INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The inactivation of tumor-suppressor genes plays a crucial role in cancer (7, 19, 33, 38, 65, 77, 79, 81). Gene inactivation pathways can be subdivided into genetic and epigenetic mechanisms. The genetic mechanisms by which a gene can be inactivated include gene loss (19, 77, 81), gene truncation, and base pair deletions, insertions, and substitutions (33, 65). Loss of heterozygosity (LOH) is an operational term that has gained widespread use to describe the result of various mechanisms of gene loss, including large or small deletions spanning the gene (or parts of it), as well as chromosomal loss with or without reduplication of the other chromosome, and nonreciprocal recombination events, such as gene conversion (19, 77, 81). Intragenic mutations can be subdivided into truncating mutations and frame-preserving mutations. Truncating mutations include base substitutions resulting in nonsense mutations and frameshift mutations caused by insertions and deletions. Frame-preserving mutations include small in-frame insertions and deletions, as well as missense mutations.

A gene can also lose its functionality by epigenetic mechanisms, such as transcriptional or posttranscriptional downregulation (7, 38, 55, 79). Promoter DNA methylation and the associated chromatin changes are important epigenetic mechanisms of transcriptional silencing in eukaryotic cells (67, 68), but cytosine-5 methylation has also been implicated in various genetic mechanisms of gene inactivation (12, 25, 44, 65). In this study, we have dissected the relative contribution of DNA methylation to various competing gene inactivation pathways by using gene targeting of the predominant DNA methyltransferase, Dnmt1.

Most studies of gene inactivation have utilized either endogenous genes such as HPRT (coding for hypoxanthine phosphoribosyltransferase) (50, 56) or exogenously introduced counterselectable markers, such as herpes simplex virus thymidine kinase (HSV-TK) (10) or HPRT (50) cDNAs. In this study, we have used a counterselectable marker encoding a fusion protein between HSV-TK and neomycin phosphotransferase (71). Counterselection of the TK activity with drugs such as 1-(2-deoxy-2-fluoro--D-arabinofuranosyl)-5-iodouracil (FIAU) or ganciclovir should allow detection of any of the gene inactivation events described above. In addition, the unique feature of a fused positive selectable carboxy-terminal marker provides several added advantages. First, selection for functional neomycin phosphotransferase (Neo) activity with the drug G418 can serve to exclude cells with preexisting gene inactivation events before the start of the mutation assay. Second, simultaneous selection against TK and for Neo provides a novel strategy for detecting missense mutations that disrupt TK activity, but leave the Neo activity intact (see Fig. 1A). Finally, we have further modified the basic counterselection strategy by generating head-to-tail concatemeric repeats of the counterselectable marker, thereby providing a preferential detection of LOH events (Fig. 1A).

The role of DNA methylation in gene inactivation has been investigated by Chen et al., using Dnmt1 gene targeting in embryonic stem (ES) cells with a slightly different strategy of HPRT and HSV-TK counterselection (12). Chen et al. conclude that DNA hypomethylation results in an increased rate of rearrangements and gene loss by mitotic recombination (12). We have used an approach similar to that of Chen et al. (12), but with different results. In addition, we have investigated alternative mechanisms of gene inactivation, including gene silencing by promoter methylation and gene mutation. It is widely acknowledged that methylation of promoter regions is associated with reduced transcriptional activity and altered chromatin structure (8, 32, 39, 67). Therefore, we expect gene silencing by promoter methylation to be very sensitive to manipulation of DNA methylation levels.

There is a large body of literature implicating cytosine-5 DNA methylation in transition mutations at CpG dinucleotides in vertebrates (17, 45, 62, 76, 85). There are four separate observations that suggest that 5-methylcytosine undergoes mutation at a higher rate than the 4 unmodified bases. First, organisms with CpG methylation show evidence of evolutionary loss of the dinucleotide CpG (84), resulting in a depletion of CpG in the genome (76). Second, CpG transition mutations represent the single most common type of somatic point mutation of the TP53 gene in human cancer (27, 31, 33, 65). Third, CpG transition mutations are responsible for approximately one-third of all human hereditary disease mutations (17). Fourth, CpG transition mutations are the most common type of point mutation found in mutation assays in vivo and in vitro (34, 35, 62, 63). In all four of these examples, the evidence for a role of DNA methylation is inferred from the overrepresentation of transition mutations observed at CpG dinucleotides, rather than from direct experimental evidence for the involvement of DNA methylation. However, the concept that CpG hypermutability in vertebrate genomes is directly attributable to 5-methylcytosine is widely accepted, since it has strong mechanistic support from the observation that spontaneous hydrolytic deamination of 5-methylcytosine occurs at high rates in vitro (80). yielding thymine as a result. The combination of Dnmt1 gene targeting in ES cells and the unique missense mutation assay that we have developed has provided the first opportunity to directly test this hypothetical mechanism in an experimental system. In addition, work with bacterial cytosine-5 methyltransferases has shown that the enzyme itself can contribute to deamination of cytosines in the target recognition sequence under conditions involving a limiting supply of the methyl donor S-adenosylmethionine (SAM) (6, 73, 89). As part of this study, we have investigated whether methyl group-deficient growth conditions result in Dnmt1-dependent elevated missense mutation rates.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Plasmids. A 2.7-kb BglII-SalI TKNeo fragment was generated from pTNFUS69 (71) and pPGKPuro (78) and ligated into the BamHI-SalI backbone of the pSL301 cloning vector (Invitrogen). The puromycin marker-containing Dnmt1-targeting vector, pMC-PURO, was constructed from the insertion-type targeting vector pMWT-PURO (78) by digestion of a unique ClaI site upstream of the conserved active site and treatment with Klenow DNA polymerase (New England Biolabs) to create a frameshift. Analysis of the resulting clone showed that a 125-bp stretch of DNA had been duplicated at the ClaI site, resulting in a frameshift and termination of the reading frame 46 codons downstream of the ClaI site. The hygromycin marker-containing Dnmt1-targeting vector pMC-HYG was constructed from pPWL512 (37) and pMWTPURO (78), resulting in the deletion of Dnmt1 sequences downstream of the ClaI site.

