Precise CRISPR/Cas9 and Cas12 Correction Using Lipoplexes in Retinal Models Derived from Patients with Inherited Retinal Dystrophies

Abstract

Gene editing, particularly CRISPR/Cas technology, represents a promising approach for the treatment of rare genetic diseases, including inherited retinal dystrophies, for which effective therapies are largely unavailable. Despite extensive research investigating gene editing across a wide range of cell types, transient delivery of CRISPR/Cas components and efficient homology-directed repair (HDR) in differentiated cells remain challenging. In this study, we employed hiPSCs derived from patients with Stargardt disease or Best disease, carrying pathogenic variants in ABCA4 or BEST1, respectively, to explore gene editing in human models. CRISPR/Cas9 and Cas12 nucleases were delivered into hiPS-derived retinal pigment epithelium (RPE) and retinal organoids using lipoplexes and compared with electroporation. We evaluated transfection efficiency, sgRNA-mediated DNA cleavage, and HDR-based correction. Precise repair of the pathogenic BEST1 variant was successfully achieved in hiPS-derived RPE cells using both nucleases, with Cas12 yielding the highest efficiency, exceeding 10% of HDR correction. Edited RPE cells preserved normal morphology and expressed specific maturity markers. In contrast, retinal organoids exhibited moderate transfection efficiency but showed no detectable CRISPR/Cas-induced DNA cleavage, highlighting the need for further optimization of gene editing in more complex cellular tissues. This study demonstrates, for the first time, precise correction of a single-nucleotide mutation in patient-derived RPE using CRISPR/Cas9 and Cas12 delivered using lipoplexes. These findings underscore the therapeutic potential of CRISPR/Cas-based strategies for inherited retinal dystrophies and provide a proof of concept for future clinical approximations.

1. Introduction

Inherited retinal dystrophies (IRD) comprise a heterogenous group of rare ocular disorders that cause progressive vision loss resulting from degeneration of retinal cells, including photoreceptors and the retinal pigment epithelium (RPE) [1]. IRD are caused by mutations in more than 300 genes involved in retinal function [2]. Among these, ABCA4 and BEST1 explain a big number of IRD cases since they are responsible for two of the most prevalent macular dystrophies: Stargardt disease and Best disease, respectively [3]. Rare diseases affect 3.5–5.9% of the global population, and, in many cases, their pathogenesis and treatment remain poorly understood [4]. Although several therapeutic strategies are currently under investigation, most are gene- or mutation-specific, which limits their broad applicability [5,6]. Furthermore, the large number of causative genes and the phenotypic heterogeneity make their therapeutic management challenging.

Gene therapy has transformed the landscape of personalized medicine and represents an important strategy for the treatment of genetic disorders [5]. In this context, clustered regularly interspaced short palindromic repeats (CRISPR)/Cas system is, by far, the most robust and used approach for genetic engineer and arises as one of the most promising approximations for the treatment of genetic diseases. In recent years, several Cas variants, such as Cas12 and Cas13, have been developed to enhance DNA cleavage and expand genome editing versatility [7,8]. Nevertheless, despite its potential, considerable efforts are ongoing to improve CRISPR-based approaches in terms of efficiency, safety, and therapeutic applicability. In particular, the genetic modification of differentiated cells remains a key challenge for in vivo therapeutic applications. Moreover, gene editing-based therapies strongly rely on effective delivery systems capable of targeting the desired cells.

Several published studies explored the efficiency of gene-editing technologies in ocular diseases using both murine and human models, most of which rely on viral delivery vectors. A recent study evaluating several Cas13 in murine photoreceptors reported a gene-editing success rate of 2.04%, accompanied by retinal thinning following adeno-associated virus (AAV)-mediated delivery [9]. In the case of hiPS-derived RPE, CRISPR/Cas9-mediated disruption of VEGF has been reported using lentiviral delivery [10]. Base editing has also shown promising results in the mouse retina, achieving significant photoreceptor editing while partially preserving retinal structure and function [11]. Notably, Muller et al. reported ABCA4 base-editing in mice, non-human primates, and in vitro human retinal models using gene therapy vectors [12].

Several pathologies, including Duchenne muscular dystrophy, lymphomas, and beta-thalassemia are currently being evaluated in CRISPR-based clinical trials [13,14,15,16]. For eye diseases, only a limited number of approved and ongoing clinical trials exist, such as the CRISPR-based therapy EDIT-101 for Leber congenital amaurosis (LCA), which relies on viral delivery [17,18,19]. Notably, the eye is an advantageous organ for in vivo applications due to its accessibility and immune-privileged environment.

