Optimization of HITI-Mediated Gene Insertion for Rhodopsin and Peripherin-2 in Mouse Rod Photoreceptors: Targeting Dominant Retinitis Pigmentosa

Abstract

Purpose: Among the genome-editing methods for repairing disease-causing mutations resulting in autosomal dominant retinitis pigmentosa, homology-independent targeted integration (HITI)-mediated gene insertion of the normal form of the causative gene is useful because it allows the development of mutation-agnostic therapeutic products. In this study, we aimed for the rapid optimization and validation of HITI-treatment gene constructs of this approach in developing HITI-treatment constructs for various causative target genes in mouse models of retinal degeneration.

Methods: We constructed the Cas9-driven HITI gene cassettes in plasmid vectors to treat the mouse Rho gene. A workflow utilizing in vivo electroporation was established to validate the efficacy of these constructs. Single-cell genotyping was conducted to evaluate allelic donor gene insertion. The therapeutic potency of HITI-treatment plasmid and adeno-associated virus (AAV) vectors was examined by section immunohistochemistry and optomotor response (OMR) in Rho^+/P23H mutant mice. We also targeted mouse Prph2 to examine the workflow.

Results: The optimized HITI-treatment constructs for mouse Rho genes achieved gene insertion in 80% to 90% of transduced mouse rod photoreceptor cells. This construct effectively suppressed degeneration and induced visual restoration in mutant mice. HITI-treatment constructs for the Rhodopsin gene demonstrated efficacy in AAV vectors and are adaptable for the mouse Prph2 gene locus.

Conclusions: The study showcases a workflow for the rapid optimization and validation of highly effective HITI-treatment gene constructs against dominant-negative inheritance in inherited retinal dystrophy. These findings suggest the potential utility of this approach in developing HITI-treatment constructs for various target genes, advancing gene therapy products for diverse genetic disorders.

Retinitis pigmentosa (RP) is the most prevalent form of inherited retinal dystrophy (IRD), and has a prevalence of 1 in 3000 to 4000 worldwide.^1–4 RP affects primarily the outer retinal layer and causes gradual degeneration of the photoreceptor cells.^1,5,6 Genetic diagnoses have uncovered more than 250 causative genes associated with IRD (see RetNet; http://sph.uth.tmc.edu/RetNet/ provided in the public domain by the University of Texas Houston Health Science Center, Houston, TX, USA). These genes play pivotal roles in the function of photoreceptors and cells of the retinal pigment epithelium (RPE), and the mutant gene expression leads to impairment and degeneration.⁷ 

To treat patients with RP, several gene-targeted therapeutics have been proposed and tested,^8–11 depending on the inheritance patterns of disease-causing mutations.⁷ Recessive mutations occur when both alleles of a gene are mutated. In these cases, the gene or protein expressed by the mutated gene allele is nonfunctional but does not influence the function of the wild-type (WT) gene allele in heterozygous carriers. In individuals harboring a recessive mutation, gene therapy, also known as replacement therapy, is frequently used to introduce a normal gene from an external source.¹² On the other hand, dominant mutations constitute a type of inheritance (heterozygous) in which a mutation in one of the two alleles precipitates disease development. Symptoms resulting from these mutations emerge in two primary ways: haploinsufficiency and dominant inhibition. Haploinsufficiency arises when the gene product expressed from one of the remaining alleles cannot sustain the original function, which thereby compromises the overall function because of reduced gene product levels. Dominant inhibition involves inhibition of the function of the WT gene or protein expressed by a mutated gene allele, which thereby disrupts normal cellular processes. Gene therapy for diseases caused by dominant mutations requires the supplementation of normal genes and repair of causal gene mutations.¹³ 

Genome-editing technology using gene-editing tools^14–16 is used to repair genomic mutations that cause dominant inhibition and is therefore expected to provide fundamental treatment.^17,18 The homology-independent targeted integration (HITI) method is a gene-editing technique that allows for the insertion of normal genes into specific genomic loci.^19,20 When a normal full-length sequence is introduced as a donor gene into a mutated target allele, the normal gene is expressed instead of the mutated gene. This method is considered to be “mutation agnostic” by enabling the same genome-editing target and donor cassette to treat all mutations, including novel mutations.^21,22 However, there is a paucity of studies reporting on the development and validation of highly efficient HITI-treatment constructs specifically designed for targeted gene insertion in rod photoreceptor cells. Such constructs would ideally stop mutant gene expression while ensuring the inserted normal gene is expressed at levels comparable to the endogenous allele in treated photoreceptors. This gap in the literature underscores the need for our current study. 

Here, we report an efficient method for validating gene constructs for HITI-mediated exogenous gene insertion. We focused on targeting the rhodopsin (RHO) gene, the leading cause of autosomal dominant retinitis pigmentosa (AdRP).^23,24 Using Streptococcus pyogenes Cas9 (SpCas9) as a gene-editing tool and in vivo electroporation as a gene-transfer technique, we identified SpCas9 guide RNA (gRNA) target sequences with high insertion efficiency in vivo, although the efficiency did not correlate fully with the scores obtained from the in silico gRNA finders. The inserted normal genes from the donor vectors expressed with chimeric introns (intron of human beta globin and IgG) and the 3′-untranslated region (UTR) of the target genes showed therapeutic efficacy in Rho-mutant mice. Importantly, these HITI constructs maintained their efficacy when transferred to adeno-associated virus (AAV) vectors, demonstrating the versatility of our approach. Furthermore, we applied the same methodology to develop HITI-treatment constructs targeting the mouse Prph2 gene, which also proved effective. These results suggest that our workflow for constructing HITI-mediated gene insertions may be broadly applicable for various forms of RP. 

Experimental Procedures

Plasmids

The details of the DNA constructs used in this study are listed in Supplementary Table S1. 

