TβR-I(6A) Is a Candidate Tumor Susceptibility Allele¹

Unrestricted cell growth due to a lack of growth inhibitory activity appears as the most important of the possible consequences of a defect in TGF-β function. This prediction was recently fulfilled by the discovery that mice with decreased TGF-β levels have an increased susceptibility to tumor development (1). Inactivating mutations of the type II TGF³-β receptor (TβR-II) have been reported in gastrointestinal cancers with microsatellite instability (2, 3, 4). Two repetitive dinucleotide sequences within the coding region of TβR-II are prone to replication errors that encode truncated, inactive receptors. Missense mutations elsewhere in TβR-II have been described in T-cell lymphomas and head and neck carcinomas (5, 6). Homozygous deletion of TβR-I was reported in pancreatic and biliary carcinomas (7). Additionally, in several human cancer cell lines that lack active TGF-β receptors, restoration of functional receptors reverses the transformed phenotype of the cell lines (8, 9).

Targeting of the TGF-β pathway in cancer is further demonstrated by the identification of inactivating mutations in Smad4/DPC4 and Smad2. Smad4/DPC4 was originally cloned as a tumor suppressor gene that is deleted or mutated in half of human pancreatic carcinomas (10). Smad4/DPC4 mutations have also been found in carcinomas of the colon, esophagus, breast, ovary, lung and head, and neck cancer (11, 12, 13). Missense and other mutations of Smad2 occur in colon cancer (14). Recent reports indicate that germ-line mutations of TβR-II (15) and Smad4/DPC4 (16) may predispose to the development of hereditary nonpolyposis colorectal cancer and juvenile polyposis, respectively.

Reduced expression of TβR-I has been noted in colon and prostate cancer (17, 18), and a tumor-specific mutation was recently reported in breast cancer (19). We have previously described a type I TGF-β receptor (TβR-I) polymorphic allele, Tβ R-I(6A), which has a deletion of three alanines from a nine-alanine stretch (20). We had observed a higher than expected frequency of TβR-I(6A) homozygotes in both tumor and nontumor DNA samples from patients with a diagnosis of cancer (20), suggesting that TβR-I(6A) might contribute to cancer development.

To test the hypothesis of an association of this mutant allele with neoplasms, we performed a case-control study to determine the relative frequencies of this allele in the germ-line of patients with various malignancies and controls of identical gender, geographical, and ethnic background. We found nine TβR-I(6A) homozygotes among 851 patients with cancer. In comparison, there were no TβR-I(6A) homozygotes among 735 healthy volunteers (P < 0.01). We also observed an excess of TβR-I(6A) heterozygotes in cancer cases compared to controls (14.6% versus 10.6%; P < 0.02, Fisher’s exact test). A subset analysis revealed that 4 of 112 patients with colon cancer were TβR-I(6A) homozygotes (P < 0.01).

To further investigate these observations, we stably transfected mink lung epithelium cell lines devoid of functional TβR-I with Tβ R-I(6A) and TβR-I. TGF-β growth-inhibition experiments showed that, compared to TβR-I, TβR-I(6A) was impaired as a mediator of antiproliferative signals. We conclude that TβR-I(6A) acts as a tumor susceptibility allele that may contribute to the development of cancer, in particular, colon cancer. Up to 9% of colorectal neoplasia may be attributed to heterozygosity or homozygosity of this allele in the general population.

As part of institutional review board-approved protocols, samples from 851 patients were analyzed for TβR-I genotype. These included 80 nontumor DNA specimens from patients with cancer reported previously (20) and 771 specimens consecutively ascertained from patients admitted to the Memorial Sloan-Kettering Cancer Center. DNA was extracted from peripheral blood after the completion of diagnostic studies on these samples. Information regarding sex, age, and ethnic status (race and religion) was recorded; all other personal identifiers were permanently removed. Detailed clinical and family history information was available only from the initial 80 specimens, of which 40 were from patients with colorectal cancer (20). A population of 617 male and female healthy individuals with a well-defined geographic and ethnic background from the New York University Medical Center who had donated blood for various reasons (predominantly prenatal screening for noncancer hereditary disease) as well as 118 normal blood donors described previously (20) constituted the control group. None of the controls had any history of cancer at the time of the blood donation. This was ascertained by a questionnaire filled by each control. All cases and controls resided in the New York metropolitan region, and both groups were similar with respect to the proportion of Caucasian (80%), African-American (5%), Hispanic (10%), or Asian (5%) origin. Cases and controls matched for gender and ethnic status. However, the mean age of the cases was 56 years, and the mean age of 376 of the controls on which age data were available was 35 years. Additionally, 50 controls and 347 histologically confirmed cancer cases from the Turin metropolitan area without any additional identifiers were available for screening.