Viral vectors remain the most widely used delivery system for CRISPR-mediated gene editing and other conventional gene therapies because of their high transduction efficiency and long-term expression. However, their use raises significant concerns, including immune responses, limited cargo capacity, dose-dependent toxicity, and prolonged transgene expression. Consequently, transient delivery of gene-editing tools is preferred to minimize adverse effects. Non-viral approaches, such as nanoparticles, offer increased cargo capacity and improved stability [20]. Particularly, lipid nanoparticles (LNPs) have shown promising results in ongoing clinical trials for CRISPR delivery in non-ocular diseases [21,22,23]. For instance, the first in vivo CRISPR/Cas9 clinical trial for hereditary transthyretin amyloidosis with cardiomyopathy demonstrated favorable safety, tolerability, and efficacy outcomes with mild side effects [24], while the NTLA-2002 clinical trial for hereditary angioedema also showed good tolerability in single-dose phase II, demonstrating clinical efficacy [25]. Similarly, other lipid-based particles like lipoplexes, which are complexes formed by electrostatic interactions with cationic lipids, are being explored for transient intracellular delivery [26,27].

Regarding ocular diseases, several studies investigated the feasibility of base-editors and gene editing in murine models [28,29]. Particularly, LNP-mediated genome editing has achieved detectable DNA editing in mouse RPE cells, although efficiency in photoreceptors remains limited [30]. In this study, we evaluated CRISPR-mediated gene editing in hiPS-derived retinal models as a proof of concept for future therapeutic applications. Specifically, we compared Cas9 and Cas12 nucleases and assessed different transfection methods for CRISPR delivery into hiPS-RPE and retinal organoids (RO), which represent valuable three-dimensional in vitro retinal models. Remarkably, we observed efficient Lipofectamine-mediated transfection and sgRNA-induced DNA cleavage in hiPSC-RPE cells, whereas gene editing levels remained low in RO, highlighting the need for improved delivery strategies. Finally, we achieved precise correction of a heterozygous single-nucleotide mutation in hiPS-derived RPE from a patient with Best disease using both CRISPR/Cas9 and Cas12 systems. Collectively, these findings contribute to the development of CRISPR-based therapeutic strategies and support lipid-based transfection as a suitable approach for transient gene editing delivery in human retinal tissues.

2. Materials and Methods

2.1. Human iPSC Culture and Differentiation into Retinal Cell Models

Human iPS cell lines FRIMOi004-A and FRIMOi006-A, derived from adult donors affected by IRD, are summarized in Table 1 and were generated in a previously reported work [31,32]. Briefly, dermal fibroblasts were obtained from skin biopsy samples collected from donors after written informed consent, in accordance with protocols approved by the institutional ethics committee. Fibroblasts were reprogrammed into iPSC using a non-integrative Sendai virus system. Approximately three weeks after transduction, emerging iPSC colonies were manually picked and expanded. For selected experiments, wild-type hiPSCs derived from control individuals without ophthalmologic disease and lacking genetic variants associated with retinal dystrophies were used. hiPSC colonies were maintained in StemFlex medium (Thermo Fisher Scientific, Waltham, MA, USA) and cultured on Matrigel-coated dishes (Merck, Bedford, MA, USA). hiPSCs were differentiated into RPE and RO as previously described [33,34]. Briefly, for RO differentiation, hiPSCs were cultured in induction medium (IM) composed of Dulbecco’s modified Eagle’s medium/nutrient mixture F-12 (DMEM/F12), 5% fetal bovine serum, 0.1 mM nonessential amino acids, 2 mM GlutaMax, 1% N2, 1% B27 (Thermo Fisher Scientific), 10 ng/mL recombinant human IGF1 (Peprotech, Rocky Hill, NJ, USA), 10 mM Nicotinamide (Merck), 10 ng/mL human Noggin (R&D Systems, Minneapolis, MN, USA), 10 ng/mL human DKK1 (Peprotech) and 10 ng/mL human bFGF (R&D Systems). At day 30, neuroretinal vesicles were manually isolated and cultured in low-attachment plates, where they were cultured in retinal differentiation medium until maturation, as described in Gonzalez-Cordero et al. RPE differentiation was performed as described in Regent et al. Briefly, hiPSCs were cultured in basal medium consisting of high-glucose DMEM, 1x non-essential amino acids, 50 µM β-mercaptoethanol, and 20% of knockout serum (Thermo Fisher Scientific) sequentially supplemented with 10 mM Nicotinamide (Merck), 100 ng/mL Activin A (Stem Cell Technologies, Vancouver, BC, Canada) or 3 µM CHIR99021 (Merck) at defined time points [35]. Finally, mature hiPS-RPE cells were maintained in basal medium containing 4% of knockout serum and dissociated using TrypLE (Thermo Fisher Scientific). Cells were visualized using a Zeiss Axiovert microscope or an EVOS XL Core cell Imaging system (Thermo Fisher Scientific). hiPS-RO and hiPS-RPE cells derived from at least three independent differentiation batches or experiments were used for data generation. The number of RO or RPE wells analyzed is indicated in each figure or presented as individual values in bar charts.

