Retinal guanylyl cyclase activation by calcium sensor proteins mediates photoreceptor degeneration in an rd3 mouse model of congenital human blindness

Deficiency of RD3 (retinal degeneration 3) protein causes recessive blindness and photoreceptor degeneration in humans and in the rd3 mouse strain, but the disease mechanism is unclear. Here, we present evidence that RD3 protects photoreceptors from degeneration by competing with guanylyl cyclase-activating proteins (GCAPs), which are calcium sensor proteins for retinal membrane guanylyl cyclase (RetGC). RetGC activity in rd3/rd3 retinas was drastically reduced but stimulated by the endogenous GCAPs at low Ca2+ concentrations. RetGC activity completely failed to accelerate in rd3/rd3GCAPs−/− hybrid photoreceptors, whose photoresponses remained drastically suppressed compared with the WT. However, ∼70% of the hybrid rd3/rd3GCAPs−/− photoreceptors survived past 6 months, in stark contrast to <5% in the nonhybrid rd3/rd3 retinas. GFP-tagged human RD3 inhibited GCAP-dependent activation of RetGC in vitro similarly to the untagged RD3. When transgenically expressed in rd3/rd3 mouse retinas under control of the rhodopsin promoter, the RD3GFP construct increased RetGC levels to near normal levels, restored dark-adapted photoresponses, and rescued rods from degeneration. The fluorescence of RD3GFP in rd3/rd3RD3GFP+ retinas was mostly restricted to the rod photoreceptor inner segments, whereas GCAP1 immunofluorescence was concentrated predominantly in the outer segment. However, RD3GFP became distributed to the outer segments when bred into a GCAPs−/− genetic background. These results support the hypothesis that an essential biological function of RD3 is competition with GCAPs that inhibits premature cyclase activation in the inner segment. Our findings also indicate that the fast rate of degeneration in RD3-deficient photoreceptors results from the lack of this inhibition.

We reasoned that if the rd3/rd3 photoreceptors rapidly degenerate because of the low cyclase activity, then a complete removal of GCAPs would additionally suppress activation of the remaining cyclase, further lower cGMP production and thus exacerbate degeneration. Conversely, if the rd3/rd3 photoreceptor degeneration mainly resulted from aberrant, nonre-stricted by RD3, activation of the cyclase by GCAPs, then deletion of GCAPs could protect rd3/rd3 photoreceptors. Indeed, deletion of GCAPs further reduced cGMP production in rd3/ rd3GCAPs Ϫ/Ϫ hybrids. Their RetGC content did not increase (Fig. 1A, inset, lane 3), and the RetGC activity not only remained much lower than in WT in the presence of Ca 2ϩ (0.01 Ϯ 0.003 nmol/min/retina, 3, p Ͻ 0.0001), but also, in contrast to rd3rd3, now failed to accelerate even at low Ca 2ϩ concentrations because of the lack of the endogenous GCAPs (0.011 Ϯ 0.003 nmol/min/retina, 3, p Ͻ 0.0001). Consistent with the deficiency of cGMP production, dark-adapted electroretinography (ERG) response to a saturating bright flash in both rd3/rd3 mice and the rd3/rd3GCAPs Ϫ/Ϫ hybrids remained drastically suppressed compared with the WT (Fig. 1B). At 1 month of age, the amplitude (mean Ϯ S.E., n) of the ERG a-wave-the negative corneal voltage deflection produced by closure of cGMP-gated channels in photoreceptors-was strongly reduced compared with the WT (p Ͻ 0.0001), from 404 Ϯ 40 V, 12, in WT to 117 Ϯ 16 V, 10, in rd3/rd3 and 47 Ϯ 7 V, 9, in rd3/rd3GCAPs Ϫ/Ϫ .
Nonetheless, despite further reduction in their RetGC activity, degeneration of photoreceptors in rd3/rd3GCAPs Ϫ/Ϫ hybrids was dramatically decelerated when compared with rd3/ rd3. In rd3/rd3, photoreceptor nuclei count rapidly declined within several months (Fig. 2, A and B). After 6 months, only ϳ3% of nuclei were identifiable in what remained of the photoreceptor layer (marked yellow), and the outer segments stratum was completely absent. In contrast, the rd3/rd3GCAPs Ϫ/Ϫ hybrids at that age displayed much lesser (Յ30%, p Ͻ 0.0001) reduction in photoreceptor count and nearly normal stratification of the retina: ϳ70% of the photoreceptors were still present and retained identifiable photoreceptor outer segments layer (Fig. 2, C and D). Therefore the evidence argued that RD3 ability to block activation of the cyclase by GCAPs was essential for preventing fast degeneration of photoreceptors, but the cellular compartment where RD3 acted in that capacity remained to be clarified.

