Frequent Frameshift Mutations of the TCF-4 Gene in Colorectal Cancers with Microsatellite Instability¹

Microsatellite repeats are widely distributed throughout the genome. By failing to repair spontaneous errors that occur during replication, a subset of tumors accumulates frequent deletion and insertion mutations in these repetitive DNA sequences (1, 2, 3). Approximately 15% of sporadic colorectal carcinomas referred to as MSI-H³ exhibit this phenotype (4). Most MSI has thus far been described in noncoding DNA within introns or intergenic regions of the genome. Only a relatively small number of genes have been shown to contain alterations in mononucleotide repeat sequences within coding regions, and these mutations are probably involved in the progression of MSI-H tumors. The first target sequence identified within a coding region was a poly(A) nucleotide tract of the TGFβ-RII gene (5). Other mutational targets of MSI-H tumors include repetitive sequences within genes involved in the regulation of cell growth, such as IGFIIR (6), and BAX, which promotes apoptosis (7). In addition, MSI-H tumors contain frameshift mutations in repeat sequences located within the DNA mismatch repair genes hMSH3 and hMSH6 (8).

A systematic search of human sequence databases for mononucleotide repeats contained within coding regions identified an (A)9 nucleotide tract in the TCF-4 transcription factor. TCF-4 is the most intensively expressed member of the TCF/lymphoid enhancer factor (LEF) family in colon tumor cell lines (9) and up-regulates the expression of c-MYC (10) as well as several other genes. Transcriptional activity of TCF-4 is enhanced by binding to β-catenin. This can occur either by inactivating mutations of APC that lead to the release of β-catenin or by activating mutations of β-catenin that make it insensitive to APC-mediated degradation (11). Because the APC/β-catenin/TCF pathway is likely to play an important role in colorectal tumor progression, a search for mutations in the TCF-4 repeat was initiated in a large series of cell lines and primary colorectal tumors with and without MSI-H.

Twenty-four cell lines were obtained from Dr. B. Sordat (Institut Suisse de Recherches Expérimentales sur le Cancer, Epalinges, Switzerland) and Dr. A. Zweibaum (Villejuif, France) or purchased from the American Type Culture Collection. A total of 105 primary colorectal tumors from patients undergoing surgery for colorectal cancer were analyzed. These were selected to increase the number of MSI-H tumors and comprised 81 right-sided and 24 left-sided tumors. None of the patients included in this series had a known family history suggestive of hereditary nonpolyposis colorectal cancer. All cases were histopathologically confirmed as adenocarcinomas.

The MSI status of the 24 colorectal cancer cell lines has been described previously (12). Microsatellite alterations of the 105 primary colorectal tumors were analyzed by PCR. MSI-H tumors were defined as those with somatic deletion mutations in the mononucleotide repeats BAT-26 in intron 5 of hMSH2 and BAT-25 in intron 16 of c-kit that have been shown to establish MSI status with an accuracy of greater than 99.5% (13, 14).

Total RNA from colorectal tumor cell lines was reverse-transcribed using murine leukemia virus reverse transcriptase. PCR was performed with primers P13S (5′-AACTGGTGCGGCCCTTGCAG-3′) and P13AS (5′-TGGGCTGAGGCAGCTGCCTT-3′) at an annealing temperature of 66°C. The RT-PCR products were separated on a 7 m urea/32% formamide/7% polyacrylamide gel, transferred overnight on a Hybond N+ nylon membrane, and hybridized with an internal ³²P-labeled probe, P22AS (5′-GTGCGTTCGCTACATACAAGGT-3′).

The CEPH human BAC library was screened for TCF-4 sequences with primer sets specific for exons 5 and 10 (sequences available on request). Subclones in pUC9 of a positive BAC were screened for exon 10 sequences containing the (A)9 repeat by hybridization with ³²P-labeled P22AS primer and sequenced with primer 13AS, allowing a partial intron 9 determination. Genomic DNA was amplified with primers P40S (5′-GCCTCTATTCACAGATAACTC-3′) in intron 9 and P40AS (5′-GTTCACCTTGTATGTAGCGAA-3′) in exon 10 with an annealing temperature of 56°C. PCR products were then analyzed for mutations as described for RT-PCR products.

The RT-PCR products were purified and sequenced with primers P13S and P13AS using a terminator cycle sequencing kit (Perkin Elmer).

