Efficacy, biodistribution and safety comparison of chemically modified antisense oligonucleotides in the retina

Abstract

Antisense oligonucleotides (AONs) are a versatile tool for treating inherited retinal diseases. However, little is known about how different chemical modifications of AONs can affect their biodistribution, toxicity, and uptake in the retina. Here, we addressed this question by comparing splice-switching AONs with three different chemical modifications commonly used in a clinical setting (2′O-methyl-phosphorothioate (2-OMe/PS), 2′O-methoxyethyl-phosphoriate (2-MOE/PS), and phosphorodiamidite morpholino oligomers (PMO)). These AONs targeted genes exclusively expressed in certain types of retinal cells. Overall, studies in vitro and in vivo in C57BL/6J wild-type mouse retinas showed that 2-OMe/PS and 2-MOE/PS AONs have comparable efficacy and safety profiles. In contrast, octa-guanidine-dendrimer-conjugated in vivo PMO-oligonucleotides (ivPMO) caused toxicity. This was evidenced by externally visible ocular phenotypes in 88.5% of all ivPMO-treated animals, accompanied by severe alterations at the morphological level. However, delivery of unmodified PMO-AONs did not cause any toxicity, although it clearly reduced the efficacy. We conducted the first systematic comparison of different chemical modifications of AONs in the retina. Our results showed that the same AON sequence with different chemical modifications displayed different splicing modulation efficacies, suggesting the 2′MOE/PS modification as the most efficacious in these conditions. Thereby, our work provides important insights for future clinical applications.

Graphical Abstract

Graphical Abstract.

Introduction

Inherited retinal diseases (IRDs) are a heterogeneous group of genetic disorders that lead to progressive photoreceptor death resulting in visual impairment. Globally, IRDs affect approximately 1 in 2,000 individuals. Currently, pathogenic variants in over 250 genes have been associated with IRDs (https://sph.uth.edu/retnet). Around 15% of these genetic defects alter the pre-mRNA splicing process (1), by either affecting the canonical splice sites or by creating new splice sites in exonic or intronic regions, leading to (partial) exclusion of regular exons or insertion of aberrant intronic regions (pseudoexons). This often results in a disturbed reading frame and subsequent reduction of protein, thereby affecting cell function (2). The high genetic heterogeneity of these diseases impairs the development of a common treatment strategy. However, in the last decade, the development of novel personalized approaches has significantly impacted the field.

One of these treatment strategies is based on antisense oligonucleotides (AONs) and aims to correct splicing defects. Amongst other functions, AONs can modulate splicing, and thereby are capable of restoring protein levels, for example by hampering the accessibility of splicing factors to their recognition sites (3–5). Therefore, AONs have been postulated as promising molecules to correct aberrant splicing in IRDs (6). This is supported by the significant increase of preclinical and clinical proof-of-concept studies to correct these splicing defects, e.g. for CEP290 (7), ABCA4 (8–15), USH2A (16,17), CHM (18) and OPA1 (19).

Although the use of AONs has provided encouraging results as promising treatments to correct aberrant splicing-related subtypes of IRD, several questions still need to be addressed to increase their current application in the eye. Most of these remaining challenges are related to delivery, uptake, longevity, and off-target effects. Particularly, little is known about the effect of the chemical modifications of these AONs on these parameters when they are delivered to the retina, a complex multi-layered neuronal tissue. Furthermore, no comparison between chemically modified AONs in the retina has been published so far, and very few studies have investigated the biodistribution, uptake, and toxicity in this tissue.

In this study, we have conducted a systematic comparison to evaluate the specific effect of AONs in the retina by using three commonly used AON chemical modifications for clinical application: 2′O-methyl-phosphorothioate (2′OMe/PS), 2′O-methoxyethyl-phosphoriate (2′MOE/PS), and phosphorodiamidate morpholino oligomers (PMO), which in this manuscript will be referred to as 2′OMe, 2′MOE and uPMO (unmodified PMO) or ivPMO (in vivo PMO), depending on whether they are unconjugated or conjugated, respectively (Figure 1). We identified and assessed AON-mediated splicing modulation capacity for several genes only expressed in specific retinal cell types, using in vitro cellular systems employing midigenes as well as in vivo model systems (wild-type mice). Our results provide novel insights on how different chemical modifications of AONs behave in the retina, contributing to improving the development of AON-based treatments for IRDs.

Figure 1.

General overview of the chemical modifications employed for the in vitro and in vivo experiments performed in this study using a single dose. The schematic representation indicates the chemical modification, linkage and the presence of conjugates (if applicable). Furthermore, the two columns on the right indicate the concentration employed and the delivery method in each study. Asterisk (*) means that this condition was only employed with Grm6 and SON oligonucleotide sequences.

Materials and methods

Selection of genes and AON design

To assess the possible preferential uptake of the antisense AON sequences by specific retinal cells, we selected two genes exclusively expressed in four different retinal cell types involved in IRDs: photoreceptor (cones and rods), bipolar cells and retinal pigment epithelium (RPE) cells. In total 8 genes were selected (Supplementary Table S1). For AON design, exons divisible by three were selected as targets to maintain the open reading frame of the transcript, preventing the degradation of the newly-generated transcript by nonsense-mediated decay (NMD) and the underestimation of the splicing modulation efficacy. Only when no eligible in-frame exon was possible, we selected an out-of-frame exon. AON design was performed as previously described (20–22). In total, between 2 and 6 different AON sequences targeting specific exons per gene were designed. A scrambled oligonucleotide (SON) was also included as a negative control (Supplementary Table S2). Screening of these AON sequences was first conducted in vitro using the 2′OMe chemical modification. Subsequently, the most promising AON per gene was selected for comparison with the other two chemical modifications. 2′OMe and 2′MOE AONs for in vitro studies were obtained from Eurogentec (Liège, Belgium), and for in vivo studies from ProQR Therapeutics (Leiden, the Netherlands). PMOs were acquired from Gene Tools, LLC (Philomath, OR, USA) for both in vitro and in vivo studies.

Design of the midigenes

We first analyzed the endogenous expression of all selected genes in several mouse cell lines. All the genes were either not expressed or expressed at very low levels, limiting our readout measurement at the RNA level (Supplementary Figure S1). Therefore, midigenes (splice reporter vectors) harboring the genomic region of interest of each gene were generated using the Gateway Cloning system (Thermo Fisher Scientific, Waltham, MA, USA), following the instructions of the manufacturer. Briefly, mouse genomic DNA was used as a template to amplify the different regions of interest of the eight selected genes using primers with attB-tails (Supplementary Table S3). Subsequently, amplified regions were cloned into the pDONR™201 plasmid (Invitrogen, Waltham, MA, USA). Finally, the inserts were either introduced in the pcDNA3 or pCI-Neo-RHO splicing vectors, depending on whether they included exon 1 (Rho, Gnat2, Opn1sw and Rdh5) or another genomic region of the respective gene (Pde6a, Best1, Grm6 and Prkca). The use of pcDNA3 allowed the expression of the gene from the CMV promoter, despite that exon 1 was not fully spliced (it presents a splice donor site but not an acceptor site). In contrast, pCI-Neo-RHO plasmid was employed as a splicing vector due to the exons 3 and 5 of the RHO gene flanking the region of interest (10,23–25). All midigenes were verified by restriction analysis and Sanger sequencing. Primer sequences are provided in Supplementary Table S3. Midigenes were further validated by transfecting them in mouse cell lines and analyzing the splicing pattern. To allow further AON validation on the Opn1sw midigene, it had to be modified by site-directed mutagenesis (Supplementary Methods).