Mice

All mouse experiments were conducted with approval from the RIKEN Center for Developmental Biology Ethics Committee (No. AH18-05-23). The following mouse strains were used in this study: CD1, C57BL/6J, Rho^P23H/P23H (homozygous), Rho^+/P23H (heterozygous), and Rho^+/AcGFP knock-in (KI) mice. Timed pregnant CD1 (Charles River Laboratories, Japan) and C57BL/6J (CLEA, Japan) mice were maintained and provided by the Laboratory for Animal Resources and Genetic Engineering (RIKEN Biosystems Dynamics Research [BDR]). Rho^P23H mutant mice were purchased from Jackson Laboratory. Rho^+/P23H heterozygous mice were obtained by crossing Rho^P23H/P23H homozygous mice with WT C57BL/6J mice. Rho^AcGFP KI mice were generated by the Laboratory for Animal Resources and Genetic Engineering, RIKEN BDR, according to a previously reported procedure.²⁵ Briefly, the donor vector had 5′- (1000 bp) and 3′- (1000 bp) homology arms for intron 2 of the mouse Rho locus and an expression cassette containing an adenovirus splicing acceptor, stop codons, an internal ribosome entry site, AcGFP, and a rabbit beta-globin poly(A) sequence. The crRNAs for intron 2 (mRhoIntron2-crRNA; 5′-UAG AGA GCA UUG CCG UUA CUG UUU UAG AGC UAU GCU GUU UUG-3′) and tracrRNA (5′-AAA CAG CAU AGC AAG UUA AAA UAA GGC UAG UCC GUU AUC AAC UUG AAA AAG UGG CAC CGA GUC GGU GCU-3′) were purchased from FASMAC (Japan). Zygotes generated by in vitro fertilization were microinjected with 100 ng/µL Cas9 protein (A36497; Thermo Fisher Scientific), 50 ng/µL mRhoIntron2-crRNA, 100 ng/µL tracrRNA, and 10 ng/µL donor vector. 

Single-Strand Annealing Assay

HEK293T cells (RCB2202; RIKEN BRC) were maintained in Dulbecco's modified Eagle's medium (DMEM; 043-30085; Fujifilm Wako) supplemented with 10% fetal bovine serum (FBS; 553-36315; BioSera). Plasmids of 100 ng of pCAG-SpCas9, 100 ng of pBAsi-U6-mRho-gRNA, 300 ng of pCAG-EGxxFP, and 100 ng of pCAG-mCherry were transfected into 80% confluent HEK293T cells in 24-well plates using FuGene6 (E2693; Roche), according to the manufacturer's protocol. After 48 hours of incubation, EGFP and mCherry expression was observed using a fluorescence microscope, and cleavage efficiency was evaluated by EGFP:mCherry ratios calculated from Pixel intensity values of GFP-positive and mCherry-positive areas measured by ImageJ. 

In Vivo Electroporation and Eyecup Preparation

In vivo electroporation was performed at postnatal day (P)0 on CD1 WT mice, Rho^P23H/P23H mutant mice, and Rho^+/AcGFP KI mice, as previously described with minor modifications.²⁶ In brief, 0.4 µL of plasmid DNA solutions of 0.16 µg/µL pRho2k/300bp-SpCas9, 0.16 µg/µL pLeaklessIII/pAAV donor cassettes, 0.06 µg/µL pBAsi-U6-mRho/mPrph2-gRNAs, and 0.02 µg/µL pCAG-mCherry were injected into the subretinal space of P0 mouse pups with a 33G blunt-ended microsyringe (MS*E05 33G 30 mm 90 degrees; ITO Corporation) and transferred into the retina by an NEPA21 Super Electroporator (Nepagene) with tweezer-type electrodes (CUY650P10; Nepagene). The parameters were as follows: voltage = 120 V; pulse length = 30 ms; pulse interval = 470 ms; number of pulses = 3; decay rate = 10%; polarity + as poring pulse; and voltage = 20 V; pulse length = 50 ms; pulse interval = 50 ms; number of pulses = 3; decay rate = 40%; and polarity + as transfer pulse. Electroporated retinas were harvested at P14, P21, P50, and/or P56 for immunohistochemical (IHC) analysis. 

AAV Construction and Subretinal Injection

The pAAV or pscAAV plasmids carrying the SpCas9, donor cassettes (which include AcGFP as a marker for successful HITI events), or gRNAs listed in Supplementary Table S1 were constructed. We deliberately chose not to include a separate infection marker (such as mCherry) to maximize the infection efficiency of our therapeutic constructs. All AAV constructs were packaged into AAV8 serotypes^20,27 by the Gene Transfer, Targeting, and Therapeutics Viral Vector Core of the Salk Institute or VectorBuilder. Subretinal injection of AAVs was performed as previously described with minor modifications.²⁸ In brief, 1- or 2-month-old C57BL/6J and Rho^+/P23H mice were anesthetized with ketamine (7.7 mg/100 g; Daiichi Sankyo) and xylazine (0.92 mg/100 g; Bayer), and 1 µL of 2.0 × 10¹² vg/mL AAV solution of 0.8 × 10¹² vg/mL AAV8-300bp-SpCas9, 0.8 × 10¹² vg/mL AAV8-donor cassettes, or 0.4 × 10¹² vg/mL AAV8-U6-gRNAs was injected into the subretinal space. Retinas were harvested 1 and 2 months after injection and subsequently used for whole mount and section immunohistochemical analysis to evaluate AcGFP expression. 

Section Immunohistochemistry and Retinal Flatmounts

Retinal eyecups, in which the sclera and cornea were dissected, were fixed with 4% paraformaldehyde in PBS (Nacalai Tesque, Inc.) for 1 hour at room temperature. After washing with PBS, mCherry- and/or AcGFP-positive eyecups were used for section IHC and retinal flatmounts. Immunostaining and imaging of the retinal sections²⁶ and flatmounts²⁹ were performed as previously described. Prior to immunostaining, the sections were blocked for 1 hour at room temperature with 5% heat-inactivated horse serum (26050-070; Thermo Fisher Scientific) and 0.5% Triton X-100 (Nacalai Tesque, Japan) in PBS. Primary antibodies used in this study were directed against GFP (rat, 1:1000, Nacalai Tesque, Inc., 04404-84), mCherry (rabbit, 1:1000, MBL, PM-005), and rhodopsin (mouse, 1:1000, Merck, MAB5316). Secondary antibodies used in this study were Donkey anti-Rat Alexa Fluor 488 (A21208; Thermo Fisher Scientific), Donkey anti-rabbit Alexa Fluor 594 (A21207; Thermo Fisher Scientific), and Donkey anti-mouse Alexa Fluor 594 (A21203; Thermo Fisher Scientific), all diluted 1:1000 in blocking solution. Nuclei were counterstained with DAPI (D1306; Thermo Fisher Scientific, 1:5000). Fluorescence images were acquired using a confocal microscope (LSM 700; Zeiss) and a BZ-9000 fluorescence microscope (Keyence). To count AcGFP- and mCherry-positive cells, 4 to 6 images of a 320 µm × 320 µm field were obtained using confocal microscopy. 

