Mutation Rates of TGFBR2 and ACVR2 Coding Microsatellites in Human Cells with Defective DNA Mismatch Repair

Microsatellite instability promotes colonic tumorigenesis through generating frameshift mutations at coding microsatellites of tumor suppressor genes, such as TGFBR2 and ACVR2. As a consequence, signaling through these TGFβ family receptors is abrogated in DNA Mismatch repair (MMR)-deficient tumors. How these mutations occur in real time and mutational rates of these human coding sequences have not previously been studied. We utilized cell lines with different MMR deficiencies (hMLH1−/−, hMSH6−/−, hMSH3−/−, and MMR-proficient) to determine mutation rates. Plasmids were constructed in which exon 3 of TGFBR2 and exon 10 of ACVR2 were cloned +1 bp out of frame, immediately after the translation initiation codon of an enhanced GFP (EGFP) gene, allowing a −1 bp frameshift mutation to drive EGFP expression. Mutation-resistant plasmids were constructed by interrupting the coding microsatellite sequences, preventing frameshift mutation. Stable cell lines were established containing portions of TGFBR2 and ACVR2, and nonfluorescent cells were sorted, cultured for 7–35 days, and harvested for flow cytometric mutation detection and DNA sequencing at specific time points. DNA sequencing revealed a −1 bp frameshift mutation (A9 in TGFBR2 and A7 in ACVR2) in the fluorescent cells. Two distinct fluorescent populations, M1 (dim, representing heteroduplexes) and M2 (bright, representing full mutants) were identified, with the M2 fraction accumulating over time. hMLH1 deficiency revealed 11 (5.91×10−4) and 15 (2.18×10−4) times higher mutation rates for the TGFBR2 and ACVR2 microsatellites compared to hMSH6 deficiency, respectively. The mutation rate of the TGFBR2 microsatellite was ∼3 times higher in both hMLH1 and hMSH6 deficiencies than the ACVR2 microsatellite. The −1 bp frameshift mutation rates of TGFBR2 and ACVR2 microsatellite sequences are dependent upon the human MMR background.


Introduction
The DNA MMR system consists of proteins that act in concert to recognize and coordinate repair of nucleotide base mismatches and slippage mistakes at microsatellite sequences on newly synthesized DNA [1]. In humans, MMR activity requires the proper functioning of hMutSa and hMutSb to recognize defects, and hMutLa to coordinate repair. hMutSa (heterodimer of hMSH2 and hMSH6) recognizes single nucleotide interstrand mispairs and insertion/deletion loops (IDLs) containing 1 or 2 looped nucleotides, whereas hMutSb (heterodimer of hMSH2 and hMSH3) recognizes IDLs containing 2 or more looped nucleotides that occur at microsatellite sequences [2]. The hMutS complexes interact with the hMutLa protein complex (heterodimer of hMLH1 and hPMS2) to coordinate excision and repair of the mispair or IDL [3][4][5]. Loss of any of the components of the MMR system inactivates or attenuates DNA repair, and is the cause of microsatellite instability (MSI) [6,7]. Patients with germline mutations of hMSH2, hMLH1, hMSH6, or hPMS2 have Lynch syndrome (formerly known as hereditary nonpolyposis colon cancer or HNPCC), the most common familial form of colorectal cancer [8][9][10][11]. Epigenetic inactivation of hMLH1 through promoter hypermethylation occurs in 15-20% of sporadic colorectal cancers [12,13]. In either instance, the resulting colorectal cancers display the phenotype of MSI observed as novel length mutations at microsatellites [7].
Microsatellites are nucleotide repeat sequences that are ubiquitous throughout the genome [14]. Rarely but significantly, microsatellites are present in the coding regions (exons) of critical growth regulatory genes and are targeted for frameshift mutation when DNA MMR is defective [15]. These frameshift mutations, which occur due to non-repair of exonic IDL, are thought to drive the pathogenesis of colorectal cancers and other MSI tumors. The type II receptor for transforming growth factor b (TGFBR2) has an A 10 microsatellite within exon 3. Frameshift mutation of this polyadenine sequence truncates TGFBR2, making it nonfunctional for TGFb signaling [16]. In 70-90% of colorectal cancers with MSI, TGFBR2 is frameshift mutated at both alleles [17]. This mutation allows the tumor to escape the growth suppressive effects of TGFb-SMAD signaling. TGFBR2 mutation appears to be a late event in MSI adenomas and tightly correlated with progression of these adenomas to malignant carcinomas [18].
