Download - Structure of Paramecium tetraurelia calmodulin at 1.8 Å resolution

Protein Science (1993), 2, 436-447. Cambridge University Press. Printed in the USA. Copyright 0 1993 The Protein Society

Structure of Paramecium tetraurelia calmodulin at 1.8 A resolution

S.T. RAO,' S . WU,' K.A. SATYSHUR,' K.-Y. LING,2 C. KUNG,' AND M. SUNDARALINGAM' I Laboratory of Biological Macromolecular Structure, Department of Chemistry & Biotechnology Center, The Ohio State University, Columbus, Ohio 43210 Laboratory of Molecular Biology, University of Wisconsin at Madison, Madison, Wisconsin 53706

(RECEIVED August 18, 1992; REVISED MANUSCRIPT RECEIVED November 19, 1992)

Abstract

The crystal structure of calmodulin (CaM; M, 16,700, 148 residues) from the ciliated protozoan Paramecium tetraurelia (PCaM) has been determined and refined using 1.8 A resolution area detector data. The crystals are triclinic, space group P1, a = 29.66, b = 53.79, c = 25.49 A , CY = 92.84, /3 = 97.02, and y = 88.54' with one molecule in the unit cell. Crystals of the mammalian CaM (MCaM; Babu et al., 1988) and Drosophila CaM (DCaM; Taylor et al., 1991) also belong to the same space group with very similar cell dimensions. All three CaMs have 148 residues, but there are 17 sequence changes between PCaM and MCaM and 16 changes between PCaM and DCaM. The initial difference in the molecular orientation between the PCaM and MCaM crystals was =7' as determined by the rotation function. The reoriented Paramecium model was extensively refitted using omit maps and refined using XPLOR. The R-value for 11,458 reflections with F >3a is 0.21, and the model consists of protein atoms for residues 4-147, 4 calcium ions, and 71 solvent molecules. The root mean square (rms) deviations in the bond lengths and bond angles in the model from ideal values are 0.016 A and 3', respectively. The molecular orientation of the final PCaM model differs from MCaM by only 1.7". The overall Paramecium CaM structure is very similar to the other calmodulin structures with a seven-turn long central helix connecting the two terminal domains, each containing two Ca-binding EF-hand motifs. The rms deviation in the backbone N, Ca, C, and 0 atoms between PCaM and MCaM is 0.52 A and between PCaM and DCaM is 0.85 A. The long central helix regions differ, where the B-factors are also high, particularly in PCaM and MCaM. Unlike the MCaM structure, with one kink at D80 in the middle of the linker region, and the DCaM structure, with two kinks at K75 and 185, in our PCaM structure there are no kinks in the helix; the distortion appears to be more gradually distributed over the entire helical region, which is bent with an apparent radius of curvature of 74.5(2) A. The different distortions in the central helical region probably arise from its inherent mobility.

Keywords: calcium binding; calmodulin; crystal structure; Paramecium; structure comparison

Calmodulin (CaM) regulates a variety of calcium-dependent intracellular processes (Klee & Vanaman, 1982; Means, 1988) inchding activation of regulatory enzymes such as certain kinases, phosphatases, cyclases, phos- phodiesterases, and ATPases (Cohen & Klee, 1988). CaM is present in all eukaryotic cells with a highly conserved amino acid sequence, whereas more sequence variations are found in fungal CaM. It is a small acidic protein (M, = 16,700, 148 residues) and contains four calcium-

Reprint requests to: M. Sundaralingam, Laboratory of Biological Macromolecular Structure, Department of Chemistry & Biotechnology Center, The Ohio State University, 1060 Carmack Road, Columbus, Ohio 43210.

binding sites. The crystal structure of mammalian CaM (MCaM) has been determined and refined at 2.2 A reso- Iution (Babu et al., 1988). The crystal structure of recombinant CaM from Drosophila melanogaster (DCaM) has been determined and refined, also at 2.2 A resolution (Taylor et al., 1991). The overall structure of CaM is similar in both cases and consists of two globular calcium- binding domains, each containing two calcium binding regions with the characteristic EF hands (Kretsinger, 1980), connected by a long central helix of nearly seven turns.

Paramecium tetraurelia is a ciliated protozoan. Natu- ral mutations in Paramecium CaM (PCaM) have been shown to be responsible for abnormal motor functions of

436

Crystal structure of P. tetraurelia caimodulin 437

Paramecium, due to changes in the CaM-dependent K+ and Na+ ion currents across the membrane (Kink et al., 1990). Interestingly, these mutations have a distinct dis- tribution on CaM: mutations reducing the inward CaM- dependent Na+ current are confined to the N-domain, whereas those reducing the outward CaM-dependent K+ current are confined to the C-domain. This suggests a functional bipartition of the two lobes of CaM and indicates their different specific interaction with the channel peptides. Studies of Paramecium also indicate that this activation is in a manner very similar to that of enzyme activation (Kung et al., 1992). Recently CaM-activated ion channels have also been found in Drosophila (Hardie & Minke, 1992; Phillips et al., 1992) but have not yet been demonstrated in mammalian tissues. In this paper, we report the crystal structure of wild type PCaM at 1.8 A resolution and compare it with the other CaM structures. This study provides the native structure and forms the basis for characterizing the structural and conformational changes that accompany the mutations. This is the highest resolution at which any CaM structure has been reported to date.

The PCaM gene has been cloned and expressed in Esch- erichia coli so that PCaM and the desired mutants can be overproduced (Kink et al., 1991). There are 17 sequence changes in PCaM compared to MCaM and 16 changes compared to DCaM (Fig. 1). Most of the differences

M a . Param. Dros .

1 1 1 1 . . 2....*....3....*....4....*..4 0 0 0 8

< - - - - G - - - - X - - - I V - - X - - - - H - " - > * * * * *

". DEEVDEMIREADIDGDGQVNYEEFVQMMTAK P a r a m . -D---------------HI------R--"s- Dros . -------------------------T---S-

Fig. 1. The primary sequence of mammalian CaM (top line) is compared with Paramecium CaM (middle line) and Drosophila CaM (bottom line). Residues that are common with mammalian CaM are indicated by -. The helical and calcium-binding sites are marked and the loop residues involved in calcium binding are marked with *,

are conservative and located in the central helix and the C-domain. In addition, in wild-type PCaM, the N-terminal residue is acetylated, K13 is dimethylated, and K115 is trimethylated. Trimethylation at K115 is commonly ob- served in wild-type CaMs but the dimethylation is unique to PCaM (Schaefer et al., 1987). The functional signifi- cance of these modifications is not known.

