Brain S100A5 Is a Novel Calcium-, Zinc-, and Copper Ion-binding Protein of the EF-hand Superfamily*

S100A5 is a novel member of the EF-hand superfamily of calcium-binding proteins that is poorly characterized at the protein level. Immunohistochemical analysis demonstrates that it is expressed in very restricted regions of the adult brain. Here we characterized the human recombinant S100A5, especially its interaction with Ca2+, Zn2+, and Cu2+. Flow dialysis revealed that the homodimeric S100A5 binds four Ca2+ ions with strong positive cooperativity and an affinity 20–100-fold higher than the other S100 proteins studied under identical conditions. S100A5 also binds two Zn2+ ions and four Cu2+ ions per dimer. Cu2+ binding strongly impairs the binding of Ca2+; however, none of these ions change the α-helical-rich secondary structure. After covalent labeling of an exposed thiol with 2-(4′-(iodoacetamide)anilino)-naphthalene-6-sulfonic acid, binding of Cu2+, but not of Ca2+ or Zn2+, strongly decreased its fluorescence. In light of the three-dimensional structure of S100 proteins, our data suggest that in each subunit the single Zn2+ site is located at the opposite side of the EF-hands. The two Cu2+-binding sites likely share ligands of the EF-hands. The potential role of S100A5 in copper homeostasis is discussed.

Calcium, a versatile second messenger of extracellular signals inside the cell, binds to a multitude of cytosolic Ca 2ϩbinding proteins each of which can, in turn, regulate several effector proteins. The S100 protein family (18 different members) constitutes the largest group of Ca 2ϩ -binding proteins of the EF-hand type (1,2). At least 13 S100 genes are clustered on human chromosome 1q21, leading to the designation S100A1 to S100A13 for the protein products of these genes (3). Their expression is cell-and tissue-specific, and different human diseases have been associated with deregulated expression (4,5). S100 proteins are non-covalent homodimers with the notable exception of the heterodimeric S100A8-S100A9. Each monomer possesses 2 EF-hands as follows: a classical C-terminal EFhand with a canonical Ca 2ϩ -binding loop of 12 amino acids and a N-terminal EF-hand with a loop of 14 amino acids which is specific for S100 proteins. This structural difference likely is the reason for the large difference in the Ca 2ϩ affinities of the N-and C-terminal EF-hands. The affinities of S100 proteins for Ca 2ϩ are in general rather low with [Ca 2ϩ ] 0. 5 1 values of 100 -300 mM (for review see Refs. 2 and 6), and in 6 out of 8 studied cases binding of Ca 2ϩ displays positive cooperativity. Zn 2ϩ binds to several S100 proteins; the Zn 2ϩ and Ca 2ϩ sites are distinct and can modify the affinity for Ca 2ϩ . S100B binds four Zn 2ϩ ions with concomitant 10-fold increases of the Ca 2ϩ affinity (7). High affinity binding of two Zn 2ϩ to the S100A12 dimer leads to induction of two high affinity Ca 2ϩ -binding sites (8). S100A3 binds eight Zn 2ϩ , but without effect on the affinity for Ca 2ϩ (9,10). S100A2 binds four Zn 2ϩ with high affinity, in a manner antagonistic to Ca 2ϩ (11). Recently it was reported that the S100B dimer can bind four Cu 2ϩ ions with a K D of 0.46 M, most of which can be displaced with 1 mM Zn 2ϩ (12). Thus, an involvement of S100 proteins in the homeostasis of transition metals has to be considered (2).
The three-dimensional structures of the Ca 2ϩ -bound and metal-free forms of S100B (13)(14)(15)(16) and S100A6 (17,18) revealed the Ca 2ϩ -sensitive and -insensitive structural elements. Both proteins contain a compact dimerization domain, an Xtype four-helix bundle, made up of the N-terminal helix of site I (helices I and IЈ in the dimer) and the C-terminal helix of site II (helices IV and IVЈ). The conformation of this socle is insensitive to Ca 2ϩ . In S100B the orientations of the helices III and IIIЈ are very sensitive to Ca 2ϩ , i.e. together with the central linkers they swing out upon binding of Ca 2ϩ . This uncovers two patches (one in each monomer) of hydrophobic residues, which may allow the protein to bind to phenyl-Sepharose and to cellular target proteins (14). In striking contrast, the Ca 2ϩbound structure of S100A6 is very similar to the metal-free form, and helices III and IIIЈ are oriented outward as in the Ca 2ϩ form of S100B (18). Thus the structural diversity is important in the S100 family. S100A5 is produced by the sixth gene in the S100A cluster on chromosome 1q21 (19,20). Except for its gene structure, no data are available on this protein so far. Initially, it was thought to possess a hydrophobic N-terminal extension of 18 residues, but we demonstrate here that S100A5 isolated from rat brain does not contain this tail. Furthermore, we describe the cation binding properties of the recombinant short form of human S100A5 and investigate the distribution of S100A5 in rat brain by immunohistochemistry with a specific antiserum, comparing it to that of S100A6, which is expressed in a large subset of neurons, glial cells, and sensory afferent fibers in brain (21).

