Ca2+ and Mg2+ Binding Properties of GCAP-1

Guanylyl cyclase-activating protein 1 (GCAP-1) is an EF-hand protein that activates retinal guanylyl cyclase (RetGC) in photoreceptors at low free Ca2+ in the light and inhibits it in the dark when Ca2+ concentrations rise. We present the first direct evidence that Mg2+-bound form of GCAP-1, not its cation-free form, is the true activator of RetGC-1 under physiological conditions. Of four EF-hand structures in GCAP-1, three bound Ca2+ ions and could exchange Ca2+ for Mg2+. At concentrations of free Ca2+ and Mg2+ typical for the light-adapted photoreceptors, all three metal-binding EF-hands were predominantly occupied by Mg2, and the presence of bound Mg2+ in GCAP-1 was essential for its ability to stimulate RetGC-1. In the Mg2+-bound form of GCAP-1 all three Trp residues became more exposed to the polar environment compared with its apo form. The replacement of Mg2+ by Ca2+ in the EF-hands 2 and 3 further exposed Trp-21 to the solution in a non-metal-binding EF-hand domain 1 that interacts with RetGC. Contrary to that, replacement of Mg2+ by Ca2+ in the EF-hand 4 moved Trp-94 in the entering α-helix of the EF-hand 3 back to the non-polar environment. Our results demonstrate that Mg2+ regulates GCAP-1 not only by adjusting its Ca2+ sensitivity to the physiological conditions in photoreceptors but also by creating the conformation required for RetGC stimulation.

In this study, we continue to develop a concept for the processes through which Ca 2ϩ sensor proteins, GCAPs, 2 regulate their target enzyme, retinal guanylyl cyclase (RetGC), in photosensitive neurons of the retina between light and dark. In darkadapted photoreceptors, cGMP keeps a fraction of cGMPgated channels open, thus allowing Na ϩ and Ca 2ϩ to enter the outer segment (1,2). Because Ca 2ϩ is continuously removed from the outer segments by a light-independent Na ϩ /K ϩ ,Ca 2ϩ exchanger, when the light closes the channels by stimulating cGMP hydrolysis, the intracellular concentration of Ca 2ϩ decreases from near 250 nM in the dark to 25 nM in the light (3)(4)(5). The change in free Ca 2ϩ dramatically affects synthesis of cGMP by retinal guanylyl cyclase (6) because GCAPs become RetGC activators at low Ca 2ϩ and RetGC inhibitors at high Ca 2ϩ (7)(8)(9)(10)(11)(12). Despite the popular view that the cyclase activity is regulated by merely binding and release of Ca 2ϩ by the apo form of GCAPs, our recent findings indicate that mutual displacement of two competing cations, Ca 2ϩ and Mg 2ϩ , defines the dynamic range of RetGC regulation in vivo (13).
GCAPs are recoverin-like neuronal calcium sensor proteins within the EF-hand superfamily (14,15). Like other recoverinlike proteins, GCAPs have four EF-hand domains, but based on their amino acid sequence only three of them are capable of binding metal ion. Although the general model of regulation requires that Ca 2ϩ be either bound or released by GCAPs, little is known about the actual Ca 2ϩ binding properties of individual GCAPs under physiologically relevant conditions or the actual conformational change in GCAPs that constitutes switching them between the activator and the inhibitor forms. A number of previous studies aimed to identify conformational changes in GCAP-1 between its activator and inhibitor forms using limited proteolysis (16 -18), chemical modification of endogenous cysteine residues (17), EPR spectroscopy (17), and tryptophan fluorescence spectroscopy (18 -21) were based on the assumption that GCAPs are only Ca 2ϩ -binding proteins and that the metalfree GCAPs are the activators of RetGC-1. Our recent finding demonstrated that Mg 2ϩ is essential for adjusting the Ca 2ϩ sensitivity of RetGC regulation by GCAPs to the actual physiological range of free Ca 2ϩ concentrations and provided the first evidence that GCAPs are both Ca 2ϩ and Mg 2ϩ sensor proteins (5,13,22) thus raising the possibility that at physiological free Ca 2ϩ and Mg 2ϩ concentrations in the light at least one of the EF-hands in GCAP-1 is in a Mg 2ϩ -bound, rather than metalfree, state. Importantly binding of Mg 2ϩ changes the intensity of the intrinsic tryptophan fluorescence of GCAP-1, and that implies that a significant structural difference exists between the Mg 2ϩ -bound and the metal-free GCAP-1 (13).
In the present study, we determined both Ca 2ϩ and Mg 2ϩ binding properties of GCAP-1 using two independent methods and a wide range of mutations introduced in its EF-hand domains. We found the following. (i) At the physiological concentrations of free Mg 2ϩ and Ca 2ϩ that exist in the vertebrate photoreceptors in the light (3,23,24), three EF-hands in GCAP-1 are predominantly occupied with Mg 2ϩ ions; hence the physiological RetGC-activating form of GCAP-1 is triple Mg 2ϩ -bound form. (ii) The same three EF-hands in purified GCAP-1 also bind Ca 2ϩ in a non-cooperative manner. (iii) Binding of Mg 2ϩ in three metal-binding EF-hands causes significant structural changes in EF1 domain, an EF-hand-like structure that cannot bind metal ions but is required for interaction of GCAPs with RetGC. (iv) Inactivation of both Ca 2ϩ and Mg 2ϩ binding in three EF-hands renders GCAP-1 incapable of activating RetGC.
