Polyglutamine repeat length in the AIB1 gene modifies breast cancer susceptibility in BRCA1 carriers



Variation in the penetrance estimates for BRCA1 and BRCA2 mutation carriers suggests that other factors may modify cancer risk from specific mutations. One possible mechanism is an epigenetic effect of polymorphisms in other genes. Genes involved in hormonal signal transduction are possible candidates. The AIB1 gene, an estrogen receptor (ER) coactivator, is frequently amplified in breast and ovarian tumors. Variation of a CAG repeat length has been reported within this gene that encodes a polyglutamine repeat in the C-terminus of the protein. Three hundred eleven BRCA1/2 mutation carriers (257 were of Ashkenazi origin) were genotyped for the AIB1 polyglutamine repeat. Relative risks (RR) were estimated using a maximum likelihood approach. The estimated breast cancer (BC) RR per average repeat length adjusted for population type (Ashkenazi vs. non-Ashkenazi) was 1.15 (95% CI = 1.02–1.30; p = 0.01) for BRCA1/2 carriers, and 1.25 (95% CI = 1.09–1.42; p = 0.001) when analysis was restricted to BRCA1 carriers. RR of BC was 1.17 (95% CI = 0.91–1.74), for individuals with 2 alleles ≥29 polyglutamine repeats and 0.78 (95% CI = 0.50–1.16) for those with at least 1 allele of ≤26 repeats, compared to individuals with the common genotypes 28;28, 28;29 or 28;30. The corresponding BC RR in BRCA1 mutation carriers was 0.55 (95% CI = 0.34–0.90) and 1.29 (95% CI = 0.85–1.96) in those with ≤26 and ≥29 repeats respectively (p = 0.025). These results indicate significant association of the risk for BC in carriers of BRCA1 mutations with the polyglutamine chain of the AIB1 gene. Longer repeat length correlates with elevated risk, whereas in carriers of a shorter AIB1 allele BC risk was reduced. The AIB1 polyglutamine length did not affect BC risk among BRCA2 mutation carriers. © 2003 Wiley-Liss, Inc.

Carriers of a mutated BRCA1 or BRCA2 gene are at an increased risk of breast cancer (BC), however penetrance estimates are different in various study populations.1, 2, 3, 4 Breast cancer risk for BRCA1/2 carriers varied from 70% by age 70 in studies based on families with numerous cases1, 2 to 37–56% in population based studies.3, 4 Modification of the risk by other genes or environmental factors clustering in families probably explains most of this difference. When considering the former possibilities genes involved in hormonal signal transduction are of particular interest in mammary carcinogenesis both in BRCA1/2 mutation carriers and in the general population.

The AIB1 gene (amplified in breast cancer-1) was cloned using chromosome microdissection and hybrid selection of the 20q region that is often amplified in BC.5 As a member of the steroid receptor coactivator (SRC-1) family, it interacts with estrogen-receptor-α (ER-α) in a ligand-dependent manner.5 It has been demonstrated that AIB1 is a phosphoprotein that can be phosphorylated by MAPKs, enhancing its activity. The stimulated protein then recruits the p300 and associated histone acetyltransferase activity.6

AIB1 is amplified and highly expressed in BC and ovarian cancer (OC), especially in ER positive tumors.7AIB1 mRNA, however, is also overexpressed in high grade tumors that are ER and PR negative, and positive for p53 and HER2/neu staining.8 Gene amplification is also found in pancreatic cancer cell lines9 and gastric cancers10 that suggests a role for AIB1 in other, non ER-related, signal transduction pathways. It has also been found to have an effect on IGF-1 response, suggesting that it might also have a role in somatic growth.11

The glutamine codons, CAG and CAA encode a glutamine rich domain in the C-terminus of the AIB1 protein.12, 13 It contains 2 polymorphic stretches of CAG repeats separated by a CAA repeat. The functional significance of this repeat is unknown. If the repeat length altered the protein function it could modulate the transcriptional activity of ER, however, and could modify BC susceptibility. Elevated BC risk associated with a longer repeat was reported in BRCA1/2 carriers14 but not in the general population.15

We assessed the effect of the polyglutamine repeat polymorphism in the AIB1 gene on BC risk in BRCA1/2 mutation carriers, mainly of Ashkenazi origin.


