The Molecular Biology of Invertebrate Voltage-Gated Ca²⁺ Channels

Voltage-gated Ca²⁺ channels perform a unique role in excitable cells. By gating Ca²⁺ influx in response to depolarization, they couple changes in membrane potential to numerous intracellular mechanisms regulated by Ca²⁺, including neurotransmitter release, muscle contraction, gene expression and cellular differentiation. Many of these responses are tuned to a specific temporal and spatial pattern of Ca²⁺ entry (e.g. Gu and Spitzer, 1995; Dolmetsch et al., 1997), which must be balanced with the potential toxicity caused by high intracellular levels of Ca²⁺. The need for carefully regulated patterns of Ca²⁺ influx in vertebrate cells is reflected by the presence of several molecular subtypes of Ca²⁺ channels, each distinct in its kinetics, pharmacology and tissue distribution. Physiological evidence suggests that invertebrate tissues also possess a variety of Ca²⁺ currents, but the molecular diversity of the channels gating these currents has not been as thoroughly studied.

A number of recent advances have now directed greater attention towards invertebrate Ca²⁺ channels. Foremost among these was the virtual completion of the Caenorhabditis elegans genome sequence in late 1998 (The C. elegans Sequencing Consortium, 1998). For the first time, we can examine the complete array of genes required for an organism possessing a nervous system, an approach that is not yet possible for a vertebrate model. Furthermore, novel C. elegans genes have led directly to the identification of new families of channel subunits (e.g. Wei et al., 1996), including the long-anticipated discovery of mammalian genes encoding low-voltage-activated T-type Ca²⁺ channels (Perez-Reyes et al., 1998). Progress has extended to other invertebrate models as well. Ca²⁺ channel subunits have been cloned from organisms in several phyla, providing a broader view of the evolution of channel genes. Functional expression of cloned subunits (Jeziorski et al., 1998; Okamura, 1999) now allows direct comparison between the molecular and physiological properties of invertebrate channels. Finally, classical genetic approaches in Drosophila melanogaster (Zheng et al., 1995; Smith et al., 1996, 1998) and C. elegans (Schafer and Kenyon, 1995; Lee et al., 1997) have correlated numerous channel mutations with their physiological effects.

The value of these findings extends well beyond the study of invertebrate physiology. In addition to elucidating how channels evolved, genes from early metazoans provide a perspective on channel structure and function that is unavailable from mammalian studies. This review summarizes the latest work on the molecular biology of invertebrate Ca²⁺ channels, with particular emphasis on their structural heterogeneity and their relationship to mammalian channel subtypes. The reader is also referred to thorough reviews of the physiology of invertebrate Ca²⁺ currents (Kits and Mansvelder, 1996; Skeer et al., 1996), the molecular biology of vertebrate Ca²⁺ channels (Stea et al., 1995; Catterall, 1995; Perez-Reyes and Schneider, 1995; De Waard et al., 1996; Jones, 1998) and the role of auxiliary subunits in channel function (Walker and De Waard, 1998).

Although Ca²⁺ currents were first identified in invertebrate tissues (Fatt and Ginsborg, 1958), the pharmacological and molecular identification of channel subtypes has been based primarily on studies in vertebrate preparations. Because the resulting criteria are commonly used to define invertebrate channels as well, a brief overview of vertebrate channel physiology is necessary to understand the present classification of invertebrate currents.

The several types of vertebrate Ca²⁺ currents that have been identified fall into two categories, high-voltage-activated (HVA) and low-voltage-activated (LVA). HVA currents can be further subdivided into L-type currents, which are sensitive to blockage or modulation by dihydropyridines (DHPs), and the various DHP-insensitive currents collectively known as non-L-type. Each current is gated by a channel that includes a large, pore-forming α1 subunit; HVA channels also contain several auxiliary subunits that have not, as yet, been identified in LVA channels. Because the α1 subunit largely determines the physiology and pharmacology of the resulting current, each of the vertebrate α1 subunits that have been cloned and functionally expressed can be correlated with a previously defined Ca²⁺ current subtype. Three L-type α1 subunits (α1S, α1C, α1D) gate DHP-sensitive currents (Tanabe et al., 1987; Mikami et al., 1989; Hui et al., 1991; Williams et al., 1992b). A fourth L-type subunit recently cloned from human retina, α1F (Bech-Hansen et al., 1998; Strom et al., 1998), has not yet been expressed, but is grouped with the L-type channels on the basis of molecular similarity. Non-L-type α1 subunits can be further distinguished by their responses to various peptide toxins (for a review, see Olivera et al., 1994). The α1B subunit (Williams et al., 1992a; Dubel et al., 1992) gates the N-type current that is blocked by the snail toxin ω-conotoxin GVIA (ω-CTx GVIA), while α1A (Mori et al., 1991; Starr et al., 1991) is responsible for the P/Q-type current that is blocked by ω-conotoxin MVIIC (ω-CTx MVIIC) and the spider toxin ω-agatoxin IVA (ω-Aga IVA). A third non-L-type subunit, α1E (Niidome et al., 1992; Soong et al., 1993), gates the toxin-resistant R-type current (Zhang et al., 1993; Randall and Tsien, 1997). LVA, or T-type, α1 subunits have only recently been cloned, but three subtypes are already known: α1G (Perez-Reyes et al., 1998), α1H (Cribbs et al., 1998) and α1I (Lee et al., 1999b). L-type, non-L-type and T-type α1 subunit genes form three phylogenetic clusters, such that proteins within a given class are significantly more similar than are those from separate classes (see Figs 1, 2).

