Mutation Research 424 Ž1999.

83–95

Lipid peroxidation—DNA damage by malondialdehyde

)
Lawrence J. Marnett
A.B. Hancock Jr. Memorial Laboratory for Cancer Research Center in Molecular Toxicology, Vanderbilt Cancer Center, Departments of
Biochemistry and Chemistry, Vanderbilt UniÕersity School of Medicine, NashÕille TN 37232, USA
Accepted 23 July 1998

Abstract

Malondialdehyde is a naturally occurring product of lipid peroxidation and prostaglandin biosynthesis that is mutagenic
and carcinogenic. It reacts with DNA to form adducts to deoxyguanosine and deoxyadenosine. The major adduct to DNA is
a pyrimidopurinone called M 1G. Site-specific mutagenesis experiments indicate that M 1G is mutagenic in bacteria and is
repaired by the nucleotide excision repair pathway. M 1G has been detected in liver, white blood cells, pancreas, and breast
from healthy human beings at levels ranging from 1–120 per 10 8 nucleotides. Several different assays for M 1G have been
described that are based on mass spectrometry, 32 P-postlabeling, or immunochemical techniques. Each technique offers
advantages and disadvantages based on a combination of sensitivity and specificity. Application of each of these techniques
to the analysis of M 1G is reviewed and future needs for improvements are identified. M 1G appears to be a major endogenous
DNA adduct in human beings that may contribute significantly to cancer linked to lifestyle and dietary factors. High
throughput methods for its detection and quantitation will be extremely useful for screening large populations. q 1999
Elsevier Science B.V. All rights reserved.

Keywords: Malondialdehyde; Pyrimidopurinone; M 1G; DNA adduct; Mass spectrometry; Postlabeling; Immunochemistry

1. Lipid peroxidation
Lipid peroxidation generates a complex variety of
products, many of which are reactive electrophiles.
Abbreviations: MDA, malondialdehyde; M 1G, pyrimido-
w1,2 a xpurin-10Ž3 H .-one; M 1 A, N 6 -Ž3-oxo-propenyl.deoxyadeno-
sine; M 1C, N 4-Ž3-oxopropenyl.deoxycytidine; 8-oxodG, 8-oxo-
7,8-dihydrodeoxyguanosine; PdG, 1, N 2-propanodeoxyguanosine;
GCrEC NCIrMS, gas chromatographyrelectron capture negative
chemical ionizationrmass spectrometry; GCrMS, gas chromatog-
raphyrmass spectrometry; PFB, pentafluorobenzyl; LCrMS, liq-
uid chromatographyrmass spectrometry
)
Corresponding author. Tel.: q1-615-343-7329; Fax: q1-615-
343-7534; E-mail: marnett@toxicology.mc.vanderbilt.edu
Some of these react with protein and DNA and as a
result are toxic and mutagenic. The basic reactions of
lipid peroxidation are outlined in Scheme 1 w1x.
Polyunsaturated fatty acids contain one or more
methylene groups positioned between cis double
bonds. The methylene groups are highly reactive to
oxidizing agents and their hydrogen atoms are re-
moved to form carbon-centered radicals Že.g., 1..
Carbon-centered radicals react with O 2 to form per-
oxyl radicals at a rate that is limited only by diffu-
sion. Thus, the initial products of polyunsaturared
fatty acid oxidation are peroxyl radicals. The fate of
the peroxyl radical depends on its position in the
carbon chain of the fatty acid. If the peroxyl radical
exists at one of the two ends of the double bond

0027-5107r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 7 - 5 1 0 7 Ž 9 9 . 0 0 0 1 0 - X
84 L.J. Marnettr Mutation Research 424 (1999) 83–95

Scheme 1. Pathways of lipid peroxidation.

