Trio-based exome sequencing arrests de novo mutations in early-onset high myopia

Significance

Because preschool children encounter fewer risks from environmental pressures, we propose that the condition of early-onset high myopia (EOHM) is driven by a genetic predisposition more than by environmental factors. In this study, we recruited 18 familial trios to decipher the genetic predisposition using whole-exome sequencing. We identified a cluster of unique genes linked to EOHM, as well as mutations in the reported genes. Notably, we showed that both rare inherited mutations and de novo mutations significantly contributed to EOHM. Expression profiling in ocular tissues and mutant mouse phenotyping demonstrated the pathogenicity of mutations in a unique gene, BSG. Our results provide insights into the genetic basis and molecular mechanisms of childhood high myopia.

Abstract

The etiology of the highly myopic condition has been unclear for decades. We investigated the genetic contributions to early-onset high myopia (EOHM), which is defined as having a refraction of less than or equal to −6 diopters before the age of 6, when children are less likely to be exposed to high educational pressures. Trios (two nonmyopic parents and one child) were examined to uncover pathogenic mutations using whole-exome sequencing. We identified parent-transmitted biallelic mutations or de novo mutations in as-yet-unknown or reported genes in 16 probands. Interestingly, an increased rate of de novo mutations was identified in the EOHM patients. Among the newly identified candidate genes, a BSG mutation was identified in one EOHM proband. Expanded screening of 1,040 patients found an additional four mutations in the same gene. Then, we generated Bsg mutant mice to further elucidate the functional impact of this gene and observed typical myopic phenotypes, including an elongated axial length. Using a trio-based exonic screening study in EOHM, we deciphered a prominent role for de novo mutations in EOHM patients without myopic parents. The discovery of a disease gene, BSG, provides insights into myopic development and its etiology, which expands our current understanding of high myopia and might be useful for future treatment and prevention.

Myopia is the most common ocular disease, with an increasing global prevalence, especially in East Asia (1–3). Uncorrected myopia is the leading cause of vision impairment worldwide, according to a report by the World Health Organization (4). High myopia (HM) is very severe myopia, which is defined as less than or equal to −6.00 diopters (D) (5). HM is clinically associated with severe ocular complications, such as macular degeneration, retinal detachment, cataract, and glaucoma, which make HM the leading cause of irreversible blindness in East Asia (1, 6).

Myopia is etiologically heterogeneous because both environmental factors and genetic factors are involved (1, 7). Epidemiological surveys show that outdoor activity reduces the prevalence of myopia, decreasing the risk of myopia associated with short-distance work (8, 9). Myopia often exhibits apparent familial aggregation (10–12), and the number of myopic parents is significantly correlated with myopic onset and progression in children (13). Twin studies and population-based epidemiological investigations show that genetic factors significantly contribute to the development of myopia (6, 14, 15), particularly HM (5). Genome-wide association studies (GWAS) and subsequent metaanalyses have identified dozens of loci and genes that are associated with general myopia or HM (16, 17). Of note, the identified genetic contributions of the dozens of loci and genes to myopia are very limited. To date, based on pedigree studies with next-generation sequencing, several disease-causing genes have been discovered, including two recessive genes, LRPAP1 (18) and LEPREL1 (19); four dominant genes, ZNF644 (20), SCO2 (21), SLC39A5 (22), and P4HA2 (23); and one X-linked gene, ARR3 (24). However, a large-scale screening of these genes in HM cohorts provided evidence that only a small proportion (<5%) of HM patients harbor mutations in these known genes, which can be attributed to as-yet-unidentified causative genes (25).

Because preschool children encounter fewer risks from environmental pressures, we proposed that the condition of early-onset high myopia (EOHM) is driven by a genetic predisposition more than by environmental factors. In this study, we recruited 18 familial trios (healthy parents and an EOHM child) to decipher the genetic predisposition using whole-exome sequencing (WES). We identified a cluster of unique genes linked to EOHM, as well as mutations in the reported genes. Notably, we showed that both rare inherited mutations and de novo mutations significantly contributed to EOHM. Expression profiling in ocular tissues and mutant mouse phenotyping demonstrated the pathogenicity of the mutations in a unique gene, BSG. Our results provide insights into the genetic basis and molecular mechanisms of childhood HM.

Results

EOHM Samples and WES.

In this study, we recruited a cohort of 54 individuals, including 18 children with EOHM and their unaffected parents. The ages at examination of all probands were less than 6, indicating EOHM. The refraction of each patient was less than or equal to −6.00 diopters (D) (Table S1).

Table S1.

The clinical information of 18 probands

Subject	Age, y	Sex	Refractive errors (DS)		BCVA		Parents’ reproductive age
			OD	OS	OD	OS	Father	Mother
H1	5	M	6.00	6.00	0.70	0.70	34	32
H5	5	F	6.00	6.00	0.60	0.60	33	31
H9	6	M	7.00	6.50	0.90	0.90	28	26
H13	6	F	9.00	8.50	0.60	0.60	36	34
H16	6	F	9.00	8.00	0.90	0.90	27	25
H22	5	M	6.00	6.00	0.30	0.30	31	29
H25	3	F	10.00	11.00	0.80	0.80	28	27
H29	3	M	7.00	8.00	0.20	0.20	32	31
H33	6	F	9.00	7.50	0.30	0.50	27	26
H39	6	M	8.75	8.75	0.80	0.80	29	27
H42	4	M	11.50	9.00	0.40	0.50	34	29
H45	4	M	8.00	8.50	0.50	0.50	36	33
H53	5	M	7.50	6.50	0.40	0.40	32	29
H60	4	F	10.00	10.00	0.80	0.80	26	25
H63	4	F	10.00	10.50	0.50	0.50	27	27
H66	5	M	7.00	6.50	0.30	0.30	28	25
H70	2	M	6.00	6.00	0.80	0.80	31	29
H74	6	M	6.00	6.00	0.80	0.80	24	23

Abbreviations are as follows: BCVA, best-corrected visual acuity; DS, diopters; F, female; M, male; ND, not determined; OD, right eye; and OS, left eye.

