Two clusters of surface-exposed amino acid residues enable high-affinity binding of retinal degeneration-3 (RD3) protein to retinal guanylyl cyclase

Retinal degeneration-3 (RD3) protein protects photoreceptors from degeneration by preventing retinal guanylyl cyclase (RetGC) activation via calcium-sensing guanylyl cyclase–activating proteins (GCAP), and RD3 truncation causes severe congenital blindness in humans and other animals. The three-dimensional structure of RD3 has recently been established, but the molecular mechanisms of its inhibitory binding to RetGC remain unclear. Here, we report the results of probing 133 surface-exposed residues in RD3 by single substitutions and deletions to identify side chains that are critical for the inhibitory binding of RD3 to RetGC. We tested the effects of these substitutions and deletions in vitro by reconstituting purified RD3 variants with GCAP1-activated human RetGC1. Although the vast majority of the surface-exposed residues tolerated substitutions without loss of RD3's inhibitory activity, substitutions in two distinct narrow clusters located on the opposite sides of the molecule effectively suppressed RD3 binding to the cyclase. The first surface-exposed cluster included residues adjacent to Leu63 in the loop connecting helices 1 and 2. The second cluster surrounded Arg101 on a surface of helix 3. Single substitutions in those two clusters drastically, i.e. up to 245-fold, reduced the IC50 for the cyclase inhibition. Inactivation of the two binding sites completely disabled binding of RD3 to RetGC1 in living HEK293 cells. In contrast, deletion of 49 C-terminal residues did not affect the apparent affinity of RD3 for RetGC. Our findings identify the functional interface on RD3 required for its inhibitory binding to RetGC, a process essential for protecting photoreceptors from degeneration.

Retinal degeneration-3 (RD3) protein protects photoreceptors from degeneration by preventing retinal guanylyl cyclase (RetGC) activation via calcium-sensing guanylyl cyclase-activating proteins (GCAP), and RD3 truncation causes severe congenital blindness in humans and other animals. The threedimensional structure of RD3 has recently been established, but the molecular mechanisms of its inhibitory binding to RetGC remain unclear. Here, we report the results of probing 133 surface-exposed residues in RD3 by single substitutions and deletions to identify side chains that are critical for the inhibitory binding of RD3 to RetGC. We tested the effects of these substitutions and deletions in vitro by reconstituting purified RD3 variants with GCAP1-activated human RetGC1. Although the vast majority of the surface-exposed residues tolerated substitutions without loss of RD3's inhibitory activity, substitutions in two distinct narrow clusters located on the opposite sides of the molecule effectively suppressed RD3 binding to the cyclase. The first surface-exposed cluster included residues adjacent to Leu 63 in the loop connecting helices 1 and 2. The second cluster surrounded Arg 101 on a surface of helix 3. Single substitutions in those two clusters drastically, i.e. up to 245-fold, reduced the IC 50 for the cyclase inhibition. Inactivation of the two binding sites completely disabled binding of RD3 to RetGC1 in living HEK293 cells. In contrast, deletion of 49 C-terminal residues did not affect the apparent affinity of RD3 for RetGC. Our findings identify the functional interface on RD3 required for its inhibitory binding to RetGC, a process essential for protecting photoreceptors from degeneration.
The RD3 (retinal degeneration-3), a 23-kDa 195-residue protein (1,2), plays two essential roles in photoreceptors. Firstly, RD3 enhances accumulation of retinal membrane guanylyl cyclase (RetGC) in rod and cone outer segments (3)(4)(5)(6) and thus enables regulation of cGMP production in the outer segment required for phototransduction. The two RetGC isozymes RetGC1 (GUCY2D) and, to a lesser extent, RetGC2 (GUCY2F) (7)(8)(9)(10), allow vertebrate photoreceptors to maintain inward ion current via cGMP-gated channels in the outer segments. After illumination, when light-stimulated phosphodiesterase activity forces cGMP-gated channels to close, RetGC in the outer segment becomes accelerated by Ca 21 /Mg 21 sensor proteins (GCAPs) (11)(12)(13)(14)(15)(16)(17) to replenish cGMP and thus expedite photoreceptor recovery from excitation and adaption to light (reviewed in Refs. [18][19][20][21]. Secondly, RD3, which is predominantly located in the inner segments of photoreceptors (5,6,22), strongly inhibits the cyclase by suppressing its basal and GCAP-stimulated activity (23,24). The inhibitory binding of RD3 competing with GCAPs for the cyclase is required for preventing photoreceptor degeneration (6), and the lack of such binding causes the congenital retinal blindness because of rapid loss of rods and cones (6,22,24). Deletions of RD3 have been linked to a severe recessive degenerative blindness, Leber's congenital amaurosis 12 (LCA12) in human patients (1,25) (Fig. 1) and rd3 retinal degeneration in mice (1). A frameshift in RD3 has also been linked to inherited canine retinal dysplasia (26). Recent studies argue that the rapid death of photoreceptors lacking RD3 primarily results not from reduction of RetGC content in the outer segment (24), but from the lack of the RD3 inhibitory activity counteracting GCAP-dependent activation of the remaining cyclase, likely in the inner segment (6,24). The molecular mechanisms of RD3/RetGC interaction, despite its importance for photoreceptor function and survival, remain unclear because of insufficient structural and functional data. The molecular structure of RetGC remains largely unknown and mutational analysis of the cyclase presents a