Results and discussion

The PCaM molecule is shaped like a dumbbell of size 65 x 30 x 30 A (Fig. 2; Kinemage 1). A long 28-residue central helix connects the two globular calcium-binding domains at either end, each of size 20 x 20 X 20 A. The seven helical regions and the corresponding residues are

G (1 18-128), and H (141-146). The calcium binding "EF- hands" consist of loop regions 1 (20-3 I ) , 2 (56-67), 3 (93- 104), and 4 (129-140), flanked by helices on either side. The last three residues of each loop are helical and continuous with the exiting helix.

The plot of the average B-values for various residues in PCaM is shown in Figure 3; the average value for all the protein atoms is 24 A'. The temperature factors for residues in the C-domain are somewhat larger than those for the corresponding residues in the N-domain. The calcium-binding loop regions have the lowest values, whereas highest thermal motion is exhibited by the residues in the long central helix region and the terminal residues. The pattern of the average B-values for PCaM is quite similar to that for MCaM. In DCaM, the central helix region has somewhat lower B-values and is not as mobile. Wa-

A (7-19), B (32-39), C (45-59, D/E (68-92), F (105-1 12),

N

Fig. 2. Ribbon diagram (Priestle, 1988) of PCaM. The helical regions A through H are marked, as well as the termini.

43 8

70 1 I - All Atoms

"Bsckbone Aloms

0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 1 160

Residue Number

Fig. 3. Plot of the ( B ) values for the PCaM residues. Solid line denotes the values for backbone atoms only, and the dashed line for all atoms of the individual residues. The regions corresponding to the four calcium-binding loops are marked with a horizontal bar near the top. The mean B-value for all the protein atoms is 24 A 2 . Note the high ( B ) values for residues in the middle of the central long helix region (75-82).

ter molecules have B-values ranging from 18 to 65 A 2 ,

with a mean value of 36 A2.

Features of the Paramecium CaM structure The overall structure of PCaM structure is similar to the previously reported MCaM and DCaM structures. The two globular domains at either end of the molecule have similar structures and are related by a rotation of 109". Because the long central helix separates the two calcium- binding domains, there are no direct interactions between them in the crystal. As in troponin C (Herzberg & James, 1988; Satyshur et al., 1988), parvalbumin (Kretsinger & Nockolds, 1973; Swain et al., 1989), MCaM, and DCaM, each calcium-binding domain in PCaM consists of a pair of EF-hand calcium-binding motifs, related by a pseudo dyad axis. At each calcium-binding site, the calcium is hepta-coordinated in a distorted pentagonal bipyramidal fashion. The ligands come from the side-chain oxygens of residues at position 1 , 3, and 5 of the loop, carbonyl oxygen from residue 7, both side-chain oxygens from the invariant Glu at position 12, and a water molecule (Fig. 4; Kinemage 3). The coordination distances are given in Ta- ble l and are similar to values found in the other struc-

S. T. Rao et al.

tures. In PCaM, sites 1 and 3 are most similar (root mean square [rms] deviation of 0.21 A for the backbone atoms; Table 2) and sites 3 and 4 are least similar (rms deviation of 0.75 A). The backbone ($I,$) angles for the four 12-residue calcium binding loops have rms deviations of 10-17" from the average values found in other structures (Strynadka & James, 1989). The calcium-binding loops in each domain are further stabilized by a pair of antiparallel @-sheet hydrogen bonds, between the eighth residues in the two loops: I27 and I63 in the N-domain and I100 and I136 in the C-domain. Both these sheets are further extended by a pair of water bridges, one on each side, as seen in the calcium-bound holoC-domain of troponin C (Satyshur et al., 1992). The distances between the calcium ions at sites 1 and 2 in the N-domain (1 1.7 A) and between sites 3 and 4 in the C-domain (1 1.2 A) are similar.

Comparison with other CaM structures

For the comparison, the following coordinates were used from the Protein Data Bank: entry 3CLN for MCaM and entry 4CLN (in pre-release form) for DCaM. The overall structures of the three CaMs are very similar. The rms deviation in the backbone atoms (residues 5-147) of PCaM with MCaM is 0.52 A , whereas with DCaM it is 0.85 A , and is significantly larger (Table 3). A plot of the deviations in the a-carbon atoms of PCaM with MCaM and DCaM is shown in Figure 5 . The largest common deviations between the three structures are at the terminal residues and in the central helical region. In addition, there are significant deviations between PCaM and DCaM in the region of the B-helix (residues 32-40) and calcium- binding sites 3 (93-104) and 4 (129-140), as well as in the linker regions in both domains. The N-domains of the three structures are more similar (rms of 0.29 A, 0.53 A for MCaM, DCaM) than the C-domains (rms of 0.57 A , 0.95 A), as is to be expected from the larger number of sequence changes (eight versus three) in the C-domain. The calcium-binding loops also compare well, with rms deviations of 0.17,0.30; 0.20,0.33; 0.22,0.55; and 0.26, 0.39 A for sites 1 , 2, 3, and 4, respectively. It is clear

Table 1. Ca2+ coordination at the four sites

Site 1 Site 2 Site 3 Site 4

20D,OD1 22D,OD1 24D,OD2 26T,O 31E,OE1 31E.OE2 W 169

Average U

2.34 A 2.50 A 2.15 A 2.19 A 2.13 A 2.41 A 2.21

2.29 A 0.14 A

56D,OD2 58D,OD2 60N,OD1 62T ~ 0 67E,OE1 67E,OE2 W 174

2.36 A 2.36 A 2.35 A 1.92 A 2.52 A 2.59 A 2.09 A 2.32 A 0.22 A

93D,OD2 95D,OD2 97N,OD1 99L,O 104E ,OE 1 104E,OE2 W171

____

2.22 A 2.50 A 2.06 A 1.93 A 2.53 A 2.87 A 2.37 A 2.31 A 0.23 A

129D,OD1 131D.OD2 133D,OD2 135H,O 140E.OE1 140E.OE2 W158

2.13 A 2.48 A 2.44 A 2.23 A 2.09 A 2.54 A 2.25 A 2.32 A 0.16 A

Crystal structure of P. tetraurelia calmodulin

Table 2. Structural comparisons in PCa“

LOOP 1 LOOP 2 LOOP 3 LOOP 4

LOOP 1 (20-31) - 0.43 0.17 0.67 LOOP 2 (56-67) 0.47 - 0.43 0.37 LOOP 3 (93-104) 0.21 0.44 - 0.65 LOOP 4 (129-140) 0.75 0.44 0.72 -

