Quantitative aspects of cGMP phosphodiesterase activation in carp rods and cones ( コイ桿体と錐体とでの cGMP ホスホジエステラーゼの 活性化効率の定量的理解 )

----------------------------------------------------------------------------3 INTRODUCTION------------------------------------------------------------------------4 EXPERIMENTAL PROCEDURES----------------------------------------------------8 RESULTS and DISCUSSION---------------------------------------------------------18 REFERENCES--------------------------------------------------------------------------31 ACKNOWLEDGEMENTS------------------------------------------------------------33 APPENDIX-------------------------------------------------------------------------------34


INTRODUCTION
Vertebrates have two types of visual photoreceptors, rods and cones, which convert light-detection signals into electrical signals. As shown in Figure 1A, both rods and cones are consisted of the outer segment and the inner segment containing ellipsoid, myoid, nucleus and synaptic terminal. Molecular mechanism that mediates detection of light and generation of electrical response (light response) in photoreceptor cells is localized in the outer segment, and the inner segment is responsible for the cell metabolism. Rods and cones are distinctive in their morphology of the outer segment.
Rods and cones are different in the properties of light responses (Figs. 1B and C). Figure 1B shows families of flash response evoked by various intensities of light flash in a carp rod (left) and a red-sensitive cone (right). A flash response in cones terminates more quickly than that in rods. As a result, the time resolution of a light stimulus is much higher in cones than in rods. Figure 1C shows a flash intensity-response relation in carp rods and  (right). The sketch shows the structure of each cell. OS, outer segment; e, ellipsoid; m, myoid; n, nucleus region; s, synaptic terminal region. (B) A flash response family of a rod (left) and a red-sensitive cone (right). Outer segment membrane current was recorded with a suction electrode by giving light flashes of various intensities at time 0 and they are superimposed. The light intensities in a rod used and expressed in the unit of % bleach were (from the dimmest intensity): 5.1 × 10 -6 % (corresponding to 8 molecules of visual pigment bleached per rod; 8 R*), 1.6 × 10 -5 % (25 R*), 5.1 × 10 -5 % (80 R*), 1.6 × 10 -4 % (250 R*) and 5.1 × 10 -4 % (800 R*). The light intensities in a red-sensitive cone used and expressed in the unit of % bleach were: 1.6 × 10 -3 % (2,100 R*), 5.2 × 10 -3 % (6,600 R*), 1.6 × 10 -2 % (21,000 R*) and 0.16 % (210,000 R*). (C) Light intensity-response relations of rods (open circles) and red-sensitive cones (closed circles) shown in (B). In (B), the responses were low-pass filtered at 50 Hz. (A) was modified from Figure 1 in Tachibanaki et al. (2007) and (B) and (C) were modified from Figure 1 in Kawamura and Tachibanaki (2008). light-sensitivity is >100-fold higher in rods than in cones. Due to this sensitivity difference, rods function in the dark, and cones in the light.
The mechanism that generates light responses is known as the phototransduction cascade (Fig. 2), a process that has been well studied in rods (Pugh and Lamb, 1993;Fu and Yau, 2007;Kawamura and Tachibanaki, 2008). Briefly, visual pigment (R) is activated by absorption of light, and activated visual pigment (R*) catalyzes the exchange of GDP for GTP in a trimeric GTP-binding protein, transducin (Tr). The GTP-bound form of Tr (Tr*) is the active form. Tr* interacts the inhibitory γ subunit of cGMP phosphodiesterase (PDE), and activates PDE (Wensel and Stryer, 1986). Activated PDE hydrolyzes cGMP, resulting in a decrease in the cytoplasmic cGMP concentration. Consequently, cGMP dissociates from a cGMP-gated cation channel situated in the plasma membrane of a rod outer segment; this closes the channel, thereby inducing a hyperpolarizing light response.
After a light stimulus, all activated species (such as R* and Tr*) are inactivated and the decreased level of cGMP is restored. R* is inactivated by phosphorylation and subsequent binding of arrestin (Chen et al., 1999;Xu et al., 1997). Tr* is inactivated by hydrolysis of Figure 2. Phototransduction cascade in rods and cones These reactions occur in the outer segment in rods and cones. Through these reactions, light-detection signals are converted into electrical signals. Red arrows show the reactions responsible for photoresponse generation and blue arrows show the reactions responsible for photoresponse termination. GRK, G-protein coupled receptor kinase; RGS9-1, regulator of G-protein signaling 9-1; GC, guanylate cyclase.
bound GTP to GDP via its intrinsic GTPase activity, which is accelerated by a photoreceptor-specific GTPase-accelerating protein, regulator of G-protein signaling 9-1 (RGS9-1) (Cowan et al., 1998). Cytoplasmic cGMP concentration is restored by synthesis from GTP by guanylate cyclase.
In the rod phototransduction cascade, enormous signal amplification takes place. In frog rods, when one molecule of rhodopsin is activated, a total of ~100 molecules of Tr are activated for 1 sec (Leskov et al., 2000). Furthermore, one molecule of PDE catalytic subunit hydrolyzes ~2000 molecules of cGMP for 1 sec (Dumke et al., 1994). These amplification reactions should contribute to high sensitivity to light in rods. In cones, a similar cascade is present, but in many animal species, a different set of players in the cascade (for example, cone-type visual pigment, Tr, PDE and cGMP-gated cation channel instead of their rod-types) is known to be present (Hisatomi and Tokunaga, 2002). It is probable that the efficiencies of the reactions in the cascade, including activation of R and Tr, and inactivation of R* and Tr*, are different between rods and cones, causing the sensitivity and time resolution to differ between rods and cones.
Previously, some of the reactions were compared between rods and cones. The efficiency of visual pigment activation by photon absorption was similar between chicken rhodopsin and iodopsin (red-sensitive cone visual pigment) (Okano et al., 1997). The catalytic activity of a single molecule of PDE was also similar between rod-type and cone-type, both purified from bovine retinas (Gillespie and Beavo, 1988). Moreover, the characteristics of cGMP-gated cation channels in catfish rods and cones were also similar (the Hill coefficient is ~2.3 for both types of channel; Haynes and Stotz, 1997). Thus, some of the reactions were already revealed to be similar between rods and cones.
To understand the mechanisms underlying the differences between the rod and cone light responses quantitatively at the molecular level, a method to purify rods and cones were established about 10 years ago in our laboratory using carp retina (Tachibanaki et al., 2001).
This method was the first report of the purification of cones to obtain in quantities large enough to do biochemical studies. The membranes prepared from purified rods and cones are suitable for the study on the reactions in the phototransduction cascade, because the proteins involved in this cascade (e.g. visual pigment, Tr and PDE) are remained in the membranes.
Using these membranes, some of the reactions in the phototransduction cascade were compared between rods and cones (Tachibanaki et al., 2001). In that report, to compare the efficiency of PDE activation between rods and cones, PDE activities to light flashes of various intensities were measured in rod and cone membranes and then, the maximum % PDE activity elicited by a light flash (peak PDE activity) was plotted against the flash intensity. In the presence of ATP, and therefore in the presence of R* phosphorylation, the light intensity required to evoke 50 % of full PDE activity was ~220-fold higher in cones than in rods (by contrast, in this study, the difference was found to be ~40-fold; see "Results and Discussion"). The molar ratio of PDE to R is similar between rods and cones (Tachibanaki et al., 2001 and this study); therefore, the aforementioned requirement of cones for a higher intensity of light means that the efficiency of PDE activation per photon is much lower in cones.
Recently, we reported that this lower effectiveness of PDE activation in cones was partly due to lower activation of Tr by R* (5-fold) in cones than in rods (Tachibanaki et al., 2012). However, the total difference in the effectiveness of PDE activation between rods and cones is not accounted for by this 5-fold difference. Therefore, in this study, I tried to determine which reactions are responsible for the much lower effectiveness of PDE activation in cones relative to rods. Firstly, I compared the efficiency of PDE activation by Tr* between rods and cones. Then I examined the contribution of Tr* and R* lifetimes to peak PDE activity elicited by a light flash to determine how their lifetimes affect the effectiveness of PDE activation in rods and cones.