major challenge because of the larger size of the enzyme and the complexity of its regulation. Another major challenge presents high propensity of RD3 to precipitate at concentrations required for structural analyses (24,27). However, the three-dimensional structure of RD3 core, an elongated bundle of four a-helices ( Fig. 1), was recently established using a soluble variant of RD3 that retained attenuated affinity for the cyclase (27). The preliminary testing of several fragments in RD3 primary structure that contained surface-exposed and buried in the core structure residues indicated that the cyclase-binding interface on RD3 includes the central portion of the helical bundle ( Fig. 1) and that the part of the bundle forming the functional interface with the cyclase involves helices 3 and 4 (24,27). Nonetheless, the identities of the residues on the surface of RD3 that are essential for its functional contact with the cyclase remained unclear, in part because the full-size RD3 remains unsuitable for structural analyses. In the present study, we functionally probed by mutations 133 residues circumventing the entire surface of the RD3 molecule, both of the central core and the predicted unstructured parts of the protein. We identified two narrow surfaceexposed clusters responsible for the inhibitory binding of RD3 to the cyclase. One of the clusters was located on the surface of a-helix 3. The other cluster critical for the inhibitory binding was found on the opposite side of the molecule, in the loop connecting a-helices 1 and 2. This study presents the first high-resolution functional map of RD3 as a guanylyl cyclase regulating protein.

Results
Mutations in two clusters of surface-exposed residues suppress RD3 ability to inhibit RetGC1 To functionally locate the surface-exposed residues critically important for the inhibitory binding of RD3 to guanylyl cyclase, we first scanned the entire surface of the molecule ( Fig. 1) with single-residue substitutions that altered the properties of the residues but did not fully revert their main properties, e.g. the hydrophilic residues were substituted with other hydrophilic residues of the opposite charge (such as Lys or Arg to Glu and vice versa) or of different sizes of the side chain, but not with strongly hydrophobic residues, to minimize the potential of affecting the overall fold of the protein. Conversely, hydrophobic side chains were changed to hydrophobic residues of different sizes. In some cases, they were replaced by more hydrophilic residues, but only if the original hydrophobic side chains were already exposed on the surface in the RD3 core three-dimensional structure (Fig. 1). We avoided making substitutions in side chains directed inside the structure (shown in black in Fig.  1) to minimize possible effects on the overall fold and/or chang-ing distances between helices in the a-helical bundle of the core by creating steric hindrances.
All expressed RD3 mutants purified from Escherichia coli were tested in a standard primary screening assay by being reconstituted with HEK293 membranes containing recombinant human guanylyl cyclase RetGC1 (6,23,24). The reason we chose to use a recombinant RetGC instead of retinal preparations in this study was 2-fold. First, the native RetGC activity in photoreceptor membranes lacking the endogenous RD3 (which could otherwise skew the effects of the added recombinant RD3), such as isolated from rd3/rd3 mice (1), is so drastically decreased (6,24) that the accurate measurements of its additional inhibition by the exogenous recombinant RD3 becomes less reliable because of larger margins of error. Second, RetGC from normal photoreceptors already containing the endogenous RD3 cannot be separated from the RD3 by being extracted in the presence of detergent without a complete loss of the cyclase regulation after the extraction (28). In contrast, the recombinant RetGC1 expressed in HEK293 cells lacking the endogenous RD3 retains its regulation by GCAP and RD3, as demonstrated earlier in multiple studies (6,7,23,24,27,29).
The recombinant human RetGC1 in the standard assay was pre-activated by 1.5 mM Mg 21 GCAP1 in the presence of 2 mM EGTA and saturating 10 mM Mg 21 (30,31). For the initial comparison of the inhibitory activity to that of the WT, the suppression of RetGC1/GCAP complex by WT RD3 and its mutants was assayed at a standard 100 nM RD3, the concentration at which the WT RD3 suppressed the activity of the cyclase Figure 1. Residues in RD3 targeted for mutational analysis. The top diagram presents the schematics of RD3 polypeptide structure comprised of four a-helices (cylinders a1 through a 4) connected to unstructured regions (straight line). Recessive mutations in RD3 linked to LCA12 include nonsense mutants R38*, E46*, Y60*, F100*, a two-base deletion causing a frameshift at the residue 46 and premature termination downstream (E46Afs83*), and an aberrant splicing resulting in deletion of the residues 38-99 (1,25,42). Positions of the residues replaced by the LCA12 nonsense mutations are also indicated in the threedimensional structure of the RD3 a-helical bundle (27) presented as a ribbon diagram (middle). Filled-spheres diagram of the structure (bottom) depicts the location of the surface-exposed (colors matching that of the corresponding helices in the primary structure diagram) versus that of the buried residues (black) in the three-dimensional structure. The surface-exposed residues were subjected to the mutational analysis as further described in this study.
by ;89% (Fig. 2 and Table 1). As a threshold for identifying mutants with a strong reduction in the inhibitory activity we selected a 2-fold reduction in the extent of the cyclase inhibition by WT RD3, i.e. the remaining cyclase activity 22%.