EXPERIMENTAL PROCEDURES
Purification of S100A5 from Rat Brain-Adult Harlan Sprague-Dawley rats (RCC Ltd. Fü llinsdorf, Switzerland) weighing 250 -300 g were anesthetized with ether and decapitated. Olfactory bulb and medulla oblongata were quickly isolated on ice, frozen in powdered dry ice, and stored at Ϫ80°C. Frozen tissues were powdered, lysed, and purified over a phenyl-Sepharose column as described (11). In a final step, S100A5 was purified by reversed-phase high pressure liquid chromatography using a Brownlee C8 column. Solvents were 0.1% trifluoroacetic acid for solvent A and 0.072% trifluoroacetic acid in 80% acetonitrile for solvent B; a gradient of 50 -90% solvent B was applied. Homogeneous S100A5 was eluted at 65% solvent B.
For mass spectrometry analysis 10 g of S100A5 was digested with modified trypsin (Roche Molecular Biochemicals) in 30 l of cleavage buffer (50 mM NH 4 HCO 3 , pH 8.5) with an enzyme/substrate ratio of 1:10 (w/w) overnight at 37°C.
Cloning and Expression of Recombinant S100A5-The coding region of the short form of S100A5, beginning at the postulated second methionine of the cDNA sequence, was amplified by polymerase chain reaction using the following specific oligonucleotides containing an NdeI site as start site of translation and a BamHI site: S100A5-5, GATCA-TATGGAGACTCCTCTGGAGAAG (5Ј-3Ј), and S100A5-3, CAGGGAT-CCGGTCACTTGTTGTCCTCTAGAAA (5Ј-3Ј). To achieve expression of the long form of S100A5, the second methionine was eliminated through site-directed mutagenesis using the following 5Ј primer, GAT-CATATGCCTGCTGCTTGGATTCTCTGGGCTCACTCCCACAGTGAG-CTGCACACTGTGGAGACTCCTCTGGAG (5Ј-3Ј). The amplified cDNAs were cloned into the prokaryotic expression vector pGEMEX (Promega, Madison, WI). The correct open reading frames were confirmed by direct sequencing of the resulting plasmids that were then transformed into the Escherichia coli strain BL-21LysS for bacterial expression. Expression of the recombinant proteins was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside at an A 600 of 0.5. After an additional 3.5 h of culture, bacteria were harvested and resuspended in lysis buffer (20 mM Tris-HCl, pH 7.5, 300 mM KCl, 10% glycerin, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 mM DTT, and 5 g/ml leupeptin). Cell lysis was performed by freezing and thawing twice and sonicating for 2 min on ice. All subsequent procedures were performed at 4°C. The suspension was ultracentrifuged at 150,000 ϫ g for 50 min, and the cleared lysate was precipitated by slowly adding ammonium sulfate to 30% followed by centrifugation at 3000 ϫ g for 15 min. The supernatant was brought to 2 mM CaCl 2 , applied to a phenyl-Sepharose affinity column, chromatographed, washed, and finally eluted with 20 mM Tris-HCl, pH 7.5, and 5 mM EGTA.
Gel Electrophoresis and Mass Spectrometry-15% SDS-Tricine gels were cast and run as described (9). For Western blotting, proteins were blotted at 50 V for 1 h onto 0.2 m nitrocellulose (Scheicher & Schuell) and cross-linked to the membrane by UV irradiation. Proteins were visualized with Ponceau S or Coomassie Blue staining. The membrane was blocked overnight in 3% bovine serum albumin and 1% fetal calf serum and then incubated with S100A5 antiserum (1:5000). The bands were visualized using a horseradish peroxidase-coupled secondary antibody (1:2000) (Sigma). Electrospray ionization mass spectra were obtained with a Perkin-Elmer SCIEX API 365 liquid secondary ion tandem mass spectrometry System mass spectrometer equipped with an atmospheric pressure electrospray ion source. Tryptic peptide masses of rat S100A5 were determined by nano-electrospray ionization mass spectrometry using a home-built nanoelectrospray ion source (54). Peptide mass fingerprinting was performed using ExPASy proteomics tools. The complete protein sequence of rat S100A5 is available from Swiss-Prot under the accession number P82540.
Production of Polyclonal Antiserum against Human Recombinant S100A5-Antiserum was raised in rabbits after immunizing four times (at 3-week intervals) by subcutaneous injection of 250 g of purified protein (500 g for the first injection) in an emulsion with Hunter's Titer Max (CytRx, Norcross, GA). The titer was determined with 50 ng and 1 g of recombinant protein on Western blots.