Single Cell Sorting and Genotyping

The P56 CD1 retinas were electroporated in vivo with plasmids containing SpCas9. Donor vectors carrying NLS-AcGFP, U6-gRNA (gRNA1), and NLS-mCherry were dissected to isolate fluorescent retinal regions and were dissociated into single cells with papain (LS003126; Worthington Biochemical) digestion, as previously described.³⁰ The dissociated cells were centrifuged, resuspended in PBS containing 0.5% bovine serum albumin (A4503; Sigma Ardrich), and filtered through a 35 µm mesh (352235; Coning). The GFP- and mCherry-double-positive cells were directly sorted into 2 µL of cell lysis buffer (VGB-301-C-50; Viagen Biotech) in a 96-well PCR plate using an SH800 cell sorter and SH800 software (version 2.1.5; Sony), according to the manufacturer's recommendations. 

Using the lysed cells as the templates, the first PCR was performed with primer pairs 5′-GCAGCAGTGGGATTAGCGTTAGTATG-3′ and 5′-AAGGGCACATAAAAATTGGGGCCCTC-3′ to amplify both the WT (350 bp) and KI (about 600 bp) fragments. Using 1:200 diluted PCR products, a second PCR (nested PCR) was performed with primer pairs to amplify the WT allele (5′-TATCTCGCGGATGCTGAATCAGCCTC-3′ and 5′-AAGGGCACATAAAAATTGGGGCCCTC-3′) and the KI allele (5′-TATCTCGCGGATGCTGAATCAGCCTC-3′ and 5′-TTCTCTGTCTCGACAAGCCCAGTTTC-3′). For direct sequencing, PCR products that were reamplified using nested PCR were collected and used. All PCR analyses were performed using PrimeStarGXL (TaKaRa) as follows: 94°C for 10 minutes and 30 cycles of 98°C for 5 seconds, 55°C for 15 seconds, and 68°C for 30 seconds. 

Quantitative Optomotor Response

A quantitative optomotor system (Phenosys GmbH, Berlin, Germany) was used to determine the visual acuity according to the manufacturer's instructions. Briefly, under an ND2 filter set for scotopic measurements, sinusoidal gratings ranging from 0.1 to 0.6 cycles per degree and at maximal contrast³¹ were displayed in randomized order for 1 minute. Data were averaged from three sessions per time point for each mouse using integrated software for the detection of head-tracking movements and projection of a virtual cylinder of revolving gratings of different spatial frequencies. Each mouse was tested for no more than 30 minutes. 

Results

Design and Validation of the Gene Constructs for Highly Efficient HITI-Mediated Insertion

Many genes responsible for RP are enriched in rod photoreceptor cells.¹³ Therefore, we attempted to establish an experimental procedure for the rapid construction and validation of a highly efficient HITI-mediated insertion of exogenous genes into rod photoreceptor cells. Because gene mutations are often contained in the first exon of coding sequences (CDS), it is desirable to insert a donor gene cassette containing the normal genes and polyadenylation (poly(A)) sequences into the 5′-UTR region. More than 100 mutations have been reported in exons 1 to 5 in of the human RHO locus.²⁴ 

First, we selected SpCas9 gRNA target sequences to introduce double-stranded DNA breaks (DSBs) into the 50 bp 5′-UTR region proximal to the Rho start codon, which is evolutionarily nonconserved (Fig. 1A). Using the gRNA Identification (grID) database, in silico SpCas9 gRNA finders³² (http://crispr.technology), we selected the 3 highest-scoring gRNAs with grID scores ranging from 600 to 800 (see Supplementary Table S1). To evaluate the cleavage efficiency of the three target sequences in vitro, we performed a single-strand annealing (SSA) assay, in which DSB efficiency was validated using SSA-mediated EGFP reconstitution of pCAG-EGxxFP plasmids carrying 5′- and 3′-EGFP fragments with the 3 gRNA target sequences (Fig. 1B). The cleavage efficiencies of gRNA1 and gRNA3 were similar, whereas gRNA2 exhibited low EGFP expression. 

Figure 1.

 

