Assessment of toxicity from AAVs expressing GFP exclusively in RPE or photoreceptors.

A) AAV vector maps for Rho::GFP and Best1::GFP. B) RPE morphology upon staining RPE flat mounts with phalloidin, which labels the F-actin enriched RPE cell junctions. Flat mounts were harvested 2 weeks post subretinal injection of 2e9 gc of either Rho::GFP or Best1::GFP into P0 pups. RPE score of Best1::GFP is 3. Scale bar: 100 µm. C) Semi-quantitative scoring of RPE cell toxicity from samples described in panel B (n=3-4 eyes per group, *p=0.03, mean ± SD, unpaired t-test). Uninjected P30 C57BL/6J RPE flatmounts were used as a control. D) Retinal morphology upon staining whole eyecup sections with DAPI, which labels nuclei. Eyecups were harvested 4 weeks post subretinal injection of 2e9 gc of Rho::GFP or Best1::GFP into P0 pups. Scale bar: 50 µm. E) Median number of ONL layers remaining for samples shown in panel D (n≥4 eyes per group, ***p=0.0003, mean ± SD, unpaired t-test). Uninjected 4 week C57BL/6J whole eyecup sections were used as a control. F) Representative whole eyecup sections from mice injected with 2e9 gc Rho::GFP or Best1::GFP at various times post infection ranging from 8-24 weeks. Green: GFP. Blue: DAPI-labeled nuclei. Scale bar: 50 µm. G) ONL thickness quantification over time after subretinal injection of Rho::GFP vs. Best1::GFP (n=3-6 eyes per group, *p=0.0001 for week 8, *p=0.0008 for week 12, *p<0.0001 for week 24, mean ± SD, multiple unpaired t-tests with Bonferroni-Dunn multiple comparisons correction).

Bulk RNA sequencing of RPE cells to reveal candidate mechanisms of ocular toxicity from Best1::GFP expression.

P0 pups were subretinally injected with 4e8 gc of Rho::GFP or Best1::GFP in contralateral eyes. Enriched RPE cells were collected at 1 week or 2 weeks post-transduction for bulk RNA-sequencing (n=4-5 eyes per group). (A) Tree plots of the top 25 Gene Ontology Biological Process terms associated with upregulated DE genes at week 1. Pathways are upregulated in the Best1::GFP relative to the Rho::GFP infection context. (B) Tree plots of the top 25 Gene Ontology Biological Process terms associated with upregulated DE genes at week 2. Pathways are upregulated in the Best1::GFP relative to the Rho::GFP infection context. (C) Read counts (normalized using DESeq2’s median of ratios method81) for selected stress and immune-related genes at 1 week post injection of 4e8 gc Best1::GFP or 4e8 gc Rho::GFP. Read counts for the EGFP transgene, the Best1 endogenous gene, some housekeeping genes (GAPDH, ACTB), and some visual cycle genes (RPE65, RGR, RDH10) are plotted for reference. The dotted lines represent read counts of 500 and 1000 and are included to help visualize expression levels on the plot (n=5 eyes per group, p<0.0001 for Rgr, Rpe65, and Rdh10, *p<0.001 for EGFP, *p<0.01 for Ifnar1, Chop, Mx1, Atf4, Rsad2, Ticam1, Rigi, and Best1, *p<0.05 for Ifit3, mean ± SD, multiple unpaired t-tests).

Knockout (KO) mouse injection screen to determine the contributions of intrinsic, innate, and adaptive immune pathways on AAV-associated toxicity.