Cell culture

Murine Inner Medullary Collecting Duct (mIMCD3, ATCC # CRL-2123) cells were cultured in Dulbecco's Modified Eagle medium (DMEM)/Ham's F12 medium (1:1 v/v). Before mixing both mediums, Ham's F12 was supplemented with 1% of alanine–glutamine. Human embryonic kidney 293T (HEK-293T, ATTC #CRL-3216) cells and 661W cells were cultured in DMEM medium. Both DMEM/F12 and DMEM mediums were supplemented with 10% fetal calf serum (FCS), 1% sodium pyruvate, and 1% penicillin–streptomycin. In addition, the 661W culture medium was also supplemented with 40 μl/l of β-mercaptoethanol. Cells were grown at 37°C and 5% CO₂.

Midigene and different modified AON/SON transfections

Approximately 300,000 mIMCD3 or 400,000 HEK293T (only for the Prkca midigene) cells were transfected with 1.2 μg of midigene in a 6-well plate using FuGENE®-HD Transfection Reagent (Promega, Madison, WI, USA) following a 3:1 ratio (FuGENE® HD Transfection Reagent:DNA). FuGENE®/DNA mixture was conducted in Opti-MEM reduced serum medium (Gibco, Waltham, MA, USA) and delivered drop by drop to the cells. On the next day, cells were split in a 12-well plate using a 1:6 dilution. After reattachment (around 4 h after splitting), cells were transfected with the AONs. A final concentration of 0.5 μM of 2′OMe and 2′MOE modified AONs/SON were transfected using FuGENE®-HD Transfection Reagent (Promega) as indicated for the midigene transfections. However, PMO antisense/scrambled oligonucleotides could not be delivered using a lipid-based agent in vitro. In this case, a final concentration of 2.5 μM unconjugated PMO-modified (uPMO) oligonucleotides were mixed with 500 μl of fresh medium and added to the cells whose medium had been previously removed. The cells were then scraped gently until all cells were mechanically detached and placed in a new well (26). Independently of the modification, transfected cells were collected 48 h after AON transfection. In the cases in which the entire gene was included and a non-multiple of three exons was targeted (Rdh5 and Gnat2), cells were treated with cycloheximide (CHX) (final concentration 100 μg/ml) for 4–6 h to block possible non-sense mediated decay (NMD) before sample collection.

Ethics statement

In vivo studies in mice were performed with approval from the Dutch Central Committee on Animal Testing (Centrale Commissie Dierproeven, AVD10300 2016 758 (RU-DEC-2016-0050) and AVD11400 2020 11085), the Radboud University Nijmegen Animal Welfare Committee (work protocols 2016–0050-16 and 2016-0050-21), as well as the combined Animal Welfare Committees of Amsterdam University Medical Centers and Radboud University Nijmegen work protocol (work protocol 2024-011-001). The research was conducted according to the regulations of the ARVO statement for the use of animals in ophthalmic and vision research. All procedures were carried out in the Netherlands.

Animals and intravitreal injection of the different-modified oligonucleotides

Given the large experimental set-up, injections were divided into three batches. The first batch included one target per cell type (Rho, Gnat2, Rdh5 and Grm6), together with the negative control (SON) and the second batch included the remaining targets (Pde6a, Opn1sw, Best1 and Prkca). The four most efficacious molecules with reference to their ability to modulate splicing were selected for a third batch of injections in which a dose-response analysis was performed.

Intravitreal injections were performed on 6-to-8-week C57BL/6J wild-type mice. Injections were performed as previously described (24,27). Briefly, analgesia (carprofen) was provided in the drinking water 24–48 h prior to the intervention and the 24–48 h post-intervention. On the day of the intervention, animals were anesthetized using isoflurane. Two microliters of each chemically modified antisense and scrambled oligonucleotide (20 μg of molecule for batches 1 and 2) were injected into the right eye, whilst 2 μl of sterile 1x PBS were injected in the left eye (27). Of note, all the AONs employed in this experiment were ordered specifically for in vivo administration and were blinded prior to delivery. In the particular case of the PMO, in vivo PMO (ivPMO) with the same sequences employed in vitro were ordered from GeneTools (Philomath, OR, USA). In this in vivo PMO molecule, the oligonucleotide is linked to an octa-guanidine dendrimer (28). Animals were examined during the first 24–72 h post-intervention. Retinas were harvested 7 days after injection. In total 13 animals per group were used, 10 of which were employed for RNA analysis (each replicate consisted of the pool of two retinas, providing five replicates), while three animals were used for morphological analyses.

For the dose-response analysis, we injected 2 μl of AON (2′MOE and 2′OMe) at variable doses (see Supplementary Table S4) in the right eye and 2 μl of sterile 1× PBS in the left eye. In this case, only RNA studies were performed using five animals per condition, except for the 20 μg condition (n = 3). Injection of the SON modified with both chemistries was used as a control. To reduce the number of animals, all SON-injected animals received 40 μg of SON (the highest dose used). Animals were sacrificed 5 days post-injection and retinas were subjected to RNA analysis.

RNA analysis

For both in vitro and in vivo samples RNA was isolated using the Nucleospin RNA kit (Machery Nagel, Duren, Germany) according to manufacturer's instructions. For the in vivo experiments, retinas were first disrupted using a polytron. Total RNA concentrations were measured using Nanodrop 2000 and 1 μg of total RNA was used for cDNA synthesis using iScript™ cDNA synthesis kit (Bio-Rad, Hercules, CA, USA) following the manufacturer's recommendations. Next, a total of 60 ng of cDNA was used as a template to amplify the region of interest in each construct. As loading control, Actb was also amplified. In addition, when pCI-Neo-RHO plasmid was used, exon 5 of RHO was amplified as transfection control. The complete list of primers is detailed in Supplementary Table S3. All PCR mixtures also contained 1× PCR buffer with MgCl₂ (Roche, Manheim, Germany), 1x Q-solution (QIAGEN, Hilden, Germany), 2.5 mM of MgCl₂, 2 μM of dNTPs, 0.2 mM of each primers and 0.5 U Taq DNA Polymerase (Roche, Mannheim, Germany). The PCR program followed was: 5 min at 94°C, followed by 35 cycles of 30 s at 94°C, 30 s at 60°C and 1 min at 72°C, with a final extension of 10 min at 72°C. This program was identical for all the tested genes except for Opn1sw and Best1 midigenes, for which the elongation time was 35 and 90 s, respectively. In addition, Gnat2 (midigene and in vivo studies) was studied by a touch-down RT-PCR which program was: 5 min at 94 °C followed by 10 cycles of 30 s at 94 °C, 30 s of progressive decreasing annealing temperature (0.5 °C/cycle) from 65 °C to 60 °C, and 1 min at 72 °C 1 min. This was followed by 25 cycles 30 s at 94°C, 30 s at 60°C and 1 min at 72°C, with a final extension of 10 min at 72°C. PCR products were resolved by electrophoresis and all the transcripts were verified by Sanger sequencing. Semiquantitative analysis was conducted using Fiji software (29).

