Mutations in the Zinc Finger Protein Gene, ZNF469, Contribute to the Pathogenesis of Keratoconus

Abstract

Purpose.: Mutations in the zinc finger protein gene ZNF469 cause recessive brittle cornea syndrome, characterized by spontaneous corneal perforations. Genome-wide association studies (GWAS) have implicated common variants in this gene as a determinant for central corneal thickness (CCT). We investigated the contribution of ZNF469 in a sample set of keratoconus patients.

Methods.: Forty-three patients with keratoconus (49% Māori or Pacific [Polynesian]) were recruited. If a family history was present, family members were recruited. Participants underwent comprehensive examination, and a DNA sample was collected. Mutational analysis of ZNF469 was undertaken using Sanger sequencing, including an ancestrally matched Polynesian control population. Bioinformatic databases of exome variation and protein prediction software were used to determine presence and frequency and the pathogenicity for each observed change.

Results.: Fourteen nonsynonymous missense single nucleotide polymorphisms (SNPs) were observed in ZNF469. Of the 43 probands, at least one probable disease-causing variant was detected in 20 (46%) (16/32 sporadic, 4/11 familial) and two variants in 5 (11.6%) (3/32 sporadic, 2/11 familial). Only heterozygous changes segregated with disease. Three “deleterious” changes observed in the Polynesian controls were removed from analysis; therefore pathogenic variants occurred in 10/43 (23.3%).

Conclusions.: Rare missense mutations in ZNF469, predicted to be pathogenic, occurred heterozygously, at a frequency of 23% in a keratoconus population. ZNF469 is associated with CCT in GWAS and is therefore likely to play a role in the synthesis and/or organization of corneal collagen fibers. The pathogenic changes observed either genetically predispose toward a “thin” cornea, which then becomes keratoconic, or are directly pathogenic.

Introduction

Keratoconus is a progressive corneal dystrophy leading to a characteristic pattern of corneal thinning and bulging (ectasia) with induced irregular myopic astigmatism. It may be markedly asymmetrical, and onset is usually in the teens, with noticeable change occurring during puberty. The incidence of keratoconus within the general population is approximately 1 in 2000, ¹ with no predilection for either sex. 

Keratoconus is a major indication for corneal transplantation in the Western world, constituting 41% of all keratoplasties performed in New Zealand via the New Zealand National Eye Bank over a 10-year period ² and 67% in the pediatric population (<14 years). ³ Comparatively, keratoconus was the clinical indication in 28.8% in a French series ⁴ and in 11.4% to 15.4% in two US series. ^5,6 This represents a disproportionate percentage of transplantation for keratoconus in New Zealand compared with the rest of the developed world. 

Many multigenerational families are identified that show autosomal dominant (AD) inheritance; of affected New Zealand Māori and Pacific individuals, 31% report a positive family history of keratoconus. ⁷ One study looking at corneal topographic values in teenagers showed a far higher incidence of suspected keratoconus in Māori/Pacific students compared with European Caucasian students (26.9% vs. 12.9% P = 0.0014). ⁸ Other investigators characterizing the mode of inheritance have reported sporadic autosomal recessive ⁹ and AD inheritance. ¹⁰ However, other studies have established linkage by using a model of inheritance that lies somewhere between these two. ¹¹  

Brittle cornea syndrome (BCS) is an autosomal recessive condition in which a thin cornea spontaneously perforates with minimal trauma, resulting in poor ocular survival. Cases in the literature arise from 1913, with early descriptions including brittle corneas, blue sclerae, keratoglobus, and keratoconus, although it is difficult to exclude an Ehlers-Danlos phenotype from the descriptors. ¹² An association with red hair was subsequently observed in consanguineous pedigrees of Arabic origin, ¹³ Greek origin, ¹⁴ and among Jews of Tunisian origin, ¹⁵ which aided identification of mutations in the ZNF469 gene in patients with BCS in 2008 ¹⁶ and has been further replicated. ^17–20 A second transcription factor gene, PRMD5, has also had mutations demonstrated in BCS. ^21,22  

Recently, genome-wide association studies (GWAS) have established a role for common variants in ZNF469 in the control of central corneal thickness (CCT) and therefore as a potential risk factor for glaucoma, ^23–27 and very recently keratoconus. ²⁸ A number of other genes are implicated in keratoconus, and these are summarized in a recent review article. ²⁹ Although there is evidence for a strong genetic contribution to keratoconus and there has been identification of specific and significant causative genes, most genes identified to date account for only a minority of cases. 