For panel B, two strain backgrounds of the TNFR1 KO mouse (Sv129 or C57BL/6, e.g., B6) were used, as indicated by the labels. For panels D and E, only the C57BL/6 TNFR1 KO strain was used. The TLR dKO mouse strain refers to Myd88/Ticam1 double KO mice. A) Schematic of the immune KO mouse screening experiment to alleviate toxicity. Neonatal mouse pups of the indicated genotype were injected with 4e8 gc Best1::GFP and eyes were harvested for RPE flatmounts, retinal flatmounts, or whole eyecup sections 4-5 weeks post injection. Figure created with BioRender. B) Median number of ONL layers in midperipheral regions of whole eyecup sections from Best1::GFP injected eyes of the indicated genotype (n=4-17 eyes per group, ****p<0.0001, **p=0.005, mean ± SD, one-way ANOVA with Dunnett’s multiple comparison’s correction). Uninjected control samples are the same as shown in Figure 1. C) Representative whole eyecup sections for C57BL/6J mice or selected immune KO strains from the plot shown in panel B. Blue: DAPI. Green: Best1::GFP expression. Scale bar: 100 µm. D) Cone quantification from retinal flatmounts (n≥4 eyes per group, **p=0.003, ***p=0.0001, mean ± SD, one-way ANOVA with Dunnett’s multiple comparison’s correction). All mice were 4-5 weeks of age at harvest. E) Semi-quantitative scoring of RPE cell toxicity (n≥3 eyes per group, mean ± SD, ****p<0.0001, ***p=0.001, one-way ANOVA with Dunnett’s multiple comparison’s correction). Uninjected control samples are the same as shown in Figure 1.

© 2026, BioRender Inc. Any parts of this image created with BioRender are not made available under the same license as the Reviewed Preprint, and are © 2026, BioRender Inc.

Analysis of expression of interferon beta (IFNB), the type I interferon receptor (IFNAR), and interferon stimulated genes (ISGs) following AAV infection.

A) Selected differentially expressed ISGs (45 total). Week 2 read counts (normalized using DESeq2’s median of ratios method81) from Rho::GFP (gray) and Best1::GFP (green) injected RPE RNA-sequencing data are plotted (n=5 eyes per group, *p<0.001 for Asns, Hmox1, Rxrg, Tuba8, Lgals9, and Casp4, *p<0.01 for Plau, Fcer1g, Csf1, Icam1, St8sia4, Fcrl2, Mlkl, Ifit1, Nupr1, Acss1, Usp18, Mx2, Mpeg1, Gdf15, Ifit2, Mx1, Lcn2, and Rsad2, *p<0.05 for Nlrc5, Tnfsf10, Ifit3, C1qa, Iigp1, Ifi44, Casp1, Agr2, Ccl6, Atf5, Cxcl10, Cmpk2, C1qc, Ccl2, Tyrobp, Apobec3, Casp12, and Ly86, mean ± SD, multiple unpaired t-tests). B) HCR-FISH for IFNAR transcripts was performed on retinal flatmounts from B6J mice injected at birth with 4e8 gc Rho::GFP (top) or 4e8 gc Best1::GFP (bottom) and harvested at P14. Scale bar: 50 µm. C) HCR-FISH for IFNAR transcripts was performed on RPE flatmounts from B6J mice injected at birth with 4e8 gc Rho::GFP (top) or 4e8 gc Best1::GFP (bottom) and harvested at P14. Scale bar: 50 µm. D) HCR-FISH for IFNB transcripts was performed on retinal flatmounts from B6J mice that were uninjected at birth (i) or injected with 4e8 gc Rho::GFP (ii) or 4e8 gc Best1::GFP (iii) and harvested at P14. Scale bar: 50 µm. E) HCR-FISH for IFNB transcripts was performed on RPE flatmounts from B6J mice that were uninjected at birth (i) or injected with 4e8 gc Rho::GFP (ii) or 4e8 gc Best1::GFP (iii) and harvested at P14. Scale bar: 50 µm.

Effect of transgenes on AAV-associated ocular toxicity.

