Activation and Inhibition of Photoreceptor Guanylyl Cyclase by Guanylyl Cyclase Activating Protein 1 (GCAP-1)

Guanylyl cyclase activating protein 1 (GCAP-1), a Ca2+/Mg2+ sensor protein that accelerates retinal guanylyl cyclase (RetGC) in the light and decelerates it in the dark, is inactive in cation-free form. Binding of Mg2+ in EF-hands 2 and 3 was essential for RetGC activation in the conditions mimicking light adaptation. Mg2+ binding in EF-hand 2 affected the conformation of a neighboring non-metal binding domain, EF-hand-1, and increased GCAP-1 affinity for RetGC nearly 40-fold compared with the metal-free EF-hand 2. Mg2+ binding in EF-hand 3 increased GCAP-1 affinity for RetGC 5-fold and its maximal RetGC stimulation 2-fold. Mg2+ binding in EF-hand 4 affected neither GCAP-1 affinity for RetGC, nor RetGC activation. Inactivation of Ca2+ binding in EF-hand 4 was sufficient to render GCAP-1 a constitutive activator of RetGC, whereas the EF-hand 3 role in Ca2+-dependent deceleration of RetGC was likely to be through the neighboring EF-hand 4. Inactivation of Ca2+ binding in EF-hand 2 affected cooperativity of RetGC inhibition by Ca2+, but did not prevent the inhibition. We conclude that 1) Mg2+ binding in EF-hands 2 and 3, but not EF-hand 4, is essential for the ability of GCAP-1 to activate RetGC in the light; 2) Mg2+ or Ca2+ binding in EF-hand 3 and especially in EF-hand 2 is required for high-affinity interaction with the cyclase and affects the conformation of the neighboring EF-hand 1, a domain required for targeting RetGC; and 3) RetGC inhibition is likely to be primarily caused by Ca2+ binding in EF-hand 4.

concentration of Ca 2ϩ from near 250 nM in the dark to near 25 nM in the light (5)(6)(7). Guanylyl cyclase activating proteins (GCAPs) 2 are Ca 2ϩ /Mg 2ϩ -sensor proteins that impart Ca 2ϩ sensitivity to retinal guanylyl cyclase (RetGC), the enzyme that supplies the photoreceptor cell with cGMP (8 -11). GCAPs become RetGC activators at low Ca 2ϩ concentrations and inhibit it at high Ca 2ϩ , such that when the Ca 2ϩ concentration drops upon the illumination, GCAPs activate RetGC to quickly restore the level of cGMP in photoreceptors and thus accelerate their recovery from excitation. Conversely, when RetGC produces enough cGMP to reopen the Na ϩ /Ca 2ϩ channels, Ca 2ϩ re-enters the outer segments and the Ca 2ϩ -bound GCAPs decelerate cGMP synthesis.
GCAPs are recoverin-like neuronal calcium-binding proteins, also referred to as the neuronal calcium sensors (NCS) family, a part of a larger superfamily of EF-hand Ca 2ϩ -binding proteins (12)(13)(14)(15)(16). Like all other members of that superfamily GCAPs have Ca 2ϩ -binding EF-hand domains of the helix-loophelix structure. The metal-binding loop in the NCS proteins is traditionally defined as 12 sequential amino acid residues, of which 6 residues provide the oxygen atoms required for the metal coordination, including the invariant first and the last coordinating residues, Asp and Glu, respectively. The N-terminal EF-hand domains in GCAP-1 and GCAP-2 lack some of the oxygen-containing groups required for coordinating the metal ion (17), but are instead required for their interaction with RetGC (18 -20). We previously found that in the conditions that mimic light-adapted or dark-adapted photoreceptors, three other EF-hands in GCAP-1 are predominantly filled by either Mg 2ϩ or Ca 2ϩ , respectively (21,22).