WES was performed for all probands and the parents of the 18 trios to investigate the genetic basis. Burrows–Wheeler transform (26) and Genome Analysis Toolkit (GATK) (27) were used for the data analyses. The detailed statistical information of the WES data from the 18 HM trios is summarized in the Table S2.

Table S2.

Detailed statistic information of the WES data from the 18 HM trios

Sample	Raw_data, Mb	Clean_data, Mb	Aligned, %	Total effective yield, Mb	Effective sequence on target, Mb	Capture rate, %	Sequencing depth	>4X, %	>10X, %	>20X, %	Duplication rate, %
H1	6,188.57	6,158.08	99.68	5,674.2	3,159.74	55.7	62.97	96.6	93.5	85.8	5.6547
H1F	4,290.07	4,270.19	99.63	3,948.61	2,176.68	55.1	43.38	95.9	90.9	77.8	5.2961
H1M	5,069.07	5,027.3	99.57	4,302.98	2,595.41	60.3	51.72	95.8	91.3	81	12.1196
H5	8,687.68	8,636.75	99.6	7,686.4	4,291.06	55.8	85.52	97.1	95.1	90.3	9.4936
H5F	6,864.29	6,820.95	99.77	6,190.2	3,541.41	57.2	70.58	96.8	94.1	87.6	2.1401
H5M	5,381.27	5,347.96	99.55	4,889.87	2,721.99	55.7	54.25	96.3	92.2	82.5	6.1083
H9	4,986.87	4,960.22	99.56	4,542.39	2,545.39	56	50.73	95.9	91.1	80.1	5.7649
H9F	5,539.92	5,506.2	99.55	5,019.89	2,819.93	56.2	56.2	96.1	92	82.5	6.4125
H9M	5,523.98	5,494.43	99.55	5,061.34	2,804.32	55.4	55.89	96.1	91.9	82.4	5.4215
H13	4,966.37	4,941.25	99.6	4,609.03	2,561.1	55.6	51.04	96	91.5	81	4.4622
H13F	5,236.36	5,209.61	99.59	4,767.53	2,721.79	57.1	54.24	96.2	92	82.2	6.1021
H13M	5,240.13	5,215.06	99.64	4,790.69	2,715.17	56.7	54.11	96.3	92.5	83	5.8597
H16	6,074.89	6,040.94	99.62	5,479.63	3,163.31	57.7	63.04	96.4	93	85.3	6.9266
H16F	6,306.13	6,275	99.61	5,757.15	3,231.67	56.1	64.4	96.6	93.5	85.9	5.6811
H16M	7,661.89	7,623.89	99.6	6,960.67	3,824.79	54.9	76.22	96.9	94.5	88.7	6.3291
H22	7,848.02	7,810.16	99.54	7,190.29	4,397.35	61.2	87.36	96.4	94.2	89.4	5.3947
H22F	10,038.38	9,990.12	99.58	9,206.96	5,609.92	60.9	111.45	96.8	95.1	91.5	5.4212
H22M	5,813.61	5,785.37	99.58	5,339.18	3,269.65	61.2	64.96	95.7	92.3	84.8	5.3391
H25	3,678.21	3,661.09	99.85	3,385.18	1,920.69	56.7	38.28	95.7	89.8	74.6	5.3405
H25F	4,277.34	4,257.23	99.86	3,784.06	2,182.36	57.7	43.49	96	91.1	78.4	8.879
H25M	4,985.69	4,962.33	99.86	4,362.59	2,533.38	58.1	50.49	96.1	92.1	81.9	10.0414
H29	4,693.86	4,671.66	99.86	4,301.64	2,472.31	57.5	49.27	96.2	92	81.5	5.6805
H29F	4,439.79	4,418.91	99.87	3,550.24	2,063.17	58.1	41.12	95.1	88.2	72.6	7.2768
H29M	4,095.63	4,076.3	99.86	3,834.09	2,225.19	58	44.35	95.9	91.1	79	3.8859
H33	4,223.78	4,203.77	99.86	3,855.05	2,220.67	57.6	44.26	95.9	91	78.9	6.2414
H33F	4,586.36	4,564.09	99.85	4,043.99	2,363.64	58.4	47.1	95.9	91.1	79.4	5.4823
H33M	4,449.31	4,428.06	99.86	4,094.93	2,340.32	57.2	46.64	96	91.5	80.1	3.9528
H39	4,900.75	4,877.46	99.85	4,302.09	2,483.68	57.7	49.5	96.2	92	81.4	9.5148
H39F	4,669.56	4,647.28	99.87	4,292.86	2,470.46	57.5	49.23	96.1	91.9	81.4	5.2823
H39M	4,086.96	4,067.44	99.86	3,505.56	2,034.55	58	40.55	95.6	90	76.1	11.8002
H42	7,649.6	7,612	99.58	7,006.19	4,333.43	61.9	86.09	96.3	93.8	88.6	5.4646
H42F	8,892.8	8,849.14	99.58	8,139.62	5,025.5	61.7	99.84	96.6	94.6	90.4	5.5364
H42M	9,856.92	9,808.76	99.6	9,025.55	5,533.45	61.3	109.93	96.6	94.9	91.2	5.5895
H45	5,215.83	5,190.89	99.85	3,032.29	1,768.14	58.3	35.24	94.9	87.7	70.5	4.4962
H45F	4,567.97	4,546.31	99.83	4,146.39	2,377.98	57.4	47.39	96.1	91.7	80.5	6.4063
H45M	4,536.14	4,513.6	99.85	4,188.58	2,429.23	58	48.41	96	91.7	81.1	5.1039
H53	6,684.85	6,652.53	99.61	6,132.08	3,757.43	61.3	74.65	96.1	93	86.6	5.4567
H53F	7,989.36	7,950.38	99.59	7,320.22	4,478.32	61.2	88.97	96.5	94.1	89.1	5.4575
H53M	7,283.67	7,248.55	99.62	6,740.25	4,152.52	61.6	82.49	96.2	93.8	88.5	4.5884
H60	8,349.31	8,308.23	99.6	7,733.53	4,766.63	61.6	94.69	96.4	94.3	89.9	4.4892
H60F	7,239.25	7,203.18	99.61	6,719.78	4,145.91	61.7	82.36	96.2	93.5	87.9	4.4525
H60M	7,488.72	7,452.39	99.58	6,922.9	4,271.09	61.7	84.85	96.2	93.8	88.7	4.6273
H63	7,791.5	7,755	99.58	7,195.62	4,322.47	60.1	85.87	96.5	94.5	89.8	4.4493
H63F	9,742.23	9,693.54	99.59	9,007.08	5,597.33	62.1	111.2	96.7	94.9	91.1	4.5376
H63M	7,307	7,271.09	99.61	6,761.53	4,183.13	61.9	83.1	96.2	93.7	88.5	4.4767
H66	10,304.11	10,252.81	99.58	9,538.47	5,972.02	62.6	118.64	96.8	95.1	91.8	4.249
H66F	9,589.37	9,542.14	99.58	8,890.81	5,501.52	61.9	109.29	96.7	94.8	91	4.1816
H66M	7,681.05	7,643.32	99.59	7,126.86	4,418.22	62	87.77	96.3	94	89	4.1142
H70	7,972.06	7,837.52	99.57	6,168.79	3,008.11	48.8	59.76	96.5	93.2	84.4	6.5063
H70F	11,315.56	11,198.43	99.56	6,515.67	3,525.6	54.1	70.04	96.4	93.3	86.2	7.536
H70M	11,260.45	11,148.82	99.51	7,361.2	3,750.88	51	74.52	96.2	93.1	86.1	6.3171
H74	12,375.53	12,259.35	99.47	7,511.69	3,879.37	51.6	77.07	96.3	93.3	86.8	9.0173
H74F	10,344.6	10,229.01	99.63	7,789.15	4,207.64	54	83.59	96.4	93.7	87.4	5.9639
H74M	9,995.07	9,871.08	99.56	6,597.1	3,227.56	48.9	64.12	96.4	93.4	85.5	6.1123
Average	6,782.18	6,740.50	99.66	5,857.42	3,409.16	57.93	67.82	96.22	92.71	84.40	5.97
SD	2,262.83	2,240.04	0.13	1,752.11	1,112.63	3.33	22.05	0.40	1.68	5.10	1.91