RD3GFP substitutes for the endogenous RD3 in rd3/rd3 rods
Localization of RD3 in different studies presented a rather controversial issue (19,20,23), evidently because of low content of RD3 and potential nonspecific cross-reactivity of different antibodies. A potential masking of the epitope in the outer segment could also influence immunochemical detection of RD3. Therefore, to establish the RD3 localization, we not only applied immunofluorescence detection as described further in the next section but also used fluorescently tagged RD3 (Fig. 3). For that, we first verified that RD3GFP expressed in Escherichia coli could substitute the nontagged WT RD3 in vitro. Similarly to the WT RD3 (21,22), the RD3GFP shifted the dose dependence of RetGC activation by GCAP to higher concentrations range and reduced the catalytic activity of the enzyme (Fig. 3A). Like a nontagged RD3, 1 M RD3GFP completely inhibited activation of the recombinant RetGC1 by GCAP1 with only a minor increase in the EC 50 (ϳ6 nM versus 2 nM) (Fig. 3B). We therefore reasoned that RD3GFP should be able to functionally substitute for the endogenous RD3 when expressed in mouse photoreceptors. The blots were probed with anti-RetGC1 (left) or anti-GAPDH antibodies (right). B, dark-adapted (scotopic) ERG recorded in WT (black trace), rd3/rd3 (red trace), and rd3/rd3GCAPs Ϫ/Ϫ (blue trace) at 1 month of age in response to a saturating 1-ms 505-nm flash of 1 ϫ 10 6 photons/rod, delivered at 100 ms. The traces are averaged (mean Ϯ S.E., error bars) from 12, 10, and 9 mice of the respective genotypes. The differences of the a-wave amplitudes (ANOVA, F ϭ 41; p Ͻ 0.0001) when compared with the WT (404 Ϯ 40 V) using Bonferroni post hoc test were significant (p Ͻ 0.0001) both in rd3/rd3 (117 Ϯ 16 V) and in rd3/rd3GCAPs Ϫ/Ϫ (47 Ϯ 7 V).