There are multiple reports that substitutions of the first and last coordinating amino acids inactivate EF-hands in NCS proteins (17,(22)(23)(24)(25). Among these are the observations that inactivation of all three metal-binding EF-hands in GCAP-2 by substitution of the last Glu in EF-hands 2 and 3 with Gln and the first Asp in the EF-hand 4 with Asn make GCAP-2 a constitutive, insensitive to Ca 2ϩ activator of RetGC (24). A similar effect was observed for GCAP-1, whose EF-hands were disabled by substitution of the last Glu in the Ca 2ϩ -binding loop with Asp (25). Those observations lead to the perception that metal-free GCAPs are the activators of RetGC and that GCAPs undergo transition from Ca 2ϩ -bound to the metal-free form between dark and light. However, our recent study of metal-binding properties of GCAPs has also indicated that preventing Mg 2ϩ binding in GCAP-1 hampers stimulation of RetGC-1 (22). We concluded from the previous study (22) that, because cation binding was not only required for inhibition of RetGC, but was also critical for RetGC activation under physiological conditions, the general view on the functioning of GCAPs as metalbinding proteins in RetGC regulation needed to be revised (5)(6)(7)26). Therefore, the functions of the specific EF-hands in GCAPs must be revisited to provide more adequate understanding of their role in RetGC regulation via Ca 2ϩ /Mg 2ϩ exchange. The main purpose of this study was to determine the role of cation binding by EF-hands 2, 3, and 4 in GCAP-1 in creating its "activator" state, such as in the light-adapted photoreceptors.

EXPERIMENTAL PROCEDURES
Recombinant GCAP-1 and Its Mutants-All mutations were incorporated into bovine GCAP-1 cDNA by PCR using a "splicing by overlap extension" technique (27). Wild type GCAP-1 and its mutants used in this study also carried a D6S substitution that creates a recognition site for the yeast N-myristoyl transferase (28), and does not interfere with the RetGC regulation by the recombinant GCAP-1 (29,30). GCAP-1 cDNA was expressed under control of the isopropyl ␤-D-thiogalactopyranoside-regulated T7 promoter in a BL21(DE3) pLysS Escherichia coli strain (Novagen/Calbiochem) harboring a pBB131 plasmid for a yeast N-myristoyl transferase expression as described (29,30). Cells were grown in standard LB medium containing 40 g/ml kanamycin and 100 g/ml ampicillin. Free myristic acid was added from a concentrated ethanol solution to the suspension of bacterial cells to a final concentration of 100 g/ml 20 min prior to the induction with 0.5 mM isopropyl ␤-D-thiogalactopyranoside. Three hours after the induction, the bacterial pellet was harvested and the recombinant GCAPs were purified as described previously in detail (22). The concentration of GCAP-1 and its mutants was determined in 20 mM Na-phosphate buffer (pH 6.5) containing 6 M guanidine hydrochloride using extinction coefficients at 280 nm computed for the individual mutants from their amino acid composition (31) utilizing the ExPASY proteomic WWW server software from the Swiss Institute of Bioinformatics (32).
Tryptophan Fluorescence Measurements-Fluorescence emission at 332 nm (excitation at 290 nm) was recorded at 23°C using 4 M GCAP-1 in 0.6 ml of 100 mM MOPS/KOH (pH 7.2), 40 mM KCl, and 1 mM EGTA. Small aliquots of concentrated MgCl 2 solution were added to obtain the desired free Mg 2ϩ concentrations. The free Ca 2ϩ and Mg 2ϩ concentrations in solution were calculated according to the method of Brooks and Stoney (33), utilizing the algorithm of Marks and Maxfield (34). The data shown are representative from three to four independent experiments producing virtually identical results.