Contrary to the common perception of GCAPs being activators of RetGC in their apo form, we also present the first direct evidence that the Mg 2ϩ -bound GCAP-1 is required for activation of the target enzyme. We conclude in this report that the regulation of the target enzyme by GCAP-1 is a result of Mg 2ϩ / Ca 2ϩ exchange in the EF-hands rather then simply binding and release of Ca 2ϩ ions by the apo form of GCAP.

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 (25). 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-myristoyltransferase (26) and does not interfere with the RetGC regulation by the recombinant GCAP-1 (27,28). GCAP-1 cDNA was inserted into the NcoI/BamHI sites of the pET11d vector (Novagen/Calbiochem) and 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 N-myristoyltransferase expression as described previously (27). Cells were typically grown in 3.0 liters of 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 30 min prior to the induction with 0.5 mM isopropyl ␤-D-thiogalactopyranoside. Three hours after the induction, the bacterial pellet was harvested by centrifugation at 8,000 ϫ g for 20 min at 4°C, and the cells were disrupted by ultrasonication. The expressed GCAP-1 and its mutants were always found in the insoluble fraction of the inclusion bodies. The insoluble material was washed twice with 50 ml of 10 mM Tris-HCl (pH 7.5) containing 2 mM EDTA and 14 mM 2-mercaptoethanol (buffer A) by centrifugation at 20,000 ϫ g for 20 min. GCAP-1 was extracted from the pellet by homogenization in buffer A containing 8 M freshly deionized urea for 30 min at 4°C and dialyzed twice against 2.0 liters of 10 mM Tris-HCl buffer (pH 7.5) containing 0.1 mM EDTA and 14 mM 2-mercaptoethanol at 4°C, and the insoluble fraction was removed by centrifugation at 20,000 ϫ g for 20 min. The concentration of Tris-HCl buffer (pH 7.5) in the supernatant was increased to 40 mM, CaCl 2 was added to a final concentration of 5 mM, the solution was then incubated on ice for 20 min, and the precipitate was removed by centrifugation at 20,000 ϫ g for 20 min. The supernatant was concentrated to 5 ml under nitrogen pressure using an Amicon YM10 membrane, centrifuged at 200,000 ϫ g for 10 min, and loaded on a Sephacryl S-200 column (2.6 ϫ 60 cm) in 20 mM Tris-HCl (pH 7.5) containing 100 mM NaCl. Fractions containing GCAP-1 were pooled and, after adding NaCl to 1 M and dithiothreitol to 5 mM, applied on a 1.6 ϫ 5.0-cm butyl-Sepharose 4 Fast Flow column (GE Healthcare) equilibrated with 20 mM Tris-HCl (pH 7.5) containing 1.0 M NaCl. The column was washed with the same buffer, and GCAP-1 was eluted with 5 mM Tris-HCl (pH 7.5). The main peak of GCAP-1 was collected, and EDTA was added to 2 mM to remove Ca 2ϩ bound to GCAP-1. The purity of GCAP-1 preparations was at least 95% as estimated by SDS gel electrophoresis. We also observed that using Chelex was not sufficient to remove all bound Ca 2ϩ from GCAP-1 (18); however, preincubation with EDTA produced virtually metal-free GCAP-1. The excess of EDTA was removed by three to four cycles of 20-fold concentration/dilution using an Amicon Ultra-15 (10,000 molecular weight cutoff) centrifugal filter in 10 mM Tris-HCl (pH 7.5) containing 20 M EDTA. We find that even in solutions made of reagents containing Ͻ5 ppb Ca 2ϩ , such as Sigma Ultra grade and Barnstead Nanopure 18.3 megaohms⅐cm water, Ca 2ϩ contaminations can be as high as 1 M. Because GCAP-1 has high affinity for Ca 2ϩ , even these low levels of contamination are sufficient to skew the measurement of the Ca 2ϩ binding (18). Therefore, the 20 M EDTA was used to prevent GCAP-1 from rebinding Ca 2ϩ from the solutions during the concentration/dilution cycles. The final concentration of GCAP-1 in stock solutions was commonly between 300 and 350 M. Concentrated protein was quickly frozen in small aliquots and stored at Ϫ70°C. The concentration of GCAP-1 and its mutants was determined in 20 mM sodium 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 (29) utilizing the ExPASY proteomic WWW server software from the Swiss Institute of Bioinformatics (30). Myristoylation was verified by electrospray mass spectrometry (supplemental Fig. 1) as described in Peshenko et al. (22).
