Update: 19 Sep 2000
The tertiary structures, as determined by X-ray crystallography and by multidimensional NMR, are rich in detail. We present drawings of the main chains of the members of 16 subfamilies such as CAM, VIS, S100, CALP, SARC and PARV and summarize the characteristics unique to that (member of) specific subfamily. Each EF-hand domain can be compared with the consensus or canonical EF-hand and, in turn, each pair of domains to the canonical lobe. For the six proteins having four domains the relationship(s) of lobe 1,2 to lobe 3,4 is important and is surely influenced by the sequence of the linker connecting domains 2 and 3 linker. The short regions preceding the first or following the fourth domains are significant. Two Structures are available for the chimetic subfamilies that have other homologous domains spliced to the EF-hands (CALP and PLC).
The canonical EF-hand consists of a~-helix F (residues 1-10) calcium-binding loop (10-21), and a~-helix F (19-29). The angle between helix E and helix F, as viewed down the axis perpendicular to and intersecting the axes of both helices, is about 100 o~; the angle for parallel strands is 180 o~, residues 2, 5, 6, and 9 of helix E, residue 17 of the loop, and residues 22, 25, 26 and often 29 of helix F have hydrophobic side chains that form the core of the lobe. The calcium-binding loop usually has the side chains of Ser, Asp, or Asn coordinating calcium at vertices X, Y, and Z. A peptide oxygen atom at vertex -Y provides the fourth calcium ligand. Usually a water molecule coordinates at vertex -X; the water may be hydrogen bonded, either directly or via a network of waters, to the side chain of residue 18 (-X). Glu at -Z binds with both oxygen atoms to make Ca2+ ion seven coordinate in a pentagonal bipyramid. InRLC-1 and SARC-1 Asp is at -Z and coordinates with both oxygens. Residues 10 (X) and 18 (-X) are axial.
Each amino acid can be numbered relative to the 29 residues in the EF-hand. For instance 1.10 is the tenth residue (and the X vertex) of the first domain; 1.- 10 is the tenth residue preceding the first domain and 1. + 10 is the tenth residue after the (29 residue) first domain but before the beginning of the second domain. In the canonical EF-hand helix E goes from residues 1 through 10 and helix F from 19-29. The actual beginnings and endings are indicated in Table 9.
In the canonical lobe the two EF-hands are related by an approximate twofold rotation axis. The nine hydrophobic side chains from each domain contribute to the hydrophobic core, which is assumed to account for the stability of that EF-hand protein. The two Ca2+ ions are 12A~ apart. The two loops pass one another to form a short stretch of antiparallel beta-sheet. The amide N of residues 1.17 contributes a proton in a hydrogen bond to the carbonyl oxygen of residue 2.17 and, as dictated by the twofold rotational symmetry, N (2.17) hydrogen bonds to C=O of 1.17.
The central helices of TNC and of CAM consist of helix F of domain 2, the 11 (TNC) or eight (CAM) residue linker, and helix E3. This linker is bent when CAM binds to myosin light chain kinase and, by inference, when CAM binds to other targets. The linkers of ELC, ~~ 9 residues, and of RLC, ~~ 10 residues, are bent as they bind to the a~-helical tail of the heavy chain of myosin. The linkers of SARC, 15 to 18 residues, and of recoverin (VIS), 7 residues, bend, although no target is bound. The linkers of CALP and SORC, only one residue, form continuous helix with helix F of domain 2 and helix E of domain 3.
Apparently, not all EF-hand proteins that have four (or more) EF-hands bind helical regions of target proteins; however, CAM, ELC, and RLC do. General characteristics of the CAM-myosin light chain kinase interaction are seen in the interactions of ELC and RLC with the heavy chain of myosin and can be anticipated in other EF-hand-target interactions. The target helix lies in a groove in the face of the lobe. This site is created by the notches between helices E and helices F of both EF-hands. The character of that groove can be modulated by the binding of calcium in one or both of the loop(s). Both smooth and skeletal MLCK fragments have Trp at position 4 in relative numbering. Many other natural and synthetic fragments have a large hydrophobic side chain at 4. At 17 there is also a hydrophobic amino acid on the opposite side of the a~-helix three and a half turns away (13 residues/3.6 turns/residue = 3.6 turns). Lobe 3,4 of CAM has, in addition to contacts with target 4, hydrophobic contacts with target 8 and target 11, one and two turns toward the C-terminal end of the target. Lobe 1,2 also has numerous hydrophobic contacts with target 17, as well as with target 13,14 and target 10, one and two turns respectively toward the N-terminal end of the target. There is a potential double asymmetry in the binding of the target in four grooves. CAM binds antiparallel to the target: lobe 1,2 to target 17/13,14/10 and lobe 3,4 to target 4/8/11. The target helix, from N- to C-terminus, contacts the EF-hands in the order 4,3/1,2. The two lobes of CAM are related by an approximate twofold axis with lobes 3 and 1 nearer the rotation axis. In contrast, the helix of the heavy chain of myosin fits into RLC: 3,4/2,1.