Immunohistochemistry-Sections from adult rat brain were prepared as described (22). They were incubated overnight in a mixture of human recombinant S100A5 rabbit antiserum (1:2000) and human recombinant S100A6 goat antiserum (1:5000; see Ref. 21) diluted in Tris buffer, pH 7.4, containing 2% normal donkey serum and 0.2% Triton. Sections were then washed, incubated for 30 min at room temperature with the corresponding secondary antibodies coupled to Cy2 and Cy3 (Jackson Immunoresearch, West Grove, PA), washed again, and coverslipped with buffered glycerol. Digital images were acquired by confocal laser-scanning microscopy using dual wavelength excitation for simultaneous recording of both fluorochromes and processed with the software Imaris (Bitplane, Zurich, Switzerland).
Metal Ion Removal and Cation Binding-S100A5 samples were precipitated twice with 3% trichloroacetic acid, dissolved with small amounts of 1 M Tris in 100 M ␤-mercaptoethanol, and passed on a Sephadex G-25 column (0.8 ϫ 40 cm) equilibrated in 50 mM Tris-HCl buffer, 150 mM KCl (buffer A) or, for all experiments involving Cu 2ϩ , 25 mM N-ethylmorpholine HCl, pH 7.4, 150 mM KCl (buffer B). The residual Ca 2ϩ contamination in the protein fractions, checked by atomic absorption, was always less than 0.05 mol/mol of protein. All the experiments were performed in buffer A or B starting with this "metalfree" solution. Total Ca 2ϩ , Mg 2ϩ , Zn 2ϩ , and Cu 2ϩ concentrations were determined with a Perkin-Elmer 2380 atomic absorption spectrophotometer. The protein concentration was determined from the ultraviolet absorption spectrum using a molar extinction coefficient at 278 nm (⑀ M,278 nm ) of 7600 M Ϫ1 cm Ϫ1 for the monomer. This extinction coefficient is much higher than the theoretical one based on the content of Tyr and Cys, which amounts to 4700 M Ϫ1 cm Ϫ1 . Therefore we recorded the spectrum of the same S100A5 solution in the presence and absence of 6 M guanidine HCl as recommended by Edelhoch (23) and Pace et al. (24). In these denaturing conditions ⑀ M,278 nm amounted to 6600 M Ϫ1 cm Ϫ1 . The reason for the unusually large extinction coefficient is unclear but was also reported for S100A4 (25).
Ca 2ϩ binding was measured at 25°C by the flow dialysis method (26) in buffer A. Protein concentrations were 20 -30 M. Treatment of the raw data and evaluation of the intrinsic metal-binding constants were as described (27). Since the Ca 2ϩ -binding isotherms display positive cooperativity, the data were analyzed by Equation 1 of Adair for two binding sites, where K 1 and K 2 are the stoichiometric association constants for the binding of the first and second Ca 2ϩ to the protein.
Zn 2ϩ binding studies were carried out by equilibrium gel filtration as reviewed previously (27) in buffer A at room temperature. For Cu 2ϩ the buffer containing 25 mM N-ethylmorpholine HCl, pH 7.4, 150 mM KCl (buffer B) was chosen for its minimal Cu 2ϩ -chelating activity. As recommended by Brown et al. (28), Cu 2ϩ was supplied as a Cu 2ϩ (glycine) 2 complex, which is stable at neutral pH and mimics the in vivo situation. In three protein-containing fractions the metal ion concentrations were determined by atomic absorption (11) and the protein concentration by UV spectrophotometry. In a control experiment it was established that binding of Cu 2ϩ , but not of Zn 2ϩ , increases ⑀ M,278 nm 1.07-fold. No binding experiments were performed at free Zn 2ϩ or Cu 2ϩ concentrations in the lower micromolar range, since we repeatedly observed that under such conditions the ratio of bound-cation/protein in the three fractions is no longer constant, likely for kinetic reasons.
Secondary Structure Monitored by Circular Dichroism-Circular dichroism (CD) spectra were acquired with a Jasco J-715 spectropolarimeter at 25°C on solutions of 0.25 mg/ml protein in buffer A and also in buffer B for experiments including Cu 2ϩ in a cell of 1-mm optical path. Cu 2ϩ was supplemented as a Cu 2ϩ (glycine) 2 complex. Ellipticities were normalized for the residue concentration using the relationship M w ϭ o M w /lc, where o is the observed ellipticity in millidegrees, M w is the average molecular weight of an amino acid in the polypeptide (116.8 for S100A5), l the path length in mm, and c the protein concentration in mg/ml.
Conformational Changes Monitored by Near-UV Absorption-Difference spectra were taken with a Perkin-Elmer LS-5B spectrophotometer at room temperature. To 70 M metal-free S100A5 in buffer A ions or denaturants were added to final concentrations of 1 mM Ca 2ϩ , 100 M Zn 2ϩ , or 4 M guanidine HCl to obtain the ion-loaded or denatured forms, respectively. Since Cu 2ϩ changes the optical properties of buffer B, the difference spectrum of the Cu 2ϩ versus metal-free form was obtained by subtraction of the spectra of metal-free protein in buffer B before and after addition of 50 M Cu 2ϩ (glycine) 2 to the protein and reference solutions.
