Purification and cation binding properties of the recombinant human S100 calcium-binding protein A3, an EF-hand motif protein with high affinity for zinc.

The calcium-binding protein S100A3 is an unusual member of the S100 family, characterized by its very high content of Cys. In order to study the biochemical, cation-binding, and conformational properties, we produced and purified the recombinant human protein in Escherichia coli. The recombinant protein forms noncovalent homodimers, tetramers, and polymers in vitro with a subunit molecular weight of 11,712. The Zn2+-binding parameters of S100A3 were studied by equilibrium gel filtration and yielded a stoichiometry of four Zn2+ per monomer with a [Zn2+]0.5 of 11 μM and a Hill coefficient of 1.4 at physiological ionic strength. The affinity for Ca2+ is too low to be determined by direct methods (KCa > 10 mM). Ca2+- and Zn2+-binding can be followed by optical methods: the Trp-45 fluorescence is high in the metal-free form and addition of Zn2+ and Ca2+, but not of Mg2+, leads to a 4-fold quenching. Ca2+ and Zn2+ promote also quite similar conformational changes in the Tyr and Trp environment as monitored by difference spectrophotometry. Fluorescence titrations with Zn2+ confirmed that there is one set of high affinity binding sites with a [Zn2+]0.5 of 8 μM and a Hill coefficient of 1.3. Binding of Zn2+ to a second set of low affinity sites induces protein precipitation. Fluorescence titrations with Ca2+ confirmed the very low affinity of S100A3 for this ion with a [Ca2+]0.5 of 30 mM and slight negative cooperativity. Mg2+ has no effect on this binding curve. Of the 10 Cys residues in S100A3, 5 only are free thiols, and accessible to 5,5′-dithiobis(2-nitrobenzoic acid); they display a high reactivity in the metal-free and Ca2+ form, but a 20-fold lowered reactivity in the Zn2+ form of S100A3. Ca2+-binding promotes the formation of a solvent-accessible hydrophobic surface as monitored by the 60-fold fluorescence increase of 2-p-toluidinylnaphthalene-6-sulfonate, whereas Zn2+ has no noticeable influence. Our data indicate that Ca2+ and Zn2+ do not bind to the same sites and that under physiological conditions S100A3 is a Zn2+-binding rather than a Ca2+-binding protein; nevertheless, very specific conformational changes are introduced by either Ca2+ or Zn2+. Since no Zn2+-binding motif of known structure was identified in the primary sequence of S100A3, the results are suggestive for a novel Zn2+-binding motif.

The S100 protein family constitutes a subgroup of Ca 2ϩbinding proteins of the EF-hand type displaying 30% or more sequence identity (Kligman and Hilt, 1988;Hilt and Kligman, 1991). Under physiological conditions their affinity for Ca 2ϩ is rather low but can be increased once S100 proteins are associated with their targets. Different S100 proteins were also found to bind Zn 2ϩ with a fairly high affinity (Baudier et al., 1986;Leung et al., 1987;Filipek et al., 1990;Dell'Angelica et al., 1994). Both intracellular roles, such as activation of enzymes, regulation of motility, and smooth muscle contraction, and extracellular roles, such as neuronal differentiation, glial proliferation, and prolactin secretion (for review, see Donato (1991), Zimmer and Dubuisson (1993), and Heizmann and Braun (1995)), have been proposed. Intriguingly, in different cases where calmodulin was thought to be the regulatory CaBP, 1 S100 proteins were finally the real activators (Bianchi et al., 1993). Most S100 proteins interact in vitro with hydrophobic matrices, with membranes, enzymes, cytoskeletal and contractile proteins, and even cell surface receptors (for review, see Donato (1991)). All of these data point to a multifunctional role of the S100 family with a particular function for each of its members. This functional specificity is supported by the fact that their expression is differentially deregulated in different types of cancer cells (Hilt and Kligman, 1991;Weterman et al., 1992;Davis et al., 1993;Pedrocchi et al., 1994aPedrocchi et al., , 1994b, suggesting participation in tumor progression. However, for none of these putative functions have the molecular details been elucidated. The protein S100A3, 2 formerly called S100E, was recognized for the first time as the product of one of the tightest gene clusters discovered in the human genome located on chromosome 1q21 (Engelkamp et al., 1993). The S100A3 gene shows a low but general transcription level in diaphragm, heart, skeletal muscle, stomach, lung, liver, fat tissue, and placenta. A YAC clone from human chromosome 1q21 has been recently isolated on which nine different genes coding for S100 proteins were localized. The clustered organization of S100 genes in the 1q21 region allowed to introduce a new logical nomenclature for these genes (Schä fer et al., 1995). The S100A3 gene product is 101 residues long and possesses one S100-type noncanonical Ca 2ϩ -binding loop of 14 residues expanding from Ala-20 to Glu-33, and one canonical EF-hand loop of 12 residues from Asp-63 to Glu-74, both flanked by two ␣-helices. In calbindin D-9k, the prototype of this S100 protein family with a resolved three-dimensional structure (Szebenyi and Moffat, 1986;Carlström and Chazin, 1993), the ␣-helices are oriented in an antiparallel fashion, thus forming a 4-helix barrel. Within the S100 subfamily S100A3 is unique for the exceptionally high number of Cys residues. Despite the Cys frequency, S100A3 does not display the classical zinc-binding motifs seen in metallothioneins (Vallee and Auld, 1990), DNA-binding proteins (Pérez-Alvarado et al., 1994), or protein kinase C (Hommel et al., 1994).
