Gene Therapy Restores Mfrp and Corrects Axial Eye Length

Hyperopia (farsightedness) is a common and significant cause of visual impairment, and extreme hyperopia (nanophthalmos) is a consequence of loss-of-function MFRP mutations. MFRP deficiency causes abnormal eye growth along the visual axis and significant visual comorbidities, such as angle closure glaucoma, cystic macular edema, and exudative retinal detachment. The Mfrp rd6 /Mfrp rd6 mouse is used as a pre-clinical animal model of retinal degeneration, and we found it was also hyperopic. To test the effect of restoring Mfrp expression, we delivered a wild-type Mfrp to the retinal pigmented epithelium (RPE) of Mfrp rd6 /Mfrp rd6 mice via adeno-associated viral (AAV) gene therapy. Phenotypic rescue was evaluated using non-invasive, human clinical testing, including fundus auto-fluorescence, optical coherence tomography, electroretinography, and ultrasound. These analyses showed gene therapy restored retinal function and normalized axial length. Proteomic analysis of RPE tissue revealed rescue of specific proteins associated with eye growth and normal retinal and RPE function. The favorable response to gene therapy in Mfrp rd6 /Mfrp rd6 mice suggests hyperopia and associated refractive errors may be amenable to AAV gene therapy.


Structural modeling of human MFRP mutations
It is not known how MFRP mutations cause nanophthalmos or other retinal diseases, its exact role in the cell unknown, and its crystal structure is unsolved, so to gain insight into how point mutations might affect MFRP function, we used structural modeling. This analysis predicted several domains related to signal transduction, proteolysis, and endocytosis: two globular cubulin domains (CUB1 and CUB2), two domains related to the lowdensity lipoprotein receptor (LDLA1 and LDLA2), and a frizzled cysteine-rich domain (CRD). Further, the MFRP N-terminal region contains a putative transmembrane domain (TM; residues 70-90), suggesting it is a type II transmembrane protein with its C-terminal region projecting out to the extracellular space. 1 In the absence of a high-resolution structure for MFRP, we used structural modeling to ascertain how our patient's mutations might alter MFRP structure and function. 2 We also predicted how other MFRP mutations reported in the published literature might affect predicted functions of MFRP (Table S1). MFRP, is composed of 4 distinct domains, so we used a domain-assembly strategy to model MFRP structure, based on homology to related structures: cubulin domains (CUB1 and CUB2) were modeled using the crystal structure of cubulin as a template; LDLA1 and LDLA2 domains were generated based on the LDLR structure; the CRD domain was modelled based on the XWnt8-bound Frizzled-8 structure ( Fig. S2A-E). These individual domains were then assembled to generate a full-length MFRP model (Fig. S3).
Using this complete MFRP model, the mutations were mapped and found in all four domains (Fig. S3), allowing predictions as to how the mutations might affect protein function. For example, cubilin (CUB) domains are extracellular motifs comprised of ten b-strands in two five-stranded b-sheets 3 that mediate protein-ligand interactions and proteolysis, like in the Tolloid proteinase family (e.g. BMP-1). 3,4 Thus, although MFRP's catalytic activity has not been demonstrated, the presence of Tolloid repeats (CUB and LDLA domains) implicate MFRP in proteolytic pathways. Mutations in these domains could affect catalytic activity (Table 2).
Likewise, MFRP's CRD contains a 40% identity to the CRDs present on SFRPs and Frizzled receptors. The CRD domain is a cysteine-rich motif in secreted frizzled receptor proteins (SFRPs) and frizzled receptors, and bind Wnt and regulate Wnt-signaling pathways. 5 Although interactions between MFRP and Wnt have not been demonstrated, its role in cell fate signaling is suggested by the presence of the CRD domain. 5,6 These predictions notwithstanding, since the mutations are scattered throughout several different functional domains, small-molecule therapy or highly targeted gene editing may not be feasible ( Fig. S3; Table S1). Instead, either a cell-based therapy or gene replacement therapy might be more effective.

