RNAi-Mediated Gene Suppression in a GCAP1(L151F) Cone-Rod Dystrophy Mouse Model

Dominant mutations occurring in the high-affinity Ca2+-binding sites (EF-hands) of the GUCA1A gene encoding guanylate cyclase-activating protein 1 (GCAP1) cause slowly progressing cone-rod dystrophy (CORD) in a dozen families worldwide. We developed a nonallele-specific adeno-associated virus (AAV)-based RNAi knockdown strategy to rescue the retina degeneration caused by GCAP1 mutations. We generated three genomic transgenic mouse lines expressing wildtype (WT) and L151F mutant mouse GCAP1 with or without a C-terminal GFP fusion. Under control of endogenous regulatory elements, the transgenes were expressed specifically in mouse photoreceptors. GCAP1(L151F) and GCAP1(L151F)-GFP transgenic mice presented with a late onset and slowly progressive photoreceptor degeneration, similar to that observed in human GCAP1-CORD patients. Transgenic expression of WT GCAP1-EGFP in photoreceptors had no adverse effect. Toward therapy development, a highly effective anti-mGCAP1 shRNA, mG1hp4, was selected from four candidate shRNAs using an in-vitro screening assay. Subsequently a self-complementary (sc) AAV serotype 2/8 expressing mG1hp4 was delivered subretinally to GCAP1(L151F)-GFP transgenic mice. Knockdown of the GCAP1(L151F)-GFP transgene product was visualized by fluorescence live imaging in the scAAV2/8-mG1hp4-treated retinas. Concomitant with the mutant GCAP1-GFP fusion protein, endogenous GCAP1 decreased as well in treated retinas. We propose nonallele-specific RNAi knockdown of GCAP1 as a general therapeutic strategy to rescue any GCAP1-based dominant cone-rod dystrophy in human patients.


Introduction
Cone-rod dystrophy (CORD, with a prevalence of 1/40,000) is a rare, highly heterogeneous class of hereditary retinal disease inherited in a dominant, recessive or X-linked fashion [1]. The disease manifests with photoaversion, reduced central visual acuity, achromatopsia at early stages, and eventually loss of peripheral vision attributed to progressive loss of first cone and then rod photoreceptors. Thus far, 27 genes have been linked to cone-rod dystrophy; of these, ten genes are associated with dominant CORD, 15 with recessive CORD, two are X-linked (RetNet, https://sph.uth.tmc.edu/retnet/). The protein products of these genes are involved in multiple aspects of photoreceptor structure and function [2]. One of the best characterized dominant CORD genes is GUCA1A encoding guanylate cyclase-activating protein 1 (GCAP1 [3]. Worldwide, about one dozen families with more than 100 affected members have been identified to date [4]. GCAP1 plays a key role in accelerating guanylate cyclase activity in retinal photoreceptors. Rod phototransduction is regulated by two guanylate cyclases (GC1 and GC2) and two GCAPs (GCAP1 and GCAP2) [5]. The two GCAP genes (Guca1a and Guca1b) are arranged in a tail-to-tail array on mouse chromosome 17 [6] while the GC genes (Gucy2e and Gucy2f) are located on different chromosomes (chromosomes 11 and X, [7]). The two GCAPs overlap partially in regulating the GCs of rods, and both contribute to rod recovery after photolysis [8][9][10][11]. In cones, only GCAP1 is involved in regulating GC1, the predominant GC of cone phototransduction. Germline deletion of both GCAPs renders GC activity Ca 2+ insensitive; flash responses from dark-adapted rods were larger and slower, and recovery to the dark state was delayed [8]. By contrast, cone ERGs recorded from GCAPs 2/2 mice had normally saturated a-wave and b-wave amplitudes and increased sensitivity of both M-and S-cone systems [12]. Both the cone driven b-wave and a-wave were delayed similarly as observed in rods. Transgenic GCAP1 could restore normal rod and cone response recovery [9,12].
The L151F mutation was identified in two independent families with cone and cone-rod dystrophy, respectively [17,23]. In the first family [17], dyschromatopsia, hemeralopia and reduced visual acuity were evident by the second-to-third decades of life and photopic electroretinographic (ERG) responses were non-recordable. In the second, five-generation family [23], 11 of 24 individuals were affected by photoaversion, color vision defects, central acuity loss and legal blindness in decades 2-3 of life. The causative GCAP1(L151F) mutation, a conservative substitution, disrupted Ca 2+ coordination at EF-hand 4 and changed Ca 2+ sensitivity. Molecular dynamics suggested a significant decrease in Ca 2+ -binding to EF-hand motifs 2 and 4, with an overall shape change of the L151F-GCAP1 molecule relative to wildtype [17].
Two mouse models of retinal dystrophy associated with GCAP1 mutations have been generated previously. The first, a transgenic mouse model expressing bovine GCAP1(Y99C) under control of the opsin promoter, produced severe and rapid rod-cone degeneration resembling retinitis pigmentosa [25]. The second model, a GCAP1(E155G) knock-in mouse, displayed late-onset and slowly progressive cone-rod photoreceptor degeneration, mimicking human CORD [26]. In this study, we generated three transgenic mouse lines expressing wildtype and L151F mutant GCAP1 with or without a C-terminal EGFP fusion. The EGFP tag was instrumental toward developing an effective knockdown RNA interference and served to distinguish transgenic GCAPs from endogenous GCAPs. All transgenes contained the complete mouse Guca1a gene, including native regulatory elements. Mutant transgenic mice develop retina pathology slowly and recapitulate features of human CORD. To develop a vector suitable for knockdown gene therapy, we generated scAAV2/8 virus that expresses a nonallele-specific shRNA that targets both mutant and native Guca1a mRNAs. By immunoblot and in-vivo fundoscopy, we show that both transgenic and endogenous GCAP1 were downregulated effectively in AAV-treated retinas. These data establish shRNA-mediated RNAi as a potential therapeutic strategy for adCORD patients carrying any EF-hand GUCA1A mutation.