ES cell culture and transfection. ES cells were maintained in HEPES-buffered (20 mM [pH 7.3]) Dulbecco's modified Eagle's medium (DMEM; JRH Biosciences) supplemented with 15% fetal calf serum (Hyclone Labs), 0.1 mM nonessential amino acids (Gibco), 0.1 mM -mercaptoethanol (Sigma), and penicillin-streptomycin (Irvine Scientific). J1 cells (49) were grown on feeder layers of gamma-irradiated embryonic fibroblast cells and supplemented with leukemia inhibitory factor (LIF) (Gibco) at 10⁶ U/ml. To construct synthetic concatemers of PGK-TKNeo-PA, the 3-kb SalI-XhoI fragment of pSLTKNeo was subjected to repeated rounds of ligation and digestion with SalI and XhoI. This reiterative procedure specifically selects for the generation of head-to-tail concatemers, since the ligation of SalI and XhoI compatible sticky ends results in the destruction of the recognition sequence for either of the two enzymes, while head-to-head (SalI-SalI) or tail-to-tail (XhoI-XhoI) ligation results in the regeneration of cleavable SalI or XhoI recognition sequences. Cells were electroporated in a mixture of 20 mM HEPES (pH 7.0), 137 mM NaCl, 5 mM KCl, 0.7 mM Na₂HPO₄, 6 mM glucose, and 0.1 mM -mercaptoethanol, with 10 to 30 µg of linearized DNA at a set voltage of 400 V and a capacitance of 25 µF, in a 0.4-cm-diameter cuvette with a Bio-Rad GenePulser II. Antibiotic selection was initiated the following day and continued for 8 to 11 days before picking. Puromycin was used at a concentration of 2µg/ml. G418 (Gibco) was used at an active concentration of 250 µg/ml. Hygromycin B (Roche) was used at a concentration of 100 µg/ml. Cells were expanded and frozen two days after picking. A parallel plate was used for DNA isolation as described previously (46). Cells for the methionine-deficient and ethionine assays were grown in DMEM without L-methionine (Gibco; catalog no. 11970-035) supplemented with 14% dialyzed fetal bovine serum (Gibco), 0.9 mM sodium pyruvate, HEPES (6.24-g/liter final concentration), Na₂CO₃ (1.2-g/liter final concentration), 0.1 mM nonessential amino acids (Gibco), 0.1 mM -mercaptoethanol (Sigma), penicillin-streptomycin (Irvine Scientific), and LIF (Gibco) at 10⁶ U/ml. This supplemented DMEM without methionine was modified to make control medium (0.2 mM methionine [Sigma; catalog no. M2893]), ethionine medium (0.2 mM methionine plus 1 mM ethionine [Sigma; catalog no. E1260]), and homocystine medium (0.005 mM methionine plus 1 mM L-homocystine [Sigma; catalog no. H4137]).

Gene inactivation assays. Before fluctuation analysis, multiple parallel vials of a common culture of cells were generated for each clone. The fraction of initial mutants present in the original samples was determined by thawing one of the parallel vials and plating the entire vial into medium with 2 µM FIAU for FIAU selection or with 2 µM FIAU plus 250 µg of G418 per ml for FIAU + G418 selection. Subsequently, the inadvertent inclusion of intitial mutants in the fluctuation analyses of the other parallel vials was avoided by choosing the number of cells plated for the expansion step to be at least 3 orders of magnitude below the number calculated from the initial mutant frequency to contain an average of one initial mutant. For the fluctuation analysis, one of the parallel vials of cells was thawed, and 12 parallel cultures of 1,000 cells per well in a six-well plate were plated onto feeders in normal medium without selection for the expansion phase of the fluctuation analysis. Counting the day of plating as day 0, medium was changed on days 3, 6, and 9. On day 10, the cells were trypsinized, and the concentration of live and dead cells was determined by hemacytometer. For FIAU selection, a defined volume (200 µl) from each of the 12 parallel expansions was seeded into another six-well plate, without feeders, in medium with 1 µM FIAU. For the FIAU + G418 selection, a defined volume (2 ml) from each of the 12 parallel expansions was seeded into 10-cm-diameter dishes, without feeders, in medium with a combination of 1 µM FIAU and 250 µg of G418 per ml. In addition, 1,000 cells were seeded in medium without selection into six-well plates to determine plating efficiency for selection. Counting the day of plating as day 0, medium was changed on days 4, 8, and 12 with medium containing 2 µM FIAU (twofold increase over initial seeding concentration) or a combination of 2 µM FIAU and 250 µg of G418 per ml. Cells for determination of plating efficiency were stained and counted on days 8 to 10. The number of colonies on the selection plates was determined by a systematic microscopic scan at a 100× magnification to ensure the validity of each colony and to ensure that small colonies were not overlooked. The counting was always performed on the same day for the different genotypes or treatment categories within a single experiment, but varied from experiment to experiment, depending on the growth rate of the colonies, and fell between days 10 and 14 for FIAU selections and between days 13 and 19 for FIAU + G418 selections. The Luria-Delbrück method of means equation was used to calculate mutation rates (40, 53). The equation r = aN ln(NCa) was used, where r is the average number of resistant clones per parallel expansion (corrected for plating efficiency at the selection phase and corrected for the fraction of live cells seeded for selection), a is the mutation rate, N is number of live cells seeded for selection, and C is the number of parallel cultures. The rates calculated from this transcendental equation were derived by a Microsoft Excel spreadsheet-assisted approximation method.