Optimization and validation of highly efficient HITI-treatment gene constructs targeting the mouse Rho locus. (A) Schematic illustration and workflow of the Cas9-driven HITI-treatment gene construction. Each donor cassette containing one of the three gRNA sequences (pink) selected from the 5′-UTR sequence of the mouse Rho locus at both ends was prepared. The targeted allele represents the gene structure when the HITI-mediated gene insertion occurs. (B) SSA assay for in vitro evaluation of the cleavage efficiency of the three gRNA sequences targeting Rho. The illustration indicates the SSA-mediated reconstruction of pCAG-EGxxFP fragments by Cas9 cleavage, and the photographs represent HEK293T cells 48 hours after transfection with pCAG-EGxxFP, pCAG-mCherry, pCAG-SpCas9, and the 3 gRNA plasmids. Numbers represent EGFP:mCherry intensity ratios measured by ImageJ. Scale bar = 500 µm. (C) Schematic illustration of HITI-mediated donor insertion by in vivo electroporation into the P0 WT (CD1) mouse retina for initial gRNA screening experiments. The plasmid cocktail injected into the subretinal space included pbRho2k-SpCas9, pLeaklessIII-Donor cassette containing one of the three gRNA sequences (mRho-gRNA1, mRho-gRNA2, or mRho-gRNA3), pBAsi-U6-gRNA, and pCAG-mCherry. The plasmid cocktail was injected into the subretinal space. (D) Immunohistochemistry sections of P21 mouse retinas electroporated in vivo at P0 with three plasmid cocktails targeting different mRho-gRNA1 (left), mRho-gRNA2 (middle), and mRho-gRNA3 (right) sequences. Electroporated cells are mCherry positive, and most cells in the outer nuclear layer (ONL) express AcGFP in retinal sections targeting mRho-gRNA1 and mRho-gRNA3 sequences. INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar = 50 µm. (E) High-power image of section immunohistochemistry with anti-rhodopsin antibody (white) of the square area shown in Figure 1D. Arrowheads indicate AcGFP-expressing cells. (F) Fraction (Frac.) of AcGFP-positive cells in mCherry-positive ONL cells from three different retinal tissue sections. Numbers, the mean ± SD (n = 6); N.D., not detected. (G) Schematic illustration of HITI-mediated donor insertion by in vivo electroporation into the P0 WT (CD1) mouse retina for experiments comparing different promoter and vector combinations. The plasmid cocktail injected into the subretinal space included pbRho2k-SpCas9 or pbRho300bp-SpCas9-synthetic pA, pLeaklessIII-Donor cassette or pAAV-Donor cassette, pBAsi-U6-gRNA, and pCAG-mCherry. (H) Representative fluorescence images of P21 mouse eyecups electroporated in vivo at P0 with 3 plasmid combinations: (left) pbRho2k-SpCas9 with pLeaklessIII-Donor cassette, (middle) pbRho300bp-SpCas9-synthetic pA with pLeaklessIII-Donor cassette, and (right) pbRho2k-SpCas9 with pAAV-Donor cassette. AcGFP (green) indicates successful HITI-mediated gene insertion, whereas mCherry (red) serves as an electroporation marker. Dashed circles outline individual eyecups. Scale bar = 1 mm.

Next, donor plasmids were designed. Rhodopsin accounts for 95% of the proteins constituting the outer segments,³³ and it is important to maintain high expression levels. Previously, the use of chimeric introns^34–36 was effective in increasing exogenous gene expression at the Rho locus in KI mice.^37,38 A poly(A) sequence derived from the mouse Rho locus was also used. Analysis of mouse Rho mRNA revealed the presence of multiple poly(A) signals in the 3′-UTR, which suggested the stability of and increase in Rho mRNA transcription.³⁹ The reverse sequence of each gRNA sequence was added to both ends of the gene cassette and inserted into pLeaklessIII, a donor plasmid vector developed to prevent leaky transcription from the plasmid backbone.^40,41 

HITI-Mediated Gene Insertion Observed in 80% to 90% of Electroporated Rod Photoreceptor Cells

We performed in vivo electroporation to rapidly evaluate the in vivo insertion efficiency of HITI-mediated gene insertion in mouse rod photoreceptor cells. We used pbRho2k-SpCas9, in which SpCas9 was expressed from a 2 kb promoter of bovine rhodopsin⁴² in differentiated mouse rod photoreceptor cells (Fig. 1C). Plasmid vectors for gRNA expression (U6-gRNA1, 2, and 3), donor constructs, and reporter expression (pCAG-mCherry) were coelectroporated in the P0 WT CD1 mouse retina, in which the plasmid DNAs can be transferred into the nucleus because the retinal cells of the future photoreceptor layer (outer neuroblastic layer) are in an undifferentiated dividing state. After these cells differentiate into rod photoreceptor cells in a nondividing state, SpCas9 is expressed from the rhodopsin promoter. This in vivo electroporation method allowed for rapid assessment of HITI efficiency without the need for AAV production. The use of the Rho promoter ensured that SpCas9 expression was limited to differentiated photoreceptor cells. We observed that this approach resulted in successful gene insertion in post-mitotic rod photoreceptors, as evidenced by the expression patterns described below. 

In the P21 eyecups, Aequorea coerulescens GFP (AcGFP) expression was observed in gRNA1- and gRNA3-treated cells (Supplementary Fig. S1), which indicated HITI-mediated KI of the donor cassettes in the correct direction. Sections IHC of these retinas showed that AcGFP-positive cells were localized at the photoreceptor layer, whereas mCherry signals, driven by the ubiquitous CAG promoter, were observed in both the outer nuclear layer (ONL) and the inner nuclear layer (INL) where other retinal neurons were developing (Fig. 1D). The AcGFP-expressing cells were also coimmunostained with anti-rhodopsin antibodies, and the IHC images showed that mRhoCDS and AcGFP in the donor plasmids were inserted and expressed from intrinsic Rho promoters in the genome (Fig. 1E). By contrast, the eyecups electroporated with gRNA2 were mCherry positive, but AcGFP fluorescence was not detectable (see Fig. 1C, Supplementary Fig. S1), which suggested that gRNA2 was ineffective in vivo. The efficiency of each HITI-mediated donor insertion was estimated from the percentage of AcGFP-positive cells to the number of mCherry signals localized in the photoreceptor layer (Fig. 1F). The percentage was 80% to 90%, which suggested that HITI-mediated gene insertion occurred in most electroporated photoreceptor cells. Based on these results demonstrating the high efficiency of gRNA1, we selected this guide RNA for use in all subsequent experiments in this study. 

Given that the bRho2k-Cas9 cassette is about 7 kb in length, which is greater than the AAV packaging limit (typically 4.7 kb including inverted terminal repeats [ITRs]), we examined HITI efficiency using a 300 bp rhodopsin proximal promoter. For these experiments, we utilized gRNA1, which demonstrated the highest efficiency in our initial screens (as shown in Fig. 1F). The AcGFP fluorescence of P21-electroporated retinal eyecups was similar to that of the bRho2k promoter (Fig. 1G), which indicated that the 300 bp promoter was available. However, when the plasmid backbone of the donor vector was switched to pAAV instead of pLeaklessIII, AcGFP fluorescence was not detected. This result indicated that the efficiency of HITI-mediated insertion was affected by the choice of plasmid backbone when using plasmid vectors. When we switched from pLeaklessIII to pAAV, we noticed a decrease in AcGFP fluorescence, indicating reduced HITI efficiency (see Fig. 1G). This observation highlights the importance of plasmid backbone design in plasmid-based HITI experiments, but does not predict the efficiency when switching to AAV vectors. 

