Interactions within the Coiled-coil Domain of RetGC-1 Guanylyl Cyclase Are Optimized for Regulation Rather than for High Affinity*

RetGC-1, a member of the membrane guanylyl cyclase family of proteins, is regulated in photoreceptor cells by a Ca 2 1 -binding protein known as GCAP-1. Proper regulation of RetGC-1 is essential in photoreceptor cells for normal light adaptation and recovery to the dark state. In this study we show that cGMP synthesis by RetGC-1 requires dimerization, because critical functions in the catalytic site must be provided by each of the two polypeptide chains of the dimer. We also show that an intact a -helical coiled-coil structure is required to pro-vide dimerization strength for the catalytic domain of RetGC-1. However, the dimerization strength of this domain must be precisely optimized for proper regulation by GCAP-1. We found that Arg 838 within the dimerization domain establishes the Ca 2 1 sensitivity of RetGC-1 by determining the strength of the coiled-coil interaction. Arg 838 substitutions dominantly enhance cGMP synthesis even at the highest Ca 2 1 concentrations that occur in normal dark-adapted photoreceptor cells. Molecular dynamics simulations suggest that Arg 838 substitutions disrupt a with respect to sliding along the helix axis while aligning nonpolar and complementary-charged groups at the interface. An antiparallel coiled-coil was also investigated, but complementarity of the residues along the interface was not as good as in the parallel arrangement, resulting in a ; 37 kcal/mol higher (less favorable) potential energy after 1000 steps of in vacuo minimization. Amino acid mutations were made to the minimized parallel structure to yield mutant homodimers: R838C and R838S. The proteins were then each solvated using a box extending at least 8 Å in all directions, resulting in the addition of ; 3870 water molecules. The water density was set to the experimental value at 25 °C of 0.997 g ml 2 1 , and periodic boundary conditions were employed to reduce edge effects. The solvent was then subjected to 1000 cycles of minimization, 2 ps of MD, and another 1000 steps of minimization. This water prep-aration was followed by 1000 steps of minimization of the protein in the field of the solvent and then 1000 steps of minimization of the full protein-solvent system. After the preparatory steps described above, the system was heated to 25 °C. Each independent simulation was carried out for 5 ns using a 2-fs integration time step (5000 ps, or 2.5 3 10 6 iterations). An 8-Å smooth, nonbonded cutoff was used, and the nonbonded list was updated every 2 cycles; this procedure has been shown to yield stable and reasonable simulations (35, 36). 25,000 structures were saved for analysis (5 per ps) from each trajectory.

Photoexcitation of vertebrate photoreceptor cells reduces the intracellular concentration of cGMP and closes cGMP-gated cation channels in the plasma membrane of the photoreceptor cell outer segment. This hyperpolarizes the photoreceptor cell and lowers the free Ca 2ϩ concentration in the outer segment (1,2). Guanylyl cyclase is stimulated by the lowered [Ca 2ϩ ] to synthesize cGMP and enhance photoreceptor cell recovery and light adaptation. There are two forms of photoreceptor cell membrane guanylyl cyclases in humans, RetGC-1 and RetGC-2. Immunofluorescence studies suggest that RetGC-1 is primarily expressed in cones and to a lesser extent in rods (3,4). Mutations in the RetGC-1 gene have been linked to Leber congenital amourosis, a severe retinopathy inherited in an autosomal recessive mode (5).
RetGC-1 belongs to the membrane GC 1 family that includes natriuretic peptide receptor-GCs (NPR-A/GC-A and NPR-B/ GC-B), heat stable enterotoxin (StaR/GC-C), and sea urchin sperm GCs (6 -8). Each member of the membrane GC family is a type I transmembrane protein that has an extracellular domain linked by a single transmembrane domain to an intracellular catalytic domain. In contrast to most other membrane GCs that are stimulated by the binding of a ligand to extracellular domain, RetGC-1 is stimulated intracellularly by a Ca 2ϩbinding protein, GCAP-1 (3,9,10). Both GCAP-1 and a related protein, GCAP-2, have four EF-hand-like domains, including three functional EF-hands that bind Ca 2ϩ . At low levels of Ca 2ϩ , the Ca 2ϩ -free form of GCAP-1 activates photoreceptor cell membrane guanylyl cyclase, but at high (ϳ1 M) Ca 2ϩ concentrations, the Ca 2ϩ -bound form of GCAP-1 inhibits it. This mechanism of regulation was established by demonstrating that mutations that prevent Ca 2ϩ binding to the EF-hands turn GCAPs into constitutive activators of photoreceptor cell membrane guanylyl cyclases (11,12). Therefore, the effect of Ca 2ϩ on photoreceptor cell membrane guanylyl cyclases is mediated solely through Ca 2ϩ binding to the GCAP protein and not through a direct interaction of Ca 2ϩ with the cyclase. This type of Ca 2ϩ -sensitive regulation of photoreceptor cell membrane guanylyl activity by GCAPs occurs both with photoreceptor cell homogenates and with recombinant RetGC-1 expressed in HEK-293 or COS-7 cell lines (9,13,14).
Several lines of evidence suggest that membrane guanylyl cyclases function as dimers. A recent report showed that the structure of the extracellular domain of NPR-A is a dimer (15), and modeling studies of guanylyl cyclases based on the known structure of adenylyl cyclases suggest that the catalytic do-mains function as dimers (16 -20). A cross-linking study by Yu et al. (21) also suggested that dimerization of RetGC-1 may be an essential step in regulation by GCAPs.
A putative coiled-coil structure that is near the catalytic domains of all membrane GCs enhances dimerization of GC-A (22). The ␣-helical coiled-coil is the most widespread subunit oligomerization motif found in proteins (23). However, the role of the coiled-coil domain in regulation of RetGC-1 has not been previously established.
Recent genetic linkage studies have shown that mutations at the C-terminal end of the coiled-coil domain of RetGC-1 cause a form of autosomal dominant cone-rod dystrophy (adCORD), a human disease in which cones and then rods degenerate (24 -26). Those studies identified four independent mutations in RetGC-1 linked to adCORD. Each of them involves a substitution at Arg 838 , highlighting the critical role of this residue. We showed previously that one of these mutations, R838C, alters the Ca 2ϩ sensitivity of GCAP-1-stimulated RetGC-1 activity (27). Here we report additional biochemical and structural analysis of this and other Arg 838 mutations. Surprisingly, our analyses reveal that Arg 838 may be part of a critical network of salt bridges that normally destabilizes the active state of the cyclase. Arg 838 substitutions disrupt this network. Consequently, the coiled-coil structure is extended in the mutant enzymes resulting in abnormally enhanced stability of the dimerized catalytic domain. In other words, the normal molecular interactions within this structure are in fact not optimized for maximum extension and stability of the coiled-coil, and instead Arg 838 represents a stop signal for dimerization. Thus, the coiled-coil domain of RetGC-1 appears to be fine-tuned for optimal regulation by physiological concentrations of Ca 2ϩ .

