The properties of Ca2+ clearance mechanisms in the IS determine the speed with which photoreceptor output responds to light (Duncan et al., 2006). Mechanisms that clear Ca2+ from IS cytosol consist of (1) plasma membrane Ca2+ adenosine triphosphatase (ATPase) transporters (PMCAs) (Krizaj et al., 2002; Duncan et al., 2006); (2) SERCA transporters (Krizaj et al., 2003, 2004; Krizaj, 2005b); and (3) mitochondrial Ca2+ uniporters (Krizaj et al., 2003) (Fig. 3).

Because both PMCAs and SERCAs are driven by metabolic energy in the form of adenosine triphosphate (ATP), Ca2+ homeostasis in photoreceptor ISs is closely linked to energy metabolism derived from glycolytic and oxidative phosphorylation pathways.

Steady-state [Ca2+]i amplitude and spatiotemporal characteristics of Ca2+ signals maintained by primary Ca2+ influx and clearance mechanisms are additionally shaped by (1) Ca2+-mediated Ca2+ channel inactivation (Corey et al., 1984; Rabl and Thoreson, 2002);

(2) powerful Ca2+-activated chloride and potassium conductances (Barnes and Hille, 1989; Cia et al., 2005; Xu and Slaughter, 2005); (3) glutamate transporter-mediated chloride conductances (Rabl et al., 2003); (4) Ca2+ buffering proteins such as calcium-binding protein (CaBP), parvalbumin, recoverin, and calbindin (Sokal et al., 2000; Haeseleer et al., 2004); (5) retrograde feedback from horizontal cells (Verweij et al., 1996, 2003; Hirasawa and Kaneko, 2003; Vessey et al., 2005); (6) protons released from synaptic vesicles or via metabolism (Kleinschmidt, 1991); (7) Ca2+/H+ exchange (Krizaj and Copenhagen, 1998); (8) chloride modulation (Thoreson et al., 2003); or (9) surface charge (Piccolino et al., 1999; Cadetti et al., 2004).

Ca2+-mediated signaling cascades are either confined to specific domains of the IS or are expressed at different densities across several IS regions. As a result, the amplitude and frequency modulation of light-evoked [Ca2+]i levels is highly region specific, allowing for tuning a wide array of Ca2+ signaling systems that use sensors with differing affinities located in different IS domains. For example, voltage-gated Ca2+ channels (VGCCs) and PMCAs are strongly expressed in the terminal and weakly in the ellipsoid region (Rieke and Schwartz, 1996; Nachman-Clewner et al., 1999; Morgans et al., 1998; Krizaj et al., 2002), resulting in large-amplitude, fast [Ca2+] onset and offset kinetics in the terminal (Fig. 4). Ellipsoid [Ca2+] levels are typically at least threefold lower than in the synaptic terminal due to low density of L-type VGCCs and mitochondrial Ca2+ uptake (Szikra and Krizaj, 2006). Ca2+ imaging studies suggested that there may be significant diffusion between the terminal and the cell body as well as between the cell body and the ellipsoid regions (Krizaj and Copenhagen, 1998; Steele et al., 2005; Szikra and Krizaj, 2006).

VOLTAGE-OPERATED CALCIUM CHANNELS PLAY A CENTRAL ROLE IN INNER SEGMENT CALCIUM REGULATION

The VGCCs represent the central element of Ca2+ regulation in the IS because they possess biophysical properties conferring susceptibility to voltage, heteromeric G proteins, Ca2+ released from stores, and Ca2+ buffers such as calmodulin. The VGCC is a multisubunit protein complex consisting of the principal α1 subunit (which forms the Ca2+ channel pore and senses the membrane potential) and auxiliary β− and α2δ-subunits. In mammalian rods, the α1-subunit belongs to the Cav1.4 (Cav1.4a1q; α1F) subtype transcribed from the L-type family CACNA1F gene (Bech-Hansen et al., 1998). Currents

