A strabismus susceptibility locus on chromosome 7p

Abstract

Strabismus has been known to have a significant genetic component, but the mode of inheritance and the identity of the relevant genes have been enigmatic. This paper reports linkage analysis of nonsyndromic strabismus. The principal results of this study are: (i) the demonstrated feasibility of identifying and recruiting large families in which multiple members have (or had) strabismus; (ii) the linkage in one large family of a presumptive strabismus susceptibility locus to 7p22.1 with a multipoint logarithm of odds score of 4.51 under a model of recessive inheritance; and (iii) the failure to observe significant linkage to 7p in six other multiplex families, consistent with genetic heterogeneity among families. These findings suggest that it will be possible to localize and ultimately identify strabismus susceptibility genes by linkage analysis and mutation screening of candidate genes.

Strabismus (misalignment of the eyes; also referred to as “squint”) is one of the most common ocular disorders in humans, affecting 1–4% of the population (1). It is frequently associated with amblyopia (uniocular visual neglect), a leading cause of visual impairment in children and young adults. The familial clustering of strabismus has been recognized since antiquity. For example, Hippocrates stated that “children of parents having distorted eyes squint also for the most part” (1).

Numerous twin and family studies point to a significant genetic component in the etiology of strabismus (summarized in refs. 1–4). Summing the data from 11 published twin studies shows that, among 206 monozygotic and 130 dizygotic twin pairs in which one member of the twin pair had strabismus, 73% of monozygotic twin pairs were concordant for strabismus, whereas only 35% of dizygotic twin pairs were concordant for strabismus. We note that the degree of concordance among dizygotic twin pairs is higher than the ≈10–15% typically reported for siblings; this may reflect the overall higher incidence of strabismus among premature and low-birth-weight infants.

The overall incidence of strabismus and the incidence of specific types of strabismus show appreciable differences between racial groups, further supporting the relevance of genetic factors. Two studies have documented a lower incidence of all types of strabismus among Africans or African Americans (0.5% and 0.6%, respectively) relative to Americans of European ancestry (2.5%; refs. 5 and 6). Moreover, the majority of African, African American, and Asian strabismics are exotropes, whereas the majority of Caucasian strabismics are esotropes (5–7).

With respect to overall heritability, the relative risk for first degree relatives of an affected individual is estimated to be between 3 and 5. Crone and Vezeboer (8) found that 13% of parents of strabismic probands had strabismus vs. a 3% incidence in the general population. Hu (9) found a 9% incidence among first-degree relatives and a 2.2% incidence among second degree relatives, versus a population incidence of 0.6%. Richter (10) found that the incidence of strabismus or the various strabismus-associated ocular anomalies among siblings of an affected proband was ≈20% if both parents were unaffected and 30–40% if one or both parents were also affected, versus a population frequency of 4%. To our knowledge, the lowest published estimate of relative risk for a sibling is 3 (for esotropia; ref. 11). Strabismus affects males and females equally.

These heritability values are likely to be underestimates for the following three reasons: (i) ≈15–20% of strabismus is clearly associated with nonocular disorders, including low birth weight and global CNS defects (12), which are included in the population incidence in most studies; (ii) there is incomplete detection of esophorias and exophorias, microtropias, and the monofixation syndrome (reduced stereoacuity to worse than 60 s of arc), all of which occur at higher frequencies in families of strabismics and probably represent the incompletely penetrant phenotype (13, 14); and (iii) similar or identical subtypes of strabismus tend to cluster within a single family and between monozygotic twins and this would lead to greater heritability estimates if strabismus subtypes were considered separately (15). As an example of the second point, Scott et al. (14) found that the monofixation syndrome was present in 7.7% of first degree relatives of patients with infantile esotropia, compared with a prevalence of 1% in the general population.

