Voltage-Gated Calcium Channels: Key Players in Sensory Coding in the Retina and the Inner Ear

Abstract

Calcium influx through voltage-gated Ca (Ca_V) channels is the first step in synaptic transmission. This review concerns Ca_V channels at ribbon synapses in primary sense organs and their specialization for efficient coding of stimuli in the physical environment. Specifically, we describe molecular, biochemical, and biophysical properties of the Ca_V channels in sensory receptor cells of the retina, cochlea, and vestibular apparatus, and we consider how such properties might change over the course of development and contribute to synaptic plasticity. We pay particular attention to factors affecting the spatial arrangement of Ca_V channels at presynaptic, ribbon-type active zones, because the spatial relationship between Ca_V channels and release sites has been shown to affect synapse function critically in a number of systems. Finally, we review identified synaptopathies affecting sensory systems and arising from dysfunction of L-type, Ca_V1.3, and Ca_V1.4 channels or their protein modulatory elements.

I. INTRODUCTION

It has been recognized for some time that exocytosis underlying neurotransmission depends on Ca²⁺ influx through voltage-gated Ca (Ca_V) channels (87, 175, 217, 282). More recently, though, it has become apparent that Ca_V channels also serve as signaling molecules at presynaptic active zones (AZs) (92, 122, 219). Here, we consider the multiple roles played by Ca_V channels at ribbon synapses (FIGURE 1), the specialized synapses found in auditory and vestibular hair cells, including those in the lateral line organ of teleost fish, in retinal photoreceptors, and bipolar cells.

FIGURE 1.

Presynaptic voltage-gated Ca channels of sensory ribbon synapses. Synaptic transmission at ribbon-type active zones (AZs) is driven by Ca²⁺ influx through voltage-gated L-type Ca channels, with Ca_V1.3 being the predominant Ca_V in the receptor cells of the inner ear [cochlear inner and outer hair cells (IHCs and OHCs; B) and vestibular hair cells (VHCs; A)], and Ca_V1.4 in synaptic terminals of photoreceptors (C) of the retina. In mammalian and non-mammalian VHCs, additionally Ca_V3.1 and Ca_V2 channels were detected, respectively (A). The identity of Ca_V channels in bipolar cells is unclear (Ca_V1.X; D), but according to various data all Ca_V1 isoforms may be present. Ca_V channels are located in close vicinity to the ribbon, a large synaptic vesicle (SV)-tethering presynaptic structure common to synapses in all depicted cell types.

All of these cells signal in an analog fashion, via graded rather than action potentials; ribbon synapses therefore support continuous exocytosis, generally at high rates (22, 48, 111, 121, 123, 143, 176, 178, 331). Thus ribbon synapses might be designed to speed the delivery of vesicles to release sites (113, 274), although there is also evidence that the ribbon slows vesicle delivery to prevent depletion (159) or unneeded release in inactive retinal synapses (385). Furthermore, ribbons may support molecular priming (335), promote synchronous exocytosis of several vesicles (129), and, perhaps, enable homotypic fusion of vesicles before exocytosis (274, 341).

The ribbon is an organelle composed largely of the vertebrate-specific protein RIBEYE, which was identified and purified first from bovine retinal synaptic ribbons (313). RIBEYE comprises a unique A domain that forms the core of the ribbon by mediating self-aggregation (222, 313, 422) and a B domain that is almost identical to COOH-terminal binding protein 2 (CtBP2; Ref. 313). The B domain also strongly resembles NADH dehydrogenase and is thus likely enzymatically active (319). Proteins interacting with the RIBEYE scaffold are largely conserved between sensory cell types [e.g., piccolino (292), bassoon (84, 178), RIM (RAB3-interacting molecule; Ref. 170), and RIM-binding protein (RIM-BP; Ref. 195) have been found at all ribbon AZs]. Given the unique function of each type of ribbon synapse, the full complement of proteins involved in the vesicle cycle varies between cell types, as we will discuss here.

Exocytosis at all ribbon synapses, though, results from influx of Ca²⁺ through Ca_V1-containing Ca channels, which mediate L-type Ca currents (Refs. 15, 25, 95, 225, 280, 349; FIGURE 1). At conventional synapses, in contrast, the presynaptic Ca_V channels are largely of the Ca_V2 group (e.g., reviewed in Ref. 55). Ca_V1 channels are arranged in clusters in the presynaptic density beneath the synaptic ribbon and, depending on the cell type, can be coupled tightly to docked vesicles; herein, we discuss how the physical proximity of Ca_V channels and vesicles results in so-called “nanodomain” control of exocytosis by highly localized changes in intracellular [Ca²⁺] (22, 43, 160, 367). Several regulatory molecules that modulate the kinetics of Ca_V channel gating and (in)activation have been found to influence synaptic transmitter release, and mutations affecting their function may have deleterious consequences for hearing, balance, or vision. In this review, we discuss structural and functional specializations of Ca_V channel clusters mediating transmitter release at ribbon synapses of sensory receptor cells with the goal of articulating how the first step in synaptic transmission at ribbon synapses permits sensory information to be encoded reliably.

II. PROPERTIES OF VOLTAGE-GATED CALCIUM CHANNELS INVOLVED IN SENSORY CODING AT RIBBON SYNAPSES

A. Ca_V Channels in Hearing, Balance, and Vision

Ca_V channels are macromolecular complexes comprising a number of subunit proteins. The core Ca²⁺-selective channel pore is formed by 1 of 10 α₁-subunit isoforms, and accessory subunits contributing to the complex include 1 of 4 different β-subunit isoforms and 1 of 4 different α₂δ subunits (FIGURE 2; for review see Refs. 47, 94). The α₁ subunit determines the majority of a channel's biophysical and pharmacological properties; accessory subunits modify these properties and in addition influence the abundance and localization of Ca_V channels, as discussed in sections II and III of this review. Ca_V channels are grouped according to the characteristics of the currents they mediate: L-type currents arise from channels in the Ca_V1 family, Ca_V2 family channels give rise to P/Q-, N-, and R-type currents, and T-type currents are mediated by Ca_V3 family channels.

FIGURE 2.

Major known Ca_V1 channel interaction partners at the ribbon synapses of the ear and the eye. The pore-forming Ca_Vα₁ subunit composed of four homologous domains (I–IV), each containing six transmembrane segments in yellow with the voltage sensor marked in green. Key interaction partners are depicted with their Ca_V-related roles indicated in parentheses. Both Ca_V1.3 and Ca_V1.4 COOH terminus carry a COOH-terminal modulatory domain (CTM) with interacting proximal and distal COOH-terminal regulatory domains (PCRD and DCRD). The EF-hand motif together with the preIQ and the IQ motif comprise the CDI machinery. Protein kinase A (PKA) phosphorylation of the Ca_V1.4 COOH terminus is depicted by the red dot. GPI, glycosylphosphatidylinositol anchor of the α₂δ subunit; VWA, von Willebrand factor-A domain; AID, alpha interaction domain in the I–II linker region that binds Ca_Vβ subunit; SH3, HOOK, and GK, src-homology-3, HOOK-, guanylate kinase-like-domain of the β subunit; NSCaTE, NH₂-terminal spatial Ca²⁺ transforming element; RIM, RAB3-interacting molecule; C₂B and PxxP, C₂B domain and proline-rich region of RIM; RIM-BP, RIM-binding protein; CaBPs, Ca²⁺-binding proteins; CaM, calmodulin; CB, CR, and PVα, mobile intracellular Ca²⁺ buffers calbindin, calretinin, and parvalbumin α, respectively; GIPC3, GAIP-interacting protein, COOH terminus 3; Bsn, bassoon; I_Ca, whole cell Ca²⁺ current; i, single channel current; P_o, open probability; VDI and CDI, voltage- and Ca²⁺-dependent inactivation.

Here, our focus is on Ca_V1 channels (for review see Ref. 419), which generate the L-type currents and are well suited to mediate neurotransmission evoked by graded potential changes at ribbon synapses. In particular, the gating properties of these channels at ribbon synapses permit rapid responses to fluctuations in membrane voltage elicited by changes in sensory stimulus intensity. The pharmacology of Ca_V1 channels is well-established, and many clinically relevant compounds, e.g., dihydropyridines (DHPs), are Ca_V1 modulators.

There are four members of the Ca_V1 family: Ca_V1.1, Ca_V1.2, Ca_V1.3, and Ca_V1.4. Both Ca_V1.3 and Ca_V1.4 channels activate at relatively hyperpolarized potentials, within the range of graded potential changes in receptor cells, and they inactivate slowly (23, 190, 191, 408). Thus they are highly appropriate for mediating tonic and/or periodic Ca²⁺ influx necessary for transmission at ribbon synapses.

Ca_V1.3 channels are expressed most abundantly in inner and outer hair cells of the cochlear (IHCs and OHCs, respectively) and in vestibular hair cells (VHCs; FIGURE 1), but they also are present in several other tissues (IUPHAR target id 530; for review see Ref. 419). The unusual electrophysiological L-type current properties [negative activation threshold, fast activation, and depending on the splice variant (sect. IIB), relatively slow inactivation of Ba and Ca currents], were first described in chick basilar hair cells (424) and later in mouse IHCs (280) and in nonmammalian auditory hair cells (e.g., see Refs. 112, 189, 314). The pharmacological and physiological properties of L-type Ca currents recorded in OHCs are consistent with these channels containing Ca_V1.3α₁ subunits, too (184). Interestingly, Ca_V1.3 channels expressed in heterologous expression systems or several tissues outside the inner ear (e.g., in the sinoatrial node) show more pronounced inactivation than do channels in hair cells (190, 224, 280, 408), indicating tissue-specific channel specializations and motivating study of alternative splicing, auxiliary subunits, and protein binding partners (see sect. IIB).

While the vast majority of Ca_V current in IHCs is mediated by Ca_V1.3 channels (∼90%, Refs. 95, 280), Ca_V1.3 channels contribute only ∼50% of the total depolarization-evoked Ca current in VHCs in cristae ampullaris of semicircular canals and maculae of otolithic gravity organs (sacculus and utriculus) (20, 95). This might explain why Ca_V1.3 knockout (KO) animals are deaf but do not suffer from a significant balance impairment (95). The residual whole-cell Ca current in these mice implies that hair cells in both sensory systems may express multiple Ca_V channels (280). The presence of Ca_V3.1 channels with somewhat atypical biophysical properties was reported in the mouse inner ear and chicken basilar papilla (202, 203, 260). Due to their low-voltage activation and usually very rapid and strong inactivation, Ca_V3-mediated Ca²⁺ influx is responsible for spontaneous activity in neurons and pacemaker cells (for review see Ref. 419). The transient presence of the Ca_V3.1-mediated Ca current during early hair cell development and/or upon ototoxic drug exposure also suggested its requirement for hair cell maturation and regeneration (202, 203). It has to be noted, however, that other Ca_V channels also may contribute to maintaining the vestibular synaptic function in Ca_V1.3 KO animals. Notably, nimodipine-insensitive Ca currents in the hair cells of lower vertebrates were suggested to be mediated via Ca_V2 channels (230, 304, 350).

Single-channel recordings helped determine the identity of Ca_V channels in hair cells and showed that Ca_V1.3 channels display very rapid voltage-dependent activation and deactivation (within <1 ms; e.g., see Refs. 303–305, 416–418), which enables short delays in synaptic transmission. Together with fluctuation analysis of whole-cell Ca_V currents (e.g., Refs. 43, 111, 129, 382), they gave insights into the elementary biophysical properties of Ca_V1.3 and, in the retina, Ca_V1.4 (366) channels. The reported values vary greatly, depending largely on the experimental conditions (e.g., single-channel conductances for Ca_V1.3 between 3.5 and 16 pS in hair cells of different species and channels expressed in heterologous expression system; Refs. 39, 129, 304, 416), hampering comparisons among studies. Combining data from recordings in mouse apical IHCs (43, 382, 418) suggests the Ca_V1.3 single-channel current of approximately -0.14 pA (assuming 1.3 mM extracellular [Ca²⁺]) and a maximal open probability of 0.2–0.4 (in the absence of BAY K 8644). For Ca_V1.4, a large discrepancy in the single-channel conductance (3.7 vs. 22 pS) was observed despite similar recording conditions in two studies (i.e., 100 vs. 82 mM Ba²⁺ as the charge carrier; Refs. 90, 366), which is worthy of further investigation.

Slow inactivation is most pronounced in Ca_V1.4 channels (23, 191, 236), which are expressed predominantly in the retina, particularly at photoreceptor terminals (FIGURE 1) where they mediate the sustained Ca²⁺ entry needed for continuous release of neurotransmitters in dark (333, 386). The identity of L-type Ca channels in retina generally and in bipolar cells specifically is controversial. Immunohistochemistry and in situ hybridization suggest the presence of all Ca_V1 subunits at retinal ribbon synapses (177, 250, 360, 406). Additionally, Ca currents in rod and cone photoreceptors are modulated differently by a variety of neuromodulators, suggesting diversity in Ca_V channel composition and function (5, 193, 339, 343), but detailed assessment of channel composition is lacking (33, 50, 251, 299).

