JBC PeproTech; Our Business is Cytokines!

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bandaru, B.
Right arrow Articles by Bhagwat, A. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bandaru, B.
Right arrow Articles by Bhagwat, A. S.
Social Bookmarking
Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?

Volume 271, Number 13, Issue of March 29, 1996 pp. 7851-7859
©1996 by The American Society for Biochemistry and Molecular Biology, Inc.
Overproduction of DNA Cytosine Methyltransferases Causes Methylation and C T Mutations at Non-canonical Sites (*)

(Received for publication, September 25, 1995; and in revised form, January 3, 1996)

B. Bandaru ,  Jaishree Gopal ,  Ashok S. Bhagwat (§)

From the Department of Chemistry, Wayne State University, Detroit, Michigan 48202

ABSTRACT
INTRODUCTION
EXPERIMENTAL PROCEDURES
RESULTS
DISCUSSION
FOOTNOTES
ACKNOWLEDGEMENTS
REFERENCES

ABSTRACT

Multicopy clones of Escherichia coli cytosine methyltransferases Dcm and EcoRII methylase (M. EcoRII) cause 50-fold increase in C T mutations at their canonical site of methylation, 5`-CmeCAGG (meC is 5-methylcytosine). These plasmids also cause transition mutations at the second cytosine in the sequences CCGGG at 10-fold lower frequency. Similarly, M. HpaII was found to cause a significant increase in C T mutations at a CCAG site, in addition to causing mutations at its canonical site of methylation, CCGG. Using a plasmid that substantially overproduces M. EcoRII, in vivo methylation at CCSGG (S is C or G) and other non-canonical sites could be detected using a gel electrophoretic assay. There is a direct correlation between the level of M. EcoRII activity in cells, the extent of methylation at non-canonical sites and frequency of mutations at these same sites. Overproduction of M. EcoRII in cells also causes degradation of DNA and induction of the SOS response. In vitro, M. EcoRII methylates an oligonucleotide duplex containing a CCGGG site at a slow rate, suggesting that overproduction of the enzyme is essential for significant amounts of such methylation to occur. Together these results show that cytosine methyltransferases occasionally methylate cellular DNA at non-canonical sites and suggest that in E. coli, methylation-specific restriction systems and sequence specificity of the DNA mismatch correction systems may have evolved to accommodate this fact. These results also suggest that mutational effects of cytosine methyltransferases may be much broader than previously imagined.

INTRODUCTION

There is good correlation between the presence of methylation at position 5 of cytosines in DNA and transition mutations. Several years ago Coulondre et al.(1) showed that two cytosine methylation sites in the lacI gene of Escherichia coli were hot spots for spontaneous C T mutations. These cytosines mutated to thymine at frequencies many times higher than the frequencies of mutations at any other base in the gene(1) . More recently, a cytosine methylation site in the cI gene of a phage lysogen (2) was also shown to be a hot spot for spontaneous C T mutations. In vertebrates, methylation of cytosines predominantly occurs within CpG dinucleotides, and cataloging of sequence changes that cause human genetic diseases has revealed that a disproportionately high fraction of these involve transition mutations at CpG sites(3) . There is also a striking correlation between some types of cancers and mutations at CpG dinucleotides in the tumor suppressor gene p53 (reviewed in (4, 5, 6) ). In addition, 53% of all the germline mutations found in Li-Fraumeni syndrome are C:G to T:A mutations within CpGs(5) .

Spontaneous hydrolytic deamination of 5-methylcytosine (5-meC) (^1)in DNA to thymine (7, 8) has traditionally been proposed (1) as the explanation of this phenomenon. Recently several alternate hypotheses for the occurrence of such mutational hot spots have been proposed and studied. These include error-prone copying of 5-meC by DNA polymerases(9) , cytosine methyltransferase (C5 MTase)-mediated C U (10) and 5-meC T (11) conversions, excision of 5-meC in DNA followed by error-prone repair (12) , and inhibition of mismatch correction systems by C5 MTases(13, 14) .

To help choose between these alternative hypotheses, we have developed two genetic systems in E. coli that can quantitate C T mutations at sites of cytosine methylation. The assay involves scoring of kanamycin-resistant (Kan^R) revertants from kanamycin-sensitive (Kan^S) alleles in which Leu (TTG) at codon 94 of the kan gene was replaced with Pro (CCG or CCA) mutations. Replacement of the second C in the codon with T restores a different Leu codon and is scored as Kan^R ( Fig. 1and Refs. 15 and 16). The two Kan^S alleles will be referred to as kanS-H94 (codon 94; CCG) and kanS-D94 (CCA), respectively. While the former system detects C T mutations within CpG sequences, the latter detects mutations within the sequence context of cytosine methylation in E. coli. EcoRII methyltransferase (M. EcoRII) is part of a plasmid-borne restriction-modification system found in a clinical E. coli isolate(17, 18) . M. EcoRII methylates position 5 of the second cytosine within the sequence 5`-CCWGG-3` (W is A or T; Refs. 19 and 20). The chromosome of E. coli K-12 also codes for a C5 MTase called Dcm(21) , and Dcm and M. EcoRII have identical methylation specificities(22, 23) . As a result, both Dcm and M. EcoRII methylate within codon 94 of kanS-D94 (Fig. 1).

Figure 1: Wild-type, mutant, and revertant kan sequences. DNA sequences surrounding codons 70 and 94 are shown. For codon 94, the wild-type sequence is shown in the middle, and the two mutant alleles containing CCA and CCG sequences and their revertants are, respectively, shown above and below the wild-type sequence. The Dcm/EcoRII site and HpaII sites in the sequences are underlined. In the case of codon 70, the pentanucleotide sequence within which Dcm is expected to methylate is underlined.


Presence of a cognate MTase in cells containing one of the kan alleles results in a 40-100-fold increase in the Kan^S to Kan^R reversion frequency(15, 16) . These studies provide direct evidence that an action of C5 MTases, presumably methylation itself, is the cause of mutational hot spots at sites of cytosine methylation. While Dcm or M. EcoRII were used as the cognate MTases with kanS-D94 in these studies, the methyltransferase in the HpaII restriction-modification system (M. HpaII) was used with kanS-H94. M. HpaII methylates the second cytosine within the sequence CCGG(24) . We have used restriction mapping and DNA sequencing to confirm that the Kan^R revertants obtained during these studies contain the expected C T mutation at codon 94(15, 16) . This shows that the occurrence of second-site revertants is low in the system and that the system is well suited to study the effect of methylation within specific sequences on C T mutations.

An unexpected finding of these experiments was that the presence of non-cognate methylases in the cells also increased Kan^R reversion frequency, although to a lesser extent. We describe below this phenomenon, discover its cause, and explore its implications to the structure and biology of C5 MTases.

