Cardiac myocytes are characterized by distinct structural and functional entities involved in the generation and transmission of the action potential and the excitation-contraction coupling process. Key to their function is the specific organization of ion channels and transporters to and within distinct membrane domains, which supports the anisotropic propagation of the depolarization wave. This review addresses the current knowledge on the molecular actors regulating the distinct trafficking and targeting mechanisms of ion channels in the highly polarized cardiac myocyte. In addition to ubiquitous mechanisms shared by other excitable cells, cardiac myocytes show unique specialization, illustrated by the molecular organization of myocyte-myocyte contacts, e.g., the intercalated disc and the gap junction. Many factors contribute to the specialization of the cardiac sarcolemma and the functional expression of cardiac ion channels, including various anchoring proteins, motors, small GTPases, membrane lipids, and cholesterol. The discovery of genetic defects in some of these actors, leading to complex cardiac disorders, emphasizes the importance of trafficking and targeting of ion channels to cardiac function. A major challenge in the field is to understand how these and other actors work together in intact myocytes to fine-tune ion channel expression and control cardiac excitability.
Cell excitability is central to cardiac physiology, such as the capacity of the heart to beat spontaneously, to function as a pump, or to secrete natriuretic peptides. Many cardiac ion currents and their molecular correlates have been identified and their roles in shaping action potentials (AP) described. In addition, progress in genetic and molecular biology has provided precise information on the structure-function relationships of most channels, pumps, and exchangers expressed in cardiac myocytes. Nowadays, much effort in the field focuses on unraveling the mechanisms that regulate targeting to and formation of channels and their complexes in the plasma membrane of cardiac myocytes. As in other excitable tissues, activation of the electrical signal and its transmission as a depolarizing wave in the whole heart requires highly organized myocyte architecture and cell-cell contacts. In addition, complex trafficking and anchoring intracellular machineries regulate the proper surface expression of channels and their targeting to distinct membrane domains. An increasing list of proteins, lipids, and second messengers can contribute to the normal targeting of ion channels in cardiac myocytes. However, their precise roles in the electrophysiology of the heart are far from understood.
Targeting and trafficking of membrane proteins, including channels and receptors, have been extensively investigated in other excitable tissues, such as the nervous system and epithelial tissues. These studies provide a precious body of knowledge that can provide insight into the regulation of cardiac excitability. However, due to the distinct structural and functional characteristics of cardiac myocytes, there are specificities of the trafficking and targeting of ion channels in this cell type that are very probably not shared with other cell types. Cardiac myocytes are highly differentiated cells specialized in excitation-contraction (EC) coupling. They possess well-developed electrical and mechanical machineries. Three crucial structural and functional entities of these cells can be identified: 1) the EC coupling domains composed of the t-tubule system/terminal cisternae of the sarcoplasmic reticulum and sarcomere, 2) the intercalated disc (ID) at the cell-cell junction, and 3) the remaining lateral membrane domain.
The functional unit for contraction, called the sarcomere, spans the area between Z lines and is made of three types of filaments: thin (actin), thick (myosin), and elastic (titin or connectin). It constitutes the central element of myocyte architecture and is responsible for their striated appearance under the microscope, a consequence of alternating A (thick filaments) and I (thin filaments) bands. The EC coupling process is generated at the level of the Z band, where the t-tubule system allows the functional coupling of electrical activity with the release of calcium by the sarcoplasmic reticulum, triggering contraction. Cardiac myocytes are rod-shaped cells that are electrically coupled mainly along the longitudinal axis, via the ID where the AP is transmitted in an anisotropic mode between two adjacent myocytes. The lateral membrane contains focal adhesions and costamere structures. The costameres connect proteins associated with the Z line of the sarcomere to the sarcolemma and ensure the linkage of the sarcomere with the extracellular matrix. Costameres allow cardiomyocytes to sense and transmit mechanical forces acting on the myocardium during repeated systoles and diastoles. These structural and functional entities exhibit distinct and highly organized ion channel complexes that contain also pumps, exchangers, neurohormonal receptors, and second messengers. Examples include the various electrogenic systems involved in the EC coupling process, organized as a true “calcium synapse,” and the gap junctional channel complex at the ID. Therefore, important questions include the following: How are the specialized plasma membrane domains of cardiac myocytes organized? How are channels targeted to these platforms? How are they anchored and aggregated to form large protein complexes? How are they dynamically recruited and recycled, and in response to which stimuli and via posttranscriptional mechanisms are these processes regulated?
This review provides an update on the current state of our knowledge in relation to these questions. Since the mechanisms regulating folding, quality control, and ER export of ion channels have been the subject of review articles (97, 411), this article focuses largely on the targeting of cardiac ion channels to specific membrane domains, their organization in the plasma membrane, and their ultimate fates. Cardiac myocytes are structurally highly polarized cells, a necessary specialization for the anisotropic propagation of the depolarization wave in the whole myocardium. The main concept promoted in this review is that the cell biology processes regulating ion channel targeting, trafficking, tethering, and recycling are crucial to this polarization. Also, we propose that the emerging role played by abnormal channel surface expression during cardiac electrical disorders points to the importance of these processes in cardiac electrophysiology and pathophysiology.
II. SPECIALIZED MEMBRANE DOMAINS FOR CHANNEL EXPRESSION IN CARDIAC SARCOLEMMA
In cardiac myocytes, ion channels are localized in three major functional and structural entities: the ID, the lateral sarcolemma, and the EC coupling domains (see FIGURE 1 AND 2). TABLE 1 summarizes the known subcellular sarcolemmal localizations for a number of ion channels expressed in the myocardium. TABLE 1 relies on immunofluorescence studies and, thus, should to be viewed with some degree of skepticism as true negative controls (i.e., use of knockout animals for the ion channel of interest) have not been performed.
A. Intercalated Disc
The extensive interconnection of cardiomyocytes confers on the myocardium a syncytium-like function. The numerous junctions present in the ID ensure the rapid and coordinated propagation of AP along the entire length of the cardiac muscle. Three types of intercellular adhesion structures are present in the ID: fascia adherens junctions, desmosomes, and gap junctions (74) (see FIGURE 2).
Fascia adherens junctions and desmosomes are the two types of mechanical junctions responsible for intercellular adhesion in myocardium. Both are composed of intercellular adhesion molecules that cross the sarcolemma at the ID. They bind both intracellular proteins of the cytoskeleton and proteins from the neighboring cell in the extracellular space, thereby connecting two adjacent cardiomyocytes. Mutations in genes coding for these junctional proteins have been observed in human cardiomyopathies such as Naxos disease and Carvajal syndrome (13, 189, 269, 304, 338, 343, 368). In the working myocardium, both fascia adherens and desmosomes are directly involved in sensing and regulating mechanical stresses acting in the longitudinal axis. In mammals, a mixed type of junctional structure termed “hybrid adhering junction” or “area composita” exists in the cardiac intercalated disc as well, suggesting an evolutionary mechanism leading to strengthen cardiac mechanical coupling (47, 133).
The fascia adherens junctions are the main cell-cell mechanical junctions located at the ends of sarcomeres. The transmembrane adhesion protein in these structures is N-cadherin (165–166, 208, 442). N-cadherin is connected to sarcomeric actin by intracellular linker proteins such as α- and β-catenin (165, 329), ARCVF armadillo-repeat proteins (192), and vinculin (141) in addition to plakoglobin (or γ-catenin), the latter of which is found also in desmosomes (82, 202).
Desmosomes, the other class of mechanical junction, are small and discrete structures composed of the desmosomal cadherins desmogleins (24, 204, 379, 381) and desmocollins (306). Desmosomal cadherins are associated with desmin, the intermediate filament component of the cytoskeleton, via the linker proteins desmoplakin (81, 134, 289) and plakophilin (272) as well as to plakoglobin, the fascia adherens junction protein. Many of these proteins are mutated in arrhythmogenic right ventricular cardiomyopathy (371).
Gap junctions constitute the third intercellular adhesion structure of the ID. Each gap junction is constituted of two connexons (or hemichannels) formed from six connexins spanning the lipid bilayer on two adjacent cardiomyocytes (466). These junctions allow nonselective diffusion of ∼1,000-Da molecules (213, 242) including ions and small molecules like second messengers, thereby ensuring electrometabolic coupling. Interestingly, gap junctions exhibit remarkably high plasticity, and both the turnover and distribution of connexins are highly regulated processes (369–370). Mechanical forces can induce changes in gap junction density at the ID affecting the subsequent intercellular electrical conduction velocity. For instance, Saffitz et al. (370) showed that pulsatile mechanical stretch markedly upregulated Cx43 in cultured rat neonatal cardiomyocytes. Thus remodeling of gap junctions induced by either chemical mediators, hypertrophy/dilatation, mechanical forces, or any alteration in cell-cell contacts would directly modulate the electrical pacing of the heart.
Gap junction regions are rigid structures, vulnerable to mechanical stresses. Compensating for this, gap junctions in cardiomyocytes are always located in close proximity to other intercellular adhesion structures of the ID, notably the adherens junctions. The importance of this association of gap junctions with adherens/mechanical junctions has been shown in human cardiomyopathies such as Naxos disease and Carvajal syndrome, which result from mutations in plakoglobin and desmoplakin (368), and also in vitro (113, 142). Geisler et al. (142) followed the successive events leading to the reorganization of an ID in cultured adult rat ventricular myocytes. The formation of cell-cell contacts initially starts with the aggregation of transmembrane adhesion proteins followed by the recruitment of gap junction proteins beneath the sarcolemma. Immunofluorescence and electron microscopy experiments showed that in nascent adherens junctions, N-cadherin, α-catenin, and plakoglobin, are present while Cx43 is absent. It is only when adherens junctions mature that a Cx43 signal becomes detectible. These observations suggest that formation of adhesion junctions is a prerequisite for gap junction formation within the intercalated disc (142). Finally, several lines of evidence indicate that altered mechanical coupling has a large impact on electrical coupling, whereas impaired electrical coupling does not affect mechanical coupling (for review, see Ref. 303).
B. Lateral Membrane
Under physiological conditions, cardiomyocytes are not electrically coupled in the transverse axis, but the AP propagates laterally along this axis of the sarcolemma to the ID. Two major structures can be distinguished in the lateral membrane of cardiomyocytes, the costamere and the t-tubule system. The costamere links cardiomyocytes to the extracellular matrix (ECM) and guarantees the maintenance of the three-dimensional organization of the myocardium. The t-tubule system is linked to sarcoplasmic reticulum extensions and initiates EC coupling (see FIGURE 1).
A major consequence of muscle contraction is cell deformation and shortening. During this process, the contractile machinery of sarcomeres must remain connected to both the sarcolemma and to the extracellular matrix to properly coordinate contraction within the three-dimensional organization of cardiac muscle layers. Costameres are transverse, riblike structures of the sarcolemma that overlie the Z lines of the sarcomere (321, 322). They function as focal adhesions in non-muscle cell types (114). By constituting physical links between Z-lines, sarcolemma, and ECM, the costameres sense and transmit bidirectionally both intrinsically generated and externally applied mechanical forces (89, 254, 266, 372, 391). Two distinct macromolecular protein complexes play major regulatory roles at the costamere level: the integrin complex and the dystrophin-glycoprotein complex (DGC).
Integrins are heterodimers composed of α and β chains. Adult cardiomyocytes express α-3B, α-6A, α-6B, α-7B, α-7C, α-7D, and α-11 subunits and the β-1A, -1D, -3, and -5 cardiac isoforms (354, 390). Each chain possesses a long extracellular domain, which binds extracellular laminin, a membrane-spanning domain and a short cytoplasmic tail that links to the actin cytoskeleton via interactions with talin, vinculin, and the nebulin-related protein N-RAP, as well as sarcomeric actinin (41). In addition to their function as adhesion molecules, integrins also are mechanotransducers. Indeed, signal transduction through the integrin complex has been extensively studied in non-muscle cells at focal adhesions (277), as well as in cardiomyocytes (354). The cytoplasmic domain of integrin constitutes a major anchor for a number of signaling molecules including non-receptor type tyrosine kinases such as focal adhesion kinase (FAK), src-family tyrosin kinases, and integrin-linked protein kinase (ILK) as well as signal mediators such as small GTPases (rac, rho, Cdc42) (114, 278, 354, 372). The integrin complex is important for cardiac physiology. For instance, inactivation of the β1-integrin gene in ventricular cardiac myocytes in mice results in dilated cardiomyopathy (390), and heterozygous knockout mice for the vinculin gene (Vin+/−) show predisposition to stress-induced cardiomyopathy (470).
Also important to the maintenance of cardiomyocyte integrity is the DGC, similarly composed of transmembrane, cytoplasmic, and extracellular proteins (226). In the myocardium, the transmembrane proteins are β dystroglycan, α, β, δ, γ, ϵ, and ζ sarcoglycans, and sarcospan. β Dystroglycan is attached to the extracellular α dystroglycan that, in turn, binds to laminin in the extracellular space. On the cytoplasmic side, dystrophin binds to syntrophin, dystrobrevin, and nitric oxide synthase (NOS), and also attaches the entire macromolecular complex to the actin cytoskeleton. The physiological roles of the DGC have been identified using animal models of Duchenne and Becker muscular dystrophies (for reviews on animal models, see Refs. 15, 108). The DGC is essential in stabilizing the sarcolemma upon physical stresses (78). Indeed, the dystrophin-deficient mdx mouse exhibits membrane fragility (78, 226). In addition to its mechanical and structural functions, the DGC plays a role in cellular communication through interactions with signaling molecules such as NOS and Grb2 (50, 149, 461). The involvement of the DGC in cardiac ion channel regulation is discussed in sections IV and VII.