B. Regulation of Ca_V1.3 and Ca_V1.4 Properties

1. Alternative splicing

A key mechanism for regulating the functional properties of Ca_V1 channels is alternative splicing. Among L-type Ca channels, alternative splicing in the COOH terminus of the Ca_V1.3α₁ subunit is best understood (39, 332, 358). Similar to Ca_V1.2 channels, full-length Ca_V1.3 and Ca_V1.4 channels carry a proximal and a distal COOH-terminal regulatory domain (Refs. 156, 332, 333; FIGURE 2), both of which are putative α-helices that form a noncovalent interaction. This so-called COOH-terminal modulatory domain (CTM) competes with calmodulin (CaM) in binding to the channels’ IQ-domain (216) and thus weakens CaM-mediated Ca²⁺-dependent inactivation (CDI). It also shifts the voltage dependence of channel activation to more positive voltages and reduces channel open probability (39, 211, 332, 333).

Alternative splicing can affect the presence of one or both components of the CTM or elements in proximity and thereby change channels’ biophysical properties. Alternative splicing of the Ca_V1.3 COOH terminus generates ‟longˮ (Ca_V1.3_L also termed Ca_V1.3₄₂) and ‟shortˮ channels with different functions in different tissues (332, 358). In cochlear IHCs, containing long and short Ca_V1.3 splice variants, alternative α₁-subunit splicing may contribute to large functional heterogeneity of the channel clusters (Refs. 312, 326, 381, see also Ref. 188 for hair cells from chicken basilar papilla, but see Ref. 262). OHCs preferentially express Ca_V1.3 channels lacking the IQ motif (Ca_V1.3_IQ∆), which likely explains the almost complete absence of CDI in these cells (326), since the IQ motif acts as the interaction site for CaM (see below).

In human retina, at least 19 splice variants of Ca_V1.4α₁ have been identified (359). Notably, alternative splicing in the IVS3-S4 linker, which may influence S4 voltage sensor function because of its close proximity to the S4 segments (see FIGURE 2 for channel structure), is common in Ca_V1 channels and likely supports their functional diversity (171). For example, the common Ca_V1.4_Δex32, in which skipping of exon 32 shortens the IVS3-S4 linker, shows altered channel activation (359). Even more abundant is Ca_V1.4_ex16a-, a splice variant with alternative splicing of exon 16. It results in a frameshift and consequently a premature stop codon at IIS6 domain, which was predicted to lead to the formation of a hemi-channel (359). If paired with an alternative splicing at the following exon (17a+₁), which in itself also causes a frameshift, the reading frame should be restored and functional channels generated (359). It is unknown whether the splicing of Ca_V1.4 is cell-type specific, developmentally regulated, and whether it is affected under pathological conditions.

2. Ca_V channel interacting partners

In numerous ways, Ca_V1 behaviors are modified via protein-protein interactions (52). Here, we review interactions between Ca_V1 and the Ca²⁺-binding proteins CaM and CaBP in sensory receptor cells.

a) calmodulin.

CaM is an important Ca²⁺ sensor that powerfully regulates Ca_V channels. The structural and functional highlights of this ‟calmodulationˮ have been reviewed recently (31, 82). Briefly, Ca²⁺ entering through L-type Ca channels binds to apoCaM preassociated with an interface that includes the COOH-terminal IQ domain (Refs. 29, 99, 108, 179, 356; FIGURE 2). Ca²⁺-bound CaM then undergoes conformational changes that promote CDI. Ca²⁺ binding to either lobe of CaM modulates channel function, but the C- and N-lobes show different affinities for Ca²⁺ (6 times higher for the CaM C- as compared with N-lobe; Ref. 100). Consequently, the two CaM lobes may display different ‟spatial Ca²⁺ selectivity.ˮ The C-lobe binds Ca²⁺ entering through the associated channel and releases it very slowly (100), thus displaying ‟local Ca²⁺ selectivityˮ (356). For the N-lobe, which both binds and releases Ca²⁺ rapidly (k_on of ∼10⁹ to 10¹⁰ M⁻¹s⁻¹, k_off of ∼10⁴ to 10⁵ s⁻¹; Ref. 100), it has been suggested that brief changes in [Ca²⁺] cannot shift CaM from the apo state. Therefore, the N-lobe is thought to display a ‟global Ca²⁺ selectivityˮ for the relatively low changes in [Ca²⁺] that arise from Ca²⁺ entry at relatively distant sites (83, 356). Intriguingly, however, the N-lobe was reported to switch its Ca²⁺ selectivity depending on the presence of the short linear motif, NH₂-terminal spatial Ca²⁺ transforming element (NSCaTE) in the Ca_V channel (83, 356). NSCaTE, which is only found on the NH₂ terminus of Ca_V1.2 and 1.3 channels (357), binds N-lobe of Ca²⁺-CaM and endows it with local Ca²⁺ selectivity. This may be relevant for establishing Ca²⁺ nanodomains (see sect. III).

CaM also serves as the Ca²⁺ sensor for Ca²⁺-dependent facilitation (CDF), but the clear presence of CDF in hair cells has not been demonstrated. There is, however, evidence for paired pulse facilitation of cone Ca currents (379), and mechanisms of voltage-dependent facilitation (VDF) have been studied at photoreceptor synapses (192).

b) calcium binding proteins.

CaM-dependent CDI in L-type Ca channels can be antagonized by calcium binding proteins (CaBPs), which are EF-hand Ca²⁺-binding proteins with ∼50% homology to CaM (for review, see Ref. 141). They consist of two independent lobes, each containing two EF-hand Ca²⁺-binding domains, although in CaBPs only two or three are functional (137, 138, 249, 268). Unlike CaM, CaBPs are found primarily in central neurons (CaBP1/caldendrin; Ref. 180), retina (CaBP1, CaBP2, CaBP4, CaBP5; Refs. 136, 138, 299, 334), and the inner ear (CaBP1, CaBP2, CaBP4, CaBP5; Refs. 73, 411).

Although a noncompetitive action of CaBP1 on CaM-dependent CDI has been proposed (265), most models of Ca_V channel modulation involve a competition between CaBPs and CaM. It has been hypothesized that Ca²⁺-bound CaBP displaces CaM from the COOH-terminal IQ motif of the α₁ subunit (73, 273, 409, 410). For CaBP1, such competitive behavior is well supported by a number of biochemical studies on isolated proteins (106, 107, 399). Despite this evidence, it is unclear how Ca²⁺-bound CaBPs could displace CaM because Ca_V channels exhibit a greater affinity for the latter.

This open question is most relevant in the brain, where CaM is far more abundantly expressed than CaBPs (410); the relative concentrations of these proteins in photoreceptors and hair cells are not yet known. One hypothesis is that the competition should occur between the apo-states of the sensors: in the Ca²⁺-free condition, CaBPs have an advantage over CaM in binding to the Ca_V IQ domain (107). An alternative, allosteric scheme in which CaM and CaBP4 together bind to Ca_V1.3 channels also has been proposed (410). Either of the latter mechanisms, though, could permit low concentrations of CaBPs to modulate Ca_V channels effectively in the presence of higher CaM concentrations. For further discussion on this topic, readers are referred to other recent reviews (e.g., Refs. 30, 31, 105).

c) calcium binding proteins affect sensory neuron function.

In cochlear IHCs, regulation of Ca_V1.3 channels by CaBPs may be important for synaptic transmission (411). Of the multiple CaBPs detected in IHCs, CaBP2 so far is the only one whose disruption is known to significantly impair hearing in humans (OMIM 607314; Refs. 277, 318; see sect. VI). The lack of CaBP4 affects hair cell physiology only mildly; along with unperturbed hearing in the mouse KO model, this indicates that CaBP4 is only a very modest suppressor of Ca_V1.3 CDI in IHCs (73).

Milder changes in Ca_V1.3 CDI can also be observed upon deletion of EF-hand mobile Ca²⁺ buffers calbindin, calretinin, and parvalbumin-α (270). The mechanisms by which these proteins affect Ca_V1.3 channels remain elusive, especially as enhanced CDI in mutant IHCs lacking these three buffers could not be reversed upon addition of synthetic Ca²⁺ chelators. For at least one of Ca²⁺ buffer proteins, calretinin, a direct interaction with Ca_V2.1 (but not 1.2) channels has been demonstrated (69), opening the door to future studies.

In the retina, CaBP4 in photoreceptor synaptic terminals interacts with Ca_V1.4 to regulate Ca²⁺ influx (136). In addition to reducing channel inactivation, CaBP4 shifts the voltage dependence of Ca_V1.4 activation to more hyperpolarized potentials and thus increases channel sensitivity to graded membrane potential changes. The effect of CaBP4 on Ca_V channel activation (but not inactivation) is only observed in the channels containing the CTM (322). Thus it was proposed that CaBP4 forms a collapsed structure around the Ca_V1.4 IQ motif to disrupt interaction between the IQ domain and the CTM (273). For example, human splice variants Ca_V1.4_ex43* and Ca_V1.4_Δex47 (lacking all or part of the CTM, respectively) exhibit robust CDI and a hyperpolarized activation range in a heterologous expression system (139, 359). The strong CDI (but not a shift in activation range) in Ca_V1.4_Δex47, however, is suppressed in the presence of CaBP4 (139). In photoreceptors, splice variants with the COOH-terminal truncation therefore might show little CDI, similar to full-length channels containing the CTM. Also, the activation range of full-length and truncated Ca_V1.4_Δex47 channels in photoreceptors may be similar due to the counteracting effects of CTM and CaBP4 on the activation range of the full-length channels. This way, both channels with or without CTM would support sustained neurotransmission from photoreceptors in darkness (139). Mutations in CaBP4 (OMIM 608965) cause autosomal recessive congenital stationary night blindness type 2 (CSNB2; also known as incomplete CSNB or iCSNB; Refs. 420, 421; see also sect. VI). This might be explained by the inability of mutant CaBP4 to regulate Ca_V1.4 channel activation and inactivation effectively (322).

CaBP5 is expressed in some types of retinal bipolar cells, in which it acts to modulate Ca_V channels, likely Ca_V1.2 (299). Although CaBP5 KO mice display normal electroretinograms (ERGs) and unperturbed gross retinal morphology, the absolute sensitivity of their rod-mediated ganglion cell light responses is reduced; this indicates that CaBP5 acts rather downstream of photoreceptor synapses (299). CaBP5 also interacts with Munc18–1 and myosin VI; as both are involved in synaptic vesicle cycling and trafficking, it may have additional roles in neurotransmitter release (336). CaBP1 and 2 also modulate synaptic transmission in the inner retina, as ganglion cell responses to light are altered in mice lacking these proteins (334).

In summary, hearing seems to mostly rely on the presence of CaBP2, whereas for vision CaBP4 might be the most relevant. Ca_V channel function in the sensory organs is likely also modulated by other CaBP members, which however may not be as obvious due to potential functional redundancy among certain protein subfamily members and consequently compensation in the respective single CaBP KO animals.

The function of Ca_V channels can further be modulated by phosphorylation via various protein kinases. Ca_V1.4 channels contain a protein kinase A (PKA) phosphorylation site in the distal CTM motif that is involved in the inhibition of the CDI (FIGURE 2). Phosphorylation of this region promotes the occupancy of CaM thus increasing channel open probability and CDI of Ca_V1.4 full-length channels (311). Indeed, virtually every known neuromodulator affects Ca currents in photoreceptors in various systems (e.g., dopamine, cannabinoids, somatostatin, retinoids; Refs. 5, 162, 339, 343, 380), raising the question of how transmission is maintained over long periods when Ca currents can be affected in so many ways. It has been suggested that PKA-mediated modulation of Ca_V1.4 channels may contribute to setting the visual sensitivity during the day-night cycle as the PKA activity in retina may also follow circadian rhythm (311). In contrast, no significance of Ca_V1.3 channel phosphorylation for the function of hair cells has been described yet. Inhibition of PKA activity was shown to have minor effects on Ca_V1.3-mediated Ca currents in IHCs (397).

3. Modulation of Ca currents by protons released during neurotransmission

Upon synaptic activity, a transient reduction in Ca current is observed at some hair cell, retinal, and electroreceptor ribbon synapses (66, 67, 79, 154, 160, 269, 309, 352, 396). This reduction reflects H⁺-dependent inhibition of presynaptic Ca_V channels by the acidic contents of exocytosed vesicles (or by the action of Na⁺-H⁺ exchanger, as suggested for the horizontal cell-to-photoreceptor feedback; Ref. 396). Binding of protons to Ca_V channels partially blocks the channel pore (64, 281) and shields membrane-bound charges, thereby reducing the Ca_V channel conductance and shifting activation to more positive potentials (183, 194, 281).

Proton-mediated modulation of Ca_V channels, by limiting excessive Ca²⁺ influx and subsequent exocytosis, may serve as a mechanism of activity-dependent synaptic plasticity, as suggested for auditory hair cells (66). At the calyceal synapses of the type I VHCs, the released protons may additionally act as nonquantal neurotransmitters (148). Interestingly, H⁺-mediated block of Ca_V channels in mammalian IHCs has not been described; this could reflect differences in experimental conditions, the abundance of carbonic anhydrase surrounding hair cells that may oppose excessive acidification of the synaptic cleft, rapid diffusion, rapid clearance of protons away from tiny postsynaptic boutons (66), and/or that IHC synaptic vesicles may be less acidic than others (257).