Molecular analyses. For Southern blot analysis, 10 µg of genomic DNA was digested with the appropriate restriction enzyme, separated on a 0.7% SeaKem ME agarose gel, and blotted onto a Zetabind nylon membrane (Cuno Laboratory Products). Southern blot hybridization was performed as described previously (15, 22). Blots were washed at 65°C with 0.5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-1% sodium dodecyl sulfate. The following probes were used. HV is a 0.6-kb HindIII-EcoRV fragment of genomic DNA that includes the second intron of Dnmt1; the minor satellite centromeric repeat probe is a 150-bp PstI fragment of pMR150 (11); and the Neo probe is a 637-bp PstI fragment derived from the repaired neomycin phosphotransferase II sequence (88). LOH analysis of TK was performed by PCR with the the HSV1-TK primers 5'-GCA TGC CTT ATG CCG TGA CCG ACG-3' (TK4B) and 5'-GCC AGG CGG TCG ATG TGT CTG TC-3' (TK5B). The Mlh1 control primers used were 5'-AGG AGC TGA TGC TGA GGC-3' (OL 117/sense) and 5'-TTT CAT CTT GTC ACC CGA TG-3' (OL 118/anti-sense) (4). Screening of the TKNeo gene for mutations was performed by amplification of a 946-bp region spanning all previously described hot spots for TK mutation (2, 9, 21, 42, 57). The primers used for primary PCR amplification were 5'-CGA CCA GGC TGC GCG TTC TCG-3' (TK-MC1) and 5'-CCA GGA TAA AGA CGT GCA TGG AAC GG-3' (TKMC-2). The nested primers used for the sequencing reaction were 5'-GGC CAT AGC AAC CGA CGT ACG G-3' (TK-MC1C/forward) and 5'-CGT TTG GCC AAG ACG TCC AAG GC-3' (TK-MC2B/reverse). Methylation analysis of the TKNeo gene was performed by bisulfite genomic sequencing (16). Primers 5'-CATCTACACCACACAACAC-3' (forward primer) and 5'-GGGGTTATGTTGTTTATAAGG-3' (reverse primer) were used to amplify a 254-bp PCR product from the bottom strand of bisulfite-converted DNA, and the PCR product was cloned into plasmid vectors with the TA cloning kit (Invitrogen). Primer TA-5' (5'-CAGTGTGCTGGAATTCGGC-3') was used to sequence DNA from individual colonies. Methylation analysis of the Pgk1 promoter was performed by methylation-sensitive single nucleotide primer extension (Ms-SNuPE) assay (26). The sequences of the primers used for amplifying a 375-bp PCR product from bisulfite-treated DNA were 5'-CAAATAAAAATAACACATCTCACTAATCT-3' (5'-Pgk) and 5'-TTATTAAGATTTAGATGGATGTAGGT-3' (3'-Pgk). PCR products were gel purified with the Qiaquick gel extraction kit (Qiagen), and the template was resuspended in 30 µl of H₂O. The sequences of the primers used for the Ms-SNuPE reactions were GAGTAAAGTTGTTATTGGT (SNUPE 4) and ATGTAGGTCGAAAGGTT (SNUPE 6). Ms-SNuPE primer extension reactions were performed in a single cycle of 2 min at 95°C, 2 min at 50°C, and 1 min at 72°C in a 10-µl total volume containing 4 µl of Qiaquick eluent, 20 mM Tris-HCl (pH 7.5), 2.5 mM MgCl2, 100 mM KCl, a 0.5 µM final concentration of each primer, 0.5 U of Taq polymerase and 1 µCi of either [³²P]dCTP or [³²P]dTTP. The reaction mixtures were combined with 4 µl of stop solution before being denatured at 95°C for 5 min and loaded onto a 15% denaturing polyacrylamide gel (7 M urea). Quantitation of methylation levels was performed on a Molecular Dynamics PhosphorImager. All expression analyses were performed by real-time fluorescence-based reverse transcription (RT)-PCR (TaqMan) as described previously (20). Total RNA was isolated from a confluent T25 flask and lysed by a single-step guanidinium isothiocyanate method (13). Two micrograms of total RNA was reverse transcribed with random hexamers, deoxynucleoside triphosphates, and SuperScript II reverse transcriptase (Gibco-BRL) in a 25-µl reaction mixture as described previously (24). The TaqMan primer and probe sequences are listed below. In all cases, the first primer listed is the forward PCR primer, second is the TaqMan probe, and the third is the reverse PCR primer: Dnmt1 (GGCTTGGGCAGCCTATGAA, 6FAM5'-AGCATCTCCTCATCGATGCTCACCTTCTG-3'TAMRA, GGAATGACCGAGACGCAGTC), histone H4 (TCTCCGGCCTCATCTACGAG, 6FAM5'-CACCTTCAGCACACCGCGGGT-3'TAMRA, CGGATCACGTTCTCCAGGA), and TK (ATACCGACGATATGCGACCTG, 6FAM5'-CGCGCACGTTTGCCCGATC-3'TAMRA.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Model system to study gene inactivation pathways. The TKNeo gene encodes a fusion protein that has both HSV-TK and neomycin phosphotransferase (Neo) functions (71) and therefore has dually selectable and counterselectable properties. This fusion protein formed the basis of a selection system to measure gene inactivation pathways. An overview of this selection strategy is shown in Fig. 1A. Cells expressing the TKNeo fusion protein are sensitive to FIAU, but resistant to G418. Loss of the entire fusion gene, nonsense and frameshift mutations within TK, or methylation of the phosphoglycerate-kinase (Pgk1) promoter upstream of TK would confer FIAU resistance and G418 sensitivity to affected cells. Missense mutation events within TK would also confer FIAU resistance, but the cells would remain G418 resistant. Each of these pathways was expected to be readily detectable for single gene integrations of the TKNeo marker. However, head-to-tail concatemeric repeats of TKNeo were expected to result in the preferential detection of deletions spanning all copies of the marker, since point mutations and promoter methylation would require the concomitant inactivation of all copies.

View larger version (35K): [in this window] [in a new window]   FIG. 1.   Gene inactivation assay. (A) Schematic describing the use of drug selection for the detection of various mechanisms of gene inactivation at either a single-copy HSV-TKNeo or a multicopy HSV-TKNeo concatemer. HSV-TKNeo encodes a fusion protein between the HSV-TK gene and the Neorgene. Cells carrying this marker (before selection) are FIAU- sensitive (FIAUS) and G418 resistant (G418R). Counterselection of TK with FIAU can detect LOH, mutation, and methylation events. Simultaneous selection with FIAU and G418 is expected to select primarily for missense mutations in TK that leave the Neo reading frame intact. The effects of HSV-TKNeo copy number on the expected contribution by each of these gene inactivation pathways are indicated. (B) Identification of single-copy or concatamerized TKNeo ES cell clones. Restriction maps of integrated single-copy and HSV-TKNeo concatemers show the ApaI (A) sites within TKNeo and the flanking BamHI (B) sites. Southern blot analysis of ApaI- and BamHI-digested DNA from a single-copy TKNeo integrant (BA-30) using a Neo probe (small filled rectangles) identified single fragments in each digest. TKNeo concatamerization in a multicopy HSV-TKNeo clone (BA-10) was evident from an intense 2.8-kb ApaI head-to-tail fragment and a weaker large flanking fragment.

