical elution profiles were obtained for all S100A2 mutants on an S100HR gel filtration column. Purified recombinant S100A2 and mutant proteins and their reactivity with the monoclonal anti-S100A2 antibody is shown in Fig. 2A. Additional evidence for the structural integrity of all mutant proteins is the characteristic UV absorption spectrum that was identical for all mutants (Fig. 2B). Moreover, the virtually identical Ca 2ϩ binding characteristics indicated that none of the mutations influenced the dimeric state and/or EF-hand flexibility of the mutants compared with wild type S100A2. The half-maximal calcium saturation concentrations for all mutants were in the range of c(Ca) 50 ϭ 310 Ϯ 30 M (Fig. 2C). The value of the mutant lacking all four cysteine residues (⌬1234) showed the highest affinity for Ca 2ϩ (c 50 ϭ 280 M). This value may be taken as the estimated Ca 2ϩ affinity of the native S100A2 wild type protein in the cell, as it may be completely reduced in the cytoplasm of living cells. For freshly reduced wild type protein, similar values could be determined, but heavy oxidation was measured when monitoring the DTNB reactivity during Ca 2ϩ titration. After several hours of incubation in the Ca 2ϩ -free state (and shorter in the Ca 2ϩ -bound state) wild type S100A2 harbors less than one free cysteine per monomer, and the Ca 2ϩ affinity is in the range of 1 mM (data not shown). Therefore, all measurements were carried out immediately after complete reduction of the proteins in dithioerythritol (see "Materials and Methods"). The state of reduction was measured as a function of the DTNB reactivity. The different cysteines contribute to this oxidation in different amounts as documented by the differential sensitivity to the reaction with DTNB. For example, the enhanced reactivity of Cys 2 (22) upon Ca 2ϩ binding is most likely elicited through exposure of this residue to the solvent (see also Fig. 6).
We employed ESI-MS to test whether S100A2 or mutants oligomerize during Zn 2ϩ -binding and to calculate the population of S100A2 dimers (or oligomers) that are bound to one or more zinc ions at a certain concentration. ESI-MS has been used before to analyze the Ca 2ϩ binding properties of parvalbumin (26) and also to visualize the noncovalent dimers formed by S100A8 (27). Although recombinant S100A2 and its mutant proteins are present as dimers under conditions used for the other tests, we were unable to obtain significant amounts of the native dimeric protein in ESI-MS analysis. The formation of S100 dimers is driven almost exclusively by hydrophobic interactions of several highly conserved residues located within helix 4 and helix 1, respectively (28). Thus, the organic solvent used in the ESI-MS will disfavor the formation and stabilization of the dimer. However, we could show that the monomers of recombinant human S100A2, forming the major particle population under these conditions, are sufficient to bind to two Ca 2ϩ ions per monomer although with lower affinity, due to the low pH and the absence of cooperativity, mediated only by the intact dimer. No significant affinity for Zn 2ϩ could be detected for monomers.

Zn 2ϩ Binding and Associated Conformational Changes Are Impaired by Serial Substitution of Cysteines
Wild Type S100A2-The addition of Zn 2ϩ to the wild type S100A2 molecule led to a strong increase in tyrosine fluorescence. Precise determination of the binding characteristics was restricted due to gradual precipitation of wild type S100A2 before a clear end point of the titration was reached. The estimated half-maximal Zn 2ϩ saturation concentration in the absence of Ca 2ϩ was c(Zn) 50 ϭ 150 M. When Zn 2ϩ was added in the presence of 500 M Ca 2ϩ (70 -75% saturation according to the Ca 2ϩ titration), this value increased about 5-fold to c(Zn) 50 ca ϭ 30 M (compare Table I).  2. Mutant S100A2 proteins. A, Coomassie Blue-stained SDSpolyacrylamide gel (C) and corresponding Western blot (W) of S100A2 mutants used in this study probed with the monoclonal anti S100A2 antibody. B, UV absorption scan from 250 -300 nm was taken in buffer C1 at room temperature and used to determine purity and concentration of all S100A2 mutants. C, Ca 2ϩ titrations monitoring the increase of internal tyrosine fluorescence carried out with 0.25 mg/ml protein at 22°C in buffer C1 results in similar fluorescence characteristics for all mutants. wt, wild type.

