In vertebrate rods and cones, photon absorption by rhodopsin or cone visual pigments triggers a phototransduction cascade that hydrolyzes cGMP, resulting in the closure of cGMP-gated cation channels on the plasma membrane that causes membrane hyperpolarization [see reviews (Pugh et al., 1997, 1999)]. To reset the resting dark-state of retinal rods (known as visual recovery), cGMP levels are replenished very quickly (Burns and Baylor, 2002) by retina-specific guanylyl cyclases (RetGCs), a membrane enzyme present in rod and cone outer segments (Dizhoor et al., 1994; Lowe et al., 1995). RetGC is a Ca²⁺-regulated enzyme (Koch and Stryer, 1988; Koutalos and Yau, 1996) whose activity is controlled by intracellular domains (Laura et al., 1996; Duda et al., 2005) that interact with soluble EF-hand Ca²⁺ sensor proteins, called guanylyl cyclase activating proteins (GCAPs: GCAP1 and GCAP2; Dizhoor et al., 1994, 1995; Gorczyca et al., 1994, 1995; Koutalos et al., 1995).

Light-induced channel closure in photoreceptor cells causes a decrease in the cytosolic free Ca²⁺ concentration (Gray-Keller and Detwiler, 1994), in mammals from 250 nM in the dark to 25 nM in the light (Woodruff et al., 2002). The catalytic activity of RetGC in the dark is negatively controlled by Ca²⁺-bound GCAPs (Dizhoor and Hurley, 1996; Dizhoor et al., 1998; Burns and Baylor, 2002), whereas the release of Ca²⁺ from GCAPs at low Ca²⁺ levels in light-activated photoreceptors causes activation of RetGC (Dizhoor et al., 1994, 1995; Gorczyca et al., 1994, 1995; Dizhoor and Hurley, 1996; Mendez et al., 2001). Light stimulation of the rod cell causes a more than 10-fold increase in cGMP production due to the activation of RetGC by GCAPs (Hodgkin and Nunn, 1988; Burns and Baylor, 2002) and is a critical step for controlling the recovery rate of a single-photon response (Pugh et al., 1999; Burns and Baylor, 2002) as well as the cone response to stronger light stimuli (Sakurai et al., 2011). In mouse rods, GCAPs have been demonstrated to have different Ca²⁺ sensitivities (Dizhoor et al., 1998; Hwang et al., 2003) and therefore contribute to the recovery by activating guanylyl cyclase at different steps of excitation and recovery, thus imparting proper recovery kinetics to the rod response (Mendez et al., 2001; Makino et al., 2008, 2012).

Guanylyl cyclase activating proteins activate RetGC at low Ca²⁺ levels (less than 50 nM) and only in the presence of physiological Mg²⁺ levels (Peshenko and Dizhoor, 2004, 2006, 2007; Dizhoor et al., 2010). This Mg²⁺ requirement for RetGC activation by GCAPs initially suggested that Mg²⁺ binding to GCAPs might be important for their activation of RetGC. Indeed, Mg²⁺ was shown to bind directly to at least two of the EF-hands in GCAP1 (Lim et al., 2009), and NMR studies showed that Mg²⁺ binding to GCAP1 at EF2 and EF3 was needed to stabilize the overall tertiary fold of the protein (Lim et al., 2009). By contrast, the Ca²⁺-free/Mg²⁺-free GCAP1 (apo-state) forms a molten globule-like structure, that contains regular secondary structure (Dell’Orco et al., 2010) but does not form a stable tertiary fold (Peshenko and Dizhoor, 2004; Lim et al., 2009). The flexible and unstructured molten-globule apo-protein could explain in part why GCAPs do not activate RetGC in the absence of Mg²⁺ (Dizhoor et al., 2010). Thus, Mg²⁺ binding to GCAP1 stablizes its protein structure in order to bind and activate RetGC (Dizhoor et al., 1994; Peshenko and Dizhoor, 2004), whereas Ca²⁺ binding to GCAP1 stabilizes a distinct structure important for the inhibition of RetGC (Dizhoor et al., 1998).

