Calcium-binding by p26olf, an S100-like protein in the frog olfactory epithelium Naofumi Miwa, Yukiko Shinmyo and Satoru Kawamura From the Department of Biology, Graduate School of Science, Osaka University, Japan Frog p26olf is a novel S100-like Ca 21 -binding protein found in olfactory cilia. It consists of two S100-like domains aligned sequentially, and has a total of four Ca 21 -binding sites (known as EF-hands). In this study, to elucidate the mechanism of Ca 21 -binding to each EF-hand (named EF-A, -B, -C and -D from the N-terminus of p26olf), we examined Ca 21 -binding in wild-type p26olf and also in its mutants in which a glutamate at the –z coordinate position within each Ca 21 -binding loop was substituted for a glutamine. Flow dialysis experiments showed that the wild-type binds nearly four Ca 21 per molecule maximally, while all the mutants bind approximately three Ca 21 . Although EF-B and -D are p26olf-specific EF-hands and their role in Ca 21 -binding is not known, the result unequivocally showed that they actually bind Ca 21 . The overall Ca 21 -binding affinity decreased in the three mutants. The decrease was very large in the mutants of EF-A and -B, which suggested that the Ca 21 -affinities are high in EF-A and -B in the wild- type. Assuming the presence of four steps of Ca 21 -binding, we determined the dissociation constant of each step in wild-type p26olf. To assign which step takes place at which EF-hand, we measured the antagonistic effect of K 1 on each step, as the effect of K 1 is thought to be a function of the number of the carboxyl groups in an EF-hand. Although the actual Ca 21 -binding mechanism may not be so simple, this study together with the mutation study suggested a tentative Ca 21 -binding model of p26olf: the order of Ca 21 -binding to p26olf is EF-B, EF-A, EF-C and EF-D. Based on these results, we speculate that similar Ca 21 -binding takes place in an S100 dimer. Keywords: calcium-binding; p26olf; S100. We have previously isolated a novel Ca 21 -binding protein, p26olf, from the frog olfactory epithelium [1]. This protein localizes in the cilia of olfactory epithelium and interacts with a frog b-adrenergic receptor kinase (bARK)-like protein in a Ca 21 -dependent manner [2]. Through the bARK-dependent phosphorylation, p26olf has been suggested to have some role(s) in olfactory signal transduction. The amino-acid sequence of p26olf is most similar to a pair of S100 proteins aligned in tandem [1,3]. The S100 protein family, one of the subgroups of proteins that contain EF-hands, is known to be involved in many types of biological function such as cell-cycle progression [4,5], differentiation [6,7] and regulation of enzyme activity [8]. Abnormal expression of S100 proteins is thought to be a cause of a number of diseases including cancer [9] (reviewed in [10,11]). There are two Ca 21 -binding sites, known as EF-hands, in S100 [12]: one in the N-terminal half is S100-specific showing low affinity for Ca 21 (dissociation constant, K d ¼ 200–500 mM) and the other in the C-terminal half is a typical EF-hand showing high affinity (K d ¼ 20– 50 m M) [13]. In p26olf, however, this typical EF-hand is somewhat modified. The amino-acid sequence of this site fulfills the requirement of the typical EF-hand but there is a four-amino-acid-insertion between the E and F a helix, which characterizes this EF-hand as p26olf-specific [3]. In p26olf therefore there are four EF-hands (named as EF-A, -B, -C and -D from the N-terminus of p26olf), and two of them (EF-A and -C) are S100-specific and the other two (EF-B and -D) are p26olf-specific. S100 proteins form a homo- or a heterodimer, which is the functional form [14,15]. In the previous studies, Ca 21 - binding was examined in S100 dimers and it was found that the apparent dissociation constants in four steps of the binding in a dimer are 20–100 m M [16,17]. As the S100- specific site shows low affinity for Ca 21 , this result suggests the presence of a cooperative interaction in and/or between S100 monomers. The mechanism of this cooperative interaction, however, is not known. In a homodimer of S100, it is difficult to determine which site is responsible for the nth binding, because there are two identical sites in a homodimer and one cannot be certain which subunit is under consideration. As for a heterodimer, the isolation of a heterodimer itself is difficult, because one cannot be sure whether only a heterodimer is present. It may be the case that homodimers are formed after dissociation of a hetero- dimer. As p26olf is a protein composed of a single peptide, there is no uncertainty about a dimer state. In the present study, we measured Ca 21 -binding in wild- type p26olf, and determined the dissociation constants of the first to fourth binding of Ca 21 by assuming the presence of four steps of sequential and ordered binding. To identify the site of the EF-hand corresponding to the nth binding of Ca 21 , we measured the affinity for K 1 in the nth binding. As the affinity for K 1 is suggested to be dependent on the number of the carboxyl groups in an EF-hand in calmodulin [18] and actually the number is different in each EF-hand in Correspondence to S. Kawamura, Department of Biology, Graduate School of Science, Osaka University, Machikane-yama 1-1, Toyonaka, Osaka 560 0043, Japan. Fax: 1 81 6 6850 5444, Tel.: 1 81 6 6850 5436, E-mail: kawamura@bio.sci.osaka-u.ac.jp (Received 15 June 2001, revised 7 September 2001, accepted 12 September 2001) Abbreviations: bARK, b-adrenergic receptor kinase. Eur. J. Biochem. 