ABSTRACT
In order to understand the action of the chemotherapeutic drug cisplatin, it is necessary to determine why some types of cisplatin-DNA intrastrand crosslinks are repaired better than others. Using cell
extracts and circular duplex DNA, we compared nucleotide excision repair of
uniquely placed 1,2-GG, 1,2-AG, and 1,3-GTG cisplatin-crosslinks, and a 2-acetylaminofluorene lesion. The 1,3 crosslink and
the acetylaminofluorene lesion were repaired by normal cell extracts
~
15-20 fold better than the 1,2 crosslinks. No evidence was found for
selective shielding of 1,2 cisplatin crosslinks from repair by cellular
proteins. Fractionation of cell extracts to remove putative shielding proteins
did not improve repair of the 1,2-GG crosslink, and cell extracts did not selectively inhibit access of
UvrABC incision nuclease to 1,2-GG crosslinks. The poorer repair of 1,2 crosslinks in comparison to the
1,3 crosslink is more likely a consequence of different structural alterations
of the DNA helix. In support of this, a 1,2-GG-cisplatin crosslink was much better repaired when it was opposite
one or two non-complementary thymines. Extracts from cells defective in the hMutS
[alpha]
mismatch binding activity also showed preferential repair of the 1,3 crosslink
over the 1,2 crosslink, and increased repair of the 1,2 adduct when opposite
thymines, showing that hMutS
[alpha]
is not involved in the differential NER of these substrates
in vitro
. Mismatched cisplatin adducts could arise by translesion DNA synthesis, and
improved repair of such adducts could promote cisplatin-induced mutagenesis in some cases.
Cisplatin is used with varying success for the treatment of human cancers.
Although 90% of testicular cancers can be cured by cisplatin chemotherapy, a
significant problem is the development of resistant tumours (
1
,
2
). The basis for the therapeutic effectiveness of cisplatin is not fully
understood but its cytotoxic action against tumour cells is thought to be
mediated through the formation of cisplatin-DNA adducts which may inhibit DNA replication and/or transcription (
3
). Cisplatin forms primarily 1,2-intrastrand crosslinks between adjacent purines in DNA, and also
introduces other adducts including 1,3 crosslinks, interstrand crosslinks and
monoadducts. The main mechanism for removing intrastrand crosslinks is
nucleotide excision repair (NER), but the efficiency of removal varies among
different intrastrand crosslinks both
in vivo
(
4
,
5
) and
in vitro
(
6
-
9
). NER in human cells involves recognition of damage by factors that include the
XPA and RPA proteins, incision by the structure-specific endonucleases XPG on the 3' side of a lesion and ERCC1/XPF on the 5' side, and repair DNA synthesis mediated by a PCNA-dependent DNA polymerase (
10
,
11
).
In order to better understand the toxic and mutagenic properties of cisplatin,
it is important to investigate the reasons for the differential repair of
intrastrand cisplatin-DNA crosslinks. Several explanations have been proposed. One hypothesisis
is that the specific binding of cellular proteins, notably HMG-box proteins, to 1,2-intrastrand cisplatin-DNA crosslinks could shield these lesions from recognition by
NER (
3
,
12
). Alternatively, 1,2 crosslinks might titrate essential DNA-binding proteins such as the transcription factor hUBF away from their
natural sites of action (
13
). The simplest explanation, however, is that the extent of DNA structural
alteration caused by a particular lesion dictates the efficiency of damage
recognition (
6
,
14
). We report here that a more distorting 1,3-intrastrand d(GpTpG)-cisplatin crosslink is repaired much more readily by human whole
cell extracts than a less distorting 1,2-intrastrand d(GpG)-cisplatin crosslink. Furthermore, placement of non-complementary thymine residues opposite the platinated
guanines of a single 1,2-GG crosslink increases the efficiency of nucleotide excision repair of the
adduct. These results argue that the specific structural alterations of the
helix caused by intrastrand cisplatin-DNA crosslinks are the primary determinants of damage recognition and
repair efficiency for the human NER machinery. Furthermore, the processing of
1,2-intrastrand cisplatin-DNA crosslinks in cells by mutagenic translesion DNA replication may
modulate the extent to which 1,2-intrastrand cisplatin-DNA crosslinks are removed from DNA. Improved repair of a cisplatin
crosslink after misincorporation of a base opposite may facilitate fixation of
the mutation in some instances.