The four EF-hands in the GCAPs have quite distinct divalent metal binding properties that control whether GCAPs can activate or inhibit RetGC. The first EF-hand (EF1) does not bind to either Ca²⁺ or Mg²⁺ because the residue at the 3-position in the EF-hand binding loop (Cys29 in GCAP1, see Figure 1) is not suitable for ligating either Ca²⁺ or Mg²⁺. Ca²⁺ binds to GCAP1 at the other three EF-hands (EF2, EF3, and EF4) in an independent fashion (Lim et al., 2009) in contrast to the cooperative binding of two Ca²⁺ to recoverin (Ames et al., 1995a). The apparent dissociation constant for Ca²⁺ binding to GCAPs is in the submicromolar range (Lim et al., 2009; Dizhoor et al., 2010), whereas Mg²⁺ binds with ~1000-fold lower affinity (Gifford et al., 2007) in the sub-millimolar range (Peshenko and Dizhoor, 2004, 2007; Lim et al., 2009). These binding affinities imply that three Ca²⁺ bind per mole of GCAP1 in dark-adapted rod cells, which have relatively high cytosolic Ca²⁺ levels [Ca²⁺]_free = 250–500 nM [(Woodruff et al., 2002; Matthews and Fain, 2003) and [Mg²⁺] ~1 mM (Chen et al., 2003)]. Light-activation of the rod cell causes a dramatic lowering of the cytosolic Ca²⁺ level [Ca²⁺]_free = 5–50 nM (Gray-Keller and Detwiler, 1994; Sampath et al., 1998; Woodruff et al., 2002) while the Mg²⁺ level remains fixed at [Mg²⁺]_free ~1 mM (Chen et al., 2003). Therefore in light-adapted rods, GCAPs do not bind Ca²⁺ but instead bind to at least two Mg²⁺. Thus, Ca²⁺-free/Mg²⁺-bound GCAPs activate RetGC in light exposed rods (Dizhoor et al., 1994, 1995; Gorczyca et al., 1995), whereas Ca²⁺-bound GCAPs (with Ca²⁺ bound at EF2, EF3, and EF4) inhibit RetGC in dark-adapted rods (Dizhoor and Hurley, 1996; Dizhoor et al., 1998).

Various point mutations in the EF-hand motifs of GCAP1 that weaken Ca²⁺ binding (but do not affect Mg²⁺ binding) cause GCAP1 to constitutively activate RetGC in rods and cones, which is genetically linked to various retinal diseases (Jiang and Baehr, 2010). These mutations in the EF-hand motifs [Y99C (Dizhoor et al., 1998; Payne et al., 1998) and E155G (Wilkie et al., 2000, 2001)] weaken the Ca²⁺ binding affinity beyond the photoreceptor Ca²⁺ concentration and cause the Ca²⁺-free/Mg²⁺-bound GCAP1 activator state to persist even at high Ca²⁺ levels in dark-adapted rods, which causes persistent activation of RetGC (Sokal et al., 1998; Olshevskaya et al., 2004). The GCAP mutants that constitutively activate RetGC then cause elevated cGMP levels in photoreceptor cells that promotes apoptosis and disease (Olshevskaya et al., 2004, 2012; Woodruff et al., 2007).

Mutagenesis studies of the individual EF-hands in GCAP1 have revealed that Ca²⁺-binding to EF4 is critical for controlling Ca²⁺-dependent activation of RetGC (Peshenko and Dizhoor, 2007). Mutants of GCAP1 that weaken or abolish Ca²⁺ binding to EF4 but retain Ca²⁺ binding at EF2 and EF3 [D144N/D148G (Peshenko and Dizhoor, 2007 and E155G Wilkie et al., 2000, 2001)] are constitutively active even in the presence of high Ca²⁺ levels in dark-adapted photoreceptors. Furthermore, these mutants that disable Ca²⁺ binding to EF4 (but not EF2 and EF3) are unable to inhibit RetGC at high Ca²⁺ levels in the dark (Peshenko and Dizhoor, 2007). In summary, GCAP1 can activate RetGC even if Ca²⁺ is bound to EF2 and EF3 (but not bound to EF4). Also, Ca²⁺ binding to EF4 is essential for having Ca²⁺-induced inhibition of RetGC. Therefore, Ca²⁺ binding to EF4 is critical for controlling whether GCAP1 can activate or inhibit RetGC.

