Guanylate cyclase–activating protein 2 contributes to phototransduction and light adaptation in mouse cone photoreceptors

Light adaptation of photoreceptor cells is mediated by Ca2+-dependent mechanisms. In darkness, Ca2+ influx through cGMP-gated channels into the outer segment of photoreceptors is balanced by Ca2+ extrusion via Na+/Ca2+, K+ exchangers (NCKXs). Light activates a G protein signaling cascade, which closes cGMP-gated channels and decreases Ca2+ levels in photoreceptor outer segment because of continuing Ca2+ extrusion by NCKXs. Guanylate cyclase–activating proteins (GCAPs) then up-regulate cGMP synthesis by activating retinal membrane guanylate cyclases (RetGCs) in low Ca2+. This activation of RetGC accelerates photoresponse recovery and critically contributes to light adaptation of the nighttime rod and daytime cone photoreceptors. In mouse rod photoreceptors, GCAP1 and GCAP2 both contribute to the Ca2+-feedback mechanism. In contrast, only GCAP1 appears to modulate RetGC activity in mouse cones because evidence of GCAP2 expression in cones is lacking. Surprisingly, we found that GCAP2 is expressed in cones and can regulate light sensitivity and response kinetics as well as light adaptation of GCAP1-deficient mouse cones. Furthermore, we show that GCAP2 promotes cGMP synthesis and cGMP-gated channel opening in mouse cones exposed to low Ca2+. Our biochemical model and experiments indicate that GCAP2 significantly contributes to the activation of RetGC1 at low Ca2+ when GCAP1 is not present. Of note, in WT mouse cones, GCAP1 dominates the regulation of cGMP synthesis. We conclude that, under normal physiological conditions, GCAP1 dominates the regulation of cGMP synthesis in mouse cones, but if its function becomes compromised, GCAP2 contributes to the regulation of phototransduction and light adaptation of cones.

dark-adapted state after a transient light stimulus or preventing a closure of all cGMP-gated channels during continuous illumination.
It is believed that mouse rods express both GCAP1 and GCAP2, whereas mouse cones express only GCAP1 in their outer segments (9,10). In rods, GCAP1 and GCAP2 regulate the cGMP synthesis in a relay fashion. Early in the photoresponse or at dim background light, when the Ca 2ϩ level is only slightly lower than in darkness, GCAP1-mediated feedback dominates. Later in the photoresponse or at brighter background light, when Ca 2ϩ drops to lower levels, GCAP2-mediated feedback is also engaged (19,27). This model is consistent with the higher Ca 2ϩ affinity of GCAP2 compared with GCAP1 (18,28,29). Although previous studies have suggested that GCAP2 may not be substantially present in normal or Nrl Ϫ/Ϫ mouse cone outer segments (10,30,31), direct genetic and functional approaches have not been used to test whether GCAP2 has any physiological role in mouse cones. Here, we aimed to determine the contribution of GCAP1 and GCAP2 in mouse cone phototransduction and light adaptation by using a comprehensive electrophysiology, genetic, biochemistry, and single-cell immunohistochemistry study.

GCAP1 and GCAP2 are expressed in mouse cones
GCAP1 is expressed in outer segments of vertebrate rods and cones from zebrafish to human (8,9). However, the expression pattern of GCAP2 varies among different species (9). Previous studies have shown contradicting results regarding its presence in mouse photoreceptors (9,10,32). Thus, we sought to determine the expression pattern of GCAP2 in mouse cones by single-cell immunohistochemistry in retinas from wildtype (WT) control, Gcap1 Ϫ/Ϫ , Gcap2 Ϫ/Ϫ , and GCAP1/2 double knockout (Gcaps Ϫ/Ϫ ) mice. The top two panels of Fig. 1 demonstrate expression of GCAP2 in WT control and Gcap1 Ϫ/Ϫ cones based on the colocalization of mouse cone arrestin (mCAR; green) and GCAP2 (red) antibodies. Additional GCAP2 signal around the cones is from rod photoreceptors that sometimes surrounded the cones even after the mechanical cell isolation (see "Experimental procedures"). We observed overlap between the cone arrestin and GCAP2 signals in both WT control and Gcap1 Ϫ/Ϫ cones, suggesting that GCAP2 is expressed in mouse cones. As expected, the GCAP2 signal was not observed in Gcap2 Ϫ/Ϫ or Gcaps Ϫ/Ϫ cones (Fig. 1, bottom two panels), thereby confirming the specificity of the GCAP2 antibody (33). Together, these results clearly demonstrate that GCAP2 is expressed in mouse cones.

