Conceived and designed the experiments: LMEvK NMJ FR AH AGU RCWW PTVMdJ BAO JRV HGL CCWK CMvD. Performed the experiments: LMEvK WDR MKI DDGD JJMW-A RK MAC SFJ NA MS YSA AABB. Analyzed the data: LMEvK WDR CMvD. Contributed reagents/materials/analysis tools: PGH FC TLY CJH SM AWH GWM TLY DAM SMH ADP ACV CB PMcG PJF FT PM JJW TYW SFJ JBtB PTVMdJ AABB FP NW MZ CYM EG UW-L AR. Wrote the paper: LMEvK WDR CMvD. Contributed to the final version of the manuscript: all authors.
¶ Membership of WTCCC2 is provided in the Acknowledgments.
The authors have declared that no competing interests exist.
Intraocular pressure (IOP) is a highly heritable risk factor for primary open-angle glaucoma and is the only target for current glaucoma therapy. The genetic factors which determine IOP are largely unknown. We performed a genome-wide association study for IOP in 11,972 participants from 4 independent population-based studies in The Netherlands. We replicated our findings in 7,482 participants from 4 additional cohorts from the UK, Australia, Canada, and the Wellcome Trust Case-Control Consortium 2/Blue Mountains Eye Study. IOP was significantly associated with rs11656696, located in
Glaucoma is a major eye disease in the elderly and is the second leading cause of blindness worldwide. The numerous familial glaucoma cases, as well as evidence from epidemiological and twin studies, strongly support a genetic component in developing glaucoma. However, it has proven difficult to identify the specific genes involved. Intraocular pressure (IOP) is the major risk factor for glaucoma and the only target for the current glaucoma therapy. IOP has been shown to be highly heritable. We investigated the role of common genetic variants in IOP by performing a genome-wide association study. Discovery analyses in 11,972 participants and subsequent replication analyses in a further 7,482 participants yielded two common genetic variants that were associated with IOP. The first (rs11656696) is located in
Primary open-angle glaucoma (hereafter referred to as glaucoma) is a progressive optic neuropathy responsible for 12.3% of global blindness
Intraocular pressure (IOP) is the major risk factor of glaucoma and existing glaucoma therapies are exclusively aimed at lowering IOP. An elevated IOP (>21 mmHg) influences both the onset and the progression of glaucoma
To identify genetic determinants of IOP, we performed a GWAS in 11,972 participants from 4 independent population-based studies in The Netherlands, and we replicated our findings in 7,482 participants from 4 additional independent cohorts of Caucasian ancestry. We investigated whether the IOP associated SNPs were also related to glaucoma in 1,432 glaucoma cases. Lastly, we examined expression levels of the identified candidate genes in human ocular tissues. We identified common variants in
Genotypic and IOP data were available for 11,972 participants from the Rotterdam Study cohort I (RS-I), RS-II, RS-III, and the Erasmus Rucphen Family (ERF) Study (
Characteristic | RS-I | RS-II | RS-III | ERF |
Participants with valid data (N) | 5,794 | 2,102 | 2,041 | 2,035 |
Age (y), mean ± SD (range) | 68.8±8.9 (55–100) | 64.4±8.0 (55–95) | 55.7±5.8 (45–97) | 48.8±14.4 (18–86) |
Male gender (%) | 41.2 | 45.7 | 43.9 | 43.3 |
IOP (mmHg), mean ± SD (range) | 14.7±3.4 (5–59) | 14.4±3.4 (7–32) | 13.6±3.0 (5–30) | 15.3±3.1 (6–33) |
IOP≥22 mmHg (%) | 3.3 | 3.3 | 1.9 | 1.2 |
Participants with IOP lowering treatment (%) | 2.4 | 3.9 | 1.5 | 0.9 |
Vertical cup-disc ratio, mean ± SD (range) | 0.50±0.14 (0.00–0.89) | 0.50±0.14 (0.05–0.87) | 0.42±0.17 (0.00–1.00) | 0.43±0.16 (0.00–0.83) |
Disc area (mm2), mean ± SD (range) | 2.42±0.48 (0.58–5.44) | 2.32±0.48 (1.06–6.20) | 1.92±0.45 (0.70–7.20) | 1.90±0.35 (1.07–3.95) |
IOP = intraocular pressure; SD = standard deviation; RS = Rotterdam Study; ERF = Erasmus Rucphen Family study.