Monoallelic HITI-Mediated Insertion of Rod Photoreceptor Cells

HITI-mediated gene insertion occurs in either a monoallelic or a biallelic manner in each rod photoreceptor. To compare the percentages of monoallelic and biallelic insertions in electroporated WT mouse rod photoreceptor cells, we performed single-cell genotyping using dissociated HITI-treated rod photoreceptor cells. For this experiment, we added the nuclear localization signals (NLSs) to AcGFP and mCherry (Fig. 2A) because AcGFP and mCherry localized in dissociated rod outer segments increased the error in cell sorting. At P56, we collected 48 cells from the AcGFP- and mCherry-double-positive fractions directly into 96-well plates and amplified the WT and KI fragments using primers specific to both sequences (Fig. 2B). Next, we performed nested PCR using primers to amplify either the WT or KI fragments (see Fig. 2B). The cell samples in which only the KI band was amplified showed biallelic insertion, whereas those in which both the WT and KI bands were amplified were monoallelic (Fig. 2C, Supplementary Fig. S2). 

Figure 2.

 

Validation of monoallelic and biallelic insertions in HITI-treated rod photoreceptor cells. (A) Schematic illustration of HITI-treatment gene construction for cell sorting. Nuclear localization signals (NLS) were added to the fluorescent markers in the donor and reporter plasmids. (B) Schematic illustration of single-cell sorting and PCR genotyping. The fluorescent region of P56-electroporated retinas was trimmed and dissociated, and 48 cells were sorted using FACS from the GFP- and mCherry-double-positive fractions. WT and KI alleles were amplified from the first PCR, and each allele was amplified using nested PCR. (C) Electrophoresis gel images of the nested PCR of WT and KI alleles of sample numbers 13 to 17 of the 48 samples tested. (D) Direct sequencing of the reamplified WT (upper) and KI (lower) PCR products. The sequences around the mRho-gRNA1 target (purple shading) are shown. The orange shaded region indicates NHEJ-mediated deletions in the gRNA1 cleavage site. (E) Schematic illustration of the de novo validation of the donor vector insertion. (F) P56-electroporated Rho^+/AcGFP retinal sections. AcGFP-positive and -negative cells were observed in HITI-treated mCherry-positive photoreceptor cells. The white small solid-box indicates an area with mCherry-positive, AcGFP-negative cells, representing monoallelic knock-in into the AcGFP allele or biallelic knock-in. The white small dotted-line box shows an area with mCherry-positive, AcGFP-positive cells, indicating monoallelic knock-in in the WT allele. Scale bar = 50 µm. (G) Fraction (Frac.) of mCherry- and/or AcGFP-positive cells counted in retinal sections. mCherry+ only represents cells with HITI-mediated insertion only, whereas mCherry+ GFP+ indicates cells with both HITI-mediated insertion and endogenous GFP expression. Data represents means ± SD (n = 4).

We detected longer (e.g. sample no. 18 in Supplementary Fig. S2) and shorter (e.g. sample no. 16 in Fig. 2C) PCR fragments in some samples, possibly due to intermolecular annealing of the uncleaved donor plasmids. We then reamplified the PCR fragments and sequenced them (Fig. 2D). Using direct sequencing of the WT PCR fragments, we detected deletions in the gRNA1 cleavage region. Using direct sequencing of the KI PCR fragments, we detected nonhomologous end joining (NHEJ) between the Rho genomic target and donor cassettes. The electropherogram of each sample was uncontaminated, which indicated that the PCR fragments were amplified from a single allele. 

Next, to verify the allele-specific HITI-mediated KI in vivo, we performed HITI insertion into Rho^+/AcGFP heterozygous mice in which AcGFP was inserted into intron 2 at the Rho locus (Supplementary Fig. S3). Because the KI mice expressed AcGFP in all rod photoreceptor cells, we performed HITI-mediated KI using donor cassettes containing mCherry (Fig. 2E). In the P56 electroporated Rho^+/AcGFP retinas, we observed that HITI-treated mCherry-positive cells in the ONL and that nonelectroporated cells maintained AcGFP expression (Fig. 2F). Some mCherry-positive cells lost AcGFP expression, which indicated insertion of a biallelic donor cassette to both alleles or monoallelic donor cassette insertion into the AcGFP KI allele. The presence of mCherry-positive cells that coexpressed AcGFP indicated the insertion of a monoallelic donor cassette into the WT allele (see Fig. 2F). The percentage of mCherry- and AcGFP-double-positive cells was about 80% (Fig. 2G), which suggested that HITI-mediated KI was more monoallelic. 

Normal Expression of Rho and Suppressed Rod Photoreceptor Degeneration of HITI-Treated Rod Photoreceptor Cells in Rho Mutant Mice

We confirmed the coexpression of mRhoCDS and AcGFP in electroporated WT mouse retinas (see Figs. 1D–F). We then performed HITI-mediated donor cassette insertion into the Rho^P23H/P23H homozygote mutant mouse retina to examine whether mRhoCDS expression suppresses rod photoreceptor cell degeneration (Fig. 3A). Because AdRP mutations cause photoreceptor degeneration, even in heterozygotes, we used P23H homozygous mice to quickly observe the suppressive effect against rod photoreceptor degeneration^43,44 and to confirm the expression and localization of the inserted normal Rho gene. 

Figure 3.

 

HITI-mediated normal rhodopsin insertion suppress rod photoreceptor degeneration in Rho^P23H/P23H mice. (A) Schematic illustration of HITI-mediated gene repair of the mouse Rho^P23H locus. (B) Low power image of section immunohistochemistry (IHC) of HITI constructs electroporated in vivo into Rho^P23H/P23H mice. Retinal sections from P21 mice immunostained with AcGFP (upper) and rhodopsin (lower) antibodies were shown. The white dotted line demarcates the boundary between the region with HITI-mediated gene insertion (left side) and the region without gene insertion (right side). Scale bar = 50 µm. (C) Section IHC of P14, P21, and P50 retinas from P23H mice coimmunostained with AcGFP (green) and rhodopsin (red) antibodies. Scale bar = 50 µm.

