The selectivity for Ca2+ over Na+, PCa/PNa, is higher in cGMP-gated (CNG) ion channels of retinal cone photoreceptors than in those of rods. To ascertain the physiological significance of this fact, we determined the fraction of the cyclic nucleotide–gated current specifically carried by Ca2+ in intact rods and cones. We activated CNG channels by suddenly (<5 ms) increasing free 8Br-cGMP in the cytoplasm of rods or cones loaded with a caged ester of the cyclic nucleotide. Simultaneous with the uncaging flash, we measured the cyclic nucleotide–dependent changes in membrane current and fluorescence of the Ca2+-binding dye, Fura-2, also loaded into the cells. The ratio of changes in fura-2 fluorescence and the integral of the membrane current, under a restricted set of experimental conditions, is a direct measure of the fractional Ca2+ flux. Under normal physiological salt concentrations, the fractional Ca2+ flux is higher in CNG channels of cones than in those of rods, but it differs little among cones (or rods) of different species. Under normal physiological conditions and for membrane currents ≤200 pA, the Ca2+ fractional flux in single cones of striped bass was 33 ± 2%, and 34 ± 6% in catfish cones. Under comparable conditions, the Ca2+ fractional flux in rod outer segments of tiger salamander was 21 ± 1%, and 14 ± 1% in catfish rods. Fractional Ca2+ flux increases as extracellular Ca2+ rises, with a dependence well described by the Michaelis-Menten equation. KCa, the concentration at which Ca2+ fractional flux is 50% was 1.98 mM in bass cones and 4.96 mM in tiger salamander rods. Because Ca2+ fractional flux is higher in cones than in rods, light flashes that generate equal photocurrents will cause a larger change in cytoplasmic Ca2+ in cones than in rods.
In rod and cone photoreceptors of the vertebrate retina, a continuous membrane current flows into the outer segment in the dark that light suppresses. The dark current is carried by the simultaneous inward flux of Na+, Ca2+, and Mg2+ ions (Hodgkin et al. 1985; Cervetto et al. 1988) flowing through cyclic nucleotide–gated (CNG) ion channels. Understanding the fraction the dark current specifically carried by Ca2+ ions is important because Ca2+ influx, as a component of the dark current, is in a dynamic balance with its efflux through a Na+/Ca2+,K+ exchanger, and this balance determines the free cytoplasmic Ca2+ in the outer segment of rods and cones (Yau and Nakatani 1985; Miller and Korenbrot 1987; Lagnado et al. 1992). Moreover, the quantitative features of both influx and efflux differ between rods and cones, and these differences contribute to explaining the difference in transduction signals between the two photoreceptor types (Korenbrot 1995).
The fraction of the dark current carried by any given ion depends on the ion selectivity of the CNG channels and the concentration and driving force of that ion. CNG channels in rods and cones are more permeable to Ca2+ than to Na+, but they differ in their relative selectivity of Ca2+ over Na+ (PCa/PNa). Under saturating concentrations of cGMP, when the probability of channel opening is ∼0.9, and in the presence of simple bionic solutions, PCa/PNa in cones is ∼21.7, but in rods it is ∼6.5 (Picones and Korenbrot 1995; Wells and Tanaka 1997; Hackos and Korenbrot 1999). The difference in PCa/PNa between CNG channels of rods and cones is also observed in recombinant channels formed from alpha subunits alone (Frings et al. 1995). The value of PCa/PNa changes with channel gating: at the cGMP concentrations expected in intact cells, when probability of channel opening is ∼0.03, PCa/PNa in cones is ∼7.4× larger than that in rods (Hackos and Korenbrot 1999).
To understand the functional significance of the differences in ion selectivity between rod and cone CNG channels, the fraction of the dark current carried by Ca2+ under physiological solutions needs to be computed from the known values of PCa/PNa. This task, however, is complex. It may be accomplished through theoretical models of ion permeation. For example, Wells and Tanaka 1997 elegantly developed an Eyring-type permeation model for CNG channels with three barriers, two wells, and four sites (each for Na+, K+ Ca2+, and Mg2+) that simulates fractional currents and permeability ratios. A comparably complete computation has not been reported for cone CNG channels.
The fraction of the dark current carried by Ca2+ in intact rod has been experimentally estimated using an indirect method. The method depends on measurements of a membrane current generated by the electrogenic transport of cations by the Na+/Ca2+,K+ exchanger, and requires the separation of this electrogenic current from a changing conductive current. The method has limitations that have been thoroughly considered by those who have used it (Nakatani and Yau 1988). Despite its possible limitations, the method in several hands has provided a reproducible estimate that in rods, between 10 and 20% of the dark current is carried by Ca2+ (Nakatani and Yau 1988; Lagnado et al. 1992; Gray-Keller and Detwiler 1994; Younger et al. 1996). The Na+/Ca2+,K+ exchanger in cones also generates an electrogenic current. This current, however, is at least an order of magnitude faster than that measured in rods (Nakatani and Yau 1989; Hestrin and Korenbrot 1990; Perry and McNaughton 1991). Separating the electrogenic from the photocurrent, hence, is fraught with error (our unpublished observation) and analysis of the Ca2+ fractional current in cones through this method is not entirely reliable.
Recently, Neher and his collaborators developed a method to assess the fraction of current carried specifically by Ca2+ in several voltage- and ligand-gated channels (reviewed in Neher 1995). The method depends on the simultaneous measurement of membrane current and cytoplasmic Ca2+ after synchronous activation of the channels of interest. We have previously extended this method to an investigation of the fractional Ca2+ flux in recombinant CNG channels formed by alpha subunits of olfactory, rod, and cone receptor cells (Dzeja et al. 1999; Frings et al. 2000). We report here on the application of this method to studies in rod and cone photoreceptors that have allowed us to directly determine the fraction of the dark current carried by Ca2+ in intact cells isolated from several species.
MATERIAL AND METHODS
Striped bass (Morone saxitilis) and catfish (Ictalurus punctatus) were obtained from Professional Aquaculture Services and maintained in the laboratory for up to 6 wk under 10:14-h dark:light cycles. Tiger salamanders (Ambystoma tigrinum) were received from Charles Sullivan and maintained in an aquarium at 6°C under 12:12-h dark:light cycles. Gecko (Gecko gecko) were obtained from local shops and maintained at ambient temperature under normal daylight cycles. The UCSF Committee on Animal Research approved protocols for the upkeep and killing of the animals.