In order to begin to understand the role of S100A3 and the molecular mechanisms by which it exerts its function, we characterized in this study the Ca 2ϩ -and Zn 2ϩ -binding properties of recombinant human S100A3 under physiological conditions. We monitored the cation-dependent changes in the environment of the Trp and Tyr residues, probed the thiol/disulfide state and the cation-dependent reactivity of the thiols, and finally monitored the solvent-exposed hydrophobic surface. The results suggest that under physiological conditions S100A3 is a Zn 2ϩ -binding rather than a Ca 2ϩ -binding protein.

EXPERIMENTAL PROCEDURES
Materials-A protein fusion and expression system was obtained from New England Biolabs. Isopropyl-thio-␤-D-galactopyranoside and restriction endonucleases were from Boehringer Mannheim. Concentrated T4 DNA ligase was obtained from New England Biolabs. EDTA, ampicillin, and lysozyme were purchased from Fluka. Hydroxylapatite (Bio-Gel HTP), polyacrylamide, and electrophoresis equipment were from Bio-Rad.
Cloning of Human S100A3 into a Prokaryotic Expression System-Human cDNA of S100A3 was amplified by the polymerase chain reaction (PCR) as described earlier (Engelkamp et al., 1993), using the primers S100A3-M and S100A3-B. The resulting PCR product comprised the complete coding region of S100A3 beginning with the starting codon ATG. The PCR product was blunt end ligated into the XmnIdigested expression vector pMal-c2 downstream the malE gene, which encodes maltose-binding protein (MBP). Colonies of transformed Escherichia coli TB1 strains expressing MBP-S100A3 fusion protein were analyzed for their correct DNA sequence inserted into the expression vector by the dideoxynucleotide chain termination method.
Expression and Purification of Recombinant S100A3-Luria-Bertani medium (1000 ml) containing 0.2% glucose and 100 mg/liter ampicillin was inoculated with a 10-ml overnight culture of an E. coli TB1 clone expressing MBP-S100A3 fusion protein. At an A 600 of 0.5, 1 mM isopropyl-thio-␤-D-galactopyranoside was added to induce the expression of the fusion protein. The culture was further incubated for 3 h. Cells were harvested by centrifugation at 4000 ϫ g (20 min, 4°C), resuspended in a 50-ml cold column buffer (20 mM Tris/HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA, 1 mM NaN 3 ) and frozen at Ϫ20°C overnight. Cells were lyzed by thawing, incubation with 1 mg/ml lysozyme for 20 min on ice, and sonication for 1.5 min (on ice). This suspension was centrifuged at 9000 ϫ g (30 min, 4°C). The crude extract (supernatant) was diluted to 2.5 mg/ml and applied to an amylose resin affinity column (length, 8 cm; diameter, 2.5 cm). The fusion protein was eluted as described by the manufacturer's manual (New England Biolabs) and incubated with 0.2% (by mass) protease factor Xa for 2-4 days at room temperature. The protease cleavage site is located exactly in front of the aminoterminal methionine of S100A3. The completeness of the cleavage was controlled by SDS-Tricine-PAGE (Fig. 1). S100A3 was separated from MBP by ion exchange chromatography followed by a second amylose resin affinity chromatography. Purity and concentration of recombinant S100A3 was controlled by SDS-Tricine-PAGE ( Fig. 1) and amino acid analysis.
