Introduction

Single base substitutions in DNA typically occur by misincorporation of nucleotides during DNA synthesis. Single base substitutions left behind by the replication complex can be corrected by the mismatch repair (MMR) system1. MMR is an evolutionarily conserved mechanism which can function in a relatively short time interval after replication. In case of the E. coli MMR system, mismatches are recognized by the MutS protein. After mismatch recognition, the system scans the DNA to find a signal for discrimination of the newly synthesized strand and the template strand. In the widely accepted model, the MMR system uses methylation of adenines at 5′-GATC-3′ sequences as discrimination signals, which is performed by the DNA adenine methylase (Dam) after replication2. The MutS protein together with MutL activates the MutH endonuclease, which creates a nick on the unmethylated strand at a hemimethylated GATC located nearby the mismatch. The misincorporated nucleotide is corrected by excision and resynthesis of the DNA strand nicked by MutH. This process requires DNA helicase II (UvrD), single strand DNA binding protein (SSB) and DNA polymerase III1. The steps between mismatch recognition and MutH mediated strand incision are much debated1,3. Communication between the mismatch site and the GATC site may involve translocation of MutS along the DNA (cis model) or DNA loop formation between the two sites (trans model).

In the cis model the time frame available for proper action of the MMR system depends on how fast the daughter strand is methylated, on the distance of the discrimination signal from the mismatch and also on the distance on DNA that the MutS protein needs to scan for locating the discrimination signal.

The reported half-life values for hemimethylated DNA behind the replication fork vary from seconds to several minutes depending on the experimental system used and also on the action of specific proteins that can hinder methylation of certain DNA regions4,5. Assuming that the migration speed of the replication fork is about 1000 bp/s6, there may be a few thousand to several hundred thousand bases of hemimethylated DNA available for the MMR system.

Previous studies demonstrated that GATC sequences affect the repair efficiency of G-T mismatches on artificial bacteriophage heteroduplexes in E. coli in a number and distance dependent manner. A single hemimethylated GATC site could serve as a strand discrimination signal for the MMR system as long as it's distance from the mismatch was less than 1 kb7,8. The MMR system was not sensitive to the relative orientation of the mismatch to the hemimethylated GATC site, i.e. it could efficiently repair the error on the unmethylated strand from both directions7,9. However, on the E. coli chromosome single nucleotide deletion mismatches could be efficiently repaired even if the closest GATC sequence was 2 kb away and the chromosomal context had a larger influence on the frameshift mutation rate than the local GATC content10.

In this work we investigate the inconsistency of the above results by studying the effect of local chromosomal GATC content on the rate of single base substitutions (SBS). These are recognized by MutS typically less efficiently than single nucleotide deletion mismatches11 and therefore their repair may be more sensitive to the distance to the nearest GATC site. Our results suggest that the strands can be discriminated in the absence of GATC methylation, similar to other organisms where the MMR system is not methyl directed12.

Results

The effect of the distance between the mismatch and GATC sites on MMR activity on the chromosome

The expected average distance between GATC sites on DNA is 256 bp but the actual distances on the E. coli chromosome vary from 4 to 4840 bp. Here we asked whether the occurrence of single base substitutions depend on the GATC context on the E. coli chromosome. To address this question we created a chromosomal reporter system in E. coli (Figure 1) using a mutant β-galactosidase (lacZ) gene which has an early TAA stop codon (codon 7). The TAA stop codon serves as the most abundant stop codon in E. coli (~60%).

Figure 1
figure 1

Structure and chromosomal context of the reporter constructs used in this study.

The reporter constructs (A–E) contained the zeocin resistance cassette and the lacZ gene which was inactivated by the C20A substitution resulting in a stop codon (red line). The positions of GATC sequences in the different constructs are indicated by vertical lines. Arrowheads indicate the direction of transcription. The replication fork proceeds from left to right in this region. The local sequence context of the stop codon is shown on the top. Measured mutation rates (M) and 95% confidence intervals (95% CI) are shown on the right. The mutation rates and 95% confidence intervals observed upon mutS deletion were 3.1 (2.2–4.1), 13.3 (11.2–15.6), 3.8 (2.8–4.9) and 13.6 (11.1–16.3) × 10−9/generation for strains ‘A’–‘D’, respectively.

