Transforming Growth Factor β Receptor 1 Polyalanine Polymorphism and Exon 5 Mutation Analysis in Breast and Ovarian Cancer

Members of the TGF-β² superfamily are potent regulators of cell growth and differentiation and have been shown to play important roles in the development of cancer. Evidence suggests that most of the activities of TGF-β are mediated by binding to the TβR-I and TβR-II receptors, which are transmembrane serine/threonine kinases. TGF-β inhibits the proliferation of normal epithelial cells, but the majority of neoplastic cell lines are resistant to these effects. TGF-β resistance can occur via different mechanisms, including changes in the expression levels of functional TGF-β or inactivation of the type I and II receptors and other downstream components of the signaling cascade. Somatic mutations in TβR-II have been identified in many tumor types, including colorectal cancers (1), endometrial cancers (2), and gliomas (3). Mutations in TβR-I are less common, although a low frequency of mutations have been identified in prostate cancer cells (4), breast cancer (5), and ovarian cancer (6). Recently Wang et al. (7) reported the identification of a recurrent CTCTGG→CTGCGTGG insertion mutation in exon 5 of TβR-I in >30% of ovarian cancers. If verified, this would indicate that the inactivation of TβR-I is a key step in the development of ovarian cancer, perhaps second only to the inactivation of TP53. In light of this data, the existence of a common germ-line sequence variant of TβR-I with impaired mediation of TGF-β antiproliferative signaling was particularly interesting. The polymorphism involves the deletion of three alanines from a nine-alanine stretch that encompasses the predicted signal sequence and the cytoplasmic region of TβR-I. Pasche et al. (8) speculated that the six-alanine variant, TβR-I (6A), might represent a cancer predisposing allele based on a higher-than-expected number of TβR-I (6A) homozygotes among cancer patients compared with controls. A subsequent larger study by the same investigators confirmed a significantly higher-than-expected frequency of TβR-I (6A) homozygotes and TβR-I (6A)/TβR-I (9A) heterozygotes among 851 cancer cases compared with 735 controls (9). The most striking association was observed among colorectal cancers, in which 4 of 112 cases were TβR-I (6A) homozygotes compared with none among the 735 controls. Pasche et al. (9) also studied a small number of breast and ovarian cancers and these, too, showed an elevated frequency of TβR-I (6A) alleles, although this did not reach statistical significance.

In the present study, we investigated the possible influence of the TβR-I (6A) allele on cancer risk in a case-control study of 355 breast cancer cases, 304 ovarian cancer cases, 98 endometriosis cases, and 248 controls. In addition, we analyzed 55 epithelial ovarian cancers and 33 breast cancers for somatic mutations in exon 5 of TβR-I and, in particular, for the presence of the CTCTGG→CTGCGTGG insertion mutation.

All of the cases and controls used for genotyping were residents in and around Southampton, United Kingdom and were of similar ethnic and geographical background. The endometriosis, breast cancer, and ovarian cancer cases represent a sequential series of patients. Incident cases of ovarian carcinomas (n = 304) and endometriosis (n = 98) were ascertained from women undergoing primary surgery for these diseases in hospitals in and around Southampton between 1993 and 1998. In each case, the histological diagnosis was confirmed by a specialist gynecological pathologist. The ovarian cancers studied consisted of 130 serous, 40 mucinous, 93 endometrioid or clear cell, and 41 undifferentiated adenocarcinomas. The age range of the ovarian cancer patients was 23–92 with a mean of 62 years, and for the endometriosis patients, the range was 20–78 with a mean of 42 years. The age range of the serous ovarian cancers was 23–87 (mean, 61 years), 32–91 years (mean, 62 years) for the mucinous cancers, 31–92 years (mean, 63 years) for the endometrioid/clear cell cancers (mean, 63 years), and 40–76 years (mean, 61 years) for the undifferentiated cancers. Three hundred fifty-five breast cancer cases were selected on the basis of an age at onset under 40 years, a family history of breast cancer (defined as two or more cases of breast cancer in a first- or second-degree female relative) irrespective of age at onset, or bilateral breast cancer irrespective of family history or age at onset. The age range of the breast cancer cases was 19–79 with a mean age of 38 years. All of the controls (n = 248) were white female volunteers who were either staff at the Princess Anne Hospital (n = 117) or outpatients (n = 131) attending for obstetric-related conditions. The age of the controls ranged from 18 to 84 with a mean age of 39 years. For all of the groups, normal genomic DNA was prepared from blood lymphocytes. Epidemiological data, such as reproductive factors, oral contraceptive use, smoking, and obesity, were not available for either the cases or the controls. However, both control and cancer groups were drawn from the same geographical area, which is predominantly an Anglo-Saxon population. Genomic DNA from 33 breast tumors was kindly provided by Dr. Nick Hayward (Queensland Institute of Medical Research, Brisbane, Australia).