Cys 1(3) Is Sufficient for the Aspect of Zn 2ϩ Binding That
Can Be Detected by Tyrosine Fluorescence Changes Cys 1(3) -Cysteine 1 is situated at position 3 in the amino acid sequence of human S100A2. All mutants containing Cys 1(3) (⌬2, ⌬3, ⌬4, ⌬234), exhibited high affinity Zn 2ϩ binding, which was associated with an increase in tyrosine fluorescence as a result of conformational changes in the molecule upon metal ion binding. When Cys 1(3) was replaced by serine, no Zn 2ϩ binding was detectable in all corresponding mutants (⌬1, ⌬12, ⌬123, ⌬124, ⌬134, ⌬1234). Fig. 3 shows two representative curves for this series of experiments. A typical curve indicating Zn 2ϩ binding associated with increased tyrosine fluorescence is demonstratd by the measurements for the ⌬2 mutant. The "Cys-null" mutant ⌬1234, by contrast, displays no Zn 2ϩ binding before the protein starts to precipitate. All mutants missing Cys 1(3) lacked specific Zn 2ϩ binding accompanied by conformational changes. In the absence of cysteines 2, 3, and 4, however, Zn 2ϩ binding was restored in the presence of Ca 2ϩ by the sole presence of the first cysteine only (mutant ⌬234) (Fig. 4). From these data, we conclude that Cys 1(3) is essential for the aspect of Zn 2ϩ binding that induces a conformational change leading to an increased tyrosine fluorescence and that this is strongly enhanced (or even induced) only after Ca 2ϩ binding of S100A2. A summary of all half-maximal concentrations for Ca 2ϩ and for Zn 2ϩ in the presence and absence of Ca 2ϩ for all mutants is given in Table I.
Cys 2(22) -Cys 2 (22) is situated within the helix-loop-helix motif of the first S100-specific EF-hand of S100A2 between the first helix and the calcium-binding loop. Substitution of Cys 2 (22) with serine (⌬2) did not cause any detectable changes in the Zn 2ϩ binding characteristics of this mutant compared with the wild type protein as determined in conformational titrations in the presence of Ca 2ϩ . But in this mutant, the affinity for Zn 2ϩ in the absence of Ca 2ϩ was remarkably increased (see also  Table I). Notably, mutant ⌬2 was also more resistant to oxidation and precipitation during the time course of the experiments than the wild type molecule or any of the mutants carrying Cys 2 (22) (⌬1, ⌬3, ⌬4). When only Cys 2 (22) was present (⌬134), no specific Zn 2ϩ -binding could be detected monitoring the tyrosine fluorescence both in the presence and absence of Ca 2ϩ (Fig. 4). However, in the radioactive 65 Zn 2ϩ binding studies we observed a signal at higher Zn 2ϩ concentrations (Fig. 5). With the S100A2 mutant ⌬134, we could also determine an enhanced reactivity to DTNB upon the addition of Ca 2ϩ , which was significantly reduced upon the addition of Zn 2ϩ (Fig. 6). Thus, this residue appears to become exposed to the solvent phase upon Ca 2ϩ binding, thereby increasing its sensitivity to oxidation.
FIG. 3. Zn 2؉ binding of recombinant S100A2 mutant proteins in the Ca 2؉ -free state. Zn 2ϩ binding curves were obtained from 0.1 mg/ml completely reduced protein in buffer C1 at 22°C. Two types of curves were observed. All mutant S100A2 proteins harboring Cys 1 (3) showed an increase of internal tyrosine fluorescence before unspecific Zn 2ϩ binding and precipitation of the sample occurred. As a representative example for all mutants binding to Zn 2ϩ , the mutant ⌬2 (gray circles) is shown, and for those exhibiting no binding in this assay the mutant ⌬1234 (black circles) is shown.
