Download Site- and strand-specific mismatch repair of

Carcinogenesis vol.18 no.7 pp.1311–1318, 1997
Site- and strand-specific mismatch repair of human H-ras genomic
DNA in a mammalian cell line
Loretta Arcangeli1,2, Josephine Simonetti1,
Catherine Pongratz1 and Kandace J.Williams1,3
1Biomedical
Program and Department of Biological Sciences, University of
Alaska, Anchorage, Alaska 99508, USA
2Present
address: Hopkins Marine Station, Stanford University, Oceanview
Boulevard, Pacific Grove, CA 93950, USA
3To
whom correspondence should be addressed
Defective mismatch repair has recently been implicated
as the major contributor towards the mutator phenotype
observed in tumour cell lines derived from patients
diagnosed with hereditary non-polyposis colon cancer
(HNPCC). Cell lines from other cancer-prone syndromes,
such as xeroderma pigmentosum, have been found to be
defective in nucleotide excision repair of damaged bases.
Some genetic complementation groups are defective specifically in transcription-coupled excision repair, although
this type of repair defect has not been associated with
cancer proneness. Mechanisms contributing to the high
incidence of activating point mutations in oncogenes (such
as H-ras codon 12) are not understood. It is possible that
novel mechanisms of misrepair or misreplication occur at
these sites in addition to the above DNA repair mechanisms.
In this study, we have compared the rate of strand-directed
mismatch repair of four mispairs (G:A, A:C, T:C and G:T)
at the H-ras codon 12, middle G:C position. Our results
indicate that, although this location is not a ‘hot spot’ for
bacterial mismatch repair, it is a ‘hot spot’ for decreased
repair of specific mismatched bases within NIH 3T3 cells.
NIH 3T3, unlike Escherichia coli, have an extremely low
repair rate of the G:A mispair (35%), as well as the A:C
mispair (58%) at this location. NIH 3T3 also have a
moderately low repair rate of the T:C mispair (80%) at
the codon 12 location. Conversely, NIH 3T3 repair of
G:T (100%) is comparable to E.coli repair (94%) of this
mismatch. These results demonstrate that a mismatch
containing an incorrect adenine on either strand at the
H-ras codon 12 middle base pair location is most likely to
undergo a mutational event in NIH 3T3 cells. Conversely,
a mismatch containing an incorrect thymine in the transcribed strand is least likely to undergo a mutational event.
Introduction
The three primary gene products required for strand-specificity
of mismatch repair in Escherichia coli, MutS, MutL and MutH,
have been well described (1). These three enzymes work in a
coordinated fashion to make a single-strand nick in the new
strand of unmethylated DNA opposite the nearest methylated
d(GATC) site, leading to either a 39 → 59 or 59 → 39 gap
excision ‘long patch’ repair process to replace the incorrect
*Abbreviations: HNPCC, hereditary non-polyposis colon cancer; RER1,
replication errors; bp, base pair; ds, double strand.
© Oxford University Press
base. A series of recent studies have resulted in the identification of several human genetic homologues of bacterial DNA
mismatch repair enzymes, such as hMSH2, hMLH1, hPMS1,
hPMS2 and GTBP, although the exact biochemical function
of each gene product has yet to be identified. Available evidence
suggests that both the MSH2-GTBP (hMutSα) heterodimer and
the MLH1-PMS2 heterodimer (hMutLα) can act as mismatch
recognition complexes (1). Recent in vitro experiments have
indicated that at least seven different biochemical activities
are required for mammalian mismatch repair (2). Mechanisms
of ‘new’ strand discrimination prior to mismatch repair within
mammalian cells have not yet been elucidated. However,
experiments using E.coli or human cell extracts have demonstrated strand-specificity of mismatch repair by the creation of
a pre-existing nick on the same strand of DNA as the incorrect
base, located either 59 or 39 from the mismatch (3–6).
Within mammalian cells, there is more than one repair
system for G:T or G:U mispairs putatively arising from
endogenous 5-methylcytosine or cytosine deamination events,
respectively (7,8). One well-described repair mechanism for
the G:T mispair is a base excision repair process using a
specific thymine glycosylase that does not require stranddiscrimination (9). In addition, it has recently been reported
that the GTBP polypeptide of the hMutSα complex appears
to bind preferentially with G:T mispairs (10). One more
mismatch repair complex, hMutLα (composed of hMLH1 and
hPMS2) can restore strand-discriminatory repair activity for
all eight mismatches in nuclear extracts of a human tumour
cell line (6). It is now evident that all combinations of
mismatches can be repaired in mammalian cells; however, G:T
mispairs appear to be corrected with higher efficiency than
other mispairs, perhaps due to a wider repertoire of repair
mechanisms (1,11).
