Heritability of Lenticular Myopia in English Springer Spaniels

Abstract

Purpose.: We determined whether naturally-occurring lenticular myopia in English Springer spaniels (ESS) has a genetic component.

Methods.: Streak retinoscopy was performed on 226 related ESS 30 minutes after the onset of pharmacologic mydriasis and cycloplegia. A pedigree was constructed to determine relationships between affected offspring and parents. Estimation of heritability was done in a Bayesian analysis (facilitated by the MCMCglmm package of R) of refractive error in a model, including terms for sex and coat color. Myopia was defined as ≤−0.5 diopters (D) spherical equivalent.

Results.: The median refractive error for ESS was 0.25 D (range, −3.5 to +4.5 D). Median age was 0.2 years (range, 0.1–15 years). The prevalence of myopia in related ESS was 19% (42/226). The ESS had a strong correlation (r = 0.95) for refractive error between the two eyes. Moderate heritability was present for refractive error with a mean value of 0.29 (95% highest probability density, 0.07–0.50).

Conclusions.: The distribution of refractive error, and subsequently lenticular myopia, has a moderate genetic component in ESS. Further investigation of genes responsible for regulation of the development of refractive ocular components in canines is warranted.

Introduction

Myopia (nearsightedness) is a form of refractive error in which an image of a distant object is focused in front of rather than on the retina. Two main forms of myopia have been identified in humans: axial and refractive. Axial myopia is defined as a dysregulated elongation of the globe. Refractive myopia occurs secondary to increased corneal and/or lenticular power. While human juvenile onset myopia certainly is due to excessive axial length, not all large eyes are myopic. Myopia, more specifically, is due to an elongation of the eye that becomes independent of compensatory changes in the crystalline lens at the onset of myopia. ¹ The thinning, flattening, and loss of lens power typical of eyes that remain emmetropic over time abruptly stops at the onset of juvenile myopia. ¹ A better understanding of the regulation of the relationship between axial elongation and crystalline lens power may provide novel insights into the development of canine and human refractive error. Naturally-occurring models for lenticular based myopia are lacking, making the study of human lenticular myopia more challenging. 

While numerous experimental models of animal myopia have been reported (tree shrew, ^2–8 chick, ^9–19 guinea pig, ^20–24 marmoset, ^25–30 and primates ³¹ ), there is a dearth of reports of naturally-occurring myopia in animals. Recent reports of naturally-occurring myopia have been identified in dogs, ^32,33 guinea pigs, ³⁴ and a purpose-bred line of chickens with reduced vision. ³⁵ Dogs as a naturally-occurring model of myopia have the added benefit of large numbers of offspring, and the ability to obtain refractive data from numerous dogs due to a relatively short life span. ³³ Previous research has shown dogs, on average, are emmetropic. ³² A small number of breeds have been identified to be myopic on average, or emmetropic, but with distinct subpopulations of myopic animals. Approximately 20% of the Labrador retriever and English Springer spaniel (ESS) populations have been reported to be myopic. ³² The Labrador retriever has been shown to be affected by a vitreous chamber elongation-based myopia, which is a heritable component, with potential to serve as a model for axial myopia in humans. ^33,36  

Recently, ESS were determined to have a lenticular myopia associated with increased lens thickness and lens power. ³⁷ This naturally-occurring lenticular myopia in the ESS may provide a model for this refractive form of myopia, and for elucidating interactions between lenticular and axial development. 

The purpose of this study was to determine if there is a genetic component for naturally-occurring lenticular myopia in ESS. 