through Cav1.4 channels are detected only when α1F subunits are coexpressed with α2δ1− and β2/β3-subunits, indicating that subunit interactions are critical for normal Ca2+ entry into the rod IS (Koschak et al., 2003; Baumann et al., 2004). Deletion of the β2-subunit (but not other β-subunits) alters the IS expression of α1F, resulting in loss of neurotransmission (Ball and Gregg, 2002; but see McRory et al., 2004). Visual deficits in patients with X-linked congenital stationary night blindness (CSNB2; Strom et al., 1998; BechHansen et al., 1998) are caused by mutations in the CACNA1F gene. The mutations underlying the CSNB2 phenotype encompass at least 75 variations. Of these mutations, 23 result in truncated α1-subunits and loss of function; in another 25 cases, various missense mutations were reported in which there still may be some residual function (Hoda et al., 2006). In some mutations, no change in gating was observed; rather, the expression of Cav1.4 was affected (McRory et al., 2004; Hoda et al., 2006), suggesting deficits in protein misfolding or trafficking. Cav1.4 knockout mice are characterized by decreased depolarization-evoked Ca2+ entry into photoreceptor terminals and failure to establish proper targeting/anchoring of presynaptic ribbon complexes in rods (Mansergh et al., 2005). Surprisingly, however, loss of the α1F-subunit resulted in total suppression of both rod and cone neurotransmission (Mansergh et al., 2005), challenging the belief that Cav1.4 are selectively localized to rods. In a recently described gain-of-function phenotype similar to CSNB2, there may be an increase in Ca2+ entry into rod ISs with similarly deleterious effects on the phenotype (Hemara-Wahanui et al., 2005).

A number of functional questions need to be addressed with respect to molecular interactions between IS VGCCs and cytosolic signals, extracellular modulators and the membrane potential. To maintain normal function in tonically depolarized cells such as photoreceptors, it is imperative to tame the positive feedback inherent in VGCC activation. A regenerative cycle between Ca2+ influx and depolarization would impede the light response by breaking the relationship between light stimuli and [Ca2+]i; it also would trigger oscillations of about 3–4Hz in the membrane potential and potentially overload the IS with Ca2+ (Akopian et al., 1997). Photoreceptor ISs employ several strategies to achieve this aim.

First, the physiologically relevant range of the I–V relationship is shifted toward negative potentials, minimizing the contribution of Ca2+ current to the steady-state “dark” membrane potential. Hence, in darkness, Ca2+ influx through VGCCs is small (Heidelberger et al., 2005) and amply compensated by counterion fluxes. The α1F-subunit-contain- ing channels in vivo activate at substantially more negative potentials than the large majority of other “high-voltage-activated” Ca2+ channels (Cox and Dunlap, 1992; but see Awatramani et al., 2005). Mammalian photoreceptor VGCC currents activate at −60 mV (primate: Yagi and MacLeish, 1994; tree shrew: Morgans et al., 1998; pig: Cia et al., 2005). Amphibian VGCC currents are first measurable at about −45 mV (Corey et al., 1984; Barnes and Hille, 1989); however, it remains to be determined whether small depolarizations to −60 mV increase [Ca2+]IS in amphibian photoreceptors without evoking detectable Ca2+ currents (Awatramani et al., 2005; Akopian et al., 1997; Bader et al., 1982). Effective voltages measured in vivo are 10–15 mV more negative than α1F-currents observed in heterologously expressing HEK cells (Baumann et al., 2004; McRory et al., 2004), suggesting that the voltage sensitivity and other gating characteristics of indigenous Ca2+ channels are modulated by cytosolic modulators. One such modulator is CaBP4, which shifts the activation curve of the

Cav1.4 channel to hyperpolarized potentials (Haeseleer et al., 2004). The significance of this modulation is indicated by the finding that loss of CaBP4 results in a 100-fold reduction of sensitivity of mouse rod and cone neurotransmission without affecting OS Ca2+ dynamics (Haeseleer et al., 2004; Maeda et al., 2005). The I–V relationship of IS VGCCs is also modulated by calmodulin, protons, negative feedback from postsynaptic cells, and metabotropic glutamate modulation (Barnes et al., 1993; Verweij et al., 1996; Hosoi et al., 2005; Higgs and Lukasiewicz, 2002).

Second, the membrane potential is stabilized at the threshold of VGCC activation through shunts provided by concomitant activation of powerful Ca2+-activated potassium and chloride conductances (Corey et al., 1984; Maricq and Korenbrot, 1988; Barnes and Hille, 1989; Xu and Slaughter, 2005). The delayed rectifier K+ current (reversal potential ~ −80 mV; Beech and Barnes, 1989) and Cl− (reversal potential near −20 mV; Thoreson et al., 2002) counterion fluxes are instrumental for shunting the membrane potential to prevent Ca2+ spiking (Akopian et al., 1997) and Na+ spiking (Kawai et al., 2005; Ohkuma et al., 2006) in rods and cones. Activation of chloride fluxes may also provide negative feedback from postsynaptic horizontal cells (e.g., Verweij et al., 1996, 2003) and directly modulate VGCCs (Thoreson et al., 2003).

