There is growing evidence that common variants of the transforming growth factor-β (TGF-β) signaling pathway may modify breast cancer risk. In vitro studies have shown that some variants increase TGF-β signaling, whereas others have an opposite effect. We tested the hypothesis that a combined genetic assessment of two well-characterized variants may predict breast cancer risk. Consecutive patients (n = 660) with breast cancer from the Memorial Sloan-Kettering Cancer Center (New York, NY) and healthy females (n = 880) from New York City were genotyped for the hypomorphic TGFBR1*6A allele and for the TGFB1 T29C variant that results in increased TGF-β circulating levels. Cases and controls were of similar ethnicity and geographic location. Thirty percent of cases were identified as high or low TGF-β signalers based on TGFB1 and TGFBR1 genotypes. There was a significantly higher proportion of high signalers (TGFBR1/TGFBR1 and TGFB1*CC) among controls (21.6%) than cases (15.7%; P = 0.003). The odds ratio [OR; 95% confidence interval (95% CI)] for individuals with the lowest expected TGF-β signaling level (TGFB1*TT or TGFB1*TC and TGFBR1*6A) was 1.69 (1.08-2.66) when compared with individuals with the highest expected TGF-signaling levels. Breast cancer risk incurred by low signalers was most pronounced among women after age 50 years (OR, 2.05; 95% CI, 1.01-4.16). TGFBR1*6A was associated with a significantly increased risk for breast cancer (OR, 1.46; 95% CI, 1.04-2.06), but the TGFB1*CC genotype was not associated with any appreciable risk (OR, 0.89; 95% CI, 0.63-1.21). TGFBR1*6A effect was most pronounced among women diagnosed after age 50 years (OR, 2.20; 95% CI, 1.25-3.87). This is the first study assessing the TGF-β signaling pathway through two common and functionally relevant TGFBR1 and TGFB1 variants. This approach may predict breast cancer risk in a large subset of the population.

Transforming growth factor-β1 (TGFB1) is a potent growth suppressor of mammary epithelial and breast carcinoma cells and acts as a negative autocrine growth inhibitor (1, 2). In normal cells, TGF-β is a potent growth inhibitor. On the other hand, it is now well appreciated that metastasis of most tumor types requires TGF-β activity and that, in advanced disease, TGF-β is clearly pro-oncogenic (3, 4). It seems therefore that for every action of TGF-β there is a counteraction that TGF-β is capable of performing (5).

Mouse models have allowed demonstration that increased TGF-β signaling is associated with decreased cancer incidence. This association has been found with both transgenic mice expressing higher TGFB1 levels (6, 7) and mice with increased TGF-β signaling because of constitutively active TGFBR1 receptor (8). On the other hand, the same sets of experiments have documented that the growth of tumors is fueled by increased TGFB1 levels and by increased TGF-β signaling. Hence, these in vivo experiments indicate that higher TGFB1 levels serve as a surrogate of increased TGF-β signaling and lower TGFB1 levels a surrogate for decreased TGF-β signaling. This dichotomy of function for TGF-β serves as the basis of the hypothesis to be tested in this study.

Five polymorphisms have been identified in the TGFB1 gene to date (9). Two are in the promoter region (C-509T and G-800A), one within the signal sequence (Leu10Pro), one within exon 1 (Arg25Pro), and one within exon 5 (Thr263Ile). The C-509T single nucleotide polymorphism (SNP) is not contained within a known consensus sequence for a promoter regulatory element and does not affect breast cancer risk (10), but one report indicates that it may modify TGFB1 expression (11). There is evidence that the Arg25Pro polymorphism modifies TGFB1 circulating levels (1214), but it has not been associated with breast cancer risk thus far. The Leu10Pro polymorphism has been extensively studied in relation to breast cancer risk. The CC (TGFB1*CC) genotype (Leu10Pro polymorphism) was found by one group of investigators to be associated with a 64% decreased breast cancer risk in a cohort study of 3,075 White American women over age 65 years at recruitment (15). In contrast, in a pooled analysis of three European case-control studies that included 3,987 cases and 3,867 controls, the CC genotype was associated with a 21% increased risk of breast cancer. In the same study, the investigators found that the C-509T and the T29C SNPs but not the G-800A were in strong linkage disequilibrium and that they were both significantly associated with increased incidence of invasive breast cancer in a recessive manner. In a hospital-based study of 232 cases and 172 controls conducted in Japan, there was no significant overall association between the CC genotype and breast cancer. However, the CC genotype was associated with a 65% reduced risk of breast cancer in comparison with the TT genotype among premenopausal women [odds ratio (OR), 0.45; 95% confidence interval (95% CI), 0.20-0.98; ref. 16]. A German study of 500 cases and 500 controls did not find any statistically significant association between either TGFB1*CC and TGFB1*CT genotypes and breast cancer (17). Most recently, a large multiethnic case-control study of 1,123 breast cancer cases and 2,314 controls from Los Angeles and Hawaii also did not find any association between the TGFB1*CC genotype and breast cancer risk (18).

Of major interest in this regard is the recent report that patients with a diagnosis of breast cancer that carry the TGFB1 C variant that results in higher circulating TGFB1 levels have a significantly decreased survival compared with noncarriers (19). This polymorphism is represented by a SNP at position 29, which results in the substitution of leucine to proline at the 10th amino acid position (Leu10Pro). The leucine-to-proline substitution results in significantly higher TGFB1 in vivo levels among TGFB1*CC carriers (2022). The Leu10Pro signal peptide substitution is well characterized for its effects on the regulation of TGFB1 secretion. Transfection of HeLa cells with constructs encoding either proline or leucine forms of TGFB1 and driven by the cytomegalovirus promoter show that the signal peptide with proline (C variant) at residue 10 causes a 2.8-fold increase in secretion compared with the leucine (T variant) form (10). Hence, there is both in vitro and in vivo evidence to support the conclusion that the T29C SNP is the most relevant SNP that modifies the amount of secreted and circulating TGFB1 and affects overall TGF-β signaling. The function of the other TGFB1 polymorphisms and their relevance to breast cancer risk remains to be further characterized.