Whole cell extracts were prepared as described previously (
15
) from the following cell lines: HeLa S3, CHO-9, GM2249 (XP-C), GM2485 (XP-D), CHO 27-1 (XP-B/ERCC3). Fractionation of HeLa whole cell
extracts and preparation of purified XPA, TFIIH (Heparin-5PW fraction) and XPC-HHR23B proteins was as described previously (
16
-
18
).
Oligonucleotides containing a single 1,3-intrastrand d(GpTpG)-cisplatin crosslink, 1,2-intrastrand d(GpG)-cisplatin crosslink and a single 2-acetylaminofluorene adduct were prepared as
described previously (
6
,
7
,
19
). The 25mer oligonucleotide 5'-TCTTCTTCTCTAGTACTCTTCTTCT-3' containing a 1,2- intrastrand d(ApG)-cisplatin crosslink was synthesized in a
similar way and was a kind gift of K. J. Yarema and J. M. Essigmann (MIT).
Covalently closed circular DNA containing a single
1,3-intrastrand d(GpTpG) cisplatin crosslink (Pt-GTG or Pt-GTGx), a single 1-2-intrastrand d(GpG)-cisplatin crosslink (Pt-GG), a single 1,2-intrastrand d(ApG)-cisplatin crosslink (Pt-AG) and a single 2-acetylaminofluorene (AAF-G)
adduct were produced by priming the plus strand of M13mp18GTG (
20
) or M13mp18GTGx (
7
), M13mp18GG (
6
), M13mp18AG and M13mp18AAF (
19
) respectively, with appropriate single lesion oligonucleotides as described (
6
,
7
,
19
). Control DNA substrates (Con-GTG or Con-GTGx, Con-GG, Con-AG and Con-AAF) were produced by the same method using the
appropriate non-damaged oligonucleotides. The vector M13mp18AG was constructed by
replacing the 38 base pair (bp)
Eco
RI-
Sal
I
fragment of M13mp18 with a sequence formed by annealing the oligonucleotides 5'-AATTCCTGGAGAAGAGAGTACTAGAGAAGAAGACCTGG-3' and 5'-TCGACCAGGTCTTCTTCTCTAGTACTCTTCTTCTCCAGG-3'.
Covalently closed circular DNA containing a single 1,2-intrastrand d(GpG)-cisplatin crosslink opposite either TpC (Pt-GG.TC), TpT (Pt-GG.TT) or CpT (Pt-GG.CT) in the direction 3' to 5' on the non-damaged DNA strand were
produced as described above using the plus strand of M13mp18GG.TC, M13mp18GG.TT
and M13mp18GG.CT respectively. These M13 vectors were formed by replacing the
179 bp
Eco
RI-
Pvu
I restriction fragment of M13mp18GG with DNA duplexes obtained by PCR
amplification of the replicative form of M13mp18GG using the primers 5'-AACTGTTGGGAAGGGCGATCGGTGCGGGCCTCTTCGCTAT-3' and either 5'-CCATGATTACGAATTCCTGGAGAAGAAGAAGGCTTAGAAG-3' (for M13mp18GG.TC), 5'-CCAT-
GATTACGAATTCCTGGAGAAGAAGAAGGTTTAGAAG-3' (for M13mp18GG.TT) or 5'-CCATGATTACGAATTCCTGGAGAAGAAGAAGGTCTAGAAG-3' (for M13mp18GG.CT).
Reaction mixtures (50 [mu]l, or multiples thereof) contained 100-300 ng of the appropriate single lesion or control DNA substrate and
100-200 [mu]g whole cell extract (
15
) in a buffer containing 45 mM HEPES-KOH (pH 7.8), 70 mM KCl, 7.5 mM MgCl
2
, 0.9 mM DTT, 0.4 mM EDTA, 2 mM ATP, 5-20 [mu]M each of dATP, dCTP, dGTP and TTP, 40 mM phosphocreatine (di-Tris salt), 2.5 [mu]g creatine phosphokinase (type I, Sigma), 3.4% glycerol and 18 [mu]g bovine serum albumin. Cell extract was preincubated with
buffer at 30oC for 5 min, DNA substrate added, and reactions incubated at 30oC for the times indicated.