Full size image

First we created four versions of the lacZ gene, which differed in their GATC content (Figure 1, A–D). Constructs ‘A’ and ‘B’ carry the wild type lacZ gene containing 14 GATC sites. In constructs ‘C’ and ‘D’ all these sites are eliminated by same-sense mutations. Constructs ‘A’ and ‘C’ contain a 12 bp insertion right upstream of the lacZ gene. This 12 bp sequence carries a GATC site.

Because appearance and repair of mismatches may depend on their local DNA context and chromosomal location10,11,13, all constructs were placed at the same chromosomal location. Mutation rates were calculated from the occurrence of single base substitutions in this stop codon which restored a functional lacZ gene (Table 1). We sequenced such functional lacZ genes in 25 colonies which grew on a minimal lactose plate and found mutations in all three positions of the stop codon (Table 2).

Table 1 The effect of mutS, mutL and dam deletions on the reversion rate of the stop codon in constructs C and D. 95% confidence intervals are shown in parentheses, while fold changes relative to the corresponding wild type constructs are underlined. Reversion rates observed in the mutL deletion strains reflect that at the last two positions of the stop codon reversion can arise only by transversions, which are less enriched than transitions in the absence of MutL22
Full size table
Table 2 Possible point mutations at the TAA stop codon and their occurrences in 25 revertants
Full size table

In construct ‘B’ the lacZ gene contained the same GATC sites as found in the wild type lacZ gene, with the closest one being about 115 bp from the stop codon. In construct ‘D’, in which all GATC sites were eliminated, the closest GATC site was located upstream in the ybbP gene 2433 bp from the stop codon (the closest downstream GATC site was 5566 bp away). Importantly, we observed the same rate of reversion of the stop codon in ‘D’ as in the wild type construct (‘B’). These results suggest that within these limits (115 to 2433 bp), the distance between the mismatch and the closest GATC site does not affect MMR efficiency on the chromosome.

If MMR fully depends on the availability of hemimethylated GATC sites, then either the MMR system acts faster or hemimethylated GATC sites are available for a longer time on the chromosome than on bacteriophage heteroduplexes. Hemimethylated GATC sites are typically available for about 1–2 minutes on the chromosome after replication4. The time required for recognition of mismatches by MutS is probably not a limiting factor because MutS is associated with the replication complex14.

ATP-bound MutS can diffuse along naked DNA at 0.1 μm2/s in vitro and spends about 10 minutes on the DNA having closed boundaries15. Therefore, it could find a site located ~2.5 kb away in about 10 seconds. However, diffusion of MutS along the E. coli chromosome in vivo is most likely obstructed by other DNA binding proteins16. For example, about one HU dimer is present per 100 bp on the chromosome on average and slow dissociation of HU dimers from DNA17 would be a substantial barrier for MutS diffusion to longer distances. That is, prokaryotic MutS proteins face a similar difficulty in reaching a distant site on DNA as eukaryotic MutS homologues do due to the presence of nucleosomes18. Nucleosomes dissociate from DNA at a comparable rate to dissociation of HU from the E.coli chromosome. The eukaryotic mismatch recognition heterodimer hMSH2-hMSH6 is able to facilitate the disassembly of nucleosomes, however, the process requires a relatively long time (t½ of 23 to 117 minutes, depending on the modification of the nucleosome)19. The E. coli chromosome is covered by DNA binding proteins of diverse nature. Therefore, it is unlikely that MutS could facilitate their dissociation universally.

DNA bound proteins obstruct MutS diffusion depending on their dissociation rate from DNA. We have simulated the potential effect of slower MutS diffusion rate on MMR efficiency (Figure 2). Using different MutS diffusion rates, we computed the probability that MutS reaches a site located 2400 bp away in any of the directions within 90 seconds (typical lifetime of a hemimethylated GATC site). We found that already at a 5-fold slower diffusion rate MutS would not reach the distant GATC site in 10% of the cases, which would result in a detectable increase in the observed mutation rate. For example, assuming that only 1% of mismatches are left uncorrected by the MMR system in the close vicinity of GATC sites, the above 10% increase in the uncorrected errors would result in an about 10-fold increase in the observed mutation rate.

Figure 2
figure 2

Efficiency of mismatch repair as function of MutS diffusion rate along the DNA.