Mice (Ethics Statement)
Procedures for the animal experiments of this study were IACUC-approved by the University of Utah and conformed to recommendations of the Association of Research for Vision and Ophthalmology (ARVO). Transgenic and wildtype (WT) mice were maintained under 12-hour cyclic dark/light conditions.

Cloning of mGCAP1 Genomic Constructs
A 14,832 bp mouse GCAP1 genomic sequence (mG1) was modified to generate three transgenes which expressed either wildtype GCAP1 fused to EGFP (G1-GFP), or one of two mutant proteins, G1(L151F) and G1(L151F)-GFP. To generate a GFP fusion mGCAP1 transgenic construct, G1-GFP, a chloramphenicol-resistant cassette flanked by AsisI and AscI sites was first inserted into the wildtype mGCAP1 genomic construct right before the stop codon using a homologous recombination method, termed ET cloning [27]. We amplified by PCR the AsisI-AscI chloramphenicol-resistant cassette with primer pair, G1_CmRa-sisIF: 59-GCGAACACGAGGAGGCAGGCACCGGC-  GACCTGGCAGCGGAGGCTGCGGGTGCGATCGCagcattacacgtcttgagcgattgt, and G1_CmRascIR: 59-AC-CGCACGGGGCCAGCCCTCAGCAGGCAGAAGCCA-CAGGGTGAATGCTCAGGCGCGCCCacttaacggctgacatgggaatta. Relevant to primer design are regions homologous to 59 and 39 flanking sequences (50-bp, shown in black) of the mGCAP1 stop codon (bold-faced) in the transgene, AsisI and AscI restriction sites (underlined), and regions complementary to the 59 and 39 sequences of a chloramphenicol-resistant cassette (lower case), respectively. The purified PCR product of the chloramphenicolresistant cassette was co-electroporated with the mGCAP1 transgenic construct (containing an ampicillin-resistant cassette) into competent cells containing an inducible Red recombinase. Ampicillin and chloramphenicol 'double'-resistant colonies were selected, in which a Red homologous recombination occurred and indicated that the chloramphenicol-resistant cassette with flanking AsisI and AscI sites was inserted immediately before the mGCAP1 stop codon. Subsequently, we replaced the chloramphenicolresistant cassette in the recombinant mGCAP1 genomic construct with an AsisI-AscI EGFP cassette.