Loops 1 and 2 vs. Loops 3 and 4: 0.42 A (a-carbons) 0.47 A (backbone atoms)

N-domain (10-73) vs. C-domain (83-146): 0.77 A (a-carbons) 0.82 A (backbone atoms)

aValues are rms deviations (A) between atoms in the four Ca2+ binding loops. The upper triangle contains numbers for the a-carbon atoms and the lower triangle for the backbone atoms.

PCaM structure is more similar to MCaM than to DCaM, with consistently smaller rms deviations. There are many regions in DCaM that have large deviations from PCaM compared to MCaM. It is difficult to understand this, because the sequence changes from PCaM are at the same sites and are nearly the same in number: 17 changes in MCaM and 16 in DCaM. A comparison of MCaM with DCaM (with only three sequence changes between them) also shows similar large differences in the same regions (Taylor et al., 1991).

The backbone conformation can be conveniently rep- resented by the angle of rotation (&) needed to superpose each peptide unit ( i ) on the next peptide unit (i + 1) in the protein, with a common Ca at residue i. This angle is a composite of the two backbone angles, 4(i) and $(i) and represents the local helical twist in the helical regions

12

439

(Srinivasan et al., 1991) The differences between the corresponding 4s values in the pairs of structures PCaM/ MCaM and PCaMIDCaM provide a measure of the backbone conformational differences between them (Fig. 6). The rms deviation in the differences from PCaM is 8” for MCaM and 15 O for DCaM. As already seen comparing the atomic coordinates, the larger differences are found for the residues at the N-terminal end and in the long central helical region. Additional deviations between PCaM and DCaM, noted above, persist. The side-chain conformations differ significantly at 17 residues between PCaM and MCaM, most of which lie on the surface of the molecule.

Long central helix Though the three crystal structures of CaM are very similar to each other, they show interesting deviations in the central helical region. The central helix is composed of the D-helix of the N-domain and the E-helix of the C- domain; the linker region between them (residues 76-82) also assumes a helical conformation, making a continuous long helix. In PCaM, the central helix appears to have no “kinks” and the distortion is uniformly distributed. In MCaM, there is a kink at residue D80, in the middle of the linker region between the D and E helices, where the O(76) and N(80) are too far apart to form the i . . . ( i + 4) a-helical hydrogen bond. Furthermore, the backbone torsion angles deviate significantly from their normal a-helical values. In DCaM, no kink is seen at D80, but two kinks are seen at K75 and I85 on either side and near the ends of the D/E linker helix. There are no such large deviations in the 0. . . N distances in PCaM (Table 4).

This is further illustrated by finding the local helix axis

Fig. 4. Stereo diagram superposing the four calcium-binding loops in PCaM using the backbone atoms of the 12 residues. The backbone is an a-carbon trace. The side chains at positions 1 , 3 , 5 , and 12, the carbonyl group at position 7, and the water- all coordinating to the Ca2+-are included. An octahedron, fitted to the five mono- dentate ligands of the first loop, is shown in dashed lines.

440 S.T. Rao et al.

Table 3. Comparison of PCaM with MCaM and DCaM structures

MCaM DCaM

a-Carbon Backbone a-Carbon Backbone Region (A) (A) (A) (A)

LOOP 1 (20-31) Loop 2 (56-67) Loop 3 (93-104) LOOP 4 (129-140)

Loops 1 and 2 Loops 3 and 4 Loops 1,2, 3, and 4

N-domain (5-73) C-domain (93-147)

All (5-147)

