Gene Therapy of Dominant CRX-Leber Congenital Amaurosis using Patient Stem Cell-Derived Retinal Organoids. Stem Cell Reports, 16(2), 252-263.

Mutations in the photoreceptor transcription factor gene cone-rod homeobox (CRX) lead to distinct retinopathy phenotypes, including early-onset vision impairment in dominant Leber congenital amaurosis (LCA). Using induced pluripotent stem cells (iPSCs) from a pa-tientwith CRX -I138fs48mutation,weestablishedan in vitro modelof CRX -LCAinretinalorganoidsthatshoweddefectivephotoreceptor maturation by histology and gene proﬁling, with diminished expression of visual opsins. Adeno-associated virus (AAV)-mediated CRX gene augmentation therapy partially restored photoreceptor phenotype and expression of phototransduction-related genes as determined by single-cell RNA-sequencing. Retinal organoids derived from iPSCs of a second dominant CRX -LCA patient carrying K88N mutation revealed the loss of opsin expression as a common phenotype, which was alleviated by AAV-mediated augmentation of CRX. Our studies provide a proof-of-concept for developing gene therapy of dominant CRX -LCA and other CRX retinopathies.


INTRODUCTION
Leber congenital amaurosis (LCA) constitutes a group of rare early-onset retinal dystrophies with severe clinical manifestations, resulting in vision loss during infancy (den Hollander et al., 2008). LCA is genetically heterogeneous with mutations identified in at least 25 genes; a vast majority of these are crucial for photoreceptor development and/or function (Kumaran et al., 2017). Most patients with LCA exhibit autosomal recessive inheritance. Mouse models have been valuable for elucidating disease etiology (Veleri et al., 2015) and for preclinical therapy development, leading to the first U.S. Food and Drug Administration-approved adeno-associated virus (AAV)based gene therapy of recessive LCA caused by RPE65 mutations (Apte, 2018). However, it is widely recognized that available animal models do not fully capture complexities of human disease, slowing further clinical translation. No treatment currently exists for other, especially dominant, forms of LCA.
Cone-rod homeobox protein CRX is essential for development of photoreceptors in the retina (Furukawa et al., 1999). CRX controls expression of most rod and cone photoreceptor genes through its interaction with bZIP transcription factor NRL (Mitton et al., 2000) and/or an intricate network of regulatory proteins (Hennig et al., 2008). Heterozygous mutations in CRX cause early-onset retinal dystrophies with extensive phenotypic heterogeneity, with rare reports of biallelic variants in LCA (Huang et al., 2012;Hull et al., 2014;Ibrahim et al., 2018;Rivolta et al., 2001;Swaroop et al., 1999). Notably, a majority of reported dominant LCA can be attributed to CRX. Disease-causing CRX mutations can either reduce the transcriptional activity or impart a gain of function effect (Tran et al., 2014). Studies in animal models revealed a severe phenotype associated with dominant frameshift mutations in Crx compared with loss-of-function alleles (Roger et al., 2014;Tran et al., 2014).
We previously described clinical presentations of two pediatric patients carrying c.G264T (p.K88N)-or c.413delT (p.I138fs48) dominant CRX mutations manifested as LCA, with a complete loss of light-evoked responses in electroretinogram recordings (Nichols et al., 2010). Importantly, retinal imaging revealed preservation of outer nuclear layer, suggesting presence of viable but nonfunctional photoreceptors, which could be potential targets for gene therapy. Molecular analysis of mutant alleles demonstrated differential interaction with NRL in vitro, suggesting interference with gene regulation in rod photoreceptors (Nichols et al., 2010). Although animal models phenocopy human CRX retinopathies (Roger et al., 2014;Tran et al., 2014) and can provide proof-of-concept for treatment (Watanabe et al., 2013), there are significant differences in development, cell-type composition, and molecular profiles of human retina (Hoshino et al., 2017). Therefore, we decided to take advantage of patientderived induced pluripotent stem cells (iPSCs) to create a human model for developing therapeutic paradigms for CRX-LCA.
Directed differentiation of iPSCs into three-dimensional (3D) retinal organoids has enabled the modeling of retinopathies in patient-specific genetic background (Kruczek and Swaroop, 2020). Next generation sequencing technologies have provided a detailed comparison of developing retinal organoids with human retina (Collin et al., 2019;Cowan et al., 2020;Hoshino et al., 2017;Kaya et al., 2019;Kim et al., 2019) and permit evaluation of major retinal cell types in a disease context. In this study, we established a retinal organoid model of CRX-LCA from patient-derived iPSCs and demonstrate perturbation in molecular phenotype of photoreceptors, including diminished expression of visual opsins, in concordance with clinically observed loss of light responses. We also show partial restoration of both rod and cone gene expression by delivering correct CRX transgene driven by a human CRX promoter via an AAV vector. Our studies provide a path forward for treatment of dominant CRX-LCA by gene augmentation.