Ca 2ϩ Binding Measurements-To determine the stoichiometry and the affinity of GCAP-1 mutant for Ca 2ϩ , each protein sample containing a known amount of GCAP-1 or its mutant was incubated at a known total concentration of CaCl 2 , and the free Ca 2ϩ concentration ([Ca 2ϩ ] f ) in the sample was then determined using Ca 2ϩ fluorescent indicator dyes (BAPTA-2, Fluo-3, BAPTA-6F, and Fluo-4FF, all from Invitrogen/Molecular Probes) to calculate the concentration of Ca 2ϩ bound to the protein. The Ca 2ϩ indicators used in each particular experiment were selected to optimally cover the range of the free Ca 2ϩ in the reaction mixture. Fluo-3 or BAPTA-2 was used when [Ca 2ϩ ] f was between 0.02 and 2.0 M; Fluo-4FF or BAPTA-6F was used for [Ca 2ϩ ] f above 1 M. The readings of the fluorescence intensity were corrected for dilution caused by addition of CaCl 2 . Free Ca 2ϩ in the reaction mixture was calculated using the following formula:  Table 1). We used a set of Ca 2ϩ /EGTA buffers (31) to determine the K d for Fluo-3 and BAPTA-2, a direct titration of Ca 2ϩ indicator with CaCl 2 in the case of Fluo-4FF, and a combination of both methods for BAPTA-6F.
Each GCAP-1 mutant was diluted from 300 -350 M stock solution to 20 -40 M final concentration in 0.6 ml of 100 mM MOPS/KOH (pH 7.2), 40 mM KCl, 1 mM dithiothreitol, 0 or 1 mM MgCl 2 , and 0.5 M appropriate fluorescent Ca 2ϩ indicator dye (Fluo-3, BAPTA-2, Fluo-4FF or BAPTA-6F). The mixture was assembled in a plastic cuvette and titrated at 23°C with addition of 3-l aliquots of CaCl 2 solution with known concentration. The fluorescence data were fitted by the equation: ] bound is the concentration of Ca 2ϩ bound to GCAP-1 calculated as [Ca 2ϩ ] bound ϭ [Ca 2ϩ ] total Ϫ [Ca 2ϩ ] free , N is the number of Ca 2ϩ ions bound per molecule of GCAP-1 at saturation, K d is the apparent affinity of GCAP-1 for Ca 2ϩ , and n H is the Hill coefficient. The trace amount of EDTA introduced into the assay from the stock solutions of the proteins was negligible compared with the protein concentration in the assay. The data shown are representative from three to five independent experiments producing virtually identical results.
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, 1 mM EGTA, and specified concentrations of MgCl 2 . Small aliquots of concentrated CaCl 2 solution were added to obtain the desired free Ca 2ϩ concentrations. The free Ca 2ϩ and Mg 2ϩ concentrations in the solution were calculated according to the method of Brooks and Storey (32) utilizing the algorithm of Marks and Maxfield (33). The data shown are representative from three to four independent experiments producing virtually identical results.
It was recently suggested (18) that one of the three metalbinding EF-hands can retain Ca 2ϩ irreversibly (18). Although we isolated GCAP-1 in the presence of high Ca 2ϩ concentrations (see "Experimental Procedures"), we found that all bound Ca 2ϩ could be efficiently removed by dialysis in the presence of EDTA, and we always observed binding of three Ca 2ϩ per GCAP-1 both in the absence (N ϭ 2.86 Ϯ 0.08, n ϭ 3) and in the presence (N ϭ 2.92 Ϯ 0.08, n ϭ 5) of magnesium in the direct Ca 2ϩ binding assay using independent preparations of GCAP-1 (Fig. 1C). The affinity of myristoylated GCAP-1 for Ca 2ϩ in our experiments strongly depended on Mg 2ϩ and was 0.1 Ϯ 0.025 M (mean Ϯ S.D., n ϭ 3) and 0.29 Ϯ 0.045 M (n ϭ 5) in the absence or in the presence of 1 mM free Mg 2ϩ , respectively. Much to our surprise, in these and the subsequent experiments, we found little cooperativity for the Ca 2ϩ binding to GCAP-1: Hill coefficient was 1.04 Ϯ 0.04 and 1.06 Ϯ 0.07 in the presence or in the absence of Mg 2ϩ , respectively (compare with the theoretical curve for the cooperative n H ϭ 2, binding in Fig. 1C, dashed line).
To identify Ca 2ϩ /Mg 2ϩ -specific EF-hand(s) in GCAP-1, individual EF-hands were disabled by point mutations designed to prevent metal binding within the 12-amino acid Ca 2ϩ -binding loop of different EF-hands (Fig. 1D). It has been established that the first Asp, the oxygen-containing side chains in the third, fifth, and ninth positions, and the last Glu of the loop are essential for coordinating Ca 2ϩ (34,35). Their substitution with corresponding amides or Gly inactivates EF-hand and hampers Ca 2ϩ binding. Substitution of the last Glu with Asp reduces the affinity of EF-hand for Ca 2ϩ in favor of Mg 2ϩ binding as it transforms the coordination sphere from 7-fold Ca 2ϩ coordination to 6-fold Mg 2ϩ coordination through switching from bidentate ligation to monodentate (36 -38). Such mutations were introduced into each metal-binding EF-hand of GCAP-1, either separately or in combination, to disable one, two, or all three metal-binding EF-hands.
Substitutions in EF-hand 2, EF(2) Ϫ -Unlike wild-type GCAP-1 that displays a single class of high affinity binding sites to which three Ca 2ϩ can bind per GCAP-1, consistent with the three active EF-hands (Fig. 1C), in the E75Q mutant the number of high affinity Ca 2ϩ -binding sites is reduced by one ( Fig. 2A). In addition to the high affinity class sites similar to the wild type (K d ϭ 0.24 Ϯ 0.04 M and 0.52 Ϯ 0.05 M in the absence or in the presence of 1 mM [Mg 2ϩ ] f , respectively) corresponding to the remaining active EF-hands 3 and 4, a low affinity binding site for the third Ca 2ϩ ion could be observed, corresponding to inactivated EF-hand 2. Similar results were observed in the case of mutations D64N and D68G/E75Q (not shown). Importantly in all EF2 mutants the K d for the high affinity binding in EF3 and EF4 was reduced in the presence of Mg 2ϩ .