Rare Inherited Mutations in EOHM.

Rare inherited mutations cause HM in an autosomal recessive, dominant, or X-linked manner. Based on the sporadic EOHM patients used in this study, we first tried to identify biallelic mutations using mirTrios (28). We identified two known HM candidate genes (LEPREL1 and GRM6), three oculopathy-related genes (FAM161A, GLA, and CACNA1F), and a further possible gene (MAOA) in six different individuals, which accounted for one-third of the EOHM samples (Dataset S1 and Table S3).

Table S3.

Summary of identified variations associated with HM

Patients	Origin	Gene symbol	Cytoband	Mutational effect	Function and disease
De novo mutations
H1	DN	TENM4	11q14.1	c.6173G>A, p.R2058H	Neural development
	DN	ATP8B1	18q21.31	c.86C>T, p.T29I	Transport of aminophospholipids
H13	DN	BSG	19p13.3	c.889G>A, p.G297S	Development and maturation of the retina
H29	DN	FOXP4	6p21.1	c.1286C>T, p.S429F	Regulation of tissue- and cell type-specific gene transcription
H33	DN	TPSG1	16p13.3	c.550G>A, p.V184I	Heparin-stabilized tetramers
H42	DN	EPHB2	1p36.12	c.667C>T, p.R223W	Subpopulation of retinal ganglion cell axons
H70	DN	CSMD1	8p23.2	c.1789T>C, p.S597P	Schizophrenia
	DN	HIST1H3B	6p22.2	c.226G>T, p.A76S	Nucleosomes wrap and compact DNA into chromatin.
Homozygous
H16	F, M	LEPREL1	3q28	c.1046T>C, p.L349P	Collagen chain assembly, stability, and cross-linking, HM
H33	F, M	GRM6	5q35.3	c.2124G>T, p.Q708H	G-protein–coupled receptor for glutamate, HM, night blindness
H45	F, M	FAM161A	2p15	c.904C>T, p.Q302X	Involved in microtubule stabilization, autosomal recessive retinitis pigmentosa-28
Hemizygous (X-linked)
H1	M	MAOA	Xp11.3	c. 53T>A, p.V18E	Oxidative deamination of amines, such as 5-hydroxytryptamine, which involved in development retinal ganglion cells
H9	M	GLA	Xq22.1	c.647A>T, p.Y216F	Hydrolyses the terminal α-galactosyl moieties. Fabry disease, which accompany eye phenotype, corneal dystrophy
H29	M	CACNA1F	Xp11.23	c.3145C>T, p.R1049W	Eye disease, including congenital stationary night blindness type 2A, cone–rod dystrophy, night blindness
H42	M	ABCB7	Xq13.3	c.1294G>A, p.V432M	ATP-binding cassette (ABC) transporters, mitochondrial iron accumulation and isodicentric
H45	M	PLXNA3	Xq28	c.196C>T, p.H66Y	Development of retina and neuron

In proband H16, we detected a damaging biallelic mutation (p.L530P) in LEPREL1, which is involved in collagen chain assembly, stability, and cross-linking. Mutations in this gene have been reported in patients with HM in western Asia (19, 29) and China (30). Leprel1 knockout (KO) mice with abnormal collagen chemistry partially recapitulate the myopic changes (31). Proband H33 carries a homozygous mutation (p.Q708H) in the GRM6 gene. Mutations in GRM6 are reported in HM (32) and nyctalopia (33). In addition to these two known genes, we identified a unique candidate gene, FAM161A, which is involved in microtubule stabilization (34, 35). Proband H45 harbors a nonsense mutation (p.Q302X) in FAM161A. Loss-of-function mutations in this gene are reported to cause autosomal recessive retinitis pigmentosa (36, 37). Interestingly, HM is coupled with these diseases in patients (38).