One-and Two-dimensional PAGE and Molecular Weight Determination-15% SDS-Tricine gels were cast and run according to the method described by Schä gger and von Jagow (1987). For two-dimensional PAGE, the precast Immobiline gels with an immobilized pH gradient from pH 3 to 10.5 (Promega) were used under denaturing conditions according to the producer's instructions.
The native apparent molecular weight of the metal-free, Ca 2ϩ and Zn 2ϩ forms of S100A3 was determined by gel filtration on a 1 ϫ 70-cm column of Sephadex G-75 in 50 mM Tris buffer, pH 7.5, 150 mM KCl, 1 mM dithiothreitol (buffer A) containing either no divalent cations, 100 mM CaCl 2 , or 100 M ZnCl 2 . The column was standardized with the calibration mixture of Bio-Rad.
Mass Spectrometry-Electrospray ionization mass spectra were obtained with a Sciex Api III and a Finnigan TSQ 700 instrument equipped with an ion-spray source. The protein molecular mass was determined from the acquired spectra with ESI deconvolution software from Finnigan.
Amino Acid Analysis-Amino acid analysis was performed by gasphase HCl hydrolysis, conversion with dansyl chloride and subsequent evaluation of the derivatized amino acid products with a Beckman System Gold HPLC instrument.
Direct Zn 2ϩ -binding Studies-For removal of contaminating metal ions, S100A3 was precipitated with 3% trichloroacetic acid and then passed through a 1 ϫ 40 cm Sephadex G-25 column equilibrated in the assay buffer. The protein concentration was determined from the UV absorption spectrum using a molar extinction coefficient at 280 nm of 14,500 M Ϫ1 cm Ϫ1 for metal-free S100A3. These values were measured on protein stock solutions in bidistilled water whose concentrations were determined by quantitative amino acid analyses.
Zn 2ϩ binding was measured at room temperature by the equilibrium gel filtration method of Hummel and Dryer (1962). A Sephadex G-25 column (0.7 ϫ 50 cm) was equilibrated in buffer A containing variable concentrations of Zn 2ϩ . 0.5-1 ml of 50 -200 M metal-free protein was applied to the column. In the eluant Zn 2ϩ concentrations were determined by atomic absorption with a Perkin-Elmer 2380 atomic absorption spectrophotometer. For the atomic absorption measurements EDTA up to 1 mM was added to all solutions, including the standards (Titrisol, Merck). Protein concentrations were measured by ultraviolet absorption.
Optical Methods to Probe the Environment of Aromatic Residues-Emission fluorescence spectra were taken with a Perkin-Elmer LS-5B spectrofluorimeter. The measurements were carried out on 1 M trichloroacetic acid-treated metal-free S100A3 at room temperature with excitation and emission slits of 5 nm. The excitation wavelength was 280 nm. 100 mM CaCl 2 , 30 mM MgCl 2 , or 200 M ZnCl 2 was added to obtain the metal-free, Ca 2ϩ , Mg 2ϩ , and Zn 2ϩ forms, respectively. The denatured form was obtained by addition of 4 M guanidine-HCl. Ca 2ϩ and Zn 2ϩ titrations were done with 2 M S100A3 in buffer A.
UV absorption spectra and difference spectra were measured with a Perkin-Elmer Corporation 5 UV/VIS spectrophotometer. Difference spectra were taken at room temperature on solutions with an optical density at 280 nm close to 1.