The various α1 subunits differ in both their tissue and subcellular distribution. The α1S subunit is found in skeletal muscle, and the α1C subunit in heart and other tissues, while the α1D and the non-L-type α1 subunits are confined primarily to brain. Within neurons, α1A and α1B are localized to synaptic terminals (Westenbroek et al., 1992, 1995), where they gate the Ca²⁺ influx necessary for transmitter release, whereas α1E, along with neuronal L-type channels, is found in the soma and proximal dendrites (Hell et al., 1993; Yokoyama et al., 1995).

Ca²⁺ channels are part of a superfamily of voltage-gated cation channels that also includes Na⁺ and K⁺ channels. The pore of a cation channel is surrounded by four homologous domains, which are contained within the single α1 subunit in Ca²⁺ channels and α subunit in Na⁺ channels, but are formed by four separate subunits in K⁺ channels. Each domain comprises six transmembrane segments, denoted S1–S6; the S4 segments, which contain positively charged residues at every third or fourth position, are believed to function as voltage sensors. The external loop between S5 and S6 of each domain projects into the lumen of the channel to form part of the pore. Within each pore region in HVA α1 subunits, a glutamate residue (E in single letter code) is found at a conserved position. Together, the four glutamates (EEEE) form a selectivity gate (Heinemann et al., 1992; Yang et al., 1993); the pore of LVA α1 subunits differs only slightly, with aspartates in place of glutamates at two sites (EEDD). Non-conservative substitutions for glutamate, such as non-polar alanine (A) or basic lysine (K) residues, decrease the selectivity of the channel for divalent over monovalent cations (Ellinor et al., 1995), as reflected by the DEKA motif found at the corresponding sites in most cloned Na⁺ channels. Thus, the presence of acidic residues in the four key positions is regarded as essential to a Ca²⁺-selective channel. The intracellular domains of HVA α1 subunits also contain highly conserved regions that are implicated in protein interaction. The Ca²⁺ channel β subunit binds to a specific region between domains I and II (Pragnell et al., 1994) that is well conserved among all vertebrate HVA α1 subunits. A portion of the II–III domain of the skeletal muscle α1S subunit is implicated in excitation–contraction coupling (Tanabe et al., 1990; Nakai et al., 1998; Grabner et al., 1999), while a segment of the I–II loop of non-L-type α1 subunits binds Gβγ subunits (De Waard et al., 1997; Zamponi et al., 1997), and calmodulin mediates Ca²⁺-dependent inactivation in L-type channels via a site in the proximal carboxyl tail (Zühlke et al., 1999). The α1 subunit also contains the binding sites for DHPs (Tang et al., 1993; Grabner et al., 1996) and other L-type channel blockers (Hockerman et al., 1995; Kraus et al., 1998) as well as peptide toxins (Ellinor et al., 1994). Conservation among mammalian α1 subunits is greatest within the transmembrane segments, the pore-forming loops and the intracellular interaction domains.

While the α1 subunit essentially determines the type of current gated, the auxiliary subunits have important modulatory roles in HVA channels. The β subunit aids in trafficking of the α1 subunit to the plasma membrane (Chien et al., 1995; Tareilus et al., 1997) and also influences the kinetics of the current, leading to activation at more hyperpolarized potentials as well as modulation of the rate- and voltage-dependence of inactivation (Wei et al., 1991; Lacerda et al., 1991; Varadi et al., 1991). The four β subtypes (β1–β4) that have been cloned from mammals thus far (Ruth et al., 1989; Perez-Reyes et al., 1992; Castellano et al., 1993a,b) differ in their effects upon α1 subunits (Olcese et al., 1994). Strong conservation within the interaction domains of α1 and β subunits allows several β subtypes to associate with native α1A (Liu et al., 1996), α1B (Scott et al., 1996) and neuronal L-type α1 subunits (Pichler et al., 1997), leading to potential heterogeneity in the resulting currents. The α2δ subunit (Ellis et al., 1988) consists of the heavily glycosylated extracellular α2 domain linked by disulfide bonds to the membrane-spanning δ segment, both derived from the same gene. The α2δ subunit also modulates the α1 subunit (Singer et al., 1991), although less dramatically than does the β subunit. A second and third mammalian subtype of the α2δ subunit have recently been reported (Klugbauer et al., 1999). The γsubunit (Jay et al., 1990) was once thought to be present only in skeletal muscle Ca²⁺ channels. Recently, a γ homologue, named stargazin on the basis of its associated mutation in mice, has been discovered in brain (Letts et al., 1998) and is able to modulate the inactivation of heterologously expressed neuronal α1 subunits.