GCAP1 and GCAP2 regulate the kinetics and sensitivity of mouse cone phototransduction
To determine the specific roles of GCAP2 and GCAP1 in mouse cone phototransduction, we compared light responses of dark-adapted cones from WT control, Gcap1 Ϫ/Ϫ , Gcap2 Ϫ/Ϫ , and Gcaps Ϫ/Ϫ mice using ex vivo electroretinography (ERG) recordings. To isolate the cone photoreceptor component of the ex vivo ERG signal, we used synaptic blockers and Ba 2ϩ to remove b-and c-waves and rod-saturating background light.
Key experiments and the light adaptation studies were also done in a Gnat1 Ϫ/Ϫ genetic background to remove the rod component of the ERG signal. As has been shown previously (19), simultaneous deletion of GCAP1 and GCAP2 slowed down light response recovery and increased the sensitivity of cones to light flashes (see Fig. 2d and Table 1). Removal of GCAP1 alone increased time to peak (t p ) of the responses elicited by dim light (Fig. 2, d and h, and Table 1) and increased the sensitivity of cones almost as much as the deletion of both GCAP1 and GCAP2 ( Fig. 2d and Table 1). However, the recovery kinetics of the late tail phase of the responses in Gcap1 Ϫ/Ϫ cones was not decelerated for both dim flashes (Fig. 2d) and bright saturating flashes (Fig. 2e). Afterdepolarization, or response recovery overshoot, which was often present both in control and GCAP-deficient cones, prevented us from fitting an exponential function to the late tail phase of the responses to estimate the response recovery time constant ( rec ). However, the faster overall kinetics of Gcap1 Ϫ/Ϫ cone dim flash responses as compared with that of Gcaps Ϫ/Ϫ cones was demonstrated by their shorter integration time when compared with Gcaps Ϫ/Ϫ mice (Table 1). Isolating the cone component of the response by using Gnat1 Ϫ/Ϫ mice confirmed that the responses from GCAP1-deficient cones are still substantially faster than these of cones lacking both GCAP1 and GCAP2 (Fig. 2, f and g). These results suggest that GCAP2 can shape the light response kinetics specifically in brighter light, at least in the absence of GCAP1. In contrast, the sensitivity of dark-adapted cones to dim light flashes appears to be mediated mainly by GCAP1 ( Fig.  2h and Table 1). We also recorded light responses from Gcap2 Ϫ/Ϫ mice ( Fig. 2c) but did not find any significant changes of response kinetics or light sensitivity of GCAP2-deficient cones compared with WT controls (Fig. 2, d, e, and h,   Figure 1. GCAP2 is expressed in the mouse cones. Dissociated retinal cells from mice of the indicated genotypes were incubated with the mCAR antibody (green) to label cones followed by incubation with GCAP2 antibody (red). Arrowheads point to the positions of the cones in each field. Nuclei were stained with DAPI (blue). Images shown are representative of 38 cones from 14 fields (C57), 26 cones from 10 fields (Gcap1 Ϫ/Ϫ ), 10 cones from seven fields (Gcap2 Ϫ/Ϫ ), and nine cones from three fields (Gcaps Ϫ/Ϫ ). Scale bar, 20 m.

GCAP1 and GCAP2 regulate mouse cone phototransduction
and Table 1). Thus, we conclude that GCAP1 can support normal cone photoresponses in the absence of GCAP2.