Four SNPs on chromosome 17p13.1 were significantly associated with IOP in the discovery meta-analysis (p<5×10−8;
SNP | Chrom | Position | MA | MAF | Gene region | #SNPs |
Beta | SE | P-value |
rs11656696 | 17p13.1 | 9974404 | A | 0.43 |
|
4 | −0.26 | 0.05 | 9.8E-09 |
rs7894966 | 10q23.2 | 88608604 | G | 0.04 |
|
8 | 0.67 | 0.13 | 1.6E-07 |
rs216146 | 5q32 | 149426114 | T | 0.39 |
|
2 | 0.22 | 0.05 | 1.4E-06 |
rs2117760 | 3p13 | 70933151 | A | 0.32 |
|
1 | 0.22 | 0.05 | 4.1E-06 |
rs7555523 | 1q24.1 | 163985603 | C | 0.12 |
|
11 | 0.30 | 0.07 | 5.7E-06 |
rs1826598 | 16q23.1 | 76130456 | A | 0.11 |
|
1 | 0.32 | 0.07 | 6.0E-06 |
rs9841621 | 3p24.3 | 18384081 | G | 0.01 |
|
5 | −0.81 | 0.18 | 8.9E-06 |
SNP = single nucleotide polymorphism; Chrom = chromosome; MA(F) = minor allele (frequency); SE = standard error.
number of SNPs with p<10−5 in the region.
According NCBI build 37.1, rs11656696 is located at position 10033679 in the growth-arrest-specific gene
We examined at least 416 KB of the chromosomal regions spanning the known disease genes
Replication of the IOP association was done in 4 additional cohorts from the TwinsUK study (N = 2,235), the Australian Twin study (N = 1,807), the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications study (DCCT/EDIC; N = 1,304), and the Wellcome Trust Case-Control Consortium 2 / Blue Mountains Eye Study (WTCCC2/BMES; N = 2,136) (
Replication analyses | Joint analysis of discovery and replication cohorts | ||||||||||||||||||||||
TWINS-UK | Australian Twins | DCCT/EDIC | WTCCC2/BMES | ||||||||||||||||||||
SNP | MAF | Beta | SE | P-value |
|
MAF | Beta | SE | P-value |
|
MAF | Beta | SE | P-value |
|
MAF | Beta | SE | P-value |
|
Beta | SE | P-value |
rs11656696 | 0.42 | −0.32 | 0.11 | 3.9E-03 | P | 0.42 | −0.11 | 0.10 | 2.9E-01 | G | 0.42 | 0.04 | 0.11 | 6.8E-01 | G | 0.42 | −0.06 | 0.09 | 4.9E-01 | G |
|
|
|
rs7894966 | 0.02 | −1.15 | 0.37 | 1.7E-03 | I | 0.03 | −0.32 | 0.29 | 2.6E-01 | I | 0.04 | −0.11 | 0.29 | 6.9E-01 | I |
|
|
|
|||||
rs216146 | 0.39 | 0.00 | 0.11 | 9.8E-01 | I | 0.37 | −0.08 | 0.11 | 4.8E-01 | I | 0.43 | −0.08 | 0.11 | 4.7E-01 | G | 0.41 | −0.08 | 0.09 | 3.4E-01 | I |
|
|
|
rs2117760 | 0.30 | 0.12 | 0.11 | 2.9E-01 | I | 0.30 | −0.03 | 0.11 | 7.7E-01 | I | 0.33 | 0.05 | 0.11 | 6.5E-01 | I |
|
|
|
|||||
rs7555523 | 0.12 | 0.24 | 0.15 | 9.6E-02 | I | 0.12 | 0.23 | 0.16 | 1.4E-01 | I | 0.11 | 0.18 | 0.17 | 2.9E-01 | G | 0.12 | 0.30 | 0.13 | 1.8E-02 | I |
|
|
|
rs1826598 | 0.11 | 0.15 | 0.15 | 3.4E-01 | P | 0.10 | 0.10 | 0.17 | 5.6E-01 | G | 0.11 | 0.20 | 0.18 | 2.7E-01 | G | 0.12 | 0.01 | 0.13 | 9.3E-01 | G |
|
|
|
rs9841621 | 0.02 | −0.42 | 0.36 | 2.4E-01 | P | 0.02 | −0.34 | 0.33 | 3.0E-01 | I | 0.02 | −0.31 | 0.40 | 4.3E-01 | G | 0.02 | −0.15 | 0.31 | 6.3E-01 | G |
|
|
|
DCCT/EDIC = Diabetes Control and Complications Trial / Epidemiology of Diabetes Interventions and Complications study; WTCCC2/BMES = Wellcome Trust Case-Control Consortium 2 / Blue Mountains Eye Study; SNP = single nucleotide polymorphism; MAF = minor allele frequency; SE = standard error;
Column indicates whether the SNP has been genotyped (G), imputed (I), or partly (P) genotyped, i.e. genotyped in 2/3 of the participants.