Study population

Blood samples from 325 BRCA1/2 carriers were identified through 2 centers: 243 were collected through the oncology department and the cancer genetic clinic in the Hadassah Medical Centre in Jerusalem, Israel and 82 through the cancer genetic clinic in the Royal Marsden NHS Trust, London, UK. Individuals were tested on the basis of a family history of breast or ovarian cancer or on the basis of their Ashkenazi origin. All but 1 of the cases from Jerusalem (that carried a BRCA2 mutation; 8558delA), were carriers of 1 of the 3 Ashkenazi founder mutations (BRCA1 185delAG: 134 cases, BRCA1 5382insC: 76 cases, BRCA2 6174delT: 32 cases). The UK carriers included 23 carriers of Ashkenazi founder mutations (1 individual carried both a 185delAG and 6174delT) and 59 other mutations (48 BRCA1; 11 BRCA2; for list of mutations see Appendix).


Genomic DNA was extracted according to standard protocols, and used as a template for the PCR reaction. The region of the AIB1 gene that includes the polymorphic CAG repeat was amplified using primers, forward: 5′-TCCGACAACAGAGGGTGGCTATG-3′ and reverse: 5′-TTAGGAGGTGGGCTGAAGGCCTG-3′of which the forward primer was labeled fluorescently. Typically 25 ng of genomic DNA was used in each PCR reaction of 15 μl. PCR cycling consisted of 35 cycles of 94°C, 64°C and 68°C for 30 sec for each stage after an initial 10 min at 95°C. Each reaction contained 2.0 mM MgCl2 and 1 U of Amplitaq Gold (Perkin-Elmer, Norwalk, CT). After PCR amplification the repeat length was analyzed using an ABI 377 automated DNA Sequencer. The size of the amplified products was determined relative to size standards using Genescan and Genotyper Analysis software (Perkin-Elmer). Genotyping was carried out by 2 independent personnel who were blinded to the case-control status of the samples. Polymorphic polyglutamine repeat length containing 20–31 repeats of the trinucleotide glutamine codons (CAA, CAG) was scored as described previously.14, 15 Several samples were sequenced to confirm the repeat length as established by the ABI 377 Genotyper. Blinded quality control samples were inserted to validate genotype scoring accuracy.

Statistical analysis

Six of the 325 carriers were excluded from all analyses; 2 for whom the date of birth was unknown, and 4 individuals recorded as affected but for whom the age at diagnosis was unknown. A further 8 individuals could not be genotyped so that the final analysis included data from 311 mutation carriers.

The analyses presented here examined the association between breast cancer risk and genotype. For simplicity the analyses were based on the first occurring cancer (i.e., all individuals were censored at the first cancer). With this classification there were 195 breast cancer cases (18 also had also ovarian cancer) and 116 individuals unaffected with breast cancer at censoring (39 had ovarian cancer). Individuals unaffected with breast cancer were censored either at date of ovarian cancer diagnosis, date of prophylactic surgery or date of last follow up.

The effects of AIB1 genotypes on breast cancer risk in mutation carriers were evaluated using a maximum likelihood approach, based on the likelihood of the observed phenotypic and genotypic data conditional on the phenotypic data. This approach provides consistent estimates of the risk parameters despite the non-random ascertainment of subjects with respect to phenotype. In this analysis, the incidence of breast in individuals with genotype Category j was assumed to be given by:

equation image

where exp(βj) is the rate ratio for breast cancer in Category j and λ0(t) are the age-specific incidence rates in the baseline category. We estimated the rate ratios for different genotype categories. λ0(t) was chosen so that the overall breast cancer incidence rates in carriers, averaged over all genotypic categories, agreed with previously derived incidence rates for BRCA1 and BRCA2 carriers. These estimates were taken from the recent meta-analysis of cancer risks in BRCA1 and BRCA2 carriers from population-based studies.16 In our study, breast cancer estimated risks are moderately lower than the estimates derived by the BCLC studies of high risk families1, 2 and also slightly lower than the estimates of a population based study conducted in the Ashkenazi population.3 These risk estimates is applicable to our carrier sample that is mainly composed of a population based series. Two types of analyses were carried out. In the first analysis we treated allele size as a continuous covariate, and estimated a single log-risk ratio parameter for the increase in risk per unit length. In the main analysis, we used the average of the allele lengths on the 2 chromosomes as the covariate (in effect assuming an additive effect of the 2 alleles) but we also carried out analyses based on the longer and shorter allele lengths. In the second type of analysis, we classified AIB1 alleles according to the shorter allele into 3 categories: ≤26 repeats, 28 repeats and ≥29 repeats, and estimated risk ratios based on the category of the shorter repeat length. In each analysis, we estimated the log-risk ratio parameters together with population allele frequencies. The latter were estimated separately for the Ashkenazi and non-Ashkenazi groups, to allow for possible differences in AIB1 allele frequencies. These analyses were carried out using the program MENDEL.17 We also carried out analysis using the previously described score test18 that gave similar results (not shown).