The P21 Rho^P23H/P23H mouse retinas electroporated in vivo with HITI plasmids at P0 are shown in Figure 3B. The AcGFP-positive ONL was thicker than the surrounding ONL, where AcGFP was negative, which suggested that electroporated rod photoreceptor cells were protected against degeneration (see Fig. 3B). The electroporated rod photoreceptor cells were coimmunostained with anti-AcGFP and -RHO antibodies. RHO protein was preferentially localized outside the ONL, whereas RHOP23H mutant proteins were mislocalized (see Fig. 3B). The IHC sections obtained at several developmental time points showed that HITI-treated rod photoreceptors were maintained from P14 to P50, whereas non-electroporated retinas showed progressive photoreceptor degeneration (Fig. 3C). Notably, the ONL thinning was suppressed in HITI-treated retinas from P21 to P50, demonstrating the protective effect of our approach. 

Recapitulation of HITI-Mediated KI into the Mouse Rho Locus Using AAV Vectors

Because electroporation has difficulty in transferring genes directly into cell nuclei in the nondividing state, AAV is often used as a vector for in vivo gene therapeutic agents. We examined whether the HITI-treatment constructs that worked in the electroporation-based validation could be switched into the AAV vectors (see Fig. 1). Based on our initial results (see Fig. 1), we selected gRNA1 for all subsequent experiments, including our AAV constructs. We used the type 8 AAV serotype transduced into adult mouse photoreceptor cells.^20,27 First, we packaged the HITI gene constructs into two AAV8 vectors, in which the donor and U6-gRNA1 cassettes were integrated into a single AAV8, according to a previous HITI study²⁰ (Supplementary Fig. S4). AAV8 cocktails were injected into 2-month-old C57BL/6 mice. No AcGFP signals were observed in the eyecups after 1 month (see Supplementary Fig. S4). 

We next packaged the donor gene and gRNA expression cassette into separate AAV8 capsids to construct three AAV8 capsids (Fig. 4A). Because the U6-gRNA cassettes are about 0.4 kb, which is small compared with the AAV package size, two U6-gRNA cassettes were placed side by side across the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) and packaged into self-complementary AAV (scAAV). We inserted WPRE because AAV plasmids without the WPRE sequence could not be obtained in sufficient quantities using Midi/Maxiprep plasmid purification. The three AAV8 vectors were injected subretinally into 2-month-old C57BL/6 WT mice and retinal flatmounts were made 1 and 2 months after the injection to observe AcGFP expression. The retinas at 1 month after the injection showed a strong AcGFP signal only around the AAV-injected area. The AcGFP region spread in the retinas at 2 months after injection (Fig. 4B) and increase in AcGFP-positive cells was observed in experimental retinas injected with AAV8-SpCas9, AAV8-Donor, and scAAV8-U6-gRNAx2, in which the number of AcGFP-positive cells was similar to that observed after in vivo electroporation. At 2 months in the retinas of the negative control group, sporadic AcGFP signals were detected around the injection area. Because the ITR sequences of AAVs have transcriptional activity,⁴⁵ the AcGFP signals may indicate leakage of ITR-mediated transcription. The IHC sections of retinas at 2 months after injection showed AcGFP-positive cells predominantly in the ONL, indicating successful HITI-mediated gene insertion in photoreceptors (Fig. 4C). The AcGFP signal was stronger and more widespread in retinas treated with SpCas9 + Donor + gRNA compared to those treated with SpCas9 + Donor only, demonstrating the importance of the gRNA for efficient gene insertion. 

Figure 4.

 

Effectiveness of HITI-treatment gene constructs packaged into AAV8 vectors. (A) Schematic illustration of the HITI-treatment AAV constructs. The mRho-gRNA1 was used as the target. The nucleotide length of each AAV construct is shown on the right. (B) Flatmounts of subretinally injected retinas with (right) or without (left) mRho-gRNA1 expressing AAV (scAAV8-U6-gRNA×2). AAV cocktails were injected into 2-month-old C57BL/6J mice, and the retinas injected after 1 (upper) and 2 (lower) months are shown. Orange dotted lines indicate the AAV-injected area in the retina. Scale bar = 1 mm. (C) Immunohistochemistry sections of retinas at 2 months after injection immunostained with anti-GFP antibody. Scale bar = 50 µm. (D) Schematic illustration of measurement of the optomotor response (OMR). The OMRs from the left and right eyes were recorded using clockwise and counterclockwise moving patterns, respectively. The HITI-treatment AAV cocktail was injected into the left eye of 1 month Rho^+/P23H mice. (E) Fluorescent fundus image 2 months after injection. (F) OKR measurements in the left and right eyes. The graph shows representative measurements from untreated (left) and AAV-treated (right) Rho^+/P23H mice. The horizontal and vertical axes represent the spatial frequency of the moving patterns and the OKR response, respectively. OKR was measured in the left (red) and right (black) eyes. Data represents means ± SD (n = 3) with individual experimental points. *P < 0.05.

To evaluate the possible therapeutic effects of AAV in mutant mice, we examined the pattern perception of HITI-treated mice using the optomotor response (OMR), which is observed as the head movement of mice synchronous to moving patterns of various spatial frequencies. Because Rho^P23H/P23H homozygous mice (see Fig. 3) showed complete degeneration of photoreceptor cells before AAV reconstitution, Rho^+/P23H heterozygous mice were used. AAV cocktails were injected into the left eye of 1-month-old mice (Fig. 4D). We chose to inject the AAV cocktails into 1-month-old Rho^+/P23H heterozygous mice because these mice exhibit progressive retinal degeneration. By initiating treatment at this age, we could assess the therapeutic effect of our HITI approach on an actively degenerating retina. One month after injection, the mice with AcGFP-positive retinas were selected using fluorescent fundus imaging (Fig. 4E). The number of AcGFP-expressing cells was lower in these mice than in WT mice, possibly because of the progression of photoreceptor degeneration. 