The majority of patients who have been diagnosed with
hereditary non-polyposis colon cancer (HNPCC*) have germline mutations in one or more of the human mismatch repair
homologues. Studies using tumour cell lines developed from
these patients have demonstrated a distinct mutator phenotype
that appears to have a substantial increase in replication errors
(RER1) and therefore, may be involved in subsequent tumour
progression. RER1 cells commonly exhibit microsatellite
instability, an increase in HPRT gene mutability, insertiondeletion mutation events and tolerance to alkylation base
damage. Furthermore, extracts from these cells have an inability
to perform nick-directed mismatch correction (2). DNA mismatch repair in these cells is blocked prior to the gap excision
stage. In addition, human mismatch repair deficiencies have
recently been implicated in deficiencies of transcriptioncoupled nucleotide excision repair of damaged nucleotides (12).
Progress in understanding the link between mismatch repair,
the mutator phenotype and cancer in human cells has been
phenomenally rapid. Issues yet to be resolved include biochemical interactions between mismatch repair complexes and DNA,
and differential repair of specific mismatched bases in sensitive
1311
L.Arcangeli et al.
genomic locations. Virtually all mismatch repair information
obtained to date has resulted from experiments using synthetic
or bacterial DNA sequences and bacterial assays or mammalian
cell extracts. Therefore, information about the differential
repair of specific mammalian genetic sequences within mammalian cells is essentially unknown.
Oncogenes and tumour suppressor genes containing distinct
‘hot spots’ of mutation are of high interest in regard to accuracy
of DNA repair processes as well as fidelity of DNA replication.
Activating mutations in these genes consistently found in
human tumours may be due partly to a genetic defect in one
of the DNA repair pathways, such as mismatch repair or
nucleotide excision repair (13). However, to account for the
high specificity and frequency of mutations at these hot spots,
it is probable that a yet to be identified mechanism of
misreplication or misrepair plays an additional role at these
sensitive locations in the genome. Clearly, etiology of cellular
transformation has yet to be elucidated in regard to the high
mutation rate in these susceptible genomic locations.
The subject of this work has been to examine differences
in both rate and accuracy of in vivo site-, strand- and basespecific mismatch repair located at a well known ‘hot spot’ in
the human H-ras oncogene. The objectives of this research
have been to determine differences, if any, in the frequency
and accuracy of mammalian mismatch repair proficiency at a
sensitive oncogenic location (H-ras codon 12, middle nucleotide) depending on type and orientation of each mismatched
base. Mismatch repair rates have also been compared between
mammalian and bacterial organisms to establish a baseline
understanding of differences in mismatch repair mechanisms
that may exist between these diverse species. We use a unique,
highly sensitive assay previously developed in this laboratory
for studying mammalian cellular repair of DNA damage at a
relevant location in regard to cellular transformation (14).
Materials and methods
Enzymes
T4 DNA ligase, calf intestinal alkaline phosphatase (CIP), KpnI, XhoI
and BfrI were purchased from Boehringer Mannheim Biochemicals. T4
polynucleotide kinase was purchased from Promega. Replitherm DNA polymerase was purchased from Epicentre Technologies. BamHI and HpaII were
purchased from New England Biolabs. HindIII was purchased from United
States Biochemical. Proteinase K was purchased from Sigma Chemical Co.
Shrimp alkaline phosphatase I (SAPI), exonuclease I (ExoI) and thermosequenase were purchased from Amersham.
Cells, plasmids, oligonucleotides and other reagents
NIH 3T3 cells were obtained from American Type Culture Collection (ATCC).
The construction of plasmids p220.pbc and p220.pbc1H/B has been described
previously (14; Figure 1). M13mp18, M13mp19, pUC19, DH5α competent
E.coli, LipofectAMINE, and Opti-MEM were purchased from Life Technologies, Inc. E.coli strain NR9161 (-mutL) was a kind gift from Roel Schapper
(NIEHS, Research Triangle Park, NC). All synthetic oligonucleotides were
purchased from Operon Technologies, Inc. Radioactively labelled nucleotides
were purchased from Dupont–New England Nuclear. Agarose for electrophoretic separation and purification of DNA was purchased from FMC
Bioproducts. Gene Clean II was purchased from Bio 101, Inc. Dulbecco’s
Modified Eagle’s Medium (DMEM; 4.5 g/l glucose) and bovine calf serum
were purchased from Hyclone Laboratories, Inc. Hygromycin B was purchased
from Calbiochem Biochemicals. All other reagents were purchased from
Sigma Chemical Company unless otherwise noted.
Formation of p220.pbc1H/B containing H-ras codon 12 site-specific mismatch
A 2 kb BamHI–KpnI H-ras segment from p220.pbc1H/B (containing codon
12) was ligated into the polylinker region of pUC19 (prasBK2.0), M13mp18
(M13ras18.9) and M13mp19 (M13ras19.1), as described previously (14). The
plasmid prasBK2.0 was digested with AflIII and PvuI to release a 2.5 kb
fragment containing the inserted H-ras segment. The purified 2.5 kb DNA
1312
fragment was then digested with HindIII and BfrI to remove the original 30base pair (bp) segment containing codon 12 of H-ras (Figure 1, step 1).
Mismatch oligonucleotides complementary to the 30 bp region spanning
codon 12 (middle dG replaced with dT or dA for non-transcribed strand
mismatch oligomer, or middle dC replaced with dT or dA for transcribed
strand mismatch oligomer) were phosphorylated and purified by G-50
Sephadex column as described previously (14).