Protein Reduction and Thiol Reactivity-The influence of cations on the thiol reactivity was assayed on S100A3 samples that were previously reduced by overnight incubation with 100 mM DTT at pH 8.5 and chromatographed on a Sephadex G-25 column (0.7 ϫ 35 cm) equilibrated in nitrogen-saturated buffer A. The thiol reactivity was assayed by monitoring spectrophotometrically at 412 nm the kinetics of the reduction of Ellman's reagent according to Riddles et al. (1983). The reaction was initiated upon mixing the protein solution with 10 l of DTNB to a final concentration of 0.3 mM. Titration of Exposed Hydrophobicity-The Ca 2ϩ -and Mg 2ϩ -dependent changes in hydrophobic matrices of S100A3 was followed by monitoring the fluorescence properties of 2-p-toluidinylnaphthalene-6-sulfonate (TNS) as described by FIG. 1. Expression and purification of recombinant human S100A3. Coomassie Blue-stained 15% SDS-Tricine-PAGE under reducing conditions, showing induction and purification steps of MBP-S100A3 fusion protein and of S100A3. Lane 1, 40 g of crude extract of E. coli; lane 2, 40 g of flow-through following loading onto amyloseresin column; lane 3, 5 g of eluate from amylose-resin column: MBP-S100A3 fusion protein; lane 4, 25 g of fusion protein after factor Xa cleavage; lane 5, 3 g of MBP; lane 6, 2.5 g of finally purified S100A3 (monomer). McClure and Edelman (1966). Solutions of 10 M metal-free S100A3 and 0.5 M TNS were exited at 328 nm and the emission spectrum recorded with slits of 10 nm.

RESULTS
Expression and Purification of Recombinant S100A3-To express and isolate S100A3 in large amounts, human S100A3 cDNA was cloned into the prokaryotic expression vector pMal-c2. A PCR product of human S100A3 cDNA was introduced at the protease factor Xa cleavage site behind the malE gene of pMal-c2 to generate a MBP-S100A3 fusion construct. The amount of expressed MBP-S100A3 fusion protein corresponded to about 12% of total cell protein of a bacterial culture. After isolating the fusion protein by amylose resin affinity chromatography S100A3 was cleaved off from MBP by the protease factor Xa. S100A3 was finally purified by ion exchange chromatography and a second amylose resin affinity chromatography. The correct cleavage of the fusion protein and the purity and concentration of S100A3 was controlled by amino acid analysis and SDS gel electrophoresis. Fig. 1 shows the expression and purification of the fusion protein and of S100A3.
Biochemical Properties of Human Recombinant S100A-After classical treatments of S100A3, such as dialysis, ultrafiltration, and freezing, the S100A3 protein is partly insoluble, but dissolves quickly and completely when solid DTT up to 50 mM is added. SDS-PAGE clearly shows that different disulfidelinked oligomers are formed, which are reduced to the monomer after DTT treatment (data not shown).
To verify recombinant S100A3 for correct synthesis in bacteria we determined its exact mass by electrospray ionization mass spectrometry. Before desalting with butyl-300 microbore reversed-phase HPLC it was again necessary to mix the protein probe with 50 mM DTT to prevent precipitation on the column and to obtain any mass signal. In acidic solvent a molecular weight of 11,712 Ϯ 1.7 was obtained, which is in good agreement with the calculated molecular weight, including the amino-terminal methionine, of 11,713.3 (Fig. 2). This result shows the correct expression of S100A3 in E. coli TB1, including an unprocessed amino-terminal methionine.
SDS-PAGE after reduction of S100A3 with 10 mM DTT for 30 min at 37°C also yielded a band with a molecular mass of 11 kDa. Determination of the apparent molecular mass by gel filtration on Sephadex G-75 after thorough reduction yielded different values depending on the presence of divalent cations: 22.4 kDa in the absence of divalent cations, 24.9 kDa in the presence of 100 M Zn 2ϩ , and 39.0 kDa in the presence of 100 mM Ca 2ϩ . Thus, as in other members of the S100 family, S100A3 forms a noncovalent homodimer. Moreover, 100 mM Ca 2ϩ promotes formation of a higher order oligomer (likely tetramers). It is not clear if the latter phenomenon is to be attributed to the specific binding of Ca 2ϩ or to an ionic strength effect, which is known to stabilize hydrophobic interactions.
We assessed the isoelectric point of S100A3 by two-dimensional gel electrophoresis under reducing conditions using an immobilized pH gradient ranging from pH 3.5 to 10. In contrast to the calculated pI of 4.53, the determined pI of the denatured protein was found to be 5.5 (data not shown). This divergence may be caused by the experimental conditions and not by any modifications of the protein as the measured mass of recombinant S100A3 was found to be identical to the calculated value.
