Postlabeling methods for carcinogen-DNA adduct analysis.

Radioactive carcinogens have provided most of our present knowledge about the chemistry of interactions between carcinogens and biological systems. The requirement of radioactive carcinogens has restricted carcinogen-DNA binding studies to chemicals that are readily available in isotopically labeled form, i.e., a minute fraction of all potentially mutagenic or carcinogenic chemicals. To extend the scope of carcinogen-DNA binding studies, an alternative method, which does not require radioactive test chemicals, has been developed. In this approach, radioactivity (32P) is being incorporated into DNA constituents by polynucleotide kinase-catalyzed [32P]phosphate transfer from [gamma-32P]ATP after exposure of the DNA in vitro or in vivo to a nonradioactive, covalently binding chemical, and evidence for the alteration of DNA nucleotides is provided by the appearance of extra spots on autoradiograms of thin-layer chromatograms of digests of the chemically modified DNA. Quantitation of adduct levels is accomplished by scintillation counting. The sensitivity of the technique depends on the experimental conditions for 32P-labeling and on the chemical structure of the adducts. Greater sensitivity may be achieved if adducts can be separated as a class from the normal nucleotides. This is the case for an estimated 80% of all carcinogens, giving rise to bulky and/or aromatic substituents in DNA. Under the present conditions, one such adduct in 10(9) to 10(10) normal nucleotides can be detected. A total of approximately 80 compounds has been studied thus far Binding to DNA of rodent tissues was readily detected by the 32P-postlabeling assay for all known carcinogens among these compounds, and adducts were detected in DNA from human placenta of smokers. ImagesFIGURE 5.FIGURE 7.FIGURE 8.


Introduction
Any chemical capable of forming covalent bonds with DNA of somatic and reproductive mammalian cells in vivo is a potential mutagen, carcinogen, and teratogen. Since such genotoxic chemicals may be of natural or man-made origin, exposure to them cannot be completely eliminated, but human contact with them must be minimized. Methods enabling one to detect and quantify DNA binding potential of chemicals directly in vivo should be of great value in the detection of gene-altering chemicals in the environment, provided they can be applied to a large number of chemicals of diverse structure and their sensitivity is sufficient to detect low DNAbinding activities. Because covalent DNA binding of chemicals in experimental animals may range from one adduct in 103 normal nucleotides to one adduct in 109 to 1010 nucleotides, such methods should ideally be capable *Department of Pharmacology, Baylor College of Medicine, Texas Medical Center, Houston 77030. of detecting extremely low binding of the order of a single adduct per diploid mammalian genome of about 1.2 x 1010 DNA nucleotides.
As pointed out by Lutz (1,2), the covalent binding index (CBI), defined as ,umole of chemical bound per mole of DNA nucleotide/millimole of chemical administered per kilogram body weight of animal, exhibits a good quantitative correlation with the hepatocarcinogenic potency of chemicals of diverse structure. Therefore, methods for the analysis of adducts in DNA should not only detect but also quantitate DNA binding. The subject of this paper is to review our recent efforts to develop such an ultrasensitive method involving 32p_ postlabeling of adducts (3)(4)(5)(6) and to describe results obtained by applying the method to a number of genotoxic chemicals of diverse structure.

Materials and Methods
The sources of materials used in the 32P-postlabeling procedure as well as safety precautions and chromatographic and autoradiographic procedures have been re-  (6). Screen-intensified autoradiography was performed at -80°C. For in vivo modification of DNA, female BALB/c or CD-1 mice (25 g) and male Sprague-Dawley or Fischer rats (200 g) were maintained on standard laboratory diet and water ad libitum. A list of chemicals used for animal treatments is given in Table 1. For adduct detection in mouse skin DNA, the backs of mice were shaved with clippers 3 days prior to treatment, which was performed by topical application of four doses of 1.2 ,umole each of test compounds in 200 ,uL of solvent. Compounds were dissolved in acetone, except for azo dyes, which were dissolved in acetone/water (7:3, v/v). The treatments were at 0, 24, 48, and 72 hr for arylamines and derivatives, azo compounds, and nitro compounds, and at 0, 6, 30, and 54 hr for polycyclic aromatic hydrocarbons. Control mice were given 200 pL of solvent alone. DNA was isolated 24 hr after the last treatment. For adduct detection