For repair synthesis assays 2 [mu]Ci [[alpha]-
32
P]dATP (3000 Ci/mmol), 2 [mu]Ci [[alpha]-
32
P]dCTP (3000 Ci/mmol) and 2 [mu]Ci [[alpha]-
32
P]TTP (3000 Ci/mmol) were added to reaction mixtures. Purified DNA was digested
in a 30 [mu]l volume with 15 U
Bst
NI (New England Biolabs) at 60oC for 4 h prior to electrophoresis in non-denaturing 12% polyacrylamide gels. Phosphorimager (Molecular
Dynamics) counts were used to calculate repair synthesis, taking account of the
base composition of the damaged DNA strand of each restriction fragment.
Incision assays using Pt-GTGx DNA were as described previously (
7
,
22
). Briefly, DNA purified from incision reactions was cleaved with
Hin
dIII and
Xho
I prior to separation on denaturing 12% polyacrylamide gels. DNA was then
transferred to a nylon membrane (Hybond-N+, Amersham) by capillary transfer for 90 min. Fixed membranes were
incubated for 16 h at 42oC in hybridisation buffer containing 7% SDS, 10% PEG 8000, 250 mM NaCl, 130
mM potassium phosphate buffer (pH 7.0) and 100 pmol of the 27mer
oligonucleotide 5'-GAAGAGTGCACAGAAGAAGAGGCCTGG-3' labelled with [
32
P] at the 5' terminus. Membranes were washed twice for 10 min in 1* SSC, 0.1% SDS before exposure to X-ray film with intensifying screens or a phosphorimager
screen. Incision assays using Pt-GG DNA (and derivatives Pt-GG.TC, Pt-GG.TT and Pt-GG.CT containing mispaired thymines on the non-damaged DNA strand) were performed as described
above except that purified DNA was incised with
Hin
dIII and
Eco
RI prior to denaturing gel electrophoresis and hybridisation was performed using
a
32
P-labelled 25mer oligonucleotide 5'-GAAGAAGGCCTAGAAGAAGACCTGG-3' (Probe 1) complementary to the sequence of the
damaged DNA strand flanking the 1,2-intrastrand d(GpG)-cisplatin crosslink. Analysis of the non-damaged DNA strand of Pt-GG DNA substrates was performed using a
32
P-labelled 24mer oligonucleotide 5'-TCTTCTTCTAGGCCTTCTTCTTCT-3' (Probe 2).
Purified
Escherichia coli
UvrA, UvrB, and UvrC proteins were used as described (
6
) to incise damaged DNA at the site of the lesion. For pre-incubation with UvrABC, 100 ng DNA was incubated in reaction buffer (in
the absence of cell extract) for 15 min at 30oC with 36 nM UvrA and 29 nM UvrB. UvrC was then added to 3.4 nM and the
mixture was incubated for an additional 15 min. Finally, 200 [mu]g of cell extract protein was added and the reactions were incubated for 3 h
at 30oC. Alternatively, the same amounts of UvrA, UvrB and UvrC proteins were
added after 15 min pre-incubation with the cell extract.
Circular duplex DNA substrates containing a single 1,2-intrastrand d(GpG)-cisplatin crosslink (Pt-GG), a single 1,2-intrastrand d(ApG)-cisplatin crosslink (Pt-AG), a single 1,3-intrastrand d(GpTpG)-cisplatin crosslink (Pt-GTG) or a single 2-acetylaminofluorene
adduct (AAF-G) were constructed (Fig.
1
a). Each adduct is located in a unique restriction endonuclease target site and
resistance to cleavage by the appropriate restriction enzyme is diagnostic for
the presence of the adduct (
7
,
23
). DNA substrates containing single lesions were refractory to cleavage by the
relevant enzyme (Fig.
1
b). The small amount of nicked circular and linear DNA formed during these
reactions indicates either that some unmodified DNA is present in the
preparations, or that a restriction enzyme is occasionally able to cleave one
or both DNA strands at the site of the adduct. This background nicking was most
evident after incubation of DNA containing a single 1,2-intrastrand d(ApG)-cisplatin crosslink with
Sca
I. This is the only case where one of the phosphodiester bonds cleaved by the
restriction enzyme is not immediately adjacent to an adducted base (Fig.
1
a), which probably explains why this particular modified DNA is cleaved more
efficiently. Control substrates lacking adducts were completely linearised by
the diagnostic restriction endonucleases. Both adducted and non-modified DNA substrates were completely linearised by
Sal
I at a target site ~15 bp away from the lesions (Fig.
