Anophthalmia/microphthalmia (A/M) represent severe developmental ocular malformations. Currently, mutations in known genes explain less than 40% of A/M cases. We performed whole-genome copy number variation analysis in 60 patients affected with isolated or syndromic A/M. Pathogenic deletions of 3q26 (SOX2) were identified in four independent patients with syndromic microphthalmia. Other variants of interest included regions with a known role in human disease (likely pathogenic) as well as novel rearrangements (uncertain significance). A 2.2-Mb duplication of 3q29 in a patient with non-syndromic anophthalmia and an 877-kb duplication of 11p13 (PAX6) and a 1.4-Mb deletion of 17q11.2 (NF1) in two independent probands with syndromic microphthalmia and other ocular defects were identified; while ocular anomalies have been previously associated with 3q29 duplications, PAX6 duplications, and NF1 mutations in some cases, the ocular phenotypes observed here are more severe than previously reported. Three novel regions of possible interest included a 2q14.2 duplication which cosegregated with microphthalmia/microcornea and congenital cataracts in one family, and 2q21 and 15q26 duplications in two additional cases; each of these regions contains genes that are active during vertebrate ocular development. Overall, this study identified causative copy number mutations and regions with a possible role in ocular disease in 17% of A/M cases.
Anophthalmia/microphthalmia (A/M) is characterized by small (microphthalmia) or absent (anophthalmia) eye globes; anophthalmia or microphthalmia is reported in 1 out of 10 pediatric patients with visual impairment with an overall frequency of 1–3/10,000 births [1-3]. A/M is associated with additional non-ocular anomalies in approximately one third of cases; both isolated and syndromic cases can be caused by genetic factors, including chromosomal anomalies and single gene disruptions [4, 5]. Mutations in SOX2, OTX2 and FOXE3 transcription factors were shown to play major roles in A/M phenotypes with some contribution from CHX10, RAX, PITX3, PAX6, GDF6, STRA6, FRAS1, CHD7, BMP4 and other factors [4, 5]. At this point, mutations in known genes explain less than 40% of this condition ; therefore, additional causative factors need to be discovered.
Analyses of chromosomal abnormalities such as translocations or deletions/duplications of chromosomal material have been widely utilized to discover genomic regions associated with human disorders and to subsequently identify causative gene(s). This approach resulted in the identification of numerous human ocular disease genes including PITX2 for Axenfeld–Rieger syndrome , SOX2 for syndromic anophthalmia , B3GALTL for Peters Plus syndrome , and CHD7 for CHARGE syndrome . Copy number variation (CNV) analysis or array comparative genomic hybridization (array-CGH) allows high resolution screening for structural alterations in human genomes . CNVs are deletions or duplications of various sizes that can be inherited or occur as dominant de novo mutations. Recessive disease-causing mutations can be clinically observed if a deleted allele happens to be combined with a mutation-carrying allele [8, 11].
In order to better understand the mechanisms of these debilitating human phenotypes, identification of novel genetic factors involved in human ocular syndromes will be essential. In this article, we report results of the array-CGH analysis of 60 patients affected with A/M that identified pathogenic CNVs as well as several candidate regions/genes that can now be further investigated.
Materials and methods
This human study was approved by the Institutional Review Boards of Children's Hospital of Wisconsin and Albert Einstein Healthcare Network with written informed consent obtained for every subject. Genomic DNA was extracted using standard procedures from blood or buccal samples. Previously reported patients were excluded [12, 13]. Sixty individuals with A/M were included in this study; 35 patients had syndromic A/M and the remaining 25 had isolated ocular defects. Previous screening excluded mutations in SOX2, OTX2, PITX2, PITX3, FOXE3, and BMP4 [12, 14-17].
