Role of proofreading and mismatch repair in maintaining the stability of
nucleotide repeats in DNA
Role of proofreading and mismatch repair in maintaining the stability of nucleotide repeats in DNA
Bernard S.
Strauss*
,
Daphna
Sagher
and
Sonia
Acharya
Department of Molecular Genetics and Cell Biology, The University of Chicago,
Chicago
, IL 60637,
USA
Received October 18, 1996;
Revised and Accepted December 15, 1996
ABSTRACT
The role of the proofreading exonuclease in maintaining the stability of multiply repeated units in DNA was studied in
Escherichia coli
. Reversion of plasmids in which the
[beta]
-galactosidase
[alpha]
complementing sequence was moved +2 out of frame by inserts containing (CA)14
, (CA)5
, (CA)2
or (TA)6
or +1 by creating a run of 8 C was compared in
mutS
and
mutSdnaQ
strains. Proofreading corrects at least half of the frameshift errors for all
the plasmids and at least 99% of the errors in the (CA)2
plasmid. The (CA)2
plasmid reverts mostly by +1 frameshifts in the restriction sites flanking the
insert. With the (CA)14
, (TA)6
, (CA)5
and 8C plasmids, reversion is mainly by loss of a repeat unit. The data support
the hypothesis that the
dnaQ
gene product recognizes frameshifts close to the DNA growing point. Frameshifts
distal to the growing point are mainly corrected by mismatch repair
.
We speculate that mismatches in mononucleotide repeats are susceptible to proofreading because they can either migrate to a point where
they are recognized by the exonuclease or, alternatively, because single nucleotide distortions are more readily detected than dinucleotides.
INTRODUCTION
Levinson and Gutman (
1
) showed that mismatch repair-deficient strains of
Escherichia coli
(
mutS
and
mutL
) produced increased numbers of frameshifts in (CA) repeat sequences (microsatellites) inserted into M13 bacteriophage DNA. Strand
et al
. (
2
), working with yeast, confirmed the observation that instability was associated with a deficiency in mismatch repair. Yeast mutants deficient in mismatch
repair deleted a (CA) repeat unit several hundred fold more frequently than the
wild-type. The frequency of (CA) deletion was either unchanged or increased a
modest 10-fold in yeast strains deficient in the proofreading exonuclease of either
of the two DNA polymerases as compared with the wild-type. They suggested that either the proofreading exonuclease did not
detect loops or bubbles that form away from the growing point or that
heterologies of >1 bp were not corrected (
2
). Although both yeast DNA polymerase mutants are efficient mutators [in
contrast to
E.coli
pol I mutants; (
3
)], the effect of double mutants, deficient in both replicative exonucleases, on
frameshift mutation was not reported.
We previously demonstrated a role for proofreading exonuclease activity in UV-induced frameshift mutagenesis at repeated sequences
in vitro
(
4
). There is also
in vitro
evidence that proofreading exonuclease is important in the surveillance of spontaneous frameshift
mutations (
5
). Since pol III is the major
E.coli
replication polymerase, the use of mutants deficient in the exonuclease subunit
of this enzyme offered an interesting additional test of the possible role of proofreading in microsatellite instability. The experiments reported here with
dnaQ
mutants show that, as in yeast (
2
), proofreading has only a minor role in the surveillance of frameshifts in long
repeated sequences. The role of proofreading depends on the size and
composition of the repeats. Since the exonuclease acts at DNA growing points,
whereas mismatch repair occurs at a distance from the replication fork, these
experiments imply that the events in frameshift mutation at highly repeated
sequences are not limited to the DNA growing point.
MATERIALS AND METHODS
Except where otherwise indicated, the methodology and the media used are as
described by Miller (
6
).
Strain construction
The selectable marker
met D
is located close to the replicative polymerase (
dnaE
) and proofreading (
dnaQ
) loci. We constructed a strain which carried
metD
-
metB
-
along with a deletion in the
lac
region. This strain, termed BS40 [
metD
-
metB
-
ara
+
pro
-
[Delta](
lac
)
Str
r
] was the progeny of a cross of S90C [F-
Str
r
metB
+
ara
-
metD
+
[Delta](
pro lac
)] * Hfr CD4 (
Str
s
metB
-
ara
+
met D
-
proA
-
) selected on arabinose + streptomycin-containing plates with subsequent screening for a
pro
-
colony unable to utilize D-methionine. This strain must carry a deletion in the
lac
region. We have not determined whether the deletion comes from the CD4 or S90C
parent. The newly isolated mutators (see below) were transferred into BS40 by
P1 transduction, selecting for
metD
+
and replica plating onto rifampicin plates to identify mutators. The fertility
factor of XL-1 blue
(F'::Tn
10 proA
+
B
+
lacI
q
lacZ
[Delta]
M15
) was transferred into BS40 and its derivatives to allow for M13 growth. The
mutS
(
mutS215
::Tn
10
) and
mutL
(
mutL218
::Tn
10
) strains prepared by E.Siegel were obtained from the
E.coli
Genetic Stock Center in a
thy
-
background. Since these strains were susceptible to thymineless death, we
transduced
mutS
and
mutL
into strain S90C by selection with streptomycin and tetracycline. S90C
mutH
was obtained from Dr J.Miller. A list of the strains used in this investigation
is available in the on-line edition of this Journal.