Generation of Genomic Transgenic Mice
The three transgenes were released from pBSKS (+) vector by NotI digestion and microinjected into the pronuclei of fertilized mouse FVB/N oocytes to produce transgenic mice at the University of Utah core facility. Transgenic mice were identified by PCR genotyping with primers G1T_F: 59-ATAGGGCGTC-GACTCGATCACGCAGC, and G1T_R:59-TAAGGGCG-GAAGATCACGGAGGTAGC). The primers are located across the 39 boundary of the transgene and the pBSKS (+) backbone. A diagnostic fragment of 540-bp discriminates the transgenes from the endogenous GCAP1 gene. The C to T point mutation (L151F) was confirmed by DNA sequencing in the G1(L151F) and G1(L151F)-GFP transgenic founder mice. Transgenic mice were outbred to C57BL/J mice.

Immunocytochemistry
Mouse eyes were dissected and immediately immersion-fixed with 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 2 hours on ice. After removal of the anterior segment, the eyecups were equilibrated sequentially with 15% and 30% sucrose in phosphate buffer for cryoprotection, and then embedded in OCT. Subsequently, retina cryosections of 12 mm thickness were cut and used for direct fluorescence confocal microscopy or immunohistochemical analysis. For immunohistochemistry, retina sections were incubated with primary antibody (UW101, 1:2,000, 4uC) overnight, followed by fluorescence-conjugated secondary antibody (1:300, 21uC) for 1 hour in a humidified chamber. Sections were washed with phosphate buffer for 5 minutes (X3) between incubations. To label nuclei, 49,6-diamidino-2-phenylindole (DAPI, 1:5,000) was added to the solution containing secondary antibody; retina sections were incubated with DAPI for one hour if imaging by direct fluorescence microscopy. After applying a drop of anti-fade agent, the sections were imaged with an Olympus Fluoview (model FV 1000) inverted confocal microscope.

Histology
Mouse eyeballs were fixed by immersion in 2% glutaraldehyde-1% paraformaldehyde in 0.1 M cacodylate, pH 7.4, overnight at 4uC. Following removal of the anterior segment, the eyecups were postfixed in 1% osmium oxide in the same buffer for 1 hour, dehydrated with an ascending series of ethanols, infiltrated with ethanol:plastic mixtures and embedded in Spurr's resin (Ted Pella, Inc., Redding, CA). Retina sections (1 mm) passing through the optic nerve were cut with an ultramicrotome and contrasted with Richardson's stain. Bright-field images of retina histology were acquired using a Zeiss Axiovert 200 (Carl Zeiss Inc., Thornwood,  Retinal morphology of the three GCAP1 transgenic lines and age-matched controls were analyzed with plastic retinal sections at ages 2, 9 and 12 months (A). Retinas from G1 (L151F) and G1(L151F)-GFP mutant transgenic mice show no morphological change at age 2 months (top, left two panels) compared to control retinas (top, right two panels). At 9 months of age, a thinned photoreceptor layer (including OS, IS and ONL) becomes evident with ,10% reduction (middle, left two panels). Photoreceptor layer thinning became more obvious at 12 months with ,20-30% reduction (bottom, left two panels). In contrast, G1-GFP and age-matched wildtype photoreceptor layers are normal at all three ages (bottom, right two panels). Abbreviations: OS, outer segment; IS, inner segment; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, the inner nuclear layer. B. Quantitation of the photoreceptor outer nuclear layer (ONL) thickness in three GCAP1 transgenic mice and age-matched controls at 12 months of age. ONL thickness was measured every 250 mm moving from the optic nerve head (ONH) to the periphery, both inferiorly and superiorly. ONL thickness was reduced in retinas of G1(L151F) and G1(L151F)-GFP transgenic mice, but not in G1-GFP transgenic or wildtype mice. (n = 3 each group). doi:10.1371/journal.pone.0057676.g004 NY) microscope and 62x objective. Images across entire retina sections were acquired with a 20x objective and imported into image J software with scaling of 3.8 pixels/mm. Outer nuclear layer (ONL) thicknesses were measured at 250 mm intervals from the optic nerve head.