GCAP2 promotes cGMP synthesis in low Ca 2؉ in mouse cones
Biochemical experiments have demonstrated that GCAP proteins activate cGMP synthesis of RetGCs in low Ca 2ϩ (1,6). Here, we asked whether GCAP2 could promote cGMP synthesis in intact mouse cones. To assess the Ca 2ϩ -mediated acceleration of cGMP synthesis in cones, we determined the change of the maximal saturated cone photoresponse amplitude (r max ) when the retinas were switched from normal perfusion solution with1.2 mM [Ca 2ϩ ] o to low ϳ30 nM [Ca 2ϩ ] o in ex vivo ERG experiments. Such a treatment causes rapid reduction in the level of Ca 2ϩ in photoreceptor outer segments and the subsequent GCAP-mediated up-regulation of cGMP synthesis (41). The r max is proportional to the cGMP-gated channel current, and thus, increased cGMP concentration caused by accelerated cGMP synthesis rate is expected to increase r max . We determined r max from saturated cone responses elicited by periodic bright test flashes in dark-adapted mouse retinas before and after low-Ca 2ϩ exposure. In control Gnat1 Ϫ/Ϫ retinas, r max increased about 4-fold after a low-Ca 2ϩ exposure (Fig. 3, black squares), demonstrating the up-regulation of cGMP synthesis and subsequent opening of the cGMP-gated channels. However, the cells could not maintain such a high cGMP-gated (CNG) channel channel current for long, and eventually r max declined under low Ca 2ϩ . When cones lacking both GCAP1 and GCAP2 (from Gcaps Ϫ/Ϫ Gnat1 Ϫ/Ϫ retinas) were exposed to low Ca 2ϩ , a much more subtle increase of r max was observed (Fig. 3, green squares), consistent with the lack of up-regulation of cGMP synthesis in low Ca 2ϩ in the absence of both GCAPs. Notably, when we exposed Gcap1 Ϫ/Ϫ Gnat1 Ϫ/Ϫ retinas to low Ca 2ϩ , we observed substantial increase in r max that was comparable with that in control Gnat1 Ϫ/Ϫ mice (Fig. 3, red squares).  Table 1 Light (flash) response parameters from WT, Gcaps ؊/؊ , Gcap1 ؊/؊ , and Gcap2 ؊/؊ mouse cones All recordings were from Gnat1 ϩ/ϩ mice except for the light adaptation parameters I 0 and n, which were obtained from Gnat1 Ϫ/Ϫ mice. Retinas were exposed to constant 70,000 photons (530 nm) m Ϫ2 s Ϫ1 background light to suppress the rod component of the response except in the light adaptation experiments that were from Gnat1 Ϫ/Ϫ mice. r max , saturated photoresponse amplitude (I F ϭ 183,000 photons m Ϫ2 at 530 nm); t p , time to peak (I F ϭ 220 photons m Ϫ2 at 530 nm); t i , integration time defined as an area under a dim flash response divided by the amplitude of the response; I1 ⁄ 2 , light flash intensity eliciting a response with peak amplitude r ϭ 0.5r max determined by fitting Equation 1 to the response amplitude data; I 0 , background light intensity at which cone sensitivity is 50% of that in darkness determined by fitting Equation 2 to the light adaptation data; n, steepness factor determined by fitting Equation 2 to the light adaptation data. * and † indicate statistically significant difference as compared with the WT control and Gcaps Ϫ/Ϫ mouse cones, respectively (p Ͻ 0.05, two-tailed Student's t test). NA, not available.

GCAP1 and GCAP2 regulate mouse cone phototransduction
These results demonstrate that Ca 2ϩ feedback mediated by GCAP2 can promote acceleration of the cGMP synthesis in intact mouse cones in the absence of GCAP1.