Exon skipping in Prkca, Pde6a, Best1 and Opn1sw genes was further studied using qPCR with the GoTaq Real-Time Quantitative PCR Master kit (Promega). Reactions were performed in triplicate on the Applied Biosystem QuantStudio 5 Digital system (Applied Biosystem, Waltham, MA, USA). Each sample was normalized to the housekeeping gene (Gusb) and compared to the untreated condition (PBS-injected eye) of the same animal. Primer details are listed in Supplementary Table S3.

Morphological and immunohistochemical analysis

Dissection of the eyes was performed as described elsewhere (24,27,30). Briefly, eyes were enucleated and subsequently fixed in 2% PFA for 10 min at RT. The cornea and lens were removed and eyecups were fixed in 2% PFA during 2 h at RT. After that, samples were washed at least three times in abundant 1× PBS and subsequently incubated in 10% sucrose for 1 h at 4°C, followed by 1 h at 4°C in 20% sucrose solution, and finally overnight at 4°C in 30% sucrose solution. After that, eyecups were correctly oriented and embedded in OCT (Tissue-Tek, Sakura Finetech, Torrance, CA, USA) (27). Sections of 7 microns were employed for morphological studies and immunohistochemical analyses. Morphology was assessed by staining the sections with toluidine blue. For immunostaining, cells were permeabilized for 20 min in PBS-Tween 0.01% and then blocked for 30 min with blocking buffer (1× PBS, 0.1% ovalbumin (w/v) (Sigma-Aldrich) and 0.5% fish gelatin (w/v) (Sigma-Aldrich)), and incubated overnight with the primary antibodies. After washing 4 times with 1× PBS, cells were incubated with the secondary antibodies and DAPI 1 h (RT) and washed again before mounting them with prolong gold antifade kit (Life Technologies, Carlsbad, CA, USA). Employed antibodies are indicated in Supplementary Table S5. Images were taken with an upright trinocular fluorescence microscope (Zeiss Axio Imager Z1 Fluorescent; Zeiss, Aalen, Germany). Outer segment layer (ONL) thickness was measured with Fiji software. A comparative analysis was performed on the treated and untreated eyes of the same animal.

microRNAscope

To evaluate the distribution and uptake of the AONs, Grm6 AONs were used as a proof-of-concept. Grm6 AON was selected as it was the most suitable sequence to generate a probe with no potential off-targets, providing a cleaner readout. MicroRNAscope (Bio-Techne, Minneapolis, MN, USA) analysis was conducted following the manufacturer's instructions. Briefly, samples were post-fixed in 4% PFA, followed by ethanol gradient (from 50 to 100%) and 12% formaldehyde. Then, samples were permeabilized with hydrogen peroxide. To allow the probe to target the entire tissue, a retrieval step of 15 min was conducted. After that, proteinase III was added for 30 min at 40°C. Positive, negative, and Grm6 probes were added and incubated for 2 h at 40°C. The signal was amplified six times followed by Fast-Red staining and DAPI counterstain. Samples were mounted with prolong gold antifade kit (Life Technologies). The cell imaging was taken on a Zeiss Axio Imager Z1 fluorescent microscope.

Statistical analysis

The total value of the different transcripts was used to normalize the percentage of expression of all transcripts in each condition. A one-way ANOVA test followed by Bonferroni correction was used to assess significant differences between conditions in the RT-PCR studies. As the ONL thickness can vary between animals (due to e.g. weight or sex), a paired t-test was employed to compare ONL thickness between treated and untreated eyes from the same animal. This was applied to all animals injected with the different chemical modifications. All the statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). P-values <0.05 were considered statistically significant.

Results

The selected genes were lowly expressed in regular mouse cell lines

Our objective was to assess the influence of AON-specific chemical modifications on splicing modulation in photoreceptors (rods and cones), RPE, and bipolar cells. For that purpose, the selected genes were: Pde6a and Rho for rod cells, Gnat2 and Opn1sw for cone cells, Best1 and Rdh5 for RPE cells, and Grm6 and Prkca for bipolar cells. We subsequently evaluated their endogenous expression levels in murine 661W and mIMCD3 cell lines. However, most of these genes showed very low levels of expression (Supplementary Figure S1), hampering the assessment of exon skipping levels using the endogenous transcripts. Therefore, we employed a system based on the use of midigenes, to express (partial) transcripts of the target genes to be able to evaluate the splicing modulation potential of each AON robustly (12,31,32).

Successful exon-skipping using midigene-based splicing assays was achieved with 2′OMe-AONs

The midigene system has shown to be an excellent tool for AON screening, even when the entire genomic context is missing (12–14). Midigene constructs for the selected genes (Supplementary Table S1) were generated and transfected into mIMCD3 or HEK293T (only for Prkca midigene) the splicing pattern was compared to the one of the mouse retina. In some cases, midigenes contained the full gene (Rho, Opn1sw, Gnat2and Rdh5, Figure 2, upper panel), in other cases a specific region of interest of the target gene (Pde6a, Best1, Grm6and Prkca, Figure 2, lower panel). Once all midigenes reported the expected splicing patterns (see Supplementary Results and Supplementary Figure S2), an AON screening to identify successful exon skipping levels was conducted with 2′OMe-AONs.

Figure 2.

Comparison of the in vitro splicing modulation effect mediated by AONs with different chemical modifications. In the upper panel, the schematic representation and results of the four full genes cloned into pcDNA3 construct under the CMV promotor. In all cases, 2′OMe and 2′MOE oligonucleotides were transfected at a final concentration of 0.5 μM while uPMO oligonucleotides were mechanically delivered at a final concentration 5 times higher (2.5 μM). On the panel below, the schematic representation and results of the four midigenes containing partial genomic regions of each of the targets cloned in between exons 3 and 5 of RHO gene using the pCI-Neo-RHO vector. In all cases the schematic view of the construct is provided with the AON location and the amplification primers (black arrows). Below the map, a representative electrophoresis picture of the RT-PCR analysis upon delivery of modified AON or scrambled oligonucleotide (SON) is shown. # indicates that the transcript has been classified as a PCR artifact. This classification came after verifying the sequencing results, which showed that these transcripts do not correspond to a gene/transcript of interest. Below the gel, semi-quantification of the RT-PCR products is provided as the percentage of wild-type (WT) and exon skipped (ES). ES values includes all splicing-related events, including partial insertions (Pi) or exon elongations (indicated as + nucleotides (nt)). MQ indicates the negative control of the PCR. Actb was amplified as loading control. Exon 5 of RHO was amplified as transfection control when the pCI-Neo-RHO vector was used. Each bar is represented as the mean ± SD. Statistical significance with respect to the untreated condition (UT) is indicated as * P < 0.05, ** P < 0.01 and *** P < 0.001 using one-way ANOVA followed by Bonferroni correction. ‘ns’ means ‘no statistical significance’. In the Grm6 midigene, AONs target exon 8. However, untreated midigene already produced a transcript in which this exon is naturally skipped, similar to what occurs endogenously in the mouse retina. After AON delivery, an increase in this existing transcript, rather than the appearance of a new transcript, as seen for other midigenes.