1
b). The presence of a single lesion on each DNA substrate was further confirmed
by primer extension analysis of the damaged DNA strand [data not shown, see (
7
)].
It has been proposed that 1,2-intrastrand cisplatin-DNA crosslinks are poorly repaired because they may be shielded from
repair enzymes by cellular proteins which can bind to these adducts (
6
,
12
,
24
). Fractionation of human cell extracts should deplete or remove putative
shielding proteins. We tested whether the use of fractions instead of whole
cell extracts would increase the repair efficiency of Pt-GG DNA relative to Pt-GTG DNA. HeLa whole cell extract was fractionated (Fig.
3
a) as described previously (
17
) and Pt-GTG or Pt-GG DNA were incubated with various combinations of cellular
fractions and purified repair proteins (Fig.
3
b). We previously showed that fractions IIa, IIc, and IV were unnecessary for
repair, that fraction I (FI) could be replaced with purified RPA and PCNA
proteins, and that fraction IId could be replaced with purified XPG protein (
17
). However, the poor repair of Pt-GG DNA relative to Pt-GTG DNA by HeLa whole cell extract (lanes 1 and 2) did not improve
after completely omitting either FI (lanes 3 and 4) by replacement with RPA and
PCNA, FI and FIV (lanes 9 and 10), IIa and FIV (lanes 15 and 16) or IIa, FIV
and IId (lanes 17 and 18) by replacing IId with XPG. The difference in repair
efficiency between Pt-GTG and Pt-GG remained at least 15-fold in all combinations tested. It is possible that some
shielding proteins were present in Fractions IIb or III, but these results make
it more likely that the repair of the 1,2-intrastrand cisplatin-DNA crosslink is inefficient because it is an intrinsically poorer
substrate for human nucleotide excision repair proteins.
To obtain an independent measure of the repair efficiency of Pt-GTG and Pt-GG DNA, the extent of dual incisions occuring at each of these
lesions was measured by the formation of platinated oligonucleotides (Fig.
5
). Dual incision of Pt-GTG DNA produced a characteristic pattern (
7
) of 24-30mer platinated oligonucleotides (Fig.
5
, lane 3). To obtain a comparable signal intensity from the oligonucleotides
formed during dual incision of Pt-GG DNA, it was found necessary to pool the products from 20 incision
reactions (Fig.
5
, lane 1). Purified DNA from one incision reaction is shown for Pt-GTG DNA (Fig.
5
, lane 3). The formation of 24-30mer platinated oligonucleotides containing the
1,2-intrastrand d(GpG)-cisplatin crosslink was ~15-fold less efficient than observed for the 1,3-intrastrand d(GpTpG)-cisplatin crosslink, consistent with the difference in
repair synthesis measured for these two lesions.
Structural studies (
25
-
28
) provide good evidence that the 1,3-GTG lesion distorts duplex DNA more than a 1,2-GG adduct. This distortion includes significant unwinding of the DNA
helix, by 23o for a 1,3-intrastrand d(GpTpG)-cisplatin crosslink in comparison to 13o for a 1,2-intrastrand d(GpG)-cisplatin crosslink (
25
). We postulated that the placement of mispaired thymines on the non-damaged DNA strand directly opposite the platinated guanines of a 1,2-intrastrand cisplatin-DNA crosslink might lead to increased structural distortion
and improve the nucleotide excision repair efficiency of this DNA lesion.
Three additional circular duplex M13 DNA substrates were synthesized, containing
a single 1,2-intrastrand d(GpG)-cisplatin crosslink but with thymines in the non-damaged DNA strand opposite either the 5' platinated guanine (Pt-GG.TC), the 3' platinated guanine (Pt-GG.CT), or both platinated guanines
(Pt-GG.TT) (Fig.
6
a). Incubation of the mispaired Pt-GG DNA substrates with HeLa cell extract generated incision products of
the same size range and pattern as was observed for Pt-GG DNA in Figure
5
(Fig.
6
b, left, lanes 3, 5 and 7) with 27mers predominating. However, repair of the
mismatched crosslinks was substantially more efficient than repair of the Pt-GG DNA substrate. A similar result was observed using CHO-9 cell extract (Fig.
6
c).
