Article Text

Download PDFPDF

ADVIRC is caused by distinct mutations in BEST1 that alter pre-mRNA splicing
  1. R Burgess1,
  2. R E MacLaren2,3,
  3. A E Davidson1,
  4. J E Urquhart1,
  5. G E Holder2,3,
  6. A G Robson2,3,
  7. A T Moore2,3,
  8. R O’ Keefe4,
  9. G C M Black1,5,
  10. F D C Manson1,5
  1. 1
    Genetic Medicine, University of Manchester, Manchester Academic Heath Science Centre, Central Manchester University Hospitals NHS Foundation Trust, St. Mary's Hospital, Manchester, UK
  2. 2
    Moorfields Eye Hospital, London, UK
  3. 3
    Institute of Ophthalmology, University College London, London, UK
  4. 4
    Faculty of Life Sciences, The University of Manchester, Manchester, UK
  5. 5
    Manchester Royal Eye Hospital, University of Manchester, Manchester Academic Heath Science Centre, University Hospitals NHS Foundation Trust, Manchester, UK
  1. Correspondence to Dr F Manson, AV Hill Building room 1,002. The University of Manchester, Oxford Road, Manchester, M13 9PT, UK; forbes.manson{at}


Autosomal dominant vitreoretinochoroidopathy (ADVIRC), a retinal dystrophy often associated with glaucoma and cataract, forms part of a phenotypic spectrum of ‘bestrophinopathies’. It has been shown previously that ADVIRC results from BEST1 mutations that cause exon skipping and lead to the production of shortened and internally deleted isoforms. This study describes a novel ADVIRC mutation and show that it disrupts an exonic splice enhancer (ESE) site, altering the binding of a splicing-associated SR protein. As with previous ADVIRC mutations, the novel c.704T→C mutation in exon 6 altered normal splicing in an ex vivo splicing assay. Both this and another exon 6 ADVIRC-causing mutation (c.707G→A) either weakened or abolished splicing in an ESE-dependent splice assay compared with a nearby exon 6 mutation associated with Best disease (c.703G→C). Gel shift assays were undertaken with RNA oligonucleotides encompassing the ADVIRC and Best disease mutations with four of the most commonly investigated SR proteins. Although SC35, SRp40 and SRp55 proteins all bound to the wild-type and mutated sequences with similar intensities, there was increased binding of ASF/SF2 to the two ADVIRC-mutated sequences compared with the wild-type or Best disease-mutated sequences. The exon skipping seen for these two exon 6 ADVIRC mutations and their affinity for ASF/SF2 suggests that the region encompassing these mutations may form part of a CERES (composite exonic regulatory elements of splicing) site.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The BEST1 gene encodes bestrophin-1, a transmembrane protein of the basolateral membrane of the retinal pigment epithelium (RPE) that acts as a chloride channel when expressed heterologously.1 2 Two models exist for the membrane topology of bestrophin-1. The main difference between them relates to the membrane insertion of transmembrane domains 5 and 6.3 4 Pathogenic mutations in BEST1 are associated with a range of retinal disorders including Best disease (OMIM 153700), autosomal dominant vitreoretinochoroidopathy (ADVIRC) (OMIM 193220) and autosomal recessive bestrophinopathy (ARB) (OMIM 611809).5 6 7 8

ADVIRC is a rare condition characterised by a peripheral retinal circumferential hyperpigmented band, punctuate white retinal opacities, fibrillar condensation of the vitreous, vascular abnormalities and neovascularisation9 found in families with an abnormal electro-oculogram (EOG) light rise.7 In all cases the mutations (c.256G→A p.V86M, c.707A→G p.Y236C, c.715G→A p.V239M) caused exon skipping (of exons 4, 6 and 7 respectively) and resulted in the production of bestrophin-1 isoforms containing in-frame deletions.7 It is hypothesised that all three mutations disrupt exonic splice enhancer sequences.

Exonic splice enhancer and silencer (ESE and ESS) sites play an important role in defining pre-mRNA exon/intron boundaries for recognition by the spliceosome and are bound by serine/arginine-rich (SR) proteins and heterogeneous nuclear ribonucleoproteins. Single-base substitutions within ESE or ESS sites and their consequent effect on splicing are now recognised to be an important cause of human genetic disease (eg, Disset et al 10, Cartegni et al11).