For each target, at least two AON sequences per targetable exon were designed and tested (Supplementary Figures S3 and S4). In all cases, at least one AON successfully led to partial or complete skipping of the targeted exon. The best AON sequence (Supplementary Table S6) for each target was used to establish the comparison between the different chemistries in vitro and in vivo. Of note, the AON sequence selected for Grm6 targets exon 8, which can be naturally skipped, as shown in both the midigene system and the mouse retina.

Splicing modulation capacity varied amongst chemical modifications in vitro

Upon selection of the best sequence, we proceeded with a systematic comparison of each chemically modified AON for each individual target (Figure 2). Overall, the 2′MOE AONs seemed to be the most efficacious molecules in almost all conditions except for Pde6a (Table 1A, Supplementary Material). The 2′OMe AONs showed similar efficacies, however some differences were observed for some targets (Figure 2). In general, the uPMO-modified AONs showed the least efficacy, although the concentration used was higher due to the associated delivery issues when using uPMO in vitro. While Endoporter, a weak-base amphiphilic peptide (33), has shown improved delivery for cultured cells (34,35), in our hands it did not work. However for Gnat2 and Rdh5, the efficacies of the uPMO AONs were in line with the exon skipping promoted by the 2′MOE and 2′OMe-modified AONs. For some target genes, the interpretation of the results was challenging due to the multiple splicing patterns detected in the overexpression midigene assay (Gnat2 and Rdh5). Furthermore, for Gnat2 uPMO-AON and uPMO-SON delivery increased the transcript skipping from exon 3 to exon 7, indicating a potential chemistry-related effect for this particular target under our conditions. Similarly, for the Rdh5-midigene, the uPMO-SON resulted in a unique splicing pattern (Figure 2). Overall, these differences could be attributed to the difficulties in obtaining a clear readout or to technique's preference for certain transcripts, rather than a clear chemistry-related influence. Analysis of the sequence of the SON did not reveal any potential sequence similarity with the respective targets. In the other six cases, the chemically-modified scrambled SON did not show any difference compared to the non-transfected conditions.

Table 1.

Overview of the splicing modulation capacity of the different chemically-modified AONs in vitro (A) and in vivo (B)

A	In vitro studies			B	In vivo studies
	Splicing modulation capacity				Splicing modulation capacity (at 20 μg). Treated eye
Midigene	2′ OMe-AON	2′MOE-AON	uPMO-AON	Target gene	Retinal cell in which target gene is expressed	2′ OMe-AON	2′MOE-AON	ivPMO-AON
Rho	93.3 ± 5.5% ES	99.9 ± 0.10% ES^	57.0 ± 1.78% ES	Rho	Rod cells	19.1 ± 0.30% ES^	18.4 ± 0.68% ES^	7.7 ± 0.42% ES
Gnat2	45.7 ± 0.81% ES	76.7 ± 0.58% ES^	46.3 ± 0.46% ES	Gnat2	Cone cells	63.4 ± 0.23% ES	86.7 ± 1.03% ES^	17.6 ± 0.06% ES
Grm6	41.5 ± 0.07% ES^	43.5 ± 0.21% ES^	41.0 ± 0.42% ES^	Grm6	Bipolar cells	49.9 ± 0.01% ES	74.7 ± 0.62% ES^	25.3 ± 0.25% ES
Rdh5	90.5 ± 0.35% ES^	89.0 ± 0.71% ES^	84.0 ± 0.71% ES^	Rdh5	Retinal pigment epithelium cells	92.4 ± 0.05% ES^	84.3 ± 0.10% ES	25.4 ± 0.33% ES
Pde6a	11.5 ± 0.05% ES^	3.0 ± 0.01% ES	2.0 ± 0.01% ES	Pde6a	Rod cells	No ES detected. 10.9 ± 0.92% of WT reduction	No ES detected. 20.6 ± 0.48% of WT reduction^	N/A
Opn1sw	38.1 ± 0.22% ES	52.4 ± 1.14% ES^	23.3 ± 1.01% ES	Opn1sw	Cone cells	48.3 ± 0.75% ES	55.9 ± 0.83% ES^	N/A
Best1	17.3 ± 0.03% ES	47.2 ± 0.52% ES^	0.8 ± 0.11% ES	Best1	Retinal pigment epithelium cells	No ES detected. 69.5 ± 1.36% of WT reduction^	No ES detected. 62.4 ± 0.18% of WT reduction^	N/A
Prkca	30.0 ± 2.0% ES	39.9 ± 1.02% ES^	2.7 ± 0.06% ES	Prkca	Bipolar cells	No ES detected. 39.4 ± 1.24% of WT reduction^	No ES detected. 20.5 ± 0.21% of WT reduction	N/A

Values highlighted with a circumflex (^) indicate the most potent chemical modification for each midigene (A) or targeted gene (B) based on the statistical analysis represented in Figure 2 and Figures 6 and 7, respectively. When multiple chemistries are highlighted with (^) for the same target, it indicates no statistical differences and therefore similar efficacies are considered. In section B, the values for ivPMO-AON splicing modulation capacity are estimated in retinas showing toxic effects, therefore this value may not be representative of the true potency of the ivPMO-AON molecule itself. ES indicates exon skipping. WT means wild-type transcript. N/A indicates not applicable.

In vivo evaluation of AONs with different chemical modifications displayed different efficacy and safety profiles

After the in vitro characterization, we proceeded to the in vivo analysis. For that, wild-type C57BL/6J animals were injected with the different chemically-modified AONs. As indicated in the methods, due to the large experimental set-up, the different conditions were divided into two batches of animals: Batch 1: Rho, Gnat2, Rdh5, Grm6 and SON, and Batch 2: Pde6a, Opn1sw, Best1, and Prkca.

AONs were delivered intravitreally and animals were followed up during the subsequent 72 h. In the first batch, five out of fifteen groups of animals showed consistent presence of dilated pupils, accompanied in some cases with cataracts only in the injected eye (Figure 3). These pupils were unresponsive to light exposure and remained dilated during the entire course of the experiment, even when exposed to bright light, indicating possible damage to the retina and/or sympathetic nervous system (36). All the other groups did not show any external phenotype. When samples were unblinded, all affected animals belonged to the in vivo PMO group. These results showed that the ivPMOs when delivered to the retina are somehow toxic independently of the target, as the SON also showed the same phenotypic characteristics. Furthermore, the PBS-treated eye and all the other 2′MOE/2′OMe-treated eyes did not lead to any visible external alteration. These observations were further supported by the fact that enucleated ivPMO-treated eyes for RNA isolation or morphological studies were smaller and whiter compared to the normal retinas (Supplementary Figure S5).