The activin type II receptor, ACVR2, contains polyadenine tracts at both exons 3 and 10 but only its exon 10 A 8 tract is mutated in ,85% of colorectal cancers with MSI [19,20]. The biallelic frameshift mutation causes ACVR2 protein loss, and is associated with histologically poor grade tumors and significantly larger volume tumors [20,21]. Restoration of ACVR2 in colon cancer cells causes growth suppression [22].
Colon cancer cell models highlight the relationship between defective DNA MMR and TGFBR2 and ACVR2 frameshift mutations. Both genes commonly have a 21 bp frameshift mutation with defective MMR. Restoration of wild type (WT) TGFBR2 and ACVR2 by chromosome transfer reveals growth suppression in the cells and slower growth in xenografts in nude mice [22,23]. Interestingly, HCT116+chr3 cells, which have two mutant hMLH1 and two mutant TGFBR2 alleles plus one WT hMLH1 and one WT TGFBR2 allele, express ,33% WT TGFBR2 mRNA and ,67% mutant TGFBR2 mRNA (unpublished data). On the other hand, HCT116+chr2 cells, which have two mutant hMLH1 and two mutant ACVR2 alleles plus one WT ACVR2 allele, express ,20% WT ACVR2 mRNA [22], suggesting a slow but steady mutation of the transferred ACVR2 allele in hMLH1 deficiency.
Determining mutation rates of actual human coding genes in human MMR deficiency has not been previously performed, although model systems using noncoding sequences with human cell and yeast MMR systems have been utilized [24][25][26][27]. To test the hypothesis that TGFBR2 and ACVR2 frameshift mutations are dependent on the human MMR background, we constructed EGFP plasmids in which a 21 bp frameshift mutation at coding microsatellites of TGFBR2 exon 3 and ACVR2 exon 10 was detected by EGFP expression in human colon cancer cells with MMR deficiency. With this new cell model, we were able to calculate a human gene mutation rate in each human MMRdeficient background, and could directly compare the mutation rate between TGFBR2 and ACVR2.

Results
Successful cloning and stable transfection of pIREShyg2-TGFBR2-EGFP and pIREShyg2 ACVR2-EGFP plasmids The plasmid pIREShyg2-EGFP allows the expression of EGFP under the control of a constitutive cytomegalovirus promoter, which is active throughout the cell cycle [25]. We inserted portions of exon 3 of TGFBR2 or exon 10 of ACVR2 as outlined in Fig. 1 after the translation initiation codon of the EGFP gene, either inframe of an EGFP (IF) or +1 bp out of frame of the EGFP (OF) in pIREShyg2-EGFP. For experimental plasmids, TGFBR2 or ACVR2 sequences were cloned +1 bp OF in pIREShyg2-EGFP and thus a 21 bp frameshift mutation at the coding microsatellite would shift the EGFP gene into the proper reading frame to allow Figure 1. pIREShyg2-TGFBR2-EGFP and pIREShyg2-ACVR2-EGFP plasmids. Portions of exon 3 of TGFBR2 and exon 10 of ACVR2 sequences containing coding microsatellites were inserted immediately after the start codon of the EGFP gene, being in frame with EGFP (IF) or +1 bp out of reading frame with the EGFP (OF) in pIREShyg2-EGFP. Mutation resistant (MR) plasmids were constructed by interrupting the coding microsatellite sequences (A 10 to A 2 CA 2 GA 2 CA in TGFBR2 and A 8 to A 3 GA 4 in ACVR2), preventing frameshift mutation. Deletion of 1 bp in OF plasmids, experimental plasmids, restores the proper reading frame and allows EGFP expression. MR IF and MR OF plasmids were used for positive and negative control for EGFP expression, respectively. doi:10.1371/journal.pone.0003463.g001 EGFP expression. Mutation resistant (MR) counterpart plasmids were constructed by interrupting the coding microsatellites (A 10 to A 2 CA 2 GA 2 CA in TGFBR2 and A 8 to A 3 GA 4 in ACVR2) and would be resistant to frameshift mutation. The MR TGFBR2 and MR ACVR2 plasmids were placed OF (+1 bp) and IF to be used as negative and positive controls for EGFP expression, respectively. Transient transfections of these plasmids into the hMLH1 2/2 background were performed to verify their EGFP expression in cells and detected by a fluorescence microscope before stable cell lines with different MMR deficiency were established. Positive controls expressed EGFP whereas negative controls did not express EGFP in the cells.
