Introduction

The retinoblastoma gene (RB1) encodes retinoblastoma-associated protein (pRb), a central regulator of the cell cycle and well-known tumor suppressor. Loss of RB1 tumor suppressor function is commonly observed in many cancer types, including those of the skin, bones, bladder, lungs. and eyes1,2,3,4,5,6. To better understand RB1 function in cancer development or progression, researchers have developed model cell lines or organisms with RB1 mutations, which have provided useful information about RB1 function in cell cycle control, genome stability, and retinal developments7,8,9,10,11,12. However, many of these studies have adopted deletion-7,8 or indel-based7,8,9,10,11 mutation models rather than nonsense mutations, which is the most frequent type of mutation detected in patients13,14.

Among various available CRISPR-based genome editing strategies, we used base editors to introduce a nonsense mutation relevant to cancer and correct this mutation in RB1. Two types of base editors have been invented: cytosine base editors (CBEs)15,16,17 and adenine base editors (ABEs)16,18,19and Precise conversion of C·G to T·A, or vice versa, can be achieved by introducing CBEs or ABEs into cells with high efficiency20,21,22,23,24,25,26. Applications of base editors enable therapeutic gene correction, as well as disease modeling and study, through BE-mediated gene silencing (CRISPR-STOP)27,28 and PTC bypassing (CRISPR-Pass)29.

In this study, we aimed to establish RB1-mutant cell lines carrying nonsense mutations to provide a reliable protocol that enables the selection of mutant cell colonies even with low editing efficiency. To enable cellular-level investigation of cancer development due to RB1 loss, we introduced retinoblastoma, a rare ocular tumor that usually occurs when biallelic genetic mutations occur in RB1 in susceptible retinal cells3,13,14,30,31,32. We created retinoblastoma model cell lines by inducing a premature stop codon at arginine 552 of RB1, a common mutation among patients with retinoblastoma33,34,35. We also tested whether ABE-mediated gene rescue could treat this pathogenic type of retinoblastoma.

Materials and methods

Plasmid construction

Plasmids expressing cytosine and adenine base editors were obtained from the non-profit plasmid distribution service Addgene (Watertown, MA, USA) (pcDNA3.1_pCMV-nCas-PmCDA1-ugi pH1-gRNA[HPRT)]; #79620, pCMV-ABEmax; #112095, ABE8e; #138489). The Target-AID-encoding sequence was obtained using the NotI and BmtI restriction digests of the pcDNA3.1_pCMV-nCas-PmCDA1-ugi pH1-gRNA(HPRT) plasmid and cloned into a mammalian expression plasmid backbone under the control of a CMV promoter. The backbone was amplified using KOD Multi & Epi PCR kits (Toyobo, Osaka, Japan) from the p3s-SpCas9 plasmid, which was a gift from Professor Jin-Soo Kim (#104171; Addgene). Fragments were gel-purified and Gibson assembled using NEBuilder HiFi DNA Assembly master mix (New England Biolabs, Ipswich, MA, USA) for 1 h at 50 °C and transformed into chemically competent Escherichia coli (DH5α; Enzynomics, Daejon, Republic of Korea). The gRNAs were cloned into a BsaI-digested pRG2 vector (#104174; Addgene). For gRNA cloning, oligonucleotides containing the spacer (Tables S1 and S2) were annealed to form double-stranded DNA fragments with compatible overhangs and ligated using T4 ligase (Enzynomics). All plasmids used for the transfection experiments were prepared using NucleoBond Xtra Midi Plus EF kits (Macherey-Nagel).

Cell culture

BJ-5ta, a human fibroblast cell line, was purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). The cells were maintained in a mixture of Dulbecco’s modified Eagle’s medium (LM001-05; Welgene, Gyeongsan, Republic of Korea) and medium 199 (LM006-01; Welgene) at a ratio of 4:1. Media were supplemented with 10% fetal bovine serum (FBS) (16000-044; Gibco, Waltham, MA, USA), 0.01 mg/mL penicillin-streptomycin (15140-122; Gibco), and 0.01 mg/mL hygromycin B (10687010; Invitrogen, Waltham, MA, USA).