Ocular alignment relies on complex sensory and motor pathways in the retina, thalamus, visual cortex, and brainstem, and on the proper development and functioning of the extraocular muscles and orbit. In keeping with this multiplicity of components, current evidence indicates that the inheritance patterns of the common forms of strabismus are complex (16–18). For example, in a study of 173 pedigrees with infantile nonaccommodative esotropia, Maumenee and colleagues (18) found that the disease best fit a model with either two autosomal dominant genes with incomplete penetrance or multifactorial inheritance. Not included in this analysis are the many genetic conditions that are associated with strabismus via their affect on the development and function of the CNS or the anatomy of the eye and orbit (summarized in ref. 4). These include various chromosomal anomalies, Williams syndrome, Waardenburg syndrome, de Lange syndrome, and fragile-X syndrome. A further complication comes from the observation that uncorrected refractive error is a risk factor for strabismus. In particular, anisometropia is a general risk factor for strabismus and hypermetropia is a risk factor specifically for esotropia. As refractive error has a significant heritability (2), genes affecting refraction per se may act as strabismus susceptibility loci.

Recent progress in understanding genetic mechanisms relevant to ocular motility has come from the study of rare Mendelian disorders that lead to congenital fibrosis of the extraocular muscles (CFEOM; summarized in ref. 19). The CFEOM syndromes, together with Duane syndrome, most likely reflect primary defects in some combination of cranial nerves III, IV, or VI, with subsequent ocular muscle dysfunction and fibrosis. Three CFEOM loci and two Duane syndrome loci have been mapped, and the gene for CFEOM2 has recently been identified as the ARIX/PHOX2A homeobox gene (20), previously implicated in the development of brainstem motor nuclei in the mouse (21).

Despite these advances in understanding oculomotor development, the genetic mechanisms involved in the pathogenesis of the common nonsyndromic forms of strabismus and amblyopia remain little studied and poorly understood (22). To date, there have been no published reports of systematic, large-scale efforts to map genes that predispose humans to the common forms of strabismus. In this paper we describe our initial efforts in this direction, including the identification of families with multiple affected members, the results of whole genome linkage scans in seven of these families, and high-resolution genotyping that implicates a recessive locus on chromosome 7p that predisposes to strabismus in one large family.

Materials and Methods

Recruitment of Subjects. Subjects were recruited in accordance with the guidelines set by The Johns Hopkins University Joint Committee on Clinical Investigation. The vast majority of subjects were recruited in the U.S. and most reside in the Mid-Atlantic region. Most appear to be of European ancestry. Patient records were surveyed at 15 pediatric ophthalmology practices to identify families with multiple affected members. Patients (or parents of patients) with a positive family history were informed of the study by letter and invited to contact the study coordinator. Respondents were interviewed to identify affected and unaffected relatives, and pedigrees and phenotypes were further confirmed by telephone interviews with multiple family members, and for some subjects, by retrospective analysis of medical records. Subjects were considered to be affected if they had a history of any of the following: (i) readily apparent ocular misalignment, (ii) ocular surgery to correct a misalignment, (iii) patching of one eye, (iv) orthoptic exercises, or (v) a diagnosis of strabismus or amblyopia. The approximate age of onset and the type of ocular deviation was recorded when this was known, including esotropia vs. exotropia, intermittent vs. constant deviation, and alternate vs. unilateral fixation. The following were also recorded where possible: age at which subjects first began wearing glasses; refractive error and/or visual acuity; and failure of any vision screening tests at school, in the military, or for driving an automobile. To minimize the effect of environmental variables and to exclude subjects with syndromic sources of strabismus, the following were considered cause for exclusion from the study: strabismus secondary to trauma, craniofacial anomalies, anomalies of the orbit or extraocular muscles, mental retardation, CNS infection, chromosomal anomalies or aneuploidies, prematurity (<35 weeks gestation), low birth weight (<1.8 kg), nerve palsies, Duane syndrome, Williams syndrome, oculocutaneous albinism, orbital inflammatory disease, strabismus after retinal reattachment surgery, sensory deprivation strabismus (e.g., secondary to congenital cataract), and myasthenia gravis.