Generation of specific mismatched heteroduplex DNA (HD DNA) was
made according to choice of mismatch 30-mer oligo and single-strand (ss)
M13ras complementary DNA. Use of ss M13ras18.9 results in a 30-bp
double-strand (ds) region containing either dT or dA mismatched codon 12
middle nucleotide in the coding strand of H-ras (non-transcribed strand).
Inversely, use of ss M13ras19.1 results in a 30-bp ds region containing either
dT or dA mismatched codon 12 middle nucleotide in the non-coding strand
of H-ras (transcribed strand).
The phosphorylated mismatch oligo was annealed with the respective ss
M13ras DNA at a 50:1 molar ratio and a total DNA concentration of
50 ng/µl in TE (10 mM Tris, pH 8.0, 1 mM EDTA) at 65°C for 15 min,
resulting in an M13ras with a 30 bp mismatch insert on one strand
(Figure 1, step 2A). H-ras fragments digested and purified from prasBK2.0
complementary to the regions flanking the 30-bp mismatch region were heatdenatured at 100°C for 8 min and annealed with the ss M13/mismatch oligo
HD DNA at a 1:2 molar ratio at 65°C for 30 min in annealing buffer (40 mM
Tris, pH 7.5, 20 mM MgCl2, 50 mM NaCl) (Figure 1, step 2B). This partially
ds M13ras molecule was cleaved with XhoI and KpnI to produce a 1.8 kb ds
fragment (Figure 1, step 3), that was then purified, gently dephosphorylated
(to discourage 1.8 kb concatemer ligation; 0.01 U CIP per pmol of 59 termini
for 1 h at 37°C) and ligated to the 13.5 kb XhoI–KpnI vector portion of
p220.pbc at a 4:1 molar ratio, and additional phosphorylated mismatch oligo
at a 200:1 molar ratio, at 16°C overnight. This final step produces the
p220.pbc1H/B plasmid, containing a site- and strand-specific mismatch at Hras codon 12, middle base pair (Figure 1, step 4).
Figure 2 depicts a comparative NarI restriction digestion of a purified 1.8 kb
XhoI/KpnI segment of control H-ras DNA and the same segment that has
been obtained after preparation of site- and strand-specific mismatch DNA, as
described above (Figure 1, steps 1–3). Although some of the NarI cleaved DNA
segments apparently reflect missing NarI sites within intronic regions of the
2.0 kb segment of H-ras after cloning into the M13 bacteriophage, the two bands
representing the distances between the NarI site at codons 10/11 (nucleotide no.
1693 of human genomic H-ras sequence) to the adjacent NarI site downstream
of exon 1 (no. 1280), as well as to the KpnI site upstream of exon 1 (no. 1967),
appear identical in size for both DNA segments, 413 bp and 274 bp, respectively,
indicating that H-ras exon 1 (nos 1664–1774) is intact and the annealed mismatch
30-mer is in place within the site- and strand-specific mismatch DNA.
Transfection, selection and analyses of human H-ras DNA from ampicillin
resistant E.coli and hygromycin resistant NIH 3T3 cells
Competent DH5α E.coli were transformed with mismatched plasmid DNA
following the manufacturer’s protocol (Life Technologies, Inc.). NR9161 E.coli
were made competent and transformed by the calcium chloride procedure (15).
Bacteria were grown overnight at 37°C on LB agar plates containing 75 µg/ml
carbenicillin. NIH 3T3 cells were grown in DMEM, 10% calf serum at 37°C,
5% CO2. Cells were seeded at 13106 per 100 mm plate in preparation for
transfection experiments 16–18 h later. Plasmid DNA (50 ng per plate) was
transfected into NIH 3T3 cells using LipofectAMINE reagent as described by
the manufacturer (Life Technologies, Inc.). NIH 3T3 hygromycin resistant cells
were subsequently selected by the addition of 125 U of hygromycin per millilitre
of media, starting 36 h after transfection. At the end of 3 weeks, hygromycin
resistant colonies on each plate were methanol (97%) fixed.
DNA was purified from each E.coli and NIH 3T3 colony and amplified by
PCR, yielding an amplified DNA product of 129 bp containing the entire exon
1 of human H-ras plus several human intronic nucleotides surrounding exon 1,
as described previously (14). A 20-µl aliquot of the amplified DNA was then
digested with HpaII restriction enzyme. Only PCR amplified DNA containing
wild-type human H-ras codon 12 sequence (GGC) is cleaved by HpaII, yielding
an 88 bp and a 41 bp band from the original 129 bp sequence (Figure 3A). All
PCR amplified DNA not completely cleaved by HpaII was treated with the
enzymes SAPI and Exol and cycle-sequenced, as described by the manufacturer
(Amersham).
All contaminating sequences, either from original p220.pbc used for the
13.5 kb vector DNA, or wild type transfection control DNA (codon 12; middle
G:C), or from p220.T24 used for activated transfection control DNA (codon
12; middle T:A), were detected by observing sequences at codons 6 and 15,
both of which have been altered to create either a unique HindIII site (codon
6) or BfrI site (codon 15) in the final p220.pbc1H/B plasmid containing the
site-specific mismatch (14). Any sequences not containing the altered sequences
at codons 6 (CTG → CTT) and 15 (GGC → CTT) were discarded.