All mice in this figure were harvested at 4-5 weeks post injection. B6J controls (either injected with 4e8 gc Best1::GFP or uninjected) were previously plotted in Figure 3. Data for the “tracer” dose of Best1::GFP (2e7 gc) is replotted here from Figure S2. For reference, the read counts (normalized using DESeq2’s median of ratios method81) for the EGFP transgene is ∼10^6, whereas the counts for the endogenous transcripts CTSD, B2M, and RPE65 are ∼10^5, ∼10^4, and ∼10^3, respectively. A) Quantification of the number of ONL layers remaining at 4 weeks for a panel of Best1::transgene expressing AAVs, subretinally injected at birth and harvested 4-5 weeks post injection. Each Best1::transgene expressing AAV was administered at 4e8 gc. The Best1::GFP tracer was co-injected at 2e7 gc to track nonfluorescent transgenes. Uninjected, age-matched C57BL/6J eyes were plotted for comparison (n=4-17 eyes per group, mean ± SD, ****p<0.0001, ***p=0.0003, one-way ANOVA with Bonferroni’s multiple comparisons correction).

Model for how induced ocular toxicity is generated.

A) Following subretinal injection of AAV, which infects both RPE and photoreceptor cells, cell stress responses in the RPE arise in response to expression of specific transgenes and/or their expression levels. The stress response in the RPE kills them and leads to secondary photoreceptor death. Maladaptive type I interferon responses, possibly triggered by cGAS sensing of cytosolic double-stranded DNA, contribute to retinal toxicity. Infected RPE and photoreceptor cells respond to infection by expressing and secreting type I interferon molecules, e.g., IFNb. Type I IFN molecules can activate the expression of interferon stimulated genes (ISGs) in cells that express the Type I IFN receptor (IFNAR), including RPE, photoreceptors, and possibly myeloid cells. Myeloid cells include microglia and peripheral infiltrating macrophages. Myeloid cells are recruited to the ONL and RPE layers in response to toxic AAV injection and may participate in interferon signaling. Myeloid cells also undergo pyroptosis as visualized by ASC speck formation, which is indicative of inflammasome assembly. However, pyroptosis itself does not contribute to the observed toxicity.

© 2026, BioRender Inc. Any parts of this image created with BioRender are not made available under the same license as the Reviewed Preprint, and are © 2026, BioRender Inc.

Representative wild type C57BL/6J mouse RPE flatmount and eyecup sections.

A) Top: Representative RPE flatmount from a wild type C57BL/6J mouse at P30. RPE flatmount is stained with phalloidin (red). Scale bar: 100 µm. Bottom: RPE semiquantitative scoring method. B) Representative eyecup section from a wild type C57BL/6J mouse at P28. Section is stained with DAPI (blue). Scale bar: 50 µm. To count the number of ONL layers, 3 lines were drawn spanning the ONL and the number of nuclei layers was counted. In this example, the left orange line spans ∼12 layers, the middle line spans ∼13 layers, and the right line spans ∼12 layers. The median of these values was reported for the sample, which in this case is ∼12 layers.

Dose response characterization of Best1::GFP.

All mice in this figure were harvested at 4 weeks of age. The uninjected C57BL/6J mouse control image is the same as the one shown in Figure S1B. A) Top row: Representative eyecup section from an uninjected C57BL/6J mouse (left) or from C57BL/6J mice injected with 2e7 gc Best1::GFP (middle) or 2e8 gc Best1::GFP (right). Bottom row: Representative eyecup sections from C57BL/6J mice injected with 4e8 gc Best1::GFP (left) or 2e9 gc Best1::GFP (right). Sections are stained with DAPI (blue) and the GFP channel is overlaid. Scale bar: 50 µm. B) Median number of ONL layers remaining from samples injected with the indicated dose of Best1::GFP. Uninjected eyes serve as age-matched controls (n=4-17 eyes per group, ****p<0.0001, mean ± SD, one-way ANOVA with Bonferroni multiple comparisons correction). Uninjected control samples are the same as shown in Figure 1.

Principal components analysis on bulk RNA-sequencing data from AAV-infected RPE.

This dataset is the same as the one described in Figure 2. P0 pups were subretinally injected with 4e8 gc of either Rho::EGFP or Best1::EGFP. Enriched RPE cells were collected at 1 week and 2 weeks post-transduction for bulk RNA-seq (n=5). A) PC1-2 (36% and 14% of variance) after PCA of normalized counts for all genes. B) PC2-3 (14% and 9% of variance) after PCA of normalized counts for all genes. C) The cumulative amount of variance captured by the addition of each principal component from PCA.

