ABSTRACT
Previous work showed that melphalan-induced mutations in the
aprt
gene of CHO cells are primarily transversions and occur preferentially at G-G-C sequences, which are potential sites for various bifunctional
alkylations involving guanine N-7. To identify the DNA lesion(s) which may be responsible for these
mutations, an end-labeled DNA duplex containing a frequent site of melphalan-induced mutation in the
aprt
gene was treated with melphalan, mechlorethamine or phosphoramide mustard. The
sequence specificity and kinetics of formation of both interstrand and intrastrand crosslinks were determined. All mustards selectively formed two base-staggered interstrand crosslinks between the 5
'
G and the G opposite C in the 5
'
G-G-C sequence. Secondary alkylation was much slower for melphalan than
for the other mustards and the resulting crosslink was more stable.
Mechlorethamine and phosphoramide mustard induced intrastrand crosslinks between the two contiguous Gs in the G-G-C sequence in double-stranded DNA, but melphalan did not. Molecular dynamic
simulations provided a structural explanation for this difference, in that the
monofunctionally bound intermediates of mechlorethamine and phosphoramide mustard assumed thermodynamically stable conformations with the second arm in a position appropriate for intrastrand crosslink
formation, while the corresponding melphalan monoadduct did not.
The nitrogen mustards are among the oldest cancer chemotherapeutic agents and remain the drugs of choice for treatment of many human cancers,
including Hodgkin's disease, multiple myeloma, Burkitt's lymphoma and, in combination chemotherapies, various carcinomas (
1
). Clinically useful mustard analogs include melphalan (L-phenylalanine mustard), the prototypical mustard mechlorethamine (HN2) and cyclophosphamide, which is converted to its active form, phosphoramide mustard, via the cytochrome P450 pathway (Fig.
1
). In aqueous solution, each of the chloroethyl side chains of these compounds can spontaneously cyclize to form an aziridinium
ion capable of adding to a nucleophilic site in DNA. The resulting monoadduct can form a second aziridinium ion which can simply react with solvent or can add to another nearby
nucleophilic site, resulting in a crosslink either between DNA and protein or
between two DNA bases (reviewed in
2
). While DNA interstrand crosslinking has been demonstrated for several nitrogen
mustards, the abundance of bifunctional adducts in mustard-treated DNA (
3
) suggests that intrastrand crosslinks may also be formed, however, the
existence of such lesions in defined sequence DNA has not been demonstrated.
A serious complication of treatment with nitrogen mustards is the increased risk
of a secondary leukemia in long-term survivors (
4
). The occurrence of characteristic karyotypic changes, particularly deletion of a 5q or 7q arm, in therapy-related acute myelogenous leukemias (
5
,
6
) suggests that specific genetic loci are consistently being altered. The exact mechanisms of these putative drug-induced genetic alterations remain largely unknown. If the therapeutic and
mutagenic effects of these drugs result from different lesions in the genome,
the two effects may be separable.
The predominant mutations induced by the aromatic nitrogen mustards melphalan
and chlorambucil in the shuttle vector pZ189 replicated in human 293 cells were
A[middot]T -> T[middot]A transversions, apparently resulting from adenine N-3 alkylations (
7
,
8
). In contrast, however, most melphalan-induced base substitutions in the
aprt
gene of CHO cells occurred at G[middot]C base pairs; G[middot]C -> T[middot]A and G[middot]C -> C[middot]G transversions
predominating (
9
). Many of the base substitutions were at G-N-C (particularly G-G-C) sequences, which are potential interstrand crosslink sites, at least for
the prototypical mustard mechlorethamine (
10
). Surprisingly, however, many of these substitutions occurred at the internal
base pair (3'
Phosphoramide mustard was obtained from Dr V.N.Narayanan (Developmental Therapeutics Program, National Cancer Institute). Since phosphoramide mustard is highly reactive in aqueous solution (
11
), it was freshly dissolved in reaction buffer (10 mM Tris-HCl, 0.1 mM EDTA, pH 8) at a concentration of 100 mM immediately before
use. Melphalan and mechlorethamine (Sigma, St Louis, MO) were dissolved in 0.1
N HCl at a concentration of 10 mM, stored at -70oC and diluted in reaction buffer before use.