In another three unrelated patients, we detected mutations in three candidate genes, including MAOA, GLA, and CACNA1F. A boy (H1) harbored a hemizygous mutation in MAOA (p.V18E), which encodes an oxidative deaminase for amines. It is reported that 5-hydroxytryptamine is involved in the development of retinal ganglion cells (39, 40). In addition, we identified a hemizygous mutation (p.Y216F) in the galactosidase α (GLA) gene in proband H9. This gene is a known candidate gene for Fabry disease with an ocular pathology (41) and corneal dystrophy (42). Furthermore, a hemizygous mutation in the CACNA1F gene (p.R1060W) was discovered in proband H29. CACNA1F mutations are reported in patients with HM, congenital stationary night blindness type 2A (43), cone–rod dystrophy (44), and nyctalopia (45).

Contribution of the de Novo Mutation to EOHM.

With the exception of the rare inherited mutations described above, we propose that de novo germline mutations may contribute to the genetic architecture of EOHM, which has not been fully studied. Using the Burrows–Wheeler Aligner (BWA)/GATK/mirTrios, we identified a total of 29 de novo single-nucleotide variants (SNVs) within the coding regions. We confirmed that 20 of the 29 de novo SNVs were genuine de novo mutations by direct PCR sequencing, and 17 were identified as nonsynonymous mutations (Dataset S2). Overall, 13 of the 18 probands (72%) carried at least one de novo mutation, and 7 probands harbored (39%) more than two de novo mutations.

The overall de novo mutation rate in the probands (1.11 events per proband on average) was consistent with a background de novo mutation rate of ∼0.91–1.07 that was estimated from previous studies (46–48). To determine whether the EOHM probands had elevated de novo mutations compared with the controls, we obtained the de novo mutation rates in the normal individuals from the NPdenovo database (49). As a result, we found an increased trend of the overall de novo mutation rate in the HM patients (1.11 events per proband on average) compared with that in the normal individuals (0.74 events per individual on average) with an HM/control rate ratio (RR) of 1.51 (P = 0.05) (Fig. 1A). Interestingly, we observed a significantly elevated de novo missense mutation rate in the patients compared with that in the normal individuals (RR = 1.98, 0.94 vs. 0.48, P = 0.008), and this difference was even greater (RR = 3.74, 0.39 vs. 0.1, P = 0.004) when only the damaging de novo missense mutations were considered. In addition, the number of de novo SNVs in each proband was significantly correlated with the paternal age (r = 0.491, P = 0.019) (Fig. 1B) using a Pearson correlation analysis, which is consistent with previous findings (50, 51). We correlated the number of de novo mutations detected and the degree of myopic refraction in each eye to analyze the possible direct contributions of the de novo mutations to the HM phenotypes. We observed a trend of a higher degree of myopia as the number of de novo mutations increased (0, one, and two) (Fig. 1 C and D).

Fig. 1.

Patterns of de novo mutations in HM patients and their contribution to disease risk. (A) Plot of the mean de novo mutation rate of HM patients (HM) and normal individuals (control). The de novo mutation rate for normal individuals was calculated based on 982 normal individuals from the NPdenovo database (www.wzgenomics.cn/NPdenovo/). The statistical significance of the differences in the de novo mutation rates between the HM patients and the controls was tested using a two-sample Poisson rate test. (B) The relationship between the number of de novo mutations and the paternal age. (C) The relationship between the number of de novo mutations in the proband and the diopter sphere–oculus dexter (DS-OD). (D) The relationship between the number of de novo mutations in the proband and the diopter sphere–oculus sinister (DS-OS). (E) A scatter diagram of the total damaging scores and the expected de novo mutation rate (expected DNMR) of the genes with de novo mutations. The total damaging score was calculated by 14 generic functional prediction tools, and the expected DNMR was used for each gene DNMR average from the mirDNMR database (www.wzgenomics.cn/mirdnmr/).

Candidate Genes with Damaging de Novo Mutations.

The detection of recurrent de novo mutations is a commonly used method to identify disease-causing genes. However, in this study, we found that the de novo mutations occurred in different genes in all cases, which prevented us from performing a statistical analysis of any of the specific genes. Therefore, we used 14 bioinformatics tools to predict the damaging effects of all missense de novo mutations detected and identified mutations that were more likely to confer a disease risk (Fig. 1E). One de novo missense mutation in the EPHB2 gene was identified in proband H42, and the mutation was predicted to be damaging by 10 bioinformatics tools. The EPHB2 gene is involved in retinal axon projections via interactions with ephrin-B proteins (52). In addition, it was reported that the growth cone collapse and axon retraction of retinal ganglion cells could be induced by EPHB2 gene expression (53). Therefore, the direct evidence of the contribution of the EPHB2 gene to retinal axon projections suggests that the EPHB2 mutations may be a possible cause of the optical problems observed in the proband. One de novo missense mutation in the CSMD1 gene was identified in proband H70, which is related to several neuron function-related disorders, such as schizophrenia, autism, sclerosis, etc. (54). One de novo missense mutation in the TENM4 gene was identified in proband H1. Notably, the TENM4 gene is also associated with neuron function-related disorders based on the genome sequencing of cases and controls (55) and a GWAS study (56). In addition, the TENM4 gene is essential for embryonic mesoderm development in mouse model studies (57). One de novo missense mutation in the BSG gene was identified in proband H13. The BSG gene encodes a photoreceptor-specific transmembrane protein, Basigin, which cross talks with rod-derived cone viability factor (RdCVF) (58, 59). The BSG gene will be discussed further in the subsequent sections as a unique candidate gene for EOHM.