Differential gene expression analysis for RPE samples injected with Rho::EGFP vs. Best1::EGFP.

Differential expression analysis was performed by DESeq2. This dataset is the same as the one described in Figure 2. P0 pups were subretinally injected with 4e8 gc of either Rho::EGFP or Best1::EGFP. Enriched RPE cells were collected at 1 week and 2 weeks post-transduction for bulk RNA-seq (n=5). A) Heatmap of differentially expressed genes. For this heatmap, the thresholds for differential expression were an adjusted p-value < 0.05 and a log2 fold-change > 1.

GO terms for downregulated differentially expressed genes from bulk RPE RNA-sequencing dataset.

This dataset is the same as the one described in Figure 2. P0 pups were subretinally injected with 4e8 gc of either Rho::EGFP or Best1::EGFP. Enriched RPE cells were collected at 1 week and 2 weeks post-transduction for bulk RNA-seq (n=5). A) Tree plots of the top 25 Gene Ontology Biological Process terms associated with downregulated DE genes at week 1 (top) and week 2 (bottom). Pathways are downregulated in the Best1::GFP relative to the Rho::GFP infection context.

Survey of immune related gene abundances in unstimulated microglia.

Data from panel A was generated in this study. Data from panel B was previously published and represents unstimulated microglial immune-related counts per million reads (CPMs).46 (A) CPM for selected immune-related genes at 2 weeks of age. Mice were injected with 4e8 gc Best1::GFP or uninjected. Genes were selected on the basis of being an AAV transgene (EGFP), a housekeeping or visual cycle gene, an immune related receptor/sensor, a cell stress-related gene, or an interferon stimulated gene. B) Plot of immune-related transcript CPMs. Cx3cr1 and Tmem119 are included as reference microglial marker genes.

Representative RPE flatmounts from the KO mouse screen.

All mice were harvested 4-5 weeks post infection. Each image represents an individual eye. Scale bar: 20 µm. A) RPE flatmounts from uninjected C57BL/6J mice (n=3 shown). 16 regions of 4 RPE flatmounts from 2 animals were scored, with the average score = 0. B) RPE flatmounts from C57BL/6J mice injected with 4e8 gc Best1::GFP (n=9 shown). 38 regions of 16 RPE flatmounts from 10 animals were scored, with the average score = 2.2. C) RPE flatmounts from IFNAR1 KO mice injected with 4e8 gc Best1::GFP (n=9 shown). 22 regions of 9 RPE flatmounts from 8 animals were scored, with the average score = 1.9. D) RPE flatmounts from IFNGR1 KO mice injected with 4e8 gc Best1::GFP (n=4 shown). 15 regions of 4 RPE flatmounts from 3 animals were scored, with the average score = 2.8. E) RPE flatmounts from CGAS KO mice injected with 4e8 gc Best1::GFP (n=9 shown). 30 regions of 9 RPE flatmounts from 7 animals were scored, with the average score = 1.7. F) RPE flatmounts from MAVS KO mice injected with 4e8 gc Best1::GFP (n=4 shown). 9 regions of 4 RPE flatmounts from 2 animals were scored, with the average score = 2.3. G) RPE flatmounts from RAG1 KO mice injected with 4e8 gc Best1::GFP (n=8 shown). 25 regions of 8 RPE flatmounts from 5 animals were scored, with the average score = 2.4. H) RPE flatmounts from IL2RG KO mice injected with 4e8 gc Best1::GFP (n=12 shown). 38 regions of 12 RPE flatmounts from 8 animals were scored, with the average score = 2.2. I) RPE flatmounts from TNFR1 KO mice injected with 4e8 gc Best1::GFP (n=10 shown). 30 regions of 10 RPE flatmounts from 6 animals were scored, with the average score = 2. J) RPE flatmounts from C3 KO mice injected with 4e8 gc Best1::GFP (n=3 shown). 12 regions of 3 RPE flatmounts from 3 animals were scored, with the average score = 2.6. K) RPE flatmounts from CHOP KO mice injected with 4e8 gc Best1::GFP (n=3 shown). 6 regions of 3 RPE flatmounts from 3 animals were scored, with the average score = 1.1. L) RPE flatmounts from TLR double KO mice injected with 4e8 gc Best1::GFP (n=5 shown). 19 regions of 5 RPE flatmounts from 5 animals were scored, with the average score = 2.4.