RESULTS
Skin biopsies of two LCA patients carrying dominant c.G264T (p.K88N) or c.413delT (p.I138fs48) mutation in the CRX gene ( Figure 1A), as well as unaffected familial controls (Table S1), were used to derive iPSC lines, which demonstrated normal karyotype and typical features of stem cells ( Figure S1A, S1B, S5A, and S5B; Table S2). Retinal organoids were differentiated using a previously published protocol (Kaya et al., 2019;Zhong et al., 2014) (Figure 1B). Pairwise comparisons of organoids derived from each patient to respective healthy familial control were performed in all presented data. Both patient and control cell lines formed morphologically similar retinal neural epithelia ( Figures S1C and S5C).
Given that a frameshift mutation showed a particularly severe retinal phenotype in Crx Rip mouse model (Roger et al., 2014), we first examined the organoids derived from c.413delT(p.I138fs48) iPSCs. Retinal organoids exhibit three stereotypical stages of differentiation (Capowski et al., 2019;Kaya et al., 2019) corresponding to developmental epochs in human fetal retina (Hoshino et al., 2017). At stage 1, neuroepithelia of control and patient organoids were indistinguishable (Figures S1C-S1E). During stage 2, CRX-expressing photoreceptors aligned at the apical side of neural retina in both control and patient organoids ( Figure 1C) and stained positive for early photoreceptor marker, Recoverin ( Figure 1D). Immunoblotting of day 90 organoid protein extracts revealed a truncated CRX protein (referred as I138fs) in patient samples ( Figure 1E). Quantification of the protein bands indicated higher levels of total CRX in patient samples with a significant proportion contributed by the mutant allele ( Figure 1F). The phenotypic distinction between the control and patient organoids was evident at stage 3, when photoreceptors began to form outer segment-like structures visible as an apical ''brush'' in light microscopy. While clearly detectable in control organoids, formation of these structures was abrogated in CRX-I138fs patient organoids (Figure 2A; details of sampling size used for quantifications and description of statistical analyses are provided in methods and figure legends). As early as day 125, control organoids showed patches of developing rods as revealed by immunostaining of the rod visual pigment Rhodopsin, a transcriptional target of CRX. However, Rhodopsin staining was virtually absent in CRX-I138fs organoids ( Figure 2B). Similarly, cone photoreceptor opsins displayed altered expression. S Opsin immunostaining was delayed compared with the control ( Figures S2A-S2F), whereas the number of L/M Opsin + cones was substantially reduced ( Figure 2C). 3D renderings of organoids at day 200 confirmed the diminished opsin staining ( Figure 2D, Videos S1 and S2). Another outer segment component, Peripherin2, was also dramatically reduced in CRX-I138fs organoids ( Figure 2E). In contrast, developing ribbon synapses appeared unaffected (Figures S2G and S2H). Transcriptome analysis at key stages of organoid differentiation, days 90, 125, 150, and 200, identified differential expression of multiple transcripts associated with photoreceptor function ( Figure 2F), consistent with the established role of CRX (Furukawa et al., 1999;Hennig et al., 2008). Rod and L/M cone opsins (RHO, OPN1MW) were the most significantly diminished transcripts in CRX-I138fs organoids ( Figure 2F), in accordance with immunostaining results ( Figures 2B and 2C). Thus, histological and transcriptome analyses of retinal organoids showed aberrant photoreceptor development due to the I138fs mutation in CRX.
Next, we asked whether gene augmentation with normal CRX might alleviate the disease phenotype by competing with the mutant isoform. First, we tested transduction with two commonly used AAV capsid serotypes AAV2 and AAV8 with cytomegalovirus (CMV) promoter driving a GFP reporter. AAV2 capsid showed much higher proportion of cells expressing both GFP and CRX in organoids at day 150, 10 days post vector addition at day 140 (Figure 3A). We then tested human CRX promoter sequences for driving transcription in retinal cells that normally express the CRX gene. Three promoter fragments were combined to generate a shorter composite promoter (see Table S5), which was placed upstream of GFP and packaged into AAV2 capsid ( Figure 3B). As compared with (C) CRX expression in developing photoreceptors at day 90. CRX antibody labeled apically aligning developing photoreceptors in both control and patient organoids. Scale bar, 50 mm. Note a more diffuse pattern in the patient sample. Nuclei were stained with DAPI (4 0 ,6diamidino-2-phenylindole). (D) Developing photoreceptors of both genotypes also express the marker protein Recoverin. Scale bar, 20 mm. (E) Immunoblot analysis of organoid protein extracts at day 90. Molecular mass markers (kDa) are indicated on the left. Protein samples from control and patient stem cell-derived organoids (n = 3 biological replicates containing three organoids each) were run in separate lanes. Tubulin was used as a loading control. A smaller molecular weight band, which corresponds to the mutant CRX-I138fs48 isoform, is evident in the patient samples. (F) Densitometry quantification of CRX protein bands normalized to the Tubulin loading control. Total CRX protein is significantly more abundant in CRX-I138fs patient samples. Mean ± SD plotted, *p < 0.05, ***p < 0.001, one-way ANOVA.