At each value of the diffusion rate, MutS located at the mismatch is released to perform one-dimensional random walks along the DNA. The simulation was repeated 10000 times. The efficiency of the mismatch repair is scored as the fraction of the released MutS that reach either a site 2400 bp downstream of the mismatch or a site 5550 bp upstream of the mismatch within 90 seconds. These 90 seconds correspond to the average lifetime of a hemi-methylated GATC. The shown behavior was reproduced (within a factor 2 in diffusion constant) in a more elaborate model where many independently methylated GATC sites (methylated in 90 seconds on average) are placed at 256 bp intervals outside the −2400 to 5550 bp region. In that more complicated model the repair efficiency was scored as the probability that MutS reached any of these sites in a hemimethylated state.

Full size image

The reversion rate of the stop codon is affected by a GATC site located 50 base pairs away

Constructs ‘A’ and ‘C’ contain a 12 bp GATC containing sequence about 50 bp upstream of the lacZ gene. Construct ‘A’ had a lacZ gene with the wild type GATC pattern, while in ‘C’ all GATC sites were eliminated from lacZ, corresponding to ‘B’ and ‘D’, respectively. We found similar mutation rates of the stop codon in case of ‘A’ and ‘C’, however, these rates were about 4-fold lower than the rates observed in case of constructs ‘B’ and ‘D’.

To determine whether the observed 4-fold difference in the mutation rate is specific to the stop codon used in our reporter system, we compared the single nucleotide substitution rate of the endogenous rpoB gene in strains A–D. We found that mutations in the rpoB gene causing rifampicin resistance appeared with similar probabilities in all the four strains (~3.5 × 10−8/generation). The above results suggest that in strains ‘A’ and ‘C’ the 12 bp sequence provides a short range protection against the occurrence of single nucleotide substitutions.

To test whether the GATC site is responsible for this protection, we created two sequence variants of construct ‘C’. In one of the variants the GATC sequence was replaced by GTTC and in the other one it was moved three base pairs closer to the stop codon (Table 3). In the absence of the GATC site the mutation rate increased to a similar level as was observed in construct ‘D’, while shifting the GATC site did not change the mutation rate.

Table 3 Mutation rates of sequence variants of construct ‘C’. The 12 bp sequence insertion (bold-faced) and its sequence context in construct ‘C’ is shown on the top. Sequence changes in the two sequence variants are marked red. Measured mutation rates (M) and 95% confidence intervals (95% CI) are shown on the right
Full size table

Comparison of constructs ‘E’ and ‘F’, in which the lacZ gene was placed in a reverse orientation compared to the other constructs, showed that the protection provided by the 12 bp insertion is independent of the direction of replication (Figure 1). However, we observed about 50% increase in the mutation rates in both of these ‘reverse’ construct (‘E’ and ‘F’) compared to the corresponding ‘direct’ constructs (‘B’ and ‘C’, respectively). This difference is most likely due to the unequal fidelity of leading strand and lagging strand synthesis20.

Effect of mutS and mutL deletions on the reversion rate of the stop codon

To test whether the protection provided by the 12 bp insertion resulted from an increased MMR efficiency, we created mutS deletion derivatives of the four strains carrying the constructs placed in the ‘direct’ orientation (Figure 1, A–D). We observed about 20-fold increase in the reversion rate of the stop codon in all the four cases as a result of mutS deletion (see Figure 1 legend and Table 1). Similar results were obtained when we compared mutL deletion derivatives of strains carrying constructs ‘C’ and ‘D’ (Table 1).

Mismatch repair is directed in the absence of Dam methylation

The 12-bp insertion present in constructs ‘A’ and ‘C’ contains a GATC site, which can be methylated by the Dam methylase and which can serve as a strand discrimination site. To test the role of GATC methylation in mismatch repair occurring at the reporter stop codon, we created dam deletion derivatives of cells carrying constructs ‘C’, which has a single GATC sequence in the reporter region and ‘D’, which has none. Elimination of GATC methylation resulted in a relatively small increase (~4-fold) in the reversion rate of the stop codon compared to the 20-fold increase observed in the cases of mutS and mutL deletions (Table 1).