Patients were screened for structural genomic variation using the Affymetrix Genome-Wide Human SNP Array 6.0 (Santa Clara, CA) including over 1.8 million genetic markers as previously described . The copy number analysis was performed using the Canary Algorithm and Affymetrix Genotyping Console 4.1.1 software, protocols and workflows. Standard settings for intensity quality control were utilized with a median absolute pairwise difference (MAPD) greater than 0.4 and criteria that a CNV region of interest must contain consistent information from a minimum of five markers and/or be at least 150 kb in size. Copy number probe data were subjected to custom region analysis to obtain information for all RefSeq (NCBI build GRCh37/hg19) genes. In addition, a second custom region analysis of 203 genes known to be involved in developmental ocular disorders was performed and included all exonic/intronic regions as well as 200 kb of surrounding sequence. Regions identified as polymorphic/benign (>1% frequency) when compared to the Database of Genomic Variants (DGV; http://projects.tcag.ca/variation/) were excluded from further analysis . The remaining copy number variants were classified as ‘pathogenic’ if the CNV was repeatedly associated with the ocular findings in prior reports, as ‘uncertain clinical significance; likely pathogenic’ if the CNV involved regions with a known role in human disease but not clearly associated with the specific ocular findings, and as ‘uncertain clinical significance’ if the CNV was novel or rarely seen (<1%) in the DGV but not associated with human disease in the literature .
SOX2 deletions less than 100 kb were validated by TaqMan Copy Number assay using the Hs02675353_cn probe (Applied Biosystems, Carlsbad, CA) located in the exon of SOX2; available family members were tested with the same probe. TaqMan analysis of family members for the PAX6 duplication was completed using the Hs00240871_cn probe (Applied Biosystems).
Pathogenic copy number mutations
Deletions of 3q26.33 including the SOX2 gene were identified in four independent patients (patients 1–4; Table 1; Fig. 1). The deletion size ranged from 3.6 Mb (patient 1) to ∼30 kb (patient 4) (Figure 1). All four patients were affected with bilateral anophthalmia/severe microphthalmia, mental retardation/developmental delay, and growth deficiency. Three patients had an ataxic gait (patients 1, 2 and 3) and one (patient 4) was reported to have an unsteady gait. Two patients (patients 1 and 2) had micropenis and another (patient 4) had hypoplastic uterus and ovaries. Parents were unaffected in all cases; the SOX2 deletions in patients 1 and 3 were determined to be de novo (parental samples were not available for testing in remaining cases). Deletion of this region is not present in control populations reported in the DGV.
Severe mental retardation, short stature, hypoplastic uterus and ovaries, unsteady gait
SOX2 anophthalmia syndrome: see above
Copy number variants of uncertain significance (including likely pathogenic)
Six additional CNVs of uncertain significance with a possible role in ocular disease were identified (Table 2; Fig. 2); these CNVs were either not seen or rarely seen (<1%) in control populations reported in the DGV. Three of the CNVs affect regions with a known role in human disease and were considered to be likely pathogenic and the remaining three represent potential novel loci.
Table 2. Copy number variants of uncertain significance (including likely pathogenic)
Levy-Shanske (tetrasomy): facial dysmorphism, craniosynostosis and other defects
PCSK6 is enriched in the lens in mouse embryos
Patient 5 is an 18-year-old male with isolated bilateral severe microphthalmia and normal development. CNV analysis identified a 2.2-Mb duplication of 3q29 (Fig. 2a) which is not seen in the general population (DGV). This CNV is consistent with 3q29 microduplication, a syndrome which includes A/M or anterior segment ocular anomalies in some cases, typically in association with developmental delay/mental retardation and variable other anomalies [20, 21].
Patient 6 is a 6-year-old female with bilateral microphthalmia, congenital cataract, glaucoma diagnosed at 2 months of age, abnormal pupil (failure to dilate), optic nerve hypoplasia, and retinal detachment. Non-ocular anomalies include borderline microcephaly, hypotonia with history of gross motor delays (resolved), pituitary hypoplasia, and attention deficit hyperactivity disorder (ADHD). The patient was found to carry an 877-kb duplication of 11p13 including the PAX6 gene (Fig. 2b); this duplication was not observed in the general population (DGV). The patient's mother was found to carry the same duplication and is affected with bilateral congenital cataract, glaucoma diagnosed at 11 years of age, myopia, abnormal pupil (failure to dilate), and right corneal opacity; non-ocular anomalies include unilateral sensorineural hearing loss and short stature. The patient's unaffected father and sister were found to have normal copy number at this region. There is a strong family history of congenital cataracts with or without glaucoma: the patient's maternal uncle, maternal first cousin, maternal grandmother, three maternal great-aunts and uncles, the maternal great-grandmother, and maternal great-great-grandmother are all affected but were unavailable for testing.
Patient 7 is a 10-month-old male with Neurofibromatosis type 1 (NF1) along with right microphthalmia, iris coloboma, and corneal leukoma as well as bilateral clinodactyly and redundant periumbilical skin. His mother and sister are also affected with NF1 with no structural ocular anomalies. He was found to have a 1.4-Mb deletion of 17q11.2 including the NF1 gene (Fig. 2c) that was not observed in the general population (DGV).