EXPERIMENTAL PROCEDURES
Bacterial Strains, Bacteriophage, and Media-KH54, imm 21 , and the Escherichia coli strain AG1688 were kindly provided by Dr. J. C. Hu (Texas A&M University). AG1688, which carries the lacI q gene for maintaining low expression level of the fusion protein, was used to assess immunity to the tester phage KH54. KH54 carries a deletion of cI repressor. imm 21 contains a substitution of the immunity region of lambda such that lambda repressor is not able to repress its growth. The E. coli strain XL-1 blue (Stratagene, La Jolla, CA) was used for cloning of the plasmids. E. coli strains used in this study were grown routinely in Luria broth (LB) (28) supplemented with appropriate antibiotics. Antibiotics were added to cultures at the following concentrations: ampicillin (100 g ml Ϫ1 ) and kanamycin (30 g ml Ϫ1 ). All strains were stored at Ϫ70°C in LB containing 20% (v/v) glycerol.
Expression in Human Embryonic Kidney (HEK) 293 Cells-4 -5 g of the constructs was transiently transfected into 50 -60% confluent HEK-293 cells using Fugene-6 reagent (Roche Molecular Biochemicals, Indianapolis, IN). The cells were grown in Dulbecco's modified Eagle's medium/nutrient mixture F-12 (50/50) (Life technologies Inc., Rockville, MD), 10% heat-treated fetal calf serum, 10 mM HEPES, pH 7.3, in 100-mm tissue culture dishes. Cells were harvested after 48 h, washed with phosphate-buffered saline, and removed from dishes by rocking for 10 min at room temperature in phosphate-buffered saline containing 0.2% EDTA. Cells were pelleted by centrifugation and were swollen in hypotonic lysis buffer (10 mM Tris, pH 7.5/5 mM MgCl 2 /1 mM ATP/5 mM 2-mercaptoethanol) for 10 min on ice. Cells were lysed by passing four times through a 26-gauge needle. The lysates were pelleted at 2000 rpm at 4°C in a Beckman tabletop centrifuge to remove large debris and unbroken cells. The supernatant from this spin was pelleted at 14,000 rpm for 10 min, resuspended in lysis buffer, frozen on dry ice, and stored as aliquots at Ϫ70°C. Protein content was measured in the presence of 0.1%SDS using Bio-Rad protein reagent with bovine serum albumin as a standard.
Western Blotting-HEK-293 membranes containing equal amounts of total membrane protein (40 g) were electrophoresed on a 7.5% SDS-PAGE gel and transferred to nitrocellulose membranes. The membranes was were blocked in TTBS (Tris-buffered saline, 0.05% Tween 20) containing 5% dry milk for 1 h at room temperature. For detection of RetGC-1 protein, membranes were probed using a mix of rabbit polyclonal CAT-Ab (corresponding to Met 747 -Ser 1052 of RetGC-1) and KHD-Ab (corresponding to Met 443 -Ser 746 of RetGC-1) for an hour in blocking buffer (31). After extensive washing with TTBS, the membrane was probed for 1 h at room temperature with a donkey anti-rabbit antibody coupled to horseradish peroxidase (Amersham Pharmacia Biotech, Piscataway, NJ). After several washes with TTBS, the membrane was developed using chemiluminescent reagents (Amersham Pharmacia Biotech). Hemagglutinin (HA)-tagged RetGC-1 was detected using a horseradish peroxidase-linked Anti-HA (3F10) from Roche Molecular Biochemicals.
GC Assays-Membranes containing equal amounts of total protein were resuspended in GC buffer (100 mM KCl/50 mM MOPS, pH 7.4/7 mM 2-mercaptoethanol/10 mM MgCl 2 /8 mM NaCl/0.5 mM ATP). All reactions also contained the appropriate Ca 2ϩ -EGTA buffers. Measurement of guanylyl cyclase activity was carried out at 30°C for 20 min as described previously (27). Stimulated reactions contained recombinant human GCAP-1 or bovine GCAP-2. Assays measuring Mn 2ϩ /Triton X-100 activity contained 1% Triton X-100 and 10 mM MnCl 2 instead of MgCl 2 . All experiments shown were repeated at least twice from independent transfections with similar results.
Ca 2ϩ Buffer-Ca 2ϩ -EGTA buffers were prepared from solutions of EGTA (Sigma) and EGTA saturated with CaCl 2 (Fluka, Milwaukee, WI) by pH titration as described previously (32). Free Ca 2ϩ concentrations under the assay conditions were calculated using a multifactor program (33) and verified by Ca 2ϩ electrode analysis and by titration with Rhod-2 fluorescent dye (Calbiochem, La Jolla, CA).
Lambda Repressor (cI) Fusion Constructs-All repressor chimeric proteins were expressed from a lacUV5 promoter on a pBR322-based plasmid called pJH391 (34). The wild-type, coil2 (L824S,Y827S) and coil3 (L831S,L834S) RetGC-1 gene were each amplified by PCR from amino acid position 817-844 using pRC-CMV RetGC-1, pRC-CMV L824S,Y827S RetGC-1, and pRC-CMV L831S,L834S RetGC-1 as template, respectively. The oligonucleotide primers used in polymerase chain reaction introduced a unique 5Ј-SalI restriction site, a 3Ј-BamHI restriction site, and a stop codon at the C-terminal end. After digestion with SalI and BamHI, the PCR products were ligated into the vector pJH391 to create an in-frame fusion to the N-terminal 117 amino acids of the Lambda phage cI protein, generating plasmids pJH391-DDwt, pJH391-DDcoil2, and pJH-DDcoil3. The negative control, pKH101, expresses only the first 101 amino acids of the DNA binding domain of the lambda repressor. The positive control expresses a fusion between the repressor and the leucine zipper of GCN4 (34).
Repressor-derived Dimerization Tests-Bacterial cells expressing different lambda repressor N-terminal fusions were assessed for immunity to KH54 by cross-streak analysis. KH54 lysate (ϳ10 9 plaqueforming units), prepared as described by Miller (1992), was spread down a corner of a Petri dish, and bacterial colonies were streaked across the phage. Immunity was scored as the ability to grow after contact with the phage on the far side of the streak. A second control phage lysate, imm 21 , with different immunity was spread down the Petri dish adjacent to phage KH54. All strains of bacteria, including those that express oligomeric repressor fusions would be sensitive to imm 21 . The purpose of using two different host ranges was to ensure discrimination of cells that are immune because of repressor dimerization from cells that have lost the lambda receptor.