Electroretinography (ERG)
Scotopic and photopic ERG responses of control and transgenic mice were recorded as described [4]. Mice were dark-adapted overnight, anesthetized by intraperitoneal injection of 10 mg/ml ketamine-1 mg/ml xylazine (10 ml/g body weight) in PBS under dim red illumination, and kept comfortable on a heating pad. After dilating pupils by applying a drop of 2.5% phenylephrine (Akorn, Inc., Decatur, IL), ERG responses were recorded from 3-5 mice of each genotype per time point using a UTAS E-3000 system (LKC Technologies,Inc., Gaithersburg, MD). For scotopic ERG, mice were tested at intensities ranging from 240 decibels (db) (23.4 logcds m -2 ) to 25 db (2.9 log cds m -2 ) without initial rod saturation. For photopic ERG, a rod saturating background light of 10 db (1.48 log cds m -2 ) was applied for 20 minutes before and during recording. Single flash responses were usually recorded at stimulus intensities of 210 db (20.6 log cds m -2 ) to 25 db (2.9 log cds m -2 ). Five or fewer flashes were averaged per intensity level, with longer flash intervals with increasing intensity. Tripleantibiotic ointment was routinely placed on the eye to prevent infection after ERG testing.

OptoMotry
Transgenic and control mice were tested for spatial visual acuity using an Optomotry apparatus which facilitates rapid screening of functional vision [28]. Briefly, mice were placed singly on a platform centered within a quad-square formed by four inward-facing computer screens, and observed by an overhead video camera. Vertical sine-wave gratings were presented at  photopic luminance levels on screens rotating 12 deg/sec (speed) either clockwise or counterclockwise, as determined randomly by OptoMotryß software. The mouse tracks the grating with reflexive head movements when the rotating grating is perceptible. By observing mouse tracking, spatial frequency thresholds were quantified by increasing the spatial frequency of the rotating grating (usually between 0.03-0.35 grating cycles/degree) until finding a maximum frequency at which the mice track.

Constructs Expressing Mouse GCAP1-EGFP and GCAP1_shRNAin vitro
Using shRNA prediction tools provided online by Whitehead, Ambion, Invitrogen and Dharmacon, we designed four putative shRNAs (mG1hp1-4) specifically targeting mouse GCAP1 and a mismatch control (mG1hp4m2) carrying the central twonucleotide mutation in the mG1hp4 guide strand (GG to CC).
To avoid possible off-target effects, an NCBI BLAST homology search was performed for the candidate shRNAs and no significant homology with other mouse genes or sequences was identified. A short-hairpin RNA (shRNA) expression vector, pmC_hH1, containing a shRNA expression promoter (human H1 promoter, hH1) and a CMV promoter-driven mCherry reporter, was used to generate anti-mGCAP1 shRNA constructs expressing mG1hp1-4 and mG1hp4m2 [4]. The construct mG1-EGFP, expressing EGFP fusion mouse GCAP1, was generated by cloning the mouse GCAP1 cDNA into pEGFP-N1 vector (Clontech, Mountain View, CA) downstream of the CMV promoter.

Knockdown Efficiency Assay in Cell Culture
HEK293 cells (ATCC, CRL-1573) were cultured in DMEM supplemented with 10% fetal calf serum and 100 u/ml of penicillin-streptomycin (Invitrogen, Carlsbad, CA). In 12-well plates, HEK293 cells were cotransfected with mG1-EGFP plasmids and anti-mGCAP1 shRNA expression plasmids, mG1hp1-4 or mG1hp4m2, at different mass ratios (1:3-3:1) with a total amount of 1.5 mg/well (normalized by pBS vector DNA). At 48 hours post-transfection, expression of mG1-GFP and mCherry reporter in the transfected HEK293 cells was directly analyzed by live cell fluorescence microscopy using a Zeiss Axiovert 200 microscope. At 50 hours post-transfection, the cotransfected cells were harvested to analyze protein levels of mG1-GFP by immunoblotting; b-actin was used as an endogenous control.
Preparation of scAAV2/8 Virus Expressing Anti-mGCAP1 shRNA Self-complimentary AAV2/8 (scAAV2/8) was used to express mG1hp4 and its mismatched control, mG1hp4m2, in the mouse retina. The scAAV2/8 packaging constructs were generated by cloning hH1 promoter-driven mG1hp4 and mG1hp4m2 cassettes into a shuttle vector pscAAV-CAG-mCherry, which was modified from pscAAV-CAG-hGFP (provided by W.W. Hauswirth). The two plasmid cotransfection method was used to produce scAAV2/ 8 viral particles expressing mG1hp4/4m2 in collaboration with the University of Florida [29]. Vector genome-containing viral particles were titered by real-time PCR and resuspended in a balanced salt solution (Alcon Laboratories, Fort Worth, TX) containing 0.014% Tween-20 at a concentration of ,1.0 610 12 vector genomes per milliliter (vg/ml).