Fig. 1. Construction of p220.pbc1H/B containing strand- and site-specific mismatch at H-ras codon 12, middle position, as described in Materials and methods.
Defective mismatch repair
1313
L.Arcangeli et al.
Results
Although it is not clear why this large plasmid will not transfect
into cells efficiently if it is not in the ds circular form, it does
serve as a control to exclude simple ‘gap repair’ results from
true mismatch repair results.
To determine experimentally if this method would, indeed,
bias mismatch repair to the incorrect base by nicked-strand
specificity, each mismatch experiment was performed by
transfecting aliquots of the same mismatched ligation products
into both E.coli [DH5α as mismatch repair competent strain;
NR9161 (-mutL) as mismatch repair defective strain] and NIH
3T3 cells. Table I demonstrates the overall high rate and high
fidelity of correct mismatch repair (to G:C) at this site for all
four of the mismatches tested by mismatch repair competent
E.coli (DH5α). Table I also illustrates 100% negative mismatch
repair by mismatch repair defective E.coli (NR9161). Results
from NR9161 experiments indicate the high purity of the sitespecific mismatch ligation products used in these experiments,
although this may be due, in part, to cellular selectivity for ds
(non-gapped) circular plasmids. Table II contains results of
NIH 3T3 mismatch repair of aliquots of the same ligation
products for each mismatch. All experiments were repeated
with separately prepared mixtures of ligation products to ensure
reproducibility of results. This assay system therefore appears
to represent a true comparison of mismatch repair capabilities
of this oncogenic ‘hot spot’ by a mammalian cell line and by
a prokaryotic organism for each of the four mismatches tested.
Fidelity of cellular mismatch repair at a ‘hot spot’ of oncogenic
activation has been investigated using an Epstein–Barr virus
(EBV) plasmid vector containing genomic human H-ras DNA
(14). Specific base pair mismatches are located at the middle
nucleotide position of codon 12: either dG of the coding strand
or dC of the transcribed strand has been replaced with dA or
dT. Site- and strand-specific mismatches are constructed by
annealing complementary 30 base oligomers and surrounding
H-ras fragments to a ss M13ras plasmid to produce heteroduplex DNA that is then cleaved by restriction digestion and
religated into the remaining portion of the original plasmid
vector (Figure 1). This methodology does not facilitate hemimethylation d(GATC) directed mismatch repair by E.coli. It
is also unlikely that the initiating mechanism of mammalian
strand-directed mismatch repair is facilitated. However, this
procedure does produce a ‘pre-ligation’ total of four ‘nicks’
on the DNA strand containing the incorrect mismatched
nucleotide (each end of the 1.8 kb ds fragment plus each end
of the ss 30-mer mismatch oligo) versus two ‘nicks’ on the
opposite DNA strand (each end of the ds 1.8 kb fragment
only) during preparation of mismatched DNA (Figure 1). The
highly unlikely event of 100% in vitro ligation efficiency
would require DNA ligase to ligate each end of the ss 30-mer
oligonucleotide containing the incorrect base as well as both
strands of each end of the ds 1.8 kb fragment of DNA into
the remaining 13.5 kb portion of p220.pbc, for a total of six
ligations per plasmid (Figure 1, step 4). Therefore, this protocol
was determined to be appropriate for biasing the selection for
plasmids containing unligated ‘nicks’ on either end of the ss
30-mer containing the incorrect base and/or on the same DNA
strand on either end of the 1.8 kb insert. In addition, we have
repeatedly observed that this 15.3 kb plasmid has an ~50-fold
higher transfection efficiency in NIH 3T3 cells in the ds
circular form than when it is linear or gapped (missing 30mer oligo). We have been unable to transform E.coli successfully with this plasmid when it is in the linear or gapped form.
Mismatch repair analyses
Figure 3A illustrates HpaII restriction digestion and sequencing
results of ‘activated’ and ‘wild-type’ H-ras codon 12 PCR
amplified DNA. The PCR amplified 129 bp DNA band that
did not cut in the presence of HpaII (‘activated’) was sequenced
to reveal a GGC → GTC activation mutation at codon 12.
The PCR amplified DNA that was cleaved by HpaII (‘wildtype’) to yield an 88 bp and 41 bp band was sequenced to
reveal the expected GGC sequence at codon 12.
Figure 3B illustrates the results of a typical experiment in
which E.coli were transformed with p220.pbc1H/B containing
a codon 12, middle G:A mismatch. H-ras DNA from individual
carbenicillin resistant colonies was PCR amplified and cleaved
with HpaII. Lanes 1, 3, 4, 7 and 8 do not appear to be cleaved
by HpaII. Sequencing the amplified DNA from these colonies
reveals a GGC → GTC transversion mutation. Lanes 5 and
12 contain amplified DNA that appears to be mostly or
completely cleaved by HpaII. Subsequent sequencing revealed
the expected GGC wild type sequence. The remaining lanes
in Figure 3B contain DNA that is only partially cleaved by
HpaII. Subsequent sequencing revealed a mixture of GGC and
GTC at codon 12.
