Ryanodine stores and calcium regulation in the inner segments of salamander rods and cones

Transcription

1 J Physiol (2003), 547.3, pp DOI: /jphysiol The Physiological Society Ryanodine stores and calcium regulation in the inner segments of salamander rods and cones David Krizaj, F. Anthony Lai* and David R. Copenhagen Departments of Ophthalmology and Physiology, UCSF School of Medicine, San Francisco, CA 94143, USA and *Cell Signalling Laboratory, Wales Heart Research Institute, University of Wales College of Medicine, Cardiff CF14 4XN, UK Despite the prominent role played by intracellular Ca 2+ stores in the regulation of neuronal Ca 2+ homeostasis and in invertebrate photoreception, little is known about their contribution to the control of free Ca 2+ concentration ([Ca 2+ ] i ) in the inner segments of vertebrate photoreceptors. Previously, caffeine-sensitive intracellular Ca 2+ stores were shown to play a role in regulating glutamate release from photoreceptors. To understand the properties of these intracellular stores better we used pharmacological approaches that alter the dynamics of storage and release of Ca 2+ from intracellular compartments. Caffeine evoked readily discernible changes in [Ca 2+ ] i in the inner segments of rods, but not cones. Caffeine-evoked Ca 2+ responses in cone inner segments were unmasked in the presence of inhibitors of the plasma membrane Ca 2+ ATPases (PMCAs) and mitochondrial Ca 2+ sequestration. Caffeine-evoked responses were blocked by ryanodine, a selective blocker of Ca 2+ release and by cyclopiazonic acid, a blocker of Ca 2+ sequestration into the endoplasmic reticulum. These two inhibitors also substantially reduced the amplitude of depolarization-evoked [Ca 2+ ] i increases, providing evidence for Ca 2+ -induced Ca 2+ release (CICR) in rods and cones. The magnitude and kinetics of caffeine-evoked Ca 2+ elevation depended on the basal [Ca 2+ ] i, PMCA activity and on mitochondrial function. These results reveal an intimate interaction between the endoplasmic reticulum, voltage-gated Ca 2+ channels, PMCAs and mitochondrial Ca 2+ stores in photoreceptor inner segments, and suggest a role for CICR in the regulation of synaptic transmission. (Resubmitted 8 November 2002; accepted after revision 23 December 2002; first published online 24 January 2003) Corresponding author D. Krizaj: Department of Ophthalmology, UCSF School of Medicine, Box 0730, Room K-140, 10 Kirkham Street, San Francisco, CA , USA. krizaj@phy.ucsf.edu Cellular Ca 2+ homeostasis is a dynamic process that relies on a precise balance between Ca 2+ influx, sequestration, buffering and extrusion. Maintenance of steady-state [Ca 2+ ] i therefore requires concerted co-ordination of Ca 2+ channels, transporters and soluble Ca 2+ -buffering proteins (Pozzan et al. 1994; Berridge et al. 2000; Delmas & Brown, 2002). Most neurones utilize two main sources of Ca 2+ for initiating Ca 2+ -dependent processes within the cytoplasm: Ca 2+ entry across the plasma membrane via voltageand/or ligand-gated Ca 2+ channels and Ca 2+ release from internal stores. One important class of Ca 2+ stores belongs to the ryanodine receptor (RyR) family localized to the endoplasmic reticulum (ER). Ca 2+ release from these stores is often studied by using the methylxanthine compound caffeine (Neering & McBurney, 1984; Akaike & Sadoshima, 1989; Sitsapesan & Williams, 1990). Ca 2+ released from caffeine-sensitive stores has been shown to control a wide variety of neuronal processes, including development (Hong et al. 2000), exocytosis (Smith & Cunnane, 1996; Narita et al. 2000) and both short- and long-term synaptic plasticity (Llano et al. 2000; Emptage et al. 2001; Sabatini et al. 2001). Ca 2+ release from caffeine- and ryanodinesensitive stores typically amplifies the [Ca 2+ ] i changes induced by influx of Ca 2+ through voltage-sensitive Ca 2+ channels through the process of Ca 2+ -induced Ca 2+ release (CICR; McPherson et al. 1991; Friel & Tsien, 1992; Hernandez- Cruz et al. 1995; Verkhratsky & Shmigol, 1996). Photoreceptors can be divided into two general classes: (1) rods, which reliably signal single-photon absorptions, are relatively slow in response to light stimulation and exhibit a high sensitivity to light, and (2) cones, which are less sensitive to light, faster, more noisy and operate optimally in bright daylight. It has been suggested that many key differences in rod and cone function can be accounted for by different features of Ca 2+ homeostasis in these two classes of photoreceptor (Korenbrot, 1995; Fain et al. 2001; Krizaj & Copenhagen, 2002). For example, the dynamic range of Ca 2+ homeostasis in cone outer segments (OSs) may be more than three times greater than in the rods (Sampath et al. 1999). Rods and cones differ in the fraction of Ca 2+ in the current through cyclic nucleotide-

2 762 D. Krizaj, F. A. Lai and D. R. Copenhagen J Physiol gated (CNG) channels, Ca 2+ sensitivity of CNG channels, the rates of Ca 2+ clearance from the OS and the inner segment (IS) and the kinetics of synaptic transmission at rod and cone synapses (Schnapf & Copenhagen, 1982; Korenbrot, 1995; Krizaj & Copenhagen, 1998; Sampath et al. 1999; Ohyama et al. 2002). There are, however, significant gaps in our understanding of light-dependent changes in free Ca 2+ in rods and cones. Little, for example, is known about the contribution of intracellular compartmentalization and Ca 2+ buffering to [Ca 2+ ] homeostasis in photoreceptor ISs and OSs. Previously we have shown that caffeine-induced Ca 2+ release from internal stores regulates glutamate release from rod photoreceptor synaptic terminals (Krizaj et al. 1999). However, Ca 2+ release from intracellular stores in the rod OS was not addressed in that study, nor did we investigate internal Ca 2+ stores in cone ISs. Moreover, the mechanisms of Ca 2+ release from the ER and the mitochondria in the IS are not yet understood. This present study was designed to characterize the mechanisms underlying the release of Ca 2+ from internal compartments in both rods and cones. Our results suggest that Ca 2+ release via RyR-gated channels is tightly controlled by the uptake into the ER (via the sarcoplasmic endoplasmic Ca 2+ ATPases, SERCAs), by sequestration into mitochondrial Ca 2+ stores and by Ca 2+ extrusion across the plasma membrane via plasma membrane Ca 2+ ATPases (PMCAs). We also show that rods and cones exhibit marked differences in their responses to Ca 2+ -releasing agents such as caffeine, and provide evidence suggesting that signalling domains that link the plasma membrane to the ER differ between rods and cones. METHODS Preparation of isolated cells Larval-stage tiger salamanders (Ambystoma tigrinum) were decapitated and pithed using procedures approved by the UCSF Committee for Animal Care. Retinas were dissected from enucleated eyes, and cells were dissociated in 0 Ca 2+ /papain (7 U ml _1 ; Worthington, Freehold, NJ, USA) saline for 25 min at room temperature (20 22 C) and then plated onto coverslips coated with IgG/IgM (Jackson ImmunoResearch, West Grove, PA, USA) and the Sal-1 antibody (a kind gift from Dr Peter MacLeish) or 0.2 mg ml _1 Concanavalin A (Sigma, St Louis, MO, USA). All these procedures were performed in bright light. As a consequence, the photopigment in rod OSs was bleached. In the cone photoreceptors studied here, the OSs were removed during the dissociation procedure. Thus, under our experimental conditions, voltage changes in rod and cone ISs were unlikely to be influenced by photocurrents from the OSs. Coverslips were inserted into a perfusing chamber with a volume of 90 ml (Warner Instruments, Hamden, CT, USA). The chamber was superfused via a multi-inlet manifold (MP-8, Warner Instruments) allowing complete solution exchange within 10 s. The control saline solution contained (mm): 97 NaCl, 2 KCl, 2 CaCl 2, 2 MgCl 2, 10 Hepes, 20 glucose, 1 pyruvic acid, 2 lactic acid, 0.3 ascorbic acid and 1 glutathione at 240 mosmol l _1. The ph was adjusted to 7.6 with NaOH. [Ca 2+ ] measurement and data acquisition These methods are described fully elsewhere (Krizaj & Copenhagen, 1998). Briefly, photoreceptors were loaded with 3 5 mm fura-2 acetoxymethylester (fura-2 AM; Molecular Probes, Eugene, OR, USA) for 10 min and subsequently washed for 20 min. The fluorescence signals were acquired on an inverted microscope (Nikon Eclipse 200) using a dry w 40 objective (NA = 0.8). In all experiments, the signals were averaged over a region of interest (ROI) encompassing the perikarya (cell nucleus and the surrounding cytoplasm), the ellipsoid or the synaptic region of the IS; only the results obtained from the cell body are illustrated in this paper. Image acquisition was generally binned at 5 w 5 and was run at Hz by a cooled 12 bit digital CCD camera using a Kodak KAF1400 chip (PXL, Photometrics, Tucson, AZ, USA). The camera and the shutter (Lambda 10 2, Sutter Instruments, Novato, CA, USA) were controlled by commercial software (Metafluor 4.1; Universal Imaging, West Chester, PA, USA). Ratios between the 340 and 380 nm excitation wavelengths were calculated after subtraction of the background fluorescence. Free Ca 2+ levels were calibrated in vivo with 10 mm ionomycin using the standard relationship (Grzynkiewicz et al. 1985) and the K d for Ca 2+ binding to fura-2 was taken to be 224 nm (Grzynkiewicz et al. 1985). The calibration was not completed for all cells, therefore data for some cells are presented as 340/380 nm ratios; the relatively low acquisition rate of the CCD camera may have resulted in an underestimation of peak [Ca 2+ ] i in the calibrated cells. For confocal line-scan analysis, cells were loaded with 5 mm fluo-4 AM (Molecular Probes), incubated for 10 min and washed for another 10 min. Confocal images were collected in the line-scan mode using a LSM 5 Pascal confocal microscope (Zeiss, Jena, Germany) and a w 63 waterimmersion objective (NA = 0.8). Fluo-4 fluorescence was excited with the 488 nm band of the Ar laser with transmission set at 1%. The scan interval was ms. These cells were not calibrated and the data are presented on an intensity scale in arbitrary units (a.u.). Immunohistochemistry and confocal image acquisition Salamander eyecups were immersion-fixed for 1.5 h in 4 % (w/v) paraformaldehyde in phosphate buffer (PB; 0.1 M; ph 7.4). The formaldehyde was obtained as a 16 % solution in sealed ampoules from Electron Microscopy Sciences (Fort Washington, PA, USA). The retinas were rinsed three times in PB and cryoprotected in 30 % sucrose overnight at 4 C. Pieces of retina were mounted in OCT, sectioned vertically at 14 mm thickness on a cryostat, collected on Super-Frost Plus slides (Fisher, Pittsburgh, PA, USA) and stored at _20 C until use. For immunohistochemistry, retinal sections were washed in PB for 15 min then permeabilized and blocked in a solution containing 0.5 % Triton X-100 and 10 % goat serum. As secondary antibodies we utilized the Alexa 488 goat anti-rabbit IgG (H+L) conjugates (Molecular Probes, Eugene, OR, USA), diluted at 1:1000. After incubation, sections were washed in PB and mounted in Vectashield (Vector, Burlingame, CA, USA). Negative controls were performed for every set of experiments by omitting the primary antibody. Immunofluorescence and bright field Nomarski fields of view were obtained using a confocal microscope (Zeiss LSM 5 Pascal) at 10 % power for the 488 nm argon line and 100 % power for the 543 nm He/Ne line. Acquired images were processed with Adobe Photoshop (version 7.0) software. Caffeine, thymol, chlorocresol, ryanodine and cyclopiazonic acid (CPA) were obtained from Sigma (St Louis, MO, USA). SERCA antibodies were a generous gift from Professor Frank Wuytack (Katholieke Universiteit, Leuven, Belgium).

3 J Physiol Calcium signalling in rods and cones 763 RESULTS Caffeine releases more Ca 2+ from intracellular stores in rods than in cones Spatially averaged Ca 2+ signals from rod and cone ISs were measured in fura-2 AM-loaded rod and cone photoreceptors. In every cell, the region of interest was drawn around the perikaryal region of the IS comprising the cell nucleus and the surrounding cytoplasm. To test whether rods and cones possess intracellular Ca 2+ storage compartments, we exposed them to caffeine, a methylxanthine commonly used to stimulate Ca 2+ release from RyR-gated intracellular compartments (Neering & McBurney, 1984; Sitsapesan & Williams, 1990). Caffeine, at a concentration known to evoke half-maximal activation of Ca 2+ release from caffeinesensitive stores (10 mm; Akaike & Sadoshima, 1989; Uneyama et al. 1993), has been shown previously to increase [Ca 2+ ] in rod ISs (Krizaj et al. 1999). Our impression from these previous experiments was that caffeine was much less effective in cone ISs. Here, we systematically compared rod Figure 1. Caffeine evokes an increase in free [Ca 2+ ] i in the inner segment (IS) of rods but not cones a, simultaneous [Ca 2+ ] measurement from salamander rod and cone ISs loaded with the Ca 2+ indicator fura-2. Note that due to limited loading of the fura-2 ester into the small outer segment (OS) cytoplasmic volume, the OS Ca 2+ signal is not detected at the 340/380 nm excitation used in this particular experiment. The cell body and the ellipsoid of the rod and cone ISs, respectively, are marked with white arrows. b, 10 mm caffeine evoked an increase in [Ca 2+ ] i in the rod IS but not the cone IS. The [Ca 2+ ] i increase was prominent in the cell body (yellow arrowheads) and modest in the ellipsoid (red arrowhead). e, caffeine washout. f, subsequent exposure to 20 mm KCl raised [Ca 2+ ] i in both cells. The increase was again less pronounced in the ellipsoid regions of both cells (red arrowheads in g). h, [Ca 2+ ] i returned to baseline levels 2 min following KCL washout. Scale bar = 20 mm.