EXPERIMENTAL PROCEDURES

Bacterial Strains, Plasmids, and Oligonucleotides

E. coli strains DH10B (Delta(mrr-hsdRMS-mcrBC) mcrA 80dlacZDeltaM15 DeltalacX74 deoR recA1 endA1 araD139 Delta(ara, leu)7697 galU galK rpsL nupG), JH140 (HfrC DeltalacZ dinD1::MudI1734 (Kan^Rlac) phoA(Am) tonA22 garB10 ompF627 relA1 pit-10 spoT1 T2^R), BH143 (= DH10B recA Delta(dcm-vsr) hisG::Tn10) and RP4182 (Delta(dcm-vsr) trp rpsL) have been described before(25, 26, 27) . Plasmids carrying kanS-D94 and kanS-H94 have also been described before(15, 16) . Plasmid pR300 is pR300(+)Cys(28) , and CCAG-pSV2-neo(13) , pDCM72 (16) , pSS55(27) , and pHpa23 (15) have been described before. PERM1 was constructed by cloning a HindIII fragment from pMB1 (29) into pACYC184. It contains the ecoRIm gene and a part of the ecoRIr gene. To construct pR400, M13 phage derivative BM1909 containing the wild-type ecoRIIm gene (28) was cloned into pUC119 as a PstI-EcoRI fragment. For reasons not related to this work, several nucleotides in this gene were changed to introduce an SpeI site (ACTAGT) and an AatII site (GACGTC) in the gene without changing the coded amino acids. (^2)The changes in the sequence were confirmed by DNA sequencing. PR234 (^3)was constructed by cloning a HindIII fragment containing ecoRIIm gene from pSS114 (30) into the P promoter vector pKK223-3(31) . The oligonucleotides 5`-GGCACTTCGCCTGGTAGCAGCCAGTC and 5`-GGCACTTCGCCCGGGAGCAGCCAGTCC, and their respective complements were synthesized by Genset Corp. (La Jolla, CA). They were purified using NENsorb Prep columns (DuPont NEN).

Methyltransferase Assays

To determine MTase activity in cells, 5-ml cultures were grown of BH143 containing the appropriate plasmids. Cells were concentrated by centrifugation and suspended in 1 ml of sonication buffer (10 mM Tris-HCl, pH 7.8, 10 mM 2-mercaptoethanol). EDTA was added to a concentration of 1 mM, and lysozyme was added to 1 mg/ml. After 15 min on ice, the cells were freeze-thawed and then sonicated. Cell lysate was centrifuged in a microcentrifuge for 20 min, and supernatant was suspended in 50% glycerol. Five-µl aliquots of this preparation were used in a 50-µl reaction mixture. Methyltransferase activity was measured as transfer of methyl groups from S-[methyl-^3H]adenosyl-L-methionine (DuPont NEN) to chromosomal DNA from the strain RP4182. Unincorporated SAM was removed from the reaction using Sepharose CL-6B (Pharmacia Biotech Inc.) spin columns. The concentration of the protein in the extracts was determined using the protein assay kit from Bio-Rad, and the enzyme activity was calculated as the initial velocity of the reaction per milligram of extract used.

Duplexes I and II were methylated with M. EcoRII purified to apparent homogeneity (specific activity 9.8 pmol of methyl groups/min/µg of protein) using S-[methyl-^3H]adenosyl-L-methionine (DuPont NEN) (0.078 µM, 85 Ci/mmol) as the methyl donor in methylase buffer (100 mM Tris-HCl, pH 7.8, 20 mM EDTA, pH 8.0, 0.4 mM dithiothreitol). The reaction volume was 50 µl, and the reactions were carried out at 37 °C and terminated at various times by the addition of 2 µl of 10% SDS. The samples were purified by extraction with phenol-chloroform and passed through Sephadex G-50 (Pharmacia Biotech Inc.) spin columns to remove unincorporated radioactive label. The incorporated radioactivity was quantitated by scintillation counting. When duplex I was used, it was at concentrations between 10 and 200 nM and the enzyme was at 2.1 nM. When duplex II was used, it was at concentrations between 1.0 and 20.0 µM and the enzyme was at 0.454 µM. Steady-state kinetics of methyl transfer was analyzed using the Statview Student package for the Macintosh, and the kinetic constants were calculated.

C T Mutation Assay

The procedure for the reversion assay has been described before(15, 16) . Briefly, three independent coloniesa were picked from each transformation and 5-ml cultures were grown to logarithmic phase (A 0.3) in LB medium containing appropriate antibiotics. The culture was then concentrated and resuspended in 1 ml of LB broth and plated on LB plates without antibiotic or with kanamycin (50 µg/ml) to detect the reversion.

beta-Galactosidase Assay

The assay was carried out according to the procedure outlined by Miller(32) . A 50-ml culture of JH140 containing pNK627 and pR234 was grown from a single colony in LB medium containing ampicillin (100 µg/ml) and tetracycline (10 µg/ml) at 37 °C. Starting at A = 0.03, the turbidity of the culture and the beta-galactosidase activity was monitored. At A = 0.065, the culture was divided in two halves and IPTG was added to one culture to a concentration of 100 µM. At 30-min intervals, 5-ml samples of the culture were removed to measure optical density and a part of this culture was used to determine beta-galactosidase activity. Samples (0.5 ml) were used for the enzyme assay at the low cell densities, and progressively smaller volumes (to 50 µl) were used as the turbidity of the culture increased. For each assay, the sample was diluted to 1 ml with Z buffer and cells lysed with chloroform and SDS as described by Miller(32) . 2-Nitrophenyl-beta-D-galactopyranoside (0.2 ml of 4 mg/ml) was added to the lysates to start the reaction. The reaction tubes were maintained at 28 °C and monitored for development of color and stopped by the addition of 0.5 ml of 1 M Na(2)CO(3) when a light yellow color was seen. The solutions were subjected to a clearing spin, and A was measured. The units of beta-galactosidase activity are defined as (1000 times A)/(t times v times A), where t is the time of reaction in minutes and v is the volume of the sample used for the assay in milliliters.

RESULTS

Dcm and M. EcoRII Cause C T Mutations within a CCGGG Site

When E. coli containing kanS-D94 allele was transformed by a plasmid that codes for Dcm (pDCM72), Kan^S to Kan^R reversion frequency increased by a factor of about 50 (Table 1). This increase is somewhat lower than the increase reported previously (100-fold; (16) ). The differences in the two sets of numbers may be due to the use of different E. coli strains in the two experiments. Unexpectedly, pDCM72 also caused a significant increase in reversion frequency with the kanS-H94 allele (Table 1). Although the extent of increase in revertants in this case was substantially smaller, it was reproducible. In different experiments it varied between 2- and 4-fold above background (data not shown). To assure that these increases were not the result of day-to-day variations in growth conditions of cells, the experiments with pDCM72 and the plasmid used as the negative control, pACYC184, were always done in parallel. Triplicate cultures were used in all experiments, and, although results from only one experiment are reported, each experiment was repeated to confirm its reproducibility. Similar precautions were taken for all the reversion assays reported below.

When the gene for M. EcoRII (ecoRIIm) was cloned into pBR322 and introduced into the test strains, it also caused an increase in Kan^R reversion frequency with the kanS-H94 allele (Table 2). Once again, the level of increase in the reversion frequency was substantially lower than that with the kanS-D94 allele. Together, these data suggest that when dcm or ecoRIIm genes are cloned into medium copy number plasmids, there is an increase in the frequency of C T mutations at the non-canonical site CCGGG by a factor of up to 4.