The results in Fig. 2 and Table 1 demonstrated that although substitutions of vast majority of the surface residues in RD3 had little effect on the ability of RD3 to suppress the cyclase activity, two narrow clusters, each containing only several surface-exposed side chains, presented a stark exception. The first cluster included previously untested part of the molecule occupying central portion of the loop that connects helices a1 and a2 in the central helical bundle structure. This part of the loop presented several hydrophobic side chains exposed on the surface: Tyr 60 , Trp 62 , and Leu 63 . Even substitutions with other hydrophobic residues, such as Y60A or W62A, were most detrimental for the ability of RD3 to inhibit the cyclase. The second cluster critical for the cyclase inhibition was located on the surface of the helix a3 (Fig. 2) and included fully exposed highly hydrophilic side chains Arg 99 , Arg 101 , and Gln 102 . This helix was shown previously to be an essential part of RD3 threedimensional structure that folds the interface for binding to the cyclase (24,27). Ile 97 , closely adjacent to the surface-exposed residues in this cluster, also had a profound negative effect on the cyclase inhibition by RD3 when substituted with a large Tyr residue (Fig. 2). Nonetheless, the Ile 97 is only partially exposed on the surface and remains mostly buried inside the helical bundle, hence the substitution of this side chain could also create a structural hindrance, altering the proper alignment between the a3 and a1 helices, similarly to previously tested residues buried in the core structure (24,27), and thus affect the neighboring surface interface indirectly. Therefore, we cannot draw a conclusion that Ile 97 is directly involved in contacting the cyclase. The same relates to Pro 90 , a residue that could potentially affect not only interaction with the cyclase, but also the proper fold of the interface. In contrast, Arg 99 , Arg 101 , and Gln 102 do not directly push on the backbone of other helices. Hence, the reduction of the RD3 inhibitory activity after mutating those residues indicated that they most likely constitute the surface part of the interface directly interacting with the cyclase.
The two clusters of surface-exposed residues define the high affinity of RD3 for the cyclase The reduction in the inhibitory capacity of RD3 caused by mutations in the two surface-exposed clusters demonstrated in the standard assay ( Fig. 2) could have two different explanations, either being a result of RD3 losing its binding affinity for the cyclase or a result of RD3 binding to RetGC without inhibiting the cyclase activity. Therefore, we tested the apparent affinities of various mutants for the cyclase by comparing the dose dependence of the cyclase inhibition by individual mutants to that of the WT RD3 (Fig. 3). We identified three fairly distinct groups of the mutants demonstrating different apparent affinities for the RetGC based on their dose dependence curves. Consistent with the results of the initial mutational scanning of the surface of the molecule in a standard assay at a fixed 100 nM RD3 (Fig. 2), the vast majority of the RD3 mutants (from the pool shown in plain text in Table 1) did not demonstrate reduction of the apparent affinities for the cyclase (Fig. 3, A and B). For a large group of such 62 tested mutants, the dose dependence of the cyclase inhibition was not shifted to the Figure 2. Point mutations in RD3 reveal two surface-exposed clusters affecting inhibition of RetGC1/GCAP1 complex. Purified 100 nM human RD3 containing mutations (see Table 1) in the corresponding positions of the residues numbered according to their sequence in the RD3 primary structure, beginning with Met 1 , was added to the assays containing membranes isolated from HEK293 cell-expressing human recombinant RetGC1 reconstituted with 1.5 mM GCAP1 in the presence of 2 mM EGTA and 10 mM Mg 21 . The cyclase activity (mean 6 S.D., three measurements) remaining in the presence of RD3 is plotted as percentage of the activity measured in the absence of RD3. The thick dashed line corresponds to the average cyclase activity remaining in the standard assay after inhibition by 100 nM WT RD3. The mutations at the surface-exposed side chains causing the strongest loss of inhibition ( 22% residual activity criterion threshold, thin dashed line) are shown in red, except for the Ile 97 side chain only partially exposed on the surface and mostly located inside the core between helices 1 and 3. The top diagram presents the location of the respective clusters in the primary structure (red); the large double-sided arrow above the diagram indicates the span of the primary structure forming the interface-containing helical core of the bundle (24,27).
