GUCY2D mutations in retinal guanylyl cyclase 1 provide biochemical reasons for dominant cone–rod dystrophy but not for stationary night blindness

Mutations in the GUCY2D gene coding for the dimeric human retinal membrane guanylyl cyclase (RetGC) isozyme RetGC1 cause various forms of blindness, ranging from rod dysfunction to rod and cone degeneration. We tested how the mutations causing recessive congenital stationary night blindness (CSNB), recessive Leber's congenital amaurosis (LCA1), and dominant cone–rod dystrophy-6 (CORD6) affected RetGC1 activity and regulation by RetGC-activating proteins (GCAPs) and retinal degeneration-3 protein (RD3). CSNB mutations R666W, R761W, and L911F, as well as LCA1 mutations R768W and G982VfsX39, disabled RetGC1 activation by human GCAP1, -2, and -3. The R666W and R761W substitutions compromised binding of GCAP1 with RetGC1 in HEK293 cells. In contrast, G982VfsX39 and L911F RetGC1 retained the ability to bind GCAP1 in cyto but failed to effectively bind RD3. R768W RetGC1 did not bind either GCAP1 or RD3. The co-expression of GUCY2D allelic combinations linked to CSNB did not restore RetGC1 activity in vitro. The CORD6 mutation R838S in the RetGC1 dimerization domain strongly dominated the Ca2+ sensitivity of cyclase regulation by GCAP1 in RetGC1 heterodimer produced by co-expression of WT and the R838S subunits. It required higher Ca2+ concentrations to decelerate GCAP-activated RetGC1 heterodimer—6-fold higher than WT and 2-fold higher than the Ser838-harboring homodimer. The heterodimer was also more resistant than homodimers to inhibition by RD3. The observed biochemical changes can explain the dominant CORD6 blindness and recessive LCA1 blindness, both of which affect rods and cones, but they cannot explain the selective loss of rod function in recessive CSNB.

Numerous mutations in GCAP1, RD3, and RetGC1 have been linked to congenital blindness (reviewed in Refs. 9, 21, and [28][29][30][31][32]. Mutations in GUCY2D gene coding for a human RetGC1 cause various forms of blindness, ranging from selectively disabling rod responses to dim light to complete blindness via degeneration of rods and cones. Substitutions of Arg 838 in the RetGC1 dimerization domain cause autosomal dominant cone-rod dystrophy type 6 (CORD6), a rapidly progressing loss of vision caused by the degeneration of functional cones and rods (33)(34)(35). The CORD6-linked mutations reduce the sensitivity of the GCAP/RetGC1 complex to deceleration by Ca 21 (34)(35)(36). Consequently, the excessive production of cGMP in the dark elevates the influx of Na 1 and Ca 21 in photoreceptors and provokes their apoptosis (37)(38).
Various nonsense and missense GUCY2D mutations in RetGC1 cause recessive blindness, Leber's congenital amaurosis type 1 (LCA1) (30, 31), where most rods and cones remain alive but are dysfunctional from birth (39). The LCA1-linked mutations in the intracellular portion of the cyclase eliminate or strongly reduce the activity of RetGC1 in vitro (39). Recessive GUCY2D mutations were also recently reported to cause congenital stationary night blindness (CSNB) (40). Unlike LCA1, CSNB patients retain daylight (cone-specific) vision but lack dim light (rod-specific) vision. Stunkel et al. (40) identified four CSNB GUCY2D alleles, three of which coded for mutations in a cytoplasmic portion of RetGC1. Notably, in every reported case the recessive CSNB allele was accompanied by a recessive GUCY2D LCA1 allele.
In the present study, we analyzed how the activity and regulatory properties of RetGC1 were affected by substitutions in the cytoplasmic portion of the enzyme, causing different types of blindness (Fig. 1). As expected, two LCA1 RetGC1 mutants completely lacked GCAP-stimulated activity. Surprisingly, however, the three tested CSNB-linked mutants were also inactive when expressed individually or co-expressed with the LCA1 variants accompanying them in CSNB patients. We also found that the CORD6 substitution, R838S, dominates Ca 21 sensitivity in the cyclase heterodimer containing a WT subunit. This can further explain the strong dominant phenotype of the mutations, leading to the abnormal Ca 21 sensitivity of the cGMP production that is typical for this disease (33)(34)(35)(36).