We investigated the associations of the
In a first study of expression levels in human ocular tissues, we observed moderate to high expression of
Quantitative determination of
Gene | CB-PE | CB-NPE | Choroid | RPE | Photoreceptors | TM |
|
55 (1.3) | 57 (2.0) | 73 (5.1) | 76 (1.7) | 78 (8.6) | 78 (3.1) |
|
93 (1.5) | 93 (1.0) | 86 (2.5) | 88 (1.9) | 88 (2.4) | 88 (1.5) |
The two genes are ranked by increasing expression, calculated by the mean percentiles (SD) of the expression levels. Gene expression of CB-PE and CB-NPE (n = 4), choroid (n = 3), photoreceptors (n = 3) and RPE (n = 6) were performed on Agilent Human 44k microarray of post-mortem donor eyes without glaucoma or any other ocular diseases.
Data from Liton et al., performed on Affymetrix Human U133 microarray, showing mean percentiles (SD) of human gene expression levels in TM tissue from 3 healthy eyes
CB-PE = ciliary body, pigmented epithelium; CB-NPE = ciliary body, non-pigmented epithelium; RPE = retinal pigment epithelium; TM = trabecular meshwork.
We identified rs11656696 in
Ingenuity diagram of biochemical and functional interactions between the newly identified GAS7 and TMCO1 disease genes implicated in elevated IOP and glaucoma, and previously known glaucoma disease genes (WDR36, MYOC, OPTN, CAV1). Functional relationships in the knowledge database Ingenuity (
The second variant that we found to be associated with IOP and glaucoma was rs7555523 in
No previous GWASs of IOP have been conducted to date. When comparing our findings to those of association studies of glaucoma, we found an overlap with 3 regions. First, we replicated the association with the
Our study design had three potential limitations. First, we did not measure central corneal thickness (CCT) in the majority of the participants of the discovery cohorts. CCT is an important determinant of IOP measurements and may be an IOP-independent risk factor for glaucoma
Second, in the gene discovery analyses, the initial IOP levels were not known for the participants who received IOP lowering medication or who had a history of IOP lowering surgery. We imputed these IOPs, because (particularly in the elderly population of RS-I) participants with extreme IOPs, which are likely to be genetically determined, are otherwise excluded. Similar approaches have been applied to research of blood pressure, where an analogous problem occurs: those with the higher blood pressures are otherwise excluded
Third, some replication cohorts differed from the discovery cohorts with respect to their age, sex or disease status. The participants of the Australian Twin study and DCCT/EDIC were evidently younger than the participants of the other cohorts were. Aging has previously been associated with an increase in the accumulation of extracellular material in the trabecular meshwork, as well as a decrease in trabecular meshwork cells
In conclusion, this genome-wide association study in 8 independent Caucasian cohorts identified rs11656696 in
All participating studies adhered to the tenets of the Declaration of Helsinki and were approved by their Medical Ethics Committees. Written, informed consent was obtained from all participants.
