Characteristics of EF-hands and EF-hand proteins

Functions of EF-hand proteins

In the most simple scheme a calcium modulated protein is in the apo form in a quiescent cell with [Ca2+] ~10-7.2 M. Following cell stimulation, [Ca2+] rises to ~10-5.8 M and the protein binds a Ca2+ ion in its EF-hand(s). This induces a change in conformation enabling the EF- hand protein to bind and activate a target enzyme. There are many known, and certainly more yet to be discovered, variations or elaborations on this scheme.
1. By far the most egregious approximation is the simple model of pCaout ~ -2.8 , pCacyt ~ 7.2 in a quiescent cell, then a stimulus with [Ca2+]rising to pCacyt ~5.8, then falling to ~7.2 with relaxation. Many cells maintain a tonic activity - seldom maximally stimulated or fully quiescent. Further, [Ca2+] varies across the cytosol with distance and time. Of special relevance, there are constant leaks of calcium into the cytosol through the membranes of cell surface and of the endoplasmic reticulum balanced by ever active calcium extrusion pumps. This results in the [Ca2+] being higher near these membranes than the average value in the cytosol.
2. Although we speak of apo and calcium forms, the situation is more complex. In most EF- hand domains the affinity for calcium is about 104 times stronger for Ca2+ than for Mg2+. With [Mg2+] in the cytosol near constant, pMg ~2.8, this EF-hand domain would be primarily in the apo form in the quiescent cell. In contrast if the pKd(Ca2+) ~ 6.8 and pKd(Mg2+) ~ 2.8, then this EF-hand would be at least half magnesium in the resting cell. At higher affinity for divalent cations pKd(Ca2+) ~ 7.4 and pKd(Mg) ~ 3.4 the EF-hand would be partially in calcium and partially in magnesium forms in the resting cell. For instance EF-hands 3 and 4 of troponin C (TNC) appear to be mostly in the calcium form in the resting muscle cell and serve a structural function while EF-hands 1 and 2 are mostly apo and, following stimulation, function to transduce information. The actual functions of calcium modulated proteins are exquisitely dependent on their pKd(Ca2+)'s and pKd(Mg2+)'s and on the spatial and temporal distributions of the calcium pulse. This fine tuning is not captured in simple on/off diagrams.
3. The targets of the EF-hand, calcium modulated protein need not be an enzyme. Proto-oncogene Cbl protein (CBL) binds to DNA. alpha-actinin (ACTN), fordrin (FDRN), and fimbrin (FIMB) interact with thin filaments. Trichohyalin (HYFL) interact with keratin based intermediate filaments.
4. A single EF-hand protein may interact with multiple targets. Calmodulin (CAM) regulates over twenty enzymes or structural proteins.
5. An EF-hand protein may be free in the cytosol in the apo form and associate with its target in the calcium form. Or the EF-hand protein, such as calcineurin B (CLNB) may bind to its target calcineurin A (a protein phosphatase) in both apo and calcium forms.
6. This permanent attachment to a target may be taken one step further. Thirty-one of the 66 subfamilies are heterchimeras; the gene encoding EF-hands has fused with the gene encoding other domains. Of these 31, functions are known for only fourteen (four kinases, a phosphatase, lipase, dehydrogenase, protease, three that interact with the cytoskeleton, and one each that bind to a ryranodine receptor, repressor, and DNA.
7. All of the cytosolic EF-hand proteins are inferred to be calcium modulated. However, they need not be directly involved in information transduction. Several appear to be involved in temporal buffering of calcium, such as parvalbumin (PARV), and/or calcium transport, ICBP. In a resting muscle cell PARV, with relatively high affinity for divalent cations, is (primarily) in the magnesium form. EF-hands 1 and 2 of TNC have low affinity for divalent cations and are primarily apo. Following stimulation the influx of calcium binds first to TNC, which has lower affinity for calcium than does PARV, because it takes a few milliseconds for the Mg2+ to dissociate from the PARV. The Ca2+ then diffuses to the apo sites on the PARV thereby facilitating and sharpening the relaxation process and queueing the Ca2+ for the calcium extrusion pump [3].
8. The preceding overview of functions assumes the existence of EF-hands as calcium modulated proteins in the cytosol. Reticulocalbin (RTC) has a leader sequence typical of proteins found within the lumen of the endoplasmic reticulum and glycerol phosphate dehydrogenase (GPD) is found on the outer surface of the inner mitochondrial membrane. Osteonectin (BM40) and QR1 are extracellular in an environment whose pCa is near constant ~ 2.9. One EF-hand of BM40 binds calcium, supposedly with high affinity, but not with any modulation. This appears to be an example of Nature's having taken a protein initially "designed" for one function and put it to use in extracellular stabilization. S100 and PARV are sometimes found extracellularly. Whether this reflects a normal function or pathology has yet to be determined.
9. Although bacteria extrude calcium to maintain pCa ~ 7.0 in the cytosol, there is no evidence of their using calcium as a cytosolic messenger. There is one example of an EF-hand protein (CMSE) in a prokaryote, Saccharopolyspora, perhaps transduced from a eukaryote. An EF- hand protein (MSV) is encoded by a virus, Entomopoxvirinae.
A great deal of variation in structure is realized from this seemingly simple, thirty residue domain. When convoluted with the finely shaped pulse(s) of messenger calcium, a rich tapestry of information can be transduced by EF-hand proteins.