in mouse liver DNA, animals were given a single IP dose (150 mg/kg) of methylating agent in 0.1 mL of 0.9% NaCl; control mice received 0.9% NaCl alone. DNA was extracted 3 hr after administration. Alkenylbenzenes were given to mice by IP injection of compound (400 mg/kg) in 0.2 mL tricaprylin, and liver DNA was isolated 24 hr after treatment; control mice received tricaprylin alone. Dibenzo(c,g)carbazole was given SC to female CD-1 mice at a dose of 12 mg/kg in tricaprylin. For adduct detection in rat liver DNA, animals were given a single IP dose of test chemical (40 mg/kg) in 0.3 mL of dimethyl sulfoxide; control rats received vehicle alone. DNA was isolated 4 hr after administration. Treatment with mycotoxins is detailed below. 32P-postlabeling analysis of DNA adducts was performed as described previously (4,6) using 1 to 2 p,g of DNA for enzymatic digestion and 0.15 to 0.3 ,ug of DNA nucleotides for 32P-labeling. For adduct intensification, the labeling reaction mixture contained 400 ,uM DNA-P and 1.7 puM [y-wP]ATP (9120 Ci/mmole), and apyrase treatment (4) was omitted. For estimation of adduct levels, spots were excised from replicate maps and counted by Cerenkov assay. Appropriate blank areas of the chromatogram were also assayed and their count rates subtracted from the sample count rates. In the case of aromatic adducts, the amounts of 32P-labeled digest applied to the thin-layer chromatograms were 150 to 460 ,uCi for counting of adducts and 0.25 ,uCi for assaying normal nucleotides. Calculations were done according to the equations shown in Figure 1. Under the conditions of excess ATP (4,6), the relative adduct labeling

Results
For the 32P-postlabeling scheme to serve as a test for the capacity of chemicals to bind to DNA, it should be applicable to most or all covalently binding chemicals; thus a major question to be answered in the initial phase ofthe development ofthe method was whether a greater number of adducts of diverse structure was amenable to 32P-postlabeling by T4 polynucleotide kinase-catalyzed phosphorylation. We have thus far studied a total of ca. 80 chemicals comprising arylamines and derivatives, azo compounds, nitroaromatics, polycyclic aromatic hydrocarbons, heterocyclic polynuclear compounds, alkenylbenzenes, mycotoxins, and methylating agents. In every case, 32P-labeling of carcinogen-DNA derivatives could be readily detected, indicating that the 32P-postlabeling method can be applied to a very large number of chemically diverse adducts. Many of the compounds studied are listed in Table 1.   bisphosphates (I). 32P-labeling of nucleotides may be conducted under standard conditions (employing excess ATP over DNA-P) or under adduct intensification (ATP-deficient) conditions, as shown in Figure 2. The latter conditions afford the preferential labeling of many aromatic carcinogen-DNA adducts, which results in a greatly increased sensitivity of adduct detection (7). In addition to PEI-cellulose TLC, reversed-phase TLC on octadecylsilane (C18) layers was found to separate the normal nucleotides from the adduct nucleotides (6). To remove the labeled normal nucleotides and resolve the [32P]adducts, a four-directional (4-D) anion-exchange PEI-cellulose TLC system was developed (Fig. 3). In this procedure, freshly prepared labeled digest (14)(15)(16)(17)(18) ,uCi/,L) is being applied slowly to the origin (Fig. 3, OR) located close to the center of the thin-layer sheet (20 x 20 cm). Development is begun immediately after   sample application without drying of the origin area. Conditions for the various developments have been indicated in the legend of Figure 3. This procedure has been applied successfully to aromatic carcinogens having two to six aromatic rings; in the case of less aromatic carcinogens (such as alkenylbenzenes, sterigmatocystin, and aflatoxin B1), removal of normal nucleotides from the adducts was best accomplished by C18 reversed-phase TLC (see Fig. 4). The lower limit of adduct detection was found to be 1 adduct in (3.5-6) x 107 nucleotides for the standard 4-D PEI-cellulose TLC procedure, while one adduct in about 105 nucleotides could be detected if removal of normal nucleotides from the adduct nucleotides was carried out by reversed-phase TLC, the increase in sensitivity being due to reduced background radioactivity (6). The technique entails the simultaneous labeling of normal and adduct nucleotides, thereby enabling the accurate quantitation of DNA adduct levels. It is also possible to isolate the adducts first and then label them in the virtual absence of normal nucleotides, a technique that affords an increase in sensitivity of detection to one adduct in about 1010 normal nucleotides (K. Randerath and E. Randerath, manuscript submitted).