Mechanism of rd3 retinal degeneration
The transgenic expression of the RD3GFP was guided to mouse retina rods using a 4.2-kb rod opsin promoter (24) (Fig.  3C). Following PCR detection of the transgene in the progeny of the RD3GFP-positive founders (Fig. 3C, right panel), they were subsequently bred into a homozygous rd3/rd3 background, and the expression of RD3 was verified by Western immunoblotting of the retinal extracts (Fig. 3D). In contrast to the WT, the endogenous RD3 was undetectable in rd3/rd3 and rd3/ rd3RD3GFP ϩ littermates, but the latter had clearly detectable GFP-tagged RD3. RD3GFP expression strongly, to the levels typical for the WT, elevated the expression of both RetGC1 and RetGC2 in rd3/rd3RD3GFP ϩ , barely detectable in their nontransgenic rd3/rd3 littermates.
The RD3GFP expressed in mouse rods fully substituted for the endogenous RD3 in biochemical and physiological tests. RetGC activity in the rd3/rd3RD3GFP ϩ retinas became nearly normal. The maximal activity (mean Ϯ S.D., n) of the cyclase in the presence of EGTA in rd3/rd3RD3GFP ϩ retinas at 1 month of age increased 14-fold when compared with the nontransgenic rd3/rd3 siblings, to 0.62 Ϯ 0.03 nmol cGMP/min/retina (p Ͻ 0.0001, ANOVA/Bonferroni post hoc), not significantly lower than the WT (0.77 Ϯ 0.18 nmol cGMP/min/retina; p ϭ 0.2) (Fig. 3E, inset). Ca 2ϩ sensitivity of the cyclase regulation in rd3rd3RD3GFP ϩ retinas was also similar to the WT retinas (Fig. 3E). The RD3GFP rescued rod photoresponse in living animals as well (Fig. 3F). The ERG a-wave to saturating flash in dark-adapted WT mice presents mostly rod response with only a small contribution from cones (Ͼ97% of all photoreceptors in a mouse retinas are rods) (25). The maximal amplitude (mean Ϯ S.E., n) of the dark-adapted ERG a-wave in rd3/rd3 RD3GFP ϩ mice at 1 month of age (416 Ϯ 48 V, 14) became dramatically larger than in rd3/rd3 siblings (95 Ϯ 9 V, 24; p Ͻ 0.0001) and matched the WT (387 Ϯ 40 V, n ϭ 9; p ϭ 1). The rd3/rd3 rods expressing RD3GFP also became resistant to degeneration (Fig. 4). Dark-adapted ERG a-wave (Fig. 4A) in rd3/rd3 RD3GFP ϩ mice after 4 months of age remained similar to the WT (436 Ϯ 36 V, 14, and 400 Ϯ 34 V, 12, respectively; p ϭ 1), both drastically different from the rudimentary a-wave of the rd3/rd3 mice in that age group (20 Ϯ 9 V, 13; p Ͻ 0.0001). The preservation of the photoreceptor function, however, was less efficient in cones, consistently with the RD3GFP being expressed under primarily a rod-specific promoter. In light-adapted mice, conditions when rods are saturated and only cone ERG remains detectable, the cone a-wave amplitude ( Fig. 4B) was rudimentary in rd3/rd3 (3.8 Ϯ 1 V) as compared with WT (18.7 Ϯ 1.4 V; p Ͻ 0.0001). It slightly improved in rd3/rd3RD3GFP ϩ (10.4 Ϯ 1.9 V) compared with the rd3rd3 (p ϭ 0.012) but remained significantly lower than in WT (p ϭ 0.0013). The morphology of the rd3/rd3RD3GFP ϩ retinas after 4 months of age showed normal thickness of photoreceptor layer, and in a stark contrast to their rd3/rd3 littermates, their photoreceptor nuclei count did not significantly differ from the  WT (Fig. 4,C and D). No evidence of retinal degeneration was seen in three rd3/rd3 RD3GFP ϩ mice even past 6 months (not shown).

RD3 localization in the retina
Polyclonal antibody raised against purified recombinant mouse RD3 indicated that RD3 was present in the inner rather than the outer segment of photoreceptors (Fig. 5A). The immunofluorescence in the photoreceptor layer probed with the anti-RD3 antibody was evident in the inner segments and cell bodies but virtually undetectable in the outer segments. The antibody also nonspecifically cross-reacted with inner retina neurons, both in WT and in rd3/rd3 (not shown), but the signal in photoreceptors was not observed in homozygous rd3/rd3 mice, arguing for its specificity in photoreceptors (Fig. 5A, lower panel).
The RD3GFP fluorescence of the transgenically expressed product was present only in photoreceptors and most brightly between the external limiting membrane and the outer segment layer. RD3GFP fluorescence (Fig. 5, B-E) was predominantly localized to the rod inner segment, cell bodies, and synaptic terminals (Fig. 5B), but very little of the RD3 fluorescence was detectable in rod outer segments ( Expression of the RD3GFP transgene in cones, if any at all, was much lower than in rods (Fig. 5, F and G). Cones, identified using counterstaining with peanut agglutinin, appeared as dark silhouettes surrounded by the fluorescing rods (Fig. 5F). Even though some fluorescence seemingly appeared in the narrow posterior part of sporadic cones after stronger ␥ correction of the image (Fig. 5G), it could be a result of the much brighter fluorescence emission bleeding from the surrounding rods. Moreover, even after the stronger ␥-adjustment of the fluorescence images, RD3GFP was observed in the adjacent rod cell bodies but never inside the boundaries of cone cell bodies marked by rhodamine-peanut agglutinin (Fig. 5G). Hence, although we cannot exclude that a low-level "leaky" transgene expression in rd3/rd3RD3GFP ϩ cones could occur, the better preservation of cone function (Fig. 4D) was more likely due to better protection of cones by averting massive obliteration of the dominating the outer neuronal layer of the retina rods (25). Notably, after deletion of GCAPs by breeding the RD3GFP- , and rd3/rd3 RD3GFP ϩ (filled circles) retinas; the differences in activities (ANOVA, F ϭ 58; p Ͻ 0.0001) when compared using Bonferroni post hoc test were significant between rd3/rd3 (0.046 Ϯ 0.01 nmol cGMP/min/retina) and both rd3/rd3RD3GFP ϩ (0.62 Ϯ 0.03 nmol cGMP/min/retina) and WT (0.77 Ϯ 0.18 nmol cGMP/min/retina) (p Ͻ 0.0001), but not between rd3/rd3RD3GFP ϩ and the WT (p ϭ 0.2). The cyclase activity was assayed in the dark-adapted retinas under IR illumination as described under "Experimental procedures." F, dark adapted ERG (1-ms 505-nm flash of 1 ϫ 10 6 photons/rod) at 1 month of age in wild type (black trace), rd3/rd3 (red trace), and rd3/rd3RD3GFP ϩ (green trace).

