The Role of Calcium-binding Sites in S-modulin Function*

S-modulin controls rhodopsin phosphorylation in a calcium-dependent manner, and it has been suggested that it modulates the light sensitivity of the photoreceptor cell. S-modulin binds to the ROS membrane at high Ca2+concentration, and N-terminal myristoylation is necessary for this property (the calcium-myristoyl switch). S-modulin has four EF-hand motifs, of which two (EF-2 and -3) are functional. Here, we report on the roles of EF-2 and -3 in S-modulin function (calcium binding, membrane association, and inhibition of rhodopsin phosphorylation) by site-directed mutants (E85M and E121M). Surprisingly, E121M, which has a mutation in EF-3, neither binds Ca2+ nor inhibits phosphorylation. In contrast, E85M binds one Ca2+ and has the same membrane affinity as wild-type S-modulin, but has lost the ability to inhibit rhodopsin phosphorylation. It is suggested that the binding of Ca2+to EF-3 is probably required for EF-2 to be a functional Ca2+-binding site and to induce exposure of the myristoyl group; and that the binding of Ca2+ to EF-2 is important for the interaction with rhodopsin kinase.

In vertebrate rod photoreceptors, cGMP-gated cation channels are opened in the dark-adapted state (1,2). Light activates rhodopsin and triggers the phototransduction cascade, which results in the closure of cation channels in the rod outer segment (ROS) 1 and blocks the influx of Ca 2ϩ . As intracellular Ca 2ϩ is continuously pumped out by a Na ϩ -K ϩ /Ca 2ϩ exchanger in the outer segment (3,4), the cytoplasmic Ca 2ϩ concentration decreases in light-adapted photoreceptors. This decrease of Ca 2ϩ concentration is the underlying mechanism of light adaptation of vertebrate photoreceptors (5,6).
Phosphorylation of rhodopsin plays a role in shutting off the activation of transducin (7,8), and the efficiency of phosphorylation is regulated, in a Ca 2ϩ -dependent manner, by a Ca 2ϩbinding protein, S-modulin, in frogs (9,10) or recoverin, its bovine homologue (11). At high Ca 2ϩ concentrations (darkadapted state), S-modulin inhibits phosphorylation of lightactivated rhodopsin, but does not interfere at low Ca 2ϩ concentrations (light-adapted state). Therefore, S-modulin and recoverin contributes to increased light sensitivity in the darkadapted state.
S-modulin and recoverin are Ca 2ϩ -binding proteins which contain covalently attached fatty acyl groups at their N terminus (12). The binding of Ca 2ϩ to these proteins induces exposure of the fatty acyl groups, which enables them to associate with ROS membrane (13,14). This property is the so-called "Ca 2ϩ -myristoyl switch" (13). There are four EF-hand motifs in S-modulin and recoverin, but only two of them (EF-2 and -3) are thought to be able to bind Ca 2ϩ (15). Therefore, S-modulin function (inhibition of frog rhodopsin phosphorylation) is mediated by the binding of Ca 2ϩ ions to EF-2 and -3.
We made site-directed mutants of S-modulin that lack Ca 2ϩ binding ability in their EF-2 or -3. The present study describes the Ca 2ϩ binding properties, membrane-association, and inhibitory effects on rhodopsin phosphorylation of wild-type S-modulin and these mutants. The results suggest that EF-3 first binds Ca 2ϩ , which enables S-modulin to associate with ROS membrane and to bind Ca 2ϩ at EF-2. Subsequent EF-2 binding of Ca 2ϩ ions probably causes a conformational change permitting interaction of S-modulin with rhodopsin kinase, which inhibits rhodopsin phosphorylation.
Expression and Purification of Recombinant Proteins-The procedures for expression and purification of recombinants follow Hisatomi et al. (16). Briefly, the expression vectors, pET-Smd (16), pET-E85M, and pET-E121M were transfected to Escherichia coli BL21DE3 (Novagen) with (ϩmyr) or without (Ϫmyr) pBB131, an expression vector of N-myristoyltransferase. The recombinant proteins were expressed by the addition of 1 mM isopropyl-␤-D-thiogalactopyranoside. Recombinant proteins were solubilized in 8 M urea buffer, refolded by dialysis, and applied to a DEAE-Sephadex column. The fraction containing recombinant S-modulins was dialyzed against the buffer containing 5 mM Ca 2ϩ and applied to a phenyl-Sepharose column. Recombinant S-modulins were eluted with 5 mM EGTA.