transduction with ubiquitously expressed CMV promoter, CRX promoter-driven GFP signal localized to a distinct apical lamina corresponding to the photoreceptor layer in both control and CRX-I138fs patient organoids ( Figure 3B). The final therapeutic construct contained this CRX composite promoter driving human CRX expression (Figure 3C). This AAV-CRX vector was added to patient organoids at day 120, and the samples were analyzed at days 150 and 180. Transduction with the vector increased CRX mRNA and protein in treated organoids ( Figures S3A-S3D). Rhodopsin and L/M Opsin expression were used as surrogates for treatment, given their near absence in patient organoids. Retinal organoids transduced with 1310 11 vector genomes (vg) began to show rescue of Rhodopsin expression by day 150 (Figures S3E and S3F), which became more widespread by day 180 ( Figure 3D). Quantification of Rhodopsin immunolabeling intensities confirmed higher fluorescence compared with untreated  patient iPSC-derived organoids with both 1310 11 vg and 3310 11 vg doses and was almost half of signal intensity measured for the healthy control ( Figure 3D). Fluorescence intensity levels were similar at both vector doses, suggesting that 1310 11 vg per organoid is sufficient for substantial rescue of Rhodopsin expression. Treatment also partially restored L/M Opsin expression ( Figures 3E, S3G, and S3H). Rescue effect persisted at least for 6 months after treatment (day 300) with both Rhodopsin and L/M Opsin-positive cells still detectable ( Figures 4A and 4B). Long-term expression of CRX did not lead to activation of the apoptotic marker Caspase3 in AAV-treated organoid neural retina ( Figure 4C).
To decipher and validate the impact of CRX-I138fs mutation on specific cell types within retinal organoids, we performed single-cell RNA-sequencing (scRNA-seq) using 10x Genomics platform. Control, untreated CRX-I138fs and AAV-treated organoids were dissociated using a papainbased method (Fadl et al., 2020) at day 200 yielding 40,712 single-cell transcriptional profiles. Data processing using Seurat package identified cell clusters, which were assigned to known retinal cell types and visualized using UMAP dimension reduction ( Figure 5A and S4A-S4C). In this representation, major retinal cell classes (apart from ganglion cells) emerge from centrally located undifferentiated cells ( Figure 5A). Cell-type distribution was similar across the three sample origins (Figures 5A and S4B). Rods and cones formed well-defined differentiation trajectories in this manifold and were identified by expression of both common (CRX, RCVRN) and subtype-specific markers (rod: GNGT1, GNAT1; cone: ARR3, PDE6H; Figures 5B, S4D,and S4E). As predicted, CRX transcripts increased in photoreceptors after AAV-CRX transduction ( Figure 5C). Rod and cone expression profiles could be clearly separated based on control or patient sample origin, whereas AAVtreated cells occupied the space in between ( Figures 5D,  5E, 5G, and 5H). This shift was particularly evident by plotting the origin of most cells across hexagonal bins ( Figures  5E and 5H). ScRNA-seq detected partially rescued expression of opsins following AAV treatment (RHO, OPN1MW3, Figures 5F and 5I; OPN1MW3 was the most significantly dysregulated of three medium wavelength opsin genes OPN1MW1-3), as well as of other rod-and cone-specific transcripts ( Figures S4D and S4E; for each gene adjusted p value <0.05, nonparametric Wilcoxon rank-sum test with Bonferroni correction; minimum percent expressed = 10% cells, minimum log fold change = 0.25). CABP4, a retinal disease gene and direct transcriptional target of CRX (Assawachananont et al., 2018) showed a similar trend ( Figures S4F and S4G). Thus, single-cell analysis confirmed treatment effect of AAV-mediated overexpression of normal CRX. To determine whether the observed phenotypes and rescue were mutation-specific or could be generalized to other cases of dominant CRX-LCA, we examined organoids with the CR X K8 8N mutation. As for the frameshift mutation, no differences in morphology and expression of key retinal markers were evident at stage 1 or 2 of organoid differentiation ( Figures S5C-S5E). However, outer segmentlike structures were less developed at stage 3 in patient iPSC-derived organoids compared with the control ( Figures  6A and 6B). Immunostaining showed the presence of CRX and Recoverin, but severely diminished Rhodopsin and L/M Opsin staining in CRX-K88N organoids ( Figure 6C). Transcriptome analyses at days 120 and 200 revealed delayed upregulation of many photoreceptor-specific genes ( Figure 6D) and confirmed the loss of Rhodopsin and L/M Opsin expression (RHO, OPN1MW2; Figure 6D). Treatment of CRX-K88N organoids with AAV-CRX vector increased CRX levels ( Figure S6A), partially rescued Rhodopsin and L/M Opsin expression ( Figures 6E-6G) and reduced abnormal S Opsin levels ( Figure S6B), with a modest induction of rod visual arrestin and synaptic proteins (Figure S6B). Thus, CRX-K88N organoids showed a similar phenotype to CRX-I138fs, and AAV gene therapy was able to restore expression of CRX target genes.

