Mutational Inactivation of Transforming Growth Factor β Receptor Type II in Microsatellite Stable Colon Cancers¹

TGF-β⁴ inhibits the growth of epithelial cells in general (1, 2) and can inhibit growth and/or induce apoptosis in nontransformed colon epithelial cells (3, 4). TGF-β signaling is transduced by a heteromeric receptor complex composed of type I and type II components, both of which are serine/threonine-directed kinases (5). A role for RII as a human colon cancer tumor suppressor gene was demonstrated by the discovery of inactivating RII mutations in colorectal cancers that show MSI due to defects in DNA mismatch repair (6, 7, 8). These cancers with MSI account for 13% of all colon cancers (6, 7, 9). Furthermore, the restoration of wild-type RII in cell lines from colon cancers with MSI abolishes their tumorigenicity in athymic mice (10). RII mutations in colon cancers with MSI usually result in frameshifts clustered in a naturally occurring 10-bp microsatellite-like polyadenine tract in the 5′ coding half of the gene (BAT-RII; Refs. 7 and 8). In a few colon cancers with MSI, inactivation of one of the RII alleles occurs via non-BAT-RII mutations that alter the RII kinase domain (7, 9, 11), demonstrating an underlying selective advantage for RII inactivation, irrespective of whether this occurs via BAT-RII or non-BAT-RII mutational events. We hypothesized that the RII mutation should function similarly as a tumor suppressor gene in MSS colon cancers. Accordingly, we examined the RII sequence in 19 MSS colorectal cancer cell lines. These Vaco cell lines have been matched to antecedent tumor and normal tissues and have been extensively characterized as a model for human colon cancer (7, 12, 13, 14). Three of the 19 cell lines were shown to express mutant-only RII transcripts. Functional analysis showed that the RII mutations inactivated RII signaling in each case. Moreover, an additional 11 of these cell lines demonstrated a loss of TGF-β responsiveness, apparently through post-RII signaling defects. Therefore, TGF-β signaling in general and RII in particular are targets for inactivation in MSS colon cancers as well as in colon cancers with MSI.

The establishment of the panel of the Vaco colon cancer cell lines has been described previously (7, 12). The Vaco8-2 cell line was established from a stage IV cecal colon adenocarcinoma from a 56-year-old male. Vaco400 was established from a liver metastasis from a 54-year-old man. Vaco410 was derived from a stage IV colon adenocarcinoma metastasis in the right lobe of the liver of a 35-year-old woman. These cell lines were maintained as described previously (12). Genomic DNA from the original paraffin-embedded formalin-fixed tumor blocks was extracted as described previously (15). DNA from matched normal tissue for each colon cancer was obtained when available.

The colon cancer cell lines were plated at clonogenic density (100–500 cells/well in 24-well plates for adherent cell lines and 1000–5000 cells/well in 6-well plates for collagen-dependent adherent cell lines) and treated with TGF-β1 (10 ng/ml) 3 h after plating. The cell lines were grown until discrete microscopic colonies appeared (usually for 7–10 days), and then the number of colonies in each well was manually counted. Mean and SEs of the means were calculated from experiments performed in triplicate wells and repeated in at least nine independent determinations.

RNA from the cell lines was prepared by extraction with guanidine isothiocyanate (7). RT-PCR and cloning of the reaction products and manual sequencing of the cloned DNA were performed as described previously (7), or the reaction products were sequenced by automated sequencing using a ABI 377 DNA Sequencer. For amplification of RII from genomic DNA, sense primer 1871 (5′-GGTGTGTGAGACGTTGACTGAGTG-3′) was paired with antisense primer 2001 (5′-AATCTTCTCCTCCGAGCAGCTC-3′) for 30 cycles of 95°C for 30 s, 58°C for 1 min, and 70°C for 1 min. All RII mutations were confirmed to be present in at least two independently amplified PCR reactions compared with the reference sequence (GenBank accession number M85079).

The cell lines were transiently transfected following the manufacturers’ protocols using Lipofectin (Life Technologies, Inc., Gaithersburg, MD), Superfect (Qiagen), or FuGENE (Boehringer Mannheim) with a TGF-β-responsive firefly luciferase reporter plasmid (p3TP-Lux; Ref. 16) and an internal control reporter plasmid containing cDNA encoding R. reniformis luciferase under a thymidine kinase promoter (pRL-TK) or a cytomegalovirus promoter (pRL-CMV; Promega, Madison, WI). The p3TP-Lux plasmid was kindly provided by Dr. Joan Massagué (Memorial Sloan-Kettering Cancer Center and Howard Hughes Medical Center, New York, NY). After transfection, the cell lines were exposed to TGF-β1 (10 ng/ml) for 72 h and then assessed for reporter activity. The luciferase activity was measured using the Dual-Luciferase Reporter Assay System following the protocol included with this kit (Promega). The samples were assayed on a Turner Dual Injector Luminometer (Promega) or a MLX Multiplate Luminometer (Dynex Technologies).