Substitutions in EF-hand 3, EF (3)  To determine which of the two non-mutated EF-hands corresponds to the only remaining high affinity site, we complemented inactivation of the EF3 by inactivation of either EF2 or EF4, i.e. EF(2,3) Ϫ and EF (3,4) Ϫ GCAP-1 (Fig. 2C), respectively. In the EF(3,4) Ϫ mutant, the remaining EF-hand 2 bound Ca 2ϩ similarly to the high affinity site found in the EF(3) Ϫ GCAP-1 (K d ϭ 0.033 Ϯ 0.007 M and K d ϭ 0.13 Ϯ 0.03 M in the absence or in the presence of 1 mM [Mg 2ϩ ] f , respectively), but the high affinity site of the EF(3) Ϫ was eliminated in the EF(2,3) Ϫ mutants (D68G/E75Q/D100N/D102G, Fig. 2C, and E75Q/E111Q, D64N/D100N and E75D/E111D, data not shown). Thus, the EF-hand that retained the high affinity site for Ca 2ϩ in the EF(3) Ϫ mutants is EF2, whereas the EF4 turned into a low affinity binding site as a result of the EF3 inactivation.
The effect of EF3 inactivation on the EF4 did not allow estimation of the affinity for Ca 2ϩ of individual EF4 in the wild-type GCAP-1. Nevertheless EF2 and EF4 both bind Ca 2ϩ and Mg 2ϩ because the dependence of Ca 2ϩ binding in the individual EFhands 2 and 4 on Mg 2ϩ remained evident (Fig. 2, B and C).
Substitutions in EF-hand 4, EF(4) Ϫ -The EF4 domain was completely disabled by two point mutations, D144N and D148G. The resultant EF(4) Ϫ mutant bound two Ca 2ϩ ions (in EF-hands 2 and 3). Consistent with the ability of both EF2 and EF3 to bind Mg 2ϩ , the affinity for Ca 2ϩ of this mutant was lowered in the presence of 1 mM Mg 2ϩ from K d of 0.2 Ϯ 0.03 to 0.4 Ϯ 0.08 M (Fig. 2D).
We found that high affinity binding of a single Ca 2ϩ ion in EF3 is preserved when both EF-hands 2 and 4 were inactivated by substituting the first Asp of the loop with Asn (D64N/ D144N, Fig. 2E). The EF3 in this mutant bound Ca 2ϩ with an apparent K d of 0.4 Ϯ 0.07 and 1.0 Ϯ 0.2 M in the absence or in the presence of 1 mM free Mg 2ϩ , respectively (Fig. 2E). The important conclusion from these series of experiments was that all three active EF-hands in GCAP-1 can bind both Ca 2ϩ and Mg 2ϩ .  of the Ca 2ϩ -free GCAP-1 so that only phase II was observed in the presence of Mg 2ϩ (Fig. 3A). Importantly the completion of the conformational change occurred only when all three Ca 2ϩ were bound to GCAP-1 both in the presence and in the absence of Mg 2ϩ . This argues against one Ca 2ϩ not being able to dissociate from GCAP-1.
To attribute the Ca 2ϩ -dependent spectral changes in GCAP-1 to metal binding in individual EF-hands, we first tested the GCAP-1 mutants in which only one EF-hand was disabled: EF(2) Ϫ , EF(3) Ϫ , and EF(4) Ϫ (Fig. 4, A, B, and D). In the absence of Mg 2ϩ , EF(2) Ϫ mutant exhibited phase I, which was saturated by addition of 10 mM Mg 2ϩ , and the conformational change corresponding to phase II, which was clearly revealed in the presence of Mg 2ϩ (Fig. 4A). Notably Ca 2ϩ -induced changes in Trp fluorescence of the EF(2) Ϫ mutants remained biphasic in the absence of Mg 2ϩ , thus indicating that one of the remaining EF-hands in the EF(2) Ϫ GCAP-1 is responsible for phase I, and the other is responsible for phase II. Similar results were obtained for another EF(2) Ϫ mutant, E75Q (not shown).
In the EF(3) Ϫ mutant, either D100N (Fig. 4B) or E111Q (not shown), change of Trp fluorescence as a function of Ca 2ϩ also remained clearly biphasic (Fig. 4B). However, higher concentrations of Ca 2ϩ were required to produce phase II in the case of the EF(3) Ϫ GCAP-1. This is in agreement with the results from the direct Ca 2ϩ binding experiments showing that when the EF3 was inactivated, the affinity of EF4 for Ca 2ϩ was greatly reduced compared with the wild type (Fig. 2B). Taken together, these two observations strongly argue that the initial decrease in fluorescence (phase I) of the EF(3) Ϫ GCAP-1 is a result of Ca 2ϩ binding to the EF-hand 2, and the subsequent increase in fluorescence (phase II) is a result of Ca 2ϩ binding to the low affinity EF-hand 4 in this mutant. Indeed disabling both EF-  hands 2 and 3 effectively prevented the sharp initial drop in fluorescence corresponding to phase I (Fig. 4C), and therefore no subsequent phase 2 could be observed for this mutant either.