0.16 0.15 0.16 0.19

0.17 0.19 0.26

0.27 0.48

0.5 1

~~~

0.17 0.22 0.30 0.20 0.29 0.33 0.22 0.50 0.55 0.26 0.32 0.39

0.19 0.32 0.36 0.25 0.49 0.53 0.29 0.51 0.61

0.29 0.48 0.53 0.57 0.79 0.95

0.52 0.76 0.85

direction relating successive peptide residues in the long central helix and plotting the angle between the first and the succeeding steps. If the helix is nearly linear, the an-

t I I

u - ~

0 20 40 60 EO 100 120 140

Residue Number

Fig. 5. Plot showing deviations in the a-carbon atoms of PCaM with MCaM ( ~ ) and DCaM ( - - - -). Note that there are more regions of large deviations in DCaM than in MCaM.

60

-40 - . < .

60 ' " " ' " " ' ' ' " ' " " " " ' ' ' ' ~ '

0 25 50 75 100 125 150 0 25 50 75 100 125 150

Step Number

Fig. 6. Plot of the differences in the angle +.s for various residue steps. ~, PCaM and MCaM; - - - - , PCaM and DCaM. For defini- tion of 6s. see text.

gles will be close to zero. If it is uniformly bent, the angles will increase linearly as one progresses down the helix, and kinks in the helix will appear as discontinuities. Such a plot for all three CaM structures is shown in Figure 7, where the effective bending of the helix is 30-35". The

Table 4. O(i) . . . N(i + 4) distances and C=O(i) . . . N(i + 4) angles in the [ong central helix of PCaM

0 . . . N C = O . . . N Residue Residue distance angle ( 0 ( i + 4)

65 F 69 L 66 P 70 S 67 E 71 L 68 F 72 M 69 L 73 A 70 S 74 R 71 L 75 K 72 M 76 M 73 A 77 K 74 R 78 E 75 K 79 Q 76 M 80 D 77 K 81 s 78 E 82 E 79 Q 83 E 80 D 84 E 81 s 85 L 82 E 86 I 83 E 87 E 84 E 88 A 85 L 89 F 86 I 90 K 87 E 91 V 88 A 92 F

Mean rms

2.84 3.05 3.24 2.91 3.01 3.00 2.80 2.86 3.15 3.10 2.95 3.13 3.22 3.07 3.04 2.94 3.23 2.92 3.01 3.08 3.14 3.04 3.10 3.43

3.05 0.14

166 156 1 49 I56 158 I54 165 149 162 157 159 159 165 161 148 173 154 169 160 149 148 155 166 154

158 6.8

-62 -50 -51 -38 -77 -34 -70 -34 -69 -46 -60 -47 -54 -58 -57 -45 -61 -50 -62 -38 -68 -37 -70 -38 -58 -45 -59 -43 -66 -29 -66 -54 -46 -42 -63 -47 -58 -42 -61 -44 -73 -34 -68 -46 -55 -53 -56 -42

-62 -43 7.0 7.0

17 17 14 19 24 31 35 40 40 46 49 51 50 51 49 46 41 45 45 40 30 29 33 25

36 11.7

Crystal structure of P. tetraurelia calmodulin 44 1

8 I m

40

65 70 75 80 05 95 Step Number

C C ._ ._ m ._ c

5

B

3

-2 -

& -3 ' " " " " ~ " " ~ " " ~ " " ~ "

65 70 75 80 85 90 Step Number

Fig. 7. Plot of the angle between the first and successive steps in the long Fig. 8. Plot of the deviation in the radial distance of the axis point, in central helix as a function of the step number. 0, PCaM; 0, MCaM; multiples of the rms value, for various steps in the long central helix of 0, DCaM. Note the nearly monotonic rise of the curve for PCaM, and PCaM (e), MCaM (0), and DCaM (0). Notice how the axis dips to- the sharp discontinuities in MCaM and DCaM. ward the center of curvature for the region around residue step 85.

curve for PCaM almost monotonically rises, indicating a nearly uniformly bent helix (see Kinemage 2). The curve for MCaM displays the largest deviation around the kink at D80, whereas that for DCaM has sharper discontinuities at the two kink regions at K75 and 185. The "axis points," equidistant from the centers of mass of the two peptide units being related, define the local helix axis and are nearly coplanar, with rms deviations from the mean plane of 0.13,0.20, and 0.20 A for PCaM, MCaM, and DCaM, respectively. Fitting the local axis points to a sphere gives apparent radii of curvature of 74.5(2), 68.4(3), and 86.7(3) A for PCaM, MCaM, and DCaM, respectively. The rms values indicate that the central helix in PCaM is more regular than in MCaM and DCaM. The angular excursion of the local axis is different in the three CaM structures, and yet the N- and C-domains are in register.

A plot of the ratio of the normalized deviation of the axis points from the mean radius along the central helix shows an interesting common feature (Fig. 8): the E-helix axis (residues 83-88) is pulled in toward the center with a local minimum at L85. In a bent helix, the hydrophobic residues generally line the concave side (Blundell et al., 1983). The hydrophobic residues of the D-helix (up to residue 73) are on the concave side (with smaller radial distances, as expected). Then the hydrophilic linker inter- venes before the first hydrophobic residue L85 on the E-helix. L85 and the remaining hydrophobic residues lie on the convex side (outside)-it is interesting that the helix buckles inward at this point.

In the central helical region, there are seven sequence changes between PCaM and MCaM, but there are no sequence changes between MCaM and DCaM. Table 5 sum- marizes the hydrogen bonding interactions involving these residues in all three cases, excluding the a-helical hydrogen bonds. The sequence changes are conservative except

R(86)I, where a charged residue in MCaM is replaced by a hydrophobic residue in PCaM. This results in a loss of a salt bridge in PCaM, but the R(90)K mutation gains an intermolecular hydrogen bond between NZ of K90 and the carbonyl 0 of D129. The T(70)S mutation results in the loss of side-chain hydrogen bonds to the carbonyls of E67 (MCaM) and P66 (DCaM). Mutation D(78)E main- tains the side-chain hydrogen bond to the carbonyl of R74 in both PCaM and DCaM, bracketing the kink at residue 75 in DCaM. In PCaM, the side chain of E78 is longer and no kink in the helix is seen. These and other changes in the interactions among the three structures may be related to the different central helix distortions.