Expanded Screening Identified BSG Mutations.

A mutation in the BSG gene (c.889G>A, p.G297S) identified in the EOHM patient (Fig. 2) showed strong pathogenicity, according to computational predictions. Moreover, it is completely absent in Exome Variant Server (EVS) and 1000 Genomes Project (1000G) and exhibits a very rare frequency in Exome Aggregation Consortium (ExAC) (1/115742, 8.64e-06). We further screened the entire coding region of the BSG gene in a large cohort of 1,040 unrelated patients with HM, none of which had mutations in the known genes, to determine the replication of the BSG mutations. Interestingly, we also identified one different missense mutation (c.661C>T, p.P221S), one nonsense mutation (c.205C>T, p.Q69X), and one splicing mutation (c.415+1G>A) in the BSG gene (Table 1 and Fig. 2) in a total of four unrelated families. All of these mutations were absent in the ExAC database and either led to a protein coding change (c.205C>T, p.Q69X; c.415+1G>A) or displayed strong pathogenicity according to the computational assessment (c.889G>A, p.G297S; c.661C>T, p.P221S). Furthermore, both of the missense mutation (G297S and P221S) sites are located in highly conserved amino acids across different species (Fig. 2). However, because the parental DNA was unavailable, it is not clear whether these mutations are de novo mutations. Taken together, these results confirmed the recurrence of the BSG mutations by expanded screening in an additional HM cohort, which supported the pathogenicity of this gene for HM.

Fig. 2.

Identification of mutations in the BSG gene. (A) Identification of mutations in the BSG gene in five unrelated patients. (B) Schematic of the BSG gene and its domains with the sites of the variants identified in this study. (C) Both missense mutations (G297S and P221S) are located in highly conserved regions.

Table 1.

Summary of BSG mutations and the associated phenotypes identified in this study

Patient ID	Mutation (zygosity)	ExAC	EVS	1000G	Type (damaging score*)	Refractive errors (DS)		BCVA
						OD	OS	OD	OS
H13	c.889G>A, p.G297S (het)	1/115742	None	None	Missense (12/14)	−9.00	−8.50	0.6	0.6
T100	c.661C>T, p.P221S (het)	None	None	None	Missense (9/14)	−6.00	−7.00	0.8	0.8
HM850	c.415+1G>A (het)	None	None	None	Splicing	−11.50	−12.00	0.3	0.3
M487	c.205C>T, p.Q69X (het)	None	None	None	Nonsense	−11.00	−9.25	1.0	1.0
M813	c.205C>T, p.Q69X (het)	None	None	None	Nonsense	−7.25	−9.00	1.0	1.0

BCVA, best-corrected visual acuity; DS, diopters; EVS, Exome Variant Server; ExAC, Exome Aggregation Consortium; 1000G, 1000 Genomes Project; OD, right eye; OS, left eye.

^*Damaging score: Damage prediction of missense mutation using 14 online tools (Polyphen2_HDIV, Polyphen2_HVAR, MutationTaster, SIFT, LRT, MutationAssessor, FATHMM, RadialSVM, LR, VEST3, CADD, GERP++, phyloP100way, and SiPhy_29way).

Bsg Mutant Mice Display Typical Myopic Phenotypes in the Axial Length.

We generated knockin mice (Fig. S1) with a c.901G>A mutation corresponding to the c.889G>A mutation identified in the EOHM patient to further investigate the functional impact of the BSG mutation. The total axial length (AL) and vitreous chamber depth (VCD) were measured in variant ages (4, 6, 8, and 10 wk) of the mutant mice and wild-type (WT) siblings. The results showed that the ΔAL significantly changed with group (F = 51.26, P = 1.63e-10) and time (F = 42.36, P = 6.50e-14) overall, and there were no interactions between group and time (F = 2.35, P = 0.1012) (Fig. 3). The heterozygous mutant group had an increased AL in the subsequent 2 wk compared with the WT group (Tukey multiple comparison, Δmean = 0.015 mm, P < 1e-50). The ΔAL in the subsequent 2 wk also changed with time (peaks at 6 wk, and then the ΔAL decreased slightly). However, there were no significant differences with group (F = 0.47, P = 0.49) and time (F = 1.86, P = 0.16) in ΔVCD (Fig. S2). The trend of the AL and VCD of the WT mice was consistent with the previous studies as follows: AL increases during postnatal development, whereas the VCD decreases (60, 61).

Fig. S1.

Bsg knockin mutant mice were generated by a homologous recombination approach and genotyped by PCR using tail genomic DNA.

Fig. 3.

Clinical features of the Bsg mutant mice. Comparisons of the ALs in the WT and mutant mice at each time point [week 6 (w6)–w4, w8–w6, w10–w8]. HET, heterozygous mutant mice; WT, wild-type mice.

Fig. S2.

Comparisons of the ALs in the WT and mutant mice at each time point [week 6 (w6)–w4, w8–w6, w10–w8]. HET, heterozygous mutant mice; WT, wild-type mice.

To test whether retinal function was affected in the mutant mice, we performed an electroretinogram (ERG). The results showed that both the photopic and scotopic ERG responses of the mutant mice were normal compared with those of their WT siblings (Fig. S3). This result indicated that the retinal function was not affected by the Bsg mutation, which was consistent with the clinical manifestation in the patients. Taken together, the results showed that the Bsg mutant mice displayed typical HM phenotypes with a longer AL but no retinal dysfunction.

Fig. S3.

The electroretinogram responses in the mutant mice were normal compared with their WT siblings. (A) The scotopic ERG response of the WT and mutant mice. (B) The photopic ERG response of the WT and mutant mice.