E. coli Cellular Fractionation and Western Blotting of Repressor Fusions-E. coli (AG1688) expressing different repressor fusions were grown in 3 ml of Luria broth overnight. Cells were harvested, and soluble proteins were extracted with 300 l of B-Per Bacterial protein extraction reagent, according to the manufacturer's protocol (Pierce, Rockford, IL). Equal amounts of total soluble protein (40 g), as determined by a Bio-Rad assay, were electrophoresed on 15% SDS-PAGE gels and transferred to nitrocellulose membranes. Lambda repressordimerization domain fusions were detected by an affinity-purified DD antibody. The antibody was raised against Met 747 -Ser 1052 catalytic domain of RetGC-1 and affinity-purified using a glutathione S-transferase dimerization domain (amino acids 817-857) RetGC-1 fusion (31). Western blots were performed as described earlier.
Molecular Dynamics Simulations-MD simulations were performed using the program ENCAD (35). The potential energy function and MD protocols have been described elsewhere (36,37). An all-atom representation was used for both the protein and solvent. Side-chain protonation states were chosen to represent neutral pH (Lys and Arg residues were positively charged and Asp and Glu were negatively charged). Because a crystal structure is not available for the dimerization domain of RetGC-1, the initial starting structures for the simulations were modeled as parallel coiled-coils spanning residues 823-857. The models were designed to keep the residues in register with respect to sliding along the helix axis while aligning nonpolar and complementarycharged groups at the interface. An antiparallel coiled-coil was also investigated, but complementarity of the residues along the interface was not as good as in the parallel arrangement, resulting in a ϳ37 kcal/mol higher (less favorable) potential energy after 1000 steps of in vacuo minimization. Amino acid mutations were made to the minimized parallel structure to yield mutant homodimers: R838C and R838S.
The proteins were then each solvated using a box extending at least 8 Å in all directions, resulting in the addition of ϳ3870 water molecules. The water density was set to the experimental value at 25°C of 0.997 g ml Ϫ1 , and periodic boundary conditions were employed to reduce edge effects. The solvent was then subjected to 1000 cycles of minimization, 2 ps of MD, and another 1000 steps of minimization. This water preparation was followed by 1000 steps of minimization of the protein in the field of the solvent and then 1000 steps of minimization of the full protein-solvent system. After the preparatory steps described above, the system was heated to 25°C. Each independent simulation was carried out for 5 ns using a 2-fs integration time step (5000 ps, or 2.5 ϫ 10 6 iterations). An 8-Å smooth, nonbonded cutoff was used, and the nonbonded list was updated every 2 cycles; this procedure has been shown to yield stable and reasonable simulations (35,36). 25,000 structures were saved for analysis (5 per ps) from each trajectory.

RESULTS
The purpose of this study is to investigate the structural basis by which a coiled-coil dimerization domain plays an essential role in regulation of RetGC-1 by Ca 2ϩ . In the first step of our analysis we establish that the catalytic domain of RetGC-1 must form a dimer to synthesize cGMP. We show this by demonstrating that the catalytic site is made up of essential functional groups contributed separately by each polypeptide chain of the dimer. We then demonstrate that the coiled-coil domain is essential to bring the subunits of the catalytic domain together. Finally, we analyze how mutations linked to adCORD alter the Ca 2ϩ sensitivity of cGMP synthesis by affecting the dimerization strength of the coiled-coil domain.
Guanylyl Cyclases Function as Dimers-Molecular modeling studies based on the structure of adenylyl cyclase suggest that synthesis of cGMP by RetGC-1 requires an intramolecular nucleophilic attack by a deprotonated 3Ј-hydroxyl of GTP on the ␣-phosphate (P ␣ ) of GTP (16,38,39). A metal ion-A (Mg A 2ϩ ) appears to be involved in the activation of the 3Ј-OH group (Fig.  1a). This is followed by stabilization of the transition state by two metal ions (Mg A 2ϩ and Mg B 2ϩ ) (39 -41). A recent report of an adenylyl cyclase crystal structure in complex with metal ions conclusively shows that two metal ions bind in the active site (20). Two highly conserved aspartic acid residues from the catalytic C1a domain of adenylyl cyclase coordinate both metal ions (16,20,42).
A structure of the RetGC-1 guanylyl cyclase catalytic domain has not yet been determined, but a model for it has been proposed (16,43). In this model the two metal ions essential for activation and conversion of GTP to cGMP in RetGC-1 are coordinated by aspartates (Asp 885 and Asp 929 ) from one subunit of the RetGC homodimer (Fig. 1a). Residues that interact with the guanine ring appear to be on the opposite subunit (16). This proposed mechanism predicts that dimerization of the cyclase catalytic domains is necessary for cGMP formation.
To confirm this dimeric model experimentally, we used two types of RetGC-1 mutations that inactivate separate functions of the catalytic site: 1) We substituted Asp 885 and Asp 929 with alanine to prevent Mg 2ϩ binding. We then expressed the mutant cyclases in human embryonic kidney cells (HEK-293) and assayed their abilities to produce cGMP from GTP. Both mutants, D885A and D929A failed to synthesize cGMP either in the presence or absence of GCAPs, consistent with the role of these amino acids in coordinating essential metal ions ( Fig. 1, b and c). These mutants also were inactive in the presence of Mn 2ϩ and Triton, which constitutively activates normal membrane GCs and has been used to quantitate general catalytic ability (44). This finding confirmed that the mutations affect the catalytic site and do not specifically interfere with GCAP-1-mediated activation. Other studies using adenylyl cyclase reached similar conclusions about the essential role of these aspartates in cyclic nucleotide synthesis (16,20,40). 2) Guanylyl cyclases can be converted to adenylyl cyclases by two specific amino acid substitutions (19,43). In RetGC-1 this can be achieved by combining E925K and C997D substitutions to produce a protein with virtually no guanylyl cyclase activity ( Fig. 1, b and c) (43). However, we know that these mutants are folded properly, because they have appreciable adenylyl cyclase activity and retain responsivity to GCAPs (43). Similar types of mutations have also been studied using a soluble guanylyl cyclase isoform (19).
The homodimer model of RetGC-1 predicts that there are two active sites in the catalytic domain ( Fig. 1b, panel i). At each site, one polypeptide chain provides the aspartates required for metal binding and the other chain provides the residues required for guanine specificity. The model predicts that a heterodimer of an aspartate mutant polypeptide with an adeninespecific polypeptide would produce one functionally active site (Fig. 1b, panels iv and v). We confirmed this prediction by co-expressing either D885A or D929A RetGC-1 together with the adenine specific form (AC) of RetGC-1. Expression of either type of mutant alone produced no GC activity (Fig. 1b, panels ii, iii and Fig. 1c). But co-expression created a heterodimer that synthesizes cGMP and retains GCAP sensitivity (Fig. 1b, panels iv, v and Fig. 1c). The co-expressed mutants (D885A/AC or D929A/AC) also had appreciable Mn 2ϩ /Triton-stimulated activity confirming that the inactivating effects of the mutations do not represent specific effects only on sensitivity to GCAPmediated activation. In summary, in the heterodimer, the D885A or D929A GC-1 chain provides the guanine ring binding site, and the partner chain (AC) provides aspartates that coordinate metal ions needed for catalysis. The result of this study confirms experimentally that dimerization of the catalytic domain is absolutely required for synthesis of cGMP by RetGC-1.