Preparation of Rods and Cones
Carp (Cyprinus carpio: 25-30 cm in length) were purchased from Hirose Carp (Fukushima, Japan). Animal care was conducted according to the institutional guidelines.
Carp rods and cones were isolated as described previously (Tachibanaki et al., 2007) with some modifications. All manipulations were carried out in complete darkness with the aid of an infrared image converter (NVR 2015, NEC)  and then 3,000 × g for 4 sec). With this manipulation, the red blood cells were disrupted and their cytoplasm was removed as the supernatant. In contrast, rods and cones retain their morphology. Although the membrane fractions of red blood cells remained in the cone sample, I found that the red blood cell membranes do not influence the PDE activation. The resultant purified rods and cones were frozen immediately in liquid nitrogen and stored at -80C. These purified cells contained their outer segments and ellipsoids, but did not contain the nuclei and synaptic terminals.

Preparation of Rod and Cone Membranes
Purified rods and cones once frozen were thawed and washed 3 times with K-gluc buffer to remove their cytoplasmic fractions (centrifugation for 15 min at 100,000 × g at 4C). The precipitated membranes were suspended in K-gluc buffer and an aliquot of the suspension was used to quantify the amount of each type of visual pigment in rod and cone membranes. Quantification of each visual pigment was performed by assuming that the molar absorption coefficient is 40,000 M -1 cm -1 as described previously (Tachibanaki et al., 2001).
The membrane suspension thus obtained was stored at -80C until use.

Light Source
A light flash (Sunpak Auto 25SR) or a 150-W tungsten-halogen lamp was used to bleach visual pigments. In either case, a cut-off filter was used to pass light with wavelengths greater than 410 nm.

Purity of GTPγS
Commercially available GTPγS is contaminated with GDP (>20 % of GTPγS). To quantitate the amount of GTPγS as accurately as possible, the purity of each sample of purchased GTPγS was checked by HPLC using a Mono Q column (SMART System, GE Healthcare) (Hartwick and Brown, 1975). Elution was carried out with a linear gradient of KCl starting with buffer Qa (7 mM KH 2 PO 4 , pH 4.0) and finally with buffer Qb (250 mM KH 2 PO 4 , 500 mM KCl, pH 4.5). The eluted nucleotides were monitored by the absorbance at 254 nm. The purity of GTPγS was calculated from each peak area of eluted guanine nucleotides (most abundant nucleotide was regarded as GTPγS, and other nucleotides were identified with use of standards). The concentration of GTPγS indicated is its actual calibrated concentration, and that of GDP is also the actual concentration (for both GDP added to the sample and GDP present as a contaminant in the GTPγS reagent).

Analysis of Nucleotides
During the course of this study, I realized that nucleotides are hydrolyzed in my membrane samples. To know how much ATP or GTP remained in the sample at a certain incubation time, the time courses of ATP and GTP hydrolysis were measured. Each sample (25 μl) containing rod or cone membranes and various kinds of nucleotides was mixed with an equal volume of 10 % (w/v) trichloroacetic acid (TCA) to quench the GTPase and ATPase activities at a desired time. The sample was centrifuged for 15 min at 20,400 × g at 4C. A portion (40 μl) of the supernatant was mixed with 92.4 μl of 0.13 N NaOH for neutralization.
The amount of each nucleotide of the sample was quantified by HPLC using a Mono Q column or a Mini Q column as described above. The initial rate of the ATP or GTP hydrolysis was determined by fitting the time course with an exponential curve, where Y is the concentration of a remaining nucleotide (mM), A is the concentration of nucleotide at time 0 (mM), B is the initial rate of nucleotide hydrolysis (mM/sec), k (in sec -1 ) is a rate constant and t (in sec) is the time after the addition of nucleotides.

Preparation of GTPγS-bound Form of Rod Transducin
The stably active form of rod Tr with GTPγS-bound (rTr*-GTPγS) was used to activate PDE. (GTPγS or GTP binds to the α subunit of Tr, and the GTPγS-or GTP-bound form of the α subunit is responsible for the activation of PDE. Here, for simplicity, the abbreviation Tr* was used for indicating the molecular species that activates PDE.) The GTPγS-bound form of rod Tr (rTr*-GTPγS) was prepared and purified as described previously (Tachibanaki et al., 1997), with some modifications. Rods purified from 60-100 carp retinas were homogenized and fully bleached for 5 min on ice by using a 150-W tungsten-halogen lamp. Next, the homogenized rod membranes were washed twice with K-gluc buffer. The membranes were further washed with a low-ionic strength buffer (buffer A: 5 mM HEPES, 0.5 mM MgCl 2 , 1 mM DTT, pH 7.5) supplemented with 0.2 mM EDTA (ethylenediaminetetraacetic acid), and subsequently washed three times with buffer A. The washed membranes were suspended in buffer A containing 10 μM GTPγS to form rTr*-GTPγS, and then centrifuged to obtain rTr*-GTPγS in the supernatant. This extraction was repeated three times. Extracted rTr*-GTPγS was concentrated, and the buffer was replaced with K-gluc buffer by ultrafiltration using a VIVASPIN20 10,000 MWCO filter (GE Healthcare). The amount of rTr*-GTPγS was quantitated by the Coomasie Brilliant Blue (CBB) staining after SDS-PAGE, using bovine serum albumin (BSA) as a standard. Only three bands, at molecular masses corresponding to those of the Tr α, β and γ subunits, were visible in the CBB staining.

