Biological significance of DNA adducts: summary of discussion of expert panel
Increasingly, data on DNA adducts are being used as part of the hazard and risk assessment process for chemicals. Because there is no comprehensive database to demonstrate a specific biological consequence for some DNA adducts, a workshop was convened to consider how adduct data could be most appropriately integrated into the entire database for a given chemical.
Advances in analytical methodology, particularly in sensitivity and specificity of methods, have resulted in some cases in the ability to quantify one DNA adduct per 109 nucleotides. With the use of 14C-labelled compounds measurements of levels down to one DNA adduct per 1011 nucleotides is achievable. In comparison, the haploid genome in humans contains 4.6 × 109 nucleotides.
Standard biological assays of effects, such as mutagenicity, are currently not nearly as sensitive. Typical background levels of mutation frequencies in mammalian cell mutation assays mostly range from one mutation per 105–106 cells. Additional measures, such as preselection of clones with reduced background mutation frequency, or large numbers of cells assayed, can reduce background mutation frequency to one per 107 cells. The sensitivity of detection of mutations resulting from an exogenous carcinogen is limited by the presence of these background mutations. The limited availability of selective systems capable of detecting rare mutations and the practical factors involved in undertaking large scale studies, require doses at least two orders of magnitude greater than those sufficient to produce a detectable increase in adducts, in order to result in a detectable increase in mutations.
Advances in our overall understanding of the molecular processes leading to an expressed mutation, the biological endpoint of concern, have raised additional questions regarding the biological significance of DNA adducts. Starting with the formation of a DNA adduct (which may or may not escape repair), replication/fixation (or not), and eventual expression of a mutant protein (or not), there are many opportunities for a DNA adduct not to be expressed as a mutation, despite the presence of measurable increases in these adducts. The approach sometimes used in hazard and risk assessment, of equating every DNA adduct with a mutation, is not supported by our current understanding of the molecular mechanisms that result in expressed mutations.
There is a growing debate about the biological significance of low levels of DNA adducts, engendered by remarkable increases in sensitivity of the analytical tools used for their detection. In light of this debate, a group of academic and industrial scientists participated in a workshop to address some specific questions designed to provide a perspective on the relevance of low levels of DNA adducts to human hazard and risk assessment.
The three questions under discussion were as follows:
? What does an increase in chemical-specific DNA adducts indicate?
? With particular reference to low levels of DNA adducts, what data demonstrate a consequence or otherwise for human health?
? What status should DNA adduct measurements have in overall hazard and risk assessment?
1. Report on session 1
1.1. Question: What does an increase in DNA adducts indicate?
The first session, chaired by Drs. Jim Swenberg and David Phillips, addressed the role of DNA adducts in mutagenicity and carcinogenicity. A short introduction on what types of DNA adducts are produced under different chemical treatments and how these might produce mutations was discussed. The N7-alkylguanine adducts were identified as the most commonly formed and analyzed adducts (see below). DNA adducts are recognized by cellular repair mechanisms that have evolved to operate in concert to restore the original base sequence. For example, in base excision repair the adducted base is first cut out by a glycosylase enzyme to form a single base gap in the affected DNA strand (an apurinic or AP site). A new base is then inserted by polymerase enzymes and the strand is repaired with a DNA ligase. The presentation was illustrated with examples.
According to Dr. Swenberg, all human tissues contain DNA adducts, AP sites, and strand breaks at a level of around 1 lesion in 106 nucleotides (nt). This endogenous background of DNA lesions, caused by low levels of naturally occurring depurination, endogenous alkylation, and reactive oxygen species, is continuously repaired by normal cellular defence mechanisms. Exogenous agents such as radiation, environmental chemicals, and changes in life-style (i.e., tobacco use) add to the cellular DNA adduct burden. The consequences of this will depend upon what types of adducts are formed, the quantity, how rapidly adducts are lost or repaired, the efficacy/fidelity of the repair, and the speed and extent of cell proliferation.
A DNA adduct is produced after an electrophilic chemical covalently reacts with a DNA base. This new structure, if not repaired, could lead to an error during replication and subsequently to a mutation. All four bases in DNA can be adducted but one adduct that is very commonly measured is N7-alkylguanine. This adduct, because of its shape and location, does not distort the DNA helix and, distinct from the others, the N7 adduct is unlikely to cause an error in replication if present. Evidence supports this hypothesis. Dr. Swenberg showed the proportions (%) of DNA adducts produced after treatment of DNA in vitro with three different alkylating agents, as follows:
Of these adducts, O6-guanine, O2-thymine, and O4-thymine adducts can lead to a mutagenic event if present when the DNA is copied. However, it is important to recognize that in vitro adduct formation can be quite different from steady state levels that are present after multiple, in vivo exposures. In rat liver, O6-alkylguanine is rapidly removed, while O2- and O4-thymine adducts accumulate.
