Association of the FBXO11 gene with COME/ROM in the Minnesota COME/ROM Family Study

Abstract

Objective

The FBXO11 gene is the human homolog of the gene mutated in the novel deaf mouse mutant Jeff (Jf), a single gene model of otitis media. We have evaluated single nucleotide polymorphisms (SNPs) in the FBXO11 gene for association with chronic otitis media with effusion/recurrent otitis media (COME/ROM)

Design

A total of 13 SNPs were genotyped across the 98.7 kb of genomic DNA encompassing FBXO11. Data was analyzed for single SNP association using generalized estimating equation (GEE1) and haplotypes were evaluated using Pedigree Disequilibrium Test (PDT) methods.

Subjects

The Minnesota COME/ROM Family Study, a collection of 142 families (619 subjects) with multiple affected individuals with COME/ROM.

Results

The FBXO11 SNPs are contained in a single linkage disequilibrium (LD) haplotype block. Ten of the 13 SNPs were sufficiently polymorphic in the sample to permit analysis. In univariate genetic analysis, one SNP (rs2134056) showed nominal evidence of association to COME/ROM (P=0.017) and 2 SNPs approached significance (rs2020911, P=0.055; rs3136367, P=0.087). In multivariable analyses, including known risk factors for COME/ROM (gender, smoking exposure, daycare, lack of breastfeeding, and allergies), the evidence of independent association was reduced for each SNP (e.g., rs2134056, from P=0.017 to P=0.076). In subsequent analyses using the PDT, the association of FBXO11 SNP rs2134056 (P=0.056) with COME/ROM was confirmed. Incorporating multiple SNPs in 2- and 3-locus SNP haplotypes, those haplotypes containing rs2134056 also exhibited evidence of association of FBXO11 and COME/ROM (P-values ranging from 0.026 to 0.099).

Conclusions

We have observed evidence consistent with an association between polymorphisms in FBXO11, the human homolog of the Jeff mouse model gene, and COME/ROM.

INTRODUCTION

Chronic otitis media with effusion (OME) and recurrent otitis media (OM) are relatively common conditions, affecting 10-30% of children, they appear to be increasing over time,^1-4 and often result in hearing loss⁵ and middle ear sequelae.^{6, 7} Disease rates are highest in the preschool years, and decline as children get older.¹ These conditions are often treated with tympanostomy tubes to ventilate the middle ear, which is the most common surgical procedure performed in children in an ambulatory setting.⁸ Epidemiologic studies have shown that a family history of these conditions increase a child’s personal risk^{2, 9, 10} and twin studies demonstrated a high degree of heritability for recurrent OM and OM duration.^{11, 12} We conducted a linkage study (the Minnesota COME/ROM Family Study) which revealed. evidence for linkage to COME and ROM on chromosome 10q and suggestive evidence for linkage on chromosome 19q.¹³ Several case control studies have identified relationships between candidate genes and COME/ROM, including HLA,^{14, 15} cytokine genes,^{16, 17} and genes involved in ability to clear infectious agents implicated in OM.^{18, 19} To date, however, few detailed molecular genetic analyses of genetic associations between candidate genes and COME/ROM have been reported.

The FBXO11 gene is an intriguing candidate for association with COME/ROM. The Jeff deaf mouse mutant represents a single gene model of chronic proliferative otitis media,²⁰ characterized by the development of otitis media in pathogen-free conditions in the absence of any other inflammatory pathology. Jeff animals present with fluid and pus accumulation in the middle ear cavity, with diffuse mucosal inflammation, disruption of the Eustachian epithelium, and reduction of the lumen due to abnormal growths.²⁰ The Jeff mutation appears to be fully penetrant and has been mapped to the distal region of mouse chromosome 17.²⁰ A missense mutation has been identified in Fbxo11, a gene located on this region of mouse chromosome 17 (R.E. Hardisty-Hughes and S.D.M Brown, unpublished data). The human ortholog, FBXO11, is located on chromosome 2 and encodes a 141-residue protein that belongs to the F-box family of genes. FBXO11 is characterized by the presence of a 50-amino acid sequence, the F-box, which functions in protein-protein interactions.²¹ The FBXO11 gene therefore represents a strong candidate gene for otitis media. This manuscript examines polymorphisms in the FBXO11 gene for association in the human disease pathway for COME/ROM.