In the retina, liberation of protons during synaptic transmission could act to match Ca current activation threshold to the photoreceptor resting membrane potential and to set a rate of steady-state exocytosis (21, 79). Acidification of the synaptic cleft by synaptic activity also may be the mechanism of lateral inhibition of rods and cones by horizontal cells (reviewed in Ref. 149). In this scenario, a light stimulus hyperpolarizes horizontal cells, which could drive the influx of protons into the horizontal cells. Alkalinization of synaptic cleft, observed at cone synaptic terminals using a genetically encoded pH indicator (392), would consequently shift the voltage dependence of cone Ca_V channels towards hyperpolarized potentials, evoking more Ca current at a given membrane potential. Similar mechanisms were suggested for horizontal cell-to-rod feedback (11, 361). In the “reverse operation” of photoreceptors (i.e., they depolarize in darkness and hyperpolarize in light), this would lead to inhibition (that is, affected photoreceptors would respond as if it were darker; Ref. 149). Ca currents also are modulated by other ions in the extracellular environment. These include Zn²⁺ (7, 403), Cl⁻ (362, 364–366), and K⁺ (407). Such modulation might be particularly significant at invaginating photoreceptor synapses, at which diffusion might be restricted and the concentrations of ions might change significantly.

III. CALCIUM CHANNEL SPATIAL ORGANIZATION AS IT RELATES TO EXOCYTOSIS

A. [Ca²⁺] Microdomain Versus [Ca²⁺] Nanodomain Control of Exocytosis

The abundance and distribution of Ca_V channels determines the kinetics and efficiency of exocytosis, the metabolic cost of synaptic activity, and the precision of postsynaptic spike timing. At some synapses the opening of one or at most a few Ca_V channels triggers the fusion of a nearby synaptic vesicle; such synapses include vestibular and low- to medium-frequency-tuned auditory hair cell ribbon synapses (43, 123, 129, 169, 176), retinal photoreceptor and bipolar cell ribbon synapses (22, 160, 367), as well as some conventional central nervous system (CNS) synapses (reviewed in Ref. 98). In such instances, exocytosis is said to be under control of a “[Ca²⁺] nanodomain,” a term usually employed when the distance between a Ca_V channel and a vesicular fusion sensor amounts to less than 100 nm and the opening of a single channel generates a [Ca²⁺] sufficient to activate the vesicular sensor. In such a scenario, exocytosis of the readily releasable pool of vesicles (RRP) is linearly dependent on the number of open Ca_V channels. This is in contrast to “[Ca²⁺] microdomain” control of exocytosis, in which Ca_V channels are >100 nm from vesicles and Ca²⁺ influx pooled from multiple channels is necessary to elevate [Ca²⁺] sufficiently to elicit exocytosis (42; reviewed in Ref. 98). In the case of “microdomain” control, exocytosis shows high cooperativity and highly nonlinear dependence on the number of open channels.

Microdomain control of exocytosis reduces jitter and noise introduced by stochastic channel openings and therefore enhances the “signal-to-noise” ratio of AP-dependent signaling (232). As well, by permitting multiple channels to shape [Ca²⁺] at release sites, microdomain control has the potential to provide a mechanism by which Ca_V channel modulation regulates Ca²⁺-dependent short-term plasticity of transmitter release (317). Nanodomain control of exocytosis, on the other hand, provides increased speed and efficacy of synaptic transmission and greater sensitivity and dynamic range of the response (reviewed in Refs. 98, 232). A [Ca²⁺] nanodomain might further be advantageous for synchronizing multivesicular release (129, 160, but see Refs. 238, 252, 331). Therefore, it is ideal for encoding graded sensory signals with a large dynamic range in the retina and cochlea.

B. Topology of Ca_V Channels and Synaptic Vesicles at the (Inner) Hair Cell Ribbon-Containing AZ

The spatio-temporal profile of an intracellular [Ca²⁺] increase around a single Ca_V channel depends on its conductance, the concentration and binding kinetics of the intracellular Ca²⁺ buffers, and the potential spatial constraints (e.g., diffusion barriers; Ref. 129) (232, 254, 317). As the intrinsic Ca²⁺ dependence of exocytosis is highly nonlinear, with Ca²⁺ sensors of exocytosis requiring a high degree of cooperativity of bound Ca²⁺ to trigger transmitter release (35, 40, 41, 143, 145, 316, 402), the release of a synaptic vesicle strongly depends on its distance to the Ca_V channel.

Direct anatomical study of the topology of Ca_V channels and synaptic vesicles is difficult and initial estimates of distances between channels and release sites often come from analysis of the effects of the slow Ca²⁺ chelator, EGTA, on exocytosis (255). Consistent with nanodomain coupling, EGTA has small effects on transmission from auditory hair cells in a number of species [mice (34, 169), frogs (169), turtles (315)]. In the apical IHCs after the onset of hearing, the estimated average distance between a Ca_V channel and a release site is 15–17 nm (270; in line with Refs. 43, 123, 402), which is essentially half of a vesicle diameter and thus likely the limit to how close a vesicle can be placed next to a Ca_V channel.

This estimate is supported by a number of imaging studies. Serial section electron micrographic (EM) analysis of the mature apical IHC synapse demonstrates presynaptic densities occupied largely by the presynaptic ribbon (∼97%) and covering ∼420 × 80 nm² (111, 402). On average, one ribbon tethers 70 vesicles (taking into account that the tomogram typically contains a half of the ribbon) (170, 402; FIGURE 3); 14 of these are found near the center of the AZ, 20–50 nm from the plasma membrane, and likely constitute the readily releasable pool, or RRP (170, 344, 384). Roughly 40% of the RRP appears to be tethered to the plasma membrane, with a tether length of ∼20–25 nm (170, 383). RRP vesicles can be released by brief (<20 ms) depolarizations to potentials eliciting maximal currents (with 2–10 mM extracellular [Ca²⁺]; Refs. 34, 73, 253, 271).

FIGURE 3.

Active zone (AZ) models of ribbon synapses. A–H: cochlear inner hair cell (IHC) ribbon synapses after the onset of hearing: a section of an EM tomogram (A, E) and 3D model reconstruction (B and C, F and G). C and G: view on the AZs from the top. Ribbon and ribbon-associated vesicles were omitted for clarity. D and H: illustrations depicting the three reconstructed AZs and showing a possible arrangement of Ca_V channels. Note that the EM section typically contains a half of the ribbon, consequently the reconstructions in F and G lack a part of an AZ. All scale bars: 100 nm. [A–C modified from Strenzke et al. (344). E–G courtesy of Rituparna Chakrabarti.] I and J: EM analysis of rod bipolar (RB) terminals. I: transmission electron micrographic image of a RB terminal illustrates two ribbons (arrows) at dyad synapses, presynaptic to AII and A17 amacrine cells. Scale bar: 500 nm. J: illustration of an AII dendrite contacting a single RB terminal varicosity. Note that the AII is postsynaptic to vesicles on ribbons as well as vesicles near (<24 nm) the plasma membrane but not on ribbons. Illustration is based on EM reconstruction of a retinal volume imaged by scanning block-face electron microscopy (46, 237). Scale bar: 1 µm. [J modified from Mehta et al. (237); I is unpublished data acquired by J. H. Singer.]

Ca_V channels are clustered below each ribbon, as shown by wide-field, confocal, STED, and electron microscopy (43, 111, 158, 258, 302, 374, 402). These clusters appear as one to three stripes flanked by the presynaptic scaffold protein bassoon (111) and are colocalized with ribbons, such that each ribbon may be decorated by a single elongated Ca_V channel cluster (258). Estimates of the number of channels in a hair cell come from electrophysiological measurements of macroscopic and single-channel currents, and from these, the total number of Ca_V channels in a hair cell is estimated to be 1,000–3,000 depending on species (43, 111, 129, 302, 382, 416). How much of the macroscopic Ca current is carried through channels located outside synapses is not entirely clear. But, it appears that in the IHCs after the onset of hearing, few or no Ca_V channels may be extrasynaptic (43, 402), so dividing the number of channels by the number of ribbons yields an estimate of 40–180 channels per AZ (see TABLE 1).

Table 1.

Estimated average number of Ca_V channels per an AZ of hair cells in different organs and species

Hair Cell Type, Species	Whole Cell No. of Channels	No. of Synapses/Ribbons per Hair Cell	Approx. Average No. of Channels per an AZ
Mature mouse IHC	1,450–1,850 (43, 111, 382)	12–16 (2–3 wk old apical-mid region) (43, 111, 382)	80–120
Gerbil IHC	2,800 (416)	14–22 (basal adult IHC) (167, 416)	115–180
P5–7 mouse type I VHC	1,175 (382)	~8 (96, 382)	160
Frog saccular and papilla hair cell	1,800–~3,000 (129, 302)	19 (302), 55 (129)	90–100, 40–50

It was assumed that there are a few (500; Ref. 43, 10%; Ref. 416) or no extrasynaptic channels. IHC, cochlear inner hair cell; VHC, vestibular hair cell; AZ, active zone.

Mathematical modeling has revealed possible IHC AZ topographical arrangements (402). A model with 36 randomly distributed as well as a few “private” Ca_V channels (i.e., channels in direct contact with Ca²⁺ sensors on synaptic vesicles, thus molecularly coupled) reproduced well the estimated apparent Ca²⁺ cooperativity of IHC RRP exocytosis at smaller AZs. Such arrangement might represent the realistic mature IHC AZ topography (270, 402). For larger AZs, it seems like an additional spatial constraint around the “private” channels (i.e., an “exclusion zone”) has to be introduced to keep the control of exocytosis under experimentally observed nanodomain regime (Supplemental Figure 9 from Ref. 402).

It is worth considering the molecular mechanisms that might promote nanodomain coupling by regulating the distance between synaptic vesicles and Ca_V channels. In the immature calyx of Held, for example, it was suggested that septin 5, a filamentous GTP/GDP binding protein, may form the physical barrier preventing vesicle docking to the AZ, thus also maintaining a microdomain control of exocytosis (412). Upon development, septin molecules reportedly undergo a spatial reorganization, being less frequently detected at the mature AZs, thus allowing a tighter docking of synaptic vesicles to the AZ and the movement of Ca_V channels closer to release sites. At the IHC synapse, a developmental tightening of the vesicle-channel coupling also has been observed to accompany maturational changes in AZ arrangement (Ref. 402; see also sects. IIID3 and V), but a potential role for septin or a protein with similar function in determining the coupling distance has not been addressed. Septin itself might not be found at IHC synapses, which appear to lack neuronal syntaxins (261), the binding partners of septin at conventional synapses (27).

Future work using high-resolution microscopy or SDS replica labeling might be able to validate models of AZ organization, including the distribution of individual proteins necessary to maintain nanodomain coupling. Indeed, a recent study incorporating confocal microscopy, optical fluctuation analysis, and 3D-STED nanoscopy has provided the most accurate measurements of single mouse IHC AZs to date: 30–360 Ca_V channels are found at each AZ (of an apical IHC after the onset of hearing), far greater variability than suspected previously (258). Another issue to address is the anatomical basis of observed nanodomain [Ca²⁺] heterogeneity (110). It is interesting to consider how AZ shape and size and tonotopic variability in Ca_V channel properties (e.g., channel inactivation, open probability; Refs. 168, 244, 417) could affect signal coding at individual synapses. For example, Ca_V channel-vesicle coupling in IHCs may be looser at the base of the cochlea, such that exocytosis there is under [Ca²⁺] microdomain control, a difference also observed between high- and low-frequency-tuned auditory bullfrog hair cells (169). It was proposed that microdomain coupling may be more suited for the high-frequency-tuned IHCs of the basal cochlea, which cannot follow the frequency components of the sound, to accurately encode timing and intensity, whereas nanodomain coupling of the low-to-medium frequency-tuned IHCs would enable accurate phase-locking to stimuli in the mid- to apical cochlea (169). Tonotopic differences were also observed in the elementary properties of IHC Ca_V1.3 channels in the gerbil IHCs (416, 417) and, to a lesser degree, mouse IHCs (244). Reported differences in open probability may be due to tonotopical differences in channel modulation or contributing Ca_V1.3 splice variants (39, 417) and suggest that Ca_V channel properties along the cochlear tonotopic axis may be optimized to best suit the sound coding requirements (i.e., phase-locking to low-frequency sounds towards the cochlear apex; Ref. 417).

Understanding the organization that gives rise to nanodomain coupling has the potential to solve one of the mysteries of ribbon synapse physiology generally and hair cell synapse function specifically. At mammalian IHC synapses, two quite distinct types of postsynaptic events are observed: very large monophasic excitatory postsynaptic currents (EPSCs) and small multiphasic EPSCs (58, 121, 123, 126, 308). The two were suggested to reflect highly and poorly synchronized multivesicular release, respectively (121). At frog hair cell synapses, which operate at lower frequencies than mammalian IHCs, EPSCs are far more uniform (129, 176, 204).

