While there is fairly strong evidence that otitis media has a strong genetic component, to our knowledge there has been only one prior family-based linkage study of otitis media. This first study provided evidence of linkage of COME/ROM to chromosome 10q26.3 at marker D10S212 and to chromosome 19q13.43 at marker D19S254 [20
]. We report here on our genome-wide linkage scan for risk genes for otitis media. In our study, we chose to focus on severe disease, collecting affected sib pair families where at least two full siblings had undergone tympanostomy tube insertion. By narrowing the disease definition, we hoped to identify a more severe form of OM that is perhaps due to a small number of loci. This approach has proven remarkably successful in several complex diseases, such as breast cancer and Alzheimer's disease, where subdivision according to age of onset contributed to the mapping and cloning disease genes [40
]. Additionally, it has been shown that there are statistical advantages in using a "narrow-phenotype" approach [42
], because the power of an affected sib pair approach increases as the population prevalence of the trait decreases.
The rate of tympanostomy tube insertion in second siblings may be increased due to the parents' wishes to have the procedure performed earlier or for less significant disease than the first child because of prior positive experiences in the first sibling. This could potentially make it difficult to separate the genetic effect of OM from the parents' influence on the treatment. However, we have reason to believe such an effect, if it exists, will be small in our sample. Firstly, the majority of our sample was treated at the Children's Hospital of Pittsburgh, where our criteria for insertion for tubes are consistent and very stringent. Secondly, at entry we obtained, for each subject, a history of middle-ear disease prior to tube insertion to assess the subject's eligibility.
Since the majority of our families were Caucasian, we present two linkage analyses, one where we analyzed only the 403 Caucasian families by themselves, and a combined analysis where we jointly analyzed both the 403 Caucasian and 26 African American families (Table and Figure ). The Caucasian-only analyses generate a strong linkage signal on chromosome 17q12, as well as four other potentially interesting peaks (10q22.3, 7q33, 6p25.1, 4p15.2). The combined analyses strengthen the evidence for the 10q22.3 peak. Our scan does not provide evidence for linkage in the previously reported regions of 10q26.3 and 19q13.43 [20
The very tip of our 17q12 linkage peak occurs in AP2B1
(adaptor-related protein complex 2, beta 1 subunit), which plays a role in Nef-mediated CD8 down-regulation [43
], and children with recurrent otitis media had low numbers of CD8+-producing IFN g cells in adenoids [44
]. However, the chromosome 17 peak also contains a cluster of CCL
(chemokine C-C motif ligand) genes, several of which were highlighted as possible candidates by the GRAIL analyses. CCL5
, also known as RANTES
, is 18 kb from the linkage peak, and has been previously associated with otitis media [45
is an eosinophil chemoattactrant that is thought to play a role in the accumulation of eosinophils often observed in middle ear effusions of OM with allergy.
The combined 10q22.3 linkage peak is 1.2 Mb from a previously implicated candidate gene SFTPA2
. This peak also contains a strongly associated SNP rs1437803 (Additional file 2
, P-value 0.0005, rank 5), which is 513 kb from SFTPA2
. The human surfactant protein A (SP-A) is expressed in the Eustachian tube, plays a role in innate host defense, upregulates phagocytosis of many OM risk pathogens (including Streptococus pneumoniae, Haemophilus influenzae
, and respiratory syncytial virus), and consists of two very similar functional genes SFTPA1
located 5 kb apart on chromosome 10. Ramet et al [53
] reported that the frequency of specific SP-A haplotypes and genotypes differ between children who experience their first episode of acute otitis before age 6 months and the general population. Pettigrew et al [54
] also found that polymorphisms within the SP-A loci were protective for otitis media among white infants at risk for asthma.
In the remaining linkage peaks, the majority of the genes highlighted by GRAIL as possible candidates do not have prior evidence of involvement with OM.
While none of the association results (Additional file 2
) are significant after correction for multiple testing, the level of evidence required is less for previously implicated genes. Thus, in addition to the associated SNP 513 kb from SFTPA2
mentioned above, it is noteworthy that we have associated SNPs 48 kb from interferon-γ (IFNG) on chromosome 12. Genetic variants at IFNG
are associated with risk for otitis media in infants infected with the respiratory syncytial virus (RSV) [55
]. When RSV-induced and non-RSV-induced otitis media cases were compared, significantly higher levels of IFNG were found in the RSV-induced cases [56
]. We also observed two associated SNPs 870 kb from tumor necrosis factor (TNF
) on chromosome 6. A TNF
-308 polymorphism was associated with both otitis media susceptibility and placement of tympanostomy tubes [57
]. Variants in the promoter region of TNF
were associated with being otitis-prone [58
The gamete competition test of association used here relies on transmission distortion within families, and so should detect association signals at larger distances than the conventional case/control association test in unrelateds. Even so, the proximity of some association signals to previously implicated candidate genes mentioned above may not be that meaningful, as these regions contain many other plausible candidate genes.