We were concerned that the observed increases in Kan^R reversion frequency caused by these MTases may be due to mutations at a site other than the CCWGG site at codon 94. To eliminate this possibility, plasmid DNA was extracted from 10 revertants obtained in experiments involving Dcm and kanS-H94. These DNA preparations typically contained three plasmids: the plasmid carrying the MTase gene, the plasmid with the original kanS-H94 allele, and the plasmid with the revertant. The latter plasmid was separated from the other plasmids by retransformation into a new host and selecting for Kan^R phenotype. DNA was isolated from the transformants and analyzed by restriction digests. When codon 94 is CCG, a C T mutation at the second position in this codon (but not at the first position) eliminates a SmaI site (CCCGGG) and creates a BstNI site (CCTGG, Fig. 1). Fig. 2shows the restriction pattern for two such revertants. As expected, both the plasmids had lost the SmaI site (lanes 6 and 7) and gained a BstNI site (lanes 9 and 10). Furthermore, the sizes of the newly created BstNI fragments in the revertants were consistent with the sizes expected if a new BstNI site were to be created at codon 94 (not shown). The remaining eight revertants also showed a similar pattern of restriction sites. These results confirm that Dcm causes the expected sequence change at codon 94 of kanS-H94.

Figure 2: Restriction mapping of revertants. pKanS-H94 DNA and DNAs of two independent revertants were digested with different restriction enzymes, the products were electrophoresed on a 0.7% agarose gel and stained with ethidium bromide. Lanes 2, 5, and 8 contain pKanS-H94 DNA. Lanes 3, 6, and 9 contain DNA from one of the revertants, and lanes 4, 7, and 10 contain DNA from the second revertant. Lane 1, X174 HaeIII digest; lanes 2-4, uncut DNA; lanes 5-7, SmaI digest; lanes 8-10, BstNI digest; lane 11, bacteriophage BstEII digest. Reversion causes disappearance of SmaI fragment marked A (lane 5) and a smaller fragment (not seen), and appearance of a new fragment (marked B in lanes 6 and 7). The same mutation creates a BstNI site at codon 94, causing the disappearance of fragment C (lane 8) and appearance of two new fragments (D and E in lanes 9 and 10).


It seemed conceivable that somehow the presence of a DNA methyltransferase (MTase) in the cells or the mere interruption of the tetracycline resistance (Tet^R) gene in the vector causes a small increase in general mutation frequency in the host and that this phenomenon was responsible for the observed increase in Kan^R reversion frequency. To eliminate this possibility, EcoRI MTase was cloned into pACYC184 and the plasmid was introduced into the test strains. M. EcoRI is an MTase that methylates the second adenine in the sequence GAATTC(20) . When the Kan^R reversion frequency in these strains was compared with that in strains with pACYC184, no increase the frequency of mutations was observed with either of the two test plasmids (Table 3). Therefore, the apparent increase in reversion frequency due to the C5 MTases at non-canonical sites is not the result of some nonspecific effect of the presence of an MTase in the host on a multicopy plasmid or of the interruption of the tetracycline resistance gene in pACYC184.

M. HpaII Causes C T Mutations at a CCAG Site

We wished to determine whether this phenomenon was restricted to the closely related MTases Dcm and M. EcoRII or whether other C5 MTases could also cause mutations at non-canonical sites. To test this, gene for the MTase in the HpaII restriction-modification system was cloned into pACYC184 and the resulting recombinant was tested for its effect on Kan^R reversions with test plasmid carrying the CCAGG site at codon 94 (kanS-D94). With this plasmid, M. HpaII would be expected to encounter a CCAG site, instead of its canonical site, CCGG(24) .

M. HpaII caused a 4-fold increase in Kan^R reversion frequency with the kanS-D94 allele (Table 4). Once again, although this increase was substantially less than that seen at the canonical site, it was quite reproducible. In different experiments the enhancement in mutagenesis caused by M. HpaII at the non-canonical site has been found to vary between about 3- and 4-fold above background (not shown). Plasmid DNA was isolated from nine such revertants, and restriction analysis of the DNA was performed. In this case, the previously existing BstNI site (CCAGG) at codon 94 of kanS-D94 was found to be lost in 8 out of the 9 revertants (not shown). The remaining revertant had suffered DNA rearrangements and was not studied further. These results show that in nearly every case, the increase in the Kan^R reversion frequency caused by M. HpaII was the result of C T change at codon 94 and was not due to second-site revertants. Based on these results, we conclude that M. HpaII is also capable of causing mutations at a non-canonical sequence.

M. EcoRII Methylates CCGGG Sites at a Slow Rate

A possible simple explanation for the results described above is that the MTases methylate the non-canonical sites at a certain rate and that the resulting 5-methylcytosine deaminates to give rise to C T mutations. This seemed particularly likely in our reversion assay because the non-canonical sites differ from the canonical sites only in one position. Thus a modest ``relaxation'' of MTase specificity causing methylation of the CCGGG site could explain the result. M. DsaV, a C5 MTase that methylates the second cytosine in the sequence CCNGG (27, 33) illustrates this point. When a plasmid carrying the gene for this MTase was introduced into the test strains, it caused roughly equal amounts of increase in the reversion frequency for the two kan alleles (Table 5).

We tested the ability of M. EcoRII in vitro to transfer methyl groups to a 27 bp DNA duplex containing CCGGG sequence (Duplex I) and compared the kinetic parameters for this reaction with those for methyl transfer to duplex containing a CCAGG sequence (Duplex II). The DNAs used in these experiments contained the sequences surrounding codon 94 of kanS-D94 or kanS-H94. For this reason, they were considered to be good models for understanding methylation at codon 94 of kan by M. EcoRII in vivo.

The enzyme methylated the non-canonical DNA sequence at a low rate (Table 6). Interestingly, the principal difference in the interaction of the enzyme with the two substrates was reflected in differences in K (Table 6). While the K for methyl transfer to the CCGGG-containing substrate was lower than that for the canonical substrate by a factor of 2.0 times 10^3, K(m) for the non-canonical substrate was higher by only a factor of 27 (Table 6). If the K(m) values for the two substrates are taken to reflect K(S) values, M. EcoRII can be said to discriminate between the two substrates more at the level of catalysis than at the level of DNA binding.

Although the rate of methylation of the non-canonical duplex by M. EcoRII is poor, methylation does take place at the expected site. We demonstrated this by methylating P-labeled Duplex I with excess M. EcoRII and challenging the DNA with HpaII endonuclease. The digested DNA was separated from resistant DNA by gel electrophoresis, and the extent of protection against HpaII was quantitated. While HpaII digested 93-96% of the untreated DNA, it consistently cut the M. EcoRII-treated DNA less well. Analysis of the gel using a PhosphorImager revealed that 7.0% (S.D.= ±2.4%; n = 3) more of the total DNA was resistant to HpaII as a result of reaction with M. EcoRII, than without it. Because HpaII is inhibited by methylation of either cytosine in its recognition sequence(34, 35) , we can only conclude that M. EcoRII must have methylated one of the cytosines in the sequence CCGGG to render it resistant to HpaII.

Overproduction of M. EcoRII Causes Methylation at Codon 94 of kanS-H94 and Higher Reversion Frequencies

If occasional methylation of the CCGGG site by Dcm and M. EcoRII was responsible for the increase in Kan^R reversion frequency, it should be possible to detect such methylation by challenging pKanS-H94 DNA isolated from these cells with HpaII and separating the products on agarose gels. Occasional methylation at these sites should create one or more partial digestion products that could be identified by an appropriate detection technique. However, when pKanS-H94 was isolated from cells that also contained pDCM72 and was digested with HpaII, no partial digestion products could be seen by ethidium bromide staining of the gel (not shown). We reasoned that this may be partly due to the fact that HpaII has >30 sites in each plasmid making the task of identifying partial digestion products difficult. We also reasoned that the increase in reversion frequency of kanS-H94 allele caused by pDCM72 or pR300 is small and may indicate that in cells containing these plasmids only a small percentage (<10%) of pKanS-H94 molecules are methylated at codon 94. This should make the detection of partial digestion products even harder.

