I.  INTRODUCTION

Top Previous Next References

External signals often lead to transient increases in intracellular Ca²⁺ (13). Cells have elaborated exquisite ways of controlling and utilizing the gradient of ion concentration across the plasma membrane and have developed an intracellular messenger system by adopting Ca²⁺ signaling. Increases in the cytoplasmic Ca²⁺ concentration lead to binding of Ca²⁺ by intracellular regulatory proteins (33), an event that initiates a wide variety of cellular processes. One of many Ca²⁺-binding regulatory proteins, calmodulin (CaM), is a ubiquitous, multifunctional protein that can bind to at least 30 different target enzymes and proteins (see Tables 1 and 2) including Ca²⁺-transport ATPase, phosphodiesterase, myosin light chain kinase (MLCK) and other CaM-dependent protein kinases, calcineurin (phosphoprotein phosphatase 2b), and nitric oxide synthase (61, 93, 141). Although much of the research focus has been on the Ca²⁺-bound form, cellular Ca²⁺-binding proteins such as CaM can exist primarily in two states, the other being effectively Ca²⁺ free. Either state can bind a different (or overlapping) set of target proteins and in that way signal cellular responses.

Calmodulin was discovered as an activator of cyclic nucleotide phosphodiesterase in brain and heart (24, 65). It was subsequently rediscovered several times, since many Ca²⁺-dependent cellular processes were eventually shown to involve the same Ca²⁺-binding protein. As a result, there was confusion in the nomenclature in the early literature, since the protein was referred to by several different names (e.g., Ca²⁺-dependent regulator, modulator protein, Ca²⁺-dependent modulator, activator protein, troponin C-like protein). However, the name CaM, first suggested by W. Y. Cheung, has found general acceptance. Calcium-free CaM is referred to as apocalmodulin (ApoCaM).

Calmodulin is widely distributed in nature and is ubiquitous in eukaryotic cells. Calmodulins isolated from diverse organisms are remarkably similar in biological, chemical, and physical properties (78, 92). There is 100% identity in its amino acid sequence among vertebrates, with multiple genes encoding identical CaM (43). This high degree of conservation may be essential for the maintenance of interaction with a diverse family of CaM-binding proteins (30). It is a small (16.7 kDa), very acidic (isoelectric point ~4), relatively stable, and heat-resistant protein. It contains four Ca²⁺-binding sites per molecule (53, 69). One of the properties of CaM, which was discovered earlier and is by now extremely well documented (155), is the conformational change that takes place upon binding Ca²⁺. It transmits the Ca²⁺ signal by binding to and activating numerous enzymes central to cellular regulation (27). Although CaM could potentially exist in 16 different Ca²⁺-bound states (i.e., ApoCaM and CaM with 1, 2, 3, or 4 Ca²⁺ bound at various combinations of the 4 sites), most enzymes that are activated by Ca²⁺-CaM require that three or four Ca²⁺ be bound (29). Thus much of our discussion can be simplified by considering primarily two states of CaM: ApoCaM and "Ca²⁺-CaM," the form containing three or four Ca²⁺ that activate certain enzymes. What makes CaM more interesting is not merely how many, but also how varied are its binding proteins. Binding is not to a conserved motif: Ca²⁺-CaM binding domains in different target proteins show very little similarity in primary sequence. Instead, these binding sites are amphipathic -helices, typically ~20 residues in length, with basic and hydrophobic residues intermingled. Typically, one or more aromatic residues are also found near the amino-terminal end of this region, which are important to binding. This is an unusual sort of specific molecular recognition (157). Interestingly, the binding sites for ApoCaM show greater homology.

Table 1 lists Ca²⁺-CaM binding proteins that have been reasonably well characterized. With only two exceptions (phosphorylase b kinase and syntrophin), we do not discuss the proteins in Table 1 further. The interested reader should consult the numerous current reviews of Ca²⁺-CaM (19, 30, 91, 93, 100, 106, 141, 142). Among these reviews are an issue of a journal (19) and a volume of a series (91) that give details of individual CaM-protein interactions. Calmodulin can also interact with proteins in a Ca²⁺-independent manner (118). This review discusses the structure of ApoCaM and how it differs from Ca²⁺-CaM, the roles of ApoCaM in the eukaryotic cell, and those proteins for which there is substantial evidence of ApoCaM binding. These ApoCaM-binding proteins are listed in Table 2.

The binding of CaM to proteins in the presence of Ca²⁺ chelators (e.g., EGTA or EDTA) is frequently referred to as ApoCaM binding, although the metal-bound state of CaM is often not known. This may be a significant problem. In Figure 1 is illustrated a simple model for the interaction of a binding protein (E) and CaM in the presence and absence of Ca²⁺.¹ It is a rather simple matter to show that in such a scheme, K₁K₂ = K₃K₄, where K is a dissociation constant. This equivalence means that all of the various binding equilibria are linked to one another. This concept of linkage is an old one (149). Although the overall average Ca²⁺ affinity of CaM (K₁) is probably ~11 µM under conditions approaching physiological (29), this is not necessarily the Ca²⁺ affinity of CaM bound in a complex with another protein. If a protein (E) preferentially binds Ca²⁺-CaM (i.e., K₂ < K₄), this binding will increase CaM's Ca²⁺-binding affinity in the complex. Because of this effect, CaM binding to enzymes may occur at extremely low Ca²⁺ concentrations, concentrations possible even in the presence of a chelator. Thus whether binding is to Ca²⁺-CaM or to ApoCaM must be directly demonstrated and cannot always be inferred from the presence or absence of a chelator.

View larger version (14K): [in this window] [in a new window]   Fig. 1. CaM, calmodulin; E, binding protein; K, dissociation constants.

A few specific examples may be helpful. Taking simply the concentration of Ca²⁺ required for half of CaM's binding sites to be occupied, MLCK shifts CaM's apparent affinity for Ca²⁺ from the ~11 µM mentioned above to ~0.6 µM. When the effect on individual Ca²⁺-binding sites was calculated using a current model, the binding of one site was changed by nearly 3,000-fold (29). Inducible nitric oxide synthase (iNOS) is isolated from macrophages in buffers containing EGTA and is found tightly bound to CaM. A CaM binding site has been demonstrated between amino acids 504 and 532 in the primary structure (2, 120, 140). A synthetic peptide of this sequence binds CaM, and the binding is highly Ca²⁺ dependent as shown by the concentration at which half-maximal binding to dansyl-CaM occurs (2). The affinity of the peptide for CaM in 0.1 mM Ca²⁺ (K₂) is ~1,000-fold greater than it is in the virtual absence of Ca²⁺ (K₄). These peptide binding studies and the scheme in Figure 1 suggest that the binding of CaM to these sequences could increase CaM's affinity for Ca²⁺ by 1,000-fold: from ~11 µM (see above) to about 11 nM Ca²⁺. When a chimera was constructed in which these iNOS sequences were used to replace the normal CaM-binding sequences of neuronal NOS (nNOS), half-maximal activation occurred at ~4 nM Ca²⁺ instead of the 200-300 nM found for the wild-type nNOS (120). Inducible NOS has additional sequences that also contribute to CaM binding (120), and the shift in CaM's Ca²⁺ affinity may result from a complex linkage of multiple binding events. However, this shift nevertheless does occur in the direction predicted by the scheme in Figure 1, and of the predicted magnitude. Inducible NOS is maximally active at concentrations of Ca²⁺ calculated to be as low as 0.1 nM (120), yet Ca²⁺ and EGTA do affect activity in other studies (140). Thus whether CaM or ApoCaM is bound by iNOS cannot really be known with any certainty until the Ca²⁺ composition of pure, active enzyme is finally measured.

Another valuable point to be made from this linkage is that a protein's affinity for Ca²⁺-CaM (relative to its affinity for ApoCaM) can cause changes in the Ca²⁺ affinity of the CaM-target protein complex. Thus, as target proteins have evolved to bind preferentially one form of CaM over another, the same process also alters the Ca²⁺ concentrations to which the protein will respond. Calmodulin-dependent processes can thus respond to different Ca²⁺ concentrations (dissociation constant values of 10⁸ to 10⁵ M), that are determined by a combination of CaM's intrinsic affinity for Ca²⁺ and its affinity for a particular binding protein.