On the other hand the EF(4) Ϫ demonstrated a highly Ca 2ϩsensitive phase I but no phase II (Fig. 4D). When both EF-hands 3 and 4 were disabled, the single remaining Ca 2ϩ -binding site, EF2, produced only a change that corresponds to the high affinity Ca 2ϩ binding, phase I (Fig. 4E). Taken together, these data demonstrate that Ca 2ϩ binding in both EF-hands 2 and 3 causes the conformational change corresponding to phase I, whereas Ca 2ϩ binding in EF4 can produce only phase II.
The properties of the EF(2,4) Ϫ GCAP-1 further support this model. The EF(2,4) Ϫ mutant exhibited only phase I, corresponding to the Ca 2ϩ binding to the intact EF-hand 3 (Fig. 4F).
Mg 2ϩ -dependent Changes in Trp Fluorescence of EF-hand Mutants-Unlike Ca 2ϩ , Mg 2ϩ binding results only in a decrease of Trp fluorescence in both wild-type GCAP-1 and EF-hand mutants. However, one would expect that some of the mutations used to disable Ca 2ϩ binding in EF-hands would be less effective when it comes to disabling the Mg 2ϩ binding. It has been reported that substitution of the last Glu in the metalbinding loop of other EF-hand proteins with Asp hampers Ca 2ϩ binding but at the same time increases the affinity of the EFhand for Mg 2ϩ (36 -38). Therefore, it is important to identify the mutations in the EF-hands that would also reliably inactivate Mg 2ϩ binding. As a template for the mutagenesis, we used the D100N/D102G/D144N/D148G mutant, EF(3,4) Ϫ , because (a) that mutant produced a well defined phase I upon binding either Ca 2ϩ or Mg 2ϩ (Fig. 4E) and (b) even with both EF-hands 3 and 4 disabled, it still demonstrated high affinity metal binding to the remaining EF2 (Fig. 2C). We expected that two simultaneous mutations, Asp to Asn and Asp to Gly, in two different coordinating positions in each metal-binding loop for EF-hands 3 and 4 would disable metal binding to both these EF-hands. In such a case the Mg 2ϩ -dependent decrease in Trp fluorescence of the D100N/D102G/ D144N/D148G mutant should be attributed solely to the Mg 2ϩ binding in the intact EF2.
We then generated four different EF(2,3,4) Ϫ mutants by introducing the following additional point mutations in the EF2 of the EF(3,4) Ϫ GCAP-1: D64N, E75Q, E75D, or D68G/E75Q. As expected, in all four mutants the Ca 2ϩ sensitivity of phase I was reduced by at least 2 orders of magnitude compared with EF(3,4) Ϫ GCAP-1 (Fig. 5A), but these mutants had different sensitivity to Mg 2ϩ (Fig. 5B). The addition of the D64N mutation to the EF(3,4) Ϫ GCAP-1 completely prevented Mg 2ϩ -dependent changes in fluorescence. The D68G/E75Q double mutation also dramatically reduced the apparent affinity of the EF2 for Mg 2ϩ (Fig. 5B). Thus the remaining activity of the EF2 in Mg 2ϩ binding can be suppressed by either a double mutation, D68G/E75Q, or even a single point mutation, D64N. Contrary to that, E75D substitution not only failed to disable EF2 but, if anything, slightly increased its affinity for Mg 2ϩ ([Mg 2ϩ ]1 ⁄ 2 ϭ 0.25 Ϯ 0.02 mM, n ϭ 3; Fig. 5B). The E75Q substitution also preserved the Mg 2ϩ binding with the apparent affinity slightly decreased compared with EF(3,4) Ϫ ([Mg 2ϩ ]1 ⁄ 2 of 1.1 Ϯ 0.08 mM, n ϭ 3, and 0.42 Ϯ 0.02 mM, n ϭ 4, respectively). Thus, the substitution of the first Asp in the metal-binding loop of EF-hand with its amide can effectively disable both Ca 2ϩ and Mg 2ϩ binding, whereas substitution of the last Glu with Asp or Gln only decreases the affinity of the EF-hand for Ca 2ϩ but not for Mg 2ϩ .
Finding the mutations that disable Mg 2ϩ binding enabled us to selectively inactivate individual EF-hands in GCAP-1 and determine, by using the Trp fluorescence, the Mg 2ϩ binding properties of the remaining unaltered EF-hands. To determine the affinity of the EF-hand 3 for Mg 2ϩ , we used mutations (D64N/D144N and D68G/E75Q/D144N/D148G) that completely disabled both Ca 2ϩ (Fig. 2) and Mg 2ϩ (Fig. 5) binding in EF-hands 2 and 4. We found that the affinity of individual EFhand 3 for Mg 2ϩ was sensitive to the mutations in the neighboring EF-hands 2 and 4 (Fig. 6A). The D64N/D144N mutations in those EF-hands preserved Mg 2ϩ binding in the EF3 with affinity comparable to the wild type ([Mg 2ϩ ]1 ⁄ 2 of 0.54 Ϯ 0.06 mM, n ϭ 3, and 0.25 Ϯ 0.025 mM, n ϭ 6, for D64N/D144N mutant and wild-type GCAP-1, respectively). Strangely enough, another EF(2,4) Ϫ mutant, D68G/E75Q/D144N/ D148G, exhibited noticeably reduced affinity of EF-hand 3 for Mg 2ϩ ([Mg 2ϩ ]1 ⁄ 2 ϭ 6.0 Ϯ 0.6 mM, n ϭ 3; Fig. 6A). These results demonstrate that the EF-hand 3 in GCAP-1 efficiently binds Mg 2ϩ , although to reveal that, special care needs to be taken in selecting mutations affecting the metal binding in the neighboring EF-hands 2 and 4.