For the gene discovery phase, we combined data of 11,972 participants derived from 4 large, independent population-based cohort studies in The Netherlands: the Rotterdam Study cohort I (RS-I), RS-II, RS-III, and the Erasmus Rucphen Family (ERF) Study. Replication of the findings was sought in 4 independent populations: the TwinsUK Adult Twin study, the Australian Twin Study, the Diabetes Control and Complications Trial / Epidemiology of Diabetes Interventions and Complications study (DCCT/EDIC)
The RS-I is a prospective population-based cohort study of 7,983 residents 55 years of age and older living in Ommoord, a suburb of Rotterdam, The Netherlands
In all discovery cohorts, the IOP was measured with Goldmann applanation tonometry (Haag-Streit, Bern, Switzerland), which is the international standard for IOP assessment in ophthalmic research and clinical practice. A drop of fluorescein sodium was instilled in each eye. The tonometer was set at 10 mm Hg, and the prism was carefully applied to the corneal surface of the right eye. Without looking at the scale, the examiner rotated the dial until the inner margins of the two semicircles touched each other. The examiner then moved the slit lamp away from the eye and read the IOP. The tonometer was set at 10 mm Hg, and the measurement was repeated. If the two measurements differed, a third measurement was performed, and the median value was recorded. The procedure was repeated for the left eye
In the RS-I, RS-II and RS-III cohorts, DNA was genotyped with the Illumina Infinium II HumanHap550 chip v3.0 array. In the ERF study, DNA was genotyped on 4 different platforms (Illumina 6k, Illumina 318K, Illumina 370K and Affymetrix 250K), which were then merged. Genotype data were imputed by using HapMap CEU build 35 as the reference population, resulting in over 2.5 million SNPs. For details please see
SNPs showing strongest association in the discovery phase were carried forward and assessed for association with IOP in 2,235 participants from the TwinsUK Study, 1,807 from the Australian Twin Study, 1,304 from the DCCT/EDIC Study, and 2,136 from the WTCCC2/BMES Study. The TwinsUK , Australian Twin and WTCCC2/BMES were also population-based studies, and participants were ascertained regardless of their phenotypes or clinical status. The DCCT/EDIC study comprised only patients with type 1 diabetes included in a preventive trial. Descriptions of the study populations, clinical examinations, and genotyping methods of the replication cohorts are provided in
SNPs showing the strongest associations in the discovery and replication phase were also evaluated in 4 series of glaucoma patients. The first series included 188 participants from RS-I in whom the technician measuring IOP was completely ignorant of the presence of glaucoma. Controls were healthy participants of RS-I. The second case-control study was an independent series of 104 glaucoma cases from an isolated population (Genetic Research in an Isolated Population [GRIP] study), with the ERF population as a control group. The third study included 152 cases and 141 controls recruited from all over The Netherlands as part of the Amsterdam Glaucoma Study (AGS). The last case-control study comprised a series of 988 glaucoma cases and 378 controls ascertained in Erlangen and Tübingen, Germany. Details of the clinical evaluation and glaucoma diagnosis in these studies are described in
Analyses were performed for the mean IOP of both eyes or for one eye if data on the other eye were missing. In the gene discovery analyses, IOP levels were imputed for those who received IOP lowering medication or had a history of IOP lowering surgery, because the initial IOP levels were unknown. Based on a reported average of a 30% IOP reduction caused by IOP lowering medication, estimated in a meta-analysis, IOP values of those receiving this medication were divided by 0.7 to estimate pre-treatment IOP
Associations between IOP and genome-wide loci were assessed with linear regression models under the assumption of an additive model for the effect of the risk allele. Analyses were adjusted for age and sex. In the ERF study, the analyses were also performed with additional adjustment for the time of the IOP measurement. Genomic inflation factors (λ) were calculated to evaluate any population stratification. Analyses were performed with the ProbABEL package from the ABEL set of programs (
The results from the 4 cohorts were subjected to an inverse variance meta-analysis. Genomic control was used to correct the standard errors of the effect estimates before pooling
Results of the discovery meta-analysis were also used to explore regions in the immediate vicinity of the known glaucoma genes (
Loci which were suggestive (p<1×10−5) of association with IOP in the discovery meta-analysis were taken forward to the replication phase. If two or more significantly associated SNPs within a locus were in linkage disequilibrium (LD), only the SNP with the best probability of association (lowest p-value) was selected. Linear regression analyses adjusted for age and sex were performed under the assumption of an additive effect of the risk allele. The results from the discovery and replication cohorts were combined by using an inverse variance meta-analysis (METAL software).