Amplification of the genomic sequence containing the GCG repeat was performed as described by Pasche et al. (8) using the exonic primers 5′-ccacaggcggtggcggcgggaccatg-3′ and 5′-cgtcgcccccgggagcagcgccgc-3′. The forward primer was labeled with a fluorescent dye (FAM) to enable detection. The alleles were separated on a 6% denaturing polyacrylamide gel and were detected using a scanning laser fluorescence imager (Bio-Rad Molecular Imager FX). Control samples with a known genotype (confirmed by direct sequencing) were included in each PCR batch. In addition, genotyping of the TβR-I polymorphism was repeated in 20% of cases and 20% of the controls to assess the consistency of the PCR assay.

LOH across the TβR-I locus at 9q22 was assessed using the microsatellite markers D9S283 (9q13–22) and D9S127 (9q31). PCR and LOH analyses was carried out as described previously (10). The intensities of the alleles were quantitated using a phosphorimager (Bio-Rad Molecular Imager FX) and Quantity One software (Bio-Rad, Hercules, CA).

TβR-I exon 5 amplification was carried out above in the presence of 0.05 mCi of [α-³²P]dATP using primers 5′-atggtctgcagcccaacc-3′ and 5′-gcctccaccttctattttc-3′. The samples were analyzed by SSCP/HD analysis through a 0.5× mutation detection enhancement gel matrix (BioWhittaker, Rockland, ME), as described previously (10).

PCR products were purified using the Wizard PCR cleanup system (Promega) and sequenced with a Thermo-sequenase Cycle Sequencing kit (Amersham, Little Chalfont, United Kingdom).

Frequencies were analyzed (compared) using Fisher’s exact test. ORs and 95% CIs were calculated using the relevant 2 × 2 contingency tables. All of the statistical calculations were two-sided and performed using InStat version 3.01 (GraphPad Software, Inc., San Diego, CA).

An example of the GCG repeat analysis is shown in Figure 1. The genotyping was repeated for a minimum of 20% of the cases and controls, and no discrepancies in the genotyping were revealed. The genotype distribution among the cases and controls is shown in Table 1. There was no departure from Hardy-Weinberg equilibrium distribution within the control (P = 0.99), breast cancer (P = 0.79), ovarian cancer (P = 0.92), or endometriosis groups (P = 0.99). There was no evidence of an age-related variation in the TβR-I (6A) allele frequency, with 18.2% of control individuals ages <35 carrying at least one TβR-I (6A) allele compared with 17.9% among those ages >35.

Among the breast cancer cases, there was a statistically significant increase in the number of cases heterozygous for the TβR-I (6A) allele compared with controls (OR, 1.6; 95% CI, 1.1–2.5). Homozygosity for the TβR-I (6A) allele was also associated with an OR of 1.5, although there was a wide 95% CI because of the small number of cases and control with this genotype. The number of cases that were either TβR-I (6A) homozygotes or heterozygotes was also significantly higher among the breast cancer cases with an OR of 1.6 (CI, 1.1–2.5). The majority of the breast cancer cases have been screened for germ-line BRCA1 and BRCA2 mutations, and a total of 17 BRCA1 and 7 BRCA2 mutations were detected. The distribution of the TβR-I alleles among the mutation carriers was not significantly different from that among the breast cancer cases as a whole, with 83% having the 9A/9A genotype and 16% having the 9A/6A genotype. The distribution of the TβR-I alleles did not differ significantly between the cases selected on the basis of age of onset <40 years, bilateral breast cancer, or a family history of breast cancer (not shown); and, in particular, the TβR-I (9A)/TβR-I (6A) heterozygote frequency was 23.8, 21.9, and 25.0%, respectively.