2.2. CRISPR/Cas9 and Cas12 Gene Editing

Cas9 sgRNAs and ssODNs were previously designed and validated (Thermo Fisher Scientific) [36]. The ssODN used for correcting c.229C>T in BEST1 was the following: 5′-AAACTGACTCTGTATTGCGACAGCTACATACAGCTCATCCCCATTTCCTTCGTGCTGGGTGAGTTCCCCCTTCT-3′ with phosphorotioate modifications at the ends of the template. Cas12 sgRNAs (Synthego, Redwood City, CA, USA) were designed using CRISPOR or CRISPR RGEN Tools [37,38]. Approximately 50% confluent cell cultures were transfected with 1 µg mRNA-Express GFP transcript (System Biosciences, Palo Alto, CA, USA), 150 ng of sgRNA, and/or 200 ng ssODN (Thermo Fisher Scientific) using Lipofectamine 3000, Lipofectamine Stem or Lipofectamine CRISPRMAX (Thermo Fisher Scientific) according to manufacturer’s instructions. The sgRNA:Cas ratio was maintained at 2:1. In selected experiments, the amounts of ssODN and Cas9 were increased to 250 ng and 1 µg, respectively. For transfections in 12-well plates, 2.5 µL of Lipofectamine reagent was used per well. Cellular electroporation was performed using the Neon transfection system (Thermo Fisher Scientific) with the following parameters: 1400 V, 20 ms, 2 pulses.

For HDR-mediated correction, hiPS-RPE were cultured in medium supplemented with 10 µM ROCK inhibitor (Merck), 10 µM of the HDR activator L755507 (Merck), and 0.5 µM of the NHEJ inhibitor M3814 (Selleckchem, Houston, TX, USA) for 24 h. Then, cells were supplemented only with ROCK inhibitor during single-cell cloning until colony growth. After approximately 2 weeks, genomic DNA was extracted from viable hiPS-RPE clones for Sanger sequencing. Off-target prediction was done using CRISPR RGEN Tools allowing up to three mismatches in the prediction algorithm [38]. Off-targets with up to two mismatches in homology were further analyzed by Sanger sequencing (Table S2).

2.3. PCR Amplification and Sanger Sequencing

PCR amplification of the target genomic region was performed and run on a gel to confirm a single DNA band alongside a negative control. PCR products were subsequently purified using 96-well Acroprep Advance plates (Pall Corporation, Ann Arbor, MI, USA) with a vacuum manifold (Pall Corporation). Purified products were Sanger sequenced using forward and reverse primers (Macrogen Inc., Madrid, Spain). All primer sequences used in this study are listed in Table S3. Sequencing data was downloaded from the manufacturer’s platform and aligned and analyzed using ApE Plasmid Editor v3.1.8.1 for contig assembly [39]. Gene-edited clones were additionally screened for the presence of uncommon variants to confirm hiPS cell line purity.

2.4. Off-Target Prediction and Analysis

Off-target prediction was performed using the online tool Cas-OFFinder [38], allowing up to three mismatches. All predicted off-target sites located within exonic regions or intronic regions in close proximity to exons were analyzed by Sanger sequencing, following the same criteria as previously reported studies [36,40,41].

2.5. Genomic Cleavage Detection Assay

DNA cleavage was assessed using the GeneArt Genomic Cleavage Detection kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, cells were transfected with sgRNA, and after four days, the target locus was PCR amplified. A single band was confirmed by agarose gel electrophoresis. Next, the PCR product was subjected to several rounds of denaturation and re-annealing to generate mismatches, which were subsequently cleaved by the Detection Enzyme. Resulting fragments were visualized by agarose gel electrophoresis using the iBright CL1000 system (Thermo Fisher Scientific). Band intensities were quantified using iBright Analysis Software version 5.5.0 (Thermo Fisher Scientific) to estimate Cas nuclease activity.