Figure 3C demonstrates sequencing results from a typical
mismatch experiment in which NIH 3T3 cells were transfected
with the same plasmid DNA preparations containing a G:A
mismatch. Sequences 3, 4, 5, 6, 7, 9 and 10 have been corrected
to wild-type (GGC). The remaining five sequences are G/T
mixtures at the middle nucleotide of codon 12, similar to lanes
2, 6, 9, 10 and 11 in Figure 3B. This could indicate that either
DNA replication has occurred before mismatch correction or
the presence of more than one plasmid per cell, with each
mismatch repaired differently. Either situation could result in
a mixture of wild-type and activated H-ras DNA. However,
because subsequent results during NIH 3T3 experiments
revealed a wide range of frequency of mixtures (0–65%),
depending on the mismatch, it is more probable that DNA
Fig. 2. NarI restriction cleavage and agarose electrophoresis to compare
control ds 1.8 kb XhoI/KpnI DNA fragments with G:A mismatch ds 1.8 kb
XhoI/KpnI DNA fragments. The two DNA bands, 413 bp and 274 bp,
represent the distance between NarI recognition sites nos 1280 and 1693
(codons 10/11 location) and between NarI recognition site no. 1693 and
KpnI site no. 1967. Exon 1 of H-ras spans nos 1664–1774. 1 kb ladder is
shown on the left for size determinations. Sequence diagram on right not
drawn to scale.
1314
Defective mismatch repair
Fig. 3. PCR amplification and HpaII restriction digestion of H-ras mismatch plasmid DNA after replication in NIH 3T3 cells and E.coli. (A) HpaII cleavage
and sequencing results of PCR amplified ‘activated’ (codon 12 GTC) and ‘wild-type’ (codon 12 GGC) control plasmid DNA after replication in NIH 3T3
cells. (B) HpaII cleavage and sequencing results of PCR amplified H-ras G:A mismatch plasmid DNA after replication in E.coli. (C) Sequencing results of
PCR amplified H-ras G:A mismatch plasmid DNA after replication in NIH 3T3 cells.
replication has occurred before repair in these mixtures, and
therefore competent cellular mismatch repair is dependent on
the specific type of mismatch at this location (Table II).
Table I (E.coli) and Table II (NIH 3T3) are compiled results
of several experiments examining frequency of correct repair
back to G:C, incorrect repair to A:T or T:A or putative
‘unrepair’ by the presence of mixtures of G:C and A:T or T:A
for four different mismatches (G:A, A:C, T:C, G:T) at the
human H-ras codon 12 ‘hot spot’ of mutation.
Table I contains results from a total of 306 DH5α colonies
and 23 NR9161 (-mutL) colonies. Of the four mismatches
examined after replication in DH5α, G:A mismatches were
repaired correctly back to G:C at the lowest rate of 87% (78/
90) and repaired incorrectly at the highest rate of 8% (7/90).
1315
L.Arcangeli et al.
Table I. E.coli mismatch repair of human H-ras codon 12 ‘hot spot’
Repaired to
Mismatch
(DH5α)a
Correctly repaired
(Mismatch → G:C)
(Total assayed)
Incorrectly repaired
(Mismatch → A:T or T:A)
(Total assayed)
Unrepaired
(Mismatch → Mixturesb)
(Total assayed)
(NR9161)c
Unrepaired
(Mismatch → Mixturesb)
(Total assayed)
G:A
A:C
T:C
G:T
87%
(78/90)
97%
(84/87)
100%
(69/69)
94%
(56/60)
8%
(7/90)
2%
(2/87)
0%
(0/69)
3%
(2/60)
5%
(5/90)
1%
(1/87)
0%
(0/69)
3%
(2/60)
n.d.
n.d.
100%
(7/7)
100%
(16/16)
aDH5α are wild-type E.coli
bMixtures are G:C and A:T
cNR9161
with no mismatch repair defect.
or T:A, depending on mismatch.
are mismatch repair defective E.coli (-mutL).
Table II. NIH 3T3 mismatch repair of human H-ras codon 12 ‘hot spot’
Repaired to
Correctly repaired
(Mismatch → G:C)
(Total assayed)
Incorrectly repaired
(Mismatch → A:T or T:A)
(Total assayed)
Unrepaired
(Mismatch → mixturesb)
(Total assayed)
Mismatch
G:A
A:C
T:C
G:T
Gappeda
35%
(24/69)
58%
(25/43)
80%
(36/45)
100%
(67/67)
100%
(23/23)
0%
(0/69)
0%
(0/43)
0%
(0/45)
0%
(0/67)
0%
(0/23)
65%
(45/69)
42%
(18/43)
20%
(9/45)
0%
(0/67)
0%
(0/23)
aGapped
results from transfection of complete circular p220.pbc1H/B
plasmid without 30-mer mismatch oligomer.
bMixtures are G:C and A:T or T:A, depending on mismatch.