Genomic DNA (200 ng) were amplified as described previously (20). All genotypes except for TβR-I/TβR-I and TβR-I/TβR-I(6A) were confirmed by either direct sequencing or restriction enzyme digestion of the product. We used BssSI because this enzyme recognizes a unique site within the amplified fragment and yields a common 66-bp fragment and fragments of 44-, 53-, and 56-bp for TβR-I(6A), TβR-I, and TβR-I(10A), respectively.

Cells devoid of TβR-I were transfected with vectors encoding TβR-I, TβR-I(6A), and TβR-I(10A). To facilitate the quantitation of the transfected receptor, the TβR-I, TβR-I(6A), and TβR-I(10A) constructs encoded an influenza virus HA epitope at the COOH-terminus of the receptor. This epitope does not alter the signaling activity of TβR-I (21). Growth-inhibition assays were performed over a wide range of TGF-β concentrations.

pCMV5 vectors encoding TβR-I-HA, TβR-I(6A)-HA, or Tβ R-I(10A)-HA were prepared as described previously (20). The pUHD10-3 hygromycin vectors were generated by inserting each of the HA-tagged alleles into the linearized plasmid. The proper alignment of the construct was verified by sequencing.

To analyze the signaling activity of the various alleles, R1B/L17 cells containing the pUHD15-1 tetracycline repressor plasmid (22) were stably transfected with the pUHD10-3 hygromycin vectors using the lipofectin procedure according to the manufacturer’s protocol (Life Technologies, Inc.). This cell line is derived from Mv1Lu mink lung epithelial cells by chemical mutagenesis (23) and is defective in TβR-I and completely insensitive to TGF-β1 transcriptional and antiproliferative responses (24, 25). Clones were selected by ring cloning in 1 mg/ml of G-418 (Geneticin, Life Technologies, Inc., Life Technologies, Inc.), 0.3 mg/ml of hygromycin (Calbiochem), and 2 μg/ml of tetracycline. To examine the expression of the exogenous TβR-I, the clones were grown to 50% confluency. The culture medium was then switched to medium with or without 2 μg/ml of tetracycline. After 18 h, the cells were lysed in hypotonic buffer, and TβR-I expression was assessed by Western immunoblotting of whole-cell extracts (10 μg) using the 3F10 anti-HA antibody (Boehringer Mannheim). The expression of the exogenous TβR-I was not repressible in the presence of tetracycline in most of our clones.

Clones with identical levels of expression in the absence of tetracycline were selected for TGF-β growth-inhibition assays. TGF-β growth inhibition was performed in medium containing 10% fetal bovine serum as described before (24). All experiments were performed in triplicates with four different clones from each of the TβR-I, Tβ R-I(6A), and TβR-I(10A) cell lines.

TGF-β receptor affinity-labeling assays using ¹²⁵I-TGF-β1 and disuccinyl suberimidate were performed as previously described (26). Stably transfected cell lines with a comparable high level of receptor expression were placed in methionine/cysteine-deficient medium for 1 h. Pulse-chase labeling of the cells was performed with ³⁵S-methionine/cysteine (Amersham) for 15 min. After the removal of the pulse medium, the cells were washed extensively and chased in regular DMEM supplemented with 10% FCS for up to 8 h. At each time point, the cells were lysed in TNE buffer supplemented with protease and phosphatase inhibitors. Immunoprecipitation using the 3F10 anti-HA antibody (Boehringer Mannheim) was performed. The immunoprecipitates were subsequently washed and separated by SDS-PAGE, followed by autoradiography.

TβR-I(6A) homozygotes were observed in 9 of the 851 cases and in none of the 735 controls (P < 0.01, Fisher’s exact test). Of the 851 cancer cases, 124 were TβR-I(6A) heterozygotes [1 TβR-I(6A)/TβR-I(10A) and 123 TβR-I/TβR-I(6A); 14.6%] compared to 78 among 735 controls (10.6%; P = 0.02; Table 1). Of the 112 colon cancer cases analyzed (Table 2), four were TβR-I(6A) homozygotes, compared to 735 controls in which no TβR-I(6A) homozygotes were observed (P < 0.01). The over-representation of TβR-I(6A) homozygotes in colon cancer cases was confirmed in an analysis of a subset of the study population presumed to be in a Hardy-Weinberg equilibrium. Two TβR-I(6A) heterozygotes and two TβR-I(6A) homozygotes were observed in 25 colon cancer cases of Ashkenazi background. Based on a TβR-I(6A) heterozygote frequency of 12% observed in Ashkenazi controls, the finding of two homozygotes is highly significant (P < 0.01).