2.6. Immunofluorescence Staining

For immunofluorescence analysis, RPE cells were seeded on Matrigel-coated ibidi slides (ibidi GmbH, Gräfeling, Germany) and fixed with 4% paraformaldehyde (Thermo Fisher Scientific) for 15 min at room temperature. Next, cells were permeabilized with 0.25% Triton X-100 in phosphate-buffered saline (PBS) and blocked for 1 h at room temperature in blocking solution (5% fetal bovine serum, 4% bovine serum albumin, and 0.5% Tween-20 in PBS). Fixed cells were incubated overnight at 4 °C with primary antibodies against ZO-1 (ZO1-1A12), EZRIN (3C12), or RPE65 (401.8B11.3D9) (Thermo Fisher Scientific). Cells were then incubated with anti-mouse Alexa Fluor-488 secondary antibody (A32723, Thermo Fisher Scientific) and counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (62248, Thermo Fisher Scientific). Immunofluorescence imaging was performed using a Zeiss Axiovert and Axiocam 503 mono (Carl Zeiss Inc., Jena, Germany), and images were processed with ImageJ software version 1.53k (NIH) [42]. Representative images are shown. GFP transfection efficiency was quantified using ImageJ and represented as positive fluorescent areas relative to the total area with cells.

2.7. Statistical Analysis

Statistical analyses were performed using Prism 10.1.2 (GraphPad Software, La Jolla, CA, USA). Statistical significance was assessed using the non-parametric Mann–Whitney U test and set at values of p  ≤  0.001 (***), p  ≤  0.01 (**), p  ≤  0.05 (*) levels, or non-significant (ns) for p  >  0.05. Bar graphs represent mean ± SD.

3. Results

3.1. Transfection of hiPS-RPE and hiPS-RO for Gene Editing of IRD-Associated Pathogenic Variants

To investigate gene editing in retinal models, we used hiPSCs derived from two patients affected by Best disease or Stargardt disease [31,32]. Specifically, the patient with Best disease carried a missense mutation in BEST1, and the one with Stargardt disease harbored four potential pathogenic variants in ABCA4. Among these, only the frameshift mutation, which is the only variant on one allele, was selected for gene editing due to its recessive inheritance pattern (Table S1). hiPS cell lines were differentiated into RPE and RO and subsequently subjected to CRISPR/Cas9- or Cas12-mediated gene editing to achieve HDR-based correction of the IRD-associated pathogenic variants (Figure 1A). Notably, both immature (day 100 of differentiation) and mature retinal organoids (day 250), which display a fully organized multilayered retinal structure, were used in this study (Figure 1B).

We first examined transfection efficiency in each differentiated cellular model and compared it with that of undifferentiated hiPSCs. For this purpose, we transfected an mRNA encoding the green fluorescent protein (GFP) using either electroporation with the Neon system or Lipofectamine-based transfection. For lipofection, we tested Lipofectamine 3000, which is optimized for hard-to-transfect cell lines, and Lipofectamine Stem because it is suitable for transfecting cells in suspension, like RO.

Differentiated retinal cells exhibited moderate levels of GFP+ cells compared with undifferentiated hiPSCs, which reached nearly 90% fluorescence following electroporation and more than 50% after lipofection (Figure 1C,D). In turn, hiPS-RPE and immature RO showed approximately 30% of GFP+ cells after transfection with Lipofectamine 3000 (Figure 1C,D). Unfortunately, transfection efficiency in mature retinal organoids was markedly lower, with GFP expression detected only in limited areas following Lipofectamine Stem treatment (Figure 1C,D). Notably, electroporation did not improve transfection efficiency compared with Lipofectamine-based transfection in hiPS-RO (Figure 1D). It is also worth noting that inefficient translation of the delivered GFP mRNA in differentiated models may have contributed to the reduced fluorescence-based detection of transfected cells. Overall, lipoplexes emerged as a suitable approach for retinal cell transfection, although efficiency was substantially lower in mature RO.

3.2. sgRNA-Mediated DSB in Retinal Cells Using Lipoplexes and Preserving Cellular Integrity

Preservation of cellular integrity and function is a critical consideration after transfection and gene editing. Actually, some studies previously reported cellular toxicity or adverse consequences after gene editing [28,43,44]. Thus, we analyzed hiPS-RPE and hiPS-RO morphology after transfection of CRISPR system. In these experiments, we used Lipofectamine CRISPRMAX and not Lipofectamine 3000 because it is optimized for sgRNA and Cas transfection.

hiPS-RPE did not display significant alterations in their appearance or behavior following any of the transfection methodologies tested, compared with untreated controls (Figure 2A). In contrast, hiPS-RO, which are highly structured three-dimensional models, exhibited pronounced disorganization of cellular layers and signs of apoptosis after electroporation (Figure 2A). Conversely, they preserved their morphology after treatment with Lipofectamine, particularly following CRISPRMAX delivery, maintaining a conserved structure comparable to untreated RO (Figure 2A). Notably, previously published data reported the degradation of outer segments (OS) in RO after Lipofectamine 2000 treatment [44]. Remarkably, we observed conservation of OS integrity four days after treatment using Lipofectamine Stem and CRISPRMAX (Figure 2B).