Assessment of IFNAR KO strain on CMV::GFP toxicity phenotype.

All mice in this figure were harvested at 4-5 weeks of age. A) RPE and retinal flatmounts were harvested from B6J or IFNAR KO mice injected at birth with 4e8 gc CMV::GFP. The RPE toxicity scores (n=4-6 eyes per group, mean ± SD, unpaired t-test) and median cone counts (n=4-8 eyes per group, *p=0.02, mean ± SD, unpaired t-test) are plotted. Uninjected control samples are the same as shown in Figure 3.

Baseline RPE, cone, and ONL layer quantifications for the mouse strains used in this paper.

All mice were 1-2 months of age at harvest. A) Quantification of cone cell counts (n≥3 eyes per group, mean ± SD, *p=0.03, one-way ANOVA with Dunnett’s multiple comparisons correction). All samples quantified in this plot are uninjected. All mice were 1-2 months of age at harvest. B6J uninjected control samples are the same as shown in Figure 3. B) Median number of ONL layers (n≥3 eyes per group, *p=0.01, mean ± SD, one-way ANOVA with Dunnett’s multiple comparisons correction). Unless labeled as (uninj) for uninjected samples on the plot, all samples were injected with 4e8 gc Rho::GFP. B6J uninjected control samples are the same as shown in Figure 3.

Pharmacologic depletion of microglia to assess the role of microglia in AAV-associated ocular toxicity.

For panels B-D, the Best1::GFP AAV dose was 4e8 gc. For panels E-G, the Best1::GFP AAV dose was 4e9 gc. A) Experimental workflow for depleting microglia in pups. Neonatal mice were subretinally injected with Best1::GFP at a dose of 4e8 or 4e9 gc. At 1 week of age, the lactating mother was given chow with 1200 ppm PLX5622 (or fed normal chow as a control) incorporated to initiate microglial depletion. The dam and her pups remained on the chow until P28, when the pups were harvested. B) RPE toxicity scoring (n≥4 eyes per group, mean ± SD, **p=0.007, unpaired t-test). Best1-GFP (4e8 gc) and uninjected control samples are the same as shown in Figure 3. C) Quantification of cone counts (n≥4 eyes per group, mean ± SD, unpaired t-test). Best1-GFP (4e8 gc) and uninjected control samples are the same as shown in Figure 3. D) Median number of ONL layers from sectioning retinal flatmounts (n≥3 eyes per group, mean ± SD, unpaired t-test). Uninjected control samples are the same as shown in Figure 3. E) RPE toxicity scoring (n≥4 eyes per group, mean ± SD, **p=0.007, unpaired t-test). Uninjected control samples are the same as shown in Figure 3. F) Quantification of cone counts (n≥4 eyes per group, *p=0.02, mean ± SD, unpaired t-test). Uninjected control samples are the same as shown in Figure 3. G) Median number of ONL layers from sectioning retinal flatmounts (n≥4 eyes per group, mean ± SD, unpaired t-test). Uninjected control samples are the same as shown in Figure 3.

Quantification of pharmacologic depletion of microglia (AAV Best1::GFP dose: 4e8 gc).