Targeted mutation of the Dnmt1 gene. To assess the effect of DNA methylation on the various gene inactivation pathways, sibling ES cell lines were generated with identical TKNeo integrations, but with different Dnmt1 genotypes (Fig. 2). Figure 2A shows a pedigree of the cell lines that were used in this study. First, one allele of the Dnmt1 gene was disrupted in J1 wild-type cells (49) by using an insertion-type vector containing a puromycin resistance marker (P) to generate the Dnmt1^+/P cell line BA (see Materials and Methods and reference 79). Next, the TKNeo gene was introduced into the BA cell line either as a single-copy integration (BA-30) or as a multicopy concatemer (BA-10). Finally, a second gene targeting step with an insertion-type vector containing a hygromycin resistance marker (H) gave rise to either Dnmt1 heterozygous cells (Dnmt1^+/PH) or Dnmt1 homozygous knockout cells (Dnmt1^P/H), depending on whether the hygromycin vector had recombined at the wild-type allele or at the previously targeted puromycin allele. Southern blot analysis was used to determine which of the two events had taken place (Fig. 2B).

View larger version (56K): [in this window] [in a new window]   FIG. 2.   Targeted mutation of the Dnmt1 gene. (A) Pedigree describing cell lines established following two Dnmt1 targeting events and TKNeo transfection. Dnmt1 wild-type clones are shaded black, Dnmt1 heterozygous clones are shaded gray, and Dnmt1 homozygous mutant clones are in white. Dnmt1 targeting of wild-type (J1) ES cells by using a puromycin-containing plasmid (P) generated the Dnmt1+/P cell line BA. This cell line was then transfected with TKNeo (white- and black-striped bar) to yield Dnmt1+/P cell lines containing either a single-copy TKNeo gene (BA-30) or concatamerized TKNeo gene (BA-10). The concatemers were generated in vitro by ligation (see Materials and Methods). These cell lines underwent another round of gene targeting at Dnmt1 with a hygromycin-containing plasmid (H) to generate sibling single-copy TKNeo Dnmt1+/PH (BF-14, BF-23) and Dnmt1P/H (BF-16, BF-19) cell lines and sibling multicopy TKNeo Dnmt1+/PH (BI-6, BI-18) and Dnmt1P/H (BI-4, BI-7) cell lines. (B) DNA from isolated clones was digested with the restriction enzyme KpnI and hybridized with a genomic Dnmt1 probe spanning the HindIII to EcoRV sequence of intron 2 (HV probe [small solid squares]). Wild-type alleles of Dnmt1 produced 7-kb fragments following KpnI digestion. Insertion of the puromycin-containing vector into the wild-type allele increased the size of the KpnI fragment to 22-kb. Targeting of the hygromycin-containing vector into the resulting Dnmt1+/P cells can occur at either the wild-type or the previously targeted allele. Dnmt1+/PH clones were generated by insertion into the previously targeted allele, yielding 36- and 7-kb bands following KpnI digestion (clone BI-6). The 36-kb band contains three copies of the probe and therefore gives a stronger hybridization signal than the 7-kb band. Dnmt1P/H clones were generated following insertion into the remaining wild-type allele, yielding 21- and 22-kb bands (clone BI-7). (C) Dnmt1 expression analysis. TaqMan RT-PCR was used to measure Dnmt1 expression levels in Dnmt1+/+, Dnmt1+/PH, Dnmt1P/H, and Dnmt1c/c ES cells. The Dnmt1 cDNA products for each of the alleles are shown, as are the ClaI and BstEII sites used to generate the mutant alleles. The cDNAs present in each cell type are also expressed schematically below each bar, with the black section representing the area of the cDNA up to the truncation caused by the mutation. The P, H, and PH alleles are the mutated Dnmt1 alleles generated by gene targeting as described for panels A and B. The C allele is a previously described Dnmt1-null mutant (47). The TaqMan RT-PCR region analyzed is indicated by the dashed lines and spans the ClaI site-mutated P, H, and PH alleles. The BstEII site mutated in the C allele is located downstream of the amplicon used in the TaqMan RT-PCR. Therefore, the RT-PCR amplicon is intact and transcribed in the C allele (although the mRNA product may be less stable). DnmtI expression values were first normalized to histone H4 expression values before comparison to wild-type expression levels. The number of independent clones analyzed for expression is indicated below each bar. The error bars indicate the standard error of the mean. (D) Methylation analysis of wild-type and targeted clones. Genomic DNA was digested with the methylation-sensitive enzyme HpaII (H) and its methylation-insensitive isoschizomer, MspI (M) and hybridized on a Southern blot with a minor satellite centromeric repeat probe, pMR150 (11). Extensive HpaII digestion of DNA from Dnmt1P/H and Dnmt1C/C cells, indicating hypomethylation of the repeat sequences, was not observed in the wild-type and Dnmt1+/PH cells.

Effect of Dnmt1 deficiency on gene inactivation rates. To quantitate the effects of DNA hypomethylation on gene inactivation rates, we performed fluctuation analyses with the Luria-Delbrück method of means equation (40, 53). We first analyzed the effects of Dnmt1 deficiency on the combined rate of inactivation of the TKNeo gene, by subjecting ES cell clones with either a single-copy TKNeo gene or multicopy TKNeo concatemer to counterselection with just FIAU. In each experiment, sibling pairs of Dnmt1^+/PH and Dnmt1^P/H clones were tested side by side under identical experimental conditions. We performed five independent fluctuation analyses of pairs of single-copy integrant clones and four independent fluctuation analyses of pairs of multicopy concatemers. Each fluctuation analysis consisted of 12 parallel expansions. Therefore, the data shown in Fig. 3B are based on a total of 216 [2 × (4 + 5) × 12] separate expansions. The average rate of collective TKNeo inactivation by all mechanisms in the single-copy Dnmt1^+/PH cells was found to be 4.8 × 10⁴, whereas the rate measured for the single-copy Dnmt1^P/H cells was 2.6 × 10⁴. The gene inactivation rates for the multicopy concatemeric insertion were slightly lower at 1.6 × 10⁴ in the Dnmt1^+/PH cells and 0.9 × 10⁴ in the Dnmt1^P/H cells. The lower rate of gene inactivation for the multicopy concatemers compared to that with the single-copy insertion is consistent with the fact that some gene inactivation mechanisms are not likely to result in the concomitant inactivation of all copies of TKNeo, as outlined in Fig. 1A. For both the single-copy and multicopy insertions, the gene inactivation rates were consistently lower in the Dnmt1-deficient hypomethlated cells. This difference was statistically significant (P = 0.0326) for the single copy, and close to significant (P = 0.0663) for the concatemer by a two-sided paired t test. The combined data for all nine pairs of fluctuation tests of Dnmt1-proficient and Dnmt1-deficient cells gave a P value in a paired t test of 0.0108. We conclude that Dnmt1 deficiency causes a reduction in the rates of gene inactivation at these two loci. These results are not consistent with the observation by Chen et al. of an enhanced rate of gene inactivation in Dnmt1-deficient cells (12).