268, 6029–6036 (2001) q FEBS 2001 p26olf, we can possibly identify the site of the EF-hand responsible for the nth binding of Ca 21 in p26olf. In addition, to examine how each EF-hand contributes to Ca 21 -binding in p26olf, we measured Ca 21 -binding in four mutants each of which has a Glu !Gln mutation at the –z coordination position. Based on these results, we will discuss the order of Ca 21 -binding to p26olf. EXPERIMENTAL PROCEDURES Ca 21 -binding studies Recombinant p26olf was prepared as described [1]. Protein concentration was determined by densitometry using a laser densitometer (Molecular Dynamics). To measure the stoichiometry of Ca 21 -binding to p26olf, we performed flow dialysis experiments as previously described [2]. Briefly, we incubated p26olf (final concentration; 10 m M) in a 20-mM Tris/HCl buffer (pH 7.5) at 20 8C with various concen- trations of 45 CaCl 2 (6.2 Â 10 2 Ci : mmol 21 ) at KCl concen- trations that depended on the type of the experiment. Unless otherwise stated, the Tris buffer always contained 100 m M KCl in the present study. The reaction mixture was placed in a prewashed microconcentrator (Microcon, Amicon) and then it was centrifuged briefly. We counted the activities of 45 Ca in 4 mL portions of both the reaction mixture and the filtrate using a scintillation cocktail (Clearzol; Nacalai, Kyoto, Japan). By comparing the activities of 45 Ca in the reaction mixture and the filtrate, the amount of Ca 21 bound to p26olf was calculated. Blank experiments without p26olf were performed to correct for nonspecific binding of Ca 21 to the membrane of microconcentrators. The Ca 21 concentration in each reaction mixture was calibrated with aCa 21 electrode using diluted solutions of a 20-mM CaCl 2 solution (Sigma). The raw data were analyzed with the Hill equation [19], the Scatchard plot [20], and then, the Adair equation [21]. The Adair equation is represented as follows. R ¼ 1 K 1 x 1 2 1 K 1 1 K 2 x 2 1 ::: 1 n 1 K 1 1 K 2 ::: 1 K n x n  = 1 1 1 K 1 x 1 1 K 1 1 K 2 x 2 1 ::: 1 1 K 1 1 K 2 ::: 1 K n x n  Where R is the molar ratio of bound Ca 21 to p26olf at a free Ca 21 concentration x, and K 1 , K 2 , and K n are the macro- scopic dissociation constants for the binding of one and two, and n Ca 21 to p26olf in a reaction of CD measurement CD spectra of wild-type p26olf and mutants (final concentration; 10 m M each) were measured with a Jasco J-720 W spectrophotometer with 1-mm light path in Tris buffer containing 100 m M KCl. The Ca 21 concentration was varied using a Ca/EGTA buffering system [2] and was calibrated fluorometrically with the following two calcium indicators: below 10 m M Ca 21 , we used fluo-3 (Dojin, Kumamoto, Japan), and over 10 m M Ca 21 , fluo-3FF (Calbiochem). The CD spectra were recorded in the region of 200–250 nm at 0.1 nm interval, scan speed of 10 nm : min 21 , response time of 4 s, and at 20 8C. The spectrum of each sample was measured twice and they were averaged. Construction of mutated expression vectors and expression of mutants of p26olf To generate a series of mutants of p26olf, we did site- directed mutagenesis according to the methods of Tachi- banaki et al. [22]. We used the following oligonucleotides and the complementary oligonucleotides as PCR primers: AACTTCAAA CAGTTTGAGCAG for EF-A-mutation, GACTTTCAA CAGTTTCTCAAC for EF-B-mutation, GAT TACACA CAGTTCGAGGCA for EF-C-mutation, AATTT CCAG CAGTTCATGAAC for EF-D-mutation. The under- lined codons show the sites of mutations, replacing glutamate (E) with glutamine (Q). After PCR reactions, the mutated DNA fragments were digested with Nde I and Bam HI and inserted into pET3a (Novagen), and these recombinant plasmids were introduced into Escherichia coli BL21 pLysS (Novagen). After induction of the expression of p26olf by isopropyl thio-b- D-galactoside, p26olf-mutants were purified according to the method of purification of native p26olf [1]. RESULTS Ca 21 -binding to p26olf In p26olf, two S100 homolog domains are located sequen- tially in the molecule [1,3]. S100 proteins are generally known to form a homo- or a heterodimer in solution [14,15], but in some experiments the recombinant S100 protein formed a trimer [17]. Therefore, it could be the case that our recombinant p26olf prepared here might form a dimer. If it is present, interpretation of the results of Ca 21 -binding may be complicated. Therefore, to exclude this possibility we performed gel filtration column chromatography and confirmed that our p26olf exists as a monomer; purified p26olf eluted through the column as a single peak with a molecular mass of < 28 kDa in the absence of Ca 21 , and < 16 kDa in the presence of Ca 21 (data not shown). Because the calculated molecular mass of p26olf is < 24 kDa [1], this result indicated that p26olf exists as a monomer in our solution. In order to determine the affinity for Ca 21 and the cooperativity of Ca 21 -binding to p26olf, we next analyzed the stoichiometry of Ca 21 -binding to p26olf by flow dialysis in the Tris buffer. Ca 21 -binding to p26olf as a function of 6030 N. Miwa et al. (Eur. J. Biochem. 