Genotyping. Blood samples were collected and DNA was prepared by proteinase K digestion followed by equilibrium centrifugation in CsCl. Genotyping was performed with samples from 133 subjects at the Center for Inherited Disease Research by using a modification of the Cooperative Human Linkage Center (CHLC, University of Iowa, Ames) version 9 marker set (386 microsatellite markers; average spacing of 9 cM; and average heterozygosity of 0.76). For the whole genome scan, eight masked duplicates of the 133 samples, eight positive controls, and two negative controls were genotyped. Allele calls were reviewed by two technicians, masked to family structure and phenotype. Binning of all allele sizes for each marker was performed by using the FASTCLUS procedure in sas. Calls falling outside of bins and Mendelian inconsistencies (identified by using gas; ref. 23) were reviewed for laboratory error. Family relationships were analyzed by using relcheck (24). In addition, simwalk2 (25) was used to calculate the posterior probability of a genotyping error for each individual in the data set. Those genotypes with a posterior probability of being incorrect of >75% were removed from the analysis, as recommended by Douglas et al. (26). Overall, 2,755 of 54,426 (5.1%) of the genotypes were not obtained or were judged unusable. Two of 2,436 pairs of genotypes from the eight masked duplicate samples produced different calls, giving an overall error rate of 0.04%. The remaining number of Mendelian inconsistencies was 83 of 54,426 (0.15%), and these presumably represent a combination of genotyping errors and microsatellite instability. For high-resolution mapping, the 133 subjects were genotyped for four additional markers on chromosome 7 and two additional markers on chromosome 9. Additional details related to genotyping methodology can be found at www.cidr.jhmi.edu.

Ten additional tandem repeat markers were developed for family A by scanning chromosome 7 sequences using the Tandem Repeats Finder web site (ref. 27; http://c3.biomath.mssm.edu/trf). PCR primer sequences and map locations are listed in Table 3, which is published as supporting information on the PNAS web site, www.pnas.org.

Results

Identification of Strabismus Pedigrees. Most individuals with strabismus manifest the trait in infancy or early childhood. Therefore, we focused our recruiting efforts on pediatric ophthalmology patients. As described in detail in Materials and Methods, phenotyping was largely by history. The success of this approach rests on the efficient detection of ocular misalignment or uniocular visual dysfunction by family members, childcare and school personnel, and pediatricians. The limited number of strabismus treatment options (surgical realignment, patching of one eye, and orthoptic exercises) further facilitates the assignment of phenotype based on treatment history. Although this method of assigning phenotype is efficient and permits the inclusion of relatives who reside in remote locations, it is unlikely to identify the more subtle manifestations of strabismus such as microtropia and the monofixation syndrome. To date, we have recruited 209 families with a median of three to four affected members per family (Table 1).

Table 1. Number of families and number of affected individuals per family.

Number of affected subjects per family	Number of families
1	13
2	54
3	55
4	30
5	19
6	14
7	10
8	7
9	2
>10	5
Total	209

Seven families were chosen from the first 150 families for whole genome linkage analysis; the corresponding seven pedigrees are shown in Fig. 1. Family A is noteworthy for a sibship in which 8 of 14 individuals are affected. All of the affected members of this sibship were treated by a single clinician (M.P.); all eight affected individuals developed esotropia in infancy or childhood, seven of eight had some degree of hypermetropia, and four of eight underwent surgical realignment.

Fig. 1.

Seven pedigrees analyzed by whole genome scanning. The assignment of affected individuals (filled symbols) is described in Materials and Methods. Dots indicate those individuals for whom genotypes were obtained. Haplotypes are shown for family A with the 20 high-resolution chromosome 7p markers (Fig. 2B and Table 3) arranged in order with the most telomere-proximal marker (D7s3056) at the top and the most centromere-proximal marker (D7s1791) at the bottom, as listed to the left of subject 203. For each individual, the paternal haplotype is on the left and maternal haplotype is on the right. Numbers indicate the size of the PCR products in bp. The haplotype for individual 203 was reconstructed from the haplotypes of his children; the phase of the markers for individual 205 cannot be determined with the available data. The four chromosomes in individuals 203 and 204 are color coded.

Statistical Analysis of the Whole Genome Scan. Because the mode of inheritance of strabismus has not been clearly defined, two types of analyses were performed: the model-dependent logarithm of odds (LOD) score method using linkage (28, 29) and the model-free method using genehunter (25). Allele frequencies were estimated by using sib-pair (30). Both two point and multipoint parametric LOD score calculations across the whole genome were performed under a strong genetic model with a phenocopy rate of 0.0001, penetrance of 0.9, and a frequency of disease alleles of 0.001.

Two-point model-based analyses were performed by using linkage with both recessive and dominant models using the parameters described above. Markers D7S513 and D7S1802 gave a maximum single-point LOD score of 2.65 and 1.74, respectively, under the assumption of a recessive mode of inheritance. In addition, D9S1838 gave a maximum single-point LOD score of 2.31 under a dominant model.