MATERIALS AND METHODS

Participants

All families and individuals in this study were recruited to participate in the genetic linkage study.¹³ Families that were excluded from the linkage study because they lacked an affected sibling pair with sufficient DNA were evaluated in this study. Study participants included families from previous studies of OM conducted by the Otitis Media Research Center at the University of Minnesota, the general public who responded to fliers posted around the Academic Health Center, and Otolaryngology clinic patients. Prospective families from previous studies were identified using OM history data collected in the original study on the participant (proband), parents and siblings. For the families recruited from the general public and clinic patients, mothers were interviewed about the OM history of all family members to determine family eligibility. Data were collected at the study visit to determine phenotype (history of chronic or recurrent OM) including findings from the otomicroscopic exam, multifrequency tympanometry, reported OM history and medical record¹³. An individual was considered affected if at least two of these sources were positive, or if one source was positive, and middle ear or tympanometric findings presumed to be definitive evidence of a history of COME/ROM (e.g. tympanosclerosis, atrophy, tympanogram with high static admittance) were present. Subjects were examined by a neurotologist, and subjects with obvious craniofacial anomalies were excluded. Subjects were not examined by a dysmorphologist. It should be noted that anatomic, morphologic and environmental factors in OM pathogenesis are highly interrelated making difficult to impose specific criteria for inclusion/exclusion. Exclusion criteria included Down syndrome, cleft palate, or adopted siblings.

One hundred forty two families (619 individuals) participated in the candidate gene study. This includes 132 families who were included in the genome scan and linkage studies¹³, and 10 families who were recruited using the same ascertainment criteria, and included only in this candidate gene study. DNA extracted from blood (94%) or buccal (6%) samples was used to assess candidate genes.

Genotyping

Thirteen (13) single nucleotide polymorphisms (SNPs) in FBXO11, selected from the dbSNP public database, were genotyped in this study. SNPs were chosen to cover the entire FBXO11 genomic sequence, including 5’ promoter region, exons, introns, and 3’-UTR and included rs2020911, rs3136367, rs3136371, rs3771285, rs3732191, rs330787, rs2651767, rs2134056, rs960106, rs2937345, rs874869, rs4952896 and rs7582252. Of these SNPs, rs7582252, rs2937345 and rs3136371 were not polymorphic and were not included in subsequent analyses.

Genotyping of the FBXO11 SNP polymorphisms was performed using the MassARRAY SNP genotyping system (Sequenom, Inc., San Diego, CA) using a standard protocol.²² Total genomic DNA was purified from whole blood samples obtained from the IRASFS subjects using PUREGENE® DNA isolation kit (Gentra, Inc., Minneapolis, MN). DNA was quantitated using standardized fluorometric readings on a Hoefer DyNA® Quant® 200 fluorometer (Hoefer Pharmacia Biotech Inc., San Francisco, CA). Each sample was diluted to a final concentration of 5 ng/μL.

Statistical Genetic Analyses

Each pedigree was examined for evidence of incorrect family relationship by using genome scan data and PREST software.²³ For each of the SNPs genotyped in this study, Mendelian inconsistencies in their genotype assignment were examined using PEDCHECK software.²⁴ Any genotypes inconsistent with Mendelian inheritance were converted to missing (9 total, 0.001%). Following pedigree and genotype correction, maximum likelihood estimates of allele frequencies were computed using the largest set of unrelated individuals. All SNPs were then tested for departures from Hardy-Weinberg proportions. With multiple SNPs genotyped in the FBXO11 gene, the haplotype block structure and linkage disequilibrium (LD) between markers was assessed. Estimates of LD (D’ and r²) were computed also using the largest set of unrelated individuals and Dprime software. For families in which no founder was genotyped, the first affected offspring was used to estimate the statistics.

Association between individual SNP and COME/ROM status was performed using a series of generalized estimating equation models.²⁵ The correlation between subjects within a pedigree was adjusted for in the analyses by assuming exchangeable correlation among siblings within a pedigree and computing the sandwich estimator of the variance. The sandwich estimator is also denoted the robust or empirical estimator of the variance and is robust to misspecification of the correlation matrix because it estimates the within pedigree correlation matrix from the first and second moments of the data. Broadly speaking this method is comparable to a logistic regression analysis with adjustment for familial correlation among family members. As such the unaffected members of the pedigrees provide allele counts that correspond to control genotype frequencies. For each SNP, the two degree of freedom overall test of genotypic association was performed and, if the overall genotypic association was significant, three individual contrasts defined by the a priori genetic models (dominant, additive and recessive) were computed. If the overall genotypic association was not significant, the a priori contrasts were examined after adjusting for the three comparisons using a Bonferroni adjustment, consistent with the Fishers protected least significant difference multiple comparison procedure. Tests of association between SNP and COME/ROM were computed adjusting for gender, number of smokers in the household, prior breastfeeding, presence of allergies and past daycare attendance.