We have previously identified TGFBR1*6A, a common variant of TGFBR1. TGFBR1*6A has a deletion of three GCG triplets coding for alanine within a nine alanine (9A) repeat sequence of TGFBR1 (TGFBR1*9A) exon 1, resulting in a six alanine (TGFBR1*6A) repeat sequence (23). The 9-bp deletion that differentiates TGFBR1*6A from TGFBR1 is located within the predicted signal sequence cleavage region. Two separate groups of investigators have shown that TGFBR1*6A mediates TGF-β growth inhibitory signals significantly less effectively than TGFBR1 (24, 25). TGFBR1*6A is a candidate tumor susceptibility allele that is associated with an increased incidence of various types of cancer (26, 27). The first report of an association between breast cancer and TGFBR1*6A was a case-control study of 355 women with breast cancer and 248 controls by Baxter et al. (26). In that study, TGFBR1*6A was associated with a 60% increased risk of breast cancer. These results are further supported by our recent meta-analysis of eight case-control studies that included 1,420 breast cancer cases and 1,823 controls, which showed that TGFBR1*6A predisposes to the development of breast cancer (OR, 1.38; 95% CI, 1.14-1.67; ref. 28).

Thus, differences in TGF-β signaling, whether mediated by ligand or receptor variants, have been associated with risk for breast cancer. Various genotypic combinations may theoretically have either opposite or synergistic effects on breast cancer risk.

It is known that TGFB1 and TGFBR1 map to different chromosomes, 19q13.1 and 9q22, respectively. Thus, they are independently inherited. This has led us to hypothesize that the TGFB1 T29C and the TGFBR1*6A variants have a functional interaction with respect to breast cancer risk. Furthermore, based on genotyping of these two variants, we hypothesized that individuals with the combination of these two variants resulting in the highest predicted levels of TGF-β signaling have the lowest breast cancer risk and conversely those with the lowest levels of TGF-β signaling the highest breast cancer risk.

To test the hypothesis that a combined assessment of TGF-β pathway signaling variants may predict breast cancer risk more accurately than each variant alone, we genotyped 660 women with a diagnosis of breast cancer and 841 healthy female controls for the two most common and biologically relevant variants of the TGF-β signaling pathway, TGFB1 T29C and TGFBR1*6A.

Study participants. As part of institutional review board–approved protocols, we collected blood samples from 660 consecutive patients admitted to the Memorial Sloan-Kettering Cancer Center (New York, NY) with a diagnosis of breast cancer. DNA was extracted from peripheral blood following completion of diagnostic studies on these samples. Information regarding sex, age, age at breast cancer diagnosis, and ethnic status was recorded. In a subset of 99 and 97 patients, information on estrogen receptor (ER) and progesterone receptor (PR) status as assessed by immunohistochemistry was collected at the time of case collection. However, in the rest of the samples, due to destruction of personal identifiers, it was not possible to retrieve the information retrospectively. All breast cancer cases were histologically confirmed at the Memorial Sloan-Kettering Cancer Center. A population of 841 healthy female controls ages 20 to 87 years with well-defined ethnic background who had donated blood for various reasons (predominantly prenatal screening for noncancer disease) constituted the control group. Controls were of similar gender, ethnicity, and geographic location as the breast cancer cases. None of the controls had any personal history of cancer at the time of blood donation. This was ascertained by a questionnaire completed by each control. Although age categories were obtained for all participants, exact age information was not available for some controls because it was not collected prospectively. All personal identifiers were permanently removed from both cases and controls. A fraction of cases (n = 493) and controls (n = 330) included in this report have been genotyped for TGFBR1*6A and included in a recent meta-analysis (28).

DNA isolation. DNA from whole blood lymphocytes was extracted using the QIAamp DNA blood mini kit and stored at −20°C until use for genotyping.

TGFB1 genotyping. The first variant is a TGFB1 T-to-C point mutation at position 10 resulting in a leucine-to-proline substitution. Part of the TGFB1 gene was amplified by PCR according to the following conditions: initial denaturation at 95°C for 10 minutes followed by 35 cycles at 93°C for 20 seconds, 65°C for 30 seconds, and 72°C for 30 seconds. The last extension step was prolonged to 3 minutes. The reactions were carried out in a total volume of 50 μL containing 100 to 300 ng genomic DNA, 1× standard PCR buffer without MgCl2, 1.5 mmol/l MgCl2, 6% DMSO, 25 pmol of primers, 200 μmol/L of each deoxynucleotide triphosphate, and 1 unit Taq DNA polymerase. The primers used were 5′-TGCCGCCCTCCGGGCTGCGGCTGCGGC-3′ and 5′-TCTTGCAGGTGGATAGTCCCGCGGTCGG-3′. The PCR product is 102 bp long. The PCR product was cleaved with HaeIII overnight. The resulting fragments from the HaeIII digestion were separated on a 12% gel in 0.5× Tris-borate EDTA buffer and visualized with ethidium bromide. Digestion by restriction enzyme HaeIII generates polymorphic fragments of 69, 43, and 26 bp, respectively, and a 33-bp nonpolymorphic fragment. The various genotypes were confirmed by direct sequencing.

TGFBR1 genotyping. The second variant is a 9-bp deletion within a stretch of nine GCG repeats coding for alanine (29). PCR amplification was done using intronic primers flanking TGFBR1 exon 1: 5′-GAGGCGAGGTTTGCTGGGGTGAGGCA-3′ and 5′-CATGTTTGAGAAAGAGCAGGAGCGAG-3′. Genotyping was done as described previously (24). Briefly, PCR amplification was done according to the following conditions: initial denaturation at 95°C for 1 minute followed by 35 cycles at 94°C for 30 seconds and 68°C for 3 minutes. The last extension step was prolonged to 3 minutes. The reactions were carried out in a total volume of 25 μL containing 100 to 300 ng genomic DNA using the Advantage GC-genomic kit (Clontech, Palo Alto, CA). For single-strand conformational polymorphism analysis, PCR product (5 μL) was mixed with 10 μL formamide dye. The solution was heated at 95°C for 5 minutes, placed on ice water for 1 minute, and then loaded on the single-strand conformational polymorphism gel. The single-strand DNA fragments were resolved on a 20% Tris-borate EDTA acrylamide gel. Results for the different polymorphisms were confirmed by direct sequencing.

High, intermediate, and low signalers. Previous studies have shown that TGFBR1*6A is a hypomorphic form of TGFBR1 (24, 25). Hence, all TGFBR1*6A carriers were classified as low signalers, with the exception of individuals carrying both TGFBR1*6A and the TGFB1*CC genotype that were classified as intermediate signalers (Fig. 1). The TGFB1 T-to-C substitution has been shown to result in higher TGFB1 secretion (10). TGFB1*CC carriers have significantly higher TGFB1 circulating levels than carriers of the TGFB1*CT and TGFB1*TT genotypes (30). Individuals carrying the TGFB1*CC genotype and with two copies of the wild-type TGFBR1 (i.e., a TGFBR1 that transduces TGF-β signals more efficiently than TGFBR1*6A) were classified as high signalers. Individuals with other genotypic combinations were classified as intermediate signalers (Fig. 1).