Subsets with or without the TβR-I(6A) allele were similar with respect to the male/female proportion and the proportion of various ethnic backgrounds. Among the subset of cases with colorectal cancer, the median age of onset of disease was 56 years in the four TβR-I(6A) homozygotes, compared to 58 years in 107 patients who were wild type for at least one allele. Of the four TβR-I(6A) homozygotes with colon cancer, all were Caucasian; two were of Ashkenazi Jewish ancestry, and two were non-Jewish. One of the TβR-I(10A) cases was an African-American female with colon cancer, and the others were a female of Ashkenazi Jewish origin who had ovarian cancer and a Caucasian non-Jewish male with germ cell cancer.

Whereas two TβR-I(10A) heterozygotes and a single TβR-I(10A)/TβR-I(6A) compound heterozygote were observed in patients with germ cell cancer, ovarian cancer, and colon cancer, respectively, no TβR-I(10A) heterozygotes or compounds were observed in the 735 subjects unaffected by cancer. However, in the control group, there were two TβR-I(8A) heterozygotes and one TβR-I(5A) heterozygote characterized by a stretch of eight and five alanines, respectively. One TβR-I(8A) heterozygote was of African descent from Jamaica, the other was of Hispanic (Puerto Rican) origin. The only TβR-I(5A) heterozygote was also from Puerto Rico. Hence, individuals of varied ethnic backgrounds were found to carry a total of five TβR-I alleles: TβR-I(5A), TβR-I(6A), Tβ R-I(8A), TβR-I, and TβR-I(10A). Although the TβR-I(10A) allele was noted in three patients with a diagnosis of cancer, small numbers preclude epidemiological correlation. Tβ R-I(6A) homozygotes were also noted in single cases of lymphoma, non-small cell lung cancer, and ovarian cancer, and two cases of germ cell cancer (Table 1).

A separate group of patient samples from Northern Italy was investigated. One TβR-I(6A) homozygote and one TβR-I(10A) heterozygote were observed in cancer patients but not in controls. We found 1 TβR-I(6A) homozygote among 48 cases of breast cancer. No TβR-I(6A) homozygotes were observed among 50 controls and 65 cases of colon cancer. There was one TβR-I/TβR-I(10A) but no TβR-I(6A) homozygotes among 234 Italian bladder cancer cases. Similarly, no TβR-I(6A) homozygotes were found among 77 New York patients with a diagnosis of bladder cancer.

There was a small but significant difference in TGF-β growth inhibition when TβR-I(6A) clones were compared with TβR-I(10A) or TβR-I clones. TGF-β growth-inhibition assays using four different sets of TβR-I, TβR-I(6A), and TβR-I(10A) clones showed similar results (Fig. 2). Each set was chosen based on comparable levels of receptor expression as assessed by Western immunoblotting (Fig. 3). Both high and low expresser clones were chosen for each TβR-I, Tβ R-I(6A), and TβR-I(10A), and the percentages of growth inhibition for a given TGF-β concentration were found to be similar among clones of a same allele, i.e., independent of the amounts of receptors expressed (data not shown). ANOVA showed that TGF-β inhibited TβR-I(6A) clones less effectively than TβR-I clones (P < 0.01) and TβR-I(10A) clones (P < 0.01).

TGF-β-receptor interactions were analyzed using increasing concentrations of ¹²⁵I-TGF-β1 and a cross-linking reagent to visualize ¹²⁵I-TGF-β1 interactions with the transfected TβR-I and the endogenous type II receptor (TβR-II) in these cell lines. The binding affinity was assessed by adding unlabeled TGF-β1 (Fig. 3). No differences in TGF-β affinity were observed between TβR-I, TβR-I(6A), and TβR-I(10A) (Fig. 3). The half-life of TβR-I and TβR-I(6A) proteins in the stably transfected cell line, which was assessed by performing pulse-chase metabolic labeling experiments using ³⁵S-methionine/cysteine, showed no difference in the metabolic stability of TβR-I and TβR-I(6A). Both had a half-life of ∼1.5 h (data not shown), as compared with a reported half-life of 2.5–3 h for endogenous TβR-I from the parental Mv1Lu cell line (27).

This study confirms the hypothesis that TβR-I(6A) homozygosity is associated with cancer. A predisposition to cancer, predominantly of the large intestine, was observed both in TβR-I(6A) homozygotes and in TβR-I(6A) heterozygotes. One possible interpretation of these findings is that these genotypes result in a decreased function of the receptor. The decrease in function might be small in TβR-I(6A) heterozygotes but may become more pronounced in TβR-I(6A) homozygotes. It is consistent with the findings that defects in the TGF-β receptor system causing a limited loss in the growth inhibitory effect of TGF-β lead to tumorigenesis (1). Such a functional alteration is compatible with normal development into adult age but is likely to contribute to tumorigenesis in colon epithelial cells and other tissues. The fact that this alteration was over-represented among patients with a diagnosis of colon cancer suggests that TβR-I(6A) homozygous colon epithelial cells may derive an advantage from decreased TGF-β signaling during tumor development. TβR-I(6A) functional alteration with respect to growth inhibition corroborates the recent finding that TβR-I(6A) mink lung epithelial cell lines are less effectively growth-inhibited by TGF-β when using the firefly luciferase reporter gene, pSBE4²⁸.