higher concentrations compared with the WT (note that because of the multitude of the mutants in that group only the combined family of their dose dependence curves are plotted in Fig. 3B in gray, without specifying the symbols for individual mutants). Even drastic changes in some of the well-conserved parts of the molecule did not produce a discernible reduction of the apparent binding affinity. Mutations in conserved N-terminal Trp residues or a large, 49-residue-long, deletion of the C-terminal fragment ( Fig. 3A) (as well as deletion of an N-terminal Leu 3 -Ser 5 fragment, not shown), did not alter the apparent affinity of the RD3 inhibitory binding to the cyclase. The respective IC 50 values for the WT, W6L/W9L, and 146ter RD3 were 3.6 6 0.52 nM (mean 6 S.D., n = 19), 4.5 6 1.2 (n = 3), and 3.4 6 0.8 nM (n = 3) (not statistically significant by Student's t test: p = 0.28 and 0.8, respectively). In contrast, mutations of the residues located in clusters 1 and 2 and their vicinity prominently shifted the dose dependence curves to higher concentrations ( Fig. 3, C and D). The loss of the apparent affinity was strongest for substitutions of the residues occupying the central positions in each cluster (Fig. 3C). Mutations in Tyr 60 , Trp 62 , and Leu 63 (cluster 1) or Arg 101 and Gln 102 (cluster 2) caused a drastic, up to 240-fold, reduction in the apparent affinity of RD3 for the cyclase compared with the WT (Fig. 3C and Table  2, also marked red in Fig. 4) (ANOVA p , 0.0001, F = 789; Bonferroni post hoc test at CL = 99% and a = 0.01 yields p values between ,0.0001 and 0.0235). Along with that, mutations neighboring the centers of the clusters, Asp 59 , Ser 61 , and Ser 74 adjacent to Tyr 60 -Leu 63 in cluster 1 or His 89 , Gly 94 , Arg 99 , Ala 105 adjacent to Arg 101 -Gln 102 in cluster 2 (Fig. 3D, also marked orange in Fig. 4), reduced the apparent affinity ;3to 10-fold (ANOVA p , 0.0001, F = 107; the Bonferroni post hoc test p values for differences from WT between ,0.0001 and 0.016). For the majority of the surface mutations within the clusters 1 and 2, the increase in IC 50 was also highly significant when compared with WT using Student's t test (Table 2). Notably, the critical residue in cluster 2, Arg 101 , was highly sensitive to both the reversal of a charge (R101E) and the replacement with a small hydrophobic side chain (R101A) ( Table 1 and 2), suggesting that this side chain is essential for the direct contact with the cyclase. In contrast, Ala 105 , a less critical residue proximal to the Arg 101 , was sensitive to replacement with a larger Table 1 RetGC1 Activity in the presence of RD3 mutants. RetGC1 pre-activated by 1.5 mM Mg 21 GCAP1 was assayed as described in "Experimental procedures" in the presence of 100 nM RD3, and cyclase activity was normalized as percentage of its activity measured in the absence of RD3. The RD3 mutants for which the remaining cyclase activity was at least 2-fold higher than in the WT are highlighted in bold; all highlighted mutants were significantly different from the WT (P < 0.0001, Student's t test)  side chain, Tyr, but not a more hydrophilic residue, Glu, which may indicate that this side chain is more important for proper presenting the neighboring Arg 101 than for making a direct contact with the cyclase itself. Taken together, the highly differential reduction of the apparent affinity of RD3 for the cyclase after mutating the large number of surface-exposed residues indicated that the functional interface for the RD3 inhibitory binding to the target enzyme requires two clusters located on the opposite sides of the central a-helical bundle surface (Fig. 4).
Mutations in both clusters had cumulative effect in reducing the RD3 inhibitory binding (Fig. 5). Compared with the single mutations (Fig. 3), second mutations in each cluster exacerbated the loss of the inhibitory activity. R101A/Q102L (cluster 2) shifted the IC 50 ;1200-fold, to 4.5 mM from 3.6 nM in WT, stronger than either mutation alone (Figs. 3C and 5 and Table  2). In case of W62A/L63R (cluster 1) the effect was even more pronounced. Only rudimentary inhibition was observed by the double mutant at the RD3 concentrations exceeding 1 mM. Simultaneous inactivation of the two clusters of the interface (W62A/L63R 1 R101A/Q102L) completely eliminated the ability of the resultant Int(2) RD3 mutant to inhibit RetGC1 in the conditions of the assay (Fig. 5).

Inactivation of the binding interface on RD3 prevents colocalization of RD3 with RetGC
Based on the previous studies (24,27), even RD3 mutants with reduced apparent affinity for the RetGC1/GCAP complex inhibition in vitro (IC 50 increase up to ;1 mM) can co-localize with RetGC1 in HEK293 cells, evidently, because in cyto the absence of competing GCAPs (6, 23) helps RD3 bind the cyclase. However, more severe reduction in the apparent affinity of RD3 for RetGC1 in the in vitro inhibition assay also suppresses RD3 co-localization with the cyclase in cyto (24,27).  Table 2.

Regulation of retinal guanylyl cyclase by RD3 protein
To further verify that the two clusters detected in the RetGC inhibition assay constitute the RD3 interface for binding to the cyclase, we co-expressed fluorescently tagged RD3-GFP variants with mOrange-tagged RetGC in HEK293 cells. Tagging RD3 by the fluorescent protein at the C terminus and RetGC1 with a tag replacing a portion of its N-terminal "extracellular" domain does not affect the function of either protein in vitro (24,(32)(33)(34)(35). When expressed separately in HEK293 cells, RD3-GFP demonstrates a uniformly diffused distribution throughout the cytoplasm and the nuclei (except for nucleoli and vacuoles (24,27,35); mOrange-RetGC1, a transmembrane protein, displays clearly defined membrane localization primarily in the endoplasmic reticulum (32)(33)(34)(35). Both proteins, however, co-localize in the membranes (Fig. 6A) when co-expressed using the conditions of the in cyto assay described in detail previously (24,27,35). Consistent with the functional assays presented in Figs. 2 and 3, the distribution of the GFP tag attached to WT RD3 or RD3 lacking 47 C-terminal residues (D 148-195 ) across the cells coincided with the mOrange tag of the cyclase. The respective Pearson's correlation coefficients (PCC, mean 6 S.D.), 0.90 6 0.061 and 0.89 6 0.051, confirmed strong colocalization with RetGC1 of the two RD3 variants, without a significant difference between them (Fig. 6, A and B) (note that PCC = 1.0 is the theoretical value for complete coincidence for both tags whereas PCC 0.5 indicates the lack of co-localization) (36). In stark contrast to the D 148-195 RD3, the Int(-) RD3 mutant (Fig. 6C), in which the interface for the cyclase on the surface of the molecule was inactivated by point mutations in the two clusters (W62A/L63R and R101A/Q102L), failed to co-localize with the cyclase (PCC = 0.32 6 0.18; p , 0.0001 compared with the WT by t test).