G:A mismatches also had the highest rate at 5% (5/90) of
unrepaired mismatch before replication (resulting in a mixture
of wild-type and mutated end product). The other three
mismatches examined at this site (A:C, T:C and G:T) were
repaired correctly back to G:C by DH5α at overall high rates
of 97% (84/87), 100% (69/69) and 94% (56/60), respectively.
Although only T:C and G:T mismatches were used to transform
NR9161 (-mutL), 100% of the colonies examined contained
unrepaired H-ras DNA as evidenced by mixtures of wild-type
and mutated sequences at the codon 12 location.
Table II contains results of PCR amplified human H-ras
DNA from a total of 247 hygromycin resistant NIH 3T3
colonies. Only 35% (24/69) of G:A mismatches were repaired
correctly back to G:C with the remaining unrepaired. Only
58% (25/43) of A:C mismatches were repaired correctly back
to G:C, with the remaining unrepaired. A total of 80%
(36/45) of the T:C mismatches were repaired correctly back
to G:C, again with the remaining unrepaired. A total of 100%
(69/69) of G:T mismatches were repaired correctly back
to G:C. Also, 100% (23/23) of gapped plasmids (circular
1316
p220.pbc1H/B without 30-mer mismatch oligomer) were gapfilled/repaired correctly to produce a G:C base pair at codon 12.
An overall comparison of Tables I and II demonstrates that
for specific mismatches at this location, NIH 3T3 cells are
considerably less capable of performing mismatch repair than
mismatch repair competent DH5α. G:A mismatch repair occurs
with lowest fidelity of all four mismatches examined in both
organisms, although correct G:A mismatch repair to G:C in
DH5α (87%) is still significantly higher than in NIH 3T3
(35%). Interestingly, G:A and A:C mismatch repair in NIH
3T3 (35 and 58%, respectively) are the lowest repair rates of
all four mismatches at this location, with T:C (80%) and G:T
(100%) repair rates increasingly improved, respectively. The
repair rates of A:C (58%) and T:C (80%) are somewhat in
contrast to our previous report of A:C (87.5%) and T:C (67%)
repair rates, in which we used a different method to produce
the site- and strand-specific mismatch (14). It is not clear what
may have caused this difference in results, perhaps there are
differences in distribution of residual ‘nicks’ after construction
by each method. Current experiments indicate that NIH 3T3
cells do not appear to have any increased efficiency of
mismatch repair of transcriptional versus coding strand at this
site (G:T . T:C .. A:C .. G:A), but NIH 3T3 cells do
appear to have significantly decreased repair efficiency of
adenine (35%, 58%) as compared to thymine (80%, 100%),
regardless of which DNA strand contains the incorrect adenine.
Interestingly, in contrast to DH5α, when mismatch repair does
occur in NIH 3T3, it is consistently accurate repair back to
G:C, as evidenced by 0% mutation rates (incorrectly repaired)
for all mismatches tested in Table II (provided that mixtures
are truly indicative of unrepaired plasmids).
Discussion
The experimental system described in this paper has been
developed to study the frequency and accuracy of mammalian
DNA repair when a strand- and site-specific lesion is introduced
at a major ‘hot spot’ of mutation in the human H-ras oncogene.
The mammalian expression vector used for these studies
contains the entire human H-ras genomic sequence and a
hygromycin resistance gene for selection of mammalian clones
expressing this plasmid. In addition, this EBV vector replicates
synchronously with the cell cycle, maintains a low spontaneous
mutation rate, and a low plasmid copy number per cell (14).
Our methodology permits site-specific incorrect base insertion either on the transcribed or non-transcribed (coding)
strand. One limitation to our approach, however, has been our
inability to create strand-specific nicks as clearly defined as
in vitro systems currently in the literature (3–6,10,16). Instead,
we have relied on a strand-biasing approach that involves
placing four of the initial six nicks on the same DNA strand
as the incorrect base during plasmid preparation. Subsequent
in vitro ligation conditions are unlikely to approach 100%
efficiency, therefore an unligated nick in the same strand as
the incorrect base is highly favoured. This approach appears
feasible when observing results from DH5α transformation
experiments from several different plasmid and mismatch
preparations (Table I). Also, results obtained in this laboratory
agree with those of other investigators in regard to the high
rate and fidelity of strand-directed DH5α mismatch repair
(17,18). In addition, two separate mismatch experiments using
mismatch deficient NR9161 E.coli indicate that each plasmid
preparation contains the mismatch at the time of transformation,
Defective mismatch repair
as amplified H-ras DNA from each of these colonies remains
unrepaired.