Transfection

Stable R552X homozygous and heterozygous cell lines were generated from BJ-5ta fibroblasts (ATCC CRL-4001). Next, 1.0 × 105 BJ-5ta cells were electroporated with plasmids encoding Target-AID (675 ng) and sgRNA (225 ng) via a Neon Transfection System (Invitrogen) with the following parameters: pulse voltage, 1,400 V; pulse width, 20 ms; pulse number, 2. Transfections were performed in 24-well plates in a volume of 500 µL cell growth media without antibiotics and the media were changed to fresh media containing antibiotics at 1 day post-transfection. The cells were harvested at 3 days post-transfection, and genomic DNA was prepared using NucleoSpin Tissue (Macherey-Nagel & Co., Düren, Germany) for next-generation sequencing (NGS) analysis.

To rescue the established PTCs in the R552X model cell line, ABE-mediated genome editing was performed. We electroporated 3.0 × 105 cells of each BJ-5ta R552X cell line with plasmids expressing ABEmax or ABE8e (1.5 µg) and sgRNA (600 ng). Electroporation was performed using the Neon Transfection System (Invitrogen) with the same parameters as described above. Transfections were performed in 6-well plates in a volume of 2 mL of cell growth medium without antibiotics, and the medium was replaced with fresh medium containing antibiotics 24 h post transfection. Five days after each transfection, cells were harvested, genomic DNA was extracted using NucleoSpin Tissue (Macherey-Nagel & Co.), and total RNA was prepared using NucleoSpin RNA Plus (Macherey-Nagel & Co.) for NGS analysis.

Selection of edited cells

After confirming the base conversion rate of CBE-edited bulk cells, the cells were split and seeded at 10 cells/well in 96-well plates to increase the possibility of selecting cell colonies containing the desired base substitution. In each screening step, cells were cultured in 96-well plates in 100 µL cell growth media. The 10-cell colonies were analyzed using deep sequencing, and the colony that showed the highest targeted conversion rate was seeded again at 1 cell/well in 96-well plates. After enrichment, targeted deep sequencing was performed to verify the base conversion rate of each single-cell clone. The clones were sorted into homozygous and heterozygous cell lines.

Targeted deep sequencing

The purified total RNA was reverse transcribed using ReverTra Ace -α- (Toyobo). Segments of the synthesized cDNA and genomic DNA that encompassed the base editing target sites were amplified using KOD Multi & Epi PCR kits (Toyobo) for sequencing library generation (Table S3). These libraries were sequenced using MiniSeq with a TruSeq HT Dual Index system (Illumina) as described previously21. Briefly, equal amounts of PCR amplicons were subjected to paired-end read sequencing using an Illumina MiniSeq platform. After MiniSeq, the paired-end reads were analyzed by comparing the wild-type and mutant sequences using a BE analyzer (http://www.rgenome.net/be-analyzer/).

Western blotting

To extract the proteins, the cells were first washed with ice-cold PBS. Following centrifugation at 14,000 x g for 15 min at 4 °C, the cell pellets were resuspended RIPA lysis buffer (RV2002-050-00; Biosesang, Yongin, Republic of Korea) at 1 × 106 cells/mL. After 20 min of incubation on ice, the samples were centrifuged and the supernatants were collected. A protease inhibitor cocktail (P3100; GenDEPOT, Altair, TX, USA) was added to prevent protein degradation. After protein extraction, primary antibodies were added and incubated overnight. The sections were then incubated with secondary mouse anti-rabbit IgG-HRP antibodies (sc-2357; Santa Cruz Biotechnology, Dallas, TX, USA) or goat anti-mouse IgG-HRP antibodies (sc-2005; Santa Cruz Biotechnology). Then, the HRP conjugates were transferred to a PVDF membrane (IPVH00010; Millipore, Burlington, MA, USA) and detected with ECL solution (34580; Thermo Fisher Scientific, Waltham, MA, USA). The staining was normalized to that of β-actin. The primary antibodies used were as follows: N-terminal: MAB6495, R&D Systems, Minneapolis, MN, USA. Central: 9313, Cell Signaling Technology, Danvers, MA, USA; LS-C83412, LS-C26527, LS Bio, Seattle, WA, USA. C-terminal: ab181616, Abcam, Cambridge UK; 9309, CST; β-actin: A2066, Sigma-Aldrich, St. Louis, MO, USA.

Quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) analysis

Total RNA was extracted from the cells using TRIzol (15596018; Ambion, Austin, TX, USA), followed by the addition of chloroform (C2432; Sigma). Extracted RNA was incubated overnight in isopropanol at − 20 ℃. After incubation, the extracted RNA was washed with 70% ethanol. RNA was reverse-transcribed into cDNA using a High-Capacity RNA-to-cDNA kit (4387406; Applied Biosystems, Waltham, MA, USA) according to the manufacturer’s protocols. A total of 1 µg of RNA was reverse transcribed to cDNA. PCR was performed using TaqMan Fast Advanced Master Mix (4444965, Applied Biosystems) according to the manufacturer’s protocol. Quantification was performed by normalizing levels to those of glyceraldehyde 3-phosphate dehydrogenase (GADPH), 18s rRNA, hypoxanthine-guanine phosphoribosyltransferase (HPRT), and glucuronidase beta (GUSB). The probe identity numbers are listed in Table S4.

Fluorescence-activated cell sorting (FACS) – Propidium iodide (PI) staining

The total cells (1.0 × 106) were counted using an automated cell counter. The cells were washed with PBS and resuspended in binding buffer, followed by PI staining (P4170; Sigma) at a final concentration of 50 mg/mL. RNase A solution (R-6513; Sigma) at a final concentration of 0.5 µg/mL was added to prevent RNA staining. After an overnight incubation at 4 ℃, the stained cells were then analyzed with a flow cytometer (C6 Flow Cytometer; Becton Dickinson Accuri, Ann Arbor, MI, USA) to analyze the cell cycle status.

Cell proliferation assay

Cell proliferation was examined after 3, 24, 48, and 72 h of incubation using the WST-1 kit EZ Cytox (EZ-1000; Dogen, Seoul, Republic of Korea) according to the manufacturer’s protocols. BJ-5ta cells were seeded in 96 wells at a density of 3 or 5 × 103 cells/100 µL per well. After the designated time had passed, 10 µL of cell viability assay reagent was added into each well followed by 1 h of additional incubation. The absorbance was measured at 450 nm using a colorimetric reader. Experiments were performed in triplicate or six replicates.

Statistical analysis

The results of western blotting and FACS were analyzed using the Kruskal–Wallis test and post analysis using the Mann–Whitney U test by comparing all possible pairs in each experiment. IBM SPSS Statistics 25.1 software (IBM) was used for these two processes. The FACS results were analyzed using the Kruskal–Wallis test and post-analyzed using Dunn’s multiple comparison test. The qRT-PCR results were analyzed using one-way analysis of variance (ANOVA) and Tukey’s test for post hoc analysis. The mRNA expression of the RB1 wild-type (WT) was set to 1.0 and the other data were adjusted proportionally. GraphPad v5.0 software (PRISM) was used for these two analyses. All data are shown as mean ± s.e.m. The cut-off for statistical significance was set to P ≤ 0.05.

Results

Generation of the RB1 gene mutated model cell lines.