In this regard, it is also important to realize the limits of chelators. Although the dissociation constant of EGTA⁴ (the fully deprotonated form) for Ca²⁺ is an impressive 10 pM, the actual binding is highly pH sensitive, and this value would only be approached at quite alkaline pH (20). At pH 7, the calculated dissociation constant is actually only ~0.2 µM (calculation not shown), and CaM complexes may actually have a greater Ca²⁺ affinity in many experiments.

Thus the affinity of CaM for Ca²⁺ can be much greater in the presence of its target proteins. Because it is difficult to ensure adequate sequestration of Ca²⁺ under some experimental conditions, it can be expected that some of the binding attributed to "ApoCaM" may actually turn out to be Ca²⁺ dependent. Nevertheless, it is clear that true ApoCaM binding does occur. For example, neuromodulin binds CaM in EGTA, and this binding is disrupted by Ca²⁺ (4). Because neuromodulin binds preferentially in low Ca²⁺ or its absence (i.e., K₄ < K₂), it would be expected to decrease CaM's Ca²⁺ affinity, and the EGTA used (4) should be adequate to ensure Ca²⁺-free CaM. Thus, in this case, it is clear that ApoCaM binding is a neuromodulin property.

The reader should be aware though that the Ca²⁺-bound state of ApoCaM is not determined in the literature reviewed here. There is not a single case we are aware of in which a protein-ApoCaM complex was shown to be Ca²⁺ free. Lacking these crucial data, we have been forced here to utilize less certain criteria for ApoCaM binding. The criteria used are as follows.

1) The first criterion is when a protein is purified through a number of steps utilizing adequate amounts of chelator and CaM remains associated. This criterion applies to phosphorylase b kinase, iNOS, and several other proteins discussed in sections IIIB and IVC.

2) The second criterion is when the protein-binding affinity in the presence of chelator is greater than or equal to that measure in the presence of Ca²⁺. This is to be contrasted to the case with most Ca²⁺-CaM binding proteins where chelator usually diminishes affinity by several orders of magnitude. This criterion applies to neuromodulin, neurogranin, and many others.

3) The third criterion is when the affinity for ApoCaM is such that binding may be physiologically important. For example, the Bordetella pertussis adenylate cyclase binds CaM under conditions that are even less favorable than the low intracellular Ca²⁺ concentrations present in the host eukaryotic cell (52).

These criteria are adequate for the purposes of this review, but the reader should approach this topic critically. To aid in this, our discussion of the ApoCaM binding proteins (see sect. IV) provides details of the evidence that binding is Ca²⁺ independent.

The immediate physiological consequences of CaM action have been demonstrated in several cell types. In smooth muscle cells, Ca²⁺/CaM-dependent activation of MLCK initiates muscle contraction (128). In mammalian skeletal muscle and liver, Ca²⁺-CaM activation of phosphorylase kinase induces glycogenolysis (27). The discovery that immunosuppressive drugs such as cyclosporin cause inhibition of the Ca²⁺/CaM-dependent protein phosphatase calcineurin suggests a role for CaM in T-cell activation (80). Hemenway et al. (56) reported that calcineurin plays a role in the regulation of cation transport and homeostasis with their study on vph6 mutants of Saccharomyces cerevisiae. In plants, CaM activates NAD kinase and may play an essential role in regulating photosynthesis (62). In Paramecium, CaM regulates motile behavior (70).

Although many of the above activities require Ca²⁺, the Ca²⁺ dependence of many other cellular roles of CaM is less clear. To discuss this, some background of the genetic analysis of CaM must be given. These studies show CaM to be essential for cell viability and that ApoCaM serves an essential role(s).

Mutational analysis of S. cerevisiae was carried out to explore the function CaM plays during cell growth and division. Yeast CaM has all the attributes expected of a Ca²⁺ receptor. Calmodulins from yeast and vertebrates have structural, biochemical, and biophysical similarity (125) and are functionally conserved (35, 101). In addition to having a high affinity for Ca²⁺ (84) and changing conformation to reveal hydrophobic surfaces upon binding Ca²⁺ (36), yeast CaM activates both mammalian and yeast enzymes in a Ca²⁺-dependent manner (84, 105). Also, vertebrate CaM can substitute for yeast CaM in vivo (35, 101, 112). A Ca²⁺/CaM-dependent protein kinase (82, 94, 108) and a Ca²⁺/CaM-dependent protein phosphatase (31) have been identified and purified from yeast. Harris et al. (55) identified residues critical for CaM action in yeast cell growth by gain-of-function mutations in a human CaM-like protein.

Disruption of the unique gene encoding CaM in budding yeast (36), fission yeast (133), or Aspergillus nidulans (116) is lethal. Yeast cells with mutations that cause a complete loss of CaM function suffer many defects simultaneously. As a result, it is difficult to tell which are direct and which are secondary consequences of the loss. Furthermore, the more dramatic effects of the loss of CaM function may mask defects that are more subtle or slow to develop. Ohya and Botstein (103) used the techniques of yeast genetics to generate temperature-sensitive mutations that abolish single functions of CaM, while leaving the others intact. Examination of 14 temperature-sensitive yeast mutants bearing one or more phenylalanine to alanine substitutions in the single essential CaM gene of yeast (CMD1) revealed diverse essential functions. Their mutations were classified into four intragenic complementation groups. Each group showed different characteristic functional defects in actin organization, CaM localization, nuclear division, or bud emergence.

Calmodulin performs two essential functions in S. cerevisiae (16, 34, 103, 104, 129). Calmodulin plays an important role in polarized yeast cell growth via an interaction with an unconventional type V myosin, Myo2p (17). An essential mitotic function requires an interaction between CaM and a 110-kDa protein component of the spindle pole body, the yeast equivalent of the microtubule organizing center (47, 126). Very recently, Moser et al. (95) reported that CaM is localized to the spindle pole body of Schizosaccharomyces pombe and performs an essential function in chromosome segregation.

The terminal phenotype of a CaM-deficient S. cerevisiae indicated difficulty in nuclear division (102). In S. pombe, a CaM R116F substitution results in a proteolytic attack of this mutant CaM, and resultant CaM shortage causes difficulties in sporulation (132).

The above studies show that in yeast CaM is essential and serves important roles in the cell cycle. The same is also true in other organisms and cell types. Increased CaM concentration transiently accelerates proliferation, and antisense RNA-induced decrease in CaM transiently arrests cell cycle progression in mouse C127 cells (115). Calmodulin concentration doubles at the G₁/S boundary in cultured mammalian cells (21). Braam and Davis (14) found that the transcription of CaM and similar genes increased dramatically and rapidly upon mechanical stimulation of Arabidopsis and related this induction to thigmomorphogenesis. Depletion of Aspergillus nidulans CaM causes a failure to progress from a nimT23 cell cycle block at the G₂ to M phase boundary (83).

Diversity of functional consequences of CaM mutations has been observed previously among CaM mutants of Paramecium (70). Kink et al. (70) examined CaM and its gene from the wild-type and viable mutants of Paramecium tetraurella. The mutants selected for their behavioral aberrations had little or no defects in growth rate, secretion, excretion, or motility. They could be grouped according to whether they underreact or overreact behaviorally to certain stimuli, reflecting their respective loss of either a Ca²⁺-dependent Na⁺ current or a Ca²⁺-dependent K⁺ current. Their results indicated that the sites defined by these mutations are important in membrane excitation but not in other biological functions.

Thus CaM is essential in a disparate group of organisms, but is this due to a role of Ca²⁺-CaM or ApoCaM?

Mutant S. cerevisiae CaM that do not bind detectable levels of Ca²⁺ support normal rates of cell growth and division (48). They also localize in a manner identical to wild-type CaM throughout the cell cycle (16). Vertebrate CaM with analogous mutations also support normal growth of S. cerevisiae, indicating a Ca²⁺-independent function fundamental to the CaM molecule (48). Moser et al. (96) did an analysis of CaM in the fission yeast S. pombe to explore the universality of a Ca²⁺-independent function for CaM during cell proliferation. They assessed the abilities of both Ca²⁺-binding site mutants and heterologous CaM to support proliferation. Saccharomyces pombe requires at least one intact Ca²⁺-binding site in contrast to S. cerevisiae. Substitution of a valine for the conserved glutamic acid in position 12 of all four of the EF-hand Ca²⁺-binding sites of CaM yielded a protein that failed to support proliferation. The same mutant S. pombe protein allowed the growth of S. cerevisiae. The importance of Ca²⁺ binding may reflect essential Ca²⁺-dependent CaM functions in S. pombe not present in S. cerevisiae. Alternatively, essential functions that are not Ca²⁺ independent in S. cerevisiae may be Ca²⁺ dependent in S. pombe.