Mouse Subretinal Injection
G1(L151F)-GFP mice were injected subretinally with scAAV2/ 8 on postnatal day 30 (P30) as described [4]. Briefly, after anesthesia with a ketamine-xylazine solution (see ERG protocol), 0.5% proparacaine solution (Alcon Laboratories Inc., Fort Worth, TX) was applied to the cornea as a topical anesthetic. A small puncture hole was made through the cornea with a 30K -gauge beveled needle (Becton Dickinson & Company, Franklin Lakes, NJ). A 33-gauge blunt needle attached to a microliter syringe (Hamilton Bonaduz, Switzerland) was introduced tangentially through the hole, and 1 ml of scAAV2/8 virus was delivered subretinally. Following injection, a triple-antibiotic ointment containing bacitracin, neomycin sulfate and polymyxin B (Taro Pharmaceuticals, Inc., Hawthorne, NY) was placed on the eye to prevent infection. Injections were performed in left eyes only, leaving the right eyes as untreated controls.

In-vivo Fluorescence Imaging of the Mouse Retina
Fluorescence live imaging of mouse retinas was used to detect the expression of GFP fusion GCAP1 transgenes, G1-GFP and G1(L151F)-GFP, in mice 1-2 months of age, as well as in the knockdown of mG1(L151F)-GFP in transgenic mice one-month posttreatment with scAAV2/8. Mice were anesthetized with isoflurane gas (2%-3%) and their pupils were dilated with 2.5% phenylephrine (Akorn, Inc., Decatur, IL). After applying 0.5% proparacaine solution (Alcon Laboratories Inc., Fort Worth, TX) to the cornea as topical anesthetic, a glass coverslip was placed on the mouse eye to maintain corneal moisture and reduce refractive error. The mouse was stabilized during imaging on a customdesigned heated aluminum stage specifically fit for the microscope (Zeiss LSM 700; Carl Zeiss, Inc., Thornwood, NY). Fluorescent fundus images were acquired with a 5x air objective and Zeiss image acquisition software. Triple-antibiotic ointment was placed on the eye to prevent infection after imaging.

Statistical Data Analysis
ONL thicknesses and the ERG response amplitudes were analyzed by two-way ANOVA, while the Optomotry data was analyzed by single-way ANOVA. The level of statistical significance was set at p = 0.05. Figure 7. Dose-dependent and sequence-specific knockdown of mGCAP1 by mG1hp4. A-D. Fluorescence microscopy of HEK293 cells 48 hrs after cotransfection with mGCAP1-EGFP and mCherry_mG1hp4 (A, B) or its mismatched control mCherry_mG1hp4m2 (C, D) at five different mass ratios. GFP signal in row A is gradually diminished as mG1hp4 increases to a saturation level, indicating knockdown dose-dependency from 1:3 to 2:1. In row C there is no obvious change of GFP signal in cultures from 1:3 to 2:1 since the mutant mCherry_mG1hp4m2 is inactive for suppression of mGCAP1. Panels of rows B and D show increase in mCherry signal corresponding to increased mass ratios, left to right. E. Representative western blot evaluating mGCAP1-GFP protein levels in the cotransfected cells (50-hr post-transfection); mGCAP1-GFP was detected by anti-GCAP1 antibody, UW101, with b-actin immunostaining as the loading control. F. Quantification of relative mGCAP1-GFP levels shown in (E). The intensities of the mG1-GFP specific bands in western blot were measured by using ImageJ, and normalized against b-actin. The relative level of mG1-GFP in each cotransfected cell sample was averaged from three tests, representing the knockdown efficiency of each shRNA at different concentrations. doi:10.1371/journal.pone.0057676.g007