A comprehensive survey of the physiology of invertebrate Ca²⁺ currents is beyond the scope of this review. Instead, we will attempt to summarize certain findings that consistently emerge from the literature (for reviews, see also Kits and Mansvelder, 1996; Skeer et al., 1996). There is considerable electrophysiological and pharmacological evidence for diversity of invertebrate Ca²⁺ currents (e.g. Pelzer et al., 1989). Both LVA and HVA currents have been recorded from neurons (Haydon and Man-Son-Hing, 1988; Yamoah and Crow, 1994) and other invertebrate tissues (Hagiwara et al., 1975; Arnoult and Villaz, 1994). Evidence for more than one neuronal LVA current in cockroach (Grolleau and Lapied, 1996) and mollusc (Kits and Mansvelder, 1996) has been reported. Similarly, invertebrate HVA currents may be separated into components distinguished by kinetics or by responses to channel blockers. Mammalian L-type currents are blocked by three major classes of L-type channel ligands, DHPs, phenylalkylamines (PAAs) and benzothiazepines, that bind to distinct but overlapping sites on mammalian L-type channels (for a review, see Hockerman et al., 1997). The potency of these ligand classes in blocking invertebrate Ca²⁺ currents appears to correlate less closely than in vertebrate tissues. PAAs are more potent than DHPs in blocking neuronal current components in Drosophila (Pelzer et al., 1989; Morales et al., 1999), cockroach (Mills and Pitman, 1997; Wicher and Penzlin, 1997) and honeybee (Schäfer et al., 1994), whereas DHPs such as nifedipine are more effective in Drosophila muscle (Gielow et al., 1995) and Lymnaea neurons (Dreijer and Kits, 1995) and also block currents in crayfish muscle (Araque et al., 1994) and Aplysia tissues (Nerbonne and Gurney, 1987; Brezina et al., 1994; Laurienti and Blankenship, 1997; Fieber, 1998). The DHP agonist BayK 8644 stimulates a current in crustacean muscle (Erxleben and Rathmeyer, 1997), but not the nifedipine-sensitive current in Aplysia neurons (Nerbonne and Gurney, 1987; however, see Trudeau et al., 1993). Of the HVA currents that are DHP-insensitive, certain components can be blocked by peptide toxins. A current in crayfish neurons is blocked by ω-Aga IVA, but not by ω-CTx GVIA or L-type ligands, leading to its identification as P-type (Araque et al., 1994; Wright et al., 1996; Hong and Lnenicka, 1997; García-Colunga et al., 1999); ω-Aga IVA also specifically blocks Ca²⁺ current components in squid neurons (Llinas et al., 1989; McFarlane and Gilly, 1996), locust neurons (Bickmeyer et al., 1994) and sea urchin eggs (Vogel et al., 1999). In neurons from cockroach (Wicher and Penzlin, 1997) and Aplysia (Fossier et al., 1994), ω-Aga IVA and ω-CTx GVIA appear to block distinct HVA current components, suggesting multiple non-L-type channels. In certain instances, a neuronal HVA current is insensitive to both L-type and non-L-type ligands (Richmond et al., 1996; Hong and Lnenicka, 1997).

These studies, as well as some reviews (Olivera et al., 1994; Skeer et al., 1996), have frequently concluded that invertebrate Ca²⁺ currents do not correspond precisely to the profiles of defined vertebrate current subtypes. While invertebrate currents are often identified as L-type, non-L-type or T-type, many authors emphasize the significant biophysical or pharmacological differences from vertebrate currents. Furthermore, an invertebrate Ca²⁺ current that does resemble a particular vertebrate class may not necessarily be gated by a channel of the corresponding molecular subtype. A possible example is the current in arthropod neurons described as P-type on the basis of its sensitivity to ω-Aga IVA, but not to ω-CTx GVIA. The greater potency in invertebrate tissues of ω-Aga IVA and other spider toxins (Llinas et al., 1989; Leung et al., 1989) is not surprising, because ω-CTx GVIA and other toxins generated by fish-hunting snails of the genus Conus are designed for optimal effect in vertebrate tissues, whereas spider toxins are targeted towards insects. Thus, the invertebrate currents blocked by ω-Aga IVA, a toxin that binds to the α1A subunit in mammalian tissues, may be gated by α1 subunits of less specific molecular identity that are nonetheless sensitive to certain toxins.

For these reasons, invertebrate Ca²⁺ channels are best examined by a combination of physiological and molecular approaches. Studies in vertebrate tissues have benefited from such a synthesis, demonstrating that a wide range of currents can result from small variations in channel structure. Given the evidence for comparable current heterogeneity in invertebrate tissues, the composition of invertebrate Ca²⁺ channels is of paramount interest.

In contrast to other classes of channels, such as voltage-gated K⁺ channels (Papazian et al., 1987), that were first cloned from invertebrates, no invertebrate Ca²⁺ channel genes were identified until several mammalian channel subunits had already been sequenced and expressed, and a classification framework had been established. Recent progress in the cloning of invertebrate Ca²⁺ channels has been rapid, however (Table 1). To date, full-length L-type homologues have been isolated from the ascidian Halocynthia roretzi (Okamura, 1999), the arthropods Drosophila melanogaster (Zheng et al., 1995) and Musca domestica (Grabner et al., 1994a), the nematode C. elegans (Lee et al., 1997) and the cnidarians Cyanea capillata (Jeziorski et al., 1998) and Stylophora pistillata (Zoccola et al., 1999). Non-L-type α1 subunits have been cloned from Drosophila (Smith et al., 1996) and the mollusc Loligo bleekeri (Kimura et al., 1997), as well as a nearly complete sequence from C. elegans (Schafer and Kenyon, 1995). In addition, the coexistence of L-type and non-L-type channels has been established in the arthropod Blattella germanica (Inagaki et al., 1998), the mollusc Aplysia californica (White and Kaczmarek, 1997) and the platyhelminth Bdelloura candida (Greenberg et al., 1995) by the amplification of multiple cDNA fragments. (No invertebrate T-type α1 subunits, with the exception of the C. elegans gene discussed in the following section, have yet been reported.) As is true for other classes of voltage-gated and ligand-gated ion channels, Ca²⁺ channel α1 subunits are well conserved between vertebrates and invertebrates. Similarity is highest in the transmembrane domains and in the pore regions, where the EEEE motif is invariably found. The extended intracellular domains tend to vary more, except in regions that have been implicated in channel regulation, such as the I–II loop and the proximal carboxyl tail.

Examination of invertebrate Ca²⁺ channel genes, either by phylogenetic analysis (Fig. 1) or by direct measurement of amino acid identity (Fig. 2), reveals numerous clues to the pattern of channel evolution. Each α1 subunit that has been cloned from an invertebrate can be unambiguously identified as either L-type or non-L-type, but is not aligned with a specific mammalian subtype within the classes (Fig. 1), suggesting that the various subtypes arose during the evolution of vertebrates. For instance, the Drosophila subunits initially identified as α1A and α1D are now believed to have appeared well before the divergence of individual subtypes (Smith et al., 1996; Peixoto et al., 1997). Subunits from different classes are no more similar to each other in early invertebrates than in mammals, indicating that the separation of channel families occurred very early in, or prior to, the emergence of metazoans. In addition, no convergence is evident between invertebrate Ca²⁺ channels and Na⁺ channels, even in the jellyfish Cyanea, a member of the earliest extant phylum in which a nervous system is expressed.