Next, we evaluated the ability of lipoplexes and electroporation to deliver the CRISPR system into retinal cells and assessed sgRNA-mediated induction of DNA double-strand breaks (DSBs). To this end, we used control sgRNAs targeting HPRT or B2M to assess the activity of Cas9 and Cas12 nucleases, respectively. In hiPSC-derived RPE cells, both nucleases induced significant levels of DNA DSBs, with cleavage efficiencies ranging from approximately 15% to 30% following Lipofectamine-based delivery, comparable to those achieved by electroporation (Figure 2C,D). For Cas9 nuclease, the highest HPRT cleavage was obtained using Lipofectamine Stem reaching almost 30%, slightly lower than that obtained in undifferentiated hiPSCs (Figure 2C and Figure S1A). In turn, Lipofectamine CRISPRMAX showed lower levels of cleavage, though not significantly different (Figure 2C). Similarly, comparable sgRNA-mediated cleavage efficiencies were observed using CRISPR/Cas12 across all delivery methods tested (Figure 2D and Figure S1B). Unfortunately, very low or undetectable levels of cleavage were observed in hiPS-RO using either Cas9 or Cas12 (Figure 2C–E). As expected, control transfections using either sgRNA or Cas nuclease alone did not generate detectable DNA DSBs (Figure S1C).

Collectively, these results indicate efficient transfection and sgRNA-induced DNA cleavage in hiPSC-derived RPE cells, along with preservation of cellular morphology following lipofection. However, we only observed limited transfection of hiPS-RO as assessed by GFP expression, while sgRNA-mediated DSB could not be detected using either Cas9 or Cas12 at the tested conditions (Figure 2F).

3.3. Specific DNA Cleavage Targeting Pathogenic BEST1 or ABCA4 Variants in hiPS-RPE

The results described above demonstrated substantial sgRNA-induced DSB in hiPS-RPE, reaching similar ratios to those observed in undifferentiated hiPSCs and exceeding efficiencies previously reported in the literature [28]. Based on these findings, we decided to focus on hiPS-RPE for the correction of pathogenic variants causing IRD. For that purpose, we designed specific sgRNAs for each Cas nuclease targeting the disease-causing mutations in BEST1 and ABCA4 (

Table 1. hiPS cell lines used for gene editing.

ID Cell Line	hiPS Cell Line	Gene	Mutation	sgRNA_ID	Cas	PAM a	Sequence b	Distance DSB/Mut. (nt)
hiPS BEST1	FRIMOi006-A	BEST1	c.229C>T	sgBEST1_4	Cas9	TGG	CGAAGGAAAUGGAGAUGAGC	−5
				sgBEST1-Cas12_1	Cas12	TTG	UGCGACAGCUACAUCCAGCU	−7
				sgBEST1-Cas12_2	Cas12	TTTG	AGAAACUGACUCUGUAUUGC	−24
				sgBEST1-Cas12_3	Cas12	TTG	CGACAGCUACAUCCAGCUCA	−6
hiPS ABCA4	FRIMOi004-A	ABCA4	c.3211_3212insGT	sgABCA4_7	Cas9	CGG	GCAUGCAGAGAAAGCUGUGU	+1
				sgABCA4-Cas12_1	Cas12	TTG	UGCCUCCAGGUGGCAUGCAG	−11
				sgABCA4-Cas12_2	Cas12	TTGT	GCCUCCAGGUGGCAUGCAGA	−10

^a PAM, protospacer adjacent motif; ^b Nucleotides in blue indicate the pathogenic variant.

).

While spCas9 exclusively requires an NGG protospacer-adjacent motif (PAM), the hfCas12Max nuclease can recognize multiple PAM sequences with different predicted affinities (Figure 3A). Accordingly, we designed several guides to evaluate PAM-related affinities and to account for differences in the distance between the target mutation and the induced DSB (Figure 3B and Table 1). For Cas9-mediated editing, we employed a previously validated sgRNA that had been successfully used for gene editing in this hiPSC line [36].