Some examples of applications of the postlabeling method follow; additional examples have been described (4)(5)(6). The method was applied to skin DNA from mice treated topically with the polycyclic aromatic hydrocarbons benzo(a)pyrene (BP), 7,12-dimethylbenz(a)anthracene (DMBA), and 3-methylcholanthrene (MC), respectively. As shown by autoradiography (Fig. 5), a large number of 32P-labeled MC-DNA adducts was detected in digests of DNA obtained from mouse skin at several time points after carcinogen treatment. Four of these adducts (spots 4, 7, 9, and 10) were highly persistent (Figs. 5c, 5d). As shown in Figure 6, substantial removal of MC-DNA adducts occurred during the first 2 weeks after carcinogen application, while adducts remaining thereafter underwent little change. Analogous results were reported for BP-DNA and DMBA-DNA adducts (5), with some adducts persisting in mouse epidermis and dermis for about 1 year after a single topical carcinogen application (7). These results raise the pos-sibility that the persistent adducts occupy specific genomic sites in quiescent cells where they may not be amenable to repair because of localized conformational alterations of DNA or shielding by associated proteins. In Figure 7, application of the 3P-postlabeling method to rat liver DNA after exposure in vivo to the mycotoxins aflatoxin B, and sterigmatocystin, respectively, is illustrated. These compounds are known potent hepatocarcinogens (8,9), and in vivo DNA binding of aflatoxin B1 has been studied extensively (10,11), but no such data were available for sterigmatocystin-DNA binding in vivo. Our studies on these mycotoxins revealed that the adducts seen on 32P-labeled fingerprints (Fig. 7) were mostly oligo-(diand tri-)nucleotides, because the enzymatic digestion under standard conditions (4) of DNA containing such adducts led to the formation of oligonucleotides (mostly diand trinucleotides) rather than to the formation of the usual mononucleotide adducts (12). Sterigmatocystin-DNA adducts were detectable in rat liver DNA as late as 3.5 6

14
Days after treatment ,uL acetone each. Digest of DNA from control mice that had received acetone only did not give any of the numbered spots. Separation was by the standard procedure (Fig. 3), and spots were visualized by autoradiography.  months after injection of a single dose (9 mg/kg) of sterigmatocystin to male Fischer rats (12). When equal doses were compared, sterigmatocystin led to an approximately 10 times lower level of DNA modification than did aflatoxin B1, in accord with a close to 10-fold lower carcinogenic potency of the former compound as compared to the latter (9). We have applied the 32P-postlabeling test to the analysis of the binding of a series of carcinogenic and noncarcinogenic alkenylbenzenes (13) to adult (14) and newborn (15) mouse liver DNA. Extensive carcinogenicity studies by Miller et al. (13) in mice had shown that, among these compounds, only three, i.e., safrole, estragole, and methyleugenol, were hepatocarcinogenic in the test animals. We wished to ascertain by 32P-postlabeling assay whether a correlation existed between the DNA-binding activities of these compounds in mouse liver and their biological activity or lack of activity in assays for carcinogenicity in this organ. As illustrated in Figure 8, our results showed that the known hepatocarcinogens exhibited the strongest binding to mouse liver DNA 24 hr after IP administration of a 10-mg dose (one adduct in 10,000-15,000 DNA nucleotides or 200-300 fmole adduct/,g DNA), while the other structurally related alkenylbenzenes, except eugenol, all bound to mouse liver DNA also. but at lower levels.
In Table 1, most of the compounds we have studied thus far are listed, together with the tissues investigated, the number of adducts detected for each compound, and an estimation of total adduct level for each compound. With the exception of anthracene, pyrene, and perylene, all the compounds gave rise to 32P-labeled adducts at the levels indicated. It appears noteworthy that within the group of polycyclic aromatic hydrocarbons, a good correlation was observed between the carcinogenic potency of individual compounds for mouse skin (16,17) and their binding levels to mouse skin DNA. In particular, we failed to detect DNA binding of the noncarcinogens anthracene, pyrene, and perylene (at a sensitivity of detection of one adduct in about 109 nucleotides), while the strong carcinogens BP, DMBA, and MC, exhibited the highest levels of DNA binding.
The illustrations provided in this article and elsewhere (4)(5)(6) demonstrate that each compound gives rise to a characteristic fingerprint of 32P-labeled adduct derivatives on PEI-cellulose maps. Therefore, it is possible on the basis of such fingerprints, to identify the carcinogen to which the particular DNA had been exposed in vivo.
The adduct intensification version of the 32P-postlabeling method was recently applied to detect adducts in human DNA from oral mucosa (B. Dunn, H. F. Stich, H. P. Agrawal, E. Randerath and K. Randerath, unpublished experiments) and human placenta (E. Randerath, R. Everson, R. Santella and K. Randerath, unpublished experiments) of smokers. This work showed that the method was applicable to the detection of adducts in DNA from humans and may thus become a tool for the detection of genetic damage in cells and tissues from humans exposed to carcinogenic/mutagenic chemicals.

Discussion
In this article a recently developed 32P-postlabeling method for the analysis of carcinogen-nucleic acid adducts has been reviewed. While we have focused mainly on the analysis of carcinogen-DNA adducts, the method is also applicable to carcinogen-RNA adducts. The salient features of the new 32P-postlabeling test for covalent DNA binding of chemicals can be summarized as follows.