Mechanism of rd3 retinal degeneration
expressing transgene into a GCAPs Ϫ/Ϫ background (Fig. 6), the RD3GFP fluorescence was no longer restricted to the rod inner segment-it even became preferentially concentrated in rod outer segments.

Biological function of RD3 in photoreceptors
Two possible roles were proposed and supported by experimental evidence for the RD3 in regulation of photoreceptor guanylyl cyclase-to promote cyclase accumulation in photoreceptors (18 -20) and to inhibit the cyclase activity prior to its reaching the outer segment (21,22). Evidently, the role of RD3 in maintaining the RetGC levels is most critical for the function of photoreceptors-the phototransduction becomes inefficient when the cGMP synthesis is impaired, forcing most of the cGMP gated channels that would be normally open in the dark to become closed (17,22,26) (Fig. 1B). However, when it comes to the role of RD3 in protecting photoreceptors from degeneration, our data indicate that the inhibitory capacity of RD3 is most essential for their survival, even more essential than maintaining the RetGC levels in the outer segment. Indeed, we pre-viously observed (22) that photoreceptors in RetGC1 Ϫ/Ϫ RetGC2 Ϫ/Ϫ double knockout, completely devoid of the photoreceptor cyclase activity (27,28), degenerate much slower than rd3/rd3 photoreceptors, who retain low yet clearly detectable levels of RetGC, but are unable to suppress that residual cyclase activity by RD3. Our present study agrees with the hypothesis that the prevention of premature/aberrant RetGC activation by GCAPs, most likely in the inner segment, is crucial for the photoreceptor survival (Fig. 7). Although the levels of the cyclase activity in rd3/rd3 mice are strongly suppressed because of the compromised delivery and/or accumulation of the enzyme in the outer segment, this does not explain the fast rate and the severity of the rd3 photoreceptor degeneration. RetGC activity in rd3/rd3GCAPs Ϫ/Ϫ is not only reduced to the same low levels as in rd3/rd3 but can no longer even become accelerated by GCAPs (Fig. 1). However, the lack of GCAPs drastically rescues the rd3/rd3 photoreceptors (Fig. 2), evidently because even though there is still no RD3 to suppress the residual RetGC activity in these photoreceptors, there are also no GCAPs now to cause its aberrant activation-hence, the fast degeneration is averted (Fig. 7). Whereas the fast degeneration results from