FIG. 4. Zn 2؉ binding in the Ca 2؉loaded state. Zn 2ϩ titrations monitoring the change of internal tyrosine fluorescence were carried out in the presence of 500 M Ca 2ϩ . 0.1 mg/ml completely reduced protein was titrated with Ca 2ϩ to 500 M. Then the titration was continued with Zn 2ϩ . The fluorescent signal contributed by the presence of Ca 2ϩ is shown by an exemplary Ca 2ϩ titration (dotted line). Three different effects were observed depending on the S100A2 mutant tested. Representative for all mutants harboring Cys 1(3) , the curve of mutant ⌬2 is shown (circles), which displays the characteristic 5-fold increase in c(Zn 2ϩ ) 50 compared with the measurement in the absence of Ca 2ϩ (compare Fig. 3). Representative for mutants lacking Cys 1(3) but still harboring Cys 4(94) , mutant ⌬1 is shown (triangles). No specific conformational effect upon Zn 2ϩ -binding can be observed. In all mutants lacking both Cys 1 (3) and Cys 4(94) , a decrease of the internal tyrosine fluorescence is observed. This type of effect is represented by the titration curve of mutant ⌬124 (rhomboids).

Cys 3(87) and Cys 4(94) Are Involved in the Formation of a Zn 2ϩ -binding Site That Does Not Alter the Overall Conformation of S100A2 upon Metal Binding
Cys 3(87) located at position 87 and Cys 4(94) at position 94 are placed C-terminal of the canonical second EF-hand motif of S100A2. Substitution of Cys 3(87) alone (⌬3) did not alter the Zn 2ϩ binding characteristics of S100A2, and the sole presence of Cys 3(87) (⌬124) led to low Zn 2ϩ binding (Fig. 5). Since the complementing mutant ⌬3 displayed wild type-like Zn 2ϩ binding characteristics in all tests, we conclude that the Zn 2ϩ binding capacity attributable to Cys 3(87) alone (displayed by mutant ⌬124 in the radioactive solid phase assay) is not relevant for specific Zn 2ϩ binding of the intact S100A2 molecule. By contrast, substitution of Cys 4(94) (⌬4) led to a clear decrease in the Zn 2ϩ -binding capacity as can be seen in the 65 Zn 2ϩ -binding assay (Fig. 5), whereas the conformational change induced by Zn 2ϩ in the presence or absence of Ca 2ϩ was not affected.
In mutants where both Cys 4(94) and Cys 1(3) are missing (⌬124, ⌬134, and to a lesser degree ⌬1234), a decrease of the tyrosine fluorescence could be observed when Zn 2ϩ was added in the presence of Ca 2ϩ (Fig. 4). Cys 4(94) counteracted this conformational effect when Zn 2ϩ was added to mutant proteins missing Cys 1(3) . This effect represents a conformational change that cannot be observed in the wild type S100A2 protein. In radioactive binding studies using the mutant containing Cys 4(94) only (⌬123), 65 Zn 2ϩ binding could be detected, but in fluorescent measurements this mutant showed no increased tyrosine fluorescence upon Zn 2ϩ binding, similar to all mutants lacking Cys 1(3) (see also above).
Based on sequence alignments, it has been proposed that a Zn 2ϩ finger-like structure with at least two possible ligands for Zn 2ϩ (CX n C) is present in the C terminus of several members of the S100 protein family and that this may represent a conserved motif for Zn 2ϩ binding in this family (10). To test the hypothesis that Cys 3(87) and Cys 4(94) , which are 7 residues apart act cooperatively in the ligation of a zinc ion (or form a zinc finger-like structure), an additional mutant, exclusively harboring these two cysteine residues (mutant ⌬12) was constructed. As determined in the radioactive 65 Zn 2ϩ binding studies, this mutant exhibited Zn 2ϩ binding greater than that seen for the mutant containing only Cys 4(94) (Fig. 5), but no confor-mational change was associated with this activity.
