Molecular Vision 2016; 22:1342-1351
Received 31 August 2016 | Accepted 08 November 2016 | Published 10 November 2016
The first three authors contributed equally to the study.
1Department of Ophthalmology, University of Florida, Gainesville, FL; 2Pennsylvania College of Optometry, Salus University, Elkins Park, PA
Correspondence to: A.M. Dizhoor, Pennsylvania College of Optometry, Research, S-416, Salus University, 8360 Old York Road, Elkins Park, PA, 19027; Phone: +1 215-780-1468; FAX: +1-215-780-1464; email: firstname.lastname@example.org
Purpose: To test, in living photoreceptors, two mutations, S248W and R1091x, in the GUCY2D gene linked to Leber congenital amaurosis 1 (LCA1) that fail to inactivate the catalytic activity of a heterologously expressed
Methods: GUC2YD cDNA constructs coding for wild-type human (hWT), R1091x, and S248W GUCY2D under the control of the human rhodopsin kinase promoter were expressed in Gucy2e−/−Gucy2f−/− knockout (GCdKO) mouse retinas, which lack endogenous RetGC activity. The constructs were delivered via subretinally injected adenoassociated virus (AAV) vector in one eye, leaving the opposite eye as the non-injected negative control. After testing with electroretinography (ERG), the retinas extracted from the AAV-treated and control eyes were used in guanylyl cyclase activity assays, immunoblotting, and anti-RetGC1 immunofluorescence staining.
Results: Cyclase activity in retinas treated with either hWT or R1091x GUCY2D transgenes was similar but was undetectable in the S248W GUCY2D-treated retinas, which starkly contrasts their relative activities when heterologously expressed in human embryonic kidney (HEK293) cells. Rod and cone ERGs, absent in GCdKO, appeared in the hWT and R1091x GUCY2D-injected eyes, while the S248W mutant failed to restore scotopic ERG response and enabled only rudimentary photopic ERG response. The hWT and R1091x GUCY2D immunofluorescence was robust in the rod and cone outer segments, whereas the S248W was detectable only in the sparse cone outer segments and sporadic photoreceptor cell bodies. Robust RetGC1 expression was detected with immunoblotting in the hWT and R1091x-treated retinas but was marginal at best in the S248W GUCY2D retinas, despite the confirmed presence of the S248W GUCY2D transcripts.
Conclusions: The phenotype of S248W GUCY2D in living retinas did not correlate with the previously described normal biochemical activity of this mutant when heterologously expressed in non-photoreceptor cell culture. This result suggests that the S248W mutation contributes to LCA1 by hampering the expression, processing, and/or cellular transport of GUCY2D, rather than its enzymatic properties. In contrast, the effective restoration of rod and cone function by R1091x GUCY2D is paradoxical and does not explain the severe loss of vision typical for LCA1 associated with that mutant allele.
The GUCY2D gene (Gene ID: 3000; OMIM 600179) codes for the human retinal membrane guanylyl cyclase 1 (RetGC1) isozyme and is one of the genes most frequently associated with the early onset, inherited recessive blindness, Leber congenital amaurosis type 1 (LCA1) [1-3]. The acute loss of vision in patients with GUCY2D LCA1 results not from the severe degeneration of rods and cones  but from the inefficient production in the diseased photoreceptors of the second messenger in phototransduction, cyclic GMP (cGMP) . Of the two existing retinal membrane guanylyl cyclase isozymes, RetGC1 (GUCY2D in humans) and RetGC2 (GUCY2F) [5,6], the former catalyzes most of the cGMP production in mammalian photoreceptors [7,8]. Note that the RetGC1 ortholog in mice is coded by the Gucy2egene ; therefore, only the human RetGC1 will be further referred to as GUCY2D (wherever the difference in gene-based nomenclature between species can cause confusion, we use the biochemical names, RetGC1 and RetGC2). cGMP synthesis becomes decelerated in the dark and accelerated in the light by guanylyl cyclase activating proteins (GCAPs) . These proteins are Ca2+-sensor proteins that switch between their Ca2+- and the Mg2+-bound states following the respective increase or decrease of Ca2+ influx through the cGMP gated channels between dark and light . Another photoreceptor protein, retinal degeneration 3 (RD3) [11,12], reduces RetGC1 activity by competing with GCAPs  and is essential for the normal production of RetGC1 and its transport to the outer segment [14-16]. We previously showed, in vitro, that different mutations linked to GUCY2D LCA1 inactivate the catalytic activity of RetGC1 and/or suppress its ability to bind GCAPs and RD3 [15,17].