Figure 1.

Genotypic combinations of TGFB1 and TGFBR1 functional variants and their predicted TGF-β signaling levels. X axis, various genotypic combinations of TGFB1 and TGFBR1 genotypes. Y axis, predicted level of TGF-β signaling based on in vitro functional assays and in vivo measurements in humans. We hypothesize that carriers of the TGFB1*CC and TGFBR1/TGFBR1 genotypes (high baseline TGF-β signaling) have the lowest breast cancer risk and carriers of the TGFB1*TT or TGFB1*CT and TGFBR1*6A genotypes (low baseline TGF-β signaling) have the highest breast cancer risk.

Figure 1.

Genotypic combinations of TGFB1 and TGFBR1 functional variants and their predicted TGF-β signaling levels. X axis, various genotypic combinations of TGFB1 and TGFBR1 genotypes. Y axis, predicted level of TGF-β signaling based on in vitro functional assays and in vivo measurements in humans. We hypothesize that carriers of the TGFB1*CC and TGFBR1/TGFBR1 genotypes (high baseline TGF-β signaling) have the lowest breast cancer risk and carriers of the TGFB1*TT or TGFB1*CT and TGFBR1*6A genotypes (low baseline TGF-β signaling) have the highest breast cancer risk.

Close modal

Immunohistochemistry. ER and PR were stained on paraffin-embedded, formalin-fixed tissue. Slides were subjected to antigen retrieval using 0.1 mol/L citric buffer (pH 6.2) and to microwave treatment for 10 minutes. For ER, monoclonal antibody clone 1D5 (Beckman Coulter, Immunotech, Miami, FL) was diluted at 1:20. Clone 1A6 monoclonal antibody against PR (Novocastra, Newcastle, United Kingdom) was diluted at 1:150. Detection was achieved using a biotinylated secondary antibody (BA-2000, Vector Laboratories, Burlingame, CA) in conjunction with the ABC kit (Vector Laboratories). The cutoff point for ER and PR positivity was 5% per institutional standards.

Statistical analysis. Distributions of TGFBR1 genotypes, TGFB1 genotypes, age, and ethnicity were compared between cases and controls using χ2 tests. To examine the effect of TGFBR1 and TGFB1 genotypes adjusting for covariates, ORs for breast cancer were estimated using unconditional logistic regression models. Both crude and adjusted ORs for breast cancer were estimated comparing carriers of TGFBR1*6A and C allele versus noncarriers for TGFBR1 and TGFB1, respectively, under dominant, additive, and recessive genetic susceptibility models. Age and ethnicity were controlled for in all analyses. Crude and adjusted ORs are similar, and we reported adjusted ORs. Additionally, adjusted ORs were estimated comparing joint states of the TGFBR1 and TGFB1 genotypes by a priori categorizing subjects as high, intermediate, and low signalers based on the known functional status of the variant alleles (refs. 10, 24, 25; Fig. 1). Trend statistics were conducted to assess whether risk of breast cancer increases according to this a priori categorization. A small P of the trend test indicates that breast cancer risk is related to the scoring based on TGFBR1 and TGFB1 genotypes. Whether the effects of TGFBR1 and TGFB1 on breast cancer differ by age was evaluated by stratified analysis and tests for multiplicative interaction. A small P indicates that interaction of age and gene is statistically significant on the multiplicative scale.

Sensitivity analysis was conducted to evaluate the effect of the fact that the exact age of a large number of controls is not known. In the subgroup of those with known exact age, ORs were estimated with or without adjusting age as a continuous variable. In the overall study population, ORs were estimated not only with or without adjusting age as a categorical variable but also under different scenarios by assigning controls with unknown exact age with values of lowest and highest extremes (i.e., all 20 or all 40). Analysis was also conducted for extreme scenarios where all carrier controls were assumed to be 40 years old and all noncarrier controls 20 years old and vice versa. An additional analysis was done to identify possible differences in TGFBR1 and TGFB1 allelic frequencies among the various age groups of controls for whom exact age was known. Because a subset of cases had data on tumor prognostic characteristics available, we conducted polytomous logistic regression to calculate ORs by comparing each of the case groups with the total controls with respect to TGFB1 and TGFBR1 status.

There were 658 cases and 841 controls genotyped for TGFB1 and 611 cases and 691 controls for TGFBR1. The mean age of cases was significantly higher than that of controls (P < 0.01; Table 1). The allelic frequencies of TGFB1 and TGFBR1 variants among controls were identical in all age groups.

Table 1.

Study population

TGFBR1 study population*
TGFB1 study population
Cases (n = 611), n (%)Controls (n = 690), n (%)PCases (n = 658), n (%)Controls (n = 841), n (%)P
TGFBR1 genotype       
    *9A/*9A 515 (84.3) 612 (88.7) 0.03 — —  
    *9A/*6A 92 (15.1) 77 (11.2)  — —  
    *6A/*6A 4 (0.6) 1 (0.1)  — —  
TGFB1 genotype       
    TT — —  200 (30.4) 240 (28.5) 0.23 
    TC — —  339 (51.5) 419 (49.9)  
    CC — —  119 (18.1) 182 (21.6)  
Age (y)§       
    20-40 89 (14.6) 394 (57.1) <0.01 97 (14.7) 534 (63.5) <0.01 
    41-50 166 (27.2) 82 (11.9)  181 (27.5) 84 (10.0)  
    51-60 168 (27.5) 110 (15.9)  178 (27.1) 112 (13.3)  
    61-70 120 (19.6) 69 (10.0)  127 (19.3) 75 (8.9)  
    71+ 68 (11.1) 35 (5.1)  75 (11.4) 36 (4.3)  
    Mean (SD)§ 54.0 (12.7) 55.3 (11.2)  53.9 (12.9) 55.4 (11.1)  
Race       
    White 512 (83.8) 541 (78.4) <0.01 544 (82.7) 649 (77.2) <0.01 
    Black 44 (7.2) 43 (6.2)  53 (8.1) 51 (6.1)  
    Hispanic 25 (4.1) 80 (11.6)  27 9 (4.1) 110 (13.1)  
    Asian 18 (3.0) 22 (3.2)  20 (3.0) 26 (3.1)  
    Unknown 12 (1.9) 4 (0.6)  14 (2.1) 5 (0.5)  
TGFBR1 study population*
TGFB1 study population
Cases (n = 611), n (%)Controls (n = 690), n (%)PCases (n = 658), n (%)Controls (n = 841), n (%)P
TGFBR1 genotype       
    *9A/*9A 515 (84.3) 612 (88.7) 0.03 — —  
    *9A/*6A 92 (15.1) 77 (11.2)  — —  
    *6A/*6A 4 (0.6) 1 (0.1)  — —  
TGFB1 genotype       
    TT — —  200 (30.4) 240 (28.5) 0.23 
    TC — —  339 (51.5) 419 (49.9)  
    CC — —  119 (18.1) 182 (21.6)  
Age (y)§       
    20-40 89 (14.6) 394 (57.1) <0.01 97 (14.7) 534 (63.5) <0.01 
    41-50 166 (27.2) 82 (11.9)  181 (27.5) 84 (10.0)  
    51-60 168 (27.5) 110 (15.9)  178 (27.1) 112 (13.3)  
    61-70 120 (19.6) 69 (10.0)  127 (19.3) 75 (8.9)  
    71+ 68 (11.1) 35 (5.1)  75 (11.4) 36 (4.3)  
    Mean (SD)§ 54.0 (12.7) 55.3 (11.2)  53.9 (12.9) 55.4 (11.1)  
Race       
    White 512 (83.8) 541 (78.4) <0.01 544 (82.7) 649 (77.2) <0.01 
    Black 44 (7.2) 43 (6.2)  53 (8.1) 51 (6.1)  
    Hispanic 25 (4.1) 80 (11.6)  27 9 (4.1) 110 (13.1)  
    Asian 18 (3.0) 22 (3.2)  20 (3.0) 26 (3.1)  
    Unknown 12 (1.9) 4 (0.6)  14 (2.1) 5 (0.5)  
*