Discussion

In this work we studied the effect of local chromosomal GATC content on the rate of single base substitutions (SBS). We found that within the limits of 115 to 2433 bp, the distance between the mismatch and the closest GATC site does not affect MMR efficiency on the chromosome. This observation is in agreement with the findings of Martina et al., who reported that frameshift mutation rate on the E. coli chromosome is independent of the distance to GATC sites located about 200 to 2000 bp away10. However, we found that a GATC site located about 50 base pairs from the stop codon could provide a short range protection from single base substitutions. This protection was independent of Dam methylation, MutS and MutL. Although we do not yet understand the mechanism underlying this observation, our results suggest that the GATC content on the chromosome may influence the mutation rate at different locations. Such regulation could become important under conditions where MutS is depleted and therefore the point mutation rate is higher21. However, the potential protective function of GATC sites is counteracted in E. coli because methylated bases are mutational hotspots on the chromosome22.

In agreement with previous reports23,24,25, we observed that dam mutants have weaker mutator phenotype than mutL and mutS mutants. If GATC methylation is the only strand discrimination signal for MMR, then the dam mutation is expected to have the same effect on the mutation rate as mutS or mutL mutations. There are two possible explanations for this discrepancy. One is that some of the dam mutant cells are lost due to cell death and the other is that the MMR system is able to correct the majority of replication errors in the absence of Dam methylation26,27.

Mismatch repair creates double strand breaks in dam mutant cells due to MutH mediated cleavage of both unmethylated strands at GATC sequences located nearby mismatches24,27, which may result in cell death. To account for the lower mutation frequency of Δdam cells compared to ΔmutS cells, about 4 out of 5 mutants (~80%) must be lost due to the above process. However, the misincorporation rate at replication is roughly one per replicated genome28. Therefore, loss of 80% of mismatches in the dam mutant would result in a substantial increase in the population doubling time, which was not observed29,30. Also, double strand breaks are most likely repaired efficiently and do not persist for a long time in dam mutants31.

Although previous experiments showed that the old and new strands can be discriminated by the MMR system based on their methylation status, this does not mean that it is the only signal used to discriminate the strands. Claverys and Méjean demonstrated that in a GATC free plasmid system, 50–70% of replication errors occurring at a TAG stop codon can be corrected in a mutSL dependent way26. Our experimental results and theoretical simulations support this model, i.e. that the MMR system can correct replication errors in the absence of Dam methylation. In our system, we observed only about 4-fold higher reversion rate at the stop codon in the dam mutants compared to wild type cells, while mutS deletion gave a 20-fold increase in reversion. The effect of dam deletion could be explained in part by the random replication initiation process in the dam mutants29. Because of the random initiation, replication forks can follow each other at a shorter interval and thus uncorrected errors can be copied in a fraction of cells. This could result in an increase in the observed mutation rate and would further decrease the contribution of GATC methylation to MMR.

The major difference between the bacteriophage heteroduplex and the chromosomal or plasmid studies is that in the latter cases mismatches are generated by the replication machinery. Because single-strand breaks can direct repair in the absence of MutH32, MMR may function independently of GATC methylation on the lagging strand, which is synthesized in fragments (using the single strand breaks as signals).

However, at this point we can only speculate how mistakes could be corrected on the leading strand independently of GATC methylation. One possibility is that MutS binding to the mismatch is oriented by contacting the sliding clamp of the replication complex14 and this positional information propagates to the nearest GATC site to signal for MutH to cleave the newly synthesized strand, regardless of the methylation status of the template strand.

Methods

Plasmid and strain construction

The kanamycin resistance gene from plasmid pRFB11033 was PCR amplified using primers 5′-GGGTAACGCCAGGGTTTTCCCAGTCACGACGTTGTAAAAC GACGGCCAGTGAATCAGAGTCCCGCTCAGAAGAACTCGTC-3′ and 5′-ACATAATGGA TTTCCTTACGCGAAATACGGGCAGACATGGCCTGCCCGGTTATTATTATTAGAGTCCCGCTCAGAAGAACTCGTC-3′ and inserted between chromosomal positions 362459 and 365560 in E. coli MG1655 (GenBank NC_000913.2) by recombineering34 (to replace the region containing the lacZ gene and the lac O1 operator site). In this strain (SEM3106) the kanamycin resistance gene and the lacYA genes are transcribed constitutively from the lac promoter.