To optimize the target C-to-T mutation efficiency, we utilized Target-AID, a cytosine base editor with an editing activity window from 16 to 19 positions proximal to the protospacer adjacent motif17 (Fig. 1A, B). High-throughput sequencing data from bulk BJ-5ta cells, an hTERT-immortalized fibroblast line, transfected with Target-AID- and guide RNA-encoding plasmids showed 3.1% efficiency of the targeted base conversion (c.1654 C > T). To establish single-cell colonies that were RB1 R552X mutation bi-allelic (homozygous) or mono-allelic (heterozygous), we enriched edited cell populations by sequential screening, which consisted of spreading, sequencing, and selection procedures (Fig. 1C). After the first screening, the cell populations with C to T mutations were greatly enhanced in 10-cell colonies (Fig. 1D). The two 10-cell colonies that showed the highest editing efficiencies (indicated with red dots in the first screening) and two 10-cell colonies that showed the third and fourth highest editing efficiencies (blue dots in the first screening) were spread out as single cells to sort R552X homozygous or heterozygous cells, respectively. High-throughput sequencing data from single-cell colonies showed that the R552X homozygous (red dots in the second screening) and heterozygous (blue dots in the second screening) single-cell clones were successfully established (Fig. 1D, E).

Fig. 1
figure 1

Generation of the RB1 gene-mutated human fibroblast cell lines using the cytosine base editor (CBE). (A) Schematic diagram displaying the induction of the retinoblastoma-causing mutation, R552X, in RB1 exon 17. The protospacer adjacent motif, Target-AID editing window, and targeted cytosine are highlighted in blue, green, and red, respectively. This type of retinoblastoma occurs when arginine 552 changes to a stop codon. (B) Schematic diagram of the CBE, Target-AID. A nickase Cas9 (D10A)-sgRNA RNP complex forms an R-loop at the target site. The fused cytidine deaminase, PmCDA1, converts an exposed cytidine into a thymidine. (C) The procedure of base editing and sequential screening with high-throughput sequencing for the establishment of homozygous and heterozygous RB1-mutated cell lines are shown. (D) Edited frequencies (%) of bulk, first, and second screening populations, as measured by next-generation sequencing (NGS). Green dots represent bulk cell populations which were used for first screening. Blue and red dots represent cell populations used in the second screening process for heterozygous and homozygous cell lines, respectively (mean ± SEM). (E) Edited frequencies (%) of established R552X homozygous and heterozygous cell lines measured by NGS.

Full size image

Confirmation of RB1 gene knock-out in R552X cell line

We confirmed RB1 gene expression in homozygous and heterozygous R552X model cell lines (hereafter Ho-R552X and He-R552X, respectively). To investigate whether the expression level of RB1 mRNA fluctuated after introducing the R552X mutation via nonsense-mediated decay or alternative splicing products, we analyzed the mRNA expression levels using 17 different primer sets covering most of the exons in the RB1 gene (Fig. 2A). Quantitative PCR data showed that the expression level of RB1 mRNA decreased depending on the number of R552X-carrying alleles, not on the type of primer set for analyzing individual exons (Fig. 2B and Supplementary Fig. 1). This observation provides direct evidence that the introduction of the R552X mutation triggers nonsense-mediated decay without altering splicing products. The expression of RB1 mRNA was reduced by an average of 55% and 88.7% in He-R552X and Ho-R552X cells, respectively, compared to WT. Western blotting analysis using four different antibodies recognizing spatially separated epitopes showed a decrease in RB1 protein levels in both He- and Ho-R552X cell lines, with similar changes in mRNA expression levels (average 39.4% and 80.2% reductions compared to WT, respectively) (Fig. 2C, D). These results confirmed that the R552X mutation, especially Ho-R552X, completely disrupts RB1 gene expression resulting in no detectable mutant mRNA.