Thiol Reactivity and Covalent Modification by IAANS-The thiol reactivity was assayed on a protein sample that was previously reduced by overnight incubation with 100 mM dithiothreitol at pH 8.5 and passed over a Sephadex G-25 column equilibrated in de-gassed and nitrogen-saturated buffer A. The kinetics of the thiol reactivity were monitored at 412 nm (29).
A 1-ml solution of 1 mg/ml reduced S100A5 was incubated for 3 h at room temperature with 1.8 mM IAANS. 15 mM DTT was added, and the mixture was passed over a desalting column of Sephadex G-25. The amount of ANS associated with the protein was determined by UV absorbance at 325 nm using a molar extinction coefficient of 24,900 M Ϫ1 (30). The protein concentration of the covalent adduct (ANS)S100A5 was determined with the Lowry assay (31) with recombinant S100A5 as standard.
ANS Fluorimetry-ANS fluorescence was measured at 25°C on 3 M (ANS)S100A5 in buffer B after direct excitation of the ANS group at 326 nm or after excitation of Tyr at 278 nm in order to monitor the fluorescence resonance energy transfer to the ANS group. The Cu 2ϩ titration experiments were carried out at 25°C after direct excitation at 326 nm. Given that the apparent dissociation constant is around 4 M (see "Results"), a good estimate of the affinity was possible by titration of 0.4 M protein adduct (affinity titration); a good estimate of the stoichiometry was possible by titration of 20 M protein (stoichiometric titration). The slits were set at 10 and 2.5 nm in the affinity and stoichiometric titrations, respectively. In the affinity titration the data points were normalized between 0 (apoprotein) and 1 (Cu 2ϩ -saturated protein).

Expression and Purification of Human
Recombinant S100A5-To express and isolate S100A5 in large amounts, the cDNA (19) was cloned into the prokaryotic expression vector pGEMEX. Since two potential translational initiator codons are present in-frame, both cDNA forms were initially cloned. However, we were not able to express protein from the first ATG per se (data not shown). To achieve expression from this first ATG, the second in-frame initiation codon was eliminated by site-directed mutagenesis. However, the bacteria expressing this protein grew very slowly, and only a very limited amount of protein could be purified by passing the bacterial lysate over a phenyl-Sepharose affinity column. In contrast, the protein starting at the second Met could be purified in large quantities. Purity and concentration of both forms of S100A5 was controlled by SDS gel electrophoresis (Fig. 1) and amino acid analysis. To allow histological localization of S100A5, the short recombinant protein was used to produce a polyclonal antiserum in rabbits. Fig. 2 demonstrates that antibodies could be obtained that specifically recognize the long and the short forms of S100A5 with no obvious cross-reactivity seen against other members of the S100 protein family.
Distribution of S100A5 in Brain-By using the specific antibodies, S100A5 immunoreactivity was detected in only three brain regions where a prominent labeling of axons and terminals was evident. In the olfactory bulb, S100A5 staining was intense in olfactory nerves and their terminals in a subset of glomeruli. Comparison with S100A6, which is also very prominent in olfactory nerves, revealed that a majority of glomeruli contained terminals expressing S100A5 only (about 60%), followed by glomeruli with a mixture of both proteins (30%), and a small number of glomeruli with terminals expressing only S100A6 (10%). In all cases, however, S100A5 and S100A6 staining was found in distinct nerve fibers, as illustrated for a small glomerulus in Fig. 3A. In the brainstem, S100A5 immunoreactivity was detected in a subset of axons in the nucleus of the solitary tract. These fibers formed numerous varicosities suggestive of terminal buttons and were intermingled with axons of similar morphology intensely stained for S100A6 (Fig.  3B). Only occasionally was a given axon double-labeled (Ͻ5%, as assessed in images recorded at high magnification). Finally, a few S100A5-positive axons were seen in the spinal trigeminal tract, extending rostrally up to the level of the nucleus principalis and caudally into the spinal cord. Throughout the brain, no cross-reactivity with other S100 proteins (S100A6 and S100B) was detected, and S100A5 was not found in any glial cell or neuronal soma in brain.