Purification of GTPγS-bound Form of Rod Transducin α Subunit
From the rTr*-GTPγS sample prepared as described above, I purified GTPγS bound form of rod Tr* α subunit (rTr*α-GTPγS) by using a Blue Sepharose CL6B column (GE Healthcare) as described previously (Heck and Hofmann, 1993) with some modifications.
Before loading the sample, the column was equilibrated with buffer B (10 mM HEPES, 5 mM MgCl 2 , 1 mM DTT, pH 7.5). The rTr*-GTPγS sample was filtered using a 0.22 μm membrane filter, and the filtrate was loaded on the column and proteins in the sample were eluted using a 0-1 M NaCl gradient. The eluted fractions of rTr*α-GTPγS were detected with SDS-PAGE and concentrated, and the buffer was replaced with K-gluc buffer by ultrafiltration as described above. The amount of rTr*α-GTPγS was quantitated by CBB staining after SDS-PAGE, using BSA as a standard.

Quantification of Activated Transducin
When necessary, I quantitated the amount of Tr* by using a [ 35 S]GTPγS filter-binding assay, as described previously (Tachibanaki et al., 2012).
Rod or cone membranes were first suspended in K-gluc buffer supplemented with 150 nM GDP to reduce the binding of GTPγS in the dark (see "Results and Discussion"). This suspension was irradiated for 1 min by using a 150-W tungsten-halogen lamp with neutral density filters (typically, 6-8 % rhodopsin was bleached), and the suspension was mixed with a solution containing [ 35 S]GTPγS, GDP, cGMP, and EGTA to obtain the GTPγS-bound form of Tr (Tr*-GTPγS). cGMP was added to quantitate the amount of Tr* under the same conditions used in the PDE activity measurements (see below). The final sample had a volume of 12 μl and contained 1.5 μM visual pigment, 5 mM cGMP, 0.8 mM EGTA, and 0.6 nM-1.8 μM [ 35 S]GTPγS. GDP was also added at a concentration four times higher than that of GTPγS; this was necessary because in the measurement of the dark control, GTPγS-binding increased significantly in cone membranes when GDP was not added (see "Results and Discussion"). GTPγS-binding in the dark was terminated 60 sec (rods) or 10 sec (cones) after the addition of GTPγS by adding 100 μl of K-gluc buffer containing both 50 mM NH 2 OH to inactivate R* and 10 mM cold GTP to terminate the apparent [ 35 S]GTPγS binding. In the light, a similar manipulation was performed to terminate the reaction 60 sec (rods) or 10 sec (cones) after the addition of GTPγS. Separate studies confirmed that GTPγS-binding in the light was completed by these times in both rod and cone membranes.
The samples were then filtered through a pre-rinsed nitrocellulose membrane by using a vacuum manifold. The material remaining on the filter membrane was washed three times with K-gluc buffer containing 25 mM MgCl 2 to remove unbound GTPγS, and the nitrocellulose membrane was then dried. The radioactivity of [ 35 S]GTPγS remaining on the membrane was quantitated by using an image analyzer (BAS 2000, Fuji). The manipulations described above were carried out at 20C. I also measured the time courses of Tr* activation by R* in rod and cone membranes. A rod or cone membrane suspension in K-gluc buffer (6 μl) was mixed with 6 μl of K-gluc buffer containing GTPγS. The final sample contained 3.0 (rods) or 1.0 (cones) μM visual pigment, 5 mM cGMP, 0.8 mM EGTA, 0.1 or 0.4 mM GDP and 0.1 mM GTPγS labeled with [ 35 S]GTPγS. After pre-incubation for 20 sec, the samples were irradiated by a light flash (Auto 25SR, Sunpak) with neutral density filters and a cut-off filter passing >410-nm light. By this irradiation, 0.027 % (rod) or 0.17 % visual pigment (cone) was bleached. Then, the reaction was quenched by adding 100 μl of K-gluc buffer containing both 50 mM NH 2 OH and 10 mM cold GTP at a desired time. The timing of addition was strictly controlled by using a rapid-quench apparatus that could terminate the reaction in less than 0.1 sec . After that, the amounts of bound GTPγS to Tr were measured as described above. The initial rate of the Tr activation was determined by fitting the results with an exponential curve, where Y is the amount of GTPγS bound to Tr per R* (GTPγS bound/R*), A is the initial rate of GTPγS binding (GTPγS bound/ R*•sec), k (in sec -1 ) is a rate constant and t (in sec) is the time after a light flash.

PDE Activity Measurement
PDE activity was measured by the pH assay method, as described previously (Tachibanaki et al., 2007). The pH drop caused by hydrolysis of cGMP was monitored with a combination glass microelectrode (MI-410, Microelectrodes). The concentration of cGMP hydrolyzed was calibrated by using the pH drops caused by full hydrolysis of known concentrations of cGMP. The range of pH drops during measurements was less than 0.2 pH units.
Four types of PDE activity measurements were performed under slightly different experimental conditions, as described below. All of the measurements were carried out at room temperature in a V-vial ® containing 100 μl of sample.
Without the addition of rTr*-GTPγS or rTr*α-GTPγS, PDE was not activated even in the light. In this type of measurement, the PDE activity was a function of added rTr*-GTPγS (rTr*α-GTPγS), so that to obtain the half-maximal concentration of rTr*-GTPγS, the relation between the amounts of added rTr*-GTPγS and the PDE activities observed was fitted with a Hill equation, where V is the PDE activity expressed as cGMP hydrolyzed per visual pigment present per second (cGMP/R•sec), V max is the maximum PDE activity (cGMP/R•sec), [S 1/2 ] is the concentration of rTr*-GTPγS giving 50 % of the maximum PDE activity (μM), [S] is the concentration of added rTr*-GTPγS (μM) and n is the Hill coefficient.