Plasmid construction
Repeat sequences (Fig.
1
) were introduced into the [beta]-galactosidase gene in both M13mp2 and in a plasmid, putting the gene out of frame
and yielding colorless plaques or colonies when grown in appropriate strains.
Reversions were detected as blue colonies or plaques. The plasmid used was p205-GTI (
7
) This 8.6 kb plasmid has an SV40 origin and a G418 resistance marker for
selection in mammalian cells, a pBR322 origin for replication in bacteria and
an Amp resistance marker. The LacZ [alpha] peptide differs from the wild-type in codons 2-5 and is preceded by the
E.coli tet
promoter. The sequence at codons 3 and 4 was changed to create a
Bam
HI site for insertion of the different oligomers containing the PyA repeats.
These oligomers were based on the sequence of pSH31 (
8
), with
Bam
HI sites at the ends. Modifications of the
lacZ
sequence (creation of the
Bam
HI site or the 8C stop sequence) were by
in vitro
mutagenesis in uracil-containing M13 (
9
). The 8C stop sequence (Fig.
1
) was prepared using the oligonucleotide CACCCCCCCCTTCGCTAG to replace the wild-type CATCCCCCTTTCGCCAG at codons 30-35 of the modified
lacZ
. M13 RF was then prepared, digested with
Bcl
I and
Bgl
II, separated on an agarose gel and the modified sequence isolated from the gel
and ligated into a similarly digested plasmid. The M13 used for these
manipulations was derived from M13mp2 by replacing the
Ava
II-
Pvu
I fragment (positions 5914-6351) with the
Bgl
II-
Bcl
I fragment from plasmid p205-GTI (modified
lacZ
and
tet
promoter). All experiments described in this study were done with the same
basic sequence in both the phage and the plasmid. We designate the M13 phage containing the (CA)14
insert M13is(CA)14
. The modified plasmids containing the different repeats are called (CA)14
, (CA)5
, (CA)2
, (TA)6
[since a (TA)5
is placed next to a TA in the basic insert] and 8C stop.
Mutagenesis of P1 phage
P1 phage were grown on strain CD4 transduced to
metD
+
. A phage suspension (4 * 1013
p.f.u./ml) was mutagenized with 1.0 M hydroxylamine solution (
10
). Phage survival was 2 and 6% in different runs measured against a mock-treated control. The mutagenized phage preparation was used for
transduction (
6
). Transductants were isolated on minimal agar plates containing D-methionine (10 [mu]g/ml), 2 mM phenyl-[beta]-D-galactoside (P-gal), 150 [mu]M 5-bromo-4-chloro-3-indolyl-[beta]-D-galactopyranoside (X-gal), streptomycin (125 [mu]g/ml) and citrate (20 mM, reduced to 5 mM
in later experiments;
6
). This selective medium is designed to identify mutants which form blue
papillae (
11
).
Sequencing
dnaQ
mutants
Three overlapping fragments of cellular DNA were amplified with primer pairs
based on sequences generated with the MacVector program. Numbering starts at the first nucleotide of the GenBank
dnaQ
sequence (Locus ECORNHQ, accession no. K00985M3020). The primer pairs were: F10 (283-304) 5'-TGATACCCTGGCGGACATACTG-3', B9 (842-820) 5'-CATGAACTCATCGGCTACTTCGG-3'; F12 (577-600) 5'-CCGCTATGAGCACTGCAA-
T TACAC-3', B11 (1103-1080) 5'-AACTTCCGCAAGGATCTGGGCATC-3'; F23 (893-916) 5'-CGGCTTTATGGACTACGAGTTTTC-3',
B32 (1529-1508) 5'-AGTGAATAGTGGCGGA- ACGGAC-3'. PCR products were sequenced using the
Promega fmol TM*1 Sequencing System with the amplification primers and the
additional following internal primers: B3 (626-603) 5'-GGTTTCGGTATCGAGAACGATCTG-3'; B15 (1168-1145) 5'-TGCTGTTGTGTCTCTCCTTCCATC-3'; F10b (367-386) 5'-CCATCAACTCCATACGGGTTG-3'; F17b (678-696) 5'-ATT-
GGTGCCGTTGAAGTGG-3'; F22b (1011-1029) 5'-AGCCTCGATGCGTTATGTG-3'; F27b (1219-1236) 5'-GCGTTGTTTTTGCGACAG-3'.