Dr. Swenberg moved on to discuss ethylene oxide (EO) and propylene oxide (PO). All tissues investigated so far have demonstrated quantifiable levels of N7-hydroxyethylguanine (HEG) adducts, the same adduct formed following experimental exposures to EO. O6-Hydroxyethylguanine reached steady state concentrations by 2 weeks, while N3-hydroxyethyladenine adducts attained steady state levels in a few days. Thus, the number of HEG:N-3 HEA:O6-HEG became 250:1:1 after 4 weeks of exposure to 300 ppm EO. With EO-exposed rats and mice, he showed that after 4 weeks of exposure, N7 HEG adducts were significantly increased over the control range of 1–3 HEG in 107 nt in liver, spleen, lung, and brain at exposure levels between 3 and 100 ppm EO. At 10 ppm EO and above, HEG adduct levels were higher in brain and spleen than in liver and lung and in all tissues there was a linear increase in HEG adducts with increasing exposure. This was not consistent with the incidence of tumors in these studies, where lifetime exposure of rats to 33 or 100 ppm EO, but not to 10 ppm EO, resulted in an increased incidence of mononuclear-cell leukemia and brain tumors, respectively.
In rats exposed to 500 ppm PO for 20 days, N7 HPG adducts were determined in a number of tissues and the rate of adduct repair was measured as the rate of depurination or apurinic (AP) sites. In nasal respiratory epithelium (NRE), the target tissue for carcinogenicity, the depurination rate increased from a spontaneous rate of 1.6 AP sites/106 nt/day in controls to 16 AP sites/106 nt/day; NRE was the only tissue in which tumors developed following lifetime inhalation exposure to PO. In spite of an increased rate in depurination and markedly increased levels of N7 HPG adducts in NRE, there was no evidence of an increase in the number of AP sites. The conclusion drawn was that base excision repair intermediates such as AP sites are repaired with great efficiency. Mutations may reflect formation of low numbers of promutagenic DNA adducts.
Dr. Swenberg concluded that endogenous adducts, AP sites, and strand breaks are present in cells at 1 in 106 nt and that, when <1 chemical-specific DNA adduct in 108 nt is present, they may induce DNA repair enzymes without causing mutations. This level represents a practical threshold. In addition, he concluded that not all DNA adducts are of equal concern and N7-alkylguanine, though the most common and easy to measure, may be of the least concern.
Dr. David Phillips built on the points made by the previous speaker. He showed that, in large mouse bioassays for some carcinogens, both tumor formation and DNA adduct formation show linearity at low doses. However, for 2-acetylaminofluorene, DNA adduct formation was linear at low doses in liver and bladder, while tumor induction was linear in liver and highly non-linear in bladder. Therefore the requirements for carcinogenesis should be considered to be:
Adduct formation.
Adduct persistence/accumulation.
Cell division.
Technology has advanced so rapidly in recent years that it is now possible to readily detect adduct levels at 1 in 109 bases whereas the frequency of tumors likely to result from adducts at this level will probably be too low to detect in bioassays with standard numbers of animals, thus:
While there may be adduct levels at which there is no observable biological effect, there are at present insufficient data on which to set a threshold level for biological significance.
The point of this was to demonstrate that not all carcinogens produce adducts and that the presence of adducts does not always mean tumors. Tamoxifen for example produces DNA adducts and tumors in rat liver, but although DNA adducts are detected in mouse liver, no increases in tumors were observed. Toremifene, an analogue of tamoxifen, without demonstrable carcinogenicity, produces 100-fold fewer adducts in rat liver (tamoxifen 116 adducts/108 nt; toremifene 0.85 adducts/108 nt). There are examples of other compounds that produce low levels of adducts but no significant increases in tumors above the background in the animal model used. Anethole, elemicin, myristycin, dill apiol, and parsley apiol are some examples.