To improve the chances of detecting methylation at non-canonical sites by the MTases, two changes were made in the procedure. First, the gene for M. EcoRII was cloned into a high copy number plasmid, pUC118. Cells containing the resulting plasmid, pR400, were found to contain approximately 50 times as much methyltransferase activity as those containing pR300 (Table 7). Because pBR322-based plasmids have 30-50 copies/cell, while pUC-based plasmids have several hundred copies per cell, the level of expression of M. EcoRII may be the result of gene dosage effect. We reasoned that the overproduction of the MTase should result in greater methylation at non-canonical sites and increase the likelihood of its detection by ethidium bromide staining. This plasmid was introduced in cells containing kanS-H94 in pACYC184 (pKanS-H94/ACYC), and the ability of pR400 to methylate at codon 94 of kan was studied. Second, SmaI was used to detect methylation at codon 94 instead of HpaII. Codon 94 is within a SmaI site (Fig. 1), and C-5 methylation of the innermost cytosine in the SmaI recognition sequence, CCCGGG, inhibits this enzyme(35) . Furthermore, as pR400 contains no SmaI sites (Fig. 3, lane 4), SmaI restriction pattern of plasmids from these cells containing pR400 and pKanS-H94/ACYC consists of linear fragments from pKanS-H94/ACYC.

Figure 3: Protection of pR400 against SmaI. Plasmid DNAs were digested with SmaI and the products separated by gel electrophoresis. Lane 1, uncut pKanS-H94; lane 2, pKanS-H94 cut with SmaI; lane 3, uncut pR400; lane 4, pR400 cut with SmaI; lane 5, uncut pR400 + pKanS-H94; lane 6, SmaI-cut pR400 + pKanS-H94; Lane 7, bacteriophage BstEII markers.


When pKanS-H94/ACYC was isolated from a strain lacking pR400 and was digested with SmaI, the DNA was cut to completion revealing three bands on the agarose gel (Fig. 3, lane 2). In contrast, when pKanS-H94/ACYC was isolated from a strain containing pR400, the former plasmid was found to be partially protected against SmaI. In this case, a significant fraction of pKanS-H94/ACYC DNA appeared to be uncut (Fig. 3, compare lanes 5 and 6). As a result, although 3 times as much DNA was loaded in lane 6 compared to lane 2 of the gel, bands corresponding to complete SmaI digest were more intense in the latter lane (Fig. 3). In addition, three partial digestion products could be seen on the gel (Fig. 3, lane 6),^2 two of which had sizes consistent with the sizes of expected partial digestion products containing codon 94. These results directly demonstrate that when M. EcoRII is overproduced in cells, the second cytosine in codon 94 of kanS-H94 is methylated in some molecules.

The higher levels of M. EcoRII in cells containing pR400 also caused higher frequencies of Kan^R reversion. When pR400 was introduced in cells containing pKanS-H94, the Kan^R reversion frequency increased by a factor of 18 (Table 2). This was more than four times higher than the increase caused by pR300 at this site. This suggests that the ability of M. EcoRII to cause mutations at the CCGGG site may be directly related to the ability of the enzyme to methylate this site.

Methylation and Mutagenesis Caused by M. EcoRII at Other Non-canonical Sites

We interpreted the presence of uncut DNA in the SmaI digest of pKanS-H94/ACYC to mean that all SmaI sites in the plasmid were partially methylated by M. EcoRII. Furthermore, because SmaI sites contain CCSGG sequences (S is C or G), it seemed likely that the reason for the observed partial resistance of pKanS-H94/ACYC against SmaI was that M. EcoRII was methylating most CCSGG sites. We tested this hypothesis by attempting to cut pR400 with ScrFI, which recognizes the sequence CCNGG and is inhibited by C-5 methylation of the second cytosine in its sequence(25, 36) . If M. EcoRII were to methylate only the CCWGG sites in a plasmid, ScrFI should cut this DNA only at CCSGG sites and hence its fragment pattern should be identical to that of NciI. The latter enzyme cuts at CCSGG sites regardless of C-5 methylation of the inner cytosine(27, 33) . Instead, if M. EcoRII were to methylate some of the CCSGG sites in the plasmid, the ScrFI digestion pattern of the plasmid should contain a number of partial digestion products and the restriction pattern should be distinct from the NciI pattern. The ScrFI digestion patterns of pBR322 and pR300 appear to contain no incomplete digestion patterns (Fig. 4, lanes 9 and 10) and are similar to the corresponding NciI digestion patterns (lanes 6 and 7).

Figure 4: Protection of pR400 at CCSGG sites. Lane 1 contains bacteriophage BstEII markers. Lane 2 contains uncut pBR322 DNA. Lanes 3-11 contain different DNAs digested with different enzymes. Lanes 3, 6, and 9, pBR322; lanes 4, 7, and 10, pR300; lanes 5, 8, and 11, pR400. Lanes 3-5, EcoRII; lanes 6-8, NciI; lanes 9-11, ScrFI. The positions of partial digestion products in ScrFI digest are marked by brackets on the right side of lane 11.


As expected, the ScrFI digestion pattern of pR400 was different than its NciI digestion pattern and contained a number of partial digestion products (Fig. 4, compare lanes 8 and 11). It is clear from these data that M. EcoRII produced by pR400 methylates several CCSGG sites in addition to its methylation of CCWGG sites (Fig. 4, lane 5). We have further shown that this result is not restricted to pR400, but is also true of other overproducers of M. EcoRII. Plasmid DNAs from other overproducers of M. EcoRII including one other pUC-based overproducer and those based on overexpression of the gene from P or P promoters were partially protected at CCSGG sites from ScrFI digestion (see below and data not shown). In each case the DNA was completely sensitive to BstNI and NciI, demonstrating that the lack of cutting by ScrFI was unlikely to be due to inhibition by contaminants in the DNA.

We were interested in finding out whether sequences other than CCNGG were protected by M. EcoRII. In particular, it seemed possible that M. EcoRII may also recognize other ``four-out-of-five'' (4/5, for short) sites such as NCWGG and CCWGN in DNA and methylate the second base in the sequences. To test for such methylation, the sequence of pR400 was scanned and a PstI site (GCCTGCAGGT) in the pUC polylinker that overlaps with four 4/5 sites was identified. PstI is known to be inhibited by C-5 methylation within its recognition sequence(36, 37) . When pR400 DNA was digested with PstI, about 10% of its DNA was found to be resistant to PstI, confirming the partial methylation of this site (not shown).

Using pKanS-H94 it was not possible to determine whether M. EcoRII caused an increase in the rate of C T mutations sites at the other non-canonical sites. However, we noticed that Jones and colleagues had fortuitously constructed a mutant of kan with a 4/5 site at codon 70 ( Fig. 1and (13) ). If this site were to be methylated by Dcm or M. EcoRII, the mutant would be expected to revert at a higher frequency. To test this, pDCM72 was introduced into cells containing this kan allele (pKanS-D70) and the reversion frequency was determined. Presence of Dcm in the cells more than doubled the reversion frequency (Table 1, lines 5 and 6). A plasmid carrying the dcm was used in these experiments instead of pR300 or pR400, because the latter two plasmids are incompatible with the plasmid containing the kan allele. These data further support our earlier conclusion that C5 MTases can increase rates of C T mutations at sites related to their canonical sites.