EF-hands and Ca2+ binding

Most functions of EF-hand proteins involve, or are inferred to involve, calcium binding. However, 81 of the 247 domains of the 66 subfamilies, do not bind calcium. In some domain subfamilies, indicated in table 1, by +/- some representative do and some do not bind calcium. For a few, one is reluctant to predict, e.g. ?/+. Four subfamily domains have non canonical Ca2+ coordination, indicated by - a, b, c, or d -in table and illustrated in figure 1. The important point is that one should not assume that all EF-hand domains bind calcium nor that one can always predict calcium binding, let alone affinity, from amino acid sequence.
The consensus sequence shown in figure 1 is a valuable heuristic for illustrating and understanding the main characteristics of an EF-hand. We emphasize, however, that the most sensitive search involves testing a candidate protein against a large data base, not against the heuristic. Such standard searches involve several complications. One EF-hand is so short that an heterochimeric protein will find homologs with higher significances based its non-EF-hand regions. The portions of the test sequence that are identified as similar to other proteins should be deleted and the remaining sequence subject to a new search. A second caution is that the proteins in the data base that are most similar to the test sequence (or to its remainder) may not be indicated in the data base descriptor as containing EF-hands, hence the value of table. Conversely, one can search the data base including the test sequence of interest with a known EF-hand. Again because one EF-hand is so short and because there is such a range of EF-hand sequences, one is well advised to search with several disparate EF- hands and/or to search with pairs of EF-hands. It is important to compare the significance of any alignment with the distribution of alignment scores from the entire data base.
The canonical calcium binding loop of the EF-hand is best regarded as a reference point for evaluating the numerous variations. Positions X (residue 10), Y (12), Z (14), -Y (16), - X(18), and -Z (21) represent the vertices of an octahedron. However the Ca2+ ion is seven coordinate in a pentagonal bipyramid, with major axis along X. Six amino acids are involved in Ca2+ coordination. Since both oxygen atoms of the carboxylate group at -Z bind the coordination number is seven. Usually Asp or Asn are found at X and Y; Asp, Asn, or Ser at Z; the carbonyl oxygen of a variety of residues is a -Y; -X is more variable but usually Asp, Asn, or Ser; and usually Glu at -Z. Although one can make a scheme that correlates calcium affinity with distribution of residues, we know of no scheme that predicts affinities for EF- hand loops that were not included in making the scheme. This is because the free energy of the system depends on the (change in) conformation of the entire protein, not just the loop in question.
Little is known of the coordination of the Mg2+ ion in EF-hand loops. However, given the many precedents [4] from the structures of small molecule, one can safely infer that the Mg2+ ion will be six coordinate. Probably the bidentate carboxylate at -Z rotates to become monodentate and the other oxygen atoms are ~0.2 _ closer to the Mg2+ than to the Ca2+.
Three (a, b, and c of table 1) of the non-canonical calcium coordinations involve several carbonyl oxygen atoms in place of the usual oxygen containing side chains. None of these coordinations were predicted; hence caution when considering the prediction (without crystal structure) of non-binding, "-", in table 1. The fourth exception, d in table 1 for CBL, is the use of the side chain of a Glu from a nearby _-helix instead of the Ser at the -X vertex in the second EF-hand loop, which by sequence appears to be canonical.
The angle between helix E and helix F varies between apo and calcium bound. This is important because the targets, discussed in the next two sections, of these calcium modulated proteins, like CAM, are alpha-helices that fit into the groove between the two EF-hands of a pair. The accessibility and hydrophobicity of this groove depends critically on the interhelical angles of the two EF-hands. To first approximation the interhelical angle, see figure 2 and legend, is closed in the apo form and more open in the calcium form. However, we emphasize four points: The angles associated with closed forms vary over a broad range as do the angles of open conformations. Second, the change in orientation associated with calcium binding varies. Third, several EF-hands of ELC and of RLC do not bind calcium. Their conformations are invariant to [Ca2+] but important to their functions. They cannot be simply classified as open or closed [5]. Finally, many of these EF-hands are, as will be discussed, in the magnesium, not the apo form in the quiescent cell; we have little information about the conformations of magnesium EF-hands.