Dr. Werner Lutz presented some data available to quantitatively relate DNA adduct levels and tumor incidence in animal bioassays. The list comprised 27 chemicals, of all major structural classes of carcinogens. For the correlation with tumor incidence, the DNA adduct numbers measured at a given dose were normalized to the dose that resulted in a 50% tumor incidence under the conditions of a two year bioassay (TD50 dose). In rat liver, the calculated adduct concentration correlating with a 50% hepatocellular tumor incidence spanned from 53 to 2083 adducts per 108 nucleotides, for aflatoxin B1, tamoxifen, IQ, MeIQx, 2,4-diaminotoluene, and dimethylnitrosamine (ordered by increasing number of adducts). In mouse liver, the respective figures were 812–5543 adducts per 108 nucleotides, for ethylene oxide, dimethylnitrosamine, 4-aminobiphenyl, and 2-acetylaminofluorene, which are carcinogens with a variety of target organs. The observed span (40-fold in rats, seven-fold in mice) may reflect differences between the varied potential for different DNA adducts to lead to critical mutations. If additional carcinogens fit in this narrow range, the measurement of DNA adduct levels in target tissue may have the potential to be not only an exposure marker but an individual cancer risk marker. However, for toremifene and styrene, low levels of DNA adducts were detected in rat liver at the end of negative long-term bioassays which showed no evidence of liver carcinogenicity. For a cancer risk assessment at low levels of DNA damage, treatment-related increments must be discussed in relation to the background DNA damage and its inter- and intra-individual variability.
Dr. Kari Hemminki shared a list of human cancers for which the target tissues were analyzed for chemical-specific DNA adducts, and the levels of DNA adducts that were associated with an increased risk of cancer. These included lung tissue from smokers (B(a)P DNA adducts; 10/108 nt), and colon tissue (PhiP-G DNA adducts; 3/108 nt).
All participants agreed that, with exposure of animals to DNA-reactive carcinogens, DNA adducts are frequently seen in tissues where tumors do not develop, and they concluded that DNA adducts alone are not sufficient to cause cancer.
Looking towards future ways of exploring this issue, Dr. Orphanides gave a brief presentation on the application of genomic and proteomic techniques and their applicability in exploring changes that occur after chemical treatment. These approaches allow a holistic analysis of molecular responses to toxicants with potential to reveal novel insights into cellular responses to genotoxic chemicals. Among the potential benefits they offer are: (1) an appreciation of cellular pathways involved in adaptive and adverse responses to DNA damage, (2) an understanding of qualitative and quantitative differences in biological response and correlation with DNA adducts, and (3) an appreciation of differences in biological effects induced by different adducts.
1.2. Consensus statements from Question 1
? The presence of DNA adducts alone is not evidence of mutation.
? "While there may be [DNA] adduct levels at which there is no observable biological effect, there are at present insufficient data on which to set a threshold level for biological significance".
? In the absence of any other toxicological data, the formation of chemical-specific DNA adducts should be considered an adverse effect, i.e., one which potentially compromises the organism.
? Other appropriate scientific data provides necessary context to the interpretation of DNA adduct data and when available must be considered in combination with DNA adduct data if one is proposing to use the latter in the hazard and risk assessment process.
2. Report on session 2
2.1. Question: With particular reference to low levels of DNA adducts, what data demonstrate a potential consequence or otherwise for human health?
The session, chaired by Drs. Gary Williams and Werner Lutz, started by defining what might be considered as a ‘low level of adducts,’ and what health effects are related to adduct formation. The participants agreed that for the purpose of the discussion, a ‘low level of adducts’ could be established as one adduct per 109–1010 nucleotides (i.e., 10?10–10?9), and the principal health effects related to adduct formation are mutations, heritable effects, and cancer, as well as cell ageing. As to what data demonstrate a role for DNA adducts, it was agreed that an ‘effect’ must be a quantifiable, expressed mutagenic response, and that in order to demonstrate a heritable effect in germ cells, it is necessary to demonstrate such a response, not just the presence of adducts.
An increase in the number of adducts increases the response, and well-conducted animal studies have demonstrated that when there is an increase in tumor formation, there is often an increase in the number of adducts. The human data are more variable. For example, the data associating tumors with particular tobacco-specific carcinogens are difficult to analyze due to the presence of promoting agents as well as DNA-reactive chemicals in cigarette smoke, and divergent results have been found depending upon the tumor type under investigation. For instance, in normal pharynx and larynx tissue obtained from cancer patients at surgery, levels of bulky, PAH-related DNA adducts correlated with the level of smoking but not with age. Since age at diagnosis of a smoking-related tumor could be regarded as an indicator of an individual cancer risk (the higher the susceptibility, the younger at tumor manifestation), DNA adduct levels did not appear to be of predictive value for time of disease onset.