The exact age was not known for 360 controls in the lowest age category (20-40 years).

The exact age was not known for 500 controls in the lowest age category (20-40 years).

P for χ2 or Fisher's exact test (comparing proportions).

§

Average age for controls was calculated based on those with exact age available.

Transforming growth factor-β signaling assessment.TGFBR1 and TGFB1 genotyping was done on 608 cases and 690 controls. There was a significantly lower proportion of high signalers (TGFB1*CC and TGFBR1) among cases (15.1%) than among controls (21.4%; P = 0.003). The proportion of intermediate signalers (72.0% versus 68.8%) and low signalers (12.8% versus 9.7%) was similar among cases and controls, respectively. Individuals defined as low TGF-β signalers had a significantly higher risk of breast cancer than those defined as high signalers (OR, 1.69; 95% CI, 1.08-2.66). The results were similar when ethnic status and age were adjusted (Table 2). Breast cancer risk incurred by low signalers was most pronounced among women ages ≥50 years (OR, 2.05; 95% CI, 1.01-4.16; Table 3).

Table 2.

Adjusted ORs of breast cancer by TGFBR1, TGFB1 genotypes, and TGF-β predicted signaling status

Gene/genotypen (cases/controls)OR (95% CI) for breast cancer risk*OR (95% CI) for breast cancer risk
TGFBR1    
    Dominant model    
        9A/9A 515/612 1.00 1.00 
        9A/6A or 6A/6A 96/78 1.46 (1.06-2.02) 1.50 (1.07-2.11) 
    Additive model    
        9A/9A 515/612 1.00 1.00 
        9A/6A 92/77 1.42 (1.03-1.96) 1.46 (1.04-2.06) 
        6A/6A 4/1 4.75 (0.53-42.66) 4.40 (0.48-40.52) 
    Recessive model    
        9A/9A or 9A/6A 607/689 1.00 1.00 
        6A/6A 4/1 4.54 (0.51-40.73) 4.19 (0.46-38.48) 
TGFB1    
    Dominant model    
        TT 200/240 1.00 1.00 
        TC/CC 458/601 0.91 (0.73-1.14) 0.98 (0.77-1.25) 
    Additive model    
        TT 200/240 1.00 1.00 
        TC 339/419 0.97 (0.78-1.23) 1.02 (0.79-1.32) 
        CC 119/182 0.79 (0.58-1.06) 0.89 (0.63-1.21) 
    Recessive model    
        TC or TT 539/659 1.00 1.00 
        CC 119/182 0.80 (0.62-1.03) 0.86 (0.65-1.14) 
TGF-β predicted signaling status    
    High signalers    
        CC/9A9A 92/148 1.00 1.00 
    Intermediate signalers    
        TT/9A9A, CC/9A6A, CC/6A6A, or TC/9A9A 438/475 1.48 (1.11-1.98) 1.27 (0.93-1.74) 
    Low signalers    
        TT/6A6A, TT/9A6A, TC/9A6A, or TC/6A6A 78/67 1.87 (1.23-2.84) 1.69 (1.08-2.66) 
    P for trend  0.02 0.02 
Gene/genotypen (cases/controls)OR (95% CI) for breast cancer risk*OR (95% CI) for breast cancer risk
TGFBR1    
    Dominant model    
        9A/9A 515/612 1.00 1.00 
        9A/6A or 6A/6A 96/78 1.46 (1.06-2.02) 1.50 (1.07-2.11) 
    Additive model    
        9A/9A 515/612 1.00 1.00 
        9A/6A 92/77 1.42 (1.03-1.96) 1.46 (1.04-2.06) 
        6A/6A 4/1 4.75 (0.53-42.66) 4.40 (0.48-40.52) 
    Recessive model    
        9A/9A or 9A/6A 607/689 1.00 1.00 
        6A/6A 4/1 4.54 (0.51-40.73) 4.19 (0.46-38.48) 
TGFB1    
    Dominant model    
        TT 200/240 1.00 1.00 
        TC/CC 458/601 0.91 (0.73-1.14) 0.98 (0.77-1.25) 
    Additive model    
        TT 200/240 1.00 1.00 
        TC 339/419 0.97 (0.78-1.23) 1.02 (0.79-1.32) 
        CC 119/182 0.79 (0.58-1.06) 0.89 (0.63-1.21) 
    Recessive model    
        TC or TT 539/659 1.00 1.00 
        CC 119/182 0.80 (0.62-1.03) 0.86 (0.65-1.14) 
TGF-β predicted signaling status    
    High signalers    
        CC/9A9A 92/148 1.00 1.00 
    Intermediate signalers    
        TT/9A9A, CC/9A6A, CC/6A6A, or TC/9A9A 438/475 1.48 (1.11-1.98) 1.27 (0.93-1.74) 
    Low signalers    
        TT/6A6A, TT/9A6A, TC/9A6A, or TC/6A6A 78/67 1.87 (1.23-2.84) 1.69 (1.08-2.66) 
    P for trend  0.02 0.02 
*

Crude ORs.