Cys Null Mutant-Finally, we designed a mutant in which all four cysteines were substituted for serine (⌬1234). No Zn 2ϩ binding could be detected in any of the two assays employed in this study. This result further supports the assumption that the highly conserved histidine in position 18, His 1 (18) , or His 2(40) alone is not sufficient for Zn 2ϩ binding.

Association of Endogenous and Recombinant S100A2 with Membranes
To test if the different conformational effects induced in S100A2 by the addition of Ca 2ϩ and Zn 2ϩ in vitro are also mirrored by altered subcellular compartmentalization, cell fractionation experiments were carried out monitoring the behavior of endogenous porcine S100A2, recombinant wild type, and the cysteine null mutant ⌬1234. Endogenous S100A2 from porcine LLC-PK1 cells was found in the microsomal fraction of porcine LLC-PK1 cells in a Ca 2ϩ -dependent manner. The protein was efficiently removed from this fraction by further addition of EGTA. The sole presence of 50 M Zn 2ϩ was not sufficient for significant targeting of S100A2 to membranes (Fig. 7A). These tests were also carried out with extracts of human A431 cells that contain only low detectable levels of endogenous S100A2. We selected these cells for the assay because a cell line negative for endogenous S100A2 could be devoid of the correct binding partners. Cell extracts were mixed FIG. 5. Solid phase 65 Zn 2؉ binding assay. X-ray films were analyzed densitometrically. The different capacities of the S100A2 mutants to bind 65 Zn 2ϩ were determined in each test by calculating the value for each column in relation to the wild type calibrated to 100%. The S.D. is taken from six assays in doublets (n ϭ 12). One representative x-ray film is shown below. All values were corrected by subtraction of the with recombinant human wild type S100A2 (Fig. 7B) or mutant ⌬1234 (Fig. 7C) and treated similarly to LLC-PK1 extracts. This assay revealed targeting of S100A2 to the microsomal fraction in a strictly Ca 2ϩ -dependent manner. The behavior of mutant ⌬1234 in this assay was indistinguishable from human recombinant wild type and endogenous porcine S100A2, demonstrating that Ca 2ϩ is the prime determinant for membrane association of S100A2 and that Zn 2ϩ is not involved in this mechanism.

DISCUSSION
In this study, we aimed at determining the Zn 2ϩ binding characteristics of human S100A2 and its dependence on Ca 2ϩ binding to the EF-hands. Detailed knowledge of these features of the protein is important for a thorough understanding of the potential involvement of this protein in signal transduction and transcriptional regulation or in events related to apoptosis. For the latter, several indications point toward a direct causal relation between programmed cell death and the Zn 2ϩ binding of S100 proteins. The role of Zn 2ϩ binding in the regulation of p53 by S100B is currently under debate (29 -32), and for S100A2 a correlation of its nuclear localization (16), Zn 2ϩ binding (4), and transcriptional regulation by p53 (15) has been documented. Since the sites responsible for Zn 2ϩ -binding in S100 proteins are still ill defined, we purposely selected S100A2 for this study because it harbors potentially important residues in critical positions, i.e. the very N and C terminus.
Human S100A2 is highly sensitive to oxidation due to the presence of four cysteines in each monomer, affecting the affinities for both Ca 2ϩ and Zn 2ϩ . The reduction of the Ca 2ϩ affinity in human S100A2 after oxidation is most likely a result of the formation of disulfide bonds impeding EF-hand flexibility. The individual cysteines contribute to this oxidation in different amounts as shown in the altered reactivities to DTNB in the presence or absence of metal ions (see also Fig. 6). Especially, residue Cys 2 (22) in the sequence of human S100A2 appears to be involved when oxidation affects the metal binding in vitro (see below).