The reporter construct was first assembled in plasmid pEM7/zeo (Invitrogen). We have made a C->T substitution at position 2299 to eliminate the GATC site from the zeocin gene. The different versions of the lacZ gene, differing in their GATC content, were placed downstream of the PEM7 promoter and the zeocin resistance gene as shown in Figure 1. All versions contained the C20A substitution in lacZ resulting in a stop codon at codon 7. The GATC-free versions of lacZ were synthesized by GeneArt (Invitrogen). All GATC sites were eliminated by silent substitutions. The sequences containing the zeocin gene and lacZ were inserted into the rhsD gene on the E. coli chromosome by recombineering34, between positions 523237 and 523641. The sequences of the chromosomal constructs and their ~300 bp flanking regions (see Supplementary Material) were verified (Eurofins MWG Operon).

In the ΔmutS strains the mutS gene was replaced by a chloramphenicol resistance gene (CmR). The CmR cassette in plasmid pRFB12233 was PCR amplified using primers 5′-ATGAGTGCAATAGAAAATTTCGACGCCCATACGCCCATGATGCAGCAGTA TCTCAGGCTGAAAGCCCAGCATCCCGTGCCGTTACGCACCACCCCGTC-3′ and 5′-TTA CACCAGGCTCTTCAAGCGATAAATCCACTCCAGCGCCTGACGCGGGGTGAGTGAATCCGGATCAAGATTTAATTACGCCCCGCCCTGCCACTC-3′ and inserted between positions 2855190 and 2857604 to replace the MutS coding sequence on the chromosome. The ΔmutL and Δdam strains were created in a similar way, inserting the same CmR cassette between positions 4395504 and 4397215 (ΔmutL) and 3513866 and 3513165 (Δdam). Deletions were confirmed by PCR and elimination of GATC methylation was further confirmed by digestion of genomic DNA extracts by Ksp22I (SibEnzyme), which is inhibited by GATC methylation.

Determination of mutation rates

Mutation rates and 95% confidence intervals were determined by fluctuation analysis35. Strains were analyzed using up to 30 independent cultures. Each parallel culture was started by inoculating 3 ml LB medium with 2 × 106 cells. The medium contained 30 μg/ml kanamycin. Cultures were incubated in a shaking incubator overnight at 37°C. 1 ml cell suspension (3 × 109 cells) were plated on M63 agar plates containing 0.4% (w/v) lactose, 20 μg/ml X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside), 30 μg/ml kanamycin and 0.004% (w/v) vitamin B1. Plates were incubated at 37°C for 3 days and blue colonies were counted. The occurrence of rifampicin resistant cells was determined by plating 10 μl of cell suspension on LB plates containing 30 μg/ml kanamycin and 30 μg/ml rifampicin and counting the colonies which appeared after incubation of plates overnight at 37°C. Mutation rates were determined from the distribution of the number of mutants in the cultures by the MSS-Maximum Likelihood Estimator Method, using the FALCOR web tool36.

Model of MutS diffusion along the DNA

To simulate the mismatch repair efficiency through MutS mediated mismatch localization, we constructed a model where MutS is released from the mismatch site at time zero and moves along the DNA by random diffusion. We assume that MutS locates the mismatch immediately after the replication fork has left the site. For a given assumed diffusion constant D, the MutS molecule subsequently step a distance l = 15 nm right or left for each time step dt = l2/2D. The nearest GATC upstream is 2433 bp and the nearest GATC downstream is 5566 bp. This correspond to position −55 and +126 in units of the lattice spacing of l = 15 nm. The GATC site is reached if and only if the released MutS passes either of these two positions within the time frame of 90 seconds. The 90 seconds mimics the average lifetime of hemi-methylated GATC sites on the DNA. For each value of the diffusion constant D, 1000 MutS releases were examined. Figure 2 shows the probabilities of reaching a GATC site calculated from the simulations.

Importantly, we also examined an extended model that includes all GATC sites beyond the −2433 bp and the +5566 bp positions and in addition, also includes the fact that each hemi-methylated site becomes fully methylated with a rate of 1/90 seconds. The resulting curve for probability to reach a GATC is similar to the one shown in Figure 2, apart from being displaced by about a factor 2 to the left (GATC localization can be done with half of the shown diffusion constant).