Two months after injection, the OMR was measured. Visual stimuli were displayed under a filter set for scotopic measurements, in which rod photoreceptor cells were evoked predominantly. In the WT mice, OMR-driven head movement was observed by moving patterns of spatial frequency between 0.1 c/degree and 0.3 c/degree, and Rho^+/P23H mice without AAV treatment responded to the moving patterns. In HITI-treated mice, the OMR response was significantly increased in the left eye, which suggested increased sensitivity of rod-derived perception because of HITI-mediated mRhoCDS expression instead of RhoP23H expression. 

Rapid Validation of Highly Efficient HITI-Mediated Gene Constructs for the Mouse Prph2 Locus

To examine whether our workflow for HITI-mediated gene construction can be applied to other genes, we targeted mouse Prph2, the second leading cause of AdRP^46,47 using WT CD1 mice. In the grID database in silico finder, 3 gRNA target sequences with grID scores ranging from 800 to 900 were selected around the 100 bp region of the proximal sequence from the Prph2 5′-UTR (Fig. 5A). We then constructed plasmids for the HITI-mediated mPrph2CDS insertion targeting the three gRNA sequences and performed in vivo electroporation. In the donor plasmid, mRho cDNA was replaced with mPrph2CDS and the 3′-UTR was replaced with the Prph2 3′-UTR (see Fig. 5A). In P21 mouse retinas, HITI constructs with mPrph2-gRNA1 and mPrph2-gRNA3 showed adequate AcGFP fluorescence signals (Figs. 5B, 5C). Most mCherry-positive cells in the ONL coexpressed AcGFP, which suggested that these HITI-treatment constructs were highly effective. However, the HITI construct with mPrph2-gRNA2 showed low EGFP expression even though the grID score was as high as 900. 

Figure 5.

 

Optimization and validation of highly efficient HITI-treatment gene constructs targeting the mouse Prph2 locus. (A) Schematic illustration of the Cas9-driven HITI-mediated gene insertion into the mouse Prph2 locus. Each donor cassette containing one of the three gRNA sequences (pink) selected from the 5′-UTR sequence of the mouse Prph2 locus at both ends was prepared. The targeted allele represents the gene structure when the HITI-mediated gene insertion occurs. (B) Schematic illustration of HITI-mediated donor insertion by in vivo electroporation into the P0 WT (CD1) mouse retina for initial gRNA screening experiments. The plasmid cocktail injected into the subretinal space included pbRho2k-SpCas9, pLeaklessIII-Donor cassette containing one of the three gRNA sequences (mPrph2-gRNA1, mPrph2-gRNA2, or mPrph2-gRNA3), pBAsi-U6-gRNA, and pCAG-mCherry. The plasmid cocktail was injected into the subretinal space. (C) P21 mouse eyecups electroporated in vivo at P0 with 3 plasmid cocktails targeting the mPrph2-gRNA1 (left), mPrph2-gRNA2 (middle), and mPrph2-gRNA3 (right) sequences of the mouse Prph2 locus. The dashed circles represent each eyecup. Scale bar = 1 mm. (D) Immunohistochemical analysis of HITI-treated P21 mouse retinas targeting three mPrph2 gRNAs. Electroporated cells were mCherry positive, and most cells in the ONL expressed AcGFP in retinal sections targeting gRNA and gRNA3 sequences. Scale bar = 50 µm.

Discussion

In this study, we established an electroporation-based workflow for the rapid validation of highly efficient HITI-treatment gene constructs in mouse differentiated photoreceptors. This approach is expected to contribute to conducting proof-of-mechanism studies of HITI-mediated gene therapy products in the ophthalmology field, especially for photoreceptor cells, and to shorten the development time. 

The advantage of the HITI-based genome-editing technique is the ability to insert exogenous DNAs in the desired orientation at the target site in nondividing cells. This is due to the insertion of foreign DNA through NHEJ, a genome DNA repair mechanism that occurs at high frequency in non-dividing cells. However, electroporation-based plasmid gene transfer in neonatal mouse retinas, that is, in dividing cells undergoing differentiation, is technically not inherently compatible with HITI genome editing. To induce genome editing in differentiated photoreceptors, our approach ensures that genome editing occurs in these cells due to the use of the Rho promoter. As demonstrated in Figure 2, the NHEJ-mediated gene insertion is confirmed in post-mitotic photoreceptors. Furthermore, using a similar electroporation method, our study achieved a gene insertion efficiency of 80% to 90% (see Fig. 1F). In contrast, when using donor gene constructs designed for homology-directed recombination (HDR)-mediated insertion that contain homologous sequences, the insertion efficiency was 10%.⁴⁸ Although the electroporation method has limitations in quantifying HITI efficiency in adult retinas, it offers a rapid and efficient way to validate the proof of mechanism of HITI concepts in differentiated rod photoreceptors without the time and resource investment required for AAV production. 

When SpCas9 is used as a genome-editing tool, three gRNA sequences selected by the in silico gRNA finder would be sufficient for determining a good gRNA sequence for in vivo electroporation, although there is no significant correlation between in silico scores and the cleavage efficiency tested in vitro and in vivo. Testing this initial validation using AAV would be time and cost intensive. Plasmids with appropriate backbones are key to successful HITI-mediated insertion via in vivo electroporation. In pLeaklessIII plasmids, triple poly(A) sequences are inserted into donor cassettes to prevent anomalous transcription from the plasmid backbone.⁴⁰ Our observations on the effect of plasmid backbone on HITI efficiency underscore the importance of vector design in plasmid-based HITI experiments. The difference in efficiency between pLeaklessIII and pAAV backbones demonstrates that transcriptional leakage from the plasmid backbone can significantly impact HITI-mediated insertion when using plasmid vectors. Switching the donor plasmid backbone to pAAV resulted in no HITI-mediated donor cassette insertion because there would be transcriptional activity from both the plasmid backbone and AAV ITR sequence.⁴⁵ For plasmid-based HITI efficiency validation, our results suggest that using pLeaklessIII or similar leak-prevention designs is preferable. 