Methods

Study Population

Refractive data collected from 1996 through 2008 were available on 337 purebred English Springer spaniels from the Midwest United States. From this population, 20% of dogs were identified as being myopic in both eyes (≤−0.5 diopters [D]). We determined exact relationships between affected dogs, their littermates, and their parents (dam/sire), which resulted in construction of a large pedigree, including 226 related field trial ESS. Pedigree information was obtained from local breeders to discern ancestral connections. All procedures adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Refractive Error

All ESS had received full ophthalmic examinations using slit-lamp biomicroscopy (Kowa SL-14 portable slit lamp; Kowa Company Ltd., Tokyo, Japan) along with indirect ophthalmoscopy with a Heine binocular indirect ophthalmoscope (Heine Optotechnik, Herrshing, Germany) using a Pan Retinal 2.2 lens (Volk Optical, Inc., Mentor, OH). Any ESS with significant ophthalmic abnormalities (i.e., cataract, retinopathy) were excluded from the study. The ESS then were refracted by one of two authors (CJM, DOM) using streak retinoscopy (Welch-Allyn streak retinoscope, Model No. 18100; Welch-Allyn, Inc., Skaneateles Falls, NY) at a 67-cm working distance approximately 30 minutes after instillation of cyclopentolate (cycloplegic and mydriatic agent [Cyclogyl] or 1% tropicamide [Mydriacyl]; Alcon Laboratories, Inc., Fort Worth, TX). Mean spherical equivalents were determined by averaging the results for vertical and horizontal meridians. Astigmatism was defined as a ≥0.5 D difference between refractive errors for the vertical and horizontal meridians within one eye. 

Data Analysis

For computing the prevalence of myopia, myopia was defined as ≤−0.5 D spherical equivalent. Subsequently, hyperopia was defined as ≥+0.5 D and emmetropia as >−0.5 D, but <+0.5 D. We used a repeated measures model that treated each refractive error (D) as a repeated measure, while estimating the potential for a difference between right and left eyes. A correlation coefficient in excess of 0.95 would provide strong evidence that the components (whether genetic and/or environmental) that influence refractive error for the right eye are the same set of factors that influence refractive error for the left eye. 

Specifically, we assumed that refractive error was a normally distributed random variable, such that the i-th observation, y_i, can be modeled as Display Formula where μ is a constant common to all animals and x_i is a set of recorded nongenetic contributors to refractive error (i.e., sex, coat color, and right or left eye measurement). We did not include litter size in this analysis, as some of the offspring information was incomplete and exact litter size was not known in all cases. We defined β_i as the unknown effects associated with these nongenetic contributors, a_i the additive genetic contribution to refractive error, pe_i the unique environmental contribution to refractive error, and e_i the unknown residual variances. In addition a_i was assumed to be sampled from a multivariate normal density, with mean zero and covariance Display Formula , with Display Formula being the numerator relationship matrix, the pe_i to be sampled from a multivariate normal density, independent of a_i, with mean zero and variance of the unique environmental contribution Display Formula and the residuals to follow from a normal density, with mean zero and variance of the temporary or unknown environmental contribution Display Formula . The objective was to estimate the unknown vector β_i, along with the unknown variances σ²_a, σ²_pe, and σ²_e. With that, the narrow sense heritability of refractive error also can be estimated (h² = σ²_a/[σ²_a + σ²_pe + σ²_e]; genetic variance as a proportion of total variance) and the phenotypic changes that we observed were affected by genes or genetic factors (single allele or multiple alleles) and/or by nongenetic factors (residual, environment, and so forth). The correlation of refractive error between the two eyes of the same animal also can be estimated (r = [σ²_a + σ²_pe]/[σ²_a + σ²_pe + σ²_e]).  

A Bayesian framework, a strategy of considerable power and plasticity, was used to evaluate these unknown values. ³⁸ The public domain package MCMCglmm, ³⁹ available through the language R, ⁴⁰ allowed us to analyze our large pedigree, which included multiple mates and consanguinity. The prior distributions for the putative fixed effects were independent normal densities, with null means and variances of e ¹⁰; that is, vague or “flat.” The prior distribution for the variance components (σ ² _a , σ ² _pe , and σ ² _e ) was assumed to be an inverse-Wishart density, with a small degree of belief (i.e., nu = 0.002), also representing a vague or uninformed prior. As implied by the use of the MCMCglmm package, estimates of the posterior density for the unknown parameters were generated through a Monte Carlo Markov chain. No estimations were performed to investigate a dominant model for broad sense heritability. Each variable (coat color, age at refraction, sex, and refractive error) was evaluated with 3 chains, each starting at disperse values for the unknown variances. Convergence of the chains was examined through trace plots making use of the R package coda ⁴¹ and computation of the Gelman-Rubin statistic. ⁴² Each chain was run a total of 500,000 rounds with a burn-in of 50,000 rounds and a thinning interval of 100 (creating a single chain sample of 4500 values). In this scenario, all variables had chains that mixed well (with all Gelman-Rubin statistics between 1.00 and 1.03), with the absolute value of all autocorrelations for parameters below 0.07. It is important to note that the Gelman-Rubin statistics for all the unknown parameters of the model had the value of 1.0; a value indicating no significant difference between the three Gibbs sampling chains. Moreover, a test of the auto-correlations between subsequent values in the thinned samples, for all unknowns, was well below the threshold of 0.07 (the absolute value being below 0.02). Accordingly, we combined the three chains into one, providing a Gibbs chain of 13,500 values from which to evaluate the posterior density of the unknowns. 