4 764 D. Krizaj, F. A. Lai and D. R. Copenhagen J Physiol and cone responses to caffeine. An example of caffeineinduced changes in [Ca 2+ ] i in rod and cone ISs is shown in Fig. 1. A rod and cone were imaged in the same field of view using a high-resolution CCD camera. The data represent a spatial average of the fluorescence signal encompassing the perikaryal compartment (the cell body). The cells were identified by the typical shape of their cell bodies and ellipsoids (Townes-Anderson, 1985; Krizaj & Copenhagen, 1998; Krizaj et al. 1999); the rod photoreceptor consists of a somatic and an ellipsoid compartment attached to a large cylindrical OS. As seen in Fig. 1, caffeine triggered a [Ca 2+ ] i elevation in the rod IS but not in the cone IS. The [Ca 2+ ] i increase was most prominent at the edge of the rod soma (Fig. 1b and c), from where it spread into the interior of the IS (Fig. 1c and d). Note the comparatively smaller [Ca 2+ ] i increase in the mitochondria-containing ellipsoid region of the rod IS (red arrowhead). To test whether both the rod and the cone possess operational voltage-gated Ca 2+ influx pathways, we subsequently depolarized the cells with 20 mm KCl. Exposure to high K + resulted in an elevation in [Ca 2+ ] i in the rod and in the cone IS. The magnitude of evoked [Ca 2+ ] i increase was again less in the ellipsoid compared with the rest of the IS (red arrowheads). These results indicate that Ca 2+, released from caffeine-sensitive stores, differs in the extent of its contribution to steadystate [Ca 2+ ] i in rods and cones. Furthermore, our results suggest that pronounced differences in Ca 2+ homeostasis exist between the cell body and the ellipsoid regions of the IS. Simultaneous optical recordings of [Ca 2+ ] i in a rod and cone IS are illustrated in Fig. 2 for a different rod cone pair. [Ca 2+ ] i is plotted as a function of time. Caffeine puffs were intertwined with puffs of high K + (stars) used to load the intracellular Ca 2+ stores. As illustrated in Fig. 2A and B, 10 mm caffeine evoked a transient [Ca 2+ ] i increase in the rod but not the cone IS. [Ca 2+ ] i in the rod did not stay elevated in the continued presence of caffeine, but returned back to the baseline with a time constant of 44 s. In a sample of 42 rods tested with fura-2, superfusion with 10 mm caffeine in control saline resulted in a 182 ± 31 nm increase in peak [Ca 2+ ] i, whereas caffeine convincingly raised [Ca 2+ ] i similarly in one out of 39 cone ISs tested. Caffeine-evoked responses were blocked by ryanodine, a specific antagonist of Ca 2+ release from caffeine-sensitive stores in rod ISs (Krizaj et al. 1999). Ryanodine (20 mm) elevated baseline [Ca 2+ ] i in rods by ~30 nm, but had no effect on cone [Ca 2+ ] i (n = 5; data not shown). These two results are consistent with the idea that ryanodine stores are much reduced or masked in cones. In order to test whether caffeine evokes fast changes in [Ca 2+ ] i that might be undetectable with the slower CCD system, we performed Ca 2+ measurements at high temporal resolution using the line-scan function of the laserscanning confocal microscope. In these experiments cells were loaded with the single-wavelength optical dye fluo-4. Caffeine evoked a [Ca 2+ ] i increase in all rods studied (29/29). In contrast, the majority of cones (38/44) showed no response to caffeine under these conditions (Fig. 2B). We did, however, detect small [Ca 2+ ] i elevations in response to caffeine in six cones. One illustrative example of caffeineevoked Ca 2+ increase in a cone IS is shown in Fig. 2C. The scan line was placed across the perikaryal regions of a rod and a cone IS simultaneously. Exposure to caffeine triggered a prominent [Ca 2+ ] i elevation in the rod and a slow, smallamplitude increase in the cone. Based on our measurements using high K + in fura-2-loaded cells, we estimate that caffeine-evoked [Ca 2+ ] i increases in cone ISs never exceeded Figure 2. Time course of caffeine effects on intracellular free [Ca 2+ ] in rod and cone ISs A, simultaneous measurement of [Ca 2+ ] from salamander rod and cone ISs using a CCD camera. Cells were loaded with fura-2 AM and stimulated with brief puffs of 90 mm KCl from a nearby pipette (indicated by the asterisks) followed by superfusion with 10 mm caffeine (indicated by the horizontal bar). Caffeine evoked a transient [Ca 2+ ] i increase in the rod, but not cone, IS. B, simultaneous line-scan measurement of Ca 2+ signals from salamander rod and cone ISs loaded with fluo-4 AM. Caffeine (10 mm) evoked a large fluorescence increase in the rod IS but did not change the fluorescence in the cone. C, caffeine triggered an increase in [Ca 2+ ] i in a subset of cone ISs. Simultaneous recording of fluo-4 fluorescence from rod and cone IS. Caffeine (10 mm) triggered a relatively rapid large-amplitude increase in the rod Ca 2+ signal. Concomitantly, a small Ca 2+ signal with a slow rise time was observed in the cone IS. a.u. = arbitrary units.