Given the association of pathogenic mutations in ZNF469 causing autosomal recessive BCS, and the mounting evidence that ZNF469 plays a role in determination or regulation of CCT, this is an ideal candidate gene for the keratoconus phenotype. The high incidence and familial aggregation of cases in the Polynesian population suggest a mode of inheritance that could be AD with reduced penetrance, or autosomal recessive with pseudodominance assuming a high carrier frequency of potentially pathogenic mutations given a small migratory population and founder effect. This study examines whether ZNF469 is a candidate gene in the pathogenesis of keratoconus in a New Zealand population, including people of Polynesian ancestry. 

Methods

Patient Recruitment

Patients were recruited from the Department of Ophthalmology, Greenlane Clinical Centre, Auckland District Health Board, with a clinical diagnosis of keratoconus, and were reviewed in the University Clinic, Department of Ophthalmology, University of Auckland. The protocol of this study adhered to the tenets of the Declaration of Helsinki with Institutional Ethics and Māori Research Review Board approval (Ministry of Health NTX/06/12/161 and ADHB A+3657). Family members were recruited where possible when a positive family history was reported. 

Clinical

All subjects underwent extensive clinical examination including Snellen visual acuity, autorefraction, corneal topography, and pachymetry using a combined Placido/slit-scanning elevation tomography system (Orbscan II; Bausch & Lomb Surgical, Rochester, NY, USA) and/or Pentacam Schiempflug analysis (Oculus, Wetzlar, Germany), slit-lamp examination and photography, and laser scanning in vivo confocal microscopy (IVCM) using the HRTII (Heidelberg Retina Tomograph II, Rostock Corneal Module [RCM]; Heidelberg Engineering GmbH, Heidelberg, Germany). 

DNA Collection

Following informed consent, biological samples (peripheral venous blood or saliva specimen) were collected for DNA extraction using the salt extraction method from blood ³⁰ and according to the manufacturer's instructions for saliva kits (Oragene; DNAGenotek, Ottawa, ON, Canada). For controls, two populations were used. Deoxyribonucleic acid samples were collected from randomly selected and ancestrally matched individuals attending the Department of Ophthalmology who did not exhibit any clinical evidence of corneal abnormality in terms of appearance or topographic parameters. The other control population with self-reported Māori ancestry was recruited as previously described with ethical approval from a regional Ethics Committee (OTA/99/11/098). ³¹  

Mutational Analysis of ZNF469

Deoxyribonucleic acid samples were screened for mutations in all coding exons of ZNF469 (NM_001127464.1) and PRMD5 (NM_018699.2) including intron–exon boundaries; further details of primers and polymerase chain reaction (PCR) conditions are provided in Supplementary Table S1. Following column purification with HighPure PCR purification kit (Roche Diagnostic, Mannheim, Germany), the product was sequenced directly according to protocols accompanying the ABI BigDye terminator kit v3.1 (Applied Biosystems, Inc., Foster City, CA, USA). Bidirectional sequencing of amplicons was undertaken on an ABI 3700 prism genetic analyzer (Applied Biosystems, Inc). Nucleotide sequences were compared with the published ZNF469 and PRMD5 sequences using CodonCode Aligner (CodonCode Corporation, Centreville, MA, USA) and polymorphic variation data in electronic databases to determine pathogenicity. Thirty-one Polynesian (New Zealand Māori = 14, Samoan = 15, Tongan = 1, Niuean = 1) control individuals (62 alleles) underwent full sequencing of the coding regions of ZNF469. A further 15 control participants with self-reported Māori ancestry had exomes sequenced at 50× depth by Illumina technology using an Agilent (Santa Clara, CA, USA) exome enrichment kit. Sequence data were aligned, analyzed, and managed as previously described. ³²  