Pairs of EF-hands

With only a few exceptions EF-hands occur in adjacent pairs. The fifth EF-hands of mili and micro calpains (CALP), and probably those of the close homolog, sorcin (SORC), pair to form a dimer. The N-terminal domain of PARV, indicated as #2 in table 1, covers the hydrophobic surface of the 3, 4 pair.
Just as there is a great range of angles between helix E and helix F of a single EF-hand so there is also a range of relationship between the first (ODD numbered) and second (EVEN) EF-hand. Both, either, or neither might bind calcium. When both bind calcium they may do so cooperatively but this is still subject to debate.
There are three general patterns among the fourteen subfamilies of known structure. The EF-hand protein is involved in information transduction and an _-helix of the target protein lies in the hydophobic groove between two EF-hands. This groove, in CAM and in TNC is more exposed when the two EF-hands are in the calcium forms and their respective helices E and F are more open. In the second pattern an _-helix of a chimeric protein lies within the ODD, EVEN groove of the same protein. This self binding is inferred to be, at least in some instances, a component of the information transduction pathway. In the third pattern the EF-hand protein is involved in calcium buffering or transport and the ODD, EVEN groove is either covered as in PARV by EF-hand 2 or partially occluded by the 1,2 loop as in ICBP. One can safely anticipate that other patterns of function will be revealed as more structures are determined.

Congruence within and among EF-hand subfamilies

Congruence is an important characteristic in assigning proteins to subfamilies and in grouping subfamilies such as CTER and CPV. By definition all members of a subfamily must be congruent, as illustrated in the dendrogram computed from the 56 domains of fourteen TNC's from ten different species, figure 4. The dendrogram illustrates the two essential characteristics of congruence. All domains 1 are more closely related to one another than to other domains; correspondingly all domains 2 cluster together as do domains 3 and 4. Second, the distribution of domains within each of the dendrograms of the four subdomains is (nearly) identical. An additional interesting characteristic, not inherent to the concept of congruence, is that the cluster of domains 1 is most closely related to domains 3 and the domains 2 are more closely related to domains 4. This reflects a gene duplication and fusion of an ODD, EVEN pair of EF-hands in the ancestor of all animals that have TNC.
Correspondingly, if two subfamilies are congruent, the dendrograms computed from their constituent proteins will be (nearly) identical to those computed from the corresponding domains of those constituent proteins. For instance all ten subfamilies within CTER are inferred to have evolved from a common, four domain precursor by gene duplication and subsequent divergent evolution figure 5.