An Intact Coiled-Coil Domain Is Essential for RetGC-1 Activation by GCAPs-Because dimerization is required for RetGC-1 catalytic activity, we analyzed the region in RetGC-1 most likely to stabilize catalytic domain dimerization. A highly conserved hinge region between the kinase homology domain and the catalytic domain GC-A (NPR-A) is responsible for dimerization of the GC-A intracellular domain (Fig. 2a) (22). This region is also required for GC-A enzymatic activity (22). The role of the corresponding hinge region in RetGC-1 has not yet been established (45). Structure prediction analysis suggests that this region of RetGC-1forms an amphipathic ␣-helical coiled-coil structure (Fig. 2b) (46). Coiled-coils share a characteristic heptad repeat, (abcdefg) n , in which the first (a) and fourth (d) positions are usually hydrophobic residues. Analysis of the hinge region sequence from RetGC-1 by the COILS2 secondary structure prediction program (46), suggests that it has a very high probability to form four heptads from amino acid residues 817-844 (Fig. 2b). To test the role of these coils in RetGC-1 activity, we selectively disrupted either the second (L824S,Y827S) or third (L831S,L834S) heptad repeat by changing conserved hydrophobic "a" and "d" residues to serines. We then expressed the mutant and wild-type cyclases in HEK-293 cells. As shown in the Western blot, both the mutants and wild-type were expressed at similar levels ( Fig.  2c). Both expressed mutants (coil2 or coil3 mutant) could not be stimulated by hGCAP-1 (Fig. 2c) or by bGCAP-2 (results not shown) to synthesize cGMP. To be certain that the mutations did not cause nonspecific aggregation of the expressed proteins, we performed centrifugation of the HEK-293 membranes through a 50% sucrose cushion. We found both coil mutant forms of RetGC-1 to be associated with membranes to the same extent (90% of total RetGC-1) as wild-type RetGC-1 (results not shown) (47). Therefore, the lack of guanylyl cyclase activity in coil2 or coil3 mutant is not due to its nonspecific aggregation into insoluble complex.
To confirm the ability of the hinge region of RetGC-1 to mediate protein dimerization, we used an E. coli system that was first developed to demonstrate that the leucine zipper motif of GCN4 is responsible for dimerization of that protein.
Since then, this system has also been used to extensively study the sequence requirements for ␣-helical coils from various molecules, including the leucine zipper of yeast GCN4 and coiledcoil domains of the APC protein (product of the APC (adenomatous polyposis coli) gene) (34,48). This method involves using the N terminus of phage repressor as a reporter for the dimerization capabilities of various amino acids that are fused downstream. The repressor consists of an N-terminal domain that binds DNA and a C-terminal domain that mediates dimerization (49). Heterologous domains that are able to dimerize can functionally substitute for the C-terminal domain. A functional dimeric repressor blocks the transcription of genes that are responsible for the phage lytic cycle and confers immunity to lambda phage.
We fused the putative dimerization domain (DD) from amino acid residue 817 to 844 of RetGC-1 (DD-wt), coils 2 (DD-coils2), and coils 3 (DD-coils3) mutants to the N-terminal DNA binding domain of repressor. We then assessed the ability of the hybrid protein to render E. coli immune to lambda phage AC mutant (4 g) alone or in combination as indicated (each construct 2 g). The AC mutant is E925K,C997D RetGC-1 (43). The immunoblot with antibody to RetGC-1 kinase homology domain and catalytic domain shows equivalent levels of expression (inset). The control lane labeled pRC-CMV contained membranes from HEK-293 cells transfected with the plasmid pRC-CMV that had no insert. Equal amounts of membrane protein were assayed for 20 min. for basal-, GCAP-1-, or GCAP-2-stimulated activities (43). For GCAP-stimulated samples, 3.4 M GCAP-1 or GCAP-2 and 1 mM EGTA were added. Mn 2ϩ /Triton X-100 samples contained 1% Triton X-100 and 10 mM MnCl 2 instead of MgCl 2 .

FIG. 1. Dimerization is essential for the catalysis of GTP by
RetGC-1. a, model for mechanism of guanylyl cyclase. Adapted from adenylyl cyclase structure (18,41). The A or B after the amino acid designation represents the polypeptide chain of the dimer. b, schematic diagrams depicting possible catalytic sites in RetGC-1. i, two catalytic sites formed by homodimer of wild-type RetGC-1, D represents the aspartates at positions 885 or 929 that are essential for coordinating the metal ions, Gua represents the guanine ring binding site in RetGC-1. ii, D885A or D929A mutants that have lost both the catalytic sites. iii, adenine (Ade)-specific form of GC-1 (E925K,C997D) that has lost its ability to bind GTP. iv and v, heterodimer formed with D885A or D929A mutants and adenine-specific form of GC-1 showing a single functional catalytic site. c, measurement of basal and stimulated cyclase activities. HEK-293 cells were transiently transfected with D885A, D929A or the infection by formation of active dimeric repressor protein.
The E. coli strains expressing the repressor fusions were cross-streaked against the KH54, a phage carrying a deletion in the functional repressor that prevents lysogenization. Immunity to phage was determined visually (Fig. 2e). As a negative control, an E. coli strain expressing just the repressor DNA binding domain (pKH101) was used (50). A well-characterized repressor-leucine zipper fusion (GCN4) was used as a positive control (34).
As revealed by cross-streak analysis, bacteria expressing repressor hybrids DD-wt and GCN4 (positive control) were clearly more immune to KH54 infection than the negative control (Fig. 2e). These results indicate a successful reconstitution of DNA binding activity promoted by the normal dimerization domain (DD-wt) of RetGC-1. We also confirmed that the immunity to KH54 depends specifically on the functional repressor and is not a result of interference with other cellular pathways (e.g. lack of receptor). This was demonstrated by showing that the cells immune to KH54 are still permissive to lysis by imm 21 (Fig. 2e). imm 21 is a related phage that lyses E. coli by the same mechanism as KH54. However, it can specifically escape the control exerted by the repressor protein.
In contrast, both the DD-coils 2 and DD-coils 3 mutants behaved like the repressor DNA binding domain alone (pKH101) and were unable to confer immunity to cells. We also analyzed repressor activity quantitatively by measuring plating efficiency of phage KH54 on cells expressing the different repressor fusions. The DD-coils2 and the DD-coils3 were 1000 times less efficient than the DD-wt of RetGC-1 at restoring repressor activity (Fig. 2e). Western blot analysis using an affinity-purified antibody against the dimerization domain (DD) of RetGC-1 showed that the soluble repressor fusion proteins were present at nearly identical levels (Fig. 2d). By comparing the RetGC-1 immunoreactivity of these antibodies with antibodies against the catalytic domain we confirmed that the DD antibodies recognized the wild-type, coils2, and coils3 RetGC-1 mutants equally well (data not shown). These immunological controls rule out the possibility that the inability of DD-coils2 and DD-coils3 to dimerize repressor is merely due to low level expression.