system Že.g., 2., it is reduced to a hydroperoxide.
Conjugated diene hydroperoxides such as 4 are the
simplest products of lipid peroxidation and are rela-
tively stable in the absence of metals. However,
metal complexes and metalloproteins are abundant in
cells and they rapidly reduce all fatty acid hydro-
peroxides by one electron to alkoxyl radicals, which
undergo multiple chemical reactions to generate a
broad range of products w2x. Thus, even the simplest
initial lipid peroxidation products produce a complex
range of epoxides, hydroperoxides, and carbonyl
compounds, inter alia.
In phospholipid membranes, the molecules that
reduce peroxyl radicals to hydroperoxides are either
another molecule of fatty acid or vitamin E. If
another molecule of fatty acid reduces the peroxyl
radical, a new carbon-centered radical is generated
which propagates the fatty acid oxidation. Through
this radical chain process, one oxidizing agent can
cause many molecules of fatty acid to become oxi-
dized. The actual lengths of the radical chains that
occur in cells are sensitive to many factors but in
pure chemical systems approximately 60 molecules
of linoleic acid and 200 molecules of arachidonic
acid are consumed per initiation event w3x. The major
determinant of the length of radical chains in vivo is
the concentration of vitamin E in the phospholipid
bilayer w4x. Vitamin E reduces peroxyl radicals which
normally breaks the radical chain and slows the rate
of lipid peroxidation. However, under conditions of
 L.J. Marnettr Mutation Research 424 (1999) 83–95
very low oxidant, vitamin E has actually been shown
to initiate another radical chain w5x.
If the peroxyl radical is at an internal position in
the fatty acid chain Že.g., 3., cyclization to an adja-
cent double bond will compete with reduction to a
hydroperoxide. The cyclization product Ž5. is a cyclic
peroxide adjacent to a carbon-centered radical. This
carbon-centered radical has two fates. The radical
can couple with O 2 to form a peroxyl radical which
is reduced to a hydroperoxide Ž6. as described above.
Alternatively, carbon-centered radical 5 can undergo
a second cyclization to form a bicyclic peroxide
which after coupling to O 2 and reduction yields a
molecule Ž7. structurally analogous to the prosta-
glandin endoperoxide, PGG2 , albeit without the
stereochemical control exhibited in the enzymatic
reaction w6x. Compound 7 serves as a common inter-
mediate for the production of isoprostanes and
malondialdehyde ŽMDA. through chemical conver-
sion of the bicyclic peroxide group w6,7x. The frag-
mentation that produces MDA generates a 17-carbon
fatty acid as the other product w8x. MDA has been
used for many years as a convenient biomarker for
lipid peroxidation because of its facile reaction with
thiobarbituric acid to form an intensely colored chro-
mogen w9x. However, the thiobarbituric acid reaction
is notoriously non-specific which has led to substan-
tial controversy over its use for quantitation of MDA
from in vivo samples. Recently, isoprostanes have
become popular biomarkers of lipid peroxidation and
their use has been validated in several in vivo mod-
els w7x.
All of the hydroperoxide intermediates depicted in
Scheme 1 and their many regioisomers and
stereoisomers can be reduced by metal ions to alkoxyl
Scheme 2. Oxidation of amino acids to aldehydes by myeloperoxidase.
85
radicals that decompose to a range of products.
Among these are many carbonyl compounds includ-
ing hexanal, 4-hydroxynonenal, and other saturated
and a ,b-unsaturated aldehydes and ketones w10x.
Many of these compounds react with protein and
DNA and are toxic and mutagenic. MDA appears to
be the most mutagenic product of lipid peroxidation
whereas 4-hydroxynonenal is the most toxic w11x.
2. Alternate pathways to aldehydes linked to
inflammation
Another pathway has been described recently for
the formation of aldehydes that is independent of
lipid peroxidation but which may be important for
the generation of gentoxic compounds. Hazen et al.
have demonstrated that myeloperoxidase in the pres-
ence of chlorine generates hypochlorous acid which
oxidizes amino acids to chloramines w12x. The latter
derivatives lose of chloride and CO 2 to form imines;
the imines hydrolyze to aldehydes ŽScheme 2.. In the
case of threonine, the hydroxyaldehyde product un-
dergoes dehydration to form acrolein. This may be
an important source of the acrolein that damages
DNA endogenously in human beings and which has
been assumed to arise from lipid peroxidation w13x.
3. Genotoxicity of MDA
MDA was first shown to be carcinogenic in 1972
following topical administration to mice w14x. This
was an unusual result because the tumors arose in a
very short time and were not the type of tumors
86 L.J. Marnettr Mutation Research 424 (1999) 83–95

typically seen in a mouse skin assay. Subsequent

attempts to confirm this skin carcinogenesis result
were unsuccessful w15x. Mukai and Goldstein re-
ported the mutagenic activity of MDA toward
Salmonella typhimurium in 1976 w16x. Since MDA is
unstable, the material used for the assay was pre-
pared by hydrolysis of tetraethoxypropane. Some
impurities generated during tetraethoxypropane hy-
drolysis are nearly 30-fold more mutagenic than
MDA and account for a significant part of the activ-
ity Ž; 50%. detected in the assay w17x. However,
assay of highly purified MDA prepared by three
independent means revealed that this compound is,
indeed, mutagenic w18x. The carcinogenic activity of
MDA was established in a 2-year rodent bioassay
w19x. Highly purified preparations of MDA induced
thyroid tumors in rats but not mice.
Several other aldehydes, including some a ,b-un-
saturated aldehydes are mutagenic in Salmonella if
steps are taken to minimize their toxicity w20x. This
consists of adding high concentrations of glutathione
shortly after administration of the aldehyde to the
bacteria to strip aldehydes from cysteinyl groups of
proteins which are putative targets for their toxic
action. In fact, 4-hydroxynonenal was originally dis-
covered as a highly toxic product of lipid peroxida-
tion and both it and its glutathione conjugate were
used as chemotherapeutic agents in some small hu-
man clinical trials w21x. The toxic activity of the
4-hydroxynonenal appeared to be due to its reactivity
with sulfhydryl proteins w22x.
Scheme 3. Formation of MDA–DNA adducts.