In this study we describe a novel missense mutation, c.704T→C p.V235A, in BEST1 that is associated with ADVIRC and show that it too alters pre-mRNA splicing in vitro. Using ESE-dependent splice assays and gel shift assays, we found altered SR protein binding. These data help to elucidate the mechanism responsible for the altered splicing of BEST1 seen in ADVIRC.


The study was approved by the Research Ethics Committees at Moorfields Eye Hospital and informed consent was obtained from all participants before investigation.

Both affected family members reported in this study underwent a full ophthalmic examination, consisting of dilated fundoscopy, ultrasound biometry and orbscan corneal topography, and electrophysiological testing according to International Society for Clinical Electrophysiology of Vision (ISCEV) standards.12 13 Genomic DNA was extracted from venous blood using standard procedures.

Patient details

Patient 1

The proband (II-1 in fig 1), who had childhood-onset myopia, was first seen in 1988 aged 38 years when she reported difficulty with reading. Her visual acuity was 6/9 in each eye and there was bilateral microcornea, microspherophakia with mild lens opacities, shallow anterior chambers and myopic optic discs. Intraocular pressure was normal. There was peripapillary chorioretinal atrophy and a post-oral band of atrophy and pigmentation running circumferentially through 360 degrees (fig 2A).

Figure 1

Pedigree of the two patients reported here. The proband (patient 1) is II-1 and patient 2 is II-2. Family members I-1, II-3 and II-4 were not examined for this study but are reported to have some characteristics consistent with autosomal dominant vitreoretinochoroidopathy. The clinical findings are described more fully in the text.

Figure 2

Left fundus of the proband (patient 1) and sequence electropherograms. (A) Composite image showing peripapillary atrophy around optic disc and demonstrating peripheral circumferential hyperpigmentation characteristic of ADVIRC (arrows). (B) Proband with c.704T→C mutation (arrow) in exon 6 of BEST1 (top), wild-type control (bottom).

In 1994, the patient underwent a right pars plana lensectomy and 3 years later developed acute pupil-block glaucoma in the left eye, which responded to treatment. She later underwent a successful left phacoemulsification and lens implant. In 2006, her intraocular pressure was 36 mmHg in the right eye and 29 mmHg in the left, which fell to 28 mmHg and 20 mmHg, respectively, after treatment with Xalatan eye drops. Visual acuity is good (6/12 in each eye) and there has been no change in appearance of the fundus.

Electrodiagnostic testing showed an undetectable pattern electroretinograms (ERG) in the right eye and markedly reduced responses in the left. Full-field ERG found evidence of rod and cone dysfunction. The subject was unable to cooperate with EOG testing.

Patient 2

The brother (II-2 in fig 1) of the proband was diagnosed aged 47 years with low-tension glaucoma and had noticed increased problems with night vision over the preceding 3 years. Visual acuity was 6/12 in the right eye and 6/6 in the left. Fundus examination found bilateral optic-disc cupping and a broad band of post-oral pigmentation and atrophy in each eye. Electrodiagnostic tests found a normal pattern ERG in each eye but the full-field ERG found mild abnormality of rod and cone function. The EOG light rise was 200% in each eye, which is within normal limits.

Family history

The father (I-1 in fig 1, now deceased) of the proband was known to have very poor night vision and ‘macular degeneration’, and had undergone cataract surgery. The mother had normal sight. The proband’s four siblings include patient 2 and another brother (II-3) with similar ‘small eyes’ to his sister; he had undergone a retinal detachment repair and bilateral angle closure glaucoma. He subsequently had cataract surgery but was not available for study. The proband’s two sisters were not examined for this study. One (II-4) had pre-senile cataracts and underwent surgery aged 49 years.

DNA sequencing

The exonic sequences and 50–100 bp of flanking intron sequence of BEST1 were analysed by direct sequencing from PCR amplicons (primers and conditions available on request). The putative pathogenic sequence alteration was tested for co-segregation in another family member. The absence of putative pathogenic sequence alterations in 210 control chromosomes was confirmed by single-stand conformation polymorphism/heteroduplex analysis.