Cycle sequencing reactions were done with 32
P-end- labeled primers for 30 cycles or with an ABI prism automatic
sequencer at The University of Chicago Cancer Research Center.
Reversion analysis
Single plaques of M13mp2 phage containing the appropriate constructs were picked
into 1.5 ml LB medium and grown at 37oC overnight. The supernatants were assayed on strain JM103. Plasmids were
introduced into host cells by CaCl2
-mediated transformation and the cultures were grown overnight or, for
mutSdnaQ
mutants, until visible growth was obtained. Plasmid preparations were
introduced into strain JS5 by electroporation with a BioRad Gene Pulser with
Pulse Controller set at 2.5 kV and with 25 [mu]FD capacitance at 400 [Omega] using 0.2 cm cuvettes. The medium for the detection of reversion to
complementation of the [beta]-galactosidase genes is LB + `A salts' (
6
) supplemented with 100 mM IPTG, 150 [mu]M X-gal and ampicillin (100 [mu]g/ml). Blue colonies were scored after at least 18 h incubation at
37oC.
Revertants were colony purified. (CA)2
revertants segregated colorless colonies even after several subcultures and
many of the plasmid preparations from this construct were eventually sequenced as mixtures. Plasmid DNA was isolated from a 5 ml overnight culture of purified
colonies. Sequences were obtained either by cycle sequencing with the ABI Prism
377A Sequencer (courtesy of the UCCRF-DSF) or manually with the USB Sequenase Version 2.0 Kit for dideoxy Sanger
sequencing with [[alpha]-35
S]dATP (>1000 Ci/mmol; Amersham) and dITP mixes. The primer for the automatic
cycle sequencing reactions was 5'-CTGCGTGCAATCCATCTTGT-3', which primes 135 bp away from the insert. The primer for the Sanger sequencing
was 5'-TGGGTAACGCCAGGGTT-3', which primes 35 bp from the insert.
Quantitative PCR
Quantitative PCR was carried out on total DNA using 32
P-end- labeled primers designed to amplify either a 265 bp target in the
dnaQ
gene (chromosomal marker, primers 5'-CATGAACTCATCGGCTACTTCGG-3' and 5'-AACTTCCGCAAGGATCTGGGCATC-3') or a 460 bp target in the
plasmid [p205 (CA)5
] extending from a section of the kanamycin resistance gene on one side (5'-CCTGCGTGCAATCCATCTTGTTC-3') to the eukaryotic thymidine kinase on the other (5'-TCCACTTCGCATATTAAGGTGACG-3'). Bacterial cultures (1.5
ml) were spun down, washed in water and then resuspended in 50 [mu]l water, mixed with 150 [mu]l 5% Chelex (Chelex 100 Resin, 100-200 Mesh, sodium form; BioRad) in water. The suspension was
boiled for 10 min, chilled on ice, centrifuged for 15 min at top speed in a
microfuge and the supernatant diluted in water. Diluted supernatant (5 [mu]l) containing 6-60 pg DNA (for plasmid amplification) or 60-150 pg (for chromosomal amplification) was subjected to PCR in the presence of 2 pmol
32
P-end-labeled forward primer and 4 pmol backward primer using
Taq
polymerase. Following 5 min at 94oC, amplification was for 20 cycles of 1 min at 57oC (annealing), 1 min at 72oC (extension) and 1 min at 94oC, with a final 10 min extension. Samples were separated on
a 7.2% acrylamide gel and the amplified bands quantified with a phosphorimager
(STORM 860; Molecular Dynamics). The DNA amounts used were such that a linear increase in
concentration gave a linear increase in the intensity of the signal. For each target, only one primer was end-labeled. The primers for the different targets were labeled with the same [32
P]ATP mix so a direct comparison of number of molecules could be made from the
relative radioactivity of the target bands.