ii) PDE activation by endogenous Tr*
In the second type of measurements, PDE was activated by known concentrations of endogenous Tr* in rod and cone membranes. Rod or cone membranes suspended in K-gluc buffer were illuminated for 1 min with a continuous light (bleaching 6-8 % of the visual pigment per min), and then GTPγS was added at various concentrations (0.6 nM-1.8 μM) ranging from a concentration much lower than that of endogenous rod or cone Tr to a concentration much higher, in order to produce various concentrations of the stably active form of Tr (Tr*-GTPγS). When GTPγS concentration was very low and lower than the concentration of endogenous Tr, the amount of Tr*-GTPγS was limited by the amount of added GTPγS (Bruckert et al., 1994). In other words, in this case, the concentration of Tr*-GTPγS did not exceed the concentration of added GTPγS. The sample was kept in the light for 1 min, and the PDE activity measurement was initiated by adding a solution containing cGMP. The reaction mixture contained 1.5 μM visual pigment, 5 mM cGMP, 0.8 mM EGTA, 0.6 nM-1.8 μM GTPγS (previously added), and GDP as a contaminant at a concentration 0.67 times higher than that of GTPγS. PDE activity was determined from the rate of cGMP hydrolysis. Full PDE activity elicited by light was measured in a sample of the same batch used for each measurement in the presence of 10 μM GTPγS.

iii) PDE activation in the presence of GTP or GTPγS
In the third type of measurements, PDE activity was compared in the presence of GTP and in the presence of GTPγS in rod and cone membranes. PDE activity was measured by giving a light flash of various intensities in the presence of 0.1 mM GTP or GTPγS, and in the presence and absence of ATP. When necessary, the HEPES concentration was reduced (1.5-2.2 mM) to expand pH-drop signals. The reaction mixture contained 0.75 μM visual pigment, 2.5 mM cGMP, 0.8 mM EGTA, 0.4 mM GDP, 0.1 mM GTP or GTPγS, and (when necessary) 0.25 mM ATP in K-gluc buffer.
In this type of measurements, the pH drop caused by hydrolysis of ATP was corrected.
During incubation with cone membranes, hydrolysis of ATP induced a pH decrease of the sample solution. Because hydrolysis of cGMP was monitored by the pH assay method, the pH decrease caused by ATP hydrolysis recorded in a control measurement was subtracted from the pH record of the measurement of cGMP hydrolysis when ATP was present. To minimize ATP and/or GTP hydrolysis, nucleotides other than cGMP were added 10-15 sec before a light flash. Control studies showed that when the PDE activity reached its peak (1.5-3 sec in cone membranes and 3-8 sec in rod membranes after a light flash), ATP was present at a sufficient concentration (>0.13 mM, Fig. 13; see "Appendix"). In the presence of GTP, the PDE activity gradually increased after a light flash, reached its peak, and then declined to the dark level. The peak amplitude of the PDE activity obtained in the presence of GTP was a function of the flash intensity, and is referred to as peak PDE activity in the following. When GTPγS was used, cGMP hydrolysis increased linearly with time, so that this steady PDE activity was determined 10-20 sec after the flash in cone membranes, and 45-55 sec after the flash in rod membranes. Because PDE activity measured in the presence of GTPγS was also a function of the flash intensity, I also refer to this steady activity as peak PDE activity. Full PDE activity was measured by exposing the sample to light of saturating intensity.

iv) PDE activity measurement with trypsin activated PDE
In the fourth type of measurements, we compared the full PDE activity elicited by a

Elution of Activated Transducin in Rod Membranes
Membrane concentration-dependent elution of Tr*-GTPγS was measured in rod membranes with SDS-PAGE. Rod membranes were suspended in 5-100 μl of K-gluc buffer by varying the pigment (membrane)

Phosphorylation Assay in Rod Membranes
Phosphorylation assay was performed as described previously (Tachibanaki et al., 2007) with some modifications. Rod membranes were mixed with K-gluc buffer containing where Y is the number of phosphates incorporated per visual pigment bleached (Pi/R*), A is the initial rate of visual pigment phosphorylation (Pi/ R*•sec), k (in sec -1 ) is a rate constant and t (in sec) is the time after a light flash.

Construction, Expression and Purification of Recombinant PDE6Gs and PDE6H
To quantitate the amount of PDE γ subunits in rod and cone membranes by immunoblot, I expressed recombinant PDE γ subunits in E.coli as standard proteins. All clones were obtained from carp retinal cDNA library constructed previously . This cDNA library was constructed by using ZAP-cDNA Synthesis Kit (STRATAGENE), therefore, cDNAs were inserted between the EcoRI and XhoI restriction sites of pBluescript SK-vectors. The full lengths of the DNA sequences of carp PDE γ subunits are already known. PDE6G1 (GenBank accession number AB180754) and PDE6G2 (AB180755) were deposited as a nucleotide sequence of rod-type PDEγ. PDE6H (AB180756) was deposited as a nucleotide sequence of cone-type PDEγ.
Whole coding region of each carp PDE γ sequence was inserted into the NdeI/BamHI sites of pET 3a expression vector (Novagen).
Expression and purification of PDEγ were performed as described previously (Artemyev et al., 1998) with some modification. BL21(DE3) cells carrying each pET 3a-PDEγ plasmid were grown at 37C overnight in 3 ml of LB medium (10 g/l tryptone, 5 g/l yeast extract, 5 g/l NaCl) containing 50 μg/ml ampicillin. The overnight culture was added to one litter of bacterial culture (16 g/l tryptone, 10 g/l yeast extract, 5 g/l NaCl) with 50 μg/ml ampicillin in an environmental shaker at 37C until its OD 600 reaches to 0.8. Subsequently, 0.7 ml of 1 M IPTG was added and incubated for 3 h at 30C. Cells were collected as a pellet by centrifugation at 2,500 × g for 10 min at 4C, and washed two times with 100 ml of Wash buffer ( Purified PDEγ was then lyophilized, dissolved in 20 mM HEPES buffer (pH 7.5), and stored at -80C until use.