2. t-Tubule system
t-Tubules are characteristic narrow, tubular invaginations of the lateral sarcolemma observed in adult mammalian ventricular cardiomyocytes (315, 426). These digitations face tightly the adjacent sarcoplasmic reticulum (SR) and the short intermembrane space between the sarcolemma and the SR constitutes a dyadic cleft for the calcium communication that drives EC coupling. The T-tubule system also exists in atrial cardiomyocytes, but the transverse-axial tubular network is more rudimentary in these cells (315, 443). It has been proposed that this loose organization in the atrial cells results in an absence of EC coupling in the center of the atrial myocyte under basal conditions and, consequently, a less robust contraction than occurs in ventricular cells (46, 163, 250).
III. TRAFFICKING AND SORTING OF ION CHANNELS
The intracellular trafficking of ion channels involves complex processes. Synthesized, and assembled in the endoplasmic reticulum (ER), ion channels must thereafter be transported to the Golgi apparatus, further processed there, and then targeted to the membrane subdomains where they are functional. Each of these subprocesses involves a large number of highly coordinated, interacting facilitators and regulators. While little is known directly about how these mechanisms operate in cardiomyocytes themselves, a great deal has been learned about how cardiac ion channels are trafficked when expressed in heterologous cell systems and, occasionally, cardiomyoblast cell lines. As basic trafficking mechanisms are conserved across eukaryotes (387), there are reasons for optimism that many if not most of these data reflect mechanisms occurring in cardiomyocytes.
A. Export Mechanisms From ER to Golgi
1. Early events in the ER and quality control
The life of an ion channel begins in the rough endoplasmic reticulum. Although it has not been directly established for any cardiac-expressed channel, assembly of Kv1.3, and very probably most channels, occurs concurrently with synthesis (205–207, 245, 350, 431). The cell employs quality control mechanisms to ensure that only properly assembled channels are exported from the ER. For example, the conservation of a specific threonine residue in S1-S2 linker is necessary for ER export of Kv1.4 and Kv3.2 (268).
Channels that have failed to fold or assemble correctly likely inappropriately expose hydrophobic residues to the aqueous solvent. Because of their aggregation with other misfolded proteins, these channels may be retained in the ER (e.g., CFTR, Ref. 106). Such hydrophobic exposure to solvent has been linked also to the degradation of misfolded proteins by the proteasome (28, 273). Many trafficking-defective hERG (Kv11.1) channel mutants are dealt with in this way (22, 135, 147–148, 276). Interestingly, trafficking of some of these mutants out of the ER can be rescued by hERG-binding drugs (123, 148, 276, 323, 355, 477), temperature decrease (132, 265, 276, 331, 341), and/or chemical chaperones (22, 477), probably by promoting the proper folding and/or assembly of the channel. Strongly suggesting that the hERG-binding drugs act by stabilizing the core structure of the channel, their effectiveness in promoting ER exit of a trafficking-deficient pore mutation, G601S, is directly related to their potencies in blocking the channel (123). A mutation in the inner pore drug binding domain that reduced drug block also reduced rescue. Also supporting this model, rescue of another mutant, N470D, by incubation with the E-4031 hERG blocker is associated with a reduction in trypsin sensitivity of the mutant protein, such that its sensitivity more closely matches that of the wild-type channel (147). In the same study, association of the mutant channel with a chaperone, calnexin, was also found to be reduced by E-4031. Thus the folding of the mutant channel into a wild-type conformation is very probably promoted by the drug. That some pore-binding drugs rescue the trafficking of specific hERG mutants at concentrations well below those required for channel block (341) indicates, however, that additional mechanisms for rescue by hERG blockers may exist. Not surprisingly, chaperones like Hsp70/Hsc70, Hsp90, and calnexin have been shown to facilitate ER exit of hERG (122, 147) and Kv1.2 (253). Although, in the case of Hsp70/Hsc70, there may be a role in facilitating vesicular trafficking and membrane fusion as well (75, 326, 481), this effect is also likely to be generally a simple function of the proper folding/assembly of the channels. β-Subunits, KChIPs, KChAP, and other accessory proteins also bind to their target channels in this locale, promoting forward trafficking via chaperone-like activities (30, 216, 335, 396, 452).
2. Active export control
A number of additional mechanisms have evolved to regulate the export of ion channels from the ER. Among the first to be discovered was the RXR motif, first identified in the KATP channel (473) and subsequently shown to be operative in CFTR (61), as well as in hERG (215), in the major cardiac inward rectifier, Kir2.1 (248), and in other channels (255, 409). These motifs are hidden in properly folded and assembled channels but promote retention in the ER when they are exposed on the channel surface. An introduced pair of RXRs in a naturally occurring KCNQ1 mutant similarly causes retention of that channel in the ER (373). RXR motifs are also present in many voltage-gated (Kv) channels, although, with the exception of hERG (215), it has yet to be established that they indeed act as retention signals. Other motifs, such as dilysine repeats and other dibasic signals (159, 310, 472), also serve to retain channels in the ER, at least in part by promoting the retrograde transport of “escaped” channels back to the ER (139, 233). At least a subset of RXR and other dibasic motifs bind 14–3-3 proteins (271, 310, 399, 468), a family of molecules that function in part to promote surface expression of membrane proteins (399). While at first glance this seems inconsistent with a retention role for the RXR motifs, Yuan et al. (468) have shown that the affinity of the 14–3-3 proteins for the RXR motifs is dramatically higher for tetrameric than for monomeric constructs. Thus most likely, 14–3-3 proteins promote the export of properly assembled channels. That coat-associated protein I (COPI), an essential component for the recycling of membrane proteins from the Golgi to the ER (25, 307), was shown to compete with 14–3-3 for RXR strongly suggests that misassembled channels, with their lower 14–3-3 affinity, are returned to the ER by COPI (275, 468). Binding of COPI to a dibasic motif in the twin-pore KCNK3 potassium channel is indirectly inhibited by 14–3-3β. 14–3-3 binds to an adjacent “release site,” thereby driving dissociation of the COP protein from the channel (310). Another ER retention signal, KDEL (473), binds to the KDEL receptor in the transport vesicles, targeting proteins harboring the signal for Golgi to ER recycling (56, 475).
In addition to ER retention signals, ion channels also harbor motifs that promote export from the ER. These forward trafficking signals are quite diverse. Among the simplest motifs shown to be important in the active export of ion channels from the ER is the diacidic Aspartate-Alanine-Aspartate (“DAD”) motif identified in CFTR (444). This motif is required for physical interaction with Sec23/Sec24 and, combined with a demonstrated requirement for Sar1 function, is, thus, thought to promote the COPII-dependent exit of the channel from the ER. Mutation of this motif to DAA dramatically slows but does not eliminate the export of CFTR to post-ER compartments (361, 444). Efficient exit of TASK-3 from the ER also requires an intact diacidic motif (482), as does KATP export (420). More complex acidic residue-containing motifs have been shown to be necessary for the normal ER exit of other channels. An intact acidic cluster motif (DDXXDXXXI) is required for ER export of the large-conductance calcium-activated potassium (BKCa) channel (66). In Kir2.1, FYCENE serves such a function (248, 414). Nonacidic motifs have been implicated in the ER exit of other channels. A VXXSL signal promotes efficient exit of Kv1.4, and a similar but less effective VXXSN sequence is operative in Kv1.5 (236, 478). The cyclic nucleotide-binding domains of HERG, ERG3, and HCN2 may also act as forward-trafficking signals (9), and the Kv1.4 pore appears to harbor a pore-based forward trafficking signal (447).
B. Molecular Motors and Rails
1. Microtubule cytoskeleton
Properly folded and assembled cardiac ion channels must be transported to the sarcolemma to perform their functions. This trafficking of cardiac ion channels through the ER, the Golgi apparatus, and the cytoplasm is dependent on the microtubule cytoskeleton and associated motors. Little is known about the specific transport of ion channels to the ER exit sites, but generally the trafficking of newly synthesized proteins through the different cellular compartments and on to the plasmalemma has been shown to be microtubule-dependent (FIGURE 3) (reviewed in Ref. 26).
Microtubules are long heterodimers of α- and β-tubulin, (∼250 Å in diameter), aligned mainly along the longitudinal axis of adult cardiomyocytes (209). While the microtubule cytoskeleton is difficult to discern in these cells (80), it is essential to both the compartmentalization of intracellular organelles and to their normal structure (353, 448). Manipulation of microtubule integrity with microtubule depolymerizing agents such as colchicine and nocodazole has been shown to affect the surface expression of both potassium and sodium channels in cardiomyocytes. In atrial myocytes, nocodazole treatment increases the surface expression of Kv1.5 channels and the amplitude of the corresponding current IKur (72). Similar results were reported for Kv4.2 and hERG channels, respectively mediating Ito and IKr (241). The inward rectifier IK1 current, the functional expression of Kir2.1, is substantially reduced (241), whereas KCNQ1 surface expression is not influenced by microtubule depolymerizing or stabilizing agents (301). KCNQ1 interaction with the microtubule cytoskeleton is nevertheless important to its regulation, since its modulation by protein kinase A (PKA) is abolished by microtubule depolymerization. Colchicine-treated neonatal rat cardiomyocytes also showed an increased INa, although this appears to be in large measure due to an indirect effect on Nav1.5 channel kinetics (288). In the same cells, stabilization of the microtubule cytoskeleton reduces INa density via a mechanism that reduces channel expression at the sarcolemma (59). While changes in channel properties have been reported in some studies after manipulation of microtubule integrity, a growing body of evidence rather shows interference with channel trafficking. And, while the microtubule cytoskeleton is structurally important, it is also essential for normal vesicular traffic from the ER through the Golgi and on to the sarcolemma. This traffic is driven by the dynein and kinesin motors that track along microtubules. Kinesins comprise a large family of related anterograde motors; in contrast, only three major dynein isoforms exist in eukaryotes, two of which have roles in intracellular trafficking.
Dynein I, the major retrograde motor tracking along microtubules in the cytoplasm, is comprised of two identical 520-kDa heavy chains that contain the movement-generating ATPase, along with a number of intermediate and light chains, the functions of which are still being defined (reviewed in Ref. 190). The dynein motor on its own is incapable of moving cargo and association with the dynactin complex is required to link the motor to the cargo. Dynactin integrity is quite sensitive to the cytoplasmic stoichiometry of its subunits. Overexpression of the p50 subunit, for example, causes the dissociation of the dynactin complex, decoupling the dynein motor from its cargo (109). This method has become a preferred means to block dynein function (53, 109, 437). In one example, endogenous ClC-2-dependent currents were dramatically increased by p50 overexpression or pharmacological block of dynein function by ENHA in COS-7 cells (98). Immunofluorescence experiments and biotinylation assays showed that this increase in ClC-2-dependent currents was due to reduced internalization of the surface-expressed channel. Dynein plays a similar role in Kv1.5 expression. The increase in Kv1.5 expression associated with microtubule disruption is recapitulated by p50-mediated interference with dynein function in heterologous cells (72). Kinetic analysis as well as immunocytochemistry and surface proteolysis confirmed a dramatic increase in Kv1.5 channel residence at the cell surface. Like ClC-2, Kv1.5 interacts directly with the dynein/dynactin complex. Under dynein blockade, channels fail to internalize leading to supernormal levels of channels at the cell surface. Moreover, direct blockade of internalization through treatment with a dynamin inhibitory peptide mimicked the effect of p50 overexpression on the channel. The role(s) of dynein in the expression of other channels has been less thoroughly investigated. Loewen et al. (241) demonstrated that Kv4.2 surface expression in heterologous cells is increased by p50 overexpression, correlating well with its increased expression after microtubule polymerization in rat ventricular myocytes. Similarly, hERG-, Kv2.1-, and Kv3.1-dependent currents and surface expression of the channels themselves are increased by p50 overexpression (241).
The motors required for the delivery of new channel through the Golgi apparatus and on to the cell surface are the kinesins. In mammals, kinesins are encoded by 45 genes, with many splice variants. Most kinesin motors travel in the anterograde direction along microtubules (reviewed in Ref. 171). The kinesin-14 family moves, like dynein, in the retrograde direction (19). Very much smaller than dynein and more structurally diverse, these proteins nevertheless “walk” along the microtubules in a similar ATP-driven manner (27). Unlike dynein, it is thought that the diversity of cargoes carried is more dependent on the individual kinesin motor than on accessory subunits per se.