However, it should be noted that the measurement of adducts at the time of diagnosis may not always be a true reflection of the adduct levels in previous years, when initiating mutations may have occurred, and there are also some good examples of prospective studies in which adduct levels are of predictive value in subsequent development of cancer, e.g., aflatoxin adduct levels in urine and liver cancer, and smoking-related bulky adducts in blood and lung cancer.
The role of a specific adduct in the etiology of a genetic response may be assessed by introducing it into cellular DNA via a shuttle vector. Such studies have demonstrated that not all non-repaired adducts give rise to mutation. For example, there is some evidence that some N7-alkylguanine adducts are not mutagenic. Even for the adducts which are promutagenic, their potencies to induce mutations vary widely, and none gives rise to a 100% risk of increased mutation rate.
A level of 10?8 promutagenic adducts is considered to represent a risk factor, however, this is unlikely to be the case for all adducts, i.e., not for the non-promutagenic adducts. Therefore, risk will depend on the level of adducts and also on the qualitative adduct profile (promutagenic vs. non-promutagenic). Consideration must be given to examples of positive bioassays with low levels of promutagenic adducts (benzo[a]pyrene) and negative bioassays with high levels of non-promutagenic adducts, such as some N7-alkylguanine adducts (ethylene). Furthermore, although an overall adduct incidence of 10?8 conveys a level of risk, we cannot draw strong conclusions about lower incidences. However there are no data supporting such risk for non-mutagenic DNA adducts.
2.2. Consensus statements from session 2
? Not every type of DNA adduct results in mutation (promutagenic vs. non-promutagenic).
? Lower DNA adduct levels imply lower risk.
? The quantitative and qualitative nature of the DNA adduct profile is important in defining risk.
? As the biological consequences of different adducts vary considerably, it will be necessary to consider which adducts are present in order to determine any relationship to risk.
N.B. Some participants felt that an adduct frequency of 1 promutagenic adduct in 108 nucleotides may be suggestive of cancer risk, depending on the adduct profile, whereas others thought that it was not possible to set such a level of adducts, as the biological consequences of different adducts can vary considerably.
3. Report on session 3
3.1. Question: What status should DNA adduct measurements have in overall hazard and risk assessment?
The session was chaired by Drs. Jim Parry and Len Levy. The previous sessions have indicated that it is important to distinguish between ‘potential hazard’ and ‘actual risk’ when evaluating the safety of chemicals. However, some regulatory regimes are based primarily upon hazard assessment, rather than on the determination of actual risk under likely exposure scenarios.
For example, even where data are available demonstrating the absence of effect at low doses, a ‘threshold’ concentration is not defined below which no toxic effect is produced without a clear understanding of the mechanism that explains the observed response. Therefore, in the absence of sufficiently detailed mechanistic data, the conservative ‘linear’ approach is used in the hazard assessment process.
It is important to distinguish between "potential hazard" and "actual risk" when evaluating the safety of chemicals. The group agreed that the detection of DNA adducts provides vital data to assist in the interpretation of mechanistic data and as an indicator of exposure. The presence of DNA adducts alone should not be taken as evidence of mutagenicity in the hazard and risk assessment process. DNA adduct data can extend the observable range of data to improve confidence regarding the shape of the dose–response. It is, however, important to also look at other contributing factors such as cell proliferation, tumor incidence, and/or mutation incidence.
The relationship between DNA adduct formation and possible mutagenicity/genotoxicity is an important point to consider as these resultant effects may have implications both in terms of germ cell (possibly leading to heritable mutation) and/or somatic cell effects (possibly leading to cancer). These toxicological end-points are crucial in terms of overall hazard assessment process and can have important regulatory sequelae. In some regions the mere identification of DNA adducts in testis after a chemical exposure, even in the absence of any evidence of subsequent heritable genetic damage, may be enough to result in the chemical attracting a regulatory ‘classification’ for possible heritable effects.
As discussed previously during the workshop, it is reasonable to postulate that, despite the formation of DNA adducts as a result of a chemical exposure, there may nevertheless be threshold levels for effects from DNA adducts below which no significant increases in induced mutations may be detectable. There are a number of factors which may influence the existence of a threshold for genotoxic activity, including spontaneous depurination, DNA repair (eliminating DNA adducts before being processed into mutation), and multiple targets (such as spindle fiber effects). Each of these factors may modify the risk associated with the exposure.