The fact that S. pombe does require Ca²⁺ binding by CaM was exploited to assess the importance of each Ca²⁺ binding site in vivo. Moser et al. (96) also analyzed the Ca²⁺-binding properties of S. pombe CaM by NMR and showed that the relative affinity of each site for Ca²⁺ does not parallel the functional importance of that site. Furthermore, despite the observed differences in the Ca²⁺-binding properties of CaM from S. pombe and vertebrates, CaM from vertebrate sources can substitute for that of S. pombe. Thus, although CaM serves an essential role, it seems likely that ApoCaM may sometimes be what is essential and Ca²⁺ binding, while required for many CaM functions, may not always be prerequisite to life itself.

There are certainly many candidates for essential cellular functions of CaM that may not require Ca²⁺ binding to the protein. Numerous studies suggest that CaM is required for microtubule function, and it has been demonstrated that the ability of CaM to bind to the mitotic apparatus is not dependent on the presence of Ca²⁺ (131). Endocytosis in yeast requires CaM but is apparently Ca²⁺ independent (75). Calmodulin may regulate a protein phosphorylation cascade that results in chromosome segregation (37, 68, 117). However, the Ca²⁺ transients proposed to trigger this cascade were later found not to be required for chromosome segregation (67). Calmodulin may also regulate a nuclear contractile system that includes actin in proliferating liver cells (8). In contrast to Ca²⁺-CaM, ApoCaM binds to and can regulate a variety of proteins whose target sequences are far more restricted. Neuromodulin, neurogranin, and unconventional myosins (e.g., intestinal brush-border myosin I) and several enzymes interact with CaM in the absence of Ca²⁺ (1, 10, 40, 45) binding CaM tightly at low Ca²⁺ concentration or in its absence (22, 148).

Thus Ca²⁺ binding is not necessary to some of the vital roles CaM serves in the cell, and understanding the roles of cellular ApoCaM is necessary to our understanding of the cell cycle and viability itself.

Calmodulin is a major constituent protein of the cytoplasm and membranes and is present in or on organelles (54, 62, 66, 77, 114). In the microscopic localization, CaM is found associated with membrane structures including the plasma membrane, nuclear membrane, and on the outer (cytoplasmic) surface of mitochondria and the endoplasmic reticulum. In the rat cerebellum, these structures were found to contain much of the CaM-antibody staining and, at the synapse, the postsynaptic membrane was found to contain abundant CaM (77).

These localization studies have also been complemented with cell fractionation experiments (62, 66, 134). One advantage of cell fractionation is that it allows testing the Ca²⁺ dependence of binding to these cell constituents. The particulate fraction of tissues often contains much of the total CaM present in the cell, an observation that supports the microscopic localization to membranes. This has been particularly well characterized in the brain (134). Homogenization of brain in Ca²⁺-containing buffers followed by centrifugation at 105,000 g results in recovery of ~42% of the cell's CaM in the particulate fraction. Further fractionation reveals that the majority of this CaM is in the microsome fraction, with a sizable portion also found along with mitochondria. Extraction of this microsome fraction repeatedly with EGTA shows that ~70% is reversibly bound in a Ca²⁺-dependent manner but that the remaining 30-35% is not extractable with EGTA and is only solubilized after treatment with nonionic detergents such as Lubrol (66, 134) or Triton (124). Overall, in brain, ~15% of the cell's CaM is bound to membranes in this Ca²⁺-independent bound fraction (134). Calmodulin is ~0.5% of the total protein present in brain; thus bound ApoCaM is a major brain protein (134). Because this fraction is not released from membranes after repeated extraction with EGTA or EDTA, we refer to it here as particulate ApoCaM.

When a variety of tissues were assayed, the percentage of total CaM found as particulate ApoCaM was shown to vary from a low of ~11% in testis to 63% in spleen. Several observations argue that this is an important pool of cellular CaM. It probably does not represent entrapment, since cytoplasmic marker enzymes have shown to not copurify with it (124). By several criteria, the purified, particulate ApoCaM is identical to soluble CaM in its properties, suggesting that it is the same protein tightly bound to the membrane in a Ca²⁺-independent fashion (124). This particulate ApoCaM may represent storage of CaM that is released as part of the cell cycle. The fraction of the cellular CaM in this particulate CaM pool varies with cell density in tissue culture and is different in cancerous and noncancerous cells. Chinese hamster ovary cells were grown to a density of 2-9 × 10⁶ cells/dish and showed that as cell number increased there was an increase in soluble CaM. Soluble CaM increased twofold, whereas the particulate fraction decreased by one-third (41). More soluble CaM is extracted from cancer cells than from noncancerous cells (122, 138, 145), whereas the particulate ApoCaM fraction is more abundant in noncancerous cells than cancerous ones (138). In the case of certain hepatomas, the ratio of soluble CaM to particulate ApoCaM is changed from 0.8 in normal liver to 9.8 in hepatoma (138). These results are consistent with the hypothesis that soluble CaM, necessary for rapid cell growth, is derived to some extent from the particulate CaM pool.

The binding site for this ApoCaM has not been rigorously identified; however, in brain it seems likely this binding may be primarily the activity of two proteins, neuromodulin and, to a lesser extent, neurogranin. Neuromodulin is a 57-kDa protein found only in neural tissue that binds CaM in both the presence and absence of Ca²⁺. Cimler et al. (25) found that the particulate fraction of bovine cerebral cortex contained 61 pmol neuromodulin, and Kakiuchi et al. (66) found 129 pmol particulate ApoCaM/mg membrane protein. Thus this area of brain contains enough neuromodulin to bind about one-half of the CaM that remains after extensive EGTA washing. Neurogranin is present in brain at only 2-5% of the mass of neuromodulin (10) but is about sevenfold lower in molecular mass so it could bind about one-third as much CaM as neuromodulin. Neurogranin and neuromodulin are both confined to nerve tissues, both have a quite similar IQ motif CaM binding site, both are phosphorylated by protein kinase C, and in both cases, the phosphoprotein no longer binds CaM.

It is clear that regulation by CaM may be exerted both independently and dependently of Ca²⁺ (118). The transformation from Ca²⁺-independent to Ca²⁺-dependent binding is apparently due to subtle substitutions in the recognition motif. Calmodulin can therefore perform under highly localized, fluctuating Ca²⁺ concentrations that require Ca²⁺-dependent and -independent interactions for membrane cycling, movement, and regulation. Detailed studies on ApoCaM promise to yield insights into the mechanism of activation for the entire family of EF-hand Ca²⁺-binding proteins. This review deals with the structural evaluation, target proteins, and recognition motifs of ApoCaM.

	II.  STRUCTURE OF APOCALMODULIN

Top Previous Next References

Calmodulin is an internally homologous protein. The complete sequence can be divided into roughly four segments, and each one-fourth of the structure is homologous to the others. The first (amino terminal) quarter is most similar to the third, and the second is most similar to the fourth (carboxy terminal). This structure is thought to have arisen by gene duplication of an ancestral single Ca²⁺-binding domain protein to give a two domain ancestor first, which again duplicated to give the current four domain protein (144). Each of these four domains consists of an -helix, a loop where Ca²⁺-binding occurs, and then a second -helix. The helices orient the loops for Ca²⁺ binding. This structure is the "EF hand" Ca²⁺-binding motif named for the Ca²⁺-binding site between the E and F helices of parvalbumin (73). Thus CaM having four EF-hand Ca²⁺ binding sites would be expected to have a total of eight -helical segments. This is essentially correct except that the fourth and fifth helices combine to form a long "central helix."