Mutation rate calculations and statistical treatment of the data
Mutation rates ([mu]) were calculated from the median frequency by reiteratively solving the
equation: [mu] = 0.4343
f
/log(
n
[mu]), where
f
is the median mutation frequency and
n
is the population size (
12
). The gene being studied is located on a plasmid and population size should
properly be related to plasmid copy number. Although we do not have absolute
values, we show below that the
mutS
and
mutSdnaQ
strains used contain similar plasmid copy numbers. Since the calculated
mutation rate is relatively insensitive to changes in the value of
n
, we used viable count rather than estimated plasmid number as a measure of
population size. There is only a 1.3-fold difference in calculated mutation rate for a 100-fold difference in population size (Table
3
). The mutation rate is a calculated value and we use the experimentally determined median mutation
frequency for statistical analysis. For this analysis we employed the Wilcoxon rank sum test, a non-parametric statistical method which does not depend on a normal distribution of the data. Comparison of the ratio of
mutSdnaQ/mutS
frequencies among the different plasmids was then performed using analysis of variance
after a log transformation across the different groups (plasmids) to stabilize
the variance (
13
).
RESULTS
dnaQ
mutants
We took advantage of the proximity of
dnaQ
and
dnaE
to the
metD
locus to isolate mutants by the localized mutagenesis technique (
10
). Mutagenized P1 phage were used to transduce a
metD
-
metB
-
[Delta](
proAB lac
)
Str
r
strain carrying the CC101 F' factor which reverts from
lac
-
to
lac
+
by an A:T -> C:G transversion (
14
).
Met
+
transductants were selected and putative mutators were recognized among the transductants by the presence of numerous blue
papillae against a colorless background. Approximately 30 putative mutators (
dnaQ-A
, -
B
,...), all from separate plates and from different transductions, were isolated.
The isolates were purified, checked for a mutator phenotype on rifampicin
plates and the mutation transferred by P1 transduction to strain BS40,
selecting for
metD
+
. Transductants were first transferred onto LB plates and then replica plated to rifampicin plates for classification of mutator activity.
f
, frequency of rifampicin-resistant mutants. The value for the
dnaQ
+
strain (BS40) is 5 * 10-9
.
Total DNA was extracted from the different isolates and
dnaQ
gene segments were amplified and sequenced in both directions as described in
Materials and Methods. Out of 15 isolates studied there were eight different
dnaQ
alleles (Table
1
).
DnaQ-N
and
dnaQ-O
are identical to the mutation described as
mutD51
(
15
). Five alleles are within the ExoI, ExoII and ExoIII regions defined as conserved in 3' -> 5' exonucleases (
16
). Three, including mut
D51
, are clustered midway between the ExoII and ExoIII domains. We have not found a
secondary mutation elsewhere in the genome to account for the lower mutability
of
dnaQ-N
(Table
1
). Most of the experiments described below utilize
dnaQ-G
(codon 162) or
dnaQ-E
(codon 167). Plasmids carrying either the wild-type allele of
dnaQ
(pMM5;
17
),
dnaE
[pOPPE (
18
,
19
) or a control insert p(TA)6
(this paper)] were introduced into
dnaQ-E
and
dnaQ-G
to test complementation by lowered rifampicin or nalidixic acid resistance.
Only the
dnaQ
+
plasmid showed complementation of
dnaQ-E
and
dnaQ-G
, consistently lowering mutation frequency in the
dnaQ
mutants by about two orders of magnitude (data not shown). The mutants used in
these studies had been transferred twice by P1 transduction and
metD
selection since their original isolation by P1 mutagenesis. It is therefore
unlikely that they carry any additional mutator mutations outside the
dnaQ
locus (
20
).
Experiments with bacteriophage M13
We determined the effect of both
dnaQ
and mismatch repair mutations on reversion of a (CA)14
sequence incorporated into the
lacZ
region of a modified M13mp2. We found a frequency of reversion to lac+
of ~1% when the (CA)14
-containing phage was grown in wild-type bacteria (Table
2
). This wild-type frequency was higher than previously reported (
1
) for a longer repeat sequence [(CA)21
]. When the (CA)14
-containing phage were grown on
mutH
,
mutL
or
mutS
strains an increased frequency of reversion, although not as large as
previously reported for the longer sequence (
1
), was observed.
DnaQ
mutants also displayed a higher than wild-type reversion frequency (Table
2
).
Experiments with plasmids
In order to obtain an experimental system in which the magnitude of the effect
of mismatch repair deficiency was closer to that observed in yeast (
2
), we prepared a series of repeated sequences contained in plasmid molecules.
Mutation rates were compared in wild-type,
mutS
and
mutSdnaQ
strains in order to reduce the uncertainty in interpretation caused by the
functional mismatch repair deficiency of
E.coli
dnaQ
mutants (
21
,
22
).