GCAP2 contributes to mouse cone light adaptation in bright background light
GCAP-mediated Ca 2ϩ feedback dominates the regulation of rod and cone photoreceptor sensitivity in response to fast increments or decrements of background light (19,34). However, the distinct contributions of GCAP1 and GCAP2 to the light adaptation capacity of mouse cones is not known. To address this question, we determined how the sensitivity of cones is regulated by background light in isolated retinas from control mice expressing both GCAPs and from mice lacking either both GCAPs or only GCAP1. All mice were on a Gnat1 Ϫ/Ϫ background to eliminate rod signaling and facilitate the quantification of cone light adaptation. When mouse cones are exposed to a step of light, they produce an initial hyperpolarizing response peak followed by partial relaxation to a plateau ( Fig. 4, a-c). This relaxation was attenuated after removal of both GCAPs (Fig. 4b), consistent with the dominant role of the GCAP-mediated feedback in cone light adaptation. Notably, GCAP1-deficient cones exhibited prominent relaxation after the peak of the response comparable with that in control cones, indicative of efficient light adaptation. We quantified the relaxation magnitude and kinetics by fitting a sum of two exponential functions from the peak to the plateau of the step responses using Equation 3 (see Fig. 4, a-c). Although we used a two-exponential function, the relaxation was dominated by the exponential term with the faster of the two time constants ( 1 ). Thus, we used 1 to assess the kinetics of relaxation and the amplitude from peak to the plateau (A) normalized by the peak amplitude (r 0 ) to assess the magnitude of relaxation (see Equation 3). The A/r 0 was similar between control (76 Ϯ 1%) and Gcap1 Ϫ/Ϫ (80 Ϯ 1%) mice but significantly smaller in the absence of both GCAP1 and GCAP2 (27 Ϯ 5%). This result demonstrates that the expression of GCAP2 in GCAP1-deficient cones was sufficient to promote robust light adaptation as demonstrated by the substantial relaxation of their response in steady background light. However, the kinetics of the relaxation was decelerated significantly by the deletion of GCAP1 alone (from 165 Ϯ 30 ms in control to 495 Ϯ 20 ms in Gcap1 Ϫ/Ϫ mice), whereas the value for 1 was not statistically significantly different in Gcaps Ϫ/Ϫ cones (423 Ϯ 30 ms) as compared with that in Gcap1 Ϫ/Ϫ cones. This result is consistent with the dominant role of GCAP1 in driving the rapid light adaptation of mouse cones.
To quantify the efficiency of light adaptation, we measured the sensitivity of cones to light flashes at 4.5 s after the step onset at different background light intensities. The sensitivity normalized to the sensitivity in darkness declined more steeply in Gcaps Ϫ/Ϫ cones than in control or Gcap1 Ϫ/Ϫ cones (Fig. 4d). As expected, based on their higher sensitivity in darkness, the operating range of Gcap1 Ϫ/Ϫ cones was shifted to dimmer background light intensities. However, the slope of the adaptation curve was not changed, and the adaptation capacity was clearly better in Gcap1 Ϫ/Ϫ mice than in Gcaps Ϫ/Ϫ mice. These results indicate that GCAP2 can contribute to the light adaptation of mouse cones in the absence of GCAP1. We did not have Gcap2 Ϫ/Ϫ Gnat1 Ϫ/Ϫ mice. Thus, in an effort to investigate the role of GCAP2 in light adaptation, we compared light adaptation between Gcap2 Ϫ/Ϫ and WT mice (on Gnat1 ϩ/ϩ background). In those experiments, we did not observe any change in light adaptation caused by the deletion of GCAP2 (data not shown), consistent with our flash response data showing only negligible phenotype in GCAP2-deficient cones (Fig. 2, c-e and h).