No Ca²⁺ channel gene has yet been unambiguously identified in any organism predating metazoans. However, Cch1, a cloned yeast protein that has been identified as a homologue of Ca²⁺ channel α1 subunits (Paidhungat and Garrett, 1997), may have evolved from an ancestral Ca²⁺ channel. The predicted sequence of Cch1 displays certain hallmarks of the pore-forming subunits of both Ca²⁺ and Na⁺ channels, including a four-domain structure, similar membrane topology and positively charged residues within the S4 domains. Significantly, the glutamate residues conserved in the pore of HVA Ca²⁺ channels are found at appropriate positions within three of the four pore loops of Cch1, with an asparagine present in domain I (NEEE). Replacement of the corresponding glutamate in domain I of the rabbit α1C subunit with glutamine, which is chemically similar to asparagine, diminishes the affinity of the pore for Ca²⁺ only twofold (Ellinor et al., 1995). Yeast strains carrying a mutant form of the Cch1 gene exhibit decreased levels of intracellular Ca²⁺ (Paidhungat and Garrett, 1997), but the protein has not yet been directly shown to respond to voltage or to gate Ca²⁺ influx selectively. A second yeast gene also encodes a four-domain protein of unknown function that has been labeled a Na⁺ channel homologue. However, the two yeast proteins share the NEEE motif in the pore region and are more similar to each other than to any other member of the respective ion channel families; neither subunit exhibits preferential similarity to either Ca²⁺ or Na⁺ channels (Fig. 2).

Information regarding auxiliary subunits of invertebrate Ca²⁺ channels is not as extensive as that for α1 subunits. β subunits have been cloned from Musca (Grabner et al., 1994b) and Cyanea (Jeziorski et al., 1999). Neither protein is specifically similar to any of the four mammalian β isoforms; thus, both diverged from an ancestral β subunit prior to the appearance of the mammalian subtypes (Fig. 3). The Musca β subunit, when expressed in oocytes, increases the amplitude of the endogenous non-L-type Ca²⁺ current and also increases DHP binding to COS7 cells expressing the mammalian α1C subunit (Grabner et al., 1994b). The Cyanea β subunit modulates the cloned Cyanea L-type α1 subunit in a manner similar to that reported for mammalian α1–β interactions. Coexpression of the β subunit in oocytes increases the current produced by the α1 subunit, shifts its voltage-dependence to more negative potentials and accelerates its rate of inactivation (Jeziorski et al., 1999).

An advantage of invertebrate models such as Drosophila and C. elegans is their genetic tractability, which has revealed that mutations in Ca²⁺ channel genes are responsible for a variety of physiological defects. Mutant alleles have been identified for both known α1 subunit genes in Drosophila. The non-L-type gene maps to loci implicated in visual impairment (nightblind-A) and altered courtship song (cacophony) as well as a lethal variant (Smith et al., 1996). Two mutations of the L-type subunit have been characterized at the genetic level (Eberl et al., 1998). One, a lethal form, is truncated after the IVS4 segment. A second, partial loss-of-function phenotype results from a single nonconservative substitution within the IS1 segment, which slows activation of the channel (Ren et al., 1998). Changing the corresponding residue in a mammalian channel similarly alters its kinetics, establishing the importance of the IS1 segment in channel activation. Mutations have also been identified in two C. elegans α1 subunits. The unc-2 mutation in C. elegans lies in a gene encoding a non-L-type α1 subunit (Schafer and Kenyon, 1995). The product of the unc-2 allele lacks the IVS2 transmembrane segment, probably leading to a severely altered topology and a nonfunctional channel. Animals homozygous for unc-2 do not desensitize to the inhibitory effects of dopamine upon motor activity and the stimulation of egg-laying induced by serotonin, which implies a role for the intact channel in modulating neurotransmitter release. The C. elegans egl-19 mutation is found in a gene encoding an L-type channel α1 subunit (Lee et al., 1997). At least 26 mutations, representing myotonic (gain-of-function), flaccid (reduction-of-function) and lethal (loss-of-function) mutants, are localized to the egl-19 gene. Of the gain-of-function alleles that have been identified, two contain point mutations in or near the IS6 segment and two encode conservative changes in IIIS4. Each alteration is thought to affect the rate of channel inactivation, leading to extended depolarization of muscle cells.

The tissue distribution of invertebrate channel subtypes is generally similar to that in mammals, with L-type subunits expressed in both muscle and neural tissue, but non-L-type α1 subunits localized primarily to neurons. An example is seen in Aplysia, in which both types are present in neurons, but only the L-type is found in muscle. Within ganglia, the L-type channel is uniformly distributed throughout neurons and glia, while the non-L-type channel is restricted to somata, processes and growth cones, exhibiting a punctate labeling that suggests localization in vesicle membranes (White and Kaczmarek, 1997). In C. elegans, levels of expression of the L-type (Lee et al., 1997) and non-L-type (Schafer and Kenyon, 1995) α1 subunits are higher in muscle and neurons, respectively, although each transcript is found in both cell types. Transcripts encoding both the L-type (Zheng et al., 1995) and non-L-type α1 subunits (Smith et al., 1996) in Drosophila embryo are found primarily in brain, but functional studies have demonstrated that the L-type channel gates the majority of the Ca²⁺ current in larval muscle (Ren et al., 1998); the L-type transcript is more widely expressed in the adult (Zheng et al., 1995). The Musca L-type channel transcript is detected in larval muscle, but not neurons (Grabner et al., 1994a), while the ascidian L-type α1 subunit is found in both neural and muscle tissue (Okamura, 1999). In contrast, the non-L-type channel from squid is expressed throughout the nervous system, but not in heart, muscle or stomach (Kimura et al., 1997).