268) q FEBS 2001 free Ca 21 concentration is shown in Fig. 1. Because p26olf tends to aggregate at high concentrations, we used 10 m M p26olf in a Ca 21 -binding experiment. Due to this limitation of the concentration of p26olf, we considered Ca 21 -binding to p26olf below < 200 m M Ca 21 . Above this concentration of Ca 21 , the Ca 21 -binding signal (at most 40 mM, see below) was within a noise level of the total Ca 21 concen- tration and therefore, the result would not be reliable. The binding data showed that the maximum number of Ca 21 -binding to p26olf is approximately four per molecule (Fig. 1), and therefore we fitted the data with the Hill equation by assuming that p26olf has four Ca 21 -binding sites. The fitted curve (solid line in Fig. 1) showed a reasonable fit to the experimental binding data with the K Ca d value of 22.3 mM and the Hill coefficient of 2.0 [in order to distinguish the K d value determined by the Ca 21 -binding studies and that determined by the CD measurement (see below), we added the superscript Ca or CD to the term of K d ]. The convex curve observed in the Scatchard plot of the binding data clearly indicates the presence of a positive cooperativity (see inset in Fig. 1). To estimate the dissociation constants in the Adair equation (see Experimental procedures), we performed the curve fitting by assuming the presence of four steps of Ca 21 -binding. The theoretical curve (dashed line in Fig. 1) fitted well to the binding data with K 1 ¼ 83.3 mM, K 2 ¼ 6.3 mM, K 3 ¼ 22.2 mM and K 4 ¼ 50 mM, for example, but due to the experimental errors of the measurement, we could not determine the four dissociation constants uniquely. However, after many trials, we reasonably con- cluded that the dissociation constants had a tendency of K 2 , K 3 ø K 4 , K 1 . The Adair dissociation constants obtained were slightly larger than those obtained in S100 dimers [16]. This is because in the study in Fig. 1, Ca 21 -binding was measured in the presence of 100 mM K 1 that is known to reduce the affinity for Ca 21 (see below). Effect of K 1 on Ca 21 -binding to p26olf The antagonistic effect of K 1 on Ca 21 -binding was observed in many of Ca 21 -binding proteins such as S100 and calmodulin [16,18]. Interestingly, Haiech et al. assumed that the affinity for K 1 is a function of the number of the carboxyl groups in an EF-hand [18], and their rationale is in good agreement with the actual sequence of Ca 21 -binding [23]. Although their assumption should be tested in other Ca 21 -binding proteins, we here assumed that a similar analysis can be applied to p26olf. Thus, we determined the affinity for K 1 in each step of the Adair equation by changing the K 1 concentration in order to identify which Adair dissociation constant (K 1 –K 4 ) corresponds to which EF-hand. As shown in Fig. 2, the maximum number of Ca 21 -binding was nearly four per molecule at all the K 1 concentrations tested. Assuming the presence of four Ca 21 -binding sites, the data were first fitted by the Hill equation. As a result, we obtained K d Ca values of 22, 15.0 and 12.1 mM, and the Hill coefficients of 2.0, 1.9 and 1.7 at 100 (Fig. 1), 20 and 0 m M K 1 , respectively (inset in Fig. 2). Because K 1 increased the K Ca d values in the Hill equation, K 1 competed with Ca 21 for the EF-hands of p26olf similarly as in the cases of S100 and calmodulin [16,18]. The data were then fitted by the Adair equation and the determined dissociation constants (K 1 –K 4 ) are summarized in Fig. 3. To calculate the dissociation constant for K 1 in each step, step 1 for example, we fitted the data of K 1 at different K 1 concentrations with the equation of K iapp ¼ K i (1 1 [K 1 ]/k i ) [18]. In the equation, K iapp is the calculated Adair dissociation constant at step i in the presence of K 1 , K i is the Adair dissociation constant at 0 mM K 1 , and k i is the intrinsic dissociation constant for K 1 at step i. The calculated result showed that the affinity for K 1 is high in steps 1 and 4, and low in steps 2 and 3 (see the k values in Fig. 1. Ca 21 -binding to p26olf. The amount of Ca 21 bound to p26olf in the Tris buffer containing 100 m M KCl is shown as a function of free Ca 21 concentration. The experimental points represent the average of 16 different