Mutations That Inhibit Cation Binding in EF-hands of GCAP-1 and May or May Not Result in Constitutive Activation
of RetGC-In our previous study we found that disabling all three metal-binding EF-hands in GCAP-1 by mutations that nonselectively hampered both Ca 2ϩ and Mg 2ϩ binding failed to produce a constitutive, Ca 2ϩ -insensitive activation of RetGC-1 (22). That was contrary to what one would expect if the apo form of the GCAP were the activator form and argued that the substitution of Ca 2ϩ by Mg 2ϩ , rather than merely loss of Ca 2ϩ , was required to make GCAP-1 undergo the inhibitor-to-activator transition in the light. If this hypothesis were true, then to make GCAP-1 a constitutive activator of RetGC it should be made unable to bind Ca 2ϩ , yet still retain the ability to bind Mg 2ϩ in its EF-hand(s). We found that the replacement of the last Glu in Ca 2ϩ -binding loops with Gln ( Fig. 1) prevented binding of Ca 2ϩ , but had little effect on Mg 2ϩ binding (22). Predictably, EF(2,3,4) Ϫ mutant generated by such a substitution in all three metal-binding EF-hands, GCAP-1(E75Q/E111Q/ E155Q), activated RetGC-1 at low Ca 2ϩ and continued to activate it even at high concentrations of Ca 2ϩ , which turn wild type GCAP-1 into a RetGC-1 inhibitor (Fig. 2). Contrary to that, a different EF(2,3,4) Ϫ mutant, GCAP-1(D64N/D100N/D102G/ D144N/D148G), which binds neither Ca 2ϩ nor Mg 2ϩ (22), failed to activate RetGC within the same free Ca 2ϩ range (Fig.  2). Evidently, for GCAP-1 to become a RetGC-1 activator under the conditions typical for light-adapted photoreceptors one or more of its EF-hands must be filled with Mg 2ϩ .
EF-hands 2 and 3 Require Mg 2ϩ Binding for Activation of RetGC under Light-adapted Conditions-To identify the EFhands that need to bind Mg 2ϩ to maintain the activator state of GCAP-1, we disabled Mg 2ϩ binding in the individual EF-hands by point mutations and tested their ability to activate RetGC-1 in the absence of Ca 2ϩ in comparison with the wild type Mutations introduced in EF-hands of GCAP-1: D64N, D100N/D102G, and D144N/D148G to disable both Ca 2ϩ and Mg 2ϩ binding to EF-hands 2, 3, and 4, respectively; E75Q, E111Q, and E155Q to disable only Ca 2ϩ binding to EF-hands 2, 3 and 4, respectively.

and Ref. 22).
A substitution in EF-hand 2, D64N, that prevents both Ca 2ϩ and Mg 2ϩ binding (22) was compared with the wild type and E75Q substitution, which selectively preserves Mg 2ϩ binding (22). Whereas the GCAP-1(E75Q) showed dose dependence (K 1/2 ) and the maximal activity of RetGC-1 stimulated by GCAP-1 (A max ) similar to the wild type, disabling of Mg 2ϩ binding dramatically reduced the apparent affinity of GCAP-1(D64N) for RetGC-1, but had little effect on the A max (Fig. 3A, Table 1).
Unlike EF-hands 2 or 3, disabling Mg 2ϩ binding in EF-hand 4 by a double substitution, D144N/D148G (22), had no effect on Table 1). These data strongly indicate that among the three metal-binding EF-hands of GCAP-1, binding of Mg 2ϩ in EF-hands 2 and 3 is important for RetGC-1 activation, whereas Mg 2ϩ in EF-hand 4 does not significantly contribute to creating the activator state of GCAP-1.
Indeed, simultaneous disabling of Mg 2ϩ binding in both EFhands 2 and 3 in GCAP-1(D64N/D100N/D102G) diminished its A max and K 1/2 even stronger than in the individually disabled EF-hands (Fig. 3D, Table 1). Unlike that, a different EF(2,3) Ϫ mutant, E75Q/E111Q, that retained Mg 2ϩ binding in EF-hands 2 and 3 (22) demonstrated A max and K 1/2 similar to the wild type ( Fig. 3D, Table 1). However, because the value for K 1/2 was difficult to determine more accurately due to the low activity of the mutant, we additionally verified that the GCAP-1(D64N/ D100N/D102G) lost its ability to interact with RetGC-1 in a competition experiment, where RetGC-1 was activated by 3 M wild type GCAP-1 at low free [Ca 2ϩ ] in the presence of increasing concentrations of the GCAP-1(D64N/D100N/D102G) (Fig.  3E). Even at 20-fold excess of the GCAP-1(D64N/D100N/ D102G) over the wild type we find no evidence for its interference with the activation of RetGC-1. We therefore conclude that once GCAP-1 is lacking Mg 2ϩ binding in both EF-hands 2 and 3, it becomes unable to properly interact with RetGC-1.