The Brookhaven database currently has 30 structure files relevant to CaM. A complete discussion of all this structural data is beyond the scope of this review, but some generalizations may help the reader appreciate the overall structure and the effects of Ca²⁺ binding and protein binding on structure. Figure 2 shows the structure of Ca²⁺-CaM and ApoCaM in similar orientation. The overall shape of CaM is often referred to as a "dumbbell." Each half (consisting of 4 -helices and 2 Ca²⁺-binding sites) folds into a compact globular structure (where the weights of a dumbbell would be) and the fused central helix makes a connecting "bar" (7). This coarse description is sufficient for many purposes, but a more useful description is of a dumbbell with two ends shaped more like soup ladles, the bowl of the ladle ends forming a hydrophobic "cup" in Ca²⁺-CaM (42, 74) and connected by a "flexible tether" instead of a bar (109-111). This description is accurate only for Ca²⁺-CaM, and the differences in ApoCaM are significant.

View larger version (46K): [in this window] [in a new window]   Fig. 2. Comparison of Ca2+/calmodulin (CaM) and apocalmodulin (ApoCaM). Left: rat testis Ca2+/CaM [Brookhaven National Laboratory protein database (PDB) file 3CLN]. Right: Xenopus ApoCaM (PDB file 1CFC). (Species differences are unimportant, since all vertebrate CaM are identical.) Structures were analyzed using Swiss PDB Viewer software. Structures are shown with amino-terminal (N) domains in similar orientation. C, carboxy terminal.

The differences between Ca²⁺-CaM and ApoCaM are numerous, but a few useful generalizations can be made.

1) Apocalmodulin is overall a more compact structure. Calcium-CaM has an open, "relaxed" structure. Figure 2 shows a depiction of both forms with the amino-terminal globular ends in similar orientation.

2) The position of the hydrophobic residues in the two structures is perhaps most striking. In ApoCaM, these hydrophobic residues predominantly pack together between -helical segments in each globular end and form a hydrophobic core in the interior of the structure. In contrast, in Ca²⁺-CaM, these residues swing out from the space between helical segments and form a hydrophobic cup structure at each ladle end (facing out, adjacent, and perpendicular to the central helix) accessible from the outside.

3) In ApoCaM, nearly half a dozen residues in the amino and carboxy terminus (e.g., from 8 to 15 and 140 to 146), which form the first and last -helices of CaM, are roughly parallel to the central helix of CaM, providing a "second bar" to its dumbbell. Furthermore, the central helix is interrupted in the center (at Ser-81) by a nonhelical region that is bent. In contrast, these terminal peptides are swung out away from the central helix in Ca²⁺-CaM, and the central helix is not bent in the crystal structure (see Fig. 2). This allows access to the ladle "hydrophobic cups" at each end of Ca²⁺-CaM that were blocked off by these terminal segments in ApoCaM. The two globular ends of ApoCaM are also twisted by about 180° relative to each other when compared with their position in Ca²⁺-CaM (e.g., Gly-40 and Leu-112, residues in different globular ends, are on the same side of the molecule in ApoCaM and roughly 180° apart in Ca²⁺-CaM). This twisting of the two ends and the position of the terminal helices is illustrated in Figure 2.

4) In summary, ApoCaM has a hydrophobic core, which becomes exposed as two hydrophobic faces in Ca²⁺-CaM.

Calcium-CaM binds to and activates smooth muscle and skeletal muscle MLCK and CaM kinase II. In all three cases, the region of the enzyme sequence that binds CaM has been identified (59, 89, 90). The structure of the complexes of Ca²⁺-CaM with peptides representing each of these sequences has been solved and shows the significance of these differences between Ca²⁺-CaM and ApoCaM. When each peptide binds, the hydrophobic faces and surrounding sequences at each end of Ca²⁺-CaM contact the peptide and have many favorable interactions with it. To do this, the central helix bends. Indeed, a useful concept is that the central helix serves as a flexible tether between these two globular ends (110), positioning them for favorable interactions with enzyme binding sites. Thus the inaccessibility of these hydrophobic residues in ApoCaM and the blockage by the amino- and carboxy-terminal helices mentioned above probably account for why Ca²⁺-CaM binds to and activates these enzymes while ApoCaM is ineffective.

What happens when ApoCaM is bound by proteins adapted for its binding is unknown. The structure of a fragment representing the regulatory domain of scallop myosin with two light chains (one essential and the other regulatory) bound has been reported (151). Because these light chains are homologous to CaM and remain associated when myosin is isolated from EGTA solution, this binding may be analogous to ApoCaM binding. Indeed, there are similarities between the regulatory light chain and both the Ca²⁺-CaM and ApoCaM structures. For example, the regulatory light chain is in a conformation that resembles a state intermediate between Ca²⁺-CaM and ApoCaM in that many of the hydrophobic residues still form a core structure at each end, but some are also directed toward and in contact with the myosin sequences. The bound structure is more relaxed as in Ca²⁺-CaM, and the terminal "blocking" sequences are oriented out away from the central helix in a nonblocking position. The bound region of myosin is -helical, which is also true of peptides when bound by Ca²⁺-CaM (59, 88, 89); however, in the case of myosin, the region of peptide -helix in contact with myosin light chains is longer and the contacts more extensive. Recently, Houdusse and Cohen (57) presented a model for ApoCaM binding based on these scallop myosin results. The scallop sequence contains an IQ motif, and IQ motifs do indeed bind ApoCaM (23, 118). The model shows a more open CaM structure with the blocking termini in a nonblocking position away from the central helix. Apocalmodulin is shown as interacting over a longer length of protein sequence than do the Ca²⁺-CaM binding models and with sequences within the IQ motif itself and other sequences both amino and carboxyl terminal to it. The authors also suggest that Ca²⁺ dependence or independence of binding may result from interactions with the carboxy terminal half of the IQ motif (57). Because the model (57, 58) is theoretical and based on homologous myosin light chain binding data, its accuracy is not known, although it seems plausible. However, other studies have shown that CaM can interact with proteins in recognizably different ways. More than a single sequence can bind to CaM at the same time (32, 76), and CaM does not always assume the same conformation while bound to different recognition sequences (136). Thus bound (Apo)CaM may assume conformers about which we currently know little. Although the Houdusse and Cohen model is clearly a first step toward understanding ApoCaM binding, there are almost certainly bound states of CaM about which we currently know little.

	III.  KNOWN APOCALMODULIN BINDING MOTIFS

Top Previous Next References

Currently known ApoCaM binding falls into two classes: 1) the IQ motif found in myosins, neuromodulin, neurogranin, and other proteins; and 2) a concerted binding exemplified by CaM binding to the catalytic () subunit of glycogen phosphorylase b kinase (PbK) and probably also the prokaryotic adenylyl cyclase toxins. Although these represent the currently known themes, there are likely others, since proteins are known that bind ApoCaM with no obvious adherence to either (99).

The sequence IQXXXRGXXXR (1-letter amino acid abbreviations where X denotes any amino acid) defines this motif, but the motif is rather loosely adhered to in many cases. For example, the isoleucine in the first position is frequently a different branched chain amino acid (Leu or Val) or, rarely, a methionine. The arginines in both the sixth and the terminal position are sometimes lysine or histidine, and the seventh position glycine is frequently not conserved. Despite this lack of strict conservation, there is little doubt that this is a recognizable protein motif that binds CaM or homologous proteins including troponin C and myosin light chains. An extensive compilation of observed IQ motifs, many known only by homology, is given by Rhoads and Friedberg (118). IQ motifs were first identified as ApoCaM binding sites; however, it is now clear that IQ motifs can show different levels of Ca²⁺ dependency. Several selected IQ motifs of different types are shown in Table 3.

The IQ motif is repeated one to six times in muscle and nonmuscle myosins (23). Normal muscle myosins (i.e., myosin II) contain a globular ATPase domain connected to a helical tail region by a short "neck" domain. Nonmuscle myosins (i.e., myosin I) have a similar head and neck structure, although the tail domain may be nonhelical and possess binding properties not normally associated with myosin. IQ domains are found in this neck region of myosins. In normal muscle myosin, the IQ motifs provide binding sites for essential and regulatory myosin light chains, and thus this neck region is thought to be critical for regulation of myosin's biological function. Because myosin light chains, troponin C, and CaM all belong to the same gene family and have similar structures, binding by alternate members within this family has been observed.