Taken together our guanylyl cyclase assays and E. coli repressor fusion analyses of the DD-coils2 and DD-coils3 mutants show that an intact coiled-coil domain is required to produce a functional dimeric state of the catalytic domain in RetGC-1.
adCORD Mutants Are More Sensitive to Activation by GCAP-1-Several naturally occurring mutations at the end of the putative coiled-coil domain in RetGC-1 have been linked to adCORD in humans (Fig. 3a) (24 -26). This dominantly inherited disease is characterized by early degeneration of cones, decreased visual acuity, central vision defects, loss of color vision, and photophobia. At later stages, rods also degenerate,   (46). Secondary structure, as predicted by DSC, SOPM, and SIPMA algorithms, are shown below the alignment (h ϭ ␣-helix and c ϭ random coil) (68). The cyclase sequences shown include the human (h), rat (r), and bovine (b) RetGC-1, the human, rat (r) and bovine (b) RetGC-2, the olfactory guanylyl cyclase (GC-D), the atrial natriuretic peptide receptor (GC-A), and the heat-stable enterotoxin receptor (GC-C). The amino acid residues in boldface are conserved among these membrane GCs. c, GCAP-1 titrations of coils 2 mutant (L824S,Y827S) and coils 3 (L831S,L834S) and wild-type RetGC-1. HEK-293 cells were transiently transfected with 4 g of wild-type Ret GC-1, Coils 2, or Coils 3 mutant constructs. Equal amounts of membrane protein (40 g) were assayed for cyclase activity at the indicated concentrations of hGCAP-1. Inset, an immunoblot probed with an antibody against KHD and CAT regions of RetGC-1. d, expression of repressor chimeric proteins containing RetGC-1 wildtype dimerization domain (DD-WT), Coils 2 mutant (DD-Coils2), and resulting in progressive night blindness and peripheral visual field loss.
A previous study reported that a conservative E837D substitution in RetGC-1 was linked to adCORD (26). However, reappraisal of the original human mutation showed that it is instead a double mutant (E837D,R838S) in all affected family members (25). To identify the relevant biochemical phenotype of this and other RetGC-1 mutations that cause adCORD, we first expressed the adCORD double mutant (E837D,R838S, "ER-HA") in HEK-293 cells. We also expressed the wild-type GC-1 ("GC1-HA") and an adCORD triple mutant (E837D,R838C,T839M, "ERT-HA"). Each protein also contained an N-terminal hemagglutinin tag (HA). Western blots using both anti-HA and RetGC-1 antibodies showed that the mutants ERT-HA, ER-HA, and wild-type GC1-HA are expressed at similar levels (Fig. 3b, lanes 1-3). The activity of ERT-HA in the absence of GCAP and the Mn 2ϩ /Triton activity were low (2-fold reduction) compared with wild-type GC-1, whereas the intrinsic activity of ER-HA was unaltered (Fig. 3c). Both mutants were analyzed for their ability to be activated by increasing amounts of human GCAP-1 at low [Ca 2ϩ ]. The con-centrations of GCAP-1 required for half-maximal activation (K1 ⁄2 ) were reduced for both mutants (Fig. 3d, Table I). The increased apparent affinity toward RetGC-1 mutants was specific for GCAP-1. The sensitivity to stimulation by bGCAP-2 was unaltered (data not shown).  Table I. e, GCAP-1 titrations of HA-tagged ERT, ER, and wild-type GC-1 with untagged GC-1 co-expressed membranes. HEK-293 cell membranes were transfected with 2 g of HA-tagged ERT, ER, or wild-type GC-1 with 2 g of RetGC-1. Equal amounts of membrane protein were assayed for cyclase activity at the indicated concentrations of hGCAP-1. The concentration of free calcium in these titrations is 13.3 nM. The K1 ⁄2 values derived from the curve fit are shown in Table I. The mutations in the coiled-coil domain of RetGC-1 cause a dominant disease phenotype in humans. To determine if the increased apparent affinity for hGCAP-1 is also dominant, we co-expressed either the HA-tagged ER or ERT mutant with untagged GC-1 at a 1:1 ratio. As a control, we also co-expressed HA-tagged GC-1 with untagged GC-1. Comparison of immunoblots using antibodies against the HA-tag and against RetGC-1 revealed that HA-tagged and untagged cyclase proteins are present at similar levels when co-expressed (Fig. 3b).

TABLE I Biochemical parameters of GC-1 wild-type and cyclase mutants
When ERT-HA was expressed alone, its basal activity was low, but in the co-expression experiments both ERT and ER mutants had levels of basal and Mn 2ϩ /Triton stimulated activity that were the same as wild-type RetGC-1 (Fig. 3c). It appears that co-expression with the wild-type RetGC-1 enhances the folding or stability of the mutant proteins. The co-expressed mutants, like the ERT and ER mutants expressed alone, are more sensitive to GCAP-1 (Fig. 3e, Table I) confirming that this biochemical phenotype is dominant.
Ca 2ϩ Sensitivity Is Altered in adCORD Mutants-About 3% of the cGMP-gated Na ϩ /Ca 2ϩ channels in photoreceptor cell plasma membranes are open in darkness (51). Because these channels allow Ca 2ϩ to flow into the photoreceptor cell, intracellular [Ca 2ϩ ] is highest in the dark. Under those conditions, Ca 2ϩ binds to GCAPs, which inactivate RetGC to slow cGMP synthesis. When light stimulates phototransduction, the channels close and there is a net efflux of Ca 2ϩ . At lower levels of Ca 2ϩ , Ca 2ϩ dissociates from GCAPs, which then stimulate RetGC to increase production of cGMP, reopen the channels, and restore sensitivity to light (52,53). The free Ca 2ϩ concentrations in rods and cones of humans are not precisely known, but, based on studies with other species, they probably vary within the sub-micromolar range (54).
We showed previously that one of the adCORD mutations (R838C) shifts the Ca 2ϩ sensitivity of RetGC-1 (27). Therefore, we also examined this property for the ERT and ER adCORD mutants. Although the Ca 2ϩ concentration (IC 50 ) required for half-maximal inhibition by 3.4 M GCAP-1 was 580 nM for wild-type GC-1, both mutants, ERT and ER, required significantly higher concentrations of Ca 2ϩ for inactivation (Fig. 4a, Table I). The Ca 2ϩ sensitivities were the same when 3.4, 6.8, or 17 M GCAP-1 was used (Fig. 4a, results not shown, Fig. 4b and Table I). The mutant cyclases also retained nearly 15-30% of their maximal activity at Ca 2ϩ concentrations higher than 10 M (Fig. 4, a and b). This biochemical effect, like the change in apparent affinity for GCAP-1, was dominant (Fig. 4, c and d, and Table I). The co-expressed cyclases also retained nearly 15-30% of their maximal cyclase activity at Ca 2ϩ concentrations higher than 10 M (Fig. 4, c and d). These effects of altered Ca 2ϩ sensitivity were also specific for GCAP-1.
