Rescue of the MERTK phagocytic defect in a human iPSC disease model using translational read-through inducing drugs

Inherited retinal dystrophies are an important cause of blindness, for which currently there are no effective treatments. In order to study this heterogeneous group of diseases, adequate disease models are required in order to better understand pathology and to test potential therapies. Induced pluripotent stem cells offer a new way to recapitulate patient specific diseases in vitro, providing an almost limitless amount of material to study. We used fibroblast-derived induced pluripotent stem cells to generate retinal pigment epithelium (RPE) from an individual suffering from retinitis pigmentosa associated with biallelic variants in MERTK. MERTK has an essential role in phagocytosis, one of the major functions of the RPE. The MERTK deficiency in this individual results from a nonsense variant and so the MERTK-RPE cells were subsequently treated with two translational readthrough inducing drugs (G418 & PTC124) to investigate potential restoration of expression of the affected gene and production of a full-length protein. The data show that PTC124 was able to reinstate phagocytosis of labeled photoreceptor outer segments at a reduced, but significant level. These findings represent a confirmation of the usefulness of iPSC derived disease specific models in investigating the pathogenesis and screening potential treatments for these rare blinding disorders.

Inherited retinal dystrophies represent a heterogeneous group of diseases, which are now the most common cause of registered blindness in the working age population in the UK 1 . Few naturally occurring disease models exist for this group of disorders. Furthermore, access to affected retina and retinal pigment epithelium (RPE) is restricted, providing a very limited amount of human tissue to study. With the advent of induced pluripotent stem cell (iPSC) technology 2 disease specific human models can be generated in vitro with a theoretically limitless supply of cells that can be differentiated to the tissue of interest to study these diseases and test drugs or genetic therapies 3,4 .
The RPE is a cellular monolayer that has an essential supporting role to the photoreceptors in the retina, being involved in retinol cycling, nutrient transport, growth factor production and phagocytosis of the discarded photoreceptor outer segments 5 .
MER receptor tyrosine kinase (MERTK) is a tyrosine kinase that is expressed at the apical surface of the RPE. It is a key regulator of the recognition and internalisation of photoreceptor outer segments (POS) during the process of phagocytosis by RPE cells 6 . The MERTK protein is a 999 amino acid polypeptide with a molecular weight of 110 kDa, which is subject to post-translational modifications such as N-linked glycosylation 7 . MERTK deficiency results in a form of retinitis pigmentosa, RP38, a rod cone dystrophy for which there is no effective treatment 8 .
Recently, the results of a gene therapy trial with AAV2 mediated delivery of MERTK to RP38 individuals have been published, with mixed results 9 . This highlights the requirement for new strategies to test potential therapeutics for these genetic disorders. One potential method is to target a specific variant using small molecules. In the case of RP38, several of the known human MERTK variants result in a premature stop codon [10][11][12][13][14] . In these cases, a potential strategy is to employ translational readthrough inducing drugs (TRIDs), which stimulate the bypass of the premature stop codon 15 . This allows continuation of translation, restoring a full-length protein. TRIDs have been shown to be effective in restoring protein and function in previous studies of iPSC-RPE from an individual with X-linked retinitis pigmentosa caused by a mutation in RP2 16 .
In this study, we have generated fibroblasts cells from a skin biopsy of an individual with compound heterozygous variants in the MERTK gene. These cells were then reprogrammed to iPSC and differentiated into mature RPE cells with an aim to characterise this iPSC derived model of MERTK retinitis pigmentosa, and employ TRIDs to restore MERTK function.