Preparation of Polyclonal Anti-sera
Anti-PDE6G1, PDE6G2 and PDE6H anti-serum were raised in mice against the glutathione S-transferase (GST)-fused proteins. To obtain the GST-fused PDE γ subunit, whole coding region of each carp PDEγ sequence was inserted into the EcoRI/SalI sites of pGEX 5X-1 vector (GE Healthcare).
For expression and purification of GST-fused PDEγ, BL21(DE3) cells carrying each pGEX 5X-1-PDEγ plasmid were grown at 37C overnight in 3 ml of LB medium containing 50 μg/ml ampicillin. The overnight culture was added to 500 ml of LB medium containing 50 μg/ml ampicillin in an environmental shaker at 37C until its OD 600 reaches to 0.5-1.0.
Subsequently, 0.25-0.5 ml of 1 M IPTG was added and incubated for 6 h at 37C. Cells were collected as a pellet by centrifugation at 2,500 × g for 10 min at 4C, and suspended in 50 ml of PBS buffer (137 mM NaCl, 2.7 mM KCl, 8.1 mM Na 2 HPO 4 , 1.5 mM KH 2 PO 4 , pH7.4) containing 1 % (w/v) TritonX-100. The suspension was sonicated 5 times for 1 min (UD200, TOMY), followed by centrifugation at 12,000 × g for 15 min at 4C. The supernatant was filtered through a 0.45 μm filter and loaded on a Glutathione Sepharose 4B column (GE healthcare) equilibrated with PBS buffer. The bound proteins were eluted using an elution buffer (10 mM glutathione in 50 mM Tris-HCl, pH 8.0).
Mice were immunized by intraperitoneal injections of 50-100 μg of each protein in Freund's complete adjuvant (Sigma). The immunization was repeated two times at an interval of 2 weeks with injection of the same amount of the protein in Freund's incomplete adjuvant.
The mice were bled 6-8 days after the final immunization.

Quantification of PDE6G1 in Rod Membranes by Immunoblot
An expression level of PDE6G1 in rod membranes was determined by immunoblot as described previously  using recombinant PDE6G1 as molar standards (see above). The amount of recombinant protein was quantitated by CBB staining after SDS-PAGE, using BSA as a standard.
PDE6G1 was probed immunologically and detected by Chemi-Lumi One L (nacalai tesque) using HRP labeled anti-mouse IgG antibodies as secondary antibodies. The content of a protein was expressed as the ratio to that of visual pigment.

PDE Activation by Exogenous Rod Transducin in Rod and Cone Membranes
To examine the activation efficiency of PDE by the active form of Tr in rod and cone membranes, firstly, I purified rods and cones from carp retina (Fig. 3). Briefly, rods and cones were brushed off the retina (Tachibanaki et al, 2001; Fig. 3A). The cells were purified by using a stepwise Percoll density gradient. Rods were collected at 45/60 % (w/v) interface (Fig.   3B), and cones were collected at 75-90 % (w/v) interface (Fig. 3C). Purified rods and cones were freeze-thawed and washed with K-gluc buffer to remove the cytoplasmic fractions. The samples were suspended with K-gluc buffer and used as a membrane suspension. slightly higher affinity for rTr*α-GTPγS than rod PDE, the difference was not significant.
The result indicated that rod Tr* activates cone PDE as effectively as it activates rod PDE.
The full PDE activities determined in the presence of saturating concentrations of rTr*α-GTPγS were similar in rod and cone membranes: 15.1 ± 0.6 (n = 8) cGMP hydrolyzed per visual pigment present per second (cGMP/R•sec) in rod membranes, and 15.0 ± 0.9 (n = 6) cGMP/R•sec in cone membranes. These activities were very similar to that of trypsin-activated PDE in rod membranes (17.4 ± 0.7 cGMP/R•sec, n = 12).
Because the specific activities of single molecule of rod and cone PDE are similar C (Gillespie and Beavo, 1988), the similar maximum PDE activities in rod and cone membranes described above indicated that the number of PDE molecules per visual pigment is similar in rod and cone membranes, as shown previously (Tachibanaki et al., 2001).
To examine the effect of Tr βγ subunits in rTr*-GTPγS, I purified rTr*α-GTPγS from rTr*-GTPγS sample which contained α, β and γ subunits of Tr (Figs. 5A and B). I measured PDE activities elicited by this purified rTr*α-GTPγS in rod membranes (Fig. 5C). Although PDE activation efficiency by rTr*α-GTPγS was modestly lower than that by rTr*-GTPγS, the difference was not significant. Therefore, I considered that the effect of Tr βγ subunits in rTr*-GTPγS on the PDE activation was negligible.
The result shown in Figure 4C indicates that the number of PDE molecules activated by a single molecule of exogenous rTr*α-GTPγS was similar in rod and cone membranes. In other words, this result strongly suggested that rod Tr* activates rod PDE and cone PDE with A similar efficiency. Therefore, I next tried to compare the efficiency of activation of rod and cone PDE by endogenous Tr*.

PDE Activation by Endogenous Transducin in Rod and Cone Membranes
To examine the efficiency of PDE activation by Tr* in rod and cone membranes, PDE in rod and cone membranes was activated by known amounts of endogenous Tr*, which was produced by adding limited amounts of GTPγS in the light. In most of these measurements, the concentration of GTPγS used was significantly lower than that of endogenous Tr molecules; therefore, under this condition, the number of Tr molecules activated (Tr*-GTPγS) was limited by the amount of added GTPγS. Because the sample contained 1.5 µM visual pigment (R) and the Tr:R ratio was ~1:10 (Tachibanaki et al., 2012), the Tr concentration in the sample was estimated to be ~150 nM.
Firstly, I calibrated the amount of the GTPγS-bound form of Tr. Figure  It was reported previously that some Tr* molecules were eluted from the membranes (Heck and Hofmann, 2001;Arshavsky et al., 2002). It was possible that eluted Tr* was less effective to activate PDE bound to the membrane. Accordingly, I examined how many Tr* molecules were eluted, and found that at low pigment concentrations I used, a significant proportion of Tr* was eluted from the membranes in rods (see Fig. 10 in "Appendix"). Although I could not detect the elution of Tr* in cone membranes, the reduction in the light-induced PDE activity, probably due to the elution of Tr*, was similar in rod and cone membranes at 0.75-1.5 μM pigment, the concentration I used (Fig. 11 in "Appendix"). For this reason, I believe that the extent of Tr* elution would have been similar between rod and cone membranes.
Based on the estimation of Tr*-GTPγS concentration in Figure 6A, I plotted PDE activities as a function of total Tr*-GTPγS (Fig. 6E). Because I was not certain whether the GTPγS-binding in cone membranes in the dark represented binding to Tr or to other proteins, the amount of Tr*-GTPγS in cone membranes was estimated in both ways: closed triangles are taken from the total GTPγS-binding in the light, and open triangles from the binding calculated from the light-dark differences in Figure 6A. Because GTPγS-binding in rod membranes in the dark was almost negligible, only the binding in the light is shown in Figure   6E (rod data; open black circles). The full PDE activity elicited by a light flash varied slightly in each measurement, so that the PDE activity expressed on the ordinate was normalized to the full activity in each measurement, but the average of the activity in rod membranes was similar to that in cone membranes. It is evident from Figure 6E that the efficiency of PDE activation, and therefore the number of PDE molecules activated by a single molecule of endogenous Tr*-GTPγS, was similar between rods and cones. In other words, the photon-capture signal is transmitted to PDE from Tr at a very similar efficiency between rods and cones in the carp retina.

