Identification of a Novel Locus for Autosomal Dominant Primary Open Angle Glaucoma on 4q35.1-q35.2

Abstract

Purpose.: Primary open angle glaucoma is the most prevalent type of glaucoma and the leading cause of irreversible blindness worldwide. The genetic basis is poorly understood. Of 14 loci associated with this disease, only two genes have been identified, accounting for approximately 4% of cases. The authors investigated the genetic cause of primary open angle glaucoma in a large four-generation family with an apparent autosomal dominant mode of inheritance.

Methods.: Twenty-three family members underwent comprehensive phenotyping by a single ophthalmologist, and the MYOC gene was sequenced in all affected family members for whom DNA was available. Parametric genomewide linkage analysis was performed on 10 affected family members and one unaffected family member. Within the critical region, mutation analysis of candidate genes LRP2BP, CYP4V2, and UFSP2 was carried out by direct sequencing.

Results.: No mutations were identified in MYOC. Genomewide linkage analysis generated one significant LOD score of 3.1 (maximum affected-only LOD score of 2.8) centered on chromosome 4 at 4q35.1-q35.2, a critical region that does not contain any of the previously reported primary open angle glaucoma loci. A 1.866-Mb (7.2 cM) region was identified containing 17 known or hypothetical genes. No mutations were identified in the candidate genes LRPB2BP, CYP4V2, and UFSP2.

Conclusions.: This study identifies a new primary open angle glaucoma locus, GLC1Q, in a region on chromosome 4 not previously associated with glaucoma.

Glaucoma is characterized by progressive loss of retinal ganglion cells resulting in a loss of peripheral vision and a characteristic excavative atrophy of the optic nerve, resulting in blindness if untreated. ^1,2 Primary open angle glaucoma (POAG) is characterized by onset during adulthood, an open angle of normal appearance, and glaucomatous optic nerve head damage in the absence of an identifiable secondary cause. ² Increased intraocular pressure (IOP) secondary to reduced aqueous outflow through the filtration angle is the most important risk factor for the development of POAG and is frequently, although not invariably, associated with the disease. ^{3 –5} Increased IOP is the only modifiable factor. ^{6 –8} Other risk factors for POAG include African ethnicity, positive family history, increasing age (older than 40 years), thin central corneal thickness, myopia, and low diastolic perfusion pressure. ^5,9,10 Glaucoma is the leading cause of irreversible blindness worldwide, ¹¹ with POAG the most prevalent form and the second leading cause of blindness in the United States and Europe. ^4,12 On average, POAG affects 2% of people older than 40 years with an incidence that increases with age, reaching a prevalence of 6% in white populations, 16% in black populations, and 3% in Asian populations in persons older than 70. ¹³ In England and Wales, POAG is responsible for 10% of the certification for visual impairment and 11% of the certification for severe visual impairment, and it accounts for more than 1 million hospital visits per year in the National Health Service of England and Wales. ^14,15 Approximately 50% of POAG patients have a positive family history, and first-degree relatives of an affected person are at threefold to ninefold increased risk for the disease. ^1,5,16 Pedigrees with familial POAG displaying an autosomal dominant pattern of inheritance with incomplete penetrance and variable expressivity have been described. ¹⁷ The variable and complex phenotype suggests POAG has a multifactorial etiology and is likely to involve the interaction of one or more genes with environmental factors. ¹⁸ Sixteen loci have been reported for POAG (GLC1A-P), ¹⁹ although not all have been replicated and only two causal genes—myocilin MYOC (GLC1A) and optineurin OPTN (GLC1E)—have been identified. ³ Sequence variants in WD-repeat domain 36 WDR36 (GLC1G) have been reported as a cause of POAG ^20,21 but have been excluded by others, ^22,23 and the locus remains controversial. ²⁴ Similarly, sequence variants in NTF4 (GLC1O) have been reported ^25,26 and excluded ^27,28 as a cause of POAG. Mutations in MYOC are the most common cause of POAG but still only account for 4% of adult-onset cases and 6% to 36% of juvenile-onset cases. ^29,30 Mutations in OPTN are a rare cause of POAG and probably account for <1% of cases. ¹⁸  

In this report, we describe a genomewide linkage scan in a large four-generation British family with clinically diagnosed POAG and an apparent autosomal dominant mode of inheritance. We identified a novel locus, GLC1Q, for adult-onset POAG on chromosome 4q35.1–4q35.2. 