Rod and cone function can be sustainably restored in mice that lack endogenous orthologs of RetGC1 (Gucy2e−/−) and RetGC2 (Gucy2f−/−; GCdKO)  by delivering functional GUCY2D cDNA with adenoassociated viral (AAV) vectors [19,20]. Because the majority of patients with GUCY2D LCA1 exhibit preservation of the retinal structure, an opportunity exists to consider a human gene therapy to rescue vision in this disease [3,21]. Among the reported recessive alleles linked to GUCY2D LCA1 [1-3], the properties of two mutations in GUCY2D in vitro would not explain the virtually complete lack of photoresponses in patients with LCA1 . These mutations include a missense mutation, Ser248→Trp, in the N-terminal extracellular domain of the enzyme (Figure 1) , and a nonsense mutation, R1091x [1,3], that truncates a short C-terminal region, downstream from the catalytic domain (Figure 1). Unlike other LCA1-linked GUCY2D mutants shown in vitro [3,17,22] to completely inactivate the enzymatic activity and regulation of the RetGC1 isozyme by Ca2+ and GCAPs, the biochemical properties remained substantially preserved in R1091x and remained completely normal in S248W.
By using subretinal delivery via AAV in the present study, we expressed in mouse retinas that lacked endogenous cyclase activity variants of human GUCY2D cDNA that harbor the two mutations linked to LCA1 and tested their physiologic effects in living photoreceptors. We found that S248W GUCY2D, despite its maintenance of full catalytic capacity as a recombinant protein in vitro, was completely unable to sustain rod function and failed to accumulate in rod outer segments in GCdKO mice. S248W GUCY2D appeared in only a small number of sparse GCdKO cone outer segments, thus mediating barely detectable cone responses. In contrast, AAV-mediated R1091x GUCY2D accumulated in GCdKO photoreceptor outer segments and enabled well-detectable rod and cone photoresponses.
The mouse Gucy2e−/−gene knockout line  that originated from the D. Garbers laboratory (University of Texas) and the Gucy2f−/− gene knockout line  that originated from the W. Baehr laboratory (University of Utah) were bred into a homozygous (Gucy2e−/−Gucy2f−/−) state. All experiments were approved by the University of Florida and Salus University’s Institutional Animal Care and Use Committee and were conducted in accordance with the Association for Research in Vision and Ophthalmology’s (ARVO’s) Statement for the Use of Animals in Ophthalmic and Vision Research and with National Institutes of Health regulations. Animals were maintained at the University of Florida Health Science Center Animal Care Services Facility under a 12 h:12 h light-dark cycle. Food and water were available ad libitum. Animals were shipped to Salus University for terminal assays.