Mg 2ϩ Binding in EF-hand 2 Affects Conformation of EF-hand-like Domain 1-Although both EF-hands 2 and 3 are involved in activation of RetGC-1 at low free Ca 2ϩ , it appeared that Mg 2ϩ binding in EF-hand 2 was the most critical for GCAP-1 affinity for the cyclase. There is a significant conformational difference between the Mg 2ϩ -bound and metal-free GCAP-1 (21,22). In particular, binding of Mg 2ϩ by EF-hands 2 and 3 causes structural changes in EF1, an EF-hand-like domain that cannot bind metal ion but is required for interaction of GCAPs with RetGC (18 -20). These changes can be monitored by a change in fluorescence of Trp 21 located in the middle of the GCAP-1 EF1 domain (22). Because in addition to the Trp 21 GCAP-1 contains two more Trp residues, Trp 51 and Trp 94 , we replaced them with Phe and used the resultant GCAP-1(W51F/ W94F) mutant (22) as a template for creating two other mutants, GCAP-1(D64N/W51F/W94F) and GCAP-1(E75Q/ W51F/W94F), both with additional mutations affecting EF-hand 2. Each contained a single Trp 21 residue, but only the latter mutant was able to bind Mg 2ϩ in EF-hand 2. The GCAP-1(W51F/W94F) exhibited a prominent Mg 2ϩ -dependent decrease in Trp 21 fluorescence caused by conformational changes in the EF 1 domain (Fig. 4 and Ref. 22). Similar results were observed with the GCAP-1(E75Q/W51F/W94F), whose EF-hand 2 was modified by the E75Q mutation that preserved Mg 2ϩ binding (Fig. 4). Contrary to that, disabling Mg 2ϩ binding in EF-hand 2 by the D64N substitution in GCAP-1(D64N/ W51F/W94F) prevented decrease in the Trp 21 fluorescence. These results suggest that Mg 2ϩ binding in EF-hand 2 creates proper conformation of the neighboring EF-hand-like domain, known to be required for GCAPs interaction with RetGC-1 (18,19). The exact reason why Trp 21 fluorescence does not stay completely flat, but slightly increases in the D64N mutant is unknown, but one possible explanation can be that when EFhand 2 is in its apo (non-physiological) form, cation binding to EF-hands 3 and/or 4, through conformational change in the rest of the molecule, affects the environment for Trp 21 and thus produces the small increase in fluorescence. It is therefore only possible to see this effect when the EF-2 is inactivated and the rest of the Trp residues in the molecule are removed. In the presence of the functional EF-hand 2 the influence from the rest of the molecule on the environment of Trp 21 is strongly opposed by cation binding in the neighboring EF-hand 2 that results in a decrease of fluorescence.