Multipoint parametric linkage analyses were performed for all chromosomes for the data generated by the genome scan using genehunter with the parameters described above. LOD scores considering genetic heterogeneity were also computed. Results are presented in Fig. 3, which is published as supporting information on the PNAS web site. Even though the multipoint analyses did not give any homogeneous LOD scores above 1, the heterogeneity LOD score was 1.21 for D7S513 on chromosome 7 (for a recessive model) and 1.35 for D9S1826 on chromosome 9 (for a dominant model). Interestingly, the multipoint nonparametric analyses supported the location on chromosome 9 and revealed a new location on chromosome 11. The highest nonparametric LOD scores obtained were 3.87 (P = 0.006) on 9pter with marker D9S1826 and 3.88 on chromosome 11p with marker D11S2362 (P = 0.006).

Higher-Resolution Mapping. The two regions with multipoint heterogeneity LOD scores above 1.0, located on chromosomes 7 and 9, were investigated further by genotyping additional markers in all seven families. For the region on chromosome 7p, we genotyped an additional four markers: D7S2201, D7S2200, D7S2557, and D7S1791. In the initial multipoint analysis using genehunter, some of the genotyping data in the larger pedigrees was removed from the analysis because of limitations in the genehunter software. Therefore, for fine mapping we performed model-free and model-based multipoint analyses using simwalk2 (version 2.82; ref. 31), which allows us to use information from all genotyped individuals. This analysis revealed a heterogeneity LOD score of 3.02 at the marker D7S1802 under the recessive model described above (Fig. 2A), and visual inspection shows that this score is largely referable to family A. The proportion of linked families estimated by simwalk2 was 30%. Interestingly, the nonparametric analyses, which make no assumptions regarding a genetic model or any level of genetic heterogeneity, also gave a significant signal at D7S513 (P = 0.035 with statistic A).

Fig. 2.

Fine mapping of markers on chromosome 7p. (A) Eight-point LOD scores calculated with the initial set of high-resolution 7p markers. The plot shows recessive LOD scores for all families (LOD), heterogeneous recessive LOD scores for all families (HLOD), and recessive LOD scores for family A alone (LOD, family A). (B) Multipoint recessive LOD scores for family A using the 20 high-resolution 7p markers shown on the horizontal axis. As noted in the legend of Table 1, subjects 333, 430, and 431 were omitted from the LOD score calculation. For all LOD score calculations, parameters were as follows: phenocopy rate = 0.0001; penetrance = 0.9; frequency of disease alleles = 0.001.

For family A alone, the maximal eight-point recessive LOD score calculated by simwalk2 for the eight markers shown in Fig. 2 A is 3.86 at marker D7S1802 by using the parameters described above for the whole genome scans. Moreover, for family A, nonparametric linkage analysis gave a significant P value (0.0015) at the marker D7S513 with statistic A. To further refine the region of chromosome 7p showing apparent linkage in family A, 12 additional short tandem repeat markers were analyzed within a region of ≈20 Mb (Figs. 1 and 2B and Table 3). The presence of affected individuals in each of three generations in family A is suggestive of a dominant mode of inheritance. However, multiple examples are observed in which different chromosome 7 haplotypes are transmitted from one affected individual to different affected offspring, making a dominant mode of inheritance unlikely. Moreover, unaffected parents who have affected offspring invariably transmitted different chromosome 7 haplotypes to their affected vs. unaffected offspring, consistent with a model in which these unaffected parents are heterozygous for a recessive susceptibility allele. A recessive mode of inheritance in this family would imply that the strabismus phenotype is incompletely penetrant in individual 309, and that seven apparently unrelated spouses are carriers of the susceptibility allele at 7p. Fig. 2B shows that recessive multipoint LOD scores >4.0 extend over a broad region of ≈5 Mb that is centered around marker D7s3051. We note that the data may be consistent with more complex models, such as semidominant inheritance.

To examine the effect of penetrance and allele frequency, we calculated the multipoint recessive LOD score for family A by using the 20 high-resolution markers with disease allele frequency varying from 0.0001 to 0.1 and with penetrance varying from 50% to 90% (Table 2). This analysis revealed a maximal LOD score of 4.51 at marker 1911/1912 under a model in which disease allele frequency is 0.001 and penetrance is 90%. LOD scores >4.0 were obtained with disease allele frequencies spanning the range 0.1 to 0.001, and penetrance values of 60–90%.