ORs were adjusted for ethnic groups and age as categorical variables.

P < 0.05.

Table 3.

Adjusted ORs of breast cancer by age groups (>50 or ≤50 years)

Gene/age groupsGenotypesn (cases/controls)OR (95% CI)*P for testing multiplicative interaction
TGFBR1     
    Age ≤50 y 9A/9A 217/417 1.00 0.09 
 9A/6A or 6A/6A 38/59 1.18 (0.75-1.84)  
    Age >50 y 9A/9A 298/195 1.00  
 9A/6A or 6A/6A 58/19 2.20 (1.25-3.87)  
TGFB1     
    Age ≤50 y TT or TC 223/477 1.00 0.99 
 CC 55/141 0.85 (0.57-1.29)  
    Age >50 y TT or TC 316/182 1.00  
 CC 64/41 0.87 (0.56-1.35)  
Joint status of TGFBR1 and TGFB1     
    Age ≤50 y High signalers 44/112 1.00 0.65 
 Intermediate signalers 177/314 1.33 (0.84-2.10)  
 Low signalers 32/50 1.49 (0.77-2.87)  
 P for trend  0.19  
    Age >50 y High signalers 48/36 1.00  
 Intermediate signalers 261/161 1.23 (0.76-1.98)  
 Low signalers 46/17 2.05 (1.01-4.16)  
 P for trend  0.06  
Gene/age groupsGenotypesn (cases/controls)OR (95% CI)*P for testing multiplicative interaction
TGFBR1     
    Age ≤50 y 9A/9A 217/417 1.00 0.09 
 9A/6A or 6A/6A 38/59 1.18 (0.75-1.84)  
    Age >50 y 9A/9A 298/195 1.00  
 9A/6A or 6A/6A 58/19 2.20 (1.25-3.87)  
TGFB1     
    Age ≤50 y TT or TC 223/477 1.00 0.99 
 CC 55/141 0.85 (0.57-1.29)  
    Age >50 y TT or TC 316/182 1.00  
 CC 64/41 0.87 (0.56-1.35)  
Joint status of TGFBR1 and TGFB1     
    Age ≤50 y High signalers 44/112 1.00 0.65 
 Intermediate signalers 177/314 1.33 (0.84-2.10)  
 Low signalers 32/50 1.49 (0.77-2.87)  
 P for trend  0.19  
    Age >50 y High signalers 48/36 1.00  
 Intermediate signalers 261/161 1.23 (0.76-1.98)  
 Low signalers 46/17 2.05 (1.01-4.16)  
 P for trend  0.06  
*

ORs were adjusted for ethnic groups and age within age strata.

P < 0.05.

Low signalers were those with TT/6A6A, TT/9A6A, TC/9A6A, or TC/6A6A; intermediate signalers were those with TT/9A9A, CC/9A6A, CC/6A6A, or TC/9A9A; and high signalers were those with CC/9A9A.

TGFBR1*6A and breast cancer risk.TGFBR1*6A allelic frequency was significantly higher among cases (0.082) than among controls (0.057; P = 0.03; Table 1). Breast cancer risk was significantly higher for TGFBR1*6A carriers under both a dominant (OR, 1.50; 95% CI, 1.07-2.07) and a recessive (OR, 1.46; 95% CI, 1.04-2.06) model (Table 2). Breast cancer risk for TGFBR1*6A carriers ages >50 years was almost twice higher (OR, 2.20; 95% CI, 1.25-3.87) than among women ages ≤50 years (OR, 1.18; 95% CI, 0.75-1.84; Table 3). To assess the possibility that the effect of TGFBR1*6A on breast cancer risk is nullified by the TGFB1 T29C variant, we restricted the analysis to TGFB1*CC and TGFB1*TC carriers. Breast cancer risk remained significantly higher for TGFBR1*6A carriers (OR, 1.58; 95% CI, 1.04-2.39).

TGFB1*CC and breast cancer risk. About 18.1% of cases and 21.6% of controls carried the TGFB1*CC genotype (Table 1). There was a trend toward decreased breast cancer risk for carriers of the TGFB1*CC genotype, but it did not reach formal significance (OR, 0.89; 95% CI, 0.63-1.21). The results were similar under either an additive or a recessive model. Adjustment for age and ethnic status did not change the results (Table 2). The observed trend toward decreased breast cancer for TGFBR1*CC carriers was similar for women above or below age 50 years (Table 3).

Estrogen receptor, progesterone receptor, and stage at diagnosis. Information on ER and PR status and stage at diagnosis was available for 152 patients. There was no association between TGFBR1*6A and ER and PR status and stage at diagnosis (Table 4). On the contrary, both TGFB1*CT and TGFB1*CC were more likely to have advanced stage as assessed by the presence of lymph node metastasis and stage III or IV at diagnosis (Table 4). Patients carrying the TGFB1*CC genotype were more likely to have ER-negative and PR-negative tumors (Table 4).

Table 4.