Haplotypic associations between the COME/ROM and the FBXO11 SNPs were also assessed. Haplotypes were constructed and the analyses of the 2- and 3-marker haplotypes were completed using a GEE1 analysis as described above, except the quasi-likelihood was weighted by the probability for each possible haplo-genotype for an individual. Each individual enters into the GEE1 analysis once for each haplo-genotype possibility, weighted by the haplo-genotype probability. Thus, the weight for each individual sums to one. The weighted GEE1 analyses were completed as above using the sandwich estimator of the variance to account for the within cluster correlation.

In order to test for allelic association while accounting for potential population stratification, the Pedigree Disequilibrium Test (PDT) was employed ²⁶. The PDT was performed using both single SNP as well as 2- and 3-marker haplotypes within FBXO11. The PDT operates on the same principle as the transmission disequilibrium test (TDT): alleles transmitted to cases create genotype frequencies for cases while alleles that are not transmitted comprise allele frequencies for controls. The PDT is more powerful than the TDT and it allows analysis of transmission of alleles with all available family data. The PDT is a valid test of association even in the presence of population substructure, and maintains the analysis of transmission within families²⁶.

RESULTS

Participants

One hundred forty two families (619 individuals) participated in the candidate gene study. Of the participants in this study, 40% were classified as affected (two or more sources of data positive for COME/ROM, or definitive exam findings), and 60% were unaffected. The majority (94%) was white and non-Hispanic (99%). Risk factors were prevalent, most had been in daycare, about one-fifth reported allergies, and exposure to smokers was common (Table 1).

Table 1.

Descriptive data for the Minnesota COME/ROM Family Study

Description	value	observations
Number of pedigrees:	142	–
Total subjects	619	–
Gender (% female):	53	–
Number of smokers in the home (mean ± SD and (median))	0.75 ± 0.85 (0.0)	618
Exposed to smoking in household (% exposed)	49	618
Daycare (% attending)	52	617
Any breastfeeding (% breastfed)	57	591
Allergies (% have allergies)	22	618

Genotyping and LD structure

The FBXO11 gene consists of 23 exons spanning >98 kb of genomic sequence arranged in a tail-to-tail configuration with the MSH6 G/T mismatch repair gene, resulting in an overlapping sequence of 32 bp in their 3’ untranslated regions; see Figure 1.²⁷ Thirteen SNPs were genotyped across the 98.7 kb of genomic DNA encompassing the FBXO11 gene in 619 individuals from 142 families in the Minnesota COME/ROM Family Study. Figure 1 also shows the location and distribution of the SNPs relative to the genomic structure of the gene. FBXO11 is a complex gene and the 13 SNPs were chosen to systematically cover the genomic region with a density of 1 SNP per 7.6 kb. A large gap of 31.5 kb exists between exons 2-15 in which no SNPs were identified that were polymorphic in the study subjects. The current set of 13 SNPs is, however, contained in a single linkage disequilibrium (LD) haplotype block. Only one of the SNPs is in an exon (rs3136371), but this SNP is in the non-coding 3’-UTR. Bioinformatic analysis did not identify SNPs in other exons. After genotyping, all SNPs were evaluated for Hardy-Weinberg equilibrium and pairwise LD. SNPs with minor allele frequencies greater than 0.10 exhibited genotypic frequencies that were consistent with Hardy-Weinberg proportions. D’ values were greater than 0.95 for all SNPs with minor allele frequencies greater than 0.10, suggesting that the entire FBXO11 gene is encompassed by a single LD haplotype block (data not shown).

FIG. 1.

Genomic map of FBXO11. The grey boxes represent coding exons and the ruler along the bottom shows the relative location and spacing of genotyped SNPs in kilobases. FBXO11 and the neighboring MSH6 (MutS 6, homolog of) genes are oriented in a tail-to-tail fashion.

Single SNP association analysis was performed using general estimating equation methods. Ten of the 13 SNPs were sufficiently polymorphic in the sample (i.e., had minor allele frequency greater than 0.10) to permit analysis. In analysis of the COME/ROM phenotype summarized in Table 2, SNP rs2134056 (located in the distal portion of intron 1, Figure 1) exhibited nominal evidence of association to COME/ROM (P=0.017). Two SNPs in FBXO11 (rs2020911, P=0.055; rs3136367, P=0.087) approached significance. As these initial analyses did not adjust for recognized COME/ROM risk factors (covariates), further analyses were performed. Addition of gender as a covariate in the analysis did not significantly change the evidence for association. In an analysis that included other recognized risk factors for COME/ROM (gender, smoking exposure, daycare, breastfeeding, and allergies), evidence of association was reduced for each SNP (e.g., rs2134056, significance reduced from P=0.017 to P=0.076).

Table 2.