Injections were performed in mice at birth. A) P28 representative RPE flatmounts stained with IbaI (magenta) to label microglia. These mice were fed regular chow throughout the experiment. Scale bar: 100 µm. B) P28 representative RPE flatmounts stained with IbaI (magenta) to label microglia. These mice were fed PLX5622-containing chow from P7-P28. Scale bar: 100 µm. C) Analysis of IbaI+ depletion efficiency. The number of IbaI+ cells were counted in 3-4 boxes (500 µm2) around each RPE flatmount and the median count was plotted (n≥14 eyes per group, mean ± SD, ***p=0.0002, unpaired t-test). Efficiency was ∼91% for this experiment.

Quantification of pharmacologic depletion of microglia (AAV Best1::GFP dose: 4e9 gc).

Injections were performed in mice at birth. A) P28 representative RPE flatmounts stained with IbaI (magenta) to label microglia. These mice were fed regular chow throughout the experiment. Scale bar: 100 µm. B) P28 representative RPE flatmounts stained with IbaI (magenta) to label microglia. These mice were fed PLX5622-containing chow from P7-P28. Scale bar: 100 µm. C) Analysis of IbaI+ depletion efficiency. The number of IbaI+ cells were counted in 3-4 boxes (500 µm2) around each RPE flatmount and the median count was plotted (n≥10 eyes per group, mean ± SD, ****p<0.0001, unpaired t-test). Efficiency was ∼98% for this experiment.

Quantification of pharmacologic depletion of microglia in uninjected animals.

For all images in this figure, microglia are labeled in green. A) Representative retinal flatmounts from P19 Cx3cr1-GFP heterozygous reporter mice. The lactating dam and pups were given access to PLX5622-containing chow from P8-P19. B) Representative retinal flatmounts from P19 Cx3cr1-GFP heterozygous reporter mice. The lactating dam and pups were given access to regular chow from P8-P19. C) Representative retinal flatmounts from P38 Cx3cr1-GFP heterozygous reporter mice. The mice were given access to PLX5622-containing chow from P31-P38. D) Representative retinal flatmounts from P38 Cx3cr1-GFP heterozygous reporter mice. The mice were fed regular chow throughout their life. E) Quantification of the images presented in panels A-D (n=4 eyes per group, *p<0.0001, mean ± SD, multiple unpaired t-tests with Bonferroni-Dunn multiple comparisons correction).

Assessment of myeloid cell pyroptosis on AAV-associated ocular toxicity.

For all bar plots in this figure, the B6J controls (either injected with 4e8 gc Best1::GFP or uninjected) were previously plotted in Figure 3. A) C57BL/6J mice were injected with 4e8 gc Best1::6xSTOP-mutGFP (left), 4e8 gc Best1::GFP (middle), or 4e9 gc Best1::GFP (right) and harvested at 4-5 weeks of age. Retinal flatmounts were stained for ASC (magenta). Scale bar: 100 µm. B) Cx3cr1-GFP heterozygous reporter mice were injected with 4e8 gc CMV::null and harvested at P19. Retinal flatmounts were stained for ASC. Left image is the ASC channel (magenta), middle image is the Cx3cr1-GFP channel (green), and the right image shows both channels merged (colocalization of signals is depicted in white). Scale bar: 100 µm. C) GSDMD KO mice or C57BL/6J controls were injected at birth with 4e8 gc Best1::GFP and harvested 4 weeks post injection. RPE flatmounts (n=16 B6J eyes, n=8 GSDMD KO eyes, mean ± SD, ****p<0.0001), retinal flatmounts (n=22 B6J eyes, n=8 GSDMD KO eyes, mean ± SD, unpaired t-test), and the number of ONL layers (n=17 B6J eyes, n=7 GSDMD KO eyes, mean ± SD, unpaired t-test) were quantified. D) Uninjected GSDMD KO mice or C57BL/6J controls were harvested 4 weeks post injection. RPE flatmounts (n=4 B6J eyes, n=5 GSDMD KO eyes, mean ± SD, unpaired t-test), retinal flatmounts (n=4 B6J eyes, n=7 GSDMD KO eyes, mean ± SD, unpaired t-test), and the number of ONL layers (n=4 B6J eyes, n=3 GSDMD KO eyes, mean ± SD, unpaired t-test) were quantified.