Subsequently, twenty-four stable cell lines with differing MMR genetic backgrounds (Table S1) were established with hygromycin B selection after transfection, as outlined in Table S2. After selection, colonies from each cell line were initially pooled and cultured for mutation analysis. After 5 weeks, the proportion of fluorescent cells within each cell line was measured by flow cytometry. All eight cell lines containing MR TGFBR2 IF or MR ACVR2 IF sequence showed fluorescence between 88% and 100% (median 99.1%), indicating robust selection efficiency of the MR IF stable cell lines (Fig. 2). In hMLH1 2/2 and hMSH6 2/2 cells containing TGFBR2 OF or ACVR2 OF sequences, newly fluorescent cells were observed ranging between 0.14% and 1.22% (median 0.32%) net fluorescence over counterpart cell lines containing MR TGFBR2 OF or MR ACVR2 OF sequences. The hMLH1 2/2 cells with TGFBR2 OF sequence showed the highest net fluorescence (1.22%) among cell lines with TGFBR2 OF or ACVR2 OF sequences (Fig. 2). On the other hand, hMSH3 2/2 and MMR proficient stable cell lines did not show any net fluorescence.
TGFBR2 exon 3 and ACVR2 exon 10 coding microsatellite mutations are dependent on the MMR background To determine mutation frequencies of the TGFBR2 and ACVR2 coding microsatellites in cells with different MMR backgrounds (Table S1), nonfluorescent cells containing either MR TGFBR2 OF, TGFBR2 OF, MR ACVR2 OF or ACVR2 OF were sorted and exponentially grown for 7 to 35 days. At specific time points (day 7, 14, 21, 28, and/or 35) three cultures of each cell line were analyzed in parallel for EGFP expression by using flow cytometry to detect 21 bp frameshift mutations. Three different populations were identified according to their EGFP fluorescence intensity (Fig. 3). The population with no fluorescence was named M0, the population with low fluorescence M1, and the population with high fluorescence M2. EGFP histograms of MR TGFBR2 OF and TGFBR2 OF cells in different MMR backgrounds at day 21 are shown in Fig. 3E, in which hMLH1 2/2 TGFBR2 OF cells showed 2 distinct EGFP cell populations, M1 and M2. M2 cells from hMLH1 2/2 TGFBR2 OF showed brighter EGFP expression compared to M1 cells (Fig. S1).
The 21 bp mutation frequency at each time point was expressed as a fold change using the following formula: (EGFP positive cells/total live cells from TGFBR2 OF or ACVR2 OF cells)/(EGFP positive cells/total live cells from MR TGFBR2 OF or MR ACVR2 OF cells) (Fig. 4). The M2 population accumulated over time (most dramatically with TGFBR2 and ACVR2 sequences in hMLH1 2/2 background) whereas the M1 population showed little change (Fig. 4), indicating that M1 and M2 are distinct populations. The M1 and M2 populations were plotted separately for analysis of mutation frequency.
In the M1 population, mutation frequency of the TGFBR2 sequence in hMLH1 2/2 cells was higher than other cell lines and increased over time (highest at day 35, 21-fold change), although the increase over time was small when compared to the increase in the M2 population (Fig. 4A). There was no consistent increase in the M1 population in other cell lines over time except hMSH3 2/2 TGFBR2 cells that showed a slow increase in mutation frequency up to day 35 (5-fold change). In the M2 population, the TGFBR2 sequence in the hMLH1 2/2 background demonstrated the highest mutation frequency (highest at day 21, 240-fold change) over time compared to other different MMR deficient backgrounds as predicted in EGFP histograms (Fig. 3E). Mutation frequency of the TGFBR2 sequence in the hMSH6 2/2 background also increased over time and showed the highest mutation frequency on day 35 (14-fold change), although this frequency is much lower than TGFBR2 sequence in hMLH1 2/2 cells (Fig. 4B). At day 35, frameshift mutation of ACVR2 sequence in the hMLH1 2/2 background was 79-fold higher than the negative control whereas ACVR2 sequence in the hMSH6 2/2 background showed a 4-fold change in mutation frequency. In all hMSH3 2/2 stable cell lines, M2 cells were rare events (average 0.009%) at all time points and there was no significant difference in numbers of the M2 population between MR TGFBR2 or MR ACVR2 OF and TGFBR2 or ACVR2 OF cells, respectively, as shown in Fig. 3E. Thus, fold changes in the M2 population were 1 at all time points for both TGFBR2 and ACVR2 sequences in hMSH3 2/2 cell lines (Fig. 4B).