Analysis of TKNeo LOH. Mechanisms of gene loss that result in LOH collectively represent an important pathway of gene inactivation (51, 77, 81). We used PCR analysis of a 266-bp sequence in the central region of the TK gene to investigate what fraction of FIAU resistant clones, resulting from the fluctuation analyses presented above, had undergone loss of TKNeo gene sequence. Figure 3B shows representative PCR results obtained with 10 single-copy TKNeo and 10 multicopy TKNeo FIAU-resistant colonies, and the data are summarized in Fig. 3C. The vast majority of clones had lost this central section of the TKNeo gene. Of the 384 clones analyzed that showed amplification of the Mlh1 control gene, 367 did not amplify the TK gene sequence. Interestingly, we found no occurrences of TKNeo retention in Dnmt1-deficient cells, suggesting that some cases of TKNeo gene retention could perhaps be attributed to gene silencing by promoter DNA methylation in Dnmt1-proficient cells. The frequency of retention of TKNeo in these cells was higher in single-copy cells than in the multicopy TKNeo cells (Fig. 3C), consistent with the expectation that inactivation of multicopy concatemers would occur primarily by LOH (Fig. 1A).

View larger version (25K): [in this window] [in a new window]   FIG. 3.   Gene inactivation rates. (A) Rates of TK gene inactivation following FIAU counterselection. The method of means equation from Luria-Delbrück fluctuation analysis was used to calculate the rate of events leading to FIAU resistance per cell division. For the pair of single-copy TKNeo clones, the average rates were calculated from five independent fluctuation analyses representing a total of 120 independent parallel cultures. For the pair of multicopy concatemer clones, the average rates were calculated from four independent fluctuation analyses representing a total of 96 independent parallel cultures. The bars represent the mean value obtained for these fluctuation analyses. Error bars indicate the standard error of the mean. (B) PCR analysis of FIAU-resistant clones. Representative PCR analysis of 10 single-copy and 10 multicopy FIAU-resistant TKNeo clones with primers against TK and Mlh1 are shown. PCR amplification of a 266-bp region of TK was used to assess loss of this central portion of the TK gene. PCR amplification of the Mlh1 gene served as a control for the quantity and the integrity of the DNA samples. Loss of TK was scored when lack of TK amplification was accompanied by Mlh1 amplification. Single-copy clone no. 5 and multicopy clone no. 17 are two examples of clones that have retained TK. (C) Summary of the contribution of LOH in the inactivation of TK. The relative contributions of retention and loss of TK to the gene inactivation rates shown in panel A were calculated based on the frequency at which TK was retained in resistant clones. The fractions listed under each bar indicate the number of clones that lost TK/ total number of FIAU-resistant clones as determined by PCR.

Analysis of gene silencing. Although gene loss was the major mechanism by which TK expression was eliminated in the cells, the relative contributions of the other gene inactivation pathways could be determined in the FIAU-resistant clones with retention of TKNeo. The most likely mechanism for the inactivation of the retained concatemers in the multicopy insertion of TKNeo is gene silencing by promoter DNA methylation, since concerted, linked point mutations are expected to be rare. We therefore analyzed both TKNeo gene expression and Pgk1 promoter DNA methylation levels in all parental cell lines and in a FIAU-resistant clone derived from one of the concatemeric lines (Fig. 4). The methylation analysis was performed by MS-SNuPE (26) at two CpG dinucleotides: one upstream and one immediately downstream of the transcription start sites of the Pgk1 promoter (1) (Fig. 4). Gene expression analysis was performed by real-time RT-PCR. The five single-copy parental cell lines all expressed TK and all had corresponding low levels of promoter methylation (23% or lower). Good expression of TK and low levels of promoter methylation were also seen in the multicopy concatemeric parental cell lines. However, a FIAU-resistant clone derived from the BI-18 multicopy parent showed a lack of detectable TK gene expression and an accompanying high level of promoter DNA methylation (95%), suggesting that promoter methylation was responsible for TK inactivation in this particular clone.

View larger version (35K): [in this window] [in a new window]   FIG. 4.   TK gene expression and corresponding Pgk1 promoter methylation. (A) CpG density map of the murine phosphoglycerate kinase (Pgk1) promoter (1). The locations of the CpG dinucleotides interrogated by the MS-SNuPE primers 4 and 6 are indicated by vertical arrows. The locations of the two transcription start sites (1) are indicated by horizontal arrows. (B) Autoradiographs of four representative MS-SNuPE analyses. Incorporation in the C lane is indicative of CpG methylation in the genomic DNA before bisulfite conversion, while incorporation in the T lane is indicative of lack of methylation (26). (C) Summary of methylation and expression results. Every cell line in the pedigree shown in Fig. 2A and two FIAU-resistant mutants derived from the BI-6 and BI-18 cell lines were analyzed for TK expression and Pgk1 promoter methylation. TK gene expression was measured by TaqMan RT-PCR. The expression values were expressed as a ratio to histone H4 gene expression to control for the amount of input cDNA. The BI-4 cell line had the highest level of TK expression and was therefore used as the reference for comparison of TK expression levels in the other cell lines. The methylation levels of two CpG sites in the Pgk1 promoter were measured by Ms-SNuPE analysis.