RetGC1 variants harboring GUCY2D CSNB, LCA1, and CORD6 mutations
The primary structure of RetGC1 includes several domains homologous to other membrane guanylyl cyclases (Fig. 1A). The "extracellular" domain, located in the intradiskal space of the photoreceptor disks, connects with the cytoplasmic portion of the cyclase via a short transmembrane region. The cytoplasmic part of the enzyme includes a protein kinase homology domain, a catalytic domain, and a short dimerization domain located between the kinase homology and the catalytic domains (2,4,32). The GUCY2D CSNB alleles encode substitutions: R666W or R761W, in the RetGC1 kinase homology domain or L911F in the catalytic domain (40). The CSNB alleles R761W and L911F were both accompanied by the LCA1 allele R768W, and the R666W allele was accompanied by the LCA1 allele G982VfsX39 (also known as c.2943delG or p.S981del1bp) (30, [39][40][41]. Unlike the recessive LCA1 and CSNB mutations, the CORD6-linked R838S substitution in RetGC1 dimerization domain causes degeneration of heterozygous photoreceptors (33)(34)(35). All these RetGC1 variants in our study were expressed in HEK293 cells, and immunoblotting of the membrane fractions isolated from the transfected cells confirmed the presence of RetGC1 polypeptides of ;105 kDa in the case of G982VfsX39 and of 115 kDa for all other variants (Fig. 1B).

GCAP1 and RD3 binding to RetGC1 mutants in cyto
We tested how the LCA1 and CSNB mutations affected RetGC1 association with its regulatory proteins. Biochemical testing of such protein complexes by conventional pulldown or co-immunoprecipitation techniques could not be applied to studying the RetGC-GCAP complex, because it disintegrates in detergents (42). Therefore, we employed a previously characterized method of co-localization in living HEK293 cells (43)(44)(45) using co-expression of mOrange-tagged RetGC1 with its GFP-tagged regulatory proteins (Figs. [2][3][4][5]. GCAP1-GFP and RD3-GFP expressed in the absence of RetGC1 are diffusely spread throughout the cytoplasm and the karyoplasm (43)(44)(45), but when co-expressed with the mOrange-RetGC1, they colocalize with the membrane cyclase, mostly in the endoplasmic reticulum (43)(44)(45). In contrast to WT ( Fig. 2A), the LCA1 R768W RetGC1 (Fig. 2B) failed to bind GCAP1 in cyto (Table  1). However, GCAP1 co-localized with the G982VfsX39 LCA1 RetGC1 in a manner very similar to the WT RetGC1 ( Fig. 2C and Table 1). CSNB mutations also differentially affected the ability of RetGC1 to bind with GCAP1 in cyto (Fig. 3). The colocalization of GCAP1 with R666W and R761W RetGC1 was visibly compromised (Fig. 3, A and B, and Table 1), whereas its co-localization with the L911F RetGC1 was not significantly different from the WT RetGC1 ( Fig. 3C and Table 1). The ability to associate in cyto with RD3 also varied between different LCA1 and CSNB mutants. It was clearly presented in the R666W RetGC1 but became completely undetectable in R768W or strongly compromised in other mutants, such as R761W, L911F, and G982VfsX39 (Figs. 4 and 5 and Table 2). The CORD6 substitution, R838S, did not compromise the in cyto RetGC1 binding with GCAP1 or RD3 (Tables 1 and 2).
Surprisingly, none of the mutants coded by CSNB-specific alleles was efficiently activated by GCAPs (Fig. 6, A-C). A low level of activity was detectable in GCAP1-stimulated R666W RetGC1, but the apparent affinity for GCAP1 was strongly reduced ([GCAP1] 1/2 = 7.25 6 0.83 mM versus 1.13 6 0.04 mM in WT; p = 0.006, Student's t test). The R761W and L911F RetGC1 were completely inactive in the presence of all three GCAP isoforms. The lack of activity in the L911F RetGC1 ( Fig.   6A) appeared to be at odds with the clearly defined GCAP1 colocalization pattern in cyto (Fig. 3C). To ensure that the GCAP1-GFP binds with the L911F RetGC1 in cyto specifically via GCAP1 rather than GFP moiety, we used V77E GCAP1 lacking the ability to bind WT RetGC1 (43) as a control. The L911F RetGC1 did not co-localize with the V77E GCAP1 in cyto (PCC = 0.33 6 0.12, n = 27) (Fig. 7).