The oligonucleotides 5'-
OH
CCACACTC
To obtain a 3'-end-labeled duplex, oligonucleotide 2 (0.64 [mu]g) was incubated for 3 h at 10oC in the presence of 0.96 [mu]g complementary oligonucleotide 3, 200 [mu]Ci [[alpha]-
32
P]dATP (3000 Ci/mmol; New England Nuclear), 0.2 mM each dGTP, dCTP and dTTP and 10 U Klenow fragment
(New England Biolabs) in the buffer provided plus 0.1 M NaCl and the labeled
duplex was purified on a non-denaturing gel.
To obtain a single-stranded 3'-end-labeled substrate, oligonucleotide 2 was treated with terminal transferase (US
Biochemical Co.) and 100 [mu]Ci [[alpha]-
32
P]ddATP (3000 Ci/mmol; Amersham) according to the manufacturer's instructions, ethanol precipitated and purified on a denaturing gel.
Reaction mixtures (10 [mu]l) contained 80 [mu]g/ml calf thymus DNA, single- or double-stranded
32
P-end-labeled DNA oligonucleotide (<10 [mu]g/ml), reaction buffer (10 mM Tris-HCl, pH 8, 0.1 mM EDTA) and the indicated concentrations
of nitrogen mustards. After an initial 30 min incubation at 37oC, the DNA was ethanol precipitated in the presence of 100 [mu]g/ml carrier tRNA, in order to remove free drug. The DNA was
redissolved in 50 [mu]l reaction buffer and split into five 10 [mu]l aliquots, which were incubated for 0, 3, 6, 12 or 24 h at 37oC and then frozen at -70oC. At 24 h, the samples were thawed, ethanol precipitated, dried,
dissolved in 0.1 N sodium hydroxide and incubated for 1 h at 37oC in order to convert N-7 adducts to the more stable formamidopyrimidine (FAPY) derivatives (
13
). After a final precipitation to remove NaOH, each sample was dissolved in 20 [mu]l reaction buffer and split into two aliquots to be either frozen at -70oC or treated with 1 M piperidine at 90oC for 30 min in a volume of 100 [mu]l in order to cleave both N-7 adducts and their FAPY derivatives (
14
). Aliquots of 10 [mu]l loading solution (80% formamide plus 20 mM EDTA) were added to untreated and lyophilized piperidine-treated samples and DNA was denatured for 1 min at 90oC immediately before loading. The samples were electrophoresed in
a 20% denaturing polyacrylamide/bisacrylamide (20:1) gel. The wet gels were
autoradiographed at -70oC for 10 h.
[beta]-Emissions from bands in the gel corresponding to radioactive
crosslinked DNA or broken DNA fragments were quantitated using a Betascope 603 Blot Analyzer (Betagen Inc.). Radioactivity in each band was normalized to the total radioactivity in the lane to give the
percent DNA in each band. Each assay was reproduced a minimum of three times;
the mean and standard error of each time point was calculated and plotted. An
ANOVA (
15
) was performed to assess the statistical significance of time-dependent changes in alkylation at each site.
Following Betascope analysis, the slower migrating crosslinked DNA was excised
from the gel. Crosslinked DNA bands for all the time points for each
concentration were pooled and eluted for 16 h. After removal of gel fragments
by centrifugation, the DNA was ethanol precipitated and heated in 100 [mu]l 1 M piperidine for 1 h at 90oC to cleave the stabilized FAPY interstrand crosslinks. Following lyophilization, the samples were analyzed on denaturing gels as above.
The extent of intrastrand crosslinking between the two adjacent Gs in the G-G-C sequence was calculated by comparison of cleavage patterns of 5'- and 3'-end-labeled DNA as follows.