Association of TGFB1 and TGFBR1 genotypes with ER, PR, and stage at diagnosis

TGFB1 genotype, adjusted OR (95% CI)
TGFBR1 genotype, adjusted OR (95% CI)
n (cases/controls)TTTCCCn (cases/controls)9A/9A9A/6A or 6A/6A
ER status        
    ER positive vs controls 72/841 1.00 1.27 (0.70-2.29) 0.94 (0.44-2.03) 56/690 1.00 1.51 (0.67-3.42) 
    ER negative vs controls 27/841 1.00 0.26 (0.10-0.64)* 0.36 (0.12-1.10) 21/690 1.00 2.07 (0.67-6.44) 
PR status        
    PR positive vs controls 53/841 1.00 1.10 (0.59-2.13) 0.90 (0.39-2.11) 41/690 1.00 1.54 (0.61-3.85) 
    PR negative vs controls 44/841 1.00 0.55 (0.28-1.09) 0.51 (0.21-1.28) 35/690 1.00 1.86 (0.73-4.76) 
Joint status of ER and PR        
    Both are positive vs controls 51/841 1.00 1.37 (0.62-3.05) 0.87 (0.36-2.11) 39/690 1.00 1.65 (0.65-4.17) 
    Either is negative vs controls 21/841 1.00 0.89 (0.30-2.65) 1.44 (0.40-5.16) 18/690 1.00 1.13 (0.25-5.09) 
    Both are negative vs controls 25/841 1.00 0.97 (0.25-3.84) 0.28 (0.08-1.00)* 19/690 1.00 2.36 (0.75-7.44) 
Stage        
    Stage I/II vs controls 117/841 1.00 0.83 (0.53-1.30) 0.62 (0.34-1.14) 90/690 1.00 1.61 (0.85-3.06) 
    Stage III/IV vs controls 30/841 1.00 0.46 (0.21-1.00)* 0.26 (0.07-0.93)* 24/690 1.00 0.81 (0.18-3.57) 
Lymph nodes        
    0 48/841 1.00 1.34 (0.64-2.80) 1.47 (0.62-3.45) 32/690 1.00 1.28 (0.43-3.82) 
    ≥1 44/841 1.00 0.44 (0.22-0.86)* 0.35 (0.14-0.90)* 40/690 1.00 1.57 (0.63-3.93) 
TGFB1 genotype, adjusted OR (95% CI)
TGFBR1 genotype, adjusted OR (95% CI)
n (cases/controls)TTTCCCn (cases/controls)9A/9A9A/6A or 6A/6A
ER status        
    ER positive vs controls 72/841 1.00 1.27 (0.70-2.29) 0.94 (0.44-2.03) 56/690 1.00 1.51 (0.67-3.42) 
    ER negative vs controls 27/841 1.00 0.26 (0.10-0.64)* 0.36 (0.12-1.10) 21/690 1.00 2.07 (0.67-6.44) 
PR status        
    PR positive vs controls 53/841 1.00 1.10 (0.59-2.13) 0.90 (0.39-2.11) 41/690 1.00 1.54 (0.61-3.85) 
    PR negative vs controls 44/841 1.00 0.55 (0.28-1.09) 0.51 (0.21-1.28) 35/690 1.00 1.86 (0.73-4.76) 
Joint status of ER and PR        
    Both are positive vs controls 51/841 1.00 1.37 (0.62-3.05) 0.87 (0.36-2.11) 39/690 1.00 1.65 (0.65-4.17) 
    Either is negative vs controls 21/841 1.00 0.89 (0.30-2.65) 1.44 (0.40-5.16) 18/690 1.00 1.13 (0.25-5.09) 
    Both are negative vs controls 25/841 1.00 0.97 (0.25-3.84) 0.28 (0.08-1.00)* 19/690 1.00 2.36 (0.75-7.44) 
Stage        
    Stage I/II vs controls 117/841 1.00 0.83 (0.53-1.30) 0.62 (0.34-1.14) 90/690 1.00 1.61 (0.85-3.06) 
    Stage III/IV vs controls 30/841 1.00 0.46 (0.21-1.00)* 0.26 (0.07-0.93)* 24/690 1.00 0.81 (0.18-3.57) 
Lymph nodes        
    0 48/841 1.00 1.34 (0.64-2.80) 1.47 (0.62-3.45) 32/690 1.00 1.28 (0.43-3.82) 
    ≥1 44/841 1.00 0.44 (0.22-0.86)* 0.35 (0.14-0.90)* 40/690 1.00 1.57 (0.63-3.93) 
*

P < 0.05.

TGFBR1 and TGFB1 map to different chromosomes but are essential components of the same signaling pathway. We have hypothesized previously that a combined assesment of the functionally relevant common variants may help better characterize cancer risk and may predict disease aggressiveness (31, 32). In a previous study, TGFRB1*6A was associated with increased breast cancer risk among women with a mean age at diagnosis of 38 years. These women had been selected based on age at onset under 40 years, a family member with breast cancer irrespective of age at diagnosis, or bilated breast cancer irrespective of family history or age at disease onset (26). The mean age at diagnosis of breast cancer cases presented in this study is similar to the mean age at diagnosis of breast cancer in the general population. Our findings of a significant association of TGFBR1*6A with breast cancer suggest that the effect of TGFBR1*6A can be extended beyond familial, early-onset, and bilateral breast cancer. The data add to the growing body of evidence that TGFBR1*6A may act as breast tumor susceptibility gene (2628). Interestingly, a subset analysis of our results indicate that TGFBR1*6A effects are most prominent among women ages >50 years. The hypomorphic TGFBR1*6A allele was not associated with more aggressive tumor behavior, which is in agreement with a recent report indicating that transgenic mice with decreased TGF-β signaling have an increased incidence of breast cancer but a reduced incidence of breast cancer metastases (8).

TGFB1 exerts pleiotropic effects in the oncogenesis of breast cancer in a contextual manner (i.e., it suppresses tumorigenesis at an early stage by direct inhibition of angiogenesis and tumor cell growth). However, overproduction of TGFB1 by advanced tumors may accelerate disease progression through indirect stimulation of angiogenesis and immune suppression (33). Evidence also exists that TGF-β signaling contributes to the metastasis in breast cancer (34, 35) and that TGF-β signaling blockade inhibits mammary tumor cell viability, migration, and metastasis (36). The contribution of TGF-β to the malignant phenotype of breast cancer cells is particularly prominent in cell lines that retain the TGF-β signal transduction system but have lost TGF-β-induced growth inhibition. Such is the case in breast cancer cells with a hyperactive Ras pathway (35, 37, 38).

Both stromal and epithelial cells from TGFB1*CC are likely surrounded by more TGFB1 than cells from TGFB1*TT individuals given the fact that the TGFB1 Leu10Pro (T29C) results in higher in vitroTGFB1 secretion and the TGFB1*CC genotype is associated with higher TGFB1 serum levels (10, 39). In our report, the higher frequency of lymph node metastases and the more advanced stage at diagnosis in carriers of the TGFB1*CC and TGFB1*CT genotypes are in agreement with these laboratory findings and with the recent report of a significantly reduced 5-year survival among patients with breast cancer that carry the TGFB1*CC and TGFB1*CT genotypes (40).

Taken together, these results are additional evidence that increased TGF-β signaling due to a naturally occurring variant is associated with a more aggressive tumor behavior. If confirmed in larger studies, TGFB1 genotype may become a new prognostic marker for women diagnosed with breast cancer and the TGF-β signaling pathway may become a molecular target for therapeutic interventions. The additional finding of an increased proportion of ER-negative and PR-negative tumors among carriers of the TGFB1*CC genotype points to the fact that increased TGF-β signaling results in more aggressive tumor behavior in the absence of ER and PR overexpression.