The combined effects of increased sensitivity to activation by GCAP-1 and decreased sensitivity to inhibition by Ca 2ϩ are most clearly demonstrated when the cyclase activities of WT and mutants are compared at low (100 nM) and high (1.6 M) Ca 2ϩ concentrations (Fig. 4e). Both ERT and ER have higher activity than GC-1 at micromolar levels of Ca 2ϩ (Fig. 4e). This difference in is even more pronounced when the co-expressed ERT/GC-1 and ER/GC-1 are compared with GC-1 (Fig. 4e). Taken altogether, our results show that the Ca 2ϩ feedback that controls cyclase activity through GCAP-1 is altered in adCORD patients with mutations in RetGC-1.
Arg 838 Is Critical for Proper Regulation-The only amino acid that is altered in all four RetGC-1 adCORD mutations is Arg 838 . R838C alone has been linked to adCORD (26) and produces a similar biochemical phenotype (27). Both R838C and R838S GC-1 mutants have enhanced sensitivity to hG-CAP-1 and reduced Ca 2ϩ sensitivity (data not shown, 27). To determine if neighboring amino acid residues also contribute to this phenotype, we made the E837D and T839M single mutants. E837D behaved identically to wild-type GC-1 in our assays (data not shown). The E837D substitution was initially described to cause adCORD (26). But our biochemical studies led to a reinvestigation of the original mutation, which was subsequently redefined as a double mutation: E837D,R838S (25). T839M had a higher basal activity (2-fold) and increased sensitivity (2-fold) to both hGCAP-1 and GCAP-2. However, the Ca 2ϩ sensitivity with hGCAP-1 in the co-expressed T839M mutant was only slightly altered (IC 50 , 694 nM). Collectively, these results show that altered Ca 2ϩ sensitivity is linked to adCORD and that only mutations that alter Arg 838 cause shifts in Ca 2ϩ sensitivity.

A Network of Salt Bridges in the Dimerization Domain May Be Required for Proper Regulation of RetGC-1 by GCAP-1-
Mutations that disrupt the coiled-coil domains of structural proteins, such as keratin and spectrin, have also been linked to autosomal dominant human diseases (55,56). Most substitutions in keratin that cause the skin disorder, epidermolytic hyperkeratosis, occur at the beginning or end of a coiled-coil rod domain (57). This finding is consistent with another study showing that proline mutations at the end of the rod domain in keratin are more disruptive than internal mutations (58). Similarly, it is interesting to note that the crucial arginine at residue 838 in RetGC-1 is predicted to lie exactly at the edge of the third heptad of the coiled-coil domain (Fig. 2a).
To understand the relationship between structural changes caused by arginine mutations in RetGC-1 and the observed altered regulation of cyclase activity, we modeled the sequence fully as a dimeric coiled-coil and then performed molecular dynamics simulations of wild-type and the R838C and R838S mutants in water (Fig. 5a). Molecular dynamics simulations of the WT peptide show that the residues after Arg 838 splay apart, separating the C-terminal regions of the two ␣-helices (Fig. 5, a  and b). In contrast, the ␣-helices stay together in the R838C and R838S mutants (Fig. 5, a and b). Separation of the helices in the WT peptide can be attributed to inter-molecular salt bridges (a-gЈ and a-dЈ type) between arginine and a highly conserved glutamic acid from the opposite chain (Fig. 5c). A similar arrangement of salt bridges was recently observed in the crystal structure of the coiled-coil "trigger site" of cortexillin I (59).
In WT RetGC-1, the Arg 838 salt bridges pull the coiled-coil domains together and constrain the mobility of the helices in the C-terminal end. Consequently, acidic side chains at the dimerization interface are forced to interact so that electrostatic repulsion between them drives apart the polypeptide chains downstream from the Arg salt bridges (Fig. 5, a-c). Note, however, that the chains separate to relieve charge repulsion while remaining close enough to prevent a major influx of water to the interface, thereby retaining a relatively nonpolar environment for the leucines. This phenomenon is similar to that recently observed in the atomic structure of the rod domain of the intermediate filament protein vimentin, where the C-terminal consensus motif (YRKLLEGEE) bends away from the coiled-coil axis in the region of three conserved glutamic acids (60). This conserved motif has been shown to control the filament width, and any minor alteration in that domain causes severe epidermolytic diseases in humans (61).
When arginine is replaced by cysteine or serine in RetGC-1, the constraint produced by the salt bridge interaction is lost. The C-terminal glutamates are not pulled toward the interface and instead are free to orient themselves toward the water; compare the Glu residues in the wild-type and R838S structures in Fig. 5c. The mutant helices better optimize nonpolar FIG. 4. Altered Ca 2؉ sensitivity of ERT and ER mutants to GCAP-1. a, Ca 2ϩ titration of ERT, ER, and wild-type RetGC-1 stimulated with 3.4 M GCAP-1. GC activity was measured in transiently transfected membranes expressing ERT, ER, or wild-type RetGC-1 as described in Fig.  3d. The half-maximal [Ca 2ϩ ] for inhibition (IC 50 ) was 580 nM for wild-type GC-1, 800 nM for ERT-HA, and 950 nM for ER-HA. b, Ca 2ϩ titration of ERT, ER, and wild-type RetGC-1 stimulated with 17 M GCAP-1. GC activity was measured in transiently transfected membranes expressing ERT, ER, or wild-type RetGC-1, as described in Fig. 3d. The half-maximal [Ca 2ϩ ] for inhibition (IC 50 ) was 620 nM for wild-type GC-1, 770 nM for ERT-HA, and 910 nM for ER-HA. c, GC activity measured in co-expressed membrane, as described in Fig. 3e. All reactions contained 3.4 M GCAP-1. The half-maximal [Ca 2ϩ ] for inhibition (IC 50 ) was 580 nM for wild-type GC-1(HA)/GC-1, 1000 nM for ERT-HA/GC-1, and 1190 nM for ER-HA/GC-1. d, GC activity measured in co-expressed membrane, as described in Fig. 3e. All reactions contained 17 M GCAP-1. The half-maximal [Ca 2ϩ ] for inhibition (IC 50 ) was 600 nM for wild-type GC-1(HA)/GC-1, 820 nM for ERT-HA/GC-1, and 930 nM for ER-HA/GC-1. e, GC activity measured at low calcium (124 nM) representing the maximal activity and at high calcium (1680 nM). All reactions contained 3.4 M GCAP-1. Assays were performed as described in Fig. 1. The results shown are representative of at least two independent experiments, each with duplicate data points.
packing interactions at the dimerization interface (Fig. 5b). Ser 838 satisfies its side-chain hydrogen bonding potential via interactions with the carbonyl oxygen of Leu 834 on the previous turn of the helix (Fig. 5c).