HPLC-Purified recombinant proteins were injected onto a reversephase C-18 column. Recombinant S-modulin was eluted with a linear gradient of 0 -80% acetonitrile (1.3%/min) in 0.1% trifluoroacetic acid at a flow rate of 1.5 ml/min, monitoring absorbance of the eluate at 280 nm.
Ca 2ϩ Binding Assay-Binding of Ca 2ϩ ions to S-modulin or mutant proteins was evaluated by ultrafiltration (17). Purified proteins were extensively dialyzed against 25 mM Tris-HCl (pH 8.0) to remove EGTA and calcium, then 20 mol of each calcium-free protein in 1 ml of 25 mM Tris-HCl (pH 8.0) were placed in a Centricon-10 concentrator (Amicon). 1 ml of 0.2 mM CaCl 2 solution in 25 mM Tris-HCl (pH 8.0) was then added, and the solution was thoroughly mixed. The calcium-protein mixtures were then centrifuged, and the amounts of calcium in the filtrated fractions were measured by atomic absorption (Shimazu AA-660).
Tryptophan Emission Spectrum-Spectroscopic measurement was carried out as described by Hisatomi et al. (16). Briefly, fluorescence emission spectra were recorded from 300 to 400 nm with a fluorescence spectrophotometer (Hitachi, F-4500) at an excitation wavelength of 290 nm, in a mixture containing 2 M recombinant protein, 100 mM KCl, 5 mM 2-mercaptoethanol, 1 mM EGTA, and 100 mM HEPES (pH 7.0). The free Ca 2ϩ concentration was adjusted by adding 1 M CaCl 2 .
Ca 2ϩ -dependent Membrane Association of Recombinant Proteins-Frog ROS were isolated by flotation with 45% sucrose in gluconate buffer (40 mM potassium gluconate, 2.5 mM KCl, 2 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EGTA, and 10 mM HEPES, pH 7.5), and washed with gluconate buffer containing 4 M urea to eliminate endogenous S-modulin, s26 (cone homologue of S-modulin) and other peripheral proteins. Urea-stripped ROS membranes were mixed with gluconate buffer containing 1% bovine serum albumin to prevent nonspecific binding of the recombinant proteins to the ROS membrane and tube. After washing with gluconate buffer containing various concentrations of Ca 2ϩ (Ca 2ϩ gluconate buffer), the ROS membranes were resuspended in Ca 2ϩ gluconate buffer containing recombinant proteins (120 pmol). The mixtures were incubated at room temperature for 30 min, and the soluble and membrane fractions after centrifugation (37,000 ϫ g for 5 min) were analyzed by SDS-polyacrylamide gel electrophoresis. The integrated densities of Coomassie Brilliant Blue-stained bands of the recombinant proteins were quantified by a two-dimensional densitometer (The Discovery Series, pdi Inc.).
Phosphorylation Assay-Phosphorylation of rhodopsin was measured by the methods of Kawamura (9) and Sanada et al. (18). For the phosphorylation assay, ROS were isolated in phosphorylation buffer (115 mM potassium gluconate, 2.5 mM KCl, 2 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EGTA, 10 mM HEPES, pH 7.5) in complete darkness, and washed with the buffer to eliminate endogenous S-modulin, s26, and ATP. The reaction was carried out in 25 l of the mixture containing 10 M (final concentration) rhodopsin and various concentrations of Smodulin and/or its mutants in phosphorylation buffer. The free calcium concentration in the mixture was adjusted by adding 1 M CaCl 2 solution. The reaction mixtures were exposed to light for 2 min, and the reaction was initiated by addition of a mixture of ATP (0.   After 2 min of incubation at room temperature, the reaction was terminated by adding 150 l of 10% trichloroacetic acid. After centrifugation (10,000 ϫ g for 5 min) of the reaction mixture, the precipitates were washed with 500 l of phosphorylation buffer and subjected to SDSpolyacrylamide gel electrophoresis. The amount of 32 P incorporated into rhodopsin was quantified by using an image analyzer (BAS 2000, Fuji Film).