Let
A
be the the apparent ratio of 5'-G/3'-G cleavage when oligonucleotide 2 is 5'-end-labeled,
B
be the the apparent ratio of 3'-G/5'-G cleavage when oligonucleotide 2 is 3'-end-labeled,
x
be the fraction of adducts at the G-G site which are either monoadducts at the 5' G or interstrand crosslinks involving the 5' G,
y
be the fraction of adducts at the G-G site which are monoadducts at the 3' G and
z
be the fraction of adducts at the G-G site which are intrastrand crosslinks between the 5' and 3' Gs. The total of all adduct fractions, 1, is equal to the
sum of the fractions of monofunctional adducts, interstrand crosslinks and
intrastrand crosslinks at the G-G site.
1 =
x
+
y
+
z
When an intrastrand crosslink occurs between the 5' and 3' Gs and the backbone is cleaved at both sites by piperidine, the
lesion will appear on a sequencing gel as a single break at the G nearest the
labeled end. Thus, the apparent ratio of 5'-G/3'-G cleavage when oligonucleotide 2 is 5'-end-labeled is:
A
= (
x
+
z
)/
y
= (1 -
y
)/
y
or
y
= 1/(
A
+ 1)
Similarly, the apparent ratio of 3'-G/5'-G cleavage when oligonucleotide 2 is 3'-end-labeled is:
B
= (
y
+
z
)/
x
= (1 -
x
)/
x
or
x
= 1/(
B
+ 1)
Thus,
z
= 1 - (
y
+
x
) = 1 - [1/(
A
+ 1) + 1/(
B
+ 1)]
Molecular modeling studies were performed on a Silicon Graphics Crimson VGXT workstation. The 9mer GCA
To determine the positions of interstrand crosslinks, the double-stranded DNA duplex was 5'- or 3'-end-labeled in the G-G-C-containing strand or was
5'-end-labeled in the complementary strand. Each labeled duplex was exposed to melphalan,
mechlorethamine or phosphoramide mustard. Following drug removal by precipitation, the
samples were incubated for 24 h to allow formation of interstrand crosslinks.
It is generally accepted that guanine N-7 adducts are responsible for interstrand crosslinks, however, these alkylated adducts easily depurinate (
16
). Therefore, following the 24 h incubation, samples were treated with alkali to convert N-7 adducts, including interstrand crosslinks, to the more stable, imidazole
ring-opened FAPY derivatives (
13
). When either strand was end-labeled, each drug produced interstrand crosslinks in a dose-dependent manner, as evidenced by the presence on the denaturing gel
of a slower migrating band (shown for melphalan in Fig.
2
). This band was not seen with samples of either untreated DNA or drug-treated single-stranded DNA. Although the samples were heat denatured for 1 min at
90oC and electrophoresed on a denaturing gel, the crosslink was apparently
stable under these conditions. The location of the interstrand crosslink was
determined by isolating the slower migrating fragments and heating them for 1 h
at 90oC in piperidine, conditions known to hydrolyze most guanine N-7 adducts and their FAPY derivatives (
14
). The products were resolved on a denaturing gel (Fig.
3
). For all three drugs, the major alkylated base in the crosslinked DNA was the
5'
When the 20 bp duplex was treated with melphalan, very little crosslinking
occurred during the initial 30 min drug treatment. Rather, following drug
removal, crosslinks formed very slowly, reaching a maximum between 6 and 12 h (
t
1/2
= 3 h) (Fig.
4
). When total alkylation at individual guanines in each strand (as judged by
piperidine-mediated cleavage) was examined as a function of time (Fig.
5
), it was apparent that alkylation of
Melphalan-induced mutations in the
aprt
gene of CHO cells were found preferentially at potential interstrand crosslink
sites (
9
), but many of these mutations occurred at the internal
The highly reactive mechlorethamine, which produced the largest number of
interstrand crosslinks, also generated the most intrastrand crosslinks in both
single-stranded (13%) and double-stranded (28%) DNA. No increase in the number of 3' or 5' breaks at the G-G site during a 24 h time period was detected,
indicating rapid crosslink formation. Phosphoramide mustard produced detectable intrastrand
crosslinking (8%) only in double-stranded DNA.
