Calcium dynamics and regulation in horizontal cells of the vertebrate retina: lessons from teleosts

Abstract

Horizontal cells (HCs) are inhibitory interneurons of the vertebrate retina. Unlike typical neurons, HCs are chronically depolarized in the dark, leading to a constant influx of Ca²⁺. Therefore, mechanisms of Ca²⁺ homeostasis in HCs must differ from neurons elsewhere in the central nervous system, which undergo excitotoxicity when they are chronically depolarized or stressed with Ca²⁺. HCs are especially well characterized in teleost fish and have been used to unlock mysteries of the vertebrate retina for over one century. More recently, mammalian models of the retina have been increasingly informative for HC physiology. We draw from both teleost and mammalian models in this review, using a comparative approach to examine what is known about Ca²⁺ pathways in vertebrate HCs. We begin with a survey of Ca²⁺-permeable ion channels, exchangers, and pumps and summarize Ca²⁺ influx and efflux pathways, buffering, and intracellular stores. This includes evidence for Ca²⁺-permeable α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors and N-methyl-d-aspartate receptors and for voltage-gated Ca²⁺ channels. Special attention is given to interactions between ion channels, to differences among species, and in which subtypes of HCs these channels have been found. We then discuss a number of unresolved issues pertaining to Ca²⁺ dynamics in HCs, including a potential role for Ca²⁺ in feedback to photoreceptors, the role for Ca²⁺-induced Ca²⁺ release, and the properties and functions of Ca²⁺-based action potentials. This review aims to highlight the unique Ca²⁺ dynamics in HCs, as these are inextricably tied to retinal function.

horizontal cells (HCs) are inhibitory neurons that provide the first layer of neural processing in the vertebrate retina. They share a long history as model central neurons and have been studied for over one century (Ramón y Cajal 1909). Teleost HCs, in particular, are especially large, lending themselves well to electrophysiology, and were the first cells of the retina from which intracellular recordings were obtained (Dowling 1987; Svaetichin 1953). Their graded responses to light demonstrated, for the first time, that neurons could encode and relay information without using action potentials (APs). Today, the membrane properties and ionic currents of teleost HCs are well characterized (DeVries and Schwartz 1992; Dixon et al. 1996; Jonz and Barnes 2007; McMahon and Brown 1994; Shingai and Christensen 1986; Tachibana 1981, 1983), yet their Ca²⁺ dynamics are only beginning to be clarified. In neurons, Ca²⁺ plays a central role in vesicular neurotransmitter release (Wadel et al. 2007), synaptic plasticity (Zucker 1999), gene transcription (Lyons and West 2011), and second-messenger pathways (Clapham 2007). In addition, high levels of cytosolic Ca²⁺ are toxic to neurons (Choi 1992; Lipton 1999), and therefore, Ca²⁺ sequestration and efflux mechanisms must be present to maintain cell viability. Therefore, the understanding of the dynamics and effects of Ca²⁺ in a given neuron is often crucial to understanding its function.

In the outer retina, HCs provide inhibitory feedback to photoreceptors (PRs), modulating synaptic gain and mediating a center-surround antagonism that allows for edge detection (Baylor et al. 1971; Thoreson and Mangel 2012). Although there is no consensus as to how this feedback occurs, three mechanisms have been proposed in vertebrates: 1) HCs may release less GABA onto PRs during light stimulation (GABA disinhibition) (Hirano et al. 2005, 2016; Liu et al. 2013; Wu and Dowling 1980) by vesicular release in mammals and largely through an unconventional transporter mechanism in nonmammalian vertebrates (Schwartz 2002); 2) hemichannels at HC dendrites may alter PR membrane potential by an ephaptic mechanism (Byzov and Shura-Bura 1986; Fahrenfort et al. 2009; Gardner et al. 2015; Kamermans et al. 2001); and/or 3) HCs may release protons or regulate cleft acidity (Vroman et al. 2014) to reduce the conductance of PR voltage-gated Ca²⁺ channels (VGCCs) and shift their activation to more depolarized voltages (Barnes 2003; Hirasawa and Kaneko 2003; Vessey et al. 2005).

Ca²⁺ signaling in HCs would likely affect all three mechanisms. In rats, HC VGCCs have been shown to be necessary for vesicular GABA release and for feedback in a pH-dependent fashion (Liu et al. 2013), and GABA release from HCs in fish is probably partially Ca²⁺-dependent (Ayoub and Lam 1985; Lasater and Lam 1984). HC hemichannels are regulated by extracellular Ca²⁺ concentrations ([Ca²⁺]_o) (DeVries and Schwartz 1992; Jonz and Barnes 2007). There is also evidence that intracellular Ca²⁺ concentration ([Ca²⁺]_i) affects extracellular pH and may therefore affect pH at the HC-PR synapse (Kreitzer et al. 2012). In light of this evidence, a model for HC Ca²⁺ dynamics will be instrumental in understanding HC-to-PR feedback and neural processing in the outer retina in general.

This review will explore the experiments that have characterized Ca²⁺ pathways in teleost and mammalian HCs. Each section below begins by discussing what is known in teleost models and offers a comparison with mammals. First, the major Ca²⁺-permeable ion channels, pumps, and exchangers will be described, including influx and efflux pathways, as well as Ca²⁺ buffering, extrusion, and sequestration mechanisms. Subsequent sections will review what is known about Ca²⁺-based APs in teleost HCs. These have been reported in a variety of experimental preparations but largely remain a mystery. Finally, we will summarize the research thus far on Ca²⁺ homeostasis in HCs and highlight putative roles for Ca²⁺ in HC function. These details are summarized in Table 1, and a working model of Ca²⁺-permeable pathways and their regulation in teleost HCs is presented in Fig. 1.

Table 1.