Notably, it required more than a single mutation in the binding interface to disrupt RD3 association with the RetGC1 in cyto (Fig. 7). Neither W62A nor L63R alone, despite their marked increase of the IC 50 in the inhibition assay in vitro ( Fig.  3 and Table 2), were able to abolish co-localization RD3 with the cyclase in living cells, albeit in the case of W62A co-localization was less clearly defined than in WT and in the case of L63R it became heavily compromised, evidenced by diffusion of RD3 nonanchored to the cyclase through the cytoplasm and karyoplasm (the respective PCC = 0.83 6 0.074 and 0.65 6 0.13, p , 0.001) (Fig. 7, A and B). Inactivation of the cluster 1 with the double mutation, W62A/L63R (Fig. 7C), disrupted co-localization (PCC = 0.48 6 0.13, p , 0.0001) nearly as efficiently as in Int(2) RD3 (Fig. 6C) and consistently with a severe loss of inhibition by the double mutant in the cyclase inhibition assay (Fig. 5).

Discussion
RD3, through binding to the cyclase, evidently promotes RetGC trafficking from the inner to the outer segment (2-5) to maintain the proper level of cGMP synthesis there, yet, the molecular mechanism of this process is poorly understood. The second role of RD3 in photoreceptor physiology, to inhibit RetGC and its activation by GCAPs in the inner segment, is essential for the survival of photoreceptors (6,22,24); therefore, RD3 deficiency associates with LCA12, a congenital form of human blindness (1,25), and rapid degeneration of photoreceptors in rd3 mouse strain (1). Notably, some mutations in RetGC1 and GCAPs associated with dominant retinopathies weaken inhibition by RD3 of the RetGC1/GCAP1 complexes containing the cyclase or GCAP1 coded by the disease-coding alleles (29,37). So the high-affinity inhibitory binding of the RD3 to the cyclase documented in previous studies (6,23,24) is the necessary step in both preventing the photoreceptor death and enabling its normal function.
Although much has been learned about the biochemical properties and regulatory pathways related to RetGC function in photoreceptors (17)(18)(19)(20)(21), much less is known about the tertiary and quaternary molecular structures of the complexes between RetGC and its regulatory proteins. The structures of the two main protein regulators for the cyclase, GCAPs and RD3, have been largely established (27,(38)(39)(40). The functional interface for the cyclase at a single-residue resolution was previously identified for GCAPs using mutational testing (33,41). In the present study, we have now identified the surface-exposed residues on RD3 that are essential for its inhibitory binding to RetGC.
Our findings indicate that the interface for the high-affinity inhibitory binding of RD3 to the cyclase evidently involves two narrow clusters, one in the loop connecting helices 1 and 2 and the other on the surface of helix 3 (Figs. 3 and 4). The location of the second cluster on helix 3 was not very surprising, the Table 2 The change in IC 50 of the RetGC1 inhibition by RD3 caused by mutations in clusters 1 and 2. The cyclase activity was assayed in the presence of 1.5 mM GCAP1, 2 mM EGTA, 10 mM Mg 21 and increasing concentrations of RD3 as described in Fig. 3, C and D, and "Experimental procedures." The IC 50 values were obtained from the fits assuming sigmoidal function, A % = 100% essential role of helices 2, 3, and 4 in forming the central part of the interface was demonstrated by previous low-resolution mutational testing (24,27). Yet, the location of cluster 1, which was missed during the previous lower-resolution search for the interface, was somewhat surprising. The main reasons why this part of the molecule was not previously tested for being a potential surface interface based on the RD3 primary structure (1,14,24) is that the critical hydrophobic residues of the cluster 1, Tyr 60 , Trp 62 , and Leu 63 , would more likely appear as a part of the molecule's hydrophobic core. However, after the threedimensional structure of RD3 core was established recently (27), it became evident that these side chains in loop 1/2 are, despite their hydrophobic nature, largely exposed on the surface (Figs. 1 and 4), which prompted us to include these residues in the present mutational analysis. This cluster is reminiscent of the cyclase-binding interface on the surface of GCAP1, which includes, in addition to hydrophilic residues, several surfaceexposed strongly hydrophobic side chains (33). However, dissimilarly to the cyclase interface on GCAP1, which forms a single patch on one side of the molecule, the two clusters on RD3 responsible for the high-affinity cyclase binding locate on two opposite sides of the molecule (Fig. 4).