This may explain the conflicting results of the Ziv et al. (10), Dunning et al. (15), and Le Marchand et al. (18) studies. Among premenopausal women, the proportion of tumors overexpressing ER and PR is lower than among postmenopausal women. Hence, the growth of ER-negative and PR-negative tumors from premenopausal women may benefit from increased TGFB1 levels, whereas growth of the predominantly ER-positive and PR-positive tumors from postmenopausal women may not be similarly affected by higher TGFB1 levels. Subjects enrolled in the Ziv et al. study were ages ≥65 years with a mean age of 70 years. Subjects in the Le Marchand et al. study were predominantly postmenopausal with a mean age of 63 years. Conversely, the Dunning et al. study reported a combined analysis of three case-control studies, including breast cancer cases with a mean age of 50 years, and showed that TGFB1*CC and TGFB1*CT carriers had a slight but significant increased risk of invasive breast cancer. Although not significant, our results show a trend similar to that of the former study. The mean age of our cases was 5 years higher than the Dunning study, 15 and 10 years lower than the Ziv et al. and Le Marchand et al. studies, respectively. Given that the mean age at menopause in the general population is 50 years, the proportion of postmenopausal women in our population is likely higher than in the Dunning et al. study and may explain the nonsignificant trend toward a TGFB1*CC protective effect. The findings that differences in TGF-β signaling effects are more pronounced among postmenopausal women further support this explanation. However, another plausible explanation for these conflicting results is the functional interaction between TGFBR1*6A and the TGFB1 T29C polymorphism shown in this report.

Our study has several limitations. Due to destruction of personal identifiers, we only had exact age information in a subset of our controls. For the remainder of controls, only the age range was available. Furthermore, cases and controls were not matched for variables known to be associated with breast cancer. However, we did analyses controlling for age. We also conducted sensitivity analyses using hypothetical models to show that the effect of the lack of detailed age information in a portion of our samples was negligible. It is possible that age differences in cases and controls affected the allele frequencies observed. Nonetheless, this would be expected to create a bias toward the null hypothesis because it would overestimate the deleterious allele frequency in controls given that a fraction of younger women who would have developed breast cancer were not removed from the control group. Thus, all the younger mean age of controls could have resulted in a bias toward the null hypothesis, resulting in a weaker association. In addition, an additional limitation of our study is the lack of analysis of other TGFB1 polymorphisms, some of which may modify TGFB1 circulating levels (41).

Another potential drawback is the lack of complete pathologic information in our cases. Due to the destruction of the personal identifiers, we could not retrospectively collect data on ER/PR and the lymph node status in a subset of patients. However, taking into account the limited number of cases with complete histopathologic information, our significant results in the advanced stage and ER-negative population merit further investigation in a large prospective study.

The results presented here reflect a nonselected population of patients with breast cancer. TGFBR1*6A carriers (15.7% of cases and 11.3% of controls) and high TGF-β signalers (15.1% of cases and 21.4% of controls) make up >30% of the population of both cases and controls. This is evidence that variants of the TGF-β signaling pathway are likely to modify breast cancer risk in a large subset of the population. Studies are in progress to determine the contribution of the TGF-β signaling pathway variants to familial and sporadic breast cancer.

The combined analysis of high versus low signalers is the first indication in humans that altered TGF-β signaling modifies breast cancer risk. It identifies high signalers as a subgroup of individuals with increased TGF-β signaling and decreased breast cancer risk, although, taken separately, the proportion of TGFB1*CC and TGFBR1/TGFBR1 carriers was similar among cases and controls. It also shows that combination of naturally occurring TGF-β signaling pathway variants probably result in functional differences large enough in vivo to modify breast cancer risk, similar to what has been shown recently in transgenic mouse models (8). These results warrant validation in well-designed case-control studies to explore further the role of TGF-β signaling pathway variants with respect to breast cancer risk and outcome.