The IQ motif is also sometimes adjacent to or overlapping phosphorylation sites phosphorylated by protein kinase C and dephosphorylated by calcineurin and phosphorylation inhibits CaM binding in vitro (5).

The Ca²⁺ sensitivity of CaM binding by IQ domains appears to be highly variable. For example, neuromodulin binds to CaM-Sepharose in the absence of Ca²⁺ and is eluted with Ca²⁺-containing buffers (4). Thus neuromodulin binds preferentially to ApoCaM. In other cases such as the IQ-GAP2 proteins and the Myr-4 myosin, binding may be Ca²⁺ dependent or Ca²⁺ independent, sometimes at adjacent sites within the same protein.

Another type of ApoCaM binding is that exemplified by PbK and probably the prokaryotic adenylyl cyclase toxins. Phosphorylase b kinase in skeletal muscle is activated by cAMP or Ca²⁺. The active enzyme phosphorylates glycogen phosphorylase b and converts it to an active form that catalyzes the phosphorolysis of glycogen to glucose-1-phosphate. Activation by cAMP was characterized early (72), but the activation by Ca²⁺ was not well understood until CaM was discovered to be a subunit of this enzyme (28). Glycogen phosphorylase b kinase has the subunit structure ()₄, where  is CaM. The -subunit CaM remains tightly associated even in the presence of chelators and requires strongly denaturing conditions to remove CaM. The rate of exchange with ¹⁴C-labeled CaM is only ~15% per week at 4°C and 2 mM EDTA (113) and is also indicative of tight, Ca²⁺-independent association. This CaM subunit is tightly bound by the catalytic -subunit in the holoenzyme. The -subunit has an amino-terminal presumptive catalytic domain (residues 20-276) homologous to other protein kinases and a carboxy-terminal domain (residues 276-386) that is presumed to be a regulatory region of the enzyme. This putative regulatory region was synthesized in its entirety as a series of 18 overlapping 25-mer peptides, and each was assayed for its ability to bind CaM in the presence of Ca²⁺ and inhibit a MLCK assay or to inhibit the CaM-facilitated renaturation of denatured -subunit (32). On the basis of these criteria, CaM binding was found to reside in two noncontiguous regions of sequence centered at two peptides named PhK5 (residues 342-366) and PhK13 (302-326). Both peptide sequences are shown in Table 3. The binding of either peptide is Ca²⁺ dependent, and both peptides can bind CaM simultaneously (32). Interestingly, antibodies against PhK5 and PhK13 have the ability to activate and inhibit, respectively, PbK holoenzyme and the isolated -subunit complex (143). Low-angle scattering data also suggest different modes of binding of each to Ca²⁺-CaM. It should be remembered from a previous section where CaM's structure is discussed that when CaM binds peptides representing binding domains of myosin light chain kinase (MLCK) or CaM kinase II, CaM changes from an extended dumbbell to a more compact structure as the central helix bends and the two globular ends of CaM associate with the peptide. This contraction upon binding can be observed by low-angle scattering. PhK5 causes a contraction of CaM similar to that found with the MLCK peptide, but PhK13 binds to an extended conformer of CaM. When both peptides are bound, CaM is in an extended conformer (136). Thus the binding of CaM to PbK appears to be fundamentally different than its binding to MLCK or CaM kinase II.

Thus, although high-affinity binding to either PhK5 or PhK13 (i.e., 20 nM and 6.5 nM, respectively) (32) is Ca²⁺ dependent, binding of the two is synergistic, and although the individual affinities in the absence of Ca²⁺ may be low, the synergism of binding makes binding to the -subunit resistant to changes in Ca²⁺. For example, if two independent binding events occur with different regions of sequence and binding is noncooperative, the individual affinities would multiply. Thus, even if each region binds in the absence of Ca²⁺ with quite low affinities (e.g., 10⁴ M), the product of two such independent binding events (e.g., 10⁸ M) could be substantial and yield a quite stable complex making strong denaturation necessary for dissociation. Thus Ca²⁺ binding can still affect each individual binding event and regulate activity while maintaining a stable complex with the -subunit independent of Ca²⁺.

A similar mechanism may pertain to the Ca²⁺-independent binding of CaM to the prokaryote adenylyl cyclase toxins produced by Bordetella pertussis and Bacillus anthracis, the causative agents of whooping cough and anthrax, respectively. Both toxins bind CaM in the presence and absence of Ca²⁺. Because prokaryotes do not produce CaM, presumably the requirement for CaM here serves the purpose of ensuring that catalytic activity is maximal only within the eukaryotic host cell where these toxins function. These two adenylyl cyclases are homologous to one another and markedly different from mammalian adenylyl cyclases. More is known about the B. pertussis enzyme. The protein secreted by the bacteria has a molecular mass usually reported to be ~200,000 Da. This form of the enzyme has been highly purified from an overexpressing B. pertussis strain, the molecular mass is 176 kDa, and it is activated half-maximally by 0.3 nM CaM in the presence of Ca²⁺ and by 20 nM in the absence of Ca²⁺ (49). This form of the enzyme is capable of entering the eukaryotic cell in a process that requires extracellular Ca²⁺ (49). Because Ca²⁺ binds to the enzyme and affects its physical and chemical properties (88), extracellular Ca²⁺ may be required to produce the conformer that crosses into the cell. Posttranslational modification is apparently also required, since expression of the recombinant gene in Escherichia coli produces a protein which, while catalytically active and CaM dependent, cannot enter lymphocytes while the recombinant gene expressed in B. pertussis can (119). Apparently, two different forms of the enzyme are produced by B. pertussis, which are both ~200 kDa. Both are catalytically active, CaM activated, and immunologically similar but can be separated by gel filtration chromatography. One of these forms enters lymphocytes, whereas the other cannot (119).

Why the adenylyl cyclase can bind ApoCaM (as opposed to Ca²⁺-CaM) may be adaptive to the conditions inside the eukaryotic cell. The CaM concentration inside eukaryotic cells is often in the range of 1-10 µM, and unstimulated cells typically have free concentrations of Ca²⁺ of 0.1 µM or less; thus the unstimulated cell would have abundant ApoCaM, and it is under these conditions that the toxin adenylyl cyclase must be active. Because prokaryotes like B. pertussis do not produce CaM, the CaM requirement ensures that the enzyme obtains high activity only when it leaves the bacteria and enters the host. This is an interesting system in which eukaryotic CaM serves as a "fail safe" mechanism reminiscent of zymogens that ensures that only the exported protein becomes fully catalytically active.

The domain organization of the enzyme shows two CaM binding sequences reminiscent of PbK. Although the intact adenylyl cyclase is ~200 kDa, shorter secreted forms (43-61 kDa) have been frequently isolated which are CaM activated and catalytically active, and derived from the amino terminus of the holoenzyme. The 43-kDa form has been particularly well studied. It binds Ca²⁺-CaM with an apparent dissociation constant of 0.14 nM and binds CaM in the absence of Ca²⁺ (i.e., in 2 mM EGTA, pH 8) with an apparent dissociation constant of 15 nM. Both of these apparent binding constants are so much lower than the CaM concentrations usually measured in eukaryotic tissues, and the enzyme would likely bind CaM independent of Ca²⁺ in vivo. When this 43-kDa form is trypsin-digested in the presence of CaM, cleavage gives rise to a 25- and 18-kDa fragments. Both fragments remain associated with CaM, and this trimer is catalytically active (76). The 25-kDa fragment is from the amino terminus of the 43-kDa form and thus from the amino terminus of the holoenzyme. It has a low catalytic activity in the absence of the 18-kDa carboxy-terminal fragment and represents much of the catalytic domain of the enzyme (50). Trypsin digestion of the enzyme in the absence of CaM rapidly inactivates the enzyme and gives rise to a 24-kDa fragment which, unlike the 25-kDa fragment, does not apparently bind CaM (76). The 18-kDa form (residues 236 or 238-399) has no catalytic activity and represents a regulatory region of the enzyme. It contains Trp-242, which has been shown by site-directed mutagenesis to be essential for high-affinity CaM binding. The fragments associate only in the presence of CaM to restore much of the original catalytic activity of the undigested enzyme (50).