Twenty-four colorectal cancer cell lines (8 of which were MSI-H) showed two DNA fragments upon RT-PCR of TCF-4. These fragments were excised from agarose gel, reamplified with the same primer set, and sequenced. The smaller product was 75-bp long and contained the 5′ primer in a putative exon 9, as deduced from comparison with the closely related TCF-1 gene (15), followed by a putative exon 10 containing the (A)9 repeat. The larger, 148-bp product contained an alternative, putative exon 9′ that will not be discussed in further detail in this report. Four of the eight MSI-H cell lines showed size alterations in both the 75- and 148-bp cDNA products (Fig. 1, A and B). Sequencing of these altered PCR products showed a deletion of one A of the (A)9 repeat. A normal TCF-4 allele was absent in the Lovo and TC71 cell lines but present in the TC7 and LS174T cell lines (Fig. 2, A–C). With the exception of the Colo320 cell line, all MSI-negative cell lines had only normal 75- and 148-bp DNA fragments without any size variation (Fig. 1 A). The Colo320 cell line showed two additional RT-PCR products, which, when sequenced, revealed a new 25-bp sequence (5′-TCTTTGAATTTGGAATATTACAATG-3′) adjacent to the 3′ ends of exon 9 and exon 9′, respectively.

Using the above-mentioned primer set flanking the (A)9 repeat, we were unable to amplify genomic DNA, suggestive of the presence of an intronic sequence close to the repeat. To analyze genomic DNA, the CEPH human BAC library was screened for TCF-4 sequences. One positive BAC was subcloned and partially sequenced to obtain information on intron 9 of the TCF-4 gene. This allowed us to choose a sense primer which, together with an antisense primer in exon 10, was able to amplify the (A)9 repeat from genomic DNA. To avoid unsystematic, nontemplated nucleotide addition leading to problems in allele calling, we added a G to the 5′ ends of each primer (16). Identical (A)9 size variation observed in the RT-PCR experiments was found in the cell lines with improved accuracy, confirming the specificity of the primers (Fig. 3,A). DNA from 105 primary colorectal tumors (49 of which were MSI-H) was then analyzed. One of the 56 non-MSI tumors showed a 1-bp insertion, whereas a 1-bp deletion was observed in 19 of the 49 (39%) MSI-H tumors (Fig. 3 B). Due to the presence of normal cells, it was not possible to determine for the majority of tumors whether these mutations were homozygous or heterozygous, although in two cases, normal (A)9 allele appeared to be totally absent.

Taken together, alterations of the (A)9 repeat within the TCF-4 coding region were observed in 4 of 8 (50%) and 19 of 49 (39%) MSI-H cell lines and primary tumors, respectively, compared to 0 of 16 and 1 of 56 non-MSI cell lines and tumors, respectively (P = 0.0001, Fisher’s exact test). In contrast, an (A)8 repeat and an (A)9 repeat in the coding region of the BRCA2 gene and epidermal growth factor-stimulated protein kinase p56 gene, respectively, remained intact in all eight MSI-H colorectal cell lines and in nine of the MSI-H tumors analyzed (data not shown). MSI-H cell lines and tumors were also analyzed for size alterations in coding mononucleotide repeats of the TGF-βRII (46 of 49 mutations, 94%), IGFIIR (8 of 30 mutations, 27%), and BAX (20 of 55 mutations, 36%) genes (Refs. 17 and 18).⁴ Frameshift mutation of TCF-4 and mutations in these latter genes may be of comparable functional significance in that frameshift mutation is selected for during tumorigenesis of MSI-H colorectal tumors.

In the APC/β-catenin/TCF pathway, TCF-4 transcriptional activity is increased after alterations to the APC or β-catenin proteins (9, 11) and results in enhanced expression of c-MYC (10) and other genes. In a previous study, Sparks et al. (19) did not find TCF-4 mutations in three MSI-H cell lines known to have intact APC and β-catenin. Of the five cell lines found to be mutated for TCF-4 in the present study, four (Lovo, TC71, TC7, and Colo320) are mutant for APC, and one (LS174T) is mutant for β-catenin.⁵ Our results suggest that mutations of TCF-4 stimulate its transcriptional activity synergistically with APC or β-catenin gene alterations. This hypothesis awaits confirmation by an exhaustive screening for TCF-4, APC, and β-catenin alterations in primary colorectal MSI-H tumors.

The (A)9 repeat is localized downstream of both the β-catenin binding domain and the HMG box DNA-binding region of the TCF-4 protein (9); therefore, mutations are unlikely to interfere with these activities. A number of alternatively spliced mRNA coexist for the TCF-4 gene (9, 19);⁶ consequently, the (A)9 repeat can normally code for two different COOH-terminal ends (the L and M isoforms shown in Fig. 4). These could serve as modulators of transcription and have either agonist or antagonist activities, as proposed for the closely related TCF-1 gene (20).

In MSI-H tumors, the frequently observed deletion of one nucleotide in the (A)9 repeat alters the reading frames such that the short and medium (S and M) isoforms are encoded (Fig. 4). Whether this mutational event exacerbates TCF-4 transactivating properties in colon tumors remains to be determined.

We thank A. Billault, H. Bui, F. Chareyre, L. Gressin, and C. Valette for help in screening the BAC library and the sequencing of subclones.