There was an increased risk of endometriosis among TβR-I (6A) heterozygotes (OR, 1.6) and TβR-I (6A) homozygotes (OR, 2.8), although in both instances, the OR spanned 1. Similarly, among the ovarian cancer cases as a whole, there was a nonsignificant increase in risk associated with carrying one (OR, 1.4) or two (OR, 2.6) TβR-I (6A) alleles. Stratifying the ovarian cancers according to histological subtype revealed a highly significant increase in the TβR-I (6A) allele frequency among the endometrioid and clear cell cancers versus the control group (0.172 versus 0.088; P = 0.002). The ORs for heterozygous and homozygous carriers of the TβR-I (6A) allele were 1.8 (95% CI, 1.0–3.1) and 7.8 (95% CI, 1.5–41.4), consistent with a gene-dosage effect. Overall, individuals who were either homozygous or heterozygous for the TβR-I (6A) allele had an OR of 2.1 (95% CI, 1.2–3.6) compared with those homozygous for the wild-type TβR-I (9A) allele. The TβR-I (6A) allele frequency among the mucinous tumors was higher than among the controls, but the numbers were small, and the increase was not statistically significant (P = 0.15). The TβR-I (6A) allele frequency and the genotype distribution among the serous and undifferentiated ovarian cancers were very similar to those of the controls.

The frequency of LOH on chromosome arm 9q among the 57 ovarian cancers informative for either D9S283 or D9S127 was 68% (39/57). Matching tumor DNA was available for 21 of the 62 ovarian cancer cases that were TβR-I (9A)/TβR-I (6A) heterozygotes. Three of the 21 heterozygotes showed LOH of the TβR-I (6A) allele, and 3 showed LOH of the TβR-I (9A) allele.

Fifty-five ovarian cancers (26 serous, 6 mucinous, 19 endometrioid, and 4 undifferentiated) and 33 breast cancers were analyzed for somatic mutations in exon 5 of TβR-I and, in particular, for a CTCTGG→CTGCGTGG insertion mutation that has been reported to occur in 30% of epithelial ovarian cancers (7). We did not detect any aberrant SSCP/HD band shifts in exon 5 among either the breast or the ovarian cancers. Direct genomic sequencing of exon 5 was performed on 5 serous, 5 endometrioid, and 5 mucinous ovarian cancers. An additional six ovarian cancers (four serous, one endometrioid, and one mucinous) were sequenced with the GT tracks only, to identify any CTCTGG→CTGCGTGG insertion mutations. Seven of the nine serous cancers and five of six endometrioid cancers showed LOH across 9q. We obtained clear sequence for all of the cases, and only wild-type sequence was detected.

Functional germ-line allelic variants of genes with a known role in cancer development are plausible cancer predisposing factors. The TβR-I (6A) allele fulfills both of these criteria because somatic mutations in TβR-I have been reported in a variety of malignancies (4, 5, 6, 7) and the TβR-I (6A) variant has reduced TGF-β antiproliferative effects (9) compared with the common TβR-I (9A) variant and the rare TβR-I (10A) variants. Heterozygotes for the TβR-I (6A) allele have been reported in ∼10% of eastern United States population controls (9, 11); but among a variety of cancer types, the heterozygote frequency is ∼15% and TβR-I (6A) homozygotes, to date, have been observed only among cancer patients. Pasche et al. (9) has reported the largest study of the TβR-I (6A) variant among 851 cancers and 735 controls. The strongest association was observed with colorectal cancers, in which 4 of the 112 cases were homozygous for the TβR-I (6A) allele compared to none amongst 735 controls (P < 0.01). The study also included 48 ovarian and 152 breast cancers, and, although the TβR-I (6A) allele frequency was higher than among the controls, the number of cases was too small to provide statistically meaningful data. The aim of our study was to provide additional data concerning the possible role of the deletion TβR-I (6A) variant in breast and ovarian cancer predisposition.

The frequency of TβR-I heterozygotes among our 248 British controls was similar to that reported by van Tilborg et al. (12) in a Dutch control group, consisting of 148 people who donated blood to screen for recessive hereditary noncancer diseases (16 versus 17%). A lower frequency of TβR-I heterozygotes has been reported in an eastern United States-based population (16 versus 10%), and this may reflect ethnic differences in the distribution of the TβR-I alleles (9, 11). In our study, two controls were homozygous for the TβR-I (6A) allele, which is in contrast to previous studies, which did not report homozygotes among control individuals. Nevertheless, among our breast cancers, the number of TβR-I (6A) homozygotes and heterozygotes was significantly higher compared with those in the controls (OR, 1.6; 95% CI, 1.1–2.5). The number of TβR-I (6A) homozygotes among the epithelial ovarian cancers as a whole was higher than among the controls (2.0 versus 0.8%, respectively); however, the difference in allele frequency failed to reach statistical significance (P = 0.062).