The cell lines were infected with the MFG-RII or MFG-CAT retroviruses (17). Cells were plated in 6-well plates at a density of 200,000 or 400,000 cells/well; on the following day, they were incubated in MFG-RII or MFG-CAT viral supernatant produced by a virus producer cell line and polybrene (4 μg/ml) for 4 h. The cells then were rinsed three times in PBS. G418 (600 μg/ml) was added 2 days after the viral infection. TGF-β1 (10 ng/ml) was added to designated subsets of the infected cell lines 3 h after infection to assess for reconstitution of TGF-β-induced growth inhibition. The growth inhibition assays were performed as described above. All of these experiments were performed in triplicate and repeated at least three times.

We initially determined whether TGF-β growth-inhibitory responses were intact in 19 colon cancer cell lines that had been previously determined to be MSS (7). TGF-β growth inhibition was assayed by determining the ability of TGF-β1 to inhibit the colony formation of cells plated at clonogenic density. Cell lines in which TGF-β1 (10 ng/ml) reduced colony formation at 7–10 days after plating by >25% were considered to be responsive to TGF-β-mediated growth inhibition. Five cell lines (26%) were determined to demonstrate sensitivity to TGF-β-mediated growth inhibition, with TGF-β-mediated suppression of colony formation ranging from 42–79%. However, 74% (14 of 19) of the MSS cell lines were found to be resistant to the growth-inhibitory effects of TGF-β.

Ninety percent of colon carcinoma cell lines with MSI demonstrate both TGF-β resistance and BAT-RII mutations that inactivate the RII receptor (7, 9). Accordingly, we initially examined the MSS colon cancers for BAT-RII mutations using our previously described assay (8, 9). Consistent with our prior results, no BAT RII mutations were detected in any of these cell lines (7). Therefore, we determined the complete RII cDNA sequences in RT-PCR cDNA pools amplified from each of the 14 MSS cell lines that demonstrated TGF-β resistance. In three cell lines (15%), only mutant RII sequences were obtained, and no wild-type RII was expressed. In the remaining cell lines, RII was expressed and was wild-type in sequence. RII mutations were previously demonstrated to be ubiquitous among the approximately 13% of colon cancers that show MSI (7, 9, 18). These new findings therefore double the number of colon cancers in which RII mutations are pathogenetic to approximately 28% of all colon adenocarcinomas and show that such RII mutant cases occur in both the MSI and MSS subtypes.

The mutations detected in MSS cell lines Vaco8-2, Vaco400, and Vaco410 are located within conserved regions of the kinase domain of the RII gene (Vaco8-2, Vaco410, and Vaco400 cells) or in the 5′ extracytoplasmic portion of the receptor (Vaco400 cells). In Vaco8-2 cells, a missense mutation (GAC→AAC) changed an aspartic acid to an asparagine at codon 522 (Fig. 1; Ref. 19), altering the charge of a residue within subdomain XI of the serine/threonine kinase domain of the receptor. In Vaco410 cells, a missense mutation (CGT→CAT) changed an arginine to a histidine at codon 528 (Fig. 1; Ref. 19). This arginine, which is also within kinase subdomain XI, is strictly conserved among all serine/threonine protein kinases (20). In Vaco400 cells, two mutations were observed in the pooled cDNA products; both were apparently heterozygous (Fig. 1). Sequencing individual RT-PCR cDNA clones revealed that this was due to separate mutations present on each of the two expressed Vaco400 alleles. One Vaco400 allele carried a missense mutation (TAT→GAT) at codon 470 that changed a tyrosine to an aspartic acid (19) and altered the charge of an amino acid in subdomain IX of the receptor kinase domain. The second Vaco400 allele carried a missense mutation (AAA→ACA) at codon 52 that changed a lysine to a threonine in the extracytoplasmic region of the receptor.

To confirm that these base changes did not merely arise during the establishment of the cell lines, we analyzed genomic DNA from the tumors from which Vaco8-2, Vaco400, and Vaco410 were established. In each case, the RII mutations were demonstrated to be present in the antecedent tumors from which the cell lines were derived. Normal tissue was also available from individuals matched to Vaco410 and Vaco400. In both instances, only wild-type RII sequences were demonstrated in the normal tissues, confirming that in these tumors, RII mutations arose somatically and were selected for during carcinogenesis.