Kv4.2 was the first cardiac-expressed ion channel shown to be trafficked by a specific kinesin isoform. In rat cortical neurons, transfection with a Kif17 (kinesin-2) dominant negative (DN), but not Kif5a, Kif5b, KifC2, or Kif21b dominant negatives, almost completely blocked transport of the channel from the cell body to the dendrites (73). Thirty amino acids at the COOH terminal of Kv4.2 were found to be responsible for this interaction with Kif17. Since this kinesin was recently shown to be expressed in murine heart (57), its role in Kv4.2 (or Kv4.3) trafficking in cardiomyocytes remains an open question. In the immortalized atrial cell line HL-1, expression of a Kif17 dominant negative construct decreased Kv1.5 channel movement and the surface expression of the channel, whereas a kinesin-1 (KHC) DN had no effect (57). This latter result contrasts strongly with those obtained by Zadeh et al. (469) who demonstrated the involvement of the ubiquitously expressed kinesin-1 isoform Kif5b in Kv1.5 trafficking in H9c2 cardiomyoblasts and HEK293 cells. Interestingly, when expression of the Kif5b DN and Kv1.5 were initiated at the same time, the Kif5b DN actually increased surface expression of the channel. This increase was related to an indirect block of channel internalization by the Kif5b DN, very probably via inhibition of the dynein motor activity (469). Indeed, it is well established that inhibition of anterograde trafficking inhibits also retrograde trafficking, and vice versa (157, 437). Moreover, other pools of preexisting channels are available for plasma membrane insertion, even in the absence of microtubule-based trafficking, probably in the subsarcolemmal actin cytoskeleton (see below). In any case, further experiments are clearly required to determine which kinesin isoforms are truly operative in ion channel trafficking in native cardiomyocytes.
4. Actin cytoskeleton and myosin motors
Constructed from “nonmuscle” β- and γ-actins, the cortical actin cytoskeleton in cardiomyocytes is associated with costameres, t-tubules, and SR (reviewed in Ref. 194). It has been implicated in the maintenance of t-tubule integrity (228). Disruption of the actin cytoskeleton by cytochalasin D affects the functional expression of sodium, potassium, and L-type calcium channels in adult cardiomyocytes, although results are mixed for the latter (reviewed in Ref. 58). A direct linkage to the actin cytoskeleton via the actin-binding protein Ahnak has been demonstrated for the L-type calcium channel (155–156, 173), and indirect linkages via α-actinin2 have been demonstrated for L-type, Kv1.5 (84, 260), and Nav1.5 channels (479).
Manipulations of the actin cytoskeleton affect ion channel function at least partly by affecting channel trafficking. For instance, cytochalasin D treatment, which destabilizes actin filaments, quadrupled the Kv1.5-associated current in HEK293 cells. This increase was due to an enhanced delivery of Kv1.5 channels to the plasma membrane; channel kinetics were unchanged (260). Because α-actinin2-antisense RNA produced similar results, it seems likely that this disruption favors the release of intracellular pools of channels from anchors that otherwise prevent channel insertion into the plasma membrane. In contrast to Kv1.5, Ito has been shown to be attenuated by cytochalasin D and augmented by phalloidin, an actin filament stabilizer (463). The effects of actin cytoskeleton manipulations on the sodium current appear more likely to be purely kinetic than to be mediated by changes in Nav1.5 trafficking (435). Thus the actin cytoskeleton plays contrasting roles in the functional expression of ion channels, depending on the channel type.
In addition to anchoring the channels, either at the cell surface and/or internally, the actin cytoskeleton is essential for the near-plasmalemma trafficking of those channels. Intimately involved in this trafficking are the actin-tracking motors myosin V and myosin VI. They are expressed in a wide variety of tissues (reviewed in Ref. 96), including the heart (366, 465), although little work on the function of these myosins has been done in cardiomyoctyes. Myosin V is a class of three motors, myosin Va, -b and -c, respectively, involved in vesicular transport of secretory vesicles, endosomes, and membranous organelles, as well as in a number of other cellular processes (reviewed in Refs. 96, 364). Myosin V has been shown to work with kinesin to mutually increase the processivity of both motors (14). Myosin VI similarly carries vesicular traffic through the cortical actin cytoskeleton but in the opposite direction (reviewed in Ref. 418). It is unique among myosins in moving towards the minus end of the actin filament (451). While there has been no direct demonstration of the involvement of either of these motor types in the trafficking of cardiac ion channels, there are good reasons to believe that both are essential to it. In melanocytes, myosin V has been demonstrated to capture vesicles at the cortical actin cytoskeleton after their delivery from the cell interior by the microtubule-dependent system and to transport them to the periphery (457). Myosin VI has been shown to be important in the forward trafficking of AMPA receptors in hippocampal neurons (296) in a process involving the MAGUK protein SAP97, and has been implicated in the exocytosis of CFTR in endothelial and other cell types (20, 79). It is involved also in clathrin-mediated endocytosis (54–55, 287, 340, 408) and is important in the endocytosis of CFTR (419). Myosin VI has also been shown to be involved in recycling of internalized plasma membrane proteins back to that locale (70) and has recently been implicated in secretory vesicle fusion with the plasma membrane (44).
C. Early Transport of Cargo Proteins/Vesicle Trafficking
Both myosins V and VI are known to interact with Rab GTPases, key regulators of vesicle trafficking (reviewed in Ref. 96). Rab GTPases comprise a family of over 60 small GTPases intimately involved in the regulation of vesicle trafficking in all eukaryotic cells (151, 474). Kinesins and, at least indirectly, dyneins also interact with these GTPases (376) (reviewed in Refs. 151, 412).
Involved in the trafficking of membrane proteins from the ER to the Golgi compartments are the Rab GTPases Rab1 and Rab2 and the ARF (ADP-ribosylation factor) protein Sar1. Sar1 regulates the formation of vesicles carrying most cargo from ER to Golgi by recruiting the molecular coat COPII (reviewed in Ref. 101). Rab1 ensures fusion between the vesicle budded from the ER to Golgi compartments (16, 17), and Rab2 controls retrograde transport of post-ER, pre-Golgi intermediates and COPI-coated to the ER (333, 424–425).
Through the Golgi apparatus and beyond, a plethora of Rab GTPases are involved in trafficking to the plasma membrane (reviewed in Refs. 96, 474). Only a few have been investigated for their roles in ion channel trafficking, and even less is known about their roles in the trafficking of ion channels expressed in the heart. While we are only beginning to unravel the roles of Rab GTPases in the trafficking of cardiac-expressed ion channels to the cell surface, it is becoming clear that ion channels do not necessarily traffic by entirely conventional means. CFTR was the first ion channel demonstrated to associate with Rab GTPases for its normal trafficking (143, 449). Surprisingly, this channel is largely absent from the Golgi stack in CHO cells and epithelial cells of salivary gland or duodenum (33). Nevertheless, the transport of CFTR out of the ER exit sites seems to be dependent on COPII-coated vesicles and Sar1, but it is independent of Rab1/Rab2 and syntaxin 5, involved in the conventional trafficking pathway from ER to Golgi. Moreover, processing is partially dependent on Rab6, a GTPase involved in the retrograde transport of various Golgi enzymes, and on the late endosome associated SNARE syntaxin 13 (467). Consistent with such an unconventional pathway in BHK cells, Okiyoneda et al. (312) showed that newly synthesized CFTR quickly colocalized with EEA1, a marker for early endosomes. Whether an unconventional, Rab1-independent trafficking process is operative for CFTR trafficking in cardiomyocytes remains unknown.
Newly synthesized hERG channels, heterologously expressed in HEK293 cells, also seem to traffic via an unconventional pathway to the cell surface (94). Dominant negative isoforms of Rab11, a GTPase associated with the recycling endosome (see sect. VB), dramatically reduced hERG surface expression and corresponding current. Interestingly, Delisle et al. (94) report that currents carried by the inward rectifier Kir2.1 channel were not similarly affected by a Rab11 DN. Indicating that Kir2.1 likely traffics via a more conventional pathway, functional expression of this channel was greatly reduced by the Arf1 DN and only modestly affected by the Rab11 DN. Trafficking of newly synthesized Kv4.2, too, appears to be unconventional, at least in neurons and when coexpressed with KChIP1. In this case, however, the pathway appears to be quite different from that traversed by CFTR and hERG. ER to Golgi trafficking of newly synthesized Kv4.2 channels expressed in HeLa cells has been shown to involve KChIP1 and to be independent of Sar1 but nevertheless dependent on Rab1 function (128, 160). In these cells, ER to Golgi traffic of the Kv4.2 channel is dependent on COP1-coated vesicles but not COPII. This KChIP1-dependent trafficking depends on at least two SNARE proteins, Vtila and VAMP7, that are not involved in the conventional ER to Golgi pathway and with which KChIP1 colocalizes (128). Silencing of either of these SNAREs blocked traffic of Kv4.2 channels to the plasma membrane. KChIP1, however, is a neuronal isoform (336) and is poorly expressed in the heart. In experiments testing whether these siRNAs could block trafficking of Kv4.2 when coexpressed with the major KChIP isoform expressed in the heart, KChIP2, Schultz et al. (383) found that the channels trafficked normally to the cell surface. Thus whether Kv4.2 traffics at least in part by this or a similar unconventional pathway in the heart is very much an open question.
Data on other cardiac-expressed channel types are sparse. Although it has not been shown whether delivery of the major cardiac sodium channel, Nav1.5, to the sarcolemma is Rab-dependent, the delivery of the epithelial sodium channel ENaC has been shown to be enhanced by overexpression of Rab11a (191). Interestingly, Brefeldin A, an inhibitor of ER-Golgi transport, completely abrogated that effect, suggesting that the channels delivered were likely newly synthesized. This control of the vesicular trafficking of newly synthesized ion channels by small GTPases is of great interest. If the evidence from other cell types holds true, even in part, in the heart, a great deal will be learned, not only about ion channel trafficking but also about trafficking mechanisms in general, from the study of these processes in cardiomyocytes.
IV. INSERTION AND ANCHORING OF ION CHANNELS INTO THE PLASMA MEMBRANE
The final steps of forward trafficking are delivery/insertion of ion channels from a cargo vesicle into the sarcolemma and their subsequent anchoring/stabilization. Complementing the Rab proteins that act as membrane organizers by ensuring the correct sorting and targeting of newly synthesized or recycled channels (474), SNARE proteins are specialized in membrane fusion processes. SNAREs are involved in all membrane fusion steps, allowing the passage of cargo proteins from one compartment to the other throughout the whole secretory pathway (45). In this section, we review the recent progress gained in our understanding of regulated and constitutive exocytosis in the myocardium.
A. Fusion Process and Membrane Delivery
1. Fusion machinery
The SNARE complex [SNARE: soluble N-ethylmaleimide-sensitive fusion (NSF) protein attachment protein (SNAP) receptor] (405) consists of a large superfamily of proteins that vary in structure and size (182, 222). They possess a stretch of 60–70 amino acids containing 8 heptad repeats, i.e., the SNARE motif (51), that exhibits a high propensity to form coiled coils (349). Each fusion step requires the interaction of four SNARE motifs, usually provided by three different SNAREs. One SNARE is carried by the vesicle (v-SNARE) while the target membrane, e.g., the plasma membrane, usually contains three SNAREs (t-SNAREs). V-SNAREs are represented by the vesicle-associated membrane proteins (VAMP-1, -2, or -3) or synaptobrevins (Sb-1, -2, or -3) and t-SNAREs are represented by syntaxin (Sx) and synaptosome-associated proteins of 25 kDa (SNAP-25 or, its homolog SNAP-23). The formation of a SNARE complex between the vesicle and the acceptor membrane pulls the facing membranes tightly together and overcomes the fusion barrier by destabilizing the lipid bilayers (77, 225). Many proteins regulate the activity of SNAREs. For example, Rab3 orchestrates the initial contact between membranes destined to fuse and ensures that only appropriate organelles are tethered (222, 474). Cellular factors such as Sec/munc (SM proteins) are responsible for preparing the contact between SNAREs (222, 349). Other proteins such as synaptotagmins and complexins act as calcium sensors. They bind to SNAREs after the initial Ca2+ entry that triggers regulated exocytosis in neurons (51, 63, 348, 349, 416).
2. SNARE complexes in the myocardium
Relatively few studies have explored the expression patterns of SNARE proteins in the myocardium. Moreover, most of them have been aimed at determining the identity and function of specific SNAREs involved in the regulated exocytosis of atrial natriuretic peptide (ANP) in the endocrine heart, i.e., the atrium (62, 121, 327), or in the regulated exocytosis of the glucose transporter-4 (GLUT-4) in response to insulin (130, 131, 400).
Multiple SNARE protein isoforms are expressed in the adult atrial myocardium (for a complete review of transcripts, see Refs. 121, 327). Transcripts for six VAMP isoforms have been found (VAMP-1B, -2B, -3, -4, -5, and -8), corresponding to different events such as trafficking through the trans-Golgi network, maturation of secretory granules, fusion of late endosomes, and constitutive secretion (6, 136, 446, 471). VAMP-1B and VAMP-2B transcripts are involved in the regulated exocytosis of synaptic vesicles and secretory granules (29, 182, 417, 450). Two transcripts for SNAP proteins have been detected (SNAP-23 and SNAP-25A and -25B), as well 10 for syntaxins (Sx-1A, -2c, -3, -4, -5, -6, -8, -12, -13, and -17) and 4 for the Ca2+ sensor synaptotagmin (synaptotagmin-1, -4, -6, and -11). In the adult myocardium, two SNARE complexes have been shown to be expressed at the protein level (327). The first complex, constituted of VAMP-1, VAMP-2, syntaxin-4, and SNAP-23, is known to participate in GLUT-4 trafficking (130–131) and ANP secretion in cardiac myocytes (121). The second complex is composed of VAMP-1, VAMP-2, syntaxin-1, and SNAP-25 and was previously characterized as specific for synaptic transmission in neurons (238) (see FIGURE 4).