Fig. 2
figure 2

Depletion of RB1 mRNA and protein expression level caused by gene knock-out in R552X cells. (A) Schematic of performed qRT-PCR using 17 primer sets, black rectangles, for amplifying exon junctions on RB1 cDNA synthesized from RB1 mutated mRNA in R552X cells. Nine independent measurements analyzed for each primer set. (B) Overall RB1 mRNA expression level of WT, He-R552X, and Ho-R552X cell lines measured by qPCR using the 17 primer sets. The average expression level of WT cells was normalized to 1, and the expression levels of He- and Ho-R552X were adjusted accordingly (mean ± s.e.m.). (C) Representative images of western blot analysis using four anti-pRb antibodies recognizing different epitopes. (D) Protein expression level of pRb measured by three or four independent western blot analysis (n = 3 or 4; mean ± s.e.m.). Four different antibodies used to recognize spatially separated epitopes. The average expression level of WT cells was normalized to 1, and the expression levels of He- and Ho-R552X were adjusted accordingly. P values were obtained using one-way analysis of variance (ANOVA) and Tukey’s test for post hoc analysis: *P < 0.05, **P < 0.01, ***P < 0.001. (E) Overall pRb expression level of WT, He-R552X, and Ho-R552X cells. The average expression level of WT cells was normalized to 1, and the expression levels of He- and Ho-R552X were adjusted accordingly (mean ± s.e.m.).

Full size image

Change of cell behavior induced by RB1 gene disruption

RB1 regulates the cell cycle by binding to E2F transcription factors and repressing expression of E2F-dependent genes36,37,38. Quantitative PCR analysis showed that the expression of E2F1, one of the principal targets of RB136,37,38, was not altered by RB1 disruption (Fig. 3A). To elucidate the molecular mechanism underlying increased cell cycle progression, we examined cell cycle regulation-engaged gene expression, including that of the E2F1 transcription factor. The expression levels of CCND1 and CDK4, which regulate RB1-E2F activity39were also slightly decreased following RB1 disruption, albeit this was not statistically significant (Fig. 3A). This suggests that growth-related genes are activated in RB1 knockout cell lines via the transcriptional activity of RB1-free E2F140. On the other hand, expression levels of CDKN1A, an inhibitor of cyclins and CDKs that progress the cell cycle, were lower in both RB1 knockout cell lines (Fig. 3A). In addition, cell proliferation was higher in both He- and Ho-R552X cells than in the WT cells (Fig. 3B). To investigate the effect that RB1 disruption has on cell cycle control in R552X cell lines, we stained the cells with PI and analyzed cell cycle progression by flow cytometry. The data showed a decreased G0/G1 phase population and an increased S phase population in Ho-R552X cells than in WT cells (Fig. 3C, D). These results demonstrate that the R552X cells exhibits tumor-like behavior compared to the WT.

Fig. 3
figure 3

RB1 gene disruption changes cell behavior and regulation. (A) Expression level of cell cycle regulation-engaged genes in WT and R552X cells measured by qPCR from nine independent experiments (n = 9; mean ± SEM). The average expression level of WT cells was normalized to 1, and the expression levels of He- and Ho-R552X were adjusted accordingly. (B) Relative cell proliferation rate. Cell expansion rate of each cell line was calculated using 15 independent measurements (n = 15; mean ± SEM) up to day 3. (C,D) Measurement of cell cycle progression changes in RB1-disrupted cell lines. Each cell line was stained with propidium iodide and each phase in the cell cycle were analyzed proportionally using flow cytometry across four independent experiments (n = 4; mean ± SEM). P values were obtained using one-way analysis of variance (ANOVA) and Tukey’s test for post hoc analysis: *P < 0.05, **P < 0.01, ***P < 0.001.

Full size image

ABE-mediated gene rescue of the RB1 mutation in the model cells

We attempted to recover disrupted RB1 function by repairing the PTC to a normal codon using ABEs, following a method presented in a previous study known as CRISPR-pass29. To this end, we designed a gRNA to convert the TGA stop codon into TGG (c.1656 A > G; p.X552W) to enable translational read-through, reduction of NMD, and production of full-length pRb (Fig. 4A, B). Two versions of ABEs, ABEmax and ABE8e, were individually used to treat R552X cells with gRNA. High-throughput sequencing data showed over 20% of stop codons were converted to tryptophan in the ABE8e treated Ho-R552X cells, with minor bystander editing (H555R) of genomic DNA. Furthermore, ABEmax treatment showed up to about 8% stop codon editing with negligible bystander editing (Fig. 4C). We then checked the ABE editing efficiency at the mRNA level. Notably, high-throughput sequencing data showed that over 80% of edited RB1-encoding transcripts were produced in the ABE8e-treated Ho-R552X cells, which showed higher editing efficiency than that of gDNA-treated ABE8e (Fig. 4D). The increasing population of edited mRNA compared with that of gDNA may result from NMD activation, which degrades mRNA carrying nonsense mutations.