Functional expression of cloned invertebrate α1 subunits has allowed a direct assessment of the relationship between the molecular identity and pharmacology of an invertebrate channel. The sequence of the α1 subunit isolated from Cyanea (Jeziorski et al., 1998) is significantly more similar to that of mammalian L-type α1 subunits than to that of non-L-type or T-type subunits. Nevertheless, expression of the Cyanea channel in Xenopus laevis oocytes produces a rapidly inactivating current that is insensitive to L-type channel ligands as well as to peptide toxins. DHPs and PAAs that fully block vertebrate L-type currents at low micromolar concentrations have only weak effects upon the Cyanea channel (<50 % inhibition by each agent at 100 μmol l⁻¹), and BayK 8644 fails to potentiate the current. The L-type α1 subunit from the tunicate Halocynthia is similarly insensitive to block or enhancement by DHPs when expressed in oocytes (Okamura, 1999). These findings are in accord with the indirect measurement of the DHP sensitivity of the Musca L-type α1 subunit. Three domains of mammalian L-type α1 subunits (IIIS5, IIIS6, IVS6) have been implicated in forming the DHP binding site; transfer of these domains from the Musca channel to the DHP-insensitive mammalian α1A subunit resulted in a current that could be blocked by the DHP isradipine (65 % inhibition at 10 μmol l⁻¹), but was unstimulated by BayK 8644 (Sinnegger et al., 1997).

The above results demonstrate that the close association between the pharmacological and molecular identities of mammalian α1 subunits is not maintained in invertebrate channels. Such a distinction can be problematic. L-type currents, for example, are classically defined by their sensitivity to DHPs, so the description of a DHP-insensitive L-type channel may be viewed as contradictory. However, the similarity of the Cyanea and Halocynthia α1 subunits to mammalian L-type channels is approximately 10 % greater than that to non-L-type channels, whereas the DHP-binding pocket in L-type α1 subunits is formed by only nine residues, less than 1 % of the molecule (Mitterdorfer et al., 1996; Peterson et al., 1996; Sinnegger et al., 1997). Furthermore, the sites to which DHPs or toxins bind have no known endogenous ligands, implying that the formation of such sites is incidental to channel function. We therefore propose that the molecular definition of channel subtype is the more rigorous one, and that invertebrate currents that have previously been defined by pharmacological criteria may be reclassified once the peptide sequence of the corresponding channel is known.

The sequencing of the C. elegans genome provides an unprecedented view into the molecular neurobiology of an organism. The genes within an animal possessing a nervous system and other differentiated tissues can now be examined in full; for this reason, the Ca²⁺ channel subunit homologues identified in C. elegans are considered here separately. Furthermore, the C. elegans genome project has led to the discovery of novel genes that could not be detected using conventional homology-based cloning techniques (for a review, see Bargmann, 1998). Two algorithms, TBLASTN (Altschul et al., 1990), which searches the genome using reverse translation of a peptide sequence, and Genefinder (P. Green and L. Hillier, unpublished), which detects potential exons on the basis of consensus boundaries in C. elegans, have greatly aided the identification of putative channel genes.

Five C. elegans genes encode proteins that exhibit structural similarity to Ca²⁺ channel α1 subunits. Two of these, the egl-19 and unc-2 genes noted above, have been independently cloned by groups investigating mutant C. elegans strains (Schafer and Kenyon, 1995; Lee et al., 1997) and produce channel subunits corresponding to the L-type and non-L-type classes, respectively. The third α1 subunit gene in the C. elegans genome was first found in the c54d2 cosmid and is labeled c54d2.5. Although the function of c54d2.5 was unknown at the time of its discovery, the subsequent cloning of three mammalian homologues, each of which encodes a T-type channel (Perez-Reyes et al., 1998; Cribbs et al., 1998; Lee et al., 1999b), implies that the C. elegans subunit also forms an LVA channel. The reported open reading frame of c54d2.5, based on detection of potential exons, probably lacks at least one exon comprising domains IIIS3–IIIS4 (a candidate exon for this region is found between bases 29 066 and 28 825 on the antisense strand of c54d2). Elements that appear to be specific to the primary structure of LVA channels, such as an EEDD ring within the pore, are also found in the c54d2.5 peptide sequence. Significantly, egl-19, unc-2 and c54d2.5 are the sole C. elegans genes that fall within the respective L-type, non-L-type and T-type classes.

Two additional C. elegans genes encode novel proteins that exhibit distant similarity to both Ca²⁺ and Na⁺ channel subunits. One gene, comprising an open reading frame with a proposed length of 1410 amino acid residues, is found in the c27f2 cosmid (c27f2.3; accession number AAA81424), and the other, encoding a protein of 1581 residues, lies within c11d2 (c11d2.6; accession number AAC02575). The two proteins are equally similar to the pore-forming subunits of Ca²⁺ and Na⁺ channels, but display somewhat greater resemblance to one another. The sequences of the two nematode genes were used to clone a rat homologue of unknown function, named Rb21, that is expressed in brain and heart (Lee et al., 1999a); a short homologous fragment has also been found in the Schistosoma mansoni genome (accession number AQ400292). Both the C. elegans and Rb21 proteins have the four-domain structure common to voltage-gated Ca²⁺ and Na⁺ channels. The regions corresponding to the S4 segments possess some, but not all, of the positively charged amino acids characteristic of voltage-gated channels, and the key residues within the pore regions at positions that influence ionic selectivity (EEKE) differ from those in both Ca²⁺ (EEEE in HVA subunits or EEDD in LVA subunits) and Na⁺ (DEKA in vertebrate subunits) channels. We have cloned the cDNA corresponding to the c27f2.3 gene product from a C. elegans cDNA library (M. C. Jeziorski, unpublished results). The cDNA includes four additional exons beyond the 3′ end of the reported sequence, extending the open reading frame from 1410 to 1687 codons. However, the cloned 5′ end, which matches the sequence proposed for the cDNA in the C. elegans database, appears to lack a segment encoding the N terminus, such that the open reading frame is initiated within the S1 segment of domain I. Efforts to obtain an alternative clone for the 5′ end have not yet succeeded and, like Rb21 (Lee et al., 1999a), the c27f2.3 product fails to generate a voltage-gated cation current when expressed in oocytes (M. C. Jeziorski, unpublished results).