Patient 8 is a 2-year-old female with bilateral microphthalmia, microcornea, congenital cataract, and nystagmus along with microcephaly, developmental delays, mild hyperextensibility, and short stature. She was found to have a 545-kb duplication of 2q14.2 (Fig. 3a). The full duplication is not seen in the general population (DGV); one case of a partial duplication of PCDP1 was noted (DGV). There is a strong family history with the patient's brother affected with microphthalmia, microcornea, congenital cataracts, and nystagmus along with microcephaly and speech delay whereas the patient's mother has microcornea, congenital cataract, nystagmus, glaucoma diagnosed at 16 years of age, and two missing lateral incisors; both the brother and mother carry the same 2q14.2 duplication (Fig. 3a). There is no established phenotype associated with 2q14.2 duplication.
Patient 9 is a 1-week-old female with bilateral anophthalmia, choanal atresia, broad nasal bridge, and indented nasal tip. She was found to have a 374-kb duplication of 2q21.1 (Fig. 3b), which is seen in <1% of the general population (DGV); CNV of this region is not well characterized.
Patient 10 is a 40-year-old female with isolated right severe microphthalmia. She was found to carry a 285-kb duplication of 15q26.3 (Fig. 3c), seen in <1% of the general population (DGV). Duplication and tetrasomy for the q26.1-q26.3 region are previously reported, associated with overgrowth, craniofacial dysmorphism, and severe developmental delay, but ocular anomalies have not been noted.
Our results are consistent with previous CNV studies of ocular developmental anomalies (including A/M, coloboma, congenital cataract, and anterior segment dysgenesis) which identified causative CNVs in 3–15% [22-24]. Two of these studies also reported novel candidate regions; there was no overlap between the candidate regions presented in these studies and those identified here [22, 23].
SOX2 disruption is the most common cause of A/M, accounting for 10–20% of cases , so the high frequency of SOX2 deletions (4/60; 7%) in our study is not surprising. The phenotypes reported here are typical for SOX2 anophthalmia syndrome (MIM:206900) ; while hypoplastic uterus/ovaries are not commonly noted, this phenotype has been reported previously  and may be under recognized. Similar to previous studies, there is no correlation between the severity of the disease and deletion size, suggesting that other genes in the region are not contributing to the phenotype. Review of the other genes included in the deletions in patients 1 and 2 supports this conclusion. Only two genes have been associated with human disease in heterozygous form, neither of which is due to haploinsufficiency: heterozygous germline gain-of-function mutations in PIK3CA are associated with Cowden syndrome and somatic mosaic missense mutations in this gene are seen in other conditions  while heterozygous dominant-negative mutations in GNB4 cause Charcot–Marie–Tooth . Three genes are associated with recessive metabolic or ciliary dyskinesia phenotypes (CCDC39, DNAJC19, and MCCC1) [28-30]; the remaining genes have no reported human phenotype (http://omim.org/).
Ocular defects, including microphthalmia and anterior segment defects, have been occasionally reported in patients with 3q29 microduplications (MIM:611936), but most patients also have mental retardation/developmental delay and variable other anomalies [20, 21]. The duplication seen in patient 5 overlaps the critical region defined by Goobie et al.  extending from TFRC to BDH1; while the full duplication was not seen in controls, the centromeric (TFRC) and telomeric (BDH1) portions of this region do appear to be polymorphic (DGV). Patient 5 reported here is the first case with isolated ocular anomalies and provides further evidence for the importance of this region in ocular development. DLG1 has been proposed as a likely candidate for the ocular findings ; DLG1 is expressed in the developing lens and retinal pigment epithelium . Mutant mice with Dlg1 deficiency were found to have increased proliferation in the lens, along with growth deficiency, craniofacial abnormalities, kidney defects, and neonatal lethality [31, 32].