GCAP1 and GCAP2 regulate mouse cone phototransduction GCAP1 and GCAP2 compete for activation of RetGC1 in low Ca 2؉
Our results demonstrate that GCAP2 is expressed in mouse cones and that it can contribute to the Ca 2ϩ -dependent activation of RetGCs and phototransduction feedback in their outer segments. However, it remained unclear whether the expression level of GCAP2 in WT cones is sufficient to contribute to the overall Ca 2ϩ feedback. A simple biochemical model (see "Experimental procedures" for details) predicts that a quite small concentration of GCAP2, ϳ0.1-0.5 M in the cone outer segment, could explain the ϳ4-fold increase of r max in low Ca 2ϩ observed in Gcap1 Ϫ/Ϫ cones (see Fig. 3). Based on the model prediction, we designed a biochemical experiment to assess the extent of activation of the native RetGC1 (the predominant guanylate cyclase isozyme expressed in the cones (30,31,35)) by recombinant GCAP1 and GCAP2. We used photoreceptor membranes from Gcaps Ϫ/Ϫ RetGC2 Ϫ/Ϫ mouse retinas retaining only RetGC1 isozyme to measure cGMP synthesis by RetGC1 in low Ca 2ϩ at normal physiological 0.9 mM Mg 2ϩ (36). Consistent with our model, the low basal activity of RetGC1 was significantly increased by addition of either 0.5 M GCAP2 (derived from the biochemical model) or 3 M GCAP1 (the estimated GCAP1 concentration in mouse rods (29)) (Fig. 5). Next, we assessed whether GCAP2 can contribute to the regulation of RetGC1 activity in the presence of GCAP1. To test this, we used M26R GCAP1, a mutant form that can bind to RetGC1 like the WT GCAP1 but does not activate it (37,38). In the presence of 3 M GCAP1, addition of M26R GCAP1 started to decrease RetGC1 activity at ϳ0.3 M (Fig. 5, red circles), its near-physiological concentration (28,29), and reached halfmaximal inhibition at 1 M. At the same concentration of M26R GCAP1, the activation of RetGC1 by 0.5 M GCAP2, which in the absence of GCAP1 would be sufficient to effectively accelerate RetGC1 in vivo (Figs. 2 and 3), was almost completely suppressed (Fig. 5, black circles).

Ca 2؉ -dependent regulation of cGMP synthesis by GCAP2 in mouse cone photoreceptors
Our experiments clearly demonstrate that GCAP2 is expressed in mouse cones (Fig. 1). To address the possible functional role of GCAP2 in cones, we investigated its ability to up-regulate cGMP synthesis in low Ca 2ϩ and to mediate light adaptation in cones lacking GCAP1. As previous studies have suggested that GCAP1 and RetGC1 dominate the synthesis of cGMP in the mouse cone outer segments, we expected that Gcap1 Ϫ/Ϫ retinas would respond to low-Ca 2ϩ exposure similarly to Gcaps Ϫ/Ϫ retinas (9,10,30,31,35). However, Gcap1 Ϫ/Ϫ cones were able to boost their maximal response amplitude in low Ca 2ϩ as much as control WT cones (Fig. 3). Based on our model presented under "Experimental procedures," as low a concentration as 0.1 M GCAP2 in the outer segments of Gcap1 Ϫ/Ϫ cones could explain the ϳ4-fold increase of their maximal response amplitude in low Ca 2ϩ . This concentration is more than 10-fold lower than the known GCAP1 or GCAP2 concentration in mouse rod outer segments (28,29). The quantitative power of these experiments might be limited due to the cooperativity of the CNG channel for cGMP (39,40) or the transient nature of the increase in photoreceptor response amplitude in low Ca 2ϩ (41-45). However, despite the quantitative limitations of our study, our results clearly demonstrate that GCAP2 can activate RetGC in mouse cone photoreceptor cells when GCAP1 has been deleted.
Similarly, when we examined light adaptation in Gcap1 Ϫ/Ϫ cones, we found that the slope of the light adaptation curve was comparable with that in control cones. In addition, the adaptation capacity of GCAP1-deficient cones was substantially better than that of Gcaps Ϫ/Ϫ cones (Fig. 4). Together, these results demonstrate that GCAP2 is able to up-regulate cGMP synthesis and to mediate light adaptation in cones in the absence of GCAP1.

The role of GCAP1 and GCAP2 in cone phototransduction and light adaptation
The relative contribution of GCAP1 and GCAP2 in rod physiology has been established in mouse rod photoreceptors (27). There, GCAP1 is more important in determining the peak amplitude of the dim flash response, whereas GCAP2 shapes the response recovery kinetics after the peak amplitude. These results are consistent with the known biochemical properties of GCAP1 and GCAP2. Namely, GCAP2 has a higher affinity to Ca 2ϩ (K Ca ϭ 50 nM) as compared with GCAP1 (K Ca ϭ 130 nM) (18,28,29). In darkness, Ca 2ϩ concentration in mouse rod outer segment is ϳ250 nM, and it declines to ϳ20 -50 nM in bright light (46). Hence, after a dim flash, Ca 2ϩ dissociates first from GCAP1, and the GCAP1-mediated feedback dominates over the GCAP2-mediated pathway. Later, when Ca 2ϩ has dropped to a lower level, it can also dissociate from GCAP2, up-regulating the GCAP2-mediated feedback to contribute to the recovery phase kinetics of the dim flash response. Notably, the primary target for GCAP1 in mouse photoreceptors is RetGC1, whereas regulation of the ancillary isozyme RetGC2 is carried out mostly by GCAP2 (47). Hence, in mouse rods, acti-

