Association of COL1A1 and TGFB1 Polymorphisms with Otosclerosis in a Tunisian Population


Ayda Khalfallah, Equipe Procédés de Criblages Moléculaires et Cellulaires, Laboratoire de Microorganismes et de Biomolécules, Centre de Biotechnologie de Sfax, Route Sidi Mansour Km 6, Sfax, Tunisie. Tel: +216 99 827 739; Fax: +216 74 875 818; E-mail:


Otosclerosis is a condition characterized by an abnormal bone metabolism in the otic capsule, resulting in conductive and/or sensorineural hearing loss. Otosclerosis is a common disorder in which genes play an important role. Case-control association studies have implicated several genes in the abnormal bone metabolism associated with otosclerosis: COL1A1, TGFB1, BMP2, and BMP4. To investigate the association of these genes with otosclerosis in the Tunisian population, we examined nine single nucleotide polymorphisms (SNPs) in 159 unrelated otosclerosis patients and 155 unrelated controls. We found an association of rs11327935 in COL1A1 with otosclerosis that was shown to be sex specific. The coding polymorphism T263I in TGFB1 was also associated with otosclerosis in the Tunisian population. The effect sizes of both the associations were consistent with previous studies, as the same effect was found in all cases. The association of BMP2 and BMP4 was not significant. However, a trend towards association was found for the BMP4 gene that was consistent with earlier reports. In conclusion, this study replicates and strengthens the evidence for association between polymorphisms of COL1A1 and TGFB1 in the genetic aetiology of otosclerosis.


Otosclerosis is a disease caused by an abnormal remodelling in the otic capsule, characterized by alternating phases of bone resorption and formation. It leads to slowly progressive conductive or mixed hearing loss (HL). The incidence of otosclerosis is much lower in males than females that may be attributed to hormonal factors. This disease is a frequent cause of HL in Europe with frequency of 0.3%–0.4% (Declau et al., 2001). The prevalence of otosclerosis in Tunisia was even higher than in Europe with an estimation of 0.4%–0.8% (Ben Arab et al., 2001). The Tunisians, a North African Mediterranean population, originated from the admixture of several ethnic populations (Berbers, Phoenicians, Carthaginians, Romans, Vandals, Byzantines, and Arabs). Otosclerosis is more frequent in towns on the coast of the Mediterranean sea. These towns were very open to Mediterranean (European) influences.

Many theories about the aetiology of otosclerosis have been postulated. However, the origin of otosclerosis remains unclear. Environmental factors such as measles virus infection have been implicated as the causative agent (Niedermeyer et al., 2007). Karosi et al. (2007) reported increased expression of measles virus receptor CD46 in otosclerotic lesions. Furthermore, four additional splice variants were found only in otosclerotic patients (Karosi et al., 2008).

Genetic investigations in large families segregating autosomal dominant otosclerosis identified seven loci (OTSC1–5, OTSC7, OTSC8) (Tomek et al., 1998; Van Den Bogaert et al., 2001; Chen et al., 2002; Van Den Bogaert et al., 2004; Brownstein et al., 2006; Thys et al., 2007b; Bel Hadj Ali et al., 2008), but none of the corresponding genes has been found. However, a recent study suggests the implication of the T-cell receptor β locus as the causative gene in the OTSC2 region (Schrauwen et al., 2010). Unlike the monogenic form of otosclerosis, sporadic cases without familial background are very common. Candidate gene studies have shown an association of otosclerosis with polymorphisms in COL1A1 (Collagen Type 1 alpha 1) (Chen et al., 2007), TGFB1 (transforming growth factor-{beta}1) (Thys et al., 2007a), BMP2, BMP4 (Schrauwen et al., 2008), ACE, and AGT (Imauchi et al., 2008). However, the association of COL1A1, ACE, and AGT is conflicting (Rodriguez et al., 2004; Schrauwen et al., 2009c). A recent genome wide association study identified the Reelin gene (RELN) as being associated with otosclerosis (Schrauwen et al., 2009a). The genetic association of the RELN seems to be the strongest and the most convincing association until now, as shown by meta-analysis performed in eight different populations with European and non-European origin (Schrauwen et al., 2009a; Schrauwen et al., 2009b; Khalfallah et al., 2010). In the current study, we performed a replication study of the positively associated genes (COL1A1, BMP1, BMP4, and TGFB1) with otosclerosis in the Tunisian population.