Mechlorethamine, melphalan and phosphoramide mustard adducts in double-stranded DNA were subjected to molecular dynamic simulations in an attempt
to determine a structural basis for differences in intrastrand crosslink
formation. The movement of the aziridinium ion intermediate was evaluated in
terms of the distance from each aziridinium carbon to either the guanine N-7 in the opposite strand, corresponding to a potential
Table 1
In each of these simulations, equilibrium was reached within 10 ps, after which both the energy and the adduct orientation remained essentially
constant (data not shown). Prior to reaching equilibrium, all three adducts
briefly sampled conformational space to within ~4.5-5.3 Å of the guanine N-7 corresponding to the interstrand crosslink site.
Although this distance has been considered sufficiently small to allow
interstrand crosslinking (
22
,
23
), the equilibrium conformations of both mechlorethamine and phosphoramide mustard adducts appeared to favor intrastrand crosslink formation. The average distance from the nearest aziridinium carbon
atom to the adjacent guanine N-7 was 3.14 (mechlorethamine, Fig.
7
, left) or 3.36 Å (phosphoramide mustard, not shown), with ~0.5 Å deviation from these positions and direct van der Waals
contacts between these two moieties during the simulation. Thus, modeling
studies predict that mechlorethamine and phosphoramide mustard, when adducted
to DNA, assume relatively stable conformations wherein the second mustard arm
is appropriately positioned for intrastrand crosslink formation at G-G sequences.
Nitrogen mustards produce a variety of monofunctional and bifunctional DNA
lesions and a diverse spectrum of mutations, including substitutions at both G[middot]C and A[middot]T base pairs, as well as large deletions (
2
). In the endogenous
aprt
gene of CHO cells, the most frequent mutations induced by the aromatic mustard
melphalan were G[middot]C -> T[middot]A and G[middot]C -> C[middot]G transversions and these
occurred preferentially at G-G-C, G-G-C-C and to a lesser extent G-A-C sequences. Stereochemical
considerations suggest that G-G-C sequences could be potential sites for formation of both
interstrand and intrastrand crosslinks involving guanine N-7. Since such crosslinks are likely to be more difficult to repair than
monofunctional alkylations, they represent prime candidates for premutagenic
DNA lesions. The present study was undertaken to determine the propensity of
several nitrogen mustards to crosslink specific guanine residues in a G-G-C[middot]G-C-C sequence.
The results show that both melphalan and phosphoramide mustard induce
interstrand guanine-guanine crosslinks with the same two base-staggered
At a G-G-C sequence in double-stranded DNA, the proportion of intrastrand crosslinks formed
by various mustards differed by a factor of >20-fold, with mechlorethamine > phosphoramide mustard > melphalan. These
differences appear to be due at least in part to differences in the geometry of
the monofunctionally bound drug in the major DNA groove, since they were less
pronounced in single-stranded DNA. Moreover, molecular modeling predicted that mechlorethamine
and phosphoramide mustard would assume monoadduct conformations consistent with
intrastrand crosslinking, while melphalan, due to hydrogen bonding of its
phenylalanine moiety to the DNA backbone, would not. It should be noted that
the lack of intrastrand crosslinking by melphalan cannot be attributed to the
mere bulkiness of the phenylalanine side chain, as additional modeling studies (not shown) showed that a melphalan molecule bifunctionally linked to N-7 positions of adjacent guanines can easily fit into the major groove with
essentially no distortion in DNA structure.