One of the proposed mechanisms for multivesicular release in hair cells is the synchronization of release sites (126, 129). Considering the absence of a synchronizing electrical signal like an action potential (AP), and given the presence of [Ca²⁺] nanodomain control of exocytosis, it is difficult to envision how the graded receptor potential of an IHC could synchronize the simultaneous opening of several Ca_V channels to synchronize fusion of up to six vesicles. Indeed, the mammalian IHC AZ topology as currently understood does not seem to permit such synchronization by Ca_V channels (258, 402). According to some studies, the resting open probability of Ca_V1.3 may be too low for a number of them opening simultaneously to trigger synchronized vesicle fusion in the absence of stimulation (129, 416). Additionally, several lines of experimental data and theoretical analysis do not support the presence of release site coordination at the mammalian hair cell synapse: 1) a disproportion of adapted spiral ganglion neuron (SGN) sound-driven spike rates as measured in vivo (i.e., 200–400 Hz) and as predicted from presynaptic sustained vesicle release rates [maximally 120 Hz, obtained when considering a maximal sustained release rate of 700 Hz (271), a fusion of 6 vesicles per one EPSC (126) and each triggering one spike (308)], 2) the apparent lack of an mEPSC and the inability to reconstruct multiphasic release events from combining several desynchronized theoretical mEPSCs, and 3) the persistence of EPSC amplitude and shape heterogeneity in the near absence of extracellular Ca²⁺ or upon IHC hyperpolarization, which should both generate conditions preventing synchronization of events by Ca²⁺ (58).

Other explanations for coordinated multivesicular release from mammalian IHCs are equally unsatisfactory: there is little experimental evidence for homotypic fusion before a single release event, and homotypic fusion seems incompatible with the theoretical considerations based on available experimental findings (58). Experimental exocytic quantal size distributions could only be well reproduced in a model where homotypic fusion rate decreased not only with the distance from the AZ but also with the size of the fused vesicles. A mechanism that would prevent large vesicles from further homotypic fusion (i.e., homotypic fusion “ad perpetuum”) and enable fusion with the plasma membrane, however, is not known, and expected vesicle size distribution in such case shows a strong mismatch to the actual EM data (58). Thus it has been argued that what appears to be multivesicular release is actually quantal release (58). Experimental data in favor of that hypothesis include the finding of 1) comparable EPSC charges observed among EPSC of different shapes and of 2) big clusters of AMPA receptors, which according to the mathematical modeling should be able to support large EPSC events (58).

At frog hair cell synapses, on the other hand, evidence for quantal and multiquantal transmission is quite strong (e.g., the presence of mEPSCs and disappearance of large EPSCs upon hair cell hyperpolarization) (204). Modeling studies have suggested that a single Ca_V channel opening could generate a large enough change in [Ca²⁺] in the restricted space below the ribbons to synchronize vesicle fusion and evoke multivesicular release (129). This is an interesting hypothesis, which needs the following considerations: 1) to what degree a ribbon can act as a diffusional barrier for Ca²⁺ needs to be further addressed (Ca²⁺ imaging experiments show no exclusion of Ca²⁺ from the ribbon volume, at least not in mammalian IHCs upon longer stimulation; Refs. 110, 262, 402), and 2) at conventional synapses (81, 157) temporal jitter in the process of exocytosis occurs downstream of Ca²⁺ entry and binding to molecular sensors. Conflicting observations made at mammalian versus frog hair cell ribbon synapse as well as retinal ribbon synapses raise a general, and important, question of whether there is a prototypical ribbon synapse.

C. Ca_V Channel-Vesicle Coupling at Retinal Ribbon Synapses

Like hair cell synapses, retinal ribbon synapses are capable of exocytosis on a very rapid time scale: several vesicles/AZ/ms (22, 142, 143, 331). As in hair cells, exocytosis from retinal ribbon synapses studied in slice preparations can be evoked by the opening of as few as one channel per AZ (22, 160), whereas exocytosis from isolated neurons—goldfish bipolar cells in particular—appears to require Ca²⁺ influx through multiple Ca_V channels (70). This could arise from species differences, or it could reflect alterations in AZ geometry arising from disruption of extensive interactions between presynaptic proteins (e.g., Ca_V channels) and elements in the extracellular matrix and on postsynaptic cell(s) (28). Additionally, coordinated multivesicular release is observed at retinal ribbon synapses, particularly mammalian rod bipolar cell synapses, at which it closely resembles coordinated multivesicular release at frog hair cell synapses (160, 205, 238, 331).

In the absence of diffusion barriers that confine free Ca²⁺ (e.g., Refs. 22, 129), depolarization-evoked increases in [Ca²⁺] within a nanodomain are relatively small, implying that exocytosis from intact retinal ribbon synapses is mediated by vesicular sensors with relatively high affinity for Ca²⁺ (effective K_D in the 1–10 µM range; Refs. 97, 160). Indeed, direct measurements of sensor K_D at photoreceptor AZs support this assertion (301, 367), although sensor K_D in isolated bipolar cells appears to be far higher (∼200 µM; Ref. 143). Thus activity-dependent alterations in Ca_V channel distribution that occur as membrane is redistributed during the cycle of exo- and endocytosis have the potential to affect the efficiency of Ca_V channel-release site coupling (240, 241).

The Ca²⁺ sensor mediating fast modes of exocytosis from retinal ribbon synapses is likely synaptotagmin 1 or 2 (144, 376), and the efficiency of the release process at these synapses appears to be enhanced by the action of complexin 3, a member of the complexin family of proteins, which facilitate and synchronize evoked neurotransmitter release by stabilizing partially assembled SNARE complexes (252, 369). Indeed, fast, nanodomain-evoked exocytosis is impaired significantly at bipolar cell synapses lacking complexin 3 (252, 378). Complexin 3 does not affect the distribution of vesicles at the AZ; rather, it appears to modulate the efficiency of the release machinery itself (252). Complexin 4 also can function at photoreceptor synapses, but its expression and significance seem to be limited to circumstances in which complexin 3 is absent (10, 294).

It is important to note that not all transmission from retinal ribbon synapses is driven by [Ca²⁺] nanodomains. Indeed, at the same synapses that exhibit fast, nanodomain-evoked release, sustained presynaptic depolarization generates sufficient Ca²⁺ influx to saturate endogenous buffers and elevate global [Ca²⁺] significantly, in some instances with the assistance of Ca²⁺ released from intracellular stores (CICR), to evoke tonic exocytosis (9, 51, 60–62, 237). Slower modes of exocytosis do not require complexin 3, indicating that [Ca²⁺] is elevated sufficiently to activate release machinery operating in low efficiency modes (252). Release driven by elevated global [Ca²⁺] seems to be especially relevant to rods, in which CICR-mediated neurotransmitter release from sites away from synaptic ribbons might underlie more than half of the sustained postsynaptic response (60); this release mode could promote tonic release under dim light conditions.

D. Regulation of Ca_V Channel Membrane Expression

The abundance of Ca_V channels at the plasma membrane depends on several factors that influence their trafficking, localization, and degradation. The cellular mechanisms that underlie these processes are only partially known. In this review we will highlight some of the mechanisms involved.

1. Trafficking of Ca_V channels to the plasma membrane

Classically, Ca_V trafficking to the plasma membrane is believed to depend on the presence of auxiliary β (36) and α₂δ subunits (54; FIGURE 2). The β subunit may act to promote protein trafficking out of the endoplasmatic reticulum (ER) by masking a retention signal residing in the I–II loop of the α₁ subunit (Ca_V2.1; Ref. 36) or in the COOH-terminal region (Ca_V1.2 and Ca_V2.2; Ref. 6). Alternatively or additionally, the β subunit may guide protein folding, prevent ubiquitination and degradation of Ca_Vα₁ subunits (6, 267, 388), and/or mediate weaker secondary interactions with other channel regions (6; reviewed in Refs. 47, 93, 94). The presence of a putative ER retention signal in the COOH-terminal region of the α₁ subunit also suggests that CaM may regulate Ca_V trafficking (390).

For the α₂δ subunit, its metal ion-dependent adhesion site (MIDAS) motif within the von Willebrand factor-A (VWA) domain might act as a chaperone to prevent unfolding and/or facilitate channel trafficking towards the plasma membrane (32, 54, 115). Alternatively, an intact MIDAS motif could be essential for proper coupling of Ca_V channels and exocytosis by increasing channel density near AZ release sites (153). Further molecules that may affect trafficking of Ca_V channels are CaM kinase (329) and GIPC3 (GAIP-interacting protein, COOH terminus 3). The latter was suggested to contribute to asymmetrical abundance of Ca_V channels in the AZs along the modiolar-pillar axis of each IHC (262; see sect. IV).

Some IHCs express multiple β subunits (β_2, β₃ and β₄) (197), but β₂ (likely β_2a) is the dominant isoform (256). Whereas deletion of β₃ or mutation of β₄ subunit leaves hearing unperturbed and shows minimal effects on Ca_V channel abundance in the IHC plasma membrane (197), the absence of β₂ severely affects hearing (256). Animals lacking β₂ show strongly reduced IHC synaptic Ca²⁺ influx and exocytosis, attributable to a ∼60% reduction in plasma membrane Ca_V channel expression (256). This in turn is most likely due to defective channel trafficking to the plasma membrane (256). β₂-Deficient IHCs further display a lag in development, supporting the notion of the importance of sufficient prehearing Ca spiking for proper IHC maturation (see sect. V). For the function of OHCs, β₃ may be of greater importance (197).

In contrast to β subunits, it is less clear which isoforms of α₂δ are present in cochlear hair cells. It was first suggested that α₂δ₃ may be required for IHC function because its absence leads to impaired acoustic startle reflex in mice. But, detailed investigation of the α₂δ₃ KO mouse model revealed that hearing impairment is related to dysfunction of presynaptic SGN signaling in the cochlear nucleus rather than a defect in hair cell function (279). In fact, IHCs showed no presence of α₂δ₃ mRNA in the adult stage and very low levels before the onset of hearing. Recently, the cochlear function of a mutant mouse line (ducky mice) with truncated α₂δ₂ subunit was investigated and revealed decreased Ca currents and voltage sensitivity in IHCs, suggesting that this is the predominant α₂δ subunit of the IHC Ca_V1.3 (101). Those mice also display defects in cochlear amplification, which together with affected Ca currents likely explains an observed mild hearing impairment.

In the retina, all β isoforms are expressed, and individual isoforms are partially segregated in different layers (17, 185). β₂ seems to be the predominant accessory subunit (185), and especially the outer plexiform layer (OPL) is strongly stained by the antibodies against β₂ (17). The staining pattern resembles the immunostaining of synaptic ribbon markers, implying this is the β subunit of photoreceptor terminals (17), which also contain Ca_V1.4 (see above). Recently, a novel, retina-specific β₂ splice variant, named β_2X13, was identified. It was shown that Ca_V1.4 properties are differently modulated by the two splice variants, β_2X13 and β_2a (199), which vary in their HOOK domain (see FIGURE 2) containing key determinants controlling Ca_V channel inactivation (298). Both splice variants were detected in photoreceptor terminals, suggesting the presence of functionally heterogeneous Ca_V1.4 channels. Intriguingly, the labeling pattern of Ca_V1.4 channels was altered in retinas of the mice lacking β₂, pointing to the importance of the β₂ subunit for proper retinal targeting of the Ca_V1.4α₁ subunits. Furthermore, the phenotype of these mice was in many aspects similar to that seen in CSNB2 patients (see sect. VIB), supporting the role of this subunit in retinal synaptic transmission; in contrast, CNS-β₁-, β₃-, and β₄-deficient mice show normal ERGs (18). In heterologous expression systems, coexpression of Ca_V1.4α₁ and α₂δ₁ with β₂ (as compared with β₃) subunit significantly impacted VDI (191) such that Ca_V1.4 current even better resembled the L-type Ca currents of the native photoreceptors.

Several lines of evidence support the idea that Ca_V1.4 channels form a complex together with β₂ and α₂δ₄ accessory subunits. Similarly to β₂, antibodies directed against α₂δ₄ label the OPL of the retina (75), and α₂δ₄ was shown to colocalize with Ca_V1.4α₁ and β₂ subunits at structures resembling photoreceptor synaptic ribbons (199). Current densities of Ca_V1.4 channels heterologously expressed with either α₂δ₄ or α₂δ₁ subunits were comparable, a finding that was independent from the coexpression with β₂ subunit (12, 199), suggesting that regions of the α₂δ protein involved in Ca_V1.4 membrane targeting are largely conserved between α₂δ₄ and α₂δ₁. A spontaneous frameshift mutation in mice resulted in a truncation of the α₂δ₄ protein, which was expressed at very low levels (404). Corresponding to the defective retinal signal transmission (loss of ERG b-wave), the photoreceptor synaptic layer of the mutant mice is significantly thinner. This may be explained by a recent observation in rods, demonstrating that α₂δ₄ is essential for establishing synaptic contacts (393). In humans, two mutations in exon 19 and 25 of the CACNA2d4 gene that encodes α₂δ subunits have been associated with slowly progressive cone dystrophy (8, 405). Interestingly, inclusion of an alternative exon 25b was recently reported to result in an α₂δ₄ isoform that completely lacks the δ peptide, holding the α₂δ subunit at the membrane (12). Lack of the membrane anchor should result in secretion of the remaining part of the protein, which could exert different, also non Ca_V channel related targeting functions in the CNS (for review see Ref. 94). Immunohistochemistry revealed α₂δ₃ immunoreactivity in cell bodies of the ganglion cell layer and the inner nuclear layer (INL) as well as the inner plexiform layer (IPL) and OPL processes. The expression of α₂δ₃ in multiple cell types including ganglion, amacrine, and bipolar cells and photoreceptors, but not horizontal cells, suggests that this subunit might widely participate in retinal Ca_V channel signaling (276). So far however, no functional data from α₂δ₃-deficient animals supporting this hypothesis have been reported.