SNPs that were genome-wide significantly associated with IOP in the meta-analysis of the discovery and replication cohorts were assessed in the 4 glaucoma case-control studies. Logistic regression analyses adjusted for age and sex were performed (SPSS version 15.0 for Windows; SPSS, Chicago, IL) and a pooled effect estimate was calculated (Rmeta software [
Protein pathway analysis was conducted in Ingenuity Knowledge Base (Ingenuity Systems,
Two independent expression studies were performed. In the first, retinal expression data were obtained essentially as described by Booij and colleagues
In the second expression study, ocular tissues were obtained for quantitative real-time PCR from four donor eyes (age: 81.2±4.5 years, 2 female, 2 male) without any known ocular disease. These eyes were obtained at autopsy and were processed within 8 hours after death. Informed consent to tissue donation was obtained from the donors or their relatives, and the protocol of the study was approved by the local Ethics Committee and adhered to the tenets of the Declaration of Helsinki for experiments involving human tissue. Total RNA was extracted from various ocular tissues by using the RNeasy kit (Quiagen, Hilden, Germany) including an on-column DNase I digestion step. First strand cDNA synthesis was performed by using 0.1 µg of total RNA, 200 U Superscript II reverse transcriptase (Invitrogen; Karlsruhe, Germany), and 500 ng oligo dT primers (Roche Diagnostics; Mannheim, Germany) in a 20 µl reaction volume. Quantitative real-time PCR was performed by means of the MyIQ thermal cycler and software (Biorad, Munich, Germany). PCR reactions (25 µl) contained 2 µl of first-strand cDNA, 0.4 µM each of upstream- and downstream-primer, 3.0 mM MgCl2, and 1× SsoFast EvaGreen Supermix (Biorad). All samples were analyzed in duplicates by means of a program with an initial denaturation step of 95°C for 3 minutes and 40 cycles of 95°C for 5 seconds, and 64°C (GAS7 and TMCO1) or 62°C (GAPDH) for 15 seconds. Gene-specific primers (Eurofins, Anzing, Germany) were designed to anneal with sequences located in different exons by means of Primer 3 software (
QQ-plots for the observed versus expected p-values for the individual discovery cohorts and the discovery meta-analysis.
(DOC)
Regional association plots of loci associated with IOP (5×10−8<p-value<1×10−5) in meta-analysis.
(DOC)
Regional association plots of
(DOC)
Loci associated with IOP with p-values<10−5 after meta-analyses: results of individual cohorts. SNP = single nucleotide polymorphism; Chrom = Chromosome; MAF = minor allele frequency; SE = standard error; RS = Rotterdam Study; ERF = Erasmus Rucphen Family study.
(DOC)
All SNPs associated with IOP with p-values<10−5 after meta-analyses. MA(F) = minor allele (frequency).
(DOC)
Association results for SNPs identified in previous association studies. SNP = single nucleotide polymorphism; Chrom = Chromosome; SE = standard error; RS = Rotterdam Study; ERF = Erasmus Rucphen Family study.
(DOC)
Characteristics of the replication cohorts. * not measured. **available for subset of 843 TwinsUK participants only, mean age 56 years. IOP = intraocular pressure; SD = standard deviation; DCCT/EDIC = Diabetes Control and Complications Trial / Epidemiology of Diabetes Interventions and Complications study; WTCCC/BMES = Wellcome Trust Case-Control Consortium / Blue Mountains Eye Study
(DOC)
Characteristics of the glaucoma case-control studies. * not measured. IOP = intraocular pressure; SD = standard deviation; RS = Rotterdam Study; GRIP = Genetic Research in Isolated Populations; AGS = Amsterdam Glaucoma Study.
(DOC)
PCR-primers used for the expression study. Tan, annealing temperature.
(DOC)
(DOC)
The authors thank Ada Hooghart, Corine Brussee, Riet Bernaerts-Biskop, Patricia van Hilten, and Lidian van Amsterdam for the ophthalmic data collection; Pascal Arp, Mila Jhamai, Dr. Michael Moorhouse, Jeannette Vergeer, Marijn Verkerk, and Sander Bervoets for their help in creating the GWAS database. The authors are grateful to the study participants, the staff from the Rotterdam and ERF Studies and the participating general practitioners and pharmacists.
The Australian Twin Study authors are grateful to Dr Camilla Day and staff for their help in genotyping. The Australian genotyping data were generated and processed by Grant W. Montgomery, Nicholas G. Martin, Scott D. Gordon, Dale R. Nyholt, Sarah E. Medland, Brian P. McEvoy, Margaret J. Wright, Anjali K. Henders, Megan J. Campbell. The Australian Twin Study authors additionally like to thank Jane MacKinnon, Shayne Brown, Lisa Kearns, Jonathan Ruddle, Paul Sanfilippo, Sandra Staffieri, Olivia Bigault, Colleen Wilkinson, Jamie Craig, Yaling Ma, and Julie Barbour for assisting with clinical examinations.