The M1 population represents pass-through heteroduplexes, while the M2 population represents fully mutant sequences To confirm that fluorescence from the M1 and M2 populations was driven by 21 bp frameshift mutation at the coding microsatellites of TGFBR2 OF and ACVR2 OF, at day 21 after being plated as nonfluorescent cells, cells from the M1 and/or M2 populations of hMLH1 2/2 TGFBR2 OF or ACVR2 OF cells, hMSH6 2/2 TGFBR2 OF or ACVR2 OF cells, and hMSH3 2/2 TGFBR2 OF or ACVR2 OF cells were sorted and expanded for sequencing analysis. DNA from each cell line was amplified by PCR, sub-cloned into a TA cloning vector and single cell clones were individually sequenced to assess for frameshift mutation at the coding microsatellites of TGFBR2 exon 3 and ACVR2 exon 10. As expected, nearly all DNA clones (86-100%) from the M2 population of all cell lines with hMLH1 2/2 and hMSH6 2/2 revealed 21 bp frameshift mutation (A 9 at TGFBR2 and A 7 at ACVR2), indicating fully mutant sequences inducing EGFP expression (Fig. 5A). Rare A 8 sequences were observed in M2 clones from hMSH6 2/2 TGFBR2 OF cells. In particular, all clones from the M2 population of hMLH1 2/2 TGFBR2 OF cells revealed frame shift mutations (A 9 ) with no wild type A 10 sequence (Fig. 5A). This observation correlates with the highest mutation frequency of hMLH1 2/2 TGFBR2 cells over time (Fig. 4B). In comparison, clones from the M1 population of hMLH1 2/2 TGFBR2 OF and hMLH1 2/2 ACVR2 OF revealed 84 and 69% of mutant (A 9 and A 7 ) microsatellites, respectively, and clones from the M1 population of hMSH6 2/2 TGFBR2 OF and hMSH6 2/2 ACVR2 OF cells expressed 50 and 53% of mutant (A 9 and A 7 ) microsatellites, respectively (Fig. 5A). A rare A 11 sequence was also observed in M1 clones from hMSH6 2/2 TGFBR2 OF cells. In the M1 population, clones of hMLH1 2/2 TGFBR2 OF cells showed a 21 bp frameshift mutation in 84%, corresponding to the highest increase in mutation frequency in the M1 population of all cell lines over time (Fig. 4A). Only five percent (1/20) of M1 clones from hMSH3 2/2 TGFBR2 OF cells revealed a mutated microsatellite sequence (A 9 ) (Fig. 5A). The M1 population from hMSH3 2/2 ACVR2 OF cells did not show any frameshift mutation (data not shown) and thus sub-cloning was not done for sequencing analysis. As expected, all MR stable cell lines did not show frameshift mutations at microsatellites with MMR-deficiency. In addition, all MMR proficient HT29 stable cell lines did not show any frameshift mutations at microsatellites of TGFBR2 and ACVR2.
To determine the nature of the mutations observed in the M1 population, we analyzed pooled cells as well as single cell clones. In pooled samples, unlike fully mutant sequences observed in the M2 populations, M1 population sequences often revealed two overlapping sequences, suggestive of heteroduplexes (A 9 /T 10 in TGFBR2 and A 7 /T 8 in ACVR2) (Fig. 5B). Single cell clones revealed the presence of both WT and 21 bp frameshift mutants (Fig. 5B), consistent with heteroduplexes that weakly drive EGFP expression.
Overall, our data indicate that hMLH1 deficiency has the highest susceptibility for frameshift mutation at the coding microsatellites of TGFBR2 exon 3 and ACVR2 exon 10 of the three different MMR deficiencies (hMLH1 2/2 , hMSH6 2/2 and hMSH3 2/2 ). In addition, the coding microsatellite of TGFBR2 exon 3 has a higher susceptibility to a 1 bp frameshift mutation than that of ACVR2 exon 10 in hMLH1, hMSH6, (and hMSH3) deficiencies.
