Instabilotyping: Comprehensive Identification of Frameshift Mutations Caused by Coding Region Microsatellite Instability¹

MSs,³ repetitive DNA sequences consisting of oligonucleotide units, are distributed widely throughout the human genome. Tumors with defective DNA mismatch repair frequently show MS length alterations (MSI). Tumors with very frequent MSI (MSI-H tumors) are fairly common, particularly in the colon, stomach, endometrium, and in anatomic sites associated with the familial syndrome, HNPCC. MSI occurring within the coding region (coding MSI) causes frameshift mutations and loss of protein function. This coding MSI is believed to represent a key mechanism underlying tumorigenesis in MSI-H tumors (1). In fact, coding MSI has been observed in tumor suppressor or tumor-related genes including TGFBR2, IGF2R, BAX, hMSH3, hMSH6, BRCA1, and BRCA2 (2, 3, 4, 5, 6, 7, 8).

We performed a large-scale genome-wide search for coding MSI by screening genes containing MSs for mutation in MSI-H tumors. A large list of genes containing coding MSs was generated from an online DNA sequence database by using a computer script designed for this purpose. In this paper, we report the results of mutational screening of 152 coding MSs in 46 MSI-H primary colorectal tumors.

Homopolymeric tracts consisting of A(n), C(n), G(n), or T(n), where n ≥ 6, were identified from the online database Unigene⁴ using a Practical Extraction and Report Language script (9). Searches were performed on Hs.seq.uniq.Z, a Unigene file containing the clone from each Unigene cluster with the longest region of high-quality sequence data. The output from Practical Extraction and Report Language consisted of GI, GenBank accession number, locus definition, exact nucleotide contained within the repeat, length of the repeat, and position within the defined coding region. A total of 21,001 homopolymers were found, including 1115 homopolymers within the coding region. Of these 1115 tracts, 300 were homopolymers of eight or more nucleotides within known or putative protein-encoding sequences. The sequence and coding region localization of each of these 300 tracts was verified manually. Furthermore, we hypothesized that mutation rates in UTRs would be higher than mutation rates in coding region because the latter mutations might be lethal to the cell. Therefore, to test this hypothesis, 45 homopolymeric repeats consisting of eight or more nucleotides located in 3′ UTRs were analyzed as well.

DNAs from 45 MSI-high colorectal cancers (9 HNPCC-associated and 36 sporadic; 4 Dukes’ A, 29 Dukes’ B, 8 Dukes’ C, 2 Dukes’ D, and 2 unstaged) and one MSI-high colon adenoma (HNPCC) were collected from 44 patients at the Conjoint Gastroenterology Lab, Royal Brisbane Hospital Foundation Cancer Research Center, Brisbane, Australia. Two of the 44 patients each had two synchronous colon cancers. Genomic DNA was extracted from paired normal and cancerous colorectal tissues that had been frozen in liquid nitrogen after surgical resection (10). MSI status for each sample was confirmed by analyses of MSI at five consensus loci (BAT25, BAT26, D2S123, D5S346, and D17S250) according to criteria from a National Cancer Institute workshop in 1998 (11). All of the tumors used in this study met the MSI-H criterion of mutation in ≥two of the five loci.

MSI at each locus was determined by analyses of the length of each PCR-amplified MS. These primer sequences are available on request. One primer of each pair was labeled with a fluorescent dye, i.e., Hex, Fam, or Tet. PCR reactions were performed in a total volume of 10 μl containing 20 ng of genomic DNA, 0.1 μm of each primer, 1X Taq DNA polymerase buffer 20 mm Tris-HCI (pH 8.4) 50 mm KCl (Life Technologies, Inc., Gaithersburg, MD), 0.4 mm of each deoxynucleotide triphosphate, 1.5 mm of MgCl₂, and 0.5 IU of Taq DNA polymerase (Life Technologies). Conditions were as follows: an initial denaturation step at 94°C for 4 min, followed by 35 cycles of denaturation at 94°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 30 s. After these steps, a final extension was performed at 72°C for 4 min. Products were analyzed on an automated DNA sequencer (ABI 377 or 3700; PE Biosystems, Foster City, CA) using the software programs GeneScan and Genotyper (PE Biosystems). We classified a tumor-specific alteration as MSI only when it caused a change of >50% in peak height in the tumor sample compared with the corresponding normal sample.