CSNB and LCA1 mutations in RetGC1 do not complement each other
The inactive CSNB RetGC1 mutants could hypothetically create functional heterodimers in CSNB patients using the  Table 1. The other details are described under "Experimental procedures." respective GUCY2D LCA1 allele products (40). Therefore, we co-expressed in HEK293 cells the pairs of CSNB and the LCA1 RetGC1 variants, mimicking their allelic combinations in CSNB patients: R666W 1 G982VfsX39, R761W 1 R768W, and L911F 1 R768W (Fig. 8). No cyclase activity was detected in any of the three combinations, arguing against functional complementation between the LCA1 and CSNB alleles.

Ca 21 sensitivity in heterodimers containing R838S RetGC1
R838S substitution has a dominant degenerative phenotype in CORD6 patients and in transgenic animals (33,35,37,38). When the WT and R838S RetGC1 are co-expressed in a cell culture or in the retina, deceleration of the GCAP-stimulated RetGC activity requires higher than normal free Ca 21 concentrations (35)(36)(37)(38). However, it remains unclear whether the overall Ca 21 sensitivity decreases only because of the presence of the R838S RetGC1 homodimer or whether the heterodimers containing WT (Arg 838 ) and CORD6 (Ser 838 ) subunits also contribute to the change. Their co-expression produces the Arg 838 : Ser 838 heterodimers mixed with Arg 838 :Arg 838 and Ser 838 :Ser 838 homodimers. Those dimers cannot be biochemically separated to test their individual Ca 21 sensitivities, largely because the detergents required for extraction of RetGC1 from the membrane destroy cyclase regulation (42). Instead, we tested the Ca 21 sensitivities in the Arg 838 :Ser 838 heterodimers using a modification of the functional complementation method originally developed in J. Hurley's laboratory (35,51) (Fig. 9).
In the RetGC1 catalytic domain structure (52), the active site of the cyclase binds two Mg 21 GTP molecules, each held by the opposite subunits via two different binding pockets (Fig. 9A). Each subunit (subunits a and b) coordinates the guanine base (the g site) of one GTP molecule and the ribose-5-triphosphate (the r site) of the other GTP molecule (Fig. 9, A and B). Substitutions E925K and C997D disable coordination of the guanine moiety (g -), and D929A disables coordination of the ribose-5triposphate moiety (r -) (35,51). Consequently, neither (gr 1 ) nor (g 1 r -) RetGC1 homodimer can bind GTP and produce cGMP (35,51) (Fig. 9C). However, cGMP can be produced when (gr 1 ) and (g 1 r -) subunits are combined as a heterodimer. In this case, the remaining functional r site of the subu-nit a pairs with the remaining functional g site on the opposite subunit b (35,45,51) (Fig. 9D). We reasoned that the functional complementation between the Arg 838 and Ser 838 subunits ( Fig.  9, E and F) should allow for selectively testing the Ca 21 sensitivity of GCAP1-activated heterodimers in the presence of the inactive homodimers.
We verified that the Ser 838 in dimerization domains permitted the functional complementation between GCAP1-activated (gr 1 ) and (g 1 r -) RetGC1 subunits. The Ser 838 did not prevent binding of GCAP1 ( Fig. 10A and Table 1), but neither Ser 838 (gr 1 ) nor Ser 838 (g 1 r -) RetGC1 produced cGMP in the presence of Mg 21 GCAP1 when expressed separately. Only their co-expression enabled cGMP synthesis through the formation of the heterodimer (Fig. 10B), making it possible to compare the regulatory properties of different dimeric RetGC1 combinations (Fig. 9). Deletion of Ala 63 through Met 434 from the extracellular domain did not prevent the complementation (Fig. 10B). It is worth mentioning that the E925K/C997D substitutions enable binding of ATP instead of GTP via the gsite (35,45,51). We only assessed the Ca 21 sensitivity of GTP to cGMP conversion, because high concentrations of Mg 21 ATP as a substrate for adenylyl cyclase activity skew the Ca 21 /EGTA buffering accuracy in the assay.
In the Arg 838 (gr 1 ):Arg 838 (g 1 r -) RetGC1 that binds only one GTP molecule per active site (Fig. 9C), the Ca 21 sensitivity of its regulation by GCAP remained identical to the WT (Fig.  11A). In contrast, it was strongly reduced in the dimers harboring Ser 838 . The [Ca 21 ] 1/2 in Ser 838 (g 1 r 1 ):Ser 838 (g 1 r 1 ) and Ser 838 (gr 1 ):Ser 838 (g 1 r -) dimers was increased ;2.5-fold compared with the WT, also regardless of whether they bound two or just one GTP per active site (Fig. 11, A and B). The [Ca 21 ] 1/2 increased even further, to ;6.5-fold higher than the WT, in the Arg 838 :Ser 838 RetGC1 heterodimers (Fig. 11, A and B). In contrast to the single-residue substitution, R838S, a deletion of ;40-kDa fragment from the cyclase extracellular domain did not alter the Ca 21 sensitivity of the Ser 838 (g 1 r -):Arg 838 (gr 1 ) heterodimer (Fig. 11A), because the extracellular domain does not contribute to the regulation of RetGC1 by GCAP (57).