The frameshift mutation rates at the coding microsatellites of TGFBR2 exon 3 and ACVR2 exon 10 are dependent on the MMR background The M2 population (full mutants) was used to calculate the mutation rates at microsatellites of TGFBR2 exon 3 and ACVR2 exon 10 by the ''method of the mean'' [28] (Table 1). A single mutation rate was calculated by taking a weighted average of the mutation rates at the different time points, the weights of which were chosen to minimize the variance of the estimate as previously described [25]. As predicted, the mutation rate at the microsatellite of TGFBR2 in the hMLH1 2/2 background was highest: 5.91610 24 61.26610 24 . Mutation at the A 10 microsatellite of TGFBR2 is ,3 times more frequent than mutation at the A 8 microsatellite of ACVR2 in hMLH1 deficiency (P,0.01). In addition, mutation at the microsatellite of TGFBR2 is ,4 times higher than mutation at the microsatellite of ACVR2 in hMSH6 deficiency. Furthermore, mutations at the microsatellites of . Mutation frequencies of TGFBR2 exon 3 and ACVR2 exon 10 are dependent on the MMR background. Nonfluorescent cells were analyzed for EGFP expression by flow cytometry at 7, 14, 21, 28, and/or 35 days after being sorted and cultured, and EGFP analysis was performed as described in Fig. 3. Mutation frequency at each time point was expressed as a fold change using the following formula: (EGFP positive cells/total live cells from TGFBR2 OF or ACVR2 OF cells)/(EGFP positive cells/total live cells from MR TGFBR2 OF or MR ACVR2 OF cells). Overall, the M1 population showed little change in mutation frequency whereas mutation frequency of the M2 population accumulated over time, indicating that M1 and M2 populations are distinct populations. In the M2 population, note that the hMLH1 deficiency conferred higher mutation frequencies in both ACVR2 and TGFBR2 sequences compared to hMSH6 and hMSH3 deficiencies and that TGFBR2 sequence showed a higher mutation frequency than ACVR2 sequence in identical MMR deficiencies. Cell lines showing lower mutation frequencies (less than 25-fold change) were separately plotted in the right panel using a smaller y-axis scale. Data are means from three independent experiments at each time point. doi:10.1371/journal.pone.0003463.g004 TGFBR2 and ACVR2 are ,11-15 times higher in hMLH1 deficiency than in hMSH6 deficiency (P,0.01). Mutation rates from hMSH3 2/2 and MMR-proficient cell lines were not calculated due to a lack of net fluorescent M2 populations. These data confirm that hMLH1 deficiency allows a higher susceptibility for mutation at the coding microsatellites of TGFBR2 exon 3 and ACVR2 exon 10 than hMSH6 and hMSH3 deficiencies, and that TGFBR2 exon 3 has a higher susceptibility to mutation at its coding microsatellite over ACVR2 exon 10 in both hMLH1 and hMSH6 deficiencies.

Discussion
In this study, we developed an experimental model in which the actual human coding sequences of TGFBR2 exon 3 and ACVR2 exon 10 were evaluated in real time for 21 bp frameshift mutations in human cells with differing MMR genetic back-grounds (hMLH1 2/2 , hMSH6 2/2 , hMSH3 2/2 , and MMRproficient). Our aim was to test the hypothesis that the frequency and rate of targeted genes for frameshift mutation in human MSI tumors are dependent on the MMR genetic background. 21 bp frameshift mutations in exon 3 of TGFBR2 and exon 10 of ACVR2 are common in MSI tumors, and are thought to help drive the pathogenesis of colorectal cancers manifesting MSI. Although the general frequencies of TGFBR2 and ACVR2 mutations are culled from general colorectal cancer cohorts, there is no experimental data on mutation rates of these targeted genes and how MMR deficiency can influence those rates.
In this study, we made several unique observations: (1) the 21 bp frameshift mutations at coding microsatellites within human TGFBR2 exon 3 and ACVR2 exon 10 sequences were observed in real time in different human MMR deficient backgrounds; (2) both coding microsatellites of TGFBR2 and ACVR2 mutate through heteroduplex formation (M1) before full mutation (M2); (3) MMR-deficient backgrounds determine the mutation frequency and rate of the coding microsatellites of TGFBR2 and ACVR2, for which hMLH1.hMSH6.hMSH3 deficiency; (4) hMSH3 2/2 background does not generate any significant frameshift mutation in the tested sequences; and (5) the coding A 10 microsatellite of TGFBR2 mutates at a higher rate than the A 8 coding microsatellite of ACVR2 in hMLH1 2/2 and hMSH6 2/2 backgrounds.