The method enables the detection of minute amounts of adducts formed by the reaction of DNA with nonradioactive chemicals. At least in principle, therefore, DNA binding of any chemical can be assayed.
Since radioactive test chemicals are not required for in vivo studies, the assay is less expensive than assays employing radioactive chemicals.
ducts does not have to be known for DNA binding to be detected. If the fingerprint patterns obtained with pure chemicals or mixtures are known, then such fingerprints enable the identification of the binding chemical or mixture to which the DNA had been originally exposed. Small (microgram) amounts of DNA are required for analysis.
The method is highly sensitive, in the case of aromatic adducts enabling the detection of a few adducts per mammalian genome in a cell population or a tissue.
It can be used for accurate quantitation. It is potentially useful for investigating the repair, removal, and loss of adducts from cell or tissue DNA; the effects of anticarcinogens, chemopreventive agents, and/or metabolic inhibitors on adduct formation and persistence can also be studied.
The method appears applicable to DNA from humans exposed to gene-altering chemicals and thus constitutes a tool for monitoring human exposure to such compounds.
The results reviewed in this article and by others (18) suggest that chemicals that are capable of forming covalent bonds with DNA in mammalian tissues are likely to be carcinogenic. In combination with chronic bioassays for carcinogenicity, the 32P-postlabeling assay may thus be useful for answering the question as to whether covalent bond formation between a chemical and DNA in vivo, by itself, indicates carcinogenic potential of the particular chemical. More test chemicals need to be investigated to give a definite answer to this important question.
Included in our analyses were several compounds that, though being structurally related to known carcinogens, have not been found to date to be carcinogenic in chronic animal bioassays, i.e., 4-acetylaminofluorene, anthracene, pyrene, perylene, myristicin, dill apiol, parsley apiol, elemicin, anethole, and allylbenzene. With the exception of the polycyclic aromatic hydrocarbons, anthracene, pyrene, and perylene, DNA binding of these compounds in vivo was detected by 32P-postlabeling assay, but was found to occur to a much lesser extent than that of the structurally related carcinogens. For example, the binding to rat liver DNA of 4-acetylaminofluorene was below that of 2-acetylaminofluorene by a factor of about 400 fold, and the noncarcinogenic alkenylbenzenes [myristicin, dill apiol, parsley apiol, elemicin, anethole, and allylbenzene (13)] bound to mouse liver DNA 4 to 200 times less than their carcinogenic counterparts (safrole, estragole, and methyleugenol) (14) These results demonstrate that the 32P-postlabeling test is capable of detecting weakly genotoxic compounds, which either are not carcinogenic or whose carcinogenicity is not readily shown in animal bioassays. A compound such as myristicin, which binds only ca. 4 fold less than the structurally related hepatocarcinogen safrole to mouse liver DNA (14), but has not been shown to induce cancer in mouse bioassays (13), was found to exhibit substantial genotoxic activity in mouse liver in vivo. This raises the question as to whether the compound actually lacks any carcinogenic activity in spite of its DNA-damaging potential or whether bioassays conducted under different conditions would show its carcinogenicity.
Ashby (19) has recommended that a substantial proportion of the resources currently earmarked for chronic carcinogenicity bioassays might rather be employed in short-term in vivo evaluations in rodents of chemicals known to be genotoxic from various in vitro tests. On the basis of our results, the 32P-postlabeling assay appears suitable to serve as such a short-term test for the detection of carcinogen-DNA adducts in animal tissues, i.e., as an assay for potential carcinogenicity ofchemicals.
We have also pointed out in this article that the 32Ppostlabeling technique represents a tool to assay an important property of chemical carcinogens, i.e., the formation of persistent DNA adducts, which may play a crucial role in carcinogenesis. In using the test for the detection of covalent binding of chemicals to DNA, one should therefore also include an evaluation of adduct persistence. Compounds such as 7,12-dimethylbenz(a)anthracene or sterigmatocystin, which give rise in animals to highly persistent, essentially nonreparable adducts, induce irreversible toxic effects in the genetic material of mammals in vivo; it would appear prudent to preclude or minimize human exposure to such chemicals by appropriate regulatory measures. Since the 32Ppostlabeling assay makes possible the detection and quantitation of covalent binding of chemicals to DNA, as well as an analysis of the persistence of DNA lesions, it may become an important tool for risk assessment of genotoxic chemicals.
Our work on the development and applications of the 32P-postlabeling method for adduct analysis was supported by USPHS grants CA 25590, CA 30606, CA 32157, and CA 10893 (P6), awarded by the National Cancer Institute. We wish to thank a great number of colleagues and friends-too numerous to be named here-who have greatly contributed to this work by encouraging us during the development of the method and/or by providing samples used in the standardization of experimental conditions.