Ca 2ϩ -and Mg 2ϩ -dependent Changes in Trp Fluorescence of Trp to Phe Mutants-Bovine GCAP-1 contains three Trp residues. Based on the close similarity between GCAP-1 and CGAP-2, Trp-21 is likely located in the ␣-helical structure of the EF-hand 1, Trp-51 is located between the EF1 and EF2, and Trp-94 is in the entering ␣-helix of EF3 (Fig. 1A). To assign the Ca 2ϩ -and Mg 2ϩ -dependent changes in Trp fluorescence of GCAP-1 to the change in the environment of the specific Trp residues, one or two Trp residues in GCAP-1 were replaced with Phe. The following mutants were constructed: GCAP-1(W21F), GCAP-1(W51F), GCAP-1(W94F), GCAP-1(W51F/ W94F), GCAP-1(W21F/W94F), GCAP-1(W21F/W51F) and GCAP-1(W21F/W51F/W94F). In a separate set of experiments, Trp to Phe mutants were able to activate RetGC-1 (not shown; also see Ref 20). Also we found that similar to wild type, all seven GCAP-1 mutants bound three Ca 2ϩ ions with the apparent K d near 0.1 M in the absence and 0.2-0.4 M in the presence of 1 mM free Mg 2ϩ with a Hill coefficient near 1.0.
In all mutants that have a single Trp residue replaced by Phe (Fig. 7), Mg 2ϩ suppressed fluorescence, similar to the wild-type GCAP-1. Unlike that, the Ca 2ϩ -specific changes were different among the three mutants: (i) both W21F and W51F mutants exhibited a Ca 2ϩ -dependent increase in Trp fluorescence of phase II (in the absence or in the presence of Mg 2ϩ ), (ii) the fluorescence intensity of the metal-free versus Ca 2ϩ -bound form was reduced compared with the wild type, and (iii) the Ca 2ϩ -dependent decrease in fluorescence constituting phase I was weakened compared with the wild type (Fig. 7, A, B, and C). In contrast to other single mutants, no phase II was observed for the W94F (Fig. 7D), and both Ca 2ϩ and Mg 2ϩ only decrease the fluorescence of the remaining two Trp residues. In the absence of Mg 2ϩ , the GCAP-1(W94F) exhibited a prominent, highly Ca 2ϩ -sensitive decrease in fluorescence ([Ca 2ϩ ]1 ⁄ 2 ϭ 0.02 Ϯ 0.04 M, n ϭ 3). In contrast to the wild type and W21F and W51F mutants, addition of Ca 2ϩ to the Mg 2ϩ -saturated GCAP-1(W94F) further decreased its fluorescence intensity. These results indicate that upon transition of GCAP-1 from metalfree or Mg 2ϩ -bound to the Ca 2ϩ -bound state, the Ca 2ϩ -specific rise in fluorescence intensity reflects the Trp-94 moving into a less polar environment. This was further confirmed by the properties of the double Trp to Phe mutants. In the GCAP-1(W51F/W94F), Mg 2ϩversus Ca 2ϩ -specific changes around Trp-21 could be clearly distinguished by their amplitudes (Fig. 8A). The Mg 2ϩ binding to GCAP-1(W51F,W94F) resulted in only ϳ10% decrease in the Trp-21 fluorescence intensity at saturation, whereas Ca 2ϩ binding caused Ͼ50% decrease at saturation. Evidently compared with the metal-free GCAP-1, binding of Mg 2ϩ ([Mg 2ϩ ]1 ⁄ 2 ϭ 0.16 Ϯ 0.04 mM, n ϭ 3) at its physiological concentrations of 0.9 mM [Mg 2ϩ ] f forces Trp-21 to move into a more hydrophilic environment. These are conditions that correspond to the light-adapted rods (3, 23) when GCAP-1 activates RetGC. Replacement of Mg 2ϩ with Ca 2ϩ promotes further exposure of the Trp-21 to the solution (Fig. 8A, left panel). The concentration of Ca 2ϩ required for the displacement depends on the concentration of free Mg 2ϩ (Fig. 8A, left panel), but most of the conformational change specific for Ca 2ϩ ([Ca 2ϩ ]1 ⁄ 2 ϭ 0.11 Ϯ 0.014 M, n ϭ 3) at 0.9 mM free Mg 2ϩ was observed under the conditions that are consistent with the free Ca 2ϩ and Mg 2ϩ in dark-adapted rods (3,23). Similar to Trp-21, the fluorescence intensity of the Trp-51 in