Although M. EcoRII was found to methylate several different non-canonical sites, it does not methylate DNA indiscriminately. When pR400 was digested with Sau3AI (recognition sequence GATC), AluI (AGCT), HhaI (GCGC), or ApaLI (GTGCAC) and the products separated on agarose gels, the digests appeared to be complete (not shown). In addition a StuI site (AGGCCT) within ecoRIIm gene which overlaps two 4/5 sites appeared to be completely susceptible to StuI digestion by the gel electrophoretic assay. In contrast, HpaII or MspI (CCGG; inhibited by meCCGG) digests of pR400 contained some incomplete digestion products (not shown). Presumably, this is because the recognition sequences for the former group of enzymes are much less likely to overlap with 4/5 M. EcoRII sites than is CCGG.

Overproduction of M. EcoRII Causes DNA Degradation and Induction of the SOS Response

E. coli K-12 contains at least three restriction systems that cleave DNA containing methylated cytosines within specific sequences. These are McrA, McrBC, and Mrr(38, 39) , which restrict DNA methylated at CmeCGG(40) , RmeC (R is purine; Refs. 40 and 41), and meCG(38, 39) , respectively. Methylation of the second cytosine within CCWGG does not result in restriction by these systems, but if overproduction of Dcm or M. EcoRII were to cause methylation at 4/5 sites, this should lead to cleavage of cellular DNA by one or more of these restriction systems. Such restriction can be detected as presence of degraded DNA in plasmid preparations or by a genetic assay (see below). To test this, we used a P-based overproducer of M. EcoRII (plasmid pR234). (^4)The hybrid gene was repressed by maintaining pNK627, a compatible plasmid carrying the lacI^Q gene, in the cells. Under these conditions, pR234 could be stably maintained in cells containing active Mcr and Mrr systems.

Under repressed conditions, there was enough M. EcoRII produced in the cells to methylate the CCWGG sites making plasmid DNA isolated from the cells resistant to EcoRII (not shown). ScrFI digestion of the same DNA gave rise to a pattern identical to the pattern generated by NciI (not shown) and contained no indication of presence of partial digestion products (Fig. 5, lane 3). When the P promoter was induced by the addition of IPTG to the growth medium (final concentration 100 µM), M. EcoRII activity in the cells increased by a factor of 8 (Table 7). When plasmid DNA isolated from cells after 3.5 h of induction was subjected to digestion by ScrFI, a substantial fraction of the DNA appeared as partial digestion products ( Fig. 5compare lanes 3 and 5). This is consistent with our earlier conclusion that the overproduction of M. EcoRII leads to methylation of CCSGG sites.

Figure 5: Protection of pR234 at CCSGG sites. Lane 1 contains bacteriophage HindIII markers. Lanes 2 and 3 contain pR234 DNA from uninduced cells. Lanes 4 and 5 contain pR234 DNA from induced cells. Lanes 2 and 4, uncut DNA; lanes 3 and 5, DNA cut with ScrFI.


Interestingly, the DNA isolated from the induced cells also contained high molecular weight DNA that appeared as a smear in the gel or did not enter the gel (Fig. 5, compare lanes 2 and 4). Such high molecular weight DNA was not found to contaminate plasmids isolated from uninduced cells (Fig. 5, lane 2) or from cells induced for only 2 h (data not shown). Furthermore, plasmid DNA isolated after 2 h of induction was only slightly methylated at non-canonical sites (data not shown). We suggest that the contaminating high M(r) DNA is chromosomal DNA that has been degraded by the Mcr systems between 2 and 3.5 h after the addition of IPTG. It should be noted that such DNA was absent from the preparations of pR400 (Fig. 4), probably because the plasmid was isolated from an Mcr Mrr host BH143.

We confirmed the apparent degradation of DNA following M. EcoRII induction using a genetic assay. Heitman and Model (26) have described an Mcr MrrE. coli strains in which a promoter-less lacZ gene is inserted downstream from a DNA damage-inducible promoter. In these strains degradation of DNA results in the induction of the SOS response, which causes induction of beta-galactosidase. In many cases, DNA degradation is not severe enough to cause immediate cell death and hence colonies can be obtained on plates. Because the presence of beta-galactosidase can be determined by a colorimetric assay, this is a convenient system to study restriction by the Mcr and Mrr systems. Using this strain, Heitman and Model (26) showed that presence of M. HpaII and M. MspI in cells causes degradation of DNA.

PR234 and pNK627 were introduced into one such strain (JH140), and the cells were plated on LB plates containing 5-bromo-4-chloro-3-indoyl beta-D-galactoside (X-gal) alone or with IPTG. Appropriate antibiotics were also present in the plates to assure the retention of the two plasmids during cell growth. Overnight incubation of the plates without IPTG at 30 °C gave rise to light blue colonies, while the colonies that appeared on plates with X-gal and IPTG were a much deeper shade of blue (not shown). The color of the latter set of colonies was typical of strains that are LacZ. Presumably, IPTG-mediated induction of ecoRIIm gene resulted in significant damage to DNA leading to SOS induction and expression of beta-galactosidase.

To provide a quantitative measure for this phenomenon, beta-galactosidase activity from IPTG-induced cells was determined and compared to activity from uninduced cells. The results are summarized in Fig. 6. The uninduced cells contained significant levels of beta-galactosidase activity, and the activity increased by a factor of 3 as cells entered stationary phase. This is consistent with the observation mentioned above that colonies on plates without IPTG had a faint blue color (see also (26) ). But the cells to which IPTG had been added behaved differently in two significant ways. The beta-galactosidase activity in these cells increased by a factor of 17, and the bulk of this increase occurred within the 3rd hour after the addition of IPTG (Fig. 6B). Also, in contrast to the uninduced cells, these cells appeared to have stopped dividing at about 3 h following the addition of the inducer (Fig. 6A). The time course of induction of beta-galactosidase activity and the cessation of cell division correlate well with the increase in methylation at non-canonical sites and the appearance of high M(r) DNA in the plasmid preparations described above.

Figure 6: SOS response following the induction of M. EcoRII activity. Vertical arrow marks the time at which IPTG was added to one culture. Open squares, culture grown with IPTG. Open circles, culture grown without IPTG. A, turbidity of cultures monitored using optical density measurements at 600 nm. B, beta-galactosidase activity was determined for samples from the same two cultures at various times. The enzyme activity is normalized with respect to cell density as described by Miller(32) .


DISCUSSION

We have shown here that overproduction of M. EcoRII in E. coli causes significant methylation within several CCSGG sequences and within other sequences that were not known to be substrates for its catalytic action. This methylation causes a substantial increase in C T mutations at the non-canonical sites. This correlation between MTase overproduction, methylation at non-canonical sites and mutagenesis at these sites helps explain our initial observation that even at lower levels of MTase activity significant amounts of C T mutations can be detected at non-canonical sites. These mutations are likely to be caused by the deamination of 5-meC or due to other mutagenic interactions at sites of methylation (see Introduction). Because we have demonstrated the existence of this phenomenon with M. HpaII, in addition to M. EcoRII and Dcm, we expect that all C5 MTases will display such effects.