Ca 2ϩ binding is mediated through the two EF-hand motifs and displays a high level of cooperativity (4). Ca 2ϩ -mediated conformational changes in S100 proteins have been studied by monitoring 4,4Јdianilino-1,1Ј-binaphtyl-5,5Ј-disulfonate-mediated or internal fluorescence (4,9) and by NMR-based studies (28,(33)(34)(35), and these changes revealed highly similar characteristics in several S100 family members. The increase of in-ternal tyrosine fluorescence upon Ca 2ϩ binding has been interpreted as a result of the exposition of a hydrophobic target interface in human S100A2 (4). In contrast to these findings, the conformational effects induced by the addition of Zn 2ϩ to human S100A2 (but also resulting in increased internal tyrosine fluorescence) are presumed to be distinct from those induced by Ca 2ϩ , since they can also be achieved when Zn 2ϩ is added in the presence of saturating amounts of Ca 2ϩ . In addition, CD spectra showed that changes in the ␣-helical content induced by Ca 2ϩ cannot be induced by the addition of Zn 2ϩ and, moreover, that the Ca 2ϩ -induced changes are reversed when Zn 2ϩ is added in the presence of Ca 2ϩ . Because the complete substitution of all cysteine residues for serines does not affect the Ca 2ϩ binding characteristics but completely abolishes Zn 2ϩ binding, we conclude that Ca 2ϩ and Zn 2ϩ binding are separate features employing different sites in the protein. These results also support the hypothesis that Ca 2ϩ -dependent targets of S100A2 may be different from Zn 2ϩ -dependent binding partners. From several studies addressing a similar question for other S100 family members (36 -39), it can be hypothesized that the Ca 2ϩ -dependent membrane association of S100A2 is mediated by members of the annexin family, whereas Zn 2ϩ -dependent targets are almost unknown except for one peptide that binds to S100B in a Zn 2ϩ -enhanced manner (30). It has been discussed earlier that Zn 2ϩ binding to S100 family members could be antagonistic to Ca 2ϩ binding at least in respect to target recognition (5,7,8). For human S100A2, it has been suggested that Ca 2ϩ and Zn 2ϩ may act antagonistically (4). In contrast, we here report a 5-fold increase in the affinity of S100A2 for Zn 2ϩ in the presence of Ca 2ϩ .
In this work, we were able to attribute the aspect of Zn 2ϩ binding resulting in profound conformational changes in human S100A2 specifically to the very N-terminal cysteine residue, tentatively named Cys 1(3) . This cysteine residue is both necessary and sufficient for all known features of this aspect of Zn 2ϩ binding including the induction of increased tyrosine fluorescence upon the addition of Zn 2ϩ and a 5-fold increase in Zn 2ϩ affinity in the presence of Ca 2ϩ . Additional support for the hypothesis that this Zn 2ϩ binding site employs only the first cysteine (Cys 1 (3) ) residue of the four tested stems from the solid phase 65 Zn 2ϩ -binding assay. The binding capacity for Zn 2ϩ in the ⌬1 mutant is reduced by 40%, and mutant ⌬234 retains about 40% of the Zn 2ϩ -binding of the wild type molecule.
These results of an N-terminal binding site are consistent with the findings of Kordowska et al. (9), who demonstrated that the Zn 2ϩ binding of S100A6 (calcyclin) induces a conformational change in the molecule, which can be monitored by internal tyrosine fluorescence and that Zn 2ϩ binding characteristics are altered upon derivatization of the single aminoterminal cysteine. Their results further implied a Zn 2ϩ -binding site that is distinct from the Ca 2ϩ -binding EF-hands.