Ca 2ϩ Binding and Inhibition of RetGC-In photoreceptor cells, GCAP-1 undergoes transition between the Mg 2ϩ -and Ca 2ϩbound forms as the concentration of intracellular Ca 2ϩ changes between light and dark (21,22), a process that turns GCAP-1 from RetGC activator into RetGC inhibitor. The replacement of Mg 2ϩ by Ca 2ϩ occurs in all three EF-hands (22), however, it is still unclear how individual EF-hands of GCAP-1 contribute to its transition to the RetGC inhibitor form. Another group previously reported that disabling of Ca 2ϩ binding in EF-hands 3 and 4 by substitution of the last Glu with Asp in the 12-amino acid Ca 2ϩ -binding loop converted GCAP-1 into a constitutive activator of RetGC (25). That mutant showed a similar to wild type GCAP-1 dose dependence of RetGC stimulation at low free Ca 2ϩ but did not inhibit cyclase at high free Ca 2ϩ (25). However, the mutations used in Ref. 25 do not disable binding of Mg 2ϩ (22, 36 -38). Therefore, to revisit this question, we used an EF (3,4) Ϫ mutant with EF-hands 3 and 4 disabled by different mutations, D100N/D102G and D144N/D148G, respectively. These mutations prevented both Mg 2ϩ and Ca 2ϩ binding to EF-hands 3 and 4, but preserved high-affinity Ca 2ϩ and Mg 2ϩ binding in EF-hand 2 (22). At low free Ca 2ϩ the D100N/D102G/D144N/D148G EF (3,4) Ϫ mutant was able to activate RetGC-1, but much less efficiently than wild type GCAP-1 (Fig. 5 and Table 2). Surprisingly, Ca 2ϩ sensitivity of RetGC-1 regulation by this mutant was reversed: increase in Ca 2ϩ concentrations further increased RetGC-1 stimulation (Fig. 5A), instead of inhibiting it. Such an increase could only result from Ca 2ϩ binding in EF-hand 2, because EF-hands 3 and 4 in this mutant do not bind Ca 2ϩ , even at the highest concentration used in the assay (22). Its apparent affinity for RetGC-1 was also reduced at low Ca 2ϩ concentrations but increased with the rise of free Ca 2ϩ in the assay ( Table 2).
The EF(2,3,4) Ϫ mutant, E75Q/D100N/D102G/D144N/ D148G, activated RetGC-1 in the absence of Ca 2ϩ similarly to  Table 1. the GCAP-1 EF(3,4) Ϫ mutant, D100N/D102G/D144N/D148G, because the E75Q mutation preserves Mg 2ϩ binding in EF-hand 2 (22). However, it did not regulate RetGC-1 in a Ca 2ϩ -sensitive manner (Fig. 5A), because none of its EF-hands could now bind Ca 2ϩ (22). The concentration of free Mg 2ϩ in the assay was 5 mM, which saturates binding to EF-hand 2 in wild type GCAP-1 and in the EF(2,3,4) Ϫ mutant, E75Q/D100N/ D102G/D144N/D148G (22). The transition of EF-hand 2 from the Mg 2ϩ -bound to the Ca 2ϩ -bound form substantially increased the apparent affinity of the GCAP-1(D100N/D102G/ D144N/D148G) for RetGC-1 ( Table 2). The GCAP-1(E75Q/ D100N/D102G/D144N/D148G) that could not exchange Mg 2ϩ in EF-hand 2 for Ca 2ϩ exhibited almost no change in its apparent affinity for RetGC-1 between the low and high free Ca 2ϩ (Fig. 5 and Table 2). Thus, in this artificial situation, when cation binding in EF-3 and EF-4 is completely disabled, the transition of EF-hand 2 from the Mg 2ϩ -bound to the Ca 2ϩbound form does not make GCAP-1 a RetGC-1 inhibitor and can even mimic RetGC-1 activator conformation. Consistent with the earlier observations (25), this finding argues that EF-hand 2 is not essential for the Ca 2ϩ -dependent inhibition of RetGC-1 by GCAP-1. Indeed, inactivation of EFhand 2 by a mutation, E75Q, did not make GCAP-1 a constitutive activator of RetGC-1 ( Fig. 6 and Ref. 25). However, unlike the total lack of effect in the Ref. 25, we also found that disabling of Ca 2ϩ binding in EF-hand 2 by the E75Q mutation in the conditions of saturation by Mg 2ϩ to some extent affected both the cooperativity (n decreased from 2.34 Ϯ 0.1 (n ϭ 5) in wild type to 1.7 Ϯ 0.06 (n ϭ 3) in E75Q) and Ca 2ϩ sensitivity of RetGC-1 regulation by GCAP-1 (Fig. 6). One possible explanation for the difference between the two observations is that the E75Q mutation in EF-hand 2 blocks Ca 2ϩ binding more efficiently than the E75D, as we showed previously (22).