GUCY2D CSNB alleles paradoxically code for inactive RetGC1
GUCY2D LCA1-linked substitutions in RetGC1 disable rod and, even more severely, cone function, yet the vast majority of Table 1 PCC for co-localization of RetGC1 with GCAP1 variants in cyto The mOrange-RetGC1 was co-expressed in HEK293 cells with GCAP1-GFP as described under "Experimental procedures." The PCC was determined using confocal microscopy and Olympus FluoView FV10-ASW software (45,59). Statistically significant differences from wildtype (ANOVA/Bonferroni post hoc test, confidence level 99%) are highlighted in bold. Note that PCC = 1.0 for the theoretical complete co-localization of two fluorochromes, whereas PCC , 0.6 indicates lack of co-localization (67 the photoreceptors remain alive (39). The residual physiological rod, but not cone, responses can be detected in RetGC1-deficient retinas (5,39), most likely because of the presence of a rod-specific (6) RetGC2(GUCY2F) isozyme (2, 3, 7, 48), which does not form a heterodimer with RetGC1 in vivo (53). RetGC2 regulated by GCAP2 provides a smaller-scale ancillary cGMP production (7,45), helping to accelerate rod recovery from excitation (47,48,54). In contrast, cones would normally produce cGMP using almost exclusively GCAP1-regulated (and possibly GCAP3-regulated) RetGC1 (6,14,50,55). Therefore, the loss of RetGC1 activity documented previously (39,(44)(45)(46) and in the present study (Fig. 6) can explain the main biochemical rea-sons for the loss of function in GUCY2D LCA1 photoreceptors reasonably well. The role of GUCY2D mutations in selectively suppressing rod vision (40) is much more difficult to explain. Instead of finding biochemical differences in RetGC1 regulation by GCAPs and RD3, potentially explaining the unusual physiology of the affected retinas, we documented that CSNB GUCY2D alleles (40) merely code for biochemically inactive RetGC1 (Figs. 6 and 8). The low residual activity detectable in GCAP1-stimulated R666W RetGC1 (Fig. 6A) was similar to that of previously characterized LCA1 mutant R1091X (39), and its apparent affinity for GCAP1 was even 7-fold lower than in R1091X (39). , and L911F (C) were co-expressed with GCAP1-GFP and analyzed as described in Fig. 2 and under "Experimental procedures." Note the clearer colocalization pattern in the case of L911F and the less clearly defined patterns in the cases of R666W and R761W RetGC1. The PCC values are summarized in Table 1.
The two other tested CSNB mutants were completely inactive (Fig. 6). The L911F RetGC1 retains the ability to recognize and bind GCAP1 in cyto (Figs. 3C and 8 and Table 1), but even in the presence of GCAP1, it lacks catalytic activity (Fig. 6A). Activation by GCAP3, expressed exclusively in a subset of human cones (50), would have helped to explain the selective preservation of cone vision in CSNB, but none of the tested CSNB mutants was stimulated by GCAP3 (Fig. 6C).
Hypothetically, RetGC1 in CSNB photoreceptors could acquire activity through complementation with the accompanying products of LCA1 alleles (40). However, co-expression of CSNB RetGC1 with their respective LCA1 mutants failed to produce active enzyme (Fig. 8). Evidently, CSNB and LCA1 RetGC1 are unable to form heterodimers, or such heterodimers remain nonfunctional. The loss of the residual activity in R666W (Fig. 6A) after its co-expression with G982VfsX39 (Fig.  8) may reflect the formation of a nonfunctional heterodimer in which the truncated catalytic domain of the LCA1-specific subunit completely disabled the active site.
We found yet another deficiency in two CSNB/LCA1 allelic combinations, L911F/R768W and R761W/R768W (40), which would also likely make them nonfunctional in vivo; none of the individual mutants in these combinations effectively binds RD3 (Figs. 3 and 5 and Tables 1 and 2). The delivery of the cyclase in  Table 2.
the outer segment requires RD3 (21,24,25); therefore, the mutated RetGC1 would be unlikely to effectively reach the outer segment.
Based on the biochemical properties observed in this study, one would expect that the GUCYD CSNB mutations would be more likely to cause LCA1-like rod and cone blindness than to selectively preserve cone vision (40). Two hypothetical scenarios could conceivably reconcile the biochemical and clinical phenotypes: (i) some presently unknown modification(s) rescue the activities of CSNB RetGC1 variants in cones but not in rods or (ii) RetGC2 in those CSNB patients is strongly expressed in cones and produces cGMP instead of RetGC1. Neither possibility, however, can be experimentally tested, which currently leaves the field with a paradox, because the biochemistry of RetGC1 regulation directly contradicts the physiology of the GUCY2D CSNB photoreceptors.