4. Reaction of MDA with DNA

MDA reacts with nucleic acid bases to form
multiple adducts ŽScheme 3.. Both its carbonyl
equivalent adducts to N 2 and N 1 of dG with loss of
two water molecules to form a pyrimidopurinone
w23,24x. This compound is a planar aromatic molecule
that is moderately fluorescent Ž; 3% quantum yield..
The condensation products with dA and dC arise by
addition of one of the carbonyl equivalents of MDA
with the exocyclic amino groups to form an oxo-
propenyl derivative w25–27x. No evidence for cy-
clization of these products has been observed. Com-
parison of the yields of the various adducts produced
in the reaction of MDA with DNA in vitro indicates
the pyrimidow1,2 a xpurin-10Ž3 H .-one ŽM 1G. is the
major adduct followed by N 6-Ž3-oxo-propenyl.
deoxyadenosine ŽM 1 A.. The amount of M 1G is ap-
proximately five times that of M 1 A. N 4-Ž3-
oxopropenyl.deoxycytidine ŽM 1C. is formed in only
trace amounts w28x.
A complicating factor in the reaction of MDA
with nucleosides is its polymerization to form dimers
and trimers that also react with DNA. The dimer of
MDA reacts with dG to form an oxadiazabi-
cyclow3:3:1xnonene derivative whereas the trimer of
MDA reacts with dA and dC to form Ž5X ,7X-bis-for-
myl-2X H-3X ,6X - dihydro-2X ,6X-methano-1X ,3X - oxazocin-
3X-yl. derivatives of the exocyclic amino groups
 L.J. Marnettr Mutation Research 424 (1999) 83–95
w26,27x. The oligomerization of MDA is relatively
slow at neutral pH so the monomeric adducts de-
picted in Scheme 3 are the major products generated
under physiological conditions. However, there may
be conditions in vitro or possibly in vivo where these
unusual multimeric adducts are formed. The ability
of MDA to form monomeric and oligomeric adducts
is a reflection of the nucleophilic reactivity of its
enol functional group whereas the electrophilic reac-
tivity is due to its aldehyde functional group. The
existence of oligomeric deoxynucleoside adducts ne-
cessitated the adoption of abbreviations for the indi-
vidual adducts that designate the number of MDA
units in each adduct ŽScheme 3..
5. Alternate routes to M 1G
Dedon et al. have recently demonstrated that M 1G
Žand presumably other MDA-derived adducts. can be
formed independently of lipid peroxidation w29x. Di-
rect oxidation of DNA by agents that abstract the 4X
hydrogen atom of the sugar backbone Že.g.,
calicheamicin, bleomycin. initiates a cascade of reac-
tions that lead to the formation of base propenals.
Scheme 4. Reaction of oxidizing agents with DNA to form M 1G.
87
Base propenals are oxopropenyl base derivatives
structurally analogous to acroleins substituted at the
b position with good-to-moderate leaving groups
ŽScheme 4.. These compounds Žincluding M 1 A and
M 1C. transfer their oxopropenyl group to dG to form
M 1G. Treatment of DNA with bleomycin or
calicheamicin in the complete absence of lipid leads
to the formation of M 1G. M 1G also is formed by the
direct reaction of adenine propenal with DNA. This
provides another pathway linking oxidative stress to
the formation of M 1G in DNA and may explain the
occurrence of M 1G at high levels in certain human
tissues Žsee below..
6. Mutagenic potential and repair of MDA–DNA
adducts
The mutagenic potential of MDA–DNA adducts
has been evaluated using random and site-specific
approaches. Benamira et al. reacted MDA at neutral
pH with a single-stranded M13 vector containing the
lacZ a gene and scored mutations to the lacZ ay
phenotype following replication in strains of Es-
cherichia coli induced for the SOS response w30x.
Increasing concentrations of MDA led to an increase
in lacZ ay mutations coincident with an increase in
the level of the major MDA–deoxyguanosine adduct,
M 1G. The most common sequence changes induced
by MDA were base pair substitutions Ž76%.. Of
these, 43% Ž29r68. were transversions, most of
which were G ™ T Ž24r29.. Transitions accounted
for 57% of the base pair substitutions Ž39r68. and
were comprised exclusively of C ™ T Ž22r39. and
A ™ G Ž17r39.. Frameshift mutations were identi-
fied in 16% of the induced mutants and were com-
prised of mainly single base additions occurring in
runs of reiterated bases Ž11r14.. If one assumes that
all mutations are targeted to the site of adduction, the
ability of MDA to induce base pair substitution
mutations at dG, dA, and dC residues coincides with
its ability to form adducts at all three deoxynucleo-
sides. No mutations were detected at dT residues as
expected from the fact that MDA does not form
detectable adducts to dT.
Site-specific mutagenesis experiments have been
used to investigate the mutagenic potential of M 1G.
This necessitated modification of the synthetic reac-
88 L.J. Marnettr Mutation Research 424 (1999) 83–95
tions used for oligodeoxynucleotide synthesis to ac-
commodate the instability of M 1G to the base depro-
tection conditions normally used in the last step of
the synthetic pathway w31x. M 1G was positioned at a
defined site in a duplex M13 genome using a gapped
duplex approach. Unmodified and M 1G-modified
genomes containing either a cytosine or thymine
opposite the lesion were transformed into repair-pro-
ficient and repair-deficient E. coli strains. Base pair
substitutions were quantitated by hybridization anal-
ysis. Modified genomes containing a cytosine oppo-
site M 1G resulted in roughly equal numbers of M 1G
™ A and M 1G ™ T mutations with few M 1G ™ C
mutations w32x. The total mutation frequency was
approximately 1%, which represents a 500-fold in-
crease in mutations compared to unmodified
M13MB102. Transformation of modified genomes
containing a thymine opposite M 1G allowed an esti-
mate to be made of the ability of M 1G to block
replication. The Žy.-strand was replicated ) 80% of
the time in the unadducted genome, but only 20% of
the time when M 1G was present. Correction of the
mutation frequency for the strand bias of replication
indicated that the actual frequency of mutations in-
duced by M 1G was 18% w32x. The spectrum and
frequency of mutations induced by M 1G were very
similar to those observed with a structural analog
1, N 2-propanodeoxyguanosine ŽPdG. assayed in the
same system.
Repair of M 1G was assessed by transformation of
M 1G-containing genomes into E. coli strains defi-
cient in individual genes of DNA repair based on the
assumption that inactivation of a repair gene will
increase the half-life of the adduct and increase its
mutagenicity in site-specific experiments. No effect
on mutation frequency was observed when cells
were defective in formamidopyrimidine glycosylase
or 3-methyladenine glycosylase. These two glycosy-
lases have been shown previously to participate in
the repair of 8-oxo-7,8-dihydrodeoxyguanosine Ž8-
oxodG. and exocyclic adducts, respectively w32,33x.
Approximately a fourfold increase in mutation fre-
quency was observed when M 1G-containing genomes
were transformed into cells deficient in nucleotide
excision repair Žeither uÕrAy or uÕrBy . w32x. In-
creases in mutation frequency of similar magnitude
were observed with PdG-containing genomes. The
ability of PdG Žand by analogy M 1G. to be removed
by reconstituted bacterial and mammalian nucleoside
excision repair complexes was confirmed by a series
of in vitro experiments w33x. PdG appears to be
removed more efficiently by the mammalian nu-
cleotide excision repair complex than the bacterial
complex.
7. Occurrence of M 1G in genomic DNA
7.1. Postlabeling analysis
The high mutagenic potential of M 1G begs the
question of whether it is formed in human beings. As
with all endogenous DNA adducts, answering this
question requires unusually robust and rigorous as-
says because of the lack of a control DNA from an
untreated animal or human sample. By definition, the
untreated DNA is what is being analyzed for the
presence of ‘background DNA damage’. Thus, any
assay used for analysis must be highly sensitive and
very specific. One must be able to verify that the
adduct detected is actually what it is presumed to be.
This is no simple task with any DNA adduct.
32
P-Postlabeling was first employed for the detec-
tion and quantitation M 1G. Vaca and coworkers
developed a method using nuclease P1 treatment for
enrichment that was used to detect the adduct in
mouse liver DNA w34x. Treatment of mice with
MDA Ž20 mmolrkg. caused an increase in adduct
level to 6 per 10 6 nucleotides. The levels of M 1G in
untreated animals was not reported. In subsequent
studies, treatment of mice with 200 mmolrkg MDA
stimulated the formation of MDA–hemoglobin
adducts in blood and M 1G in liver. The levels of
M 1G declined slowly following MDA treatment and
demonstrated a half-life of 12.5 days w35x. This slow
half-life for removal is interesting considering the
efficient removal of M 1G and PdG by nucleotide
excision repair described above. Interestingly, in
these experiments, the levels of M 1G in the liver
increased significantly up to 4 days after administra-
tion of MDA.
Wang and Liehr utilized 32 P-postlabeling to
demonstrate a 2- to 3-fold increase in the levels of
M 1G in the kidneys of Syrian hamsters treated with
doses of diethylstilbestrol or estradiol that induce
cancer in this organ w36x. Wang and Liehr also
 L.J. Marnettr Mutation Research 424 (1999) 83–95 89