GCAP1 and GCAP2 regulate mouse cone phototransduction
vation of the cyclase after the flash of light occurs first as activation of RetGC1 by GCAP1 followed by additional activation of RetGC1 and RetGC2 by GCAP2 (27). Here, we compared dark-adapted cone flash responses from WT, Gcaps Ϫ/Ϫ , Gcap1 Ϫ/Ϫ , and Gcap2 Ϫ/Ϫ mice to understand the relative contributions of GCAP1 and GCAP2 in determining the sensitivity and response kinetics of mammalian cones (Fig. 2). We found that deletion of GCAP1 causes a comparable increase of the sensitivity and dim flash response amplitude as the deletion of both GCAP1 and GCAP2 (Fig. 2, d and h, and Table 1). Thus, just as in rods, GCAP1 seems to dominate the up-regulation of cGMP synthesis up to the peak of the dim flash response, and the sensitivity of cones is set almost completely by GCAP1.
Comparison of saturated bright flash responses from WT, Gcaps Ϫ/Ϫ , Gcap1 Ϫ/Ϫ , and Gcap2 Ϫ/Ϫ mice revealed that deletion of both GCAP1 and GCAP2 significantly delays the escape of cones from saturation, whereas the deletion of GCAP1 had a much less dramatic effect, and the deletion of GCAP2 had almost no effect at all on the recovery kinetics (Fig. 2e). Notably, the recovery kinetics of Gcap1 Ϫ/Ϫ cones were not slower than those of WT cones so that cone responses from Gcap1 Ϫ/Ϫ mice recovered to the baseline level at the same time as those of WT and Gcap2 Ϫ/Ϫ mice (Fig. 2d). Thus, it appears that both GCAP1 and GCAP2 can compensate for the lack of the other isoform in accelerating the recovery of bright flash responses (Fig. 2e). These results are also consistent with the idea that a larger drop in Ca 2ϩ caused by brighter light is required to activate the GCAP2 pathway. In support of this notion, we observed deviation between Gcap1 Ϫ/Ϫ and Gcaps Ϫ/Ϫ mouse light adaptation only at brighter background light. This, again, suggests that GCAP2 is more important under brighter illumination and at lower Ca 2ϩ .
Although our data clearly show that GCAP2 contributes significantly to the physiology of mouse cones in Gcap1 Ϫ/Ϫ mice, it is not clear whether GCAP2 plays a role in the phototransduction and/or light adaptation of healthy WT cones. Evidently, GCAP2 is present in native mouse and Nrl Ϫ/Ϫ cones at much lower levels than in rods, whereas GCAP1 expression in cones is very strong (5,10,31). However, our functional data from Gcap1 Ϫ/Ϫ mice could be explained even by a rather low 0.1-0.5 M GCAP2 concentration in the absence of GCAP1. In contrast, our biochemical experiments assessing the relative contribution of the two GCAP isoforms show that, even at equal concentrations of the two GCAP isoforms, GCAP1 effectively outcompetes GCAP2 from RetGC1 (see Fig. 5 and Refs. 28 and 38). Assuming further that GCAP2 expression in mouse cones is lower than that of GCAP1, we conclude that under normal physiological conditions GCAP1 would dominate the regulation of cGMP synthesis in mouse cones. However, if the function of GCAP1 becomes compromised, GCAP2 should be able to effectively regulate the phototransduction feedback and light adaptation of cones.

Ethical approval
All experimental procedures were in accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committees at Washington University in St. Louis, Salus University, and University of Southern California.