Each of the eight structures??? is illustrated by a tube tracing???? the main chain and by side chains involved in calcium coordination. CAM is viewed down the twofold axis that relates domains 3 and 4. The hydrophobic face of the hemispherical lobe is near the viewer, and the curved surface with two calcium-binding loops further away. This view shows the central helix (F2, linker, and E3) almost parallel to the page. The groove formed by the two EF-notches runs almost perpendicular to the central helix. TNC is also viewed down the 3,4 axis. The central helix is three residues longer; hence lobe 1,2 is rotated ~~ 300 o~ relative to lobe 1,2 of CAM. In the view of CAM bound to the MLCK helix, lobe 3,4 has the same orientation. The linker has had to bend, extend and rotate in order to bind 4,3/1,2. ICBP is viewed down its 1,2 axis and PARV is viewed down its 3,4 axis (domain 1 is inferred to have been deleted in the evolution of PARV). ELC and RLC are viewed together as complexed with the a~-helical tail of myosin, residues Met773-Ala840. The tail bends ~~ 40 o~ over residues Arg795-Lys800 between the ELC- and the RLC-binding domains. The tail bends ~~ 60 o~ between the site that binds lobe 3,4 and the site that binds lobe 1,2 of RLC. As viewed, the three stretches of a~-helix of the heavy chain are approximately in the plane of view. SARC and VIS (recoverin) are both viewed down the 3,4 axis. The orientation of lobe 1,2 relative to lobe 3,4 is different for CAM, TNC, ELC, RLC, SARC, and VIS.
Table 9 and the brief descriptions of the structures of (representatives of) the 16 subfamilies emphasize the deviations from the consensus EF-hand domain and lobe.
Crystal structures from four sources: Rattus , Homo , Drosophila , and Paramecium [6849b] are very nearly isomorphous (Figure 5); however, their linkers 2,3 differ slightly. This linker has high crystallographic temperature factors. In the NMR structure it shows higher mobility. Linker 2,3 of Paramecium CAM (MKEQDSEE) is smoothly curved, in Bos (MKDTDSEE) there is a small kink at Asp80 (2. + 5) and in Drosophila (MKDTDSEE) kinks at Lys75 (2.29) and at I1e85 (3.2).
The crystal [6849a] and solution NMR  structures of CAM complexed with a 26-residue helical section of smooth muscle (x-tal) or skeletal muscle (NMR) myosin light chain kinase (MLCK) show very similar lobes 1,2 and 3,4 as in the uncomplexed structure. As noted, the MLCK target peptide is bound by the CAM domains in the sequence 4,3/1,2. The hydrophobic patch of lobe 3,4 covers one side of the target with many van der Waals contacts to Trp4 and lobe 1,2 covers the other side with many contacts to Phe17 of the target peptide. Linker 2,3 (MKDTDSEE) plus residues 2.28 and 2.29, is bent and extended (Figure 6).
Troponin C (TNC)
TNC from Gallus gallus (chicken) [6849f] and from Melegris gallopavo (turkey)  are isomorphous. Both have calcium bound to the higher affinity sites 3 and 4. Both have no cation bound in sites 1 and 2, although sites 1 and 2 bind calcium in solution. Herzberg et al. proposed that when both domains bind calcium the angle tightens from ~~ 140 o~ to ~~ 110 o~ thereby increasing the hydrophobicity of the flat surface of the hemisphere, as seen in lobe 3,4 of TNC and in both lobes of CAM [6838a]. Four hydnggen bonds connect loop 1 to loop 2. As discussed in 'General characteristics' there are only two hydrogen bonds connecting most pairs of domains, in which both loops bind calcium. In addition to the symmetric Ile37 (1.17) N --> C=O Ile73 (2.17) and Ile73 N --> C=O Ile37, the antiparallel sheet is extended one hydrogen bond in both directions: Thr39 (1.19) --> Gly71 (2.15) and Phe75 (2.19) --> Gly35 (1.15). This tightening of the beta-sheet correlates with an opening of the loop and reduced exposure of the hydrophobic patch. Lobe 1,2 of TNC is inferred to reflect the structures of other apo lobes.
