Genomewide Linkage Analysis to Presbycusis in the Framingham Heart Study

Objective  To identify chromosomal regions that show evidence of linkage to age-associated hearing impairment (presbycusis) in humans.

Design  We evaluated the genetic linkage between quantitative measures from audiometric examinations and markers from a genomewide scan in a population-based sample ascertained without respect to hearing status.

Participants  Audiometric examinations were conducted on 2263 original cohort members and 2217 offspring cohort members of the National Heart, Lung, and Blood Institute's Framingham Heart Study. Of these, 1789 individuals were members of 328 extended pedigrees used for linkage analysis. The outcome traits for linkage analysis were pure-tone average at medium (0.5, 1.0, and 2.0 kHz) and low (0.25, 0.5, and 1.0 kHz) frequencies adjusted for cohort, sex, age, age squared, and age cubed.

Results  We found heritability (proportion of variance due to genes) of age-adjusted pure-tone average at medium and low frequencies to be 0.38 and 0.31, respectively. Genomewide linkage analysis identified several locations with suggestive evidence of linkage. Of particular interest are the regions 11p (maximum multipoint logarithm of odds [MLOD], 1.57), 11q13.5 (MLOD, 2.10), and 14q (MLOD, 1.55), which overlap with genes known to cause congenital deafness.

Conclusions  There is evidence that genetic and environmental factors contribute to hearing loss in the mature human population. Several of the chromosomal locations identified overlap with loci known to cause congenital hearing loss. Further studies are needed to determine whether the same genes cause presbycusis and congenital hearing loss.

TWENTY-FIVE MILLION Americans are estimated to have significant hearing loss, and the prevalence increases with age.¹ Studies in human populations^2-4 and animal models⁵ indicate that genetic and environmental risk factors play important roles in the development and progression of age-associated hearing impairment (presbycusis). The heritability of self-reported hearing loss was 0.40 in a study of elderly Danish twins.³ In the Framingham cohort, heritability of presbycusis phenotypes was estimated to be 0.35 to 0.55 based on findings from audiometric examinations.² Genetic factors known to affect presbycusis include mitochondrial deletions^6-8 and the age-related hearing loss (ahl) locus on mouse chromosome 10.⁹ In contrast, more is known about the genes that cause congenital deafness; approximately 50 genes have been mapped, and most are inherited in a mendelian fashion.

Pure-tone averages (PTAs)—thresholds averaged across distinct frequencies (low, medium, and high)—provide a quantitative measure of hearing. Examination of the PTA pattern across different frequencies allows definition of several clinical phenotypes.¹⁰ Sensory presbycusis is the most common and is typified by a predominantly high-frequency loss. Metabolic presbycusis, also called the strial form, is less common and is characterized by a relatively flat loss across the low-frequency spectrum with variable degrees of high-frequency hearing loss. For the study described herein, we focused on the low and medium frequencies and used PTA as a quantitative measure of hearing.

We conducted a genomewide scan to localize chromosomal regions that harbor genes that affect age-related hearing loss. The linkage analysis was performed in the National Heart, Lung, and Blood Institute's Framingham Heart Study, a large, community-based, longitudinal study. At the time of enrollment, participants were ascertained without respect to hearing status and, therefore, are representative of the community in which they live.¹¹ The genome scan identified several chromosomal locations that show suggestive evidence of linkage with presbycusis, including regions that overlap with known congenital deafness genes, including genes for Usher syndrome and recessive nonsydromic deafness.

Individuals studied are participants in the Framingham Heart Study, a longitudinal study established in 1948 with the enrollment of 5209 men and women aged 28 to 62 years residing in Framingham, Mass.¹¹ Beginning in 1971, 5124 offspring of the original participants and their spouses were recruited.¹² Between 1973 and 1975, hearing examinations were conducted on 2293 members of the original cohort, and between 1995 and 1999, identical examinations were conducted on 2262 members of the offspring cohort. All participants gave informed consent, and the study protocol was approved by the institutional review boards at Boston Medical Center and the University of Washington.

Standard pure-tone audiograms were obtained on all participants using environments and methods meeting American National Standards Institute standards. The behavioral audiometric thresholds were averaged for the low and middle frequencies for ease of analysis. The low-frequency PTA summarized the thresholds obtained at 0.25, 0.5, and 1.0 kHz. The mid-frequency PTA consisted of the thresholds at 0.5, 1.0, and 2.0 kHz.

Exclusion criteria for the linkage study included individuals with known nonpresbycusic otologic conditions, such as chronic infection, sudden deafness, or Meniere disease. Noise exposure history or an audiogram compatible with noise-induced hearing loss was not an exclusion criterion. Participants with a PTA difference between ears greater than 21 dB were excluded from analysis to further exclude nonpresbycusic causes of hearing loss. There was no age exclusion in the analysis, and the youngest individual included was 32 years (Table 1). Young individuals with normal hearing provide little information for linkage analysis when adjusting for age, but deviations from normal hearing in the third and fourth decades provide information about the early stages of presbycusis.