Peter Donnelly (Chair)1,2, Ines Barroso (Deputy Chair)3, Jenefer M Blackwell4, 5, Elvira Bramon6 , Matthew A Brown7 , Juan P Casas8 , Aiden Corvin9, Panos Deloukas3, Audrey Duncanson10, Janusz Jankowski11, Hugh S Markus12, Christopher G Mathew13, Colin NA Palmer14, Robert Plomin15, Anna Rautanen1, Stephen J Sawcer16, Richard C Trembath13, Ananth C Viswanathan17, Nicholas W Wood18
Chris C A Spencer1, Gavin Band1, Céline Bellenguez1, Colin Freeman1, Garrett Hellenthal1, Eleni Giannoulatou1, Matti Pirinen1, Richard Pearson1, Amy Strange1, Zhan Su1, Damjan Vukcevic1, Peter Donnelly1,2
Cordelia Langford3, Sarah E Hunt3, Sarah Edkins3, Rhian Gwilliam3, Hannah Blackburn3, Suzannah J Bumpstead3, Serge Dronov3, Matthew Gillman3, Emma Gray3, Naomi Hammond3, Alagurevathi Jayakumar3, Owen T McCann3, Jennifer Liddle3, Simon C Potter3, Radhi Ravindrarajah3, Michelle Ricketts3, Matthew Waller3, Paul Weston3, Sara Widaa3, Pamela Whittaker3, Ines Barroso3, Panos Deloukas3.
Christopher G Mathew (Chair)13, Jenefer M Blackwell4,5, Matthew A Brown7, Aiden Corvin9, Mark I McCarthy19, Chris C A Spencer1
1 Wellcome Trust Centre for Human Genetics, Roosevelt Drive, Oxford OX3 7BN, UK; 2 Dept Statistics, University of Oxford, Oxford OX1 3TG, UK; 3 Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge CB10 1SA, UK; 4 Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia, 100 Roberts Road, Subiaco, Western Australia 6008; 5 Cambridge Institute for Medical Research, University of Cambridge School of Clinical Medicine, Cambridge CB2 0XY, UK; 6 Department of Psychosis Studies, NIHR Biomedical Research Centre for Mental Health at the Institute of Psychiatry, King's College London and The South London and Maudsley NHS Foundation Trust, Denmark Hill, London SE5 8AF, UK; 7 Diamantina Institute of Cancer, Immunology and Metabolic Medicine, Princess Alexandra Hospital, University of Queensland, Brisbane, Queensland, Australia; 8 Dept Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London WC1E 7HT and Dept Epidemiology and Public Health, University College London WC1E 6BT, UK; 9 Neuropsychiatric Genetics Research Group, Institute of Molecular Medicine, Trinity College Dublin, Dublin 2, Eire; 10 Molecular and Physiological Sciences, The Wellcome Trust, London NW1 2BE; 11 Centre for Digestive Diseases, Queen Mary University of London, London E1 2AD, UK and Digestive Diseases Centre, Leicester Royal Infirmary, Leicester LE7 7HH, UK and Department of Clinical Pharmacology, Old Road Campus, University of Oxford, Oxford OX3 7DQ, UK; 12 Clinical Neurosciences, St George's University of London, London SW17 0RE; 13 King's College London Dept Medical and Molecular Genetics, School of Medicine, Guy's Hospital, London SE1 9RT, UK; 14 Biomedical Research Centre, Ninewells Hospital and Medical School, Dundee DD1 9SY, UK; 15 King's College London Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, Denmark Hill, London SE5 8AF, UK; 16 University of Cambridge Dept Clinical Neurosciences, Addenbrooke's Hospital, Cambridge CB2 0QQ, UK; 17 NIHR Biomedical Research Centre for Ophthalmology, Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology, London EC1V 2PD, UK; 18 Dept Molecular Neuroscience, Institute of Neurology, Queen Square, London WC1N 3BG, UK; 19 Oxford Centre for Diabetes, Endocrinology and Metabolism (ICDEM), Churchill Hospital, Oxford OX3 7LJ, UK.