Ex vivo splice assay cloning

Plasmids encoding wild-type and mutant BEST1 fragments were generated by PCR amplification from patient genomic DNA. Fragments were subcloned into the α-globin–fibronectin–extra domain B (EDB) minigene and sequenced to ensure fidelity and orientation.7 14

Ex vivo splice assay

Constructs containing the sequence of interest were transiently transfected into HEK293 cells using Lipofectamine reagent (Invitrogen, Paisley, Renfrewshire, UK). After 24 h, the cells were pelleted and RNA extracted using Trizol (Invitrogen). After DNase treatment (Promega, Southampton, Hampshire, UK), cDNA was produced by reverse transcription (RT) PCR from ∼1 μg RNA. Vector-specific primers were used to establish cDNA linearity loading controls for the experimental PCR assays, which used primers designed to the vector and an internal exonic sequence, and were compared with a wild-type control sequence.

Dsx exonic splice enhancer-dependent in vitro splice assay

This assay takes advantage of a naturally occurring ESE site in the Drosophila Doublesex gene (Dsx). Complementary oligonucleotides (Invitrogen) of 22 bp were annealed and cloned into the Dsx vector (a kind gift from Dr A Bindereif)15 linearised with EcoRI and HindIII. PCR amplicons encompassing the adjacent T7 promoter were used for RNA transcription using T7 polymerase (Promega) from 1 μg of purified PCR product in a 50 μl reaction incubated at 37°C for 1–3 h. The RNA was then treated with DNase, extracted using phenol–chloroform, and purified on agarose gels.

In vitro splice assays were carried out at 30°C for 2 h using 1 μmol/l purified RNA in a reaction containing 0.5 mmol/l ATP, 20 mmol/l creatine phosphate, 3.2 mmol/l MgCl2, 20 mmol/l Hepes–potassium hydroxide (pH 7.3), 2.6% polyvinyl alcohol and 15 μl HeLa nuclear extract (CIL Biotech, Mons, Belgium). The reaction was stopped with 3 mol/l sodium acetate and the spliced RNA was extracted by phenol–chloroform and precipitated in ethanol overnight at −20°C. Analysis by RT-PCR was carried out as described above with 1–10 μl RNA recovered from the splice assay.

Serine/arginine-rich protein purification

Purified T7-tagged SR protein was obtained by expression and purification from transfected HEK293 cells as previously described.16

Gel shift assays

Target RNA oligonucleotides (Dharmacon, Epsom, Surrey, UK) were end labelled with 32P dATP (GE Healthcare, Little Chalfont, Buckinghamshire, UK) and 25,000 dpm (∼5 nmol/l) was mixed with 2.8–3.3 μmol/l purified protein (dialysed against BC100 buffer: 20 mmol/l Tris-HCl pH 8, 100 mmol/l potassium chloride, 0.2 mmol/l EDTA pH 8 and 20% glycerol) and 10–20 U RNasin ribonuclease inhibitor (Promega). The reaction volume was made up to 10 μl with 1× BC100 buffer and was incubated for 30 min on ice. The reaction was UV cross-linked for 140 s (Stratalinker) before running the reactions on an 8% native acrylamide gel in 0.5× Tris–borate–EDTA for 2.5 h. The gel was fixed in 10% methanol/10% acetic acid and dried before autoradiography.


Molecular findings

Mutation screening of the BEST1 coding regions in both patients identified a novel c.704T→C variant in exon 6 that predicted the production of a missense alteration, p.V235A, in bestrophin-1 (fig 2B). The mutation is located 3 bp from another ADVIRC mutation (c.707A→G p.Y236C) that we have previously described. Interestingly, substitution of valine at residue 235 with either leucine or methionine (c.703G→C and c.703G→A respectively) has been described in patients with Best disease.6 17 Thymine at position c.704 is conserved down to the nematode Caenorhabditis elegans (fig 3). The alteration was not seen in 210 control chromosomes.

Figure 3

Multiple alignment of nucleotide sequences around c.704 (boxed) from 11 species using ClustalW2 ( The species are (in order): Homo sapiens, Pan troglodytes, Canis familiaris, Rattus norvegicus, Takifugu rubripes, Tetraodon nigroviridis, Danio rerio, Gallus gallus, Xenopus tropicalis, Drosophila melanogaster and Caenorhabditis elegans. The putative CERES site, bounded by two Best disease mutation sites at c.703 and c.710, is underlined in the human sequence.

c.704T→C alters splicing of bestrophin-1 mRNA

For both wild-type and the ADVIRC-causing c.704T→C variant, a region of BEST1 genomic sequence containing exons 6 and 7, and the flanking and intervening intronic sequences was cloned into the α-globin–fibronectin–EDB minigene splice assay vector. An ex vivo splice assay was then carried out in HEK293 cells as described previously,7 and products were extracted and sequenced. The wild-type construct was spliced appropriately with a single product of approximately 260 bp corresponding to the vector exons spliced to BEST1 exons 6 and 7. The mutant construct gave rise to the wild-type product and an additional altered splice product of approximately 340 bp (fig 4A–C). This larger product included an extra copy of exon 6 spliced between exons 6 and 7.