DnaQ
and
mutSdnaQ
mutants were prepared from BS40 by successive P1 transductions, transformed
with the particular plasmid and subcultured without purification of the transformants by restreaking. Whole colonies of the
mutSdnaQ
double mutants were lifted from the plates and inoculated into 10 ml LB +
ampicillin to minimize the accumulation of suppressor mutations in these
mutable strains (
20
). Cultures of the double mutants are always heterogeneous and we avoided
picking the larger colonies for transformation. It took between 24 and 36 h to
obtain dense cultures of transformed
mutSdnaQ
double mutants for plasmid preparation. About 10-20% of the picked colonies failed to grow. Plasmid preparations from all
strains were assayed for reversion by electroporation into strain JS5.
. Lac+ frequency after growth of M13is(CA)14 in mutator strains
Strain
Rifr * 107
Blue plaques (%)
Median
S90C
0.17
1.2
1.2
1.3
0.84
BS40metD+/XL(a)
0.24
1.3
1.3
1.0
4.9
BS40metD+/XL(b)
0.06
0.71
1.0
1.0
1.0
S90mutH/XL
26
3.3
3.8
3.8
4.2
S090mutL/XL
26
5.3
5.3
4.9
12.
S90mutS/XL
20
5.2
4.7
4.7
2.6
BS40dnaQ-E/XL
2000
3.2
3.5
3.5
4.2
BS40dnaQ-G/XL
2500
3.0
3.8
3.8
5.2
BS40dnaQ-N/XL
720
2.7
2.7
2.7
4.0
Single plaques were picked and grown in 1.5 ml LB medium in triplicate. Cultures were incubated overnight. The supernatants were collected and assayed on strain
JM103.
(a) and (b) represent independent isolates of BS40
metD
+
.
Rifr
* 107
, frequency of rifampicin-resistant bacteria in the different host strains.
DNA replication in plasmids with origins derived from pBR322 (ColE1) is
unidirectional and starts by the formation of a long RNA transcript which is
then replicated for 200-400 bases by
E.coli
pol I, after which replication on the leading strand switches to pol III (
23
). Lagging strand replication is always by pol III. The instability of long
trinucleotide repeats in
E.coli
is greatest when inserts are close to the origin; CTG repeats inserted 0.2-0.45 kb downstream of the origin were reported to be `much less' stable
than when inserted 1.5 kb away (
24
). Our inserts were all downstream of the initiation codon for the [alpha] peptide, which starts 725 bases from the first deoxynucleotide added in
DNA replication (
25
).
We constructed plasmids with 14, five or two (CA) repeats, an oligonucleotide
with five (TA) repeats, which when inserted into our vector gave a sequence
with six (TA) repeats, and a sequence with eight C residues followed by a stop
codon (Fig.
1
). The dinucleotide repeats shifted the [alpha] peptide out of frame by +2/-1 nt. The shift produced by the mononucleotide repeat (8C stop) is
+1. Mutation rates were calculated by the method of Drake (
12
) using the median reversion frequency in a series of independent cultures to
minimize the effect of outliers. The rate of reversion to a lac+
phenotype decreases rapidly as the number of repeated elements in the insert is
diminished (Table
3
). The rate is very much higher for plasmid molecules propagated in a
mutS
strain.
.
Mutation rates and frequencies for plasmid inserts
Strain
Plasmid insert
(CA)14
(8C stop)
(TA)6
(CA)5
(CA)2
Wild-type BS40
Mutation rate
4.9 * 10-5
7.0 * 10-7
4.2 * 10-7
9.5 * 10-7
8.6 * 10-8
Median frequency
5.9 * 10-4
5.2 * 10-6
2.8 * 10-6
7.6 * 10-6
4.7 * 10-7 a
No. of samples
12
9
10
15
17
mutS
Mutation rate
5.4 * 10-3
8.3 * 10-4
2.5 * 10-4
1.7 * 10-4
4.0 * 10-8
Median frequency
8.8 * 10-2
1.2 * 10-2
2.8 * 10-3
2.0 * 10-3
1.5 * 10-7 a
No. of samples
11
14
13
22
17
mutSdnaQ-G
Mutation rate
9.0 * 10-3
4.3 * 10-3
1.1 * 10-3
3.8 * 10-4
3.5 * 10-6
Median frequency
13.4 * 10-2
5.8 * 10-2
1.2 * 10-2
4.4 * 10-3
1.8 * 10-5
No. of samples
12
17
12
19
17
*15 of 17 wild-type and 13 of 17
mutS
cultures had 0 revertants, giving a median frequency of 0. The value given is
the total number of revertants in all 17 replicates divided by the total number
of colonies screened.