Ca²⁺-permeable ion channels, pumps, and exchangers found in teleost HCs

Ion Channel, Pump, or Exchanger	Species	Common Name	Subtype*	Reference
Ionotropic Glutamate Receptors
NMDARs	Carassius auratus	Goldfish†	H1	(Jiang et al. 2008; Shen et al. 2006; Wang et al. 2008)
	Ictalurus punctatus	Catfish	Cone	(Linn and Christensen 1992; O'Dell and Christensen 1986, 1989)
	Perca fluviatus	European perch	Cone	(Schmidt 1997)
	Roccus americana	White perch	Cone	(Schmidt 1997)
Ca2+-permeable AMPARs	Carassius auratus	Goldfish†	Cone	(Huang and Liang 2005)
			H1	(Huang et al. 2006; Sun et al. 2010)
	Not provided	Carp	Cone	(Okada et al. 1999)
Voltage-Gated Ca2+ Channels
L-type	Carassius auratus	Goldfish		(Jonz and Barnes 2007; Tachibana 1981, 1983)
	Cyprinus carpio	Carp		(Hayashida and Yagi 2002b; Murakami and Takahashi 1987)
	Roccus americana	White perch	All‡	(Lasater 1986)
	Ictalurus punctatus	Channel catfish	Cone	(Dixon et al. 1993; Linn and Gafka 2001; Shingai and Christensen 1983, 1986; Takahashi et al. 1993)
P/L-type	Roccus chrysops	White bass	All‡	(Sullivan and Lasater 1992)
			Cone	(Pfeiffer-Linn and Lasater 1996)
T-type	Roccus chrysops	White bass	All‡	(Sullivan and Lasater 1992)
Store- and Extrusion-Related Channels
Ryanodine receptors	Carassius auratus	Goldfish†		(Huang et al. 2004, 2006)
			H1	(Lv et al. 2014)
	Ictalurus punctatus	Channel catfish	Cone	(Linn and Gafka 2001; Linn and Christensen 1992)
				(Micci and Christensen 1998)
	Moroni chrysops	White bass	H2	(Solessio and Lasater 2002)
Inositol triphosphate receptors	Ictalurus punctatus	Channel catfish		(Micci and Christensen 1996)
	Ictalurus punctatus	Channel catfish	Cone	(Linn and Gafka 2001)
Store-operated channels	Carassius auratus	Goldfish†	H1	(Lv et al. 2014)
Na+/Ca2+ exchanger	Carassius auratus	Goldfish		(Hayashida et al. 1998; Yasui 1987)
	Cyprinus carpio	Carp		(Yasui 1987)
	Ictalurus punctatus	Channel catfish		(Micci and Christensen 1998)
Plasma membrane Ca2+ pump	Carassius auratus	Goldfish		(Hayashida et al. 1998; Kreitzer et al. 2012)
	Ictalurus punctatus	Channel catfish		(Jacoby et al. 2012; Kreitzer et al. 2007; Micci and Christensen 1998)
Gap Junctions and Hemichannels
Cx26	Carassius carassius	Carp		(Janssen-Bienhold et al. 2001)
	Danio rerio	Zebrafish		(Kamermans et al. 2001)
Cx35/36	Carassius auratus	Goldfish†	H1, H2	(Liu et al. 2009)
Cx43	Carassius carassius	Carp		(Dermietzel et al. 2000; Janssen-Bienhold et al. 1998)
Cx52.6	Danio rerio	Zebrafish		(Shields et al. 2007; Sun et al. 2012; Zoidl et al. 2004)
Cx55.5	Danio rerio	Zebrafish		(Dermietzel et al. 2000; Klaassen et al. 2011; Shields et al. 2007; Sun et al. 2012)
Pannexin-1	Danio rerio	Zebrafish		(Prochnow et al. 2009a, b)
Uncharacterized	Carassius auratus	Goldfish		(Jonz and Barnes 2007)
			H1	(Schmitz and Wolburg 1991)
			Cone	(Shen et al. 2003)
	Cyprinus carpio	Carp	All‡	(Kaneko and Stuart 1984; Teranishi et al. 1983; Yamada and Ishikawa 1965)
	Danio rerio	Zebrafish		(McMahon 1994; McMahon and Brown 1994)
	Devario danio	Giant danio	H1	(McMahon and Mattson 1996)
	Ictalurus punctatus	Channel catfish	Cone	(DeVries and Schwartz 1989, 1992; Dixon et al. 1996; Sakuranaga and Naka 1985)
	Roccus americana	White perch	All‡	(Lasater and Dowling 1985)
	Roccus chrysops	White bass	Cone	(Lasater 1987)
	Roccus chrysops x roccus saxitalis	Hybrid striped bass		(Sun et al. 2009; Zhang and McMahon 2000)

Cx, connexin.

^*If blank, the associated references did not specify subtypes.

^†In these papers, the authors referred to Carassius auratus as carp instead of goldfish.

^‡The ion channel, pump, or exchanger was found on all subtypes.

Fig. 1.

Ca²⁺ pathways and their regulation in teleost horizontal cells. 1: In darkness, glutamate (red circles) is released from photoreceptors and binds onto HC glutamate receptors, allowing Ca²⁺ influx through Ca²⁺-permeable AMPARs and NMDARs and leading to depolarization of the plasma membrane. 2: Depolarization also opens VGCCs, including L-type and (at least in bass) T-type VGCCs. 3: Glutamate receptors acidify the cytosol, downregulating L-type VGCCs. In contrast, group I and/or III metabotropic glutamate receptors may upregulate VGCCs (see text). 4: Additionally, glutamate triggers CICR from RyRs, further downregulating L-type VGCCs. 5 and 6: The SERCA pump then refills Ca²⁺ stores, along with SOCs and/or NCXs functioning in reverse. 7: IP3Rs have been found in catfish but do not contribute to CICR. 8: Gap junctions and hemichannels might allow Ca²⁺ flow between coupled cells or with the extracellular environment in some conditions. 9: Ca²⁺ influx from AMPARs, NMDARs, and VGCCs is extruded by the NCX and possibly to a lesser extent, the PMCA. 10: Cytosolic Ca²⁺ is strongly buffered by Ca²⁺-binding proteins. AMPAR, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; NMDAR, N-methyl-d-aspartate receptor; VGCC, voltage-gated Ca²⁺ channel; CICR, Ca²⁺-induced Ca²⁺ release; RyR, ryanodine receptor; SERCA, sarcoplasmic/endoplasmic reticulum Ca²⁺ ATPase; SOC, store-operated channel; IP3R, inositol triphosphate receptor; NCX, Na⁺/Ca²⁺ exchanger; PMCA, plasma membrane Ca²⁺-ATPase.

Ca²⁺ INFLUX IN HCs

Ca²⁺ influx in teleost HCs: Ca²⁺-permeable and -impermeable iGluRs.

In the vertebrate outer retina, PRs are tonically depolarized in the dark by an Na⁺-dominant “dark current” (Stryer 1991; Yau 1994). As a result, glutamate is constantly released onto HCs and bipolar cells (BCs) (Dowling and Ripps 1973; Miller and Schwartz 1983). PRs, HCs, and off BCs are constantly depolarized to approximately −30 to −20 mV in the dark (Dowling 1987). For HCs, this triggers prolonged Ca²⁺ influx through both ionotropic glutamate receptors (iGluRs) and VGCCs (Linn and Christensen 1992; Tachibana 1983, 1985). HCs have been shown to respond to glutamate in a diversity of teleost species, including the following: crucian carp (Carassius carassius) (Lasater and Dowling 1982; Murakami et al. 1972), common goldfish (Carassius auratus) (Ishida et al. 1984; Tachibana 1985), channel catfish (Ictalurus punctatus) (O'Dell and Christensen 1989), European perch (Perca fluviatus), and white perch (Roccus americana) (Schmidt 1997).

It is important to understand which iGluRs exist on these cells, because iGluR subtypes have different biophysical properties and physiological functions, including Ca²⁺ permeability. N-Methyl-d-aspartate receptors (NMDARs) are permeable to Ca²⁺, whereas both α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) and kainate receptors (KARs) can be either Ca²⁺ permeable or Ca²⁺ impermeable. Notably, AMPARs lacking GluR2 subunits are Ca²⁺ permeable. Ca²⁺ permeability for AMPARs containing GluR2 subunits depends on editing at a single amino acid site (Hollmann and Heinemann 1994; Peng et al. 2006). Similarly, the Ca²⁺ permeability of KARs depends on RNA editing for at least some subunits (Köhler et al. 1993).