Whereas β₂ acts as the β subunit of Ca_V1.4 channels, other β isoforms in the retina may pair with other Ca_V subtypes. The IPL is occupied mostly by β₃ and β₄ subunits; expression of β₂ in this area is diffuse, but two distinct narrow bands stained by antibodies against β₃ are seen, suggesting its presence mainly on the cell bodies of cholinergic amacrine cells and in ON and OFF synaptic layers of the IPL. Rod ON bipolar cells might express β₂ and/or β₁. Finally, β₁ subunit seems to be expressed predominantly in glial Müller cells, as it was found on cell bodies in the INL and processes within the inner and outer limiting membrane (17).

2. Targeting of Ca_V channels to the AZs and stabilization of the channels at the plasma membrane

Little is known about the mechanisms by which Ca_V1.3 and 1.4 are targeted to synapses. In contrast to the Ca_V2 channels that mediate synaptic transmission at conventional synapses, Ca_V1 channels lack the synaptic protein interaction (synprint) motifs (in the II-III loop), MINT1, and CASK interaction sites that are involved in targeting Ca_V channels to presynaptic AZs (234, 247, 297, 327, 328). Thus Ca_V channel interactions with other presynaptic proteins might be particularly important for ribbon synapse function.

In addition to their roles in trafficking, auxiliary Ca_V channel subunits also stabilize L-type and most non-L-type Ca_V channels in the plasma membrane. The α₂δ subunits appear to reduce the turnover of Ca_V channels (32, 115). The β_2a subunit, which is palmitoylated and therefore stabilized in the membrane, may even anchor channels to the plasma membrane (65). A central role in stabilizing Ca_V channels close to the AZ, however, is ascribed to the RIMs. These presynaptic scaffold proteins are believed to hold channels at the presynaptic membrane either by directly binding to the Ca_Vα₁ subunit or indirectly by binding to RIM-BP (4) and/or β subunits (reviewed in Ref. 351; FIGURES 2 and 4).

FIGURE 4.

Ca_V channel-tethering complex at the ribbon active zone (AZ). L-type Ca channels at the ribbon synapses likely bind RIM via RIM-BP and a palmitoylated auxiliary β₂ subunit. RIM and RIM-BP may bind AZ proteins (potentially CAST/ELKS and/or others) to hold Ca_V channels below the ribbon (presumable interactions of RIMs are depicted by dotted arrow). A Bassoon and/or RIBEYE were further suggested to organize presynaptic Ca_V channel clusters. Whereas Ca_V1.3 may bind RIM also directly, the molecular interaction between Ca_V1.4 and RIM is unknown. For simplicity, presynaptic release machinery, which largely differs among sensory receptor cells and is partially unknown, was omitted from the illustration.

Interestingly, whereas RIM PDZ domains interact with Ca_V2 (2.1 and 2.2) and RIM C₂ domains may bind cytosolic II-III loops in α₁ subunits of Ca_V2.2 and 1.2 (weakly), none of this applies to Ca_V1.3 channels (72, 172). But, recently, a novel interaction of RIM C₂B domain with the COOH terminus of Ca_V1.3α₁ subunit was proposed (278). Thus Ca_V channels at hair cell ribbon synapses might also bind RIM directly, however, at different interaction sites as their counterparts at conventional synapses. The interactions of RIM with Ca_V1.4 are not known. Still, the interactions of RIMs and Ca_V channels also are mediated by other proteins, like β subunits and RIM-BP. RIM C₂B domain interacts with β_4, β₃ and β_2a subunits (117, 182). And, the proline-rich sequences (PxxP) of RIM bind to RIM-BPs (395), which in turn interact with Ca_V1, 2.1 and 2.2 channels (147). RIM proteins also may interact with other synaptic proteins like ELKS/CAST, SNAP-25, synaptotagmin-1, liprin-α, and RAB3. Thus RIMs are well positioned to recruit Ca_V channels into the AZ (172), although the absence of many of these proteins from mature IHC (but not photoreceptor/bipolar cell) ribbon synapses means that their interactions with RIM may not be relevant for IHC function (e.g., synaptotagmin-1, SNAP-25 etc.; reviewed in Ref. 272).

IHCs lack RIM1, and of the other long RIM isoforms, RIM2α and β are present, with RIM2α being the more abundant one (170). Additionally, IHCs seem to express the short isoform RIM3γ and potentially some RIM2γ, although the function of these isoforms is unclear. Although RIM2α and RIM3γ both promote larger Ca²⁺ currents in heterologous systems expressing IHC-like Ca²⁺ channel complexes, the absence of RIM3γ does not significantly affect IHC synaptic function or hearing (278). Consistent with the proposed interactions between long RIM isoforms and the presynaptic AZ cytomatrix, RIM2 isoforms are detected at the base of IHC ribbons, where they are necessary for clustering of Ca_V1.3 channels (170). In the absence of RIM2α (or both RIM2α and β), Ca_V channel abundance is reduced throughout the cell but most significantly at the AZs, but Ca_V channel cluster shape and organization at the AZs remain intact. Accordingly, Ca currents and exocytosis also are reduced by about half but not eliminated (170).

Of note, additional molecular links assist in clustering Ca_V1.3 near vesicles at IHC AZs (FIGURE 4). These might include palmitoylated β_2a subunits that pin trafficked Ca_V channels to the plasma membrane where RIM-BP might hold them close to the ribbon (also independent from the interaction with RIM). Experiments with RIM-BP2-deficient mice showed that in IHCs RIM-BP2 positively regulates the number of synaptic Ca_V1.3 channels, and in addition, it supports fast vesicle recruitment into the RRP after depletion, potentially by rapidly bringing synaptic vesicles in close proximity of Ca_V channels (195). Candidate interaction partners of RIM-BP are the presynaptic scaffold protein bassoon (74; see also below) and ELKS, whose homolog Bruchpilot was detected to interact with RIM-BP at the Drosophila neuromuscular junction (214). The roles of ELKS and CAST, however, are not well understood although they can alter Ca_V channel properties, as demonstrated at inhibitory hippocampal synapses (213) and in a heterologous expression system (181). Recent data also suggest partially redundant scaffolding roles among ELKS and RIM, at least at central neuronal synapses (391).

At photoreceptor ribbon synapses (394), RIMs have been suggested to serve as AZ scaffolds to which Ca_V channels are anchored by RIM-BPs (147). Deletion of RIM1/2 reduces presynaptic Ca currents in and, consequently, exocytosis from rod photoreceptors (124). This observation is consistent with the effects of RIM2 deletion observed in IHCs (above), although at retinal ribbon synapses the Ca current reduction seems to be attributable to altered single-channel conductance as Ca_V channel abundance is not changed (124). Also similar to IHCs, photoreceptor ribbon synapses abundantly express RIM2α, but no RIM1 (124, 218).

In addition to RIM proteins, the large multidomain protein bassoon, together with the ribbon itself, appears to be central to proper Ca_V channel clustering at the AZs of hair cells and photoreceptors (111, 178, 221, 233, 258, 325). Bassoon binds RIBEYE, a major component of the synaptic ribbon (see Introduction), and anchors the ribbon to the presynaptic density via an unknown molecular link (84; reviewed in Ref. 135). Although bassoon does not interact directly with Ca_V1.3 (111), disruption of bassoon (and concomitant loss of ribbons from the AZs) results in disorganized Ca_V channel presynaptic clusters, reduced abundance of Ca_V channels causing decreased synaptic Ca²⁺ influx, and fewer docked vesicles. Together, this results in diminished synaptic exocytosis leading to increased SGN spike latencies and desynchronized neural responses to sound onset in the auditory system (48, 111), and abnormal ERG responses in the visual system (84).

Because alterations in bassoon expression result in the loss of ribbons from a majority of AZs, the consequences of bassoon manipulation on Ca_V channel organization cannot be distinguished from the effects of ribbon depletion. In IHCs, AZs of bassoon mutants that retain ribbons exhibit an intermediate phenotype between wild-type and ribbonless AZs, suggesting that bassoon and ribbon might work in concert (111, 163).

With regard to RIBEYE/Ribeye, evidence from studies of zebrafish, but not mammalian, hair cell ribbon synapses implicates it as required for proper trafficking of Ca_V channels to the AZs (220, 221, 324, 325). In morpholino zebrafish hair cells, the knock-down of ribeye b (which also significantly reduces levels of Ribeye a) or of both copies of ribeye disrupts the clustering of Ca_V1.3a channels at the presynaptic AZ and partially affects the tight apposition of postsynaptic afferent contacts, thereby perturbing hearing and vestibular function (325). Mislocalized Ca_V channels were also found in another study in mutant zebrafish neuromasts with little Ribeye a and no detectable Ribeye b immunofluorescence (221). Conversely, the overexpression of Ribeye leads to formation of ectopic AZs containing ribbons and Ca_V channel clusters, but lacking postsynaptic partners (325). Enlarged ribbons at innervated AZs, however, do not seem to recruit additional Ca_V channels, suggesting that other factors may co-determine synaptic Ca_V channel abundance (323). In immature neuromasts, the effects between Ca_V channels and ribbon are mutual. Reduction of presynaptic Ca currents (by genetic or pharmacological manipulation) leads first to enlarged and malformed ribbons followed by degradation of synapses (324). Conversely, pharmacologically increased Ca²⁺ influx decreases the ribbon size and reduces the ribbon number (324).

At mammalian hair cell ribbon synapses, RIBEYE is unlikely to be required for proper trafficking of Ca_V channels to AZs. First, the distribution of Ca_V channels in maturing IHCs is broader than that of RIBEYE, suggesting that RIBEYE does not directly recruit Ca_V channels (402). Furthermore, the abundance of synaptic Ca_V channels as well as Ca²⁺ influx are normal in the IHCs of a Ribeye KO mouse model (26, 161), but Ca_V channel properties become mildly affected in more mature IHCs (161). A depolarized shift of the Ca_V1.3 activation range in IHCs may partially underlie reduced spontaneous and evoked SGN spiking rates and increased thresholds of sound-evoked spiking in SGNs of the Ribeye KO mice (161). Still, the overall effect of a congenital RIBEYE/ribbon loss has surprisingly mild effects on sound encoding and hearing (26, 161). This may be due to compensatory mechanisms in the IHCs of the Ribeye KO animals, where larger postsynaptic GluA patches (26), and multiple small presynaptic structures resembling conventional AZs were found (161; but see Ref. 26). Likely, a more suitable model to study the role of ribbon in sensory cells would be an inducible KO of RIBEYE. Still, data on constitutive Ribeye deletion helped to partially isolate the roles that bassoon and RIBEYE play individually (26, 161). A difference in the synaptic Ca_V channel abundance in the IHCs of Ribeye KO versus bassoon mutant mice suggests that bassoon, but not the ribbon, acts to promote channel tethering at the AZ and/or stabilize channels at the AZ (see also Ref. 258); possible mechanisms for channel stabilization include slowing channel recycling and/or degradation and preventing lateral displacement in the face of high rates of membrane turnover that accompany sustained transmission (271). In line with this idea, regulation of the presynaptic ubiquitin-proteasome system (UPS) by bassoon and piccolo, an AZ protein with large similarity to bassoon (103), recently has been demonstrated at CNS synapses (387). In Ribeye KO mouse retinas, Ca²⁺ influx in bipolar cells seemed unperturbed, but presynaptic Ca_V channel clusters in photoreceptors (assessed immunohistochemically) were reduced in size (233), contrary to the observation in the hair cells. This suggests RIBEYE may fulfill partially different roles in different cell types and organs or as mentioned above the effects of ribbon loss in IHCs may be partially masked by compensation.

Finally, it also is worth considering possible roles played by other AZ proteins. Ribbon synapses express piccolino, a short isoform of piccolo that lacks several interaction domains (292). Although the absence of piccolino affects ribbon size and shape significantly (291), effects on presynaptic Ca_V channels remain to be examined. But, its localization towards the top of the ribbon suggests that piccolino might not be an interaction partner of Ca_V channels (85). Possible roles for ELKS/CAST at ribbon synapses also remain enigmatic. Retinal ribbon synapses are affected by CAST deletion—ribbons are smaller and the Ca_V channel complement is altered—but effects on Ca_V channel function have not been ascertained (367c).