Figure 4

Splicing and gel shift assays.(A-C) Ex vivo splicing assay of exon 6 ADVIRC mutation c.704T→C. (A) Wild-type or mutant exon 6, wild-type exon 7 and flanking intronic sequences were cloned into the α-globin–fibronectin–extra domain B (EDB) minigene. Arrows and arrowheads represent the primers used to PCR from the resultant mRNA. (B) Left, agarose gel of products from PCR across the minigene construct (primers represented by arrows). Products are graphically represented on the right. Right, agarose gel showing linearity of the template. (C) Left, agarose gel of products from PCR within the exon of the minigene construct (primers represented by arrowheads). The vector product is depicted graphically on the right and demonstrates equal cDNA template was used for the wild-type and mutant assays. Right, agarose gel showing linearity of the template. (D) In vitro Dsx splice assay for four exon 6 BEST1 mutations. Wild-type and Best disease sequences (c.703G→C and c.710C→G) allow splicing to occur and therefore contain an exonic splice enhancer (ESE) site. The two autosomal dominant vitreoretinochoroidopathy (ADVIRC) mutated sequences weaken (c.704T→C) or abolish (c.707A→G) this activity. (E) Gel shift assays to determine SR protein binding. RNA oligonucleotides (corresponding to wild-type and c.703G→C Best disease, ADVIRC c.704T→C and ADVIRC c.707A→G) run by native gel electrophoresis in the absence (−) or presence (+) of ASF/SF2 or SRp55. The positions of unbound and protein-complexed RNAs are indicated.

Computed analysis of exon 6 autosomal dominant vitreoretinochoroidopathy mutations

To investigate the possibility that the mutations may alter ESE site function, the wild-type and mutated sequences (ADVIRC c.704T→C, ADVIRC c.707A→G, Best disease c.703G→C) were analysed using ESE site prediction web sites using the default settings. RESCUEese ( predicted no ESE sites in any of the exon 6 sequences. In contrast, ESEFinder ( = home)19 predicted a change to SR protein binding for both exon 6 ADVIRC mutations. For c.704T→C there was a predicted increase in the probability of SRp55 and ASF/SF2 binding and for c.707A→G there was a predicted creation of a SC35 binding site (fig 5). PESX (,20 which predicts the presence and/or absence of ESE or ESS sequences suggested that both ADVIRC mutations would abolish an ESE site, but that exon 6 Best disease mutations would not (data not shown).

Figure 5

Prediction by ESEFinder software of serine/arginine-rich (SR) protein binding sites around three exon 6 mutations in BEST1. (A) Alignment of partial exon 6 nucleotide sequences for wild-type, Best disease c.703G→C, ADVIRC c.704T→C and ADVIRC c.707A→G. The altered nucleotides are boxed. (B) Graphical output from ESEFinder showing how the likelihood of binding (y axis) varies with sequence (x axis) for three SR proteins (differently shaded boxes). Wild-type and Best disease sequences are each predicted to have one SR protein binding site whereas the two ADVIRC mutated sequences are predicted to have two or three.

c.704T→C alters splicing in an ESE dependent splice assay

An alteration in ESE activity, resulting from the two ADVIRC-causing exon 6 mutations was predicted by PESX. To test this further, a 22 bp oligonucleotide encompassing the mutated base (corresponding to c.693–c.714) was cloned into an ESE-dependent splice assay (previously described by Woerfel and Bindereif15). In the assay the Dsx ESE site is replaced by the putative ESE site sequences under investigation. The occurrence of in vitro splicing is then determined by RT-PCR.