The relatively high frequency of intrastrand crosslinking induced by mechlorethamine supports the proposal of Pieper
et al
. (
27
) that selective transcription termination at G-G sequences in mechlorethamine-treated DNA may be attributable to intrastrand crosslinks. Intrastrand crosslinks could also account for at least
some mechlorethamine-induced base substitutions in pZ189, most of which occur at G-G sites (
7
). However, there is no evidence that intrastrand crosslinks correlate with
cytotoxicity, since the ratio of intrastrand to interstrand crosslinks is much
higher for mechlorethamine than for melphalan, yet, at equivalent levels of
interstrand crosslinking, melphalan is somewhat more cytotoxic (
25
).
The present data do not clearly implicate any one DNA lesion as accounting for the relatively high frequency of melphalan- induced mutations at G-G-C sequences in
aprt
. Interstrand crosslinks could account for substitutions at the first and last positions in
the sequence, but not at the internal G, which is mutated just as frequently (
9
). Intrastrand crosslinks, which have been implicated in mutagenesis by
cisplatin (
28
,
29
), remain a possible candidate lesion for melphalan-induced mutations involving the internal G. Although intrastrand
crosslinks are apparently produced by melphalan only in single-stranded DNA, the very slow kinetics of the second arm alkylation reaction
raise the possibility that they could be formed when monoadducted DNA becomes
transiently single stranded during transcription or replication. Of course, the
mutations could be the result of N-7 monoadducts, but since G-G-C sequences are not particularly strong sites for monoadduct
formation, other factors must then be invoked to account for the sequence
specificity of mutagenesis. Although all guanine adducts detected thus far in melphalan-treated DNA appear to involve N-7 alkylation (
2
), it is difficult to eliminate the possibility that other rare but highly
mutagenic adducts might also be induced at these sites (e.g. guanine N
2
adducts, which have been implicated in mutagenesis by benzo[
a
]pyrene;
30
). Finally, the possibility of untargeted or `locally targeted' mutagenesis (
29
) or of mutagenesis by some endogenous species generated in response to
malphalan treatment cannot be ruled out.
In conclusion, although interstrand crosslinks were detected at the 5' G in the G-G-C sequences, there is little indication of what lesions might
be involved in melphalan-induced mutations at the 3' G, which occur just as frequently. Similar considerations apply to
the spectrum of mechlorethamine-induced mutations in the yeast
SUP4
-o gene (
31
), where several of the most frequently mutated sites are neither expected
alkylation hotspots (
17
) nor potential sites of inter- or intrastrand crosslinks. The disparate mutation spectra produced by
nitrogen mustards in various experimental systems (
2
) suggest that rather than being attributable to any single highly mutagenic lesion, mustard-induced mutagenesis represents the cumulative effects of several types of adenine and guanine adducts, both monofunctional and bifunctional. This complexity would appear to preclude, for the present, any rational
attempts to devise less mutagenic but equally cytotoxic analogs of these drugs.
This work was supported by grant CA40615 from the National Cancer Institute.
G.B.B. was supported in part by Training Grant HL07110 from the National Heart,
Lung and Blood Institute.
*To whom correspondence should be addressed. Tel: +1 804 828 9640; Fax: +1 804
828 8079; Email: lpovirk@gems.vcu.edu
+
Present address: Department of Medicinal Chemistry, Medical College of Virginia,
Virginia Commonwealth University, Richmond, VA 23298, USA
5'-G/3'-G
b
3'-G/5'-G
b
Intrastrand
5'-end-label
3'-end-label
crosslinks
Double-stranded DNA
Melphalan
2.30 +- 0.030
0.45 +- 0.025
0.01 +- 0.011
Mechlorethamine
2.54 +- 0.058
1.28 +- 0.044
0.28 +- 0.013
Phosphoramidemustard
1.98 +- 0.026
0.70 +- 0.012
0.08 +- 0.007
Single-stranded DNA
Melphalan
0.95 +- 0.026
1.23 +- 0.020
0.05 +- 0.011
Mechlorethamine
1.07 +- 0.025
1.56 +- 0.070
0.13 +- 0.011
Phosphoramide mustard
0.78 +- 0.026
1.34 +- 00.076
0.01 +- 0.021
REFERENCES
Return