Subjects and Methods

Diagnosis and Inclusion

A large four-generation family with clinically diagnosed POAG in three generations was seen at the glaucoma clinic of the Manchester Royal Eye Hospital (Manchester, UK) (Fig. 1). Ethical approval and informed consent was obtained from all study participants. The research adhered to the tenets of the Declaration of Helsinki. 

Figure 1.

 

POAG pedigree showing an autosomal dominant pattern of inheritance. *Family members reexamined to confirm the diagnosis of POAG and used in the linkage analysis. †Family members with traumatic glaucoma. ‖Family members with anomalous discs. #Family members with subacute angle closure glaucoma. ‡Family members with end-stage POAG and registered as blind. §The most mildly affected family members.

Twenty-three family members, including all living family members with an established diagnosis of POAG, were examined by a single ophthalmologist. Existing medical records were reviewed with the patient's written consent. Detailed ophthalmic, systemic, family, and drug history was obtained, and participants were specifically asked about symptoms such as migraine, Raynaud's phenomenon, and vascular risk factors such as hypertension. Detailed ophthalmic examinations included both clinical parameters and visual field testing. Clinical examination included anterior segment examination by slit lamp biomicroscopy with evaluation of anterior chamber depth using the Von Herrick method, gonioscopy to evaluate angle anatomy, applanation tonometry with a Goldmann applanation tonometer (Haag Streit AG, Bern, Switzerland) to evaluate IOP, measurement of corneal thickness using pachymetry, and dilated stereoscopic examination of the optic discs. 

Discs were classified as either normal, abnormal (vertical cup-to-disc ratio >0.5), or glaucomatous. Glaucomatous optic neuropathy was defined as focal loss of the nerve fiber layer resulting in a notch or neuroretinal rim thinning in the inferior, nasal, and superior aspects of the disc resulting in a cup-to-disc ratio of 0.7. ²⁹ Serial visual fields (minimum of three) were obtained for all study participants using standard Humphrey 24–2 automated perimetry (Carl Zeiss Meditec, Dublin, CA). Results of previous visual field tests were reviewed from medical records in patients with very advanced visual loss who were unable to perform visual field testing in the context of the study. 

For linkage analysis, family members were labeled as affected, unaffected, or unknown. Affected members had a diagnosis of POAG based on the presence of elevated IOP at the time of diagnosis (>21 mm Hg), age of onset >40 years, and glaucomatous optic neuropathy with corresponding visual field loss in at least one eye on at least two consecutive visual field tests (paracentral defects, nasal steps on Humphrey Glaucoma Hemifield test, arcuate-shaped defects). ³¹ Unaffected members lacked glaucomatous symptoms at age older than 70 years, and members younger than 70 years with no evidence of glaucoma were labeled as unknown because POAG is a late-onset disease. 

IOP was used as an inclusion criterion for POAG in this study to exclude family members with normal tension glaucoma. Patients with a history of ocular trauma in the affected eye, narrow angles, anomalous or colobomatous discs, or other factors rendering the diagnosis of POAG difficult or suggestive of secondary glaucoma were excluded from the study. Anomalous discs tended to be large, with an irregular scleral canal shape and abnormal vessel branching patterns more typical of colobomatous discs than glaucomatous discs. In addition, the visual field test defects of the discs classified as anomalous were incongruent with disc appearance. Photographs and visual field tests, where available, were reviewed (with no other data) by a second ophthalmologist, a glaucoma subspecialist, to confirm the diagnosis. 

For each participating family member, DNA was extracted from peripheral whole blood using a purification kit (QIAampDNA Blood Maxi Kit; Qiagen, Crawley, UK) according to the manufacturer's protocol. 

Sequencing and PCR

Mutation analysis of the coding region of MYOC was carried out in all affected family members for whom DNA was available and one unaffected family member. Sequencing of candidate genes LRP2BP and CYP4V2 was performed on eight affected family members and one unaffected family member, and sequencing of UFSP2 was performed on four affected family members and one unaffected, unrelated control. 

PCR amplicons were designed to include the exonic and approximately 40 bp intronic sequence and were generated from genomic DNA. Large exons were amplified as overlapping fragments (primer sequences and PCR conditions available on request). Amplicons were purified before sequencing using antibody purification and spin columns (Montage; Millipore, Watford, UK). 