Results
Characterisation of MERTK variant clinical phenotype and iPSC derived RPE. We obtained a skin biopsy from a 26-year-old male (GC15566) with compound heterozygous variants in the MERTK gene, consisting of a single base change in the first base of intron 1 (61+1 G>A) and a nonsense variant in exon 14 (1951 C>T) 13 . Snellen visual acuity was 6/36 in the right eye and 3/60 in the left. Fundus examination revealed bilateral central retinal degeneration, consistent with an inherited maculopathy, bilateral optic nerve head pallor and narrowed and atrophic retinal blood vessels (Fig. 1A). There were no areas of hyperpigmentation noted in the peripheral retinal examination. Optical coherence tomography (Spectralis, Heidelberg) shows an increased right macular retinal thickness nasally, a mild degree of epiretinal fibrosis, general preservation of the outernuclear layer but intermittent loss of the ellipsoid zone and RPE layers (Fig. 1B). Blue light, widefield autofluorescence (Optos) shows a hypofluorescence centrally, matching the area of central macular degeneration with general preservation of the periphery apart from a few punched out lesions outside the macular area (Fig. 1C). Adaptive optics assisted flash photographs taken with the Rtx-1 fundus camera (ImagineEyes) of the right eye at 5 degrees superior to the fovea (grey box on Fig. 1A) shows disorganised, scattered reflections consistent with photoreceptor loss (Fig. 1D). Macular microperimetry (Nidek) shows reduced retinal sensitivity at the macula (Fig. 1E), with the blue collection of spots marking the fixation of the eye.
Dermal fibroblasts were expanded from the biopsy and reprogrammed into iPSC using episomal vectors in parallel with a control human BJ fibroblast cell line ( Figure S1A-C). The iPSC were differentiated into RPE using a spontaneous method of differentiation 17 . Pigmented RPE cells were manually purified and cultured as a monolayer that gradually re-pigmented after 6-8 weeks (Fig. 2).
MERTK mRNA was expressed in both the control RPE and MERTK-RPE, but at a lower level in the MERTK-RPE (Fig. 2E). However, the MERTK protein was only detectable in the control RPE, whilst the CRALBP protein, an RPE cell marker was present in both control RPE and MERTK-RPE at equal levels ( Fig. 2D,E). Analysis of the MERTK gDNA sequence confirmed a change in the first base of intron 1, as previously described 13 . Within the coding sequence of the gDNA five heterozygous variants were detected ( Figure S2), two missense, two synonymous and one nonsense. Analysis of the MERTK cDNA prepared from MERTK-RPE cells showed only the wild type sequence at the exon 1/2 border (Fig. 2F). In the MERTK cDNA from the MERTK-RPE, only one variant was observed at each of the five positions that had been noted to be heterozygous in the gDNA ( Figure S2). Of these five, only one differed from the wild type, NCBI sequence. This was the C>T variant at coding position 1951, leading to a premature stop codon in exon 14 (Fig. 2G). There was no evidence of a truncated MERTK protein by Western blot analysis, suggesting that if a truncated protein was produced it is unstable and degraded (Fig. 2D).

MERTK iPSC derived RPE are unable to phagocytose. The proposed function of MERTK in RPE
cells is to mediate the engulfment of POS during phagocytosis 6,18 . To test this hypothesis in our MERTK-RPE, we modified established phagocytosis assays 19,20 , exposing the RPE cells to isolated POS in vitro. Four hours after the addition of isolated POS to both the control and MERTK-RPE, POS are clearly visible inside the control RPE by electron microscopy (Fig. 3A), with none observed in the MERTK-RPE (data not shown). After 6 hours of POS exposure, FITC labeled POS are clearly visible inside the control RPE but were not observed in the MERTK-RPE on confocal orthogonal stacks ( Treatment with the small molecule PTC124 (10 μgml −1 ), led to the detection of a MERTK immunoreactive species of the correct mobility by Western blot analysis that was not present in vehicle-treated MERTK-RPE cells (Fig. 4C). MERTK immunoreactivity was also detectable on immunocytochemistry in the MERTK-RPE treated with PTC124 (Fig. 4D).