The observation that proteins destined for the plasma membrane or for a specific intracellular organelle use the different compartments of the secretory pathway, i.e., the ER and the trans-Golgi network (TGN), led to the hypothesis that a vesicular transport pathway for constitutive exocytosis exists (318). Indeed, the release of components of the extracellular matrix or incorporation of newly synthesized membrane proteins into the plasma membrane probably occur via fusion of a transport vesicle with the plasma membrane after passing through the same series of membrane-enclosed compartments as do those subject to regulated exocytosis of neurotransmitters, for example (174, 339, 356, 357, 380). In the myocardium, specific actors involved in the exocytosis of ion channels are not known. A few papers exist concerning the cycling of the transmembrane protein GLUT-4. GLUT-4 is synthesized in the TGN and stored in GLUT-4 storage vesicles (GSV) before being sent to the plasma membrane upon insulin stimulation (129). GLUT-4 is also stored after internalization in the recycling endosome and recruited back to the plasma membrane under basal conditions or in response to exercise (162, 187–188, 258, 476). Its recruitment to the plasma membrane in muscle and fat cells involves VAMP-2, syntaxin-4, and SNAP-23 (130). Using cyclohexamide to block the secretory pathway, the authors demonstrated that GLUT-4 enters the secretory granules via the recycling pathway in cardiomyocytes. Therefore, the recycling endosome appears to be an important storage compartment for this transporter. In the same way, we showed that Kv1.5 channels can be released from the recycling endosome (32), but the specific SNAREs involved in their vesicular release were not investigated. Further studies are needed to unravel SNARE complex involvement in the specific fusion process between the recycling endosome and the sarcolemma. Indeed, because SNARE complexes are required for all membrane fusion steps from one compartment to the other, identification of SNARE cognates for each specific compartment promises to be an exciting avenue for future research in cardiomyocyte cell biology.
3. Direct interaction between SNAREs and ion channels
Although there is no relevant information concerning cardiac myocytes, direct interactions between ion channels and proteins of the SNARE complexes have been reported in other cell types. For example, Kv1.1 channels interact with syntaxin-1A in brain synaptosomes and with SNAP-25 in insulinoma HIT-T15 β cells (124, 430). When expressed in Xenopus oocytes, both syntaxin-1A and SNAP-25 led to increased fast inactivation of the Kv1.1/Kvβ1.1-mediated current. Moreover, high concentrations of syntaxin-1A decreased the surface expression of the channel, consistent with a role for syntaxin in the regulation of vesicle trafficking (124). Syntaxin-1A also interacts with the Kv1.2 channel in smooth muscle cells and slows activation of that channel (298). In Kv2.1-expressing HEK293 cells, coexpression of syntaxin-1A inhibits channel trafficking to the surface and reduces the corresponding current (234). Kv2.1 also interacts with SNAP-25 in rat islet β-cells (234, 249, 274) and with VAMP-2 in secreting neuroendocrine cells (430). Fast inactivating channels such as Kv4.x, responsible in the myocardium for the initial repolarization phase of the cardiac AP, also interact directly with syntaxin-1. While synatxin-1A reduces Kv4.2 gating and surface expression in Xenopus oocytes (460), syntaxin-1A increases Kv4.3 trafficking to the plasma membrane and shifts the steady-state inactivation towards more negative potentials in HEK293 cells (7).
B. Anchoring and Scaffold Proteins
One fascinating but yet largely unanswered question concerns the nature of mechanisms regulating the targeting of ion channels to distinct subdomains of the sarcolemma and their tethering in large molecular complexes. This process involves a large set of proteins that regulate the delivery, anchoring, and scaffolding of newly addressed ion channels (see FIGURE 5).
1. MAGUK proteins
A family of multidomain membrane proteins called the membrane-associated guanylate kinase homologs (MAGUK) has emerged as central organizers of specialized plasma membrane domains in various cell types including cardiac myocytes. These proteins, which localize beneath the plasma membrane, regulate the surface expression of several membrane proteins. In 1995, Kim et al. (198) provided the first evidence that neuronal Shaker channels are clustered by MAGUK proteins. Since then, MAGUK proteins have been shown to be major partners of most ion channels.
It is the study of the neuronal synapses that has revealed the molecular diversity of MAGUK proteins. At the ultrastructural level, synapses are asymmetric, characterized by a dense thickening at the cytoplasmic face of the postsynaptic membrane (320), named postsynaptic density (PSD). In 1992, Cho et al. (71) isolated from an enriched PSD fraction the first MAGUK protein. This protein, which migrates as a doublet with an apparent molecular mass of 95 kDa, was named PSD-95. Subsequently, SAP90 (201) and SAP97 (292) were isolated from the presynaptic termini of a subset of inhibitory synapses in the rat cerebellum. Two other proteins, structurally related to SAP90 and SAP97, have been found in the adherens junction fraction of epithelial cells or in the intercalated disc of cardiac myocytes. They were named ZO1 and ZO2 for zona occludens (178–179). Interestingly, all of these MAGUK proteins are highly similar to the product of the Drosophila lethal(1)discslarge-1 tumor suppressor gene. The Dlg-A protein localizes to the septate junctions between epithelial cells of the imaginal discs during Drosophila larval development. It is an important determinant of septate junctions and also of the signal transduction pathway that regulates growth in the developing larva. Such a homology among mammalian and insect MAGUK proteins suggests long evolutionary relationships among neuronal proteins involved in tight junction, cadherin-based junctions and synaptic junctions.
Structurally, MAGUK proteins are characterized by three domains: a Src homology 3 (SH3) domain, a domain with homology to the enzyme guanylate kinase (GUK), and a PDZ domain. PDZ is the acronym for PSD-95, Dlg and ZO-1. PDZ domains are protein-protein interacting domains. Class I PDZ domains recognize COOH-terminal -Ser/thr-X-ψ-Val sequence (where X is any amino acid and ψ is any hydrophobic amino acid, Ref. 104), class II domains bind ψ-X-ψ-COOH and class III domains binds Asp/Glu-X-ψ-COOH (305, 406). PSD95, Dlg1, and SAP97 contain only class I PDZ domains (42). SH3 domains are 55- to 70-amino acid sequences that mediate protein-protein interaction with proteins containing proline-rich PXXP sequences. The GUK domains are homologous to the yeast guanylate kinase but are not functional. Several MAGUK proteins are expressed in the heart including SAP97, ZO-1, CASK, and MAGI3. However, precise information on the roles and on the localizations of MAGUK proteins is available only for SAP97 and ZO proteins. ZO-1 is clearly expressed in the heart, whereas controversial data exist for ZO-2 (178, 183). These MAGUKs share a common core of three PDZ, one SH3 and GUK domains but differ in their alternative splice insertion domains and in their COOH termini. The carboxy region of ZO-1 has a proline-rich domain with several PXXP motifs that can bind to SH3 domains. Biochemical data indicate that the gap junction channel connexin-43 is linked to ZO-1 and that this interaction regulates the localization of connexin-43 at the intercalated discs. It has been proposed that the transport of connexin-43 at the intercalated discs is achieved by translocation of associated ZO-1 (428). In contrast to data obtained in cell lines, ZO-1 and connexin-43 show low colocalization at the level of the ID in ventricular myocardium which is, however, enhanced following myocyte isolation and gap junction disruption, thus suggesting that ZO-1 can regulate connexin-43 internalization (34).
In the heart, two SAP97 isoforms are expressed, both containing the alternatively spliced insertions I3 and I1B, but differing in the presence or absence of I1A (145). SAP97 contains different combinations of alternatively spliced insertions that regulate the subcellular localization of the protein and its capacity to homo- or heterodimerize (247, 270). The I3 domain regulates the anchoring of SAP97 to the cytoskeleton through the protein 4.1/ERM and the E-cadherin-catenin adhesion complex (247, 346, 457).
SAP97 is expressed both in atrial and ventricular myocytes. Immunofluorescence studies performed in myocardial sections show that SAP97 is predominantly localized at the extremity of myocytes at or near the ID, but some staining can also be seen laterally along the plasma membrane (110, 146, 232, 294, 328, 330). One study found that SAP97 is localized also to the t-tubule system in isolated ventricular myocytes (232). Further evidence that cardiac SAP97 is targeted to specialized membrane domains was obtained through the use of a model of cultured adult cardiac myocytes, in which the in vitro formation of specialized myocyte-myocyte contacts such as adherens and gap junctions can be followed (142, 210, 363). In this model, both endogenous and exogenous SAP97 localizes exclusively at newly formed myocyte-myocyte contacts containing cadherins and connexin channels (1). Co-culture of cardiac myocytes and sympathetic neuronal cells has revealed another distinct localization of SAP97 at the contacts between nerve endings and myocytes, where the anchoring protein appears to be part of the β-adrenergic signaling complex (393). Taken together, these studies clearly indicate that SAP97 is not randomly distributed but is part of protein complexes involved in the formation of specialized membrane domains in cardiomyocytes.
Several cardiac ion channels have been shown to interact with SAP97 including Nav1.5, Kir2.x, Kv4.x, Kv1.5, as well as HCN-2 and -4 (110, 146, 198, 232, 328, 330). The interactions of Nav1.5 and Kv4 channels involve the PDZ domains as previously described for the interaction between neuronal PSD95 and Kv1.4 α-subunit (110, 198, 330). In addition, a combination of protein microarrays and quantitative fluorescence polarization assays confirm a high binding selectivity of PDZ domains for Kv channels (413).
The functional consequences of SAP97/cardiac ion channel interactions have been studied in heterologous expression systems by recording ion currents in cells coexpressing cloned channels and SAP97. In all of these studies, SAP97 enhanced the corresponding current (110, 330). The stimulatory effect of SAP97 on currents is not associated with changes in biophysical properties of the currents; thus major modifications of the intrinsic properties of the channels do not underlie the results. This was shown in single-channel recordings of Kir2.3 and Kv1.5 channels using the cell-attached configuration of the patch-clamp technique in myocytes and in cell lines (1, 441). In both cases, SAP97 induced no change in unitary conductance or open probability of channels but increased the density of functional channels. At the protein level too, more Kv1.5 α-subunits were detected by Western blot assay in the membrane protein fraction of cardiac myocytes overexpressing SAP97 (2).
Experiments in adult cardiac myocytes in which SAP97 was silenced have provided direct evidence for the crucial role played by endogenous SAP97 in cardiac excitability. Silencing of SAP97 in adult rat cardiac myocytes caused a drastic reduction in the outward voltage-gated potassium current (Ito) and in the sodium current (INa) (110, 330). In keeping with the interaction between SAP97 and the various endogenous Kv α-subunits underlying cardiac outward K+ currents, both the inactivating and sustained components of Ito were reduced (110). Clearly SAP97 enhances the density of functional channels in the sarcolemma. Noteworthy, indirect interactions have also been described, involving the NH2 terminus of the Kv1.5 channel, SAP97, and α-actinin and could contribute to the formation of MAGUK-channel complex (111, 262).
How SAP97 enhances the number of functional channels in the plasma membrane is not yet completely understood. The following mechanisms have been proposed: retention, increased forward trafficking, aggregation of channels, and formation of signalosomes. Several studies indicate that SAP97 retains and stabilizes channels at the plasma membrane, much as PSD95 does with neuronal Kv1.4 channels (186) and as Lin-7/CASK complex does for the retention of Kir2.3 channel at the basolateral membrane in epithelial cell lines (313). In cardiac myocytes, the overexpression of SAP97 is associated with an accumulation of Kv1.5 subunits at the sarcolemma and a reduction in the diffuse and cytosolic localization of the channel (see FIGURE 6). This diffuse localization of channels represents a subpopulation of proteins localized in vesicles involved in recycling and other trafficking (32). Thus SAP97 may well retain Kv channels at the plasma membrane by inhibiting their endocytosis and recycling. In mouse atrial cardiomyocytes (HL-1 cells), it has been shown that genomic upregulation of SAP97 results in the stabilization of Kv1.5 channels at the plasma membrane, an effect that requires the integrity of the trafficking machinery (423).
Another mechanism underlying the effects of SAP97 on channels is related to its capacity to multimerize and to form a large protein network, thereby facilitating interactions of channel subunits with accessory proteins (95, 169). Indeed, Kv1.5 current is not enhanced by the cardiac SAP97 isoform containing the I1A domain, which prevents protein oligomerization (145, 199). Oligomerization of SAP97 is regulated by inter- and intramolecular interactions (302), which determines the conformation of SAP97: compact “closed” state (preventing binding of ligands to the SH3 and GUK domains) or “open” state, allowing access to the protein binding sites (456). It is possible that in cardiac SAP97 the presence of I1A favors the “closed” state, via its binding to the SH3 domain and intramolecular interactions, in turn preventing MAGUK oligomerization. Of note, a point mutation at the tryptophan residue involved in binding to the SH3 domain of proline-rich motifs (consensus sequence PXXPR) restores the enhancing effect of the I1A-SAP97 MAGUK on the K+ current (145).
The capacity of SAP97 to interact with multiple proteins and to form networks explains its role as a linker between channels or receptors and signaling pathways. For instance, SAP97 is required for the G protein regulation of the inward rectifier K+ channel (436). Phosphorylation of β-adrenergic receptors by cAMP-dependent protein kinases depends on the interaction between SAP97 and AKAP (169). As mentioned above, SAP97 clusters with β-adrenergic receptors at the site of contact between myocytes and sympathetic neurons. In atrial myocytes, we found that SAP97 links Kv4.x subunits and the calcium calmodulin-dependent protein kinase CaMKII (110). Mutated Kv4.x channels lacking the PDZ binding domain at their COOH termini are no longer regulated by CaMKII. Complexes containing CaMKII, SAP97, and ion channels or membrane receptors have been described for Kv4.2 channels and the N-methyl-d-aspartate (NMDA) receptor (35, 137, 263).