Sequencing was performed using terminator chemistry (Big Dye 3.1; Applied Biosystems, Warrington, UK), and reactions were run on a DNA sequencer (ABI 3730; Applied Biosystems). Sequence analysis was performed using Staden software (freeware). 

Linkage Analysis and Haplotyping

Linkage analysis was performed on nine affected family members and one unaffected family member (Fig. 1) using an Affymetrix (Santa Clara, CA) 250K single-nucleotide polymorphism (SNP) array. 

Parametric genomewide multipoint linkage analysis was performed using MERLIN software and a dominant model. ³² A disease allele frequency of 0.0001 was calculated using statistical genetic analysis software (GeneHunter; Ward Systems Group, Inc., Frederick, MD) with the penetrance value set at 0.95 and phenocopy rate of 0.05. ³³  

Significant and suggestive linkage was defined according to the criteria proposed by Lander and Kruglyak. ³⁴ Thresholds of P values in the range of 10⁻³ to 5 × 10⁻⁴ (LOD 1.9–2.4) were used for suggestive linkages. A LOD score of 3 or above was used to determine regions of significant linkage for further study. 

Haplotype analysis was performed using physical map distances provided by the National Center for Biotechnology Information (NCBI) and the genetic map from the Marshfield Centre. ^35,36 Haplotypes were generated using MERLIN and were visualized using HaploPainter. ³⁷ Chromosome positions were based on the March 2006 human reference sequence (NCBI Build 36.1/hg18). 

Results

A clinical diagnosis of POAG was made in three generations of a large four-generation family of British descent with a clear autosomal dominant inheritance pattern (Table 1). The average age of onset of disease, where recorded, was 51 years (range, 40–68 years). All living members were re-examined to confirm the initial diagnosis of POAG. This excluded three members who had traumatic glaucoma, anomalous discs (vertical cup-to-disc ratio, 0.5–0.7), or subacute angle closure glaucoma (III.9, III.17, and III.25, respectively, in Fig. 1). Older members of the family had progressed to end-stage disease (grade 4, gross optic disc cupping and small residual field of vision) and were registered as blind (Fig. 1). The most mildly affected family members had grade 2 POAG (thinning of the neural retina, arcuate scotoma, mean deviation >12 dB) (Fig. 1). ³¹ All affected family members had normal central corneal thicknesses where measured, excluding the possibility that thin central corneas were causing an underestimate of the true IOP. ¹⁰ Causative mutations in MYOC were excluded as the cause of POAG in this family by sequencing it in the eight affected family members for whom DNA was available (Fig. 1; II.9, II.11, III.5, III.6, III.10, III.11, III.18, III.20) and one unaffected family member (Fig. 1; II.2). 

Table 1.

 

Clinical Features, Where Known, of Affected Family Members Used in Linkage Analysis