Direct Cation Binding Studies-Trichloroacetic acid can be used to remove ions, concentrate the protein, and prepare the sample correctly for Hummel-Dryer experiments. Gel filtration on Sephadex G-25 in 350 M free Ca 2ϩ shows very little binding (Ͻ0.2 mol of Ca 2ϩ /mol of S100A3), indicating that a binding study of this cation cannot be carried out by direct means. Zn 2ϩ binding by the Hummel-Dryer method (Fig. 3) yields an isotherm of which the maximum value is somewhat difficult to evaluate since the protein shows a tendency to aggregate above 100 M free Zn 2ϩ . The Scatchard plot, although curved upward indicating positive cooperativity, allowed a rather precise extrapolation to four binding sites per monomer (not shown). Assuming a maximal binding of 4 Zn 2ϩ per monomer, the Hill plot was calculated and yielded a Hill coefficient (n H ) of 1.4 and a [Zn 2ϩ ] 0.5 of 11 M (inset). A Hummel-Dryer experiment in the presence of both Ca 2ϩ and Zn 2ϩ indicated that there is neither competition between the cations nor reenforcement of affinity, as was described for S100A1 (Leung et al., 1987) or calgranulin C (Dell' Angelica et al., 1994).
Fluorescence Characteristics-Fluorescence spectra (Fig. 4) indicate that denatured S100A3 has a fluorescence maximum 10-fold lower than that of the metal-free form. In the latter form and in the presence of Mg 2ϩ the Trp is very well shielded with max at 340 nm. Mg 2ϩ does not affect the spectrum, suggesting that no binding occurs. Ca 2ϩ , Zn 2ϩ , or Co 2ϩ binding leads to a 3-fold fluorescence decrease, in the case of Ca 2ϩ with a 7-nm blue shift, in the case of Zn 2ϩ or Co 2ϩ without any shift. Given the very good signal change, Ca 2ϩ , Zn 2ϩ , and Co 2ϩ titrations could be carried out by fluorimetry on a 2 M solution of S100A3. Since S100A3 displays a very low affinity for Ca 2ϩ , the ratio of bound to added Ca 2ϩ is negligible. For Ca 2ϩ a smoothly increasing sigmoid was observed (Fig. 5A), indicating more than one site with different affinities or displaying negative cooperativity. This isotherm can be analyzed with [Ca 2ϩ ] 0.5 of 35 mM and n H equal to 0.76 (Fig. 5B). The same isotherm was obtained in 30 mM Mg 2ϩ , indicating that the sites are specific. For Zn 2ϩ there are two levels of signal change (Fig.  5A), one with a midpoint of 8 M and one at 600 M. The high affinity compartment, with four Zn 2ϩ binding sites (see above), shows pronounced positive cooperativity with n H equal 1.3 as analyzed in Fig. 5C. Zn 2ϩ binding to the lower affinity compartment (at 200-1000 M free Zn 2ϩ ) induces protein aggregation, as evidenced from strongly increased signals at emission wavelengths (285-290) close to the excitation wavelength (280 nm). Addition of both Ca 2ϩ and Zn 2ϩ leads to much stronger protein precipitation. Co 2ϩ binding seems monophasic with a K D of 1 mM (data not shown). Apparently this Co 2ϩ -binding compartment corresponds to the high affinity compartment of Zn 2ϩ . Difference spectrophotometry on the S100A3 Co 2ϩ complex in the 240 -800 nm zone shows that the complex does not display the peaks in the 650 -750 zone, which are so characteristic for classical zinc fingers. Thus, the absence of these bands and the comparable low affinities for Zn 2ϩ and Co 2ϩ (nM for Zn 2ϩ and M for Co 2ϩ in zinc fingers) suggest that in S100A3 there is no such motif. Difference Spectrophotometry-Difference spectra (Fig. 6) yielded rather small signal changes (10 -20% of what is usually found in CaBPs), which could be essentially attributed to changes in the Tyr, to a lesser extent also to the single Trp environment. Qualitatively the spectra resemble very much those of the calmodulin-like protein (Durussel et al., 1993), with the appearance of negative peaks at 279 and 289 nm. They point to the shift of Tyr from a hydrophobic to a polar environment. The Phe environment seems not to be sensitive to cation binding. The difference spectra of the Zn 2ϩ and Ca 2ϩ forms are quite similar, although only Ca 2ϩ binding provoked the broad negative peak, likely due to a change in the polarity of the Trp environment (Ilich et al., 1988). Zn 2ϩ concentrations above 100 M lead to progressive precipitation of the protein, resulting in a declining base line. Strongly enhanced precipitation is observed when millimolar concentrations of Ca 2ϩ were added to these Zn 2ϩ -containing samples.