Binding to the cyclase becomes completely abolished by inactivation of both clusters of the interface in Int(-) RD3 (Figs. 5 and 6). Very low-affinity inhibitory binding to the cyclase can still be detected after inactivation of either cluster (Fig. 5), but inactivation of cluster 1 has especially strong effect and also interferes with RD3 binding to the effector enzyme in cyto nearly as effectively as the Int(-) (Fig. 7), suggesting that RD3 affinity for the target imparted by the cluster 1 is the highest between the two clusters.
Notably, the LCA12-linked mutations in human RD3 (Fig. 1,  top panel) have been shown to create termination codons upstream of or within cluster 1 or cluster 2 (R38ter, E46ter, Y60ter, F100ter), as well as produce a truncating frameshift, p. E46Afs*83, or deletion in the central part of the molecule (1,25,42). Hence, all LCA12-causing mutations reported to date eliminate the cyclase-binding interface of RD3 and truncate the protein. Even truncation eliminating second cluster of the , and Int(-) RD3 (black filled triangle) as described in Fig. 3 legend. Note that only rudimentary inhibition was detectable in case of W62A/L63R (cluster 1). An empirical estimate for its IC 50 from the observed dose dependence exceeded 10 mM, but could not be determined more precisely because the inhibition failed to approach 50% even at highest concentrations achievable in the assay. The Int(-) RD3 demonstrated a complete lack of inhibition in the conditions of the assay. The IC 50 value for the R101A/Q102L (cluster 2) was 4.5 6 1.6 mM compared with 3.6 6 0.5 nM in WT (see also Table 2). Figure 4. Location of the surface-exposed residues critical for the RD3 binding to the cyclase. A, the ribbon cartoon of the RD3 tertiary structure (27) depicting the two surface-exposed clusters that contain residues required for the high-affinity inhibitory binding of RD3 to the cyclase (with their side chains shown as sticks). The residues critically involved in the inhibitory binding are marked in red and those with a moderate involvement are marked in orange; the residues not critical for the high-affinity binding based on testing by point mutations in Fig. 2 are marked in blue. The internal residues not exposed on the surface are marked in black. B, space-filled RD3 three-dimensional structure depicts the location of the two respective clusters on the surface of the molecule.
Regulation of retinal guanylyl cyclase by RD3 protein interface, F100ter (1), completely disables inhibitory binding of RD3 to RetGC in vitro (23), despite the preservation of the cluster 1 in the remaining fragment of the polypeptide. Based on our present mutational analysis, it is also possible that LCA12 could result from yet to be found single-residue substitutions, but the occurrence of such missense mutations in RD3 would likely be even rarer than deletions found to date. The vast majority of surface-exposed side chains in RD3 tolerate replacement without affecting RD3 function (Figs. 2; 3, A and B; and   4). Therefore, to cause the disease, single-residue substitutions in RD3 would have to occur directly within the narrow clusters of the surface interface or deteriorate the fold of a-helical core of the interface (27).
Without knowing the detailed tertiary and quaternary structure of the cyclase we could at this point only speculate how the two clusters possibly make a connection with the enzyme. RetGC is a homodimer in which the catalytic domains of both subunits create the active site converting GTP to cGMP (43, 44). The inhibitory binding of RD3 to the cyclase displays a negative cooperativity, indicating possible involvement of at least two RD3 molecules in the complex (23,24). Because RD3 is very prone to self-aggregation (23,27), the negative cooperativity was previously regarded as a possible artifact produced by the aggregation of RD3 upon increase of its concentrations (24). However, a more recent study showed that a highly soluble form of RD3, not prone to aggregation, still demonstrated negative cooperativity in cyclase inhibition assay (27). Therefore, it is plausible that one molecule of RD3 can interact with two subunits of RetGC at the same time to create the inhibitory complex, in which it prevents the cyclase from being activated by GCAP. However, an alternative possibility-that each RD3 molecule interacts with two distant domains on a single RetGC1 subunit-cannot be excluded and at his point appears equally likely. The negatively cooperative inhibition in either case may indicate, for example, that binding of one RD3 per two cyclase subunits reduces the affinity for binding the second RD3 by the complex.
The mechanism of RetGC inhibition by RD3 and its competition with GCAP requires further functional and structural studies. GCAPs and RD3 compete against each other for the Regulation of retinal guanylyl cyclase by RD3 protein cyclase (23,24), but based on the mutational analysis of the RetGC1 (34,35), this competition likely involves nonidentical sites on the cyclase. It is also important to emphasize that RD3 exerts its inhibitory binding on the cyclase directly, rather than, for example, through binding to and reducing the concentration of GCAPs, because (i) the RD3 inhibits the cyclase at concentrations that are ;500-fold lower than concentrations of GCAP in the assays (Table 1 and Figs. 2 and 3) (23,24), and (ii) RD3 inhibits basal RetGC activity even in photoreceptor membranes completely devoid of GCAPs (23,24). Whether or not the binding sites for GCAP and RD3 overlap in the tertiary and/or quaternary structure of the cyclase or affect each other allosterically by changing the overall shape of the cyclase in the complex remains to be further investigated.