Single SNP Association Analysis of FBXO11 SNPs with COME/ROM using GEE1

		P-value
SNP rs #	MAF1	General Association (No covariates)2	General Association (Gender)2	General Association (All covariates)2,3
rs2020911	0.324	0.055	0.053	0.14
rs3136367	0.261	0.087	0.070	0.13
rs3771285	0.005	0.25	0.30	0.36
rs3732191	0.065	0.22	0.32	0.52
rs330787	0.381	0.53	0.54	0.56
rs2651767	0.166	0.52	0.51	0.27
rs2134056	0.231	0.017	0.018	0.076
rs960106	0.402	0.190	0.20	0.41
rs874869	0.443	0.17	0.20	0.31
rs4952896	0.254	0.13	0.12	0.16

¹MAF: minor allele frequency

²bold: P < 0.05; italics: 0.05 < P < 0.10

³Analysis with all covariates: gender, smoking exposure, daycare, breastfeeding, and allergies

Pedigree disequilibrium test (PDT) analysis

In addition to single SNP associations, there is the possibility that a combination of SNPs (i.e., a haplotype) confers increased risk for COME/ROM. Haplotype association analysis was performed with the SNP data using the Pedigree Disequilibrium Test (PDT). Table 3 shows the results of analyses with P ≤ 0.10. In single SNP analysis with PDT, the rs2134056 SNP showed some evidence of association with COME/ROM (P=0.058). Haplotype analysis was performed using a sliding window approach, so that combinations of 2- and 3-adjacent SNPs were tested sequentially. Consistent with the single SNP analysis, the 2- and 3-SNP combinations that include rs2134056 (e.g., rs2651767 and rs2134056; P=0.026) reveal evidence for association with COME/ROM.

Table 3.

Single SNP and haplotype analysis of FBXO11 SNPs using PDT

SNP rs #	1 Marker	2 Marker	3 Marker
rs2020911
rs3136367
rs3136371
rs3771285
rs3732191
rs330787
rs2651767		0.026	0.099
rs2134056	0.058			0.058
rs960106
rs2937345
rs874869
rs4952896
rs7582252

DISCUSSION

FBXO11 is a member of the FBXO subfamily of proteins that have an F-box but no recognized substrate-binding region.²¹ Although the specific function of the FBXO11 protein is unknown, a missense mutation in mouse Fbxo11 gene (R.E. Hardisty-Hughes and S.D.M Brown, unpublished data) in the Jeff mouse model of chronic proliferative otitis media, make human FBXO11 a candidate for involvement in COME/ROM. Evidence for mutation of Fbxo11 leading to otitis media in the Jeff mouse model illustrates the power of modern genetic methods to identify novel pathways, which based on biological understanding alone might not be implicated in the pathophysiology of a disorder such as COME/ROM.

Some comment seems appropriate as to the relevance of the Jeff mouse model to common COME/ROM. For example, is susceptibility to otitis media in both the Jeff mouse and the study subjects could be due to subtle craniofacial abnormalities leading to Eustachian tube dysfunction. Hearing loss in the Jeff mouse²⁰ was of a mixed nature. The raised thresholds obtained for the cochlear nerve response were beyond what would be expected for a purely conductive loss. Also some Jeff mice showed an abnormally low endocochlear potential (EP) indicating involvement of the stria vascularis in the hearing loss. Humans with middle ear disease often show sensorineural components to their hearing loss, potentially as a consequence of damage to the lateral wall of the cochlea. Also in other animal models, cochlear pathology has been seen in experimentally induced OM, which is consistent with the raised EPs seen in Jeff mice^{27, 28}.

The craniofacial abnormality in Jeff mice was evident on the C3H/HeH background (the laboratory strain on which the mutant was originally maintained, and the background used in previous studies)²⁰. When on a different genetic background (C57BL/6), the penetrance of the abnormal face phenotype diminishes and is not always coincident with the presence of otitis media (Rachel Hardisty-Hughes and Steve Brown, unpublished). For these reasons, otitis media in the Jeff mouse model does not appear to be solely due to a structural abnormality resulting in dysfunction of the Eustachian tube.

We have carried out a systematic evaluation of FBXO11 in a well characterized population of COME/ROM families from the COME/ROM Family Study. Multiple polymorphisms have been genotyped in an effort to provide a comprehensive survey of gene variation in the gene. The SNP genotyping data are consistent with the FBXO11 being contained in a single linkage disequilibrium haplotype block. Univariate, multivariate, and haplotype analysis using a variety of analytical approaches provide evidence consistent with the genetic involvement of human FBXO11 in COME/ROM. We note that the magnitude of the significance for association is not dramatic, which is consistent with the observation that the FBXO11 region was not detected in our previous genome scan for linkage. However, with a strong a priori hypothesis, these results suggest additional analyses of FBXO11 in other relevant populations.