Figure 3.

Phenotypic characteristics upon delivery of 20 μg of chemically modified AON to the mouse retina. (A) Representative pictures of mice injected with the different chemically-modified AONs designed to target Gnat2. Mice injected with the 2′OMe or 2′MOE AONs (right) did not present any apparent phenotypical abnormal characteristic in comparison with the contralateral eye (left) injected with PBS. Animals injected with the ivPMO modification (linked to an octa-guanidine dendrimer) showed a clear ocular phenotype in the right eye already after 24 h. (B) Overview of the visible ocular phenotypes detected in the animals 72 h after injection. The X-axis of the graph indicates the different chemical medication of the corresponding AONs or scrambled oligonucleotide (SON). Eyes injected with either 2′OMe or 2′MOE-AONs did not present any effect (NE; no effect in white), while the eyes injected with ivPMO-AONs presented dilated pupils (orange) or cataracts (blue). Of note, the ivPMO is conjugated to an octa-guanidine dendrimer.

PMO-AON delivery resulted in significant morphological abnormalities, while 2′OMe and 2′MOE-treated eyes presented normal morphology

To assess the nature of these phenotypic characteristics and also validate no other potential toxic events that could compromise the retina, morphological studies were conducted. Toluidine blue stained sections showed normal layer distribution and proportion after injection of 2′OMe and 2′MOE AON/SON (treated eye) in comparison with the PBS control (untreated eye) of the same animals. However, in most ivPMO-based AON/SON eyes the retinal layers were, at least partially, detached from the choroid and RPE. In addition, the structure of the retinal layer was also less defined in comparison with the untreated eye and with the other chemical modifications (Supplementary Figure S6). Complementary immunohistochemical analyses were conducted to fully understand how retinal layers were affected. In these studies, no major changes in the treated and untreated retina were observed, except for the ivPMO-treated eyes. Here, most of the retinas showed changes in the layer pattern with a mislocalized staining of Rho and PNA (Rho, Gnat2, and SON, Figure 4). This effect was most severe in Gnat2 ivPMO-AON treated animals in which the complete retinal layer structure was lost. In the case of Grm6 ivPMO-AON, all layers were clear but the ONL showed a significant reduction in thickness (Supplementary Figure S7). Interestingly, the injection of the ivPMO against Rho caused damage mostly located in one half of the eye but not in the other half (Supplementary Figure S6; Rho PMO-AON lower half of the retina). This particular observation was found in two out of the three injected eyes. Therefore we could not discard that this effect could be related to the ivPMOs distribution after the injection procedure. However, this effect was not observed with the 2′OMe/2′MOE AONs. The values for the ONL thickness were obtained from the non-damaged area, as in the other sections there was no ONL to be measured. For that reason, despite observing a reduction of the ONL thickness, we did not observe any significant difference between PBS- and ivPMO-treated eyes in this specific condition. In general, we noticed a significant reduction of the ONL after ivPMO-AON treatment (Supplementary Figure S7). Given these results, we decided to not use ivPMO AONs for the injection of the second batch of animals.

Figure 4.

Immunocytochemical (ICC) analysis on mouse retinas for a subset of targets. Morphology was analyzed by ICC using several markers. Each row include samples from the same group of mice treated with the same AON or scrambled oligonucleotide (SON) sequence. Each column indicates either the chemical modification of the injected AON or the PBS injected eye (untreated). The untreated and treated eye for each chemical modification belong to the same animal. GFAP (green) marker is associated with gliosis and retinal stress. In purple, RHO marker to stain the rod photoreceptors. PNA (red) was used as a cone marker. Nuclei were stained with DAPI (blue). Scale bar represents 50 μm. Retinal layers are indicated in all PBS-injected eyes meaning retinal pigment epithelium (RPE), outer-segment of photoreceptors (OS), outer nuclear layers (ONL), inner nuclear layer (INL) and ganglion layer (GL). Schematic representation of the retinal layers based on DAPI staining is placed on the right. All treated eyes were injected with 20 μg of the chemically modified oligonucleotide. Of note, the ivPMO is conjugated to an octa-guanidine dendrimer.

Splicing modulation evaluation revealed that 2′MOE and 2′OMe-AONs are the most efficacious chemical modifications in the retina

Retinas of the different groups were subjected to RNA isolation and exon skipping evaluation (Figure 5, Supplementary Figure S8). Like in the in vitro studies, 2′OMe and 2′MOE modified AONs showed the highest splicing modulating capacity in vivo, while the ivPMO AON-treated samples showed reduced exon-skipping efficacy. In none of the tested genes the SON presented differences between the untreated (PBS) and treated (AON) eye, indicating that the effects detected in some of the midigenes previously (Figure 5B; Gnat2) were related to a possible artifact due to the overexpression of the midigene (Figure 5A–D). The only gene showing bigger discrepancies in comparison to the midigene system was Rdh5(Figure 5C, Supplementary Figure S8C). In the midigene system, there were no differences between the different modified AONs while in the in vivo situation 2′MOE, and especially 2′OMe were more efficacious than the ivPMO-modified AON.

Figure 5.

Splicing modulation evaluation in the first batch of mice treated with 20 μg of the different chemically-modified AONs. All treated eyes were injected with 20 μg of the chemically modified oligonucleotide. Semi-quantitative analysis by RT-PCR of the splicing modulation effect for four targets is depicted: (A) Rho; (B) Gnat2; (C) Rdh5; (D) Grm6. Representative electrophoresis pictures are provided in Supplementary Figure S8. On the left (A–D) the effect of the different modified AONs (2′OMe, 2′MOE and ivPMO) in the treated eye (T) and the untreated eye (U), injected with PBS. On the right, each target was amplified in the eyes treated with the scrambled oligonucleotide (SON) (T) or the contralateral eye injected with PBS (U). Graphs represent the percentage of the wild-type transcript (WT) with those transcript which present exon skipping (ES). Each bar is represented as the mean ± SD (n = 5, two technical replicates). Statistical significance with respect to the untreated condition (UT) is indicated as * P< 0.05, ** P< 0.01 and *** P< 0.001 using one-way ANOVA followed by Bonferroni correction. ‘ns’ means ‘no statistical significance’. Of note, the ivPMO is conjugated to an octa-guanidine dendrimer.

As indicated before, in the second batch (Pde6a, Opn1sw, Best1 and Prkca) PMO AONs were not delivered due to the toxicity observed in the first batch. In this case, splicing modulation was not observed in three out of the four target genes (Figure 6A–D, respectively, and Supplementary Figure S9, Table 1B). Only for Opn1sw (Figure 6B), we observed a clear splicing modulation by the 2′OMe and 2′MOE, at a similar efficacy. For the other three genes, however, we noticed a reduction of wild-type transcript only in the treated eyes (Figure 6, middle column, and Supplementary Figure S9). Semi-quantitative and quantitative analysis by RT-PCR (Figure 6, middle column) and qPCR (Supplementary Figure S10), respectively, revealed lower levels of wild-type transcript in the three cases (Pde6a, Best1 and Prkca), indicating that these AONs somehow reduced the wild-type transcript expression, perhaps by inducing NMD or other type of transcript degradation mechanism.