Spatial Expression Patterns of the Bsg Gene in Mice.

Next, we wanted to determine the Bsg expression patterns in different tissues. Therefore, we investigated the spatial expression patterns of Bsg in various mouse tissues. Interestingly, two known myopia-related genes, Sco2 and Sntb1, exhibited patterns similar to that of Bsg (Fig. S4).

Fig. S4.

Bsg expression patterns. (A) Expression patterns of the Bsg gene in different mouse tissues. (B and C) Expression patterns of known myopia-related genes in different mouse tissues.

Discussion

Both myopia and HM are etiologically heterogeneous disorders. It is commonly known that both genetic factors and environmental factors contribute to the etiology (1). Population-based epidemiological investigations found that the disease is associated with environmental risk factors, such as a close reading distance and less outdoor activity (8, 9). With the advent of next-generation sequencing, a few of disease genes have been discovered in recent years (18–24). Because myopia is dependent on both genetics and lifestyle and preschool children have less exposure to environmental risks, we designed this study using a special cohort with EOHM. Each trio has one EOHM child and two unaffected parents, with or without another unaffected sibling. Through this design, we were able to focus on the genetic cause of the newly created EOHM in each family.

Our study used a trio-based WES strategy to dissect the genetic basis of EOHM. Based on WES and the subsequent validation, we deciphered the genetic causes of 4 known genes and discovered 12 unique candidate genes. A total of 16 biallelic or de novo mutations were identified in the present study. To date, cohort-based genetic studies have identified several genes that contribute to myopic development. Jiang et al. (25) comprehensively screened the LRPAP1, CTSH, LEPREL1, ZNF644, SLC39A5, and SCO2 genes in 298 families with EOHM and identified potential pathogenic mutations in 9 patients, with a detection rate of 3.02% (9/298). Among these genes, ZNF644 was the major gene of EOHM (1.67%, 5/298), whereas no mutations were identified in CTSH and LEPREL1. Collectively, these results suggested that the genetic defects responsible for most cases remain to be determined. Strikingly, we deciphered a significant rate of the genetic causes in these trios, supporting our initial hypothesis that EOHM is mainly driven by genetic predisposition.

Among the rare inherited biallelic mutations, three mutations were identified in the known genes GRM6, CACNA1F, and FAM161A that are responsible for inherited retinal dystrophy (IRD) (Dataset S1). Interestingly, HM occurs concomitantly in IRD patients with GRM6 or CACNA1F mutations (32, 62). Our findings are consistent with a previous study showing that 23.8% (71/298) of patients with EOHM actually harbor mutations in IRD genes (38).

The role of de novo mutations in EOHM onset has never been explored. In this study, a total of 20 de novo mutations in the coding regions were validated in 12 EOHM probands. Interestingly, the de novo mutation rate was significantly elevated in the probands compared with that in the normal subjects (RR = 1.51, 1.11 vs. 0.74, P = 0.05), in particular damaging missense mutations (RR = 3.74, 0.39 vs. 0.1, P = 0.004). In addition, the de novo mutation rate was positively correlated with paternal age in this study (Fig. 1B). Children’s refractive changes decrease with parental reproductive age (63), and thus, we speculate that the increased de novo mutation rate in the subjects with aged parents may be the underlying reason for the disease. In fact, we successfully identified several EOHM candidate genes with identified de novo mutations, such as BSG, EPHB2, CSMD1, and TENM4. These findings suggest that de novo mutations contribute substantially to the genetic etiology of EOHM.

The identification of recurrent de novo mutations serves as a useful method to identify disease-causing genes. However, we found that all of the de novo mutations occurred in different genes, which prevented us from performing a statistical analysis of these genes. Then, we searched the genes carrying damaging de novo missense mutations against the literature and found that none of them was associated with HM in previous reports, which can be explained by the fact that de novo mutations are extremely rare events that cannot be identified by GWAS or a linkage analysis. We subsequently asked whether there are any functional categories or cellular pathways enriched in this study. Despite the substantial genetic heterogeneity in HM, closely interconnected protein–protein interactions (PPIs) were identified by integrating the HM risk genes obtained from this and previous studies (Table S4). A gene ontology (GO) enrichment analysis showed that 43 of 62 genes were jointly clustered in four GO biological processes (Table S5). The results suggest that these genes play important roles in disease predisposition. The PPI analysis of these genes revealed a highly connected network, implying that EOHM is associated with visual perception, transcriptional regulation, and cell morphogenesis and homeostasis (Fig. 4).

Table S4.