Thus, like PbK, the B. pertussis adenylyl cyclase apparently has at least two regions, present in the 25- and 18-kDa peptides, that bind CaM, and this probably accounts for the high affinity for CaM in the presence or absence of Ca²⁺. A significant difference, however, is that CaM should properly be considered a subunit of PbK, whereas the adenylyl cyclase must acquire its CaM from its host.

Inducible nitric oxide synthase also has at least two sequence regions that bind CaM and possibly represents a further case of this concerted binding mechanism but is discussed separately in section IVC with the other NOS isozymes.

Although IQ motifs and noncontiguous sites represent two known types of ApoCaM binding, there must certainly be others. Proteins have been discovered that bind ApoCaM for which there is no recognizable motif. One example recently encountered is the syntrophins (99).

Syntrophins were first discovered in the postsynaptic membrane of the Torpedo electric organ (44) and are constituent proteins of a membrane complex, the dystrophin glycoprotein complex (DGC), which is defective in various forms of muscular dystrophy. There are ₁-, ₁-, and ₂-syntrophins that are ~50% identical with one another in sequence. They differ somewhat in their charge properties with ₁-syntrophin being more acidic than the more basic -syntrophins (135). The -syntrophins are found in a variety of nonmuscle tissue including nerve, and syntrophins also bind to Na⁺ channels isolated from either brain or muscle (46).

Syntrophins have a complex domain structure consisting of a syntrophin unique (SU) domain at the carboxy terminus, two pleckstrin homology (PH) domains, and a single PDZ domain. The PH domains were discovered in the major protein kinase C substrate protein of platelets, pleckstrin. This domain is found in several membrane-associated proteins. The PDZ domains are found in membrane-associated proteins such as the postsynaptic density 95-kDa protein and nNOS [see Gee et al. (46) for discussion]. Syntrophins bind to CaM (87), to dystrophin (the gene product of the Duchenne muscular dystrophy gene) (39, 71), and to F-actin (60), and they also self-associate (86, 153). Calcium-CaM inhibits the syntrophin-dystrophin interaction (99). In one study (99), CaM was found to bind to syntrophin in the presence and absence of Ca²⁺. One binding site is the carboxy-terminal 24 residues of ₁-syntrophin (CBS-C, Table 3), a sequence highly conserved in all syntrophins. Calmodulin binding to this site was observed only in the presence of Ca²⁺. A second kind of binding was localized to ₁-syntrophin-(4---174), which bound either Ca²⁺-CaM or ApoCaM (99). Somewhat different results were found by another group. They found no binding within the SU domain and found two other sites at ₁-syntrophin-(1---38) and -(106---126) (see Table 3). All binding was found to be Ca²⁺ dependent (60). Several observations may be relevant to this discrepancy, some published and others not. We have found that our His-Tag syntrophin fusion protein is rapidly degraded by E. coli proteases and that the purified proteins are proteolysed primarily at the carboxy terminus. This proteolysis has been described (99), and a rapid batch purification procedure was developed (63, 85) to lessen proteolysis. Unless great care is taken, the SU domain and the carboxy-terminal CaM-binding site it contains may become truncated, and CaM binding would not be detected. The shortest SU domain construct used by Iwata et al. (60) was about 49 kDa, and truncation of the complete 24-residue CaM-binding sequence would alter the molecular mass by only ~5%, which would have been difficult to detect. Both groups agree that CaM binds within syntrophin sequences 4-174. Newbell et al. (99) did not determine the precise location or number of these sites, whereas Iwata et al. (60) localized two sites to ₁-syntrophin-(1---38) and -(106---126). The two groups disagree on whether the site(s) binds CaM only in the presence of Ca²⁺ (60) or in both the presence and the absence of Ca²⁺ (99). However, Newbell et al. (99) had presented data that syntrophins aggregate, that aggregation is affected by Ca²⁺, and that aggregation affects CaM binding. More recently, we have found that syntrophins self-associate to form large aggregates that pellet in the ultracentrifuge. Aggregation is affected by micromolar Ca²⁺ concentrations, and these aggregates no longer bind CaM in amounts stoichiometric with syntrophin monomer (Oak and Jarrett, unpublished data). Thus CaM binding can be affected by aggregation, and the Ca²⁺ dependence of aggregation can interfere with CaM binding as had been proposed by Newbell et al. (99). There may be two CaM sites in this sequence region as determined by Iwata et al. (60). However, the effect of Ca²⁺ on aggregation (which inhibits CaM binding) can make ApoCaM binding appear to be Ca²⁺ dependent. Indeed, Newbell et al. (99) showed that fusion proteins containing only these ApoCaM binding sequences bind to CaM-Sepharose in Ca²⁺ and elute in EGTA because of this aggregation phenomenon. However, in either case, the sequence(s) within syntrophin-(4---174) do not contain either an IQ motif or other readily discernible Ca²⁺-CaM binding sequences. Such Ca²⁺-CaM binding sequences are basic, hydrophobic, and frequently aromatic-rich stretches of ~20 amino acids in length. Acidic residues and prolines within this amphipathic, -helical region are rare (see Table 3). The sequences identified by Iwata et al. (60) are compared with several Ca²⁺-dependent and Ca²⁺-independent binding sites that have been identified in Table 3. Except for short stretches, these sequences differ from this Ca²⁺-dependent paradigm and from the IQ motif. Interesting is the appearance of a proline in the midst of the most basic sequence regions in each sequence and the paucity of aromatic residues. Thus either syntrophin contains a previously unknown type of ApoCaM binding site as we have suggested (99) or they contain a new type of Ca²⁺-CaM binding that is currently unknown. In either case, much will be learned about CaM binding by further study of syntrophin.

	IV.  APOCALMODULIN-BINDING PROTEINS

Top Previous Next References

Usually the binding of CaM to proteins is Ca²⁺ dependent, but ApoCaM binding occurs with various actin-binding proteins (e.g., myosins), cytoskeletal and membrane proteins, enzymes and channels, and receptors.

Unconventional myosins represent a good example of Ca²⁺-independent CaM binding. These proteins contain a "generic" motor domain that converts chemical energy into directed mechanical force by hydrolyzing ATP, binding actin, and translocating along actin filaments (9, 23). In contrast to conventional muscle myosins, where two different light chains stabilize the regulatory domain (RD) in the neck region of the molecule, unconventional myosins are a diverse group of molecular motors in which one to six regulatory light chain subunits or CaM are bound tandemly to the RD (58). Intestinal brush-border myosin I (BBMI), an unconventional myosin from vertebrates that binds three or four CaM, has been found to retain its ability to bind these molecules even in the absence of Ca²⁺ (130). Evidence that Ca²⁺ lowers CaM's affinity for BBMI came from studies showing that this ion seems to alter the proteolytic digestion of the myosin heavy chain. Calmodulin activates the Mg²⁺-ATPase activity of BBMI in the absence of actin, and it inhibits in vitro motility. Brush-border myosin I has several IQ motifs in its neck region RD that is thought to be responsible for CaM binding. There is a family of similar CaM-binding unconventional myosins in other vertebrate tissues including rat kidney brush borders, two from brain, adrenal cortex, and smooth muscle (23).

Recently, a novel subclass of myosin I (myr4) was identified from rat brain, demonstrated to bind CaM, and showed by sequence analysis to be not closely related to BBMI (9). Calmodulin's association with myr4 was demonstrated by detection of CaM in preparations of affinity purified myr4 by immunoblotting. Immunoprecipitation with myr4 antibodies also yields a band comigrating with authentic CaM. With the use of fusion proteins for myr4, encompassing various portions of the two IQ motifs and CaM overlay assays, it was found that both myr4 IQ motifs bind CaM. The amino-terminal IQ motif (amino acids 700-721) binds CaM only weakly in the presence of free Ca²⁺ but strongly in the absence of free Ca²⁺. In contrast, the carboxy-terminal IQ motif (amino acids 722-743) binds CaM strongly in the presence of free Ca²⁺, but only weakly in the absence of free Ca²⁺. Thus two adjacent IQ motifs differ in the Ca²⁺ requirements for optimal CaM binding. Studies with the myr4 fusion protein that contained both IQ motifs (amino acids 500-743) showed the binding of CaM in the absence and presence of free Ca²⁺ in a manner equivalent to the sum of the two separate IQ motifs. These results demonstrate that IQ motifs can exhibit drastically different dependencies on Ca²⁺ for CaM binding; IQ motifs can bind CaM in a Ca²⁺-independent manner, and with a few differences in sequence, become Ca²⁺ dependent (9).