Recent studies have shown that human proteins that bind mismatches in DNA also
have some affinity for cisplatin-DNA adducts. A 1,2-intrastrand d(GpG)-cisplatin crosslink with complementary cytosine bases on the
non-damaged strand was bound by the hMutS[alpha] complex consisting of hMSH2 and hMSH6/GTBP (
30
), and hMSH2 alone may also recognise this adduct (
31
). A direct comparison among several mismatched platinated molecules shows that
hMutS[alpha] recognises duplex DNA containing a 1,2-intrastrand d(GpG)-cisplatin crosslink with much higher affinity when the non-damaged DNA strand contains a mispaired thymine
opposite the 3' platinated guanine (
32
). To test whether extracts defective in hMutS[alpha] have altered NER of such adducts, we investigated repair of normally
paired and mispaired 1,2-intrastrand d(GpG)-cisplatin crosslinks by extracts from DLD-1 and LoVo cells, which are defective in the hMSH6/GTBP and
hMSH2 subunits of hMutS[alpha], respectively.
A dual incision assay (Fig.
8
a) showed that HeLa, DLD-1 and LoVo extracts incised Pt-GG DNA poorly (lanes 1, 5 and 9), but that Pt-GG.TC, Pt-GG.TT and Pt-GG.CT DNA substrates were incised significantly
more efficiently (lanes 2-4, 6-8 and 10-12). Pt-GTGx DNA was incised with an efficiency similar to
the mispaired substrates by all three extracts (data not shown). Only small and
non-systematic differences (2-5-fold) were observed in the efficiency of dual incisions in
the three mispaired substrates. These varied from 2- to 6-fold in further experiments using two separate preparations of each
platinated DNA substrate (data not shown; see also Fig.
8
c, lanes 3-5). We conclude that one or two non-complementary T residues opposite the intrastrand cisplatin
crosslink increases
in vitro
repair of this lesion, and that hMutS[alpha] is not involved in this effect. In separate experiments, extracts from
the hMLH1-defective HCT116 cell line also showed increased NER of the mispaired
substrates (data not shown).
Repair of these DNA substrates by both HeLa (Fig.
8
b) and LoVo cell extracts (Fig.
8
c) was also analysed by monitoring repair synthesis. Damage dependent repair
synthesis by HeLa cell extract in the small
Bst
NI restriction fragment containing the adduct for Pt-GTGx and the three mispaired Pt-GG DNA substrates was >17 fold higher than that seen for Pt-GG DNA (Fig.
8
b). The same relative efficiency of repair was observed for these DNA substrates
in LoVo cell extract (Fig.
8
c, lanes 1-5). We note that the background DNA synthesis in the flanking 57, 68, 99,
127 and 139 bp restriction fragments was less for both Pt-GG and Con-GG DNA relative to the DNA substrates containing mispairs (compare
Fig.
8
b lanes 2 and 7 with lanes 3-5 and 8-10 respectively). Moreover, reactions using LoVo cell extract
showed considerably less synthesis in these fragments than HeLa cell extract
(compare lanes 3-5 and lanes 8-10 in Fig.
8
b and c). This G.T mismatch-dependent and hMSH2-dependent background synthesis might represent a low level of
bidirectional mismatch repair synthesis (
33
) initated at nicks formed during incubation with cell extract.
The structural distortion caused by a particular lesion in DNA may be influenced
by the local nucleotide sequence and the topology of the substrate. To minimise
effects of sequence, three cisplatin-DNA adducts were built into a circular duplex within the same DNA sequence
except for a 6 bp region encompassing the adduct (Fig.
1
a). We found by measuring repair synthesis that a 1,2-intrastrand d(GpG) or d(ApG)-cisplatin crosslink was repaired by HeLa cell extract 15-20-fold less efficiently than a 1,3-intrastrand d(GpTpG)-cisplatin crosslink in this sequence
context. A similar difference of repair efficiency between the 1,2 and 1,3
crosslinks was found by measuring the formation of 3' and 5' incisions. Thus, the differential NER efficiency of these lesions
is mediated by reaction steps preceding or coincident with endonucleolytic
cleavage.