Mechanism of rd3 retinal degeneration
GCAP-mediated cyclase activation in the absence of RD3, the remaining-much slower-loss of photoreceptors in rd3/ rd3GCAPs Ϫ/Ϫ could result from constitutively low activity cGMP gated channels in the outer segment triggering slow degeneration (22,28,29). However, it could also be a result of incomplete inhibition of the cyclase by RD3. RD3 normally inhibits both GCAP-stimulated and the basal activity of the cyclase (21,22). Because the basal activity of RetGC is detectable in rd3/rd3 (Fig. 1A) and remains detectable even in rd3/ rd3GCAPs Ϫ/Ϫ , it is conceivable that not only GCAP-stimulated activity, but, to a lesser extent, the basal activity of the cyclase remaining in rd3/rd3GCAPs Ϫ/Ϫ when unrestrained by RD3 is sufficient to eventually cause the degeneration.
The processes triggered downstream from deregulation of the cyclase in the absence of RD3 are yet to be delineated in the future studies, but regardless of what these processes could be, they are most likely triggered in the inner segment (Fig. 7). The localization of RD3, both of the endogenous protein in WT and of the GFP-tagged transgenic product, argue that RD3 predominantly operates in the inner segment (Figs. 4 -6). The lack of RD3 immunofluorescence in the outer segment that we observed (Fig. 4A) is very unlikely to be a result of a simple epitope masking, because Zulliger et al. (20) reported a similar result using independently produced antibody. Furthermore, the intrinsic fluorescence of RD3GFP demonstrates pattern of distribution between the inner and outer segments similar to the endogenous RD3 immunofluorescence (Fig. 5). Because the transgenically expressed RD3GFP effectively substitutes for the endogenous RD3 in rd3/rd3 photoreceptors (Figs. 3 and 4), this compartmentalization of RD3GFP must reflect the normal distribution of a functional RD3 rather than accumulation in the inner segment of a misfolded inactive protein.
The disparity in RD3 distribution between the two cellular compartments documented in our study and in Ref. 20 appears to be at odds with a report (23) stating that RD3 accumulates in both outer and inner segment, as well as in any other layer of the retina. In that report, however, the specificity of the antibody was not tested using rd3/rd3 retinas as a negative control, and therefore the results of the immunofluorescence in Ref. 23 could be affected by nonspecific cross-reactivity of the primary antibody.
It is also important that the absence of GCAPs strongly alters the RD3GFP fluorescence distribution between the inner and outer segment (Fig. 6), consistent with the earlier observation by Zulliger et al. (20) that the RD3 levels in GCAPs Ϫ/Ϫ ROS become elevated. Our results favor the hypothesis of the competition between RD3 and GCAPs for the cyclase in different cellular compartments (21, 22) (Fig. 7). Based on the kinetic

Mechanism of rd3 retinal degeneration
studies (21,22), RD3 and GCAP reduce each other's apparent affinities for the cyclase. Evidently, the inhibitory action of RD3 prevents GCAPs from binding to the cyclase while in the inner segment. This trend must become reversed as the cyclase enters the outer segment, where GCAPs now replace RD3 and put the cyclase under control of the negative calcium feedback. Competition with GCAPs would normally restrict RD3 from accumulating in the outer segment, but in GCAPs Ϫ/Ϫ photoreceptors, where RD3 faces no such competition, it has a better chance of reaching the outer segment in the complex with the guanylyl cyclase.
Currently little is known about the mechanisms of RD3 interaction with the cyclase. The structures of GCAPs (30 -32) and recently RD3 (33) have been established; however, the structure of the cyclase itself still presents a major barrier for understanding the molecular dynamics of the complex. To date, the threedimensional structure of RetGC has been resolved only for its catalytic core (34,35), not for those domains that present most important regulatory parts. The regulatory binding of GCAP1 and GCAP2 to the cyclase most critically involves the kinase homology and dimerization domains of RetGC (36,37), but in case of RD3 its interaction with the cyclase appears somewhat more complex. The mutational analyses indicate that, in contrast to GCAPs (36,37), a portion proximal to the C terminus of RetGC is critical for RD3 binding (19). However, in addition to that, some point mutations in the kinase homology domain (but not dimerization domain) of the cyclase also disable the RD3 binding (36,37). RD3 and GCAP (11,17) not only have highly dissimilar protein sequences (Ͻ5% overall identity and Ͻ9% similarity in pairwise alignment; European Bioinformatics Institute EMBOS Needle on-line tool) but also drastically dissimilar three-dimensional structures. The guanylyl cyclasebinding interface on RD3 (22,33) occupies the central part of RD3 elongated bundle of ␣-helices, whereas the cyclase-binding interface of GCAP occupies portions of three helix-loophelix EF-hand domains (38). This suggests that most likely the binding sites for RD3 and GCAP on RetGC are also nonidentical. The opposite effects of GCAPs and RD3 (21,22) (Fig. 3) on the cyclase activity, however, could indicate that the binding interfaces for GCAPs and RD3 on RetGC are partially overlapped or are in a close proximity to each other in a threedimensional structure of the cyclase. Nonetheless, the mutually restrictive binding of RD3 and GCAP does not exclude a possibility of a loner-distance allosteric interference via nonoverlapping interfaces, such that binding of one regulatory protein causes, through a rearrangement of the quaternary structure of the cyclase dimer, hindrance for the binding of the other regulatory protein. The structure of the RD3-RetGC complex, the mechanisms of GCAP displacement by RD3 and vice versa, together with the pathways of the photoreceptor degeneration triggered by RetGC in the inner segment, remain the major questions for the future studies.  mouse photoreceptors as well as those in LCA12 patients, the content of RetGC in the outer segment is reduced, but the fast degeneration in this case results not from the low content of the cyclase but from the lack of the ability to block aberrant activation of RetGC by GCAPs in the inner segment. Right panel, elimination of GCAPs further aggravates the deficiency of cGMP production in rd3/rd3 photoreceptors but averts the fast degeneration, because even in the absence of RD3 the cyclase now does not become activated in the inner segment. See "Discussion" for other details.