Results

Refractive Error Distribution

Our study population consisted of 226 ESS with a median age of 0.2 years (range, 0.1–15 years). There were 123 females and 103 males. The ESS coat color had three variations: black/white (62), liver/white (156), and tricolor (1). Coat color was unknown in 7 cases. The median refractive error for ESS in this study was 0.25 D (range, −3.50 to +4.50 D). Figure 1 illustrates the distribution of spherical equivalent refractive errors (mean of the two eyes) of the 226 ESS dogs in the sample. In our ESS population, 19% (42/226) of dogs were myopic, whereas 27% (62/226) of dogs were hyperopic. The majority of dogs (54%, 122/226) were emmetropic. The prevalence of astigmatism within the ESS population was 1.3% (3/226). A linear regression of refractive error and age was performed, and there was no significant relationship observed between refractive error and age (r ² = 0.07). Figure 2 shows the pedigree of all ESS in the study, which included 33 litters. 

Figure 1

 

Histogram of the distribution of refractive state of 226 related ESS (mean spherical equivalent of the two eyes).

Figure 2

 

Pedigree showing interconnections between the refracted 226 related ESS and their ancestors.

Heritability Analysis

Table 1 presents estimates of the nongenetic components of the model, providing insight into the contribution of coat color, sex, or the eye observed for refractive error. Because all of the 95% intervals overlapped zero, we can conclude that all of the terms had no significant impact on refractive error, including the intercept. The fact that the intercept overlapped zero indicated that this sample of dogs had a mean refractive error of emmetropia. 

Table 1

 

Estimates of the Nongenetic Components of the Model, Including Mean and Standard Deviation of Coat Color Contribution, Laterality, and Sex to Refractive Error, Along With 95% Highest Probability Density for the Gibbs Samples in the Three Combined Chains

	Mean	SD	2.5% Quantile	97.5% Quantile
Intercept	−0.109	0.213	−0.537	0.307
Color, BW	−0.380	0.972	−2.275	1.579
Color, LW	0.130	0.172	−0.205	0.462
Side, OD/OS	0.037	0.024	−0.011	0.084
Sex, M/F	−0.015	0.141	−0.288	0.263

 Since all of the 95% intervals overlap zero, we can conclude that all of the nongenetic components evaluated have no significant impact on refractive error in the ESS. BW, black/white; LW, liver/white; M, male; F, female.

Table 2 presents mean estimates of the unknown variances, while the more easily interpreted statistics of heritability and interocular correlation are noted in Table 3. Specifically, we found that the heritability of refractive error, as taken in this “repeated measures” model, had a mean value of 0.29 (95% probability density ranged from 0.07–0.5). The 95% probability density was conceptually similar to confidence interval for classical statistics, in that we were 95% confidant that heritability value for refractive error was between 0.07 and 0.5. A value of 0.29 suggested a moderate genetic component (moderate heritability) along with environmental influences contributing to refractive error in the ESS. The correlation between refractive error observations from the same dog was near unity (mean estimate of interocular correlation being 0.95), and refractive error value in one eye was nearly identical to the value in the contralateral eye. 