The possibility that cellular proteins may shield 1,2-intrastrand cisplatin-DNA crosslinks from NER enzymes has been raised previously (
6
,
12
,
24
). Cellular factors that bind to the structural distortions caused by 1,2-intrastrand d(GpG)- and d(ApG)-cisplatin crosslinks have been detected in human cell extracts
(
12
,
34
-
37
), notably several proteins containing HMG-box motifs (
38
-
41
). None of the HMG-box proteins tested were able to bind a 1,3-intrastrand d(GpTpG)-cisplatin crosslink. Addition of a large excess of HMG protein
to repair reactions can inhibit repair of a 1,2 but not a 1,3 crosslink (
24
). This effect might be caused by direct shielding, or by HMG-induced changes that make the DNA structure less favourable for repair by
human NER (
3
,
42
). However, in the experiments reported here we found no evidence that proteins
at normal levels in whole cell extracts play any significant role in shielding
1,2 crosslinks from repair. Use of fractionated cell extracts and purified
proteins (to deplete or remove putative cisplatin adduct binding factors) did
not change the relative repair of 1,2 and 1,3 crosslinks. Moreover, incision of
either type of crosslink by
E.coli
UvrABC repair endonuclease was not blocked by cell extract proteins.
This conclusion is consistent with a recent report from Zamble
et al.
(
8
) that 1,2-cisplatin crosslinks are less efficiently repaired by cell extracts than
1,3-crosslinks and that the relative repair of these lesions is unchanged even
when purified NER proteins are used in the dual incision reaction. Huang
et al
. (
24
) originally reported that both 1,2 and 1,3 crosslinks were well repaired, but
the later study (
8
) concluded that the 1,3 crosslink was consistently repaired about 3-fold better than the 1,2 crosslink. We found a larger difference (15-20-fold) in the present experiments but there are some
differences between the design of the studies. Apart from a 12 bp core, the DNA
sequences flanking the lesions are dissimilar. Moreover, Zamble
et al.
used linear 156 bp DNA as substrate while we used 7.3 kbp closed circular DNA.
We tested the effect of linearizing Pt-GTG and Pt-GG DNA with
Ava
I (Fig.
2
a) prior to incubation with HeLa cell extract, but found no significant
difference in repair efficiency compared to circular DNA (data not shown). This
suggests that the topology of M13 duplex DNA does not strongly influence the
differential repair of these lesions, although the length of flanking DNA
sequences may be important. Calsou
et al.
also concluded that 1,2 crosslinks in cisplatin-damaged DNA are less well repaired than other less abundant adducts (
9
).
Our data support the most likely explanation for the better repair of 1,3 over
1,2 cisplatin crosslinks, which is that the former adducts are better
recognized by the human NER system because they cause more distortion of the
DNA structure. By analogy, amongst UV-induced DNA damage products the more distorting (6-4) photoproducts are repaired ~10-fold better than cyclobutane pyrimidine dimers (
19
), a difference that parallels the high discrimination of the damage-binding protein XPA for (6-4) photoproducts over cyclobutane dimers (
16
).
The effect of further structural alteration at the site of the lesion was
analysed by modifying the non-damaged DNA strand to contain mispaired thymine residues opposite one or
both of the platinated guanines of the 1,2-intrastrand d(GpG)-cisplatin crosslink. G.T mispairs cause a structural change of the
DNA helix (
43
) and the mispairing of platinated guanines with thymine bases could lead to a
greater structural distortion more favourable for NER damage recognition.
Consistent with this, 1,2-intrastrand d(GpG)-cisplatin crosslinks were 17-fold more efficiently repaired by HeLa cell extracts when
opposite one or two T residues instead of C residues.
We previously proposed that the precise incision locations vary for different
types of DNA lesions whilst the distance between the two incision sites remains
relatively constant (
7
). A 1,3-intrastrand d(GpTpG) crosslink was incised predominantly at the 8th and
9th phosphodiester bonds 3' and the 16th, 19th and 20th bonds 5' to the lesion, releasing 24-32mers with 26mers as the most predominant product (
7
). We now find that the 1,2-intrastrand d(GpG)-cisplatin crosslink is repaired with a similar but distinct incision
pattern that gives rise to 27mers as the predominant excision product. Primer
extension mapping together with measurement of the sizes of platinated
oligonucleotides representing uncoupled 3' incisions has revealed several preferred sites of cleavage around Pt-GG.TC, Pt-GG.TT and Pt-GG.CT DNA (5th, 7th, 8th, 10th-12th phosphodiester bonds 3' to the lesion and the 15th, 18th and 19th
phosphodiester bonds 5' to the lesion; data not shown). Thus, the incision positions around both
the 1,2- and the 1,3-intrastrand cisplatin crosslinks analysed here are different from
each other and from other lesions including a thymine dimer (
44
). The distance between the 3' and 5' incisions is, however, a constant factor for all of these types
of damage.