Although comparable with respect to gender, geographical location, and ethnic origin, cases were older than controls (mean age for cases, 56 years; mean age for controls, 35 years). It is possible that age differences in cases and controls affected the allele frequencies observed. If the TβR-I(6A) allele predisposes to a lethal malignancy such as colorectal cancer, however, its frequency could be higher, not lower, in a younger cohort. Thus, the younger mean age of controls could result in a bias toward the null hypothesis, resulting in an even stronger association than that observed.

Our findings are in agreement with a recent report investigating TβR-I(6A) frequency among 66 patients with a diagnosis of cancer of the cervix and 68 matched controls (28). Indeed, the authors found 10.3% TβR-I(6A) heterozygotes among controls and 15.2% among patients with a diagnosis of cancer or the cervix, which is almost identical to the 10.6% and 14.6% TβR-I(6A) heterozygotes we observed. Similarly, the only TβR-I(6A) homozygote was found in a patient with cancer of the cervix.

The presumed site of signal peptide sequence cleavage is within the TβR-I polyalanine stretch (29). Therefore, it is not known whether deletions and insertion within the polyalanine stretch are part of the signal peptide or part of the mature receptor. The observation that TβR-I(6A) clones were less effectively inhibited than the wild-type clones raises the possibility that TβR-I(6A) encodes a receptor that, in its mature form, lacks three alanine residues. This hypothesis is reinforced by the report that the TβR-I polyalanine tract is not necessary for membrane localization (28). This might affect ligand binding, and TβR-I(6A) might bind TGF-β less effectively than TβR-I. Although no differences in TGF-β affinity were observed between TβR-I, TβR-I(6A), and TβR-I(10A), and no difference in the metabolic stability of TβR-I and TβR-I(6A) was demonstrated, these results do not exclude potential small differences between TβR-I and TβR-I(6A) in their ligand binding, intracellular processing, and compartmentalization that escape detection in our assays. Further studies are mandated to determine the molecular mechanism of TβR-I(6A) in cancer development.

The finding of one TβR-I/TβR-I(10A) heterozygote but no TβR-I(6A) homozygotes among 234 Italian bladder cancer cases is consistent with the absence of TβR-I(6A) homozygotes among 77 New York patients with a diagnosis of bladder cancer. These observations are in apparent contrast with our previous report that 6 of 66 bladder cancer samples were TβR-I(6A) homozygotes (20). Because those data were from tumor DNA and because loss of heterozygosity of chromosome 9 occurs in >50% of cases of bladder cancer (30), it is probable that some of those samples were in fact not true homozygotes but pseudohomozygotes in which the wild-type allele was deleted. Additional studies of tumor and nontumor DNA from patients with bladder cancer are warranted to determine whether there is a preferential loss of the TβR-I over the TβR-I(6A) arm in TβR-I/TβR-I(6A) heterozygotes with bladder cancer. TβR-I(6A) homozygotes were not observed in a small number of Northern Italian colon cancer cases. If this is replicated in a larger study, this suggests the possibility that environmental or other genetic factors may contribute to the development of colon cancer in TβR-I(6A) homozygotes.

This study reveals that a common polymorphism in the TGF-β signaling pathway results in decreased TGF-β antiproliferative effects, thus resulting in increased cancer susceptibility. Assuming a population frequency of TβR-I(6A) homozygotes as predicted by the Hardy-Weinberg equilibrium, colon cancer predisposition due to homozygosity at this locus may account for as much as 3% of incident cases of the disease (4500 cases/year in the United States), with an additional 6% (9000 cases/year) attributable to the increase in risk observed in individuals heterozygous for TβR-I(6A). In contrast, the highly penetrant genetic mutations associated with hereditary nonpolyposis colon cancer and adenomatous polyposis account for 3–4% and <1% of cases of colorectal neoplasia, respectively (31). These observations, as well as prior associations of germ-line mutations of TβR-II and SMAD4 with inherited colorectal neoplasia (15, 16), suggest that alterations in the TGF-β signaling pathway, such as with TβR-I(6A), constitute important predisposing factors in colon cancer tumorigenesis and may account for a large proportion of inheritable cases of the disease. Although the data presented here provide strong evidence that TβR-I(6A) acts as a tumor susceptibility allele, additional studies are needed to determine the phenotype of homozygotes and the range and magnitude of cancer risks in mutation carriers.

We thank Dr. Raymond L. White (Huntsman Cancer Institute, Salt Lake City) for critical comments.