Isolation of Rat Brain S100A5-To investigate which form of S100A5 is expressed in vivo, the protein was partially purified from rat brain olfactory bulb and medullar oblongata. Electrospray mass spectrometry was then used to identify the calciumbinding proteins in the final fraction. By using the Swiss-Prot data base, we identified rat calmodulin (16,790 atomic mass units), acetylated S100B (10,654 atomic mass units), a protein with homology to human olfactory marker protein (19,893 atomic mass units) and finally a protein with 10,854 atomic mass units and homology to human S100A5 (Fig. 4). To unambiguously identify this protein, it was further digested with trypsin, and the resulting mixture was analyzed by nano-electrospray mass spectrometry. Five different peptides could be identified that were all identical to the sequence derived from both mouse and rat S100A5 cDNAs. Incidentally, the peptide METPLEK was found to be acetylated which corresponds to the initiator methionine of the short form of S100A5, as does the overall mass identified. Analysis of partially purified protein extract by Western blot additionally confirmed this observation (data not shown). Therefore, we conclude from these experi-  4), long form (2,5), and a crude E. coli extract induced to express the long form (lanes 4 and 6). 2. Specificity of the polyclonal antiserum against the short form of human recombinant S100A5. Cross-reactivity of the antiserum against other members of the S100 protein family was tested on a Western blot with human recombinant proteins S100A5 (lane 1), S100B (lane 2), S100A13 (lane 3), S100A6 (lane 4), S100A4 (lane 5), S100A2 (lane 6), and S100A1 (lane 7) (2 g each). Dilution of the antibody was 1:5000. D, dimer; M, monomer. ments that S100A5 from rat brain is expressed from the second potential methionine indicating that the short form occurs in vivo.
Biophysical Characterization-To verify the correct synthesis and molecular weight of the short recombinant S100A5, we determined its exact masses by electrospray ionization mass spectrometry. The deconvoluted spectrum of the short form shows two peaks of nearly equivalent intensity with molecular weights of 10,747 (very close to the theoretical value from the sequence, 10,744) and 10,878 (corresponding to the previous sequence plus one methionyl residue). Thus our recombinant S100A5 protein represents the full-length protein, but in about half of the molecules the N-terminal Met residue was not removed during expression in Escherichia coli. On SDS-polyacrylamide gel electrophoresis the short and long forms of the protein migrate in two bands (13 and 20 kDa for the short form and slightly higher for the long form); after efficient reduction only one band is shown at 13 kDa in the short form (Fig. 1A). Thus this protein can form covalent dimers (as S100A6). The native molecular weight of the short S100A5, determined by Sephacryl S200 gel filtration in buffer A containing either 1 mM Ca 2ϩ or 1 mM EGTA, amounts to 23.4 kDa for the Ca 2ϩ -loaded protein and to 25.7 kDa for the metal-free protein. It can thus be concluded that, like the other members of the S100 family (2, 13), the S100A5 protein forms a stable homodimer of 21.6 kDa.
Direct Cation Binding Studies-Since only the short form of S100A5 could be purified from rat brain as the physiological species, this recombinant protein was used for all subsequent studies. Ca 2ϩ binding by flow dialysis under physiological ionic strength conditions (Fig. 5) revealed two high affinity Ca 2ϩbinding sites per monomer plus some nonspecific binding at high Ca 2ϩ concentrations in the absence of Mg 2ϩ . Analyzed with the Adair Equation 1 for two sites the isotherm yielded intrinsic constants (K n Ј) of 6.3 ϫ 10 3 and 4.1 ϫ 10 6 M Ϫ1 for K 1 Ј and K 2 Ј, respectively, indicative of very strong positive cooperativity (n H ϭ 1.92). Whereas the high positive cooperativity had already been observed in the case of S100A2 (11), the high affinity ([Ca 2ϩ ] 0.5 ϭ 6.3 M) of S100A5 is 20-to 100-fold higher than that of the other S100 proteins studied under identical conditions (2). The presence of 2 mM Mg 2ϩ does not influence the binding isotherm, indicating that the two sites are of the so-called Ca 2ϩ -specific type, as had also been reported for the other S100 proteins (2). No significant changes in the Ca 2ϩbinding parameters were observed when the measurements were performed in the presence of 20 M Zn 2ϩ , whereas nearly no Ca 2ϩ was bound in the presence of 20 M Cu-Gly 2 (asterisks in Fig. 5).
Since equilibrium dialysis on Ca 2ϩ -free S100A5 in the presence of Zn 2ϩ or Cu 2ϩ leads to protein precipitation after 24 h, equilibrium gel filtration experiments were carried out as indicated to probe the affinity of these two cations. Table I shows that in the absence of other divalent cations S100A5 binds 1 mol of Zn 2ϩ per monomer with an estimated [Zn 2ϩ ] 0.5 in the order of 1-3 M. S100A5 also binds 2 mol of Cu 2ϩ per monomer with an estimated [Cu 2ϩ ] 0.5 in the order of 2-5 M. For kinetic reasons the method does not allow us to determine bound Cu 2ϩ at lower free cation concentrations. This stoichiometry and affinity were also found in a Cu 2ϩ -dependent fluorescence titration (see below). Most interestingly, gel filtration experiments in the presence of both Ca 2ϩ and Zn 2ϩ or Cu 2ϩ showed that S100A5 can bind simultaneously 2 Ca 2ϩ and 1 Zn 2ϩ , whereas the binding of Ca 2ϩ is strongly reduced in the presence of Cu 2ϩ (Table I), thus confirming the flow dialysis data.