The analysis of sgRNA-mediated DNA cleavage rendered similar results (10–15%) for Cas9 targeting both the BEST1 and ABCA4 pathogenic variants, independent of the delivery method (Neon electroporation or Lipofectamine) (Figure 3C,E). In contrast, Cas12-mediated cleavage exhibited more variable efficiencies, depending on both the delivery system and the sgRNA used, ranging from 5% to 30%. Specifically, sgBEST1-Cas12_3 achieved the highest cleavage efficiency for the c.229C>T BEST1 variant, reaching nearly 30% when delivered via electroporation or Lipofectamine Stem, outperforming the other two guides (Figure 3D). Overall, DNA cut in BEST1 was higher using Cas12 than using Cas9 with the guides used (Figure 3C,D). In case of ABCA4, sgABCA4-Cas12_1 performed slightly better than sgABCA4-Cas12_2, although not statistically different (Figure 3F). In this case, Cas9 and Cas12 exhibited comparable efficiencies, ranging between 10% and 20% (Figure 3E,F). Interestingly, these two Cas12 guides that yielded the highest cleavage results for each pathogenic variant (sgBEST1-Cas12_3 and sgABCA4-Cas12_1) shared the same TTG PAM sequence, suggesting a better affinity of Cas12 to this motif in our assay (Table 1).

Considering that sgBEST1-Cas12_3 has the shortest distance between the target mutation and the induced DSB, which is a key determinant for efficient HDR-mediated correction [29], and that it exhibited the highest DNA cleavage, we decided to focus only on the HDR of the c.229C>T BEST1 variant. Accordingly, we employed the Cas9 sgRNA sgBEST1_4 and the sgBEST1-Cas12_3 guide to assess the HDR-mediated correction efficiency of the two nucleases.

3.4. Precise Correction of a Pathogenic BEST1 Variant in hiPS-RPE with Enhanced Efficiency Using CRISPR/Cas12

To evaluate the efficacy of CRISPR-mediated gene editing in retinal cells as a potential therapeutic approach for IRD, we delivered sgRNA, Cas nuclease, and a single-stranded oligodeoxynucleotide (ssODN) repair template into hiPSC-derived RPE using lipofection, which is a more suitable method for a potential therapeutic in vivo application (Figure 4A). An HDR enhancer and NHEJ inhibitor were used to improve homology-directed repair and efficient genome editing, particularly for precise single-nucleotide modifications (Figure 4A) [45,46].

First, we focused on CRISPR/Cas9 and transfected the sgBEST1_4 using Lipofectamine CRISPRMAX, which is optimized for CRISPR-system delivery and showed cleavage efficiency comparable to Lipofectamine Stem (Figure 3C). We first tested an equimolar ratio of Cas9:sgRNA (Experiment A). Two days post-transfection, treated cells displayed normal proliferation and preserved the typical RPE cobblestone morphology after Cas9 or Cas12 treatment (Figure 4B,C). Then, cells were diluted and seeded onto 96-well plates for single-cell cloning as schematized in Figure 4A. hiPS-RPE are low-proliferating cells that could be affected by the single-cell cloning strategy used for genotyping the clones after gene editing. Therefore, after several weeks, we selected wells with viable cells, and wells exhibiting more than one pool of proliferating RPE cells were discarded for the screening.

Experiment A yielded a total of 10.71% gene-edited clones (3/28), with 7.14% carrying the heterozygous BEST1 mutation in homozygosis instead of corrected, and a 3.57% harboring both the mutation and PAM modification in heterozygosis (Figure 4D). These results are consistent with previous reports in hiPSCs using the same sgRNA, which showed cleavage of both mutant and wild-type alleles [36]. This new results on gene editing outcomes in Experiment A also indicated the cleavage of the wild-type allele and its repair using either the sister chromatid or the ssODN, as schematized in Figure 4E [36]. However, no indels in the mutant allele were observed.

Next, we assessed whether increasing Cas9 concentration and ssODN template levels could enhance mutant allele cleavage and HDR efficiency (Experiment B). Surprisingly, this approach yielded 15.68% gene-edited clones (8/51). Among these, 7.84% exhibited pathogenic variant correction, with 5.88% repaired via the sister chromatid and 1.96% incorporating the ssODN-mediated PAM modification (Figure 4D,E and Figure S2A). The other half displayed the BEST1 mutation and PAM modification in heterozygosis, indicating the cut of the wild-type allele and ssODN-mediated repair. Notably, the 7.84% of corrected clones in Experiment B was comparable to the proportion of clones carrying the homozygous mutation in Experiment A, which were repaired using the sister chromatid (Figure 4D). Altogether indicated that the increase in Cas9 could trigger higher percentages of DNA DSB, and that the concentration of ssODN could influence gene-editing outcomes in cases in which the sgRNA could recognize and cut the wild-type allele.