The introduction of a
dnaQ
mutation into the
mutS
strains makes for a dramatic increase in the reversion of the (CA)2
plasmid and for a 4- to 5-fold increase in the reversion rate of the (TA)6
and 8C stop plasmids. The increase in (CA)5
is smaller and the effect of
dnaQ
on reversion of the (CA)14
plasmid appears smaller still. The measured mutation frequencies rather than
the calculated mutation rates have been used for the statistical analysis of
differences between the reversion of the constructs in the different strains.
Specifically, comparison of
mutS
with
mutSdnaQ
groups using the Wilcoxon (non-parametric) rank sum test yielded
P
< 0.0001 for the (CA)2
, (CA)5
, (TA)6
and 8C stop plasmids and
P
< 0.0017 for the (CA)14
plasmid. We conclude that there is an effect of proofreading on the stability
of all the repeat tracts. The estimated ratios (actually the ratio of the
geometric means) of the
mutSdnaQ
and
mutS
frequencies were 2.15, 3.29, 3.92 and 5.39 for the (CA)14
, (CA)5
, (TA)6
and 8C stop plasmids respectively. Pairwise comparisons detected a significant difference between the 8C stop and (CA)14
ratios (
P
= 0.008), but none of the other pairwise comparisons reached statistical
significance. A test of the hypothesis that the
mutSdnaQ
/
mutS
frequency ratio is different for the 8C stop and (TA)6
plasmids combined as compared with the (CA)14
and (CA)5
plasmids (i.e. the average of the two former versus the average of the two
latter) yielded
P
= 0.020. Due to the multiple comparisons involved in the statistical analysis
we think this hypothesis needs to be confirmed by additional, independent
experiments. Our data with the
dnaQ-E
strain (data not given) are similar but not extensive enough to permit
statistical analysis.
Revertant analysis
As expected for (CA)14
, (CA)5
and (TA)6
, the major change is loss of a single dinucleotide unit (Table
4
). No +1 revertants were observed. The 8C stop plasmid was constructed so that
revertants could be obtained by loss of a single nucleotide and the 18
independent revertants sequenced had -1 changes. A more complex pattern was observed among the revertants of
the (CA)2
plasmid (Fig.
2
). One revertant, found after propagation of the plasmid in a
mutSdnaQ-E
strain, had addition of an A in the CACA repeat region. The majority of the
revertants were +1 duplications of single nucleotides, observed for the most
part in the two
Bam
HI restriction sites flanking the inserted sequence. Figure
2
, indicating the location of these mutations, has been drawn to suggest a
possible stem-loop in the vicinity of the restriction sites.
Calculation of mutation rates requires knowledge of the number of replications,
which is usually estimated by a count of viable cells. However, there is
evidence (see below) that cultures of the double mutant contain a significant proportion of non-viable cells. It seemed possible that the dead
dnaQmutS
cells accumulated intact plasmid copies or that the viable double mutant cells
had a higher plasmid copy number. The relative number of plasmid to chromosome
copies was determined by adapting a PCR method developed by S.Benson (personal
communication). BS40
MutS
and BS40
mutSdnaQ
containing the p205(CA)5
plasmid were grown for 12 and 40 h respectively. Total DNA was extracted and plasmid and chromosomal sequences in the same samples were
amplified and the products quantified after electrophoresis (Fig.
3
).
DISCUSSION
Repeated sequences in plasmids are subject to efficient mismatch repair
The frequency of reversion of the (CA)14
repeat in phage grown in mismatch repair-deficient strains was 3- to 5-fold greater than when grown in wild-type hosts (Table
2
). Levinson and Gutman (
1
) reported 16-fold increases for (CA)21
repeats in their phage. This low ratio in the bacteriophage as compared with
yeast (
2
) results from the relatively high control values of ~1% in the wild-type host (Table
2
). One possible explanation is that the large number of slippage events in many
phage DNA molecules saturates the mismatch repair system so that only part of
the frameshifts can be corrected (
21
). A second possibility is that during single-stranded replication DNA is encoated in protein before the mismatch repair
system has had time to act.
The mutation rate for (CA)14
plasmids in a
mutS
strain is 110 times higher than in the wild-type. The ratio of reversion in
mutS
compared with the wild-type is >1000 for the 8C stop plasmid. These values compare with ratios of
100 (for chromosomal) and 500-700 for plasmid repeat sequences in yeast (
2
) and indicate that frameshifts in plasmid DNA are very efficiently monitored by
the
E.coli
mismatch repair system.