Family Member (age at diagnosis, y)	Glaucomatous Damage Grade*	Visual Field Defect		Highest IOP or IOP at Diagnosis (mm Hg)	Disc Appearance†		Shaffer Grading of Angle	CCT (μm)	Medication‡	Surgery
		RE	LE		RE	LE
II.11 (54)	4	MD −27.71 dB, tiny residual field (<5° around fixation) remaining	MD −29.03 dB, tiny residual field (<5° around fixation) remaining	34 RE 38 LE	CD ratio 1.0, cup size medium, NRR thinning 360°	CD ratio 1.0, cup size medium, NRR thinning 360°	2			Trabeculectomy BE
II.12	4	MD −25.88 dB, tiny residual field (<5° around fixation) remaining	MD −27.68 dB, tiny residual field (<5° around fixation) remaining	32 BE	CD ratio 1.0, cup size medium, NRR thinning 360°	CD ratio 1.0, cup size medium, NRR thinning 360°	4			Trabeculectomy BE
III.2 (65)	1	MD −0.71 dB, paracentral scotoma and inferior arcuate	MD −0.21 dB, nil	28 RE 26 LE	CD ratio 0.85, cup size medium, lamina cribrosa visible, superior NRR thinning++	CD ratio 0.75, cup size medium	4	503 RE 517 LE	Timoptol BE, Travatan nocte BE	Nil
III.5 (60)	3	MD −12 dB, superior arcuate defect involving central 5°	MD −7.2 dB, superior arcuate defect	36 BE	CD ratio 0.85, cup size medium, NRR complete loss inferior rim, NFL layer diffuse thinning	CD ratio 0.75, cup size medium, NRR thinning inferiorly	2	552 RE 540 LE	Trusopt and Xalatan	Trabeculectomy BE
III.6 (67)	2	MD −6.42 dB, superior arcuate	MD −4.29 dB, superior paracentral scotoma	25 RE 20 LE	CD ratio 0.75, cup size medium, NRR thinning inferiorly	CD ratio 0.65, cup size medium, NRR thinning inferiorly	4		Xalatan	Nil
III.10 (55)	2	MD −4.93 dB, nasal step	MD −8.04 dB, inferior and superior arcuate defects	28 BE	CD ratio 0.9, cup size medium, pale	CD ratio 0.95, cup size medium, RAPD	4		Xalatan BE and Timoptol BE	Nil
III.11 (52)	2	MD −6.78 dB, paracentral scotoma and superior arcuate	MD −4.72 dB, Nasal step	36 RE 32 LE	CD ratio 0.85, cup size medium, inferior NRR loss and PPA	CD ratio 0.8, cup size medium, inferior and temporal NRR thinning and PPA	4	563 RE 548 LE	Xalatan BE	Trabeculectomy RE
III.18 (51)	3	MD −8.42 dB, superior arcuate with loss within central 5°	MD −6.04 dB, superior arcuate and inferior nasal step	26 RE 25 LE	CD ratio 0.9, cup size large, NRR thinning 360° and PPA, superior notch	CD ratio 0.6, cup size medium, NRR thinning inferotemporally	3		Azopt and Xalatan	Trabeculectomy BE
III.20 (40)	1	MD −3.72 dB, inferior nasal step	MD −1.75 dB, no focal loss		CD ratio 0.75, cup size large, NRR thinning superiorly	CD ratio 0.7, cup size large, wedge defect NFL, NRR thinning superiorly	3	572 RE 578 LE	Xalatan BE	Nil

 BE, both eyes; LE, left eye; MD, mean deviation; NFL, nerve fiber layer; NRR, neuroretinal rim; PPA, peripapillary atrophy; RAPD, relative afferent pupillary defect; RE, right eye.

*  Grading of glaucomatous damage: grade 1 (mild), characterized by minimal cupping, a nasal step paracentral scotomas, and a MD less than −6 dB; grade 2 (moderate), characterized by thinning of the neuroretinal rim, an arcuate scotoma, and a MD less than −12 dB; grade 3 (severe), characterized by marked cupping, extensive visual field loss including defects within the central 5°, and a MD greater than −12dB; grade 4 (end-stage), characterized by gross cupping and a small residual field.

†  CD ratio: vertical cup-to-disc ratio; cup size: small, 1.2–1.5 mm; medium, 1.6–1.9 mm; large, 2.0–2.3 mm.

‡  Timoptol: Merck Sharp & Dohme Limited, Hoddesdon, Hertfordshire, UK; Travatan nocte: Alcon (UK) Ltd. Hemel Hempstead, Hertfordshire, UK; Trusopt: Merck Sharp & Dohme Ltd.; Xalatan: Pfizer Ltd., Tadworth, Surrey, UK; Azopt: Alcon.

Genomewide Scan

A 1.866-Mb (7.2-cM) region with a maximum LOD score of 3.1 (maximum affected-only LOD score, 2.8) was identified on chromosome 4q35.1-q35.2 between bases 186, 236, 852 and 188, 102, 579 (Fig. 2). The nonparametric linkage LOD score was 2.1 (Z-score 27.1; delta 4.79). Haplotype analysis revealed that all affected family members shared a common haplotype between rs13104825 and rs1425963 (Fig. 3). No significant linkage (LOD >3.0) was detected at any other loci. The critical region contains 17 genes (Table 2). 

Figure 2.

 

Parametric multipoint genomewide LOD scores for chromosome 4. The physical chromosome position is shown on the x-axis, and the LOD score is shown on the y-axis. A 1.866-MB region was identified with an LOD score of 3.1 on chromosome 4 between bases 186, 236, 852 and 188, 102, 579.

Figure 3.

 

Haplotype analysis on chromosome 4 for the critical POAG-linked locus identified in the POAG family. The haplotype is shown for each family member for 19 markers between rs2046813 and rs7699015, as shown on the left, together with details for the allele represented by the number 1 or 2 in the haplotype. Family members are numbered as in Figure 1. The haplotype has been left blank where the data are missing. Black: inferred disease haplotype with recombinations of parental chromosomes shown in family member III.5, refining the critical region between the markers rs13104825 and rs1425963 (physical position 186, 236, 852 and 188, 102, 579).