Myristoylation of Recombinant S-modulin-It has been es-
tablished that the conserved glutamic acid at position 12 of the Ca 2ϩ -binding loop is important for coordinating Ca 2ϩ (19,20). To investigate the role of each Ca 2ϩ -binding site, EF-2 or EF-3 was inactivated by replacing glutamic acid with the hydrophobic amino acid, methionine. Myristoylated wild-type (ϩmyr) and mutant S-modulins, E85M (ϩmyr) and E121M (ϩmyr), and unmyristoylated wild-type (Ϫmyr) were expressed and purified as described in the experimental procedures. Myristoylated recombinant proteins eluted from a C-18 column at almost the same retention time (Fig. 1, b, c, and d), which is longer than that of wild-type (Ϫmyr) (Fig. 1a) and suggests that these recombinants expressed with N-myrstoyltransferase are in fact myristoylated.
Ca 2ϩ Binding of S-modulin and Mutants-The number of Ca 2ϩ ions bound to each of these proteins was quantified in the presence of 0.1 mM Ca 2ϩ (Table I). As expected, wild-type (ϩmyr) and E85M (ϩmyr) bind two and one Ca 2ϩ per a molecule, respectively. On the other hand, E121M (ϩmyr) can not bind Ca 2ϩ . These results indicate that Ca 2ϩ binding to EF-3 is necessary for EF-2 to bind Ca 2ϩ at physiological Ca 2ϩ concentrations. It has been reported that EF-3 has the conformation of a classic EF-hand, but EF-2 is rather different (21). The conformational change induced by Ca 2ϩ binding to EF-3 may be required to raise the Ca 2ϩ affinity of EF-2, in order for EF-2 to become functional.
Fluorescence Properties of S-modulin and Mutants-E85M (ϩmyr) and E121M (ϩmyr) can be purified in the same way as for wild-type S-modulin, so it seems that the structure of Smodulin is not largely disrupted by the mutagenesis. Fig. 2 shows the tryptophan emission spectra of the wild-type and mutant S-modulins. Wild-type (ϩmyr), E85M (ϩmyr) and E121M (ϩmyr) showed almost the same spectrum in the presence of 1 nM Ca 2ϩ (Fig. 2, upper panel), but different from wild-type (Ϫmyr). This suggests that E86M (ϩmyr) and E121M (ϩmyr) are in fact myristoylated in a similar way to the wildtype (ϩmyr), and that the mutations of glutamic acid to methionine in EF-2 and EF-3 do not significantly change the environment of the three tryptophan residues in the Ca 2ϩ -free form of these proteins. However, the emission spectra of mutants were different from that of wild-type at a concentration of 0.1 mM Ca 2ϩ (Fig. 2, lower panel). The spectrum of the wild-type is red-shifted by an increasing Ca 2ϩ concentration (16); that of E85M (ϩmyr) shows a smaller red shift; and that of E121M (ϩmyr) was not affected by Ca 2ϩ concentration. The red shift observed in E85M (ϩmyr), which can bind one Ca 2ϩ ion per molecule, is probably caused by the binding of Ca 2ϩ to EF-3.
Membrane Association of Wild-type, E85M, and E121M Smodulins-Each myristoylated recombinant, (wild-type (ϩmyr), E85M (ϩmyr), or E121M (ϩmyr)) was mixed with urea-stripped ROS membranes at various Ca 2ϩ concentrations and separated by centrifugation into membrane and soluble fractions. The soluble fraction (containing proteins free from ROS membranes) and the membrane fractions (containing proteins bound to ROS membranes) were subjected to SDS-polyacrylamide gel electrophoresis. The densities of Coomassie Brilliant Blue-stained bands of the wild-type and mutant Smodulins were analyzed quantitatively, and the ratio of membrane-bound protein, (membrane fraction)/(membrane ϩ soluble fraction), was plotted against Ca 2ϩ concentration (Fig. 3). This shows that E85M (ϩmyr) has almost the same membrane affinity as the wild-type. As exposure of the myristoyl group is essential for membrane binding (13,14), our results suggest that binding of Ca 2ϩ to EF-3 induces exposure of the myristoyl group. It has been reported that ejection of the myristoyl group is required for rotation at Gly-42, unclamping of the myristoyl group, and melting of part of the N-terminal helix (21). Ca 2ϩ binding to EF-3 may induce these changes until the level necessary for membrane association.