GCAP-1(W21F/W94F) also decreased upon Mg 2ϩ binding (Fig. 8B) compared with the metal-free protein ([Mg 2ϩ ]1 ⁄ 2 ϭ 0.21 Ϯ 0.04 mM, n ϭ 3), consistent with the Trp-51 becoming exposed to a more hydrophilic environment. Binding of Ca 2ϩ to metal-free GCAP-1 also decreased fluorescence of Trp-51 ([Ca 2ϩ ]1 ⁄ 2 ϭ 0.024 Ϯ 0.05 M, n ϭ 3). However, unlike Trp-21, transition of GCAP-1 from Mg 2ϩ -bound into Ca 2ϩ -bound form did not further affect the environment for the Trp-51. These results suggest that the Ca 2ϩ -specific rise in the fluorescence intensity (phase II) is solely due to the structural movement around Trp-94. Indeed in the GCAP-1(W21F/W51F), Ca 2ϩ binding sharply increased fluorescence of the remaining Trp-94 both in the presence and in the absence of Mg 2ϩ (Fig. 8C). In contrast, binding of Mg 2ϩ in the absence of Ca 2ϩ resulted in a small decrease of the Trp-94 fluorescence ([Mg 2ϩ ]1 ⁄ 2 ϭ 0.25 Ϯ 0.06 mM, n ϭ 3), and Mg 2ϩ also affected the sensitivity of the Ca 2ϩ -dependent rise in the Trp-94 fluorescence ([Ca 2ϩ ]1 ⁄ 2 of 0.12 Ϯ 0.03 M, n ϭ 3, and 0.26 Ϯ 0.04 M, n ϭ 3, in the absence and in the presence of 0.9 mM Mg 2ϩ , respectively).

Inactivation of Cation Binding in GCAP-1 Blocks Activation of
RetGC-It has already been demonstrated that inactivation of EFhands by substitution of the last Glu in the 12-amino acid Ca 2ϩ -binding loop of the EF-hand with Gln or Asp, although preventing Ca 2ϩ binding under physiological conditions, does not hamper RetGC activation by the mutant GCAPs (12,16). However, in the present study we found that such substitutions did not block Mg 2ϩ binding to the mutated EF-hands (Figs. 5 and 6). At the same time, Mg 2ϩ is the essential cofactor for the RetGC (39) and therefore is always present in the RetGC assay (12,16). Because Mg 2ϩ cannot be excluded from the reaction, we constructed three different GCAP-1 variants in which point mutations were introduced in all three EFhands ( Fig. 9) that prevented binding of not only Ca 2ϩ but also Mg 2ϩ (Figs. 5 and 6). None of these three mutants was able to activate RetGC (Fig. 9, columns b-d) in the conditions in which RetGC is saturated with Mg 2ϩ (13) and is efficiently activated by wild-type GCAP-1 (Fig. 9, column a).

DISCUSSION
All Three Active EF-hands in GCAP-1 Bind Ca 2ϩ and Mg 2ϩ -GCAP-1 has four EF-hand motifs, but the amino acid sequence of only three of them corresponds to true metal-binding domains (40), whereas the N-terminal EF-hand 1 does not completely match the consensus sequence for the coordinating loop and does not bind Ca 2ϩ (18). Contrary to the model proposed by Hwang et al. (18), our results directly demonstrate that wildtype GCAP-1, even isolated in the presence of Ca 2ϩ , can subsequently lose all bound Ca 2ϩ in non-denaturing conditions and become able to bind not two but three Ca 2ϩ ions in a direct binding assay.
Importantly we found that all three Ca 2ϩ binding EF-hands could also bind Mg 2ϩ . Moreover at 1 mM physiological concentration of Mg 2ϩ and Ͻ25 nM free concentration of Ca 2ϩ in photoreceptor in the light all three active EF-hands in GCAP-1 were nearly saturated with Mg 2ϩ . Hence the direct conclusion from our observations is that the apo form of GCAP-1, previously regarded as the activator form for RetGC, simply does not exist under physiological conditions.
Ca 2ϩ Binding Properties of GCAP-1 and RetGC Regulation-The apparent dissociation constant of GCAP-1 for Ca 2ϩ determined at 1 mM Mg 2ϩ is remarkably close to the [Ca 2ϩ ]1 ⁄ 2 value for the Ca 2ϩ effect on RetGC regulation by GCAP-1 determined in our previous publications (13,22).