In all teleost species studied, HCs respond to a variety of iGluR agonists, including glutamate, kainate, quisqualate, AMPA, and NMDA (Huang et al. 2006; Ishida et al. 1984; Lasater and Dowling 1982; Linn and Christensen 1992; O'Dell and Christensen 1986; Okada et al. 1999; Schmidt 1997). Moreover, it was found that kainate, quisqualate, and glutamate competitively bind to the same receptor (Ishida and Neyton 1985) and that both kainate and quisqualate responses are inhibited by selective AMPAR antagonists (Linn and Christensen 1992). With the use of modulators of KAR desensitization (concanavalin A and cyclothiazide) in carp HCs, Shen et al. (1999) did not find evidence for KARs; instead, they concluded that the teleost iGluR on HCs was an AMPA-preferring receptor. Interestingly, Huang and Liang (2005) proposed that Ca²⁺-permeable and Ca²⁺-impermeable AMPARs coexist on HCs. When they coapplied AMPA with the selective Ca²⁺-permeable AMPAR blocker, joro spider toxin-3, they found a decrease in [Ca²⁺]_i transients; in a similar manner, the selective Ca²⁺-impermeable AMPAR antagonist, pentobarbital (Yamakura et al. 1995), reduced AMPA-elicited responses. As further evidence for a Ca²⁺-permeable AMPAR, AMPA applied in situ elicits a rise in [Ca²⁺]_i, even when coapplied with Co²⁺ to block VGCCs (Okada et al. 1999). Similarly, kainate-induced [Ca²⁺]_i increases cannot be attributed to the Na⁺-Ca²⁺ exchange, since kainate elicits an increase in [Ca²⁺]_i when extracellular Na⁺ is replaced with N-methyl-d-glucamine (Okada et al. 1999). Therefore, Ca²⁺-permeable AMPARs likely represent a major source of steady Ca²⁺ influx into HCs. Indeed, Sun et al. (2010) recently found that Ca²⁺-permeable AMPARs account for 75% of all iGluR currents.

Early investigators were unable to find NMDARs in fish HCs in a variety of species (Lasater and Dowling 1982; Lasater et al. 1984), and some even found a hyperpolarizing response to NMDA—and only at high concentrations (Ariel et al. 1984, 1986; Rowe and Ruddock 1982). After failing to find NMDARs in carp HCs, Okada et al. (1999) proposed that Ca²⁺-permeable AMPARs might be better suited to increase [Ca²⁺]_i in response to graded potentials. AMPARs are not limited by a voltage threshold, whereas NMDARs would be blocked at most physiological voltages. The first evidence of an NMDAR eliciting a depolarizing response in teleost HCs came from the catfish retina (Linn and Christensen 1992; O'Dell and Christensen 1986). Thereafter, similar responses were shown in crucian carp (Jiang et al. 2008; Shen et al. 2006) and goldfish (Wang et al. 2008). Perhaps these later studies found greater success because they included coapplication of glycine (a necessary NMDAR coagonist) (VanDongen 2008) with NMDA or removal of extracellular Mg²⁺ (a voltage-dependent NMDAR blocker).

NMDARs may exist only on cone HCs. In most species, teleost HC subtypes are designated H1, H2, H3, and H4, in order of increasing dendritic field size, decreasing somatic size, and increasing distance from the outer retinal layers (Stell and Lightfoot 1975). These subtypes synapse with PRs selectively: H1, H2, and H3 synapse with cones, whereas H4 synapses with rods. All aforementioned studies that found NMDARs in HCs focused solely on cone HCs (predominantly H1), with one exception: a study on crucian carp looked at both rod and cone HCs but failed to find NMDARs in H4 (Shen et al. 2006).

Although the relative contribution of NMDARs and AMPARs to the [Ca²⁺]_i response has not been clearly established, both likely contribute to Ca²⁺ influx in the dark in teleosts, as will be discussed in the next section. In contrast, there is no evidence for functional NMDARs in HCs of adult mammals (Kalloniatis et al. 2004; Massey and Miller 1987; Shen et al. 2004).

Ca²⁺ influx in teleost HCs: glutamate triggers an increase in [Ca²⁺]_i through both VGCCs and iGluRs.

HC responses to glutamate are especially well characterized in Ca²⁺ imaging experiments, in which changes in [Ca²⁺]_i are quantified over time. Initially, glutamate induces a large, transient peak in [Ca²⁺]_i, which is then lowered to a sustained, elevated baseline for as long as glutamate is applied (Hayashida and Yagi 2002b). The transient peak is blocked by ryanodine and is likely due to Ca²⁺-induced Ca²⁺ release (CICR) through ryanodine-sensitive stores (Huang et al. 2004).

Several studies suggest that the sustained [Ca²⁺]_i component is due to activation of more than one Ca²⁺-permeable pathway (Hayashida and Yagi 2002b; Hayashida et al. 1998; Linn and Christensen 1992). The sustained Ca²⁺ response to glutamate (Hayashida et al. 1998) or AMPA (Huang and Liang 2005) is greatly reduced by applying nifedipine, a selective blocker of L-type VGCCs in teleost HCs (Takahashi et al. 1993). This suggests that glutamate depolarizes HCs, thereby activating L-type VGCCs. Furthermore, in concurrent voltage-clamp and Ca²⁺ imaging experiments, glutamate only subtly increased sustained [Ca²⁺]_i after a voltage step to −5 mV (Hayashida and Yagi 2002b), suggesting that once VGCCs are open by depolarization, iGluRs do not greatly increase [Ca²⁺]_i. However, this sustained component of the glutamate-elicited Ca²⁺ response cannot be completely explained by VGCC conductance; for example, nifedipine application does not abolish the sustained response to glutamate (Hayashida et al. 1998).

There is also evidence for Ca²⁺ influx through iGluRs themselves. The blocking of VGCC conductance with the L-type blocker, nimodipine, does not abolish responses to kainate in catfish HCs (Linn and Christensen 1992). Furthermore, under voltage clamp at potentials high enough to ensure that VGCCs are deactivated, iGluR agonists, such as NMDA, glutamate, quisqualate, and kainate, increase [Ca²⁺]_i to a new, elevated baseline (Hayashida et al. 1998; Linn and Gafka 2001; Linn and Christensen 1992). This suggests that both VGCC and iGluR conductances contribute to the sustained [Ca²⁺]_i elevation during glutamate application.

Ca²⁺ influx in teleost HCs: most Ca²⁺ enters the cytosol through L-type VGCCs.

L-type VGCCs in HCs have been well studied with electrophysiological techniques in a number of species, including goldfish (Tachibana 1983), channel catfish (Shingai and Christensen 1983, 1986), white perch (Lasater 1986), and white bass (Roccus chrysops) (Sullivan and Lasater 1992). These studies reported an inward conductance that activated in a range between −45 and −30 mV and peaked between 0 and 25 mV.

Many reports found that L-type VGCCs contribute nearly 50% of the Ca²⁺ current elicited by glutamate application in situ (Hayashida and Yagi 2002b; Hayashida et al. 1998; Huang and Liang 2005), and therefore, they represent a major source of Ca²⁺ influx in teleost HCs. It should be mentioned here that unlike typical VGCCs that undergo strong, Ca²⁺-dependent inactivation, such inactivation is weaker in HCs (Tachibana 1983). Instead, there is evidence that VGCCs are downregulated by low intracellular pH, which can be induced by extracellular acidification (Jonz and Barnes 2007) or by intracellular acidification caused by glutamate application (Dixon et al. 1993; Takahashi et al. 1993). In addition, VGCCs are downregulated by CICR (Linn and Gafka 2001). The available evidence suggests that CICR can raise [Ca²⁺]_i enough for substantial VGCC inactivation, but currents through VGCCs cannot.

VGCCs may also be enhanced by metabotropic glutamate receptors (mGluRs). In channel catfish, group I and III mGluR agonists expand the range of activation potentials into the physiological range, shifting the membrane potential at which peak VGCC currents can be obtained by −5 mV (Linn and Gafka 1999). Group I and III mGluR antagonists prevent this VGCC facilitation (Linn and Gafka 1999). This mGluR-mediated modulation of VGCC currents is probably widespread in vertebrate HCs, which express a range of mGluRs (Beraudi et al. 2007; Thoreson and Witkovsky 1999), although group I mGluRs may be absent in the goldfish (Peng et al. 1995).