Various biochemical and spectroscopic studies have suggested that CaM adopts a folded conformation in which the two domains are closer together by at least 10 A than is found in the crystal (Persechini & Kretsinger, 1988; O'Neil & DeGrado, 1989; George et al., 1990). A recent NMR study (Ikura et al., 1992) has shown that when CaM complexes with the M13 peptide, only the region sur- rounding the linker (residues 73-82) undergoes a conformational change, whereas the structure on either side of this region is essentially unchanged from what is seen in the crystals. Modeling studies by Sekharudu and Sun- daralingam (1993) have also shown how the two terminal domains fold to bind to a target peptide. A recent X-ray study of the CaM-peptide complex (Meador et al., 1992) showed that the bend occurs over residues 73-77. This loosening in the middle of the long helix brings the N- and C-domains closer together to form a more compact molecule. In both PCaM and MCaM, the thermal parameters for the linker region are nearly twice as large as the ( B ) for the protein, suggesting considerable mobility in the crystals. In the DCaM crystal, the linker region appears to be somewhat less mobile (Fig. 1 of Taylor et al., 1991). The different distortions in the central helical region of

442 S. T. Rao et al.

Table 5 . Interactions of residues in the central helix region in PCaM and MCaMa

Residue

65 66 67

70

71 74 75 76 77 78

79 80

81 82

83

84

85 86

87 90

91

92

Residue in PCaM/

(M/D)CaM

F P E

S/T

L/M R K M K E/D

Q/T D

S E

E

E

L/I 1/R

E K/R

V

F

Interactions in PCaM Interactions in MCaM Interactions in DCaM

N . . . W157 - N. . . D64, OD2 0 . . . W161

OEl . . . T62, 0 OEl . . . D64, N OE2.. . W154, W174 OE2.. . D58, OD2 -

. E78, OEl

. D80, OD1 NZ. . . W186 OEl . . . R74, 0

OD1.. .M76, 0 OD2. . . W210 OD2. . . E83, OE2

OE2.. .Y138, OH

OE2. . . D80, OD2 OE2. . . W210 OE2. . . W200

OEl . . . S147, Oh

NZ. . . D129, Oh R126, NHlb

R126, NHlb W 160

-

N . . . W197 - N . . . D64, OD2 0.. . W192 0. . . T70, OD1

O E l . . .T62, 0 OEl . . . D64, N OE2. . . D56, OD2 OE2. . . D58, OD1 OG1. . . E67, 0 OG1.. . W192

- - NH2 . . . W215

0 . . . D80, OD1

0 . . . W164 -

- 0. . . E84, OEl

-

OD]. . . M76, 0

OEl . . . Y138, OH OEl . . . R86, NHl OE2. . . Y138, OH

OEl . . . D80, 0

- -

NH1 . . .E82, OEl

0 . . . ~ 1 2 6 , N H I ~ - -

- 0. . . R126, NHI 0. . . R126, NH2 - -

0. . . T70. OG1 -

OEl. . .T62, 0 O E l . . . W154 OE2. . . A57, N OE2. . . D58, OD1 OG1 . . . P66, 0

- -

NH2. . . W171 NZ. . . D78, OD1

- NZ. . . W172, W173 OD2. . . R74, 0 OD]. . . K75, NZ

-

- - 0. . . E84, OEl

OD] . . . W175 OD2.. . W175

N . . . E84, OEl OE2. . .Y138, OH OE2.. .R86, NHl

- -

OEl . . . D80,O O E l . . .S81, N O E l . . . W176 OE2.. . W176 OE2.. . W175

NHI . . . E82, OE2 NH2. . .Q3, NE2 NH2. . . Q3, OEl OE2. . . E127, OElh

- -

0.. . ~ 1 2 6 , NE^

0 . . . ~ 1 2 6 , N H I ~

0 . . .E104, OElh

a Hydrogen bonds involving backbone and side-chain atoms (53.2 A) are shown. The amino acids that are different in the two are shown in bold. Invariant residues with no interactions are omitted.

Intermolecular interaction.

the three CaM structures probably arise from its inher- water bridges. Eighteen intramolecular water bridges are ent mobility. formed (Table 6) , including the four water molecules that

extend the short anti-parallel 0-sheet in each domain in

Hydration and crystal packing both directions. Thirteen water molecules engage in intermolecular water bridges (Table 6) .

Of the 71 water molecules, 62 are in the first coordina- The direct intermolecular hydrogen bonds between pro- tion shell, directly hydrogen bonded to polar protein at- tein atoms in all three CaM structures are listed in Table 7; oms, and the remaining 9 interact with the protein via the largest number of these are between interdomain res-

Crystal structure of P. tetraurelia calmodulin 443

Table 6 . Intra- and intermolecular water bridges in PCaM

Dis!. 1 Dist. 2 Res./At. 1 (A) Water (A) Res./At. 2 Transl.

T26, OG1 3.00 A128, 0 2.94 G25, 0 2.73

N53, ND2 2.67 V91,O 2.71

R37, NHl 2.85 v 5 5 , o 2.86

T34, 0 2.85 G98, 0 2.71 R106, NE 3.19 D24, OD2 3.03 N97, ND2 3.20 R106.0 2.79

E7, N 3.20 A17, 0 3.11

E45,OE2 2.80

A102, N T29, N E47, OE2 N111, ND2 N53,O E83, 0 N42, 0 D119, OD1 A102, 0 R94, NHl

3.14 2.91 2.75 2.72 2.84 3.20 3.10 2.63 3.18 2.75

D80, OD2 2.90 K77, NZ 2.75

W153 W 156 W157

W159 W160 W161 W162

W 165 W166 W168 W 169 W 172 W173

W 176 W178

W180

W189 w190 W192 W194 W196 w201 w202 W205 W213 W214

w210 W186

3.20 2.93 2.83 2.60 2.81 3.20 2.89 3.00 2.69 2.92 3.21 2.96 3.02 3.14 2.90 3.10 2.62 2.88 2.81 3.09 2.95 2.95 3.06 2.73 3.20 3.13 3.20 2.83 2.97 3.20 3.19 2.74 3.05 3.19

Dl 19, N E140, 0 F65, N G40, 0 D56, OD1 R126, NH2 E67, 0 P43, 0 E120, OEl G113, 0 Y138, N L116, 0 Dl 18, OD1 H135, NE2 E114, 0 L116, N E139, OE2 D50, OD2 N53, OD1 E45, 0 N49, NE2 G134, 0 G61, 0 D58, OD1 N60, ND2 D56, 0 E87, OEl N42, OD1 D119, 0 D118, OD2 D122, OD2 R126, NH2 E83, OE2 E7, OEl

- l , l , l

- l , l , l

o,o,-1

idues. As is to be expected, many interactions are very PCaM, the E(119)D mutation shortens the side chain and similar in all three structures. However, the sequence is hydrogen bonded to its backbone carbonyl oxygen changes between them result in some differences, the atom via a water bridge. most notable involving residue 119. In both MCaM and In the central helical region, E87 and K90 interact with DCaM, the side chain of E l l 9 makes hydrogen bonds R126 and D129, respectively, of the +a-translated mol- with the residues in the third calcium-binding domain. In ecule (Fig. 9). There are water-mediated hydrogen bonds

NA

Fig. 9. Stereo plot of the a-carbon model

bors related by translations of +c and “c

(- - - -). Notice the antiparallel arrange- ment of A/C and E/G helices of adjacent molecules.

of PCaM ( - ) showing the two neigh-

444 S. T. Rao et al.

Table 7. Hydrogen bonds between protein atoms in PCaM, MCaM, and DCaMa

PCaM MCaM DCaM

Dist. Lattice Dist . Di:t. Res./At. ( 4 Res./At. Transl. (A) Res./At. (A) Res./At.

E6, N 2.87 E139, OE2 -l,O,l 2.56 E139, OEl E7, OEl 2.77 D131, OD2 -l,O,l 2.12 W185 2.19 E139, OEl