experiments using three different preparations of p26olf. Each bar represents the standard deviation. The data were fitted to the Hill equation (solid line) and the Adair equation (dashed line). The index of the fit represented as the coefficient of correlation (r 2 ) is 1.0 in both the fitting to the Hill equation and to the Adair equation. Fig. 2. Effect of K 1 on Ca 21 -binding to p26olf. Ca 21 -binding to p26olf was examined at 100 m M KCl (circles), 20 mM KCl (triangles) and 0 m M KCl (crosses). The data points of 20 mM KCl represent the average ^ SD (n ¼ 12) of three different preparations of p26olf, and those of 0 m M KCl represent the average ^ SD (n ¼ 5) of two different preparations. The data were fitted to the Hill equation (solid line, see text). The r 2 values are 0.99 (100 mM KCl), 1.0 (20 mM KCl) and 1.0 (0 m M KCl). q FEBS 2001 Calcium-binding by p26olf (Eur. J. Biochem. 268) 6031 Fig. 3). Due to the experimental errors, we could not determine the k-values uniquely, but our best estimate was k 1 ø k 4 ,, k 2 ø k 3 . These results suggested that the steps 1 and 4 take place in EF-hands which have many carboxyl groups in the EF-hand motif (see Discussion). Effect of K 1 on Ca 21 -induced conformational change of p26olf Our previous CD measurement in the absence of K 1 showed that Ca 21 -binding to p26olf increases the negative signal at both 210 nm and 222 nm [2]. As K 1 reduces the affinity for Ca 21 , we measured CD spectra in the presence of 100 m M K 1 as a function of Ca 21 concentration. At all Ca 21 concentrations, negative CD signals increased in a Ca 21 -dependent manner (inset in Fig. 4). Figure 4 shows the percent change of CD at 222 nm (signal of a helix) plotted as a function of the free Ca 21 concentration (lower axis) and the number of the bound Ca 21 per p26olf molecule (upper axis, nCa 21 ) calculated from the binding data in Fig. 1 (circles, data at 100 m M K 1 ; triangles, data at 0 mM K 1 taken from Fig. 2 in [2]). The percent changes were fitted using the Hill equation (solid lines). This fitting for the data at 100 m M K 1 indicated the K CD d value of 8.5 mM and the Hill coefficient of 2.3. When these values were compared with those in the absence of K 1 (K CD d ¼ 2.5 mM,Hillcoef- ficient ¼ 1.5 from [2]), it is revealed that the conformational change of p26olf takes place at higher Ca 21 concentrations in the presence of 100 m M K 1 than in the absence of K 1 . Irrespective of the K 1 concentration, however, the first Ca 21 -binding induced < 80% CD change and the second Ca 21 binding < 90% change (Fig. 4). Ca 21 -binding to mutant p26olf In order to know how each EF-hand contributes to Ca 21 -binding or Ca 21 -induced conformational changes in p26olf, we prepared four mutant proteins that lacked the activity of one of the four EF-hands in p26olf. To obtain such mutants, we mutated the C-terminal glutamate (E) in each calcium binding loop (Fig. 5). This residue is well- conserved among many EF-hand type Ca 21 -binding proteins, and is believed to be important in providing two oxygen ligands to Ca 21 [24]. In order to minimize the effect of mutation on the structure of p26olf, we substituted this glutamate (E) for glutamine (Q). As this substitution leaves one ligand intact within the residue, it is possible that the Ca 21 -binding activity of the EF-hand is partially retained. However, this substitution is shown to be sufficient to suppress Ca 21 -binding to the EF-hand at 200 mM Ca 21 [25,26], which was the highest Ca 21 concentration used in the present study. The mutant proteins generated were named as DEF-A [glutamate at position 40 was replaced by glutamine (E40Q) in EF-A], DEF-B (E86Q in EF-B), DEF-C (E149Q in EF-C) and DEF-D (E194Q in EF-D). Ca 21 -binding in these mutants was measured at various Ca 21 concentrations in the Tris buffer. The maximum Ca 21 - binding in each mutant was close to three per molecule Fig. 3. Dissociation constants of Ca 21 -binding to p26olf. The dissociation constant of each Ca 21 -binding step (K 1 –K 4 )was determined by the Adair equation for the data obtained in the presence of 0 m M,20mM, and 100 mM KCl. The intrinsic dissociation constants for K 1 (k ) for each Ca 21 -binding step is also shown. Fig. 5. Four site-directed mutants of p26olf. In each EF-hand motif, the glutamate residue (black boxes) at the C-terminus of each EF-hand motif was replaced by glutamine, and the mutants thus prepared were named as DEF-A, DEF-B, DEF-C and DEF-D. The p26olf-specific insertions of four residues are shown by white bars in the upper figure and also double-underlines in the amino-acid sequence. Fig. 4. CD changes at 222 nm as a function of Ca 21 concentration. CD spectra of p26olf were measured at various Ca 21 concentrations in the presence of 100 m M KCl (inset, sample records at 12 nM (curve 1), 7.6 m M (2), 109 mM Ca 21 (3)). Data at 100 mM KCl (circles) and at 0m M KCl (triangles; taken from Fig. 2 in [1]) were fitted to the Hill equation (solid line) (r 2 ¼ 0.99 for 100 mM KCl and 0.99 for 0 mM KCl). The lower axis of the figure gives the free Ca 21 concentration, and the upper axis gives the number of Ca 21 bound per molecule of p26olf (nCa 21 ) calculated from the binding data of Figs 1 and 2. 