Each of the PCR products was cloned using the TOPO TA Cloning kit (Invitrogen, Carlsbad, CA). Sequencing reactions were performed using the BigDye Terminator Cycle Sequencing Ready Reaction kit (Applied Biosystems) and analyzed on an ABI 377 automated DNA sequencer.

The database search identified 300 homopolymeric tracts of ≥eight nucleotides located within protein coding region (233 tracts of eight nucleotides, 51 of nine, 8 of 10, 5 of 11, 2 of 12, and 1 of 13). PCR primer sets for these 300 tracts were designed and tested on human genomic DNA. Primer sets (152 of 300) worked well and were used for additional analyses. Nine of these 152 loci represented genes shown previously to undergo coding MSI. This subgroup comprised TGFBR2, IGF2R, BAX, Bcl-10, hMSH3, hMSH6, BRCA1, BRCA2, and the Bloom syndrome gene (2, 3, 4, 5, 6, 7, 8, 12, 13).

In general, mutation caused by MSI was not frequent in either putative protein-encoding (coding) or 3′UTR loci. Examples of coding MSI are displayed in Fig. 1. However, there was a wide variability in mutation frequencies among loci in each category (0–79.1% in coding loci, and 0–52.4% in 3′-UTR loci). In addition, there was a discrepancy in mutation frequency depending on the sequence of the tract: A₈/T₈ homopolymers had lower mutation frequencies but a wider variation range versus C₈/G₈ tracts (0–58.1%, mean 3.0% versus 0–37.5%, mean 8.0%, respectively, in coding loci).

Table 1 summarizes numbers of loci at each mutation frequency range in MSI-H colorectal tumors. As we expected, coding loci had lower mutation frequencies than 3′UTR loci (Fig. 2). Ninety percent of coding loci were mutated in only up to 11.9% of colorectal tumor samples, whereas 90% of 3′UTR loci were mutated in as many as 20.9% of cases. The mean and median mutation rates were lower in 152 coding loci (5.4 and 2.3%, respectively) than in 45 3′UTR MSs (9.0 and 2.8%, respectively). The geometric mean mutation rate for coding MSs was 2.04% (95% confidence limits, 1.54–2.65) versus 3.74% for 3′UTR MSs (95% confidence limits, 2.24–5.92); these values differed significantly (P = 0.027, t test for log-transformed mutation rates).

Descriptions of 43 coding loci that were mutated in >5% of colorectal tumors are listed in Table 2. As shown, 9 of the 152 coding MSs were mutated in >20% of tumors. Three genes already known to mutate in this manner in colorectal cancers, namely TGFBR2, BAX, and hMSH3, were included in this group. Their mutation frequencies were 79.1, 37.5, and 26.2%, respectively. Among six genes not known previously to undergo coding MSI, the most frequently mutated was the ACTRII gene, a member of the TGF-β receptor superfamily (mutated in 25/43, or 58.1% of MSI-H tumors), followed by SEC63, a human homologue of a yeast DnaJ-like endoplasmic reticulum translocon component protein gene (21/43 or 48.8%); AIM2 (20/42 or 47.6%); an interferon-inducible protein gene (20/42 or 47.6%); an energy metabolism gene, the NADH-ubiquinone oxidoreductase B14.5B subunit (12/43 or 27.9%); KIAA0977, a probable human homologue of the mouse embryonal protein gene cordon-bleu (10/43 or 23.8%); and PA2G4/EBP1, a homologue of the mouse cell cycle protein p38–2G4 (9/43 or 20.9%). There were no significant differences in mutation frequencies of these genes relative to clinical tumor stage or type of disorder (i.e., HNPCC-related versus sporadic colorectal cancer). A subset of the mutations was confirmed by DNA sequencing (data not shown). In addition, 10 genes were mutated in >10% of tumors: the coagulation factor VIII:C gene (14.6%), the hepatocyte growth factor receptor gene homologue SEX (14.3%), the DNA mismatch repair gene hMSH6 (14.0%), the tRNA transporter protein gene exportin t (14.0%), the methyl-CpG binding protein gene MBD4 (14.0%), KIAA0905 (14.0%), IGF2R (11.9%), the protein-tyrosine phosphatase D1 gene (11.6%), DKFZp564D0782 (11.6%), and a putative monocarboxylate transporter gene (11.1%).