Grant support: National Cancer Institute grants CA90386 and CA89018, Avon Breast Cancer Career Development Award (B. Pasche), Mander Foundation, Lomangino and Weissenbach-Southworth Family Research Funds, Lymphoma Foundation, and New York Cancer Project administered and funded by AMDeC Foundation, Inc.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Arteaga CL, Dugger TC, Hurd SD. The multifunctional role of transforming growth factor (TGF)-β on mammary epithelial cell biology.
Breast Cancer Res Treat
1996
;
38
:
49
–56.
2
Reiss M, Barcelloshoff MH. Transforming growth factor-β in breast cancer—a working hypothesis [review].
Breast Cancer Res
1997
;
45
:
81
–95.
3
Wakefield LM, Roberts AB. TGF-β signaling: positive and negative effects on tumorigenesis.
Curr Opin Genet Dev
2002
;
12
:
22
–9.
4
Akhurst RJ, Derynck R. TGF-β signaling in cancer—a double-edged sword.
Trends Cell Biol
2001
;
11
:
S44
–51.
5
Sporn MB, Roberts AB. TGF-β: problems and prospects.
Cell Regul
1990
;
1
:
875
–82.
6
Pierce DF Jr, Gorska AE, Chytil A, Meise KS, Page DL, Coffey RJ Jr. Mammary tumor suppression by transforming growth factor β1 transgene expression.
Proc Natl Acad Sci U S A
1995
;
92
:
4254
–8.
7
Cui W, Fowlis DJ, Bryson S, et al. TGFβ1 inhibits the formation of benign skin tumors, but enhances progression to invasive spindle carcinomas in transgenic mice.
Cell
1996
;
86
:
531
–42.
8
Siegel PM, Shu W, Cardiff RD, Muller WJ, Massague J. Transforming growth factor β signaling impairs Neu-induced mammary tumorigenesis while promoting pulmonary metastasis.
Proc Natl Acad Sci
2003
;
100
:
8430
.
9
Syrris P, Carter ND, Metcalfe JC, et al. Transforming growth factor-β gene polymorphisms and coronary artery disease.
Clin Sci
1998
;
95
:
659
–67.
10
Dunning AM, Ellis PD, McBride S, et al. A transforming growth factor β1 signal peptide variant increases secretion in vitro and is associated with increased incidence of invasive breast cancer.
Cancer Res
2003
;
63
:
2610
–5.
11
Grainger DJ, Heathcote K, Chiano M, et al. Genetic control of the circulating concentration of transforming growth factor type β1.
Hum Mol Genet
1999
;
8
:
93
–7.
12
Cambien F, Ricard S, Troesch A, et al. Polymorphisms of the transforming growth factor-β-1 gene in relation to myocardial infarction and blood pressure—the Etude Cas-Temoin de Linfarctus du Myocarde (ECTIM) Study.
Hypertension
1996
;
28
:
881
–7.
13
Holweg CT, Baan CC, Niesters HG, et al. TGF-β1 gene polymorphisms in patients with end-stage heart failure.
J Heart Lung Transplant
2001
;
20
:
979
–84.
14
Beranek M, Kankova K, Benes P, et al. Polymorphism R25P in the gene encoding transforming growth factor-β (TGF-β1) is a newly identified risk factor for proliferative diabetic retinopathy.
Am J Med Genet
2002
;
109
:
278
–83.
15
Ziv E, Cauley J, Morin PA, Saiz R, Browner WS. Association between the T29 → C polymorphism in the transforming growth factor β1 gene and breast cancer among elderly white women: The study of osteoporotic fractures.
JAMA
2001
;
285
:
2859
–63.
16
Hishida A, Iwata H, Hamajima N, et al. Transforming growth factor B1 T29C polymorphism and breast cancer risk in Japanese women.
Breast Cancer
2003
;
10
:
63
–9.
17
Krippl P, Langsenlehner U, Renner W, et al. The L10P polymorphism of the transforming growth factor-β1 gene is not associated with breast cancer risk.
Cancer Lett
2003
;
201
:
181
–4.
18
Marchand LL, Haiman CA, van den Berg D, Wilkens LR, Kolonel LN, Henderson BE. T29C Polymorphism in the transforming growth factor β1 gene and postmenopausal breast cancer risk: the Multiethnic Cohort Study.
Cancer Epidemiol Biomarkers Prev
2004
;
13
:
412
–5.
19
Shu XO, Gao YT, Cai Q, et al. Genetic polymorphisms in the TGF-β1 gene and breast cancer survival: a report from the Shanghai Breast Cancer Study.
Cancer Res
2004
;
64
:
836
–9.
20
Suthanthiran M, Li BG, Song JO, et al. Transforming growth factor-β(1) hyperexpression in African-American hypertensives: a novel mediator of hypertension and/or target organ damage.
Proc
2000
;
97
:
3479
–84.
21
Yamada Y, Miyauchi A, Goto J, et al. Association of a polymorphism of the transforming growth factor-β-1 gene with genetic susceptibility to osteoporosis in postmenopausal Japanese women.
J Bone Miner Res
1998
;
13
:
1569
–76.
22
Yokota M, Ichihara S, Lin TL, Nakashima N, Yamada Y. Association of a T29 → C polymorphism of the transforming growth factor-β1 gene with genetic susceptibility to myocardial infarction in Japanese.
Circulation
2000
;
101
:
2783
–7.
23
Pasche B, Luo Y, Rao PH, et al. Type I transforming growth factor β receptor maps to 9q22 and exhibits a polymorphism and a rare variant within a polyalanine tract.
Cancer Res
1998
;
58
:
2727
–32.
24
Chen T, de Vries EG, Hollema H, et al. Structural alterations of transforming growth factor-β receptor genes in human cervical carcinoma.
Int J Cancer
1999
;
82
:
43
–51.
25
Pasche B, Kolachana P, Nafa K, et al. T β R-I(6A) is a candidate tumor susceptibility allele.
Cancer Res
1999
;
59
:
5678
–82.
26
Baxter SW, Choong DY, Eccles DM, Campbell IG. Transforming growth factor β receptor 1 polyalanine polymorphism and exon 5 mutation analysis in breast and ovarian cancer.
Cancer Epidemiol Biomarkers Prev
2002
;
11
:
211
–4.
27
Kaklamani VG, Hou N, Bian Y, et al. TGFBR1*6A and cancer risk: a meta-analysis of seven case-control studies.
J Clin Oncol
2003
;
21
:
3236
–43.
28
Pasche B, Kaklamani VG, Hou N, et al. TGFBR1*6A and cancer: a meta-analysis of 12 case-control studies.
J Clin Oncol
2004
;
22
:
756
–8.
29
Pasche B, Luo Y, Rao PH, et al. Type I transforming growth factor β receptor maps to 9q22 and exhibits a polymorphism and a rare variant within a polyalanine tract.
Cancer Res
1998
;
58
:
2727
–32.
30
Yokota M, Ichihara S, Lin TL, Nakashima N, Yamada Y. Association of a T29 → C polymorphism of the transforming growth factor-β1 gene with genetic susceptibility to myocardial infarction in Japanese.
Circulation
2000
;
101
:
2783
–7.
31
Pasche B. Role of transforming growth factor β in cancer.
J Cell Physiol
2001
;
186
:
153
–68.
32
Bian Y, Kaklamani V, Reich J, Pasche B. TGF-β signaling alterations in cancer.
Cancer Treat Res
2003
;
115
:
73
–94.
33
Roberts AB, Wakefield LM. The two faces of transforming growth factor β in carcinogenesis.
Proc Natl Acad Sci
2003
;
100
:
8621
.
34
Bandyopadhyay A, Zhu Y, Cibull ML, Bao LW, Chen CG, Sun LZ. A soluble transforming growth factor β type III receptor suppresses tumorigenicity and metastasis of human breast cancer MDA-MB-231 cells.
Cancer Res
1999
;
59
:
5041
–6.
35
Yin JJ, Selander K, Chirgwin JM, et al. TGF-β signaling blockade inhibits PTHrP secretion by breast cancer cells and bone metastases development.
J Clin Invest
1999
;
103
:
197
–206.
36
Muraoka RS, Dumont N, Ritter CA, et al. Blockade of TGF-β inhibits mammary tumor cell viability, migration, and metastases.
J Clin Invest
2002
;
109
:
1551
–9.
37
Oft M, Peli J, Rudaz C, Schwarz H, Beug H, Reichmann E. TGF-β-1 and HA-ras collaborate in modulating the phenotypic plasticity and invasiveness of epithelial tumor cells.
Genes Dev
1996
;
10
:
2462
–77.
38
Oft M, Heider KH, Beug H. TGF-β signaling is necessary for carcinoma cell invasiveness and metastasis.
Curr Biol
1998
;
8
:
1243
–52.
39
Yokota M, Ichihara S, Lin TL, Nakashima N, Yamada Y. Association of a T29 → C polymorphism of the transforming growth factor-β1 gene with genetic susceptibility to myocardial infarction in Japanese.
Circulation
2000
;
101
:
2783
–7.
40
Shu XO, Gao YT, Cai Q, et al. Genetic polymorphisms in the TGF-β1 gene and breast cancer survival: a report from the Shanghai Breast Cancer Study.
Cancer Res
2004
;
64
:
836
–9.
41
Akhurst RJ. TGF β signaling in health and disease.
Nat Genet
2004
;
36
:
790
–2.