Calcium-induced Shape Change of Calmodulin with Mastoparan Studied by Solution X-ray Scattering*

Solution x-ray scattering using synchrotron radia- tion as an x-ray source was used to analyze the Ca2+-dependent shape change of pig brain calmodulin in detail. The radius of gyration of calmodulin at 10 mg/ ml was increased by 0.9 A. The increase was nearly completed when 2.5 mol of Ca2+/mo1 of calmodulin was added, whereas the radius of gyration qf calmodulin with mastoparan decreased by about 3 A with an in- creasing Ca2+ concentration up to 4 mol of Ca2+/mo1 of calmodulin. At a moderate angle of region, both scat- tering profiles from calmodulin with or without Ca” displayed clear humps at s = 0.03 A” which are char- acteristic of a dumbbell structure. However, in the presence of mastoparan, the hump in the scattering profile became obscure and later disappeared with the third and fourth Ca2+ binding to calmodulin. These findings are attributable to the Ca2+-induced shape change of calmodulin with mastoparan from a dumbbell structure to a non-dumbbell structure in which the distance between the two lobes of calmodulin calculated so1

the shape-related parameters of the protein molecule in the solution. The scattering profile is free from smearing and its high signal/noise ratio enabled us to directly compare the experimental scattering profile with the theoretical calculation. The SR method makes it possible to investigate domain topology of the protein in solution directly because of its high quality of data at a moderate angle region of the scattering. We applied this technique to study the shape change of calmodulin induced by Ca2+ binding.
Since calmodulin is a ubiquitous protein in eukaryotic cells and regulates many cellular processes in a Ca2+-dependent manner (l), Ca2+-induced conformational changes of the protein have been studied extensively by conventional methods including CD, fluorescence, and NMR (2). In those studies it was noted that the changes of secondary structure of the calmodulin induced by calcium were almost completed by the binding of the first two Ca", although the third and fourth * This work was supported by Photon Factory Advisory Committee Proposal No. 87-082, Tsukuba, Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom all correspondence should be addressed.
The abbreviations used are: SR, synchrotron radiation; SOXS, solution x-ray scattering.
Ca2+ binding were needed to activate the target enzymes, such as phosphodiesterase (3) and myosin light chain kinase ( 4 5 ) .
Recently, an NMR study showed that the local conformational change of calmodulin with mastoparan, one of the model peptides as a target protein of calmodulin (6), occurred by the third and fourth Ca2+ binding. However, this was not the case with mastoparan (5), suggesting that the Ca2+-induced conformational change of calmodulin bound on a target protein might be different from that of target protein-free calmodulin. Mastoparan consists of 14 amino acid residues of INLKALAALAKKIL in the sequence, and the binding of 1 mol of mastoparan/mol of calmodulin with a high affinity (Kd 0.3 WM) should be essential for a physiological function of calmodulin (6). Furthermore, the molecular weight of mastoparan is so small that it would not perturb the x-ray scattering profile from the calmodulin molecule. Therefore, mastoparan is the most suitable model peptide as a target protein of calmodulin on the SOXS study.
The molecular shape of calmodulin-calcium complex in its crystalline state takes on a dumbbell shape in which the Nterminal half (N-lobe) is connected with the C-terminal half (C-lobe) by a central a-helix (7). Using x-ray scattering measurements, two studies have shown that the radius of gyration (R,) of Ca2+-saturated calmodulin was about 1 8, larger than that of Ca2+-free calmodulin. However, there is disagreement in the results whether the calmodulin in solution is still a dumbbell structure or not (8,9).
In this paper, the calcium-dependent change of the R, of calmodulin with or without mastoparan was investigated in detail. The high quality of the SOXS data at a moderate angle provided us directly with information about the correlations between the two lobes of calmodulin. We propose that the shape of calmodulin differs greatly from that of a dumbbell structure with the increasing of calcium binding in the presence of mastoparan.