Our experimental model revealed two distinct fluorescent populations of mutant cells, M1 expressing dim EGFP and M2 expressing bright EGFP (Fig. 3). The M2 population accumulated over time whereas the M1 population showed little change (Fig. 4).
These observations were similar to a study that observed frameshift mutation at a noncoding (CA) 13 microsatellite in an hMLH1 2/2 background [25].
We confirmed that EGFP expression from M1 and M2 populations was driven as a result of a 21 bp frameshift mutation of TGFBR2 OF and ACVR2 OF cells by TA subcloning and DNA sequencing analysis (Fig. 5). In particular, the M2 clones from hMLH1 2/2 TGFBR2 OF cells revealed that all clones underwent 21 bp frameshift mutation, indicating that the M2 clones are fully mutant cells containing a frameshifted A 9 /T 9 microsatellite. The M2 clones from hMSH6 2/2 TGFBR2 OF, hMLH1 2/2 ACVR2 OF, and hMSH6 2/2 ACVR2 OF cells showed ,10-15% WT microsatellite sequences that are derived from the M1 cell population. Clones from M1 populations in hMLH1 2/2 and Figure 5. Frameshift mutation at coding microsatellites of TGFBR2 exon 3 and ACVR2 exon 10 in different human MMR deficient backgrounds. Cells from the M1 and/or M2 populations of hMLH1 2/2 TGFBR2 OF or ACVR2 OF cells, hMSH6 2/2 TGFBR2 OF or ACVR2 OF cells, and hMSH3 2/2 TGFBR2 OF cells were sorted and cultured. DNA from each cell line was amplified by PCR, sub-cloned and all single cell clones were individually sequenced to assess for frameshift mutation of the coding microsatellites of TGFBR2 exon 3 and ACVR2 exon 10. Sequence analysis of DNA clones from hMLH1 and hMSH6 deficiencies revealed mostly 1 bp deletion at microsatellites (A 9 for TGFBR2 or A 7 for ACVR2), shifting the EGFP gene into the reading frame and leading to its expression (A). Note that M2 clones from hMLH1 2/2 TGFBR2 OF cells revealed 100% A 9 /T 9 microsatellite sequences, termed a ''full mutant'' whereas M1 clones revealed a mixture of A 10 /T 10 and A 9 /T 9 microsatellite sequences, which suggests the presence of an A 10 /T 9  hMSH6 2/2 backgrounds revealed the coexistence of mutated (65614%, A 9 for TGFBR2 and A 7 for ACVR2) and WT (35614%, A 10 and A 8 ) microsatellite sequences, indicating the existence of intermediate mutant cells containing A 8 /T 7 or A 10 /T 9 heteroduplexes within the M1 population transferring into full mutant cells as previously observed [25]. Relative to the M2 population, the M1 population increased initially but reached a steady state as a constant supply of actively mutating cells transitioned into the M2 population. Although ,5% of cells in hMSH3 2/2 TGFBR2 revealed mutated microsatellite sequence (A 9 ) in the M1 population, none transitioned into the M2 population. This is likely due to repair by hMutSa at the IDL, for which hMutSb is not needed for repair. As we hypothesized, the mutational frequencies and rates of TGFBR2 exon 3 and ACVR2 exon 10 microsatellites are dependent on the MMR deficient background with hMLH1 2/2 .hMSH6 2/2 .hMSH3 2/2 . As TGFBR2 and ACVR2 mutations may drive the pathogenesis of colorectal cancers, our human data is consistent with the virulence of tumor formation in Lynch syndrome. Patients with germline mutation in hMLH1 may present with cancer at younger ages compared to those with a hMSH6 germline mutation [29]. Data on mutation rates for TGFBR2 exon 3 and ACVR2 exon 10 ( Table 1) showed similar results. hMLH1 2/2 TGFBR2 showed the highest mutation rate at its coding microsatellite sequence (5.91610 24 ). This mutation rate is similar to that calculated for noncoding (CA) 13 microsatellites in hMLH1 2/2 cells [25], suggesting that this coding A 10 and the noncoding (CA) 13 microsatellites are equally vulnerable to hMLH1 deficiency. The mutation rate for TGFBR2 was 3 fold higher than that for ACVR2 in both hMLH1 (5.91610 24 ) and hMSH6 (0.54610 24 ) deficiencies. The rapid rate for TGFBR2 mutation with MMR deficiency might