3.01 N137, OD1 -l,O,l 2.89 W185 (OE2) 2.74 N137, OD1 K21, NZ 3.07 D50, OD2 o,o, 1 3.14 D50, OD1 K21,O 3.02 N53, OD1 0,0,1 3.11 N53, OD1 3.05 N53, OD1 D22, 0 2.64 R106, NH2 - l , l , l 2.32 R106, NH2 2.74 R106, NH2 D22, OD2 2.61 D118, OD1 - l , l , l 3.20 W115 2.20 Dl 18, OD1 D24, OD1 2.99 Dl 18, N - l , l , l 2.60 D118, N T34, OG1 3.06 E114, OEl O,l,I - 2.37 El 14, OE2

2.88 E114, OE2 071,1 - R37, NE 2.88 E114, OEl 0,1,1 2.82 E114, OEl R37, NH2 2.93 E l 14, OE2 0,1,1 2.11 El 14, OEl N42, ND2 2.17 K115, 0 0.1,1 3.06 K115, 0

3.03 E120, OE2 0,191 2.60 K120, OE2 N42, OD1 - - 3.02 T117, OG1 R(K)94, NZ - - 3.18 E119, OE2 H107, N - 2.86 E119, OE2 2.55 E119, OEl N111, OD1 - - 2.75 E119, OEl R74, NH2 3.22b E14, OE2 o,o,-1 - - E87, OEl 2.95 S147, 0 0,0,1 - 2.95 E127, OEl K90,O 2.81 R126, NHl 0,0,1 - 3.08 R126, NE K90, NZ 2.86 D129, 0 o,o, 1 - - v91, 0 2.99 R126, NHl 0,0,1 - 2.34 R126, NHl

-

-

-

-

-

a Residues that are different are shown in bold. Good geometry, even though the distance is greater than 3.2 A.

between the side-chain hydroxyls of D80 and E83 with backbone 0 of E83 and the side chain of E87, and between NZ of K77 and the side-chain oxygen of E7 of a "c-translated molecule. The crystal packing is also further stabilized by the anti-parallel dipoles of the A/C and E/G helices of adjacent molecules related by translation along the c-axis. These lattice interactions may have fa- vored the elongated conformation for the CaM molecule, rather than the folded compact structure.

Conclusion The structure analysis of PCaM has revealed that, although the molecule adopts an overall conformation similar to MCaM and DCaM, there are differences in the central helix region. In PCaM the central helix is smoothly bent, with the a-helical hydrogen bonds preserved and only small deviations in the 4, I/ values. MCaM has a kink at D80, where the helix hydrogen bond distance is too long, and also has significantly different 4, I/ values. In DCaM, two kinks are seen at K75 and 185, but not at D80. The thermal parameters for this region are are relatively high in all three structures. The linker region assumes a nonhelical conformation when CaM binds to its target peptides, and the differences seen in the central helix of the three structures perhaps reflect its inherent mobility even in crystals.

Materials and methods

Crystallization

Crystals were obtained by vapor diffusion equilibration of 6-pL droplets hanging from siliconized coverslips placed over cell walls. The droplets consisted of 4 pL of solution containing 14.5 mg of calmodulin, 1 mL of dis- tilled water, 5 mM Ca2+ and 2 pL of solution containing 55% (v/v) 2-methyl-2,Cpentane diol (MPD) in cacodylate buffer (pH 5.0). These droplets were equilibrated against 1 mL of a solution containing 55Vo (v/v) MPD in 0.05 M cacodylate buffer (pH 5.0). After 5-7 days at 4 "C, bamboo-leaf-like crystals grew with dimensions of up to 0.5 x 0.3 x 0.05 mm. The crystals are triclinic, space group P1. The cell constants are similar to those of MCaM and DCaM, which are also triclinic (Table 8).

Data collection

A crystal of size 0.5 x 0.3 x 0.05 mm was separated from a cluster of thin blades and mounted with its length (c-axis) along the capillary tube. The intensity data were collected at the University of Virginia area detector at 4 "C with a Rigaku RU200 rotating anode generator operat- ing at 50 kV and 80 mA. A graphite monochromator was

crystal structure of P. tetraurelia calmodulin

Table 8. Crystat and refinement data for the three C a M structures

Parameter PCaM MCaM DCaM

Unit cell dimensions U 29.66 A 29.71 A 29.57 A b 53.79 A 53.79 A 53.92 A c 25.49 A 24.99 A 24.78 A a 92.84' 94.13" 93.24" P 97.02" 97.57" 97.08" Y 88.54" 89.46" 88.86"

Space group, Z PI, 1 PI, 1 PI, 1 Number of residues 148 I48 148 Sequence changes from PCaM - 17 16 Resolution range 10-1.8 A 5-2.2 A 10-2.2 A Number of reflections 11,458 6,685 5,239 Data collection method Area detector Diffractometer Diffractometer Refinement program XPLOR PROLSQ XPLOR, PROLSQ Final R-value 0.210 0.175 0.197

Final model Protein residues 4- 147 5-147 1-148 Ca2+ ions 4 4 4 Water molecules 71 69 78 rms in bond lengths from target values 0.016 A 0.016 A 0.018 A

445

used and the beam collimated to 0.5 mm. The two de- tectors, each of area 25 x 25 cm, were placed at -37" and +22.5" on 28, at a distance of 43 cm from the crystal, so that reflections up to a resolution of 1.6 A could be recorded. Two complete 4 scans at x values of -19" and -45 O were supplemented by four additional w scans a t other x values. A total of 53,853 reflections were collected and were scaled and merged with an Rmerg ( I ) of 0.07. The average intensity decreased rapidly with increas- ing resolution, and the intensities were weak due to the small size of the crystal. At 1.8 A resolution, there were 13,176 reflections in the data set, out of a possible 15,515, or 85%. Of these, 11,458 had I > 1 . k and were used in the present work.

Structure determination and refinement

The similarity in the space group and cell constants of PCaM and MCaM suggested that they might be isostruc- tural. For data between 6.0 and 3.5 A resolution (1,034 reflections), the MCaM model, with only the protein atoms, gave an R-value of 0.45. This value dropped to 0.41 after rigid body refinement with the program XPLOR (Brunger, 1990), and no further improvement in the model was possible. A rotation function using the 7-4-A data (620 reflections) with the MERLOT program (Fitzgerald, 1988) indicated a rotation of about 7" for the MCaM model. A difference map, calculated after rigid body refinement of the reoriented model (3 A, 2,561 reflections, R = 0.33), had the four highest peaks, corresponding to the four bound Ca2+ ions at heights ranging from 12 to 60 above the mean. Interestingly, the rigid body refine-