Fig. 4
figure 4

Rescue of the disrupted RB1 gene by adenine base editor (ABE)-mediated base editing in R552X cell lines. (A) Schematic of the ABE, ABEmax. Adenine deaminases, wild-type and evolved (*) TadA dimer, fused to nCas9 (D10A) and it converts an exposed adenosine into a guanosine. In the case of ABE8e, only an additionally engineered TadA* monomer is fused to nCas9. (B) Schematic diagram displaying the conversion of the retinoblastoma-causing premature stop codon (PTC) to a tryptophan. PAM, ABE editing window, and targeted adenosine are highlighted in blue, red, and green, respectively. (C, D) Frequencies (%) of adenine base editing on gDNA (C) and mRNA (D) in R552X cell lines measured by targeted deep sequencing conducted across four biologically independent experiments (n = 4; mean ± SEM). P values were obtained using one-way analysis of variance (ANOVA) and Tukey’s test for post hoc analysis: *P < 0.05, **P < 0.01, ***P < 0.001.

Full size image

We then established ABE-treated He- and Ho-R552X cell lines, using the same method as described previously, to investigate the restoration of pRb and its related functions. An ABE8e-treated He-R552X cell line, which exhibited 99.2% of target A-to-G editing (c.1656 A > G) without any bystander editing (hereafter He-R552W), and an ABE8e-treated Ho-R552X cell line, which exhibited 99.6% of target A-to-G editing with 48.9% of bystander editing (a synonymous change; hereafter Ho-R552W), were selected and further characterized (Supplementary Fig. 2). As expected, He- and Ho-R552W cells exhibited increased expression of pRb compared to their corresponding controls, respectively (Fig. 5A, B). It is also notable that cell proliferation was decreased in He-and Ho-R552W cells (Fig. 5C). These results suggest that ABE could be applied as a potential therapeutic approach through RB1 gene rescue.

Fig. 5
figure 5

Restoration of pRb function in ABE-treated R552X cell lines. (A) Representative images of western blot analysis of pRb expression in ABE-treated R552X cell lines. (B) Protein expression level of pRb measured by western blot analysis (n = 4; mean ± s.e.m.). The average expression level of WT cells was normalized to 1, and the expression levels of the other cell lines were adjusted accordingly. (C) Relative cell proliferation rate. Cell expansion rate of each cell line was calculated using 6 independent measurements (n = 6; mean ± SEM) up to day 3. P values were obtained using one-way analysis of variance (ANOVA) and Tukey’s test for post hoc analysis: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Full size image

Discussion

In this study, we established an RB1 cell line carrying the R552X mutation, which is commonly found in patients with retinoblastoma. The two-step spreading and colony selection process enabled homo- and heterozygous RB1-mutant cell lines to differentiate from bulk cell lines with low editing efficiency. A cytosine base editing system was employed to mimic the patient-like mutations rather than simply disrupting RB1 function using Cas98,9,10. Furthermore, Cas9-induced indel mutation can alter the splicing of target gene, leading to the production of aberrant mRNA and proteins, which result in incomplete disruption of the target gene41. As base editing enables targeted base conversion of a single nucleotide without DSBs, it causes less cell damage to DNA lesions than intact Cas9-based gene editing. As Cas9 induces DNA translocation and large deletions in mammalian cells by inducing DSB42,43,44,45,46 base editors are used an alternative gene editing technique that can minimize genomic instability47. Our attempts to use Cas9 and ssODN to establish an RB1 mutant cell line with the R552X mutation failed due to low cell viability and editing efficiency (data not shown). We expect that cytosine-based editing systems will provide an efficient method for installing nonsense mutations into normal cells27,28especially in DSB-sensitive cells. However, in our experiments, the initial efficiency of CBE was low, which we believe was due to using an early version of CBE. Recently, improved versions of CBE, such as eA3A-BE548 or CBE6s49that maximize editing efficiency have been developed. Utilizing these improved techniques would likely allow us to establish desired cellular models with much higher success rates.