The other ligands necessary for ligating the zinc ion remain elusive. However, assuming that the three-dimensional structure of the S100A2 dimer is closely related to other known S100 structures, one more putative ligand for this "N-terminal" Zn 2ϩ -binding site is the histidine residue His 2 (40) contributed by the other part of the S100A2 dimer. His 2 (40) is located in the hinge region between the two EF-hand motifs, a region previously inferred to be involved in Zn 2ϩ binding (10). Together with our data from ESI-MS studies, this suggests a zinc-binding site employing the entire intact S100 dimer, thus compensating for the lack of a conserved sequence motif for Zn 2ϩ binding in this protein family.
We also detected a second aspect of Zn 2ϩ binding in the very C terminus of recombinant human S100A2. We show that the FIG. 7. Ca 2؉ -but not Zn 2؉ -dependent association of S100A2 with membranes. The membranous microsomal fraction of cell extracts of porcine LLC-PK1 cells containing endogenous S100A2 (A) and a corresponding fraction of human A431 cells mixed with wild type recombinant human S100A2 (B) or the null mutant ⌬1234 (C) was pelleted in the presence of 1.5 mM Ca 2ϩ , 50 M Zn 2ϩ , or 2 mM EGTA. Note the presence of S100A2 in the membranous fraction also for the null mutant ⌬1234, which is unable to bind Zn 2ϩ . cysteine residues Cys 3(87) and Cys 4(94) of recombinant S100A2 form at least part of a second Zn 2ϩ -binding site in S100A2 and that they contribute to this binding site in different amounts. The sole presence of Cys 4(94) (⌬123) was sufficient for the binding of radioactive 65 Zn 2ϩ ; however, with lower affinity than was seen in the additional presence of Cys 3(86) (⌬12). Also, the sum of the Zn 2ϩ bound by the mutants ⌬124 and ⌬123 was always lower than the amount of Zn 2ϩ bound by mutant ⌬12 harboring both residues, pointing toward a synergism between these two cysteine residues. This assumption is also supported by the fact that the reduction of zinc binding in mutant ⌬4 is significantly higher than the binding capacity retained by the mutant ⌬123 (compare with Fig. 5).
Stable ligation of a zinc ion requires a minimum of four ligands (three amino acid residues and one solvent molecule) (40). Thus, these two cysteines are not sufficient for the complete ligation of Zn 2ϩ . From the crystal structures that were obtained from Zn 2ϩ -loaded S100A7 (psoriasin) (11), it may be concluded that other N-terminal residues like His 1 (18) act in concert on the ligation of the metal ion, since the C terminus of one monomer is in close enough proximity to the first helix of the first EF-hand motif of the second monomer in the dimer to interact with each other. Based on the model of Brodersen et al. (11), residue His 1 (18) in S100A2 would correspond to His (17) in S100A7 in one half of the dimer, and residues Cys 3(87) and Cys 4(94) of S100A2 would correspond to His (86) and His (90) of S100A7 in the other half of the dimer, respectively. This hypothesis is further supported by the finding that under assay conditions that affect the integrity of the dimer like in ESI-MS, Zn 2ϩ -binding is abolished. Thus, Zn 2ϩ -binding in S100 family proteins appears to depend on the formation of an intact dimer.
In human S100A2, all cysteines except for one are located outside the EF-hands and have no effects on the Ca 2ϩ binding characteristics. Only Cys 2 (22) is positioned next to the first Ca 2ϩ -binding loop and could be potentially involved in Ca 2ϩ binding. Cys 2 (22) is situated 4 amino acid residues (one complete helical turn) C-terminal of the highly conserved histidine residue (His 1 (18) ) that has been suggested to be involved in the Zn 2ϩ binding of S100 proteins (see above) (11). As a result, both residues are in close proximity to each other, pointing in the same direction at the beginning of the flexible Ca 2ϩ -binding loop. Our solid phase 65 Zn 2ϩ -binding studies revealed that the mutant ⌬134 harboring Cys 2 (22) is capable of Zn 2ϩ -binding in vitro. As shown in the DTNB assay, this residue becomes exposed to the surface upon Ca 2ϩ binding (compare Fig. 6), and ligation with Zn 2ϩ does not interfere with Ca 2ϩ binding. On the other hand, substitution of this residue does not alter the Zn 2ϩ binding characteristics or affinity of recombinant human S100A2. Moreover, this residue is conserved neither among members of the S100 family nor among species. At the position corresponding to Cys 2 (22) in human S100A2, all other known sequences of S100A2 and most other S100 family members contain a highly conserved glycine. From this, we conclude that Cys 2 (22) is not involved in Zn 2ϩ binding or plays a highly subordinate role.