Assessment of chemokine receptors on AAV-associated ocular toxicity.

A) CX3CR1/CCR2 double heterozygous, KO, or WT mice were injected at birth with 4e8 gc Best1::GFP and harvested 4 weeks post injection. RPE flatmounts (n=4-6 eyes per group, mean ± SD, one-way ANOVA with Dunnett’s multiple comparisons test), retinal flatmounts (n=3-5 eyes per group, mean ± SD, one-way ANOVA with Dunnett’s multiple comparisons correction), and the number of ONL layers (n=5-6 eyes per group, mean ± SD, unpaired t-test) were quantified.

Effect of Best1::GFP expression in mice subretinally injected as adults.

C57BL/6J mice (∼12-14 weeks old) were injected with 2e10 gc Best1::GFP and harvested 4 weeks later. The neonatal B6J (4e8 gc and 4e9 gc Best1::GFP) RPE and cone counts are replotted here from Figures 3 and 5, respectively. The uninjected B6J (P30) data are replotted here from Figure 1. The uninjected B6J (11 weeks) data was previously published by our lab.43 Only transduced areas of flatmounts were quantified for adult injected RPE samples. Because transduced areas of retinal flatmounts (as assessed by GFP+ signal) could not be determined for retinal flatmounts from mice injected with Best1::GFP as adults, the adult B6J counts plotted represent quantifications made from boxes drawn roughly equidistantly around each flatmount. A) Representative images of transduced (left) and untransduced (right) areas of RPE flatmounts from mice subretinally injected with Best1::GFP as adults. B) RPE toxicity scores from mice subretinally injected with Best1::GFP as adults. Neonatal B6J injections are replotted here for comparison (n=4-16 eyes per group, mean ± SD, ****p<0.0001, **p=0.007, one-way ANOVA with Bonferroni’s multiple comparisons correction). Only transduced areas of RPE flatmounts (as assessed by GFP+ signal) were quantified in adult RPE flatmounts. C) Cone counts for mice subretinally injected with Best1::GFP as adults (n=4-22 eyes per group, mean ± SD, ****p<0.0001, one-way ANOVA with Bonferroni’s multiple comparisons correction).

Effect of CMV::GFP expression in mice subretinally injected as adults.

C57BL/6J mice (16-20 weeks old) were injected with 2e9 gc CMV::GFP and harvested 4 weeks later. The neonatal B6J and IFNAR KO (4e8 gc CMV::GFP) RPE and cone counts are replotted here from Figure S8. The uninjected B6J (P30) data are replotted here from Figure 1. The uninjected B6J (11 weeks) data was previously published by our lab.43 Only transduced areas of flatmounts were quantified for adult injected samples. A) RPE scores from RPE flatmounts of mice subretinally injected with CMV::GFP as adults (n=4-6 eyes per group, mean ± SD, *p=0.0158 for adult B6J vs. neonatal B6J, **p=0.005 for adult B6J vs neonatal IFNAR KO, **p=0.001 for adult B6J vs uninjected B6J, *p=0.02, one-way ANOVA with Bonferroni’s multiple comparisons correction). B) Cone counts for mice subretinally injected with CMV::GFP as adults (n=4-8 eyes per group, mean ± SD, ****p<0.0001, one-way ANOVA with Bonferroni’s multiple comparisons correction).

Comparison of injected Ai75d heterozygous animals and C57BL/6J animals used interchangeably for experiments.

The B6J RPE and cone count controls (injected with 4e8 gc Best1::GFP) are replotted here from Figure 3. A) RPE toxicity scores from RPE flatmounts of C57BL/6J or Ai75d heterozygous pups injected at birth with 4e8 gc Best1::GFP and harvested at 4-5 weeks of age (n=6-16 eyes per group, mean ± SD, unpaired t-test). B) Cone counts for mice C57BL/6J or Ai75d heterozygous pups injected at birth with 4e8 gc Best1::GFP and harvested at 4-5 weeks of age (n=6-22 eyes per group, mean ± SD, unpaired t-test).