Generation of Transgenic Mice and Expression of Mutant GCAP1
We used the mouse Guca1a gene as a foundation to construct three transgenes. The L151F mutation was introduced in exon 4 to establish two GCAP1 transgenic mouse lines, G1(L151F) ( Fig. 2A) and G1(L151F)-GFP (Fig. 2B). The GCAP1-EGFP fusion allowed detection of transgene expression by live fluorescence microscopy, and distinction of transgenic GCAP1 (50 kDa) from native GCAP1 (23 kDa). A third line expressed a GCAP1-EGFP fusion protein containing L151 (no L151 mutation) as a WT control (Fig. 2C). The C-terminal in-frame fusion with EGFP was achieved using the homologous recombination method, Red/ET recombination cloning, which is a DNA manipulation technique that is independent of the presence of restriction sites and the size of the DNA molecule to be modified [27] (see Methods). Primers located near the 39-end of the transgene and in the multiple cloning site of the pBSKS(+) backbone allowed precise genotyping of transgenic mice (Fig. 2D). The L151F mutation in G1(L151F) and G1(L151F)-GFP transgenic mice was confirmed by sequencing (Fig. 2E).

Subcellular Distribution of Transgenic GCAP1 Proteins
We next analyzed the subcellular distribution of G1(L151F) by immunohistochemistry and confocal microscopy. When expressed on the GCAP1/GCAP2 null background, GCAP1(L151F) localized to the photoreceptor inner and outer segments and exhibited a stronger signal in cones than in rods (Fig. 3A), comparable to endogenous GCAP1 (Fig. 3B) [5]. As predicted, the GCAP double knockout mouse retina was negative for GCAP1 immunolabeling (Fig. 3C). Direct fluorescence microscopy of retinal cryosections of G1-GFP and G1(L151F)-GFP transgenic mice, contrasted with diamidino-2-phenylindole (DAPI) nuclear stain, revealed that both G1(L151F)-EGFP and G1-EGFP transgene products were expressed in the photoreceptor inner and outer segments (Fig. 3D, E). Fluorescence attributable to transgenic wildtype and mutant GCAP1s was observed in photoreceptors residing in the most sclerad three layers of the ONL, i.e., cones, where the perinuclear accumulation is most likely caused by 'overexpression' when the transgene product exists on a background of endogenous (two alleles) GCAP1. Expression of G1(L151F)-GFP and G1-EGFP in transgenic mouse retinas could be discerned easily by in-vivo fluorescence (Fig. 3F, G).

Slowly Progressive Photoreceptor Degeneration in GCAP1 Mutant Transgenic Mice
The onset of retinal degeneration was determined by histology, electroretinography and behavioral testing of live animals. Transgenic G1(L151F), G1-GFP and age-matched wildtype mouse retinas were indistinguishable early (Fig. 4A, top panels at 2 months). Further, all three GCAP1 transgenic lines showed normal scotopic and photopic ERG responses at two months (data not shown). However, slight reduction of scotopic (a-wave) and photopic (b-wave) ERG amplitudes in 9 month-old mutant transgenic mice signaled the onset of photoreceptor degeneration, and this was reflected by a modest thinning of the ONL (Fig. 4A, middle panels). Degeneration in both mutant mouse models progressed slowly, revealing ,10-20% reduction of ONL thickness compared to wildtype controls and was particularly obvious in the inferior retina (Fig. 4A, bottom panels at 12 months; Fig. 4B, Table 1).

Identification of Anti-mGCAP1 shRNAs in vitro
Germline deletion of GCAPs in mouse display neither gross functional deficiencies nor retinal degeneration [8,9]. A nonallelespecific knockdown of Guca1a mRNA in the G1(L151F) transgenic mice may therefore be a promising tool for RNAi-based gene therapy to cure dominant CORD caused by GCAP1 mutations. Using online shRNA prediction tools, we designed four anti-mGCAP1 candidate shRNAs which have no significant homology with other mouse genes. To identify an effective anti-mGCAP1 shRNA, we developed an in-vitro shRNA screening system in which disappearance of mGCAP1-EGFP fluorescence would be proportional to the knockdown efficiencies of shRNA candidates [4]. We used a shRNA expression construct with a mCherry reporter (Fig. 6A) to test four anti-mGCAP1 shRNA candidates, mG1hp1-4 ( Fig. 6B, C), in which shRNA transcription is controlled by the human H1 promoter. Target mG1-EGFP plasmids were co-transfected into HEK293 cells with each anti-mGCAP1shRNA plasmid at a 1:3 mass ratio. Live cell imaging revealed that mG1hp2 and 4 significantly suppressed mG1-EGFP expression in the co-transfected cells whereas mG1hp1 or 3 were largely ineffective (Fig. 6D). Western blotting of the cell lysates verified that mG1hp2 and 4 knocked down mG1-EGFP with 60% and 70% efficiency, respectively (Fig. 6E, F).