To test the involvement of Arg 838 in forming such an intermolecular salt bridge in Ret GC-1, we selectively mutated it to a similarly charged lysine, or to a negatively charged gluta-mate or aspartate. The R838E and R838D substitutions should destroy the salt bridges formed by arginine in wild-type Ret GC-1. This would allow the ␣-helices to stay together and stabilize the C-terminal end of the coiled-coil domain. As expected, high concentrations of Ca 2ϩ are required to inhibit R838E and R838D mutants activated by 3.4 M GCAP-1. Furthermore, the activity remains at 50% of its maximum even at FIG. 5. Ionic interactions are involved in regulation of GCAP-1 stimulated guanylyl cyclase activity. a, the main-chain fold of the 5-ns structures from the molecular dynamics simulations are displayed. All simulations were performed in water, although only the protein is displayed. The structures are colored from red at the N terminus to blue at the C terminus. The side chain of residue 838 is shown for each sequence. Secondary structure determinations were done using the Kabsch and Sander algorithm (70), and helical regions are illustrated by the ribbon. The wild-type coiled-coil splays after residue Arg 838 , leading to increases in the distance between the chain of Ͼ4 Å. b, space-filling versions of the structures in A are displayed. The orientations in A are preserved, and the atoms are colored according to standard schemes: carbons are gray, nitrogens are blue, and oxygens are red. The ␤-carbon of residue 838 is colored magenta in each structure. The tight packing of the coiled-coil is maintained in the mutants but disrupted at the C terminus in the wild-type protein. c, interactions around residue 838 in the 5-ns wild-type and R838S proteins. Only residues 834 -845 are displayed. Both Arg 838 residues form intermolecular salt bridges with Glu 837 and Glu 841 of the opposite strand (marked with green lines). This arrangement results in repulsive interactions between Glu 841 and Glu 845 in the nonpolar packing interface between the coils. In the Ser 838 mutant, the serine forms an intramolecular hydrogen bond with its main chain. As a result, Glu 837 and Glu 841 are not pulled into the packing interface and instead they point out and interact with solvent, as does Glu 845 . d, Ca 2ϩ titration of R838K, R838E, R838D, and wild-type RetGC-1 stimulated with 3.4 M GCAP-1. The half-maximal [Ca 2ϩ ] for inhibition (IC 50 ) was 600 nM for wild-type GC-1, 558 nM for R838K, 1743 nM for R838E, and 2033 nM for R838D. The maximal guanylyl cyclase activity as measured in picomoles of cGMP formed/minute/mg of total protein was 700 for wild-type GC-1, 110 for R838K, 190 for R838E, and 130 for R838D. The results shown are representative of at least two independent experiments, each with duplicate data points. e, Ca 2ϩ titration of R838K, R838E, R838D, and wild-type RetGC-1 stimulated with 17 M GCAP-1. The half-maximal [Ca 2ϩ ] for inhibition (IC 50 ) was 620 nM for wild-type GC-1, 590 nM for R838K, 1530 nM for R838E, and 1700 nM for R838D. The maximal guanylyl cyclase activity as measured in picomoles of cGMP formed/minute/mg of total protein was 1800 for wild-type GC-1, 330 for R838K, 350 for R838E, and 160 for R838D. The results shown are representative of at least two independent experiments, each with duplicate data points.
Ca 2ϩ concentrations as high as 27 M (Fig. 5d). Similar results were obtained using 17 M GCAP-1 (Fig. 5e). Conservative substitution of arginine to lysine did not alter the Ca 2ϩ sensitivity (Fig. 5, d and e). Western blotting confirmed that all RetGC-1 mutants were expressed at similar levels (results not shown).
These findings support the idea that intermolecular salt bridges formed by Arg or Lys at position 838 pull Glu 837 and Glu 841 from the other helix into the dimerization interface. Although these Glu residues interact favorably with Arg 838 , they are then in close proximity to other Glu residues in the C-terminal region of this domain and charge repulsion leads to disruption of the dimer (Fig. 5c). Arg 838 simultaneously acts to stabilize the coiled-coil interface and serves as a stop signal. Thus, the Arg 838 salt bridges may either keep the catalytic domains far enough apart or shift their orientation sufficiently so that RetGC-1 can be activated or inhibited by GCAP-1 at the physiologically appropriate Ca 2ϩ concentrations. More direct structural studies are necessary to verify the existence of the salt bridges and their role in controlling cyclase activity.

DISCUSSION
In this study we demonstrated that the native structure of a coiled-coil domain within RetGC-1 is critical for proper regulation by GCAP-1 and Ca 2ϩ . We first established that the minimal functional stoichiometry of the RetGC-1 catalytic domain is a dimer. We then demonstrated that formation of a functional RetGC-1 catalytic dimer requires the presence of a structurally intact coiled-coil. Finally, we showed how the GCAPmediated Ca 2ϩ sensitivity of cGMP synthesis is altered by disease-linked mutations that enhance the dimerization strength of the coiled-coil domain.
The RetGC-1 Catalytic Domain Functions Only as a Dimer-Previous evidence has suggested that guanylyl cyclases function as dimers (15)(16)(17)(18)(19)(20). Models of the guanylyl cyclase catalytic domain (16,18) predict that aspartates from one polypeptide chain (Asp 885 and Asp 929 in RetGC-1) coordinate essential metal ions that interact with the phosphates in GTP and catalyze phosphodiester formation. Along with metal ions, Arg 976 , Arg 1008 , and Lys 1048 help stabilize the pentavalent phosphate (Fig. 1a). The models also predict that specific residues (Glu 925 and Cys 997 in RetGC-1) on the other polypeptide chain in the dimer are required for specific interactions with the guanine ring.
In this report we confirmed these predictions experimentally by showing that RetGC-1 polypeptides with mutations at Asp 885 or Asp 929 cannot catalyze phosphodiester formation when expressed alone. We also showed that RetGC-1 polypeptides with mutations at Glu 925 and Cys 997 cannot synthesize cGMP. However, coexpression of these two types of mutant polypeptide chains reconstitutes a functional catalytic site (Fig.  1d). This can happen only if the two different mutant chains form a mixed dimer in which the catalytic function is provided by one polypeptide chain and the guanine ring binding function is provided by the other chain (Fig. 1c).
Dimerization Induced by the RetGC-1 Coiled-Coil Is Required for Catalysis-We also showed that an intact coiled-coil structure adjacent to the catalytic domain of RetGC-1 is essential for dimerization and catalytic activity. Substitution of critical hydrophobic amino acids within this structure reduced its ability to dimerize repressor (Fig. 2e) and, in the context of RetGC-1, abolished catalytic activity (Fig. 2c).