The wild-type BEST1 exon 6 sequence (lane 1, fig 4D) was confirmed to act as an active ESE site in this assay and promoted splicing of the Dsx gene in vitro. Both exon 6 ADVIRC mutations had altered splicing compared with the wild-type sequence, with the c.704T→C variant weakening and the c.707A→G mutation abolishing ESE site-dependent splicing. In contrast, the Best disease-causing mutation, c.703G→C, strengthened the ESE activity as evidenced by an increased spliced product (lane 4, fig 4D).

ADVIRC mutations affect SR protein binding

Four of the most commonly investigated SR proteins (ASF/SF2, SC35, SRp40, SRp55) were expressed in and subsequently purified from HEK293 cells. They were then tested for their ability to bind 22 bp RNA oligonucleotides that had been designed from the same ESE sequence encompassing the exon 6 mutations (ADVIRC c.704T→C and c.707A→G; Best disease c.703G→C) that had been studied using the Dsx assay. The SC35, SRp40 (data not shown) and SRp55 (fig 4E) proteins all bound to the wild-type and mutated sequences with similar intensities. However, ASF/SF2 had increased binding to the two ADVIRC-mutated sequences compared with the wild-type or Best disease-mutated sequences (fig 4E). This was abolished in competition assays (∼20 000 fold excess unlabelled RNA pre-incubated with protein for 10 min on ice) (data not shown).


This work identifies a fourth mutation in BEST1, c.704T→C, which is associated with ADVIRC. The proband has the typical phenotype seen in other reported cases: developmental anomaly of the anterior segment predisposing to angle closure glaucoma, early adult-onset cataract, and typical fundus appearance of a broad post-oral circumferential band of atrophy and pigmentation. Patient 2 has a milder phenotype but also has the typical peripheral retinal abnormality. ADVIRC may be considered as a developmental defect that is often associated with defects of anterior segment growth. The resulting microphthalmia would explain the common incidence of hyperopia seen with this disease. Such a developmental role for bestrophin-1 may explain the refractive errors (commonly hyperopic) associated with Best disease.

In both patients reported here, the full field ERG found evidence of generalised rod and cone dysfunction. EOG testing was only possible for patient 2, and in contrast to other reported ADVIRC cases, was normal. Previous patients with this rare disorder have shown a much reduced light rise on EOG, which is the usual finding in clinical phenotypes associated with mutations in BEST1.8 21 22 The normal EOG in patient 2 may reflect the very mild phenotype in this patient or may be a feature of disease associated with this mutation. Several patients with BEST1 sequence alterations associated with Best disease have normal EOGs,23 24 and a mouse knockout of Best1 has an enhanced light peak luminance responsiveness with normal amplitude.25 This has led to the suggestion that bestrophin-1 does not solely generate the EOG light peak, but may actually antagonise it, possibly via voltage-dependent Ca2+ channels (VDCCs).25 It has further been suggested that BEST1 sequence alterations that do not alter Cl conductance per se may affect the regulation of the associated VDCCs, which would cause a different phenotype from those sequence alterations that directly affect the Cl channel function.22

In common with our previously described mutations,7 the c.704T→C mutation alters pre-mRNA splicing in an in vitro system, supporting the hypothesis that altered splicing of BEST1 occurs and may be the cause of disease in patients with ADVIRC. ADVIRC is the sole phenotype that has been reported to occur as a result of altered splicing of BEST1. RPE samples from our patients were unavailable to test whether this occurs in vivo.

We have studied two mutations in exon 6 (c.704T→C, c.707T→G) that are associated with ADVIRC and have shown that each affects the activity of an exonic splice regulator site by altering the binding affinity for SR proteins. In both cases, the ADVIRC-causing mutations seem to weaken or abolish the wild-type ESE site activity and bind more strongly to ASF/SF2 compared with the wild-type RNA sequence. Although these assays are sufficient to show altered SR protein binding, it must be noted that they are crude in their approximation of the intranuclear environment and thus it is possible that the same sequence alterations in vivo alter the binding of other splicing factors that influence pre-mRNA processing.

In the ex vivo minigene splice assay, the c.704T→C ADVIRC mutation produced a transcript with a duplicated exon 6. Exonic duplication caused by trans-splicing is a very rare phenomenon that has previously only been detected in a handful of mammalian genes.26 Interestingly, trans-splicing is ESE site dependent.27 However, although we are confident this mutation is affecting splicing, we believe that the exonic duplication probably represents an artefact of the minigene system, which is recognised to produce other splicing artefacts.7 As RNA from RPE of the patients was not available, it was not possible to test this in vivo.