Animals
WT C57Bl/6J control and age-matched adult mice devoid of guanylate cyclase-activating protein 1 (Gcap1 Ϫ/Ϫ (48)), 2 (Gcap2 Ϫ/Ϫ (49)), or both (Gcaps Ϫ/Ϫ (19)) were used in this study. The mutant strains were bred to the control C57Bl/6J background for several generations but were not siblings of the control mice. For some electrophysiology experiments, Gcap1 Ϫ/Ϫ and Gcaps Ϫ/Ϫ mice were bred into a Gnat1 Ϫ/Ϫ background to remove the rod-driven light responses (50). Mice were kept under a 12/12-h light/dark cycle and had free access to regular mouse chow and clean water.

Single-cell immunohistochemistry
Freshly dissected retinas from C57, Gcap1 Ϫ/Ϫ , Gcap2 Ϫ/Ϫ , and Gcaps Ϫ/Ϫ mice were washed in Ames' medium, placed on an ice-cooled glass slide with a few drops of cold Ames' buffer, and chopped with razor blade. Dissociated cells and small cell clumps were collected into 8-chamber slides (Lab-Tek, catalog number 177445) that were precoated with wheat germ agglutinin (100 M wheat germ agglutinin was added to the wells and incubated for 1 h). After cells were collected into wells, equal volumes of formaldehyde (4% in PBS) were added. The slides were centrifuged for 10 min at 168 ϫ g to attach the cells to the glass surface. Cells were washed in PBS; blocked with 5% goat serum, 0.1% Triton X-100 in PBS for 1 h; and incubated overnight with a rabbit polyclonal anti-mCAR antibody (51) (1:700 in blocking buffer). The next day, slides were washed in PBS and incubated with a secondary anti-rabbit antibody to visualize mCAR-labeled cones. Following PBS washes, cells were incubated with biotinylated anti-GCAP2 antibody (33) (1:300 of 1 mg/ml in blocking buffer). The GCAP2 signal was visualized by Texas Red-avidin (1:200; Vector Laboratories). The cells were mounted in Vectashield with DAPI (Vector Laboratories). Fluorescence images were acquired using a Zeiss Axio Scope microscope using the same settings and exposure times for the different genotypes.

Ex vivo electroretinography
We used ex vivo ERG to assess the function of mouse cone phototransduction and light adaptation (52). Either a background light of 70,000 photons (530 nm) m Ϫ2 s Ϫ1 or Gnat1 Ϫ/Ϫ genetic background (53) was used to remove the rod component of the ERG signal. The Gnat1 Ϫ/Ϫ mouse rods do not respond to light but maintain normal morphology. The background light needed to fully saturate rods was surprisingly high and would have been expected to bleach a significant amount of pigments during our experiments. However, after about 10 min of exposure to the background light, the cone responses remained stable for up to at least 2 h (the longest experiment), potentially due to a balance between pigment bleaching and regeneration via the Müller cell (54) visual cycle pathway. Retinas were dissected from dark-adapted eyes under IR illumination and mounted to a custom-built ERG specimen holder described in Vinberg et al. (52). Flashes and steps of light GCAP1 and GCAP2 regulate mouse cone phototransduction were provided by green LEDs (530 nm; Luxeon Rebel LED SR-01-M0090) via an inverted microscope light path where the condenser was replaced by a 10ϫ objective forming a homogenous 2.35-mm spot of light at the sample. The intensity of the light stimulus was calibrated at the level of the sample by a photometer (Model 211, UDT Instruments). Retinas were perfused at 1 ml/min at 37°C with bicarbonate-buffered Locke's solution containing 112 mM NaCl, 3.6 mM KCl, 2.4 mM MgCl 2 , 1.2 mM CaCl 2 , 10 mM HEPES, 20 mM NaHCO 3 , 3 mM disodium succinate, 0.5 mM sodium glutamate, and 10 mM glucose. The solution was equilibrated with 95%O 2 and 5%CO 2 at 37°C. Low-Ca 2ϩ solution was prepared by using 0.1 mM CaCl 2 instead of 1.2 mM and adding 0.4 mM EGTA. Addition of EGTA caused acidification of the medium, and we used NaOH to equalize the pH of our normal Locke's and low-Ca 2ϩ media. We estimate that the free [Ca 2ϩ ] of the low-Ca 2ϩ medium is ϳ30 nM in the presence of 2.4 mM Mg 2ϩ (55).
A differential amplifier (DP-311, Warner Instruments) and Bessel filter (model 3382, Krohn-Hite Corp.) together with a DigiData 1440 digitizer and pCLAMP software (Axon Instruments) were used to acquire data at 10 kHz with a 300-Hz lowpass filter. Clampfit (Axon Instruments), Origin 9.0.0 (Originlab), and Excel (Microsoft) software were used to analyze and graph the data. A Naka-Rushton function was fitted to the response amplitude (r) data.
where r max is the maximal saturated response amplitude, I F is flash intensity, and I1 ⁄ 2 is the light intensity (in photons m Ϫ2 ) required to elicit a half-maximal response. A modified Weber-Fechner function was fitted to light adaptation data.
where S F is the sensitivity of cones to a flash of light (I F that elicits r Ͻ 0.2r max ) defined as r/I F , S F,D is the sensitivity in darkness, I is the background light intensity (in photons m Ϫ2 s Ϫ1 ), I 0 is the background light intensity in which S F ϭ 0.5S F,D , and n is a factor determining the steepness of the adaptation curve. A sum of two exponential functions was used to quantify the kinetics and magnitude of light response relaxation after the initial peak during light steps.
where r 0 is peak amplitude measured at t d , A is amplitude measured from the peak to the steady-state plateau of the step response, A 1 is the fraction of recovery covered by the time constant 1 , and (A Ϫ A 1 ) is the fraction of the recovery covered by the time constant 2 .