Secondary Structure Probed by Circular Dichroism-The spectra shown in Fig. 6 indicate that metal-free S100A5 is rich in ␣-helical structure with a mean residue ellipticity at 222 nm of 19,000 deg cm 2 dmol Ϫ1 , compared with 15,500 for S100A2 (11). Quantitative analysis according to Johnson (33) yielded the following estimates of secondary structure: 59% ␣-helix, 8% of ␤-pleated sheet, and 5% ␤-turn. Addition of 1 mM Ca 2ϩ or 2 mM Mg 2ϩ , 100 M Zn 2ϩ or Cu 2ϩ does not at all change the shape or intensity of the spectrum. It can thus be concluded that, as in the case of S100A2 (11) or S100P (33), the secondary structure of S100A5 as measured by circular dichroism practically does not change upon ion binding.
Near-UV Difference Spectra-The near UV spectrum (Fig. 7, right inset) of this protein shows a maximum at 278 nm with a marked shoulder at 283 nm, typical of a protein containing only Tyr but no Trp. In UV difference spectrophotometry (Fig. 7) the "Zn 2ϩ -apo" difference spectrum is very weak, indicating that Zn 2ϩ binding does not perturb the microenvironment of the aromatic residues. The spectrum of the "Ca 2ϩ -apo" form shows two positive peaks at 280 and 288 nm (Fig. 7). The spectrum of the "Cu 2ϩ -apo" form shows a similar, although blue-shifted, profile in the 275-100-nm region but also discrete bands at 259, 265, and 269 nm, indicating that the Phe environment is sensitive to binding of Cu 2ϩ but not of Ca 2ϩ . Titration of the ⌬OD signal change of 150 M S100A5 (Fig. 7, left inset) confirmed the stoichiometry of two Ca 2ϩ per protein monomer and the high affinity of the protein for Ca 2ϩ . Since titration of S100A5 with Cu 2ϩ does not yield a clear-cut plateau (data not shown), the binding parameters pertaining to this ion were probed on (ANS)S100A5.
Thiol Reactivity and Covalent Modification with IAANS-After overnight reduction each monomer contains 2 free thiols; the first reacts very rapidly in the apo and Ca 2ϩ -state, and the second is protected from DTNB but reacts instantaneously in the presence of 4 M guanidine HCl. In the presence of Zn 2ϩ the two thiols react slowly but at equal rates (t1 ⁄2 ϭ 2 min). In the presence of both Ca 2ϩ and Zn 2ϩ only one thiol reacts, but with a similar speed as in Zn 2ϩ alone.
The protocol for covalent modification of the reactive thiol with IAANS as described under "Experimental Procedures" yielded a degree of substitution of 0.7 ANS/S100A5 monomer. The fluorescence spectra registered after direct excitation of the ANS group at 326 nm (not shown) or after excitation of the Tyr residues at 278 nm (Fig. 8) were very similar, except that the intensities after direct excitation where 1.3-fold higher. Hence, there is efficient fluorescence resonance energy transfer from Tyr to ANS, suggesting a close vicinity of the ANS group and at least one Tyr. The addition of Ca 2ϩ or Zn 2ϩ has only a FIG. 3. Digital images from confocal laser scanning microscopy depicting the distribution of axons immunoreactive for S100A5 (green) and S100A6 (red) in the olfactory bulb (A) and nucleus of the solitary tract, pars commissuralis (B), as seen by double immunofluorescence staining in transverse sections of adult rat brain. Images corresponding to each fluorochrome were merged in the false-colored images. In each panel a stack of 30 consecutive confocal images spaced by 250 nm is displayed using a "simulated fluorescence projection" algorithm for three-dimensional rendering. Note that the staining of S100A5 and S100A6 is found in largely distinct axons. Scale bars, 10 m. minor effect on the fluorescence intensity and maximum, but the addition of 40 M Cu 2ϩ leads to a 2-fold decrease. In the Cu 2ϩ form the ANS group is still shielded since addition of 4 M guanidine HCl leads to a further 2-fold decrease of the fluorescence. These fluorescence properties allowed us to perform a titration with Cu 2ϩ . A stoichiometric titration of 20 M (ANS)S100A5 (Fig. 8) revealed that upon binding of two Cu 2ϩ ions per monomer the major fluorescence change was achieved; the small downward drift of the plateau value represents Cu 2ϩ quenching of the ANS group. During titration the fluorescence decrease is 3-fold stronger for binding of the first than of the second Cu 2ϩ (per S100A5 monomer), which strongly suggests that the binding of Cu 2ϩ is sequential as in the Nereis sarcoplasmic Ca 2ϩ -binding protein SCP (35). The affinity titration of 0.5 M (ANS)S100A5 yielded a [Cu 2ϩ ] 0.5 of 4 M (Fig. 6). These values confirm the stoichiometry and magnitude of the affinity of the non-modified S100A5 for Cu 2ϩ as estimated by equilibrium gel filtration (Table I).