Ca²⁺ influx in teleost HCs: other VGCC subtypes may contribute to Ca²⁺ influx in teleosts.

Several reports have found evidence for T-type VGCCs in teleost HCs but only in white bass. T-type channels are characterized by small, transient Ca²⁺ currents that activate approximately −65 mV and peak near −20 mV, a low permeability to Ba²⁺, and block by Ni²⁺ (Akopian et al. 1997; Perez-Reyes 2003). Sullivan and Lasater (1992) found a channel with most of these properties, although there were differences between this channel and archetypical VGCCs. Specifically, T-type VGCCs in white bass HCs are not blocked by millimolar concentrations of Ni²⁺. Whereas T-type channels generally increase excitability of the plasma membrane and can mediate oscillations in membrane potential (Perez-Reyes 2003), their role in white bass HCs is unknown.

Notably, there is evidence of T-type channels in HCs of the rabbit (as discussed below) and the African clawed frog (Xenopus laevis) (Akopian et al. 1997). In Xenopus, they cause voltage oscillations in the light-to-dark transition (Akopian et al. 1997), but the function of these oscillations is unclear. In addition to T-type channels, both rabbit and Xenopus HCs alone show Ca²⁺-dependent K⁺ currents (Akopian and Witkovsky 1994; Lohrke and Hofmann 1994). In Xenopus, Ca²⁺-dependent K⁺ currents activated close to the activation potential of T-type channels (Akopian and Witkovsky 1994), leading the authors to speculate that T-type channels may modulate outward K⁺ currents. One is left to wonder whether a similar situation occurs in white bass, the only other species known to have T-type channels in HCs (Sullivan and Lasater 1992).

Pfeiffer-Linn and Lasater (1996) argued that certain properties of high voltage-activated Ca²⁺ channels in white bass separate them from typical L-type channels. These channels are like L-type channels in their activation potential, single channel conductance, and sensitivity to Bay K8644, but they are antagonized by ω-agatoxin IVA (which is generally considered to be selective for P/Q-type VGCCs) instead of L-type antagonists, such as nifedipine and nimodipine. This aberrant channel, which Pfeiffer-Linn and Lasater (1996) named a P/L-type channel, has not been found in other teleost species.

Ca²⁺ influx in teleost HCs: gap junctions and hemichannels.

HCs of a given type are electrically coupled to each other by gap junctions, forming a syncytium throughout the inner nuclear layer (Kaneko and Stuart 1984; Sakuranaga and Naka 1985). These channels are permeable to ions and molecules up to ∼1 kDa (Contreras et al. 2004), including fluorescent dyes. As a result, injection of dyes (such as Procion and Lucifer yellow) into individual HCs has provided detailed illustrations of HC syncytia (Baldridge and Ball 1991; Kaneko 1971; Kaneko and Stuart 1984; Stewart 1978). Gap junctions allow for Ca²⁺ diffusion between cells (D'Andrea and Vittur 1996; Höfer et al. 2001; Nielsen et al. 2012; Saez et al. 1989). In other vertebrate cell types that form syncytia—including cardiac myocytes (Boyden et al. 2015); developing rat, cat, and ferret neurons in the neocortex and inner retina (Katz 1993; Kotak et al. 2007; Wong 1993; Yuste et al. 1992); and rat retinal glial cells (Newman and Zahs 1997)—Ca²⁺ propagates as waves. Ca²⁺ activity has been found to be synchronized in HCs, up to 1 mm apart (Murakami et al. 1995). Ca²⁺ waves, however, have not been identified in HC syncytia.

Much work needs to be done to clarify which connexins (Cx) form either gap junctions or hemichannels on HCs. There is immunohistochemical evidence for Cx35/36 in goldfish HCs (Liu et al. 2009) and Cx43 in carp (Janssen-Bienhold et al. 1998) and both immunohistochemical and in situ hybridization evidence for Cx52.6 and Cx55.5 in zebrafish (ZfCx52.6 and ZfCx55.5, respectively) (Dermietzel et al. 2000; Shields et al. 2007; Zoidl et al. 2004). Additionally, both pannexin-1 (Px1)- and ZfCx26-based hemichannels have been found in zebrafish HC dendrites, where they have been theorized to assist in feedback to cones (Kamermans et al. 2001; Prochnow et al. 2009b). Many retinal Cx and Px transcripts have recently been cloned (Zoidl et al. 2008), suggesting that more may be found in HCs in the near future.

Functionally, electrical coupling through gap junctions provides HCs with a large receptive field size (Kaneko and Stuart 1984), whereas uncoupling decreases this receptive field size and reduces center-surround antagonism (Shen et al. 2003). A number of neuromodulators have been found to adjust the extent of electrical coupling. In fish, interplexiform cells from the inner nuclear layer of the retina release dopamine onto HCs, activating D1 receptors and partially uncoupling them through a cAMP-dependent pathway (Baldridge and Ball 1991; DeVries and Schwartz 1989; Lasater and Dowling 1985; Mangel and Dowling 1987; McMahon and Brown 1994). Nitric oxide has been found to uncouple HCs through a cGMP-dependent mechanism (Daniels and Baldridge 2011; DeVries and Schwartz 1989; Lu and McMahon 1997), and external retinoic acid has been proposed to downregulate gap junctions by binding to them directly (Zhang and McMahon 2000). Low pH has also been shown to diminish teleost HC gap-junctional conductance (DeVries and Schwartz 1989; Jonz and Barnes 2007) and change connexon morphology (Schmitz and Wolburg 1991).

Ca²⁺ modulates both gap-junctional currents and coupling. [Ca²⁺]_o is known to suppress the retinal hemichannel current (DeVries and Schwartz 1992; Jonz and Barnes 2007), but inward vs. outward hemichannel currents have a different IC₅₀ for Ca²⁺. Whereas the outward hemichannel current (IC₅₀ = 0.69 mM) is mostly blocked within the expected physiological range for [Ca²⁺]_o (1–2 mM), inward currents (IC₅₀ = 1.58 mM) are less sensitive to [Ca²⁺]_o (Sun et al. 2012). The increase of [Ca²⁺]_o has also been found to decrease the percentage of cells that are coupled in cultured giant danio (Devario aquipinnatus) HCs (McMahon and Mattson 1996). In contrast to the effects of [Ca²⁺]_o, increases in [Ca²⁺]_i have not been found to affect gap-junctional conductance (DeVries and Schwartz 1989).

Ca²⁺ influx in mammalian HCs: iGluRs.

Evidence from in situ hybridization, immunohistochemistry, and electrophysiological studies supports the presence of AMPARs in all mammalian species tested (Blanco and De La Villa 1999; Brandstätter et al. 1998; Hack et al. 2001; Hankins and Ikeda 1991; Haverkamp et al. 2000; Morigiwa and Vardi 1999; Qin and Pourcho 1999; Shen et al. 2004). As with teleosts, the case is less clear for NMDARs. An in situ hybridization study found mRNA for NR1 (an NMDAR subunit sufficient to make functional channels) throughout the inner nuclear layer of the rat retina and concluded that all cells in this layer likely express NMDARs (Brandstätter et al. 1994). By contrast, an immunohistochemical study did not find NR1 subunits in HCs; instead, BCs and amacrine cells accounted for all NR1 immunoreactivity in the inner nuclear layer (Fan et al. 2013). A similar controversy exists in cats. Whereas the outer edge of the cat inner nuclear layer showed immunoreactivity for another NMDAR subunit (NR2A), it was uncertain whether the labeling was for HCs or BCs (Hartveit et al. 1994). In adults of most other species, including the mouse, rat, rabbit, and monkey, there is no immunohistochemical evidence for NMDAR subunits in HCs at all (Brandstätter et al. 1998; Hartveit et al. 1994). Furthermore, all physiological reports have failed to find NMDARs in adult mammalian HCs (Kalloniatis et al. 2004; Massey and Miller 1987; Shen et al. 2004). Notably, these receptors may be present in the developing mammalian retina (Watanabe et al. 1994), which would suggest that any NR1 mRNA found in adults is a vestige of development (Brandstätter et al. 1998).