Ca 21 sensitivity of R838S RetGC1 defines the CORD6 dominant phenotype
The gain-of-function substitutions of Arg 838 cause a dominant degenerative phenotype by hampering RetGC1 deceleration at the Ca 21 concentrations typical for dark-adapted photoreceptors (34)(35)(36)(37). The Ca 21 sensitivity of the RetGC1-GCAP complex is reduced because the affinity of RetGC1 for Mg 21 GCAP increases as a result of the mutation (34)(35)(36). We find that the guanylyl cyclase activity in heterodimers  Table 2.
comprised of the Arg 838 -and Ser 838 -containing subunits becomes even less sensitive to the inhibition by Ca 21 than in the Ser 838 :Ser 838 homodimer (Fig. 11). The very high Arg 838 : Ser 838 heterodimer resistance to deceleration by Ca 21 , even stronger than in the Ser 838 :Ser 838 homodimer (Fig. 11A), is consistent with the additional increase of the heterodimer's affinity for Mg 21 GCAP1 (Fig. 10C).
Mutational analyses of RetGC1 indicate that its dimerization domain and the kinase homology domain are most critical for GCAP binding (44,45). Previous studies of R838S RetGC1 (35)(36)(37) suggested that its Ser 838 :Ser 838 coiled-coil dimerization domain is more optimal than WT for the regulatory binding of GCAP1. We find in our study that pairing WT and CORD6specific dimerization domains is even more conducive to GCAP1 binding.
To summarize, the heterodimers formed in photoreceptors that harbor heterozygous CORD6 GUCY2D allele would likely stimulate Ca 21 influx in the dark even more than Ser 838 :Ser 838 homodimers. It would therefore seem unreasonable to expect that decreasing the fraction of the Ser 838 :Ser 838 homodimers by increasing the level of WT GUCY2D allele expression instead of reducing the expression of the CORD6 allele could alleviate photoreceptor degeneration.
Possible role of RD3 in GUCY2D CORD6 pathology RD3 suppresses RetGC-GCAP complex activity by displacing GCAP from the cyclase (22,23,37). RetGC1 binds RD3 via interface that is nonidentical to the GCAP-binding site(s) (58)(59)(60). The C-terminal portion of RetGC1 is critical for binding RD3 (24) but not GCAPs (24,45). Consistent with that, the G982VfsX39 and L911F RetGC1 can bind GCAP1 but not RD3 in cyto (Figs. 2C and 3C, Tables 1 and 2). Conversely, the cyclase dimerization domain is more critical for GCAP binding than for RD3 binding (44,45). On the other hand, some of the disease-linked substitutions in the cyclase kinase homology domain (R761W and R768W) can compromise both GCAP1 and RD3 binding (Figs. 4 and 5), thus suggesting that the GCAPbinding and RD3-binding interfaces on the cyclase could be partially overlapped or affect each other in the quaternary structure of the complex. The increased affinity for Mg 21 GCAP1 evidently causes the Arg 838 :Ser 838 heterodimer to require higher RD3 concentrations for its inhibition than the WT RetGC1 homodimer and even higher than the Ser 838 : Ser 838 homodimer (Fig. 10C). In vivo studies strongly indicate that one of the essential RD3 functions in photoreceptors prevents photoreceptor degeneration by suppressing aberrant guanylyl cyclase activation by GCAPs (27), most likely in the inner  The PCC values for co-localization of the mOrange-RetGC1 and RD3-GFP coexpressed in HEK293 cells (45,59,60). Statistically significant (ANOVA/Bonferroni post hoc test) differences from wildtype are highlighted in bold.