provided evidence for the existence of MDA adducts

to dA as well as dG w37x. The identities of these
adducts have not yet been established but they have
been found in human white cells and breast Žsee
below..
The levels of M 1G in human white blood cells
and breast tissue have been determined by 32 P-post-
labeling. Values in white cells of 2.6 " 1.2 adducts
per 10 7 nucleotides Žrange 1–5 per 10 7 . were re-
ported from a study of 26 male and female volun-
teers w38x. Parallel analysis of 7 female volunteers
indicated levels of 3.0 " 1.3 M 1G per 10 7 nu-
cleotides in normal breast tissue. Fang et al. reported
a dramatic increase in the level of M 1G in the white
cells of women Žbut not men. fed a diet high in
sunflower oil compared to a group fed a high rape-
seed oil diet w39x. The high polyunsaturated fatty acid
diet stimulated nearly a 20-fold increase in the levels
of M 1G up to levels of 28 per 10 7 nucleotides. Wang
et al. reported that the levels of M 1G and putative
MDA–dA adducts were higher Ž2- to 3-fold. in the
normal breast tissue of women with breast cancer
compared to the normal tissue of women without
breast cancer w40x. The levels of M 1G in the tumor
tissue of women with breast cancer were lower than
those in the surrounding normal tissue. Scheme 5. Steps in the analysis of M 1G by GCrEC NCIrMS.

7.2. Mass spectrometric analysis mode enables quantitation by comparison of the

Our laboratory has employed gas chromatogra-
phyrmass spectrometry ŽGCrMS. with electron
capture negative chemical ionization detection
ŽGCrEC NCIrMS. for quantitation of M 1G w41,42x.
Scheme 5 charts the course of the analytical method.
Nuclear DNA is isolated, purified and digested to
deoxynucleosides. During nuclease treatment, an in-
ternal standard of M 1G-deoxyribose labeled with two
deuteriums is added. The mixture of deoxynucleo-
sides is passed through an immunoaffinity column
containing an immobilized monoclonal antibody
raised against a periodate-cleaved M 1G-ribose conju-
gated to albumin w43x. Unmodified deoxynucleosides
pass through the immunoaffinity column then the
M 1G-deoxyribose is eluted with methanol. The de-
oxyribose residue is removed by acid hydrolysis then
the M 1G base is derivatized with pentafluorobenzyl
ŽPFB. bromide. Analysis of the PFB derivatives by
GCrEC NCIrMS in the selected ion monitoring
intensity of the base peak of the endogenously occur-
ring M 1G Ž mrz s 186. to that of the internal stan-
dard Ž mrz s 188. ŽFig. 1.. The chromatographic
trace in Fig. 1 displays two separable M 1G-PFB
derivatives corresponding to the N7 and N9 isomers.
The GCrEC NCIrMS assays display good linearity
and quantitative recovery of M 1G when oligonucleo-
tides containing this adduct are carried through the
analytical procedure w42x. The limit of sensitivity of
the assay is approximately 1 M 1G per 10 8 nu-
cleotides starting with 1 mg of DNA. Specificity is
provided by the combination of immunoaffinity
chromatography, chemical derivatization, chromato-
graphic coelution, and selected ion monitoring.
Application of this technique to human liver and
leukocyte DNA samples reveals that M 1G is present
at levels of ; 1 in 10 6 nucleosides and 6 in 10 8
nucleosides, respectively w41,42x Žsee Table 1.. The
livers were unused transplant samples and the leuko-
cyte samples were from human volunteers. The sam-
90 L.J. Marnettr Mutation Research 424 (1999) 83–95