Although not initially intended as a family study, the original Framingham cohort contained numerous siblings. After recruitment of the offspring generation, extended pedigrees among the Framingham participants were constructed. These extended pedigrees consist of 2 (or occasionally 3) generations of Framingham participants and may contain cousin and avuncular pairs in addition to nuclear families. The largest 328 families were used for linkage analysis in this study. These families include 1789 individuals with hearing data. The extended pedigrees included 1205 parent-offspring pairs, 838 sibling pairs, 306 grandparent-grandchild pairs, and 455 first cousin pairs with phenotypic data.

DNA was extracted from whole blood or buffy coat specimens.^13,14 A 10-cM genome scan (marker set 8A, average heterozygosity, 0.77) was performed on available members of the largest Framingham Heart Study families by the Mammalian Genotyping Service laboratory at the Marshfield Clinic (Marshfield, Wis). Family relationships were verified based on all available markers using the sib_kin program of the ASPEX package (ftp://lahmed.stanford.edu/pub/aspex/index.html). Mendelian inconsistencies were detected and eliminated using the GENTEST program, which was a precursor to the INFER program in the PEDSYS package (http://www.sfbr.org/sfbr/public/software/software.html).

Assuming that genetic contributions to hearing loss have bilateral effects and that unilateral hearing loss may be reflective of noise exposure, we selected the measurement from the better ear for linkage analysis. Regression analysis was used to adjust for age, age squared, and age cubed. Regression analyses were conducted separately for men and women and for original and offspring cohort members. The statistical methods underlying the linkage analysis performed herein are based on the assumption that the outcome phenotype is normally distributed. As is the case with many clinical measures, the quantitative measures of hearing loss—PTA at low and medium frequencies—deviate from a normal distribution to some degree. Because of the skewness and kurtosis of the standardized residuals (Table 2) and the potential sensitivity of the linkage analysis methods to such deviations from normality,¹⁵ normalized residuals based on ranks were also computed. To examine the robustness of the logarithm of odds (LOD) scores reported herein, standardized residuals and normalized standardized residuals (based on rank) were analyzed.

Two-point and multipoint quantitative trait variance component linkage analyses were conducted on standardized residual PTAs and normalized residual PTAs at low and medium frequencies in 328 extended pedigrees using the Sequential Oligogenic Linkage Analysis Routines (SOLAR) package.¹⁶ The pedigree-based approach of SOLAR is more powerful than sib-pair analysis when data on extended families are available. Linkage is assessed by fitting a polygenic model that does not incorporate genetic marker information and comparing it to models that incorporate genotype data at a specific marker (2-point analysis) or across a chromosome (multipoint analysis). Heritability estimates are obtained from the polygenic model. The log (base 10) of the ratio of the likelihoods of the polygenic and marker-specific models is a LOD score, the traditional measure of genetic linkage. In contrast to parametric linkage analysis, negative LOD scores cannot be obtained with a variance component approach. A variance component LOD score of 0 indicates that there is no evidence of linkage. It has been suggested that for allele-sharing linkage methods, a LOD score greater than 3.6 is significant evidence of linkage, whereas LOD scores between 2.2 and 3.6 are suggestive evidence of linkage.¹⁷ However, LOD scores lower than 2.2 may warrant further investigation,¹⁷ and, hence, all chromosomal regions yielding LOD scores greater than 1.5 are reported. Power computations based on the specific structure of the Framingham pedigrees indicate that there is 80% power to detect a LOD score greater than 2.0 for a locus that accounts for 20% of the variability of a phenotype with moderate heritability, such as age-related hearing loss.¹⁸

Population information and clinical measurements for PTAs based on all the members of the Framingham study for whom hearing test results are available are given in Table 1. The mean age at hearing examination is approximately 69 years for the original cohort and 59 years for the offspring cohort. There are more women in the study primarily because of the increased mortality rate for men in this age group. In the original cohort, women tended to have better hearing than men at the medium frequencies, despite the women being slightly older. For the offspring participants, there was no age difference between the sexes, and women had significantly better hearing at the medium and low frequencies.

Heritability estimates calculated for residual PTA at the low and medium frequencies ranged from 0.31 to 0.38 (Table 2). The estimates were slightly higher for the normalized residuals for these traits (0.36-0.45). Thus, there is substantial evidence that hearing loss in this cohort is heritable.

Linkage analysis revealed 6 chromosomal regions with multipoint LOD scores greater than 1.5 (Table 3). Three distinct regions on chromosome 11 (2, 79, and 143 cM) showed evidence of linkage (Figure 1 and Figure 2). The 3 other regions are on chromosome 10 (171 cM), chromosome 14 (126 cM), and chromosome 18 (116 cM). Given the high correlation between the measurements in the better ear for PTA at the medium and low frequencies (r = .92; P<.001), it is not surprising that there was correspondence between the areas of linkage for the 2 traits.

