Detection of adducts of deoxyribonucleic acid in white blood cells of roofers by 32P-postlabeling. Relationship of adduct levels to measures of exposure to polycyclic aromatic hydrocarbons.

Detection of adducts of deoxyribonucleic acid in white blood cells of roofers by 32P-postlabeling: relationship of adduct levels to measures of ex posure to polycyclic aromatic hydrocarbons. Scand J Work Environ Health 1990:16;135-43. To assess the utility of adducts of deoxyribonucleic acid (DNA) as biomarkers of exposure to carcinogens in an industrial population, a pilot study of roofers occupationally exposed to polycyclic aromatic hydrocar bons was conducted. DNA isolated from white blood cellsof roofersand nonoccupationally exposedcom parison subjects matched for age, sex, and smoking status was analyzed for DNA adducts with the use of 32P-postlabeling methods. Occupational exposures to polycyclic aromatic hydrocarbons were assessed by personal air sampling and skin wipes. Ten of the 12 roofers, but only 2 of the 12 comparison subjects, had detectable levels of aromatic DNA adducts in the J2P-postlabeling assay. Among the roofers, the post shift levels of polycyclic aromatic hydrocarbons in the skin wipes were correlated with the DNA adduct levels. These results suggest that 32P-postlabeling assay may be useful for monitoring internal exposures to complex mixtures of aromatic hydrocarbons in industrial adduct levels correlated with both the postshift skin levels of the total PAH (Spearman rank correl a tion coefficient 0.81, P <0.05) and with the postshift skin levels of benzo[a]pyrene (Spearman rank corre lation coefficient 0.74, P < 0.05). The preshift skin levels of PAH and benzo[a]pyrene were modestly cor related with the median adduct levels; these results were not, however, statistically significant. The adduct levels of the roofers were not correlated with the duration of employment as a roofer or with the dietary con sumption of PAH.

conducted by Hammond et al (2), wh ich found significantly elevated mortality rates for cancer of the lung , oropharynx and esophagus, stomach, and bladder, and for nonmelanoma skin cancer.
More recently , studies of PAl-l-exposed populations, including roofing workers, ha ve been conducted utili zing a variety of biological markers of expo sure to mutagenic and carcinogenic substances (3)(4)(5). Th ese markers, which include assa ys fo r carcinogeni c DNA adducts, hold promise for inco rporatio n into human epidemiologic studies, for use in exposure and risk assessment, and for the biological monitoring of groups exposed to chemical carcinogens. Measurement of carcinogenic DNA adduct formation is particularly promising. Because the fo rma tio n of chemical addition products with DNA appears to be a critical event in carcinogenesis, measurement of carcinogenic DNA adducts should provide biologically rele vant information which reflects an integration of human exposur e, ab sorption , metabolism , and DNA-adduct formation and rep air rates (6,7) . Once validated, these assays ma y allow a more timel y detection of human carcinogens in the environment a nd workplace than is currently po ssible.
32p-postlabeling is a sensitive method for the detection of hydrophobic carcinogenic DNA adducts (8)(9)(10). It has been used to study adduct formation and repair in a large number of animal studies and to monitor human exposure to environmental carcinogens (II , 12). This assay has the advantage of not requiring advance knowledge of the adduct of interest and can thus be applied to a wide range of exposure situations.
This pilot study was performed to compare the levels of carcinogenic DNA adducts among industrial workers with well-characterized exposures to PAH to levels in an appropriately matched nonoccupationally exposed group.

Population
The study group consisted of 12 roofing workers and 12 comparison subjects who were matched for sex, age (within 11 years), and smoking status and were without occupational exposure to PAH. The potential subjects were excluded if they had cancer, were known to be carriers of human immunodeficiency virus (HIV), or they reported exposure to any of the following potentially mutagenic agents (because sister chromatid exchange frequencies and urinary mutagens were being evaluated in related studies): ethylene oxide, styrene, benzene, chemotherapeutic agents, or vinyl chloride. A roofing site at which there was exposure to PAH was identified with the assistance of Local 8 of the United Union of Roofers, Waterproofers, and Allied Workers. A field survey was conducted at this site in June 1987. Of 14 roofers at the site at the beginning of the survey, one was transferred to another worksite before completion of the survey and therefore was ineligible for participation. Of the remaining 13 roofers, 12 (93 0/0) agreed to participate. The matched comparison subjects were either employees of the Mount Sinai Medical Center or patients attending the Mount Sinai Occupational Health Clinical Center. Informed consent was obtained from all the study participants.