Materials
Unless specified otherwise, nucleotides were purchased from Millipore/Sigma, chemicals (ultrapure or molecular biology grade) were from Millipore/Sigma or Fisher Scientific, restriction endonucleases were from New England Biolabs, Phusion Flash polymerase was from Thermo Scientific, and oligonucleotide primers were from Integrated DNA Technologies.

RetGC1 mutagenesis, expression, and activity assays
Human RetGC1 (GUCY2D) cDNA inserted in pRCCMV vector (Invitrogen) was modified to introduce new restriction sites for subsequent cloning without changing the encoded protein sequence. LCA1-, CSNB-, and CORD6-specific substitutions were then introduced using "splicing by overlap extension" method (62). DNA fragments amplified in a Phusion Flash DNA polymerase mixture were purified using a ZymoResearch DNA Cleanup kit, digested at the ends with the appropriate restriction endonucleases, and ligated in the RetGC1 cDNA part of the plasmid. The resultant plasmids were isolated using a Promega Wizard protocol and verified by sequencing on both strands. Where indicated, the mOrange tag (Clontech)  Table 1.

Retinal guanylyl cyclase in congenital blindness
cDNA was inserted, substituting in-frame a ;26-kDa portion of the RetGC1 "extracellular" domain, downstream from the leader peptide and upstream from the transmemebrane domain (44)(45). RetGC1 variants were expressed in HEK293 cells transfected using calcium-phosphate precipitation method (44,58), and the membrane fractions containing the recombinant RetGC1 were isolated as previously described (44). The guanylyl cyclase activity was assayed as previously described in detail (36) with the modifications described in Ref. 59. In brief, the assay mixture (25 ml) containing HEK293 membranes, 30 mM MOPS-KOH, pH 7.2, 60 mM KCl, 4 mM NaCl, 1 mM DTT, 2 mM EGTA, or 2 mM Ca 21 /EGTA buffers, Mg 21 as indicated in experiments, 0.3 mM ATP, 4 mM cGMP, 1 mM GTP, and 1 mCi of [a-32 P]GTP (PerkinElmer), 100 mM zaprinast and dipyridamole was incubated for 30 min at 30°C, and the reaction was stopped by heat-inactivation for 2 min at 95°. The [ 32 P] cGMP product was separated by TLC using fluorescently backed polyethyleneimine cellulose plates (Merck) developed in 0.2 M LiCl. The cGMP spots visualized under UV light were cut out from the plate, and the [ 32 P]cGMP was eluted with 0.5 ml of 2 M LiCl in 20-ml scintillation vials. The radioactivity was counted by liquid scintillation in 10 ml of UniverSol (MP Biochemicals). Data fitting was performed using Synergy Kaleidagraph 4 software.