The a~-helix (1.-18 to 1.-8, MTDQQAEARAF) that precedes helix El of TNC lies parallel to helix F2 and antiparallel to E1 and may effect the stability and calcium affinity of lobe 1,2, (Figure 7). It is not known whether linker 2,3 is bent in the TNC-TNT-TNI complex.
Essential light chain of myosin (ELC)
Both ELC and RLC bind to the helical tails of the heavy chains of myosin, which are about 2000 residues long depending on species and type. Limited papain proteolysis of this heterohexamer cleaves the heavy chain at Ala840, using scallop (Aquipecten irradians) numbering, producing the coiled tail and two S1 heads, each containing one copy each of ELC and of RLC. Clostripain digestion of S1 cleaves at 772 producing a shortened head, which contains the actin-binding site and ATPase activity, and the 'regulatory domain' which consists of heavy chain residues Met773 to Ala840 bound to ELC and to RLC. The heavy chain a~-helix bends about 40 o~ over the region, R795KAYKK800, between the ELC and the RLCbinding regions, both of which are on the convex surface (Figure 8). It is not known whether the two heads of myosin are related by a twofold rotation axis nor whether the ELC and RLC of one head touch the other head.
Lobe 3,4 of ELC has numerous hydrophobic contacts with the tail of myosin over the region Phe785 to Tyr798. The orientations of lobes 3,4 of ELC and of RLC relative to myosin are similar. In contrast, the outsides of helices E1 and E2 of ELC contact the heavy chain. The coordination of calcium in loop 1 is unique to molluscan ELC and is described in Figure 1. The ATPase of molluscan myosin is activated even in the absence of actin by the binding of calcium to loop 1 of ELC, hence the term regulatory domain. The binding of RLC inhibits the activity the ATPase of molluscan myosin; calcium binding overcomes this inhibition. Myosin of vertebrate striated muscle is activated by binding to actin. Loop 3,4 of molluscan RLC interacts with calcium-binding loop 1 of ELC and is required for calcium binding. Gly117 (3.+ 1) hydrogen bonds (N --> C=O) to Phe20 (1.10) and receives a hydrogen bond from Arg24 (1.12: two residues are inserted in ELC loop 1). Gly117, RLC, forms a van der Waals contact with Gly23, ELC, whose carbonyl oxygen coordinates calcium. This Gly is conserved in molluscan ELCs but is Cys in vertebrate ELC. The outside, away from the hydrophobic core, of helices E1 and F1 contacts the heavy chain a~-helix from residues R795 to L808. The near antiparallel orientation of helix E1 and F1 may reflect the rearside contact with the target rather than the lack of calcium binding by domain 2. Linker 2,3 bends in two places: 2.+3, 3.-2.
Regulatory light chain of myosin (RLC)
Both lobes of RLC are near canonical in overall shape. They bind to the heavy chain over the region Ile811 to Trp824 (Patinopecten numbering). The sequence of domain contacts, relative to the heavy chain N --> C, is 3,4/2,1 (cf. CAM 4,3/1,2). In contrast to the straight a~-helix of MLCK as bound by CAM, the helix of the heavy chains bends 60 o~ over Trp-Gln-Trp826 (Patinopecten) or WQW831 (skeletal muscle of chicken [Gallus gallus]). Linker 2,3 is nonhelical and extended about Gly82 (2.+6). The Ca2+ (or Mg2+) ligand at -Z is Asp. There is a shift in the loop similar to that seen in domain 1 of SARC and domain 4 of oncomodulin. Loop 3 has the potential ligands to bind calcium; however, none is seen in either crystal structure.
Calcineurin B (CLNB)
Recoverin (subfamily VIS)
The crystallized recombinant recoverin lacks the usual Met1?? and N-myrislyate group, which may participate in a calcium-induced partition of recoverin into lipid bilayers. Calcium is not bound in loop 2 in the crystal at (NH4)2SO4 3.2 M; Sm3+ is bound as a heavy atom derivative at the inferred calcium-binding site of loop 2. Two equivalents of calcium bind in solution. Residues 10-17 (1.- 17-1.- 10) form an helix before domain 1 and residues 180-186 (4. + 1-4. +7) and 190-195 of the 23 residue C-terminus tail form two 3/10-helices after domain 4 (Figure 10). The outer hydrophobic surface of helix F4 of a symmetry related molecule packs against the exposed, concave hydrophobic face of lobe 1,2. The hydrophobic patch of lobe 1,2 is probably exposed in solution. The 2,3 linker (SAGKNQ) bends in such a way that the outside of helix E2 touches the outside of helices E3 and primarily F3 and the outside of helix F2 touches the outsides of helix E3. This places all four loops on one side of the molecule. The hydrophobic patch of lobe 3,4 is partly covered by an c~-helix (3. + 4-3. + 10) of the 21 residue linker 3,4 and by the 3/10 helix 180-186 to the C-terminal side of domain 4 [6835a].