The copy number mutations identified in patients 6 and 7 involving PAX6 and NF1, respectively, are likely to contribute to these patients' phenotypes. PAX6 duplications (as seen in patient 6) have been previously reported in association with eye anomalies, but not congenital cataracts and glaucoma as observed in this family. Most previous reports involve larger duplications identified cytogenetically including bands 11p12 and/or 11p14 in addition to 11p13. Ocular findings in these patients are highly variable with strabismus and nystagmus most commonly reported; microphthalmia, iris hypoplasia, eccentric pupils, and retinal defects have occasionally been reported as well . More recently, a patient with a 166-kb duplication of PAX6 and the first exon of ELP4 was reported; she was affected with seizures, developmental delay, microcephaly, ADHD and autistic features, and only minor ocular features, consisting of slight hypopigmentation of the fundi, accommodative esotropia with high hyperopia, astigmatism, amblyopia, and ptosis . Significant overexpression of PAX6 in a mouse model (5–7 and 10–45 copies) resulted in ocular defects ranging from iris and corneal defects to severe microphthalmia .
Deletion of NF1 causes NF1 (MIM 162200), which was seen in patient 7 in addition to the ocular phenotype (unilateral microphthalmia, iris coloboma, and corneal leukoma). The patient's mother and sister are also affected with NF1 but without ocular anomalies. Deletion of NF1 is seen in ∼5% of patients with NF1; the 1.4-Mb deletion seen here is entirely consistent with the most commonly observed deletion (type-1). Patients with this deletion have been well-characterized with no report of structural ocular defects . At the same time, while structural eye defects are not typical of NF1 , one patient with NF1 having anterior segment anomalies (bilateral posterior embryotoxon and left Peters anomaly) and cerebral artery stenosis was reported; the authors postulated that the ocular defects may have been caused by vascular disruption during embryonic development . In addition, gonioscopic evaluation of nine NF1 patients by ultrasonic biomicroscopy suggested that almost half (8/18 eyes) had an abnormal (occludable) anterior chamber angle . Finally, homozygous deletion of Nf1 in mice results in lens dysgenesis; eye phenotypes included normal eyes (19%), anophthalmia or microphthalmia (28%), and absent lens with intact retina (53%); study of embryonic development suggested that Nf1 is required for lens vesicle formation . None of the other genes in the region are known to play a role in ocular development (http://omim.org/). Therefore, it is possible that deletion of the NF1 gene is not sufficient to produce an ocular phenotype unless combined with other ocular risk alleles, possibly inherited from the patient's father in this case. Alternatively, the patient's ocular phenotype may be due to a separate genetic (or environmental) cause unrelated to the NF1 deletion.
The 2q14.2 duplication seen in patient 8 and two affected family members is not currently associated with an established phenotype but shows evidence of being causative for the phenotype in the family. The region overlaps with the Nanophthalmos 3 locus (MIM:611897), 2q11-q14, mapped in a family with simple microphthalmia, small corneas, and high hyperopia ; there were no variations identified in the general population for this region. Several genes within the region are of potential interest: TMEM37 is enriched in the lens , ptpn4a in zebrafish showed expression in the retina during development , and dbi in zebrafish showed embryonic ocular expression . Further study is needed to determine the contribution of the 2q14.2 duplication to the ocular phenotype seen in the family. The final two variants of uncertain significance, 2q21.2 and 15q26.3 duplications, are occasionally seen in control populations (<1%) and thus may represent rare variants or potential susceptibility factors. Both regions contain genes enriched in the lens: SMPD4 in the 2q21.1 duplication and PCSK6 in the 15q26.3 duplication .
While interpretation of the clinical significance of copy number mutations that affect regions/genes of known ocular phenotypes is usually straightforward, the clinical relevance of novel structural genomic changes, or even of variants previously described in patients with a somewhat different ocular manifestation, is often uncertain without further studies. Analysis of additional family members may provide helpful information but needs to be considered with caution because many pathogenic CNVs were found to display variable expressivity or incomplete penetrance [20, 45, 46]. Additional investigation of the regions reported in this paper in human patient populations with similar phenotypes and/or animal models will allow for better understanding of their potential role in human disease and may lead to the identification of novel developmental ocular factors.
The authors are thankful to Rebecca C. Tyler for executing TaqMan assays on select regions and Rachel Lorier, Stephen Hall, Katie Felhofer and Andrea Lenarduzzie for assistance with Affymetrix array CNV analysis. The authors also gratefully acknowledge the patients and their families for their participation in research studies. This work was supported by the National Institutes of Health awards R01EY015518 (E. V. S), R21DC010912 (E. V. S) and funds provided by the Children's Hospital of Wisconsin (E. V. S), along with 1UL1RR031973 from the Clinical and Translational Science Award (CTSA) program.