In mammals, the ankyrin family of proteins is encoded by three genes (ANK1–3) distributed in many different tissues (86). The main function of ankyrin proteins is to anchor membrane ion transporters, such as ion channels and ion-transporting ATPases, to the actin and spectrin cytoskeleton. ANK1 encodes ankyrin-R, which was the first described of these proteins, initially found in erythrocyte membranes, forming the band2.1 protein (37). In the heart, both ankyrin-B (encoded by ANK2) and ankyrin-G (ANK3) have been found to be expressed (161).
Ankyrin-B is localized both at the level of the M-line and Z-line of ventricular myocytes (279, 282). In mouse cardiac cells expressing only one allele of ankyrin-B (ANK2+/− mice), it has been shown that the sodium pump, the Na/Ca exchanger, and the inositol 1,4,5-trisphosphate (IP3) receptor were all aberrantly localized (282). In humans, many genetic variants of ANK2 have been linked to diverse cardiac pathological phenotypes: LQTS type 4 (282), sinoatrial node dysfunction, atrial fibrillation, conduction slowing, and sudden cardiac death, thereby defining an “ankyrin-B cardiac syndrome” (284). This phenotypic diversity most likely reflects the important and pleiotropic roles of this anchoring protein in cardiac cells.
Ankyrin-G has been shown to play a role in the trafficking and anchoring of the cardiac sodium channel Nav1.5 (244, 285), and, similar to that channel, ankyrin-G is also expressed at the intercalated discs and in the t-tubules of adult myocytes (283). Lowe et al. (244) have shown that reduction of ankyrin-G expression in cardiac myocytes leads to reduced sodium currents and lowered Nav1.5 expression. This effect was specific for the ankyrin-G isoform, as this reduction was not observed in myocytes with reduced expression of ankyrin-B. The precise role of ankyrin-G in the anchoring and trafficking of Nav1.5 in the different compartments is not well understood. Its importance, nevertheless, is well illustrated by the observation that in one patient with Brugada syndrome a missense mutation (p.E1053K) in the Nav1.5-encoding SCN5A gene was found that disrupts the interaction between Nav1.5 and ankyrin-G (281). These mutant Nav1.5 channels remain in the cytoplasm of adult rat myocytes. In a very recent study, it has been reported that ankyrin-G is a key functional component of the intercalated disc at the intersection of three complexes: Nav1.5, gap junctions, and the cardiac desmosome (374). Hence, ankyrin-G could regulate not only the properties of the cardiac sodium current but also underlie the integrity of mechanical and electrical coupling. As such, it is likely a key molecule for impulse propagation in the heart.
3. Dystrophin-syntrophin complexes
Dystrophin is a large protein of ∼430 kDa that is mainly expressed in striated muscle cells, including cardiac cells (172). It most likely plays multiple roles, but the main function of dystrophin is similar to that of the ankyrin proteins, i.e., it links membrane proteins to the actin cytoskeleton (43). The importance of dystrophin in the heart is illustrated by the fact that patients with mutations in its gene (DMD) virtually all exhibit severe cardiac phenotype. Most Duchenne and Becker patients suffer from a dilated cardiomyopathy (127). Moreover, the X-linked form of dilated cardiomyopathy is also caused by mutations in DMD (427).
Dystrophin stabilizes a large macromolecular complex of several proteins, the DGC, at the lateral membrane of cardiac cells (43). Chief among these are the heavily glycosylated proteins dystroglycan and sarcoglycan (115). Dystrophin has multiple protein-protein interaction domains, but can also interact indirectly with additional proteins via the syntrophin adaptor proteins (12). At least three of the five syntrophin genes have been found to be expressed in cardiac tissues (180). These syntrophin proteins interact with a specific domain in the COOH terminus of dystrophin and, via its PDZ domain, with the COOH-terminal tail of various cardiac ion channels. The cardiac sodium channel Nav1.5 has a PDZ-domain binding motif in its COOH-terminal tail. This domain was found to be necessary for indirect interaction with dystrophin via syntrophin proteins (140). Note that, among the nine different voltage-gated sodium channels, only the ones expressed in striated muscle cells, i.e., Nav1.5 in cardiac cells and Nav1.4 in skeletal muscle cells, have such a PDZ-domain binding motif (140). Gavillet et al. (138) have shown that in dystrophin-deficient cardiac cells from mdx mice, Nav1.5 protein expression was decreased and that the sodium current was accordingly diminished (see FIGURE 8). It was subsequently demonstrated that this interaction between the cardiac sodium channel and dystrophin was exclusively occurring at the lateral sarcolemma of cardiac cells since both dystrophin and syntrophin are virtually absent at the intercalated discs (330) (see FIGURE 7). Interestingly, it was also recently shown that the protein utrophin, a homolog of dystrophin that is upregulated in dystrophin-deficient cells (92), also interacts with Nav1.5 (11). The details about how the dystrophin complex regulates the expression and possible localization of Nav1.5, and what its precise role is are still not known. It seems however that these mechanisms likely involve posttranslational regulatory mechanisms since, in dystrophin-deficient tissue, the mRNA for Nav1.5 was found to be unaltered (138). It may be proposed that one of the roles of the dystrophin complex is to increase the stability of the channel at the cell membrane by reducing its internalization rate.
Recent studies have found mutations in the gene coding for syntrophin α1 (SNTA1) in patients with congenital long QT syndrome (LQTS) (433, 455). The SNTA1 mutation p.A390V (433) was shown to disrupt a sarcolemmal macromolecular complex formed by neuronal NOS (nNOS) and the plasma membrane Ca2+-ATPase type 4b with syntrophin and Nav1.5. It was found that expression of the mutant syntrophin protein in cardiac myocytes increased the late Na+ current, which may explain the prolongation of the QT interval in patients harboring this mutation. The proposed mechanism is that the mutation of syntrophin disrupts the interaction between nNOS and the Ca2+-ATPase leading to increased nitrosylation of the channel and subsequently the stimulation of the Nav1.5 late current.
Other cardiac ion channels have also been found to be potentially regulated by the syntrophin/dystrophin complex (18). Using a proteomic approach, Leonodakis et al. (231) showed that, among many other proteins, the PDZ-domain binding motif of Kir2.2 interacts with dystrophin and dystrophin-complex proteins from heart lysates; however, the functional significance of this observation is not yet known. The L-type voltage-gated calcium channel was also found to be altered in absence of dystrophin, but without any direct evidence that dystrophin may directly or indirectly bind to the channel (367, 454).
Caveolin proteins are important constituents of plasma membrane invaginations called caveolae (see sect. VA), in which there is an enrichment of signaling molecules and ion channels (453). Caveolin-3 (CAV3) is the main caveolin isoform expressed in cardiac muscle. Emphasizing the role of CAV3 in cardiac function, mutations in its gene have been found in patients with congenital LQTS (439) (LQTS type 9) and sudden infant death syndrome (SIDS) (83). Mutations in CAV3 have also been linked to other pathologies such as limb-girdle muscular dystrophy, rippling muscle disease, and familial hypertrophic cardiomyopathy (76). Many cardiac ion channels have been recently shown to interact with and to be regulated by CAV3. These are the L- and T-type cardiac calcium channels (256), and the following potassium channels: KCNQ1, Kv1.5, and KATP, as well as pacemaker channel HCN4. These CAV3-dependent regulatory mechanisms have been extensively reviewed by Balijepalli and Kamp (31). In this article we focus on the regulation of the cardiac sodium channel Nav1.5. In an important study, CAV3 has been shown to coimmunoprecipitate with Nav1.5 from rat cardiac tissue (464) and when coexpressed in cell systems (439). However, to date, the site of interaction between CAV3 and Nav1.5 is still unknown. Immunohistostaining experiments using cardiac tissues clearly demonstrate that the two proteins are colocalized (398, 439, 464). Dystrophin has also been shown to be a component of caveolae (105), raising the possibility that the interaction of CAV3 with Nav1.5 could be indirect, via proteins of the DGC (138). Interestingly, coexpression of Nav1.5 with mutants of CAV3 found in LQTS and SIDS patients increased the inward persistent Na+ current (83, 439). These findings are at least consistent with the clinical phenotypes of the patients with congenital LQTS.
C. Targeting of Ion Channels in Cardiomyocytes
The question of the nature of mechanisms regulating the targeting of ion channels to distinct subdomains of the sarcolemma has been particularly addressed for the dynamic formation of gap junction channels, which are continuously renewed and recycled. The current view is that hemichannels secreted from the Golgi apparatus are trafficked along microtubules in vesicles and delivered to the cell surface in nonjunctional areas of the plasmalemma. After their insertion, the connexons then move laterally until reaching the periphery of clustered hemichannels. This accretion of new hemichannels ensures the growth of gap-junctional plaques (227). The hemichannels then move into the cluster, i.e., to the center of the plaque, to be internalized as annular gap junctions and targeted for degradation (388).
This view has been challenged, however, by the demonstration of direct targeting of hemichannels to junctional plaques through EB1, a microtubule plus-end-tracking protein (+TIP). In addition to a binding site for the distal end of growing microtubules, the EB1 dimer possesses binding sites for the p150 subunit (p150-Glued) belonging to the dynein/dynactin complex (40). p150 anchors microtubules to adherens junctions via its interaction with N-cadherin and β-catenin (64, 237). Using different imaging approaches in living cells, Shaw et al. (392) showed that EB1-capped microtubules interacted more frequently with and remained for longer period in regions containing gap junction plaques. The molecular bridge formed by microtubules, EB1, p150, and adherens junction proteins allows the delivery of hemichannels to points of cell-cell contact and, therefore, the direct targeting of gap junctions to adherens junctions (392).
The question of the targeting of ion channels or other transmembrane proteins at the level of the lateral cell membrane has poorly been investigated in cardiomyocytes so far. However, whereas microtubules are preferentially longitudinally oriented in cardiomyocytes, the actin cytoskeleton displays both a longitudinal and transversal pattern, as shown by F-actin staining with phalloidin (personal observation and Refs. 266, 422). The actin cytoskeleton is indeed connected to focal adhesion sites corresponding to the costamere in cardiomyocytes (211, 266). The sarcomeric α-actinin protein, which faces the Z-line of the myofibrils, is also aligned with the costamere system of the lateral membrane and therefore shows a transversal organization. Our preliminary electron microscopy observations (Balse and Hatem, unpublished data) unraveled the presence of many vesicles along the Z-band, therefore suggesting that trafficking also occurs along the transversal axis of cardiomyocytes, whether the direction (i.e., anterograde or retrograde) of the trafficking is not established yet. Moreover, the fact that impairment or destruction of the actin cytoskeleton (445) as well as silencing of α-actinin-2 profoundly affect the trafficking of ions channels and their subsequent density in the sarcolemma (260) is also in favor of the role of these structural proteins in the trafficking and the targeting of ion channels. Unfortunately, these works were carried out on expression systems and, clearly, the respective role of actin cytoskeleton, α-actinin, and costamere interacting proteins has to be investigated in adult cardiomyocytes to decipher their involvement in the targeting ion channels to the specialized substructures of these highly polarized cells.
In this context, the targeting of calcium channels is also of great interest. The majority of Cav1.2 channels, the prominent cardiac calcium channel, are located in t-tubules (see sect. II) and contribute more than 80% of the L-type calcium current of cardiac myocytes. However, the mechanisms by which Cav1.2, like other t-tubule targeted proteins, are localized in distinct sarcolemmal domains are still poorly understood. Recently, a hint as to how this may occur was reported. BIN1 (also known as amphiphysin) has been reported to be involved in the targeting of Cav1.2 to t-tubules in cardiomyocytes (174). BIN1 belongs to the BAR domain protein superfamily known to be involved in membrane invagination and endocytotic processes (see sect. V). In neurons, amphiphysin is expressed in zones of endocytosis at nerve terminals. Its association with dynamin, clathrin, and the adaptor protein 2 (AP2) confer its role in clathrin-dependent endocytosis and synaptic vesicle recycling (99). Moreover, its role in neurite outgrowth has also been reported (293). In skeletal muscle cells, BIN1 has been reported to initiate the formation of t-tubules (230). In mouse and human cardiomyocytes, BIN1 and Cav1.2 reside in t-tubules in very close proximity (10–50 nm). Total internal reflection fluorescence microscopy (TIRFm) imaging indicates that microtubules carrying Cav1.2-containing vesicles are preferentially tethered to BIN1-expressing membrane areas in both myocytes and in heterologous cell lines (174). In differentiated adult myocytes, BIN1-silencing reduces Cav1.2 expression at the sarcolemma and alters EC coupling. Therefore, in addition to its role in membrane deformation, BIN1 appears to be involved in anterograde trafficking and, importantly, to drive targeting of Cav1.2 channels to specific subdomains (174).
Anchoring proteins are also important factors for this targeting of ion channels to distinct plasma membrane domains. This is suggested by the existence of distinct subpopulations of Nav1.5 channels associated with distinct anchoring proteins in cardiac myocytes. While Nav1.5 channels located at the lateral membrane are predominantly anchored to the dystrophin-syntrophin complexes, the subpopulation of sodium channels localized at the ID is most likely associated with the SAP97. Hence, depending on the interacting proteins, Nav1.5 is found at least in two different pools (330). As discussed above, desmosomal proteins such plakophilin2 could contribute also to the targeting of a subpopulation of Na+ channels to the ID (374).