Two genes that encode apparent homologues of Ca²⁺ channel β subunits are also found within the C. elegans genome. One forms a protein that is strongly similar to that of known β subunits (55 % identical to the human β3 subunit) and is highly conserved in the primary interaction site. Because the protein closely resembles other invertebrate β subunits that have been cloned and expressed, it is likely to share with them the modulatory effects upon HVA α1 subunits that are typical of β subunits. The second C. elegans homologue, w10c8.1 (accession number AAB97592), is far less similar to β subunits (22 % identical to β3), but exhibits limited identity within the interaction domain and other regions that are conserved among cloned β subunits. The w10c8.1 cDNA, which we have isolated from a C. elegans library, contains an additional unreported exon of 132 bases (corresponding to bases 10 872–10 741 on the antisense strand of C. elegans cosmid w10c8) that extends the open reading frame to 357 residues, approximately 80 % of the average length of conventional β subunits. The function of w10c8.1 remains unresolved, however. Coinjection of Xenopus laevis oocytes with RNA encoding w10c8.1 fails to modulate the expressed Cyanea L-type α1 subunit, a channel that can be regulated by both vertebrate and invertebrate β subunits (Jeziorski et al., 1999). Furthermore, w10c8.1 cannot block the modulatory effects of exogenous β subunits, suggesting that the C. elegans protein does not compete for the β subunit interaction site on the α1 subunit (M. C. Jeziorski, unpublished data). The robust effects exerted upon the Cyanea α1 subunit by other β subunits occurs in spite of sequence variation within the interaction domain of the α1 subunit, however, and the w10c8.1 protein may bind more efficiently to an interaction domain that is more highly conserved or to a domain specific to a C. elegans HVA α1 subunit. Alternatively, the β homologue may modulate other channel proteins found in C. elegans, such as the T-type channel or the more distant α1 subunit homologues of unknown function.

The C. elegans genome also contains two genes encoding putative α2δ subunits. One (accession number P34374) corresponds to the unc-36 gene, mutation of which results in a phenotype similar to that for unc-2 (Schafer et al., 1996). The published sequence of 734 amino acids, which is considerably shorter than the average length of mammalian α2δ subunits (approximately 1100 residues) and lacks the hydrophobic δ segment that anchors the α2 segment to the membrane, is almost certainly incomplete. The second identified gene (accession number CAA90091) encodes what appears to be a full-length protein. A phylogenetic analysis of known α2δ subunits (truncated to match the length of the unc-36 gene product) indicates that the two C. elegans genes are approximately as distant from each other as they are from mammalian isoforms.

Although the genetic information from C. elegans is far more extensive than that for any other multicellular organism, it may not accurately reflect the full complement of Ca²⁺ channel subunits present in all invertebrates. No copy of a pore-forming α subunit gene for a voltage-gated Na⁺ channel is found in C. elegans, for instance. The absence of such a gene is probably due to its deletion during the evolution of nematodes, because Na⁺ channels have been cloned from earlier organisms such as cnidarians (Anderson et al., 1993; Spafford et al., 1998; White et al., 1998b) and platyhelminths (Jeziorski et al., 1997) as well as from higher invertebrates such as Drosophila (Salkoff et al., 1987) and squid (Sato and Matsumoto, 1992; Rosenthal and Gilly, 1993). A similar deletion may have occurred for one or more Ca²⁺ channel subunit genes, such as a gene encoding a γ subunit, which is not represented in the C. elegans genome.

In this regard, the report of two additional cDNA fragments amplified by the polymerase chain reaction (PCR) from the coral Stylophora pistillata (Zoccola et al., 1999) deserves mention. The fragments, each approximately 100 codons in size, appear to encode the domain IV pore and S6 regions of two different non-L-type α1 subunits. The two sequences have been labeled as N-type and P-type, suggesting that divergence within the non-L-type channel subfamily occurred very early in metazoan evolution. Although the limited size of the fragments precludes their definitive identification, our preliminary phylogenetic analysis places the two sequences on adjacent branches among invertebrate non-L-type α1 subunits, and not within the mammalian α1A or α1B clusters (data not shown). Nevertheless, the potential existence of multiple non-L-type α1 subunit genes in an early invertebrate demonstrates that current models of channel evolution may require modification and emphasizes the importance of the continued isolation and characterization of invertebrate channel subunit genes.

The C. elegans genome clearly lacks the variety of Ca²⁺ channel subunit genes that underlies much of the heterogeneity in vertebrate Ca²⁺ currents. The multiple mammalian isoforms of L-type, non-L-type and T-type α1 subunits are each represented by single genes in C. elegans, and perhaps in most invertebrates. Numerous potential combinations between the four vertebrate β isoforms and the HVA α1 subunits provide additional channel variation that cannot be mimicked by a single invertebrate β subunit. How might multiple Ca²⁺ currents be generated from fewer subunit genes?