Ca²⁺ binding to GCAP1 and GCAP2 does not cause ejection of the covalently attached myristoyl group (Hughes et al., 1998; Lim et al., 2009, 2013) and Ca²⁺ binding to GCAPs do not promote their membrane targeting (Olshevskaya et al., 1997; Hwang and Koch, 2002). This is in stark contrast to the Ca²⁺-induced exposure of the N-terminal myristoyl group [termed Ca²⁺-myristoyl switch (Zozulya and Stryer, 1992; Dizhoor et al., 1993)] that promotes membrane targeting of recoverin (Zozulya and Stryer, 1992; Dizhoor et al., 1993; Valentine et al., 2003), neurocalcin (Ladant, 1995), hippocalcin (O’Callaghan et al., 2003), VILIPs (Spilker et al., 1997, 2002), and NCS-1 (Hamasaki et al., 2004). Instead, the covalently attached myristoyl group of GCAP1 remains sequestered inside the protein hydrophobic core in both Ca²⁺-free and Ca²⁺-bound forms of GCAP1 (Lim et al., 2009, 2013). NMR studies on the myristate attached to recoverin (Ames et al., 1995b, 1997; Hughes et al., 1995), VILIP1 (Li et al., 2011), and NCS-1 (Lim et al., 2011) reveal that the covalently attached myristoyl group is buried inside these proteins only in the Ca²⁺-free state. Ca²⁺-binding causes protein conformational changes that lead to exposure of the fatty acyl chain in recoverin (Ames et al., 1997), VILIP1 (Li et al., 2011), and NCS-1 (Lim et al., 2011). By stark contrast, NMR studies on GCAP1 indicate that the covalently attached myristoyl group is buried inside both Ca²⁺-free and Ca²⁺-bound GCAP1 (Lim et al., 2009, 2013). Further evidence that GCAP1 lacks a Ca²⁺-myristoyl switch comes from SPR studies that show myristoylation of GCAP1 has little effect on membrane binding (Hwang and Koch, 2002). Finally, the recent atomic level structures of GCAP1 directly demonstrated that the myristoyl group is buried inside the protein in a similar environment in both the Ca²⁺-free activator (Lim et al., 2013) and Ca²⁺-bound inhibitor states (Stephen et al., 2007).

Solid-state NMR and other spectroscopic studies on GCAP2 have suggested that Ca²⁺-free GCAP2 might be targeted to cell membranes by a reversed Ca²⁺-myristoyl switch (Theisgen et al., 2010, 2011). The covalently attached fatty acyl group has been suggested to become exposed in Ca²⁺-free GCAP2 in the presence of lipid bilayer membranes (Theisgen et al., 2011), in contrast to having the myristoyl group sequestered inside Ca²⁺-bound GCAP2 (Schroder et al., 2011). However, other studies on GCAP2 indicate that the myristoyl group remains sequestered inside the protein environment in both Ca²⁺-free and Ca²⁺-bound GCAP2 (Hughes et al., 1998). Also, unmyristoylated GCAP2 activates RetGC with nearly the same potency as that of myristoylated GCAP2, and myristoylation of GCAP2 is not essential for its targeting to the membrane-bound cyclase (Olshevskaya et al., 1997; Hwang and Koch, 2002).