Field survey
The work process at the roofing site involved the removal or "tear off" of sequential sections of an old pitch roof followed by the replacement of each section with a new asphalt roof. The levels of occupational exposure to PAH among the individual roofers were assessed from air samples from the breathing zone of 10 individuals for two workdays while pitch tearoff and asphalt reapplication was being performed (day I, Thursday, and day 2, the following Monday). In addition, pre-and postshift skin wipe samples for PAH were obtained from eight roofers on day 2 (Monday) . These samples were obtained from a 9 em' area of the forehead with a Whatman smear tab . Details of the industrial hygiene and skin wipe surveys have been published elsewhere (13).
On day 3 of the field survey (Tuesday), venous blood (35 ml) was collected in coded heparinized containers and transported on ice to Columbia University. All the study participants completed a self-administered ques-136 tionnaire which was reviewed by a physician specialized in occupational medicine. The questionnaire obtained detailed information about occupational history (past and present); history of occupational, environmental, and dietary exposure to PAH, as well as to other carcinogens and mutagens; medical history; and history of prior and current use of cigarettes, pipes, and cigars.

Industrial hygiene sample analysis
The air and skin wipe samples were analyzed for individual PAH following method 5506 of the National Institute for Occupational Safety and Health (NIOSH). The individual PAH determined with highperformance liquid chromatography with fluorescence detection were anthracene, f1uoranthene, pyrene, benz[a]anthracene, benzo[b)f1uoranthene, benzojk]-f1uoranthene, benzo[a)pyrene, and benzo[ghi]perylene. No anthracene was found in the skin wipe samples. Total PAH, as reported for both the air and skin wipe samples, was the sum of the peaks listed. The personal air samples encompassed essentially the entire workday (excluding breaks), and the results have been reported as time-weighted averages for the period of sample collection.

32P-postlabeling assay
White blood cells, plasma, and red blood cells were separated by centrifugation, and the samples were frozen at -70°C until used. DNA was isolated by standard phenol/chloroform and ribonuclease treatment procedures (14). The samples were analyzed essentially as described by Reddy & Randerath (15), except that lower levels ( Heights, Illinois, United States) was added followed by 3 IJ.I (9 units) of cloned T4 polynucleotide kinase (US Biochemicals, Cleveland , Ohio, United States), and the mixture was incubated at 37°C for 30 min. The resolution of the adducts was carried out on laboratory-prepared (16) polyethyleneimine-cellulose sheets essentially as has already been described (15). Direction one (D-1) was run overnight onto a wick in I M sodium phosphate, pH 6.8. After each step the plates were washed twice in water. Direction two was developed in the same direction as 0-1 in 2.5 M ammonium formate, pH 3.5 . Direction three, run in the opposite direction of D-I, was developed in 3 M lithium formate, 8.5 M urea, pH 3.5. Direction four was run after the plate was turned 90°in 0.6 M lithium chloride, 0.5 M Tris , 8.5 M urea, pH 8.0. A final chromatography was carried out in 1.7 M sodium phosphate, pH 6.0, to decrease the background levels of the radioactivity. The adduct spots were detected by autoradiography with intensifying screens and quantitated by scintillation counting. For positive samples with multiple adducts, the modification levels have been expressed as the sum of all adducts present. In all the assays a positive control, DNA modified by benzo[a]pyrene diol epoxide (BPDE-I-DNA) (17) and diluted with unmodified DNA from calf thymus to I adduct/LO? or 10 8 nucleotides, was assayed . Unmodified DNA from calf thymus was routinely assayed as a negative control. For the mixing experiments, I ug of human DNA was mixed with 1 ug of benzo[a]p yrene 713 , 8a-dihydroxy-9a, lOa-epoxy 7,8,9,10 tetrah ydrobenzo[a]pyrene (BPDE-I-DNA) at a modification level of I adduct/TO" nucleotides. All the samples were assayed at least twice, and many samples with detectable adduct levels were assayed more than twice.