Effect of Dnmt1 deficiency on missense mutation rates. Although we were able to detect an example of a frameshift mutation in one of the single-copy FIAU-resistant clones with TKNeo retention, the predominance of gene loss and gene silencing events in this FIAU selection assay precluded a quantitative analysis of the effects of Dnmt1 deficiency on point mutation rates. Therefore, we developed a novel assay that allowed us to specifically select for missense mutation events. The assay is based on a simultaneous selection with FIAU and G418. Clones resistant to this selection will likely have inactivated the TK activity, yet retained a functional Neo^r activity. The most straightforward mechanism by which this can be accomplished is by acquiring a missense mutation in the TK segment of the gene, leaving the Neo reading frame intact and unaltered. Four pairs of fluctuation analyses were performed with the Dnmt1^+/PH and Dnmt1^P/H cells carrying the single-copy TKNeo gene, for a total of 96 independent expansions. The Dnmt1^+/PH ES cells showed an average gene inactivation rate of 5.2 × 10⁵, whereas the Dnmt1^P/H cells showed an average rate of 0.66 × 10⁵ (Fig. 5A). These results showed that DNA hypomethylation decreased missense mutation rates by approximately eightfold. This decrease was found to be highly statistically significant by a two-sided paired t test (P = 0.0045). We were unable to obtain any G418 + FIAU-resistant clones from the multicopy concatemeric insertion clones, consistent with the prediction in Fig. 1A that concomitant point mutations in all of the linked copies of TKNeo are very unlikely to occur.

View larger version (46K): [in this window] [in a new window]   FIG. 5.   Missense mutation assay. (A) Rates of TK gene inactivation in single-copy TKNeo clones following FIAU + G418 selection. The method of means equation from Luria-Delbrück fluctuation analysis was used to calculate the rate at which TK became inactivated in these FIAU + G418-resistant clones. The average rates were calculated from four independent pairs of fluctuation analyses representing a total of 96 independent parallel cultures. The bars indicate the mean value obtained for these four fluctuation analyses, with the error bar denoting the standard error of the mean. (B) Spectrum of TK missense mutations in FIAU + G418-resistant Dnmt1+/PH and Dnmt1P/H cells. The sequenced areas of TK are indicated by the shaded regions, with the more heavily shaded regions represent more clones sequenced in that area. Amino acids at which mutations were detected are shown, and the amino acid changes are indicated. The upper half of the panel shows amino acid changes observed in the Dnmt1+/PH cells, while the lower half of the panel shows the amino acids changes observed in the Dnmt1P/H cells. ORF, open reading frame. (C) Summary of missense mutations. Mutations observed in the Dnmt1+/PH and Dnmt1P/H cells were grouped according to whether the change represented a transition or transversion mutation, whether the base change occurred at an A or C, and whether the mutation occurred at a CpG dinucleotide. (D) Spectrum of TK methylation. Bisulfite genomic sequencing of 24 Dnmt1+/PH clones and 24 Dnmt1P/H clones measured the methylation levels of 23 CpG sites (dark black lines) near the center of the TK gene in each of the clones. The height of the bars represents the percent methylation observed at each CpG site. The upper half of the panel shows the percent methylation observed in the Dnmt1+/PH cells, while the lower half of the panel shows the percent methylation observed in the Dnmt1P/H cells.

Analysis of mutation spectrum. Mutations were found scattered throughout the sequenced region in both genotypes with a cluster of Dnmt1^P/H mutants in the central area of the TK open reading frame (Fig. 5B). The number of individual mutation events in each category was too small to accurately define the spectrum of mutations in cells with and without functional Dnmt1 expression (Fig. 5C). However, a high incidence of transition mutations within CpG dinucleotides (5 of 20) was found in Dnmt1-deficient cells. This result was quite unexpected, since most CpG transition mutations are thought to be attributable to deamination of 5-methylcytosine. Therefore, we investigated whether the Dnmt1 knockout had resulted in a decrease in the level of methylation in the body of the TKNeo gene. We performed bisulfite genomic sequencing of a central part of the TKNeo gene in Dnmt1^+/PH and Dnmt1^P/H clones (Fig. 5D). We sequenced a total of 24 subcloned bisulfite-treated PCR products for each of the two genotypes. The methylation frequency of individual CpG dinucleotides is shown in the histogram in Fig. 5D. This section of TKNeo is almost completely devoid of cytosine-5 DNA methylation in the Dnmt1-deficient Dnmt1^P/H cells, which nevertheless gave rise to 5 transition mutations at CpG dinucleotides out of a total of 20 mutations. This suggests that the preferential occurrence of CpG transition mutations does not necessarily depend on the presence of cytosine-5 DNA methylation. However, it should be pointed out that since the overall mutation rate of Dnmt1^P/H cells is lower, the absolute rate of mutagenesis at CpG dinucleotides is reduced in Dnmt1-deficient cells.

Effect of methyl group deficiency on mutation rates. It is difficult to envision why cells lacking significant CpG methylation in the TKNeo gene would have such a high frequency of mutations at CpG dinucleotides. One explanation could be that these mutation events are not due to spontaneous hydrolytic deamination of 5-methylcytosine, but that they represent enzyme-dependent deamination events mediated by one of the other DNA (cytosine-5-)methyltransferases (Dnmt 3a, Dnmt3b, or even Dnmt2) (6, 44, 59-61, 73, 89), with accompanying methyl transfer to yield thymine (87) or with a block of uracil base-excision repair (86). A prediction of this scenario is that methionine-deficient growth conditions could enhance this mechanism (44, 73). To test the effects of methyl group deficiency on spontaneous mutation rates, the single-copy Dnmt1^+/PH and Dnmt1^P/H cells were expanded in media with different levels of methionine (Fig. 6). Three different media were generated by reconstituting methionine-free medium with either methionine (M), or with methionine plus ethionine (an antimetabolite of methionine [ME]), or with very low levels of methionine, but with homocystine as an amino acid replacement (H). The effects of the two methyl group-deficient media (ME and H) on the mutation rates, as measured with the missense mutation assay, are shown in Fig. 6. For the Dnmt1^+/PH cells, expansion under methionine-deficient conditions reduced the mutation rates, while mutation rates were enhanced in the Dnmt1^P/H cells under the methionine-deficient conditions. The sample sizes of both series were too small to yield statistically significant differences. However, the prediction of increased rates of mutagenesis caused by Dnmt1-mediated deamination in methyl group-deficient growth conditions does not seem to be borne out by this experiment.