Candidate genes of HM in other studies

Gene symbol	Function	Disease
P4HA2	Posttranslational modifications of collagen	HM
MYCBP2	Ubiquitination and subsequent proteasomal degradation	Optic disk anomaly associated with HM
BBS4	Sorting of specific membrane proteins to the primary cilia	Bardet–Biedl syndrome associated with HM
LRPAP1	Interacts with low-density lipoprotein receptor-related protein, LRP1/α-2-macroglobulin receptor and glycoprotein 330	EOHM
COL18A1	Encodes the α-chain of type XVIII collagen	Knobloch syndrome
RPGR	Guanine-nucleotide releasing factor, plays a role in ciliogenesis	X-linked retinitis pigmentosa (XLRP) associated with severe visual impairment in women
PRIMPOL	Influence affinities for DNA and nucleotides	HM
OPN1LW, OPN1MW	Encodes for light-absorbing visual pigments of the opsin gene	Nystagmus and poor vision
ZNF644	Eye development	HM
C8orf37	Retinal function	Early-onset retinal dystrophy and HM
SLC39A5	Interference with the BMP/TGF-β pathway, controlling organismal zinc status	Nonsyndromic HM
PROM1	Retinal development	Pediatric cone–rod dystrophy with HM
LEPREL1	Collagen chain assembly	Nonsyndromic HM
LRP2	Reuptake of numerous ligands, including lipoproteins	Stickler syndrome, adult-onset ocular pathogenesis with HM
SLITRK6	Control neurite outgrowth and regulate synaptic development	Progressive auditory neuropathy associated with HM
ZNF469	Transcriptional regulation	Brittle cornea syndrome (BCS) associated with HM
SCO2	Mitochondrial cytochrome c oxidase activity	Autosomal-dominant high-grade myopia
CCDC111	Influences amino acid substitutions	HM
UHRF1BP1L, PTPRR, PPFIA2	In MYP3 genetic locus, eye development	High-grade myopia
NYX	Retinal development	HM
TGFB1	Proliferation, differentiation, adhesion, migration	HM
CYP4V2	Metabolism of fatty acid precursors	Retinitis pigmentosa (RP) associated with HM
TCF4	Initiation of neuronal differentiation	Pitt–Hopkins syndrome with HM
STK11	Metabolism, cell polarity, apoptosis, and DNA damage response	Peutz–Jeghers syndrome (PJS) associated with HM
RPGR	Ciliogenesis, photoreceptor integrity	Female RP carriers associated with HM
ADAMTS18	Retinal development	Knobloch syndrome characterized by HM
PTPN11	Ocular development	Noonan syndrome with HM
COL9A2	Structural component of hyaline cartilage and vitreous of eye	Stickler syndrome with HM
PRDM5	Extracellular matrix development and maintenance	Brittle cornea syndrome with HM
MYP1	Color visual and optic disk	Nonsyndromic HM
COL2A1, COL11A1	Embryonic development of the skeleton	Cleft palate and HM
RP2	Between the Golgi and the ciliary membrane	X-linked retinitis pigmentosa with HM
GRM6	G protein coupled receptor for glutamate	HM
OPTC	Binds collagen fibrils	HM
CLDN16	Paracellular magnesium reabsorption	Nephrocalcinosis and HM
PAX6	Development of the eye, nose, central nervous system	HM
FBN1	Regulates osteoblast maturation	HM
CACNA1F	A variety of calcium-dependent processes	HM
ZFHX1B	Eye development	HM with Hirschsprung disease

Table S5.

GO biological process enrichment of HM candidate genes

GO terms	P value	FDR
GO:0007601∼visual perception	1.41E-10	2.24E-10
GO:0050953∼sensory perception of light stimulus	1.41E-10	2.24E-10
GO:0050877∼neurological system process	4.00E-04	0.0012
GO:0050890∼cognition	2.68E-04	0.0012
GO:0000904∼cell morphogenesis involved in differentiation	3.00E-04	0.0019
GO:0007600∼sensory perception	2.79E-04	0.0022
GO:0000902∼cell morphogenesis	3.47E-04	0.0033
GO:0032989∼cellular component morphogenesis	7.25E-04	0.0080

Fig. 4.

PPI of the HM genes. PPI network of the genes related to HM identified in this study and previous studies.

Among these genes with de novo mutations, we discovered a de novo mutation in a unique gene, BSG, in patient H13, and identified three different BSG mutations in an expanded screen of 1,040 patients with HM. We further verified its functional impact by generating knockin mice carrying the same BSG mutation identified in the first EOHM patient. Strikingly, the mutant mice displayed the myopic feature of an enlarged AL. In addition, our results showed that the spatial expression pattern of Bsg is similar to other known genes, such as Sco2 and Sntb1 (64, 65). BSG encodes basigin, which is associated with retinal development and function. A previous study showed that Bsg KO mice led to defective function and photoreceptor degeneration in the retina (58, 59). Interestingly, Basigin plays an important role in mediating the binding of rod-derived cone viability factor (RdCVF) to the glucose transporter GLUT1, which increases glucose influx into cone photoreceptors (66). This evidence indicated that the retina might be one of the disease target tissues in EOHM driven by the BSG mutation. In this study, the Bsg mutant mice displayed the typical HM phenotypes with AL. As AL is responsible for myopia development (67), our results indicate that the Bsg mutation predisposed typical myopic phenotypes.

In summary, we performed a trio-based study to genetically dissect EOHM using next-generation sequencing and deciphered an important role for de novo mutations in this disease. The discovery of a disease gene, BSG, provides insight into myopia development and etiology, which expands our current understanding of HM and might be useful for future treatment and prevention.

Methods

The human subjects were recruited from The Eye Hospital of Wenzhou Medical University in accordance with a protocol approved by the Ethics Committee of the hospital. Written informed consent was provided by the parents and on behalf of their children before the peripheral blood, and clinical data were collected from the myopia patients and their parents. The experimental procedures are described in detail in SI Methods.

SI Methods

HM Human Subjects.

This study included a total of 18 unrelated trios (29 males and 25 females) of Han Chinese ancestry with HM, and all of the probands have an onset age of <6 y. The human subjects were recruited from The Eye Hospital of Wenzhou Medical University in accordance with a protocol approved by the Ethics Committee of the hospital. In total, 1,040 patients with HM were recruited for expanded screening of BSG gene. Written informed consent was provided by the parents and on behalf of their children before peripheral blood and clinical data were collected from the myopia patients and their parents. A team of HM specialists diagnosed the patients based on comprehensive analyses of the refractive errors and best corrected visual acuity. The refractive errors of all patients were greater than −6.00 diopters (D).

WES and Variant Detection.

Genomic DNA was extracted from the peripheral blood of 54 samples (18 trios) using the Genomic DNA Extraction Kit (Invitrogen). The exome DNA was enriched using the SureSelect Human All Exon Kit (Agilent), and paired-end sequencing was performed on the Illumina HiSeq 2000 sequencing system.