These different observations point to the need to study further the proteins regulating the targeting of ion channels in cardiomyocyte substructures (either intercalated disc or lateral membrane) in respect to the channel subtype considered.
V. FATE OF ION CHANNELS
Endocytosis is the process used by cells to internalize extracellular material, ligands, as well as plasma membrane constituents, such as proteins and lipids. Internalization requires membrane bending and fission as well as dynamic remodeling of the membrane actin cytoskeleton. The coupling of these processes is ensured by molecular scaffolds constituted by BAR (Bin/Amphiphysin/Rvs) domain proteins that bring together actin and invaginating membranes (for review, see Refs. 91, 344). The composition of the plasma membrane is controlled by the tight association of endocytosis with the recycling of internalized cargo and/or the delivery of newly synthesized constituents. Both processes require the integrity of the actin cytoskeleton and microtubules and their respective associated molecular motors.
1. Diversity of endocytic pathways
Endocytosis is carried out by various mechanisms but can be roughly divided into clathrin-dependent endocytosis (CDE) and clathrin-independent endocytosis (CIE). After internalization, cargo can be recycled back to the plasma membrane or sorted for degradation in specific compartments such as proteasomes or lysosomes. The outcome depends either on internalization signals carried by the cytoplasmic end of the internalized protein (also responsible for constitutive endocytosis) or on posttranslational modifications of the cargo protein such as dynamic ubiquitilation or phosphorylation (see FIGURE 8).
CDE is responsible for 40–50% of total endocytic processes in mammalian cells (220). The transferrin (Tf) and low-density lipoprotein (LDL) receptors are canonical cargoes of this endocytic pathway. Clathrin is the main constituent of specialized membrane structures, clathrin-coated pits (CCP), that concentrate membrane proteins destined to be internalized. Clathrin is constituted of three heavy chains associated via their COOH terminals, each associated with a light chain, forming a triskelion. These triskelions auto-assemble to form cages, easily recognizable in electron microscopy.
The CCP has an average depth of 100 nm, as measured by scanning surface confocal microscopy (395). The pit invaginates until the formation of a clathrin-coated vesicle (CCV) that is cleaved from the membrane upon the action of the dynamin GTPase. Following clathrin uncoating, the vesicle and its cargo are transferred to the sorting/early endosome. However, clathrin is unable to bind directly to cargo or membranes. Another group of protein mediates the interaction between cargo protein and clathrin, the so-called “endocytic adaptors” (252, 324). Endocytic adaptors themselves can be classified into two groups: multimeric (or classic heterotetrameric) adaptor proteins and monomeric (or nonclassical) adaptor proteins (345). The canonical classic adaptor protein is AP-2, the most abundant non-clathrin component of endocytic vesicles (195). The AP-2 complex is a tetramer comprised of two large subunits (α, β2), one medium subunit (μ2), and one small subunit (δ2). The large subunits are responsible for membrane targeting, clathrin binding, and the recruitment of accessory proteins (317, 429). The small subunit stabilizes the complex and binds to a specific motif on the cargo protein (103, 196), and the medium subunit possesses both lipid and cargo-binding motifs (316). The interaction of the α subunit with the lipid phosphatidylinisitol 4,5-bisphosphate of the plasma membrane triggers the recruitment of AP-2 to the plasma membrane. Cargo-binding motifs of adaptors then interact with cytoplasmic tails of the protein destined to be internalized (reviewed in Ref. 345). The phosphorylation of T156 of the μ2 subunit seems to be necessary for efficient interaction of AP-2 with membrane or cargo protein (126, 314). The other (nonclassical) family of adaptor proteins involved in endocytosis is the CLASP (clathrin-associated sorting proteins) adaptor family. These monomers or dimers contain regions usually able to interact with membrane lipid, cargo, accessory proteins, or clathrin. This nonclassical family is composed of ubiquitin-binding adaptors (epsin 1, epsin 2, and the epidermal growth factor receptor pathway substrate 15 or EPS15), cargo-specific adaptors (Numb, LDLRAP1 or ARH, and DAB), lipid-binding adaptors without cargo selectivity (AP180 or CALM or PICALM), cargo-binding adaptors lacking lipid- and clathrin-binding selectivity (EPS15), and the arrestin (β-arrestin) and muniscin (FCHO1, FCHO2, SGIP) families of endocytic adaptors (see Ref. 345).
CIE is less well characterized than CDE, yet is increasingly gaining attention. Besides phagocytosis and macropinocytosis occurring in specific cell types, CIE occurs via many means that are only beginning to be elucidated. Two main CIE pathways can be distinguished: caveolin-mediated endocytosis (nonexistent in neurons) and clathrin- and caveolin-independent endocytosis. Caveolins 1–3 are integral membrane proteins that form a hairpin loop inside the membrane, leaving both COOH and NH2 terminals oriented towards the cytosol. Caveolins are the main constituents of caveolae, small invaginations of the plasma membrane first identified by electron microscopy as “little caves” (319). These flask-shaped invaginations of the plasma membrane lack an electron-dense coat (average depth of 60–80 nm) (36, 394; for review, see Ref. 23) and are enriched in cholesterol; they are classically assimilated to lipid rafts. Caveolae constitute the route taken by integrins, albumin, GPI-anchored proteins, and simian virus 40 (SV40) to enter the cell (69, 217, 325). As with clathrin, the final internalization step of caveolin-mediated endocytosis is catalyzed by the cleavage of the caveolin-bud by the dynamin GTPase. The internalized endocytotic caveolar carrier (ECC) is next sorted to the early endosome then either routed to the late endosome or recycled back to the plasma membrane.
Another endocytic mechanism distinct from both clathrin-coated pits and caveolae, and, moreover, dynamin-independent, has been described in mammalian cells (144) involving Flotillin-1 (or reggie-2), a membrane-microdomain-associated protein that is localized in the cell membrane in different cell lines. Its function in clathrin-independent endocytosis has been elucidated in purified endosome preparations. Flotillin-1-positive endosomes accumulate both GPI-anchored proteins and cholera toxin B subunit. Flotillin resides in punctuate structures both in the plasma membrane and in specific populations of endosomes. Total internal reflection fluorescence microscopy and immunoelectron microscopy have revealed that regions of the plasma membrane containing flotillin-1 bud into the cell distinctly from clathrin-coated pits and caveolae (144). The average depth of these flotillin-1 invaginations has been estimated at 200 nm by scanning surface confocal microscopy (394).
Several questions remain to be answered. For example, it is not well established whether adaptor proteins are generally involved in non-clathrin-mediated endocytosis. However, clathrin adaptors ESP15 and epsin have been shown to be necessary for clathrin-independent internalization of ubiquitylated EGFR (193), and the neuron-specific adaptor DAB1 has also been shown to be involved in the caveolin-dependent internalization of albumin in astrocytes (39). Second, the relative contribution of the different endocytic pathways in a single cell is not known. Indeed, blocking clathrin aggregation beneath the plasma membrane by overexpression of AP180-C enhances caveolar-mediated endocytosis (300). Other studies have also shown that blockade of one type of endocytosis favors upregulation of alternative pathways (88). This emphasizes the difficulty in unraveling functional contributions of each pathway. In COS-7 cells, the relative contribution of each endocytic pathway has been estimated to be 89% for clathrin, 9% for caveolae, and the remainder for flotillin pits (394). Probably, this distribution differs from one cell type to another. Third, the internalization speeds of the different endocytic pathways have not been clearly determined, although the caveolae-dependent pathway has been reported to be poorly dynamic (421) and flotillin-associated pits seem to internalize three times slower than clathrin-coated pits (144). Finally, the possibility that a single cargo molecule may traverse different pathways depending on the cellular conditions or environment remains to be investigated.
2. Endocytosis in the myocardium
Little is known about internalization processes occurring in cardiomyocytes for the different ion channels. However, since the rate of endocytosis will directly influence the density of ion channels and connexins in the sarcolemma, it must have profound repercussions on cardiac excitability. Therefore, as discussed in section VIIB, elucidating the endocytosis pathway of specific cardiac ion channels is of particular interest for understanding how cardiac ion channel density is regulated in cardiovascular diseases.
Many studies have explored the routes taken by cardiac ion channels in expression systems. However, despite, for example, the fact that dynamin-dependent endocytosis has been shown to be operative for various ion channels such as Kv1.2 (299), Kv1.5 (72, 267, 384), Kv7.1 (459), and Cav1.2 (174), the specificity of the pathway(s) has not been characterized. hERG channels are internalized by the caveolar pathway in HEK 293 cells (261), such as the Nav1.5 channel in ventricular tissue (164). The role of clathrin in the internalization of connexins has been clearly established in both cell lines and myocardium (152, 167, 168, 332). Interestingly, the internalization pathway used by cardiac ion channels appears to be dependent on the cell type considered. For instance, in the immortalized atrial cell line HL-1, the Kv1.5 channel has been shown to be strongly expressed in cholesterol-rich microdomains where it colocalizes with caveolin (257, 267, 384). In native rat and canine myocytes however, Kv1.5 is poorly associated with caveolin-3 as demonstrated by confocal and immunoelectron microscopy (112). We (Balse and Hatem, unpublished data) also have failed to see any colocalization between Kv1.5 channel and caveolin-3 in freshly isolated or cultured rat atrial myocytes. Using whole cell patch-clamp recordings, three-dimensional immunofluorescence microscopy and biotinylation assays, we observed that Kv1.5 internalizes via the clathrin pathway (Balse and Hatem, unpublished data).
3. Transport of internalized cargoes
As noted in section IIIC, the trafficking of ion channels in the cytoplasm is coordinated by members of the Rab GTPase family. Sequential associations and dissociations of Rab GTPases couple each stage of trafficking to the next along the pathway. Several Rab GTPases have been directly involved in the trafficking of ion channels expressed in the heart after their internalization. Among them are Rab4, Rab5, Rab7, Rab9, and Rab11. Rab5 is involved in clathrin-mediated endocytosis and is essential to early endosome formation and maturation (474). Rab4 associates with early endosomes and is involved mainly in fast recycling to the plasma membrane (407). Rab11 is essential to a slower recycling pathway and is also involved in the trafficking of membrane proteins from the Golgi to the plasmalemma (67, 375). Rab9 is involved in transport from late endosome to the TGN (243), and Rab7 is also associated with the late endosome but instead directs the trafficking of its contents to the lysosome (120, 337) and to the proteasome (102) for degradation. All of these Rab GTPases are expressed in cardiomyocytes (125, 197, 239, 311).
Of cardiac-expressed channels, CFTR (143, 378, 449), KCNQ1/KCNE1 (386), and Kv1.5 (267, 469) have all been shown to be dependent on these GTPases for their endocytosis and/or postinternalization transport in at least some tissue types. Confirming a role for Rab5 in the endocytosis of ion channels, cystic fibrosis transmembrane conductance regulator (CFTR) endocytosis is greatly reduced by expression of a Rab5DN in BHK-21 cells (143). Kv1.5 endocytosis also is dependent on Rab5 in both HEK293 cells and in the H9c2 ventricular cardiomyoblast cell line. Expression of Rab5DN reduced its internalization, resulting in an increase of the corresponding current. Moreover, Rab5 colocalized with Kv1.5 in early endosomes in these cells (469). Interestingly, however, the Rab5DN was found not to be required for Kv1.5 endocytosis in the HL-1 atrial cardiomyoblast line as measured by immunocytochemistry assays (267). McEwen et al. (297) suggested that Kv1.5 could be internalized via caveolae, known to internalize independently of Rab5. Further experiments will be required to decipher whether this discrepancy results from a differential trafficking of Kv1.5 in various cell types or from the limitations of the techniques used.
Once a channel has internalized, it may have several fates. It may be held in an intracellular pool, be recycled to the plasmalemma, or be shunted to the lysosome or proteasome for degradation. Recycling to the plasma membrane occurs through two major pathways, a rapid Rab4-dependent route directly from the early endosome and a slower, Rab11-dependent pathway via the recycling endosome. The CFTR, KCNQ1/KCNE1, and Kv1.5 channels have all been shown to recycle to lesser or greater extents through both pathways (143, 267, 386, 469). The Rab11-dependent pathway may be the dominant route for CFTR (143). Expression of a Rab11DN almost completely eliminated the return of internalized CFTR; the Rab4DN had no obvious effect on CFTR recycling, although the size of the endosome pool was substantially increased. It is possible, however, that the lack of impact of the Rab4DN is due to an indirect interference with channel internalization. Indeed, Rab4DN increased Kv1.5 surface expression and corresponding IKur in HEK293 cells and H9c2 myoblasts (469). While the Rab4DN did substantially reduce the internalization of surface-labeled channel, it also decreased the return of that internalized channel to the plasma membrane. This reduction in Kv1.5 internalization may not be universal to all cell types. When expressed in the HL-1 cardiomyoblast, a Rab4DN dramatically reduced surface expression of the channel (267).