Table I shows the distribution of AIB1 polyglutamine repeat length in the Ashkenazi carriers and British non-Ashkenazi population. When genotypes are categorized by the shorter allele, the most common alleles were 26, 28 and 29 repeats with frequency of 24.3%, 60.1% and 15.6% in the Ashkenazi population and 16.3%, 54.5% and 29.1% in the British population. These frequencies are comparable to those reported in other populations.12, 13, 14, 15 Allele lengths were however, significantly shorter, on average, in the Ashkenazi populations than the U.K. carriers (ptrend = 0.0007). Extremely short repeats of 22 repeats were limited to the Ashkenazi Population (7/257; 2.7%) and a carrier of a 31 polyglutamine repeat allele was identified among non-Ashkenazi BRCA carriers.

Table I. Genotype Distribution of the Polyglutamine Repeat Polymorphism in the AIB1 Gene in Ashkenazi and Non-Ashkenazi BRCA1/2 Mutation Carriers
Polyglutamine repeat lengthAshkenazi1n (%)Non-Ashkenazi2n (%)
  • 1

    n = 256.

  • 2

    n = 55.

22/284 (1.6)0
22/293 (1.2)0
26/263 (1.2)0
26/2831 (12.1)5 (9.1)
26/2921 (8.2)3 (5.4)
26/3101 (1.8)
28/2847 (18.3)6 (10.9)
28/29105 (41.0)23 (41.8)
28/302 (0.8)1 (1.8)
29/2940 (15.6)15 (27.3)
29/3001 (1.8)

The frequencies of AIB1 genotypes according to disease status in BRCA1 and BRCA2 carriers is shown in Table II. Cut-off points that were used in the categorized risk analyses (Table III) are marked in the table.

Table II. Frequencies of Polyglutamine Repeat Lengths in the AIB1 Gene by Disease Status in BRCA1 and BRCA2 Carriers1
Polyglutamine repeat lengthBRCA1 carriers (n = 222)BRCA2 carriers (n = 88)
BC + (n = 138)BC − (n = 84)BC + (n = 57)BC − (n = 31)
  • 1

    The individual carrying a BRCA1&2 mutation is excluded. Values are n (%). BC+, affected with breast cancer; BC−, unaffected with breast cancer.

  • 2

    Cutpoints used for the risk analysis by categorised AIB1 genotypes as presented in Table III: ≤26, = 28, ≥29.

22/282 (1.4)1 (1.2)1 (1.8)0
22/2901 (1.2)2 (3.5)0
26/262 (1.4)1 (1.2)00
26/2810 (7.2)12 (14.3)8 (14)6 (19.4)
26/298 (5.8)8 (9.5)6 (10.5)2 (6.4)
26/3121 (0.7)000
28/2823 (16.7)16 (20.2)8 (14)7 (22.6)
28/2957 (41.3)30 (34.5)25 (43.9)14 (45.2)
28/3021 (0.7)2 (2.4)00
29/2934 (24.6)12 (14.3)7 (12.3)2 (6.4)
29/30201 (1.2)00
Table III. Risk for Breast cancer in BRCA1/2 Carriers Associated with Polyglutamine Repeat Length in the AIB1 Gene
 BRCA1/2 carriersBRCA1 carriersBRCA2 carriers
RR (95% CI)pRR (95% CI)pRR (95% CI)p
  • 1

    FAR confidence limits (see Methods).