A Shift in Ca 2ϩ Sensitivity of GCAP-1-mediated RetGC-1 Regulation Is Linked to adCORD but Changes in Intrinsic Activity Are Not-Mutations in the human RetGC-1 gene have been linked to adCORD, an inherited retinal dystrophy. Substitutions at Arg 838 near the end of the coiled-coil closest to the catalytic domain occur in all five forms of adCORD linked to RetGC-1 that have been reported. We analyzed effects of these mutations on overall activity, sensitivity to GCAP-1 and sensitivity to Ca 2ϩ . In agreement with reports by Duda et al. (62,63), we initially found that the "ERT" mutation linked to ad-CORD causes reduced catalytic activity when the mutant polypeptide is expressed alone. But our subsequent analyses of this and other adCORD mutants showed that the effect on intrinsic activity is not sufficient to explain the disease phenotype. Most importantly, co-expression of the ERT mutant with wild-type RetGC-1, as occurs in patients with adCORD, produces an enzyme with normal intrinsic activity.
The biochemical phenotype that is linked to adCORD is altered sensitivity to Ca 2ϩ in the presence of GCAP-1. This phenotype is apparent both when the ERT mutant is expressed alone and when it is co-expressed with normal RetGC-1. Further studies with additional adCORD mutants also confirmed that it is the shifted Ca 2ϩ sensitivity, not the altered intrinsic activity, that is linked to the adCORD mutations.
Why do Arg 838 mutations alter the intrinsic activity of the mutant enzymes when they are expressed alone? We found by sucrose density gradient centrifugation (results not shown) that only a fraction of normal RetGC-1 in the cultured cell systems, widely used for expression of guanylyl and adenylyl cyclases, associates with the plasma membrane. The remainder appears to be not transported or folded properly. Therefore, any mutation introduced into the cyclase enzyme has the potential to favor or disfavor proper folding. Because of this observation, effects on the apparent intrinsic level of activity of mutants expressed alone should be discounted. In contrast, we found that the shifted Ca 2ϩ sensitivity occurs independently of folding efficiency. It occurs at both low activity levels when some of the mutants are expressed alone and at high intrinsic levels when they are co-expressed with normal RetGC-1.
Insights into the Mechanisms by Which GCAPs Regulate RetGCs-The effect of Ca 2ϩ on RetGCs is initiated entirely by Ca 2ϩ binding to the EF-hand structures in GCAPs (11,12). Evidence suggests that RetGCs and GCAPs bind to each other independently of [Ca 2ϩ ] (11,12,31). Therefore, it is not the GCAP⅐RetGC binding interaction that is affected by Ca 2ϩ . Instead Ca 2ϩ binding to the GCAP⅐RetGC complex induces structural changes that stabilizes the inactive state relative to the active state of the RetGC catalytic dimer.
Our findings show that the dimerization strength of the coiled-coil domain must be fine-tuned for proper regulation by Ca 2ϩ . The coiled-coil poises the catalytic domain in a state where its dimerization energy is regulated by physiological Ca 2ϩ concentrations. Mutations linked to adCORD enhance dimerization, i.e. they stabilize the active state. Therefore, higher Ca 2ϩ concentrations are required to overcome the increased stability of the active state. The shifted Ca 2ϩ sensitivity in the mutants is accompanied by an apparent increase in sensitivity to GCAP-1 in the absence of Ca 2ϩ (Fig. 3, d and e). It is unlikely that this represents increased affinity of RetGC-1 for GCAP-1. Instead when GCAP-1 binds to RetGC-1 in the absence of Ca 2ϩ the active state is stabilized by the mutations so that abnormally high activity is detected. We have reported previously (27) that the apparent increase in GCAP-1 sensitivity is directly predicted from an analysis of coupled equilibria that describe regulation of RetGCs by GCAPs.
All of the adCORD mutations in RetGC-1 that have been reported include a substitution at Arg 838 . Why is this residue so critical? Our molecular dynamics-based modeling of the RetGC-1 coiled-coil strongly suggests that Arg 838 forms critical salt bridges that force the helices into a configuration that produces electrostatic repulsion between the two polypeptide chains C-terminal to Arg 838 . This causes the chains to splay apart as they extend toward the catalytic domain. This normal situation should precisely determine the dimerization strength at the junction between the coiled-coil and the catalytic domain. Mutations at Arg 838 that disrupt the salt bridges allow the coiled-coil structure in the model to extend closer to the catalytic domain. Therefore, more energy should be required to pull apart the catalytic dimer. Further structural studies will be required to confirm this model experimentally.
Based on these findings we propose in Fig. 6 a simplified model for regulation of RetGCs by GCAPs and Ca 2ϩ . In a photoreceptor cell in darkness, free [Ca 2ϩ ] is as high as 700 nM (54). In this state Ca 2ϩ binds to GCAP-1, which pulls apart the polypeptide chains of the catalytic dimer, thereby reducing cGMP synthesis. When [Ca 2ϩ ] falls after light stimulation, Ca 2ϩ -free GCAP-1 activates RetGC-1 by forcing the catalytic domains closer together (Fig. 6, top panel).
Insights into the Molecular and Physiological Basis of ad-CORD-Mutations linked to adCORD enhance the stability of the active dimer. Dimerization is facilitated by increased stability of the C-terminal end of the coiled-coil (Fig. 6, lower  panel). Inhibition of the mutant cyclase activity in darkadapted photoreceptor cells requires a higher Ca 2ϩ concentration, because dissociation of the dimer is a less favorable process. In fact, more than 15-30% of maximal activity cannot be shut off in some of the adCORD mutants even with 30 M Ca 2ϩ . We expect that this would cause constitutively accelerated cGMP synthesis in darkness in human photoreceptor cells.
Altered cGMP metabolism has also been implicated in other retinal degenerative diseases, including cone dystrophy caused by a mutation in GCAP-1 (64,65) and in autosomal recessive retinitis pigmentosa caused by loss-of-function mutations in cGMP phosphodiesterase (66). We therefore propose that dominant cone-rod degeneration associated with adCORD mutations is a consequence of increased synthesis of cGMP and/or elevated steady-state [Ca 2ϩ ] levels in darkness. Normally, about 3% of the cGMP-gated Na ϩ /Ca 2ϩ channels in photoreceptor cell plasma membranes are open in darkness (51). However, the increased levels of the cGMP synthesis by RetGC-1 mutants would open more than the usual number of cGMPgated cation channels. Consequently, accelerated Na ϩ /Ca 2ϩ influx would drive the intracellular [Ca 2ϩ ] to higher than normal levels until they are sufficient to inactivate the mutant cyclase. Previous studies have suggested that elevated [Ca 2ϩ ] plays a fundamental role in the process of apoptotic rod cell death in humans and animals during inherited retinal degenerations, retinal diseases, injuries, and chemical exposure (67).