To examine the robustness of the LOD scores to deviations from normality, standardized residual and normalized standardized residual PTAs were analyzed. In general, the linkage peaks co-localized for the 2 trait definitions, although the magnitude of the peaks differed. There was not a consistent effect of adjustment. The use of normalized residual PTAs resulted in the increase of some LOD scores and the decrease of others. Although a reduction in LOD score when normalized residual PTAs are analyzed may indicate a type I error (false positive), it could also indicate an overadjustment of the data, thereby lessening the power to detect a genetic effect. Given the exploratory nature of the adjustment for nonnormality, we consider that any regions yielding a LOD score greater than 1.5 and, in particular, the 4 regions with a LOD score greater than 2.0 are interesting regions to examine further.

This article, to our knowledge, is the first to describe the results of a genomewide scan in humans to identify chromosomal loci that predispose individuals to age-associated hearing impairment (presbycusis). The chromosomal locations we identified include 10q, 11p, 11q, 14q, and 18q. Several of these locations contain genes that are known to cause congenital forms of deafness, suggesting that the same genes may be involved in congenital and age-associated hearing loss.

We found no evidence of linkage to the murine ahl locus, a major gene responsible for age-related hearing loss in mice.⁵ The murine ahl locus is approximately 30 cM from the p-ter on mouse chromosome 10 and was first identified in C57BL/6J mice and subsequently found in 10 strains of inbred mice.⁵ The segregation of age-related hearing loss in relation to elevated auditory-evoked brainstem response thresholds is consistent with a recessive, single-gene trait. The location of ahl on mouse chromosome 10 is syntenic to human chromosomeband 10q21-q22; however, this is not the same region where we found linkage in humans (chromosome band 10q26).

Four of the 6 chromosomal loci overlapped with genes or chromosomal loci known to be associated with deafness in humans. Of these, 3 overlap with loci implicated in the Usher syndromes, a collection of disorders characterized by varying degrees of hearing loss and retinitis pigmentosa.¹⁹

On chromosome band 11q13.5, the peak of the multipoint analysis (79 cM) coincided with the location of the myosin 7A (myo7a) gene. Mutations in myo7a have been shown to cause several different phenotypes, including Usher syndrome type 1B²⁰ and nonsyndromic recessive deafness (DFNB2).^21,22 The basis of the variable phenotype seen with myo7a mutations is unknown. However, this variability suggests a potential relationship with age-associated hearing impairment. It is possible that alternative mutations in the gene result in less severe phenotype or that carriers of mutations that result in nonsyndromic recessive deafness may exhibit some hearing loss at an advanced age.

The location at 11p (2 cM) overlaps with the Usher syndrome type 1C locus (Acadian type), which has been mapped to a region slightly distal to the maximal multipoint LOD score seen in our genome scan. Mutations in the harmonin gene have been identified as the cause of Usher syndrome type 1C.²³ Harmonin is expressed in sensory hair cells of the inner ear and contains a PDZ domain, a protein-protein interaction domain that binds to specific C-terminal sequences of membrane proteins and/or to other PDZ domains.

The chromosome arm 14q location (126 cM from p-ter) overlaps with the Usher syndrome type 1A gene. This disorder was originally mapped in French families originating in the Poitou region in western France²⁴; however, the specific gene causing this disease has yet to be cloned.

The location at 11q25 overlaps with the DFNB20 location²⁵ and was identified in a consanguineous family originating from Pakistan. The deafness susceptibility gene at this location has not yet been identified. The other 2 locations, 10q and 18q, do not overlap with any known deafness genes.

Although the LOD scores generated in this study are suggestive, they do not meet the criterion for genomewide significance, which is set at 3.6 to account for multiple comparisons when examining genetic markers across the entire genome. One reason for this may be the limited power to detect genes for a complex trait, such as hearing loss. Despite the large sample size, this study had adequate power to detect only genes that have a substantial effect on presbycusis. Comparison of these results with other linkage analyses of age-related hearing loss will be an important step to confirm the chromosomal regions described herein.

In conclusion, we report, to our knowledge, the first genomewide linkage analysis for quantitative measures of hearing in an adult population. Our results indicate that the PTAs in the low and medium frequencies are heritable. We further hypothesize, based on the location of suggestive linkage peaks, that genes implicated in congenital hearing loss and Usher syndrome may play a role in presbycusis, and we will conduct additional work to address this hypothesis.

Corresponding author and reprints: Anita L. DeStefano, PhD, Department of Biostatistics, Boston University School of Public Health, 715 Albany St, T-E418, Boston, MA 02118 (e-mail: adestef@bu.edu).

Accepted for publication July 26, 2002.

This work was supported by grant R01-DC01525 from the National Institute on Deafness and Other Communication Disorders, Bethesda, Md, and by the Virginia Merrill Bloedel Hearing Research Center at the University of Washington.