6032 N. Miwa et al. (Eur. J. Biochem. 268) q FEBS 2001 (circles, Fig. 6); this indicated the lack of one of the four Ca 21 -binding sites in the mutated EF-hand. Therefore, the binding data were fitted to the Hill equation assuming the presence of three Ca 21 -binding sites (solid lines in Fig. 6). In wild-type p26olf (dashed lines), the K Ca d value was 22.3 m M, and the Hill coefficient was 2.0 (Fig. 1). In the mutants, the K Ca d values varied from 14 to 117 mM depending on the mutation (insets in Fig. 6). The result indicated that the overall Ca 21 affinity decreased in some of the mutants: the decrease was large in DEF-A and DEF-B, and small in DEF-C, and there was almost no decrease in DEF-D. One of our expectations was that the Ca 21 -binding cooperativity might be lost when the responsible EF-hand(s) was disrupted. However, this was not the case because the Hill coefficient (nH in insets) was always similar to that of wild-type p26olf (n H ¼ 2.0) (see Discussion). In the mutation studies, all of the mutants bind approxi- mately three Ca 21 . This number might be greater at a higher concentration of Ca 21 . However, as stated already, we measured Ca 21 -binding at less than 200 mM Ca 21 due to aggregation of our sample (see Experimental procedures). Ca 21 -induced conformational changes in mutant p26olf In Fig. 4, we measured the CD spectrum changes by varying the Ca 21 concentration and found that the apparent K CD d value was 8.5 mM and the Hill coefficient was 2.3. In order to examine how the Ca 21 -induced conformational changes in p26olf are affected in the mutants, we measured the CD spectra of the mutants at various Ca 21 concentra- tions. To avoid crowding, only the results at a high (200 m M) and a low (< 10 n M) concentration are shown in Fig. 7A. The overall shape of the signal of a mutant was similar to that of the wild type: the signal at both 210 nm and 222 nm increased by increasing the Ca 21 concentration. The result indicated that both the a helix and b sheet content increase by binding of Ca 21 even in the mutants. However, the content of these structures decreased somewhat in some of the mutants. The signal of DEF-A was similar to that of the wild-type, but those of other mutants were approximately 90% (DEF-C and DEF-D) and 77% (DEF-B) of the wild- type. The small signal in DEF-B was surprising, but it has been reported that a single mutation induces a signifi- cant conformational change in a calmodulin mutant [25]. Although the mechanism of this effect has not been known, similar effects of the mutation might have affected the p26olf conformation in the DEF-B mutant. The CD signal change at 222 nm was expressed as a function of the Ca 21 concentration in each mutant, and the data were fitted by the Hill equation to determine the K CD d value and the Hill coefficient (solid lines in Fig. 7B). When compared with the result in the wild-type (dashed line), the K CD d values increased greatly in mutants DEF-A and DEF-B, but the mutation effect was negligible in DEF-C and DEF-D. It should be noted that the increase in the content of a helix (222 nm signal) observed in this study was caused by binding of Ca 21 . Therefore, the shift of the K CD d value in a mutant (Fig. 7B) would coincide with the shift of the K Ca d value shown in Fig. 6. It was actually the case that both the K Ca d shift and K CD d shift were large in DEF-A and DEF-B, small in DEF-C and there were almost no shifts in DEF-D (compare the results in Figs 6 and 7B). In Fig. 4, we showed that irrespective of the K 1 con- centration, 80– 90% of the change in the a helix content was attained by binding of two Ca 21 . Interestingly, in the mutants which have only three Ca 21 -binding sites, the binding of two Ca 21 was sufficient to induce more than 70% of the change (Fig. 7B). Fig. 6. Ca 21 -binding to mutant p26olf. We measured Ca 21 -binding to four p26olf-mutants (DEF-A, DEF-B, DEF-C and DEF-D) by the flow dialysis method. The raw data (n ¼ 3) were analyzed using the Hill equation (solid lines). Dashed line represents the fitted curve of the data of wild-type p26olf obtained in Fig. 1B. The r 2 values are 0.99 (DEF-A), 0.99 (DEF-B), 0.98 (DEF-C) and 0.99 (DEF-D). Fig. 7. CD spectra of mutant p26olf. (A) CD spectra of p26olf mutants (final concentrations; 10 m M each) either in the presence of Ca 21 (< 200 mM; 1Ca 21 ), or in the absence of Ca 21 (< 10 nM; 2Ca 21 ). (B) Ca 21 -dependent changes in CD signals at 222 nm in p26olf-mutants. Data were fitted to the Hill equation (solid line). Dashed line represents the fitted data of the CD change in wild-type p26olf. The r 2 values are 0.99 (DEF-A), 0.99 (DEF-B),0.99 (DEF-C) and 0.99 (DEF-D). The lower axis of the figure gives the free Ca 21 concentration, and the upper axis gives the number of Ca 21 bound per molecule of p26olf (nCa 21 ) calculated from the binding data of Fig. 6. q FEBS 2001 Calcium-binding by p26olf (Eur. J. Biochem. 