The R552X cell line showed a drastic decrease in RB1-encoding mRNA and RB1 protein expression, similar to that of He-R552X cells. In addition, cell cycle analysis revealed a higher proportion of R552X cells in the S1 phase compared to normal cells, suggesting a G1 checkpoint deficiency due to loss of RB1 function. This implied that our R552X cell line successfully mimicked the dysregulation of the cell cycle caused by RB1 loss, which is a representative phenotype of some cancer cells, although the cell lines originated from immortalized cell lines. Furthermore, various cancer cell types with RB loss, including cutaneous melanoma, lung squamous cell carcinoma, colon adenocarcinoma, head and neck squamous cell carcinoma, cervical squamous cell carcinoma, and endocervical adenocarcinoma, maintain consistent CCND1 expression50 aligning with our findings.

One limitation of this study is that we did not identify any retinoblastoma-specific characteristics in R552X cell lines. In 2D cell culture, the various conditions that cells experience in a physiological environment are not properly reflected, leading to limitations in gene expression studies. To overcome this, using 3D spheroid culture or organoid models can better replicate physiological conditions, such as cell-cell interactions, cell-matrix interactions, mechanical stress, and hypoxia51. Some reports have detailed the generation of retinal organoids that mimic the retinoblastoma phenotype9,11,12. These studies demonstrated increased cell cycle progression and tumorigenesis in RB1-disrupted retinal organoids, which are distinct characteristics of RB1-disrupted ES cells. In future studies, we plan to create RB1-disrupted retinal organoids using our CBE-based protocol to establish a mutant cell line. iPSC-derived retinal organoids that accurately model various patient-specific RB1 mutations can offer valuable insights into the role of RB1 mutations in tumorigenesis and the therapeutic efficacy of gene rescue strategies. We anticipate that various versions of ABE or prime editor (PE)52 as well as different delivery methods such as viral vectors or lipid nanoparticles, could be tested in retinal organoids to determine the optimal therapeutic strategy for correcting RB1 mutations.

Retinoblastoma develops after bi-allelic inactivation of RB1, which is commonly caused by acquired mutations in the normal RB1 allele, while the other allele carries a congenital disrupting mutation30. We found that mono-allelic disruption of RB1 also increased cell proliferation, implying that inheritance of the mono-allelic RB1 mutation itself might attenuate RB1 function, consistent with previous reports that single-copy loss of RB1 was frequently observed in RB1-mutated cancer53,54,55. Thus, correction of the RB1 mutation may be a precautionary treatment for RB1-loss-related tumorigenesis. Here, we attempted to rescue the nonsense mutation in RB1 using ABE and successfully corrected the mutation in both mono- and bi-allelic R552X cell lines, suggesting ABE could be an appropriate method to correct RB1 mutations as homology-directed repair, a typical method used for Cas9-mediated correction, was found to be ineffective in an RB1-disrupted cell line7. While our ABE-mediated gene editing method caused bystander conversion and another missense mutation, using engineered ABEs with a narrow editing window could mitigate these issues56,57. Additionally, using PE would allow for much more accurate and precise correction compared to ABE while also minimizing the bystander effect. We also expect that this gene correction strategy can be successfully accomplished with appropriate delivery vehicles, such as adeno-associated virus20,58 virus-like particles59 or lipid nanoparticles21,60,61 which may effectively convey the ABE or PE into the target tissues in vivo. In the near future, we anticipate that BEs and PEs will play a significant role in the in vivo treatment of RB1-deficient patients.