FIG. 5. Ca 2؉ binding to S100A5 and effect of Mg 2؉ and Zn 2؉ . Ca 2ϩ binding was measured by flow dialysis at 25°C in buffer A. No additional ions (circles), 2 mM Mg 2ϩ (triangles), 20 M Zn 2ϩ (rectangles), or 20 M Cu 2ϩ (asterisks). The lines connecting the symbols were generated using Adair Equation 1 with the following intrinsic constants for K Ca1 Ј and K Ca2 Ј, 6.3 ϫ 10 3 and 4.1 ϫ 10 6 M Ϫ1 .

TABLE I
Ca 2ϩ , Zn 2ϩ , and Cu 2ϩ binding to S100A5 monitored by equilibrium gel filtration Protein samples (0.5 ml of 50 -300 M S100A5) were chromatographed on a 0.8 ϫ 40-cm Sephadex G-25 column equilibrated in buffer A (for Ca 2ϩ and Zn 2ϩ ) or B (for Cu 2ϩ ) and the indicated free cation concentrations. Bound cations were measured by atomic absorption and protein by UV spectrophotometry. The accuracy of bound cations is Ϯ90%.  4. Electrospray mass spectrum of S100A5. Mass spectrum of rat S100A5 isolated from rat brain tissue. The inset shows the deconvoluted mass spectrum indicating the molecular mass measured.

DISCUSSION
The reported primary structure of S100A5 derived from its cDNA (19) included an N-terminal extension of 18 residues that were not present in the other members of the S100 protein family. However, a second potential initiation site is present on the cDNA at the position where several other S100A proteins start. By using the full-length cDNA, E. coli expresses only the shorter protein product, as evident from the N-terminal sequence, and the molecular weight was determined by mass spectrometry. However, when eliminating the second Met by site-directed mutagenesis, bacterial growth was severely impaired, and consequently only small amounts of protein could be purified. Intriguingly, partial purification of S100A5 from rat brain and subsequent analysis by mass spectrometry and Western blotting demonstrated that the short form is also the physiological species in this tissue. Hence, immunolocalization and biochemical and biophysical properties of this short form of human S100A5 are described in this study. The protein is dimeric with a predominance of ␣-helical structure, the content of which does not change upon binding of Ca 2ϩ or other cations. The conservation of critical residues in the primary structure suggests that the protein contains the same dimerization scaffold as other S100 proteins.
The tissue distribution of S100A5 was first assessed by Northern and Western blot analysis, but no specific signals could be detected (data not shown), indicating that the expression level of the protein is either very low or restricted to a few specific cells. The latter hypothesis could be confirmed by immunohistochemical analyses which demonstrated that S100A5 is only expressed in a few distinct brain areas, namely the olfactory bulb, the brainstem, and the spinal trigeminal tract. This expression pattern is somewhat reminiscent of that of S100A3 (36,37) and clearly distinct from other S100 proteins such as S100A6 and S100B. S100A5 displays high affinity for Ca 2ϩ when compared with other S100 proteins analyzed under identical experimental conditions with the same methodology (Table II). The reason for this is not evident from differences in the primary structure. Among S100 proteins only the monomeric calbindin 9 kDa also has a very high affinity for Ca 2ϩ (38), which may be explained by the absence of the Ca 2ϩ -induced hydrophobic patch. Table II shows that only S100A7, S100A12, and S100P, studied under different conditions and with other methodologies, display comparable affinities. The very strong positive cooperativity, also observed in the case of S100A2 (11), can be attributed to an energetically expensive conformational change that occurs after binding of the first Ca 2ϩ in each monomer. Like most other S100 proteins S100A5 does not bind Mg 2ϩ to a significant The Cu 2ϩ versus the metal-free difference spectrum (⅐⅐⅐⅐) was determined as described under "Experimental Procedures." The difference optical density (⌬OD) was expressed for a protein solution with an optical density of 1.0 at 280 nm. Right inset, UV spectrum of S100A5. Left inset, Ca 2ϩ dependence of the difference spectral changes of S100A5. The means of ⌬S (280 -274 nm) and ⌬S (288 -274 nm) were taken. extent, and Mg 2ϩ does not influence the Ca 2ϩ -binding isotherm. Recently it was reported that S100P may possess a Ca 2ϩ /Mg 2ϩ switch (39), but it seems unlikely that at physiological ionic strength and Mg 2ϩ concentrations, i.e. 1.5 mM (40), the binding of Mg 2ϩ is significant in vivo.