Substitution of Cys 4(94) in mutants incapable of Zn 2ϩ binding via Cys 1(3) (⌬124, ⌬134, ⌬1234) leads to conformational changes different from those seen for the wild type protein (i.e. a decrease of tyrosine fluorescence). We conclude that Cys 4(94) plays an important role in the maintenance of the structural integrity of the protein during Zn 2ϩ binding. One possible explanation may include the prevention of the Zn 2ϩ ion from binding to Cys 2 (22) by Cys 4(94) . In conclusion, Cys 4(94) appears essential for the high affinity Zn 2ϩ binding in S100A2 not associated with increased internal tyrosine fluorescence and acting synergistically with Cys 3(87) . However, the precise role of Cys 4(94) in zinc binding leading to increased tyrosine fluorescence remains unclear.
The S100A2 mutant in which all cysteines have been converted to serine did not show detectable zinc binding in any of the assays employed in this study. However, incubation with high concentrations of ZnCl 2 (Ͼ0.5 mM) leads to precipitation of the sample solution. The inability of this null mutant (⌬1234) to bind to Zn 2ϩ may also be interpreted as a control for His 1(18) and His 2 (40) , which are not sufficient for zinc binding even in concert with additional potential ligands (like an aspartate) as has been proposed for S100A7.
Finally, our cell extraction experiments have shown that the different metal binding features in human S100A2 are also mirrored by their metal ion-dependent localization in different cellular compartments. A similar mechanism has been shown for S100A6 (41,42). Membrane association of S100A2 is strictly Ca 2ϩ -dependent and can be hypothesized to be mediated by members of the annexin family, since several S100 proteins interact with annexins in a Ca 2ϩ -dependent manner. Thus, Zn 2ϩ -dependent targets of S100A2 may be different from Ca 2ϩdependent binding partners, since Zn 2ϩ is unable to translocate S100A2 to membranous cellular compartments and as the null mutant ⌬1234 displays Ca 2ϩ -dependent membrane localization.
Conclusions-Zn 2ϩ binding to the "N-terminal" site is closely linked to conformational alterations in the dimeric S100A2 molecule, and the affinity for Zn 2ϩ increases in response to Ca 2ϩ although the accessibility of the cysteine residue Cys 1(3) itself may not be altered upon Ca 2ϩ binding. Since all S100 family members have been shown to form dimers in solution and all three-dimensional structures determined so far reveal a highly similar molecular arrangement, it appears likely that those residues unaccessible in one S100A2 monomer to form a complete Zn 2ϩ -binding site are situated in (and contributed by) the other half of the dimer. The second "C-terminal" Zn 2ϩbinding site employs Cys 3(87) and Cys 4(94) from one monomer and most likely also His 1 (18) of the other half of the dimer. In contrast to the amino-terminal site, Zn 2ϩ binding at the carboxyl terminus is not associated with conformational alterations and appears to be independent of the Ca 2ϩ -associated properties of human S100A2. Histidine and cysteine residues can be found in most S100 family members in corresponding or neighboring positions, as is the highly conserved His 1 (18) or residues in the hinge region and C-terminal histidine and cysteine residues. We thus hypothesize that two separate regions in the S100A2 monomer are responsible for two distinct types of Zn 2ϩ binding in vitro but that only the intact dimer is capable of exhibiting full Zn 2ϩ binding activity and displays the Zn 2ϩ related functional alterations.