Rab11-dependent recycling appears to be regulated by membrane lipids. The Rab11-dependent return of KCNQ1/KCNE1 to the plasma membrane has been shown to be promoted by activation of serum- and glucocorticoid-inducible kinase 1 (SGK1) in a pathway that involves the stimulation of phosphoinositol 3-phosphate 5-kinase and the formation of phosphatidylinositol 3,5-bisphosphate (386). Activation of SGK1 also increases the recycling of Kv1.5 (434) and Kv4.3 (385) to the plasma membrane, although it is not known whether Rab11 is involved in these effects. Rapid delivery of Kv1.5 to the plasmalemma of atrial cardiomyocytes after cholesterol depletion is, however, clearly dependent on Rab11 function (32). Rab9 is involved in transport of internalized membrane proteins to the TGN (347). It has been linked to the maintenance of intracellular pools of CFTR, since its overexpression caused a substantial increase in intracellular pools of CFTR whereas expression of a Rab9DN dramatically reduced the size of those pools (143).
Not all internalized channels are recycled to the plasmalemma or stored in intracellular pools. At least a substantial fraction of internalized channels not destined for recycling are instead shunted for degradation. CFTR and Kv1.5 have both been shown to traffic, in part, to the degradation pathways. A portion of Kv1.5 colocalizes with Rab7 in H9c2 myoblasts, and overexpression of Rab7 causes a substantial decrease in Kv1.5 currents in these cells (469). Expression of a Rab7DN had little effect on Kv1.5 currents in these cells, suggesting that the channel is slow to enter this degradation pathway. In BHK-21 cells overexpressing Rab7, intracellular CFTR staining is also reduced (143). Rab27, also involved in lysosomal targeting, has been implicated as well in CFTR trafficking. Overexpression of Rab27a in HT-29 colorectal epithelial cells caused a dramatic reduction in CFTR surface expression, whereas siRNA against the Rab GTPase caused an increase in its expression (377). Overexpression of some known Rab27-binding proteins prevented the decrease in channel expression associated with Rab27a overexpression, although the mechanism(s) by which this occurred is unclear.
Clearly, Rab GTPases play crucial roles in the trafficking of ion channels expressed in the heart. While our understanding of their involvement is limited to only a few channels, there is no doubt about their universal importance. The major questions remaining are the degree to which all channel types traffic through the same Rab-defined vesicles and the mechanisms by which the decisions are made to shunt individual channel proteins for recycling, storage in intracellular pools, or degradation. To what degree are decisions made stochastically, as is evidently the cases for the transferrin receptor (219) and to what degree does posttranslational modifications influence the process need further investigation.
B. Triggers for Fate
Recently, the role of two posttranslational modifications, ubiquitylation and SUMOylation, in deciphering the fate of ion channels has gained a lot attention. The proteins modifiers ubiquitin and SUMO (small ubiquitin-related modifier) are structurally related and share a multitude of functional interrelations. Both bind to lysine residues on target proteins and compete for the same attachment site, involve a multistep enzymatic cascade for activation or inactivation, and show very labile interaction with their target. Several studies reveal that whereas ubiquitylation leads to protein internalization and degradation (170), SUMOylation rather favors, among its pleiotropic activities such as cell cycle progression, epigenetic modulation, signal transduction, and DNA replication/repair, protein stabilization (185). Below are presented recent data on the physiological roles of these mechanisms in the regulation of cardiac ion channels.
1. Ubiquitylation as a signal for internalization and degradation
For many years, it has been known that ion channels can be regulated by ubiquitylation, i.e., a covalent modification of the channel polypeptide by either single ubiquitin moieties (mono-ubiquitylation) and chains of ubiquitin proteins (poly-ubiquitylation) (for a review, see Ref. 5). The epithelial sodium channel (ENaC) was the first ion channel whose surface density was shown to be regulated by its ubiquitylation (4, 410). Many of the ion channels that have been reported to be ubiquitylated and regulated are the target of a specific family of enzymes called Nedd4/Nedd4-like E3 ubiquitin ligases (358). Virtually all these enzymes target channel subunits bearing a protein-protein interaction domain called a PY-motif, with consensus sequence L/PPxY, which allows the direct interaction with one of the three or four WW-domains of the Nedd4 ubiquitin ligases (358). Very interestingly, important cardiac ion channels have been shown to harbor well-conserved PY motifs, all in their respective COOH terminus, i.e., the cardiac sodium channel Nav1.5 (3), two potassium channels involved in the repolarization of human ventricular action potential KCNQ1 (184) and hERG (10), as well as connexin-43 (235). In all of these cases, it has been shown that the density of the given channel protein was negatively regulated by the activity of one of the Nedd4 ubiquitin ligases, in particular Nedd4–2, and in some cases, the expression of the protein was found to depend on the coexpression of the ligase. A detailed discussion of these phenomena can be found in a recent review (359). It should be pointed out, however, that while it seems clear that ubiquitylation is a signal for downregulation of the membrane pool of ion channels, it is not yet understood whether this represents an internalization and/or a degradation signal. It is also not yet clear whether such ubiquitylated channels are eventually degraded by the lysosomal system or by the proteasome complex. Note that a channel lacking a PY motif, i.e., the cardiac calcium channel Cav1.2, has been recently reported to be downregulated by the Nedd4–1 ligase (360). Cav1.2, itself, was not found to be the target of Nedd4–1 ubiquitylation activity. Instead, this downregulation likely involves the sorting of newly synthesized Cav channels for degradation. Importantly, this Nedd4–1-dependent regulation required the coexpression of Cav-β subunits, which are known to promote the export of newly synthesized proteins from the ER.
Similarly to other posttranslational modifications, ubiquitylation is a reversible process, with de-ubiquitylating enzymes catalyzing the removal of ubiquitin moieties (21). Here again, ENaC was the first ion channel reported to be de-ubiquitylated by the USP2 enzyme if ubiquitylated by Nedd4–2 ubiquitin ligase (116, 365). Still, to date, most of the evidence suggesting that the density, internalization, and degradation of cardiac ion channels are regulated by their (de)ubiquitylation comes from experiments using heterologous cellular expression systems (359). A great deal remains to be done to validate these findings in native cardiac cells and in vivo.
2. SUMOylation as a regulator of channel activity and density in the sarcolemma
Only a few studies have been conducted so far on SUMOylation of cardiac ion channels, and the physiological significance of this modification is far from clear. In heterologous expression systems, SUMOylation of the cardiac potassium channel Kv2.1 accelerates inactivation and inhibits recovery from inactivation of the Kv2.1-mediated current, IK,slow2 and ISS. Therefore, the resulting reduced current could promote AP elongation (87). The atria-specific Kv1.5 channel possesses two SUMOylation consensus sites, located close to plasma membrane spanning domains, which have been shown to be targets for mono- or poly-SUMOylation (38). In Cos-7 cells, deSUMOylation of the Kv1.5 channel, either by truncation of the SUMOylation sites or by overexpression of the SUMO protease SENP2, leads to a hyperpolarizing shift in the steady-state inactivation of the related IKur current. However, no change in IKur density was reported. The authors concluded that the lability of the SUMOylation process could constitute a significant mechanism for the fine tuning of the electrical activity of cells.
The K2P1 channel that regulates the background potassium current in many cells has also been reported to be a target for SUMOylation. If the wild-type K2P1 channel is heterologously expressed at the plasma membrane of Xenopus oocytes, its deSUMOylation is necessary to generate an outward potassium current (342). However, other groups have reported that this is true only when the wild-type lysine-274 residue is mutated to glutamic acid. Indeed, when the lysine is replaced by the positively charged arginine residue, any potassium current could be recorded (119). Thus it is still a matter of controversy whether the K2P1-mediated current is controlled or not by SUMOylation. Finally, the most recent evidence for direct regulation of ion channel function by its SUMOylation in the heart has been provided by the identification of a missense mutation in the gene encoding the nonselective cation channel TRPM4 in progressive familial heart block type I patients (see sect. VII). Because SUMOylation may compete with ubiquitylation, the balance between these two antagonistic processes in cardiac cells will likely be a central topic for future research.
VI. LIPIDS AS REGULATORS OF ION CHANNELS TRAFFICKING
Lipids are another important partner for the proper surface expression of ion channels in cardiac myocytes. Recently, the role played by cholesterol as a regulator of surface expression and organization of ion channels in membrane has been garnering a great deal of attention. Several studies have demonstrated that membrane lipid composition/cholesterol content influences the density of ion channels and receptors at the cell surface, thus modulating the corresponding currents in several cell expression systems. For instance, lowering membrane cholesterol with methyl-β-cyclodextrin (MβCD), a cyclic oligomer of glucose that extracts cholesterol form cell membranes, can affect multiple steps of ion channels or receptors trafficking. Indeed, cholesterol depletion can alter endocytosis, redistribute proteins among membrane subdomains, affect the mobility of endosomes, and/or affect recycling processes (see FIGURE 9).
A. Cholesterol, Lipid Rafts, and Channels
Cholesterol is a key factor for compartmentalization and clustering of ion channels in large multiprotein complexes. One mechanism involves the role of cholesterol in the formation of lipid rafts. Cholesterol is not randomly distributed in the plasma membrane but is packed together with sphingolipids to form cholesterol-enriched domains known as lipid rafts. Given the paucity of tools for the direct study of lipid rafts, caveolin has been widely used to indirectly study the compartmentalization of proteins in cholesterol-enriched domains of the membrane. These indirect approaches consist mainly in coprecipitation and colocalization of a protein of interest with caveolin. Indeed, a number of cardiac ion channels have been found to localize into caveolae (31, 90, 251). However, some controversy exists because of differing results obtained in studies performed in heterologous expression system and in cardiac myocytes. For instance, while in CHO cells Kv1.5 channels colocalize and coprecipitate with caveolin, in the atrial myocardium Kv1.5 channels predominates at the intercalated disc, whereas caveolin-3 is located at the periphery of myocytes, not in the interacted disc (240, 464). Moreover, the two proteins extracted from cardiac samples fail to coprecipitate (112). Given the strong effect of cholesterol on Kv1.5 channels of atrial myocytes (2, 32), these results raise the possibility that Kv1.5 channels are localized in lipid rafts distinct from caveolae. It is also possible that the localization of Kv1.5 channels in lipid rafts is transient, perhaps taking place in the time between channel delivery to the plasma membrane and their recruitment into specialized domains, much as has been described for connexins (240, 382).
B. Cholesterol and Clustering of Ion Channels
Cholesterol is also part of the fence that delineates clusters of channels. Clustering is a general feature of channel organization (177) and has been observed both in heterologous expression systems and in native tissues. For instance, Kv2.1 channels cluster at the level of postsynaptic sites on the soma and proximal dendrites in the rat spinal cord (290). Kv1.5 channels also forms clusters in the atrial myocardium and in isolated myocytes (2, 100). Clusters are well delineated membrane domains where channels are delivered (309). In atrial myocytes (2) or in HEK cells (2, 309), cholesterol depletion causes an increase in size and number of clusters, a strong indication of the importance of the lipid composition of the membrane in channel segregation. With the use of the cell-attached configuration of the patch-clamp technique and confocal imaging, it has been shown that clusters also contain nonconducting channels. These nonconducting channels could have distinct signaling properties or represent a reserve of channels that can be released for function upon various stimuli or cholesterol depletion (308).
C. Cholesterol and Endocytosis
In animal cells, cholesterol depletion leads to flattening of caveolae and blocks internalization through the caveolin pathway (68). Cholesterol has been reported also to play a role in internalization through the clathrin endocytic pathway (351, 415). Indeed, cholesterol removal prevents the budding of the plasma membrane and the formation of clathrin-coated pits (351, 415). For instance, cholesterol depletion with MβCD has been reported to reduce the rate of internalization of the TfR in cell lines (415). Ultrastructural studies revealed that cholesterol depletion inhibits the budding of clathrin-coated pit and causes an accumulation of flat clathrin lattices beneath the cell surface. The resulting internalization failure of TfR was therefore attributed to the inability of clathrin to induce curvature in cholesterol-depleted membranes.
Clathrin recruitment and self-assembly is regulated by adaptor molecules such as AP2. In the proximal tubule cell line LLC-PK1, MβCD reduced ouabain-induced accumulation of the Na+-K+-ATPase α-1 subunit in clathrin-coated vesicles. A reduced interaction between this α-1 subunit, AP2, and clathrin heavy chain was also observed, suggesting that cholesterol can regulate endocytosis by modulating protein-protein interactions during the clathrin-mediated endocytosis process as well (239). While no study has reported any direct effect of cholesterol on dynamin activity, it is known that its activity is highly regulated by its binding to acidic phospholipids (239, 432). Therefore, changes in cholesterol content and in local lipid environment may affect overall internalization processes, and probably dynamin activity as well. Finally, Borroni et al. (48) have reported that cholesterol depletion can stimulate internalization of nicotinic acetylcholine receptors, resulting in a gain-of-function of the remaining cell-surface receptors in expression systems. Further studies by the same group have shown that the internalization pathways involved are both caveolin and clathrin independent (214).
D. Cholesterol and Endosome Trafficking
Cholesterol is distributed irregularly in the different intracellular compartments: the ER, mitochondrial membranes, and intracellular membranes contain little cholesterol (223–224, 401), whereas the plasma membrane contains at least half of the total cellular cholesterol (17, 224, 480). Even at the intracellular level, lipids are not randomly distributed within endosomal membranes (203, 291), the recycling endosome being particularly enriched in cholesterol (158, 175). In this context, one should expect that these storage compartments will not display the same sensitivity towards cholesterol variations and that cholesterol content could well play a role in the sorting and transport of cargo proteins. Indeed, several publications linked cholesterol levels with the regulation of endosome trafficking. For example, both recycling- and late endosomes are particularly quick to reorganize following changes in cholesterol content (32, 65, 158, 229). Decreasing cholesterol levels increases endosome motility and redistributes cargo proteins from late endosomes to early endosomes (65, 264). Rab7, the small GTPase that regulates late endosome trafficking, is directly affected by changes in cholesterol levels. Increasing cholesterol content prevents the dissociation of the GTPase from the late endosome membrane and slows its dynamic properties while having no effect on Rab5, the GTPase associated with the early (sorting) endosome (229). In cardiac myocytes, cholesterol depletion accelerates the recycling of the Rab11-associated recycling endosome without affecting early endosome turnover (32).