The original vector backbone has been previously described in detail . The recombinant AAV vector plasmids contained flanking AAV2 inverted terminal repeats, human rhodopsin kinase promoter, an SV40 splice donor-acceptor site, human GUCY2D cDNA coding for wild-type or LCA1 mutants, and a bovine growth hormone polyadenylation signal. Wild-type GUCY2D cDNA (hWT) was synthesized de novo and contained silent mutations that remove several restriction sites to facilitate molecular cloning. The S248W GUCY2D-containing vector plasmid was constructed by replacing the Eco47III-flanked fragment coding for the S248W mutation. The R1091x GUCY2D vector was constructed by replacement of the NarI/SalI fragment containing the TGA codon at position 1091. All constructs were verified with DNA sequencing and packaged using a modified AAV8 capsid protein containing a single Tyr733Phe substitution. The AAV particles were manufactured, purified, and titered as described elsewhere [25,26]. The titers of the viruses harboring the hWT, S248W, and R1091x mutants were 4.06 × 1013, 5.16 × 1013, and 7.64 × 1013 vg/ml, respectively. All were contained in a balanced salt solution (Alcon, Ft. Worth, TX) supplemented with 0.014% Tween-20.
Subretinal injections were performed as previously described . Briefly, eyes of postnatal day 30 (P30) GCdKO mice were dilated 1 h before injection with 1% atropine, followed by 2.5% phenylephrine (Akorn, Inc., Lake Forest, IL). The mice were then anesthetized with ketamine (72 mg/kg)/xylazine (4 mg/kg; Phoenix Pharmaceutical, St. Joseph, MO). Trans-corneal subretinal injections of un-diluted virus (1 μl) were performed using a 33-gauge blunt needle on a Hamilton syringe.
Treated and untreated eyes from GCdKO mice were enucleated 4 weeks post-injection, and each retina was immediately dissected from the eyecup and submerged in RNA Protect (Qiagen, Inc., Valencia, CA) and then was homogenized in RLT Buffer (RNeasy Protect Mini Kit; Qiagen, Inc.) containing 2-mercaptoethanol for 45 s. RNA was extracted and reverse transcribed as reported previously . Real-time PCR was conducted using Bio-Rad iQ SYBR Green Supermix interfaced with C1000 Touch thermal cycler (Bio-Rad Laboratories, Hercules, CA) to measure the content of hWT and S248W using a forward, 5′-ATC CCA CCC GAG CGG CGA-3′, primer and a reverse, 5′-GGC AAC TAG AAG GCA CAG TCG AGG C-3′, primer (Integrated DNA Technologies, Coralville, IA) amplifying a 103 bp region at the junction between the 3′-end of the GUCY2D cDNA and the 5′-end of the bovine growth hormone polyadenylation signal fragment. Before the experimental data were collected, the primers were validated according to minimum information for publication of quantitative real-time (QRT) PCR experiment (MIQE) standards . PCR samples were initially heated to 95 °C for 3 min followed by 40 cycles of amplification/plate reading: 95 °C for 10 s, 58 °C for 30 s, plate reading, and 72 °C for 30 s. Once finished, a melt curve was assessed from 65 ° C to 95 °C with a plate read every 0.5 °C. Measurements averaging three replicate reactions for each retina were calculated using the ΔΔCq method, with the signals normalized to glyceraldehyde 3-phosphate dehydrogenase-specific transcripts as a control and were then normalized per signal in the non-injected GCDKO eye.
Myristoylated bovine GCAP1 was expressed from a Novagen pET11d vector in BLR(DE3) Escherichia coli strain and purified to about 95% electrophoretic homogeneity as described previously in detail . In brief, inclusions bodies containing myristoylated GCAP1 were obtained by three rounds of sonication on ice and centrifugation to remove soluble proteins. GCAP1 was extracted from the inclusion bodies in the presence of urea, dialyzed to remove urea, and then FPLC-purified using butyl-Sepharose and high-resolution size-exclusion Superdex 200 (GE Healthcare Life Sciences, Pittsburgh, PA) column chromatography. Human RD3 was expressed from the pET11d vector in the BL21(DE3) CodonPlus E. coli strain (Stratagene/Agilent Technologies, Santa Clara, CA), extracted from the inclusion bodies, and purified to about 95% electrophoretic homogeneity as previously described in detail . In brief, inclusion bodies containing RD3 were purified by three rounds of sonication/centrifugation, then RD3 was extracted from the inclusion bodies in the presence of urea, dialyzed, purified by precipitation in the presence of 250 mM NaCl, washed by centrifugation, extracted with urea second time and dialyzed, after which the insoluble material was removed by centrifugation and the supernatant containing purified RD3 was collected and used in the experiments.