268) 6033 DISCUSSION In the present study, we showed that p26olf binds approxi- mately four Ca 21 with a K Ca d value of 22.3 mM, and a Hill coefficient of 2.0 (Fig. 1). The presence of K 1 antagonizes Ca 21 -binding to p26olf (Fig. 2), but regardless of the presence or absence of K 1 ,thesecondCa 21 -binding induces an almost complete conformational change of p26olf judging from CD changes (Fig. 4). In all of the mutants of p26olf generated, Ca 21 -binding affinities decreased but differently: the extent of the decrease was large in DEF-A and DEF-B, small in DEF-C, while the affinity of DEF-D was almost the same as that of the wild-type (Fig. 6). The Ca 21 -dependent CD change takes place in all mutants, and the K CD d value in each mutant was almost similar to the corresponding K Ca d value of the mutant (Fig. 7). EF-hand motifs in p26olf Our previous Ca 21 -binding study revealed that p26olf binds nearly four Ca 21 per molecule at 200 mM Ca 21 , which suggests that the four EF-hand motifs in p26olf are functional [2]. However, the presence of an EF-hand motif does not always mean the presence of the functional Ca 21 -binding site. Furthermore, in the primary structure of p26olf, the p26olf-specific EF-hand (EF-B and -D) has a four-amino-acid residue-insertion between E and F a helix and therefore, the actual Ca 21 -binding should be tested. In the present study, in addition to the binding of approxi- mately four Ca 21 in the wild-type p26olf, we observed nearly three Ca 21 -binding to each of the mutant that lacked one of the EF-hands, which unequivocally showed that all of the four EF-hands in p26olf bind Ca 21 . In the mutant studies, the Ca 21 -binding affinity decreased (Fig. 6). The decrease is large in DEF-A and DEF-B, and small in DEF-C and DEF-D; this indicated that the affinities of EF-A and EF-B in the N-terminal half of wild-type p26olf are comparatively high, and those of EF-C and EF-D in the C-terminal half are low (Fig. 8A). It is of interest to test whether this result is attributed to the intrinsic character of the EF-hands in the N- and C-terminal halves of p26olf or due to the interaction of the two halves. Future Ca 21 -binding studies using the N-terminal half and the C-terminal half of p26olf would give us a clue to solve this issue. Order of Ca 21 -binding to p26olf Generally, it is not easy to determine the order of Ca 21 - binding in a protein that has as many as four Ca 21 -binding sites such as p26olf. In the following, however, we will try to determine the order of Ca 21 -binding to p26olf under the assumptions shown below. In the present study, we analyzed Ca 21 -binding to p26olf with the Adair equation (Fig. 3). In this equation, the nth Adair dissociation constant is defined on the basis of the number (n ) of the ligand bound to a receptor. Therefore, for example, when the first Ca 21 -binding takes place at a certain EF-hand and then the second binding takes place at two other EF-hands simultaneously, the constant K 2 represents an overall dissociation constant of this multiple second bindings. However, it is highly possible that one of the dissociation constants in this multiple binding is lower than the other. In this case, K 2 represents mainly the binding to this rather specific site (second site). Similar situation can be assumed for K 3 and K 4 . Alternatively, if Ca 21 -binding takes place sequentially in an ordered manner, the nth Adair dissociation constant represents the nth binding of Ca 21 to a specific EF-hand. With these sorts of mechanisms in mind, we assumed that the nth Adair constant represents the nth binding to a specific EF-hand in p26olf. In the presence and absence of K 1 , we measured Ca 21 -binding to p26olf (Fig. 2) and calculated the affinity for K 1 in each of the four binding steps (Fig. 3). Haiech et al. [18] have suggested that the affinity for K 1 (k in Fig. 3) is dependent on the number of the carboxyl groups within an EF-hand of calmodulin. Their idea was that the more carboxyl groups are present, the higher the affinity for K 1 is. As a result, the antagonistic effect of K 1 is expected to be higher at the EF-hand loop having more carboxyl groups. The numbers of the carboxyl groups in the EF-hands in p26olf are two (in EF-A), six (EF-B), three (EF-C) and four (EF-D) and therefore the order of the affinity for K 1 is (from high to low) probably EF-B, -D, -C and -A. As our best estimate of the relation among dissociation constants for K 1 was k 1 ø k 4 ,, k 2 ø k 3 (see Results), the order of the affinity for K 1 is (from high to low) steps 1 and 4 (there were no clear differences between these two steps), and steps 2 and 3 (again, no clear differences were observed between these two steps). Therefore, most probably EF-B and -D