Contribution of Tr* Lifetime to PDE Activity
The experiments described in the previous section demonstrated that each molecule of Tr* activates PDE at a similar efficiency in rods and cones. However, those experiments were performed with a stably active form of Tr (Tr*-GTPγS). In the actual situation in living cells, Tr* molecules accumulate with time as long as the visual pigment is active, and each Tr* molecule has its own lifetime. For this reason, the maximum number of PDE molecules activated, and thus the peak PDE activity elicited by a weak light flash, could be a function of the lifetimes of individual Tr* molecules activated and inactivated at different times after the light flash. In our previous study, we found that the lifetime of Tr* was 25-fold shorter in cones than in rods, probably due to higher expression of RGS9-1 in cones (Tachibanaki et al., 2012). Although the activation efficiency of PDE by Tr* could be almost the same in rods and cones, as shown in Figure 6, the number of PDE molecules activated by the same number of Tr* molecules could be lower in cones than in rods under conditions in which Tr* is inactivated.
To test this possibility, I compared peak PDE activities by giving a light flash under two different conditions: with GTPγS, in the absence of Tr* inactivation, and with GTP, in the presence of Tr* inactivation. I assumed that the amounts of Tr* remaining in, and eluted from, the membranes were similar between rod and cone membranes.
There were two problems in this measurement (see "Appendix"). One problem was a considerable PDE activation in the dark in cone membranes (typically, >50 % of full PDE activity was elicited within 1 min after the addition of 0.5 mM GTP). The reason for this increase in PDE activity in the dark was probably because Tr was activated in the dark to some extent (see Fig. 12A in "Appendix"). Based on this idea, I was able to decrease the PDE activation in the dark by increasing GDP concentration and decreasing GTP concentration (see Fig. 12B in "Appendix").
The other problem is the considerable changes in the concentrations of nucleotides in in cones than in rods, as predicted above. Thus, shorter Tr* lifetime in cone membranes contributes a 2-fold reduction in the effectiveness of PDE activation.
In Figure 7E, the difference in the effectiveness of PDE activation in the presence of both ATP and GTPγS between rod and cone membranes (filled circles and filled triangles) reflects the difference in the number of Tr* molecules activated by R* during its lifetime, because Tr* is not inactivated in the presence of GTPγS. This difference, therefore, could be due to the multiplied effect of the differences in both the activation of Tr by R* and the lifetime of R* in the presence of ATP. The overall difference was 21-fold (filled circles and filled triangles) and the Tr activation rate is 5-fold lower in cone membranes than in rod membranes (Tachibanaki et al., 2012). When this lower activation efficiency of Tr by R* is taken into account, I can predict that in the presence of R* phosphorylation, the lifetime of R* is shorter in cones than in rods, and that the shorter lifetime of R* in cones reduces the effectiveness of PDE activation by approximately 4-fold (21/5). The effectiveness of PDE activation in the presence of both ATP and GTP was 42-fold lower in cones than in rods (open circles and open triangles), and this difference was somewhat smaller than the value reported previously (220-fold). I believe that this difference is due to the difference in the concentration of GDP (see below).

Contribution of R* Phosphorylation on PDE Activation
The contribution of the lifetime of R*, described above, was estimated in the presence of R* phosphorylation. Therefore, it was of interest to determine how R* phosphorylation contributes to PDE activation. To estimate this contribution, I also measured peak PDE activities in the absence of ATP in rod and cone membranes in the presence of GTPγS and thus in the absence of Tr* inactivation (black traces in Figs. 7A-D). The measurements were made at various intensities of light flashes. In Figure 7F, the relationship between flash intensity and peak PDE activity was plotted for rod (black circles) and cone membranes (black triangles), and these data were compared with the results obtained in the presence of both ATP and GTPγS shown in (E) (dashed lines, re-plotted from connected lines in E). Because peak PDE activities were measured in the presence of GTPγS in these studies, the difference in the effectiveness of PDE activation should be due to the difference in the amount of Tr* produced during the lifetime of R* in the presence (+ATP) and absence (-ATP) of R* phosphorylation. In the absence of R* phosphorylation, relative to its presence, the effectiveness of PDE activation increased 3.8-fold in rod membranes and 9.5-fold in cone membranes (arrows). Apart from the fact that the lifetime of a bleaching intermediate is shorter in cones than in rods in living cells (Golobokova and Govardovskii, 2006), the effectiveness of R* phosphorylation was 2.4-fold (9.5/3.8) larger in cone membranes.
R* phosphorylation is catalyzed by G-protein coupled receptor kinase (GRK) (Kawamura and Tachibanaki, 2008). GRK is a peripheral membrane protein and localizes in the membrane fraction when rod or cone membranes are centrifuged . For this, I assumed that the phosphorylation rate was not dependent on the membrane concentration, but just to be sure, I examine the membrane concentration dependency of the phosphorylation rate in rod membranes (Fig. 8). In Figure 8A, the phosphorylation time course was measured in the presence of excess amount of R* (75 % visual pigment was bleached). From the result, the initial rate was obtained by fitting the phosphorylation data with an exponential curve (Fig. 8B). At the membrane concentration of 0.75 μM visual pigment, which I used in my study (Fig. 7), the initial rate was similar to that at the highest membrane concentration examined (15 μM visual pigment). Therefore, I concluded that the phosphorylation rate was not dependent on the membrane concentration.
In summary, under a pseudo-intracellular condition in which R* phosphorylation is present, the difference in the effectiveness of PDE activation between rods and cones, determined by peak activities, was found to originate from three steps in the phototransduction cascade. Firstly, at the stage when Tr* is generated by R*, the signal amplification is 5-fold lower in cones than in rods (Tachibanaki et al., 2012). Secondly, in the presence of R* phosphorylation, the shorter lifetime of R* in cones decreases the effectiveness of PDE activation 4-fold. Thirdly, the shorter lifetime of Tr* in cones decreases the effectiveness 2-fold. Overall, due to lower amplification and shorter lifetimes of R* and Tr*, the effectiveness of PDE activation was lowered in cones to ~1/40 (1/5 × 1/4 × 1/2; multiplied effect of lower Tr* activation, faster R* phosphorylation and faster Tr*

A B
inactivation) of the effectiveness in rods.