Figure 6.

Splicing modulation evaluation in the second batch of mice treated with 20 μg of either 2′OMe or 2′MOE AONs. All treated eyes were injected with 20 μg of the chemically-modified AON. Semi-quantitative analysis by RT-PCR of the splicing modulation effect for four targets is depicted: (A) Pde6a; (B) Opn1sw; (C) Best1; (D) Prkca. Corresponding electrophoresis pictures are provided in Supplementary Figure S9. On the left graphs (A–D) the effect of the different modified AON (2′OMe and 2′MOE) in the treated eye (T) and the PBS-treated eye (U). On the middle panel, the expression value of the wild-type transcript was normalized against the Actb expression value, due to an unaltered splicing pattern, with a noticeable reduction of the WT transcript (Supplementary Figure S9). Treated samples (T) with the AON were normalized against the value of the untreated eyes (U), which was set as 1. Of note, this batch of genes did not include ivPMO-modified AONs, due to the toxic effects observed in the first batch. On the right column, each target was amplified in the eyes treated with the scrambled oligonucleotide (SON) or the contralateral eye injected with PBS (U). In the graphs of the first and last columns, the ratio between the wild-type transcript (WT) exon skipping (ES) transcript is depicted. In all graphs, each bar is represented as mean ± SD (n = 5, two technical replicates). Statistical significance with respect to the untreated condition (U) is indicated as * P< 0.05, ** P< 0.01 and *** P< 0.001 using one-way ANOVA followed by Bonferroni correction. ‘ns’ means ‘no statistical significance’.

Taking all results together, 2′OMe and 2′MOE modified AONs were similarly efficacious for all targets, with a trend favouring the 2′MOE chemistry. To further elucidate the efficacy of the two chemistries, a dose-response analysis for Rho, Gnat2, Rdh5 and Grm6 was performed. (Table 1B).

2′MOE chemical modification seems to be more efficacious in the neuroretina upon dose-response delivery of AONs

For the selected four target genes (Rho, Gnat2, Rdh5 and Grm6) a dose-response analysis was performed using the two chemistries. In total, we assessed three doses (Supplementary Table S4), in which the 20 μg dose was included as a reference and the other two doses adapted based on the results presented in Figure 5. Given the high levels of exon skipping observed already at 20 μg, for Gnat2, Rdh5, and Grm6 we opted for two lower doses (10 and 5 μg). In contrast, for Rho a higher and a lower dose (40 and 10 μg) were selected. In all cases, a dose-response effect on exon skipping was observed to some extent for both chemistries (Supplementary Figure S11). When checking the results, 2′MOE chemistry showed higher levels of exon skipping at the lowest dose for the two photoreceptor targets (Rho and Gnat2) while the 2′OMe was somehow more potent for Grm6 and Rdh5.

Delivery of unmodified PMO-AONs showed non-toxic effects but low efficacy

Given the toxic effects observed by the in vivo PMO, we aimed to elucidate if the PMO-oligonucleotide itself or the octa-guanidine dendrimer moiety that is conjugated to the in vivo PMO was responsible for the observed toxicity. For that, we decided to use two AONs in a small experiment with only two groups of animals in which we injected the unmodified PMO (uPMO) AON (without the octa-guanidine dendrimer) against Grm6 and the uPMO SON. During the first 72 h post-injection and at the end-point, no phenotypic characteristics were observed. In fact, the collected eyes and retinas were comparable in colour and size to the ones injected with PBS (Supplementary Figure S12A). Toluidine blue stained sections did also not reveal any morphological differences (Supplementary Figure S12B). This was supported by the ICC results (Supplementary Figure S12C). However, the splicing modulation capacity of the uPMO AONs was barely detectable (Supplementary Figure S11A). Semi-quantification of the bands showed a small but significant increase of exon 8 skipping only after treatment with the uPMO AON (Supplementary Figure S13B). However, this effect was much less prevalent than the one caused by the in vivo PMO AON (Figure 5D). Hence, uPMO AON was less efficacious than the in vivo PMO, but it did not lead to any detectable toxic effect.

AON uptake was detected in all retinal layers upon intravitreal delivery independently of the AON chemical modification

The microRNAscope analysis revealed that both 2′OMe and 2′MOE modified AONs can reach all retinal layers. In contrast, the in vivo PMO AON was mainly detected in the inner nuclear layer (INL) (Figure 7). When we compared the intensity of the signal between 2′OMe and 2′MOE AON in the different layers, for the 2′OMe it was slightly higher in the INL, bipolar, and RPE layer, and less intense in the ONL. No differences between layers were observed in the 2′MOE modification. Unlike the previous chemistries, the unmodified PMO (uPMO) did not present any positive staining for any of these layers, suggesting that the uptake of the uPMO-AON was below the detectable levels one week after injection.

Figure 7.

MicroRNAscope analysis of mouse retinas from animals treated with 20 μg of the different modified (2′OMe, 2′MOE, ivPMO and uPMO) antisense oligonucleotides (AON) against Grm6. Retinal layers were labelled using DAPI staining (blue). The positive signal for the probe (either positive control or the probe detecting the Grm6 AON) is visualized in red. On top, it is provided a representative image of the RNAscope negative and positive controls as suggested by the provider. In the middle panel a representative eye for each Grm6-modified AON is depicted. The bottom row shows representative pictures of the contralateral eye treated with PBS of the first batch of animals treated with Grm6-AONs (left) or from the an animal treated with the unmodified-PMO (uPMO). As expected, no signal for the probe was observed in the PBS-treated eyes. In all cases, each channel is depicted individually and as a merged image. Scale bar represents 50 μm. Retinal layers are labelled as retinal pigment epithelium (RPE), outer nuclear layers (ONL), inner nuclear layer (INL) and ganglion layer (GL). Schematic representation of the retinal layers based on DAPI staining is placed on the right. Of note, the ivPMO is conjugated to an octa-guanidine dendrimer while the uPMO is not unconjugated.

Discussion

Assessing the effect(s) of antisense oligonucleotides (AONs) in the retina encounters challenges attributed to variability among diverse cell types and target genes (37). A nuanced understanding of target gene specificities is crucial for interpreting AON readouts effectively (38). To address this issue, our study deliberately explored various genes expressed in different retinal cell types. The ultimate objective was to investigate the intricate relationship between AON efficacy and delivery to diverse retinal cells by targeting specific genes. This approach aimed to mitigate potential influences on other retinal layers, ensuring a focused examination that facilitates a more precise interpretation of our comparisons. The therapeutic potential of AONs for retinal diseases led us to further investigate the different chemical modifications to identify the most promising chemical modification to primarily target the retina. In this study, we conducted a systematic comparison of three chemically-modified AONs currently used in clinical trials for several diseases, directly delivered to the retina by intravitreal administration. In general, our results from the in vitro studies indicated that 2′OMe and especially 2′MOE modifications presented a higher efficacy to modulate splicing compared to the PMO-modified AONs, in particular in photoreceptor cells.