Table 2.

 

The 17 Genes within the Disease Locus Identified in the POAG Family

Gene	Chromosome 4 Coordinates
SLC25A4	186, 301, 411–186, 308, 532
KIAA1430	186, 317, 813–186, 362, 176
SNX25	186, 368, 278–186, 522, 114
LRP2BP	186, 285, 032–186, 300, 152
ANKRD37	186, 317, 840–186, 321, 390
UFSP2	186, 321, 443–186, 347, 068
C4orf47	186, 350, 544–186, 370, 822
CCDC110	186, 366, 336–186, 392, 913
PDLIM3	186, 422, 854–186, 456, 712
SORBS2	186, 506, 598–186, 877, 806
TLR3	186, 990, 309–197, 006, 250
FAM149A	187, 026, 231–187, 093, 816
CYP4V2	187, 112, 674–187, 134, 616
KLKB1	187, 147, 592–187, 179, 625
F11	187, 187, 118–187, 210, 835
MTNRIA	187, 454, 814–187, 476, 721
FAT1	187, 508, 938–187, 644, 987

The genes LRP2BP, UFSP2, and CYP4V2 were considered the best candidates in the region based on their known function and ocular expression. Sequencing of the coding region of LRP2BP and CYP4V2 in eight affected family members (Fig. 1; II.9, II.11, II.12, III.2, III.5, III.6, III.10, III.18) and one unaffected family member (Fig. 1; II.5) revealed no missense or nonsense sequence changes that might have been putatively pathogenic alterations. Sequencing of the coding region of UFSP2 in three affected family members (Fig. 1; II.11, II.12, III.18) and one unaffected unrelated control also revealed no putative pathogenic alterations. New and previously reported SNPs are listed in Table 3. 

Table 3.

 

SNPs Identified in Affected, Unaffected, and Unrelated Unaffected Persons for the Three Genes Sequenced in the Critical Region

Gene	cDNA Change	Predicted Protein	Heterozygous			Homozygous			dbSNP Ref. No.
			Affected	Unaffected	Control	Affected	Unaffected	Control
CYP4V2	c.235G>C	p.L22V	0/6	0/1	0/1	6/6	1/1	1/1	rs1055138
	c.555A>T	p.A185A	2/6	0/1	0/1	3/6	0/1	0/1	rs11932764
	c.810T>G	p.A270A	1/6	0/1	0/1	0/6	0/1	0/1	rs3736455
	c.823G>A	p.E275K	1/6	0/1	0/1	0/6	0/1	0/1	rs34745240
LRP2BP	c.81T>C	p.F27F	1/6	0/1	0/1	0/6	0/1	0/1	rs2030802
UFSP2	c.48G>A	p.V88A	0/1	0/0	0/1	1/1	0/0	1/1	rs10866275
	c.66C>G	p.G82A	0/1	0/0	0/1	1/1	0/0	1/1	rs41278591
	c.85C>T	p.A76T	0/1	0/0	0/1	1/1	0/0	1/1	rs2289720

Discussion

Here we have described and comprehensively phenotyped 23 members of a large four-generation family with POAG with an apparent autosomal dominant mode of inheritance. After excluding mutations in the MYOC gene, we investigated the genetic cause of POAG in this family by parametric genomewide linkage analysis and identified a novel genetic locus, GLC1Q, for adult-onset POAG on chromosome 4 (4q35.1–4q35.2). 