Third, given that most L-type channels typically experience strong Ca2+-mediated inactivation, how is it possible for IS VGCCs to reliably transduce voltage changes into [Ca2+]i during prolonged depolarizations? All VGCC α1-subunits possess a conserved IQ motif that binds calmodulin, resulting in Ca2+-dependent inactivation due to Ca2+ entering the cytoplasm via channels themselves (Imredy and Yue, 1992), Ca2+ release from the ER (Adachi-Akahane et al., 1996), or release of Ca2+ from mitochondria (Hernandez-Guijo et al., 2001). Recent molecular and physiological studies found that α1F-subunits are much less susceptible to inactivation than α1-subunits in other types of L channels (Baumann et al., 2004; McRory et al., 2004), possibly because of a gating modulator within the C-terminal tail downstream of the IQ domain that regulates calmodulin binding (A. Singh et al., 2006). This inactivation mechanism is important for normal vision because men carrying the truncated IQ domain develop CSNB2 (A. Singh et al., 2006). Whereas little Ca2+-mediated inactivation is seen in heterologously expressed Cav1.4 (Koschak et al., 2003; Baumann et al., 2004; A. Singh et al., 2006), Ca2+-dependent inactivation is more prominent, if still modest, in intact cells (Corey et al., 1984; Rabl et al., 2003). This suggests that, in vivo, α1F inactivation is modulated by cytosolic components within the IS.

Voltage-operated Ca2+ entry into ISs is additionally regulated by kinases and phosphatases (Stella et al., 2001, 2002; Akopian et al., 2000; Zhang et al., 2005), protons (Barnes et al., 1993; Hirasawa and Kaneko, 2003), nitric oxide (NO; Kourennyi et al., 2004) and neuromodulators such as dopamine (D2 and D4 receptors), somatostatin (SSA2 receptors), ATP (P2X7 receptors), adenosine (A2 receptors), GABA (γ-aminobutyric acid), insulin, glutamate transporters, and glutamate itself (mGluR group III receptors) that act through heteromeric Gi, Go, and possibly Gq proteins (Stella et al., 2001, 2002; Thoreson et al., 2002; Akopian et al., 2000; Krizaj and Witkovsky, 1993, Higgs and Lukasiewicz, 2002; Koulen et al., 2005; Hosoi et al., 2005; Tatsukawa et al., 2005; Wersinger et al., 2006). Other potential candidates for VGCC modulation are presynaptic proteins such as bassoon (Khimich et al., 2005; tom Dieck et al., 2005), neurexins that link Ca2+ channels to presynaptic PDZ proteins (Missler et al., 2003) and neuroligins

and syndecans (Song et al., 1999). Finally, the high density of actin in all domains of the IS (Cristofanilli and Akopian, 2006) together with the susceptibility of retinal actin networks to Ca2+ modulation (Akopian et al., 2006) strongly suggest that Ca2+ signaling in the IS is an integral element of developing, remodeling, and regenerating ISs (e.g., Nachman-Clewner et al., 1999). The fascinating panoply of factors that have an impact on or are impacted by [Ca2+]i regulation is a direct indicator of the central role for this second messenger in IS function.

VGCCs that drive Ca2+ entry into the IS are selectively targeted to the synaptic terminal, with lesser expression in the cell body and the ellipsoid, as seen in mammalian (Raviola and Gilula, 1975; Morgans et al., 1998; Cia et al., 2005), reptilian (Schaeffer and Raviola, 1976; Lasater and Witkovsky, 1991; Choi et al., 2005), avian (Firth et al., 2001), and amphibian (Nachman-Clewner et al., 1999; Steele et al., 2005; Szikra and Krizaj, 2006) photoreceptors. The targeting mechanisms are unknown. However, VGCCs themselves are known to have a key role in the formation and function of the synapse. Loss of Cav 1.4 α1- or β2-subunits disrupted targeting of the synaptic ribbon to the active zone, suppressed exocytosis, and caused an absence of dendritic extensions of postsynaptic horizontal and bipolar cells into invaginations within the terminal (Ball and Gregg, 2002; Mansergh et al., 2005). This indicates that Ca2+ channels themselves regulate the assembly and trafficking of the synaptic ribbon to the photoreceptor active zone.