  • 2

    2df likelihood ratio test

AIB1 polyglutamine repeat length as a continuous covariate
 Average repeat length1.15 (1.02–1.30)0.011.25 (1.09–1.42)0.0010.93 (0.78–1.12) 
 Shorter allele1.17 (1.02–1.36)0.031.35 (1.13–1.60)0.00050.86 (0.70–1.07)0.17
 Longer allele1.36 (1.00–1.85)0.051.28 (0.93–1.74)0.171.41 (0.83–2.40)0.22
Length by specific allele grouping
 1 alleles ≤ 260.78 (0.52–1.16)1 0.55 (0.34–0.90)1 1.52 (0.88–2.63)1 
 28;28, 28;29, 28;301.00 (0.80–1.25) 1.00 (0.77–1.30) 1.00 (0.69–1.45) 
 2 alleles ≥ 291.17 (0.91–1.74)0.1021.29 (0.85–1.96)0.02520.77 (0.34–1.75)0.132

The estimated BC relative risk (RR) in BRCA1/2 carriers by average polyglutamine repeat length as a continuous variable (Table II) was 1.15 (95%CI = 1.02–1.30, p = 0.01) per repeat unit. Similar effects were observed when analysis was based on the length of the longer or the shorter allele. The estimated effect was stronger for BRCA1 carriers (RR = 1.25; 95% CI = 1.09–1.42), with no significant effect in BRCA2 carriers (RR = 0.93; 95% CI = 0.78–1.12; p = 0.013 for difference in RRs).

To further assess the contribution of polyglutamine repeat length to breast cancer risk, we categorized individuals by the smaller repeat length. In comparison with individuals with the “common” 28/28, 28/29, 28/30 genotypes, RR for carriers of 2 alleles ≥29 repeats was 1.17 (95% CI = 0.91–1.74), whereas carriers with 1 allele ≤26 repeats had a reduced risk (RR = 0.78; 95% CI = 0.52–1.16). In BRCA1 carriers, BC RRs were 1.29 (95% CI = 0.85–1.96) and 0.55 (95% CI = 0.34–0.90) in individuals with 2 alleles of ≥29 and 1 allele of ≤26 repeats respectively (p = 0.025).

The mean age at onset of the BC cases was not significantly related to repeat length; the mean age of onset was 41.2 years for those with 2 alleles of ≥29 repeats, compared to 42.0 years for cases with at least 1 allele of ≤26 repeats (Table IV, ptrend = 0.11 in age at onset with average repeat length).

Table IV. Frequencies of Categorized Polyglutamine Repeat Lengths by Disease Status and Age at Breast Cancer Onset in BRCA1/2 Carriers
AIB1 repeatsBRCA1/2 carriers (n = 311)
BC + Age at BC onset2BC −
  • 1BC+, affected with breast cancer; BC−, unaffected with breast cancer. Values are n (%).

  • 2

    Average age at breast cancer onset for individuals by AIB1 CAG repeats grouping (years).

1 allele ≤ 2640 (21)/42.031 (27)
28;28, 28;29, 28;30114 (58)/41.370 (60)
2 alleles ≥ 2941 (21)/41.215 (13)


We have found a significant association between AIB1 polyglutamine repeat length and breast cancer risk in BRCA1 gene mutation carriers and a similar but non-significant association in BRCA2 mutation carriers. The association was seen with the length of both the longer and the shorter allele, but was most significant when analyzed in terms of the average allele length. The results of the categorical analysis were also consistent with an effect of allele length, with a higher risk in carriers of 2 alleles of ≥29 repeats and a lower risk of approximately 2-fold in carriers of at least 1 allele of ≤26 repeats. These results would be consistent with a recessive type model, under which the length of the longer allele determines risk. A “dosage” type model in which each allele contributes to risk cannot be ruled out.

A study by Rebbeck et al.14 has suggested a very similar association between the longer AIB1 alleles and BC risk of 1.59 (95% CI = 1.5–2.9) and 2.85 (95% CI = 1.64–4.96) in women with at least 28 or 29 repeat lengths respectively in BRCA1/2 carriers. An interaction with reproductive history of nulliparity and later age at first live birth was also found. They did not report separate analyses for BRCA1 and BRCA2 carriers.