RetGC1 activity was assayed using [α-32P]GTP (Perkin Elmer, Waltham, MA) as the substrate as previously described [3,7,29]. As previously described , membrane fractions from transfected human embryonic kidney (HEK293) cell cultures expressing RetGC1 were incubated for 40 min at 30° under ambient illumination and inactivated at 95° for 2 min. Mouse retinas for the RetGC activity measurements were excised from the dark-adapted mice under a dissecting microscope fitted with an Excalibur infrared goggles (Fogelsville, PA) , and the assays containing retinal samples were conducted under infrared illumination using Kodak number 11 (Eastman Kodak, Rochester, NY) infrared filters and were incubated for 12 min. The [32P]cGMP product of the reaction separated with thin-layer chromatography was quantified with liquid scintillation counting [7,29].
Mice anesthetized with a lethal dose of ketamine/xylazine injection were perfused through the heart with PBS (Fisher Scientific, Fair Lawn, NJ, product number BP399-1; 1X: 137 mM NaCl, 2.7 mM KCl, 11.9 mM phosphate buffer, pH 7.4) and then with 10% freshly prepared formaldehyde solution in PBS. The eyecups from the enucleated eyes were additionally fixed in 10% formaldehyde/PBS for 5 min and for 1 h in 5% formaldehyde/PBS at room temperature. The fixed eyecups were impregnated with 30% sucrose in PBS overnight and frozen in optimum cutting temperature (OCT) medium (Electron Microscopy Sciences, Hatfield, PA) at −70 °C. Cryosections (20 µm thick) were sliced using a Hacker-Bright (Hacker Instruments & Industries Inc., Winnsboro, SC) cryogenic microtome, dried for 1 h at room temperature, and stored in –70 °C. The sections were washed 3X in PBS/0.1 M glycine (pH 7.4) and blocked for 1 h at 30 °C in the same solution plus 5% bovine serum albumin and 0.1% Triton X-100. The sections were incubated overnight at 4 °C and then for 1 h at room temperature with the rabbit polyclonal antibody against the catalytic domain of RetGC1 , washed 4X for 15 min with PBS, incubated for 1 h at room temperature with the donkey anti-rabbit Alexa Fluor 568-conjugated antibody (1:400, Molecular Probes, Invitrogen, Carlsbad, CA), and washed 3X for 15 min each at room temperature. Where indicated, fluorescein-labeled peanut agglutinin (Vector Laboratories, Burlingame, CA) was added with the secondary antibody to label the cone outer segment sheaths. Sections were covered with a TOPRO3 iodide-containing or plain Vectashield mounting medium (Vector Laboratories). Confocal images were acquired using an Olympus (Center Valley, PA) FV1000 Spectral instrument controlled with FluoView FV10-ASW software (Center Valley, PA), collecting in a sequential mode emission excited by 488, 543, and 635 nm lasers, assigned respective green, red, and pseudoblue color channels.
One month post-injection, the mice were dark-adapted overnight, their pupils were dilated with 1% tropicamide and 2.5% phenylephrine ophthalmic eye drops under dim red safelight illumination, and then the mice were dark-adapted for another 15 min. Full-field ERG was performed in the dark using a Maxwellian view projection-based Phoenix Research Laboratories (Pleasanton, CA) Ganzfeld ERG 2 setup. Mice were anesthetized with 50 ml/min delivery of 1.7–1.9% Isoflurane (VEDCO) using a Kent Scientific (Torrington, CT) SomnoSuite Small Animal Anesthesia System and remained on a heated pad during the entire procedure. To evoke scotopic electroretinography (ERG) responses, 1 ms 505 nm light pulses of 1.2 × 106 photons µm−2 at the mouse retinal surface were delivered through the infrared camera-guided corneal electrode/light-emitting diode (LED) light source in 3 min intervals, and the traces from three to four trials were averaged. Constant 505 nm illumination, equivalent to 33 × 103 photons rod−1 s−1, was used as a background to record photopic ERG responses to 1 ms 505 nm and 365 nm pulses of 6 × 105 photons µm−2 at 1-s intervals, and traces were typically averaged from 15 to 25 trials for each animal.