The wild-type
dnaQ
gene product detects frameshifts
The
dnaQ
-
G
and
dnaQ-E
mutants were isolated on the basis of an inability to detect transversions. It
is likely that frameshifts occur as a result of failed interactions between
polymerase and nucleic acid substrate at one domain (
27
), whereas base substitutions arise as a result of errors at another domain,
most probably the growing point region. It might well be that different
exonuclease changes have a differential effect on the capacity to proofread
base substitution and frameshift errors. Our data show that the
dnaQ-G
mutant used in this work (and the
dnaQ-E
mutation, mapped only five codons away, for which we have less extensive data)
plays a major role in the recognition of some frameshifts.
The total number of errors made by the polymerase can be obtained from the
mutation rate in the
mutSdnaQ
strain, in which both proofreading and mismatch repair are absent. The value
for mutation rate obtained from the
mutS
strain gives the number of errors made in the absence of mismatch repair but
with efficient proofreading. The relative contribution of proofreading can be
determined by comparison of mutation rates in the
mutS
and
mutSdnaQ
strains. A plot of the rate of polymerase errors for the (CA)5
, (TA)6
, 8C stop and (CA)14
plasmids is approximately linear with an intercept near five repeat units (Fig.
4
). These data reflect experimental mutation frequencies varying from 0.4 to
13.4% with no sign of saturation. A simple hypothesis which explains these
results is that the mutations are due to slippage which is enhanced as the
number of repeats increases past four. Regardless of the nature of the repeats,
whether TA, CA or C, the frequency of errors is proportional to their number,
i.e. the 8C sequence behaves as eight repeats and (TA)6
behaves as six. In these calculations we assume that the
dnaQ-G
mutation has completely eliminated exonucleolytic proofreading. Since cells with a complete loss of proofreading activity are probably inviable (
26
), our values for the contribution of proofreading are minimal estimates. In
addition, we cannot exclude the possibility that mutated exonuclease subunits
interact with the polymerase to produce holoenzyme with altered processivity or
stability.
About 40-50% of the frameshift mutations made by polymerase replicating the (CA)14
or (CA)5
plasmid are corrected by wild-type proofreading exonuclease, as compared with ~80% of the errors in the 8C stop or (TA)6
plasmids (Fig.
4
). Based on comparisons of
dnaQmutL
and
mutL
strains, 97.5-99.5% of base substitutions are corrected by proofreading and not by
mismatch repair (
22
). The data for the (CA)2
plasmid resemble those for chromosomal base substitutions (99% corrected by proofreading). This distinction is reinforced by the different nature of the (CA)2
revertants obtained (Fig.
2
and Table
4
). We suppose that these +1 reversions occur at the DNA growing point and are
monitored by the wild-type proofreading exonuclease in a manner similar to base substitutions.
Frameshifts in highly repeated tracts are likely to occur or migrate upstream
of the growing point, where they are only susceptible to mismatch repair (
2
). The distance from the growing point at which this slippage occurs may be
related to the distance between residues in the polymerase `thumb' (
28
), which fix the newly synthesized strand and its template to the protein and
the end of the DNA recognized by the proofreading subunit. The longer the
repeated tract, the greater the probability of a frameshift and the lower the
probability that this perturbation will reach a point susceptible to nuclease
action before the frameshift is irreversibly fixed by the progression of DNA
synthesis. The decreased contribution of proofreading to the maintenance of
long repeats
in vivo
corresponds well with the
in vitro
measurements showing decreased exonuclease surveillance of longer mononucleotide runs (
29
). It is possible that the major quantitative role of the mismatch repair system, at least in eukaryotes in which the DNA contains numerous dinucleotide repeats, is not the correction of base substitutions but the maintainance of the integrity of repeated regions
and protection from recombination (
30
,
31
).
The statistical analyses above suggest that exonuclease deficiency has more of an effect on the 8C stop and (TA)6
plasmids combined than on the (CA)14
and (CA)5
plasmids (
P
= 0.02). This difference prompts us to speculate that mononucleotides are
edited more easily than dinucleotides. In addition, two different but related
mechanisms may make perturbations in the (TA)6
and 8C stop structure susceptible to nuclease. First, we suppose that the
extrahelical nucleotide(s) can migrate towards and away from the growing point in a wave-like movement and that a mononucleotide, as in the 8C stop sequence, is likely to migrate more readily than a
dinucleotide. Secondly, a TA dinucleotide is likely to melt more readily than a
CA dinucleotide, making its migration towards the primer-template end easier and subjecting it to exonuclease editing. The data of Brenowitz
et al
. (
32
) suggest that proofreading requires melting of the newly synthesized DNA.