are responsible for the first and the fourth binding of Ca 21 , and EF-A and EF-C for the second and the third binding. As the affinity for Ca 21 is suggested to be higher in EF-B than EF-D (Fig. 8A), the first step probably takes place in EF-B and the fourth step in EF-D. Similar consideration led us to suggest that the second binding takes place in EF-A (the site showing high affinity for Ca 21 ) and that the third binding in EF-C (low affinity for Ca 21 ). Mechanism of cooperative Ca 21 -binding to p26olf Of the four EF-hands in p26olf, EF-A and -C are S100-specific and potentially show low affinity for Ca 21 (200–500 mM [13]). In our present study, however, none of Fig. 8. Possible model of Ca 21 -binding in p26olf. (A) Affinity of each EF-hand in Ca 21 -binding deduced from the mutation study. (B) A possible simplified scheme of Ca 21 -binding to p26olf. 6034 N. Miwa et al. (Eur. J. Biochem. 268) q FEBS 2001 the dissociation constants measured showed this low affinity, which suggests the presence of cooperativity in Ca 21 -binding to p26olf. We will try to explain the mechanism of the cooperative Ca 21 -binding, based on the presumed affinities of S100- and p26olf-specific EF-hands. Although the Ca 21 -binding affinity to the p26olf-specific EF-hand is not known yet, based on the similarity of the amino-acid sequence, we assumed that this EF-hand shows high affinity for Ca 21 (see below). As many previous studies reported by others were performed in the absence of K 1 , we will use our result measured in the absence of K 1 for direct comparison. As discussed above, we already suggested that the order of Ca 21 -binding to p26olf is EF-B, EF-A, EF-C and EF-D. In these EF-hands, EF-B and -D are the p26olf-specific EF-hands and probably show high affinity for Ca 21 -binding, while EF-A and EF-C are S100-specific and are expected to show low affinity. From our consideration, the first binding of Ca 21 takes place at EF-B with high affinity (K 1 ¼ 21 mM; Fig. 3). The dissociation constants of the following steps were 5.7 m M (second step dissociation constant K 2 at EF-A), 17 m M (K 3 at EF-C) and 18 mM (K 4 at EF-D). From these dissociation constants, we speculate that the mechanism of Ca 21 -binding to p26olf is as follows (see Fig. 8B). (a) The first Ca 21 -binding to a high affinity site, EF-B, induces a conformational change of p26olf, which probably increases the affinity for Ca 21 of EF-A to result in the second Ca 21 -binding to EF-A. It remained uncertain whether the binding of Ca 21 to EF-B contributes to this increase in the Ca 21 -affinity in EF-A directly or through the interaction with the C-terminal half of p26olf. The major confor- mational changes (increase in the a helix content measured with CD) complete at this stage (Fig. 4). (b) The major conformational changes induced by the second Ca 21 -binding increase the affinity of EF-C to induce the third Ca 21 -binding. (c) Finally, the fourth Ca 21 -binding takes place at EF-D with high affinity. Although we assumed the presence of sequential binding of Ca 21 in the above, the mechanism may not be so simple. The reason for this is that, in all of the mutants generated, we observed three Ca 21 -binding with a positive cooperativity (Fig. 6), which cannot be explained by a simple ordered sequential binding mechanism. It is possible therefore that, in the wild-type p26olf, most of Ca 21 binds to EF-B first to induce a second cooperative binding to EF-A, but that this process is not exclusive. If this is the case, in DEF-A mutant for example, Ca 21 firstly binds to EF-B and then also EF-C (third binding site) with a positive cooperativity. In this case, our suggestion above shows the order of Ca 21 -binding of the major population in wild-type p26olf. Alternatively, another explanation for the mutant study is possible. It may be the case that the mutation from glutamate to glutamine in the mutants induced a similar EF-hand conformational change that occurs on the binding of Ca 21 in the wild-type. If this is the case, Ca 21 -binding in the wild-type can be sequential and cooperative throughout the course of the binding as suggested above, and similar behavior can be observed in the mutant. Apparently, further studies are required to understand the actual mechanism. S100 proteins are known to be functional in the form of a homodimer or a heterodimer. Although S100-specific low affinity sites are present in an S100 dimer, the Ca 21 -binding experiment showed that the calculated Adair dissociation constants are in a range of 10– 100 m M [16,17]. The loss of the low affinity site (i.e. 200–500 m M) in the Ca 21 -binding experiment in S100 dimers would arise from the cooperative mechanism that was found in p26olf and suggested above. ACKNOWLEGEMENTS This work was supported by a Grant-in-Aid (13780636) from the Ministry of Education, Culture, Sports, Science and Technology of Japan to N. M., and Research for the Future Program of Japan Society for the Promotion of Science under the Project ‘Cell Signaling (JSPS- RFTF97L00301)’ to S. K. REFERENCES 1. Miwa, N., Kobayashi, M., Takamatsu, K. & Kawamura, S. (1998) Purification and molecular cloning of a novel calcium-binding protein, p26olf, in the frog olfactory epithelium. Biochem. Biophys. Res. Commun. 