The ability of C5 MTases to methylate non-canonical sites was anticipated in one of our earlier studies(33) . In it we showed that a multicopy plasmid carrying dsaVgene is restricted by McrBC and concluded that this was consistent with the methylation by M. DsaV at sites other than its canonical sequence: CCNGG. Methylation by N^6-adenine methyltransferases at non-canonical sites has been demonstrated before(42, 43, 44, 45) , but this is the first demonstration of similar behavior by C5 MTases.

Overproduction of Native MTase Induces the SOS Response in E. coli

Another consequence of substantial overproduction of M. EcoRII in E. coli K-12 is the induction of the SOS response. There are two reasons to believe that this response is caused by the degradation of methylated DNA by the Mcr nucleases. A number of C5 MTases that methylate within sequences distinct from CCWGG have been shown to induce SOS response(46) . These include the MTases M. MspI (46) and M. HpaII.^2 Therefore, an MTase that methylates within CCSGG sequences would be expected to induce the SOS response in E. coli. Second, when DNA of plasmids that overproduce M. EcoRII was isolated from strains that were Mcr, the large M(r) DNA characteristic of chromosomal DNA degradation was not found in plasmid preparations ( Fig. 3and Fig. 4, and data not shown).

Although the SOS response is known to cause mutations, it cannot be the cause of mutations at the non-canonical sites. One of the reasons for this conclusion is that the strain used in the mutational studies was deleted for all known Mcr systems and hence it is unlikely that SOS induction occurred in this strain. Also, mutations at non-canonical sites are found to occur at levels of M. EcoRII and Dcm at which there is no evidence of SOS induction. When dcm and ecoRIIm genes are present in medium copy number plasmids such as pBR322 or pACYC184 and are expressed from their endogenous promoters, the SOS response was not observed(46) .^2 Additionally, SOS mutagenesis requires error-prone ``translesion'' synthesis and copying of 5-meC by at least E. coli polymerase I Klenow fragment is known not to be error-prone(9) . Finally, we and others (^5)have found that plasmids similar to pR300 and pDCM72 do not cause an overall increase in forward mutation frequency in E. coli as measured by rifampicin and streptomycin resistance assays.^2

Specificity of VSP Repair May Coincide with the Methylation Specificity of Dcm

Our results help explain a puzzling observation regarding a DNA repair system in E. coli. The potential mutagenic effects of 5-meC T deaminations are counteracted in E. coli by a mismatch correction process called very short patch (VSP) repair. This process repairs T:G mismatches to C:G when they are present in the sequence contexts 5`-CTAGG/3`-GGTCC or CTTGG/GGACC(47, 48, 49, 50) . As a result, if a T:G mismatch arises from the deamination of 5-meC in C5-meCWGG sequence, VSP repair restores the Dcm site. As expected, the frequency of mutations at a CCWGG site is several times higher in a (dcm) vsr strain than in a vsr strain(2, 15) .

Oddly, VSP repair also corrects T:G mismatches that lie within sequences other than CCWGG, albeit at lower efficiencies. Genetic evidence exists for the repair of T:G mismatches in NTAGG/NGTCC or 5`-CTAGN/3`-GGTCN to C:G by this system(47) . Additionally, purified Vsr protein (which is an endonuclease that nicks DNA immediately upstream of the mismatched T; (51) ) cleaves substrates that differ from the canonical Dcm sequence by one base pair(52, 53) . The most efficiently repaired mismatch is CTAGG/GGTCC, and replacement of the central A:T pair or the terminal C:G or G:C pairs by other pairs reduces the efficiency of repair to between 5 and 68% of this value (53) . In addition, two duplexes in which both the terminal base pairs had been substituted (TTAGA/AGTCT and TTAGC/AGTCG) are also repaired at a low efficiency(53) . It is not easy to understand why VSP repair has such broad sequence specificity if Dcm (and M. EcoRII) are assumed to interact with and methylate only CCWGG sequences. However, now that we have shown that Dcm and M. EcoRII can methylate non-canonical 4/5 sites under certain conditions, the imagined discrepancy between specificities of the MTases and the DNA repair process (53) may be eliminated. What remains to be done is to catalog all the sequences methylated by Dcm and M. EcoRII and compare them to the family of sequences within which VSP repair corrects T:G mismatches.

Significant levels of methylation at non-canonical sites only occur when the MTase is overexpressed. The level of Dcm protein expressed from the chromosomal copy of this gene is low enough that not all CCWGG sites are protected from R. EcoRII endonuclease cleavage (54, 55) . Like M. EcoRII (Table 6), Dcm is expected to have a strong preference for the canonical sequence; hence, under these conditions it is unlikely that a significant fraction of 4/5 sites will be methylated by Dcm. However, little is known about the regulation of dcm or the role of Dcm in E. coli, and it is possible that under certain physiological conditions, Dcm is overproduced in the cells causing complete methylation of the canonical sites and partial methylation of 4/5 sites. We suggest that it is this possibility that Vsr is designed to deal with. Deamination of 5-meC at the non-canonical sites would create T:G mismatches within 4/5 sequences, and these would then be repaired by VSP repair.

Multiple Pathways in E. coli to Avoid Mutations at Methylation Sites

In this context, the Mcr functions could also be seen as being antimutagenic in their action. If occasional overproduction of Dcm or M. EcoRII in the cells causes partial methylation of non-canonical sites, Mcr-mediated cleavage of DNA at these sites (56) should initiate recombinational and repair processes that should eventually replace the methylated DNA with unmethylated DNA. This should reduce the possibility of C T mutations at these sites. Thus, in addition to their role in excluding incoming DNA with non-native methylation patterns, Mcr enzymes may also guard against the tendency of Dcm to methylate within non-canonical sites.

It is also interesting to note that M. EcoRII is a negative regulator of its own expression(57, 58) . Although the regulation of dcm and hpaIIm genes has not been studied, mspIm gene is also under autogenous control(59) . It is possible that one of the reasons for such tight regulation of C5 MTase genes is to reduce mutagenic damage caused by these enzymes. Negative regulation by M. EcoRII appears to be related to methylation of its canonical site. The enzyme has two DNA-binding domains, one for the methylation sequence and the other for an operator sequence within its promoter(58) . Further, unmethylated CCWGG sequence, but not the methylated sequence, inhibits binding of the enzyme to the operator. Presumably, the enzyme binds the operator and shuts off transcription when all CCWGG sites have been methylated. In this way, efficient transcription of ecoRIIm gene occurs only when unmethylated CCWGG sites are present.

It is not possible from these studies to pinpoint the level of MTase activity in the cells at which methylation of non-canonical sites becomes significant. While dcm is present as a single-copy chromosomal gene in E. coli, the EcoRII genes were originally found on a low copy number natural plasmid in a clinical isolate of E. coli(17, 18) . Compared to levels of the two MTases in these strains, the pBR322- and pACYC184-based clones used in our study express the MTases at a level that is higher by at least an order of magnitude ((25) ; see above). The same may be true of the clones that express M. HpaII. This is because the HpaII restriction-modification genes were originally found to lie in H. parainfluenza chromosome(60) . Based on the genetic assay used in our studies, the level of expression of hpaIIm gene from the plasmid-based clone is sufficient to significantly methylate the non-canonical sites. Interestingly, at this level of methylation gel electrophoresis-based assay is unable to detect this additional methylation. Clearly, the genetic assay should be of considerable use in studying this phenomenon further.