Changes in the Thiol Reactivity Induced by Zn 2ϩ -Thiol titration with DTNB on a freshly reduced (250 mM DTT for 24 h at room temperature, followed by Sephadex G-25 chromatography) sample yielded 4.5-4.9 free thiols per protein monomer. Addition of 4 M guanidine hydrochloride increased these values to 5.0 and 5.3. This, together with the results of SDS-PAGE and mass spectrometry, indicates that S100A3 contains two to three intrapolypeptide disulfide bridges. The reactivity of the free thiols in metal-free and Ca 2ϩ form of S100A3 (Fig. 7) is similar and too high for measurement by classical means. Zn 2ϩ binding decreases the reactivity of these thiols by a factor of 20 and here again the presence of Ca 2ϩ does not noticeably affect the reactivity. The kinetic profiles in the presence of Zn 2ϩ do not obey to a Guggenheim equation for a pseudo-first-order rate reaction (Durussel et al., 1993). Either the 5 thiols have from the beginning a different rate constant, or the progressive blocking of thiols by bulky DTNB leads to steric hindrance. The presence of Ca 2ϩ does in no case modify the kinetics suggesting that the Zn 2ϩ -sensitive thiols are located on the opposite end of the EF-hands.
Interaction with the Hydrophobic Probe TNS-Upon binding of Ca 2ϩ , CaBPs of the activator type, such as calmodulin (Tanaka and Hidaka, 1980) and neuron-specific CaBPs (Cox et al., 1994), usually display one or more solvent-exposed hydrophobic patches on their surface, which can be monitored with particular fluorescent probes such as TNS. Fig. 8 shows that Ca 2ϩ binding induces a 30-fold increase in fluorescence enhancement, whereas no enhancement at all is observed upon binding of Zn 2ϩ . It should be noted that the development of the Ca 2ϩ -induced hydrophobic patch(es) occurs in a biphasic manner: one phase is very rapid (occurs within the time of mixing) and may well present the exposure of hydrophobic residues in each monomer; the second phase occurs over a range of 10s of minutes and may correspond to the transition of the dimer to the 38-kDa oligomer as shown by gel filtration. It should be noted that addition of even 220 mM Ca 2ϩ does not lead to noticeable aggregation, as monitored by turbidimetry (not shown). But in the presence of both 55 mM Ca 2ϩ and 95 M Zn 2ϩ the enhancement is half of that of Ca 2ϩ alone, indicating that the binding is noncompetitive. Very similar results have been obtained with the fluorescent probe 1-anilinonaphtalene-8-sulfonate (data not shown). FIG. 8. Hydrophobic exposure in S100A3 as monitored by the fluorescence of TNS after excitation at 328 nm. Protein and TNS concentrations were 5 and 0.5 M, respectively. TNS alone (thin solid line); metal-free S100A3 (⅐⅐⅐⅐⅐); S100A3 ϩ 180 mM Ca 2ϩ (OO); S100A3 ϩ 190 M Zn 2ϩ (-----); S100A3 ϩ 180 mM Ca 2ϩ ϩ 86.5 M Zn 2ϩ (-⅐-⅐-⅐-⅐); S100A3 ϩ 190 M Zn 2ϩ ϩ 90 mM Ca 2ϩ (thin dotted line).

DISCUSSION
In this study we report the biochemical characterization and cation-binding properties of S100A3, a new member of the S100 family with an unusually high content of Cys residues. The protein is a dimer and contains two EF-hand motifs per monomer. But, whereas most other S100 proteins display Ca 2ϩ dissociation constants of 0.1-1 mM, S100A3 is able to bind Ca 2ϩ only in the 10 -100 mM free Ca 2ϩ range, i.e. very far from the cytosolic Ca 2ϩ levels. Nevertheless, this binding seems specific since it is accompanied by Tyr and Trp conformational changes very similar to those caused by Zn 2ϩ binding, by a well defined exposure of hydrophobicity and an oligomerization. The reason for this low affinity for Ca 2ϩ is not clear, since its primary structure is quite classical for a S100 member. However, our data indicate that each dimer contains five disulfide bridges. This may stabilize the protein but can impose strong constraints for the efficient binding of Ca 2ϩ . Reduction of all the disulfide bridges of S100A3 under denaturing conditions and alkylation of the thiols yields a protein product which binds Ca 2ϩ with a dissociation constant of 0.8 mM, 3 i.e. an affinity close to that of most other S100 proteins. It is still possible that S100A3 displays a real Ca 2ϩ -dependent function when associated with its target or when secreted in the Ca 2ϩ -rich extracellular fluid. In contrast to calgranulin C (Dell' Angelica et al., 1994) and S100B (Baudier et al., 1986), there are no indications that Zn 2ϩ increases the affinity of S100A3 for Ca 2ϩ . But an interesting dynamic regulation may occur through zinc binding, since the affinity is rather high (K D ϳ10 M). It is estimated that 99% of the 36 mg of Zn 2ϩ per kg human wet weight is intracellular and 25% of this amount is not, or loosely bound (Vallee and Falchuk, 1993). This represents intracellular free Zn 2ϩ concentrations of 40 -400 M, depending on the tissue, with a maximum of 2 mM in the retina. It is thus very likely that in vivo S100A3 is mostly in Zn 2ϩ -bound form. Its precise role in the direct activation of response systems and/or in promotion of exchange with other important Zn 2ϩ -regulated proteins must still be evaluated.