Fig. 1. Selected ion monitoring profiles of PFB derivatives of M 1G. The upper trace represents the channel for the material endogenously
present in DNA Ž mrz s 186. whereas the lower trace represents the channel for the internal standard Ž mrz s 188..

ple size in the latter study was rather small but no
obvious differences in M 1G levels were detected
between the two smokers included in the study and a
very small but statistically significant difference was
seen between men and women. A study of M 1G
levels in unused human pancreas transplant samples
demonstrated a bimodal distribution. The levels in
approximately half the samples fell below the limit
of detection of the assay and the levels in the other
half ranged from 2–50 per 10 8 nucleosides w44x.
As indicated above, the specificity of the GCrEC
NCIrMS assay is ensured by a combination of
chemical, immunochemical and spectroscopic pre-
requisites. However, it is possible to directly verify
the presence of M 1G residues in DNA by liquid
chromatographyrmass spectrometry ŽLCrMS.. A
sample of human liver DNA that exhibited a high
concentration of M 1G by GCrEC NCIrMS analysis
was processed in the absence of an internal standard
and the M 1G-containing fractions were isolated w45x.
The isolated material was then injected into a triple
quadrupole LCrMS with electrospray detection. As
observed with most deoxynucleosides, source-in-
duced fragmentation led to loss of the deoxyribose
 L.J. Marnettr Mutation Research 424 (1999) 83–95 91

Table 1 quadrupole. The spectrum of ions generated was

Levels of M 1G in human tissues identical to that produced by tandem mass spectral
Tissue Averagea Range Technique Reference analysis of an authentic standard of M 1G ŽFig. 2..
Liver 9 Ž ns6. 5–12 GCrMS w41x This establishes unequivocally that M 1G residues are
White cells 0.6 Ž ns10. 0.5–0.8 GCrMS w42x present in human liver DNA. Similar verification has
Pancreas 3.2 Ž ns 27. - 0.1–5 GCrMS w47x
32 been carried out on human leukocyte DNA samples
White cells 2.6 Ž ns 26. 1–5 P w48x
White cells b 0.9 Ž ns 7. 0.2–2.5 32
P w39x but in that case the low levels of M 1G required a
White cells c 11 Ž ns6. 1.2–28 32
P w39x mass fragmentographic approach. The offset voltage
Breast d 0.2 Ž ns 51. 0.05–1.3 32 P w40x in the second quadrupole was programmed to pro-
Breast e 0.08 Ž ns 28. 0.02–0.19 32 P w40x gressively change the fragmentation pattern of the
32
Breast 3 Ž ns 7. 0.7–5.6 P w48x
protonated molecular ion and the expected alteration
a
All levels represent number of adducts per 10 7 nucleotides. in fragmentation pattern for M 1G was observed w42x.
b
Women fed high monounsaturated fat diet. A concern in the analysis of endogenous DNA
c
d
Women fed high polyunsaturated fat diet. adducts is the possibility that they are artifactually
Normal tissue from breast cancer patients. generated during the workup. This is especially seri-
e
Normal breast tissue from non-cancer bearing women.
ous if one is studying adducts that could be gener-
ated by oxidative processes that might occur post
mortem or ex vivo. We have used different ap-
moiety in the first quadrupole. The protonated proaches to minimize this problem in the analysis of
molecular ion corresponding to M 1G Ž mrz s 188. M 1G w41x. During tissue handling and DNA isola-
was then collisionally dissociated in the second tion, large amounts of antioxidants were added to all
quadrupole and the product ions analyzed in the third buffers to minimize post mortem lipid peroxidation.