Table 2

 

The Mean, Standard Deviation, and 95% Highest Probability Density for the Gibbs Samples of Additive Genetic, Permanent Environment, and Residual Variances in the Three Combined Chains

	Mean	SD	2.5% Quantile	97.5% Quantile
Additive genetic	0.352	0.153	0.078	0.682
Unique environment	0.774	0.128	0.548	1.055
Residual variance	0.063	0.006	0.052	0.076

 Additive genetic component contributes approximately 1/3 to refractive error, whereas other environmental variances provide over 2/3 to refractive error.

Table 3

 

Mean, Standard Deviation, and 95% Highest Probability Density for the Gibbs Samples of the Heritability and Interocular Correlation of Refractive Error (D) in the Three Combined Chains

	Mean	SD	2.5% Quantile	97.5% Quantile
Heritability	0.291	0.111	0.069	0.506
Interocular correlation	0.947	0.007	0.931	0.960

 The Gibbs samples of the estimated heritability of refractive error in ESS show a moderate heritability, and the high interocular correlation shows that refractive error for the right eye is similar to the left eye.

Approximately 1/3 of the variance in refractive error in the ESS population could be accounted for by genetic effects and 2/3 by nongenetic (residual/environmental) influences. The additional variables, including coat color, age at refraction, and sex, did not appear to influence refractive error. The inbreeding coefficient was calculated to be 0.031 or 3.1% (minimum = 0.0001 and maximum = 0.234) for the population of related dogs. 

Discussion

To our knowledge, this is the first report estimating heritability in a naturally-occurring animal model of lenticular myopia. The lenticular form of myopia appears to have a moderate genetic component. In this study, the observed phenotypic outcome of refractive error in the ESS was approximately due to 1/3 genetic factors (heritability) and 2/3 from nongenetic factors (residual or unique environmental influences). A previous report of narrow sense heritability in Labrador retrievers exhibiting axial myopia determined that equal contributions were from genetic, and environmental or unknown factors. ³³ The difference in the estimated heritability for Labrador retrievers and ESS could be influenced by multiple factors. In the Labrador retriever study, the environment for each animal was very similar, since all dogs included in the study were of a single-source origin with similar litter sizes. In contrast, the ESS population in this study was obtained from various breeders with different breeding facilities, litter sizes, and rearing practices. The environment at one facility may be very different from the other, which could result in a potential increase in the nongenetic component influencing the observed phenotypic change, resulting in lenticular myopia in these dogs. 

Refractive error often has been overlooked as an explanation for poor performance in working dogs, whereas historically poor outcomes were deemed primarily to stem from behavior issues. Previous studies have shown that optical defocus affected the visual acuity in beagles ⁴³ and performance in field trial Labrador retrievers. ⁴⁴ In field trials, imposition of myopia (placing positive contact lenses) on Labrador retrievers has been shown to increase their retrieval time, thereby impacting their overall performance. ⁴⁴ Retinoscopy is an important diagnostic modality, which still is underutilized in veterinary ophthalmology. The correction of ametropias (placement of prescription contact lenses or Doggles [Doggles LLC, Diamond Springs, CA]), specifically myopia, would improve the performance in field trials of ESS or other working dogs. 

Lenticular myopia in ESS appears to be heritable. The recent advancement in canine genome mapping provides an opportunity to investigate a possible genetic component associated with the regulation of axial elongation and crystalline lens power in myopia. Despite this potential, there are some important limitations to the use of ESS as a naturally-occurring model of human myopia. One is that the myopia is lenticular and not the predominant axial form in humans. The second is that the young age of the ESS dogs in this sample suggests the possibility that their myopia is congenital rather than developmental, as is more common in children. Future longitudinal studies of ESS dogs are needed to investigate the latter possibility. 

Acknowledgments

Supported by a University of Illinois Graduate School Travel Grant, the Magrane Ophthalmology Fund, National Science Foundation Office of Emerging Frontiers in Research and Innovation Grant 0937847, an unrestricted grant from Research to Prevent Blindness (UC Davis), and Pat Stoffers and family (in memory of Sophie). 

Disclosure: M.A. Kubai, None; A.L. Labelle, None; R.E. Hamor, None; D.O. Mutti, None; T.R. Famula, None; C.J. Murphy, None