DISCUSSION
Dominant diseases represent a considerable challenge for developing gene therapy because of the presence of a mutant allele that interferes with normal function (Ahmed et al., 2019;Collin et al., 2019). Overexpression of a normal copy of Rhodopsin or VSX2 has been shown to partially correct defects in mouse or retinal organoid models of respective dominant disease (Mao et al., 2011;Phillips et al., 2014). We previously observed reexpression of Rhodopsin following electroporation of wild-type Crx into the retina of a mouse model harboring a spontaneous dominant frameshift mutation in the Crx gene (Roger et al., 2014). Building on that observation, we tested the hypoth-esis that overexpression of a correct CRX cDNA increasing the ratio of normal to mutant protein (gene augmentation) may rescue defects in dominant human CRX retinopathies. We developed retinal organoid models of dominant CRX-LCA from patients with two different mutation types and demonstrate the feasibility of using AAV-delivered CRX as a therapeutic strategy. We note that patient iPSC-derived organoids have recently been used to provide proof-of-concept for AAV-mediated gene therapy of recessive X-linked retinitis pigmentosa caused by mutations in RP2 (Lane et al., 2020).
CRX-LCA patient organoids showed impaired maturation of photoreceptors with particularly disrupted expression patterns of visual opsins. Rhodopsin and L/M cone opsins were almost undetectable, whereas S Opsin expression was initially delayed but recovered later, consistent with findings in animal models (Roger et al., 2014;Tran et al., 2014). We observed increased total CRX protein levels in organoids of the patient with I138fs mutation, especially of the mutant isoform, whereas K88N organoids showed reduced overall levels of CRX protein. These findings suggest distinct potential downstream impacts of the dominant alleles: (1) competition of mutant protein with normal CRX for target binding sites, (2) prolonged occupancy of mutant protein at binding sites preventing the correct form from associating to the target DNA, and (3) disruption of stoichiometric interactions with other retinal transcription factors, such as NRL. The contribution of these mechanisms may vary depending on a specific mutation, but as we demonstrate, convergent molecular pathology such as disruption of opsin expression is evident.
We took advantage of scRNA-seq to determine diseaseassociated changes in photoreceptor subtypes as well as to evaluate treatment effect following gene therapy. We observed clear separation of patient cones and rods from control populations and reduced expression of cell-typespecific genes. In rods, partial shift toward control population with AAV treatment is associated with higher rescue of Rhodopsin expression in a subset of transduced cells, validating histological observations. Importantly, we detected  (legend continued on next page) continued rescue of Rhodopsin and L/M Opsin at 6 months posttreatment without any indication of apoptosis. Further investigations are needed to define optimal dosing as well as time window for gene therapy. In addition, we may require in vivo model(s) to fully evaluate the long-term impact of augmented CRX expression. In any event, single-cell transcriptomics offers a powerful approach to assess treatment approaches for retinopathies. The difference in rescue of L/M Opsin expression was evident between patient organoids of the two genotypes, suggesting differential interaction of mutant CRX with binding partners in cones. While restoration of cone gene expression was clearly evident in our study, further enhancement is likely by selecting a more potent capsid serotype targeting cones (Gonzalez-Cordero et al., 2018). Treatment of cone photoreceptors is of relevance because mutations in CRX also cause milder adult-onset cone-rod dystrophies and macular dystrophies, which could potentially benefit from gene therapy treatment described here.
In conclusion, we used retinal organoids as a platform to examine disease mechanisms and design gene augmentation therapy for dominant CRX-LCA. We suggest that our experimental strategy should be applicable in developing effective treatments for rare, and even dominant, inherited diseases of the retina and other parts of the central nervous system.