Discussion
A deeper phenotyping of this individual with variants in MERTK reveals that, outside the macula area, there is little hyperpigmentation, a sign of irreversible retinal damage. In addition, there is some preservation of the gross retinal anatomy and persistence of the retinal layers, displayed by OCT. Furthermore, while there is evidence of photoreceptor loss on adaptive optical imaging, there is some preservation of function as outlined by the microperimetry. Not only is there some ability to see a stimulus above the baseline, the individual has good fixation, throughout the test. Further evidence that this represents a less severe case of MERTK-RP comes from comparison with other individuals with variants in MERTK [8][9][10][11][12][13][14] . The results of the phenotyping displayed in Fig. 1 are thus encouraging, as any translation of therapy would have its best results in a milder disease form.
From a skin biopsy of this individual with compound heterozygous variants and predicted loss of function in the MERTK gene we derived dermal fibroblasts which were reprogrammed to iPSC and then differentiated into RPE cells. These RPE cells have many of the characteristics of primary RPE cells; they grow as a epithelial monolayer with a cobblestone appearance, contain apical pigmented melanosomes, basal nuclei and express proteins involved in the process of retinol cycling 21 . Both the control RPE and the MERTK-RPE express the MERTK mRNA as evidenced by RT-PCR but the MERTK-RPE mRNA was present at a lower level. This difference in expression was more pronounced at the protein level with total absence of detectable MERTK protein in the MERTK-RPE.
The gDNA sequence of the MERTK gene in the MERTK variant individual revealed five heterozygous variants within the coding sequence. However, at each of these positions in the MERTK cDNA from the MERTK-RPE, only one of the variants was detectable. This suggests that only one allele is expressed; this is the allele that includes the premature stop codon in exon 14. We also examined the sequence at the exon 1/2 border as the individual's MERTK genomic sequence suggested that the variant in intron 1 would likely lead to a different splicing pattern and the insertion of an extra 11 codons between exon 1 and exon 2 13 . Our analysis of the MERTK cDNA sequence from the MERTK-RPE revealed a sequence at the exon 1/2 border that matches that of the normal MERTK transcript (NM_006343.2 NCBI database). These data are consistent with the hypothesis that the allele carrying the variant in the first base of intron 1 is not expressed at the RNA level or is rapidly removed. However, the allele carrying the premature stop codon in exon 14 is detectable. This suggests that in this case, the MERTK defect could be amenable to TRIDs.
Both the control RPE and MERTK-RPE express Integrin αV, a protein required for the rhythmic activation of MERTK 18 , at the apical surface, suggesting that the cells are otherwise primed for phagocytosis. Functional phagocytosis by the control RPE cells was confirmed by electron microscopy following exposure to isolated POS. The POS were clearly engulfed and internalised by the control RPE following a 4 hour exposure to POS. In contrast, no internalised POS were observed in the MERTK-RPE cells. Internalisation of FITC-labeled POS was also observed and quantified in control RPE using confocal stack analysis. In comparison to the control RPE, no internalised POS were observed in MERTK-RPE cells. In addition, it became evident in the control RPE that a 6-hour time point was preferable to a 20-hour assay time point, as at this longer interval, the FITC fluorophore is likely degraded by the RPE cells. A comparison of the number of bound but not internalised POS was measured between the control and MERTK-RPE during phagocytosis ( Figure S3). The difference just reached significance by t-test (0.03) and this minor difference could be accounted for by the small contribution of MERTK to binding POS, rather than the major contribution of Integrin αVβ5 in binding POS. This suggests that the major difference noted in phagocytosis between the control and MERTK-RPE occurs at the internalisation phase.
Exposure of the MERTK-RPE to G418, an aminoglycoside related to the antibiotic gentamicin, was able to restore expression of a protein of similar size to MERTK that was not present in the untreated cells. However, this restoration of protein did not correlate into a functional rescue, as the MERTK-RPE were still unable to phagocytose POS. This might be due to a toxic effect of the G418 as it was noted that the control RPE had decreased levels of phagocytosis in the presence of the G418. Aminoglycosides are known to be toxic to the retina 22 , but the exact mechanism is not clear 23 . Interference with normal RPE function could be a potential mechanism. G418, one of the more toxic aminoglycosides, is thought to cause toxicity by binding to mitochondrial ribosomes which are similar to prokaryotic ribosomes. In doing so they inhibit mitochondrial protein synthesis and cause oxidative stress. RPE are known to have high levels of mitochondria in order to provide energy for processes like phagocytosis which may render them particularly sensitive the aminoglycoside toxicity 24 .
Following treatment of the MERTK-RPE with another TRID, PTC124, a small molecule which has been approved by the European Medicines Agency for Duchenne muscular dystrophy 25 , we were able to restore expression of MERTK in the MERTK-RPE. The immunoreactivity was present, however, as multiple bands. One explanation for this is that the protein is differentially post-translationally modified depending on which amino acid is used during the readthrough process. Furthermore, in a phagocytosis assay, the PTC124 treatment restored 12% of the phagocytic function to the MERTK-RPE cells. It is debatable whether an improvement of 12% would be enough in vivo to slow the progression of the disease. PTC124 has been used to treat Duchenne muscular dystrophy 26 and cystic fibrosis 27 . In the preclinical studies of PTC124 for treatment of Duchenne muscular dystrophy (DMD), the drug was able to restore 20% of the protein in a mouse model of the disease and this partially restored protein was shown to localise to the correct cellular compartment in all muscles biopsied. In addition, the treated mice exhibited improved protection from induced injury, an established assay of DMD progression in the model. In phase 1 clinical trials, PTC124 had a very good safety profile and the functional rescue of protein translated, with 34% of subjects taking the drug showing qualitative increase in immunofluorescence for the protein in post treatment muscle biopsies. Furthermore, in phase 2 trials, patients treated with the drug were able to walk further in standardised timed tests than those taking placebo, and while this did not meet significance, other endpoints did, including slowing of disease progression 26 . This level of rescue is comparable to the results shown herein, suggesting that only partial recovery of the protein may facilitate a functional improvement in vivo.
The mechanism of action of TRIDs is not completely understood but the drugs are thought to interfere with ribosomal fidelity so that a near-cognate aminoacyl tRNA can bind at the site of a premature stop codon resulting in the incorporation of another amino acid in its place, allowing translation to continue 15 . It is also postulated that the readthrough effect of TRIDs is dependent not just on the nonsense variant but also in the context of the local sequence, the particular bases that make up the stop codon and the stability of the resulting protein. The TRID PTC124 performs best when acting on the UAG stop codon followed by a cytosine 28 , which is the sequence of the MERTK premature stop codon presented here and may account for the efficiency observed in this case. We acknowledge that this is the second MERTK iPSC RPE model 29 , however it is the first to show a treatment effect and amelioration of function.
iPSC models offer an attractive alternative to animal models as they allow the study of human tissue. The Royal College of Surgeons (RCS) rat, a natural animal model that harbours a spontaneous homozygous deletion in the MERTK gene has been used for decades as a general model of retinal degeneration 30, 31 and a specific model of MERTK related retinal disease. However, the RCS genotype 32, 33 , does not recapitulate any observed in man 10 thus diminishing its use as a clinical model in the post genome age.