There are interesting implications of the fact that the incision pattern is the
same for both the correctly base-paired and mispaired 1,2-intrastrand d(GpG)-cisplatin crosslinks. Firstly, the sequences flanking the 1,2-intrastrand d(GpG)-cisplatin crosslink are not intrinsically
refractory to cleavage by the NER endonucleases. Moreover, this suggests that
the positioning of the 3' and 5' structure-specific endonucleases, XPG and XPF-ERCC1 respectively on the damaged DNA strand, probably
in an open intermediate (
45
), is not strongly dependent on the DNA sequence of the non-damaged strand opposite the DNA lesion. The incision positions may be
determined by the nature of the DNA adduct itself, whilst particular structural
alterations of the DNA helix caused by the adduct determine the efficiency of
damage recognition. A precedent for this feature of damage recognition can be
found in a recent study of
E.coli
UvrABC endonuclease in which the incision positions around a 2-acetylaminoflourene adduct did not change when present in either duplex
DNA or DNA mimicking a slipped DNA intermediate in which the bases opposite the
lesion are not complementary (
46
).
The finding that mispaired 1,2-intrastrand cisplatin-DNA crosslinks are removed more efficiently than the same DNA lesion
opposite a complementary DNA strand may have implications for cisplatin-damage processing within cells. The 1,2-intrastrand d(GpG)-cisplatin crosslinks are the most abundant lesions formed both
in vitro
and
in vivo
by cisplatin (
3
). Many 1,2-intrastrand d(GpG)-cisplatin crosslinks are known to undergo replication bypass in
cells treated with cisplatin (
47
,
48
). A recent study demonstrated that fork-like DNA templates containing cisplatin GG lesions undergo translesion DNA
synthesis very efficiently in mammalian cell extracts (
49
).
In cases where 1,2-intrastrand cisplatin-DNA crosslinks do undergo bypass with misincorporation of a base
opposite one of the adducted guanines, our results suggest that the platinum
crosslink will be more readily removed in a subsequent NER event. Repair
synthesis to fill in the gap after excision of the platinated oligonucleotide
will copy the former mismatch, fixing the mutation in both strands. The
consequence is that nucleotide excision repair would promote mutagenesis in
this situation. On the other hand, it is clear that NER normally acts to
decrease mutation frequency in damaged DNA, because NER-defective mammalian cells show a much increased frequency of mutation by
various DNA damaging agents (
50
,
51
), including cisplatin treatment (
52
). However, increased repair of mismatched adducts might significantly influence
mutagenesis at particular sites that are efficiently bypassed in an error-prone manner.
There are indications that the mismatch repair and NER pathways interact with
one another in mammalian cells. The development of cisplatin resistance in some
cases has been associated with specific defects in mismatch repair that result
in microsatellite instability (
53
,
54
). A possible explanation is that binding of mismatch repair proteins to
cisplatin lesions (
30
,
31
), particularly when opposite non-complementary bases (
32
), could interfere with NER. In this view, loss of mismatch repair proteins
might improve NER and increase resistance to cisplatin. The present study shows
that NER is in fact much more effective when a 1,2 cisplatin intrastrand
crosslink is opposite one or two non-complementary thymines. The mechanism of this increased repair does not
involve mismatch recognition proteins, because the same difference in NER was
also found with extracts from mismatch repair-deficient cells. Nevertheless, active mismatch repair in cells could
conceivably compete with NER if a mismatch occurs across from an adduct. The
present experiments have not directly explored this possibility, but such
studies would be feasible. In certain special instances, such as preferential
repair of transcribed DNA strands, an interaction of the mismatch repair system
and NER may even be cooperative. Mellon and co-workers (
55
) found defects in the removal of UV lesions from the transcribed strand of
active genes in mismatch-repair defective cells. If an
in vitro
system is developed to simultaneously carry out transcription-coupled NER and mismatch repair, the relevant mechanism can be explored.
We are very grateful to L. Grossman (Johns Hopkins University) for the
E.coli
UvrABC proteins, J.-M. Egly (IGBMC, Strasbourg) for the TFIIH sample, and K. J. Yarema and J.
M. Essigmann for providing some of the platinated oligonucleotides used in this
study. We thank the members of our laboratory for help, Elizabeth Evans and
Rafael Ariza for comments on the manuscript, and the ICRF Cell Production
department for assistance.
REFERENCES
Return