Statistical analysis
For each subject, several replications were made of the 32p-postla beling assay (from 2 to 6). The assay results were not normally distributed, and therefore nonparametric test statistics were used. For the calculation of the median assay scores, undetectable levels were ascribed a value (I x 10-9 ) , which was the midpoint between zero and the limit of detection (2 x 10-9 ) . Because of the small number of study subjects, an attempt was made to limit the number of statistical analyses performed. A matched pair analysis comparing the median levels of aromatic DNA adducts between the roofers and nonroofers was performed with Wilcoxon' s signed rank test. To evaluate a possible association between cigarette smoking and adduct formation, we tested the results of the roofers and nonroofers separately. Among the roofers the adduct levels of the smokers were compared to the corr esponding levels of the non smokers with the rank sum test for a comparison of independent samples. Among the 12 nonroofers the median adduct levels were the same for 10 individuals, and therefore no statistical tests were performed . The associations between measures of exposur e to PAH and aromatic adduct levels were tested with Spearman's rank correlation coefficient. For the roofers, correlations were sought between the median adduct levels and (i) the duration of employment as a roofer, (ii) the levels of air and skin exposure to total PAH and benzo[a)pyrene, and (iii) dietary PAH consumption. Dietary PAH exposure was estimated from the addition of the number of servings of oven-broiled, flame-broiled, charcoal-broiled , or smoked food s con-sumed in the month prior to the blood drawing, as reported on the questionnaire.

Demographic characteristics and nono ccupational exposure information
All the roofers and nonroofers were men. The median age of the roofers was 33.5 (range 19-45) years, and the median age of the nonroofers was also 33.5 (range 21-44) years. Eight(75 Il?o) of the 12 roofers and eight of the nonroofers were current cigarette smokers. Among the current smokers, the median number of cigarettes smoked dail y during the month prior to the blood collection was 25 (range 10-40) cigarettes /d for the roofers and 17.5 (range 10-20) cigarettes/d for the nonroofers. The median number of pack-years was 17.5 (ran ge 6-30) for the currently smoking roofers and 8 (range 4-26) for the currently smoking nonroofers. All the current smokers inhaled and used filter cigarettes. One of the cigarette-smoking roofers also reported occasional pipe smoking; no other roofers or nonroofers reported smoking pipes or cigars. The median number of charcoal-broiled , oven-broiled, flame-broiled, and smoked food servings in the past month was eight (range 0-34) for the roofers and three (range 0-26) for the nonroofers. The roofers consumed a median of 1.5 (range 0-12) servings of alcoholic beverages per day, whereas the respective median of the nonroofers was 0 (range 0-2) per day .

Historical occupational exposure information
None of the nonroofers reported occupational exposure to PAtIo Three (25 Il?o) of the 12 roofers had been in the indu stry five years or less, five (42 %) had worked as roofers for 6-10 years , two (16.5 %) had worked as roofers for 11-15 years, and two (16.5 %) had been employed in roofing for 15-20 years. The median hours worked per week was 35 (range 25-42). The roofers were asked how man y consecutive weeks they had worked with hot pitch or asphalt or performed pitch tear-off in the month prior to the blood drawing. Ten (83.3 %) of the 12 reported four weeks of such work, and two (16.7 (1 10) reported three weeks.