There has been more work to determine the particular iGluR subunits of the outer retina in mammals than in teleosts. There are four AMPAR subunits (GluR1-4). HCs labeled strongly with a GluR2/3 antibody in all mammals tested [cat (Morigiwa and Vardi 1999; Qin and Pourcho 2001), monkey (Haverkamp et al. 2000), mouse (Hack et al. 2001), rabbit (Pan and Massey 2007; Peng et al. 1995), rat (Hack et al. 2001), tree shrew (Peng et al. 1995)], and GluR2 labeling was found in mouse and rat (Hack et al. 2001). Similarly, GluR4 immunoreactivity has been found in all species tested (Hack et al. 2001; Haverkamp et al. 2000; Morigiwa and Vardi 1999; Pan and Massey 2007). GluR1 immunoreactivity is absent in most reports (Haverkamp et al. 2000; Morigiwa and Vardi 1999; Pan and Massey 2007; Peng et al. 1995; Qin and Pourcho 1999). Mammalian HC subtypes may differentially express iGluRs. In cat, GluR4 is present on A-type (but not B-type) HCs (Morigiwa and Vardi 1999). There is also one report on rabbits, supporting the existence of Ca²⁺-permeable AMPARs (Rivera et al. 2001). In this report, kainate and AMPA could elicit increases in [Ca²⁺]_i even after VGCC currents were abolished. However, contributions from other sources (e.g., mGluRs or Na⁺-Ca²⁺ exchange) were not examined and could not be ruled out.

For the KAR subunits, an antibody for GluK2/3 (previously called GluR6/7) (Lerma and Marques 2013) labels HCs in cat (Morigiwa and Vardi 1999; Qin and Pourcho 2001), rat (Brandstätter et al. 1997; Peng et al. 1995), and primate (Haverkamp et al. 2001) retinas. However, Blanco and De La Villa (1999) did not find evidence for KARs in electrophysiological experiments. Thus far, the only support for functional KARs in mammalian HCs is in humans (Shen et al. 2004).

Ca²⁺ influx in mammalian HCs: VGCCs.

Electrophysiological reports in cat, rat, mouse, and rabbit HCs suggest that most VGCC current is through L-type channels (Liu et al. 2013, 2016; Lohrke and Hofmann 1994; Rivera et al. 2001; Schubert et al. 2006; Ueda et al. 1992), although B-type (axonless) HCs in cats probably have lower L-type channel densities (Ueda et al. 1992). In contrast to teleosts, there is growing support for N- and P/Q-type channels in mammals. Subunits for L-, N-, and P/Q-type VGCCs have been found in the outer plexiform layer in the rat retina (Witkovsky et al. 2006; Xu et al. 2002) and were later confirmed in HCs (Liu et al. 2013; Witkovsky et al. 2006). Additionally, blockers for N- and P/Q-type VGCCs reduce HC Ca²⁺ currents in mouse and rat (Liu et al. 2013, 2016; Schubert et al. 2006). These blockers were not tested in cat (Ueda et al. 1992) or rabbit (Lohrke and Hofmann 1994; Rivera et al. 2001), and so one cannot rule out N- and P/Q-type channels in these species. Lee and Brecha (2010) proposed that VGCCs may be involved in vesicular GABA release in mammals after finding several proteins involved in Ca²⁺-dependent neurotransmitter release in guinea pig HCs.

Most reports did not find T-type channels on mammalian HCs (Schubert et al. 2006; Ueda et al. 1992). However, one report in cultured rabbit HCs (Lohrke and Hofmann 1994) identified a transient Ca²⁺ current suggestive of T-type channels. The current was insensitive to nifedipine, had an activation potential at more negative potentials than the L-type current, and peaked at −30 to −10 mV.

Ca²⁺ influx in mammalian HCs: gap junctions and hemichannels.

In mouse and rabbit HCs, Cx50 and Cx57 are present (Cha et al. 2012; Hombach et al. 2004; Shelley et al. 2006). Notably, rabbits have different Cx expression in each HC subtype: A-type contains Cx50, whereas B-type contains Cx57 (Cha et al. 2012). Functionally, this may have effects on the receptive field size. A-Type HCs have a larger receptive field than B-type, which is presumed to be due to larger gap-junctional plaques in A-type HCs and a greater unitary conductance in Cx50 compared with Cx57 (Cha et al. 2012). It is unlikely that these Cxs are directly involved in feedback to cones, as neither Cx has been localized to the triad synapse (Cha et al. 2012), and knockout mice for Cx57 still exhibit negative feedback (Shelley et al. 2006). However, Px1 hemichannels remain candidates for mediating feedback in teleosts and mammals alike. They have been found in HCs in rat and mouse (Dvoriantchikova et al. 2006) and have been localized to HC axons and dendritic tips (Kranz et al. 2013).

Ca²⁺ EXTRUSION AND BUFFERING MECHANISMS

Ca²⁺ extrusion in teleost HCs.

Because high levels of cytosolic Ca²⁺ are toxic to neurons (Choi 1992; Lipton 1999), sustained Ca²⁺ influx through iGluRs and VGCCs must be balanced by sequestration and efflux mechanisms to maintain cell viability. Although cytosolic buffering mechanisms and sequestration into intracellular Ca²⁺ stores are known in HCs (see below), these stores must be finite and would be replete with Ca²⁺ over time. Therefore, Ca²⁺ must be extruded across the plasma membrane to limit [Ca²⁺]_i. Two major Ca²⁺ extrusion mechanisms have been shown to exist in teleost HCs: 1) a plasma membrane Ca²⁺-ATPase (PMCA) and 2) an Na⁺/Ca²⁺ exchanger (NCX) (Kreitzer et al. 2007; Micci and Christensen 1998; Yasui 1987).

To demonstrate Na⁺/Ca²⁺ exchange, Hayashida et al. (1998) applied glutamate for >1 h so as to ensure that Ca²⁺ uptake and efflux reached equilibrium, both across the plasma membrane and in/out of intracellular Ca²⁺ stores. Changes in this new [Ca²⁺]_i would have therefore reflected a change in Ca²⁺ homeostasis. When extracellular Na⁺ was removed or when Li⁺ (which cannot functionally replace Na⁺ in NCX activity) was applied, [Ca²⁺]_i increased, suggesting the presence of NCX activity. Similarly, Na⁺/Ca²⁺ exchange was also inhibited by high [Ca²⁺]_o, which reduced the driving force for Ca²⁺ and made NCX exchange energetically unfavorable.

The first report of PMCAs in teleost HCs came from Micci and Christensen (1998), who found that the nonspecific P-type ATPase inhibitor, vanadate, greatly increased [Ca²⁺]_i in freshly dissociated cells. Vanadate can also inhibit Na⁺/K⁺-ATPases, which would increase intracellular Na⁺ concentration and therefore, the driving force for NCX activity. Yet [Ca²⁺]_i only slightly increased when oubain, a selective Na⁺/K⁺-ATPase inhibitor, was applied (Micci and Christensen 1998). Vanadate, however, has been shown to inhibit refilling of certain intracellular Ca²⁺ stores (Watson et al. 2003), so a vanadate-elicited increase in [Ca²⁺]_i may be caused by either sarco/endoplasmic reticulum Ca²⁺-ATPase or PMCA antagonism.