5 J Physiol Calcium signalling in rods and cones nm. To rule out the possibility that caffeine stimulated Ca 2+ influx rather than release from intracellular stores, we repeated the caffeine experiment in the absence of extracellular [Ca 2+ ]. We found that, as shown earlier (Krizaj et al. 1999), caffeine stimulated Ca 2+ release in rod ISs superfused with 0 Ca 2+ and 3 mm EGTA (n = 5/11). Caffeine also evoked a small [Ca 2+ ] i increase in one out of 12 cone ISs (data not shown). These results indicate that Ca 2+ release from caffeine-sensitive stores in rod ISs has a much larger effect on global cytoplasmic [Ca 2+ ] i than in cone ISs. Caffeine-sensitive stores are not detectable in rod OSs A key structural element of the rod OS are the stacks of intracellular cisternae ( rod disks ) that fill most of the OS cytoplasm. The OS cisternae are thought to accumulate and release Ca 2+ (Liebman 1974; Fain & Schroder, 1990); however, the mechanism of Ca 2+ release from the cisternae has not been established (Krizaj & Copenhagen, 2002). To determine whether there was any evidence for caffeinesensitive stores in light-adapted rods, we recorded [Ca 2+ ] i from ROIs in rod ISs and OSs. Figure 3A shows a simultaneous optical recording from a rod IS and OS. Stimulation with caffeine caused a large elevation in the IS [Ca 2+ ] i but had little effect on the OS [Ca 2+ ] i. Rod OSs never showed caffeine-evoked [Ca 2+ ] i responses (n = 8). Similarly, caffeine had no effect on [Ca 2+ ] i in OSs detached from the rest of the cell (n = 3; data not shown). A similar absence of caffeine-mediated [Ca 2+ ] i increase was obtained in line-scan experiments with the confocal microscope (n = 3/3 OSs/ISs). A dissociated rod with an intact OS was loaded with fluo-4 (Fig. 3B). The IS exhibited a high resting fluorescence, which increased markedly following exposure to caffeine (Fig. 3C). No fluorescence increase was observed in the OS. Note also that Ca 2+, released in the IS during caffeine stimulation, did not enter the OS, presumably due to the large diffusional barrier between the two regions of the cell. Although we observed large differences in caffeinesensitive responses between ISs and OSs, we note that these experiments were performed under circumstances when all rod pigment was bleached, and that we might have missed a potential contribution of caffeine-sensitive stores that might occur in dark-adapted rods under in vivo conditions. These experiments do demonstrate that Ca 2+ released from the IS Ca 2+ stores is excluded from the OS cytoplasm. Ca 2+ released from caffeine-sensitive stores is rapidly extruded by PMCAs During continuous superfusion with caffeine, rod IS [Ca 2+ ] i fell back to the baseline level, with an average time constant of 44±5s (n = 25). We investigated the mechanism underlying the time course of this [Ca 2+ ] i decay in the presence of caffeine. One possibility is that a negative feedback mechanism may cause RyRs to inactivate at high local [Ca 2+ ] i (Bezprozvanny et al. 1991; Györke & Fill, 1993). This is unlikely to occur under our experimental conditions because we showed earlier that raising [Ca 2+ ] i increases the amount of Ca 2+ released from the stores (Krizaj et al. 1999). Alternatively, if significant amounts of caffeine-released Ca 2+ are rapidly extruded across the plasma membrane rather than re-accumulated into the ER, caffeine-sensitive stores may gradually become depleted during prolonged exposure to caffeine, leading to a gradual decline in caffeine-evoked changes in [Ca 2+ ] i. Thus, the amount of Ca 2+ extruded from the cell should regulate both the rate of decay of caffeine-induced [Ca 2+ ] i transients and the recovery of responsiveness to caffeine. Figure 3. Simultaneous [Ca 2+ ] i measurement from an IS and OS of a rod A, rod loaded with fura-2 AM. The cell was depolarized by 20 mm KCl to maximize store loading. Caffeine (10 mm) transiently elevated [Ca 2+ ] i in the IS, but not the OS. B, confocal image of a dissociated rod photoreceptor loaded with 5 mm fluo-4. Prominent fluo-4 fluorescence is observed in the IS of the cell, whereas the OS signal is weak. C, line-scan across the IS (top) and the OS (bottom) of the cell shown in B. The abscissa represents the time axis. Exposure to 10 mm caffeine caused a large saturating increase in the fluo-4 signal in the IS but caused no fluorescence change in the OS. Note that the caffeine-evoked Ca 2+ signal is confined to the IS.

6 766 D. Krizaj, F. A. Lai and D. R. Copenhagen J Physiol The sole Ca 2+ extrusion mechanism in the plasma membrane of salamander rod and cone ISs is the PMCA (Krizaj & Copenhagen, 1998; Krizaj et al. 2002). To test whether PMCAs regulate the recovery of [Ca 2+ ] i in the continuous presence of caffeine, we superfused the cells with 1 mm La 3+, a potent blocker of PMCAs (Milanick, 1990; Krizaj & Copenhagen, 1998) and voltage-sensitive Ca 2+ channels (Reichling & MacDermott, 1991). In the presence of La 3+, caffeine evoked a sustained elevation of [Ca 2+ ] i that lasted as long as the caffeine step itself and was fully reversible (n = 25; Fig. 4). These findings implicate a role for PMCA in clearing store-released Ca 2+ in rod ISs. Inhibition of PMCA unmasks a caffeine-sensitive store in cones Interestingly, exposure to La 3+ uncovered a response to caffeine in cones. Figure 5 illustrates a cone IS challenged with caffeine in control solution and subsequently in the presence of La 3+. During exposure to La 3+, caffeine evoked a pronounced [Ca 2+ ] i elevation in this cell. In 10 La 3+ -treated cones, caffeine raised [Ca 2+ ] i by 85 ± 19 nm, suggesting that both rods and cones possess caffeine-sensitive Ca 2+ stores. We propose that the absence of a caffeine-evoked response in cones in control saline is not due to the absence of caffeine-sensitive intracellular stores, but that extrusion of released Ca 2+ by the PMCAs is more rapid in the cones (e.g. Krizaj & Copenhagen, 1998). Consequently, the released Ca 2+ is not detected by fura-2. RyRs are expressed in salamander photoreceptors The Ca 2+ -release channels targeted by caffeine are generally thought to belong to the RyR family. To determine whether RyRs are localized to salamander photoreceptors, we immunostained salamander retinal sections with an antibody against RyRs. This pan-ryr antibody recognizes RyR isoforms expressed in the whole brain, cerebellum and skeletal and cardiac muscle (Mackrill et al. 1997; Llano et al. 2000). To examine the subcellular localization of RyRs, we also enzymatically dissociated and fixed photoreceptors from the salamander retina and immunostained them with the RyR antibody. Figure 6B and F shows that RyRs are expressed in both rod and cone ISs. Colocalization with the photoreceptor synaptic marker SV2 (Krizaj et al. 2002) showed that RyR immunostaining was especially pronounced in the synaptic terminals (Fig. 6D and H). Immunostaining was also observed in the ellipsoid region of the IS, whereas moderate labelling was observed in the perikarya of rods, single cones and double cones. These results are consistent with results obtained from our physiological experiments. Figure 6I L shows a confocal image of a transverse section of the salamander retina immunolabelled with antibodies against RyRs and SV2. RyRs are expressed at high density in the outer nuclear and outer plexiform layers of the salamander retina (Fig. 6L). Similar to the isolated cells, the synaptic terminals and the cytoplasm surrounding perikarya and ellipsoids of ISs of rods and cones were all labelled. RyRs were also found in other retinal cell types. Pronounced immunolabelling was observed in Müller cells (arrowheads in Fig. 6J), consistent with a previous Ca 2+ imaging report suggesting that these cells possess functional RyRs (Keirstead & Miller, 1995). The RyR antibody also labelled calbindin-immunopositive horizontal cells, which are known to possess caffeine-sensitive Ca 2+ stores (Micci & Christensen, 1998; not shown). These results indicate that RyRs are selectively expressed in the retina, with a prominent localization to photoreceptor ISs. Caffeine-evoked increases in [Ca 2+ ] i are blocked by SERCA inhibitors The refilling of ER stores is commonly mediated by SERCAs, a family of Ca 2+ ATPases distributed widely over Figure 4. Plasma membrane Ca 2+ ATPases (PMCAs) shape the decay of caffeine-evoked [Ca 2+ ] i transients A rod IS was superfused continually with 1 mm La 3+, a blocker of PMCAs, and then stimulated with steps of 10 mm caffeine. In the presence of La 3+, caffeine evoked sustained [Ca 2+ ] i elevations. Following caffeine removal, [Ca 2+ ] i returned to baseline in the continued presence of La 3+. Figure 5. Inhibition of PMCAs unmasks a caffeine-evoked [Ca 2+ ] i response in the cone IS A cone IS that was stimulated with caffeine in control saline showed no [Ca 2+ ] i response. Subsequently, putative caffeinesensitive stores were replenished with a step of 20 mm KCl and the solution was switched to 1 mm La 3+. In the presence of La 3+, caffeine evoked a sustained increase in [Ca 2+ ] i.