Caged 8-Br-cGMP (8-bromoguanosine 3′, 5′-cyclic monophosphate, 4,5-dimethoxy-2-nitrobenzyl ester, dihydrate, axial isomer) was synthesized as previously described (Hagen et al. 1996, Hagen et al. 1998). Fura-2 was purchased from TefLabs. Enzymes for tissue dissociation were obtained from Worthington Biochemical. All other chemicals were from Sigma-Aldrich.
Under infrared illumination and with the aid of a TV camera and monitor, retinas were isolated from dark-adapted animals and photoreceptors were dissociated as described in detail elsewhere (Miller and Korenbrot 1993, Miller and Korenbrot 1994). Single cones were isolated by mechanical trituration of fish retinas briefly treated with collagenase and hyaluronidase. Chopping the retina with a blade isolated rod outer segments. Photoreceptors were isolated and maintained in a Ringer's solution specific to each species studied. Solution composition is listed in Table.
Photoreceptors were dissociated in a Ringer's solution in which glucose was replaced with 5 mM pyruvate. Suspended in this solution, cells were added to a recording chamber held on the stage of an inverted microscope. The bottom of the recording chamber was a glass coverslip derivatized with Concanavalin A (Picones and Korenbrot 1992). Solitary photoreceptors were firmly attached to the coverslip, and after 5 min the bath solution was exchanged with the normal, glucose-containing Ringer's.
Electrical Recording and Solutions
Photoreceptors in the recording chamber were observed using an inverted microscope equipped with differential interference contrast optics and operated under infrared light (850 ± 40 nm) with the aid of an IR-sensitive TV camera and monitor. We measured membrane currents under voltage clamp at room temperature using tight-seal electrodes in the whole-cell mode. Electrodes were produced from aluminosilicate glass (1.5 mm o.d. × 1.0 mm i.d., No. 1724; Corning Glassworks), and were sealed onto the side of detached rod outer segments or the inner segment of isolated single cones. Currents were measured with a patch-clamp amplifier (Axopatch 1D; Axon Instruments, Inc.). Analogue signals were low-pass filtered below 200 Hz with an eight-pole Bessel filter (Kronh-Hite), and then digitized on line at 1 kHz (FastLab; Indec).
Solutions used to fill tight-seal electrodes are listed in Table. Salt solutions were first passed over a Chelex-100 resin column (Bio-Rad Laboratories) to thoroughly remove all multivalent cations; MgCl2 was then added. Fura-2 and caged 8Br-cGMP were added to the solutions immediately before use. Even in the dark, caged 8Br-cGMP hydrolyzes over time (Hagen et al. 1998). Therefore, electrode-filling solutions were kept in darkness and discarded after 3 h.
Solutions used to measure cGMP-gated currents with Ca2+ as the exclusive net charge carrier are listed in Table. In these solutions, the permeant cations Na+ and Mg2+ were replaced by the impermeant cations choline+ or tetramethylamonium+ (Hodgkin et al. 1985). K+ was not replaced because its presence is required for the function of the Na+/Ca2+,K+ exchanger. However, its concentrations were selected such that at −35 mV the holding voltage in the experiments, the net K+ current was zero.
Single-Cell Loading and Superfusion
Photoreceptors in the recording chamber were maintained in darkness under Ringer's solution that was exchanged intermittently. For each cell type studied here, we determined the interval necessary for equilibration between electrode lumen and cytoplasmic spaces by (a) photometrically monitoring the rate of fura-2 loading, and (b) electrically monitoring the rate of change in holding current.
The extracellular solution was changed only around the cell under investigation, using a micro-superfusion system. Test solutions were selected with electronically controlled switch valves and steady flow was maintained under a positive pressure of ∼10 cm H2O. Solutions were delivered through a glass capillary 100 μm in tip diameter placed within ∼300 μm of the cell.
Simultaneously with membrane current, we measured the fluorescence of single fura 2–loaded photoreceptors. The instrument designed for these studies is schematically illustrated in Fig. 1. Fluorescence excitation light at 380 nm (selected with a narrow band interference filter from a DC powered Xenon source (150W; Osram XBO) was delivered onto the cell through the microscope objective used to observe it (Fluor 40, 40×, 1.3 NA; Nikon). To this end, we designed an epi-illumination pathway consisting of a fiber-optic light scrambler that provided extremely uniform illumination to the back aperture of the objective (Technical Video Ltd.) and a custom dichroic mirror that reflected >60% of light between 300 and 400 nm and transmitted >90% of light between 750 and 950 nm (Chroma Technology). This pathway also included a shutter and an adjustable aperture to selectively deliver fluorescence excitation light to a limited section of the solitary photoreceptors.
We designed a second epi-illumination pathway to deliver to the cell brief, powerful flashes of light to uncage 8Br-cGMP. This pathway consisted of a Xenon flashlamp, able to deliver 200 J with ∼170-μs half bandwidth (Chadwick-Helmuth Co.). Uncaging flashes were of light <400 nm, selected with the use of short-pass optical filters (Hoya Optics Inc.). Uncaging light was focused onto the cell through the microscope objective with the aid of a custom designed relay lens. With the use of an adjustable aperture placed in this pathway, the uncaging light could illuminate selected parts of the cell. In general, uniform uncaging flashes were delivered over the entire surface of detached rod outer segments, but were restricted to the outer segment alone in studies of cones.
The fluorescent light emitted by a solitary photoreceptor was captured with a custom micro-objective designed to efficiently capture photons, while allowing electrical recording with microelectrodes (Optical Design Service). The micro-objective is a water immersion, multi-element lens 5.2 mm in diameter, 0.6 NA, and 2.8-mm working distance mounted on a micromanipulator. The lens produces a well-corrected image at a focal length of 10 mm with 0.18 NA. At the focal plane of the micro-objective, we placed the end of a multimode optical fiber (365 μm core, 400 μm clad, step index 0.22 NA). The other end of the fiber was focused, through a coupling lens, onto the photocathode of a photomultiplier tube (PMT) (R943-02; Hamamatsu) operated at −20°C and sampled in 50-ms counting bins using photon-counting instrumentation (Photochemical Research Associates).