View larger version (29K): [in this window] [in a new window]   FIG. 6.   Effect of methyl group deficiency on mutation rates. Rates of TK gene inactivation in single-copy TKNeo clones grown in medium containing various levels of methionine. The method of means equation from Luria-Delbrück fluctuation analysis was used to calculate the rate of which TK became inactivated in Dnmt1+/PH and Dnmt1P/H cells grown in the presence of FIAU and G418. The average rates were calculated from three independent fluctuation analyses representing a total of 18 independent parallel cultures for each medium condition. The compositions of the M, ME, and H media shown are as follows: M medium, 0.2 mM methionine; ME medium, 0.2 mM methionine plus 1 mM ethionine; and H medium, 0.005 mM methionine plus 1 mM L-homocystine. Error bars indicate the standard error of the mean.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

This study describes the quantitative analysis of effects of DNA methylation on different gene inactivation pathways. We chose to use mouse ES cells as our model system, since these cells can be readily manipulated by gene targeting, allowing us to assess the effects of DNA methylation on gene inactivation rates in sibling clones with and without functional Dnmt1, but carrying identical TKNeo gene integrations. We generated clones with either a single-copy or multicopy concatemerized TKNeo integration. The purpose of including multicopy concatemers of TKNeo was to create cell lines with an enhanced detection of LOH and a reduced detection of point mutation events. Indeed, this appears to have been largely successful, since we were not able to generate any dually resistant clones of the multicopy concatemers in our missense mutation assay. In our system, loss of TKNeo formally represents loss of hemizygosity, rather than loss of heterozygosity, since the TKNeo gene is located on only one allele. However, we do not anticipate there to be mechanistic differences between these two types of events. One caveat to our experimental approach is that the Dnmt1-proficient cells are heterozygous for Dnmt1, rather than wild type. Although the heterozygous ES cells in our experiments appear to have methylation levels similar to those of wild-type cells (Fig. 2D), this point should not be entirely disregarded, since Dnmt1 heterozygosity has been shown to have phenotypic consequences in vivo (43). Nevertheless, our data suggest that the Dnmt1 heterozygous cells are much more similar to wild-type cells than to Dnmt1-deficient cells.

The results obtained with FIAU selection and with the FIAU + G418 selection suggest that gene inactivation and mutation rates are lower at the integrated TKNeo loci in Dnmt1-deficient ES cells than in Dnmt1-proficient ES cells. This does not imply that the Dnmt1-proficient ES cells are abnormally unstable. We have introduced an experimental condition of genomic hypomethylation and found an associated reduced rate of genomic instability. Presumably, this is just one phenotypic consequence of the DNA hypomethylation, and it cannot be concluded that severe genomic hypomethylation is beneficial for the cell or for the organism under normal circumstances. In addition, some mechanisms of genomic instability that were not investigated in this study, such as cytogenetic abnormalities, could show opposite consequences of DNA hypomethylation.

The consistency of our findings across various mechanisms of gene inactivation prompted us to investigate whether these differences could in fact be attributed to some general characteristic of Dnmt1-deficient cells, such as a difference in growth rate. A difference in growth rate could theoretically affect the number of visible colonies at the time of counting. Although we made every effort to avoid an effect of growth rate on our experiments by counting both live and dead cells before and after expansion and at the time of seeding for selection, as well as by performing all colony counts by careful, microscopic scan, this possibility could not be entirely excluded. Therefore, we investigated whether there was indeed a difference in growth rate between Dnmt1^+/PH and Dnmt1^P/H cells under the different selection conditions and found that there was no such difference (data not shown).

One caveat to our approach is that differences observed between Dnmt1-proficient and Dnmt1-deficient cells may be attributable to the presence or absence of the Dnmt1 protein, as opposed to differences in genomic DNA methylation levels. It is becoming increasingly clear that the Dnmt1 protein has important functions separate from its DNA methylating properties (14, 23, 41, 67, 69). Another caveat is that the effects of Dnmt1 deficiency on mutation rates could be indirect. For instance, changes in DNA methylation levels could alter the expression of genes involved in DNA damage and repair. It is not obvious how one could distinguish in vivo between direct and indirect effects of DNA methylation levels.

Our novel procedure to select specifically for missense mutations combines two advantages over existing types of mutation assays. On the one hand, it is selective for point mutations and relatively resistant to detection of other events, such as deletion, a feature shared by existing single-site reversion assays (74). A drawback of these single-site reversion assays is their low sensitivity due to their small target size (a single nucleotide), whereas our assay combines the selectivity of these single-site reversion assays with the much greater target size (most of the open reading frame) of gene inactivation assays, such as HPRT and HSV-TK counterselection. This unique combination of selectivity and sensitivity allowed us to provide the first direct demonstration of the contribution of a DNA methyltransferase to an increase in the missense mutation rate in mammalian cells. Although the mutational burden of DNA methylation has often been inferred from analysis of mutation spectra, this is the first experimental corroboration that mammalian cells with reduced levels of genomic DNA methylation appear to have lower rates of mutation than their sibling cells with normal levels of DNA methylation. It is also the first time that a mammalian model system has been used to indirectly assess the role of DNA-methyltransferase-mediated mutagenesis.

The discrepancy in the frequency with which we found missense mutations in the Dnmt1^P/H (59%) and Dnmt1^+/PH (22%) dually resistant clones could potentially be explained by the occurrence of FIAU + G418-resistant Dnmt1^+/PH clones that had acquired sufficient levels of promoter methylation to render the cells resistant to FIAU, while maintaining just enough expression to retain resistance to G418. If this were indeed the case, this would necessarily imply that the level of TKNeo protein in the cell at which a cell loses G418 resistance is lower than the level at which a cell reaches FIAU resistance. To this end, we investigated the Pgk promoter methylation levels of two Dnmt1^+/PH dually resistant clones in which we had not been able to detect mutations. One such clone had promoter methylation levels of 38%, compared to 23% in the BF-14 parent before selection, while the other dually resistant clone had promoter methylation levels of 49%, compared to 7% in the BF-23 parent cell line prior to selection. This increase in promoter methylation levels to modest levels in both cases is compatible with the intermediate-expression-level hypothesis proposed above. We were unable to perform TK expression analysis of these clones to confirm this hypothesis, since they had not been cryopreserved for RNA analysis.