Further evidence for PMCAs on HCs comes from studies of proton efflux in skate, catfish, and goldfish. PMCAs exchange extracellular protons and intracellular Ca²⁺ at the cost of ATP, so that changes in extracellular pH may be related to PMCA activity (Di Leva et al. 2008). With the use of self-referencing proton electrodes, several studies found that the depolarization of isolated HCs with glutamate, kainate, high K⁺ solutions, or voltage clamp increased extracellular pH (Jacoby et al. 2012; Kreitzer et al. 2007, 2012; Molina et al. 2004). This extracellular alkalization was abolished with low Ca²⁺ solutions, Co²⁺, and nifedipine or with PMCA antagonists, such as carboxyeosin and lanthanum (Jacoby et al. 2012; Kreitzer et al. 2007, 2012). Furthermore, concurrent recordings showed that extracellular pH and [Ca²⁺]_i varied synchronously, suggesting an exchange mechanism (Kreitzer et al. 2012). Collectively, these studies suggest that Ca²⁺ influx from VGCCs leads to Ca²⁺ extrusion and simultaneous proton influx through PMCAs. Along with immunohistochemical labeling in the outer plexiform layer and in dissociated cells of the skate (Molina et al. 2004), these findings strongly support PMCA activity in teleost and skate HCs.

It is unclear how physiologically important PMCAs are in HCs. In one report in goldfish and carp, orthovanadate was found to increase [Ca²⁺]_i only marginally in HCs during prolonged glutamate application and had no effect in the absence of glutamate (Hayashida et al. 1998). This presents the possibility that the PMCA plays a minimal role in maintaining resting [Ca²⁺]_i in HCs and that NCX dominates Ca²⁺ extrusion in these cells. This may be a strategy to reduce ATP expenditure. HCs are constantly taxed with Ca²⁺ influx, and so PMCA activity would be energetically costly.

Ca²⁺ buffering in teleost HCs.

Very few studies have analyzed buffering in HCs; however, the limited information available suggests that teleost HCs have some of the highest buffering capacities known of any cell type. Ca²⁺ buffering (β) is often defined as the ratio of entering Ca²⁺ ions that bind to cytosolic buffers compared with the Ca²⁺ ions that remain free in solution. As an example, β has been estimated to be 100 in rat gonadotrophs so that for every 101 Ca²⁺ ions that enter the cytosol, only 1 Ca²⁺ ion would remain free (Tse et al. 1994). A review of neuronal Ca²⁺ buffering lists typical β values between 20 and 200, with nearly all values <300 (Matthews and Dietrich 2015).

Surprisingly, the only study to examine Ca²⁺ buffering in teleost HCs measured β at 2,500 as a lower limit (Solessio and Lasater 2002). The authors also showed that this extraordinarily high buffering capacity slowed increases in [Ca²⁺]_i. Although the VGCC current reaches a maximal value within milliseconds, the resulting increases in [Ca²⁺]_i have time constants >8 s. Functionally, this extraordinarily high Ca²⁺ buffering capacity may be neuroprotective by reducing free [Ca²⁺]_i during prolonged Ca²⁺ influx.

Ca²⁺ extrusion and buffering in mammalian HCs.

PMCA isoforms have been identified with immunohistochemistry in the mouse retina (Krizaj et al. 2002), in which PMCA1-3 colocalized with calbindin, a Ca²⁺-binding protein and HC marker. There was faint, diffuse immunolabeling for PMCA1 throughout HC somata and dendrites, in addition to weak immunolabeling for both PMCA2 and -3. PMCA2 and -3 were also found on HCs in the developing rat retina (Renteria et al. 2005).

Calbindin is found in a large number of mammalian species (Cuenca et al. 2002; Haverkamp and Wässle 2000; Kantor et al. 2016; Massey and Mills 1996; Pasteels et al. 1990). In addition, calretinin is found in a variety of mammalian species (Cuenca et al. 2002; Kantor et al. 2016; Pasteels et al. 1990), and parvalbumin stains HCs in dog (Hamano et al. 1990) and humans (Kantor et al. 2016). In teleosts, these Ca²⁺-binding proteins could not be labeled with immunohistochemistry (Hamano et al. 1990; Yazulla and Studholme 2002), despite physiological evidence for high Ca²⁺ buffering capacity (Solessio and Lasater 2002).

It has been suggested that high levels of calbindin are neuroprotective in HCs. In a rat model of retinal ischemia, most parvalbumin- and calretinin-positive retinal cells were shown to succumb to ischemia and reperfusion injury, whereas calbindin-positive HCs survived (Kim et al. 2010). Likewise, calbindin-positive cells in the ganglion cell layer were more resistant to ischemic insult than parvalbumin- and calretinin-positive cells (Kwon et al. 2005), and overexpression of calbindin reduced excitotoxic cell death elsewhere in the central nervous system (Fan et al. 2007; Yenari et al. 2001). Perhaps strong buffering with calbindin protects HCs against constant Ca²⁺ influx (Kim et al. 2010).

Ca²⁺ STORES IN HCs

Most studies on teleost Ca²⁺ stores have investigated ryanodine receptors.

Nonmitochondrial intracellular Ca²⁺ stores are often categorized by the involvement of either ryanodine receptors (RyRs) or inositol triphosphate receptors (IP₃Rs). Numerous Ca²⁺ imaging experiments provide strong support for RyRs in HCs, which mediate CICR in these neurons (Huang et al. 2004; Linn and Gafka 2001; Lv et al. 2014; Solessio and Lasater 2002). In contrast, there are only two reports that suggest that IP₃Rs exist in teleost HCs (both in catfish): specifically, 1) an immunohistochemical report (Micci and Christensen 1996) and 2) Ca²⁺ imaging experiments that showed a slow increase in [Ca²⁺]_i with IP₃ application (Linn and Gafka 2001). However, IP₃Rs do not elicit CICR in these cells (Linn and Gafka 2001); furthermore, CICR does not occur when ryanodine or dantrolene is used to block RyRs in vitro (Lv et al. 2014; Micci and Christensen 1998). The role of IP₃Rs in these cells is largely unexplored, and it is unknown if IP₃Rs exist in HCs of other teleosts.

In catfish, Ca²⁺ stores were shown to be sensitive to caffeine, a RyR agonist that elicited CICR and presumably depleted Ca²⁺ from stores (Micci and Christensen 1998). If it was applied twice within a short period of time, then the second caffeine pulse induced a smaller CICR transient, which was comparably smaller than the first caffeine pulse. The authors interpreted this to mean that stores were refilling with Ca²⁺ during the interim and estimated that the refilling of stores occurred after 2 min. Refilling was shown to be dependent on extracellular Ca²⁺, even in the presence of nifedipine, to block Ca²⁺ influx through VGCCs. Interestingly, while extrusion mechanisms are blocked, HC Ca²⁺ stores can accumulate enough Ca²⁺ to double CICR amplitudes (Micci and Christensen 1998). This suggests that stores are not replete with Ca²⁺ but have an unexpectedly large capacity to accommodate additional Ca²⁺. Because HCs can be depolarized for hours at a time and undergo constant Ca²⁺ influx, large-capacity stores might buffer against otherwise excitotoxic increases in [Ca²⁺]_i.