To determine the frequency of the ZNF469 variants detected in the familial cases (p.E316K, p.R2129K, p.A2475E, p.R2879H), in addition to the Polynesian controls, a further 70 New Zealand European controls (140 alleles) were screened. Screening for the detected ZNF469 sequence variants used HRMA on the RotorGene6000 (Corbett Life Sciences, San Francisco, CA, USA), using the High Resolution Melting Master kit (Roche Diagnostic). Further details of primers are provided in Supplementary Material S1. Each HRMA reaction included a positive and negative control based on sequencing confirmation. Any sample on the melt curve that produced an equivocal reading was subject to further PCR and sequencing to confirm or exclude the presence of the sequence variation. 

For the sequence variants, homology and predicted destruction or creation of exonic splicing enhancers, or effects on splicing, were evaluated using a variety of publicly available software. PolyPhen2 (http://genetics.bwh.harvard.edu/pph2/) and SIFT (available in the public domain at http://sift.jcvi.org/) analysis were used to predict the impact of the missense variants on protein structure and function, with Gene Splicer and Human Splicing Finder ³³ for splicing signal analysis. Factors considered for determining potential pathogenicity of unclassified variants were positive family segregation, an allele frequency of <1/100 control chromosomes, homology, and/or bioinformatic prediction of biological significance. For an unclassified variant to remain in the list as a possible disease-causing variant, the criteria included present in patient population, absent in Polynesian control population, absent in population databases of human variation (Exome Variant Server [EVS; available in the public domain at http://evs.gs.washington.edu/EVS] or 1000 Genomes [1000G; available in the public domain at http://browser.1000genomes.org]), or with an allele frequency of less than 1/100 in European control chromosomes studied here, prediction of probably/possibly damaging or damaging/deleterious in at least one of the protein prediction software programs. 

For three changes (p.R2129K, p.A2475E, and p.G3415V), both protein prediction algorithms suggested that the change was deleterious/pathogenic; however, the alleles were present in the Polynesian control population and databases of human variation. Similarly, p.G3256R was not present in any population databases or in the control population, but was calculated to be benign or tolerated in terms of protein pathogenicity. These four variants were nevertheless included in the initial analysis. 

Results

Forty-tree probands were recruited: 32 were isolated sporadic cases and the remaining 11 were probands of families with more than one affected member. Demographic details are provided in Table 1. The pedigrees are shown in Figure 1. Of the families, all were Polynesian except for one (FYJ1 European). 

Figure 1

 

Pedigrees of familial keratoconus, with ZNF469 variants by genotyped individuals. Filled symbols, affected; unfilled symbols, clinically unaffected; diagonal line, deceased; plus sign, wild-type allele; asterisk, examined and DNA sample collected.

Table 1

 

Demographics of the Keratoconus Probands at Recruitment: Familial and Sporadic

Disease	Number	Age, y (Range)	Sex, Female:Male	Ethnicity (%)		Familial (%)	Sporadic (%)
Keratoconus	43	41.5 (15–83)	22:21			11 (25)	32 (75)
				Caucasian	18 (42)	1 (2)	17 (40)
				Polynesian	21 (49)	10 (23)	11 (25)
				Indian	4 (9)	0	4 (9)

ZNF469

Thirty-seven variants were observed in ZNF469, as listed in Supplementary Table S1, but these were all missense changes, with no indel, frame shift, or nonsense changes. Of the 37 variants observed, synonymous single nucleotide polymorphisms (SNPs) were excluded from further investigation, and nonsynonymous SNPs that were either absent in population databases and/or predicted to be damaging on at least one protein prediction algorithm were included for further investigation. This left 14 variants, but a further 4 were excluded as both protein prediction algorithms were benign and/or the frequency in both the case and control population was very high. Ten possibly deleterious nonsynonymous missense SNPs therefore remained, 6 of which were not previously documented in any population-based genetic variant database (Table 2). The locations of the observed variants are illustrated in Figure 2. 

Figure 2

 

Schematic of ZNF469 protein domains, with location of variants described in keratoconus cohort (Reference Transcript ZNF46-201 ENST00000437464).