The POAG family reported here represents a rare and valuable resource for the identification of a novel gene implicated in the pathogenesis of POAG. The glaucoma phenotype was entirely consistent with POAG in its most common presentation (increased IOP and age of onset after 40 years) and strongly indicated that a single gene underlay the disease in this family. In addition, the condition was inherited in an autosomal dominant manner and was highly penetrant. With three generations available for sampling and phenotypic characterization, we were able to perform a genomewide linkage analysis capable of generating a significant LOD score. Three family members were determined not to have POAG. Family member III.9 had suffered a traumatic injury leading to secondary glaucoma, and family member III.17 had anomalous discs leading to an incorrect initial diagnosis of POAG. Family member III.25 had acute angle closure glaucoma. Therefore, after careful re-examination of all living family members, we are certain that it is exclusively POAG that is segregating through this family and not another form of glaucoma. In linkage analyses, equal weight is given to affected and unaffected members, a fact that can complicate the genetic analysis of large families such as the one reported here. In such large families with an autosomal dominant disease, there is an increased likelihood of carrier persons in whom the family mutation does not result in the disease phenotype (i.e., nonpenetrance). Incorrect diagnosis of a disease state can result in false recombination calls and mask the true disease interval in a person who has undergone genetic analysis. While looking for the gene underling vitelliform macular dystrophy, nonpenetrance complicated the early linkage analyses. ³⁸ Phenocopies are another complicating factor in linkage analysis. In the search for the genetic lesion in a large family with holoprosencephaly, factors including an apparent phenocopy, incomplete penetrance, and variable expressivity confounded the identification of SIX3 as the disease gene for 15 years. ³⁹ Our use of a penetrance value of 0.95 and phenocopy rate of 0.05, together with accurate phenotyping, should reduce these potential confounding factors to a minimum, as suggested by the fact that only one LOD score above 3.0 was attained in this study. The affected-only LOD score of 2.8 was still suggestive of linkage to the region. The increased power of the linkage calculation when including unaffected persons is well recognized. ^{40 –42}  

The novel locus identified in our study is situated at 4q35.1–4q35.2 and encompasses 17 genes. Bioinformatic analysis was undertaken to refine the candidate gene search using a comprehensive literature search and using existing bioinformatic sites (Endeavor, http://www.esat.kuleuven.be/endeavorweb; Suspects, www.genetics.med.ed.ac.uk/suspects/). This approach did not identify strong candidate genes. Of the genes in the critical region, we judged CYP4V2, UFSP2, and LRP2BP to be reasonable candidates based on their known function and ocular expression. The protein encoded by CYP4V2 is a member of the cytochrome P450 hemethiolate protein superfamily and acts as a selective omega-hydroxylase of saturated, medium-chain fatty acids and is widely expressed in human tissues including the retina and the retinal pigment epithelium. ^43,44 Mutations in CYP4V2 have been strongly associated with Bietti's crystalline corneoretinal dystrophy. ^43,44 Ubiquitination is important in diseases associated with aging, such as POAG. ⁴⁵ The UFSP2 (ubiquitin-fold modifier-specific peptidase 2) protein is a recently identified ubiquitin-like protein and is present in most, if not all, multicellular organisms, including plant, nematode, fly, and mammal and is expressed in the testes, liver, and brain. ⁴⁶ Low-density lipoprotein receptor-related protein binding protein (LRP2BP) acts as an endocytic receptor and is a hypoxia-inducible factor-1 (HIF) target. ⁴⁷ Optic nerve ischemia is believed to play a role in the pathogenesis of glaucoma, ³¹ and HIF-1 regulates the transcriptional response to oxygen flux. Sequencing of these genes identified no pathogenic variants. Future work will include exonic sequencing of other genes within the candidate region. If no mutation is found, the whole critical region will be sequenced to exclude the presence of pathogenic changes in introns or regulatory sequences. 

In summary, the present study provides evidence of a novel locus, GLC1Q, for adult-onset POAG on 4q35.1-q35.2. Comprehensive phenotyping of the family members confirmed that this POAG family represents a valuable resource for the identification of a novel gene implicated in the pathogenesis of POAG with autosomal dominant inheritance and a high degree of penetrance. The inheritance pattern strongly indicates that a single gene underlies the POAG phenotype in this family. From the genes within the critical region that were not excluded by sequencing, it is apparent that the causative POAG gene in this family will have a novel function and a pathogenic mechanism compared with the previously known POAG genes. Future functional studies should, therefore, elucidate new pathways in eye homeostasis and ascribe a new function to a previously identified gene. Studies such as that reported here continue to add to our understanding of ocular function and disease and are a vital step toward our ability to consider and develop therapeutic strategies. 

Footnotes

 Supported by the National Institute for Health Research Biomedical Research Centre and Central Manchester and Manchester Children's Foundation NHS Trust, United Kingdom. EMW is funded by a Fight for Sight grant awarded to FDCM.

Footnotes

 Disclosure: L.F. Porter, None; J.E. Urquhart, None; E. O'Donoghue, None; A.F. Spencer, None; E.M. Wade, None; F.D.C. Manson, None; G.C.M. Black, None

The authors thank Rahat Perveen for technical assistance.