3. Positioning Ca_V channels close to synaptic vesicles

Once Ca_V channels are targeted to the presynaptic membrane at ribbon synapses, they are held in place very close to release sites to promote nanodomain Ca²⁺-exocytosis coupling. The ribbon itself could serve a tethering function in the retina (233), but coupling defects are not observed in IHCs of the two bassoon mutants despite defective (or lost) ribbons (111, 163). In IHCs of the RIM-deficient animals, Ca_V channel coupling to release sites is normal, as suggested by experiments manipulating the number of open channels (170), which is in contrast to the observation made at the calyx of Held (140). Thus RIM-independent molecular links must be of particular importance at ribbon synapses. Possible candidates for bringing Ca_V1 channels in the proximity of release sites include auxiliary α₂δ (153) and β subunits (117, 182) and RIM-BP (4, 74, 128) (FIGURE 4). An additional role in channel positioning may be played by actin filaments, which may hold a subset of vesicles further away from the Ca_V channels at mature IHC AZs as suggested by recent experiments (134).

Overall, it is easier to exclude than to identify constituents of the protein tethers of Ca_V1 channels and their molecular links to synaptic vesicles at ribbon synapses. Although studies of the calyx of Held (57) suggest that complexins might influence the distance between Ca_V channels and synaptic vesicles, this may not be the case at retinal ribbon synapses, where the absence of complexin 3 reduces nanodomain control of exocytosis without affecting Ca current density and presynaptic vesicle distribution (252). It, however, is difficult to determine whether the alteration in nanodomain coupling reflects subtle alterations in the relative placement of channels and vesicles or whether the observation is attributable solely to the effects of complexin 3 on the release machinery itself (252). Furthermore, complexins are not present at IHC synapses (345). Otoferlin, an essential protein for IHC synaptic transmission (307), is also unlikely to influence the Ca_V channel-synaptic vesicle coupling because otoferlin-deficient IHCs display unperturbed Ca currents (307) and electrophysiology data from otoferlin mutant IHCs suggest unchanged coupling (271, 344). Septin 5 (see sect. IIIB) and synapsin (130) act at conventional synapses to maintain the distance between Ca_V channels and release sites; however, the latter does not appear to be found in hair cells (235, 377) nor in photoreceptors and bipolar cells (223).

Lastly, it is interesting to consider how synaptic activity itself might alter Ca_V channel-release site coupling dynamically. As vesicular membrane is added to the plasma membrane during exocytosis and removed during endocytosis, the free intracellular volume of the AZ might change, as might the positions of Ca_V channels relative to vesicles. Indeed, lateral movement of Ca_V channels during exocytosis has been observed at photoreceptor AZs (240, 363), and the placement of Ca_V channels relative to synaptic ribbons can have significant effects on [Ca²⁺] nanodomains (129). It is interesting to consider the possibility that Ca_V channel-release site coupling becomes slightly less efficient during periods of sustained Ca²⁺ entry, but any decrease in coupling efficiency should be countered by accumulation of Ca²⁺ in the presynaptic terminal.

4. Control of Ca_V channel degradation

To maintain the functional integrity of the presynaptic AZ, not only must Ca_V channels be inserted into the membrane correctly, but their removal and degradation must be controlled precisely. Ca_V channels undergo constant turnover, being internalized and then recycled or degraded to be replaced by newly synthesized ones. The turnover rates for Ca_V1 channels in the plasma membranes of sensory synapses are not known. In cultured neurons, the metabolic turnover rates for Ca_V2 channels may be a few days (half-life time of on average 3 days), similar to several other (pre)synaptic proteins (71). Perhaps the most critical process regulating Ca_V channel turnover is protein ubiquitination, which regulates proteasomal or lysosomal degradation of internalized channels. It has already been noted that the β subunit allows Ca_V1.2 and Ca_V2.2 to bypass ER-associated degradation before trafficking to the presynaptic membrane (6, 388), although the mechanism is not clear.

The best model for ubiquitination of ion channels is the epithelial sodium channel ENaC, but also other ion channels, including Ca_V channels, undergo this process (6, 388). Information regarding regulation of Ca_V channel membrane density by ubiquitination however is sparse (e.g., Refs. 6, 116, 131, 226, 388). In particular, factors that regulate the rate of ubiquination and modify the process otherwise are not clear. The α₂δ subunit for example was proposed to stabilize ion channels in the plasma membrane and prevent their degradation, likely via its MIDAS motif (54). For a detailed discussion on the stabilization, lifetime, and regulation of Ca_V channel internalization and recycling, the readers are referred to other recent reviews (e.g., Refs. 118, 330).

Here, we will highlight a recently identified regulation of Ca_V1.3 degradation in hair cells by harmonin (131). Harmonin is a PDZ-domain containing scaffolding protein, required for normal hair cell mechanotransduction (133, 245) as a central organizer of the family of Usher proteins in hair cell stereocilia. It is most well-known for its defects causing the Usher syndrome type 1C, a form of hereditary deaf-blindness, in the case of the type 1C manifested as congenital profound deafness, vestibular dysfunction, and delayed onset retinopathy (retinitis pigmentosa). The mechanism of blindness is not well understood. In the inner ear, depending on the particular mutation, defects in harmonin may affect hair bundle morphogenesis (164), the amplitude of mechanotransduction current and the kinetics of its adaptation (245), or the maturation of mechanotransduction machinery (133). Harmonin might further be involved in modulation of cochlear IHC synaptic function (131, 132). In HEK cells, harmonin tags Ca_V1.3 channels for ubiquitination and alters their functional levels at the cell surface, which likely restrains their availability also at the IHC ribbon synapse (131). This function of harmonin is perturbed in a deaf-circler mouse mutant model, expressing a mutant protein lacking the central coiled-coil and PST (proline, serine, threonine-rich) domains (164) and unable to interact with Ca_V1.3 channels (131). As the protein was found at only a fraction of IHC synapses, it was also considered as one of the candidates for establishing heterogeneity of presynaptic Ca²⁺ signals, perhaps via setting a distinct Ca_V channel abundance across IHC synapses, but see section IV for further discussion on this topic. The function of this protein at the retinal ribbon synapses, also harboring harmonin (295), is not yet entirely understood (200).

Finally, it is worth considering that ubiquitination of different splice isoforms of the same channel may vary. Splice isoforms of Ca_V2.2 for example show different susceptibility to UPS-mediated degradation (226), and alternative splicing can contribute also to heart failure by promoting proteasomal degradation of Ca_V1.2 channels in the heart (155). As several splice isoforms of Ca_V1.3 in hair cells or Ca_V1.4 in the retina have been described (see sect. II), it will be interesting to examine these issues further. As well, the role of ubiquitination in the maturation of synapses is worth considering, as the abundance and localization of Ca_V1.3 changes dramatically upon maturation of IHCs (see sect. V). Heterogeneous Ca_V channel clusters, which are strictly confined to mature hair cell synapses, may be established via different mechanisms involving processes discussed above (e.g., more precise AZ targeting, differential rates of channel ubiquitination within versus outside the AZs, etc.).

IV. HETEROGENEITY OF CALCIUM INFLUX AT NANODOMAIN-COUPLED ACTIVE ZONES

A. Heterogeneity of Spiral Ganglion Neuron Spiking

The mammalian cochlea encodes sound intensities spanning six orders of magnitude. Individual SGNs, however, encode only a fraction of this range such that different intensity bands are carried by different SGNs; similar parsing of the operating range has been observed in a variety of species (208, 209, 355, 400, 413). In the cat, three functional groups of SGNs were identified: neurons with 1) low and 2) medium spontaneous rates and high thresholds preferentially innervating the neural/modiolar side of IHCs, and 3) high spontaneous rate neurons with low thresholds contacting IHCs at the abneural/pillar side, facing the OHCs (208; FIGURE 5).

FIGURE 5.

Presynaptic IHC heterogeneity supports a wide dynamic range of sound encoding. A: response characteristics of an exemplary high- (HSR; red) and low-spontaneous rate (LSR; blue) SGN fiber. Squares emphasize a difference in dynamic range of the two fibers (i.e., a difference in the stimulus level range between 10 and 90% of the maximal driven spiking rate). SGN fibers segregate into groups that primarily innervate the opposite sides of the IHCs (middle). B and C: heterogeneity of presynaptic Ca²⁺ signals [voltage sensitivity (B) and abundance of Ca_V channels (C)] may largely contribute to SGN diversification. B: fractional activation of Ca_V channels in several active zones (AZs) recorded from a single IHC. Note a hyperpolarized shift in the voltage of half-maximal activation (V_0.5) in the abneural (red) Ca_V channel clusters. C: heterogeneity in the amplitude of IHC presynaptic Ca²⁺ signals as revealed by confocal Ca²⁺ imaging. D: hypothetical composition of presynaptic IHC Ca_V clusters. Middle: fractional activation of hypothetical Ca_V channel clusters with different proportions of two channel types with distinct activation properties. Top left,: abneural AZs, which are on average smaller and contain fewer Ca_V channels, contain a higher proportion of channels that activate at more hyperpolarized potentials (magenta). Magenta and green channels may represent distinct splice variants and/or channels with distinct interacting partners that modulate their properties. Bottom left: intermediate scenario with equal proportions of channel types. Right: neural AZs may primarily contain channels with more depolarized activation (green). [Data in B are replotted from Ohn et al. (262). C adapted from Meyer et al. (244), with permission from Springer Nature.]

This diversification of SGNs may originate pre- and/or postsynaptically. One IHC contacts several SGNs (∼5–20 depending on the species and tonotopic region), but a vast majority of SGNs receive one input from a single IHC (109, 208, 210). As an IHC is believed to be isopotential, SGN diversity created by a presynaptic mechanism would require that functionally different AZs exist within one IHC (110, 126, 207, 243, 400, 415). Diversity of AZs could arise from experimentally observed heterogeneity in Ca²⁺ influx owing to Ca_V channel properties and/or density, heterogeneity in AZ anatomy and composition, and from variability in the dynamics of exocytosis including the extent of multivesicular release (110, 111, 243, 244, 258, 262, 401, 402).

B. Mechanisms Underlying Heterogeneity of Cochlear Presynaptic Ca²⁺ Signals

In cats, the SGNs with high spontaneous firing rates and low threshold that contact the pillar side of IHCs are postsynaptic to smaller ribbons than the SGNs with lower spontaneous firing rates that contact the modiolar side of IHCs (243; FIGURE 5). Although the segregation of synapse structure and function in small rodents is not as clear-cut as in the cat auditory system, generally similar observations of AZs come from studies of mouse cochlea (207, 262). Because IHC ribbon size is positively correlated with the size of presynaptic Ca_V channel cluster and with the magnitude of synaptic Ca²⁺ signal, it is tempting to consider that presynaptic mechanisms largely underlie SGN diversity (110, 262). Further support for this notion comes from the observation of heterogeneity in the voltage dependence of presynaptic IHC Ca_V1.3 channel opening (FIGURE 5), with the smaller pillar Ca_V channel clusters exhibiting slightly lower (∼1.5 mV) activation potentials (262). And, as discussed in section V, the variability of presynaptic Ca²⁺ signals increases upon development through the emergence of AZs with stronger Ca²⁺ influx (401, 402), coincident with the appearance of high spontaneous rate auditory fibers in mice (401), which however may mostly contact AZs with weaker maximal Ca²⁺ influx (see next paragraph).

Thus there seem to be at least two levels of presynaptic Ca_V regulation of individual AZs, Ca_V channel density and Ca_V channel properties. A negative shift in voltage sensitivity could partially account for preferentially lower spiking thresholds of SGNs contacting the pillar side of IHCs. It is puzzling, though, why AZs with greater voltage sensitivity are on average smaller (262). One would expect that under conditions of nanodomain control of exocytosis, smaller AZs with smaller Ca_V channel clusters would exhibit relatively low glutamate release rates. This would in turn lower the spontaneous rates and increase the threshold of SGN spiking, the opposite of what is observed in SGNs contacting the pillar AZs. And contrary, the large neural modiolar AZs would be expected to drive SGN spiking at lower thresholds, in contrast to the effect expected from their low voltage sensitivity. Potentially the larger ribbons with more Ca_V channels in the modiolar AZs are opposed to synaptic endings with a different composition of postsynaptic glutamate receptor clusters (perhaps containing many low-affinity receptors), which is worth investigating in the future. This way, the dynamic range of spiking would be broadened, as observed in low spontaneous rate, high-threshold SGN fibers (355). The larger modiolar AZs may thus serve the purpose of supporting a higher dynamic range of transmitter release (262).

What factors could (differentially) set the voltage sensitivity of Ca_V channels within individual AZs? Possible candidates include different Ca_V channel splice variants with distinct biophysical properties, preferential abundance of specific auxiliary subunits, and diverse complements of Ca_V channel modulators. As discussed in section II, IHCs express Ca_V1.3 channel splice variants with short and long COOH termini, and only the latter contain both interacting regulatory domains that determine the channel’s gating kinetics (332). The function of this COOH-terminal modulator can be interrupted by replacing part of the distal COOH-terminal regulatory domain of the long Ca_V1.3 channels with a hemagglutinin (HA) tag, making the long HA-tagged variant behave functionally like a short Ca_V1.3 splice variant in a heterologous expression system (312). However, no shift in Ca_V channel voltage dependence was observed in IHCs of the Ca_V1.3DCRD^HA/HA KI mice (containing the HA-tag in the long Ca_V1.3 variant; Refs. 262, 312), suggesting that the absence of a functional CTM in an otherwise ‟longˮ channel can still allow the interaction, e.g., with modulatory proteins and/or that alternative splicing is not involved in establishing Ca_V heterogeneity.