MATERIALS AND METHODS
Calmodulin was prepared from frozen pig brain. The supernatant of the pig brain homogenate in 20 mM phosphate buffer, pH 7.6, 126 mM NaCl, 5 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride was fractionated with ammonium sulfate. The protein precipitate of 35-70% ammonium sulfate was adjusted to 10 mM phosphate buffer, pH 6.0, 1 mM EDTA and applied to a phosphocellulose column equilibrated under the same conditions. Calmodulin was obtained from the flow-through of the column according to the methods of Yazawa et al. (IO), and it was purified by phenyl-Sepharose (Pharmacia LKB Biotechnology Inc.) column chromatography ( l l ) , followed by extensive dialysis against EDTA to remove Ca2+. Finally, the medium conditions were adjusted to 50 mM Tris-HC1, pH 7.6, and 10 p M EDTA. The first step of phosphocellulose column was effective in excluding the contamination of low molecular weight proteins in calmodulin preparation. The purity of calmodulin was checked by sodium dodecyl sulfate and/or 8 M urea-polyacrylamide gel electrophoresis (12, 13) as a single protein band, and the concentration was determined by the method of Lowry et al. (14). Mastoparan was purchased from the Peptide Institute Co. (Osaka, Japan) and used after an adjustment of pH of the solution.
The profile of SOXS using SR was obtained by the instrument BL-1OC which was installed at the Photon Factory in the National Laboratory for High Energy Physics at Tsukuba, Japan. Since the optics of SOXS is the point focusing type, the intensity data were used without slit correction. The details of the optics and instruments are given elsewhere (15). The data treatment was handled by the HITAC M682H computer of the Computer Center, Hokkaido University, Japan. SOXS profiles were obtained with 10-30 mg/ml calmodulin exposure for 20-30 min, depending on the protein concentrations, using a quartz cell of 70 pl in volume. The temperature of the scattering experiment was kept at 25 "C by circulating water through the cell holder. The medium conditions of scattering measurements were 50 mM Tris-HC1, pH 7.6,120 mM NaC1, and the molar ratios of Ca2+ denoted in this paper were added molar ratios of Ca2+/ mol of calmodulin, and Ca2+-free calmodulin was in 1 mM EDTA.
The scattering data at a very small angle region were analyzed by the Guinier method of approximation to calculate R, (16) In this paper, s is the reciprocal coordinate, 2sin8/X (28, scattering angle; X, wavelength of x-ray used, 1.448 A). Using the SR radiation as an x-ray source, we could obtain the scattering profiles of calmodulin in high quality at a moderate angle region which enables us to give direct information about domain topologies. In fact, the scattering profiles from troponin C showed a hump at about s = 0.027 A-' which relates principally to the distance between the two lobes (17). Thus, the measurement of the hump can be used as a criterion for dumbbell-shaped calmodulin. The theoretical analysis predicts that scattering profiles from a molecule consisted of two well separate domains display two minima but no zeros  (18). Since calmodulin consists of two quite separate lobes, it is reasonable to analyze the size of the lobe by the theory described above on the assumption that the two lobes of calmodulin were not too much different from that of a sphere.

RESULTS
Guinier plots (In Z(s) uersus s2) of the data obtained with calmodulin at various Ca2+ concentrations are shown in Fig.  1. The Guinier plots show high quality data, and there is no evidence of any upward curvature at low s2 value, indicating the monodispersity of the calmodulin molecule. The straight lines in the figure were obtained with data points between s2 = 1.90 x and s2 = 1.53 x using the least squares method. The R, value of Ca2+-free calmodulin calculated from the slope of the Guinier plot was 19.9 A. Wheno Ca2+ was added, as shown in Fig. 2 1.01, 2.02, and 4.04 molar excess of Ca2+, respectively. After measurement of trace 4 in panel A , EDTA was added to the sample as 7.7 mM of the final concentration, then trace 5 was obtained (the concentration of calmodulin in this case was 28 mg/ ml). In panel A , the double-headed arrow and two arrows show the larger than that without mastoparan, indicating that mastoparan might bind to calmodulin even without Ca2+.