Dcm and M. EcoRII Are Structurally Related to MTases That Methylate CCNGG Sites

It is not surprising that many of the additional sites of methylation by M. EcoRII are CCSGG sites. We (27) and others (61, 62) have pointed out that MTases that methylate CCWGG sequences are closely related to those that methylate CCNGG sequences. In particular, the ``variable'' regions of these enzymes contain two segments that are strongly conserved among these enzymes and this region of the sequence can be partially aligned with the segment of M. HhaI that is known to contact bases within its recognition sequence(27) . Therefore, the methylation of CCSGG sites by Dcm and M. EcoRII may occur because these enzymes already contain the information necessary to recognize CCSGG sequences or are able to adopt an alternate conformation that ignores the central base pair in the sequence during DNA binding. Based on the relative values of K and K(m) of M. EcoRII for the canonical and non-canonical sequences, we predict that the enzyme binds CCSGG sequences fairly well (K(d) values comparable to those for CCWGG), but is highly inefficient at performing the catalytic reaction on this substrate. The ability of M. HpaII to cause mutations at a CCAG site suggests that similar flexibility in binding DNA substrates may be found in all C5 MTases.

Mutations at Non-canonical Sites and Cancer

The ability of C5 MTases to cause mutations within non-canonical sequences may have implications for mutagenesis in general and for mutations linked to cancer, in particular. As mentioned in the Introduction, the largest class of mutations in the p53 gene are C:G to T:A transitions within CpG dinucleotides. As such, they are analogous to mutational hot spots at CCWGG sites in E. coli. It is also known that as much as 3,000-fold overexpression of cytosine methyltransferase activity is found in some tumorigenic cell lines(63) , and increases in the level of MTase gene expression are found to correlate well with progression of colon cancer(64) . Based on the results presented here, we predict that under conditions of overproduction the human MTase may also methylate non-canonical sequences related to CpG (CpA or perhaps CpN). If true, the contribution of the human MTase to mutations in pre-cancer cells may be much greater than previously imagined.

Note Added in Proof-Recently Clark et al. (65) transfected mammalian cells with DNA from E. coli and studied its methylation pattern after several generations. Their observation that some CAG, CTG, and CCG sequences that were not Dcm sites were methylated, can be explained by the results described here. Presumably, input E. coli DNA contained occasional methylation at CCWG, CWGG, and CCSGG sites. Therefore, it is unlikely that mammalian cells contain a de novo CNG methylating activity.

FOOTNOTES

*
This work was supported by National Institutes of Health Grant GM45860 (to M. Lieb (USC Medical School) and A. S. B.) and by Research Career Development Award HG00004 from the National Institutes of Health (to A. S. B.) The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore by hereby marked ``advertisement'' in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

§
To whom reprint requests should be addressed. Tel.: 313-577-2547; Fax: 313-577-8822.

(^1)
The abbreviations used are: 5-meC, 5-methylcytosine; C5 MTase, cytosine methyltransferase; M., methyltransferase restriction enzyme; IPTG, isopropyl-1-thio-beta-D-galactopyranoside; VSP, very short patch; X-gal, 5-bromo-4-chloro-3-indoyl beta-D-galactoside.

(^2)
B. Bandaru, J. Gopal, and A. S. Bhagwat, unpublished results.

(^3)
S. Gabbara, unpublished results.

(^4)
S. Gabbara and A. S. Bhagwat, unpublished results.

(^5)
M. Lieb, personal communication.

ACKNOWLEDGEMENTS

We are grateful to P. Jones (USC School of Medicine) for providing the plasmid CCAG-pSV2-neo. We thank Dr. R. Blumenthal (Medical College of Ohio) and E. Raleigh (New England Biolabs) for critical reading of the manuscript.