7 J Physiol Calcium signalling in rods and cones 767 ER membranes (Pozzan et al. 1994; Guerini & Carafoli, 1998). SERCA pumps are selectively and reversibly blocked by CPA, a mycotoxin isolated from Aspergillus and Penicillium (Thomas & Hanley, 1994). To test whether SERCAs mediate Ca 2+ sequestration into the rod ER, we exposed these cells to CPA and measured caffeine responses before, during and after that exposure. Figure 7 shows [Ca 2+ ] i from two simultaneously recorded rods. The dark trace shows the response to a caffeine-filled pipette positioned close to the IS of this cell (rod 1). [Ca 2+ ] i in the Figure 6. Ryanodine receptors (RyRs) are localized to photoreceptor ISs Confocal fluorescence images of retinal cells immunostained with antisera against RyRs and SV2. A and E, Nomarski images of rod and double cone ISs dissociated from the salamander retina. B and F, RyR immunofluorescence is prominent in the synaptic terminal and the ellipsoid region. A moderate RyR signal is also observed in the plasma membrane surrounding the perikaryon. C, G and K, SV2 immunolabels synaptic terminals of retinal neurons. D and H, RyR and SV2 signals colocalize in the synaptic terminals of the rod and the cone. I, Nomarski image from a salamander retinal section. J, the RyR antibody labels photoreceptor perikarya and synaptic terminals. A prominent signal is also observed in Müller cell bodies and processes (arrowheads), in ganglion cell bodies and in synaptic processes in the inner plexiform layer (IPL). Little staining is seen in photoreceptor OSs and in bipolar cell bodies. L, RyR and SV2 colocalize in synaptic processes of the outer plexiform layer (OPL) and the IPL. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bars are 10 mm in A H and 50 mm in I L.

8 768 D. Krizaj, F. A. Lai and D. R. Copenhagen J Physiol cell was transiently elevated by 250 ms puffs of 50 mm caffeine. As illustrated in Fig. 7, caffeine-evoked responses in rod 1 were reversibly suppressed by CPA, consistent with the hypothesis that blocking the SERCA pumps depletes the caffeine-sensitive store. We found that in most cells, CPA alone increased the basal [Ca 2+ ], ostensibly as a result of blocked Ca 2+ sequestration into the stores. This is seen clearly in a rod not exposed to caffeine (rod 2; dotted trace) that exhibited the CPA-evoked elevation in [Ca 2+ ] i. In five cells exposed to CPA, [Ca 2+ ] i rose by 21 ± 1 nm. The CPA-mediated [Ca 2+ ] i increase is consistent with an inhibition of tonic Ca 2+ sequestration into the ER mediated by SERCA pumps. In the presence of CPA, [Ca 2+ ] i declined progressively towards the baseline level (Fig. 7), presumably reflecting a gradual depletion of Ca 2+ stores (Thomas & Hanley, 1994; Lo & Thayer, 1995; Usachev et al. 2001). Similar results were also seen in four cells superfused with 1 mm thapsigargin, another SERCA pump inhibitor (data not shown). Preliminary experiments using antibody staining against all three SERCA isoforms in the salamander retina suggest that salamander photoreceptors express the SERCA 2b isoform (D. Krizaj & D. R. Copenhagen, unpublished results). The refilling of caffeine-sensitive stores depends on the magnitude of the conditioning stimulus Previously, we have shown that the amplitude of caffeineevoked Ca 2+ transients depended on the magnitude of the steady-state [Ca 2+ ] i resulting from Ca 2+ influx via voltagegated Ca 2+ channels (Krizaj et al. 1999). We next addressed the issue of Ca 2+ release from stores when plasma membrane Ca 2+ channels are closed. Specifically, we examined the relationship between the filling status of Ca 2+ stores and the magnitude of [Ca 2+ ] i transients evoked by caffeine. The stores were filled to different extents by conditioning steps of 90 mm KCl (Fig. 8, asterisks). Each KCl puff transiently elevated [Ca 2+ ] i ; longer step durations raised the [Ca 2+ ] i level higher. The range of KCl-evoked Ca 2+ transients was nm. To test whether the refilling of stores is graded with the magnitude of [Ca 2+ ] i increase, we waited until [Ca 2+ ] i returned to baseline. We then measured [Ca 2+ ] i rises in response to 10 mm caffeine. A representative experiment is illustrated in Fig. 8. The first conditioning KCl-evoked Ca 2+ transient elevated [Ca 2+ ] i by 20 nm. Subsequent exposure to caffeine raised [Ca 2+ ] i by 25 nm (i.e. ~50 % of baseline level). Larger [Ca 2+ ] i transients result in the larger amplitude of subsequent caffeine-induced responses. The final conditioning transient elevated [Ca 2+ ] i to ~800 nm, and the caffeine-evoked response was ~120 nm. This result suggests that the amount of Ca 2+ sequestered into Ca 2+ stores depends on the magnitude of the immediately prior cytoplasmic [Ca 2+ ] transients. Similar results were observed in four cells. Ca 2+ is released from the ER via CICR The predominant physiological trigger for Ca 2+ release from neuronal ryanodine-sensitive stores is thought to be Ca 2+ itself (i.e. CICR). According to the CICR hypothesis, Ca 2+ changes evoked by Ca 2+ influx through voltage-gated Ca 2+ channels are amplified by Ca 2+ release from neuronal intracellular stores (Hua et al. 1993; Verkhratsky & Shmigol, 1996). We tested whether such amplification occurs in rods by evoking increases in [Ca 2+ ] i with test depolarizations in Figure 7. Ca 2+ is sequestered into intracellular stores by sarcoplasmic endoplasmic Ca 2+ ATPases Simultaneous recordings from two rod ISs. A puffer pipette containing 50 mm caffeine was positioned close to rod 1; rod 2 was located at the opposite side of the coverslip. Brief puffs of caffeine evoked transient increases in [Ca 2+ ] i in rod 1, but had no effect on rod 2. In both cells, cyclopiazonic acid (CPA) by itself evoked a slow elevation in [Ca 2+ ] i by several tens of nanomoles. Figure 8. The magnitude of caffeine-evoked Ca 2+ release depends on the magnitude of the conditioning depolarization A rod photoreceptor IS exposed to conditioning steps of high K + followed by superfusion with 10 mm caffeine. Each asterisk represents a single 128 ms KCl puff; bars denote superfusion. Increasing the duration of the conditioning [Ca 2+ ] i step caused a subsequent increase in the caffeine-evoked [Ca 2+ ] i increase. The caffeine-evoked responses saturated when the conditioning free [Ca 2+ ] i reached ~500 nm.