Field survey (industrial hygiene)
The results of the personal air sampling performed Thursday (day I) and Monday (day 2) and of the preand postshift skin wipe samples collected on Monday (day 2) are presented in table I. On Monday (the day before the blood samples were collected) the median level of benzo[a]pyrene in the personal air samples was 0.79 (ran ge 0.60-1.39) ug/rrr 'cand the median total PAH in the personal air samples was 8.32 (range 6.0-13.77) ug/rri'. Neither the total PAH nor the benzo[a]pyrene levels in the personal air samples were correlated with those of the same individual on the fol-   BPDE-I-DNA before the digestion and labeling. The adducts in the human samples did not cochromatograph with the major BPDE-I guanine N2 adduct in either sample. Data for sample number 17 are shown in figure Id . A comparison of the median adduct levels of the matched pairs of roofers and nonroofers showed that the roofers had significantly higher adduct levels than the nonoccupationaIly exposed subjects (P < 0.01, twotailed). Among the roofers, no association was found between the adduct levels and current cigarette smoking (P > 0.10). Among the nonroofers, the two individuals with detectable adduct levels were both smokers; the remaining six smo king nonroofers all had undetectable adduct levels.
The correlations between th e measures of expo sure to PAH and the median aromatic adduct levels were evaluated for the roofers with Spearman's rank correlation coefficient. The results of this analy sis are summarized in table 3. The adduct levels of the roofers were not correlated with the levels of benzo[a]pyrene nor the total PAH in the per sonal air samples from either Thursda y or Monday. Among the eight roofers for whom the skin wipe data were available, the median adduct levels corr elated with both the postshift skin levels of the total PAH (Spearman rank correl ation coefficient 0.81, P <0.05) and with th e postshift skin levels of benzo[a]pyrene (Spearman rank correlation coefficient 0.74, P < 0.05). The preshift skin levels of PAH and benzo[a]pyrene were mode stly correlated with the median adduct levels; these results were not, however, statistically significant. The adduct levels of the roofers were not correlated with the duration of employment as a roofer or with the dietary consumption of PAH.