Materials and Methods

Population and Clinical Diagnosis

The study population consists of 159 unrelated otosclerosis patients (115 females and 44 males) aged between 22 and 77 (mean age 45.5 years). The control group consists of 155 unrelated subjects (104 females and 51 males) aged between 50 and 87 (mean age 57.5 years). Both groups were matched according to sex and ethnicity. Patients and controls were selected by the otolaryngology department of Sfax and Mahdia hospitals in Tunisia. Pure-tone audiometery was performed on both groups with air conduction at 125, 250, 500, 1000, 4000, and 8000 Hz and bone conduction at 250, 500, 1000, and 4000 Hz. HL was conductive bilateral in the majority of patients (138 cases, 87%), and unilateral only in 13%. Standard clinical immittance measurements are used to confirm (or exclude) otosclerosis, the typical pattern being a normally shaped tympanogram and absent stapedial reflexes. Temporal bone computed tomography was performed as complimentary examination in 17 patients. Otosclerosis was finally confirmed during stapedectomy.


DNA was extracted from EDTA-anticoagulated blood using a standard phenol extraction technique. Nine SNPs from the TGFB1, BMP2, BMP4, and COL1A1 genes, including the SNPs that were associated with otosclerosis in previous studies, were analysed (Chen et al., 2007; Thys et al., 2007a; Schrauwen et al., 2008) (Table 1). The genotyping of SNPs rs1800472, rs3178250, and rs17563 from TGFB1, BMP2, and BMP4, respectively, was done by Kbioscience (Hoddesdon, UK). To type the SNPs rs1107946, rs2269336, and rs1800012 from COL1A1 and rs8179181 from TGFB1, an economical and reliable polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) method was employed using different restriction enzymes (Table 2). SNP rs1982073 (TGFB1) was genotyped by generating a restriction site. One nucleotide was modified (underlined nucleotide in Table 2) in the reverse primer, so that a BsrBI restriction site is created. Finally, rs11327935 (COL1A1) was a deletion of one nucleotide. We fluorescently labelled the PCR fragments by adding a fluorescent dye [6-carboxy-fluorescine (FAM)] at the 5′ end of the forward primer. PCR products, which differed by 1 bp were then separated using electrophoresis on an ABI PRISM 3100-Avant automated DNA analyser (Applied Biosystems, Foster City, CA, USA). Genotypes were determined using the GeneScan TM and GenoTyper software (Applied Biosystems). Details of the primer sequences, PCR conditions, and restriction enzymes are described in Table 2.

Table 1.  SNPs selected for genotyping.
Genesrs numberPositionNucleotide change
COL1A1 (Chr17q21.33)rs11079465′ upstreamG/T
rs113279355′ upstreamΔT
rs22693365′ upstreamC/G
BMP2 (Chr20p12.3)rs31782503′ UTRT/C
BMP4 (Ch14q22.2)rs17563Exon4A/G
Table 2.  Primers and corresponding restriction enzymes for SNP analysis.
Genesrs numberPrimer sequences (5′–3′)Restriction enzymeSize of PCR product (bp)Temperature of annealing (°C)
  1. F = forward primer, R = reverse primer. The underlined nucleotide represented the modified nucleotide in the reverse primer (the C was replaced by G).

COL1A1 (Chr17q21.33)rs1107946F: AGGACTAGGGGGCTGAGGTBsmAI (Fermentas)21062
rs11327935F: ACCATCCAAGATTCCATTGC[5′]6-FAM 24064
rs2269336F: AGGCCTTCTCTCAGCCCTACBccI (Biolabs)24159
rs1800012F: GGCCTCCCTTCCAATCAGPflMI (Fermentas)23560
TGFB1(Ch19q13.2)rs1982073F: CGCCCTTCTCCCTGAGGACBsrBI (Fermentas)19464
rs8179181F: AGGGAGACCCAGATGGAGATMluI (Fermentas)23560

Statistical Analysis

Power calculation

A power estimate was done to find out if the size of our replication study is sufficient to detect the previous reported effects (Gordon & Finch, 2005). The power to detect an association depends on the allele frequency of the SNP and the odds ratio (OR). These were estimated using data from previous studies (Chen et al., 2007; Thys et al., 2007a; Schrauwen et al., 2008). The disease prevalence was set at 0.35% and power analysis was performed using the genetic power calculator program (Purcell et al., 2003).