9 J Physiol Calcium signalling in rods and cones 769 the absence and presence of pharmacological agents that deplete the stores (caffeine), or block either release from the stores (ryanodine) or sequestration into the stores (CPA). Short puffs of KCl evoked transient increases of [Ca 2+ ] i in ISs. These rose to several hundred nanomolar in less than 1 s, the fastest interval we could record, and returned to baseline in tens of seconds (Fig. 9). Previous work showed that KCl-evoked [Ca 2+ ] i increases disappeared in Ca 2+ -free saline and were reversibly reduced by % by 1 2 mm nifedipine, a dihydropyridine antagonist of L-type Ca 2+ channels (11/11 rods and 3/3 cones, data not shown; Corey et al. 1984; Krizaj & Copenhagen, 1998; Nachmann- Clewner et al. 1999). The findings show that Ca 2+ influx through voltage-gated Ca 2+ channels is necessary for depolarization-evoked Ca 2+ transients. It is not clear, however, whether this Ca 2+ influx can trigger further release of Ca 2+ from intracellular stores via CICR. To examine whether Ca 2+ release from caffeine-sensitive stores contributed to the magnitude of depolarizationevoked Ca 2+ transients, high K + was puffed onto cells before and during the depletion of Ca 2+ stores with caffeine, CPA or ryanodine. When Ca 2+ stores in rod ISs were depleted during continuous superfusion of cells with 10 mm caffeine (n = 4), 5 mm CPA (n = 4) or 20 mm ryanodine (n = 2), the magnitude of KCl-evoked Ca 2+ transients was markedly reduced. Figure 9B illustrates a rod in which KCl-evoked responses were reduced by more than 50 % by caffeine. Likewise, ryanodine and CPA dramatically reduced the amplitude of KCl-evoked [Ca 2+ ] i transients (Fig. 9B and C). Cones responded similarly. As illustrated in Fig. 9D, CPA transiently elevated baseline [Ca 2+ ] i in cones and caused a reduction in the magnitude of KCl-evoked Ca 2+ transients (n = 3). Note that caffeine Figure 9. Ca 2+ -induced Ca 2+ release contributes to depolarization-evoked [Ca 2+ ] i increases in both rod and cone ISs A, a rod IS in which [Ca 2+ ] i was raised periodically with 128 ms puffs of 90 mm KCl. Superfusion with 10 mm caffeine itself raised [Ca 2+ ] i by ~100 nm. However, the magnitude of the subsequent KCl puff was substantially reduced in the presence of caffeine. B, a rod IS in which exposure to CPA slightly elevated the [Ca 2+ ] i baseline and significantly reduced the magnitude of KCl-evoked [Ca 2+ ] i transients. Following recovery, caffeine itself evoked a normal [Ca 2+ ] i response. C, a rod IS in which 20 mm ryanodine irreversibly decreased the magnitude of KCl-evoked [Ca 2+ ] i transients. D, a cone IS in which CPA caused a transient elevation of [Ca 2+ ] i and reversibly reduced the magnitude of KCl-evoked transients. Note that whereas KClevoked responses recovered following CPA washout, subsequent superfusion with 10 mm caffeine did not result in elevation of [Ca 2+ ] i.

10 770 D. Krizaj, F. A. Lai and D. R. Copenhagen J Physiol itself evoked little [Ca 2+ ] i change in the cone IS, presumably because most of the released Ca 2+ was extruded by PMCAs or sequestered into the mitochondria (e.g. Figs 5 and 11). These results provide evidence that Ca 2+ release from ryanodine-sensitive Ca 2+ stores in rod photoreceptors occur via CICR, and provide evidence suggestive of the involvement of Ca 2+ stores in buffering and sequestration of cytoplasmic Ca 2+ during depolarizationinduced [Ca 2+ ] i signalling in rod and cone ISs. Caffeine-insensitive stores are a component of intracellular Ca 2+ stores in photoreceptors We estimated the amount of Ca 2+ stored in non-caffeinesensitive intracellular compartments. In order to empty the caffeine-sensitive stores, we first superfused cells with 10 mm caffeine in the absence of external Ca 2+. To visualize the amount of Ca 2+ stored in other intracellular compartments, we then exposed cells to 10 mm ionomycin. Ionomycin, which is commonly used to calibrate [Ca 2+ ] i, is an ionophore that selectively shuttles Ca 2+ across lipid bilayers. In addition to equilibrating Ca 2+ across the plasma membrane, ionomycin molecules insert themselves into the membranes of intracellular compartments, causing an equilibration of their Ca 2+ with the cytosol. In the rod shown in Fig. 10, in 0 Ca 2+ /EGTA saline, exposure to caffeine raised [Ca 2+ ] i by ~290 nm; the cell was superfused with caffeine for 2.2 min so as to completely deplete the caffeinesensitive store. Subsequent exposure to ionomycin raised [Ca 2+ ] i by 800 nm, which was followed by a [Ca 2+ ] i decline to the very low level imposed by the 0 Ca 2+ saline. In 69 rods, ionomycin raised [Ca 2+ ] i by 354 ± 45 nm. Prolonged exposure to 10 mm caffeine in the absence of extracellular Ca 2+ depletes the caffeine stores completely. We found that subsequent applications of caffeine evoke no further response (data not shown). In the absence of external Ca 2+, caffeine increased [Ca 2+ ] i by 61 ± 26 nm in 5/11 cells. In the other Figure 10. Photoreceptors possess multiple types of internal Ca 2+ stores Ryanodine stores were depleted via a prolonged exposure to caffeine in 0 Ca 2+ supplemented with 3 mm EGTA. Subsequent exposure to ionomycin caused an additional elevation of Ca 2+ by ~800 nm, consistent with the presence of caffeine-insensitive stores in the IS. six rods, no effect of caffeine on [Ca 2+ ] i was observed, possibly due to the long exposure times in 0 Ca 2+. The ionomycin experiments demonstrate that rod photoreceptors possess both caffeine-sensitive Ca 2+ stores and one or more separate, caffeine-insensitive, Ca 2+ storage compartments. Furthermore, these results suggest that the amount of releasable Ca 2+ can vary substantially due to the previous history of cytoplasmic exposure to Ca 2+ ; at low [Ca 2+ ] i, the amount of released Ca 2+ is much less than that released in normal saline or high K + due to the gradual depletion of intracellular stores. We also tested for the presence of caffeine-insensitive stores in cones. Caffeine slightly elevated [Ca 2+ ] i in only one cone out of 12, superfused with 0 Ca 2+ saline. Ionomycin by itself, however, increased [Ca 2+ ] i in 31/31 cones superfused with 0 Ca 2+ saline (163 ± 44 nm), suggesting that these cells also store Ca 2+ in intracellular compartments. Mitochondria buffer Ca 2+ released