Immunoblotting
HEK293 cell fractions containing the recombinant RetGC1 were subjected to electrophoresis in 7% SDS-PAGE and transferred to Immobilon-P membrane (Millipore Sigma) at 60 V for 20 h. The membranes were blocked using SuperBlock blend (Fisher Scientific), and probed by rabbit polyclonal antibody raised against the recombinant C-terminal fragment, Met 748 -Ser 1103 , of a human RetGC1 (RRID:AB_2877058), and the chemiluminescence image developed using a Pierce SuperSignal Femto kit was acquired using a Fotodyne Luminous FX instrument.

RD3 expression and purification
The recombinant human RD3 was expressed from pET11d (Novagen/Calbiochem) vector in BL21(DE3) CodonPlus Escherichia coli strain (Agilent Technologies), extracted from the inclusion bodies, and purified as previously described (22,59) with the modifications described in Ref. 60. The purity of the preparations was verified by SDS-PAAG electrophoresis, Coomassie Blue staining, and densitometry. The protein solution was mixed with glycerol to final 35% (v/v), then aliquoted, frozen in liquid N 2 , and stored in 270°C. The aliquots were thawed only once, immediately before use in the RetGC assay.

GCAP expression and purification
Human GCAP1 (E6S), GCAP2, and GCAP3 cDNAs were inserted into the NcoI/BamHI sites of the pET11d vector (originated from Novagen/Calbiochem). The N-myristoylated GCAPs for the in vitro assays were expressed in BLR(DE3) E. coli strain harboring pBB131 plasmid coding for a yeast N-myristoyl transferase and urea-extracted from the inclusion bodies. GCAP1 was purified by calcium precipitation, butyl-Sepharose chromatography, and Sephacryl S-100 chromatography using previously published procedure (58) modified as described in Ref. 64. The purity of GCAPs preparations verified by SDS gel electrophoresis was 90%. The human GCAP2 was purified as previously described (65) with the following modifications. The cells were grown in 0.5 liter of standard LB medium containing 50 mg/ml kanamycin and 100 mg/ml ampicillin until they reached A 600 0.6-0.7. Myristic acid was added from a concentrated ethanol solution to the suspension of bacterial cells to the final concentration of 100 mg/ml, 30 min prior to the induction with 0.5 mM isopropyl-b-D-thiogalactopyranoside. 3 h after the induction, the bacterial pellet was harvested by centrifugation at 8,000 3 g for 10 min 4°C, and frozen in 270°C. The pellet was thawed; resuspended in 25 ml of 10 mM Tris-HCl, pH 7.5, containing 2 mM EDTA and 14 mM 2-mercaptoethanol; and sonicated on ice for 2 min. The inclusion bodies from the disrupted cells were collected by centrifugation at 20,000 3 g for 20 min, 4°C. The sonication/centrifugation step was repeated two more times. GCAP2 was extracted from the pellet by homogenization in 10 ml of 10 mM Tris-HCl, pH 7.5, containing 2 mM EGTA, 14 mM 2-mercaptoethanol, 2 mM MgCl 2 , and 8 M urea for 30 min at 4°C. After centrifugation at 20,000 3 g for 20 min at 4°C, the supernatant was dialyzed at 4°C, first for 3-4 h against 2.0 liters of 10 mM Tris-HCl buffer, pH 7.5, containing 0.5 mM EGTA, 1 mM MgCl 2 , and 7 mM 2-mercaptoethanol, and then overnight against 2 liters of 10 mM Tris-HCl, pH 7.5 buffer containing 0.1 mM EGTA, 1 mM MgCl 2 , and 7 mM 2-mercaptoethanol. After the dialysis, the insoluble material was removed by centrifugation at 20,000 3 g for 20 min at 4°C, and Tris-HCl and CaCl 2 were added to the supernatant fraction to 50 mM and 10 mM, respectively. After 20 min at room temperature, the precipitated material was removed by centrifugation at 20,000 3 g for 20 min at 4°C, and the supernatant was collected, concentrated to 5 ml using Amicon Ultra-15 (10,000 molecular weight cutoff) centrifugal filter and then centrifuged for 10 min at 200,000 3 g at 4°C in a Beckman Optima ultracentrifuge and loaded on a Sephacryl S-100 column (2.6 3 60 cm) equilibrated with 20 mM Tris-HCl, pH 7.5, containing 100 mM NaCl and 0.3 mM CaCl 2 . After the main peak containing GCAP2 was collected, dithiothreitol was added to 4 mM and EDTA was added to 2 mM to remove Ca 21 bound to GCAP2. The excess of EDTA was then removed by 3-4 cycles of 20-fold concentration/ dilution in 10 mM Tris-HCl, pH 7.5, containing 30 mM EDTA using an Amicon Ultra-15 (10,000 molecular weight cutoff) centrifugal filter at 4°C. Human GCAP3 was purified using similar procedure except that 1.0 3 30-cm Superdex-200 Figure 9. A paradigm for measuring guanylyl cyclase activity in RetGC1 heterodimer Arg 838 :Ser 838 . A, three-dimensional structure of the active site in RetGC1 (52). The catalytic domains from two subunits, a (gray) and b (blue), coordinate two GTP molecules (green and orange), each guanine base, via Glu 925 and Cys 997 (the g site) and the Mg 21 ribose-5´-triphosphate moiety, via Asp 929 (the r site). B, binding of each GTP molecule requires the g and r sites located on the opposite subunits. C, inactivation of a single g or r site (gor r -) prohibits the resultant homodimers from coordinating both GTP molecules (35,45,51). D, complementation between the (gr 1 ) and (g 1 r -) subunits allows their heterodimer (gr 1 ):(g 1 r -) to coordinate one GTP molecule in the active site (the subunit coordinating guanine base is shown with dark shading) (35,45,51). E and F, to selectively measure the activity of heterodimer Arg 838 :Ser 838 , the (g 1 r -) subunit harboring WT Arg 838 (E) or CORD6 Ser 838 (F) in its dimerization domain is co-expressed with the complementing (gr 1 ) subunit that harbors, respectively, Ser 838 (E) or Arg 838 (F). Only the Arg 838 :Ser 838 heterodimers in both cases can bind and convert GTP to cGMP, whereas the homodimers Arg 838 : Arg 838 and Ser 838 :Ser 838 are inactive.
column equilibrated with 20 mM Tris-HCl, pH 7.5, and 100 mM NaCl was used for size-exclusion chromatography instead of Sephacryl S-100.