Discussion
In this pilot study, we found that aromatic DNA adducts were detect able in the white blood cells of 83 0/0 of a sample of roofers with occupational exposure to PAH but were dete cted , at very low levels, in only 17 % of matched, nonoccupationally exposed subjects. The DNA adduct levels in th e white blood cells of the roofers were sign ificantly correlated with the levels of total PAH and benzo[a)pyrene in the postshift skin wipe samples. Neither the total PAH nor the benzo[a]pyrene levels of the per sonal air samples correlated with the levels of the aro matic DNA adducts. DNA adducts hold great promise for use as biom arkers of human exposure to occupational and environmental carcinogens . Pre vious studies of populations exposed to PAH in indu st ry (such as roofers, coke oven workers, and foundry workers) utilizin g immunoassay techniques with antibodies to BPDE-I-DNA have demonstrated the presence of white blood cell adducts (3)(4)(5) . Adducts have also been demonstrated in coke oven workers and foundry workers with fluore scence methods (3,5).
In this study, we utilized 32P-postla beling methods, which can detect effects o f exposure to complex mixtures of bulky aromatic compounds such as PAH and thus are ideal for monitoring workers who are exposed to compl ex mixtures of PAH. Aromatic DNA adducts have been detected by 32P -postlabeling methods in a number of hum an tissues, including human pla centa, buccal muco sa, bone marrow , lung, and peripheral white blood cells (18)(19)(20)(21)(22). More recentl y, foundry workers, who sustain occupational expo sure to a mixture of PAH, have been studied with 32p -postlabeling methods (23). With the use of historical (nonconcurrent) industrial hygiene data and job title, foundr y workers were classified as belonging to high (benzo[a]p yrene ;::= 0.2 ug/rn ' ), medium (0.05 ug/rn' < benzo[a]pyrene < 0.2 ug/rn'), or low (benzo[a)pyrene < 0.05 ug/rn ' ) exposure groups. Aromati c DNA adduct s were found to be pre sent in the white blood cell DNA in three out of four samples from highly exposed workers, in eight out of 10 samples from the medium exposure group, in four out of 18 sa mples from the low exposure group, and in one out of nine samples from the nonoccupationaIly exposed subjects. The adduct levels among the foundry worke rs ranged from undetectable to I adduct/JO? nucleotides. Our work corroborates the findings of this study and indi cates that 32P-postlabeling methods provide a useful bioassay of internal exposure to carcinogenic PAH in the workplace.
Th is study con trolled for age, sex, and cigar ette smoking, one of the major nonoccupational sources of expo sure to PAH. Environmental sour ces of exposure to PAH were carefully characterized, and both air and dermal levels of occupational exposure to PAH were measured.
The levels of total PAH and benzo[a)pyrene in the personal air samples did not correlate with the adduct levels. Nevertheless, the air sampling data prov ide useful information about expo sur e ran ges at which addu ction occurs and, hence, provid e information which may be useful in human risk assessment. The roo fers in this study were expo sed to levels of benzo[a)pyrene which would be classified as high in the scheme utilized in the foundry worker study (23) and which would be classified as fairly high in the scheme proposed by Linstedt & Sollenberg in 1982 (24). The data from the post shift skin wipe sampl es for PAH and benzo[a]pyrene were significantly correl ated with the adduct levels of the roofers (figure 2). The finding of a correlation between th e adduct levels and the skin levels of PAH and benzo[a)p yrene but not between th e adduct levels and the air exposure dat a may be an aberrat ion caused by the small numbers since expo sure data were missing for some individual s. Alternatively, white blood cell aromatic DNA adduct levels ma y pro ve particularly useful as indi cators of dermal expo sur e to PAH. PAH are absorbed percutaneously, and skin has been demonstrated to have significant arylhydrocarbon hydroxylase activity. Pohl et al (25) (26), as well as in mouse skin treated with coal tar or asphalt (27). Work by Schok et et al (28) has also demonstrated th at the treatment of adult and fetal skin with coal tar or asphalt in sho rtterm cultur e results in the form at io n of DNA adducts which are detectable by 32P-postla beling. Howe ver, although suggestive, the skin wipe technique is relatively new, and it is not known how it relates to actu al dermal or respiratory ab sorption. Therefore , proper interpretation of the use of skin wipes as a measure of exposure requires furth er stud y and validation.
The pre shift skin wipe levels of PAH and benzola]pyrene were modestly, but not significantly, correlated with the median of the adduct levels (table 3). These preshift wipes, obtained on a Monday morning, presumably reflect levels of PAH reta ined in the skin over the weekend despit e tempo rary cessation of occupation al exposure .
We did not find that cigarett e smo king was associated with elevated adduct levels in peripheral white blood cells in the 32p-postlab eling assay. Other studies investigating the relationship between adduct levels in peripheral white blood cells and cigarette smoking have yielded similar results (21,23). However, a relationship between adducts and cigarette smoking has been found in other tissue types, such as placenta and lung (l8 , 19,22). Although th e roofers in our study tended to smoke more heavily th an their mat ched counterparts, it is unlikel y that this differential distribution affected th e findin gs of thi s pilot study since cigarette smoking does not appear to be associated with increased adduct levels. We also found no effect of dietar y PAH consumption on the median adduct levels among the roofers and therefore diffe rent levels of dietary PAH consumption is unlikel y to have resulted in confounding.
The identity of the adducts present in this study remains unknown, although they are likely to be aromatic DNA adducts. In th e two samples evaluated in a mixing experiment, the adducts did not appear to be the major BPD E-i-guanine N2 adduct; ho wever these results cannot be generalized to the samples from the othe r study participants. While 32P-postlabeling was able to discriminate between th e exposed a nd unexposed subjects of our study, some limitations of the assay must be noted. The quantitative data pre sented in table 2 have been based on the assumption that the DNA was completely digested. While this assumption is true for a number of types of carcinogen-modified DNA, some unknown adducts may be resistant to nuclease digestion and there fore not detected by the assay. In add ition, other adducts originally pre sent ma y not be resistant to nuclease PI and thus would not be labeled. In the procedure used in this study, neither are alkylated bases that chromatograph with the normal nucleotides detected. Therefore, the adduct levels given in table 2 should be considered the minimum amount present.
Nonetheless, although this study is limited by its small sample size, it suggests the utility of incorporating 32p-postlabeling assay into studies of industrial populations as a marker of internal exposure to P AH and as an adjunct to traditional methods of exposure assessment. Incorporation of this biomarker into traditional epidemiologic studies holds promise both for improving the ability to discern the degree of risk posed by occupational and environmental exposures to carcinogens and putative carcinogens and for elucidating disease mechanisms.