Scattering profiles from calmodulin solution with different amounts of Ca2+ at moderate angles are shown in Fig. 3. As shown in panel A of Fig. 3, the charactristic features of all the scattering pJofiles without mastoparan are humps at around s = 0.03 A" and broad peaks which are characterized by two miaima at s = 0.045 and s = 0.079 A-l apd peaks at s = 0.056 A-'. The peak height at s = 0.056 A" increases depending on the Ca2+ concentration. These changes are almost complete upon the addition of 2 mol of Ca2+/mo1 of calmodulin and are reversible because trace 5 can be superimposed on trace 1. In the presence of equimolar mastoparan as shpwn in panel B of Fig. 3, however, the humps at around 0.03 A" decreased and disappeared in traces 3 and 4, although the humps could still be seen in traces 1 and 2. Panel B of Fig. 3 also shows that the broad peak at about s = 0.056 A-l becomes obscure in traces 3 and 4, whereas the peak still can be perceived in traces 1 and 2.

DISCUSSION
All the scattering profiles of calmodulin without mastoparan clearly display the humps at around s = 0.03 A-1 (panel 4 of Fig. 3). Since the s value of 0.03 A" is equivalent to 33 A of Bragg spacing (s = l/d, d, Bragg spacing), the hump would be a characteristic profile of a dumbbell structure in which the N-lobe correlates with the C-lobe at a distance of 33 A. The value of 33 A seems to be reasonable with the value estimated from the crystal structure of calmodulin ( 7 ) . In the two domains model, the scattering profiles in Fig. 3 give us information of each lobe size as described under "Materials and Methods." The R,(lobe) value estimated by the equation (see "Materials and Methods") is about 12 A, which is also in good agreement with the value estimated from the crystal structure of calmodulin ( 7 ) . It is noted that the apparent differences among the height of the peaks between the two minima shown in Fig. 3 correspond to the local conformational change since the ratio of so1/so2 is close to the theoretical value of 1.74, irrespective of the Ca2+ concentrations (18). Presumably, rearrangements of the secondary or tertiary structures in the lobes occur by the binding of Ca2+.
It has been well established that the high affinity binding sites for Ca2+ are located in the C-lobe (5, 20). Therefore, the Ca2+-induced small R, increase of calmodulin without mastoparan shown in Fig. 2 indicates that Ca2+ binding to only the C-lobe can be attributed to the change in R,. The C-lobe might partially wrap around the central helix because of its rather slack structure in the absence of Ca2+. The conformation of the C-lobe, however, would become so tight by the binding of Ca2+ (3) that it might slightly move away from the central helix and increase the R,. However, the size of C-lobe did not change significantly as described above; therefore, the conformational changes of the C-lobe would be very local, and the shape of the whole calmodulin may still be a dumbbelllike structure irrespective of the Ca2+ binding.
Consequently, in the absence of mastoparan the shape of the calmodulin molecule and the structure of each lobe in the solution would be similar with those in the crystal.
On the other hand, the Ca2+-induced R, change of calmodulin with mastoparan is significant enough to consider that the whole molecular shape of calmodulin would be changing sequentially by the binding of 4 mol of Ca2+. It is noted that position at the hump and two minima described in the text, respectively. Dotted lines on traces 2-5 are sketches of trace 1. In panel B, dotted lines on traces 1-4 are sketches of corresponding traces 1-4 in panel A . the distinctive shape change of calmodulin was induced by the Ca2+ binding to the N-lobe. As we showed elsewhere (19), the infinite dilution value of R , with Ca2+-saturated calmodulin in the presence of mastoparan was 17.8 & 0.3 A; therefore, the decrease of R, dependent on Ca2+ was not caused by the concentration effect of the calmodulin-mastoparan complex. This result has led to the question as to whether the calmodulin molecule still maintains a dumbbell structure in the presence of mastoparan-Ca2+ or not.