REFERENCES

  1. Coulondre, C., Miller, J. H., Farabaugh, P. J., and Gilbert, W. (1978) Nature 274, 775-780 [CrossRef][Medline] [Order article via Infotrieve]
  2. Lieb, M. (1991) Genetics 128, 23-27 [Abstract]
  3. Cooper, D. N., and Youssoufian, H. (1988) Hum. Genet. 78, 151-155 [CrossRef][Medline] [Order article via Infotrieve]
  4. Jones, P. A., Rideout, W. A., III, Shen, J.-C., Spruck, C. H., and Tsai, Y. C. (1992) BioEssays 14, 33-36 [CrossRef][Medline] [Order article via Infotrieve]
  5. Greenblatt, M. S., Bennett, W. P., Hollstein, M., and Harris, C. C. (1994) Cancer Res. 54, 4855-4878 [Free Full Text]
  6. Laird, P. W., and Jaenisch, R. (1994) Hum. Mol. Genet. 3, 1487-1495 [Abstract]
  7. Lindahl, T., and Nyberg, B. (1974) Biochemistry 13, 3405-3410 [CrossRef][Medline] [Order article via Infotrieve]
  8. Wang, R. Y.-H., Kuo, K. C., Gehrke, C. W., Huang, L., and Ehrlich, M. (1982) Biochim. Biophys. Acta 697, 371-377 [Medline] [Order article via Infotrieve]
  9. Shen, J.-C., Creighton, S., Jones, P. A., and Goodman, M. F. (1992) Nucleic Acids Res. 20, 5119-5125 [Abstract/Free Full Text]
  10. Shen, J.-C., Rideout, W. M., III, and Jones, P. (1992) Cell 71, 1-20 [CrossRef][Medline] [Order article via Infotrieve]
  11. Yebra, M., and Bhagwat, A. S. (1995) Biochemistry 34, 14752-14757 [CrossRef][Medline] [Order article via Infotrieve]
  12. Vairapandi, M., and Duker, N. J. (1993) Nucleic Acids Res. 21, 5323-5327 [Abstract/Free Full Text]
  13. Yang, A. S., Shen, J.-C., Zingg, J.-M., Mi, S., and Jones, P. A. (1995) Nucleic Acids Res. 23, 1380-1387 [Abstract/Free Full Text]
  14. Klimasauskas, S., and Roberts, R. J. (1995) Nucleic Acids Res. 23, 1388-1395 [Abstract/Free Full Text]
  15. Bandaru, B., Wyszynski, M., and Bhagwat, A. S. (1995) J. Bacteriol. 177, 2950-2952 [Abstract/Free Full Text]
  16. Wyszynski, M., Gabbara, S., and Bhagwat, A. S. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 1574-1578 [Abstract/Free Full Text]
  17. Bannister, D., and Glover, S. W. (1968) Biochem. Biophys. Res. Commun. 30, 735-738 [CrossRef][Medline] [Order article via Infotrieve]
  18. Yoshimori, R., Roulland-Dussoix, D., and Boyer, H. W. (1972) J. Bacteriol. 112, 1275-1279 [Abstract/Free Full Text]
  19. Bigger, C., Murray, K., and Murray, N. E. (1973) Nat. New Biol. 244, 7-10 [Medline] [Order article via Infotrieve]
  20. Boyer, H. W., Chow, L. T., Dugaiczyk, A., Hedgepeth, J., and Goodman, H. M. (1973) Nat. New Biol. 244, 40-43 [CrossRef][Medline] [Order article via Infotrieve]
  21. Marinus, M. G., and Morris, N. R. (1973) J. Bacteriol. 114, 1143-1150 [Abstract/Free Full Text]
  22. May, M. S., and Hattman, S. (1975) J. Bacteriol. 122, 129-138 [Abstract/Free Full Text]
  23. Schlagman, S., Hattman, S., May, M. S., and Berger, L. (1976) J. Bacteriol. 126, 990-996 [Abstract/Free Full Text]
  24. Mann, M. B., and Smith, H. O. (1977) Nucleic Acids Res. 4, 4211-4221 [Abstract/Free Full Text]
  25. Bhagwat, A. S., Sohail, A., and Roberts, R. J. (1986) J. Bacteriol. 166, 751-755 [Abstract/Free Full Text]
  26. Heitman, J., and Model, P. (1991) Gene (Amst.) 103, 1-9
  27. Gopal, J., Yebra, M. J., and Bhagwat, A. S. (1994) Nucleic Acids Res. 22, 4482-4488 [Abstract/Free Full Text]
  28. Wyszynski, M. W., Gabbara, S., and Bhagwat, A. S. (1992) Nucleic Acids Res. 20, 319-326 [Abstract/Free Full Text]
  29. Bhagwat, A. S., and Person, S. (1981) Mol. Gen. Genet. 182, 505-507 [CrossRef][Medline] [Order article via Infotrieve]
  30. Som, S., Bhagwat, A. S., and Friedman, S. (1987) Nucleic Acids Res. 15, 313-332 [Abstract/Free Full Text]
  31. Brosius, J., and Holy, A. (1984) Proc. Natl. Acad. Sci. U. S. A. 81, 6929-6933 [Abstract/Free Full Text]
  32. Miller, J. (1992) A Short Course in Bacterial Genetics , pp. 72-74, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY _
  33. Gopal, J., and Bhagwat, A. S. (1995) Nucleic Acids Res. 23, 29-35 [Abstract/Free Full Text]
  34. Walder, R. Y., Langtimm, C. J., Chatterjee, R., and Walder, J. A. (1983) J. Biol. Chem 258, 1235-1241 [Abstract/Free Full Text]
  35. Butkus, V., Petrauskiene, L., Maneliene, Z., Klimasauskas, S., Laucys, V., and Janulaitis, A. A. (1987) Nucleic Acids Res. 15, 7091-7102 [Abstract/Free Full Text]
  36. Nelson, M., Christ, C., and Schildkraut, I. (1984) Nucleic Acids Res. 12, 5165-5173 [Abstract/Free Full Text]
  37. Dobritsa, A. P., and Dobritsa, S. V. (1980) Gene (Amst.) 10, 105-112
  38. Kelleher, J. E., and Raleigh, E. A. (1991) J. Bacteriol. 173, 5220-5223 [Abstract/Free Full Text]
  39. Kretz, P. L., Kohler, S. W., and Short, J. M. (1991) J. Bacteriol. 173, 4707-4716 [Abstract/Free Full Text]
  40. Raleigh, E. A., and Wilson, G. (1986) Proc. Natl. Acad. Sci. U. S. A. 83, 9070-9074 [Abstract/Free Full Text]
  41. Raleigh, E. A. (1992) Mol. Microbiol. 6, 1079-1086 [Medline] [Order article via Infotrieve]
  42. Woodbury, C. P., Jr., Downey, R. L., and von Hippel, P. H. (1980) J. Biol. Chem. 255, 11526-11533 [Free Full Text]
  43. Taylor, J. D., Goodall, A. J., Vermote, C. L., and Halford, S. E. (1990) Biochemistry 29, 10727-10733 [CrossRef][Medline] [Order article via Infotrieve]
  44. Reich, N. O., Olsen, C., Osti, F., and Murphy, J. (1992) J. Biol. Chem. 267, 15802-15807 [Abstract/Free Full Text]
  45. Smith, D. W., Crowder, S. W., and Reich, N. O. (1992) Nucleic Acids Res. 20, 6091-6096 [Abstract/Free Full Text]
  46. Heitman, J., and Model, P. (1987) J. Bacteriol. 169, 3243-3250 [Abstract/Free Full Text]
  47. Lieb, M., Allen, E., and Read, D. (1986) Genetics 114, 1041-1060 [Abstract/Free Full Text]
  48. Lieb, M. (1987) J. Bacteriol. 169, 5241-5246 [Abstract/Free Full Text]
  49. Jones, M., Wagner, R., and Radman, M. (1987) Cell 50, 621-626 [CrossRef][Medline] [Order article via Infotrieve]
  50. Zell, R., and Fritz, H.-J. (1987) EMBO J. 6, 1809-1815 [Medline] [Order article via Infotrieve]
  51. Hennecke, F., Kolmar, H., Bründl, K., and Fritz, H.-J. (1991) Nature 353, 776-778 [CrossRef][Medline] [Order article via Infotrieve]
  52. Gl ä sner, W., Hennecke, F., and Fritz, H.-J. (1992) in Structural Tools for the Analysis of Protein-Nucleic Acid Complexes (Lilley, D. M. J., Heumann, H., and Suck, D., eds) pp. 165-173, Birk ä user Verlag, Basel
  53. Gläsner, W., Merkl, R., Schellenberger, V., and Fritz, H.-J. (1995) J. Mol. Biol. 245, 1-7 [Medline] [Order article via Infotrieve]
  54. Bhagwat, A. S., Johnson, B., Weule, K., and Roberts, R. J. (1990) J. Biol. Chem. 265, 767-773 [Abstract/Free Full Text]
  55. Ringquist, S., and Smith, C. (1992) Proc. Natl. Acad. Sci. U. S. A. 89, 4539-4543 [Abstract/Free Full Text]
  56. Sutherland, E., Coe, L., and Raleigh, E. A. (1992) J. Mol. Biol. 225, 327-348 [CrossRef][Medline] [Order article via Infotrieve]
  57. Friedman, S., and Som, S. (1993) J. Bacteriol. 175, 6293-6298 [Abstract/Free Full Text]
  58. Som, S., and Friedman, S. (1994) Nucleic Acids Res. 22, 5347-5353 [Abstract/Free Full Text]
  59. Som, S., and Friedman, S. (1993) EMBO J. 12, 4297-4303 [Medline] [Order article via Infotrieve]
  60. Kulakauskas, S., Barsomian, J. M., Lubys, A., Roberts, R. J., and Wilson, G. G. (1994) Gene (Amst.) 142, 9-15
  61. Karyagina, A. S., Lunin, V. G., and Nikolskaya, I. I. (1990) Gene (Amst.) 87, 113-118
  62. Davis, R., van der Lelie, D., Mercenier, A., Daly, C., and Fitzgerald, G. F. (1993) Appl. Environ. Microbiol. 59, 777-785 [Abstract/Free Full Text]
  63. Kautiainen, T. L., and Jones, P. A. (1986) J. Biol. Chem. 261, 1594-1598 [Abstract/Free Full Text]
  64. El-Deiry, W. S., Nelkin, B. D., Celano, P., Yen, R.-W. C., Falco, J. P., Hamilton, S. R., and Baylin, S. B. (1991) Proc. Natl. Acad. Sci. U. S. A. 88, 3470-3474 [Abstract/Free Full Text]
  65. Clark, S. J., Harrison, J., and Frommer, M. (1995) Nature Genet. 10, 20-27 [CrossRef][Medline] [Order article via Infotrieve]
©1996 by The American Society for Biochemistry and Molecular Biology, Inc.

Add to CiteULike CiteULike