Ca 21 /EGTA buffers
The EGTA/CaCl 2 /MgCl 2 mixtures maintaining required free Ca 21 concentrations at 1 mM free Mg 21 were prepared using Tsien and Pozzan method (66) and verified by fluorescent indicator dyes as previously described in detail (63).

Co-transfection and confocal imaging
The DNA transfection followed a Promega FuGENE protocol by the manufacturer as described in Refs. 45, 59, and 60. The HEK293 cells were transfected after reaching 30-50% confluence in a LabTeck 4-well cover glass chamber with the mixtures of bGCAP-GFP or hRD3-GFP coding plasmids with mOrange-RetGC1 coding plasmid, at ;1/100 molar ratio (58-60), using 3 ml of Promega FuGENE reagent per 1 mg of DNA. Confocal images were taken after 24-30 h of incubation in 5% CO 2 at 37°C, using an Olympus FV1000 Spectral instrument with the respective 543-and 488-nm excitation for the red and the green fluorochromes in sequential mode. The images were processed, and the PCC values in whole-cell images were determined using Olympus FluoView FV10-ASW software as previously described (43)(44)(45)(58)(59)(60). No changes to the original images were made except for minor g correction applied to whole image for more clear presentation in print. Quantitative analyses were performed using the original images without corrections.

Statistics
Statistical significance of the differences was tested by ANOVA/Bonferroni post hoc at 99% confidence level or by unpaired/unequal variance Student's t test using a Synergy Kaleidagraph 4 software.

Three-dimensional molecular visualization
The image of the RetGC1 active site was created using PyMOL molecular graphics system (version 2.0, Schrödinger, LLC) utilizing coordinates of the model structure reported by Liu et al.

Data availability
The data referred to in this article are contained within the article. Unprocessed data can be available from the corresponding author (adizhoor@salus.edu) upon reasonable request. Funding and additional information-This work was supported by NEI, National Institutes of Health Grant EY11522 and by the Pennsylvania Department of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.