Kinase domain mutations in RII have previously been observed in occasional colon cancers with MSI both by us and by other investigators (9, 11) and have also been noted in two squamous cell carcinomas of the oropharynx (21) and in one T-cell lymphoma (22). The selection of these Vaco colon cancer mutations during tumorigenesis, their location within RII, and the nature of the amino acid changes produced suggested that these mutations likely inactivated the TGF-β receptor and induced the TGF-β resistance observed in the Vaco cell lines. Of note, RII mutations were not observed in an investigation of a different set of colon cancers studied by single-strand conformational polymorphism; this contrasting result is most likely a consequence of the lower sensitivity of single-strand conformational polymorphism compared to sequencing (11).

To confirm that the RII mutations detected in the MSS colon cancers inactivated the receptor, we determined the ability of wild-type RII to restore TGF-β-mediated responses in these cell lines. We initially assayed the ability of wild-type RII to restore TGF-β-mediated transcriptional responses as assayed by the TGF-β-responsive firefly luciferase activity encoded by the reporter construct p3TP-Lux (16). p3TP-Lux was transiently transfected into these cell lines accompanied by either RII expression vector, pRC/CMV-TGF-βRII, or control DNA, and p3TP-Lux-mediated luminescence was determined in the presence and absence of exogenous TGF-β1. Each determination was corrected for transfection efficiency by assaying for the luminescence from a cotransfected control construct, pRL-TK or pRL-CMV (Promega), a plasmid that expresses R. reniformis luciferase activity that is easily discriminated from firefly luciferase activity.

Consistent with the findings that Vaco8-2, Vaco400, and Vaco410 were resistant to TGF-β-mediated growth inhibition, each of these cell lines demonstrated low basal activity of p3TP-Lux and no further response of the reporter to added TGF-β1 (Fig. 2). In Vaco410 and Vaco400 cells, cotransfection of wild-type RII augmented the basal p3TP-Lux activity. p3TP-Lux activity increased further after the addition of exogenous TGF-β1, such that total the 3TP-Lux output reached 5–10-fold over baseline (Fig. 2). The increase in p3TP-Lux activity due to the introduction of wild-type RII likely reflects the activation of the restored TGF-β signaling pathway by the autocrine TGF-β produced by these cell lines.⁵ The responsiveness of p3TP-Lux to exogenous TGF-β1 clearly establishes that introducing wild-type RII restores TGF-β-mediated signaling in these cell lines. Thus, the RII mutations in Vaco410 and Vaco400 directly account for the TGF-β resistance of these cell lines. As discussed below, we hypothesized that the Vaco8-2 RII mutation also inactivated receptor signaling, but that additional progression events in this cell line interfered with the ability to reconstitute TGF-β signaling by a single gene alone.

To assess the potency of the TGF-β tumor suppressor pathway inactivated in RII mutant MSS colon cancers, we determined the ability of wild-type RII to inhibit colony formation in RII mutant cancer cells. Colon cancer cells were plated at a concentration selected to achieve clonogenic density after G418 selection and transduced with MFG-RII, a replication-incompetent retrovirus encoding wild-type RII, or an equal titer of a control MFG-CAT retrovirus, and then colonies arising in the presence or absence of TGF-β1 were counted and compared. Noninfected cells were removed from the assay by selection of the cells in G418. As expected, TGF-β1 had no colony suppression activity in any of the cell lines infected with the control MFG-CAT virus (Fig. 3). In contrast, 90% of colonies were suppressed in both Vaco400 and Vaco410 cells that were transduced with wild-type RII and treated with TGF-β1 (Fig. 3). Thus, RII is revealed as a potent suppressor gene in these cell lines, and the RII mutations in these tumors are again confirmed as the cause of their resistance to TGF-β growth inhibition. Similar to the inability of wild-type RII in Vaco8-2 to restore TGF-β-mediated transcriptional responses, we observed that wild-type RII also did not restore TGF-β-mediated growth inhibition in Vaco8-2. RT-PCR and Western blot analysis confirmed that after MFG-RII infection in Vaco8-2, wild-type RII cDNA was expressed, and RII protein expression was increased to levels similar to those obtained in MFG-RII-infected Vaco400 and Vaco410. Parenthetically, we noted that because G418-resistant Vaco8-2 colony numbers were the same in MFG-CAT- and MFG-RII-transduced Vaco8-2, there was no apparent nonspecific toxicity of the MFG-RII virus relative to the MFG-CAT control. Thus, the partial reduction of Vaco400 and Vaco410 colony numbers by the MFG-RII retrovirus alone is likely due to the autocrine TGF-β produced by these cell lines.