Interestingly, Micci and Christensen (1998) proposed that the NCX operates in reverse to refill stores. By inhibiting NCX activity with Li⁺, they found that subsequent caffeine application could not elicit CICR. Recently, additional evidence from bovine smooth muscle was presented to link NCX activity, in the reverse mode, to refilling of RyR-sensitive stores (Hirota et al. 2007); however, confirmation of this in teleost HCs awaits the use of selective NCX inhibitors.

Alternatively, it is possible that store-operated channels (SOCs) refill stores. In goldfish, it was shown that a SOC inhibitor (2-aminoethoxydiphenyl borate) prevented stores from refilling (Lv et al. 2014). Further research will need to clarify whether both mechanisms exist in both species and what their relative contributions are to refilling.

Functionally, CICR is known to occur immediately after glutamate application in HCs, followed by a sustained Ca²⁺ influx from VGCCs and iGluRs. Linn and Gafka (2001) found that Ca²⁺ from CICR is able to downregulate VGCCs. This downregulation was not seen when CICR was blocked (with ryanodine) or when stores were depleted (with caffeine). Conceivably, this would serve to decrease the sustained Ca²⁺ influx through VGCCs during glutamate application, preventing excitotoxic increases in [Ca²⁺]_i.

Ca²⁺ stores in mammalian HCs.

In mouse HCs, CICR is also triggered by caffeine (Schubert et al. 2006), suggesting the presence of RyRs. Molecular and immunohistochemical studies corroborate the presence of RyRs on HCs. An antibody for the cardiac RyR isoform, RyR2, labels HC somata in cow (Shoshan-Barmatz et al. 2005) and rabbit (Shoshan-Barmatz et al. 2007). However, RyR (from total retinal fractions) showed unique variations from its cardiac or skeletal muscle counterparts. It was Ca²⁺ independent, showed different kinetics, and was inhibited by caffeine (Shoshan-Barmatz et al. 2005). The unique properties of this isoform have not been explored in other species, and it is unclear if it plays a role in HC physiology. In contrast to teleosts, we found no reports of IP₃Rs in HCs of the mammalian retina.

Ca²⁺ APs IN HCs

Ca²⁺ APs in fish.

In teleosts (i.e., goldfish, carp, and catfish) and skates, isolated HCs are known to produce APs (Johnston and Lam 1981; Lasater et al. 1984; Murakami and Takahashi 1987; Tachibana 1981). There has been brief mention of similar regenerative membrane activity in HCs of white perch (Lasater 1986) and white bass (Sullivan and Lasater 1992). The physiological significance of APs in HCs is currently unknown.

Despite their all-or-nothing activity, these APs are quite unique. They persist in the absence of Na⁺ (Shingai and Christensen 1983) and are not abolished by Na⁺ channel blockers, such as tetrodotoxin (Tachibana 1981), although in some fish—such as skates (Malchow et al. 1990) and catfish (Shingai and Christensen 1983, 1986)—Na⁺ currents contribute to the rising phase of the AP. Instead, APs in HCs are primarily Ca²⁺ dependent. They are amplified with high [Ca²⁺]_o and diminished with low [Ca²⁺]_o (Tachibana 1981); blocked by Co²⁺, Mn²⁺, and Ni²⁺ (Johnston and Lam 1981); and enhanced and prolonged by VGCC-permeant ions, such as Ba²⁺ and Sr²⁺ (Tachibana 1981). The APs are also slow, lasting on the order of seconds to minutes (Dixon et al. 1993; Tachibana 1981).

APs have been described in HCs in a number of preparations, including the intact retina (Murakami and Takahashi 1987), cultured cells (Lasater et al. 1984; Tachibana 1981), and freshly dissociated cells (Shingai and Christensen 1983). Although all investigators could induce APs with depolarizing current pulses [e.g., Dixon et al. (1993); Lam and Ayoub (1983); Shingai and Christensen (1983)], some reports found them to occur spontaneously, especially with Ba²⁺ or K⁺ channel-blocker application (Murakami and Takahashi 1987; Tachibana 1981). Furthermore, both Ishida et al. (1984) and Lasater et al. (1984) initiated AP activity with glutamate application.

APs may be exploited to understand Ca²⁺ dynamics in HCs. There is evidence that APs coincide with extracellular acidification, possibly by activating the PMCA (Kreitzer et al. 2007, 2012). Conceivably, APs could change the [Ca²⁺] and pH of the cleft at the triad synapse, affecting HC feedback to PRs.

Ca²⁺ APs in other vertebrates.

APs occur in HCs of mammals. Ueda et al. (1992) found that A-type HCs of the cat showed spiking activity, whereas B-type cells did not. The authors demonstrated that APs were generated by VGCC current and that A-type HCs have a greater VGCC current density than B-type cells.

APs have also been recorded from HCs in eyecup preparations in other vertebrates. In the turtle (Pseudemys scripta elegans), HCs show spike-like depolarizations during the light-to-dark transition; furthermore, these spikes require extracellular Ca²⁺ and are suppressed by Co²⁺ (Akopian et al. 1991). In a similar eyecup preparation, Xenopus HCs were found to depolarize during the light-to-dark transition. This effect was enhanced by Ba²⁺ and Sr²⁺ and blocked by nifedipine (Akopian et al. 1997). This suggests that VGCCs may serve to repolarize HCs quickly to their dark membrane potential at light offset.

[Ca²⁺]_i AND FEEDBACK

GABA release can occur through Ca²⁺-dependent or Ca²⁺-independent mechanisms.

Ca²⁺ may be involved in negative feedback from HCs to PRs through a GABAergic mechanism. Both GABA and the enzyme that synthesizes it (glutamic acid dehydrogenase) have been found in HCs from fish and mammals (Brandon 1985; Deniz et al. 2011; Guo et al. 2010; Lam et al. 1980; Mosinger and Yazulla 1987; Mosinger et al. 1986), suggesting that at least some HCs are GABAergic. For teleosts, in particular, it has been proposed that only H1 cells accumulate and release GABA (Ayoub and Lam 1987; Lam et al. 1980).

Mechanisms of GABA release likely differ between mammalian and nonmammalian species. In mammals, there is growing support for a Ca²⁺-dependent, vesicular release of GABA from HCs (Cueva et al. 2002; Haverkamp et al. 2000; Lee and Brecha 2010; Liu et al. 2013; Thoreson and Mangel 2012), as occurs elsewhere in the nervous system (Wadel et al. 2007). Several essential components for vesicular GABA release have been localized to HCs (Cueva et al. 2002; Guo et al. 2010; Lee and Brecha 2010), including SNARE proteins and synaptotagmin-2, a Ca²⁺ sensor for exocytosis (Lee and Brecha 2010). In retinal slices from rat, HC VGCCs open during depolarization, allowing [Ca²⁺]_i to rise, which triggers vesicular GABA release (Liu et al. 2013).

In HCs of teleost and other nonmammalian species, only a small number of studies have found evidence for Ca²⁺-dependent release (Ayoub and Lam 1985; Lasater and Lam 1984). Instead, GABA release may occur through Ca²⁺-independent plasmalemmal GABA transporters operating in reverse [reviewed in Schwartz (2002)]. Both mechanisms (Ca²⁺-independent transporter and Ca²⁺-dependent vesicular release) might dominate in certain illumination conditions. Ayoub and Lam (1985) proposed a model in which mild depolarization would trigger Ca²⁺-dependent GABA release, whereas strong depolarization would rapidly exhaust the supply of GABA-containing vesicles while triggering transporter-dependent GABA release. Translated into light responses, this model suggests that GABA transport dominates in the dark, whereas Ca²⁺-dependent vesicular release dominates in dim lighting. There is no evidence for plasmalemmal GABA transporters in mammals (Feigenspan and Weiler 2004).