Finally, we wanted to compare whether the higher DNA cleavage achieved with CRISPR/Cas12 would translate into improved HDR-mediated correction of the pathogenic BEST1 variant. Notably, the sgBEST1-Cas12_3 used cuts the DNA near to the site where sgBEST1_4 does, enabling a more appropriate comparison between Cas9 and Cas12 systems. Moreover, we used the same Lipofectamine reagent (CRISPRMAX), but also the Lipofectamine Stem because it showed a slight increase in the DNA DSB (Figure 3D).

Significantly, we achieved successful correction of the BEST1 mutation using both Lipofectamine reagents, yielding higher HDR rates than those obtained when using Cas9 (Figure 4F and Figure S2A). The highest gene-editing efficiencies were observed with Lipofectamine Stem, in line with the highest percentages of DNA cleavage, compared to CRISPRMAX (Figure 4F). Specifically, we reached a maximum of 24% of gene editing using Lipofectamine Stem, from which the 10% were clones with the pathogenic variant properly corrected and the 14% carrying the mutation in homozygosis (Figure 4F). Nevertheless, the assay performed with CRISPRMAX rendered the best ratios of HDR correction and gene editing of the mutant allele resulting in 10.53% properly repaired clones, whereas only 2.63% harbored the pathogenic variant in homozygosis (Figure 4F). Further, expanded transfected cells exhibited normal morphology and expressed RPE-maturity markers, such as ZO-1, EZRIN, and RPE65, as compared with untreated cells (Figure 4G).

Additionally, the best results regarding on-target effects were obtained with CRISPR/Cas12 using Lipofectamine Stem similarly to Experiment B performed with Cas9 (Figure 4H). Notably, in all gene editing assays, the on-target abnormalities were below 15% of the screened clones. Regarding off-target analysis, homologous regions prediction identified only two regions with less than three mismatches for sgBEST1_4 and one for sgBEST1-Cas12_3 besides to the target sequence itself (Table S2). Notably, we found no off-target effects in edited clones, except in one clone in which we observed a single-nucleotide change in the experiment performed using Cas9 (

Table 2. Off-target analysis results.

Gene	Mutation	sgRNA ID	Cas	Off-Target ID	Chr	Position	Clones Analyzed	Off-Target Effects
BEST1	c.229C>T	sgBEST1_4	Cas9	Cas9_OT-1	chr11	65736485	12	0
				Cas9_OT-2	chr6	123292078	16	1
		sgBEST1-Cas12_3	Cas12	Cas12_OT-1	chrX	12446865	11	0

and Figure S2B,C). Finally, to discard potential contamination with another cell line, and considering that the strategy followed in the screening was single-cell cloning, we checked and confirmed the presence of a rare variant carried by the patient (Figure S2D).

Collectively, this study shows the effective and safe lipid-mediated transfection of hiPS-RPE and hiPS-RO. Moreover, we achieved substantial DNA cleavage with both Cas9 and Cas12 nucleases in hiPS-RPE and successfully corrected a pathogenic BEST1 variant, with Cas12 yielding the highest efficiency, above 10% of properly repaired clones. In addition, our results indicate a potential use of CRISPR without using exogenous DNA templates for correcting heterozygous pathogenic variants. These findings significantly contribute to the progress of gene-editing strategies as potential therapeutic approaches for inherited retinal dystrophies.

4. Discussion

Gene editing, particularly CRISPR, represents a powerful strategy for the treatment of genetic disorders, especially in case of IRD, whose genetic heterogeneity complicates clinical management. By enabling permanent modification of the genome, CRISPR-based approaches have the potential to halt disease progression independently of the specific gene or mutation involved. However, while CRISPR efficacy has been extensively investigated in proliferating cells, its in vivo potential and applicability in differentiated cell models remain insufficiently explored, particularly when using transient delivery methods. Therefore, addressing this gap is essential to advance the development of safe and effective gene editing–based therapies.

In this context, we evaluated the feasibility of transfection and gene editing in human retinal cell models and compared the efficiencies of Cas9 and Cas12 nucleases. We showed that hiPS-RPE cells exhibited considerable DNA cleavage, reaching the highest levels using Cas12. In contrast, hiPS-RO displayed only minimal transfection efficiency and sgRNA-induced DSB in immature RO with the methodologies tested, suggesting a correlation between cellular maturity and transfection efficacy. This observation highlights the importance of developing highly efficient delivery strategies for in vivo CRISPR applications. Moreover, while electroporation produced high overall transfection rates, lipoplex-mediated delivery achieved a superior balance between transfection efficiency, cellular integrity, and DNA cleavage.