Animals
All experiments involving animals were conducted in accordance with the Public Health Service guidelines and approved by the Salus University Institutional Animal Care and Use Committee. The WT C57B6J and rd3/rd3 mouse strains were purchased from JAX Research/Jackson's Laboratory. Mice of other genotypes used in this study were kindly provided by other investigators: RetGC1 Ϫ/Ϫ line GCE null (27) by Dr. David Garbers (University of Texas), RetGC2 Ϫ/Ϫ (28) by Dr. Wolfgang Baehr (University of Utah), and GCAPs Ϫ/Ϫ (13) by Dr. Jeannie Chen (University of Southern California). All mouse lines were made congenic to the C57B6 background by repetitive breeding for over 10 generations prior to conducting the experiments (22).

RD3 expression and purification
A human RD3GFP cDNA was constructed using the "splicing by overlap extension" method (39) as previously described in full detail (22). Recombinant human RD3, either untagged or tagged with eGFP (Clontech), was expressed from a pET11d vector in a BL21(DE3) Codon Plus E. coli strain (Stratagene/ Agilent Technologies) induced by isopropyl-␤-D-thiogalactopyranoside, extracted from the inclusion bodies, and purified by salt precipitation and dialysis as previously described (21,22).

RD3GFP transgenic mice
The expression construct was assembled in a Stratagene pBluescript plasmid as follows. The human RD3GFP cDNA was PCR-amplified from a pQBI-fN3 construct previously used for RD3GFP expression in HEK293 cells (37) by PhusionFlash DNA polymerase (Thermo Fisher) using 5Ј-GGGAGAAAACT-CGAGGGGCCAGGGGCTATGTCTCTCATCT and 5Ј-ATA-GAGGATCCTCAGTCGGCTTTGGGCGCCCGGAAT primers and inserted into the XhoI/BamHI sites of a plasmid harboring 4.2-kb 5Ј-UTR of a mouse rod opsin gene (KpnI/XhoI fragment) and a modified polyadenylation signal fragment of the last exon of a mouse protamine Mp1 gene (24) inserted in the BamHI/XbaI sites of pBluescript. The construct was verified using automated Sanger sequencing on both strands. The 6.3-kb fragment containing the transgenic construct excised by the PvuI/XbaI and purified from agarose gel was injected in male pronuclei of fertilized mouse eggs (service of the University of Michigan Transgenic Animal Model Core). The tail DNA samples from the F 0 founders and later of their progeny were tested for the presence of the RD3GFP transgene by PCR using a pair of primers, 5Ј-GGTCCACCTAT-GACCTCAGCCCCAT (f1) and 5Ј-TGACAGAGAACTTGTG-GCCGTTAACAT (r1), generating 0.5-kb RD3GFP-specific fragment (Fig. 3C). The transgene-positive line was then maintained by crossing to the C57B6J and to rd3/rd3 mice in the C57B6J background (22).

Mouse Rd3 gene genotyping
A 0.6-kb fragment of the gene containing part of the Rd3 exon 3 and the adjacent portion of intron 2 was PCR-amplified from tail DNA samples using high-fidelity Phusion Flash DNA polymerase (Thermo Scientific) and a pair of primers, 5Ј-CCT-CAGAACTTTGTCCCTGTCAGCCA and 5Ј-GGTGATTGC-CCTGGGTGGTCAGT, purified using a Zymo Research DNA cleanup kit, and subjected to automated Sanger sequencing using a primer, 5Ј-ATGTCACTGGCAAAGGGTGCGAT, to identify the WT, heterozygous rd3, and homozygous rd3/rd3 by the presence of a stop codon-generating C 3 T transition (17).