As E121M (ϩmyr) can not bind Ca 2ϩ at concentrations less than 0.1 mM Ca 2ϩ , it can hardly bind to the ROS membrane at all, although in the presence of 1 mM Ca 2ϩ , E121M (ϩmyr) does bind slightly. This may be explained by the binding of Ca 2ϩ to EF-2 or inactivated EF-3 at very high (more than 1 mM) Ca 2ϩ concentration, which is well above normal physiological Ca 2ϩ concentrations. Fig. 4 shows the incorporation of 32 P-labeled phosphatic acid into rhodopsin in the presence of various concentrations of wild-type (ϩmyr), E85M (ϩmyr), or E121M (ϩmyr) S-modulins. Wild-type (ϩmyr) inhibits rhodopsin phosphorylation at a high (0.1 mM) Ca 2ϩ concentration, but neither E85M (ϩmyr) nor E121M (ϩmyr) can inhibit rhodopsin phosphorylation even at high Ca 2ϩ concentrations. This suggests that Ca 2ϩ binding to EF-2 is important for the inhibitory activity of rhodopsin phosphorylation.

Inhibition of Rhodopsin Phosphorylation by Wild-type or Mutant S-modulins-
To investigate whether E85M (ϩmyr) can interact with rhodopsin kinase or not, its effect on rhodopsin phosphorylation by wild-type (ϩmyr) was investigated. Fig. 5 shows the incorporation of 32 P-labeled phosphatic acid into rhodopsin plotted against the concentration of E85M (ϩmyr) in the presence of wild-type (ϩmyr). The inhibitory effect of wild-type (ϩmyr) is not affected by E85M (ϩmyr), indicating that E85M (ϩmyr) does not compete with wild-type (ϩmyr) for binding with rhodopsin kinase. This strongly suggests that the Ca 2ϩ binding to EF-2 is important for recognition of rhodopsin kinase.

DISCUSSION
Ca 2ϩ Binding Cooperativity-Ames et al. (23) reported that unmyristoylated recoverin exhibits heterogeneous and uncooperative binding of two Ca 2ϩ ions, but these two Ca 2ϩ ions bind cooperatively to myristoylated recoverin with a Hill coefficient of 1.75. One explanation of this difference between myristoylated and unmyristoylated recoverins is that the myristoyl group accommodated within the protein moiety may reduce the Ca 2ϩ affinity of EF-3 to a level lower than that of the EF-2 of unmyrisotylated S-modulin. When Ca 2ϩ concentration is high, EF-3 binds Ca 2ϩ , which causes the myristoyl group to be exposed and EF-2 to be functional, upon which it binds Ca 2ϩ .
The Hydrophobic Region Exposed by Ca 2ϩ Binding-As with myristoylated wild-type (ϩmyr), unmyristoylated wild-type (Ϫmyr) binds to a phenyl-Sepharose column in a Ca 2ϩ -dependent manner. This suggests that a hydrophobic region of the protein moiety, in addition to the myristoyl group, is exposed by binding of Ca 2ϩ (13). In our preliminary experiments, both unmyrstoylated E85M (Ϫmyr) and E121M (Ϫmyr) lose affinity for phenyl-Sepharose. We conclude that exposure of the hydrophobic region is probably caused by Ca 2ϩ binding at EF-2.
Structural Changes of S-modulin Induced by Ca 2ϩ Binding-The upper part of Fig. 6 illustrates the conformational changes of S-modulin deduced from our present analysis. The lower part of Fig. 6 represents the corresponding three-dimensional structure of bovine recoverin during Ca 2ϩ binding (15,21,22). The structure of N-terminal region in the single Ca 2ϩbound form is different from that in the Ca 2ϩ -free form but similar to the double Ca 2ϩ -bound form. It is consistent with our model that the conformational change induced by the Ca 2ϩ binding to the EF-3 (shown in blue) may expose the N-terminal myristoyl group. The structure of EF-2 (shown in red) in the single Ca 2ϩ -bound form is also largely different from that in the Ca 2ϩ -free form. This difference is probably important for EF-2 to be a functional Ca 2ϩ -binding site.
One of the largest conformational differences is shown at the region corresponding to amino acids from 180 to 186 (shown in green) between single Ca 2ϩ -bound and double Ca 2ϩ -bound forms. In the Ca 2ϩ -free and single Ca 2ϩ -bound forms, this region appears to form an ␣-helix, which undergoes a conformational change to a random coil when Ca 2ϩ to EF-2. The conformational change of this region may induce an interaction with rhodopsin kinase.