For mammals and other vertebrates alike, it is unclear whether PRs are the primary targets for GABA, since feedback has not been blocked by GABA antagonists in goldfish (Verweij et al. 1996), monkey (Verweij et al. 2003), or turtle (Thoreson and Burkhardt 1990). In the rat, a unifying mechanism for inhibitory feedback to PRs has been proposed in which GABA may modulate pH and/or ephaptic feedback (Liu et al. 2013). In this model, Ca²⁺-dependent GABA release would activate ionotropic GABA receptors on the HCs themselves. As GABA receptors are permeable to both Cl⁻ and HCO₃⁻, they could modulate cleft pH and shunt current into HCs (Liu et al. 2013). As further support, targeted deletion of HC vesicular GABA transporters in the mouse abolished inhibition to PRs (Hirano et al. 2016). Therefore, at least in rodents, Ca²⁺-dependent GABA release from HCs is important for negative feedback to PRs.

Ca²⁺-dependent synaptic plasticity may modulate feedback to PRs.

In teleost HCs, Ca²⁺ is necessary for rapid, synaptic plasticity at the PR synapse (Wagner and Djamgoz 1993). In light, HCs form spinules at their dendritic terminals in a matter of minutes. In darkness, these spinules retract. They have only been found in teleosts and are absent in mudpuppy, turtle, and mammals, including the dog, cat, rat, and monkey (Wagner 1980). Spinules may mediate feedback responses to PRs: as the number of spinules per synaptic ribbon increases, so does the strength of feedback-dependent, color-opponent responses in HCs (Kirsch et al. 1990; Weiler and Wagner 1984). As the retina is light adapted, improved feedback allows HCs to respond to increasingly higher frequencies of light stimuli (Lasater and Lam 1984; Prince et al. 1987), and this effect has been correlated with spinule formation (Djamgoz et al. 1988). These studies have led to the conclusion that spinules are candidates for the site of negative feedback from HCs to PRs (Wagner and Djamgoz 1993).

Spinule retraction is AMPAR dependent (Okada et al. 1999; Schmitz et al. 1995) and Ca²⁺ dependent (Okada et al. 1999; Schmitz et al. 1995; Weiler et al. 1995, 1996). There is also immunohistochemical evidence of heightened calmodulin levels in dendritic terminals in light, and calmodulin antagonists block spinule retraction in darkness (Schmitz et al. 1995; Weiler et al. 1995, 1996). Collectively, these studies suggest that Ca²⁺ enters Ca²⁺-permeable AMPARs during the light-to-dark transition and induces spinule retraction through a calmodulin-dependent pathway.

Ca²⁺ may also modulate synaptic plasticity through a separate pathway. In goldfish and carp, flickering light enhances monophasic, L-type HC voltage responses to red light (Hu et al. 2000). In other words, as light is repetitively flashed, L-type HCs hyperpolarize more over time (up to ∼10% more after 20 flashes of 500 ms duration). Expanding on this work, Huang et al. (2006) found that the injection of EGTA into HCs in situ led to a smaller increase in light-induced HC hyperpolarization. This increase did not occur if Ca²⁺-permeable AMPARs were selectively inhibited, if ryanodine was applied to block CICR, or if caffeine was preapplied to deplete Ca²⁺ stores. The authors concluded that [Ca²⁺]_i entry through Ca²⁺-permeable AMPARs activates CICR, which in turn, increases the HC hyperpolarizing response in the retina. The exact mechanisms of Ca²⁺-mediated changes in voltage responses have not been proposed. Functionally, this would be expected to increase feedback onto PRs, enhancing the perception of light, during and shortly after flickering.

[Ca²⁺]_i in HCs mediates positive feedback to PRs.

In addition to negative feedback, HCs can provide positive feedback to PRs. Such positive feedback has been found in multiple species, including the anole (Anolis segrei), tiger salamander (Ambystoma trigrinum), rabbit, and zebrafish (Jackman et al. 2011). The authors propose that positive feedback is spatially constrained so that it acts only upon glutamate-releasing PRs. This means that if a cone is already depolarized and releasing glutamate, HCs would further increase that cone's [Ca²⁺]_i and facilitate more glutamate release. This would counteract negative feedback and preserve the initial PR signal.

Whereas negative feedback is voltage dependent, positive feedback is dependent on [Ca²⁺]_i in HCs. Jackman et al. (2011) found that positive feedback can be elicited by glutamate release from PRs or by uncaging Ca²⁺ in HCs. Additionally, the Ca²⁺-permeable AMPAR antagonist, philanthotoxin-74, blocked positive feedback. From these findings, the authors conclude that Ca²⁺ must enter HCs directly through iGluRs. Once [Ca²⁺]_i rises in HCs, it is unclear how it mediates positive feedback. One possibility is that PMCAs could pump protons out of the synaptic cleft, as PMCAs have been shown to cause extracellular alkalization in a Ca²⁺-dependent fashion (Jacoby et al. 2012; Kreitzer et al. 2012). However, Wang et al. (2014) found that the depolarization of HCs in vivo acidifies, instead of alkalizes, the cleft. They suggest that PMCAs may not be on HC dendritic tips and so may not regulate cleft pH or positive feedback.

CONCLUSION AND FUTURE DIRECTIONS

Teleost HCs present an extreme of Ca²⁺ homeostasis in neurons and are therefore informative models for Ca²⁺ regulation. Whereas ion concentrations are presumed to be relatively constant in typical neurons, since they are seldom depolarized for long (Wright 2004), this assumption may not be made for the tonically depolarized neurons of the outer retina (Hayashida and Yagi 2002a). The understanding of how HCs respond to elevated and sustained Ca²⁺ influx promises to provide insight into Ca²⁺ physiology. Moreover, it may provide insights into pathological conditions, such as excitotoxicity, where excessive Ca²⁺ influx is central to neuronal death (Choi 1992).

Indeed, HCs are reputedly resistant to excitotoxicity (McMahon and Ponomareva 1996; Mizuno et al. 2010). The high Ca²⁺ buffering capacity (Solessio and Lasater 2002) and large capacity of HC stores (Micci and Christensen 1998) may allow for neuroprotection against excitotoxicity, despite the prolonged depolarization that these cells endure. The understanding of these and other potentially neuroprotective methods of handling sustained Ca²⁺ influx may lead to novel approaches in treating and understanding diabetic retinopathy, strokes, and other conditions characterized by excitotoxicity.

Teleost HCs have provided many unique insights into retinal and cellular physiology, especially as it pertains to Ca²⁺. Research on HCs has contributed to many important findings, including mechanisms of neurotransmitter release (Schwartz 1987), refilling of Ca²⁺ stores (Micci and Christensen 1998), and synaptic plasticity (Okada et al. 1999). Many questions still remain for the future and will surely improve our understanding of Ca²⁺ physiology, including the nature of HC Ca²⁺ APs, the regulation and multifaceted roles of Ca²⁺ stores, and the role of NMDARs in HCs. Perhaps most importantly, the exploration of HC Ca²⁺ dynamics may clarify feedback mechanisms in the outer retina, so as to improve our understanding of neural processing in the retina.

GRANTS

Support for M. G. Jonz was provided by the Natural Sciences and Engineering Research Council (NSERC) of Canada (Grant 342303).

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

M.W.C. prepared the figure; M.W.C. drafted the manuscript; M.G.J. and M.W.C. edited and revised the manuscript; M.W.C. and M.G.J. approved the final version of the manuscript.