In NIH 3T3 cells, however, aliquots of the same Hras codon 12 site- and strand-specific mismatch plasmid
preparations do not appear to be repaired with a similarly high
rate as DH5α, using this experimental system (Table II), with
the exception of G:T → G:C (100%). Table II also indicates
a wide variation in rate of specific mismatch repair observed
with NIH 3T3 cells as compared to DH5α. There could be
several reasons for these results. Currently, mammalian cellular
DNA repair mechanisms are not as well understood as E.coli
DNA repair. Therefore, while the strand-nick bias conditions
in these experiments appear sufficient for DH5α strand-directed
mismatch repair, the same conditions may not be sufficient
for NIH 3T3 strand-directed mismatch repair. We are currently
experimenting with different methods of constructing a more
specifically defined ‘nicked’ mismatch-containing plasmid to
resolve this issue. Interestingly, however, our current experiments indicate that when mismatch repair does occur in NIH
3T3 cells, fidelity of correct repair back to G:C is a consistent
event, as opposed to incorrect repair resulting in a mutation
(Table II). Therefore, regardless of the rate of mismatch repair,
strand-directed correct mismatch repair appears to play a
significant role in NIH 3T3.
Replication errors that are not corrected by proof-reading
processes during DNA replication are thought to be corrected
by strand-discriminating mechanisms of mismatch repair at a
stage of the cell cycle other than S phase. Cell cycle synchronization studies in progress in this laboratory should define
precisely differences in specific mismatch repair rates at this
oncogenic location, during different discrete stages of the
mammalian cell cycle. Additionally, NIH 3T3 cells have a
high rate of spontaneous transformation to the malignant
phenotype (19). Perhaps these cells have some, as yet, unidentified deficiency in specific mismatched base repair, contributing
to their mutator phenotype. To test this possibility, we are
currently assessing different mammalian cell lines for efficiency and fidelity of mismatch repair.
Codon 12 of H-ras is a well-known hot spot of mutagenesis in human tumors and animal model studies (20–22).
Mechanisms (or lack thereof) contributing to the increased
rate of mutation at this, and other precisely targeted activating
locations, are not yet known. One hypothesis to explain such
precise mutagenic targeting is that specific types of DNA
damage occur more frequently at certain sites in the genome
(21). A second hypothesis is that there is random DNA damage
in the genome with decreased repair of damage at specific
sites (13,22). A third hypothesis is that cells with certain
activating mutations are selected for increased survival (23).
It is possible that none of these hypotheses are mutually
exclusive. We are in the process of comparing frequency and
fidelity of mismatch repair at similar codons near codon 12 in
H-ras that are not frequent sites of mutation to determine if
the specific types of mismatch repair observed in these studies
are decreased precisely because they occur at the codon 12
location of H-ras.
Tables I and II indicate that both DH5α and NIH 3T3 have
increased difficulties specifically with the G:A mismatch at
this location, supporting results of other investigators that this
is one of the most difficult mismatches for all organisms to
repair (1,24). This may be because G:A is not easily recognized
as a mismatch, or because of the diverse physical conformations
this mismatch can assume within the DNA molecule, perhaps
confusing enzymatic recognition (25). A:C mismatches have
been noted for high fidelity and rate of correct mismatch repair
in other systems (11,16,24,25). Our current results, using NIH
3T3 cells, do not agree with others for a high rate of repair
of this particular mismatch at this oncogenic location (58%).
Most strikingly, the same A:C mismatch plasmid preparation
is repaired with high consistency by DH5α (97%). The
mismatch at this site that is most efficiently and accurately
repaired by NIH 3T3 is the G:T mismatch (100%), which is
in agreement with the results of other investigators (1,10,11,26).
This is probably a frequently occurring mismatch at this heavy
CpG region of DNA, due to endogenous deamination of 5methylcytosine, and therefore might have more than one
enzyme complex available for accurate repair (9,10). Likewise,
the T:C mismatch has been previously demonstrated also to
be repaired by a specific G:T mismatch enzyme, thymine DNA
glycosylase (9), in addition to strand-directed mismatch repair,
which may explain the higher rate of repair of this mismatch
in these experiments.
Perhaps the most interesting aspect of NIH 3T3 mismatch
repair in this study is the significant difference in repair rates
of mismatches containing an incorrect adenine (G:A and A:C)
versus thymine (T:C and G:T), regardless of strand placement
of the incorrect base. Clearly, a mismatch containing an
incorrect adenine at this site is most likely to undergo a
mutational event. Cell cycle synchronization experiments,
mismatch repair at other site-specific locations in exon 1 of
H-ras, and experiments examining the effect of a specific
strand-break either 39 or 59 of each mismatch, currently
ongoing in this laboratory, should prove interesting to compare
with the inefficiently repaired mismatches by NIH 3T3 in this
study. Similar mismatch studies using other cell lines are also
under way.
Results in Tables I and II argue that although codon 12,
middle G:C does not appear to be a strong ‘hot spot’ for
infidelity of bacterial mismatch repair, this location does appear
to be a ‘hot spot’ for lack of repair by NIH 3T3 cells containing
a G:A or A:C mispair, and possibly a T:C mispair. Each
mismatch, not correctly repaired, results in a T:A or A:T
‘activating’ mutation frequently observed at this site in
human tumours.
The results of this study clearly demonstrate that it is
possible to obtain information about mammalian mismatch
repair in vivo at a relevant genomic location involved with
cell transformation. Future results using this system should
contribute significantly towards a clearer understanding of
mechanisms of mutation and initiation of carcinogenesis.