The precise in vivo effects of altered SR protein binding are difficult to predict as SR proteins may also bind ESS sites and can therefore alter the balance between inhibition and enhancement of exonic skipping. An example of this is an ESS site in type 1 bovine papilloma virus (BPV-1), which inhibits splice site selection by sequestering SR proteins required for normal wild-type splicing.28 As ADVIRC-associated mutations in exon 6 are predicted to lead to exon skipping, it is possible that the sequence around this region is part of an overlapping ESE/ESS site known as a CERES (composite exonic regulatory elements of splicing),29 a 7–8 bp sequence that can work either as an ESE or ESS site. If so, this may also explain the enhanced ESE activity of the c.703G→C Best disease sequence and would provide a model that allows for differences in quantitative protein–RNA interactions that affect pre-mRNA splicing. This, in turn, would explain how 3/5 consecutive bases (c.703–707) with four different nucleotide substitutions affect splicing in different ways.

Between c.703 and c.710, four nucleotides are conserved down to the C. elegans level, including both ADVIRC-associated mutations c.704 (the second position thymine in valine 235) and c.707 (the second position adenine in tyrosine 236), as well as c.706 thymine and c.710 cytosine. The conservation of these nucleotides may provide further evidence of a CERES site around nucleotide positions c.704 and c.707. Nucleotides at c.703 and c.710 are mutated in Best disease and in the Dsx splice assay, neither of these changes alters splicing (as measured by ESE site activity) compared with the wild-type sequence (fig 4D). In contrast, when nucleotides are altered between these residues splicing is altered, as is apparent for the ADVIRC mutations at c.704 and c.707. Thus the lack of disruption of splicing seen with Best disease mutations at c.703 and c.710 compared with a disruption of splicing seen with ADVIRC mutations at c.704 and c.707 gives further support to the existence of a CERES site between c.704 and c.709 (fig 3). Systematic mutation of nucleotides around these positions would be required to confirm this hypothesis.

Key points

  • We describe a fourth novel mutation (c.704T→C) in BEST1 that causes autosomal dominant vitreoretinochoroidopathy (ADVIRC).

  • As with other ADVIRC mutations, the c.704T→C mutation altered normal splicing. The mutation and another exon 6 ADVIRC-causing mutation (c.707G→A), either weakened or abolished splicing in an ESE-dependent splice assay when compared with a nearby exon 6 mutation associated with Best disease (c.703G→C). Both ADVIRC sequences, but not the Best disease sequence, had increased binding to the SR protein ASF/SF2 over the wild-type sequence.

  • We conclude that the region encompassing these mutations may form part of a CERES (composite exonic regulatory elements of splicing) site.

For reported cases in which BEST1 mutation causes ADVIRC, we predict an increase in inappropriately spliced mRNA that encodes an internally deleted isoform of bestrophin-1.The three previously reported ADVIRC mutations all led to the loss of whole exons in the α-globin–fibronectin–EDB minigene system. Whether the c.704T→C mutation causes an exon 6 duplication or deletion, it would probably escape nonsense-mediated decay and, in either case, cause an in-frame alteration of bestrophin-1. The functional consequences of such in-frame protein alterations remain undefined. The mutant isoforms may lose important domains, such as trafficking or protein-binding domains, although none has been identified in the exons affected by ADVIRC mutations. Alternatively, as bestrophin-1 exists as a dimmer,30 the ADVIRC mutations may alter the ability of the protein to dimerise. However it is perhaps notable that the ADVIRC phenotype is similar for mutations described in different exons and that all reside within predicted transmembrane domains of bestrophin-1.3 4 It is therefore possible that the putative shortened bestrophin-1 isoforms may not fold or be modified correctly and thus hinder their trafficking to the basolateral membrane of the RPE.


GCMB is a Wellcome Senior Research Fellow in Clinical Science.


View Abstract


  • Funding Wellcome Trust (GR067443MA to RB); National Eye Research Centre (SCIAD051 to AD); EVI-Genoret and Foundation Fighting Blindness USA (to ATM); Manchester Biomedical Research Centre; Manchester Academic Health Sciences Centre (MAHSC) and the NIHR Manchester Biomedical Research Centre.

  • Competing interests None.

  • Provenance and Peer review Not commissioned; externally peer reviewed.