E. Cholesterol, Exocytosis, and Recycling
As discussed in section IVA, membrane fusion is a highly regulated process requiring the participation of SNARE proteins, localized at both donor and acceptor membranes. Interestingly, as in the case of cholesterol, SNARE proteins are not distributed uniformly in endosomes or in the plasma membrane. In PC-12 cells, syntaxins are concentrated into large clusters in the plasma membrane where secretory vesicles preferentially dock, as revealed by the accumulation of and partial overlap with the synaptosomal protein SNAP25. Cholesterol depletion causes the dispersion of these clusters and a reduction of the secretion rate. Despite cholesterol sensitivity, theses clusters are not associated with caveolin-type lipid rafts (221). Nevertheless, SNAREs are to some extent associated with lipid rafts, at least as defined as detergent-resistant membranes (DRM) in several cell types. For instance, syntaxin 1A and SNAP25 are able to interact with VAMP only when associated with lipid rafts and cholesterol depletion decreases the regulated exocytosis of dopamine (60).
Ion channels and receptors constitute cargo proteins that are delivered to the plasma membrane from intracellular compartments following formation of SNARE complexes. A number of voltage-gated ion channels such as Kv4.x have been shown to be targeted to lipid raft domains together with syntaxin 1A, SNAP25 and VAMP2. Upon cholesterol depletion, Kv4.x channels and SNAREs dissociate, leading to a decrease of the A-type potassium current in pancreatic α-cells (458). A recent study has shown that TRPM8 channels are highly mobile in discrete structures residing in or just beneath the plasma membrane. Following cholesterol depletion with MβCD, TRPM8 channels accumulate instead in nonmobile clusters, and an increase in related current is seen without any change in open channel probability (440). The proposed explanation is that altered lipid composition of the plasma membrane is sufficient to stabilize TRPM8-containing vesicles that, under control conditions, continuously undergo hop-diffusion. This is reminiscent of our studies on the atrial Kv1.5 channel. Kv1.5 and its related current IKur are highly sensitive to cholesterol variations, despite a lack of association with lipid rafts in native myocytes (2, 112, 259). Depletion of membrane cholesterol in both heterologous expression systems and myocytes leads to the formation of Kv1.5 channel clusters and increases IKur density (2, 32). This current increase is generated by an accretion of active channels in the plasma membrane that resulted from an enhanced exocytosis of the recycling endosome (32) (see FIGURE 10).
VII. PATHOPHYSIOLOGY AND PHARMACOLOGY OF CHANNEL TRAFFICKING AND TARGETING
Improved knowledge of the trafficking and surface expression of ion channels has led to important breakthroughs in the understanding of pathogeneses of cardiac electrical abnormalities such as conduction blocks, sinus node dysfunction, and cardiac arrhythmias. These cardiac disorders are main causes of sudden death, stroke, and acute heart failure. They can be inherited or acquired in which case they are often associated with the development of cardiac diseases. The nature of cardiac arrhythmias is diverse, ranging from whole tissue myocardial remodeling dominated by fibrosis to single molecule defects caused by mutations in genes encoding ion channel subunits. However, in almost all cases, there is an alteration of channel function, hence the term channelopathies. Besides changes in biophysical properties, abnormal surface expression of ion channels has been recognized, too, to be a major cause of altered channel function and cardiac diseases. Genetic studies of inherited forms of congenital LQTS have revealed that several mutations are responsible for a default trafficking of LQT2-channel proteins, which are retained in the ER and degraded rather than trafficked to the cell surface. Several examples exist for LQT1, the KCNQ1 channel, and Brugada syndrome which are linked to mutations in SCN5A gene encoding Nav1.5. Trafficking-defective mutations have been described also for the inward rectifier K+ channels (93).
A. Delocalization of Connexins
Abnormal spatial distribution of connexins is a common observation during cardiopathies, heart failure, atrial fibrillation, and atrial dilation (8, 107, 168, 295, 334, 362, 389, 402, 438). In these clinical conditions, a fraction of connexins, notably connexin-43, is delocalized at the lateral membrane where they are scattered without ultrastructural evidence of junctional plaque formation. Moreover, they are not associated with N-cadherin and ZO-1, which normally regulate their aggregation in the gap junction plaque. These data suggest a defect in the targeting and/or trafficking of the channels. This is also supported by the observation that a large proportion of delocalized Cx43 are nonphosphorylated, given the role of the phosphorylation processes in the regulation of channel properties, assembly and targeting in junctional plaques (218, 404). Electron microscopic study reveals an enhanced internalization of Cx43 at the lateral membrane in the failing ventricular myocardium of dog (167). Internalized Cx43 are incorporated in multilamellar membrane structures resembling double membrane vesicles previously termed annular gap junctions and which are the result of a clathrin-dependent endocytosis of large portions of or entire gap junction plaques (117, 332). In failing ventricular myocardium, intracellular Cx43 extensively colocalizes with autophagosomes which are dynamic organelles sorting cell debris to the lysosome for degradation. Taken together, these studies suggest intense activities of trafficking, endocytosis, degradation, and probably recycling of connexins in failing ventricle. This could contribute to the plasticity of the cardiac electrical membrane properties in the context of myocyte hypertrophy and reshaping of the myocardium.
B. Ankyrin Mutations
Among the membrane-associated proteins recently shown to be involved in channelopathies, ankyrin proteins play a prominent role. The roles of ankyrin proteins in cardiac cells have been presented in section IV. ANK2, the gene encoding ankyrin-B, has been found to be polymorphic (280). The first human mutation in this gene to be described (282) was found in a large French family with congenital LQTS (LQT4) associated with marked sinus bradycardia. Thereafter, many other mutations of ANK2 were found in patients with sinus node dysfunction, sinus bradycardia, polymorphic ventricular fibrillation, and sudden death (280; reviewed in Ref. 161). All of these mutations were found to cause loss-of-function in ankyrin-B (280) by disrupting the regulation of important interacting membrane transporters, i.e., the sodium-calcium exchanger, the Na+-K+-ATPase, and/or the IP3 receptor. This wide spectrum of pathological phenotypes led to the coining of the term ankyrin-B syndrome. In recent work from the Mohler group (85), ankyrin-B has also been shown to be important in the pathogenesis of atrial fibrillation. A loss-of-function mutation in its gene was shown to be linked to atrial fibrillation, and a clear decrease in ankyrin-B expression was observed in atrial tissue from atrial fibrillation patients. The involvement of the other ankyrin gene in cardiac disease seems, so far, to be less important, aside from the interesting observation that a mutation in the domain of Nav1.5 interacting with ankyrin-G may lead to Brugada syndrome (281).
C. Endocytosis Defects
Defects in the endocytosis of ion channels, too, can be responsible for abnormal surface expression of channels. In a familial form of cardiac conduction block, a mutation in the transient receptor potential cation channel TRPM4 is responsible for a constitutive SUMOylation of the protein. The resulting deSUMOylation defect prevents TRPM4 endocytosis and leads to abnormal accumulation of the channel in the plasma membrane of the Purkinje cells where these channels are abundantly expressed (212). The electrophysiological consequence is a “gain-of-function” of TRPM4 and the increase in the corresponding inward current, causing membrane depolarization of Purkinje fibers and, in turn, conduction blocks (212).
In the same issue of the journal, Guo et al. (153) described another mechanism of abnormal channel endocytosis in the context of cardiac arrhythmias. Hypokalemia has been known for many years to be a significant risk factor for often-fatal ventricular arrhythmias such as torsades de pointes. It can result from enhanced C-type inactivation of hERG or reduced density of channels under conditions of low extracellular potassium (462). Guo et al. (153) showed that low external potassium increases hERG channel internalization resulting in the suppression of the corresponding current together with a >80% reduction in channel expressed at the cell surface. Internalized hERG trafficked via Rab5-positive early endosomes and was quickly targeted to the lysosomes. This is a clathrin-independent endocytosis process that requires the ubiquitylation of the protein (153). Hypokalemia in rabbits (mean serum [K+] <2.4 mM) displayed QT-interval prolongation and reduction in the hERG expression. Extending the period over which rabbits were maintained on a very-low-K+ diet causes a reduction not only in IKr but also in IKs, a current which is underlain by KCNQ1/KCNE1 (154).
Yet, another indication of the crucial role of channel endocytosis in cardiac excitability has been provided by Schumacher et al. (384) investigating the action of quinidine, a well-known blocker of Kv1.5 (118). The authors found that the drug promoted the rapid internalization of that channel but not of Kv4.2 or Kv2.1, channels also blocked by quinidine. Quinidine, but not its diastereomer quinine, caused a rapid internalization of a GFP-tagged Kv1.5 expressed in HL-1 cardiomyoblasts or mouse neonatal myocytes. This process is both dynamin- and dynein-dependent (72, 267). Interestingly, the amino acid residues known to be involved in quinidine block of Kv1.5 only partially overlap with those essential for the internalization response. Perhaps with better understanding of the structural requirements for induced Kv1.5 internalization, more specific drugs can be designed that will promote this endocytosis without affecting currents carried by other channels.
The results from most research efforts, to date, in the field of trafficking and targeting of cardiac ion channels have been essentially the identification and description of numerous molecular actors involved in these processes. It is certain that this list is not exhaustive and that many more partners are likely to be discovered. What appears to be the most challenging task in the field now is to understand how these actors work together to fine tune the functional expression of ion channels and control cardiac excitability. Analogously, in the central nervous system, mechanisms regulating the recruitment of NMDAR and Kv channels are central to long-term potentiation (LTP) (49, 150, 200, 246, 397, 403).
There are important limitations to our current knowledge that complicate the integration of these molecular actors into a physiological model of cardiac excitability. First, most of our current knowledge on the trafficking and targeting of ion channels comes from noncardiac expression systems, i.e., heterologous expression in cell lines or from systems such as brain and kidney. Many more studies conducted on cardiac myocytes are clearly needed. Second, experimental models allowing the study of a single molecular actor in the trafficking and targeting of ion channels in context of the global electrical activity of the heart are completely lacking. In this regard, the identification of genetic mutations leading to trafficking defects constitute precious sources of information as illustrated, for example, by the roles of ankyrin-B and TRPM4 channel mutations in complex electrical syndromes in humans (282). New approaches aiming to either silence or overexpress genes of interest in the whole heart will be crucial to deciphering the roles played by the molecular actors involved in trafficking and targeting of ion channels and, therefore, in the regulation of the global cardiac electrophysiology. In this context, the potential use of adeno-associated viral (AAV) vectors to specifically target the heart opens new possibilities for research into cardiac electrophysiology (176). The third limitation concerns the physiological tools that are necessary and/or available to visualize the trafficking and the targeting of channels in their native environment. One solution could be the use organo-culture or explant models. Indeed, such culture explant systems are widely used in neurosciences to study the development of normal or pathological tissues (352), to assess the effects of drugs as well as cell or gene therapies (for example, Refs. 52, 181, 286) and to measure strength changes between synapses during LTP or LDP, for example. The preservation of the three-dimensional organization of the myocardium as a multicellular network for at least a few days combined with the use of AAV vectors or adenoviruses would allow the study of the trafficking and the targeting of fluorescently tagged ion channels. With the use of either biphoton microscopy or TIRF microscopy, one might follow ion channel trafficking in a system very close to the in situ situation.
Despite the urgent need for more knowledge of the trafficking and targeting of cardiac ion channels in the heart, this review emphasizes that these processes could be the major source of plasticity in the myocardium, tuning its cardiac electrical activity and contributing to its adaptation to changes in cardiac working conditions. An important perspective of future research will be to determine the roles of trafficking and targeting processes in the remodeling of the electrical properties of diseased myocardium characterized by turnover of channels and profound changes in the ultrastructure of the myocardium.
This work was supported by Fondation Leducq “Structural Alterations in the Myocardium and the Substrate for Cardiac Fibrillation” (to E. Balse, A. Coulombe, and S. Hatem) and the European Union (EUTRAF; to E. Balse, A. Coulombe, and S. Hatem). The group of H. Abriel is supported by Swiss National Science Foundation Grant 310030B_135693 and the University of Bern. DF group was funded by Canadian Institutes for Health Research and by the Heart and Stroke Foundation of British Columbia and Yukon.
No conflicts of interest, financial or otherwise, are declared by the authors.
We are indebted to Jodene Eldstrom for careful reading of the manuscript and scientific advice. We express our thanks to Christine Longin and Sophie Chat (MIMA2 platform, INRA Jouy-en-Josas, France) and Catherine Rücker-Martin (INSERM UMR-S 999, Centre Chirurgical Marie Lannelongue, Le Plessis Robinson) for expert assistance with the electron microscopy imaging.
Address for reprint requests and other correspondence: S. Hatem, UMRS 956, Faculté de Médecine Pierre et Marie Curie, Paris-6, 91 boulevard de l'Hôpital, 75013 Paris, France (e-mail:).