As expected , the non-injected eyes of the GCdKO mice completely lacked photoreceptor activity. However, injection of AAV vectors harboring wild-type (hWT) (9 mice) and R1091x (10 mice) GUCY2D enabled scotopic (Figure 2A) and photopic ERG (Figure 2B) responses. The scotopic ERG amplitudes varied between different animals (Figure 2C), a result likely due to the variable efficacy of subretinal injections, but the average scotopic ERG a-wave enabled by R1091x RetGC1 expression was reduced only slightly and was not statistically significantly different from that of the wild-type GUCY2D injected eyes (Figure 2D). In contrast, S248W GUCY2D enabled only marginally detectable cone b-wave in photopic ERG (Figure 2B) in eight of the nine tested animals and no identifiable scotopic a-wave in scotopic ERG (Figure 2A,C,D).
RetGC activity, undetectable in the non-injected eyes of GCdKO mice , was imparted by the wild-type and R1091x GUCY2D transgenes but remained virtually undetectable in the retinas injected with the S248W construct (Figure 3A). The wild-type and R1091x RetGC1 protein products of the AAV-delivered GUCY2D transgenes were robust on immunoblotting (Figure 3B), but S248W was barely detectable. Despite the lack of obvious protein expression of restoration of retinal function, S248W GUCY2D cDNA transcripts were clearly present in two AAV-S248W GUCY2D treated eyes (Figure 3C).
In the wild-type and R1091x GUCY2D cDNA-treated retinas, the immunofluorescence for the AAV-mediated cyclase expression was strong in the rods and cones (Figure 4). The brightest signal in both cases was present in the outer segments. In contrast, S248W RetGC1 was occasionally detected in sparse cones but remained virtually undetectable in the rod outer segments. The overall localization patterns correlated with the presence of rod and cone ERG in the wild-type GUCY2D and R1091x GUCY2D-treated retinas and only marginal cone responses in the S248W GUCY2D-treated mice (Figure 2).
In the absence of a RetGC-inhibiting protein (RD3), mouse photoreceptors fail to accumulate RetGC1 and RetGC2 in the outer segments [14-16]. Therefore, the weak expression of transgenic S248W GUCY2D protein in living photoreceptors prompted us to test whether binding to RD3 was strongly compromised. We used the same HEK293 membrane preparations containing heterologously expressed S248W and wild-type GUCY2D that previously were shown to retain activity in vitro . Recombinant human RD3, which inhibits RetGC1 by competing with GCAP1 [13,30], completely suppressed GCAP1-activated S248W and wild-type GUCY2D activity in vitro in a similar dose-dependent manner and with nearly identical apparent affinities (Figure 5).
As a surprising finding from an earlier study, GUCY2D mutants encoded by two recessive LCA1 alleles from human patients remain partially (R1091x) or fully (S248W] active when heterologously expressed in HEK cells . Our present findings help explain why S248W GUCY2D does not sustain vision when another loss-of-function LCA1 allele is present in the same patient. The affected residue is located in the extracellular domain [5,6] of the cyclase, the part mostly exposed into the intradiscal space in the outer segments, which does not participate in regulation of cyclase activity by GCAPs on the cytoplasmic side of the enzyme [17,31]. However, S248W GUCY2D fails to properly express in photoreceptors.