Mutants
Contrary to the EF-hand 2, inactivation of both EF-hands 3 and 4 did make GCAP-1 a constitutive activator of RetGC-1, and so did the inactivation of individual EF-hands 3 or 4 ( Fig. 6 and Ref. 25). Yet, at this point we cannot determine the direct contribution of EF-hand 3 in RetGC-1 inhibition by GCAP-1. Although our present findings do not exclude that this indeed may be the case, we found that disabling of EF-hand 3 by various mutations always dramatically decreased the affinity of wild type EF-hand 4 for Ca 2ϩ (22). In other words, we were unable to find a mutation that would disable Ca 2ϩ binding in EF-hand 3 without also hampering Ca 2ϩ binding in EF-hand 4 at the same time. In addition to that, the results shown in Fig. 6 can argue that Ca 2ϩ binding in EF-hand 3 itself has a relatively small additional effect when compared with EF-hand 4. Hence, even in its Ca 2ϩbound form EF-hand 3 mostly preserves the activator form of the GCAP-1 rather than provides a conformational switch for RetGC inhibition. Therefore, it is more likely that the main role of EF-hand 3 in  The concentration of free Mg 2ϩ in the GC assay was 5 mM. The data were fitted as described in the legend to Fig. 3. The values of A max and K 1/2 are summarized in Table 2. respectively. b A max , the maximal level of RetGC-1 activation by GCAP at indicated free Ca 2ϩ in the presence of 5 mM free Mg 2ϩ (mean Ϯ S.D., n is the number of independent measurements). c K 1/2 , the concentration of GCAP required for half-maximal activation of RetGC-1 (mean Ϯ S.D., n is the number of independent measurements). JULY 27, 2007 • VOLUME 282 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 21649
Ca 2ϩ -dependent inhibition of RetGC can be indirect, through regulation of Ca 2ϩ binding to the neighboring EF-hand 4.

DISCUSSION
The Role of EF-hands in RetGC Regulation by GCAP-1-GCAP-1 contains four EF-hand structures of which three can bind either Ca 2ϩ or Mg 2ϩ under physiologically relevant conditions (Fig. 7A). We find in this study that both EF-hands 2 and 3 in GCAP-1 must be occupied by Mg 2ϩ to maintain GCAP-1 in its RetGC activator state under the conditions that exist in photoreceptors in the light (Fig. 3, Table 1). The apo EF-hands 2 and 3 do not support the GCAP-1 conformation required for the interaction with the cyclase (Fig. 3). We consider one of the most important findings of this study that the EF-hand 2, whose role in RetGC regulation was previously deemed unclear (25), is in fact a crucial element in RetGC regulation by GCAP-1, and the cation binding in EF-hand 2 is required for high-affinity interaction with RetGC (Figs. 3 and 5 and Tables 1 and 2). Another EF-hand that cannot effectively maintain the activator conformation of GCAP-1 in its apo form is EF-hand 3. It also needs to be in a cation-bound form to enhance GCAP-1/ RetGC-1 affinity, however, unlike EF-hand 2, the Mg 2ϩ binding in EF-hand 3 is also important to keep the maximal level of RetGC stimulation (Fig. 3, Table 1).
Two sites for binding of GCAP to RetGC may be required to explain why prevention of Mg 2ϩ binding in two different EFhands can produce the differential decrease in A max versus the increase of the K 1/2 shown in Fig. 3 and Tables 1 and 2. The properties of the mutants suggest that at least two types of interaction between RetGC-1 and GCAP-1 occur: a high-affinity "docking" of GCAP, supported by cation binding in the EFhand 2 (and partially EF-hand 3), and binding at a secondary site in GCAP-1, regulated by the EF-hand 3. Hence, the lack of Mg 2ϩ binding in EF-hand 2 drastically reduces the affinity for RetGC, whereas the lack of Mg 2ϩ in EF-hand 3 produces a smaller effect on the affinity, but substantially reduces the A max of the cyclase.