The fiber-optic bench incorporated an interference filter to select the emitted fluorescence in the range between 410 and 600 nm (Omega Optical). A synchronized shutter in the bench was used to insure that light from the uncaging Xe flash did not reach the PMT. The fiber-optic bench also included a dichroic mirror and lens to focus the light from an IR laser diode (830 nm) onto the back end of the transfer optical fiber (Oz Optics, Ltd.). This feature was essential because it allowed us to locate and focus the micro-objective onto the individual cell under investigation. The light from the IR laser diode, transferred through the optical fiber, was emitted by the micro-objective. The experimenter, observing through the microscope objective, could see this IR light and use it as a guide to accurately position and focus the micro-objective.
In every experiment, a calibrated photodiode (United Detector Technology) was used to measure the intensity of the 380-nm fluorescence excitation light at the position of the cells on the microscope stage. Data presented here were measured at excitation intensities between 1.1 and 1.8 × 108 photons μm−2 s−1. To quantitate cell fluorescence intensity, at the end of each experiment, we measured the fluorescence emitted by individual fluorescent beads excited identically to the cells (Fluoresbrite™ carboxy BB 4.1 ± 0.2 μm microspheres; Polyscience, Inc.). Fluorescence intensity of cells and beads was corrected by subtracting from each the signal measured in their absence and due to leakage of light between excitation and emission filter sets (cross talk). In each experiment, we averaged the fluorescence of 10 different beads (units of counts in 50-ms bin per bead) and defined this intensity as one bead unit (bu). Cell fluorescence was expressed in bu by dividing the intensity of each cell by the bead unit measured in the same experiment.
To calculate Pf, the fractional Ca2+ flux, we corrected the data to analyze only the flash-dependent changes. To this end, we averaged the value of the holding current and fluorescence for 1 s preceding the uncaging flash and subtracted this value from the data; consequently, both the holding current and the fluorescence intensity preceding the flash had an average value of zero. We then integrated the corrected current over the first 15 s after the uncaging flash. Finally, we fit the integral of the current changes to the fluorescence changes.
Selected mathematical functions were fit to experimental data with nonlinear, least square minimization algorithms (Origin; MicroCal Software). Dynamic simulations were carried out using Tutsim™ software (Actuality Corp.). Statistical error throughout is presented as SD. In some calculations of population mean, we included two, and only two, measurements from the same cell (as indicated in the text). To calculate errors on ratios of means, we computed propagation of errors as described by Bevington 1992:
where σ2 is the variance of the mean of quantities a, u and v.
The fraction of the cGMP-dependent current carried by Ca2+ ions in photoreceptors, here denominated Pf, can be determined by simultaneously measuring total membrane current with patch-clamp electrodes, and Ca2+ concentration with the Ca2+-sensitive fluorescent dye, fura-2. Neher 1995 has reviewed this analytical method and we have described in detail its application to studies of CNG channels (Frings et al. 2000). In brief, cGMP-dependent currents are rapidly activated by the sudden intracellular application of 8Br-cGMP, achieved by photolysis of caged 8Br-cGMP molecules. Because the channels exhibit only limited ion selectivity, the cGMP-dependent membrane current in normal Ringer's solution, IT, is composed of contributions from several ion species: Na+ (INa), K+ (IK), Mg2+ (IMg), and Ca2+ (ICa) ():
The Ca2+-free form of fura-2 is present in the outer segment's cytoplasm, where it binds the inflowing Ca2+. Binding Ca2+ causes a decrease in the intensity of fluorescence excited by 380 nm light (F380). The extent of decline in F380 is proportional to free Ca2+ concentration, but it measures the total amount of Ca2+ entering the cell if, and only if, a number of experimental conditions are strictly met. (a) Ca2+ enters the cell only through the channels of interest. (b) All entering Ca2+ is bound by free fura-2; none is sequestered or actively transported out of the cell. (c) Fura-2 outcompetes endogenous Ca2+ buffers because it is present at a concentration that exceeds that of the other buffers.
The change in the mole amount of intracellular Ca2+ caused by the cGMP-activated current over a time interval Δt, is given by :
where F is Faraday's constant. If fura-2 is used to monitor cytoplasmic Ca2+, intracellular Ca2+ may distribute among several competing pools: (a) free Ca2+, (b) Ca2+ bound to cytoplasmic buffers (CaB), (c) Ca2+ bound to fura-2 (CaFu), and (d) Ca2+ pumped into cellular organelles or out of the cell (P). Thus:
where Δnfree, ΔnCaFu, ΔnCaB, and ΔnP are the changes in mole amount of Ca2+ free, bound to fura-2, bound to cytoplasmic buffers, and pumped (or sequestered) out of the cytoplasm, respectively. Neher 1995 reasoned, and proved rigorously, that under certain experimental conditions, when all entering Ca2+ is bound to the fura-2 present in the cell, can be simplified. These conditions are: (a) measurements are executed with high fura-2 concentrations, 1–5 mM. Under this condition, the indicator dye, which is a very powerful buffer, outcompetes the endogenous Ca2+ buffers, rendering ΔnCaB negligible. (b) The rapid rate of Ca2+ binding by fura-2 (few milliseconds) (Kao and Tsien 1988) allows a few seconds to be sufficiently long for binding to the dye to be complete, but still short relative to the time necessary to pump Ca2+ into organelles or out of the cell; this condition makes ΔnP negligible. (c) Fura-2 exists in its Ca2+-free form. Then, all entering Ca2+ is bound by the dye and Δnfree is negligible. Then, only under these experimental conditions:
where [Ca2+], [CaFu], and [Futot] are the concentrations of free Ca2+, bound Ca2+, and total fura-2, respectively. Kd is the dissociation constant of the fura-Ca2+ complex. With this definition, considering a small incremental elevation in [Ca2+] and under the experimental conditions that hold true for , then:
where Vc is the accessible cytoplasmic volume. If we define the emitted intensity of fura-2 fluorescence excited at 380 nm (F380) as the sum of the fluorescence of its two molecular forms, one Ca2+-free, SFu, and the other Ca2+-bound, SCaFu, then:
This equation indicates that, under these conditions, the change in fluorescence is proportional to the total amount of Ca2+ entering the cell, as the ion shifts the fraction of fura-2 between its free and bound states. Combining , , and yields :
Let us define f as the ratio of the change in fluorescence, ΔF380, and the integral of the membrane current ():
Under conditions where Ca2+ is the exclusive current carrier; that is, when IT = ICa, then f takes the value fmax.
and the fraction of the IT current carried by Ca2+ under normal conditions, Pf, can be calculated if f and fmax are both measured in the same cell type.