The S100A5 dimer binds also two Zn 2ϩ ions with a [Zn 2ϩ ] 0.5 of 2 M. Zn 2ϩ binding is quite common among S100 proteins which display affinities varying from high (S100A2, A3, A5, A6, and A12) to low (S100B and S100A7) and acts in synergy, antagonism, or is without effect on Ca 2ϩ binding (Table II). For S100A5 Zn 2ϩ binding does not influence Ca 2ϩ binding, but it is likely that Zn 2ϩ enhances the interaction of the protein with targets. Indeed, Zn 2ϩ increases 30-fold the activation of twitchin kinase by the Ca 2ϩ form of S100A1 (41), and Ca 2ϩ -S100B displays a 5-fold higher affinity for the peptide TRTK-12, provided Zn 2ϩ is present (42). The location of the Zn 2ϩbinding sites in S100 proteins is still unresolved although the crystal structure of S100A7 revealed a Zn 2ϩ -binding site at the opposite side of the EF-hands in the same subunit, a site formed by three His and one Asp residues (43). This sequence motif is present in a number of other S100 members but not in S100A5. Until now Cu 2ϩ binding has been described for S100B only, with a cation/dimer stoichiometry of 4 (12) and a [Cu 2ϩ ] 0.5 value of 0.5 M. The affinity of S100A5 for Cu 2ϩ must be higher than that of S100B, since the Cu 2ϩ (glycine) 2 complex is ϳ500fold stronger (44) than the Cu(OH) 2 complex used in the study on S100B. In S100A5 Cu 2ϩ binding is strongly antagonistic to the binding of Ca 2ϩ , suggesting that Cu 2ϩ binds to some of the ligands present in the Ca 2ϩ -binding loops. The location of the Cu 2ϩ -binding sites in S100A5 is, however, difficult to assess since well characterized Cu 2ϩ -binding motifs, such as PHGGG-WGQ in prion protein (45), GMTCXXC in Menkes ATPase (46), GNHWAVGH in some neuropeptides (47) or thiol clusters in metallothionein (48), are absent in S100A5.
Whereas the secondary structure of S100A5 does not notice-

TABLE II
Ca 2ϩ -and Zn 2ϩ -binding parameters of the S100 proteins The Ca 2ϩ -binding parameters of S1008, S100A9, and S100A13 have not been reported; S10010 does not bind Ca 2ϩ . a Per dimer; NA, not applicable; ND, not determined; No, no effect. b The Ca 2ϩ -binding parameters of the indicated S100 proteins have been measured in our laboratory under identical ionic conditions and with the same methodology.
c Low affinity estimated from fluorescence changes. d Only one high affinity site. e Two sites of different affinity.
ably change upon binding of cations, the tertiary structure, probed by difference spectrophotometry, is sensitive to cations. Ca 2ϩ and Cu 2ϩ affect the near-UV spectra, whereas Zn 2ϩ has no effect. Specific to Cu 2ϩ is that it also affects the absorption of Phe residues. These data strengthen our hypothesis from direct binding that the Zn 2ϩ sites are completely different from the EF-hands, whereas Cu 2ϩ may share ligands in the EFhands. In S100A5 one thiol is protected from DTNB and the other reacts very fast. We hypothesize that the DTNB-accessible thiol is Cys 41 in the central linker since this linker is mobile and accessible in the three-dimensional structures of S100B and S100A6, whereas Cys is part of the dimerization scaffold. Interestingly, the fluorescence of ANS attached to the reactive thiol is sensitive to Cu 2ϩ binding, but not to Zn 2ϩ and Ca 2ϩ binding.
The biological significance of Cu 2ϩ binding by S100A5 may be far-reaching in the light of the "copper paradox": copper is a cofactor of several essential enzymes, but even traces of free copper lead to the generation of hydroxyl radicals and to oxidative damage (49). Therefore, virtually no free copper is found in the cytoplasm (50), and copper delivery to the vital copper enzymes occurs via a group of specific metal ion chaperones, e.g. Atx1 delivers Cu 2ϩ to Menkes protein, Lys 7 to superoxide dismutase 1, and Cox17 to cytochrome c oxidase (51). These chaperones have a moderate affinity for Cu 2ϩ with K D in the micromolar range, so that the routing to the enzymes, which have a much higher affinity for the metal ion (K D in the nanoto picomolar range), is energetically favored. Interestingly, S100A5 has a similar affinity for copper as the chaperones and could thus function in specialized cells as a routing device of cytoplasmic Cu 2ϩ . Alternatively, if S100A5 can assume an extracellular location close to the specific neuronal cells, it might protect those cells against oxidative stress by sequestering Cu 2ϩ , as was proposed for S100B (12) and cytoplasmic prion protein (28) in other cell types. Finally, sequestration of Cu 2ϩ by S100A5 may lead to a reduced infectivity of scrapie (52,53). This is in agreement with our recent study (55) demonstrating an intra-as well as extracellular localization of S100A5 in some astrocytic tumors.
From our studies we conclude that S100A5 is an unusual member of the S100 protein family in that it has a high affinity for Ca 2ϩ , Zn 2ϩ , and Cu 2ϩ and that it induces distinct structural changes. Since S100A5 expression was detected only in very restricted regions of the adult brain it might fulfill some very specialized functions in these neurons requiring binding of either Ca 2ϩ , Zn 2ϩ , or Cu 2ϩ .