Proteomic Analysis
We first identified 20 proteins that were significantly upregulated in Mfrp rd6 /Mfrp rd6 . Among these proteins were thrombospondin-1 (TSP-1), argininosuccinate synthase-1 (ASS1), and cartilage oligomeric protein-1 (COMP-1). TSP-1 is a pro-angiogenic factor produced by the RPE and is believed to modulate choroidal vascular growth. 7 Increased TSP-1 levels may explain the vascular leakage that contributes to cystoid macular edema in nanophthalmos patients (Fig. 1C). ASS1 is involved in the production of arginine and nitric oxide (NO). NO is a potent vasodilator in the central nervous system and retina. 8 We hypothesized that increased retinal vasodilation would lead to increased intravascular leakage, causing exudative retinal detachment in nanophthalmos patients. 9 COMP-1 is an extracellular matrix protein and a marker of collagen turnover. 10 Increased levels of COMP-1 may promote scleral thickening, one of the pathological features of nanophthalmos. 11 Interestingly, this protein was one of the most prominently expressed proteins in Mfrp rd6 /Mfrp rd6 mice. These proteins returned to control levels following gene therapy ( Fig. 3A; Fig S4A).
We then identified 12 proteins were downregulated in Mfrp rd6 /Mfrp rd6 compared to controls and then restored to control levels following AAV2/8-mMfrp injection. Among these proteins were EGF-containing fibulin-like extracellular matrix protein 1 (EFEMP1), peroredoxin 6 (PRDX6), centrosomal protein of 97 kDa (CEP97), and ferritin (Table S4). Interestingly, mutations in EFEMP1 are known to cause a Mendelian form of macular degeneration known as Malattia Leventinese (MLVT). Patients with MLVT often display altered RPE cell ultrastructure, sub-retinal RPE deposits, and separation between the RPE and Bruch's membrane. 12,13 Notably, our Mfrp rd6 /Mfrp rd6 mice displayed similarly altered RPE morphology in our RPE wet mount, suggesting that rescue of EFEMP1 levels contributed to restored RPE morphology (Fig. S3). 14 PRDX6 is an antioxidant protein expressed in astrocytes and Müller cells. Since oxidative stress is implicated in retinal degeneration, we hypothesize that restoration of PRDX6 levels in the RPE contributed to reduced photoreceptor cell death and normalized retinal function in AAV2/8-mMfrp mice. 15,16 CEP97 is involved in regulating cilia assembly, a process that is critical to the maintenance of photoreceptor outer segments in the retina. 17 Finally, ferritin is involved in iron storage and transport. Increased iron levels in the eye have been implicated in oxidative stress and retinal degeneration. 18 Gene therapy rescue of ferritin may play a role in restoring iron levels in the retina and decreasing toxicity.
BSG is a glycoprotein that is implicated in retinal development. Knockout of Bsg in mice has been shown to cause altered retinal function, photoreceptor degeneration, and choroidal neovascularization. 20,21 SLC16A8 is a proton-coupled monocarboxylate transporter involved in sub-retinal pH regulation. Interestingly, knockout of Scl16a8 in mice leads to a decrease in subretinal pH due to the accumulation of lactate and causes altered visual function. 22 Previous genome-wide association studies (GWAS) have implicated the SLC16A8 locus in age-related macular degeneration. 23 CSPG5 is a chondroitin-sulfate proteoglycan that has been previously found to be upregulated in mouse models of retinal degeneration (Rpe65 -/-). Although increased levels of CSPG5 have been implicated in retinal degeneration, its role in this process is poorly understood. 24

05). (B)
A total of 36 proteins were upregulated in AAV2/8-mMfrp mice that were not seen in control mice or in Mfrp rd6 /Mfrp rd6 mice. We hypothesized that these pathways were upregulated in response to the AAV vector injection (p<0.05). (C) A total of 12 proteins were restored from Mfrp rd6 /Mfrp rd6 levels following gene therapy (p<0.05). Details of the individual proteins are listed in Table S4. (D) A total of 65 proteins were not rescued to control levels following gene therapy.