Alternatively, specific Ca_V channel interaction partners could be differentially distributed among individual AZs of the IHCs. Detected at only a subset of IHC synapses and found to cause a depolarized shift in Ca_V channel activation (131, 132, 262), harmonin (see also above) seems like a strong candidate for such a function. But, Ca²⁺ imaging at individual IHC AZs of wild-type and Ush1c deaf-circler mutant animals suggested that harmonin is not required for heterogeneity of presynaptic Ca²⁺ signals (262). That study, however, identified a different candidate for diversification of Ca_V channel voltage activation or abundance: GIPC3, a PDZ-domain-containing cytosolic scaffold protein (reviewed in Ref. 174). Mutation in GIPC3 causes human deafness with audiogenic seizures and progressive hearing loss mainly due to defects in hair bundle structure, mechanotransduction, and hair cell K currents (59, 293). The lack of GIPC3 in a mouse model increases whole cell Ca current and causes an overall shift in voltage activation towards more negative values in IHCs, thereby increasing spontaneous SGN spike rates. More importantly, it disrupts the modiolar-pillar gradient in AZ Ca²⁺ signal amplitude: in the absence of GIPC3, pillar AZs on average show more Ca²⁺ influx than AZs facing the modiolus (262). Because the spatial gradient in ribbon/Ca cluster size distributions may be important for generating the broad dynamic range of spiking in low spontaneous rate SGN fibers, it is notable that disruption of GIPC3 narrows the dynamic range of SGN spiking (262). This could be explained by a shift in hearing threshold due to impaired mechanotransduction and cochlear amplification, but the effect also may result, at least in part, from the absence of large AZs at the modiolar side of IHCs. Thus GIPC3 may be one of the factors establishing the heterogeneity of IHC presynaptic AZs, influencing the spatial gradient in Ca_V channel abundance but not in the Ca current voltage-dependence by an unknown mechanism. By analogy to the related protein GIPC1, which is involved in the establishment of planar cell polarity (PCP) during inner ear epithelial development (119), GIPC3 may assist the spatially polarized trafficking of specific components of the Ca_V channel complex to or away from the AZs. In the future, it will be interesting to determine whether other proteins relevant for PCP are involved in establishing IHC synaptic heterogeneity.

Another known factor that contributes to creating heterogeneity of maximal Ca²⁺ influx is bassoon (and/or ribbon itself) as disruption of bassoon results in the absence of large presynaptic Ca clusters (163, 401). As discussed in section IIID2, this may be directly related to the mutual influence between ribbon and Ca_V channel cluster (324), although this relation may be somewhat different in hair cells of lower or higher vertebrates. Interestingly, the spatial gradient of ribbon sizes also is degraded by noise trauma that causes a temporary hearing threshold shift, and the gradient only partially recovers within a week after the insult (206). The mechanisms behind this phenomenon are not well understood but might include the same processes that establish and maintain the spatial gradients in the healthy cochlea. Most importantly, it remains to be answered how spatial gradient in the voltage dependence of presynaptic Ca²⁺ influx is established, as this seems to be the most significant determinant of the postsynaptic SGN spiking rates.

C. Role for Heterogeneity at Retinal Synapses

The vertebrate retina faces similar signal processing challenges as it encodes visual stimuli that vary over ∼9 log units in intensity (300). Retinal circuitry has been the subject of many recent reviews to which the reader may refer (e.g., Refs. 14, 77, 231, 263), so discussion here will be constrained narrowly to the issue of variability in [Ca²⁺] and in Ca²⁺-exocytosis coupling.

Like those of the cochlea, the retinal signal transducers (rod and cone photoreceptors) distribute their outputs to multiple types of projection neurons, which exhibit varying responses to a sensory stimulus (13). The retina, however, differs from the cochlea in several significant ways that are likely to make synapse-level heterogeneities far less significant for physiological signal processing.

Three features of photoreceptor synapses minimize the effect of quantal variability (both in the amplitude and in the timing of individual exocytotic events) at photoreceptor synapses. 1) Photoreceptors are depolarized in darkness and hyperpolarize in response to light. Thus, in darkness, photoreceptor terminals experience sustained Ca²⁺ influx that drives tonic exocytosis of glutamatergic vesicles; light is encoded by a decrease rather than an increase in transmitter release (opposite to the coding of sound by IHCs). Consequently, the tonic release rate in darkness is quite high so that occasional slowing of the release rate, as occurs in a stochastic process, is not read out by the downstream circuitry, which has a relatively long integration time, as the absorption of photons (68, 159, 283, 287). Additionally, the long integration times of photoreceptors (10–15 to >100 ms) minimize contributions of heterogeneities in Ca²⁺ signaling that occur on shorter time scales. 2) Postsynaptic mechanisms act to reduce the influence of individual release events. Postsynaptic signaling in ON bipolar cells is mediated by a signaling cascade that begins with a metabotropic glutamate receptor (mGluR6) and ends with a cation conductance (likely TRPM1), and saturation of the second messenger cascade within this pathway can serve to reduce or eliminate the influence of presynaptic variability on postsynaptic responses (104, 310, 368, 379a). OFF bipolar cells express ionotropic glutamate receptors (both AMPARs and KARs), and the majority of these receptors are likely to be desensitized and/or saturated by tonic glutamate release (78, 80, 125). 3) The morphology of photoreceptor synapses is such that diffusion within and outside of the synaptic cleft (actually, a large synaptic invagination) filters quantal variability (80, 284, 285, 353).

Additionally, retinal circuits are characterized by significant synaptic convergence that reduces the influence that any single synapse could have on downstream neurons. With the exception of the midget pathway in the primate fovea, in which a single cone photoreceptor contacts only one second-order cone bipolar cell that in turn makes synapses with one ganglion cell (187), bipolar cells pool the outputs of multiple photoreceptors and ganglion cells, in turn, receive synapses from many bipolar cells (340). Thus a single ganglion cell can pool the outputs of >10,000 photoreceptors as in the case of convergence of rods to ON alpha ganglion cells in the mammalian retina (76, 340, 373); clearly, heterogeneities in transmission at individual synapses will be averaged out by this circuit design. Nevertheless, Ca²⁺ signals at individual ribbon synapses in cones have been found to vary strongly, which might help encoding contrasts over a wide range of light intensities (127).

V. DEVELOPMENTAL CHANGES IN CALCIUM CHANNELS AND CALCIUM CHANNEL-EXOCYTOSIS COUPLING

A. Ca_V Channels in Development of Sensory Receptor Cells and Neural Circuits

Ca_V channels play a vital role in the development of sensory circuits. Even in the absence of external stimulation, developing sensory systems, like other developing neural circuits, are highly active. Endogenously generated patterned activity might promote neuronal survival, alter or strengthen synaptic connections, and guide the refinement of precise topographic organization of the cortical and subcortical sensory centers or sensory circuits (reviewed in Refs. 3, 38, 53, 173, 201). In the developing cochlea, correlated Ca spikes have been detected in the neighboring IHCs (166, 196, 229, 371, 372). This bursting activity results from interplay between voltage-gated Ca (and Na) and K channels (34, 44, 196, 227–229) and may be modulated by ATP released from supporting cells in the cochlear Kölliker’s organ (166, 372, 389; but see Refs. 165, 320). The firing of SGNs and downstream neurons requires synaptic input from pre-hearing cochlear IHCs (121, 370, 371). In the retina, interneurons rather than sensory receptors are responsible for generating spontaneous waves of activity that propagate across the layer of ganglion cells and into subcortical structures (2, 19, 102, 114, 239). However, in both IHCs and photoreceptors, Ca_V channels are required for proper development and maturation of ribbon synapses (44, 215, 259).

In the absence of Ca_V1.3 channels, immature IHCs lose the ability to fire Ca spikes and show a strong impairment in development (44). This includes the lack of timely expression of large-conductance Ca²⁺-activated K (BK) channels and the persistence of immature SK2 channels and direct efferent innervation of IHCs (44, 259). Furthermore, IHCs display several features of morphological immaturity: Ca_V channel cluster and ribbon size, shape, and distribution appear immature (see below); similarly altered synapses are observed in photoreceptors lacking Ca_V1.4 (see sect. VIB). In IHCs, this maturation process appears to be controlled by thyroid hormone (TH)-mediated signaling (45, 321).

B. Maturation of IHC Ca²⁺ Signaling

As IHCs mature, their spontaneous firing activity becomes more structured, exhibiting stereotyped bursts of Ca APs (320), and IHCs change from “pacemakers” that time SGN bursting (371) into sound transducers (34, 120, 196, 401). In mice, this transition occurs around postnatal days 11–12, upon opening of the ear canal and regression of Kölliker’s organ as its spontaneous activity ceases (370). It is believed to be enabled by changes in IHC ionic conductances. BK channels appear, and slow outward-rectifying K channels are upregulated to attenuate spike generation (196, 228). As well, the total number of Ca_V channels is reduced (34) and their activation kinetics speeds up (417, 418). The remaining Ca_V channels are confined to the presynaptic AZs, forming clusters (43, 401, 402). Some of those AZs now host larger Ca_V channel complements, and consequently, stronger local Ca²⁺ signals emerge (401). As revealed by immunohistochemistry, these clusters undergo intensive remodeling, transforming from spotlike presynaptic arrays to stripes flanking bassoon (402).

The stabilizing interactions with scaffold proteins that support such reorganization are not known, but the proteins bassoon, RIM2, or RIM-BP may be involved, as mouse models with mutations in these proteins show perturbed Ca_V channel spatial organization and/or abundance (as discussed above; Refs. 111, 170, 195). In the remodeled clusters, Ca_V channels develop tighter spatial coupling to vesicular sensors of exocytosis (402). Although it remains to be determined how the maturation of Ca_V channel-vesicle coupling occurs, it clearly has the effect of increasing the Ca²⁺ efficiency of exocytosis and reducing the cytosolic Ca²⁺ load accompanying synaptic transmission, thereby lowering the metabolic cost of Ca²⁺ clearance. Changes in Ca_V channel clustering are accompanied by concomitant alterations in ribbon size and structure (402). Upon development, the number of ribbons decreases as synapses either undergo fusion or pruning (again, a detailed understanding of underlying processes and mechanisms is lacking). The remaining ribbons are larger and tether more vesicles (402). At the same time, SGNs with higher spontaneous rates and auditory sensitivity can first be detected (401). AZs thus become largely heterogeneous in size and structure (110). This large variability in ribbon sizes and Ca_V channel clusters likely enables differential coding of auditory information and supports its large dynamic range (described in sect. IV).

VI. CHANNELOPATHIES RELATED TO GENETIC ALTERATIONS OF VOLTAGE-GATED CALCIUM CHANNELS AND INTERACTING PROTEINS

A. Synaptopathies in Hair Cells

To date, two forms of human Ca channelopathies affecting hearing have been identified: one arises from a mutation in the Ca_V1.3 channel itself (15) and the other from a mutated Ca_V1.3 modulator, CaBP2 (318; FIGURES 6 and 7). The pathological mutation in CACNA1D, the gene encoding the Ca_V1.3α₁ subunit, was found in a genetic screen of patients with autosomal recessive deafness (15). The affected members of two consanguineous Pakistani families carry a homozygous 3-bp insertion, c.1208_1209insGGG, in the alternatively spliced exon 8B, which is abundantly present in the cochlear IHCs and the pacemaker cells of the sinoatrial node. This in-frame introduction of an additional glycine residue within a highly conserved region near the channel pore results in nonconducting Ca_V channels (FIGURE 6). The loss of conductivity may be due to abnormal voltage-sensor movements that fail to induce proper pore opening or to occlusion of the channel pore (15). In accordance with the expression pattern of the affected Ca_V1.3 spliced variant, the major symptoms of Ca_V1.3 deficiency in human and KO mice are sinoatrial node dysfunction and deafness (15, 280); the channelopathy therefore is termed SANDD syndrome (OMIM 614896). Whether other CACNA1D mutations or polymorphisms contribute to the risk for hearing disorders or heart dysfunction is not known.

FIGURE 6.

Human Ca channelopathies caused by mutations in Ca_V channels and their interaction partners at ribbon synapses. A: pathogenic loss-of-function mutations causing congenital stationary night blindness type 2 (CSNB2) were found throughout the Ca_V1.4 structure (red), including the CTM region (orange). At the inner mouth of the channel pore, few mutations were identified to cause a gain-of-function (green circles). B: interestingly, the position of one of the mutations in the Ca_V1.4 (i.e., in the IS6) coincides with the position of the loss-of-function mutation in the Ca_V1.3 channel, causing sinoatrial node dysfunction and deafness (SANDD) syndrome. In humans, the function of ribbon presynaptic Ca_V channels can be further affected by mutations in CaBPs and α₂δ₄ subunit (red crosses). Note that one intronic CaBP2 base exchange likely leads to alternative splicing, the generation of nonsense codon, and consequently to nonsense-mediated mRNA decay. CaBPs were suggested to prevent Ca²⁺/CaM binding to the Ca_V channels and thus reduce calcium-dependent inactivation in both Ca_V1.4 (A) and Ca_V1.3 (B) channels. CaBP2 may also affect VDI, potentially shielding the Ca_V channel pore from the “inactivation lid” (Ca_Vβ-AID complex in the I–II linker).