Acknowledgement
This work was supported by NIH grant CA-57495 from the National
Cancer Institute.
References
1. Modrich,P. and Lahue,R. (1996) Mismatch repair in replication fidelity,
genetic recombination and cancer biology. Ann. Rev. Biochem., 65,
101–133.
2. Modrich,P. (1995) Mismatch repair, genetic stability and tumour avoidance.
Phil. Trans. Roy. Soc. Lond. B, 347, 89–95.
3. Umar,A., Boyer,J.C. and Kunkel,T.A. (1994) DNA loop repair by human
cell extracts. Science, 266, 814–816.
4. Risinger,J.I., Umar,A., Barrett,J.C. and Kunkel,T.A. (1995) A hPMS2
mutant cell line is defective in strand-specific mismatch repair. J. Biol.
Chem., 270, 18183–18186.
1317
L.Arcangeli et al.
5. Drummond,J.T., Li,G.-M., Longley,M.J. and Modrich,P. (1995) Isolation
of an hMSH2-p160 heterodimer that restores DNA mismatch repair to
tumor cells. Science, 268, 1909–1912.
6. Li, G.-M. and Modrich,P. (1995) Restoration of mismatch repair to nuclear
extracts of H6 colorectal tumor cells by a heterodimer of human MutL
homologs. Proc. Natl Acad. Sci. USA, 92, 1950–1954.
7. Stephenson,C. and Karran,P. (1989) Selective binding to DNA base
pair mismatches by proteins from human cells. J. Biol. Chem., 264,
21177–21182.
8. Griffin,S., Branch,P., Xu,Y.-Z. and Karran,P. (1994) DNA mismatch binding
and incision at modified guanine bases by extracts of mammalian cells:
implications for tolerance to DNA methylation damage. Biochemistry, 33,
4787–4793.
9. Neddermann,P. and Jiricny,J. (1993) The purification of a mismatchspecific thymine-DNA glycosylase from HeLa cells. J. Biol. Chem., 268,
21218–21224.
10. Palombo,F., Gallinari,P., Iaccarino,I., Lettieri,T., Hughes,M., D’Arrigo,A.,
Truong,O., Hsuan,J.J. and Jiricny,J. (1995) GTBP, a 160-kilodalton protein
essential for mismatch-binding activity in human cells. Science, 268,
1912–1914.
11. Brown,T.C. and Jiricny,J. (1988) Different base/base mispairs are corrected
with different efficiencies and specificities in monkey kidney cells. Cell,
54, 705–711.
12. Mellon,I., Rajpal,D.K., Koi,M., Boland,C.R. and Champe,G.N. (1996)
Transcription-coupled repair deficiency and mutations in human mismatch
repair genes. Science, 272, 557–560.
13. Bohr,V.A. (1995) DNA repair fine structure and its relations to genomic
instability. Carcinogenesis, 16, 2885–2892.
14. Arcangeli,L. and Williams,K.J. (1995) Mammalian assay for site-specific
DNA damage processing using the human H-ras proto-oncogene. Nucl.
Acids Res., 23, 2269–2275.
15. Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning: A
Laboratory Manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold
Spring Harbor, NY.
16. Fang,W.-H. and Modrich,P. (1993) Human strand-specific mismatch repair
occurs by a bidirectional mechanism similar to that of the bacterial
reaction. J. Biol. Chem., 268, 11838–11844.
17. Langle-Rouault,F., Maenhaut-Michel,G. and Radman,M. (1987) GATC
sequences, DNA nicks and the MutH function in Escherichia coli mismatch
repair. EMBO J., 6, 1121–1127.
18. Lahue,R.S., Su,S.-S. and Modrich,P. (1987) Requirement for d(GATC)
sequences in Escherichia coli mutHLS mismatch correction. Proc. Natl
Acad. Sci. USA, 84, 1482–1486.
19. Rubin,H. and Xu,K. (1989) Evidence for the progressive and adaptive
nature of spontaneous transformation in the NIH 3T3 cell line. Proc. Natl
Acad. Sci. USA, 86, 1860–1864.
20. Barbacid,M. (1987) ras genes. Ann. Rev. Biochem., 56, 779–827.
21. Bos,J.L. (1989) ras oncogenes in human cancer: a review. Cancer Res.,
49, 4682–4689.
22. Balmain,A. and Brown,K. (1988) Oncogene activation in chemical
carcinogenesis. Adv. Cancer Res., 51, 147–182.
23. Guerrero,I. and Pellicer,A. (1987) Mutational activation of oncogenes in
animal model systems of carcinogenesis. Mutat. Res., 185, 293–308.
24. Brown,T.C. and Jiricny,J. (1989) Repair of base-base mismatches in simian
and human cells. Genome, 31, 578–582.
25. Brown,T. (1995) Mismatches and mutagenic lesions in nucleic acids.
Aldrichimica Acta, 28, 15–20.
26. Brown,T.C. and Jiricny,J. (1987) A specific mismatch repair event protects
mammalian cells from loss of 5-methylcytosine. Cell, 50, 945–950.
Received on July 22, 1996; revised on March 14, 1997; accepted on March
17, 1997
1318