Unlike other metal-binding EF-hands in GCAP-1, the EFhand 4 does not contribute to the activation of the cyclase and is only required for switching RetGC back to the inhibited state upon increase in free Ca 2ϩ in the dark. It is also important to note that both EF-hands 2 and 3 retain the activator conformation of GCAP-1 when the inhibitory effect of Ca 2ϩ binding EF-hand 4 is blocked (Fig. 6). Although such a situation cannot be found in normal physiological conditions, where only Mg 2ϩ conformation exists in the light, and only Ca 2ϩ conformation inhibiting RetGC via EF-hand 4 is present in the dark, under certain pathological conditions the ability of EF-hands 2 and 3 to efficiently maintain the active conformation of GCAP-1 even in the Ca 2ϩ bound form contributes to the abnormal activity of the cyclase in the dark found in association with congenital blindness (39 -41). Apparently, blocking Mg 2ϩ binding in EFhand 3 can also create an artificial situation when the replacement of Mg 2ϩ in EF-hand 2 by the Ca 2ϩ mutant can partially compensate for the abnormal conformation of an apo form of EF-hand 3 (Fig. 5).
The role of cation binding in EF-hand 3 can be 2-fold: (i) to maintain the activator conformation of the GCAP-1 when filled with Mg 2ϩ ; and (ii) to contribute to the inactivation of the cyclase when bound with Ca 2ϩ . The former role is supported by the results presented in this study, and the latter was also proposed previously by another group (25). Whereas both roles are possible, we find that it is difficult to verify the direct contribution of the EF-hand 3 to inactivation of the cyclase, because a direct measurement of Ca 2ϩ binding in EF-hand 4 showed that where A is the activity of RetGC-1, A max and A min are the maximal and minimal activity of RetGC in the assay, respectively, [Ca] 1/2 is the free Ca 2ϩ concentration required for half-maximal inhibition of RetGC-1 by GCAP, n is the Hill coefficient. For other conditions of the assay see "Experimental Procedures." FIGURE 7. A, in light-adapted conditions, all three metal-binding EF-hands in GCAP-1 are occupied by Mg 2ϩ , however, only Mg 2ϩ binding in EF-hands 2 and 3 is required for RetGC stimulation. The apo forms of these EF-hands do not create the proper conformation for GCAP-1. In the dark, all three EF-hands of GCAP-1 are predominantly occupied by Ca 2ϩ , however, the main requirement for converting GCAP-1 into the "RetGC inhibitor" is binding of Ca 2ϩ in EF-hand 4. Ca 2ϩ binding in EF-hands 2 and 3 is primarily required for GCAP-1 to preserve binding to RetGC and to facilitate the high-affinity binding of Ca 2ϩ to EF-hand 4. B, Ca 2ϩ /Mg 2ϩ exchange in EF-hands of GCAP-1 between light and dark provides functional switch between its "RetGC activator" and RetGC inhibitor states. The apo form of GCAP-1 has no function. Other explanations are under "Discussion." it was always suppressed when Ca 2ϩ binding in the EF-hand 3 was affected (but not vice versa). Therefore, it is likely that EFhand 3 contributes to switching off the cyclase primarily through its indirect effect on the affinity of the EF-hand 4 for Ca 2ϩ , rather than in directly providing a conformational switch for RetGC inhibition. This would also suggest that Ca 2ϩ binding in the EF-3/EF-4 globular domain occurs sequentially, first in EF-hand 3 and only after that in EF-hand 4.