Other CaM-binding unconventional myosins are Dilute, p190, and MYO2. These proteins share certain structural features. The biochemistry of the p190-CaM complex from vertebrate brain has been the best characterized. P190, which is also present in several nonbrain tissues, was originally identified as a CaM-binding protein enriched in brain actomyosin preparations, and partially purified from EGTA extracts as a complex with regulatory light chains. To map the CaM-binding domain of p190, sequences comprising the head, tail, and neck domains were expressed as bacterial fusion proteins and probed in an ¹²⁵I-labeled CaM gel overlay assay. The results showed that the fusion protein containing the neck domain of p190 exhibits substantial CaM binding activity in the absence of Ca²⁺, whereas no detectable binding of CaM was observed for the head or the tail domain fusion proteins. These results suggest that CaM binding sites of p190 map to the IQ motifs contained within this neck region (40).

It is clear from all these studies that the IQ motifs of unconventional myosins are the regions responsible for binding of ApoCaM or Ca²⁺-CaM and that the neck region of myosins is the most variable region of the molecule. It is also clear that whether an IQ domain binds CaM in a Ca²⁺-dependent or -independent fashion depends on modifications of this basic sequence motif that are only beginning to be characterized.

Cytoskeletal and membrane proteins also bind CaM in the absence of Ca²⁺. Syntrophin, a peripheral protein of sarcolemma, is discussed in section IIIC. Neuromodulin (also referred to as P-57 or GAP-43) is a neurospecific, membrane-associated protein. Palmitoylation of two cysteine residues accounts for its membrane association (156). Neuromodulin binds to CaM with at least 10-fold higher affinity in the absence of Ca²⁺ than in its presence (25). This protein was originally detected by cross-linking experiments between azido ¹²⁵I-CaM with detergent-solubilized cerebral cortex membranes in the presence of excess EGTA (4). Radioimmunoassay experiments also demonstrated the presence of this protein in retina and spinal cord (25). Because of its unusual higher affinity for ApoCaM compared with Ca²⁺-CaM, an easy purification protocol using CaM-Sepharose affinity chromatography was developed. This involves binding of neuromodulin to CaM-Sepharose in the presence of excess EGTA, a Ca²⁺ chelator, and elution from the affinity column by buffers containing excess of Ca²⁺ (4). Purification to apparent homogeneity was accomplished by solubilization from membranes, DEAE-Sephacel chromatography to remove endogenous CaM, and two CaM-Sepharose columns (±Ca²⁺). Neuromodulin is the only protein in the membrane preparation with higher affinity for ApoCaM than Ca²⁺-CaM. Neuromodulin binds a single CaM (4). Regulation of CaM binding to neuromodulin is exerted by phosphorylation at a serine residue (Ser-41, see Table 3) in a Ca²⁺-dependent manner and phosphorylation prevents neuromodulin binding to CaM-Sepharose. Also, CaM decreases the rate of phosphorylation of neuromodulin, suggesting that a residue nearby or involved in the binding of CaM to neuromodulin is the one modified (1). Identification of the phosphorylation site by using recombinant neuromodulin gave as a result a peptide of the sequence IQASFR. Serine-41 is the residue phosphorylated, and this sequence is the amino terminus of neuromodulin's IQ motif (5), explaining why phosphorylation of neuromodulin dramatically lowers its affinity for CaM. It has been proposed that neuromodulin binds and concentrates CaM at specific sites within the cell and releases it locally in response to stimulation by phosphorylation and/or to increase in free Ca²⁺ (1, 22). Phosphorylation may be more important, since at physiological ionic strengths, the difference neuromodulin's affinity for CaM is only lessened about fivefold by Ca²⁺ (81). In addition, phosphoneuromodulin is a substrate for calcineurin (phosphatase 2B) and phosphatase 2A, indicating that the concentrations of CaM in neurons may be controlled by a phosphorylation/dephosphorylation cycle (5, 22). Characterization of the CaM binding domain to neuromodulin was also studied by constructing a deletion mutant neuromodulin, designated NM(CM), which lacks the IQ motif amino acids 39-55 and examined for binding to CaM-Sepharose (22). The wild-type neuromodulin adsorbed to CaM-Sepharose in the presence of excess of EDTA and eluted with Ca²⁺, while NM(CM) did not adsorb to CaM-Sepharose in the presence or absence of Ca²⁺. This resulted in the identification of the sequence Q³⁹ASFRGHITRKKLKGEKK⁵⁶ required for CaM binding; this domain contains the only phosphorylation site (Ser-41) that affects CaM binding to neuromodulin. The adjacent Phe-42 apparently interacts hydrophobically with CaM, an interaction that is disrupted by the introduction of negative charge to Ser-41. This proposal was also supported by studying the neuromodulin/CaM interaction after substitution of Ser-41 by Asp or Asn. The result is Asn mutant retains its affinity for CaM-Sepharose, while the Asp mutant does not bind CaM-Sepharose (22).

Neurogranin (or RC3) also binds CaM in the absence of Ca²⁺. This protein, purified from bovine forebrain, has a molecular mass of 7.8 kDa and forms dimers and higher oligomers in the absence of reducing agents (10). Database searches between the amino acid sequence of neurogranin and other protein sequences revealed a striking homology between neurogranin and the known phosphorylation site/CaM-binding domain of neuromodulin. On the basis of this homology, the possible interaction of neurogranin with immobilized CaM was tested (10). Neurogranin also binds to CaM-Sepharose in the absence of Ca²⁺ and is eluted with Ca²⁺. As was found with neuromodulin, CaM binding to neurogranin inhibits its phosphorylation by protein kinase C. Phosphorylation of neurogranin by protein kinase C and trypsin digestion results in a peptide, IQASFR, phosphorylated on the serine residue. This sequence is part of the conserved IQ motif present in both neuromodulin and neurogranin. Because of the strict conservation of the CaM binding and phosphorylation site in neuromodulin and neurogranin, the authors suggest that neurogranin might have a similar function of sequestering CaM at specific locations in the brain and releasing it in response to increase in Ca²⁺ and/or to phosphorylation by protein kinase C (10).

PEP-19, a small 7.6-kDa neuron-specific protein, which contains a consensus motif similar to the CaM binding domains of neuromodulin and neurogranin have also been demonstrated to bind CaM in a Ca²⁺-independent manner. Slemmon et al. (123) demonstrated that PEP-19 inhibits the CaM-dependent nNOS activity with an EC₅₀ of 8 µM, showing that PEP-19 binds CaM with moderate affinity. PEP-19 was shown to bind immobilized CaM in the absence and presence of Ca²⁺ with an apparent dissociation constant of 1.15 and 1.25 µM, respectively. Calmodulin binding was localized to an IQ motif at PEP-19 amino acids 36-60. Specific substitution of conserved amino acids within the consensus IQ motif of PEP-19, neurogranin, and neuromodulin abolish CaM binding. Interestingly, sequences upstream and downstream of the motif itself also affect binding. The PEP-19 IQ motif contains a Ser in the same relative position as the corresponding Ser in neuromodulin and neurogranin; however, in this case, the residue is not phosphorylated in vitro by protein kinase C, suggesting that PEP-19 constitutes a PKC-independent pathway for CaM regulation (123).

Igloo, a neuromodulin-related protein from Drosophila, binds CaM (98), but whether this interaction is preferentially Ca²⁺ dependent or Ca²⁺ independent was not directly demonstrated. Igloo and neuromodulin share some sequence similarities. Most striking are its three IQ motifs that are similar to the ones in neuromodulin. The three IQ modules of Igloo (IGL) were denoted V, T, and S for the amino acid abbreviation of the residue in the putative protein kinase C phosphorylation position found for neuromodulin. Database searches revealed additional homology of these three modules of IGL to PEP-19, neurogranin, and the cephalochordate polypeptide Ca²⁺ vector target protein. On the basis of this sequence homology, and the fact that protein kinase C phosphorylates the IQ motif regions in other proteins, recombinant fusion proteins containing glutathione S-transferase and regions of IGL were produced and tested as a substrate for PKC. Phosphorylation by PKC occurred only within the S module, and mutation of this Ser to an Ala prevented phosphorylation. Surprisingly, the threonine residue of IGL fusion protein used was not phosphorylated in vitro by protein kinase C. The GST-IGL fusion proteins were found not to bind vertebrate CaM-Sepharose. However, with the use of Drosophila CaM and the yeast two-hybrid system, interaction with IGL and with each individual IQ motif module was demonstrated.