Table 2

 

Nonsynonymous SNPs Identified

Nucleotide	Protein	Cases	Ancestry	rs Number	Polynesian Controls	MAF Polynesian Controls	Caucasian Controls	EVS	1000G	PolyPhen2	SIFT
c.946G>A	p.E316L	1 fam	Māori	-	0/92	0%	0/140	Absent	Absent	Prob dam	Tol
c.1697C>T	p.A566V	3	Caucasian, n = 2 Indian	rs181785233	0/92	0%		Absent	T0.005	Poss dam	Tol
c.6386G>A	p.R2129K	8	Tonga n, = 2 Samoan, n = 3 Caucasian, n = 2 Māori	rs13334190	13/92	14.13%	0/140	Absent	A0.063	Poss dam	Del
c.6796G>A	p.G2266A	1	Caucasian		0/92	0.00%		Absent	Absent	Benign	Tol
c.7424C>A	p.A2475E	4, including  2 fam	Family: 1 Caucasian, 1 Māori Sporadic: Caucasian, Indian	rs141218390	2/92	2.1%	6/140	5.6	T0.056	Prob dam	Dam
c.8246A>T	p.D2749V	1	Caucasian	rs3812954	1/92	1.08%	0/140 (4.3%)	5.6	T0.056	Poss dam	Del
c.8636G>A	p.R2879H	1 fam	Tongan	-	0	0.00%		Absent	Absent	Poss dam	Dam
c.9616C>T	p.P3206L	1	Caucasian		0	0.00%		Absent	Absent	Benign	Dam
c.9766G>A	p.G3256R	1	Caucasian		0	0.00%		Absent	Absent	Benign	Tol
c.10244g>T	p.G3415V	3, 1 fam	Family: CI Māori Sporadic: Māori, n = 2	-rs140056980	12/92	13.04%		Absent	T0.010	Benign	Dam

 The table provides data of minor allele frequency (MAF) in both the Polynesian control and the population databases (EVS and 1000G), protein prediction, and frequency within the disease cohort. The variants R2129K, A2475E, and G3415V occurred in the control populations, although predicted to be deleterious, and were subsequently excluded. MAF, minor allele frequency; fam, family; prob dam, probably damaging; poss dam; possibly damaging; tol, tolerated; del, deleterious.

In the 43 cases, at least one of these possibly deleterious/rare variants in ZNF469 was detected in 20 (46%). In 5 cases, two variants were observed (11.6%—three compound heterozygotes, one homozygote, and one case with two changes presumed to be in cis); all these had a severe phenotype using the Collaborative Longitudinal Evaluation of Keratoconus (CLEK) classification ³⁴ (Table 3). For the sporadic cases, 13/32 had one change and 3/32 had two changes. Of the familial cases, 2/11 probands had one change and 2/11 had two changes, although only heterozygous changes segregated with disease. 

Table 3

 

Characteristics of Individuals With Two Changes in ZNF469 (Homozygous, Compound Heterozygous, and in cis)

	Family	Mutations	Ethnicity	CLEK29 Classification Proband	CLEK29 Classification Sibling	Pedigree
1	1 fam	p.E316K	NZ Māori	Severe OU	Severe OU	TAR1
		p.A2475E	in cis
2	1 fam	p.R2129K	Tongan	Severe OU	Severe/moderate	TUT1
		p.R2879H
3		p.G2266A	Caucasian	Severe OU
		p.D2749V
4		p.A566V	Caucasian	Severe OU
		p.R2129K
5		p.G3415V	NZ Māori	Severe OU
		p.G3415V

 OU, both eyes in cis: The two changes observed in both brothers and were not present in an unaffected parent or an unaffected offspring. Therefore the two changes most likely occur on the same allele.

Three of the observed variants, although predicted to be deleterious, appeared at a frequency greater than 1% in the Polynesian controls and/or in the EVS or 1000G databases. Removing these three variants (p.R2129K, p.A2475E, and p.G3415V) left one change present in 10 probands (23.2%), but only one compound heterozygote (1/43 2.3%). 