Given the difficulty of assessing splicing modulation capacities in genes that are mainly or exclusively expressed in the retina, we used an artificial system based on splice reporter vectors (midigenes) in transfectable cells. However, due to the lack of full gene context, exon-skipping events resulting in an out-of-frame situation are not affected by nonsense-mediated decay in this system (12,39). Thus, to ensure having a proper readout for the in vivo analyses, we tried to limit ourselves to only in-frame exons when possible.

In vitro, 2′MOE oligonucleotides showed the highest efficacy in five out eight midigenes while PMO was the least efficacious. However, the neutral charge of the PMO AONs hampered the use of lipid-based agents for its delivery. As a consequence, we used a mechanical delivery method (26) in which five times more PMO AON was used, with respect to the other modifications, making difficult to establish a direct comparison. Still, the efficacy was lower than the 2′MOE and 2′OMe. Due to these differences, we cannot completely discard that lower efficacy is a consequence of the delivery method and poor uptake in cultured cells. However, the clear results obtained for Rdh5 and especially Grm6 (Figure 2) indicated that despite the delivery method, uPMO-AON capacity to modulate splicing is similar to the one reported by the other two modifications depending on the sequence.

Requirements for in vivo and in vitro delivery differ. For example, for in vitro studies, the use of transfection reagents is frequently employed. In contrast, in vivo administration requires some adjustments to assess specific targeting, prevent clearance in the body, and ensure enough cellular uptake by using some delivery vectors (40). Chemical modifications have been a subject of research to develop more efficacious and safer molecules. For instance, phosphorothioate (PS)-modified oligonucleotides are more stable and resistant to degradation, allowing for longer circulation in the bloodstream and better uptake by cells (41), limiting the risk of off-target effects and unwanted immune responses (42). In addition, the combination of a PS linkage with sugar ring modifications has shown a superior cellular uptake compared to their unmodified counterparts, making them more effective in delivering therapeutic interventions to target cells (43). For the eye, local administration of naked 2′OMe and 2′MOE with the corresponding PS nucleotide linkages has shown to be well-tolerated, discarding the need for a delivery vector (44–46). This is also accentuated by the properties of the eye, it is an isolated, contained, and immune-privileged tissue, making this organ the perfect model for therapeutics (47–50). Our in vivo results confirmed that the 2′MOE modification results in higher splicing modulation efficacies for the majority of the targets and undetectable toxic effects at the concentration used. The 2′OMe modification showed a similar safety profile, and a bit reduced splicing modulation efficacy for the majority of the targets, except for Rdh5 and Grm6. This is in line with previous studies in our group, in which we delivered 60 μg of AON (3 times more AON that what was used in this study), without any sign of toxicity (24).

As aforementioned, we focused on targeting exons divisible by three to have a proper readout for the in vivo analyses. We identified three AON sequences that were efficacious in vitro but did not cause any exon-skipping in vivo (second batch of animals). As expected, this was the case of the AON against Prkca, which was an AON targeting an out-of-frame exon. However, surprisingly, two of these cases (Pde6a and Best1) targeted an in-frame exon but we were not able to detect exon skipping. We did observe that the overall wild-type transcript levels were reduced in all three cases, despite no splicing event being detected. This could indicate that perhaps co-skipping was occurring (multiple exons skipped together) generating an out-of-frame transcript that is potentially degraded by nonsense-mediated decay. The capability of AONs to cause co-skipping was already observed in other cellular models for studying Duchenne muscular dystrophy (51) or dystrophic epidermolysis bullosa (52), supporting our hypothesis. This co-skipping was not previously detected in vitro, probably because these midigenes only included part of the gene of interest. To perform out-of-frame co-skipping, NMD would need to be inhibited. However, the NMD inhibitor CHX is only used for in vitro approaches as in vivo it can lead to DNA damage, teratogenesis, and reproductive fitness (53).

One limitation of our in vivo studies is the duration of one week, as we were targeting essential genes for visual function with detrimental long-term effects. This prevented the observation of direct metabolic effects resulting from oligonucleotide chemical modifications, which often manifest over time and may require a longer exposure for noticeable effects (54,55)

PMOs are resistant to a variety of different enzymes present in biological biofluids. This property makes PMO a highly suitable option for in vivo applications (56). Despite this, all commercially available PMOs for in vivo delivery are linked to a different moiety to promote cellular uptake of the oligonucleotide (57). The one employed in this study is linked to a dendrimer of arginine-rich peptides and it was the only commercially available modification (57). This moiety not only confers resistance to either proteases and nucleases but also reduces the degradation of the oligonucleotide and increases the uptake of the PMO-modified oligonucleotide into the cells by endocytosis (28). Both in vitro and in vivo the capacity to modulate splicing of the 2′MOE and 2′OMe modified AONs outperformed the PMO-modified AON-mediated one. However, the in vivo PMO showed severe unexpected toxicological effects in the retina, which were even detectable externally. During the first 24–72 h after the injection, an ocular phenotype appeared in most of the PMO-injected eyes (88.5%) independently of the oligonucleotide sequence, suggesting a PMO-mediated toxic effect. Subsequent observations showed that this toxicity was severely affecting the retina, as the tissue presented a whitish colour, reduced size compared to the contralateral eyes injected with PBS, and morphological abnormalities ranging from layer size reduction to complete degeneration of the retina. Furthermore, the external phenotype resembled the one caused by injecting gentamicin into the eye, where the injected eyes became pale, with degenerated neuroretina but well-preserved RPE and choroid (58). However, gentamicin primarily affected the inner retina, leaving the ONL relatively intact. In contrast, our study showed that the ONL is the most affected layer upon ivPMOs administration, indicating differences between gentamicin and in vivo PMOs' effects on the retina. Nevertheless, as the ocular phenotype did not appear at the same time in all animals, but rather in a range between 24 and 72 h after treatment, we aimed to relate the onset of the symptoms with the morphological studies (Table 2). Eyes injected with ivPMO-Gnat2 AON presented dilated pupils and white eyes either caused by a white fundus or cataracts already after 24 h. In contrast, ivPMO-Grm6 AON injected eyes showed the ocular phenotype later than other groups (most of the animal after 72 h after AON injection). These observations correlate with morphological studies. All these toxic effects were not observed when delivering an unmodified PMO sharing the same sequence. These results suggested that the observed toxicity was caused by the dendrimer linked to the employed ivPMO or due to the targeted gene itself. However, we did not find the same phenotypes for the other chemistries targeting the same genes, highlighting that the primary reason should be the oligonucleotide modification. One report has described cytotoxicity of in vivo PMOs (62). Briefly, Ferguson and colleagues showed that a pool of in vivo PMOs administrated by injection in the tail vein resulted in a high mortality rate. These studies suggested an increased cationic charge associated with the head groups of the dendrimer. As a result, blood clotting was elevated resulting in the death of the treated animals. Another potential explanation is that the ivPMOs may induce arteriolar attenuation, a phenomenon characterized by a reduction in the diameter of retinal arterioles (63,64). This narrowing of arterioles can impede the circulation of oxygenated blood, a factor associated with retinal degeneration (65). This could explain the observed smaller size and whiter colour in our ivPMO-treated retinas. Finally, we should consider that the observed toxicity might be influenced by the dose, even though the other two modifications did not show any harmful effect at the same dose in our study. Whether any of these causes is underlying our observations in the retina after in vivo PMO delivery is unknown and remains to be studied. Previously, a study comparing PMO and 2′MOE modifications delivered to the central nervous system was reported (66). Similar to our results, the 2′MOE AONs outperformed the uPMO in a severe model for spinal muscular atrophy (SMA). Indicating that probably higher doses are required for the PMO. Unfortunately, in this work, the in vivo PMO was not reported and therefore we cannot extrapolate or compare the observed toxic profiles.