Striking differences were observed in SOXS profiles of calmodu$n with mastoparan at a moderate angle. The hump at 0.03 A -I , which reflects the correlation between two lobes as described previously, decreased and disappeared in traces 3 and 4, although the hump could still be seen in traces 1 and 2 in panel B of Fig. 3. Furthermore, the two minima discussed above in calculating the lobe size are also unclear in traces 3 and 4 of panel B in Fig. 3, although the minima are still perceptible in traces 1 and 2. The ratio of SOI/SOZ equals 1.73 in both traces 1 and 2, indicating that the size of lobes did not change significantly. These findings clearly indicate that the Ca2+ binding to the N-lobe made the shape of the calmodulin molecule far from a dumbbell structure in which the distance between the two lobes might be so close that the correlation characterized as the hump disappeared. The two domains model could no longer be applied to the calmodulin. Therefore, the R, value decreased drastically by the third and fourth Ca2+ binding to calmodulin as shown in Fig. 2. Presumably, the distance between the two lobes became close enough to transmit the information of the Ca2+ binding on the N-lobe to the C-lobe by an almost direct manner ( 5 ) . We could not interpret the decreasing distance between the two lobes except for the idea of a bending in the central helix of the calmodulin (21, 22). As we reported recently (19), the calculation of pair distance distribution function, P ( r ) , from the extended scattering profiles which involves information about interdomain distance (8,9) supported the bending model.
Based on the Chou and Fasman structure prediction (23), the sequence of mastoparan is considered a high helix former. Helical wheel representation of mastoparan displays two hydrophobic regions which locate at opposite surfaces of the helix. It is, therefore, reasonable to consider that one molecule of mastoparan could interact with two lobes of Ca2+-saturated calmodulin at the same time by the interaction between hydrophobic patches of two lobes (21) and hydrophobic surfaces of mastoparan. Thus, the Ca2+-saturated calmodulin in the presence of mastoparan underwent a bending at the linker region of this moleculle. A recent report on the crystal structure of calmodulin refined at a high resolution indicates that residues 79-81, which are at the middle region of the central helix, show significant deviations from ideal a-helical geometry, and also these residues have high mean temperature factors (24). It is conceivable that a part of the linker region including residues 79-81 would be in disorder for the secondary structure in solution (25).
Since the target enzymes, such as myosin light chain kinase and phosphodiesterase, were activated by the third and fourth Ca2+ binding to calmodulin as already described, the whole molecule of calmodulin was needed for the activation (26,27). It is therefore possible to consider that the key role of the molecular mechanism of the activation would be in the struc-tural change of calmodulin from a dumbbell to a bending structure induced by the third and fourth Ca2+ binding. The protein structure of the target enzymes could be affected by the structural change of calmodulin induced by the Ca2+ binding. However, we do not as yet know the physiological significance of the Ca2+ binding to the C-lobe. In the presence of mastoparan, the Ca2+ binding to the C-lobe induced the decrease of R, (Fig. 2), although the SOXS profiles at a moderate angle still displayed the characteristic features of a dumbbell structure as described previously. Presumably, mastoparan interacts strongly with the C-lobe but weakly with the N-lobe, because the hydrophobicity of each lobe was intensified with the Ca2+ binding. Therefore, at intermediate Ca2+ concentrations, the molecular shape of calmodulin-mastoparan complex is still a dumbbell-like shape, but the shape might have a diversity in which the distance between two lobes is variable depending on the degree of hydrophobic interaction between the lobes and mastoparan. The diversity of the shape of this molecule which is caused by a pliable structure of the central helix, therefore, might be essential to interpret that the calmodulin could bind to various kinds of target enzymes. The studies on R, of calmodulin with mastoparan at various Ca2+ concentrations depending on the calmodulin concentration may give us vital information about the second virial coefficients of molecules.