PRMD5

PRMD5 mutational analysis was undertaken in 30 sporadic cases. Two of the sporadic cases included in the ZNF469 investigation had insufficient DNA remaining for analysis. 

Sequence variants observed were synonymous, except for one change, observed heterozygously in three Māori cases (Supplementary Table S2), p.S356T. SIFT predicted that this change would be tolerated, although it was called possibly damaging by PolyPhen2. Because of the low yield of changes in PRMD5, further testing was not undertaken in the familial cases. 

Discussion

The association of ZNF469 in the determination of central corneal curvature and thickness in the normal population, paired with the BCS phenotype with recessive mutations, makes ZNF469 an excellent candidate gene to study in a keratoconus population. As a candidate it also makes biological sense, as functionally ZNF469 is considered to play a role in the synthesis and/or organization of corneal collagen fibers in conjunction with PRMD5. ²² Recent work shows downregulation of key extracellular matrix (ECM) component genes [clusterin [CLU], glypican-6 [GPC6], procollagen C-endopeptidase enhancer 2 [PCOLCE2], and thrombospondin [THBS1]) in ZNF mutant fibroblasts. ³⁵  

Rare missense variants in ZNF469, predicted to be pathogenic, were detected in the heterozygous state at a frequency of up to 46% in our keratoconus population, and two variants were observed in 11.6% (three compound heterozygotes, one homozygote, one with two changes in cis). Patients with two changes generally had a severe keratoconus phenotype as defined by the CLEK classification ³⁴ (Table 3), all requiring corneal transplantation by the age of 20. Three of the observed variant changes were also present in the control Polynesian population and the population databases of human variation, although at very different frequencies. Removing these changes from analysis, 23% of patients still had one probable pathogenic variant. 

In the initial analysis we included three changes (p.R2129K, p.A2475E, and p.G3415V) observed in both the keratoconus population and the control Polynesian population, which were predicted by one or both protein prediction algorithms to be damaging (Table 2). p.G3415V occurred in the homozygous state in one patient with a severe keratoconus phenotype and in the heterozygous state in two patients. 

The p.R2129K variant was also found in combination with a second presumed pathogenic variant in two patients, both with a severe keratoconus phenotype, and the p.A2475E variant occurred in association with the p.E316K variant in two Tongan brothers with a severe phenotype (Fig. 1, Ped TAR1); however, nonsegregation suggests that the changes occur in cis. 

The presence of the p.G3415V allele in the heterozygous state in 13% of the Polynesian control population, as well as the p.R2129K in 14%, is a relatively high frequency. A likely explanation is that this is due to a founder effect given the migratory history of the Polynesian population. Ensembl (available in the public domain at http://www.ensembl.org/) frequency data for the p.R2129K are 6% in all populations, but range from 2% in the Asian compared with 15% in the African population (with the highest rate in the subpopulation American of African Ancestry in the SW USA and 16% in the Yoruba). p.G3415V is not present in most populations but is observed in the Asian population (4%–6% in Han Chinese). 

Thus p.R2129K is common, with highest prevalence in Polynesians and African Americans. The normal range of CCT in the Polynesians is not reported, but Haseltine et al. ³⁶ showed a clear ethnic variation; CCT in African Americans was lower (529.3 μm) than in Hispanic people (544.7 μm, P = 0.008) and Europeans (549.9 μm, P < 0.001). While it is feasible that common variants such as p.R2129K influence not only keratoconus but also CCT, this variant was not significant in any of the previously reported GWAS studies of CCT. ^23–27  

Although these three variants (p.R2129K, p.A2475E, and p.G3415V) were excluded in the final analysis because of the high frequency in control populations, a sound argument for leaving them in the analysis could potentially be made. However, we have documented, discussed, and removed them as the frequency in controls would render them as polymorphisms. 