Because wild-type RII did not restore TGF-β-mediated signaling in the Vaco8-2 cell line, the functional properties of the Vaco8-2 RII mutant were studied by cloning this mutation into a cDNA expression vector and transiently transfecting the Vaco8-2 RII into the RII-deficient cell line Hct116 (7). As shown in Fig. 4 A, transient transfection of wild-type RII restored TGF-β signaling in Hct116 as assayed by the p3TP-Lux reporter. In contrast, the Vaco8-2 RII was inactive in restoring RII function in Hct116. Hence, the Vaco8-2 RII mutation that was selected for during Vaco8-2 tumorigenesis is sufficient to inactivate the TGF-β receptor and to abolish TGF-β signaling.

We initially considered the hypothesis that the Vaco8-2 RII mutation might have dominant negative activity, rendering the cell line resistant to the addition of a wild-type allele. However, cotransfecting the Vaco8-2RII and wild-type RII expression vectors revealed, at best, a borderline inhibition of wild-type RII by the Vaco8-2 mutant. (Fig. 4 A); accordingly, we hypothesized that in Vaco8-2, a separate event had further inactivated the TGF-β signaling pathway.

The binding of TGF-β to its receptor activates a heteromeric complex of Smad2-Smad3-Smad4 transcription factors to translocate to the nucleus (23, 24). Previously, we and others have reported that mutations of Smad2 and Smad4 are present at a low frequency in some human colon cancers (13, 14, 25, 26). Accordingly, we examined the possibility of Smad inactivation as a separate event in the Vaco8-2 TGF-β signaling pathway. Whereas neither Smad4 nor RII transfected singly into Vaco8-2 triggered the 3TP-Lux reporter, cotransfection of RII and Smad4 triggered a 5–10-fold p3TP-Lux response, thus suggesting that the inactivation of RII and Smad4 accounted for two separate defects in the Vaco8-2 TGF-β signaling pathway (Fig. 4 B). Examination of the Vaco8-2 Smad4 gene revealed that neither Smad4 exon 1 nor exon 2 could be amplified from Vaco8-2 genomic DNA. Thus, biallelic deletion of Smad4 was indeed a second genetic event in the Vaco8-2 cell line TGF-β pathway. Due to the small amounts of residual normal tissue remaining in even the microdissected tumor antecedent to the Vaco8-2 cell line, we could not establish whether Smad4 loss occurred in the predecessor Vaco8-2 tumor or only after establishment of the Vaco8-2 cell line. Nonetheless, it is intriguing to speculate that the presence of mutations in both the RII and Smad4 genes in the Vaco 8-2 cell line, although rare, seems to suggest that RII and Smad4 may have some functions that are distinct from one another.

In summary, multiple lines of evidence now suggest that the TGF-β pathway is a potent tumor suppressor of human colorectal carcinogenesis. We have previously determined that among the 13% of human colon cancers of the MSI subtype, nearly all have inactivating RII frameshift mutations clustering at the BAT-RII tract (7, 9). Data from the Vaco colon cancer cell lines now suggest that among MSS human colon cancers, an additional 15% of tumors also bear inactivating RII mutations, and that restoring wild-type RII to these tumors is potently growth suppressive. Hence, RII mutations play a pathophysiological role in approximately 30% of human colon cancers, and RII gene therapy may have a future role in the treatment of these tumors. Analysis of TGF-β sensitivity in the Vaco colon cancer cell line panel further suggests that an additional 55% of human colon cancers demonstrate functional inactivation of TGF-β-mediated growth inhibition. Furthermore, these cancer lines have lost TGF-β-mediated transcriptional responses as well.⁶ Presumptively, these additional cancers bear defects in the TGF-β signaling pathway at points distal to RII. We have previously shown that Smad2 and Smad4 mutations are present in only a small percentage of human colon cancers (13, 14, 27). However, the tumor-promoting activity of such downstream inactivation of TGF-β signaling is suggested by recent studies in murine models showing that heterozygous Smad4 mutation promotes colon adenoma to carcinoma progression (28), and homozygous Smad3 knockout leads directly to colon cancer (29). Study of additional mechanisms leading to the inactivation of TGF-β signaling in the remaining set of wild-type RII MSS colon cancers is currently ongoing.

We thank Dr. Joseph Willis for assistance in microdissecting clinical material for this study and Kim Yonkof for technical assistance in sequencing the TGF-β RII gene.