The downfall in S248W expression in the retina most likely occurs at or after the protein synthesis stage (Figure 3): either insufficient synthesis or an accelerated degradation rate or both. The rudimentary cone ERG accompanying the S248W GUCY2D presence in a small number of cone outer segments suggests that the mutant cyclase present in the outer segments retains its enzymatic function, consistent with its unaffected regulatory properties in vitro , and therefore is unlikely to undergo global protein unfolding. However, the replacement of a strongly hydrophilic side chain with a strongly hydrophobic residue could affect the folding of the extracellular domain enough to prompt GUCY2D degradation, either as the nascent polypeptide or before it reaches the outer segment. Evidently, the ability to bind RD3 (Figure 5), implicated in facilitating RetGC1 transport to the outer segment [12,14-16], does not help S248W accumulate in the outer segments once the N-terminal portion of the enzyme becomes compromised. The presence of S248W GUCY2D–specific immunofluorescence in cell bodies along with sparse cone outer segments (Figure 4) suggests that the affected step could be post-translational rather than cotranslational.
In this study, we used Gucy2e−/−Gucy2f−/− mice, which lack RetGC1 and RetGC2 isozymes, while in patients with LCA1 RetGC2 is unaffected. However, the presence of the unaffected RetGC2 could not compensate for the defects in RetGC1 for the following reasons. RetGC2 is expressed in rods but not in cones [20,32]. In contrast, RetGC1 is the predominant (70–80% total activity) isozyme in mammalian rods and virtually the only isozyme detected in cones [7,20,23]. Consequently, even when GUCY2F remains unaffected in patients with LCA1, only the mutated GUCY2D is present in cones. Catalytic activity of RetGC requires its dimerization [33-35]; thus, in rods RetGC2 would be able to rescue S248W RetGC1, by forming a heterodimer containing the two isozymes. However, this possibility is highly unlikely. Although RetGC1 and RetGC2 can form a heterodimer when heterologously coexpressed in HEK293 cells , only homodimers, RetGC1:RetGC1 and RetGC2:RetGC2, are formed in living photoreceptors . Therefore, we reason that the use of the double-RetGC knockout in our study provides adequate conditions for testing the S248W GUCY2D. Moreover, we suggest that using the Gucy2e−/−Gucy2f−/− background in this case is preferable, because these mice retain neither biochemical RetGC2 activity nor rudimentary rod ERG [7,18,30], which are present in Gucy2e−/− single-knockout  and are capable of obscuring effects of the AAV-delivered S248W GUCY2D.
In a stark contrast to the S248W GUCY2D allele, the present in vivo study was not able to elucidate the role of the R1091x allele in LCA1 pathogenesis. We found that this mutant is expressed well in photoreceptor cells, robustly accumulates in the outer segments of living rods and cones, and effectively enables their responses to light in a fashion similar to the wild-type GUCY2D transgene. A plausible explanation for the obvious paradox between the substantial activity of the R1091x GUCY2D found in a mouse retina and the established genetic link to LCA1 [1,3] is that in human photoreceptors this allele is expressed differently or the expressed protein is delivered to the outer segment through a different pathway than in the mouse. Although little is known about the molecular mechanisms of RetGC processing and trafficking to the outer segment, two RetGC isozymes in mouse rods can evidently utilize different pathways . Therefore, it is also conceivable that trafficking of RetGC1 in primate rods could involve different protein interactions than in rodent photoreceptors and be less permissive for the GUCY2D lacking its C-terminal fragment. Perhaps expression of the human R1091x GUCY2D mutant in a GUCY2D-deficient primate animal model could resolve this paradox, but such a model is not available at this time.
The funding for this study was provided by the National Eye Institute grant R01EY011522 (AMD), Pennsylvania Department of Health CURE Formula grant (AMD), Foundation Fighting Blindness (SEB), Research to Prevent Blindness (SEB), Creed’s Cause foundation (SEB). A part of this study was presented at the 2016 ARVO annual meeting, abstract #3180. Disclosure: SEB and SLB have a US patent pending (61/327,521) related to the use of AAV for the treatment of GUCY2D-LCA1.