GCAP1 expression and purification
Myristoylated bovine GCAP1 for in vitro assays was expressed from pET11d vector in a BLR(DE3) E. coli strain (both 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 a previously published procedure (40,41). The purity of GCAP1 preparations estimated by SDS gel electrophoresis was Ն90%.

RetGC1 expression and guanyly cyclase activity assays
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 previously described (42). The guanylyl cyclase activity was assayed as previously described in detail in Ref. 42 with modification described in Ref. 22. In brief, the assay mixture (25 l) containing HEK293 membranes, 30 mM MOPS-KOH (pH 7.2), 60 mM KCl, 4 mM NaCl, 1 mM DTT, 2 mM Ca 2ϩ /Mg 2ϩ /EGTA buffers, 0.9 mM free Mg 2ϩ , 0.3 mM ATP, 4 mM cGMP, 1 mM GTP, 1 Ci of [␣-32 P]GTP, 100 M zaprinast and dipyridamole, and 10 mM creatine phosphate/0.5 unit of creatine phosphokinase (Sigma-Aldrich) 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 2 M LiCl in scintillation vials, and the radioactivity was counted using liquid scintillation. Mouse retinas for RetGC activity measurements were excised from the dark-adapted ϳ3.5-week-old mice using IR illumination (Kodak number 11 IR filters) and a dissecting microscope fitted with an Excalibur IR goggles as described, wrapped in aluminum foil, frozen in liquid N 2 , and stored at Ϫ70°C prior to their use in the cyclase activity assays conducted under IR illumination. The incubation time for the reaction was 12 min. The assay contained [ 3 H]cGMP internal standard to ensure the lack of cGMP product hydrolysis by light-activated phosphodiesterase. Ca 2ϩ /EGTA buffers at 0.9 mM physiological for the photoreceptors (43) free Mg 2ϩ were prepared using Tsien and Pozzan method (44) and verified by fluorescent indicator dyes as previously described in detail (40). Data fitting was performed using Synergy Kaleidagraph 4 software.

Electroretinography
The mice were dark-adapted overnight, their pupils were dilated by applying 1% tropicamide and 2.5% phenylephrine ophthalmic eye drops under dim red safelight illumination, and

Mechanism of rd3 retinal degeneration
then the mice were dark-adapted for another 10 min. Fullfield ERG in mice anesthetized by inhalation of 1.7-1.9% isoflurane (VEDCO) was performed in the dark as previously described in detail (46,47) using a Phoenix Research Laboratories Ganzfeld ERG2 setup. Series of 505-nm 0.1-1-ms light pulses (up to 1.1 ϫ 10 6 photoisomerizations/rod) were delivered through IR camera-guided corneal electrode/LED light source, and 3-10 traces for each flash strength in series were averaged. Constant 505-nm illumination (ϳ33 ϫ 10 3 photoisomerizations/rod/s) was used as a background light to record photopic (cone) ERG responses to 1-ms 505-nm equivalent of 5.4 ϫ 10 5 photoisomerizations/rod and delivered at 1-s intervals; traces were typically averaged from 10 to 15 trials for each animal.

Retinal morphology
The mice were anesthetized with a lethal dose of ketamine/ xylazine injection and perfused through the heart with PBS and then with 2.5% glutaraldehyde in PBS. The eyes were surgically removed and fixed overnight in 2.5% glutaraldehyde, 2.5% formaldehyde, PBS solution (Electron Microscopy Sciences) at 4°C. The fixed eyes were washed in PBS, soaked in PBS overnight, processed for paraffin embedding, sectioned, and stained with hematoxylin/eosin (AML Laboratories, St. Augustine, FL). The retinas were photographed using an Olympus Magnafire camera mounted on an Olympus BX21 microscope. The photoreceptor nuclei in the outer nuclear layer of the retina were counted from the 425-m fragment of the 5-m-thick retina section between the optic nerve and the periphery, and the density of the nuclei per 100 m was averaged for each genotype.