One possible source of distinct currents is through post-translational modification. Most invertebrate subunit sequences contain consensus sites for phosphorylation by protein kinases A and C, among other kinases, and phosphorylation has been extensively implicated in the modulation of invertebrate Ca²⁺ current amplitude (for a review, see Kits and Mansvelder, 1996). Stimulation of protein kinase C also causes the non-L-type Ca²⁺ channel in Aplysia bag cell neurons to translocate from vesicular to plasma membranes (White et al., 1998a). The ability of phosphorylation to generate multiple current components that differ in their activation or inactivation kinetics or pharmacological response has not been extensively investigated, however.

In contrast, alternative splicing of subunit transcripts, which is extensively employed in mammalian tissues to generate variant Ca²⁺ channels (for a review, see Perez-Reyes and Schneider, 1995), has been carefully examined in Drosophila (Smith et al., 1996; Peixoto et al., 1997). The non-L-type α1 subunit gene, known as Dmca1A, varies at three key sites (Peixoto et al., 1997). Two involve the S4 segments critical to determining the voltage-dependence of the channel. The loop connecting IS3 and IS4, as well as much of IS4 itself, is encoded in two separate exons, while the IVS3–IVS4 region contains a six-base insertion in some transcripts. Variation at each site may alter channel activation. Another intriguing splice site occurs within the intracellular loop between domains I and II. A portion of this region that is highly conserved in vertebrate α1 subunits forms both the primary interaction site for the β subunit and the site to which βγ subunits of G-proteins bind in certain non-L-type channels. Whereas one exon in Dmca1A contains the fully conserved interaction sites, a second exon of identical length is lacking the motif to which Gβγ subunits bind and is conserved with the canonical β subunit binding domain at only four of nine residues. Three of the remaining residues are those most critical to the α1–β interaction (De Waard et al., 1996), and the Cyanea α1 subunit, which also contains a poorly conserved I–II loop with the key residues intact, can be modulated by β subunits (Jeziorski et al., 1999). Nevertheless, variation within this interaction domain may result in subtle influences upon channel kinetics or membrane targeting. Although it has been suggested that the two exons in the I–II loop arose early in the evolution of non-L-type channels (Peixoto et al., 1997), the introns flanking the corresponding exon in the Cyanea L-type channel gene do not appear to contain an alternative exon for the β binding region (M. C. Jeziorski, R. M. Greenberg and P. A. V. Anderson, unpublished observations). Invertebrate β subunits may also be alternatively spliced, as demonstrated in the Musca gene (Grabner et al., 1994b).

An additional way in which multiple channels may be derived from one invertebrate gene is via RNA editing. In certain genes, specific adenosine residues in the genomic sequence can be converted to guanosines in the unprocessed mRNA, resulting in altered translation at discrete sites (for a review, see Simpson and Emeson, 1996). RNA editing has been extensively implicated in the regulation of mammalian glutamate channels (Sommer et al., 1991; Lomeli et al., 1994) and, in squid, functionally distinct K⁺ channels are generated by editing of transcripts from a single gene (Patton et al., 1997). RNA editing may also underlie heterogeneity of invertebrate Ca²⁺ currents. Transcripts from the Drosophila Dmca1A gene can be edited at ten different sites, resulting in an amino acid substitution at each position (Smith et al., 1996; Peixoto et al., 1997). The proportion of edited compared with unedited codons at each site is not yet known, but many of the affected residues lie within functionally critical regions such as transmembrane segments or pore-forming loops. The potential for differential editing of specific residues in different cell types or in response to developmental or environmental cues could result in significant channel diversity from a limited number of genes.

The path by which Ca²⁺ channels evolved is implied by their relationship with other members of the voltage-gated cation channel superfamily. Each domain of the pore-forming subunits of Ca²⁺ and Na⁺ channels is similar in structure to a K⁺ channel subunit, suggesting that the form of the K⁺ channel containing six transmembrane segments evolved first, then underwent two stages of concatenation to create an early four-domain protein. One proposed reason for the transition from single-domain to four-domain proteins is that an asymmetrical pore structure may be required in Ca²⁺ and Na⁺ channels (Salkoff and Jegla, 1995). The four-domain protein may have given rise to Ca²⁺-selective channels before Na⁺ channels. This pattern of emerging channel selectivity is thought to correspond to the demands imposed upon organisms as they gained greater complexity (Hille, 1992; Ranganathan, 1994; Armstrong and Hille, 1998). Early cells may have employed K⁺-selective channels as a defense against osmotic stress (Milkman, 1994; Sukharev et al., 1994). The need to control intracellular Ca²⁺ concentration perhaps resulted from the ability of free Ca²⁺ to precipitate phosphate and to disrupt the cell’s primary source of metabolic energy; the role of Ca²⁺ as a second messenger, and the subsequent need for gated Ca²⁺ influx, may have developed from such regulatory mechanisms (Clapham, 1995). The requirement for rapid signal propagation in the earliest nervous systems would then have necessitated a role for Na⁺-permeable channels.