Several recent studies investigated the feasibility of gene editing in vivo and in differentiated cell models [28,44,47,48]. Notably, most of these studies relied on plasmid- or AAV-mediated delivery systems [48,49,50]. In retinal cells, Pulman et al. explored direct RNP delivery into photoreceptors and RPE cells, analyzing indel generation and performing in vivo gene editing in the mouse retina using base editors, reaching approximately 2% of intended substitutions in RPE and around 10% in the whole neural retina [28]. Similarly, Haldrup et al., reported up to 36% of gene disruption in RPE using lentivirus-derived nanoparticles delivered by subretinal injection in the mouse retina [51].

CRISPR-based therapies strongly rely on the delivery strategy of the gene editing system. AAV vectors have been widely used and have demonstrated promising results due to their high transduction efficiency, cell-type specificity through capsid engineering, and ease of manipulation. However, several limitations constrain their applicability. Prolonged expression of CRISPR components may lead to adverse effects and increased off-target activity [28,52,53,54]. Moreover, AAV serotypes can significantly influence cellular transduction efficacy, and persistent expression is undesirable for therapeutic approaches from a safety perspective. Their limited cargo capacity may also restrict their applicability or reduce editing efficiency. In turn, in gene editing strategies, precise targeting of specific cell-types is not always required, as the therapeutic effect relies on permanent DNA correction.

Lipid-based delivery systems, such as lipoplexes and LNP, have emerged as virus-free alternatives that are gaining special clinical interest. Here, we compared several commercially available Lipofectamine reagents with electroporation obtaining promising results. Cellular electroporation facilitates intracellular delivery through transient disruption of the phospholipid bilayer. However, the required electrical pulses frequently induce cellular damage and death [55]. In this context, we observed preservation of RO outer segments and RPE morphology following Lipofectamine treatment, unlike electroporation and previously reported data using lipid-based transfection [43]. Nevertheless, LNP, although appreciated for their safety, tolerability, and biodistribution, require careful and optimized production using engineered lipid formulations under highly controlled conditions. A major limitation remains their ability to efficiently fuse with and deliver cargo into specific cell types without inducing toxicity [56]. Further improvements in delivery methodologies are therefore required, particularly for targeting specific retinal cell types such as photoreceptors.

Here we show the precise correction of the BEST1 c.229C>T mutation using CRISPR/Cas9 and Cas12, achieving approximately 10% gene correction in hiPS-RPE via lipoplex-mediated delivery. Although gene-editing efficiency was lower than that observed in hiPSCs, transfection efficiency and sgRNA-induced cleavage were also reduced in differentiated hiPS-RPE. Notably, compromised cellular viability of hiPS-RPE following single-cell cloning could prevent a more accurate analysis of the gene-edited clones, suggesting that actual correction rates may be underestimated. Importantly, on-target defects were significantly lower than those obtained in hiPSCs, ranging from 2% to 14.29%, compared with nearly 40% reported in hiPSCs [36]. It is worth noting that these alterations may not be detrimental in dominant-negative disorders such as Best disease, where putative resulting frameshift mutations could result in a wild-type phenotype through nonsense-mediated decay of the mutant allele.

Efficient and specific HDR correction is one of the key points of CRISPR-based precision medicine, particularly for heterozygous pathogenic variants, which are the most prevalent mutations in IRD. In this study, several gene-edited clones exhibited cleavage the wild-type allele followed by repair using the sister chromatid rather than the exogenous ssODN, as reported previously [36]. This finding emphasizes the importance of highly specific sgRNA design and suggests that ssODN delivery may not always be necessary, thereby reducing cargo complexity. Additionally, cells harboring homozygous pathogenic variants remain susceptible to sgRNA recognition and could potentially be corrected through subsequent rounds of gene editing.

Optimization of Cas nucleases has substantially enhanced CRISPR potential, particularly in precision medicine. Cas12 offers a broader PAM compatibility, reduced off-target activity, and uses shorter sgRNAs, thereby reducing the cargo. Consistent with these advantages, we observed increased DNA cleavage efficiency, improved HDR correction, and reduced on- and off-target effects when using Cas12 nuclease compared to Cas9. These findings underscore the potential of Cas12-mediated gene editing in retinal therapies and provide a framework for developing safe, transient, and effective delivery strategies for differentiated retinal cells.

5. Conclusions

Collectively, these findings expand current knowledge of gene editing in differentiated retinal models for disease modelling and therapeutic evaluation in IRD. We demonstrate robust DNA cleavage using both Cas9 and Cas12 nucleases and report successful HDR-mediated correction of an autosomal dominant pathogenic variant in RPE cells causing Best disease.