Atomic level structures are known for Ca²⁺-bound forms of GCAP1 (Stephen et al., 2007) and GCAP2 (Ames et al., 1999). The four EF-hands in GCAP1 (Figures 1 and 3) are grouped into two globular domains: N-domain includes EF1 and EF2 (residues 18–83) and C-domain contains EF3 and EF4 (residues 88–161). Ca²⁺ is bound to GCAP1 at EF2, EF3 and EF4, and the structure of each Ca²⁺-bound EF-hand in GCAP1 (Figure 3B) adopts the familiar open conformation of EF-hands as seen in calmodulin and other Ca²⁺-bound EF-hand proteins (Ikura, 1996). Indeed, the interhelical angles for each Ca²⁺-bound EF-hand in GCAP1 are similar to those of recoverin (Ames et al., 1997) and NCS-1 (Bourne et al., 2001). Although the internal structure of each EF-hand in GCAP1 is similar to that of recoverin, the overall three-dimensional packing arrangement and spatial organization of the four EF-hands is very different for GCAP1 vs. recoverin. Indeed, the overall root-mean-squared deviation of main chain atoms is 3.4 Å when comparing the structures of GCAP1 and recoverin (Figure 2). A unique structural feature of GCAP1 is that the N-terminal α-helix (residues 5–15) upstream of EF1 and C-terminal helix (residues 175–183) downstream of EF4 are held closely together by their mutual interaction with the N-terminal myristoyl group (Figure 2F). Thus, the covalently attached myristoyl group in GCAP1 is sequestered within a unique environment inside the Ca²⁺-bound protein, which is quite different from that of recoverin as described above in Figure 2. The myristoyl group attached to GCAP1 makes contacts with N-terminal residues (V9, L12, and F42) and the C-terminal helix (L174, V178 and I181; Figure 2F). In essence, the myristoyl group serves to bridge both the N-terminal and C-terminal ends of the protein, which explains how Ca²⁺-induced conformational changes in the C-terminal domain (particularly in EF4) might be transmitted to a possible target binding site in EF1. A Ca²⁺-myristoyl tug mechanism (Peshenko et al., 2012) has been proposed to explain how Ca²⁺-induced conformational changes in EF4 serve to “tug” on the adjacent C-terminal helix that connects structurally to the myristoyl group and EF1. This tug mechanism helps explain how Ca²⁺-induced structural changes in EF4 might be relayed to the cyclase binding region in EF1 (Lim et al., 2013). The Ca²⁺-induced structural changes involving the C-terminal helix might also be related to Ca²⁺-dependent phosphorylation of S201 in GCAP2 (Peshenko et al., 2004).

The atomic level structure of Ca²⁺-free/Mg²⁺-bound activator form of GCAP1 or GCAP2 is currently not known. The main difficulty is that Ca²⁺-free/Mg²⁺-bound GCAP proteins form dimers and higher order protein oligomers that causes considerable sample heterogeneity at high protein concentrations needed for NMR or to make crystals for X-ray crystallography (Ames et al., 1999; Lim et al., 2009, 2013). Ca²⁺-dependent dimerization of GCAP2 has been suggested to be important for activating the cyclase (Olshevskaya et al., 1999b; Ermilov et al., 2001). Protein dimerization was also reported for GCAP1, and a GCAP1 mutant (V77E) that prevents protein dimerization also abolishes its ability to activate RetGC, suggesting that dimerization of Ca²⁺-free/Mg²⁺-bound activator state might be important for activating RetGC (Lim et al., 2013). However, protein dimerization of GCAP1 is not Ca²⁺ dependent and only occurs at relatively high protein concentrations above 100 μ M (Lim et al., 2013). Future studies are needed to investigate whether GCAPs might form a functional dimer upon binding to the dimeric RetGC to form a 2:2 complex (Ramamurthy et al., 2001; Peshenko et al., 2010), and whether the dimeric quaternary structure may play a regulatory role.