Fig. 2. Mass spectrum of the collisionally dissociated mrz s 188 ion produced during the analysis of a human liver DNA sample.
92 L.J. Marnettr Mutation Research 424 (1999) 83–95
Interestingly, inclusion of butylated hydroxyanisole
Ž10 mM. or vitamin E Ž10 mM. did not lower the
levels of M 1G detected in liver tissue but inclusion
of vitamin C increased them approximately 2-fold.
We attribute this to the long-recognized ability of
ascorbic acid to support iron-catalyzed lipid per-
oxidation.
The other approach we have employed to assess
possible artifactual M 1G production is the inclusion
of isotopically labeled dG in the buffer used for
DNA digestion ŽScheme 6. w41x. Deoxyguanosine,
chemically synthesized to contain four 13 C and one
15
N atoms, was added to the digestion mixture in
amounts comparable to the amounts of dG in the
DNA sample. If any MDA was present as a result of
lipid peroxidation during tissue manipulation, it
would react with the isotopically labeled dG to form
M 1G that would yield a PFB derivative in GCrEC
NCIrMS with a mrz s 191. No such material was
detected in any of our analyses.
Analysis of rat tissues for M 1G by GCrEC
NCIrMS reveals a level in liver Ž; 6 per 10 7 nucle-
osides. comparable to that detected in human beings.
The level in 2 year old rats is nearly double that in 6
week old rats. Interestingly, analysis of testis DNA
samples from the same rats reveals that the levels of
M 1G are below the limit of detection of the assay Ž1
per 10 8 nucleosides.. When rats are exposed to the
lipid peroxidation stimulus CCl 4 by oral intubation,
Scheme 6. Isotopic probe for artifactual generation on M 1G ex
vivo.
the levels of M 1G in liver increases by nearly 100%
w41x.
8. Perspective
The available evidence indicates that M 1G, and
possibly other MDA-derived adducts, is a highly
mutagenic adduct that is present in the genomic
DNA of human beings ŽTable 1.. The levels detected
by various methods are reasonably similar consider-
ing the differences in the basic analytical procedures.
What is lacking at present is an understanding of the
importance of the detection of this adduct in people.
This is a problem facing virtually everyone who
studies the chemistry and biology of DNA adducts.
How can one relate the level of a particular DNA
adduct to a disease outcome? This is especially
difficult for a disease like cancer which is multifac-
torial and involves not only mutations in multiple
genes but also proliferation, clonal expansion, and
escape from apoptosis.
Relating the level of an adduct to an outcome is a
challenge that will be approached primarily by cor-
relative studies. Does the adduct level go up or down
in a predictable or logical way in situations that
predispose or protect from development of the dis-
ease? Given the variety of endogenous DNA dam-
age, answering this question for a single adduct will
require studies with large population groups. Thus,
the analytical assays will not only need to be robust
and specific but also amenable to high throughput
formats that can be carried out by interested but not
expert investigators. 32 P-Postlabeling and GCrEC
NCIrMS assays are time consuming and require
extensive sample manipulation by highly skilled per-
sonnel. 32 P-Postlabeling offers the advantage of high
sensitivity with minimal requirements for DNA Ž10
mg.. Immunochemical approaches offer considerable
promise for population-based studies because of their
speed and sensitivity. For example, Leuratti et al.
recently reported the development of an immunoslot
blot assay for M 1G that uses the previously described
monoclonal antibody raised against this adduct w46x.
The assay was used to examine a large number of
gastric biopsy samples to probe for a relationship
between inflammation and MDA–DNA adduction.
 L.J. Marnettr Mutation Research 424 (1999) 83–95
Although there is always concern about cross-re-
activity in immunochemical reactions, the availabil-
ity of mass spectrometric approaches to adduct iden-
tification provides a backup that can be used to
confirm adduct level and structure when an interest-
ing trend is noted. GCrMS methods can be used to
confirm adducts levels and LCrMS methods can be
used to confirm structure. In fact, one might specu-
late that LCrMS methods will eventually evolve
into high throughput, robust assays. The sensitivity
of LCrMS has increased approximately three orders
of magnitude in the last 10 years to a point where it
is within an order of magnitude of GCrMS. For
example, the limit of detection of M 1G by LCrMS
is 3 per 10 7 from 1 mg of DNA. Impressive im-
provements are being made in interfaces, detectors,
plumbing and data reduction that will certainly trans-
late into major increases in sensitivity in the next
few years. Therefore, it seems only a matter of time
before DNA hydrolyzates are processed by LCrMS
in high throughput mode for quantitation of a range
of endogenously and exogenously derived adducts.
The availability of a general method for adduct
identification and quantitation will help determine an
inventory of DNA damage in individual human be-