Visualization of shRNA-mediated Gene Silencing in-vivo
We packaged mG1hp4 and mG1hp4m2 into scAAV2/8 vectors that express shRNA rapidly and persistently in mouse photoreceptors [4]. The scAAV8-mG1hp4 virus was injected subretinally into the left eyes of G1(L151F)-GFP mice at 1 month of age, before appreciable photoreceptor degeneration has occurred. At one week post-injection, mice exhibiting mCherry expression in .50% of the retinal area (visualized by live fluorescence imaging) were retained for further analysis. Fluorescence fundus photo-graphs of both treated and untreated eyes were taken from the G1(L151F)-GFP transgenic mice one month after the injection. In treated left eyes, the extent of mCherry expression covered nearly the whole retina suggesting successful and widespread vectormediated expression (Fig. 8B). In those retinas expression of transgenic G1(L151F)-GFP was significantly suppressed (Fig. 8A) compared to that seen in untreated right eyes (Fig. 7C). In contrast, scAAV8-mG1hp4m2 treatment in the left eyes had no detectable effect (Fig. 8D-F).

Discussion
The purpose of this study was to generate mouse models of dominant CORD based on a GCAP1(L151F) mutation associated with human disease. The mouse models serve as tools to explore the efficiency of a nonallele-specific knockdown strategy suitable for human gene therapy. The transgenes consisted of a 15 kb Guca1a genomic fragment containing the native regulatory elements including endogenous promoter, all introns and endogenous polyadenylation sites. In the GCAP1-EGFP and GCAP1(L151)-EGFP fusion models, the coding region of exon 4 of the GCAP1 gene was fused to EGFP cDNA (Fig. 2B, C). Fusion of EGFP to GCAP1 at its C-terminus provided means to directly visualize GCAP1 localization by live fluorescence retina imaging (Fig. 3) and to monitor the suppression of GCAP1 by RNA interference (Fig. 8). The transgenes expressed wildtype and mutant GCAP1s in mouse photoreceptors at a level similar to endogenous GCAP1 in heterozygous mice (GCAP +/2 ). No ectopic transgene expression was observed apart from accumulation of transgenic GCAP1 in perinuclear locales of a few individual cones (Fig. 3D).
In human patients carrying GCAP1 mutations, phenotypes are noticed typically within the second or third decade and progress slowly [2]. In transgenic mice expressing GCAP1(L151F), we observed reduced thickness of the photoreceptor layer as early as 9 months of age (Fig. 4). At 12 months of age, reduction of photopic ERG b-wave amplitudes was more pronounced than that of scotopic a-waves suggesting that cones degenerate somewhat faster than rods in these mouse models (Fig. 5). As mouse cones express GCAP1 only, a more severe L151F dominant negative effect is expected in these cells. Another, more severe, EF4-hand mutation associated with CORD (E155G) causes reductions in cone function in patients starting at an average age of 16 years, about 10 years earlier than for L151F patients [16,17,23]. Consistent with this more aggressive degenerative human phenotype, far greater functional deficits in cones than rods were observed in the E155G knock-in mouse model [26]. Starting at 3 months postnatally, photopic ERG responses of E155G knock-in mice were reduced and the flicker response was severely depressed. At 12 months, the photopic ERG b-wave was reduced to 42% of normal, compared to 60% in our L151F transgenic mice. The presence of two normal Guca1a alleles in our L151F model versus just one in the E155G knock-in model may account for the phenotype differences. Further, the E155G mutation in which an essential charged residue is replaced by a neutral one represents a much more severe EF-hand structural change than seen in L151F. Nevertheless, our 'genomic transgenic' mouse yields phenotypes comparable to the 'knock-in' mouse suggesting that the 15 kb genomic fragment harboring Guca1a contains all the essential regulatory elements.