A GCAP1 mutant, D144N/D148G called EF4mut, (Lim et al., 2013) that binds Ca2+ at EF2 and EF3 (and does not bind Ca²⁺ at EF4) can activate RetGC at high Ca²⁺ concentrations (Peshenko and Dizhoor, 2007). Therefore, EF4mut (with Ca²⁺ bound at EF2 and EF3) serves as a model of the Ca²⁺-free/Mg²⁺-bound activator state. The EF4mut protein is more soluble and stable than Ca²⁺-free/Mg²⁺-bound wild type, and EF4mut exhibits NMR spectra with much higher resolution and sensitivity compared to that of Ca²⁺-free/Mg²⁺-bound wild type (Lim et al., 2009, 2013). NMR structural studies on EF4mut provide some insights for a structural model of the GCAP1 activator state (Figure 3A). The overall structure of EF4mut is similar to the crystal structure of Ca²⁺-bound GCAP1 (root mean squared deviation of main chain atoms is 1.3 Å when comparing the two structures). However, residues at the domain interface (between EF2 and EF3) are structured somewhat differently in EF4mut compared to the crystal structure of Ca²⁺-bound GCAP1. Many of the GCAP1 residues at the domain interface have quite broad NMR resonances, suggesting that these residues are conformationally dynamic (Lim et al., 2013). The corresponding residues in recoverin (Figure 1) also exhibited broad NMR resonances and ¹⁵N NMR relaxation dispersion studies reveal that these domain interface residues exhibit millisecond exchange kinetics (Xu et al., 2011). Ca²⁺-induced rearrangement of residues at the domain interface in recoverin gives rise to a 45° swiveling of the two domains (Ames et al., 1997). A structural comparison between EF4mut and Ca²⁺-bound GCAP1 suggests a related but much smaller Ca²⁺-induced structural change at the domain interface in GCAP1 (Figure 3). The most noteworthy Ca²⁺-induced structural difference in GCAP1 can be seen in the entering helix of EF3 that unravels a half turn in EF4mut, which causes a repositioning of the W94 side-chain at the domain interface. A Ca²⁺-induced change in the structural environment around W94 is consistent with previous tryptophan fluorescence and electron paramagnetic resonance studies of GCAP1 (Sokal et al., 2001; Peshenko and Dizhoor, 2006, 2007). We suggest that Ca²⁺-induced rearrangement of residues at the domain interface (V77, A78, L82, K85, and W94) plays a role in modulating Ca²⁺-dependent contacts with RetGC1.

The structural information for the GCAPs above provides insights into the activation mechanism of RetGC (Figure 4). GCAP1 residues in EF1 (Krylov et al., 1999; Olshevskaya et al., 1999a,b; Ermilov et al., 2001) and at the domain interface (Krylov et al., 1999; Lim et al., 2013; Peshenko et al., 2014) are suggested to make direct contact with RetGC (see labeled residues in Figure 4). In our model (Figure 4), Ca²⁺-induced conformational changes in EF4 (Lim et al., 2013) are transmitted to the cyclase binding site (see labeled residues in EF1, EF2, and EF3, Figure 4) by a Ca²⁺-myristoyl tug mechanism as described by (Peshenko et al., 2012). In the GCAP1 activator state under physiological conditions (Figure 4, left panel), EF2 and EF3 are bound to Mg²⁺ (blue circles in Figure 4) with EF1 and EF4 unoccupied. The Ca²⁺-free state of EF4 forms a loose and dynamic structure (Lim et al., 2013), which allows the adjacent C-terminal helix to reach all the way to the N-terminal myristoyl group (magenta in Figure 4) and thus indirectly form hydrophobic contacts with residues in EF1 and EF2. In essence, the myristoyl group forms a bridge between Ca²⁺-induced conformational changes in the C-terminal domain and the cyclase binding site in the N-terminal domain. In the GCAP1 activator state, residues in the cyclase binding site [see Figure 4, labeled residues in EF1, EF2 and EF3 (Peshenko et al., 2014)] are spatially close together and form particular contacts with RetGC that require close proximity between W94 and V77 (see arrow in Figure 4). In the Ca²⁺-bound GCAP1 inhibitor state (Figure 4, right panel), Ca²⁺ is bound to EF2, EF3, and EF4 (orange circles in Figure 4). Ca²⁺-binding to EF4 causes local conformational changes that in turn “tug” on the C-terminal helix which causes a slight reorientation of the N-terminal domain (EF1 and EF2) with respect to the C-terminal domain (EF3 and EF4). This domain swiveling causes key residues in EF3 at the domain interface (K85, W94, and K97) to move farther away from cyclase binding site residues in EF2 [F73, V77, and A78 (Peshenko et al., 2014)], which disrupts key contacts to RetGC that we suggest may be important for regulating cyclase activation.