Another aspect of the cyclase regulation by RD3 that requires further in-depth study is the apparent role of RD3 in delivering the cyclase to the outer segment (2)(3)(4). Although only a limited part of the RD3 structure is critical for making its high-affinity binding interface with RetGC (Fig. 4), other region(s) of the RD3 molecule could conceivably contribute to the intracellular trafficking of cyclase. For example, a small portion of the N-terminal or a large portion of the C-terminal unstructured regions can be removed from RD3 without any apparent loss of its high-affinity inhibitory binding to RetGC (Table 1 and Figs. 3  and 6), which appears to be somewhat at odds with the parts of these regions in RD3 primary structure being evolutionarily conserved (1,2). Therefore, it is rather tempting to speculate that some of the conserved N-and/or C-proximal portions of the RD3 molecule are specifically required for the cyclase trafficking, such as by playing the role of a recognition signal or participating in the interactions with the trafficking machinery, after the high-affinity binding of RD3 to the cyclase has occurred via the interface identified in our present study. Experimental evaluation of such a possibility may also help shed light on additional biological processes potentially involving RD3, because some recent reports suggest that RD3 (or its homologs) can also interact with other trafficked or nontrafficked proteins (45, 46), and not only in the retina but also perhaps in different organs of vertebrate and invertebrate species.

Materials
Unless specified otherwise, nucleotides were purchased from Millipore/Sigma, chemicals (ultrapure or molecular biology grade) from Millipore/Sigma or Fisher Scientific, and restriction endonucleases from New England Biolabs.

RD3 mutagenesis, expression, and purification
Mutations were introduced in a human RD3 cDNA utilizing a conventional splicing by overlap extension technique (47), as described previously (6,24,27), using the cDNA's 59-end coding and the 39-end coding primers, 59-AAGGACCATGGCTC-TCATCTCATGGCTTCGGTGGAACGAGAAATT and 59-GGATCCTCAGTCGGCTTTGGGCGCCCGGAAT, respectively, and primers that contained the designated nucleotide substitutions and at the same time generated overlapping regions for the subsequent splicing. The primary fragments were amplified in a PCR reaction using a human cDNA as a template and utilizing a Thermo Scientific Phusion Flash High-Fidelity DNA Polymerase. The fragments were then spliced in a second round of PCR using only the 59-end and the 39-end primers shown above. The spliced DNA constructs were purified using a Zymo Research DNA Clean-up Kit, digested with NcoI-HF and BamHI-HF endonucleases, inserted into the NcoI/ BamHI sites of the pET11d vector (Novagen/Calbiochem), and amplified in a Zymo Research JM109 Mix & Go! E. coli strain, and minipreps isolated using a Promega Wizard Kit from 5-ml overnight cultures grown in the presence of ampicillin were verified by the automated DNA sequencing of the entire RD3 inserts in the resultant plasmids. To enable the use of the NcoI site for ligating the 59-end of the constructs into the expression vector, the N-terminal Ser 2 was replaced by Ala in all bacterially expressed variants of RD3, a modification that does not interfere with the ability of RD3 to bind and inhibit RetGC (6,23). For expression of a human RD3 tagged at the C terminus with SuperGlo eGFP (Clontech) in HEK293 cells, the cDNA was amplified using a forward primer, 59-TTTTTTAAGCTT GGGCCAGGGGCTATGTCTCTCATCTCA and a reverse primer, 59-AGCGGCAATTGTGAGTCGG CTTTGGGCGC-CCGGAAT, digested with HindIII-HF and MfeI-HF endonucleases, and subcloned into the HindIII/EcoRI sites of a pQBIFn3 vector (Clontech). The original N-terminal residue, Ser 2 , remained preserved in these constructs. In some cases, instead of PCR amplification, the mutations were introduced into fragments of the RD3 cDNA using chemical synthesis (service of Integrated DNA Technologies, Iowa City, IA) and ligated into the vectors described above using the appropriate restriction sites. The recombinant human RD3 was expressed from pET11d (Novagen/Calbiochem) vector in BL21(DE3) CodonPlus E. coli strain (Agilent Technologies), extracted from inclusion bodies, and purified as described previously (23,27), with minor modifications as follows. Inoculated with a frozen stock, 5-ml bacterial cultures in a standard Luria broth (Thermo Fisher Scientific) were grown overnight in incubator shaker at 37°C, 200 rpm, and then in 100-ml culture to reach A 600 0.6-0.7. The protein expression was induced by 1 mM isopropyl-b-D-thiogalactopyranoside for 2 h. The bacterial cell pellet was harvested by centrifugation at 7000 rpm for 10 min in a Sorvall Fiberlite F14-6 3 250y rotor at 4°C and frozen in 270°C. The pellet was thawed, resuspended in 20 ml of 10 mM Tris-HCl, pH 7.5, 1 mM EDTA/14 mM 2-mercaptoethanol buffer solution (TEM) and sonicated on ice for 2 min using 2-s ultrasonic pulses. The inclusion bodies from the disrupted cells were collected by