ment reduced the difference in orientation between PCaM and MCaM to 2.6" from 7O, but it was then possible to refine the model.

The sequence of the model was changed to correspond to that of PCaM, and the four Ca2+ ions were included in the model. All the subsequent refinement studies were carried out with the XPLOR program, and fast Fourier transform (FFT) sampled at 4 the resolution limit was used to compute the derivatives for the X-ray term. Using 6-1 .8-A data (9,869 reflections), the model containing the coordinates for residues 5-147 and the four Ca2+ ions was subjected to simulated annealing by initially heating the system to 4,000 K with the charges on the side-chain atoms of K, R, D, and E turned off. The system was slowly cooled to 300 K, in 25 K steps, and sampled at 0.5-fs intervals. This dropped the R-value from 0.36 to 0.28, with an rms movement of 0.21 A in all the atoms.

The model was improved using 3F0 - 2Fc Sim- weighted (Sim, 1960) omit maps in which 10 residues were omitted at a time; these were then refitted into the map using FRODO (Jones, 1985). Fo - F, difference maps were used to identify solvents, and water molecules were included with full occupancy if they had at least 3a density and were within 3.2 A of potential hydrogen bonding sites on the protein or other already established solvents with good geometry. Several rounds of conjugate gradient refinement, interspersed with model refitting and selec- tion of solvent molecules and a correction for the overall thermal anisotropy in the data, led to a final R-value of 0.210 for 11,458 reflections between 10 and 1.8 A resolution. Two of the coordination distances in site 2 from Ca were rather short (1.74 A to N60,ODl and 1.75 A

446

to W174). The electron density in the omit maps had a continuous outer contour covering the Ca and these two atoms. The refinement was repeated for these residues, with the FFT sampling done at a closer interval of the resolution (0.45 A). The ligand distances improved con- siderably (Table l), though the R-value remained the same. The main uncertainty at this site was in position- ing the Ca itself, which had a plateau of electron density. In the final residual map, the prominent features of the electron density were around the four Ca sites and S atoms of Met side chains, reminiscent of anisotropic motion. Refinement of the model with anisotropic thermal parameters was not attempted due to the low data: parameter ratio. The refinement statistics are given in Ta- ble 9. The R-value for all 13,176 reflections with F > 0.0 is 0.225. The relatively high R-value is perhaps due to a large number of weak reflections in the data set and also the disorder in the N- and C-termini. The difference in the orientation of the final model with MCaM is only 1.7".

The final model consists of protein residues 4-147, four e a 2 + ions, and 71 water molecules. The bond distances and bond angles have rms deviation from the ideal values of 0.016 A and 3", respectively. The method of Luz- zati (1952) indicates an estimated error in the atomic coordinates of 0.21 A. The coordinates and structure factors have been deposited with the Brookhaven Protein Data Bank (Bernstein et al., 1977).

The electron density in omit maps was clear and continuous for the backbone region and most of the side chains (Fig. 10). The side chains of residues near the surface of the molecule (E7, K13, E14, N42, E45, E78, E83, T110, E139, and V146) had discontinuities in the electron density, particularly at the CP atoms. In MCaM, the four residues 1-4 at the N-terminus and the C-terminal residue 148 are not modeled due to weak electron density. In our

S. T. Rao et al.

Table 9. Root mean square deviations in the final model from ideal geometrya

Program used for refinement XPLOR Bond lengths 0.016 A Bond angles 3.0" Dihedral angles 23.7' "Improper" angles 2.8" Parameter file used paraml9.pro Thermal parameters (B-values)

Bonded ( I ,2) atoms 1.4 t i 2

Bond angle (1,3) atoms 2.4 A' Resolution of data 10.0-1.8 A Number of reflections 11,458 R-value 0.210

aThe model contains residues 4-147, four calcium ions, and 71 solvents.

PCaM crystal, it was possible to build residue 4 into weak electron density, but the electron density was not inter- pretable for the remaining residues. Presumably these residues are disordered, as in the MCaM crystal. In MCaM, residue 5 is modeled as part of the first A-helix. In PCaM, residues 4-6 loop out and the helix starts with residue 7. In the central helical region, the electron density in the omit maps was weak and is reflected in the high ( B ) values. However, the electron density for the backbone and most side chains was sufficiently clear for model fitting.

The dimethylated K13 is on the A-helix in the N-domain, and NZ makes a hydrogen bond with W177. Although the electron density for the methyl groups is not unequiv- ocal, the two methyl groups appear to stack on F65. The trimethylated K115 is in the linker region between sites 3 and 4 in the C-domain, at the opposite end of the molecule exposed to solvent. The side chain is tucked toward

Fig. 10. Stereo diagram of a 3F0 - ZF, omit map for residues 31-38 in the B-helix. The atoms of these residues were not included in the phasing model. The contours are at 1.50 above the mean.

Crystal structure of P. tetraurelia calmodulin 447

the interior of the molecule, and weak electron density is visible around NZ. The trimethyl amine group is probably disordered.

Note added in proof

Further refinement of DCaM structure at a higher resolution of 1.7 A (Chattopadhyaya et al., 1992) has shown that the two kinks in the central helix, seen in the earlier lower resolution study, are now absent, as in our PCaM.

Acknowledgments