EXPERIMENTAL PROCEDURES
Retinal organoids from patient-derived iPSCs Skin biopsies were obtained from two CRX-LCA pediatric patients and respective healthy parental controls with informed consent (Table S1). Dermal fibroblasts were reprogrammed to generate iPSCs using a Sendai virus-based approach. Resulting iPSC lines were of normal karyotype and free of mycoplasma contamination (Table S2). iPSCs were differentiated following an established protocol (Kaya et al., 2019) with a minor modification of culturing dissected retinal organoids individually in a 96-well plate format (Table S3). Histology was performed as described in (Shimada et al., 2017), a list of all primary antibodies used for immunofluorescent staining is provided in Table S4. Images were obtained on Zeiss 700 confocal microscope.

Next generation sequencing
RNA was isolated from three pooled frozen organoids per sample using the QIAGEN RNeasy Kit according to manufacturer's manual. RNA quality was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies). Samples (with RNA integrity number >7) were used for library generation using the TruSeq Library Preparation Kit (Illumina Inc). Sequencing and bioinformatic analysis were performed as described (Kaya et al., 2019).
For single-cell RNA-seq, organoids were dissociated following a papain-based method and subjected to single-cell isolation and sequencing as described (Fadl et al., 2020). Processing and analysis of single-cell transcriptomes using Seurat package is detailed in Supplemental information.

Cloning of a composite human CRX promoter
Three fragments from the human CRX 5 0 untranslated region (NCBI: NG_008605.1) were amplified from human genomic DNA and ligated to produce a 631 base pair promoter sequence (Table S5). In this promoter, 189 nucleotides correspond to positions 3,085-3,274, 69 correspond to 3,323-3,392, and 361 correspond to 4,808-5,169 of the reference. This promoter element contains first exon of the human CRX gene (nucleotides 4,999-5,169).

Statistical analysis
At least three independently cultured organoids (individual wells of a 96-well plate) were used for quantification. Three differentiation batches were analyzed to verify disease phenotypes. In histological analyses, data from three sections were averaged to account for regional variation within individual organoids. With respect to next generation sequencing and single-cell sequencing, three organoids were pooled to obtain one sample. GraphPad Prism version 8.0 software was used to plot data and perform statistical analysis. For two group comparisons two-tailed Student's t test was used. For multiple groups, one-way ANOVA analysis was used with either Dunnett's or Tukey's post hoc tests. Quantitative data are presented as means ± standard deviation (SD). Exact p values are reported in the figures, and p values of <0.05 were considered statistically significant.

Study approval
Skin biopsies were obtained from the patients and unaffected family members at the National Eye Institute Clinical Center under an institutional review board-approved protocol #15-EI-0128, NCT01432847.
Detailed description of all experimental methods is provided in Supplemental information.

ACCESSION NUMBERS
The accession number for the RNA-seq data reported in this paper is GEO: GSE153101.

CONFLICTS OF INTEREST
K.K., S.H., Z.W., and A.S. are inventors on a patent application ''Gene therapy for treatment of CRX-associated retinopathies,'' submitted by the National Eye Institute. Z.W. is now an employee of PTC Therapeutics. Other authors declare no conflict of interest.