251, 860–867. 2. Miwa, N., Uebi, T. & Kawamura, S. (2000) Characterization of p26olf, a novel calcium-binding protein in the frog olfactory epithelium. J. Biol. Chem. 275, 27245–27249. 3. Tanaka, T., Miwa, N., Kawamura, S., Soma, H., Nitta, K. & Matsushima, N. (1999) Molecular modeling of single polypeptide chain of calcium-binding protein p26olf from dimeric S100B (bb). Protein Eng. 12, 395– 405. 4. Calabretta, B., Battini, R., Kaczmarek, L., de Riel, J.K. & Baserga, R. (1986) Molecular cloning of the cDNA for a growth factor- inducible gene with strong homology to S-100, a calcium-binding protein. J. Biol. Chem. 261, 12628–12632. 5. Baudier, J., Delphin, C., Grunwald, D., Khochibin, S. & Lawrence, J.J. (1992) Characterization of the tumor suppressor protein p53 as a protein kinase C substrate and a S100b-binding protein. Proc. Natl Acad. Sci. USA 89, 11627–11631. 6. Lagasse, E. & Clerc, R.G. (1988) Cloning and expression of two human genes encoding calcium-binding proteins that are regulated during myeloid differentiation. Mol. Cell. Biol. 8, 2402–2410. 7. Kato, K., Suzuki, F. & Ogasawara, N. (1988) Induction of S100 protein in 3T3-L1 cells during differentiation to adipocytes and its liberating by lipolitic hormones. Eur. J. Biochem. 177, 461–466. 8. Patel, J. & Marangos, P.J. (1982) Modulation of brain protein phosphorylation by the S-100 protein. Biochem. Biophys. Res. Commun. 109, 1089–1093. 9. Ilg, E.C., Schafer, B.W. & Heizmann, C.W. (1996) Expression pattern of S100 calcium-binding proteins in human tumors. Int. J. Cancer 68, 325–332. 10. Heizmann, C.W. & Cox, J.A. (1998) New perspectives on S100 proteins: a multifunctional Ca 21 -, Zn 21 - and Cu 21 -binding protein family. Biometals 11, 383–397. 11. Donato. R. (1999) Functional roles of S100 proteins, calcium- binding proteins of the EF-hand type. Biochim. Biophys. Acta. 1450, 191–231. 12. Kretsinger, R.H. (1997) EF-hands embrace. Nat. Struct. Biol. 4, 514–516. 13. Baudier, J. & Cole, R.D. (1989) The Ca 21 -binding sequence in bovine brain S100b protein b-subunit. Biochem. J. 264, 79 –85. 14. Isobe, T., Ishioka, N. & Okuyama, T. (1981) Structural relation of two S-100 proteins in bovine brain; subunit composition of S-100a protein. Eur. J. Biochem. 115, 469–474. 15. Potts, B.C., Smith, J., Akke, M., Macke, T., Okazaki, K., Hidaka, H., Case, D.A. & Chazin, W.J. (1995) The structure of calcyclin reveals a novel homodimeric fold for S100 Ca 21 -binding proteins. Nat. Struct. Biol. 2, 790–796. q FEBS 2001 Calcium-binding by p26olf (Eur. J. Biochem. 268) 6035 16. Baudier, J., Glasser, N. & Gerard, D. (1986) Ions binding to S100 proteins. J. Biol. Chem. 261, 8192–8203. 17. Franz, C., Durussel, I., Cox, J.A., Schafer, B.W. & Heizmann, C.W. (1998) Binding of Ca 21 and Zn 21 to human nuclear S100A2 and mutant proteins. J. Biol. Chem. 273, 18826–18834. 18. Haiech, J., Klee, C.B. & Demaille, J. (1981) Effects of cations on affinity of calmodulin for calcium: ordered binding of calcium ions allows the specific activation of calmodulin-stimulated enzymes. Biochemistry 20, 3890– 3897. 19. Hill, A.V. (1910) The possible effects of the aggregation of the molecules of hæmoglobin on its dissociation curves. J. Physiol. 40, 4–7. 20. Dahlquist, F.W. (1978) The meaning of Scatchard and Hill plots. Methods Enzymol. 48, 270–299. 21. Adair, G.S. (1925) The hemoglobin system. J. Biol. Chem. 63, 529–545. 22. Tachibanaki, S., Nanda, K., Sasaki, K., Ozaki, K. & Kawamura, S. (2000) Amino acid residues of S-modulin responsible for inter- action with rhodopsin kinase. J. Biol. Chem. 275, 3313–3319. 23. Gilli, R., Lafitte, D., Lopez, C., Kilhoffer, M C., Makarov. A., Briand, C. & Haiech, J. (1998) Thermodynamic analysis of calcium and magnesium binding to calmodulin. Biochemistry 37, 5450– 5456. 24. Babu, Y.S., Bugg, C.E. & Cook, W.J. (1988) Structure of calmodulin refined at 2.2 A ˚ resolution. J. Mol. Biol. 204, 191–204. 25. Maune, J.F., Klee, C.B. & Beckingham, K. (1992) Ca 21 binding and conformational change in two series of point mutations to the individual Ca 21 -binding sites of calmodulin. J. Biol. Chem. 267, 5286–5295. 26. Babu, A., Su, H., Ryu, Y. & Gulati, J. (1992) Determination of residue specificity in the EF-hand of troponin C for Ca 21 coordination, by genetic engineering. J. Biol. Chem. 267, 15474–15496. 6036 N. Miwa et al. (Eur. J. Biochem. 268) q FEBS 2001 . Calcium-binding by p26olf, an S100-like protein in the frog olfactory epithelium Naofumi Miwa, Yukiko Shinmyo and Satoru Kawamura From the Department. 4). (b) The major conformational changes induced by the second Ca 21 -binding increase the affinity of EF-C to induce the third Ca 21 -binding. (c) Finally, the