To determine Pf, the fraction of the total cGMP-dependent current carried by Ca2+ ions, we measured flash-generated changes in membrane current under voltage clamp and fluorescence of individual rod and cone photoreceptors loaded with caged 8Br-cGMP and fura-2. We studied the cells with tight-seal electrodes filled with solutions that lacked triphosphate nucleotides and, therefore, could not sustain endogenous synthesis of cGMP (Hestrin and Korenbrot 1987; Rispoli et al. 1993). At −35 mV holding voltage, the magnitude of the membrane current preceding the uncaging flash is a measure of the stability of the electrical recordings and the completeness of diffusion exchange between electrode lumen and the cytoplasm. When the cells' cytoplasm is fully equilibrated with the electrode lumen, the holding current, Ihold, at −35 mV should be zero for detached rod outer segments (dROS) and outward for cones, which consist of both inner and outer segments. These expectations follow from the fact that CNG channels are the only channels in the dROS and they should be closed (Baylor and Nunn 1986). In cones, and rods with an inner segment, in contrast, an outward current can be expected to flow through the voltage-dependent K+ channels of the inner segment (Maricq and Korenbrot 1990; Barnes and Hille 1989). The data presented here were measured in tiger salamander dROS no less than 4.5 min after going into whole-cell mode and while Ihold was −4.6 ± 2.7 pA (n = 17); striped bass single cones at no less than 3 min after going into whole-cell mode and while Ihold was 21.1 ± 12.3 pA (n = 35); catfish cones at no less than 3 min after going into whole-cell mode and while Ihold was 14.3 ± 8.3 pA (n = 11), and catfish rods (which retain a portion of their inner segment) at no less than 4 min after going into whole-cell mode and while Ihold was 46.2 ± 16.7 pA (n = 12). Cell fluorescence of fura-2 also confirmed that, at the times indicated, equilibration between cell cytoplasm and the electrode was complete.
Typical current and fluorescence changes generated by uncaging flashes in dROS of tiger salamander and single cones of striped bass are illustrated in Fig. 2. In both receptor types, the uncaging flash activated a transient, inward current and, simultaneously, caused a decrease in fluorescence intensity. The current increase reflects the transient opening of cGMP-gated channels, while the loss of fluorescence arises from the entry of Ca2+ ions, which subsequently bind to fura-2. Shown are repeated measurements in the same cell generated by flashes of increasing intensity. Photoreceptors recovered to their initial conditions in between flashes, provided that at least 3 min elapsed between the flashes. The peak current amplitude increased with flash intensity and so did the extent of fluorescence change. Since the intensity of the flash is correlated with the amount of released 8Br-cGMP, the results indicate that current amplitude and Ca2+ influx are dependent on the cyclic nucleotide concentration. Since CNG channels in rods are more sensitive to 8Br-cGMP than those of cones (Hackos and Korenbrot 1997), it is not surprising that uncaging flashes of comparable intensity caused larger current changes in rods than in cones.
The rapid rise of the flash-generated current (time to peak ≤100 ms) (Fig. 2) is consistent with the rate of photolysis of the caged 8Br-cGMP (Hagen et al. 1998) and the rapid activation of channels by the free nucleotide (Karpen et al. 1988). However, the rapid decline from a peak towards the starting value is surprising. We observed this feature in every cell and every species we studied. The time course of the current decay from its peak was well described by a single exponential. In tiger salamander dROS, the mean decay time constant was 0.84 ± 0.23 s (range 0.61–1.27, n = 13 for peak currents ≤200 pA). In striped bass cones, the mean decay time constant was 0.89 ± 0.38 s (range 0.54–1.77 s, n = 9 for peak currents ≤200 pA). Experimental results detailed in the suggest that the rapid current decline arises primarily from the binding of free 8Br-cGMP to specific binding sites in the outer segment. The existence of cGMP-binding sites in rod outer segments has been inferred from analysis of cGMP diffusion kinetics (Olson and Pugh 1993; Koutalos et al. 1995) and measured biochemically (Cote and Brunnock 1993). Although comparable data do not exist for cones, the similarity in decay kinetics suggests that cGMP-binding sites are not different in rods and cones.
Experimental Conditions Required to Determine the Value of Pf Are Met in Isolated Photoreceptors
To calculate Pf, it is crucial to establish that the experimental conditions necessary to apply the analytical theory described above indeed hold.
Ca2+ enters the cell only through the channels of interest.
In dROS, CNG are the only ion channels present in the plasma membrane (Baylor and Nunn 1986; Hestrin and Korenbrot 1987). In cones, and rods with an inner segment, Ca2+ could potentially enter through voltage-gated Ca2+ channels in the inner segment. We assured that voltage-gated Ca2+ channels are inactive by conducting all experiments under voltage clamp at −35 mV holding potential, a condition under which the known Ca2+ channels in rod and cone inner segments are inactive (Corey et al. 1984; Maricq and Korenbrot 1988).
All Ca2+ that enters the cell through the CNG channels in the outer segment binds to fura-2.
In cones or rods with an inner segment, Ca2+ entering the outer segments can flow into the inner segment through the cilium that connects them. Flow between outer and inner segment can be expected for any diffusible molecule. Evidence of this flow is presented in Fig. 3. When 8Br-cGMP was released in the inner segment only, by restricting uncaging light to this part of the cell, currents were nonetheless generated in the outer segment. The currents, however, exhibited a delay of a magnitude well predicted for the aqueous diffusion of 8Br-cGMP from the inner segment to the outer segment. To minimize errors that may arise from intracellular Ca2+ flow, data in cone photoreceptors were measured by restricting uncaging light to the outer segment alone, but sampling fura-2 fluorescence over the entire cell. Under these conditions, cGMP-dependent Ca2+ entry occurs only in the outer segment, but it is fully sampled even if it flows to the inner segment.