FIGURE 7.

Model of CaBP2 channelopathy at the inner hair cell ribbon synapse. A: the illustration depicts a ribbon AZ in a CaBP2-deficient compared with a wild-type inner hair cell (IHC) at some time point after the stimulus onset (sine wave). CaBP2 deficiency is proposed to lead to pronounced steady-state Ca_V channel inactivation. Increased fraction of inactivated channels (magenta) does not support sufficient synaptic transmission: fewer Ca_V channels are available for triggering release of synaptic vesicles, which is reflected in the spiking pattern of auditory nerve fibers (B). B: representative examples of auditory nerve fiber spike times (wild-type, black-gray; CaBP2-KO, magenta) in response to repetitions of 50-ms tone bursts displayed in dot raster plots. Onset responses are shown with enlarged and darker symbols. Note increased latency and jitter of the first spike as well as decreased evoked spike rates. The examples were taken from data acquired for Picher et al. (277).

Two mutations in the gene coding for CaBP2 cause an autosomal recessive, moderate to severe, prelingual hearing impairment (277, 318) that underlies the deafness locus DFNB93 (354). One mutation, detected in three consanguineous Iranian families, is a splice-site mutation, predicted to cause exon-skipping and produce a truncated protein, lacking the two COOH-terminal EF-hand motifs thus leaving only one functional EF-hand (138). The second mutation, found in an Italian family, leads to nonsense-mediated RNA decay (277). Both result in impaired modulation of Ca_V1.3 channels, and the hearing defect was recapitulated in a recently generated mouse KO model (277).

CaBP2 might suppress both VDI (preferentially) and CDI (277); the exact mechanism of action however requires further investigation, as does the question of potential partial compensation among different CaBP isoforms present in the IHCs (73). In this context, it is also worth noting that CaBP1, too, has been shown to affect both CDI and VDI in Ca_V1.2 channels (266, 423) and in heterologously expressed Ca_V1.3 channels (73). The current hypothesis suggests that the lack (or truncation) of CaBP2 results in a significant steady-state Ca_V1.3 inactivation, thereby reducing the [Ca²⁺] necessary to drive efficient synaptic transmission at IHC ribbon synapses (FIGURE 7). CaBP2 might additionally act as a cytosolic buffer, although currently it is not clear if and how a defect in CaBP2-buffering would contribute to the hearing impairment (277). Although the protein is also found in OHCs, its presence does not appear to be required to permit cochlear amplification as evidenced by normal otoacoustic emissions in KO mice and young patients (277; but see Ref. 318).

B. Synaptic Pathologies in Retinal Neurons

Numerous mutations in Ca_V1.4 (CACNA1F gene; Refs. 25, 349), α₂δ₄ subunit (CACNA2d-4 gene; Ref. 405; see also sect. IIID1), and CaBP4 (Ref. 420; see also sect. IIB2) are associated with CSNB2 in humans (421; FIGURE 6). These mutations affect retinal proteins in photoreceptors and alter both ON and OFF bipolar signaling pathways. CSNB2 patients show low visual acuity and various degrees of nystagmus, strabismus, and refractive error (37). Electroretinograms as the only precise functional phenotyping and diagnosis criteria of CSNB2 in patients show that both scotopic and photopic responses are affected (421). Their similar phenotypes have recently been reviewed in detail (421), thus we focus here on mutations in the CACNA1F gene as it is most commonly affected in CSNB2.

While most of the few cases associated with CaBP4 mutations present with high hyperopia, refractive errors due to CACNA1F mutations are more variable and include myopia and hyperopia (37, 212). Visual fields in patients are normal, but daylight vision, color vision, and visual acuity can be affected (37). More than 50% of the patients suffer from photophobia (37), often seen in cone dysfunction syndromes (1). Patients with CACNA1F mutations in particular may present with few or no night vision problems (246, 421). Clearly some of the variation in the clinical manifestation of the disease might arise from the different mutations in the CACNA1F gene (FIGURE 6) that cause different channel defects. Judging from studies in various heterologous expression systems, the spectrum of Ca_V1.4 dysfunction is indeed wide (49, 146, 150, 151, 236, 275, 333). Many mutations cause severe structural changes in the protein so that functional channels are unlikely to form, and most missense mutations lead to dramatically reduced or abolished Ca_V1.4 activity (for review, see Ref. 342). A typical loss-of-function phenotype might be attributed to fewer functional channels as some of the mutations are predicted to decrease channel stability and promote misfolding (49, 342). Truncation mutations in the COOH terminus were reported to result in the loss of functional CTM and thus unmasked CDI (49, 333). Such mutations may fail to support continuous Ca²⁺ influx and limit the dynamic range of photoreceptors (e.g., when adapting to variation in illumination).

Our understanding of the pathology of CSNB2 largely stems from various mouse models. The two Ca_V1.4 KO models available (225, 338) provided the first critical clues; however, they exhibit a much stronger phenotype than human patients (225). Ca_V1.4 KO mice are functionally blind, show a substantial loss in ganglion cell responsiveness to physiological light stimuli (186), and lack visually evoked cortical activation (225). More similar to the human CSNB2 is a phenotype of a naturally occurring mouse model, nob-2 (56, 91), in which the full-length Ca_V1.4 protein (carrying a slight modification at the NH₂ terminus) is expressed at significantly reduced levels (91). These mice show detectable but reduced ERG responses, although spatial contrast sensitivity measured from optokinetic responses is rather similar to that of wild-type mice (91).

One missense mutation found in a New Zealand family (Ca_V1.4-I745T in human, mouse homolog I756T; Ca_V1.4-IT; Ref. 152) has attracted particular interest in the CSNB2 community (FIGURE 8). In a heterologous expression system, Ca_V1.4-IT gating properties show a remarkable −30 mV shift in the voltage dependence of activation, likely due to increases in Ca_V1.4 channel conductance and open probability and to significantly slower inactivation kinetics (146). Mice that possess the corresponding mutation recapitulate many aspects of the phenotype of human patients with the same mutation (185, 215, 288) and additionally show a progressive retinal degeneration (186, 288). Judged from ERG recordings, photoreceptor degeneration in the New Zealand family seems rather unlikely (152), but no long-term clinical investigation was undertaken to support this. CSNB2 patients and myopic controls, however, were found by a different study using spectral-domain optical coherence tomography to show differences in retinal layer thickness (63).

FIGURE 8.

Model of Ca_V1.4 channelopathy at photoreceptor synapse. A: photoreceptor terminals of Ca_V1.4 wild-type (WT, black) compared with mutant photoreceptors. The gain-of-function mutation Ca_V1.4-IT (IT, red) is characterized by a strong hyperpolarized shift in the current-voltage relationship (middle). Morphological changes reported in Ca_V1.4-IT retinas are indicated (right). A certain amount of functional ribbon synapses is still formed in Ca_V1.4-IT animals; however, these only allow anomalous downstream synaptic transmission (‟delayedˮ). Most synapses in Ca_V1.4-IT retinas contain immature-like ribbons that may not support synaptic transmission (‟nonresponderˮ). B: representative examples of OFF ganglion cell spike times in response to repetitive light stimuli. C: peri-stimulus time histograms show an increased latency of ganglion cell response of Ca_V1.4-IT OFF cells.

In the absence of functional Ca_V1.4 channels, photoreceptor ribbon synapses remain mostly immature, as evidenced by their roundish (or elongated) appearance (215, 286, 288, 414) in the Ca_V1.4 mutant animal models. Additionally, Ca_V1.4 dysfunction leads to dendritic sprouting of bipolar and horizontal cells, forming ectopic synapses (185, 215, 288). Similar sprouting of bipolar cell dendrites is seen in mouse models of retinal degeneration in which photoreceptors die (346–348). Disruption in the maturation of photoreceptor synaptic ribbons along with free floating ribbons also was observed in cases of changed Ca_V channel dynamics, e.g., in the presence of Ca_V1.4-IT (FIGURE 8). Accordingly, the physical integrity of proteins in the ribbon compartment as well as the arciform density/plasma membrane compartment are affected in Ca_V1.4-IT (288, 338). This phenotype is comparable to the one described for mice lacking intact bassoon, which in photoreceptors and bipolar cells links the two compartments (84, 367a, 290, 337). In addition to its role in anchoring the ribbon, bassoon is also essential in the assembly and transport of ribbon precursor spheres during early steps of photoreceptor synaptogenesis (242, 290, 367b). In some aspects, bassoon-deficient retinas resemble the nob2 (no b-wave 2) phenotype but not that of Ca_V1.4 KOs. For example, dystrophin, which is a ligand of the dystroglycan/pikachurin complex important for ribbon synapse formation (264), is correctly localized in the nob2 mutants but not in Ca_V1.4-deficient retinas (24, 338). The small fraction of remaining Ca_V1.4 channels in the nob2 retinas presumably sustains the binding of some critical interaction partners for the development and (limited) functioning of ribbon synapses. KOs of β₂ and α₂δ accessory subunits and Ca_V1.4-activating CaBP4 cause similar defects in synaptic ribbon development and formation (18, 136, 404), supporting the notion that retinal ribbon formation relies on a microdomain environment containing key AZ proteins (289).

C. Ca_V Channel Pharmacology and Limitations for Targeted Therapies

L-type Ca channel blockers have been widely prescribed for decades as treatment for hypertension and myocardial ischemia. The sensitivity of L-type Ca channels to DHP inhibitors varies between tissues, presumably due to their differential Ca_V and accessory protein expression. In heterologous expression systems, Ca_V1.3 and Ca_V1.4 exhibit ~5- to 10-fold lower sensitivity to DHPs than Ca_V1.2 at negative membrane potentials (190, 191, 408). This is consistent with the early work on Ca_v channel pharmacology of photoreceptors, which also suggested a low-affinity blockage of L-type Ca channels (398). The subtype difference might explain why therapeutic doses of DHPs do not cause side effects like changes in hearing thresholds from Ca_V1.3 inhibition in cochlear IHCs, visual impairment from Ca_V1.4 block in retinal photoreceptors, and CNS disturbances from Ca_V1.2/Ca_V1.3 block in the brain. Moreover, compared with heart tissue, DHP concentrations taken up into the brain were reported to be substantially lower (375). This suggests the uptake through the blood-retina barrier may be restricted. Pharmacological activation of mutated Ca_V1.4 channels showing a strong reduction in their expression density by L-type Ca channel activators (e.g., BAY K 8644) would not be clinically applicable to human retinal disorders due to toxic side effects expected from activation of Ca_V1.2 and Ca_V1.3 in other tissues (for review, see Ref. 419).

VII. CONCLUSIONS

In considering the future of studies of Ca_V channels and their role in synaptic transmission at ribbon synapses, the establishment and maintenance of AZ organization emerges as an important theme, particularly as “nanodomain” Ca_V channel-release site coupling has been found to exist at most ribbon synapses. Future work likely will address the following questions: What mechanisms control the trafficking of Ca_V channels to the AZ, the localization of Ca_V channels within the AZ, and the lifetimes of the Ca_V channel proteins?

Although manipulations that dramatically disrupt AZ architecture (e.g., disruption of the organizing protein bassoon and the ribbon protein RIBEYE/Ribeye; Refs. 178, 233, 325) alter surface Ca_V channel expression, the mechanisms underlying the precise targeting of Ca_V channels to functional ribbon AZs are unknown. Ca_V1 lacks a conventional synaptic protein interaction (synprint) site (247), so other, unknown protein-protein interactions must be responsible for the precise Ca_V channel-release site coupling characteristic of ribbon AZs.

With regard to the localization of Ca_V channels at the AZ, it has been demonstrated that Ca_V channels in photoreceptors move—they are displaced laterally—following exocytosis (240). It is of interest, then, to consider how Ca_V channel-release site coupling can be maintained as large quantities of membrane are added to and removed from the cell surface during tonic exocytosis, particularly as Ca_V channel mobility at ribbon synapses is affected by membrane composition (241). As well, it is important to recognize that the ability of cells, particularly hair cells, which are tethered in a rigid substrate necessary for mechanotransduction, to deform following exo- and endocytosis is limited.

Finally, although long-term synaptic plasticity at ribbon synapses has not been shown to occur, it likely does given the diurnal variability in presentation of stimuli to sensory organs (e.g., photoreceptor synapse structure changes with the light-dark cycle; Refs. 16, 248). Here, some insight perhaps can be gained from studies of photoreceptors in hibernating mammals or of hair cells exposed to noise trauma (198, 206, 238, 296), where changes in Ca_V channel abundance or function may be expected.

GRANTS

This work was supported by the German Research Foundation (DFG Priority Program 1608, PA 2769/1–1; to T. Pangrsic), National Eye Institute Grants EY 017836 and 021372 (to J. H. Singer), and Innovative Training Networks 674901 as well as Fonds zur Fӧrderung der wissenschaftlichen Forschung P26881 and P29359 (to A. Koschak).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.