by the ER We demonstrated that Ca 2+ release from ryanodine stores is much more prominent in rods than in cones (Figs 1 and 2). We also showed that both cell types possess caffeine-insensitive Ca 2+ stores. To determine whether Ca 2+ sequestration into caffeine-insensitive stores plays a role in influencing Ca 2+ release from the ryanodine store, we examined the role of mitochondria in modulating Ca 2+ release from ryanodine stores. Mitochondria are known to contribute significantly to Ca 2+ sequestration from the cell cytoplasm (Duchen, 1999; Rizzuto et al. 2000). Since mitochondria are often located near the mouths of Ca 2+ - release channels across the ER membrane (Landolfi et al. 1998), they may rapidly sequester released Ca 2+ before it is captured by cytosolic Ca 2+ dyes. We hypothesized that mitochondria may be especially effective in capturing Ca 2+ released from caffeine-sensitive stores in cone ISs. To fill the putative mitochondrial Ca 2+ stores, [Ca 2+ ] i was elevated with high K + and release from caffeine-sensitive stores was evoked in the absence or presence of blockers of oxidative phosphorylation. The protonophore p-trifluoromethoxy-phenyl hydrazone (FCCP) abolishes the membrane potential at the inner membrane of the mitochondria and thus blocks Ca 2+ uptake into the mitochondrial matrix (Gunter et al. 1994; Babcock et al. 1997). Figure 11A illustrates a caffeine-induced [Ca 2+ ] i signal from a rod IS. The test response evoked a transient increase in [Ca 2+ ] i. When exposed to 1 mm FCCP alone, basal [Ca 2+ ] i levels increased by 300 nm, presumably because Ca 2+ was released from mitochondrial stores. Subsequent application of caffeine in the presence of FCCP revealed a significant potentiation of the magnitude of the caffeine-evoked [Ca 2+ ] i transient. Similar results were observed in 9/9 rods. FCCP-evoked [Ca 2+ ] i increases were also observed in cone ISs (16/16 cells). In addition, a significant response to caffeine in cone ISs was unmasked in the presence of FCCP (n = 10/13 cones). Figure 11B shows a confocal line-scan recording

11 J Physiol Calcium signalling in rods and cones 771 from a cone IS superfused with control saline. Caffeine by itself caused little change in the intensity of fluo-4 fluorescence. FCCP by itself triggered a large increase in fluorescence, presumably reflecting an increase in cytoplasmic [Ca 2+ ] following inhibition of mitochondrial Ca 2+ uptake (Babcock et al. 1997; Duchen, 1999). When the cell was exposed to caffeine in the presence of FCCP, a significant elevation of [Ca 2+ ] i was observed (Fig. 11B). These results indicate that mitochondria contribute significantly to Ca 2+ homeostasis by buffering Ca 2+ released from the ER as well as Ca 2+ entering the cell through the plasma membrane. DISCUSSION The aim of this study was to explore the role of ryanodinesensitive stores in the regulation of intracellular Ca 2+ in rod and cone photoreceptors. We sought to understand how these stores were filled and emptied and how Ca 2+ extrusion and sequestration processes regulated Ca 2+ release from the stores. We found that although RyRs are expressed in both classes of cell, Ca 2+ released from stores in rod ISs contributes more prominently to global Ca 2+ homeostasis than that from cone ISs. We provide evidence for CICR in both rod and cone ISs and show that mitochondrial uptake and extrusion via PMCAs play important roles in removing Ca 2+ released from intracellular stores Photoreceptors possess caffeine-sensitive Ca 2+ stores We found that RyRs play an important role in Ca 2+ homeostasis within ISs of rod photoreceptors, but contribute much less to changes in the cytoplasmic free [Ca 2+ ] i of cone ISs. Several lines of evidence reveal the localization and functional roles of caffeine- and ryanodine-sensitive Ca 2+ stores in photoreceptor ISs. RyRselective antibodies labelled the ISs of rods and cones. Caffeine, theophylline, 4-chlorocresol and thymol, compounds that cause Ca 2+ release from caffeine-sensitive stores (Akaike & Sadoshima, 1989; Cseresnyes et al. 1997), all produced an increase in [Ca 2+ ] i (Krizaj et al. 1999). The effect of caffeine on [Ca 2+ ] i was blocked irreversibly by ryanodine and reversibly by CPA, which inhibit release and refilling of Ca 2+ stores, respectively (Fig. 7; Thomas & Hanley, 1994). The caffeine-evoked Ca 2+ increases observed in this study are likely to occur from release by the networks of ER cisternae that ramify within ISs and synaptic terminals (Sjöstrand & Nilsson, 1965; Ripps et al. 1976; Holtzman & Mercurio, 1980; Mercurio & Holtzmann, 1982; Ungar et al. 1984; Townes-Anderson, 1995). Ungar et al. (1984) showed that the ER in the IS accumulates large amounts of Ca 2+ deposits. We now show that this Ca 2+ can be released by caffeine and that the Ca 2+, sequestered within the ER is in dynamic equilibrium with the cytoplasmic [Ca 2+ ] i. Our results suggest that the stores in photoreceptor ISs are already partially filled at very low levels of [Ca 2+ ] i. As evidence of this we find that CPA, a specific inhibitor of Ca 2+ sequestration via SERCA pumps, itself increased [Ca 2+ ] i in rod and cone ISs (Figs 7 and 10). We interpret the CPA-mediated [Ca 2+ ] i increase as suggestive of the inhibition of tonic Ca 2+ sequestration into the ER. A continuous leak and sequestration of Ca 2+ into the ER is therefore likely to occur even in hyperpolarized rods at low [Ca 2+ ] i. It is possible that IS Ca 2+ stores in light-adapted cells are replenished by Ca 2+ influx through store-operated Ca 2+ channels, as described in hippocampal and dorsal Figure 11. Ryanodine-sensitive Ca 2+ stores communicate with the IS mitochondria in rod and cone ISs A, a rod IS loaded with fura-2 AM. The rod was depolarized by 20 mm KCl throughout the experiment. Caffeine (10 mm) evoked a transient [Ca 2+ ] i increase followed by an undershoot. p-trifluoromethoxyphenyl hydrazone (FCCP, 1 mm) itself caused an increase in [Ca 2+ ] i and a fivefold potentiation in the magnitude of the caffeine-evoked [Ca 2+ ]. B, line-scan recording from a cone IS loaded with fluo-4. No response to caffeine was observed in control saline. FCCP (2 mm) by itself caused a large increase in the Ca 2+ signal in the cone IS. In the presence of FCCP, caffeine evoked a significant elevation in [Ca 2+ ] i.