3.2. Consensus statements from session 3
In terms of the somatic cell/cancer end-point the workshop concluded:
? The detection of DNA adducts provides an indicator of exposure of somatic cells to a DNA reactive agent.
? DNA adducts provide important information on the mechanism of action of a particular chemical.
? DNA adducts need to be measured and defined (in terms of their identity and of possible mutational spectra) in the tissues where a biological response is being investigated.
? In the absence of other data, demonstration of DNA adducts in this tissue may indicate a potential carcinogenic hazard but their biological relevance needs to be assessed on a case-by-case basis.
? DNA adducts alone would not lead to classification as a carcinogen.
In terms of the germ cell/heritable effects end-point the workshop concluded:
? The detection of DNA adducts provides a valuable indicator of exposure of germ cells to a DNA reactive agent.
? DNA adduct data provides complementary and often critical information which aids in interpretation, such as providing an understanding of mechanisms of chemical and biological activity.
? Currently, there is no evidence to indicate that DNA adducts are converted to mutations at higher frequencies in germ cells than in somatic cells.
? In the case of germ cell hazard assessment the primary factor for consideration should be the availability of data assessing heritable genetic changes in germ cells, using a methodology appropriate to the chemical under consideration and by a route relevant to human exposure.
? Thus, DNA adduct data alone, at exposure concentrations that produce no elevations in mutations in vivo in somatic cells and in the absence of a positive heritable effect in an appropriate assay, are not suitable for the risk assessment of the inherited effects of chemicals.
The fundamental point is that whatever endpoint is of concern, the results of appropriate epidemiological and/or animal studies, should not be supplanted by DNA adduct data taken in isolation.
Overall the workshop agreed with the IWGTP statement that "While there may be [DNA] adduct levels at which there is no observable biological effect, there are at present insufficient data upon which to set a threshold level for biological significance". Technology is now becoming available for the refined measurement of very low levels of DNA adducts and of their possible biological effects. This in turn raises the issue as to whether and/or how the demonstration of a possible threshold of mutagenic effect for DNA adducts should influence the hazard and risk assessment of a chemical.
The workshop concluded that, current scientific data does not generally prove or disprove that the low levels of DNA adducts which can presently be measured represent a hazard. Although they may be considered as a potential hazard, the magnitude of the risk associated with these low levels of adducts is unknown and could be immeasurably small. Information on adduct profiles and the integration of adduct data with other toxicity data remain critical to determination of risks, if any, associated with low levels of DNA adducts. Further research will provide perspective on potential hazard and risk, if any, associated with low levels of DNA adducts.
4. Discussion leaders
P.B. Farmer, University of Leicester, UK
L.S. Levy, MRC Institute for Environment and Health, Leicester, UK
W.K. Lutz, University of Wuerzburg, Germany
J.M. Parry, University of Wales, Swansea, UK
D.H. Phillips, Institute of Cancer Research, Sutton, UK
J.A. Swenberg, University of North Carolina, Chapel Hill, NC, USA
G.M. Williams, New York Medical College, NY, USA
5. Participants
Dr. Peter B. Farmer, University of Leicester, UK
Dr. Kari Hemminki, Karolinska Institute, Sweden
Dr. Leonard S. Levy OBE, University of Leicester, UK
Dr. Werner K. Lutz, University of Wuerzburg, Germany
Dr. George Orphanides, Syngenta Central Toxicology Laboratory, UK
Dr. J.M. Parry, University of Wales Swansea, UK
Dr. David Phillips, Institute of Cancer Research. UK
Dr. Alain Sarasin, Laboratory of Genetic Instability & Cancer, France
Dr. Dan Segerback, Karolinska Institute, Sweden
Dr. J.A. Swenberg, University of North Carolina, USA
Dr. Kenneth W. Turteltaub, Lawrence Livermore National Laboratory, USA
Dr. Gary M. Williams, New York Medical College, USA
6. Representatives of host organisations
Dr. Marcy Banton, Lyondell Chemical Company, TX, USA
Dr. Otto Grundler, BASF, Germany
Dr. Anne P. LeHuray, American Chemistry Council, USA
Dr. Nigel Moore, BP Chemicals Ltd., UK
Mr. Mike Penman, ExxonMobil Chemical Europe, Belgium
Dr. Lynn H. Pottenger, Dow Europe, GmbH, Switzerland
Dr. Bob Priston, Shell Chemicals, UK
Mr. Mike Thomas, Lyondell Chemical Company, UK
References