In photoreceptor outer segment, entering Ca2+ is cleared by a Na+/Ca2+,K+ exchanger and the rate of clearance is faster in cones than in rods (reviewed in Korenbrot 1995). To investigate whether this transport interferes with Pf determination, we measured currents and fluorescence in a bass single cone in the presence and absence of extracellular Na+, conditions under which Ca2+ clearance through the Na+/Ca2+,K+ exchanger should be active or inactive, respectively. Typical results are illustrated in Fig. 4. Similar observations were made in four other cells. In the presence of normal Ringer, the uncaging flash activated a transient, inward current and caused a decrease in fluorescence. 3 min later, the same cell was superfused with a solution in which all Na+ was iso-osmotically replaced by guanidinium+. Guanidinium+ permeates CNG channels indistinguishably from Na+ and, therefore, the flash-activated currents, as expected, were nearly identical under both ions. Guanidinium+ does not replace Na+ in the Na+/Ca2+,K+ exchange, which, therefore, is inactive. The initial decrease in fluorescence was indistinguishable in the presence of the two cations, but ∼12 s after the uncaging flash, the time course of fluorescence distinctly diverged. It was nearly stationary in the presence of Na+, but continued to decline in the presence of guanidinium+ (Fig. 4). It was necessary to return to Na+ Ringer for the fluorescence to recover its initial value. These results indicate that, in the presence of an active exchanger, 2 mM fura-2 in the cytoplasm retains the entering Ca2+ for the first 10 s or so after channel opening. At longer times, a combination of binding saturation of fura-2 and the activity of the exchanger compromises the retention of Ca2+ by the fluorescent dye. The method of analysis presented here was applied only to data recorded over the first 6–8 s after the uncaging flash.
Fura-2 is at a sufficient concentration to outcompete endogenous buffers.
We verified that 2 mM fura-2 is a sufficient concentration by comparing the Pf results obtained in the presence of 1 and 4 mM fura-2. In both rods and cones, Pf data gathered under either concentration were the same (data not shown, but see Frings et al. 2000), indicating that the experimental condition holds. The same conclusion has been reached in previous studies with CNG channels expressed in mammalian cells (Frings et al. 1995; Dzeja et al. 1999).
Pf in Rod Outer Segments of Tiger Salamander
To determine the value of f, we measured cGMP-activated changes in current and fluorescence in dROS of tiger salamander in the presence of the solutions listed in Table and Table. To determine Pf, it was also necessary to measure fmax, the value of f when IT = ICa (). To do so, we carried out experiments using the same protocols used to measure f, but we exposed the cells to the solutions listed in Table. Cells were maintained in normal Ringer's and exposed only for ∼60 s to the extracellular test solution to determine fmax. Extracellular Na+ and Mg2+ were replaced by the impermeant cation choline, while maintaining Ca2+ at the normal 1 mM. Intracellular Mg2+ and most of the K+ ions were also replaced with choline. A small amount of K+ was retained in the intracellular test solutions to assure that the Na+/Ca2+,K+ exchanger operated in normal, Na+ Ringer's. At the intracellular K+ concentration selected and at −35 mV holding voltage, the net K+ current was zero. Under fmax conditions, Ca2+ is the only net charge carrier through the CNG channels.
In Fig. 5 are shown typical measurements of f and fmax in tiger salamander dROS. In the standard Ringer's at −35 mV, the holding current before the uncaging flash was essentially zero and the flash caused a transient inward current and a decrease in fluorescence from its initial value. In the Ringer's, where Ca2+ ions are the only net charge carriers, the uncaging flash also generated a transient inward current and a decrease in fluorescence intensity in dROS. Also shown in Fig. 5 are flash-induced changes in fluorescence and the integral of the current changes. These two data sets superimpose over a span of ∼5 s after the flash (Fig. 5, bottom). Superimposition is evidence that the conditions necessary to calculate f (see above) are met over this restricted time span. We determined the singular value of f, the ratio of the change in fluorescence to the current integral only over this restricted time span (f, bu/pC).
The mean peak photocurrent amplitude in tiger salamander rods is ∼70 pA (Miller and Korenbrot 1994). To determine the value of Pf under conditions similar to those expected in the functional retina, we restricted analysis to data sets in which uncaging flashes generated peak current amplitudes ≤200 pA, with a mean value of 115 ± 53. The average value of fmax was 0.0112 ± 0.0019 bu/pC (13 measurements in eight different cells, range 0.0101–0.0130). The average value of f was 0.00238 ± 0.0009 bu/pC (n = 17). Hence, under these conditions, the average value of Pf () in tiger salamander rods is: 0.21 ± 0.01
We also measured Pf in dROS of gecko because the Ca2+ homeostasis of this organelle has been carefully and thoroughly studied (Gray-Keller and Detwiler 1994; Gray-Keller et al. 1999). We studied this species less extensively than others and we pooled data measured in cells over a much wider range of peak current amplitudes. For currents between 210 and 1,048 pA in peak amplitude, the average f value was 0.0044 ± 0.0012 bu/pC (n = 4). Given fmax of 0.022 for these cells, the average value of Pf in gecko rods is 0.19 ± 0.05.
Pf in Single Cones of Striped Bass
In Fig. 6 are shown typical measurements of f and fmax in single cones of striped bass. Shown are electrical and photometric data both in standard Ringers solution and in one designed to have Ca2+ ions as the only net charge carrier. The experimental data are qualitatively similar to that in dROS, uncaging flashes cause a transient inward current and a decrease in fluorescence. We analyzed the experimental data to determine the values of f and fmax as described above. For currents <200 pA in peak amplitude, the average value of fmax was 0.0098 ± 0.0014 bu/pC (range 0.008–0.0125, n = 9). The average value of f was 0.0032 ± 0.0011 bu/pC (n = 40). Hence, under these conditions, the average value of Pf () in striped bass single cones is 0.33 ± 0.02.
Pf in Single Cones and Rods of Catfish
The two species described above, tiger salamander and striped bass, are the ones in which most of the electrical studies of ion selectivity have been conducted. The rod versus cone differences described, however, could represent differences between species, rather than between photoreceptor types. This is unlikely since rod/cone differences are also found in alpha subunit channels of bovine photoreceptors (Frings et al. 1995; Dzeja et al. 1999). Nonetheless, we investigated the value of Pf in rods and cones of the same species.
In Fig. 7 are shown typical measurements in rods and cones of catfish. In this species, unlike tiger salamander or gecko, the detached rod outer segment includes a small portion of the cell's ellipsoid. Consequently, at −35 mV, the holding current preceding the uncaging flash was outward both in the dROS and the single cones. In both cell types, the uncaging flash caused an inward current and a decrease in fluorescence. For uncaging flashes of the same intensity, the amplitude of the inward currents (and its concomitant change in fluorescence) was always smaller in cones than in rods. This almost certainly reflects the fact that the sensitivity for cGMP is higher in channels of rods than in those of cones (Haynes and Yau 1990; Rebrik and Korenbrot 1998).
We determined the values of f (Fig. 7) and fmax (data not shown) for rods and cones of the catfish.