centrifugation at 22,000 3 g for 10 min, 4°C, in a Sorvall Fiberlite F21-8 3 50y rotor. The pellet was resuspended in TEM, sonicated for 1 min on ice and centrifuged using the same regime. White pellet containing inclusion bodies was dissolved in 2.5 ml TEM buffer containing 2 mM EDTA and 8 M urea (BioXtra, Millipore/Sigma-Aldrich) by gentle stirring for 1 h at 4°C and centrifuged at 22,000 3 g for 10 min, 4°C. The supernatant was dialyzed in a Pierce/Thermo Scientific 3-ml 10,000 MWCO Slide-A-Lyzer Cassette G2 at 4°C against 1 liter of TEM buffer containing 0.1 mM EDTA and 7 mM 2-mercaptoethanol for 3 h without stirring, then overnight against 1 liter of TEM containing 0.1 mM EDTA and 7 mM 2-mercaptoethanol with gentle stirring and finally for 4 h against fresh 1 liter of the same buffer. RD3 was precipitated from the dialyzed protein fraction by adding NaCl to 250 mM. Precipitated protein was collected by centrifugation at 22,000 3 g, 4°C, dissolved in TEM containing 8 M urea on ice and then dialyzed as described above, except that pH of TEM buffer was 8.3. The dialyzed protein solution was centrifuged at 5000 3 g for 10 min at 4°C in a standard 2-ml Eppendorf tube, the supernatant was collected, and the protein concentration was measured by absorbance at 280 nm in 50 mM Tris-HCl, pH 7.5, containing 7 M guanidine chloride, assuming 0.1 g/liter absorbance 1.41 (calculated using a ProtParam software available online from the ExPASy server RRID:SCR_018087). The purity of the preparations was determined by SDS-PAGE in 15% gel, Coomassie Blue staining, and densitometry. For storage, the protein solutions were mixed with glycerol to final 35% v/v, aliquoted, frozen in liquid N 2 , and stored in 270°C. Each aliquot was thawed only once, immediately before use in the RetGC assay. The final RD3 concentrations in the stock solutions typically varied between 5 and 50 mM.

GCAP1 expression and purification
Myristoylated bovine GCAP1 for in vitro assays was expressed from pET11d vector in a BLR(DE3) E. coli strain (both originated from Novagen/Calbiochem) harboring a pBB131 plasmid coding for a yeast N-myristoyl transferase and purified by calcium precipitation, butyl-Sepharose, and Sephacryl S-100 chromatography using previously published procedure (31)(32)(33). The purity of GCAP1 preparations estimated by SDS gel electrophoresis was 90%.

RetGC1 expression and activity assay
Human recombinant RetGC1 was expressed from a modified Invitrogen pRCCMV vector in HEK293 cells transfected using calcium-phosphate precipitation method, and the membrane fraction containing the expressed cyclase was purified as described previously (34). The guanylyl cyclase activity was assayed as described in detail previously (30,48), with modification described (27). 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, 10 mM Mg 21 , 0.3 mM ATP, 4 mM cGMP, 1 mM GTP, and 1 mCi of [a-32 P]GTP (Perkin Elmer), 100 mM zaprinast and dipyridamole was incubated at 30°C for 30 min and the reaction was stopped by heat inactivation at 95°C for 2 min. The resultant [ 32 P]cGMP product was separated by TLC using fluorescently backed polyethyleneimine cellulose plates (Merck) developed in 0.2 M LiCl, cut from the plate and eluted with 0.5 ml 2 M LiCl in 20-ml scintillation vials, and the radioactivity was counted by liquid scintillation in 10 ml UniverSol mixture (MP Biochemicals). Data fitting was performed using Synergy KaleidaGraph 4 software.

Co-transfection and confocal imaging
HEK293 cells were transfected in LabTeck 4-well cover glass chamber with 1 mg of mOrangeRetGC1 DNA per well using 3 ml/mg DNA of the Promega FuGENE reagent following the protocol recommended by the manufacturer at ;1/100 molar ratio of RD3-GFP coding plasmid versus mOrangeRetGC1 coding plasmid as described (27,35). Confocal images were taken after 24-32 h of incubation in 5% CO 2 , 37°C, utilizing an Olympus FV1000 Spectral instrument using the respective 543 nm and 488 nm excitation for the red and the green fluorochromes in sequential mode and processed using Olympus FluoView FV10-ASW software as described previously (27,(32)(33)(34)(35). No changes to the original images were made except for minor g correction applied to whole image for more clear presentation in print. Quantitative analysis was performed using only original images, without g corrections. PCC for testing co-localization of RD3-GFP with mOrange-tagged RetGC1 in whole-cell images was calculated using Olympus FluoView FV10-ASW software as previously described (32,35).

Three-dimensional molecular visualization
The protein images were created using the RD3 coordinates recently established by NMR spectroscopy (27) (PDB ID 6DRF) using PyMOL Molecular Graphics System, Version 2.0, Schrödinger, LLC.

Statistics
Statistical significance of the differences was tested by ANOVA and unpaired/unequal variance t test using Synergy KaleidaGraph 4 software.

Data availability
All data referred to in this manuscript are contained within the manuscript. Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.