This work was supported by an Ohio Regent Eminent Scholar award to M S . and by National Institutes of Health grants GM22714 and GM36386 to C.K. We gratefully acknowledge an award of computer time on the Ohio State Cray Y-MP/864 supercomputer.

References Babu, Y.S., Bugg, C.E., & Cook, W.J. (1988). Structure of calmodu-

lin refined at 2.2 A resolution. J. Mol. Biol. 204, 191-204. Bernstein, F.C., Koetzle, T.F., Williams, G.J.B., Meyer, E.F. Jr., Brice,

M.D., Rodgers, J.R., Kennard, O., Shimanouchi, T., & Tasumi, M. (1977). The Protein Data Bank. A computer-based archival file for macromolecular structures. J. Mol. Biol. 112, 535-542.

Blundell, T.L., Barlow, D. J., Borkakoti, N., & Thornton, J.M. (1983). Solvent-induced distortions and the curvature of a-helices. Nature

Brunger, A.T. (1990). XPLOR Manual, v 2.1. Yale University, New

Chattopadhyaya, R., Meador, W.E., Means, A.R., & Quiocho, F.A.

306, 28 1-283.

Haven, Connecticut.

(1992). Calmodulin structure refined at 1.7 A resolution. J. Mol. Biol. 228, 1177-1192.

Cohen, P. & Klee, C.B. (1988). Calmodulin, Molecular Aspects of Cellular Regulation. Elsevier, New York.

Fitzgerald, P.M.D. (1988). MERLOT, an integrated package of computer programs for determination of crystal structures by molecular replacement. Acta Crystallogr. 21, 273-278.

George, S.E., Van Berkum, M.F.A., Ono, T., Cook, R.G., Hanley, R.M., Putkey, J.A.. &Means, A.R. (1990). Chimeric calmodulin- cardiac troponin C proteins differentially activate calmodulin target enzymes. J. Biol. Chem. 265, 9228-9325.

Hardie, R.C. & Minke, R. (1992). The frp gene is essential for a light- activated Ca2+ channel in Drosophila photoreceptors. Neuron 8,

Herzberg, 0. & James, M.N.G. (1988). RefinFd crystal structure of troponin C from turkey skeletal muscle at 2.0 A resolution. J. Mol. Biol.

Ikura, M., Clore, G.M., Gronenborn, M., Zhu, G., Klee, C.B., & Bax, A. (1992). Solution structure of a calmodulin-target peptide complex by multidimensional NMR. Science 256,632-638.

Jones, T.A. (1985). Interactive computer graphics: FRODO. Methods Enzymol. 115, 157-171.

Kink, J.A., Maley, M.E., Ling, K.-Y., Kanabrocki, J., & Kung, C. (1991). Efficient expression of the Paramecium calmodulin gene in Escherichia coli after four TAA-to-CAA changes through a series of polymerase reactions. J. Protozool. 38, 441-447.

643-65 1.

203, 761-779.

Kink, J.A., Maley, M.E., Preston, R.R., Ling, K.-Y., Wallen-Friedman, M.A., Saimi, Y., & Kung, C. (1%). Mutations in Paramecium calmodulin indicate functional differences between the C-terminal and N-terminal lobes in vivo. Cell 62, 165-174.

Klee, C.B. & Vanaman, T.C. (1982). Calmodulin. A&. Protein Chem.

Kretsinger, R.H. (1980). Structure and evolution of calcium-modulated proteins. CRC Crit. Rev. Biochem. 8, 119-174.

Kretsinger, R.H. & Nockolds, C.E. (1973). Carp muscle calcium-binding protein. 11. Structure determination and general description. J. Biol. Chem. 248, 3313-3326.

Kung, C., Preston, R.R., Maley, M.E., Ling, K.-Y., Kanabrocki, J.A., Seavey, B.R., & Saimi, Y. (1992). In vivo Paramecium mutants show that calmodulin orchestrates membrane responses to stimuli. Cell Calcium 13, 413-425.

Luzzati, P.V. (1952). Traitement statistique des erreurs dans la determination des structures cristallines. Acta Crystallogr. 5 , 802-810.

Meador, W.E., Means, A.R., & Quiocho, F.A. (1992). Target enzyme recognition by calmodulin: 2.4 A structure of a calmodulin-peptide complex. Science 257, 1251-1255.

Means, A.R. (1988). Molecular mechanism of action of calmodulin. Rec. Progr. Horm. 44, 223-286.

O’Neil, K.T. & DeGrado, W.F. (1989). The interaction of calmodulin with fluorescent and photoreactive model peptides: Evidence for a short interdomain separation. Proteins Struct. Funct. Genet. 6,284- 293.

Persechini, A. & Kretsinger, R.H. (1988). Thecentral helix of calmodulin functions as a flexible tether. J. Biol. Chem. 263,12175-12178.

Phillips, A.M., Bull, A., &Kelly, L.E. (1992). Identification of a Dro- sophila gene encoding a calmodulin-binding rotein with homology to the trp phototransduction gene. Neuron 8, 631-642.

Priestle, J.P. (1988). RIBBON: A stereo cartoon drawing program for proteins. J. Appl. Crystallogr. 21, 572-576.

Satyshur, K.A., Pyzalska, D., Greaser, M., &Sundaralingam, M. (1992). Refinement of chicken skeletal muscle troponin C at 1.75 A resolution. Acta Crystallogr. D, in press.

Satyshur, K.A., Rao, S.T., Pyzalska, D., Drendel, W., Greaser, M., & Sundaralingam, M. (1988). Refined structure ofchicken skeletal muscle troponin C in the two-calcium state at 2-A resolution. J . Biol. Chem. 263, 1628-1647.

Schaefer, W.H., Lukas, T.J., Blair, LA., Schultz, J.E., & Watterson, D.M. (1987). Amino acid sequence of a novel calmodulin from Par- amecium tetraurelia that contains dimethyllysine in th first domain. J. Biol. Chem. 262, 1025-1029.

Sekharudu, C.Y. & Sundaralingam, M. (1993). A model for the calmod- din-peptide complex based on the troponin C crystal packing and its similarity to the NMR structure of the calmodulin-MLCK peptide complex. Protein Sci. 2(4), in press.

Sim, G.A. (1960). A note on the heavy-atom method. Acta Crystallogr.

Srinivasan, R., Geetha, V., & Seetharaman, J. (1991). A unique or essentially unique parametric characterisation of biopolymeric structures. J. Mol. Struct. Dyn. 8, a208.

Strynadka, N.J.C. & James, M.N.G. (1989). Crystal structures of the helix-loop-helix calcium-binding proteins. Annu. Rev. Biochem. 58,

Swain, A.L., Kretsinger, R.H., & Amma, E.L. (1989). Restrained least squares refinement of native (calcium) and cadmium-substituted carp

J. Biol. Chem. 264, 16620-16628. parvalbumin using X-ray crystallographic data at 1.6 A resolution.

Taylor, D.A., Sack, J.S., Maune, J.F., Beckingham, K., & Quiocho, F.A. (1991). Structure of a recombinant calmodulin from Drosoph- ila melanogaster at 2.2-A resolution. J. Biol. Chem. 266, 21375- 21380.

35, 213-321.

13, 511-512.

95 1-998.