Recently we showed that the dynamic range for the RetGC regulation by Ca 2ϩ /GCAP in vitro is determined by both the affinity of GCAP for Ca 2ϩ and relative affinities of the cyclase for the Ca 2ϩ -free versus Ca 2ϩ -loaded GCAP (22). The model (22), used to quantitatively describe the mechanism of the abnormal Ca 2ϩ sensitivity of the R838S mutant of RetGC-1 (39,41,42), was based on an important assumption that the interaction between GCAP and RetGC did not drastically affect GCAP affinity for Ca 2ϩ . Applying that model, the [Ca 2ϩ ]1 ⁄ 2 for RetGC inhibition can be determined using the equation (22), where K A and K ACa are apparent affinity constants of RetGC for Ca 2ϩ -free and Ca 2ϩ -loaded GCAP, respectively; K d is an appar-     Fig. 1, B and C). However, another important observation is that binding of Ca 2ϩ by purified GCAP-1 in solution is non-cooperative. This apparently distinguishes GCAP-1 from GCAP-2, which also binds three Ca 2ϩ ions but with a cooperativity factor of 2.1 Ϯ 0.2 (43). Yet the cooperativity of RetGC inhibition by Ca 2ϩ / GCAP-1 remains close to 2 (Refs. 13 and 22 and Fig. 1B in this study); therefore, the cooperativity of Ca 2ϩ binding by GCAP-1 in complex with RetGC is likely affected by its interaction with the target enzyme. There are two principal possibilities why Ca 2ϩ affects RetGC in a cooperative manner. First, there is strong evidence that RetGC is a dimer (39,44,45) whose active center is formed by a dimer of catalytic domains. Hence the replacement of Mg 2ϩ by Ca 2ϩ in the EF-hands of two different GCAP-1 molecules attached to two different RetGC subunits may produce the cooperativity of the Ca 2ϩ effect observed in RetGC assay. Second, binding of GCAP-1 to the cyclase may affect GCAP conformation and thus change the interactions between EF-hands within the same molecule of GCAP-1. It is currently difficult to assess what the actual mechanism affecting Ca 2ϩ cooperativity of RetGC regulation may be because methods for isolation of the active complex between RetGC and GCAP-1 remain unavailable.
Ca 2ϩ Versus Mg 2ϩ Selectivity of EF-hands in GCAP-1-The structure of EF-hand loop in Mg 2ϩ -bound GCAP-1 can significantly differ from that of the Ca 2ϩ -bound form because Mg 2ϩ can bind to the same 12-amino acid cation-binding region in the EF-hand as Ca 2ϩ but uses different coordinating geometry (37,46). Therefore, some point mutations that effectively disable Ca 2ϩ binding by a particular EF-hand may only slightly affect Mg 2ϩ binding. For example, substitution of the first Asp in the 12-amino acid metal-binding loop of EF-hands 2 and 3 with Asn effectively disabled Mg 2ϩ binding by the EF-hands, whereas the substitution of the last Glu in the metal-binding loop with Gln or Asp had very little or no effect on Mg 2ϩ binding (Figs. 5 and 6). Double mutations in two Ca 2ϩ -coordinating positions (D68G/E75Q in EF-hand 2, D100N/D102G in EF3, and D144N/D148G in EF-hand 4) also effectively prevent binding of Mg 2ϩ (Figs. 5 and 6).
Cation-induced Conformational Changes in GCAP-1-We previously found that GCAPs are not just Ca 2ϩ but rather Ca 2ϩ /Mg 2ϩ sensor proteins, which makes Ca 2ϩ sensitivity of the cyclase operate within the proper range of intracellular Ca 2ϩ concentrations (5,13). In the present study, we found that Mg 2ϩ binding decreased GCAP-1 fluorescence, corresponding to the overall conformation of the protein becoming less compact around all three Trp residues present in GCAP-1.
The fluorescence of all three GCAP-1 Trp residues, Trp-21, -51, and -94, was affected by conformational changes between its metal-free, Mg 2ϩ -bound, and Ca 2ϩ -bound forms. Unlike Ca 2ϩ , transition of GCAP-1 between the metal-free and the Mg 2ϩ -bound forms was accompanied by movement of all three Trp residues into a more hydrophilic environment (i.e. solution). This may be a particularly interesting observation in the case of Trp-21, which is located in the N-terminal EF-hand-like structure that itself cannot bind divalent cations. This implies that a major conformational change occurs in the EF-hand 1 caused by Ca 2ϩ or Mg 2ϩ binding in the neighboring EFhand(s). The EF-hand 1 in GCAPs is essential for their ability to interact with the target enzyme, RetGC (17,18,47). Therefore the sensitivity of the EF-1 conformation to the metal-bound status of GCAP-1 may constitute a part of the mechanism by which GCAPs regulate activity of RetGC.
The Mg 2ϩ -bound (RetGC activator) form of GCAP-1 has all three Trp residues partially exposed to the polar environment; hence the conformation of the RetGC activator form around all of them appears more open compared with the metal-free conformation. GCAP-1 should be able to maintain its fully Mg 2ϩbound, more "relaxed" form in the light when the physiological concentrations of free Ca 2ϩ and Mg 2ϩ in the photoreceptor are near 20 nM and 1 mM, respectively. As free Ca 2ϩ concentrations rise in the dark, binding of Ca 2ϩ in EF-hands 2 and 3 further increases the exposure to solution of Trp-21, located in the entering ␣-helical structure of the EF-hand 1, but produces no major additional change around Trp-51 (the loop between EFhands 1 and 2). At the same time, binding of Ca 2ϩ in the EFhand 4 causes Trp-94 in the entering ␣-helix of the EF3 to move into the opposite, more hydrophobic environment, presumably formed at the junction between the entering ␣-helix of EF2 and the entering ␣-helix of EF3. This movement would likely place Trp-94 deeper into the protein structure compared with the more relaxed Mg 2ϩ -bound form of GCAP-1.
Mg 2ϩ Role in GCAP-1 Function as RetGC Regulator under Physiological Conditions-Replacement of Ca 2ϩ by Mg 2ϩ in GCAPs plays an essential role in the RetGC regulation process.