Table 2.

Overview between the features of the ocular phenotype and time of appearance in the ivPMO-injected eyes

Oligonucleotide sequence	Severity based on morphological studies	Features of the ocular phenotype	Time point
Rho	Medium	6 mice presented dilation of the right pupil	24 h
		7 mice presented dilation of the right pupil	48 h
		9 mice presented dilation of the right pupil	72 h
		The 10 retinas for RNA analysis were whiter and smaller than the non-treated retinas of the same animals	1 week (end point)
Gnat2	High	6 mice presented white fundus in the right eye	24 h
		8 mice presented dilation of the pupil of the injected eye	48 h
		10 mice presented ocular phenotype: 7 mice showed only dilation of the right pupil, and 3 presented dilation and cataract	72 h
		3 mice presented cataracts. The 10 retinas for RNA analysis were whiter and smaller than the non-treated retinas of the same animals	1 week (end point)
Grm6	Low	2 animals with pupil dilation	24 h
		5 mice presented dilation of the right pupil	48 h
		9 mice presented dilation of the right pupil	72 h
		The 10 retinas for RNA analysis were whiter and smaller than the not treated retinas of the same animals	1 week (end point)
Rdh5	Medium	4 animals with pupil dilation	24 h
		7 mice presented dilation of the right pupil	48 h
		10 mice showed dilation of the right pupil	72 h
		3 mice presented cataracts. The 10 retinas for RNA analysis were whiter and smaller than the non-treated retinas of the same animals	1 week (end point)
Scrambled oligonucleotide (SON)	Medium	3 animals with pupil dilation	24 h
		7 mice presented dilation of the right pupil	48 h
		9 mice showed dilation of the right pupil	72 h
		1 mouse presented cataract. The 10 retinas for RNA analysis were whiter and smaller than the non-treated retinas of the same animals.	1 week (end point)

Severity grade was based on the morphological studies conducted after retina collection. Of note, most of the ocular phenotypes were accumulating with time, and in any case the symptoms observed were resolved. For example, mice with white fundus could show as well pupil dilatation or cataracts. h means hours.

According to the literature, the ivPMO modification is more stable and less cytotoxic than other PMOs conjugated to a cell-penetrating peptide (CPP) for in vivo delivery (59–61). The use of conjugations might be needed as the unmodified PMO showed decreased uptake and consequently, low efficacy. Several publications have shown that CPP-PMOs are not toxic (67–69) and improve the pharmacokinetic profile, tissue uptake, and longevity in vivo (70,71). A clinical trial based on a CPP-conjugated PMO was recently approved to be dosed in retinitis pigmentosa patients carrying variants in the PRPF3 gene (VP-001 drug; NCT05902962). To date, there is no publication associated with this drug or the associated CPP, but the outcome of this trial could clarify some safety-related aspects of PMO-based therapies in retinal diseases.

To assess the uptake and biodistribution microRNAscope was performed. As highlighted in the methods section, we selected Grm6(expressed in bipolar cells) due to the characteristics of the AON sequence and compatibility with the assay. The advantage of intravitreal injections is that effectively deliver molecules to all retinal layers, ensuring the oligonucleotides reach the entire retina (72,73). This microRNAscope analysis confirmed that the uPMO-AON was barely present in any retinal layer one week post-injection, while for the other three chemically-modified AONs (2′OME, 2′MOE, and in vivo PMO), the uptake was detected. This could be related to either uptake or molecule stability defects caused by the absence of the octa-guanidine dendrimer. However, we cannot completely discard the degradation of the uPMO. Nonetheless, PMO modification has been bolstered for in vivo applications due to its high resistance to proteases, nucleases, esterases as well as a wide range of enzymes (including hydrolases) present in the body (56). For the other two chemical modifications, the microRNAscope results suggested that the AON reached all layers and the distribution was similar. Based on intensity, it seems that the 2′MOE modification might have a higher uptake in the ONL, explaining why this chemistry seems to be more efficacious for photoreceptor-specific genes. Our findings suggest that the 2′MOE chemical modification is the most efficacious for the retina, aligning with the data obtained in a comparison of 2′MOE and PMO oligonucleotides in the central nervous system. Sheng et al. indicated that MOE modification is more therapeutically effective in a severe SMA mouse model, demonstrating stronger phenotypic improvement (66).

In conclusion, our results showed that the 2′MOE chemistry seems to be the most promising chemistry in terms of efficacy, distribution, and safety for photoreceptor cells, and overall both 2′MOE and 2′OMe have a similar profile, although 2′MOE showed a trend towards higher efficacy for the majority of the targets. In contrast, PMO showed disparate results. The modified PMO for in vivo application resulted in high toxicity at the tested conditions, while the unmodified PMO did not provoke any toxic effect, but uptake was compromised. Altogether, we provided a systematic comparison of three different chemical modifications of AONs in the mouse retina. These results can provide valuable information for future oligonucleotide-based therapeutic strategies for retinal diseases.

Notes

Present address: Tamara Hoppenbrouwers, Food Quality and Design, Wageningen University and Biobased Research, Wageningen, The Netherlands.

Data availability

The data underlying this article are available in the article and its online supplementary material. The midigenes are available upon request to the corresponding author.

Supplementary data

Supplementary Data are available at NAR Online.

Funding

Foundation Fighting Blindness USA Project Program Award [PPA-0517–0717-RAD to R.W.J.C., A.G.]; the funding organizations provided unrestricted grants and had no role in the design or conduct of this research. Funding for open access charge: Department and the PI means of the corresponding author.

Conflict of interest statement. The authors would like to declare the following indirect conflict of interests: R.W.J.C. and A.G. are inventors on several filed patents for antisense oligonucleotides (WO2013036105A1, WO2018109011A1, WO2020015959A1, WO2020115106A1, WO2021023863A1), none of these molecules has been studied in this publication. R.W.J.C also declares to be the Scientific Officer of Astherna B.V., I.V.D. is currently partially employed by Astherna B.V., but all the work presented in this manuscript was conducted before her appointment. The rest of the authors declare no conflict of interest.