For currents <75 pA in peak amplitude (range 30–75 pA), the average value of fmax was 0.0087 ± 0.0031 bu/pC (nine determinations in six cells). For currents <200 pA in peak amplitude, the average f value was 0.00295 ± 0.0013 bu/pC (17 determinations in 11 cells). Under these conditions, the average value of Pf was 0.34 ± 0.06.
For currents <50 pA in peak amplitude (range 13–40 pA), the average value of fmax was 0.0124 ± 0.0028 bu/pC (seven determinations in four cells). For currents <220 pA in peak amplitude, the average f value was 0.00174 ± 0.0005 bu/pC (19 determinations in 12 cells). Under these conditions, the average value of Pf was 0.14 ± 0.01.
The values of Pf at −35 mV holding voltage, for cGMP-dependent currents <200 pA in amplitude and in the presence of 1 mM external Ca2+ for the species studied here, are compared in Fig. 8. The differences in Ca2+ fractional current between rods and cones are thus features preserved across species.
Interaction between Ca2+ and CNG Channels Are Reflected in the Value of Pf and Differs Between Channels of Rods and Cones
We measured the current and fluorescence changes caused by identical flashes delivered to the same cell bathed in varying amounts of Ca2+ (Table). Cells were superfused with test solution beginning 15 s before the flash and for the course of the measurement (30 s). Otherwise, cells were maintained in the normal Ringer's. We waited 2–3 min between flashes, time sufficient for cells to recover to their starting conditions. We analyzed only data in which these starting values differed by <10%. We tested various Ca2+ concentrations in random sequence, but always started at the normal 1 mM Ca2+.
In Fig. 9 are shown typical data measured in striped bass single cones and tiger salamander dROS. We illustrate fluorescence changes and the integral of the current scaled to optimally fit the changes in fluorescence. For both photoreceptor types, the data shown were measured at 3-min intervals in the same cell bathed with varying external Ca2+ concentrations. The signal-to-noise ratio in the photometric cone data is significantly less than in those of rods because bass cone outer segments are only about one tenth the volume of the tiger salamander's rod outer segments. In cones, the average f values we measured were: at 0.5 mM Ca2+, 0.00129 ± 0.00018; at 1 mM Ca2+, 0.00295 ± 0.00049; at 1.5 mM Ca2+, 0.00383 ± 0.00113; at 2 mM Ca2+, 0.00421 ± 0.0015; at 2.5 mM Ca2+, 0.00518 ± 0.0021; and at 5 mM Ca2+, 0.00952 ± 0.00603. Data were measured in a total of 10 cells, with any one concentration sampled in 3–10 cells. Thus, fractional Ca2+ current increased steeply with Ca2+ concentration. The dependence of Pf on Ca2+ was well described by the Michaelis-Menten function:
KCa is the value at which Pf = 0.5. We obtained the best fit to the mean of the data with KCa = 1.98 mM.
In rods, the fractional Ca2+ current also increased with external Ca2+, but much less steeply than in cones and, even at 10 mM, external Ca2+ carries only about half the current (Fig. 9). The average f values we measured were: at 1 mM Ca2+, 0.00286 ± 0.000883; at 2.5 mM Ca2+, 0.003812 ± 0.00098; at 5 mM Ca2+, 0.005713 ± 0.004314; and at 10 mM Ca2+, 0.00668 ± 0.000496. Data were measured in a total of nine cells, with any one concentration sampled in four to nine cells. We note that in this data set, the mean value of Pf at 1 mM Ca2+ was 0.25 ± 0.043. This value is slightly larger than the mean value presented above because the data set included peak 8Br-cGMP-dependent current of amplitude ≤1,000 pA, with a mean value of 454 ± 347 pA. The mean value of Pf is smaller for currents ≤200 pA because selectivity is linked to gating in rod CNG channels (Hackos and Korenbrot 1999). The best fit of the Michaelis-Menten function () to the mean of the rod data was obtained with KCa = 4.96 mM (Fig. 9).
We have directly measured the fraction of the cGMP-gated current carried by Ca2+ in the outer segment of intact rods and cones. Ca2+ carries 14–20% of the dark current in rods and 32–34% in cones. The differences in fractional Ca2+ flux between rods and cones are maintained across several species.
Our results in rods are similar to those of indirect measurements previously reported by others, and based on an analysis of the current generated by the electrogenic transport activity of Na+/Ca2+,K+ exchangers in the outer segment membrane (Nakatani and Yau 1988; Lagnado et al. 1992; Gray-Keller and Detwiler 1994; Younger et al. 1996). Such analyses require the unambiguous separation of the exchanger current from the photocurrent, as well as assumptions about transport stoichiometry. The indirect studies have set the value of Pf in rods anywhere between 0.10 and 0.18. The similarity of present and past results in rods confirms the accuracy of the indirect method and extends the observation to several species.
Our results in cones are in quantitative disagreement with those reported by Perry and McNaughton 1991, based on analysis of the electrogenic exchanger currents in tiger salamander cones. Perry and McNaughton 1991 correctly recognized that the fraction of the dark current carried by Ca2+ current is higher in cones than in rods. Their estimate of this fraction, however, is smaller than the results we have found. They estimated that Pf is 0.12 in rods and 0.21 in cones. This difference likely reflects the uncertainty in separating the exchanger from the conductive currents. The exchanger current in rods of lower vertebrates decays with an exponential time course of several hundred milliseconds or seconds time constant (Nakatani and Yau 1988; Lagnado et al. 1992; Gray-Keller and Detwiler 1994; Younger et al. 1996). In cones, in contrast, the exponential decay of the exchanger current has a time constant of tens of milliseconds duration (Hestrin and Korenbrot 1990; Perry and McNaughton 1991).
Previous electrical studies had established that the selectivity of Ca2+ over Na+, PCa/PNa, measured from reversal potentials is higher in CNG channels of cones than in those of rods (Frings et al. 1995; Picones and Korenbrot 1995). These measurements, however, are conducted under simple ionic solutions, with only Ca2+ and Na+ present. Extending them to determine the fractional Ca2+ flux under normal physiological solutions is complex and requires application of a specific model of ion permeation. Wells and Tanaka 1997 have developed a plausible model based on Eyring rate theory that successfully simulates PCa/PNa, I-V curves and the Ca2+ fractional flux in rod CNG channels. The model presumes the existence of three barriers, two wells, and four sites that interact electrostatically (each for Na+, K+ Ca2+ and Mg2+). Similar but less elaborate models were first formulated and analyzed by Zimmerman and Baylor 1992. A comparably complete computation has not been reported for cone CNG channels. However, a simpler Eyring-type model consisting of one barrier and two wells has been successfully used to analyze monovalent cation selectivity in cone CNG channels (Picones and Korenbrot 1992; Haynes 1995).