Where are the four Zn 2ϩ sites per monomer, or rather the eight Zn 2ϩ sites per native dimer located and what kind of sequence motifs in S100A3 could be responsible for Zn 2ϩ binding? Since the three-dimensional structure of most S100 proteins has not been elucidated, one can only compare with Zn 2ϩbinding motifs in proteins where the ligands responsible for binding have been identified. Zn 2ϩ sites are either of the tetradentate or tridentate type (Vallee and Auld, 1990). Tetradentate sites are found in small Zn 2ϩ -binding domains (zinc fingers) in which the cation is strongly held by four ligands composed of Cys and/or His (Schabe and Klug, 1994). These motifs have affinities of the order of 10 12 M Ϫ1 or more (Zeng et al., 1991). The tridentate type of site, found in different extracellular (Vallee and Auld, 1990) and intracellular enzymes (Perlman and Rosner, 1994), bind Zn 2ϩ with an affinity constant of 2 ϫ 10 6 M Ϫ1 (Francis et al., 1994). In these enzymes Zn 2ϩ is held by three ligands: two His residues in a typical H-E-x 2 -H or H-x 3 -H sequence implanted on a ␣-helical segment and a third coordinating Glu residue, located at a variable distance, up-or downstream, of the His motif (Vallee and Auld, 1990). S100A8, S100A9, and calgranulin C display such a motif and bind Zn 2ϩ with an affinity of at least 10 8 M Ϫ1 . The single Zn 2ϩ site is clearly distinct from the two Ca 2ϩ -binding sites. However, other S100 proteins do not possess one of the motifs frequently encountered in proteins where Zn 2ϩ binding has been proven to be functionally important. Strong binding of 8 Zn 2ϩ ions per dimer was also reported for S100B (Baudier et al., 1986). S100A1 (Leung et al., 1987) binds Zn 2ϩ with a rather low affinity. S100A6 (calcyclin) binds Zn 2ϩ with a [Zn 2ϩ ] 0.5 of about 2 mM (Filipek et al., 1990;Pedrocchi et al., 1994b). But S100A3 is the only S100 protein with 10 Cys residues, most of which are clustered at the opposite site of the two Ca 2ϩ -binding loops. This abundance of sulfur atoms and the fact that the fully reduced and alkylated protein does not bind Zn 2ϩ at all anymore 3 suggest that the Zn 2ϩ ions are bound in thiolate clusters of the Kagi and Kojima type (reviewed in Vallee and Auld (1990)). Direct binding of Zn 2ϩ to the Cys residues also would explain the strong reduction of the thiol reactivity in the presence of Zn 2ϩ , but not of Ca 2ϩ . The metallothioneins bind 7 Zn 2ϩ per mol (8 for the S100A3 dimer) to 20 cysteinyl residues in clusters of the type Zn 3 S 9 and Zn 4 S 11 . The Zn 2ϩ -thiolate cluster has recently also been observed in DNA-binding proteins (Pan and Coleman (1990). However, since Co 2ϩ -binding does not induce the characteristic absorption bands at 700 nm as it does in metallothioneins, and since half of the Cys residues in S100A3 are not in the free thiol form, it is tempting to postulate that a new type of cluster is present in the latter protein. Structural work is in progress to provide a more detailed description of this novel Zn 2ϩ -binding motif.