Associating testing

For Hardy–Weinberg equilibrium (HWE) analysis, we tested all the SNPs in the control group using the program Genotype Transposer (version 1.0) (Cox & Canzian, 2001). The statistical analyses were performed using the program SPSS 15.0 (SPSS, Chicago, IL, USA). The differences between otosclerosis patients and control subjects were tested using a likelihood ratio test in a logistic regression framework. Additive or dominant/recessive models were assumed. An additive model (or log-additive model) assumes that the odds of disease increase multiplicatively for every copy of the disease allele. We considered statistical significance at P-values less than 0.05. As the occurrence of otosclerosis is different between females and males, potential gender differences in the association signal were tested by including an interaction genotype × sex in the logistic regression model.

To confirm true replication and to test whether the effect sizes were similar to the previous studies (Chen et al., 2007; Thys et al., 2007a; Schrauwen et al., 2008), a meta-analysis was done. This tool is also particularly robust in estimating the population-wide effect of a genetic risk factor in disease. Meta-analysis and estimation of the common effect size was performed in a logistic regression framework including the main effects genotype and population. To test the homogeneity of the effect size across populations, the interaction genotype × population was added to this model. If the homogeneity test is not significant, there is no systematic difference in the genetic effect size between populations.


Genotyping was performed for nine SNPs from four different genes found to be associated with otosclerosis in previous case-control studies (Chen et al., 2007; Thys et al., 2007a; Schrauwen et al., 2008). The estimation of the power to detect the previously found effects in the Tunisian population varied from 21.3% to 79.7% (Table 3). The highest power was found for rs11327935 of the COL1A1 gene. None of the genotyped SNPs showed a deviation from HWE (data not shown). Association testing showed a significant result for rs1800472 (TGFB1) with otosclerosis under an additive model (Table 4). In addition, a significant association of rs11327935 (COL1A1) with otosclerosis was found assuming a dominant/recessive model (Table 4). Meta-analysis for these SNPs showed that effect sizes were similar across all populations studied so far (Table 4). Testing for interaction between genotype frequencies and sex-revealed differences between gender for two SNPs of the COL1A1 gene: rs11327935 and rs2269336. The SNP rs11327935 was shown to be only significant in females (P= 1.562 × 10−4) and not in males (P= 0.827). For the SNP rs2269336, testing for interaction revealed a significant effect present in males in the same direction as found previously (P= 0.048). The association of BMP4 with otosclerosis showed a trend towards association, and meta-analysis showed that it was in the same direction as was found previously (Table 4).

Table 3.  Power calculations for the Tunisian population. Power calculations were based on the data from previous studies.
Genesrs numberPrevious dataaPower calculation
  1. aPower calculations were done based on previously reported data (Chen et al., 2007; Schrauwen et al., 2008; Thys et al., 2007).

  2. bDAF = disease allele frequency, OR = odds ratio, na = not applicable, no effect was previously found.

Table 4.  Association testing of TGFB1, COL1A1, and BMP2–4 with otosclerosis in the Tunisian population.
Genesrs numberAdditiveaDominant-recessivebPop includedMeta-analysis Homogeneity testP-valueOR (95%CI)
P-valueOR (95%CI)P-valueOR (95%CI)
  1. Significant P-values are in bold. A meta-analysis was done only for the significant SNPs and the SNPs showing a trend towards association (P < 0.1). Am = American, D-B = Dutch—Belgian, FRN = French, and TUN = Tunisian.

  2. aAssuming an additive model.

  3. bAssuming a dominant-recessive model.

 rs19820730.8070.957 (0.670–1.365)      
TGFB1rs18004720.0070.274 (0.098–0.762)  D-B, FRN, TUN0.7428.61×10−80.356 (0.238–0.531)a
 rs81791810.7431.077 (0.691–1.677)      
 rs11079460.8510.958 (0.614–1.494)      
 rs113279350.0591.395 (0.985–1.974)0.0013.988 (1.570–10.128)AM, GER, TUN0.2616.82 × 10−42.820 (1.509–5.268)b
COL1A1rs22693360.4820.865 (0.576–1.298)      
 rs18000120.1431.267 (0.922–1.741)      
BMP2rs31782500.3460.845 (0.596–1.199)0.6780.848 (0.389–1.849)    
BMP4rs175630.0760.740 (0.529–1.034)  D-B, FRN, TUN0.8245.51 × 10−40.816 (0.726–0.916)a


Genetic studies involving different sets of patients allow the replication of results, and also help to eliminate false positive data (Lander & Schork, 1994). In this study, we attempted to replicate the initial finding of four previously described genes (COL1A1, TGFB1, BMP2, and BMP4) that were found to be associated with otosclerosis in a Tunisian population. All of these genes are implicated in bone metabolism. As the otic capsule is unique in its development and structure compared to other bones, it is possible that specific polymorphisms in these genes only influence the otic capsule and not the other bones.