ings, regardless of the source.
Acknowledgements
Research in the Marnett laboratory is supported
by research and center grants from the National
Institutes of Health ŽCA47479, ES00267, and
CA68485..
References
w1x N.A. Porter, Mechanisms for the autoxidation of polyunsatu-
rated lipids, Acc. Chem. Res. 19 Ž1986. 262–268.
w2x T.A. Dix, J. Aikens, Mechanisms and biological relevance of
lipid peroxidation initiation, Chem. Res. Toxicol. 6 Ž1993.
2–18.
w3x W.A. Pryor, Oxy-radicals and related species: their forma-
tion, lifetimes and reactions, Annu. Rev. Physiol. 48 Ž1986.
657–667.
w4x L.R.C. Barclay, S.J. Locke, J.M. MacNeil, J. VanKessel,
G.W. Burton, K.U. Ingold, Autoxidation of micelles and
model membranes. Quantitative kinetic measurements can be
93
made by using either water-soluble or lipid-soluble chain-
breaking antioxidants, J. Am. Chem. Soc. 106 Ž1984. 2479–
2481.
w5x A.R. Waldeck, R. Stocker, Radical-initiated lipid peroxida-
tion in low density lipoproteins: insights obtained from ki-
netic modeling, Chem. Res. Toxicol. 9 Ž1996. 954–964.
w6x W.A. Pryor, J.P. Stanley, A suggested mechanism for the
production of malondialdehyde during the autoxidation of
polyunsaturated fatty acids. Non-enzymatic production of
prostaglandin endoperoxides during autoxidation, J. Org.
Chem. 40 Ž1975. 3615–3617.
w7x J.D. Morrow, L.J. Roberts, The isoprostanes—current
knowledge and directions for future research, Biochem. Phar-
macol. 51 Ž1996. 1–9.
w8x M. Hamberg, B. Samuelsson, On the mechanism of the
biosynthesis of prostaglandins E 1 and F1a , J. Biol. Chem.
242 Ž1967. 5336–5343.
w9x D.R. Janero, Malondialdehyde and thiobarbituric acid-reac-
tivity as diagnostic indices of lipid peroxidation and perox-
idative tissue injury, Free Radic. Biol. Med. 9 Ž1990. 515–
540.
w10x H. Esterbauer, Lipid peroxidation products: formation, chem-
ical properties and biological activities, in: G. Pli, K.H.
Cheeseman, M.U. Dianzani, T.F. Slater ŽEds.., Free Radicals
in Liver Injury, IRL Press, Oxford, 1985, pp. 29–47.
w11x H. Esterbauer, P. Eckl, A. Ortner, Possible mutagens derived
from lipids and lipid precursors, Mutat. Res. Rev. Genet.
Toxicol. 238 Ž1990. 223–233.
w12x S.L. Hazen, F.F. Hsu, A. d’Avignon, J.W. Heinecke, Human
neutrophils employ myeloperoxidase to convert a-amino
acids to a battery of reactive aldehydes: a pathway for
aldehyde generation at sites of inflammation, Biochemistry
37 Ž1998. 6864–6873.
w13x R.G. Nath, F.-L. Chung, Detection of exocyclic 1, N 2-pro-
panodeoxyguanosine adducts as common DNA lesions in
rodents and humans, Proc. Natl. Acad. Sci. U.S.A. 91 Ž1994.
7491–7495.
w14x R.J. Shamberger, T.L. Andreone, C.E. Willis, Antioxidants
and cancer: IV. Initiating activity of malonaldehyde as a
carcinogen, J. Natl. Cancer Inst. 53 Ž1974. 1771–1773.
w15x S.M. Fischer, S. Olge, L.J. Marnett, S. Nesnow, T.J. Slaga,
The lack of initiating andror promoting activity of sodium
malondialdehyde on Sencar mouse skin, Cancer Lett. 19
Ž1983. 61–66.
w16x F.H. Mukai, B.D. Goldstein, Mutagenicity of malondialde-
hyde, a decomposition product of peroxidized polyunsatu-
rated fatty acids, Science 191 Ž1976. 868–869.
w17x L.J. Marnett, M.A. Tuttle, Comparison of the mutagenicities
of malondialdehyde and side products formed during its
chemical synthesis, Cancer Res. 40 Ž1980. 276–282.
w18x A.K. Basu, L.J. Marnett, Unequivocal demonstration that
malondialdehyde is a mutagen, Carcinogenesis 4 Ž1983.
331–333.
w19x J.W. Spalding, Toxicology and carcinogenesis studies of
malondialdehyde sodium salt Ž3-hydroxy-2-propenal, sodium
salt. in F344rN rats and B6C3F1 mice, NTP Technical
Report 331 Ž1988. 5–13.
94 L.J. Marnettr Mutation Research 424 (1999) 83–95
w20x L.J. Marnett, H.K. Hurd, M.C. Hollstein, D.E. Levin, H.
Esterbauer, B.N. Ames, Naturally occurring carbonyl com-
pounds are mutagens in Salmonella tester strain TA104,
Mutat. Res. 148 Ž1985. 25–34.
w21x E. Schauenstein, H. Esterbauer, H. Zollner, Malondialde-
hyde, in: E. Schauenstein, H. Esterbauer, H. Zollner ŽEds..,
Aldehydes in Biological Systems, Pion, London, 1977, pp.
133–140.
w22x H. Krokan, R.C. Grafstrom, K. Sundqvist, H. Esterbauer,
C.C. Harris, Cytotoxicity, thiol depletion and inhibition of
O 6-methylguanine–DNA methyltransferase by various alde-
hydes in cultured human bronchial fibroblasts, Carcinogene-
sis 6 Ž1985. 1755–1759.
w23x H. Seto, T. Okuda, T. Takesue, T. Ikemura, Reaction of
malondialdehyde with nucleic acid: I. Formation of fluores-
cent pyrimidow1,2-axpurin-10Ž3 H .-one nucleosides, Bull.
Chem. Soc. Jpn. 56 Ž1983. 1799–1802.
w24x L.J. Marnett, A.K. Basu, S.M. O’Hara, P.E. Weller,
A.F.M.M. Rahman, J.P. Oliver, Reaction of malondialdehyde
with guanine nucleosides: formation of adducts containing
oxadiazabicyclononene residues in the base-pairing region, J.
Am. Chem. Soc. 108 Ž1986. 1348–1350.
w25x V. Nair, G.A. Turner, R.J. Offerman, Novel adducts from the
modification of nucleic acid bases by malondialdehyde, J.
Am. Chem. Soc. 106 Ž1984. 3370–3371.
w26x K. Stone, M. Ksebati, L.J. Marnett, Investigation of the
adducts formed by reaction of malondialdehyde with adeno-