An indirect conclusion supported by our results is that strong buffering sites for cyclic nucleotides operate in intact rod and cone outer segments of all species tested. The sites are relatively fast in kinetics, their Kd is high and they are active in the cyclic nucleotide concentration range expected under normal operating conditions in the cell. Biochemical analysis suggests that, in rods, the cGMP binding sites are noncatalytic sites on phosphodiesterase (Cote and Brunnock 1993). The sites are not functionally different in rods and cones.
There are significant physiological consequences to accurately knowing the value of Pf. Cytoplasmic Ca2+ homeostasis in the photoreceptor outer segment arises from the dynamic balance of influx, efflux, and buffering by endogenous sites, with no role recognized for transport in or out of intracellular compartments. Knowing the value of Pf, therefore, is crucial to understanding Ca2+ dynamics. This is made particularly clear by computational models developed to explain in molecular detail the phototransduction process. Particularly detailed and thorough models have been put forward (Forti et al. 1989; Sneyd and Tranchina 1989; Torre et al. 1990; Nikonov et al. 1998). Every one of these models includes an explicit term for Pf, and the value assigned to this parameter strongly affects the performance of the model.
The difference in Pf between rod and cone channels implies that light-dependent changes in Ca2+ concentration caused by equal amplitude currents will be larger in cones than in rods. This fact, added to the known difference in the clearance rate of Ca2+ from the outer segment indicates that, for a given current response, Ca2+ concentration changes will be larger and faster in cones than in rods. These differences contribute to explaining the dissimilarity in the photosensitivity, kinetics, and light- and dark-adaptation characteristics of the two photoreceptor types.
Differences in Pf between rod and cone channels reflect differences in the molecular interaction between the cation and the channel protein. Seifert et al. 1999 studied homomeric channels formed from α subunits of CNG channels cloned from olfactory neurons, rod, and cone photoreceptors and Drosophila. They discovered that the interaction between extracellular Ca2+ and the channels differs among channels cloned from different cell types (rod versus cone versus olfactory), but not among channels cloned from the same cell type in different species (bovine versus chicken versus rabbit). Again, a unique CNG channel structure/function is associated with a specific cell type, regardless of the animal species in which the cell is found.
Mutagenesis analysis has demonstrated than in recombinant homomeric alpha CNG channels, Ca2+ interacts with a glutamate residue within the pore loop of the channel (E363) (Root and Mackinnon 1993; Eismann et al. 1994). Although the same residue exists in the pore of all CNG alpha channels, the thermodynamics of the interaction between Ca2+ and glutamate E363 differs among CNG channels. For example, the extracellular Ca2+ concentration that blocks channel conductance by 50% at −70 mV is 7 μM in rods, but 68 μM in cones (Seifert et al. 1999). Thus, structural features of E363 alone cannot explain the characteristics of the interaction between Ca2+ and the CNG channels. Recent work has demonstrated that the Ca2+/E363 interaction changes with the state of protonation of the residue, which, in turn, is controlled by structural features of the regions of the protein adjacent to the pore, the S5 and S6 loops (Seifert et al. 1999). The S5-pore-S6 “structural module” differs among recombinant alpha channels depending on the cellular source of the cDNA (Seifert et al. 1999).
The molecular view of the interaction between Ca2+ and the S5-pore-S6 structural module obtained from experiments with recombinant homomeric alpha CNG channels cannot simply be extended to the structure of the native photoreceptor channels. Native channels in both rods and cones are formed from alpha and beta subunits. Rod channels are formed by two alpha and two beta subunits, although the sequence of these subunits is controversial (Shammat and Gordon 1999; He et al. 2000). Two striking differences exist between native and recombinant alpha channels with respect to their Ca2+ fractional currents. (a) In the presence of 1 mM external Ca2+, Pf in native cone channels is about twice that in rods, but in alpha recombinant channels this is reversed: Pf in rods is about three times larger than that in cones. (b) The Ca2+ dependence of Pf is extremely steep in rod alpha recombinant channels, but remarkably shallow in native rod channels. In recombinant channels, the Ca2+ concentration at which Pf = 0.5 is 0.4 mM and Pf is essentially 1 when Ca is only 1 mM (Dzeja et al. 1999). In the native channel, Pf is 0.2 at 1 mM Ca2+ and even when Ca2+ is 10 mM, Pf is still only ∼0.5.
A structural understanding of the differences between recombinant alpha and native channels is not at hand, but, if we extend the information gained from the structural studies of Seifert et al. 1999, the functional differences likely reflect differences in the dielectric environment surrounding the pore glutamate. Differences in dielectric environment can arise from two possible sources or their combination. The obvious first possibility is that the beta subunits, which have a glycine instead of a glutamate in the pore loop, create a pore sufficiently unlike that in alpha channels as to explain the difference. A second possibility is that the dielectric environment is changed by Mg2+ ions. The results reported here were all measured in the presence of 1 mM external Mg2+ and 0.5 mM free cytoplasmic Mg2+. The studies of alpha recombinant channels were conducted in the absence of external Mg2+ and with 0.27 mM free cytoplasmic Mg2+ (Dzeja et al. 1999). Because the channels bind Mg2+ almost as well as Ca2+, the presence of Mg2+ must influence the interaction between Ca2+ and the channel (Wells and Tanaka 1997). The role of these two possible mechanisms needs to be resolved experimentally.
It is notable that, whereas the dependence of Pf on outside Ca2+ differs significantly between native and recombinant rod channels, this is not the case for cone channels. In both native and recombinant cone channels, the Michaelis-Menten function describes well the dependence of Pf on Ca2+, with KCa = 1.38 mM in alpha recombinant channels (Dzeja et al. 1999) and KCa = 1.98 mM in native ones. This characteristic of the channels is similar despite the fact that measurements were conducted in the absence of Mg2+ in recombinant proteins and with 1 mM outside Mg2+ in the native cells. If the presence of beta subunits is important in defining the interaction of divalent cations with native channels, our results suggest that the alpha/beta subunit interactions differ in rods and cones. Future work should provide information on the mechanisms of this difference.