Signal Amplification Difference between Rods and Cones in Other Animals
Signal amplification in the phototransduction cascade has been compared between rods and cones previously (Pugh and Lamb, 1993) is taken into account, signal amplification seems to be comparable between mammalian rods and cones but this similarity does not seem to hold in amphibian rods and cones (Pugh and Lamb, 1993). In my study, I showed that the signal amplification at the stage of PDE activation by Tr* is the same between rods and cones (Fig. 6). In our previous study, signal amplification was 5-fold lower in cones than in rods at the stage of Tr activation by R*. The hydrolytic rate of a single PDE* molecule and the number of PDE molecules per visual pigment are very similar between rods and cones (Gillespie and Beavo, 1988;Tachibanaki et al., 2001). From these results we could conclude that the signal amplification is 5-fold lower in carp cones than in carp rods. There is a report that in mice, signal amplification may be lower in cones than in rods (Nikonov et al., 2006). Although further detailed studies are absolutely necessary, it may be the case that the rate of the reaction in the phototransduction cascade is different between rods and cones, and that the extent of this difference depends on the species.
However, this does not indicate that all Tr* molecules bind to PDE γ . As shown in Fig.   6E, maximum PDE activation was observed at ~100 nM Tr*-GTPγS at the 1.5 μM visual pigment concentration. In this measurement, 66 % of Tr*-GTPγS could have been eluted from the membranes and 34 % of them remained in the membranes (Fig. 10, see "Appendix").
In case the eluted Tr*-GTPγS does not have activity to activate PDE, ~30 nM is the lower limit of the concentration of Tr*-GTPγS that contributes to activate all of the PDE molecules present in the membranes. This concentration is higher than the concentration of PDE in the membranes; assuming that the molar ratio of holo-PDE to visual pigment is 1:270 in rods (Dumke et al., 1994), at 1.5 μM visual pigment concentration, the holo-PDE concentration is estimated to be 5.6 nM or the concentration of the catalytic subunit of PDE to be 11.2 nM.
This calculation shows that molar excess of Tr*-GTPγS is required for activation of PDE. A similar observation was reported in a previous study of rod membranes (Brukert et al., 1994).
To estimate the amount of holo-PDE with a different method, I tried to quantitate the amount of PDE γ by immunoblot in rod and cone membranes. Unfortunately among the three types of PDE γ , i.e. rod-type PDE γ (PDE6G1 and PDE6G2) and cone-type PDE γ (PDE6H) (see Experimental Procedures), I obtained only anti-PDE6G1 and anti-PDE6G2 antiserum. By immunoblot using these antisera, PDE6G1 was detected in rod membranes, but PDE6G2 was not detected at all. Then, I quantified the amounts of PDE6G1 in rods (Fig. 9). By quantitative immunoblot, I determined the amount of PDE6G1 to be 0.047 ± 0.008 (mean ± SE, n = 3) per rhodopsin present (Fig. 9C). Assuming that the molar ratio of holo-PDE to visual pigment is 1:270 in rods (Dumke et al., 1994), the amount of PDE γ is estimated to be 0.0074 (2/270) per rhodopsin. My value is 6-fold (0.047/0.0074) higher than this ratio. PDE γ is exclusively localized to the outer segment of photoreceptor cells as a subunit of holo-PDE (Tsang et al., 2006), and so there should be no contribution from the inner segment. My higher value may suggest that PDE γ is expressed more abundantly than the catalytic subunit of PDE. Alternatively, it is possible that the higher value may be due to underestimation of the amount of the standard protein that was quantitated by CBB staining using BSA as a molar standard. Obviously, further careful study is needed to quantitate the content of PDE in rod and cone membranes. In this study, the effectiveness of PDE activation was 42-fold lower in cone membranes than in rod membranes in the presence of both ATP and GTP. This difference in the effectiveness of PDE activation was lower than the value reported previously (220-fold; Tachibanaki et al., 2001). The reason for this inconsistency is not clear, but it is possible that in previous study, GTP hydrolysis in cone membranes was much higher than in rod membranes; the resultant GDP, which inhibits Tr activation (Figs. 12 and 14), may have reduced the effectiveness of Tr* production in cone membranes.

Summary
I showed that the efficiency of the PDE activation by Tr* is similar between rods and cones. However, due to the shorter lifetimes of R* and Tr* in cones, PDE activation becomes less effective in cones. Although the quantitative contribution of the reduced effectiveness of PDE activation to the reduced light sensitivity in cones needs to be determined, it is evident that the reduced rate of cGMP hydrolysis tends to shorten the light response time-to-peak, i.e., the time at which the hydrolysis and the synthesis of cGMP are balanced. Because photoreceptor light sensitivity is defined as the amplitude of a peak response, shortened time-to-peak contributes to lower light-sensitivity in cones than in rods.

Effect of Tr* Elution on PDE Activation in Rod and Cone Membranes
Tr and PDE are present in the membranes in the dark. When Tr is activated by R*, some Tr* molecules dissociate from the membranes and are found in a soluble fraction (Heck and Hofmann, 2001;Arshavsky et al., 2002). This behavior of Tr molecule is consistent with the finding of the light-dependent translocation of rod Tr from the outer segment to the inner segment (Sokolov et al., 2002). It is possible that eluted Tr* activates PDE less effectively than Tr* remaining in the membranes. To examine this possibility, under the condition used in this study, I measured the elution of Tr* from the rod membranes by SDS-PAGE (Fig. 10). In Figure 10B, a significant proportion of Tr α subunit, and probably βγ subunits also, were activity) and the activity of PDE treated with trypsin (trypsin-activated PDE activity) as a measure of the total activity of PDE in the sample (Fig. 11). At low visual pigment concentrations (0.75-1.5 μM), the light-induced PDE activity was 30-40 % of the trypsin-activated PDE activity in both rod and cone membranes. The reduction of the activity was comparable with the portion of Tr* remaining in the membranes shown in Figure 10. In rod membranes, at high visual pigment concentration (15 μM), the light-induced PDE activity was close (86 %) to the trypsin-activated PDE activity. However, at the same 15 μM pigment concentration, only about half (53 %), which was slightly higher than the value at 0.75-1.5 μM visual pigment (35 %), of the Tr* remained in the membranes (Fig. 10). Therefore, there seems to be a non-linear correlation between the portion of Tr* remaining in the membranes and the light-induced PDE activity. Because the reduction in the light-induced PDE activity was similar in rod and cone membranes at 0.75-1.5 μM visual pigment, the concentration I used in this study, the extent of Tr* elution would have been similar between rod and cone membranes.