Antibodies and immunoblotting
Rabbit RD3 polyclonal antibody 10929 was produced against purified recombinant mouse RD3 expressed in E. coli as described in Refs. 21, 22; GCAP1 and RetGC1 rabbit polyclonal antibodies were characterized previously (45,48), rabbit polyclonal anti-peripherin antibody was a gift from Dr. Andrew Goldberg (Oakland University), and rabbit polyclonal anti-GAPDH antibody was purchased from an Invitrogen/Thermo Fisher Scientific (catalog no. PA1987).
Retinas from 3.5-week-old mice were excised in 20 l of 10 mM Tris-HCl, pH 7.4, containing 1:100 dilution of protease inhibitor mixture concentrate (Millipore-Sigma) and homogenized in a radioimmune precipitation assay mixture of detergents for protein extraction (Abcam) containing the protease inhibitors, 300 l/6 retinas. After 1 h of extraction on ice, the insoluble material was removed by centrifugation at 14,000 ϫ g, 4°C, and the protein extract was mixed with equal volumes of 2ϫ Laemmli SDS sample buffer (Millipore-Sigma) and subjected to electrophoresis in 15% or 7% polyacrylamide gel containing 0.1% SDS. Following the electrophoresis, the proteins were transferred overnight at 50 V to polyvinylidene difluoride Immobilon P membrane (Millipore) at 18°C using Trisglycine transfer buffer (Novex/Invitrogen). The membranes were transiently stained by Ponceau S (Sigma) dye solution in 1% acetic acid to mark the positions of molecular mass markers, destained by series of washes in water and TBS (Fisher Scientific) containing 0.5% Tween 20 (TTBS), and blocked by Super Block (ThermoFisher) solution in TTBS, probed by primary antibody, and the luminescence was developed using peroxidase-conjugated goat anti-rabbit polyclonal IgG (Cappel/MP Biomedical) and a Pierce Super-Signal Femto substrate protocol (Thermo Scientific). The images were acquired using a Fotodyne Luminous FX imager and subjected to densitometry using ImageJ software (National Institutes of Health).

Confocal microscopy
The mice were perfused with the fixative solution as described above except that 10% freshly prepared formaldehyde solution in PBS was used instead of glutaraldehyde. The eyes were then dissected, and the eyecups were placed into 10% formaldehyde/PBS for 5 min and then 5% formaldehyde/PBS for 12 h at 4°C. The fixed eyecups were washed in PBS, impregnated with 30% sucrose in PBS, and frozen at Ϫ70°C in OCT medium (Electron Microscopy Sciences). The cryosections were taken using a Hacker-Bright cryogenic microtome, dried for 1 h at room temperature, and stored in Ϫ70°C. The sections were washed three times in PBS containing 0.1 M glycine (pH 7.4), blocked for 1 h at 30°C with the same solution containing 5% BSA and 0.1% Triton X-100, incubated overnight at 4°C, incubated for 1 h at room temperature with the primary antibody, then washed with PBS solution three times for 15 min each, incubated with the 1:400 diluted donkey anti-rabbit Alexa Fluor 568 -or Alexa Fluor 488 -conjugated antibody (Molecular Probes/Thermo Fisher), and washed four times with PBS for 15 min each at room temperature. Where indicated, the cones were labeled by incubating retinal sections with rhodamine-conjugated peanut agglutinin (Vector Laboratories), 5 g/ml, overnight at 4°C, and washed three times with PBS for 15 min each at room temperature. The sections were covered with Vectashield mounting medium (Vector Laboratories), and confocal images were acquired using an Olympus FV1000 spectral instrument controlled by FluoView FV10-ASW software, collecting in a sequential mode images excited by 488-and 543-nm lasers. Fluorescence of RD3GFP was recorded using excitation at 488 nm. Where indicated, fluorescence was superimposed on a differential interference contrast (DIC) image. No changes were made to the original images except for ␥ correction applied to the whole image for clarity of presentation in the print. Fluorescence distribution, where indicated, was quantified using unmodified images.

Statistics
Statistical significance of the differences was tested by ANOVA/Bonferroni post hoc multiple-pairs comparison test (confidence interval 99%; ␣ ϭ 0.01) using Synergy Kaleidagraph 4 software.