We investigated possible mechanisms to explain the surprisingly rapid decay of the outer segment current generated by uncaging 8Br-cGMP (2). If 8Br-cGMP were not processed within the outer segment, it would be reasonable to expect a slow decay in current after the uncaging flash due to equilibration of freed 8Br-cGMP between cytoplasm and electrode lumen. To determine the rate of this equilibration, we used tight-seal electrodes filled with the usual solution (2) containing, in addition, 30 μM 8Br-cGMP, and we monitored the membrane current after membrane rupture to attain whole-cell mode. 10 illustrates a typical result in bass single cones at −35-mV holding voltage. After membrane rupture, an inward current developed without detectable delay that reached a maximum in ∼52 s. In eight different cells, the mean time to the maximum was 73.6 ± 22.3 s. Since current decay after an uncaging flash is much, much faster (above), its time course cannot be explained by the rate of cell/electrode equilibration.
Rapid current decay could reflect hydrolysis of free 8Br-cGMP by phosphodiesterase (PDE) activated by the uncaging flash. This is plausible because, while 8Br-cGMP is ∼100-fold less effective than cGMP as a substrate of rod PDE (Zimmerman et al. 1985), it can, if present at a sufficiently high concentration, be hydrolyzed by activated PDE. Indeed, bright lights generate photocurrents in rods loaded with high concentrations of 8Br-cGMP (0.5–1 mM) (Cameron and Pugh 1990). To test whether current inactivation reflects 8Br-cGMP hydrolysis by PDE, we conducted two control experiments. In the first one, we measured in single cones the current generated by uncaging cGMP in the continuous presence of 0.1 mM 3-isobutyl-1-methylxanthine (IBMX), an effective blocker of cone PDE (Gillespie and Beavo 1989). The current rapidly reached a peak and, in spite of the presence of IBMX, it decayed with an exponential time course similar to that measured in the absence of the blocker (10). In a second control experiment, we tested whether the uncaging flash generated a photoresponse in single cones loaded with 15 μM 8Br-cGMP. We do not know the exact concentration of free 8Br-cGMP generated by the uncaging flash, but considering that K1/2, the nucleotide concentration that half activates the maximum CNG current in the intact cone in the absence of Ca2+, is ∼15 μM (Rebrik and Korenbrot 1998), the concentration necessary to generate 200 pA current is ∼5.5 μM. Even in the presence of 15 μM 8Br-cGMP, the flash did not cause a photocurrent (10). We can, therefore, reject the hypothesis that current decay is caused by 8Br-cGMP hydrolysis.
Ca2+ modulates ligand sensitivity of CNG channels in both rods (Hsu and Molday 1993; Gordon et al. 1995; Sagoo and Lagnado 1996; Bauer 1996) and cones (Rebrik and Korenbrot 1998). Therefore, it is possible that current decay reflects a decrease in CNG channel sensitivity to free ligand, caused by a rise in cytoplasmic Ca2+ associated with the enhanced current. Such modulation, however, is unlikely to occur under our experimental conditions because changes in cytoplasmic free Ca2+ are vanishingly small in the first 6–8 s after the uncaging flash due to the presence of 2 mM fura-2.
We propose that the rapid current decline arises from binding of free 8Br-cGMP to specific sites in rod and cone outer segments. The existence of cGMP-binding sites in rod outer segments has been inferred from analysis of cGMP diffusion kinetics (Olson and Pugh 1993; Koutalos et al. 1995) and measured biochemically (Cote and Brunnock 1993). These sites are presumed to be the known noncatalytic nucleotide binding sites on PDE molecules (Hebert et al. 1998). To test this hypothesis, we carried out a competition experiment in the intact cell. The noncatalytic binding site in the rod PDE binds cIMP (3′-5′ cyclic inosine monophosphate) with ∼20-fold higher affinity than 8Br-cGMP (Hebert et al. 1998). On the other hand, the ligand sensitivity of the rod CNG channels for cIMP is ∼0.006-fold that for 8Br-cGMP. K1/2 is ∼8 μM for 8Br-cGMP, but ∼1.2 mM for cIMP (Tanaka et al. 1989). Based on these quantities, simple calculations indicate that 20 μM cIMP should not activate the CNG channels (<0.003%), but should significantly reduce binding of free 8Br-cGMP to the PDE. If 8Br-cGMP binding is reduced by cIMP, this should be apparent by a slowdown in the kinetics of binding of 8Br-cGMP in the presence of cIMP, when compared to the kinetics of binding in the absence of cIMP. This is indeed the case (11).
We measured in tiger salamander dROS the current activated by uncaging 8Br-cGMP in the presence of 20 μM cIMP. We filled tight-seal electrodes with the standard solution containing caged 8Br-cGMP and fura-2 (2) with an additional 20 μM cIMP. The mean holding current was unaffected, indicating that, as expected, cIMP at this concentration did not activate CNG channels. An uncaging flash caused a transient inward current of peak amplitude dependent on the intensity of the uncaging flash (11). The current activated rapidly, and then declined from the peak towards its starting value. The rate of decline, however, was much slower than in the absence of cIMP (11, also compare with 2). The time course of the current decay from its peak was well described by a single exponential with a mean decay time constant of 2.34 ± 1.1 s (range 1.32–3.98, n = 10). The decay time constant in the absence of cIMP was ∼0.84 s (see results). On average, then, the rate of decay of currents activated by uncaging 8Br-cGMP was about threefold slower in the presence of 20 μM cIMP than in its absence (11). Mathematical simulations of binding kinetics of the nucleotide to a single site (not shown) indicate that the slowdown in current is consistent with cIMP having reduced the number of available sites for 8Br-cGMP binding by ∼70%. Our experimental results are consistent with the proposition that free 8Br-cGMP binds to specific sites in the photoreceptor outer segment and that these sites select cIMP over 8Br-cGMP in a manner similar to that described for rod PDE (Hebert et al. 1998). Data on nucleotide binding are not available for cone outer segments (or cone PDE), but binding is likely quantitatively similar to that in rods since current decay after the uncaging flash was nearly the same in the two photoreceptor types.
Dr. Hackos' present address is Molecular Physiology and Biophysics Unit, National Institute of Neurological Disease and Stroke, National Institutes of Health, Bethesda, MD 20892.116
Abbreviations used in this paper: bu, bead unit; CNG, cyclic nucleotide–gated; dROS, detached rod outer segment; IBMX, 3-isobutyl-1-methylxanthine; PDE, phosphodiesterase; PMT, photomultiplier tube.