Through the study of 152 coding MS loci, we found generally rare mutation rates with a wide variability. Nucleotide sequence seemed to have some influence on mutation rate; these findings are in agreement with a recent study of intronic MSs (14). Although mutation rate was lower in coding loci than in 3′UTR loci as hypothesized, a significant number of coding loci including genes or molecular pathways not implicated previously in human cancer were frequently mutated. The validity of these novel mutations was attested to by our finding MSI at loci reported previously, such as TGFBR2, IGF2R, BAX, hMSH3, and hMSH6, at frequencies similar to those found by other groups (2, 3, 4, 5, 6, 7). Thus, at least some of the genes with frequent mutation identified in this study may be involved in the process of tumorigenesis or tumor progression and, therefore, may merit functional study or other additional analysis.

A review of the literature regarding the six novel coding MS loci mutated most frequently (i.e., in >20% of tumors) in our study suggests links between their protein products and cell proliferation and/or differentiation. For example, ACTRII is a member of the TGF-β receptor family associated with the Activin-SMAD signal transduction system, which is involved in the induction of differentiation, growth suppression, and apoptosis (15, 16, 17, 18). Both activin and ACTRII are expressed in normal intestine (19). Mutant ACTRII is reported to inhibit activin-mediated induction of differentiation (20). Furthermore, somatic frameshift mutation accompanied by loss of heterozygosity in the activin type I receptor, a binding partner of ACTRII to form the receptor complex, was recently demonstrated in pancreatic cancers (21). In our study, 5 of 25 tumors that showed mutation in the ACTRII had no detectable normal allele, and an additional 4 cases showed 70% loss of the normal allele, suggestive of loss of heterozygosity. Taken together, our data and published observations suggest that inactivation of activin receptors is associated with tumorigenesis in the gastrointestinal tract.

Other genes also have interesting characteristics. The SEC63 protein is involved in the process of protein folding and translocation, including nuclear translocation of nucleoproteins (22). AIM2 is an interferon-inducible protein expressed in normal intestine, the expression of which is suppressed in melanomas (23). Overexpression of AIM2 is reported to cause growth inhibition and an increase in cell death (24). NADH-ubiquinone oxidoreductase participates in energy metabolism, which is indispensable to the rapid growth of tumors (25). The precise function of the KIAA0977 protein has not yet been determined, but its amino-terminal portion bears 37% homology to the mouse embryonal cordon-bleu gene product, which participates in axis structure formation (26). PA2G4/EBP1 is a cell cycle-specific transcriptional enhancer; effects of PA2G4/EBP1 on cellular proliferation or differentiation have not yet been defined, but growth suppression by transfected PA2G4/EBP1 has been described (27).

In summary, our screen of 152 coding mononucleotide repeat sequences for mutation in 46 MSI-H colorectal tumors identified six novel loci showing mutation rates >20%. Coding region loci known to mutate in MSI-H tumors were also frequently mutated in our study, lending weight to our findings at novel loci. Moreover, published literature suggested potential roles for at least some of these genes in human carcinogenesis. This genome-wide strategy, therefore, appears to show promise as a means of identifying candidate genes or molecular pathways for additional study in cancer research.