Biochemical model of RetGC1 activation by GCAP2
We used the following equations to model binding of Ca 2ϩ to GCAP2 and binding/activation of RetGC1 by Ca 2ϩ -free GCAP2. The parameter values were taken from Peshenko et al. (28).
where K Ca ϭ 50 nM is the apparent dissociation constant of Ca 2ϩ from GCAP2. We model the activation of RetGC1 by GCAP2 by assuming that only Ca 2ϩ -free GCAP2 can activate the RetGC1.
where K GC1 ϭ 1.25 M and GC1 total ϭ 3.2 M. Cyclase activity (␣; in M s Ϫ1 ) can be calculated as follows.
if we assume that GTP (the substrate) Ͼ Ͼ K m(GTP-GC) (dissociation constant of the GTP from RetGC1). We assume that the basal RetGC1 activity k n1 ϭ 2.6 s Ϫ1 and for the activated RetGC1 k s1 ϭ 33 s Ϫ1 (28 where ␤ ϭ 4.1 s Ϫ1 is the spontaneous cGMP hydrolysis activity of rod PDE in darkness (56). The CNG channel current (57) can be calculated as follows.

(Eq. 11)
where J max is the CNG channel current at high [cGMP]. Assuming that [Ca 2ϩ ] is 250 nM in a dark-adapted mouse cone outer segment under normal extracellular Ca 2ϩ and declines to 25 nM during our low-Ca 2ϩ exposure (see above), as low as a 0.1 M total concentration of GCAP2 in the cone outer segment is predicted to cause a 4.4-fold increase of J cG when switched from normal (1.2 mM) Ca 2ϩ to low Ca 2ϩ .

Expression and purification of GCAPs
We used recombinant mouse myristoylated GCAP1 (E6S) and GCAP2 expressed from pET11d vector (Novagen/Calbiochem) in BLR(DE3) Escherichia coli strain harboring yeast N-myristoyltransferase as described previously (28). GCAP2 GCAP1 and GCAP2 regulate mouse cone phototransduction was purified using urea extraction from the inclusion bodies and size-exclusion chromatography (26,58). GCAP1 was purified using urea extraction and hydrophobic and size-exclusion column chromatography as described previously to reach a final protein of 95% purity by SDS-PAGE (28,59). The M26R bovine GCAP1 mutant was produced and purified as described previously (37,38).