Conclusion
We have been able to create and characterise iPSC derived RPE cells with MERTK deficiency to create an accurate in vitro model of human disease. Our results show that the MERTK-RPE cells have a defect in the engulfment of POS. Using the translational readthrough inducing drugs G418 and PTC124 we were able to restore expression of the MERTK gene to detectable full-length protein in the MERTK deficient RPE. Moreover, the PTC124 compound was able to rescue 12% of the phagocytic function of the RPE cells. It may be possible, therefore, to use TRIDs to treat retinitis pigmentosa due to nonsense variants in MERTK, which hitherto has no effective therapy. Furthermore, human iPSC derived models may also be useful to bridge the gap between preclinical animal studies and clinical trials to test other treatments such as gene therapies.

Materials and Methods
Clinical phenotyping. Fundus photos were captured with a digital camera and Topcon image capture system. Widefield images and autofluorescence images were captured using the Optos 200Tx retinal imaging system. Optical coherence tomography images and autofluorescence images were captured using a Heidelberg Spectralis. Microperimetry was performed using the Nidek MP-1 microperimeter set to Humphrey 10-2(20 db) with a white 200 ms Goldman III stimulus and a white 2° single cross fixation target. Adaptive optics (AO) images were captured using the Imagine Eyes rtx1 ™ Adaptive Optics Retinal Camera.
Generation of fibroblasts from a skin biopsy. The research adhered to the tenants of the Declaration of Helsinki and was approved by the Moorfields Eye Hospital Ethics Committee and the Health Research Authority NRES Committee London. Informed, written consent was gained and a biopsy was obtained from an individual with a MERTK variant and fibroblasts cultivated from it as previously described 16, 34 . Reprogramming of fibroblasts. Control fibroblasts (BJ line) were obtained from Stemgent. MERTK and control fibroblasts were reprogrammed to iPSC with episomal delivery of SOX2, OCT3/4, KLF4, L-MYC and LIN28 (27077, 27078, 27080, Addgene) using a protocol that has previously been described 16,35 .
These stem cells had a normal morphology, expressed markers of pluripotency, had a normal karyotype and were able to differentiate into endoderm, mesoderm and ectoderm ( Figure S1D-F).
Maintenance of iPSC, generation of RPE. iPSC colonies were cultured on Matrigel hESC-qualified matrix (354277, BD Biosciences) in Essential 8 (A1517001, Thermo Fisher). Once 80% confluent, the media was changed to Differentiation Media 17 . From here on, cultures were fed twice weekly for at least a further 8 weeks until pigmented foci were observed. These pigmented foci were isolated manually and purified, fed twice weekly with X-VIVO 10 (LZBE 04-743Q, Lonza) with gentamicin (ThermoFisher). All experiments were performed with passage 2 RPE aged 8 to 16 weeks.
Phagocytosis assay. Photoreceptor outer segments (POS) were purified from fresh sheep eyes obtained from a local abattoir (Chelmsford, UK) using a previously published protocol 19 . Some POS were labeled with Fluorescein-5-Isothiocyanate (FITC 'Isomer I' , F1906, ThermoFisher). Samples for electron microscopy analysis of phagocytosis used unlabeled POS. The labeled or unlabeled POS were added to DMEM with 4.5 g/L glucose, L-glutamine and pyruvate (41966029 ThermoFisher), 10 mM GAS6 (885-GS-050, R&D systems, MN, USA) and 10% FBS (10099141, ThermoFisher) and exposed to the iPSC derived RPE at a ratio of 10 POS per cell for 6 hours or 20 hours. At the endpoint, the cells were fixed in 4% PFA at 4 C for 20 mins, permeabilised with 0.3% triton in block solution (1X PBS, BSA 0.5%, Glycine 1%, Sodium Azide 0.1%) and stained with rhodamine phalloidin