be due partly to TGFBR2's longer polyadenine tract compared to ACVR2, as longer microsatellite tracts mutate more frequently in MMR deficiency [7]. In the case of ACVR2 exon 10, even though the mutation rate is slower than TGFBR2 exon 3, ultimately fully mutant clones accumulate. The rapid rate for TGFBR2 exon 3 mutation is probably most reflective in the M1 population, as there is a rapid heteroduplex formation particularly in hMLH1 deficiency, followed by full mutation. In ACVR2 exon 10, heteroduplex formation is relatively slower. With both TGFBR2 and ACVR2 constructs, heteroduplex formation and subsequent full mutation are slower in the hMSH6 2/2 background compared to hMLH1 2/2 background. It has been shown that MSH6 and MSH3 are redundant in regard to frameshift mutagenesis in a yeast model [30], which supports our finding that hMSH6 and hMSH3 defects have much lower frameshift mutation rates than the hMLH1 defect that completely eliminates MMR. Lower frameshift mutation rate in hMSH6 deficiency would logically predict a lower penetrance in Lynch syndrome for which no germline hMSH3 mutation has been reported.
In summary, we established and utilized a cell model in which actual human coding microsatellite sequences of TGFBR2 exon 3 and ACVR2 exon 10 were evaluated in real time for frameshift mutation in different human MMR backgrounds. hMLH1 deficiency confers a significantly higher mutation rate at the coding microsatellites of TGFBR2 and ACVR2 compared to hMSH6 and hMSH3 deficiencies. In addition, TGFBR2 mutates at a higher rate than ACVR2 in both hMLH1 and hMSH6 deficiencies. These bona-fide human genes targeted for mutation in MMR deficiency mutate at differing rates, and lose expression of their encoded proteins in colonic neoplastic cells. Understanding these targeted genes in MMR deficiency has implications in understanding the pathogenesis of MSI colorectal tumors.

Materials and Methods
Cloning of pIREShyg2-TGFBR2-EGFP and pIREShyg2-ACVR2-EGFP plasmids Plasmid pIREShyg2-EGFP was a kind gift from C. Richard Boland, MD (Baylor Univ. Med Center, Dallas, TX). Details of cloning of pIREShyg2-EGFP were previously described [24]. Portions of exon 3 of TGFBR2 and exon 10 of AVCR2 (shown in Fig. 1) were amplified by PCR from the MMR proficient human colon carcinoma cell line FET (kind gift of Michael Brattain, Ph.D. Roswell Park Cancer Inst; Buffalo, NY). New PmeI and AscI sites were created in the 59 and 39 ends of those TGFBR2 and ACVR2 sequences by PCR, respectively (primers: 59-GCGTCGTTTAAACCTGCTTCTCCAAAGTGCATTATG-39 and 59-AAGGCGCGCCAAGAAAGTCTCACCAGGCTT-39 for TGFBR2 and 59-AGCTTTGTTTAAACGACCTGTAGAT-GAATACATGT-39 and 59-AAGGCGCGCCAAACAGGCCT CTTTTTTTTATG-39 for ACVR2). The PCR products and pIREShyg2-EGFP were digested with PmeI and AscI (New England Biolabs, Ipswich, MA) and the digested PCR products were cloned into PmeI-AscI sites of pIREShyg2-EGFP to generate pIREShyg2-TGFBR2-EGFP and pIREShyg2-ACVR2-EGFP plasmids (Fig. 1). Experimental plasmids were constructed in which the TGFBR2 and ACVR2 sequences were cloned +1 bp OF in pIREShyg2-EGFP immediately after the translation initiation codon of the EGFP gene, and thus frameshift mutation of 21 bp would allow expression of EGFP (Fig. 1). As negative control plasmids for EGFP expression, mutation resistant (MR) counterpart plasmids (+1 bp OF plasmids) were constructed by changing 1 or 3 nucleotide sequences (A 10 to A 2 CA 2 GA 2 CA in TGFBR2 and A 8 to A 3 GA 4 in ACVR2) within microsatellites using a Quickchange II site-directed mutagenesis kit (Stratagene, La Jolla, CA), preventing any frameshift mutation (Fig. 1). MR IF plasmids containing portions of TGFBR2 or ACVR2 were constructed as positive controls for EGFP expression (Fig. 1). The ligation products were transformed into DH5a cells. Positive colonies were screened, and the correct sequences of TGFBR2 and ACVR2 were confirmed by sequencing in an ABI 3700 analyzer.