12 772 D. Krizaj, F. A. Lai and D. R. Copenhagen J Physiol root ganglion neurones (Garaschuk et al. 1997; Usachev & Thayer, 1999). This possibility is supported by our observation that the magnitude of caffeine-evoked Ca 2+ transients was always much larger in control Ca 2+ -containing saline than in the absence of extracellular Ca 2+. In any case, our results indicate strongly that the IS ER acts as a continuous source and sink for cytoplasmic Ca 2+ at both high and low steadystate [Ca 2+ ] i (e.g. Friel & Tsien, 1992; Thomas & Hanley, 1994; Solovyova et al. 2002). Ca 2+ homeostasis in rods and cones is regulated by the interaction between ryanodine-sensitive stores and PMCAs Rod and cone ISs respond differently to caffeine. Typically, under conditions in which robust caffeine-evoked [Ca 2+ ] i increases were seen in rods, cone ISs were unresponsive to caffeine. Basal [Ca 2+ ] i in cone ISs was almost half that measured in the rod, consistent with the notion that PMCA-mediated extrusion is more efficacious in cones and that Ca 2+ stores are not as full in cones compared to rods. Indeed, we have shown previously that plasma membrane Ca 2+ extrusion in cones is more efficient compared to Ca 2+ extrusion from rod ISs (Krizaj & Copenhagen, 1998). Our current results suggest that the Ca 2+ released from IS Ca 2+ stores is intercepted by PMCAs and by the mitochondria before diffusing into the cytoplasm, and that this mechanism is likely to be much more effective in cone than in rod ISs. The ER in cone ISs may exist as subsurface cisternae that create a restricted cytosolic space within tens of nanometres of the plasma membrane and in which Ca 2+ may accumulate without diffusing into the bulk cytosol (Berger, 1967; Baumann & Walz, 2001; Delmas & Brown, 2002). PMCAs also shaped caffeine-evoked responses in rod ISs. In control saline, [Ca 2+ ] i typically returned to baseline in the continued presence of caffeine. However, we found that exposure to extracellular La 3+ caused caffeine-evoked [Ca 2+ ] i responses to be much more sustained. La 3+ reversibly blocks Ca 2+ extrusion from ISs by its action on PMCAs, which are the exclusive mechanism for Ca 2+ extrusion from salamander photoreceptor ISs (Krizaj & Copenhagen, 1998). We thus propose that the transient nature of caffeine-evoked responses results from of Ca 2+ extrusion and a gradual depletion of Ca 2+ stores. In the absence of PMCA-mediated Ca 2+ extrusion, release of Ca 2+ via RyRs was well matched by Ca 2+ sequestration via SERCAs, as evidenced by [Ca 2+ ] i recovery following the removal of caffeine (Fig. 5). This suggests that SERCAs also play important roles in IS Ca 2+ homeostasis and that the steady-state [Ca 2+ ] i in ISs reflects a dynamic, finetuned interplay between both classes of Ca 2+ transporter (e.g. Brini et al. 2000). The role of mitochondria Although mitochondria are concentrated in the ellipsoid, they are found across all regions of the IS, including the synaptic terminal, and are often seen in intimate contact with ER cisternae (Sjöstrand & Nilsson, 1965; Holtzman & Mercurio, 1980; Mercurio & Holtzman, 1982; Brandstätter et al. 1999). In this paper we show that mitochondria in the photoreceptor IS accumulate significant amounts of Ca 2+ and buffer the Ca 2+ released from ryanodine stores. We were able to release mitochondrial Ca 2+ with the protonophore FCCP, which abolishes the driving force for Ca 2+ accumulation into the mitochondrial matrix (Gunter et al. 1994). In the presence of FCCP, the magnitude of caffeine-evoked Ca 2+ release was potentiated, suggesting that IS mitochondria act as a large-capacity Ca 2+ buffer that accumulates Ca 2+ during rapid Ca 2+ release from the internal stores. We conclude that the mitochondria are likely to constitute a significant component of the caffeineinsensitive intracellular Ca 2+ pool in photoreceptors. For example, Ca 2+ sequestration by mitochondria may be responsible for the low baseline [Ca 2+ ] i levels in the ellipsoid regions of rod and cone ISs as well as the small [Ca 2+ ] amplitude responses during depolarization and stimulation with caffeine (e.g. Fig. 1). Functional considerations for RyR signalling in photoreceptors The physiological impact of Ca 2+ stores in the IS will be determined by the spatial distribution and functional properties of signalling microdomains formed by Ca 2+ - release channels and pumps and the three-dimensional layout of the ER and mitochondria within the synaptic terminal, cell body and the ellipsoid. Our results show that Ca 2+ release can significantly amplify the Ca 2+ signal triggered by influx through voltage-gated Ca 2+ channels, suggesting a role in Ca 2+ homeostasis in depolarized cells. It remains to be determined to what extent CICR influences light-evoked responses in these subcompartments. For example, during the light response, IS ryanodine stores may function as a sink to sequester excess cytoplasmic Ca 2+, whereas CICR might help improve the kinetics of the light response during light decrements at background illumination. An added benefit of Ca 2+ sequestration might be to prevent potential neurotoxic effects of elevated Ca 2+ (Nicotera & Orrenius, 1998; Mattson et al. 2000). Both ER and mitochondria have been implicated in Ca 2+ - mediated apoptosis in many cell types, including photoreceptors (Edward et al. 1991; Duchen, 1999; Linden et al. 1999; He et al. 2000; Yang et al. 2001). Taken together, our data indicate that Ca 2+ signalling in photoreceptors is complex, involving communication between several different compartments. We show links among influx through L-type voltage-gated Ca 2+ channels, sequestration and release by the ER and by the mitochondria, and activation of the plasma membrane Ca 2+ pump by Ca 2+ released from the caffeine-sensitive compartment. Our results suggest that releasable Ca 2+ stores provide a mechanism by which diverse signalling