Functional differences in the transgenic mice at one year of age relative to age-matched wildtype mice are relatively mild (Figs. 4,5), similar to those of second and third decade GCAP1(L151F) patients. The GCAP1(L151F) mutation was identified in two unrelated, five-generation families affected with cone and cone-rod dystrophy [17,23]. In young patients, rod and cone ERGs were near normal. Loss of cone function occurs first and rod function generally persists much longer in CORD families. Although the disease phenotypes were variable among the affected members in both families, most patients experienced late-onset cone dysfunction by the second and third decade of life. Another important CORD mutation is N104K in EF-hand 3 which we investigated previously [20]. A proband carrying a GCAP1(N104K) mutation and diagnosed by fundoscopy with dominant CORD at 39 years, experienced a slow loss of acuity over the course of 12 years [20]. Considering differences in human and mouse lifetimes, our lateonset and slowly progressive transgenic mouse models reveal very similar features and compare well with the human disease. The subtle deviations may be explained by the presence of two normal Guca1a alleles in our transgenic mice and by differences in human and mouse retinal anatomy, as the mouse has neither macula nor cone-rich central fovea.
The principal goal of this research is to identify gene therapy vectors that may be used successfully to delay the onset of cone degeneration, and/or cure CORD disease. In an earlier study, we demonstrated the feasibility of shRNA knockdown using an allelespecific approach in a retinitis pigmentosa mouse model carrying the GCAP1(Y99C) mutation [4]. In that mouse line, a GCA-P1(Y99C) cDNA was expressed under the control of the mouse opsin promoter producing a retinitis pigmentosa like phenotype [25]. A therapeutic recombinant AAV robustly and persistently expressed shRNAs in photoreceptors at one week post-injection and silenced the disease-causing bovine GCAP1(Y99C) transgene with ,80% efficiency; the gene silencing was effective for nearly one year without apparent off-target interference and significantly improved rod photoreceptor survival, delayed disease onset and increased visual function [4].
Germline deletion of GCAPs in mouse delays recovery to the dark state by seconds as guanylate cyclase activity is not accelerated in low Ca 2+ [8]. However, there are no detrimental morphological consequences, nor have gross functional deficiencies or retinal degeneration been reported. Patients with GCAP1 null or GCAP1/GCAP2 null mutations have not been identified to date, perhaps because a clear degeneration phenotype is absent [30]. We therefore reasoned that a nonallele-specific shRNA knockdown of both wildtype and mutant GCAP1s could be a general therapeutic strategy to rescue the dominant degeneration caused by any of the eight known EF-hand GCAP1 mutations. The most effective anti-mGCAP1 shRNA (mG1hp4) which degraded more than 70% of mGCAP1-EGFP was identified by an in-vitro fluorescent screening assay (Fig. 6). The ''negative control'' shRNA that carried a GG to CC mutation in the center was inactive (Fig. 7). After packaging into scAAV2/8vector, the mG1hp4 proved to be a potent GCAP1 knockdown agent in-vivo as well, as shown by live fluorescence imaging (Fig. 8). One month post-injection with AAV8-mG1hp4, we observed significant suppression of GCAP1(L151F)-GFP by 70% in the treated eyes. As expected, knockdown of mutant GCAP1 occurred concomitantly with knockdown of endogenous GCAP1 (,90%) (Fig. 8J, K). As mG1hp4 was designed to specifically target mouse GCAP1 without a C-terminal GFP, AAV-mG1hp4 has shown more efficient suppression of endogenous GCAP1 gene than the GCAP1(L151F)-GFP transgene. These experiments provide proof of principle that RNAi suppression of both wildtype and mutant GCAP1 may be a potent therapeutic strategy, applicable to all GCAP1 mutations of EF-hands 3 and 4, as long as the guide strand of shRNA is located outside the disease-causing mutations. Notably, when shRNA reaches its highest suppression of the target gene, overexpression does not increase efficiency but generates offtarget silencing effects on other genes due to saturation of the intracellular RNAi machinery [31]. Thus, identification of a highly efficient shRNA may require its titration to find the lowest level efficacious for RNAi-based gene therapy in humans.