Table 2. Multipoint recessive LOD scores for markers on distal 7p for families A-G and for family A alone.

		Penetrance, %
	Disease allele frequency	50	60	70	80	90
Families A-G	0.1	2.42	2.50	2.40	2.38	2.78
	0.01	2.63	2.62	2.54	2.64	3.19
	0.001	2.85	2.91	2.92	2.80	3.02
	0.0001	2.47	2.55	2.62	2.65	2.57
Family A	0.1	3.55	3.84	3.99	4.01	3.93
	0.01	3.84	3.98	4.11	4.24	4.35
	0.001	3.93	4.11	4.30	4.47	4.51
	0.0001	3.62	3.75	3.86	3.94	3.91

LOD scores were calculated with simwalk2 using a phenocopy rate = 0.0001. For families A-G, the heterogeneous LOD score calculation is based on the eight markers used in the initial high-resolution map (see Results). For family A alone, the calculation is based on all 20 high-resolution 7p markers (Fig. 1 and Table 3). Subject 333 and her children, 430 and 431, were not included in the LOD score calculation for family A alone because subject 333, a nonblood relative, is affected.

We also typed markers D9S150 and D9S162 on chromosome 9 for all seven pedigrees. Although a five-point dominant model-based analysis including markers D9S2157, D9S1818, D9S1826, D9S158, and D9S1838 gave a maximal multipoint LOD score of 0.8, pedigree D alone gave a LOD score of 1.70 (data not shown). Model-free analyses performed by simwalk2 gave a significant P value of 0.0017, converted to a minus log(P) of 2.75 for the statistic B, which has been shown to be the most powerful statistic for detecting linkage to a dominant trait (32). We have not performed higher-resolution mapping of the candidate region on chromosome 11p because simwalk2 gave only a marginally significant P value (0.04) with marker D11S2362.

Discussion

The principal results of this study are: (i) the demonstrated feasibility of identifying and recruiting large families in which multiple members have (or had) strabismus; (ii) the linkage in one large family of a presumptive strabismus susceptibility locus to 7p22.1 with a homogeneous LOD score of 4.51 at marker 1911/1912 under a model of recessive inheritance; and (iii) the failure to observe significant linkage to 7p in six other multiplex families, consistent with genetic heterogeneity among families. These findings suggest that expanded family recruitment and genotyping efforts are likely to reveal additional strabismus susceptibility loci.

As noted in the Introduction, the common forms of strabismus present a number of technical challenges to genetic analysis. However, strabismus also presents a number of attributes that are technically advantageous for identifying large numbers of families with multiple affected individuals, as described below. First, strabismus is common and does not affect viability or fecundity. Second, the phenotype is usually apparent in early childhood. Therefore, pedigrees often show multiple generations of affected living subjects. Moreover, relatives who do not have the major manifestations of strabismus can be readily identified because late onset is highly unlikely. Third, in technologically advanced countries, affected individuals are identified efficiently and almost invariably brought to the attention of an ophthalmologist. A reliable diagnosis can be made by any ophthalmologist using tests that are simple and safe. In untreated patients, overt ocular misalignment can also be readily documented by appearance or by examining family photographs. Fourth, both diagnostic and treatment methods have remained relatively unchanged over at least 50 years. Therefore, a diagnosis of strabismus or a history of one of the standard treatments is sufficient to establish the phenotype with high confidence. Fifth, in addition to the risk of developing amblyopia, ocular misalignment is considered a major cosmetic and social handicap. As a consequence, the parents of affected children are generally motivated to seek treatment, and our experience has been that most families with multiple affected members are also motivated to participate in research that might improve diagnosis or treatment. Taken together, these attributes should facilitate the identification and recruitment of large numbers of families with characteristics favorable for linkage analysis.

The experiments reported here represent a first step in defining the genes that predispose to the common forms of strabismus. At present, the candidate region on chromosome 7p encompasses 5–10 cM. Therefore, further fine mapping and the identification of additional families will be required to narrow the interval. It is hoped that future work will result in the ultimate identification of causally relevant genes with a resulting improvement in the understanding of this disorder.