sine, Chem. Res. Toxicol. 3 Ž1990. 33–38.
w27x K. Stone, A. Uzieblo, L.J. Marnett, Studies of the reaction of
malondialdehyde with cytosine nucleosides, Chem. Res. Tox-
icol. 3 Ž1990. 467–472.
w28x A.K. Chaudhary, G.R. Reddy, I.A. Blair, L.J. Marnett, Char-
X
acterization of an N 6-oxopropenyl-2 -deoxyadenosine adduct
in malondialdehyde-modified DNA using liquid chromatog-
raphy electrospray ionization tandem mass spectrometry,
Carcinogenesis 17 Ž1996. 1167–1170.
w29x P.C. Dedon, J.P. Plastaras, C.A. Rouzer, L.J. Marnett, Indi-
rect mutagenesis by oxidative DNA damage: formation of
the pyrimidopurinone adduct of deoxyguanosine by base
propenal, Proc. Natl. Acad. Sci. U.S.A., 1998, submitted.
w30x M. Benamira, K. Johnson, A. Chaudhary, K. Bruner, C.
Tibbetts, L.J. Marnett, Induction of mutations by replication
of malondialdehyde-modified M13 DNA in Escherichia coli:
determination of the extent of DNA modification, genetic
requirements for mutagenesis, and types of mutations in-
duced, Carcinogenesis 16 Ž1995. 93–99.
w31x G.R. Reddy, L.J. Marnett, Synthesis of an oligodeoxynu-
cleotide containing the alkaline labile malondialdehyde–
deoxyguanosine adduct pyrimidow1,2-axpurin-10Ž3 H .-one, J.
Am. Chem. Soc. 117 Ž1995. 5007–5008.
w32x S.P. Fink, G.R. Reddy, L.J. Marnett, Mutagenicity in Es-
cherichia coli of the major DNA adduct derived from the
endogenous mutagen malondialdehyde, Proc. Natl. Acad.
Sci. U.S.A. 94 Ž1997. 8652–8657.
w33x K.A. Johnson, S.P. Fink, L.J. Marnett, Repair of propan-
odeoxyguanosine by nucleotide excision repair in vivo and in
vitro, J. Biol. Chem. 272 Ž1997. 11434–11438.
w34x C.E. Vaca, P. Vodicka, K. Hemminki, Determination of
X X
malonaldehyde-modified 2 -deoxyguanosine-3 -monophos-
32
phate and DNA by P-postlabeling, Carcinogenesis 13
Ž1992. 593–599.
w35x A. Kautiainen, C.E. Vaca, F. Granath, Studies on the rela-
tionship between hemoglobin and DNA adducts of malon-
aldehyde and their stability in vivo, Carcinogenesis 14 Ž1993.
705–708.
w36x M.-Y. Wang, J.G. Liehr, Induction by estrogens of lipid
peroxidation and lipid peroxide-derived malonaldehyde–
DNA adducts in male Syrian hamsters: role of lipid peroxida-
tion in estrogen-induced kidney carcinogenesis, Carcinogene-
sis 16 Ž1995. 1941–1945.
w37x M.Y. Wang, J.G. Liehr, Lipid hydroperoxide-induced en-
dogenous DNA adducts in hamsters: possible mechanism of
lipid hydroperoxide-mediated carcinogenesis, Arch. Biochem.
Biophys. 316 Ž1995. 38–46.
w38x C.E. Vaca, J.-L. Fang, M. Mutanen, L. Valsta, 32 P-Postlabel-
ing determination of DNA adducts of malonaldehyde in
humans: total white blood cells and breast tissue, Carcino-
genesis 16 Ž1995. 1847–1851.
w39x J.L. Fang, C.E. Vaca, L.M. Valsta, M. Mutanen, Determina-
tion of DNA adducts of malonaldehyde in humans: effects of
dietary fatty acid composition, Carcinogenesis 17 Ž1996.
1035–1040.
w40x M. Wang, K. Dhingra, W.N. Hittleman, J.G. Liehr, M. de
Andrade, D. Li, Lipid peroxidation-induced putative malon-
dialdehyde–DNA adducts in human breast tissues, Cancer
Epidemiol. Biomarkers Prev. 5 Ž1996. 705–710.
w41x A.K. Chaudhary, M. Nokubo, G.R. Reddy, S.N. Yeola, J.D.
Morrow, I.A. Blair, L.J. Marnett, Detection of endogenous
malondialdehyde–deoxyguanosine adducts in human liver,
Science 265 Ž1994. 1580–1582.
w42x C.A. Rouzer, A.K. Chaudhary, M. Nokubo, D.M. Ferguson,
G.R. Reddy, I.A. Blair, L.J. Marnett, Analysis of the malon-
X
dialdehyde-2 -deoxyguanosine adduct, pyrimidopurinone, in
human leukocyte DNA by gas chromatographyrelectron cap-
ture negative chemical ionizationrmass spectrometry, Chem.
Res. Toxicol. 10 Ž1997. 181–188.
w43x C.L. Sevilla, N.H. Mahle, N. Eliezer, A. Uzieblo, S.M.
O’Hara, M. Nokubo, R. Miller, C.A. Rouzer, L.J. Marnett,
Development of monoclonal antibodies to the malondialde-
hyde–deoxyguanosine adduct, pyrimidopurinone, Chem. Res.
Toxicol. 10 Ž1997. 172–180.
w44x F. Kadlubar, K. Anderson, N. Lang, P. Thompson, S.
MacLeod, M. Mikhailova, M. Chou, J. Plastaras, L. Marnett,
S. Haussermann, H. Bartsch, Comparison of endogenous
DNA adducts levels in human pancreas, Proc. Am. Assoc.
Cancer Res. 39 Ž1998. 286, Abstract.
w45x A.K. Chaudhary, M. Nokubo, T.D. Oglesby, L.J. Marnett,
I.A. Blair, Characterization of endogenous DNA adducts by
liquid chromatographyrelectrospray ionizationrtandem mass
spectrometry, J. Mass Spectrom. 30 Ž1995. 1157–1166.
w46x C. Leuratti, R. Singh, C. Lagneau, P.B. Farmer, L.J. Marnett,
D.E.G. Shuker, Quantitation of malondialdehyde–DNA dam-
age in human tissues using a sensitive immunoslot blot assay,
Proc. Am. Assoc. Cancer Res. 39 Ž1998. 286, Abstract.
 L.J. Marnettr Mutation Research 424 (1999) 83–95 95

w47x F.F. Kadlubar, K.E. Anderson, N.P. Lang, P.A. Thompson,
S.L. MacLeod, M.W. Chou, M. Mikhailova, J. Plastaras, L.J.
Marnett, S. Haussermann, J. Nair, I. Velie, H. Bartsch,
Comparison of endogenous DNA adduct levels in human
pancreas, Mutat. Res., 1998, in press.
w48x C.E. Vaca, J.-L. Fang, M. Mutanen, L. Valstar, 32 P-Post-
labeling determination of DNA adducts of malonaldehyde in
humans: total white blood cells and breast tissue, Carcino-
genesis 16 Ž1995. 1847–1851.