There is good evidence that the different histological subtypes of ovarian cancer are distinct biological entities (10); and, accordingly, the data were stratified into serous, mucinous, and endometrioid/clear cell types. On this basis, the endometrioid and clear cell cancers showed a significant increase in the TβR-I (6A) allele frequency (P = 0.002). In particular, five of the six TβR-I (6A) homozygotes in the ovarian cancer group occurred among the endometrioid and clear cell subtypes, and overall, 5.4% were TβR-I (6A) homozygotes compared with only 0.8% of the controls. We have shown in previous studies that the endometrioid and clear cell ovarian cancers are likely to arise by malignant transformation of endometriosis (10), and we were interested to see whether the TβR-I (6A) allele was also associated with this disease. The TβR-I (6A) allele frequency was indeed higher among the 98 endometriosis cases, and 2% were TβR-I (6A) homozygotes, although this failed to reach statistical significance (P = 0.09). However, our study had only a 60% power to detect an OR of 2, and larger studies into the possible association with endometriosis are warranted.

LOH analysis of the ovarian cancers demonstrated that a large proportion (68%) had lost one or both of the markers flanking the TβR-I locus. This may point to a critical involvement of TβR-I inactivation during ovarian carcinogenesis, although a more detailed analysis will be required to ascertain whether the LOH was specifically targeting the TβR-I locus. It is interesting to note that 68% of the ovarian cancers examined had LOH at 9q, but, among the 9A/6A heterozygotes, the LOH frequency was only 29%. Furthermore, we did not observe any preferential loss of the TβR-I (9A) allele among cases heterozygous for the TβR-I polymorphism. It is possible that heterozygotes, which already contain a functionally impaired allele, are less likely to undergo LOH but, in light of the small number of cases studied, this conclusion should be regarded as preliminary.

Given the potential role of germ-line variation in TβR-I in breast and ovarian cancer susceptibility, we were intrigued by a recent study reporting a recurrent CTCTGG→CTGCGTGG somatic mutation in >30% of ovarian cancers (7). Despite using both SSCP/HD and direct sequencing analysis, we failed to detect somatic alterations in exon 5 in any of the 55 ovarian or 33 breast cancers investigated. SSCP/HD analysis is efficient at identifying insertion and deletion mutations, and it is very unlikely that our analysis would have failed to identify this 2-bp insertion mutation. Indeed, SSCP was the technique used to identify the original mutation (7). Direct genomic sequencing on a subset of the ovarian cancers also failed to identify any somatic alterations. It is unlikely that differences in the tumor cohorts could explain our failure to detect the mutation, because all the major histological subtypes of ovarian cancer were represented in our study. In any case, Wang et al. (7) reported finding mutations in all of the histological subtypes, grades, and stages of ovarian cancers. The exon 5 frameshift mutation is unusual in that it involves the insertion of one guanine residue in codon 276 and another in codon 277, apparently as a simultaneous event. The DNA sequence image of the mutation given by Wang et al. (7) was unconvincing, and it is possible that the mutation may be an artifact. Nevertheless Wang et al. provided convincing evidence that the mutation was associated with absent or reduced expression of TβR-I protein. It is likely that, by whatever mechanism, the loss of TβR-I expression is an important event in ovarian carcinogenesis.

In summary, we support previous observations that the TβR-I (6A) allele represents a cancer-predisposing factor. In our study, the TβR-I (6A) allele was associated with a 1.6-fold increased risk of breast cancer. Given that our selection criteria are characteristic of women with a genetic predisposition to breast cancer, it is unclear as to what extent this finding will apply to all breast cancer cases. With respect to ovarian cancer, the TβR-I (6A) allele appears to increase the risk of the endometrioid and clear cell subtypes but not the serous or mucinous subtypes. Although our finding of an increased risk of breast and ovarian cancer associated with the TβR-I (6A) allele agrees with Pasche et al. (9), it is possible that chance or confounding factors, such as ethnicity and the influence of known reproductive risk factors such as oral contraceptive use, may have generated a false positive result. Consequently, it will be important to replicate our findings in larger, population-based, case-control studies. Nevertheless the high frequency of LOH at the TβR-I locus in ovarian cancers is consistent with an important role for this gene in tumor development; however, we have not been able to confirm a previous report of a somatic mutational hot spot in exon 5 of the gene.