Analysis of mutant cells by flow cytometry
Five thousand nonfluorescent cells expressing MR TGFBR2 OF, TGFBR2 OF, MR ACVR2 OF, or ACVR2 OF were sorted into 24-well plates on a FACS ARIA by using Diva software (Becton Dickinson Immunocytometry Systems (BDIS), San Jose, CA). During a 7 to 35 day analysis period, cultures were expanded as required to keep cells in exponential growth. Cells were trypsinized, washed in PBS, and resuspended in a total volume of 200 ml of PBS/0.5 mg/ml of propidium iodide (PI) and 3% BSA. Cell suspensions were analyzed on a FACSCalibur with CELLQUEST acquisition and analysis software (BDIS, CA). At specified time points, three cultures were analyzed in parallel. To identify EGFP-positive cells, region 1 (R1) was set in the forward/ side scatter and region 3 (R3) was set in the forward/PI scatter, and then R1 and R3 were gated by live cells. Region 2 (R2) was set in the fluorescence 1 (FL1, green)/fluorescence 2 (FL2, red) scatter. Cells from the gated R1, R3, and R2 were plotted further on a fluorescence intensity histogram, and three populations were separated. The population displaying no fluorescence was named M0, the population with low fluorescence intensity, M1, and the one with high fluorescence intensity, M2. The counts of M1 and M2 cells were expressed as percentages of R3 (total live cell number).

PCR and DNA sequencing
Total cellular DNA from stable cell lines and M1 and M2 cell populations were PCR-amplified by specific primers (59-GCG TCGTTTAAACCTGCTTCTCCAAAGTGCATTATG-39 and 59-TGCCGTCGTCCTTGAAGAAGA-39 for exon 3 of TGFBR2 and 59-GATCCGCCACCATGTTTAAACGAC-39 and 59-GCTGTTGTAGTTGTACTCCAGCTTG-39 for exon 10 of ACVR2) in a reaction containing the primers, buffer, DNA template, deoxynucleotides, and Pfu Ultra high fidelity DNA polymerase (Stratagene). The PCR products were used for DNA sequencing to identify stable cell lines and frameshift mutations at coding microsatellites. In addition, we subcloned PCR-amplified TGFBR2 and ACVR2 DNA fragments from M1 and M2 cell populations utilizing a TA cloning vector (Invitrogen Corp) as per the manufacturer's protocol. DNA clones were then individually sequenced to determine the prevalence of mutated and WT TGFBR2 and ACVR2 sequences.
Determination of 21 bp frameshift mutation rates of TGFBR2 exon 3 and ACVR2 exon 10 in human cells The mutation rate was defined as the probability of a cell undergoing a mutation in its lifetime and expressed per cell per generation. We used a ''method of the mean'' developed by Luria and Delbruck [28] to estimate mutation rate. The ''method of the mean'' is moment-based, whereby the mutation rate is estimated as a function of the sample mean of the number of mutants. The formula used in the computation is r = mN ln(mNC), where r is the mean number of mutants in a culture, C is the number of parallel cultures, m is the mutation rate, and N is the number of cells at risk of undergoing a mutation, which Luria-Delbruck assumed to be equal to the final number of cells in a culture. Three parallel cultures were used, and r was estimated as the mean of the number of mutants across the three cultures. The total number of cells N was based on averaging across cultures. The formula listed above was used to calculate mutation rates of the M2 cell population (full mutants) using data from flow cytometry analysis at each time point between day 14 and day 35. Single mutation rates were then calculated by combining and averaging time-specific mutation rates to minimize the variance of the estimate as previously described [25]. Data were expressed as mean6the standard errors of mean (SEM).

Statistical analysis
Mutation rates of cell lines were compared by T-test or one-way ANOVA.