As expected, the absolute rate of CpG mutagenesis is reduced in Dnmt1-deficient cells. However, the relative fraction of CpG transition mutations is actually increased in Dnmt1-deficient cells. This finding is unexpected, although an elevated mutation rate of CpG dinucleotides, independent of cytosine methylation, has been proposed previously (64). This result could be a statistical aberration, given the limited number of sequenced mutations. The modest effect of Dnmt1 deficiency on the relative frequency of CpG mutagenesis may be partly due to the low levels of DNA methylation observed in the TK coding region, even in Dnmt1^+/PH cells (Fig. 5D). The Dnmt1 protein could play an alternative role in protecting against cytosine deamination at CpG dinucleotides. However, this would not explain why the incidence of transition mutations in Dnmt1-deficient cells is higher at CpG dinucleotides than at other CG base pairs. Another possibility may be enzyme-mediated deamination by Dnmt3a, Dnmt3b, or Dnmt2. Enymatic deamination has been observed with the bacterial methyltransferases HpaII and EcoRII, which are able to deaminate cytosine to uracil directly when the supply of the methyl donor SAM is limiting (72, 73). The increase in missense mutation rate observed in Dnmt1^P/H cells under methyl group-deficient conditions is consistent with an involvement of Dnmt3a, Dnmt3b, or Dnmt2 in enzyme-mediated deamination. If this is indeed the case, then the same events may also occur in the Dnmt1^+/PH cells, but the effect may be masked by the higher rate of other mechanisms of gene inactivation in these cells. Alternatively, the Dnmt1 protein or the resulting methylation may actually protect against the proposed enzyme-mediated mutagenesis. It should be noted that the missense mutation assays were conducted in a wild-type background of uracil and thymine DNA glycosylase genes (29, 30, 58, 70). This would severely limit our ability to detect enzyme-mediated and spontaneous deamination events. Elucidation of the roles of the different DNA methyltransferases in the stimulation of or protection from cytosine deamination extends beyond the scope of this article and will require much further investigation before any firm conclusions can be drawn.

The role of DNA methylation in gene inactivation has been previously addressed by Chen et al., using Dnmt1 gene targeting in ES cells with a slightly different strategy of HPRT and HSV-TK counterselection (12). Chen et al. conclude that DNA hypomethylation results in an increased rate of rearrangements and gene loss by mitotic recombination (12). This observation is consistent with other studies showing an increase in chromosome 1 pericentromeric rearrangements in DNMT3B-deficient cells (28, 59, 83) and in cells treated with the DNA methyltransferase inhibitor 5-aza-deoxycytidine (36). On the other hand, the conclusion by Chen et al. (12) that Dnmt1 deficiency increases the rate of LOH is not consistent with the observation that reduced levels of functional Dnmt1 strongly suppress intestinal polyp formation in Apc^+/Min mice (18, 43), a system that requires LOH of the wild-type Apc locus (48, 52). In this study, we followed a slightly different strategy to measure rates of gene inactivation and arrived at conclusions opposite to those of Chen et al. (12). This may in part be explained by the different experimental setups of the two studies. Chen et al. used two defined loci for the genomic location of their counterselectable markers: the X-linked HPRT gene and a TK gene targeted to the Dnmt1 locus. Our system relies on random integrations of single-copy or multicopy concatemers of TKNeo. Therefore, some of the differences between the two studies may reflect differences between chromosomal locations, which may differ in their relative utilization of various gene inactivation pathways. For instance, the Hprt locus used by Chen et al. cannot sustain very large deletions or chromosomal loss in male ES cells, since it is X linked (12). Chen et al. reported a high frequency of counterselected clones (71% of the Dnmt1^+/ clones) that did not show evidence of TK gene loss and for which the mechanism of gene inactivation was unclear (12). Our results suggest that some of these clones may have acquired promoter methylation, with an accompanying transcriptional silencing. The study by Chen et al. employed several different ES cell lines, including two independently derived ES cell lines, several derivative daughter cell lines, and a single pair of sibling cell lines (12). In contrast, we performed only simultaneous comparisons between eight paired sibling ES cell clones. Therefore, while the diversity of source material used by Chen et al. provides justification for the generalization of their results, the strength of our study lies in the fact that we used very closely matched, multiple, pairwise measurements, which provide a more solid basis for statistical comparisons. As a consequence, we could perform paired t tests on 18 separate rate measurements, whereas the Chen study used a ² test on four rate measurements (12). In addition, in our rate calculations, we have used the Luria-Delbrück method of means equation, which allows for the correction of plating efficiency (40, 53). This adjustment is particularly important for ES cells, which have a reduced efficiency of plating in the absence of feeder cells. In contrast, the P₀ method of Luria-Delbrück used by Chen et al. does not provide a method to correct for the plating efficiency (40, 53). Nevertheless, Chen et al. did adjust their final calculated rates for plating efficiency (12). The problems associated with the application of fluctuation analysis to mammalian cell systems have been widely noted (3, 40). The method of means may overestimate the mutation rate, while the P₀ method may underestimate it (40). Indeed, we arrive at substantially higher rates of gene inactivation than those of Chen et al. (12). This underscores the difficulty of fluctuation analysis and the necessity to limit conclusions to side-by-side comparisons within a single study, as we did. It is not clear which, if any, of the differences outlined above can account for the opposite results obtained in these two studies. Therefore, this issue will need further investigation, before definitive and broad statements regarding the quantitative effects of DNA methylation on gene inactivation can be reached. A reduced rate of gene inactivation under Dnmt1-deficient conditions would be consistent with the observation that Apc^Min/+ mice with reduced levels of functional Dnmt1 expression have a substantially reduced polyp multiplicity (18, 43). LOH of the wild-type Apc allele is a frequent event in Apc^Min/+ polyps and may be a rate-limiting step in this model system (48, 52). Our data are also consistent with other previous reports of oncogenic effects of increased Dnmt1 expression (5, 75, 82) and tumor-suppressive effects of decreased Dnmt1 expression (54, 66).