Reduction in the Dark PDE Activity in Cone Membranes in the Presence of GTP or GTPγS
To measure the PDE activity in the presence of GTP or GTPγS, firstly I tried to repeat the previous experiment where 0.5 mM GTP was present (Tachibanaki et al., 2001). However, in the case of cone membranes, it was difficult to measure the PDE activity, because the PDE activity in the dark was significantly high even at the starting point of the measurement (>50 % of the full PDE activity). Therefore, I modified the experimental conditions. Figure 11. Membrane concentration dependent reduction of the full PDE activity in rod and cone membranes PDE activities were measured by the pH assay method at indicated membrane concentrations by giving a light flash bleaching ~10 % of the visual pigment with excess amount of GTPγS in rod and cone membranes. PDE activities were also measured by adding optimum concentrations of trypsin to activate all of PDE in a sample. PDE activities elicited by a light flash relative to those obtained in the trypsin-treated membranes were plotted against the membrane concentrations in rods (circles) and cones (triangles). Each data point indicates mean ± S.E. (n = 3).

rod cone
The reason for the increase in PDE activity in the dark was probably because Tr was activated in the dark to some extent. Indeed, GTPγS-binding in the dark was much higher in cone membranes than in rod membranes in the presence of 0.1 mM GDP (Fig. 12A). To decrease the Tr activation in the dark, I increased GDP concentration to 0.4 mM. Under this condition, GTPγS-binding in the dark decreased to 25 % in cone membranes (Fig. 12A). I checked whether this increase in the GDP concentration affects the Tr activation rate (Fig.   12B). By fitting the time course of GTPγS binding to Tr and therefore that of Tr activation with an exponential curve, the rate of Tr activation was decreased only by ~17 % in rod membranes (from 145 GTPγS bound/R*•sec, i.e. 145 Tr*/R*•sec, to 121 Tr*/R*•sec).
Although the GTPγS-binding was not measured at 0.1 mM GDP in cone membranes in my study, the rate of Tr activation measured at 0.4 mM GDP (18 Tr*/R*•sec) was close to the rate in the presence of 0.1 mM GDP measured in the previous study (30 Tr*/R*•sec, Tachibanaki et al., 2012). The difference of the rate of Tr activation between rod and cone membranes was 5 (145/30) at 0.1 mM GDP and 7 (121/18) at 0.4 mM GDP. From this result, I concluded that although GDP reduces the Tr activation rate, GDP affects Tr activation almost equally in rod and cone membranes irrespective of its concentrations within the range I used.

High Rate of Hydrolysis of Nucleotides in Cone Membranes
I examined the hydrolysis of nucleotides in cone membranes during a measurement, because purified cones retain large ellipsoid regions (Figs. 1A and 3C) that probably contain the hydrolytic enzyme(s) for ATP and GTP. These nucleotides are essential for the reactions  in the phototransduction cascade, and are added in my PDE activity measurement. If these nucleotides are hydrolyzed rapidly, interpretation of my data will be complicated. The concentration of a nucleotide was determined by HPLC. As I expected, ATP hydrolysis was significant in cone membranes in the presence of GTP or GTPγS (Fig. 13). By fitting the hydrolysis of ATP with an exponential curve, the initial rate of ATP hydrolysis was determined: it was 2.9 μM/sec in the presence of GTPγS and 9.0 μM/sec in the presence of GTP (Fig. 13F) in the cone sample containing 0.75 μM visual pigment. In Figure 13B, in the presence of GTPγS, GTP concentration increased gradually: it was probably because GTP was produced from GDP by transfer of a high energy phosphate from ATP.
I also examined the changes in the nucleotide concentration in cone membranes during a measurement in the absence of ATP and found that GTP was significantly hydrolyzed in cone membranes in the absence of ATP (Fig. 14). The initial rate of GTP hydrolysis was 3.3 μM/sec (Fig. 14B). Thus, nucleotide concentrations changed significantly in cone membranes during my measurements in cone membranes. Accordingly, I tried to complete each measurement as quickly as possible to minimize the changes in the concentration of Figure 14. GTP hydrolysis in cone membranes in the absence of ATP (A) Quantification of guanine nucleotides with Mini Q column chromatography. Cone membranes containing 0.75 μM visual pigment were mixed with GTP, and the reaction was quenched by addition of TCA at indicated times (0 sec, black trace, 60 sec, red trace; 240 sec, blue trace). After centrifugation (20,400 × g, 15 min), the supernatants were loaded onto a Mini Q column and nucleotides were detected by the absorbance at 254 nm. At the elution time of GDP, multiple peaks were observed probably because TCA was eluted at the same time. When the sample did not contain TCA, only a single peak was observed. Concentration of GTP was determined from the area of the elution peak of GTP. The area of the elution peak of GTP at 0 sec incubation was used as the initial concentration of GTP (0.5 mM). (B) The time course of GTP hydrolysis in cone membranes. The data points were fitted using an exponential curve (Y=A-B/k×[1-exp{-k×t}]). The best fitted A, B and k values were, respectively, 0.52 mM, 0.0033 mM/sec and 0.0053 sec -1 (dashed curve). Error bars show the deviation from the mean (n = 2). GDP GTP 0 sec 60 sec 240 sec A B nucleotides: in cone membranes, nucleotides except cGMP were added 10-15 sec before giving a light flash. In rod membranes, nucleotide concentrations did not change significantly so that at 30 sec before bleaching, nucleotides were mixed. In the preliminary experiment, nucleotide concentrations did not change significantly in purified cone outer segment membranes. These results suggest that the ellipsoid region in the inner segment contains substantial amounts of the hydrolytic enzymes for ATP and GTP compared with the outer segment.