With Rho promoters, HITI-mediated gene insertion is facilitated by efficient cleavage of the target genome and donor vector. Evaluation of the manner of insertion in single-rod photoreceptor cells indicated that 22% were biallelic and 78% were monoallelic. The HITI-mediated insertion was similar to that observed in the HITI experiment using Rho^+/AcGFP KI mice. Because the ratio of monoallelic insertions to the AcGFP KI allele was included in the AcGFP-negative fraction, the percentage of gene insertions was higher for the WT allele than for AcGFP allele, possibly because of the difference in genome structure between the WT and AcGFP alleles. It is possible that the donor cassette was difficult to insert because AcGFP had already been inserted. 

Most mutations in RHO are dominant inheritance mechanisms that cause dominant inhibition or haploinsufficiency^23,24; thus, the syndrome emerges in heterozygotes. In this study, the percentages of monoallelic and biallelic HITI-mediated mRhoCDS insertions were 78% and 22%, respectively. Based on these results, about 60% of HITI-treated rod photoreceptor cells were expected to be repaired when HITI was performed in mutant heterozygote individuals. That is, assuming that monoallelic insertion occurs equally in the mutant and normal RHO alleles, normal RHO was expressed for 22% of biallelic insertions and 39% of monoallelic insertions. 

It is essential that the expression level of the inserted gene is restored to the same level as that of the WT to achieve high therapeutic efficacy in treating rhodopsin mutations. Therefore, we optimized the construction of the donor cassette. When generating KI or transgenic mice expressing exogenous cDNA, inserting an intron before the cDNA increases the transcription level. Because 95% of the proteins expressed in rod outer segments are rhodopsins,³³ a high expression level must be maintained. It was previously reported that the transcription and expression levels of KI genes increase upon insertion of chimeric intron into the 5′-UTR of the mouse Rho allele.^37,38 The 3′-UTR region of the mouse Rho gene should be considered because this gene produces multiple mRNA transcripts that affect RNA transport and stability.³⁹ We used the chimeric intron and 3′-UTR in the donor cassette, and HITI-treated rod photoreceptor cells in Rho^P23H/P23H mutant mice were maintained with elongated outer segments. This result indicated that normal rhodopsin proteins expressed in optimized donor cassettes were sufficient to suppress photoreceptor degeneration. 

HITI-treated gene constructs can be used as AAV vectors. The insertion efficiency was similar to that for in vivo electroporation. In addition, AAV-treated Rho^+/P23H mutant mice showed good recovery in OMR measurements. Rho^+/P23H heterozygous mice did not exhibit progressive degeneration, but rod photoreceptor sensitivity was lower than in WT mice.⁴⁴ This recovery could be attributed to the increased sensitivity of HITI-treated photoreceptor cells expressing normal rhodopsins. In HITI-treated rod photoreceptor cells, the remaining mutant genome is considered to be activated. Considering that AcGFP expression was stopped by the HITI-mediated insertion of the mCherry-containing donor cassette into the Rho^+/AcGFP mouse retinas, the remaining mutant gene was not expressed. The use of a triple AAV system, while necessary for delivering all HITI components, presents challenges in terms of reproducibility and overall efficiency. Future efforts should focus on optimizing the vector design to potentially reduce the number of required AAVs. This could involve exploring larger-capacity AAV variants, developing more efficient promoters and regulatory elements, or investigating alternative delivery strategies that may allow for the consolidation of HITI components. To advance HITI-based approaches toward clinical applications, it will be essential to refine the AAV delivery system. This may include the development of engineered AAV capsids with improved targeting and transduction efficiency in photoreceptors, as well as the exploration of alternative gene delivery methods that could overcome the current limitations of multi-vector systems. 

We validated the HITI-treatment gene constructs for the mouse Prph2 locus from two of three gRNA target sequences. It took about 2 months for vector construction and in vivo validation, which shows that the workflow is convenient for application to other AdRP target genes. The workflow developed in this study can also be applied to human systems. For example, humanized mice carrying partially human genomic sequences can be used for in vivo validation, and electroporation-based gene transfer can be tested with human stem cell-derived organoids. This time- and cost-effective workflow may help to accelerate the development of HITI-based genome-editing reagents. Whereas our results in mouse models are promising, it is crucial to acknowledge the limitations in directly extrapolating these findings to human applications. The efficiency and safety of HITI-mediated gene insertion in human photoreceptors will require extensive further investigation. Future studies would focus on validating these approaches in human induced pluripotent stem cell (iPSC)-derived retinal organoids and explanted human retinal tissue. Additionally, careful assessment of long-term safety in nonhuman primate models will be necessary before considering clinical translation. 

Acknowledgments

The authors received generous support from all members of the Laboratory of Retinal Regeneration, RIKEN Center for Biosystems Dynamics Research. We thank the members of the Laboratory for Animal Resources and Genetic Engineering, RIKEN Center for Biosystems Dynamics Research, for providing and maintaining the mice and H. Fujiwara for supporting the FACS analyses. 

Supported in part by grants from the Japan Agency for Medical Research and Development (grant number 17bm0204002h0005 to M.T.), JSPS KAKENHI (grant numbers 24687010 and 17K11471 to A.O.), and the Charitable Trust Fund for Ophthalmic Research in the Commemoration of Santen Pharmaceutical's Founder (to A.O.). 

Disclosure: A. Onishi, VCGT Inc. (E) and Visiting Professor at Ritsumeikan University; Y. Tsunekawa, None; M. Mandai, None; A. Ishimaru, VCGT Inc. (E); Y. Ohigashi, Vision Care Inc. (E); J. Sho, Vision Care Inc. (E); K. Yasuda, VCGT Inc. (E); K. Suzuki, None; J.C. Izpisua Belmonte, Altos Labs, Inc. (E); F. Matsuzaki, None; M. Takahashi, President of VCGT Inc. and Vision Care Inc., and Visiting Professor at Ritsumeikan University