Although anecdotally, it is widely believed that keratoconus is more prevalent and aggressive in New Zealand, especially in the Māori and Pacific Island population, exact figures are not available. ^7,8 However, keratoconus is the leading indication for corneal transplantation in both adults and children in New Zealand. ^37,38 It is plausible that genetic factors are responsible for the ethnic predisposition to keratoconus in New Zealand, and the increased frequency of these probable deleterious ZNF469 variants may go some way toward explaining the higher incidence of keratoconus in the Polynesian population with the appropriate precipitating environmental stimulus. The lack of segregation may be explained by an inheritance pattern of AD with reduced penetrance, or the frequency of the common variant(s) if an environmental component to disease manifestation is not universally present. 

Brittle cornea syndrome has a spectrum of nonocular features including red hair, ¹² which informed the ZNF469 gene identification ^16,39 and has been replicated in other individuals since. ^17–20 The extraocular manifestations reported in association with BCS include scoliosis and reduced bone mineral density in the range of osteopenia and osteoporosis, as well as abnormal dentition (ranging from hypomineralization to mandibular crowding) and cardiac anomalies. Of note, the spectrum of phenotype includes absence of extraocular features, ¹⁹ individuals without corneal fragility, ²⁰ and a phenotype more consistent with Ehlers-Danlos. ¹⁷  

One of the limitations of this study was that although all patients were extensively characterized from an ocular perspective and a full medical history was taken, we did not specifically ask or examine for the BCS extraocular features described. Therefore subclinical abnormalities in bone mineral density or abnormal dentition would not have been detected by us. Unfortunately, it is not possible to retrospectively systemically phentoype these patients. Another limitation with this study was the relatively small number of individuals comprising the control Polynesian population. Genetic variation in Māori and Polynesian individuals is not represented in any population variation databases; this emphasizes one of the current limitations of these public databases, for which allele frequencies are not available for minority ancestral groups. 

The nature of the mutations in ZNF469 previously described in association with the BCS phenotype are also more severe than observed in this cohort. The mutations in BCS described to date are deletions or duplications resulting in frame shifts with premature termination of the protein, ^17,21,39 nonsense, ²¹ or missense and are located in the fourth of the five zinc finger domains. ¹⁸ This contrasts with the missense variants we report, predominantly occurring heterozygously and not directly located within the ZNF domains. This may also contribute to the ocular findings and, indeed, to extraocular features (if present) being at the mild end of the BCS spectrum. 

Burkitt Wright et al. ²² demonstrated mutations in PRMD5 in BCS and noted a milder corneal phenotype, but still reduced CCT and keratoconus in heterozygous carriers of PRMD5 mutations compared to those in homozygous mutation carriers. The authors suggested that this may be due to a dosage effect. In this population, only one nonsynonymous PRMD5 change of unknown significance was observed in three individuals. 

No clear functional studies of ZNF469 mutants are available to clearly document the pathophysiological process causing these changes. However, the high frequency of at least one potentially pathogenic variant in Polynesians compared with a control population variant is surprising. 

In conclusion, ZNF469 is a strongly associated factor in determination of CCT with recessive loss of function mutations causing BCS. This study, demonstrating an increased frequency of heterozygous, pathogenic changes in a keratoconus cohort of sporadic and familial cases of mixed ethnicities, suggests that ZNF469 contributes to the pathogenesis of keratoconus. The higher allele frequency of some of these changes in a control Polynesian population either excludes them as pathogenic or may partly account for the higher frequency of keratoconus within this population. These findings suggest that the pathogenic changes observed may genetically predispose toward a thin cornea, which given the right environmental stimulus may become keratoconic; or, alternatively, it is feasible that the changes are directly pathogenic and that a dosage effect predisposes to keratoconus. 

Replication of these data in other population cohorts may contribute to understanding the findings, as will further clarification of the role of the ZNF469 gene product in the development and regulation of normal corneal collagen structure and function. 

Supplementary Materials

doc S1 

doc S2 

Acknowledgments

We thank Amanda Richards and Bryan Hay for excellent technical assistance. Referring Ophthalmologists. 

Supported by the Save Sight Society of New Zealand, the Maurice and Phyllis Paykel Trust, and the Auckland Medical Research Foundation. 

Disclosure: A.L. Vincent, None; C.A. Jordan, None; M.J. Cadzow, None; T.R. Merriman, None; C.N. McGhee, None