Modifiers of cancer risk in BRCA1/2 mutation carriers reported previously include the H-ras minisatellite locus, rare alleles of which were reported to be associated with ovarian cancer risk.20 Longer length of the CAG repeats in the AR gene elevate breast cancer risk in BRCA1 carriers,18, 21 whereas in BRCA2 carriers BC risk was associated with a rare polymorphism in the RAD51 gene.22, 23 In these studies, however, the proportion of carriers in the high-risk category is relatively low. In contrast, the AIB1 high-risk category may be much larger. For example, the low and high risk categories in our analysis (1 ≤26 repeat allele and 2 ≥29 repeat alleles respectively) both have population frequencies of approximately 20%. The clinical implication of these findings are still limited, however it might provide additional guideline for risk estimation and consultation for BRCA1 carriers.

The functional effect of the polyglutamine chain of the AIB1 gene is currently unknown. Polyglutamine repeats are often identified in transcription factors and associated with transcriptional activity. The polyglutamine repeat length variation in several genes was associated with various neurological diseases.24, 25, 26 The polyglutamine repeat in the AR gene has been associated with prostate cancer27 and male breast cancer risk,28 perhaps due to increased level of transactivation of the androgen receptor with shorter repeat alleles.29 It has also been demonstrated that polyglutamine chains per se can activate transcription when fused to a yeast GAL4 DNA binding domain. Furthermore, there is evidence that the transactivation is affected by the length of the repeat.30 A functional role for the AIB1 polyglutamine repeat length polymorphism is suggested by 2 studies published recently. Patel et al.31 reported association with bone mineral density, implying an effect on ER signaling pathways. IGF-I blood levels were also associated with the repeat length among oral contraceptives users.32 Thus, it is reasonable to speculate that the polyglutamine repeat length in the AIB1 gene could have a significant effect on the protein function, perhaps altering the coactivation of the ER and hence modifying hormone-induced cancer risk.

Frequent amplification and overexpression of the AIB1 gene was found, especially in ER and PR positive tumors that are rarely found in BRCA1 carriers.7 Contradictory findings of higher AIB1 mRNA expression in association with ER and PR negativity, P53 and Her2/neu positivity were also reported.8 Although the majority of BRCA1 associated breast cancers are ER and PR negative,33 the role of estrogen signaling in BC formation in this group is well established.34, 35 There is evidence that connects hormonal pathways to BRCA1/2 function. BRCA1/2 gene expression is associated with steroid hormonal stimulation36, 37 and hormonal manipulation such as oophorectomy reduces BC risk in both BRCA1 and BRCA2 carriers.35, 36 Stronger evidence, however, links BRCA1 with hormonal signaling. BRCA1 was found to inhibit ER transcription activity in a breast cancer cell line,35 and in the presence of mutated BRCA1, breast epithelial cell growth becomes independent of estrogen.40 Therefore, the complex co-regulation of ER might be impaired in the presence of non-functional BRCA1 protein and may be more dependent on functional variations of other coactivators such as AIB1. This could explain the differential effect of AIB1 repeat length in BRCA1 mutation carriers vs. BRCA2 carriers.

In conclusion, our results suggest AIB1 polyglutamine repeat lengths are associated with breast cancer risk among women with mutations in the BRCA1 gene and to a lesser extent in BRCA2 carriers. If these results are confirmed in further large series, genotyping of this locus could modify the risk predictions used in counseling of BRCA1/2 mutation carriers. In addition, the differential effect we found in BRCA1 carriers compared to BRCA2 carriers may hint at the variation in molecular mechanisms underlying tumorigenesis in these 2 groups. Studies of the functional properties of the varying polyglutamine repeat lengths in the AIB1 gene is, therefore, of great interest.


Mutations in BRCA1/2 non-Ashkenazi carriers:

BRCA1 (n = 48): 4184delTCAA: 11 cases, 3450delCAAG: 6 cases, 3124delA: 3 cases, 3875delGTCT: 3 cases, 2 cases of each: 546G→T, 1014delT, 1445T→A, 2157delG, 2190delA, 5629delG and 1 case of the following: 122G→T, 1182A→T, 1224delGAT, 1454delAA, 1623delTTAAA, 1942delAGAA, 2012insT, |2313G>|T, 2371delTG, 2594delC, 2722G→C, 3034delAAAC, 5312delT.

BRCA2 (n = 11): 6503delTT: 4 cases, 7057delCTTAT: 2 cases and a case of the following: 1529delAAAG, 3173C→G, 6690delTC, 6819delTG, 9179C→G.