Finally, the dendrites of postsynaptic horizontal cells mediate retrograde synaptic interactions that control presynaptic [Ca2+]i (Verweij et al., 1996, 2003). Comparison of Boltzman functions fitted to Ca channel activation curves suggested that protonation of VGCCs exerts control over the gain of synaptic transfer by shifting the Ca current–voltage relationship (Barnes et al., 1993; Barnes, 1994; Hirasawa and Kaneko, 2003). The exquisite multiplicity of Ca2+ regulatory machinery organized around VGCCs shows that signal transmission at rod and cone synapses is highly modulatable.

Voltage-operated Ca2+ channels in amphibian and mammalian photoreceptors as well as heterologously expressed Cav1.4 channels are about sixfold potentiated by BayK 8644 and inhibited by dihydropyridine blockers such as nifedipine, isradipine, and nimodipine (Wilkinson and Barnes, 1996; Baumann et al., 2004; Szikra and Krizaj, 2006) and by the phenylkylamine verapamil (Cia et al., 2005). These characteristics confirm that CaV1.4 channels are members of the L-type VGCC family. However, sensitivity for dihydropyridine inhibition is uncharacteristically low (Hart et al., 2003; Cia et al., 2005), and some IS Ca2+ channels are reportedly susceptible to N-type Ca2+ channel antagonists such as ω-conotoxin (Wilkinson and Barnes, 1996; Kourennyi and Barnes, 2000). Unlike most L-type channels, heterologously expressed mammalian Cav1.4 subunits (Baumann et al., 2004) and endogenous mammalian VGCCs (Cia et al., 2005) are blocked by both L- and D-cis diltiazem enantiomers. This observation has potent therapeutic implications as diltiazems protect photoreceptors against Ca2+ overload and degeneration (e.g., Frasson et al., 1999; Takano et al., 2004; Vallazza-Deschamps et al., 2005).

Ca2+ Channels in Rods and Cones

There is a lot of evidence that rods and cones possess different mechanisms of Ca2+ influx and clearance, differential expression of modulatory receptors. as well as different intracellular cascades involving Ca2+ and other second messengers (Krizaj and Witkovsky,

Fig. 5. Store-operated Ca2+ channels in photoreceptors. Depletion of Ca2+ stores caused a large decrease in baseline [Ca2+]i, followed by a rebound overshoot, a diagnostic marker for storeoperated Ca2+ channels. The overshoot was blocked by the store-operated channel antagonist SKF 96365.

1993; Stella and Thoreson, 2000; Krizaj, 2000; Kourennyi et al., 2004; Zhang et al., 2005; Hosoi et al., 2005; Duncan et al., 2006). These differences are discussed at greater length in Krizaj and Copenhagen (2002). Three examples of compounds with opposite effects on rod and cone VGCCs involve dopamine (Krizaj and Witkovsky, 1993), adenosine (Stella et al., 2002), NO (Kourennyi et al., 2004), and glutamate (Hosoi et al., 2005). Depolarization-induced increases of calcium concentration in rods and cones are enhanced and inhibited, respectively, by the NO donor S-nitrosocysteine (Kourennyi et al., 2004). In contrast, glutamate acting through mGluR receptors suppresses VGCCs in amphibian cones but not rods (Hosoi et al., 2005). Finally, Ca2+ regulation in rods, but not cones, is characterized by large-scale Ca2+ release from ryanodine-based ER stores (Krizaj et al., 2003; Cadetti et al., 2006) (Fig. 5). Hence, rod and cone signals not only possess different timing and spectral sensitivity of the light response in the OS (that arise from expression of rod-specific and cone-specific phototransduction cascade elements), but also have a series of Ca2+ channels, transporters, and buffering systems in the IS that modulate, modify, and shape the final photoreceptor output, reflected in the rate of exocytosis in the synaptic terminal.

NEUROTRANSMISSION FROM RODS AND CONES

TO SECOND-ORDER RETINAL NEURONS

The IS represents a signaling bottleneck in which Ca2+ channels and transporters compress visual stimuli from the 4–5 log response range generated by the phototransduction cascade in the OS to about 0.5 log dynamic range of presynaptic [Ca2+]IS (Szikra and Krizaj, 2006) and transmitter release (Choi et al., 2005). The compression reflects a cumulative superposition of numerous positiveand negative-feedback loops inherent in biophysical and neuromodulatory properties of channels and transporters. Thus, transmission of graded light-evoked signals across the photoreceptor synapse is controlled and modulated by voltage-activated and Ca2+-activated mechanisms within the IS.