Current physiological and molecular evidence supports several elements of this evolutionary model. The structure of the single domain apparently developed before prokaryotes and eukaryotes diverged; the kch protein cloned from Escherichia coli possesses the six transmembrane segments and conserved pore region characteristic of eukaryotic K⁺ channels (Milkman, 1994). Subsequent within-gene duplication may have occurred prior to the separation of metazoans from other eukaryotes, since the four-domain structure is evident in the two yeast channels described above. Conceivably, either the single-domain or four-domain structures could have evolved independently in two separate lineages. However, homology among voltage-gated channels has typically been inferred from sequence identity, which is less likely to arise by convergent evolution than is structural similarity. The ionic selectivity of the early four-domain channels is unknown, but it has been proposed that the yeast Cch1 protein gates Ca²⁺ (Paidhungat and Garrett, 1997), and the structure of its pore is somewhat consistent with Ca²⁺ selectivity. Other physiological evidence also indirectly suggests that Ca²⁺ currents predate Na⁺ currents. The existence of Ca²⁺ currents in protists such as Paramecium (for a review, see Kung, 1989) and Stylonychia mytilus (Deitmer, 1986) has been firmly established, and Ca²⁺ currents have been described in plant cells as well (Thuleau et al., 1994). The earliest phylum in which voltage-gated Na⁺ currents have been extensively recorded is Cnidaria (Mackie and Meech, 1985; Anderson, 1987), although the report of a protist action potential that is carried in part by Na⁺ (Febvre-Chevalier et al., 1989) indicates that Na⁺-selective channels may have evolved prior to metazoans. Sequence similarity between Na⁺ and Ca²⁺ channels, which is greater than that between either class and the yeast or Rb21 proteins, argues for a relatively recent division of the two families. Phylogenetic analysis suggests that the ancestral four-domain channel gave rise first to the Rb21 family, then to a precursor of metazoan Ca²⁺ and Na⁺ channels. Na⁺ channels evolved either from this branch or from the LVA Ca²⁺ channel subfamily after it separated from HVA channels. Definitive support for this proposed series of evolutionary divergences awaits isolation of channel sequences from protists and other early eukaryotes. Furthermore, we commonly infer the type of current gated by a channel from the overall sequence identity of the protein, but minimal substitution within the pore is required to alter the selectivity of a channel. Thus, a channel subfamily associated with a particular current may have evolved initially to gate a different ion or mixture of ions before assuming its present role.

From the earliest Ca²⁺ channel, two additional phases of divergence occurred, one to generate the three (or more) major Ca²⁺ channel types, and a second to create the various subtypes found in mammals. Similarly, present models of metazoan evolution propose two major stages of genome expansion. The first, which occurred prior to the separation of parazoans (sponges) and true metazoans, permitted the development of functionally distinct isoforms of a given protein, thus facilitating differentiation of tissues within an organism. Sponges, which lack discernible tissues, nevertheless possess multiple genes for proteins such as protein kinase C, phospholipase C (Koyanagi et al., 1998), protein tyrosine kinases (Suga et al., 1999) and protein kinase phosphatases (Ono et al., 1999), isoforms that correspond to gene subfamilies found in higher organisms. The second genomic expansion occurred early in chordate evolution, when two successive genome duplications freed gene subtypes to diversify further, leading to the additional genetic variation that predated the appearance of vertebrates (Holland et al., 1994). The protein subtypes that resulted from the latter duplication event typically differ less from each other than do the subfamilies from which they emerged, and they are thought to vary more in tissue distribution than in function (Iwabe et al., 1996).

Evidence from invertebrate Ca²⁺ channel genes is consistent with the timing of the two duplication events. At least three classes of Ca²⁺ channel α1 subunit genes (L, non-L and T-type) are represented in C. elegans, and the α1 subunits cloned from cnidarians, the earliest true metazoans, fall within specific channel subfamilies, indicating that the appearance of multiple α1 subunit genes predates this phylum. A similar organization is suggested at the opposite end of the invertebrate spectrum, where the α1 subunit cloned from the invertebrate chordate Halocynthia can be identified as L-type, but not as a specific subtype. The invertebrate β subunits that have been cloned also appeared prior to the divergence of mammalian subtypes (Grabner et al., 1994b; Jeziorski et al., 1999). The earliest clear evidence for subtypes within the larger classes of subunit genes is found in nonmammalian vertebrates. Neurons of the marine ray Discopyge ommata express at least two non-L-type subtypes, α1B and α1E (Horne et al., 1993), and a β subunit cloned from Xenopus corresponds to the β3 mammalian subtype (Tareilus et al., 1997). Comparison of invertebrate and vertebrate genes supports the model of a fourfold increase in diversity prior to the evolution of vertebrates. The L-, non-L- and T-type α1 subunit genes in the C. elegans genome are represented, respectively, by four, three and three known mammalian subtypes. Similarly, the four vertebrate β subunit subtypes exhibit comparable identity to the lone conventional C. elegans β subunit gene; clustering of the vertebrate subunits into two pairs may reflect the two rounds of genome duplication that have been proposed.

The gene duplication model cannot be expected to be followed strictly by all gene families. Several potential inconsistencies are already apparent, such as the absence of C. elegans genes encoding a Na⁺ channel α subunit or a Ca²⁺ channel γ subunit, and the existence of two α2δ subunit homologues in C. elegans, each equally distant from known mammalian proteins. Nevertheless, the recent discovery of a fourth mammalian L-type α1 subunit suggests that additional non-L- or T-type channel genes may eventually be found and that the novel putative channel subunits found in C. elegans may also be revealed to have multiple vertebrate homologues.

It is clear from the advances described in this review that a survey of Ca²⁺ channel subtypes can no longer be complete without consideration of invertebrate channels. Information from the various nonmammalian genome projects may offer the most immediate and tangible benefits, but much can be learned from channels in other organisms as well. Invertebrate genes separated several hundred million years ago from the ancestral genes that gave rise to mammalian channels and, since that time, have evolved independently, within the constraints of maintaining a functional Ca²⁺ channel. The resulting variance between vertebrate and invertebrate peptide sequences more clearly defines the conserved regions essential to channel function while revealing structural variation that may underlie distinct physiological characteristics. When combined with the ability to analyze and manipulate genetic information rapidly in established model systems, the growing understanding of the molecular biology of invertebrate Ca²⁺ channels promises to lead to an integrated view of Ca²⁺ channel function and diversity.

We thank Dr Peter Okkema and Dr William Walthall for supplying C. elegans cDNA libraries, Dr Edward Perez-Reyes for contributing mammalian Ca²⁺ channel subunit cDNA clones and Dr Carlos Valverde and Dr Aurea Orozco for comments on the manuscript. This work was supported in part by NSF grant IBN-9808386 to P.A.V.A. and R.M.G.