The stability of long repeated sequences and of single nucleotides within genes are controlled by different but overlapping systems. Proofreading enzymes may correct large numbers of polymerase-produced point mutations but be relatively inefficient at correcting errors in long repeats. Mismatch repair proteins correct
errors in long repeated units (microsatellites) but in the presence of an
efficient prooofreading system cells deficient in mismatch repair may remain
relatively free of point mutations. It may be that the specificity of mismatch
repair mutations in certain types of cancer (
33
,
34
) is related to the production of specific mutations in genes with long repeated
tracts rather than to their overall effect on point mutations (
35
).
ACKNOWLEDGEMENTS
These experiments were started at the laboratory of Dr Tomas Lindahl at the ICRF
in London and one of us (B.S.) would like to thank Dr Lindahl and the group at
ICRF, particularly Barbara Sedgwick, Steve West, Rick Wood, Masahiko Sato and Bob Lloyd, for their help. We are especially grateful to Dr Barbara Bachman of the
E.coli
Genetic Stock Center and to Dr Jeffrey Miller for the bacterial strains used to
start this work and to Drs Alain Sarasin and Anne Stary for the P205-GTI plasmid. We thank Dr Charles McHenry for plasmid pOPP-E and Roel Schaaper for the
dnaQ
-containing plasmid and for helpful suggestions. Dr Theodore Karrison (Department of Medicine, The University of Chicago) provided the statistical analyses and Dr Spencer Benson suggested the use of PCR for the measurement of plasmid and chromosomal DNA. Drs
Roel Schaaper and Malcolm Winkler provided helpful comments on the manuscript.
This work was supported in part by Department of Energy grant DE-FG02-88ER60678 and by National Cancer Institute grant R37-CA32436 and by a Fogarty International Fellowship to one of us
(B.S.).
REFERENCES
1 Levinson,G. and Gutman,G. (1987) Nucleic Acids Res., 15, 5323-5338.
2 Strand,M., Prolla,T., Liskay,R.M. and Petes,T. (1993) Nature, 365, 274-276.
3 Bates,H., Randall,S.K., Rayssiguier,C., Bridges,B.A., Goodman,M.F. and Radman,M. (1989) J. Bacteriol., 171, 2480-2484.
4 Sagher,D., Turkington,E., Acharya,S. and Strauss,B. (1994) J. Mol. Biol., 240, 226-242.
5 Bebenek,K., Joyce,C.M., Fitzgerald,M.P. and Kunkel,T. (1990) J. Biol. Chem., 265, 13878-13887.
6 Miller,J. (1992) A Short Course in Bacterial Genetics. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
7 Stary,A., Menck,C. and Sarasin,A. (1992) Mutat. Res., 272, 101-110.
8 Henderson,S. and Petes,T.D. (1992) Mol. Cell. Biol., 12, 2749-2757.
9 Kunkel,T.A., Roberts,J. and Zakour,R. (1987) Methods Enzymol., 154, 367-382.
10 Hong,J. and Ames,B. (1971) Proc. Natl. Acad. Sci. USA, 68, 3158-3162.
11 Nghiem,Y., Cabrera,M., Cupples,C. and Miller,J. (1988) Proc. Natl. Acad. Sci. USA, 85, 2709-2713.
12 Drake,J.W. (1991) Proc. Natl. Acad. Sci.USA, 88, 7160-7164.
13 Snedecor,G.W. and Cochran,W.G. (1980) Statistical Methods, 7th Ed. Iowa State University Press, Ames, IA, Chapter 15, section 13.
14 Cupples,C.G. and Miller,J.H. (1989) Proc. Natl. Acad. Sci. USA, 86, 5345-5349.
15 Takano,K., Nakabeppu,Y., Maki,H., Horiuchi,T. and Sekiguchi,M. (1986) Mol. Gen. Genet., 205, 9-13.
16 Blanco,L., Bernad,A. and Salas,M. (1992) Gene, 112, 139-144.
17 Horiuchi,T., Maki,H., Maruyama,M. and Sekiguchi,M. (1981) Proc. Natl. Acad. Sci. USA, 78, 3770-3774.
18 Tomasiewicz,H.G. and McHenry,C.S. (1987) J. Bacteriol., 169, 5735-5744.
19 Tomasiewicz,H.G. (1990) The Macromolecular Synthesis II Operon of Escherichia coli. University of Colorado Medical School, Denver, CO.
20 Schaaper,R.M. and Cornacchio,R. (1992) J. Bacteriol., 174, 1974-1982.
21 Schaaper,R.M. and Radman,M. (1989) EMBO J., 8, 3511-3516.
22 Schaaper,R.M. (1993) J. Biol. Chem., 268, 23762-23765.