The IQGAP proteins are homologous to the GTPase-activating proteins (GAP) of the small G proteins (such as Ras and Rho), which also contain four apparent IQ motifs. Human IQGAP1 and IQGAP2 (a liver-specific protein) are the currently known IQGAP proteins (146). Despite the homology, there is currently no evidence that they have GAP activity, although IQGAP2 binds Cdc42 and RacI. Recently, the possible interaction of the IQGAP2 four IQ motifs with calmodulin was studied (15). Immunoprecipitation studies and the yeast two-hybrid system showed the interaction of CaM with full-length IQGAP2. This interaction was disrupted when a truncated IQGAP2 lacking the IQ motif sequences was tested. This CaM interaction with IQGAP2 occurs even in 1% Triton and high salt concentrations. IQGAP2 binds CaM very strongly in the presence of Ca²⁺ and to a lesser extent in the absence of Ca²⁺. These results demonstrate an interaction of IQ motifs with both Ca²⁺-CaM and ApoCaM and suggest that these interactions may provide a link between CaM-mediated processes and the small G protein signaling pathways.

Some properties of the CaM binding enzymes PbK, adenylyl cyclase, and iNOS have been discussed previously. Glycogen phosphorylase b kinase, an enzyme that catalyzes the phosphorylation of glycogen phosphorylase and serves a key regulatory role in glycogen metabolism, contains CaM as an integral subunit (the -subunit). The quaternary structure is ()₄. The -subunit CaM remains associated even after prolonged exposure to Ca²⁺ chelators such as EGTA and EDTA (28). Skeletal muscle PbK can also bind four additional extrinsic CaM, referred to as ', in a Ca²⁺-dependent fashion. Troponin C, a member of the CaM homology family, also binds and can substitute for ', albeit with lower affinity. The greater abundance of troponin C in skeletal muscle makes it likely this interaction is physiologically relevant (26). The linked function between Ca²⁺ and PbK binding to CaM (') has been particularly well characterized (18).

Bordetella pertussis adenylyl cyclase binds, and is activated by, CaM in a Ca²⁺-independent manner. This enzyme is the first reported example of Ca²⁺-independent stimulation of any enzyme (52). When the enzyme was rendered Ca²⁺ free by desalting the enzyme in buffers containing EGTA and including the chelator in the assay mixture, cyclase activity increased 23-fold upon addition of CaM. Half-maximal activation occurred at 24 nM of CaM in EGTA and 0.1 nM in the presence of Ca²⁺. Calmodulin activated the enzyme in EGTA to about twice the activity found in Ca²⁺. Bordetella pertussis adenylyl cyclase also binds CaM-Sepharose in the presence of 5 mM EGTA (52).

Inducible (macrophage) NOS has also been reported to be a Ca²⁺-independent CaM binding protein. The enzyme is isolated with a tightly bound CaM even in EGTA, and extrinsic CaM is not required for maximal activity. Inducible NOS is maximally active at Ca²⁺ concentrations as low as 0.1 nM in vitro (120) and is thus probably maximally active in vivo even at the low intracellular Ca²⁺ concentrations encountered in unstimulated cells. However, because Ca²⁺ activates the enzyme relative to the activity in EGTA in some studies (140), the Ca²⁺-bound state of the intrinsic CaM is uncertain. Amino acids 503-532 of the mouse iNOS consist almost exclusively of the characteristic hydrophobic/basic regions of Ca²⁺-CaM binding sites. Comparison of this sequence with the corresponding region of constitutively expressed NOS (cNOS), which is a Ca²⁺- and CaM-dependent protein, shows only 43% identity of the 21-amino acid residues that are designated as the presumptive CaM-binding site of cNOS, suggesting that these differences allow iNOS to bind CaM at low concentrations (150). Two synthetic peptides corresponding to the CaM-binding domain of iNOS [P29, NOS-(504---532), and P34, NOS-(499---532)] were subsequently used to study Ca²⁺-dependent and Ca²⁺-independent binding of CaM. Both peptides bound CaM both in the presence as well as in the absence of Ca²⁺ (i.e., in the presence of chelator EGTA). The binding affinity of CaM in the presence of Ca²⁺ was reported with a dissociation constant of ~1 nM; this binding affinity is lower, but still remarkably high (45-85 nM) in the presence of EGTA (2). A chimeric NOS, constructed by replacement of the CaM-binding sequence of endothelial NOS (residues 493-512) with aligned sequence from iNOS (residues 501-532), results in a functional protein that is activated by CaM in either Ca²⁺ or EGTA, but these changes do not produce irreversible CaM binding, suggesting that other regions of iNOS (outside of residues 501-532) are required (140). A further study to test directly whether the predicted CaM-binding region of iNOS (amino acids 501-532) accounts for its Ca²⁺ independence used a chimera protein composed of reciprocally exchanged amino acids 503-532 of iNOS and the corresponding residues 725-754 of nNOS. The results of this study showed that both chimeras required 7-10 nM free Ca²⁺ to bind CaM and that Ca²⁺ is essential for enzymatic activity for both chimeras. Additional results by using the indicator fura 2 showed that the concentration of free Ca²⁺ required for half-maximal activity (EC₅₀) are ~0 for iNOS, 200-300 nM for nNOS, and 7-10 nM for the chimeras. Thus sequences outside the 503-532 region of iNOS are essential for its apparently irreversible binding of CaM (120).

Molecular procedures for screening for CaM-binding proteins in plants using ³⁵S-labeled recombinant CaM as a probe resulted in the identification of a petunia protein of 58 kDa with a high amino acid sequence similarity to the 53-kDa glutamate decarboxylase (GAD) from E. coli (11). Recombinant GAD from petunia catalyzes the conversion of glutamic acid to -aminobutyric acid and also binds CaM, whereas the E. coli GAD does not have CaM-binding activity. The CaM-binding domain on petunia GAD resides at the carboxy end of this protein, a region that is not present in the E. coli GAD. The authors concluded that intracellular levels of Ca²⁺ via CaM may regulate -aminobutyric acid synthesis in plants (11). A related study reported the identification of a 62-kDa CaM-binding protein from fava bean seedlings with GAD activity. These authors reported Ca²⁺-dependent CaM activation of the fava bean GAD activity (79). Activation of the petunia enzyme by CaM is also Ca²⁺ dependent; however, CaM binds to petunia GAD even in the absence of Ca²⁺, and CaM-GAD cross-linking is also Ca²⁺ independent. A synthetic peptide corresponding to the petunia GAD carboxy-terminal sequence (amino acids 470-495) binds CaM as a stable complex in the presence of Ca²⁺, but not in its absence. Using carboxy terminus-deleted mutants of GAD (by removal of 2 Lys residues, Lys-494 and Lys-495) results in a reduction of cross-linked complexes in the absence of Ca²⁺, suggesting that these lysine residues might facilitate electrostatic interactions between GAD and acidic CaM residues in a role similar to that of Lys in other CaM-target sites (6). Thus CaM binding to petunia GAD occurs in the presence or absence of Ca²⁺ as is the case for other enzymes described in this review (e.g., PbK, iNOS, Ca²⁺-release channels), but Ca²⁺, presumably binding to CaM, is required for enzyme activation (as with PbK and perhaps iNOS). Apocalmodulin binding occurs in plants as well as in animals and fungi.

Guanosine 3',5'-cyclic monophosphate-dependent protein kinase has also been reported to be activated by CaM in a Ca²⁺-independent fashion (152). Enzymatic assays of cGMP-dependent protein kinase showed that CaM increased maximum binding of this enzyme in the absence of Ca²⁺ without altering the Michaelis constant for ATP. It was found that in the presence of CaM concentrations producing maximum stimulation, this enzyme was specifically activated by a low concentration of cGMP. Adenos