GCAP (subfamily VIS)
Intestinal calcium-binding protein (ICBP)
The linker 1,2 (FPSLLKGPS) forms one turn of a~-helix. Crystallographic studies show that the Gly Pro43 (1. +8) peptide bond is about equally distributed between trans and cis forms [6849g]. Although domain 1 has two residues inserted in the loop, the trace of the main chain of the four helices and loop 2 is quite similar to the canonical lobe (Figure 4).
Phospholipase C (PLC)
Sarcoplasmic calcium-binding protein (SARC)
Crystal structures are available for Nereis diversicolor (sandworm)  and for an isoform of Branchiostorna lanceolatum (amphioxus)  SARC. Although the sequences share only 12% identity overall, the main chain traces of the two are near superimposable. Seven isoforms of amphioxus differ in only nine positions between residues 20 (1.11) and 36 (1.18). No target for SARC has been found, nor is one inferred to exist. The F1 helix is drawn closer to the Ca2+ ion because the side chain of Asp is shorter than that of Glu, usually found at -Z. The interhelix angle (E1, F1) is ~~ 62 o~ in contrast to ~~ 100 o~ usually found when both (or one) loops have bound Ca2+ ions (Figure 9). The sharp bend in linker 2,3 (2.+ 3-2. +6, Asn85-Pro-Glu-Ala88 Nereis) makes helices F2 and E3 run antiparallel. We use the alignment of Cook et al., which is based on tertiary structure ; none of 29 residues in domain 2 are identical between N. diversicolor and B. lanceolatum. A shift of one residue would score 4/29 identity. Linker 1,2 is 18 residues long in N. diversicolor and 22 in B. lanceolatum. Helix F1 is extended three or four residues and helix E2 starts 11 or 13 residues earlier than the canonical EF-hand; F1 and E2 are near antiparallel. The extended F1, E2 pair of helices and especially the F2, E3 pair form a shelf upon which the hydrophobic face of lobe 3,4 rests. The extended C-terminal tail (4.+3-4. +17) covers part of the face of lobe 3,4. The hydrophobic faces of lobes 1,2 and 3,4 contact one another, but are somewhat tilted. An antiparallel beta-sheet is not formed between loop 3 and loop 4, although hydrogen bonds are formed between the two loops.
The crystal structures of two beta isoforms (Cyprinus carpio and Esox lucius [pike]), two alpha isoforms (Rattus norvegicus and Triakis sernifasciata [shark]) and oncomodulin (Rattus norvegicus) are very similar (Figure 3??). The three EF-hand domains are numbered 2, 3, and 4 because 3 and 4 are almost congruent with domains 3 and 4 of CAM*ELC*RLC*TNC. Domain 1 is inferred to have been deleted from the four-domain precursor. Most PARVs are 108 or 109 residues long and end with 4.28 or 4.29.
Domain 2 is not paired with another EF-hand. Its hydrophobic inside forms a cap covering the hydrophobic surface of lobe 3,4. The polar hydrogen bond between Arg75 (3.+5) and Glu81 (4.1) is invariant and is inferred to stabilize lobe 3,4. Helix E3 has two i ---> i-3 hydrogen bonds, as in a 3/10 helix; this causes it to curve. In domain 3 there is no bridging water to the Ca2+ ion, as is found in most EF-hand loops at vertex -X. In oncomodulin -X= Asp; helix F is shifted slightly toward the Ca2+ ion (N-ward), consistent with the side chain of Asp being one carbon shorter than the Glu usually seen at -Z. RLC-1 and SARC-1 also have Asp at -Z. In Gadus callarias (cod) Z = Glu; hence five carboxylates are inferred to coordinate calcium. Cod also has four additional residues at its C-terminus; a crystal structure is not available. The loop is stabilized by two Asx turns Asp (3.12, Y) to Set (3.14, Z) and Ser (3.14) to Phe (3.16). Loop 4 is stabilized by an Asx turn (Asp 4.14 to Lys 4.16).