The EF-hand domain 1, which cannot itself coordinate metal, has been shown to be crucial for GCAP-1 and GCAP-2 interaction with RetGC (18 -20, 30, 42) and in some other NCS proteins for interaction with their targets (43)(44)(45). We found a striking effect of EF-hand 2 occupation by Mg 2ϩ on the conformation of the EF-hand 1 domain, revealed by its fluorescence spectra (Fig. 4). It is therefore tempting to speculate that EFhand domain 1 is not only directly involved in high-affinity binding to RetGC, but also that high-affinity binding is directly controlled via the conformation of its neighboring metal-binding EF-hand 2.
To summarize, based on this and the number of previous studies (19,20,30,42), we can propose the following functions to the EF-hands in GCAP-1 (Fig. 7A). (i) EF-hand-like domain 1: no metal binding, and contributes to the high-affinity interaction with RetGC. (ii) EF-hand 2: binding of Mg 2ϩ in the light maintains the activator conformation of GCAP-1, controls the high-affinity binding of GCAP-1 to RetGC-1, presumably through involving EF-hand-like domain 1; is not essential for the maximal level of stimulating activity of GCAP-1; and replacement of Mg 2ϩ by Ca 2ϩ in this EF-hand has only a small effect on inhibition of RetGC. (iii) EF-hand 3: binding of Mg 2ϩ in the light maintains the optimal conformation of GCAP-1 for RetGC activation and, although to a lesser extent, contributes to the high-affinity GCAP/RetGC interaction; binding of Ca 2ϩ fails to create the inhibitory conformation of GCAP-1, but most likely contributes to inactivation of the cyclase indirectly, by facilitating Ca 2ϩ -binding in the neighboring EF-hand 4. (iv) EFhand 4: binding of Mg 2ϩ in this EF-hand contributes to neither binding nor activation of RetGC in the light; binding of Ca 2ϩ in this EF-hand provides a potent functional switch to turn the cyclase off.
Mg 2ϩ /Ca 2ϩ Cycle in GCAP-1 Controls RetGC-1 Regulation-Free Ca 2ϩ concentrations in rods and cones change nearly 10-fold in response to illumination, in mammals between 250 nM in the dark and 25 nM in the light (5)(6)(7). This provides a potent feedback for RetGC that accelerates the recovery of rods and cones (reviewed in Refs. 1, 4, and 46). Contrary to the previous view on GCAPs role in RetGC regulation via release and binding of Ca 2ϩ to the apo form of GCAPs, we now argue that the actual process of switching the GCAP-1 between the activator and the "inhibitor" state is instead based on Mg 2ϩ /Ca 2ϩ exchange, summarized in the schematics presented in Fig. 7B. The apo form of GCAP-1 is neither activator nor inhibitor of RetGC-1 and has no physiological role. Much like GDP/GTP exchange is required to regulate the conformation of a G-protein, the cation exchange in GCAP-1 is required to regulate RetGC-1. The fundamental differences from the GDP/GTP exchange in G-proteins is nevertheless, apparent: no external receptor is required for GCAP-1 to release or bind the divalent cations, and this can be accomplished solely as a result of the free Ca 2ϩ concentration change relative to the free Mg 2ϩ concentrations in the dark versus light.
The critical difference between the activator and the inhibitor conformation of GCAP-1 must be very subtle compared, for example, with that of recoverin (49 -51). This is not a difference between the apo protein and the metal bound form, but between the two metal-bound forms, Mg 2ϩ versus Ca 2ϩ . Moreover, EF-hands 1, 2, and 3 are even less likely to undergo a major change between Mg 2ϩ and Ca 2ϩ forms, because EF-2 and EF-3 are both capable of maintaining the high-affinity interaction with the cyclase in either Mg 2ϩ or Ca 2ϩ form. To date, only a partial Ca 2ϩ -bound structure of two other GCAPs was established (47,48), but to properly understand the mechanism and the structural basis for the RetGC regulation, one would need to compare, potentially rather subtle, differences between Mg 2ϩand Ca 2ϩ -bound GCAP-1.