Raw sequencing reads were filtered using the FastQC program (www.bioinformatics.babraham.ac.uk/projects/fastqc/) to remove reads with a sequencing quality lower than 20, and the 3′/5′ adapter sequences on each read were trimmed using the Cutadapt program (https://cutadapt.readthedocs.io/en/stable/). The clean reads were mapped to human reference genome (hg19/GRCH37) by the BWA software, and PCR duplications were removed by the Sequence Alignment/Map tools (SAMtools). Local realignments of reads, quality recalibration, and variant calling (VCF format) were performed by the Genome Analysis Toolkit (GATK). De novo and rare inherited mutations (including SNVs and small indels) in each trio were detected by the mirTrios program based on the VCF files produced by GATK. Annotation of the detected variants with respect to the consequences of their mutation on known genes (gene name, functional effect, amino acid change, etc.), mutation effect predicted by multiple computational methods (SIFT, Polyphen2, MutationTaster, etc.), and population allele frequency (from 1000 Genomes, EVS, and ExAC database) were performed using the ANNOVAR software. The sequencing data have been deposited in the figshare database (DOI: 10.6084/m9.figshare.3497660; https://figshare.com/s/0eab58f90f3b1b20a181).

Pathway and Network Analyses.

Protein interaction relationship data were obtained from the BioGrid (V33) (https://thebiogrid.org/) and STRING (V34) (string-db.org/) databases. Protein–protein interactions (PPIs) and gene coexpression networks for the HM-relevant genes were constructed using the Cytoscape software tool (V3.1 36) (cytoscape.org/).

BSG Amplification and Genotyping.

An additional cohort with HM was selected, and their DNA was submitted for Sanger sequencing. Primers were designed to amplify all coding regions and the intron–exon boundaries of the BSG gene. The PCR products were purified and sequenced on an ABI 3500 Genetic Analyzer (Applied Biosystems).

Generation of Bsg Knockin Mice.

This study was approved by the animal care and ethics committee at Wenzhou Medical University and was conducted according to the Association for Research in Vision and Ophthalmology guidelines. Bsg knockin mice were generated using a homologous recombination approach. The targeting vector was transfected into ES cells. The ES clones were then screened by Southern blotting for correct targeting and injected into C57BL/6J blastocysts; the chimeric mice were bred with C57BL/6J WT mice to produce heterozygous Bsg mice. All mice were maintained under a standard light–dark cycle in Wenzhou Medical University and had free access to food in specific pathogen-free facility. Newborn mice were genotyped by PCR using tail genomic DNA. The primer sequences were as follows: Bsg-WT-F, ATGAGTGTTGAGGTCATATACAGGAGG; Bsg-WT-R, CCAGCCAGATCAGTACTAGGCT; and Bsg-Mut-R, CAGAGGCCACTTGTGTAGCG.

Biometric Measurements.

The total axial length (AL) and vitreous chamber depth (VCD) were measured in variant ages (4, 6, 8, and 10 wk) of the mutant mice and age-matched WT siblings. The center wavelength in this system was 840 nm. The system was mounted onto a traditional slit lamp to image the mouse eyes, and two central crossing lines were set and were applied to align the scanning position. The ocular axial resolution of the spectral domain optical coherence tomography (SD-OCT) instrument was 6 μm and the prolonged scan depth was 7.2 mm. The mice were weighed and then anesthetized by i.p. injections (4 mL/kg) of a mixture of xylazine (4 mg/kg) and ketamine (75 mg/kg). Generally, the animal’s eyes were wide open naturally, with pupils pointing laterally and upward. Subsequently, the mice were placed over a self-made polystyrene box in front of the modified slit lamp. One operator adjusted the mouse eye position until the beam was roughly incident on the central cornea, and the OCT beam was visualized with an infrared sensor card. Then, another operator precisely aligned the optical axis of the image until the specular reflex on the corneal apex was observed, after which it was moved from the anterior surface of the cornea to the retinal pigment epithelium layer. The intersection of the two crossed lines with maximal intensity on the apices of the corneal surface confirmed the perpendicularity of the OCT beam to the whole eyeball. When the specular reflex was detected on all curved surfaces, the operator manually recorded the OCT image. All of the manipulations of the mice (anesthetization and orientation) were performed by a single operator, and another operator recorded the positions of the ocular reflective surfaces. The raw OCT data were exported, and the axial components were measured using custom-designed software. All distances were measured by passing through the center of the pupil on the OCT image.

Statistical Analysis.

Measurement data were expressed as mean ± SD (*P < 0.05, **P < 0.01, ***P < 0.001). The de novo mutation rates were tested by two-sample Poisson rate test (one-tail), and Pearson correlation analysis was performed between the number of de novo mutations in HM patients and other association factors like age, DS-OD, and DS-OS. In Bsg knockin mice model, P value was calculated by one-way ANOVA in each age, and post hoc pairwise comparison was tested by Tukey’s honest significant difference method. All statistical calculations were performed using R, version 3.3.0, software. A value of P < 0.05 was considered as statistically significant.

We defined ΔAL as the change of the AL in subsequent 2 wk (positive value means increase and negative value means decrease), and ΔVCD as the change of the VCD in subsequent 2 wk, and thus got three groups of ΔAL and ΔVCD at each time point (6–4 wk, 8–6 wk, and 10–8 wk). We performed analysis of variance (ANOVA) with group and time as two fixed factors. F test was used to evaluate whether there exists difference overall. The F ratio is significant at the 5% significance level. After performing the F test, “post hoc” analysis was carried out to assess the between-group differences.

Real-Time Quantitative PCR.

The expression of the Bsg gene in different mouse tissues was determined by real-time quantitative PCR using a 7500 Real-Time PCR system (ABI). The β-actin gene was used as an endogenous control. The 20-μL reaction contained 10 μL of 2× fast start universal SYBR Green Master (ROX) (Roche), 10 μM each primer, and 1 μL of genomic DNA as the template. The following thermal cycling conditions were used: 10 min of preheating at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 60 °C. For normalization, the threshold cycle (CT) values of the endogenous control were subtracted from the corresponding CT values of the candidate genes, generating ΔCT values. All relative expression values of the genes were reported as means ± SEs of the means on a 2-log scale.