The exquisite accuracy of neurotransmission at photoreceptor synapses is preserved by the sensitivity of graded transmitter release to changes in presynaptic [Ca2+]i.

This release appears to be linearly proportional to [Ca2+]IS over the full range of presynaptic membrane potentials (Rieke and Schwartz, 1996; Witkovsky et al., 1997; Thoreson et al., 2004). The linearity of the relationship between Ca2+ influx and transmitter release in photoreceptors is difficult to understand because Ca2+ action in presynaptic terminals tends to be highly cooperative (Dodge and Rahamimoff, 1967; Goda and Stevens, 1994; Bollmann et al., 2000). It is possible that linearity is caused by linear summation of independent Ca2+ entry sites (e.g., Heidelberger et al., 2005). Alternatively, the linearity may result from linear activation of Ca2+ extrusion mechanisms during the light response because the PMCA pumps that regulate the light response are activated proportionally to [Ca2+]i (Zador and Koch, 1994; Duncan et al., 2006). Finally, the linear operating range of presynaptic [Ca2+]i might be sculpted by release of Ca2+ from intracellular stores (Cadetti et al., 2006) or Ca2+ entry through store-operated Ca2+- permeable channels (Szikra et al., 2006).

A series of recent findings showed that Ca2+ release from intracellular stores plays a crucial role in rod neurotransmission (Krizaj et al., 1999; Cadetti et al., 2006; Suryanarayanan and Slaughter, 2006). The predominant role of Calcium-Induced Calcium Release (CICR) at rod synapses is unparalleled in the central nervous system. CICR represents an amplification mechanism that boosts exocytosis in the darkness, when photoreceptors are depolarized. Consistent with this, CICR inhibitors most affect glutamate release during stimulation with dim flashes (Suryanarayanan and Slaughter, 2006), and synaptic gain is steepest around the dark membrane potential (see Witkovsky et al., 1997; Belgum and Copenhagen, 1988). The magnitude of Ca2+ release from stores is set by the “basal” cytosolic [Ca2+]i (Krizaj et al., 1999) as Ca2+; entering the IS through VGCCs triggers further release of Ca2+ from the ER by activating ryanodine receptor channels within the ER (Krizaj et al., 2003). Ryanodine receptors belong to the cardiac RyR isoform 2 family (Krizaj et al., 2004). CICR contributes most to synaptic [Ca2+]i during long depolarizations with Ca2+ fluxes that cross the threshold for ryanodine receptor activation. At these potentials, global elevation of [Ca2+]i measured in synaptic terminals (400–2000 nM; Szikra and Krizaj, 2006) triggers exocytosis at rates ranging from about 250 vesicles/s in lizard cones to about 400 vesicles/s in amphibian rods (Rieke and Schwartz, 1996; Schmitz and Witkovsky, 1997; Choi et al., 2005).

The capacitance method combined with Ca2+ imaging was used to measure exocytosis directly in isolated photoreceptors. The threshold for exocytosis when [Ca2+]i was measured with high-affinity indicators fura-2, fura-FF, and fluo-3 was under 1 M (Rieke and Schwartz, 1996; Thoreson et al., 2004). However, fura-2 cannot be used to estimate the actual Ca2+ concentration required for transmitter release as [Ca2+]i close to sites of Ca2+ entry can approach tens to hundreds of micromoles, well above the about 2 M saturation for fura-2 (Neher, 1995; Sinha et al., 1997; Llinas et al., 1992a,b). Studies using the low-affinity Ca2+ indicator furaptra at ribbon synapses suggested that exocytosis at ribbon synapses has a threshold of about 10 M (photoreceptors; Kreft et al., 2003) or 50 M (teleost bipolar cells; Heidelberger et al., 1994), consistent with the idea that vesicle release is determined by local [Ca2+]i microdomains. Similar lowand high-affinity mechanisms were reported in bipolar cells following use of lowand highaffinity indicators, respectively (Heidelberger et al., 1994; von Gersdorff and Matthews, 1996; Zhou et al., 2006).