The main constituent of the bone organic matrix is type I collagen encoded by the COL1A1 and COL1A2 genes. Mutations in these genes result in a variety of diseases, such as osteogenesis imperfecta, which is characterized by fragile bones (Dalgleish, 1997). For otosclerosis, the first case-control study analysed the association with polymorphisms in the COL1A1 and COL1A2 genes, in a small population of American–European descent. An association with otosclerosis was only found for the COL1A1 gene (McKenna et al., 1998). Chen et al., 2007 reproduced the association in the same population in addition to a small German population and identified the association of specific haplotypes of COL1A1 with otosclerosis. In vitro analysis of the most frequent haplotype in patients showed an increased promoter–reporter activity. Of the four associated SNPs in the American population, only one SNP (rs11327935) could be replicated in the German population (Chen et al., 2007). Another study investigated the association of the COL1A1 gene with otosclerosis in a Spanish population, but failed to detect any significant result (Rodriguez et al., 2004). In our study, four SNPs from COL1A1 were genotyped. One SNP (rs11327935) was found to be significantly associated with otosclerosis assuming a dominant/recessive model of inheritance. Effect sizes were similar to the previous studies as shown by meta-analysis (Table 4). This SNP has the most power in our population to detect association (approximately 80%) (Table 3). In addition, we showed that rs11327935 was associated with otosclerosis only in females. Association of another SNP (rs2269336) was also detected only in males. These results show a sex-specific association of COL1A1 to otosclerosis. It is unclear whether this sex-specific association is specific to the Tunisian population, as this has not been reported before. Of note is that in Tunisia, women tend to be younger than European women when having their first child, and the number of pregnancies is higher in Tunisia.

TGFB1 is a multifunctional cytokine that is essential for maintaining homeostasis involving bone and the immune response (Watanabe et al., 2002). Mutations that increase TGFB1 activity causes Camurati-Engelmann disease (Kinoshita et al., 2000), a bone-sclerosing disorder, and those in other domains may be associated with osteoporosis (Park et al., 2003). Thys et al., 2007a investigated the implication of TGFB1 in otosclerosis and demonstrated that the amino acid changing SNP T236I was associated with the disease in Dutch–Belgian and French populations. The T allele, coding for isoleucine, was more frequent in the control population compared to the cases, suggesting a protective role for this SNP in otosclerosis. Functional analysis showed that this allele was more active than the wild type. It was hypothesised that I263 decreases susceptibility to otosclerosis by inhibiting osteoclast differentiation and activation in the first otospongiotic phase of otosclerosis (Thys et al., 2007a). In a more recent study, sequencing analysis revealed the presence of several nonsynonymous variants in the TGFB1 gene in otosclerosis patients compared to controls (Thys et al., 2009). Among the three genotyped SNPs from TGFB1, we confirmed even with an underpowered sample size (28.8%) the association of the coding polymorphism T263I (rs1800472) with otosclerosis. In addition, we performed a meta-analysis for rs1800472 including previously published data (Dutch–Belgian, French, and Tunisian populations) (Thys et al., 2007a). This showed that the effect of the association is consistent across populations and further strengthens the association of TGFB1 with the disease (Table 4).

BMP-2 and BMP-4 have been demonstrated to be potent osteotropic factors promoting bone formation (Yamaguchi et al., 1996). A larger study performed by Schrauwen et al. revealed the association of two SNPs (rs3178250 and rs17563) from BMP2 and BMP4 with otosclerosis (Schrauwen et al., 2008). These SNPs were genotyped in the Tunisian population, but we could not detect any significant results (Table 4). However, this most likely represents a lack of power to detect these effects in our population as shown by our power calculations (Table 3). For rs17563 (BMP4), a trend towards association was found that might represent a true effect. Meta-analysis also reveals that the effect size of this SNP is consistent across all populations studied so far, which is consistent with association (Table 4).

Otosclerosis is considered a bone disorder. Genes implicated in bone metabolism such as TGFB1, BMP2–4, and COL1A1 are considered excellent candidate genes to investigate in an association study with otosclerosis. In our replication study, we confirmed the previously detected associations between polymorphisms of TGFB1 and COL1A1 in a non-European population from Tunisia and showed sex-specific associations of COL1A1 with otosclerosis. Further studies should elucidate the pathophysiological mechanisms underlying these associations.


This study was supported by grants from the European Commission (FP6 Integrated project EuroHear LSHG-CT-20054-512063), the FWO (Grant G.0138.07), and le Ministère de l'Enseignement Supérieur, la Recherche Scientifique, Tunisia.