Document aJJRnkBrmN2wEaNMa7o8mamZa

Journal of Chromatography B, 778 (2002) 367374 www.elsevier.com / locate / chromb Non-linear production of benzene oxidealbumin adducts with human exposure to benzene Stephen M. Rappaporta ,*, Karen Yeowell-O'Connella, Martyn T. Smithc, Mustafa Dosemecib, Richard B. Hayesb, Luoping Zhangc, Guilan Lid, Songnian Yind, Nathaniel Rothmanb aDepartment of Environmental Sciences and Engineering, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7400, USA bDivision of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD 20892, USA cSchool of Public Health, University of California, Berkeley, CA 94720, USA dChinese Academy of Preventive Medicine, Institute of Occupational Medicine, Beijing, China Abstract Benzene is initially metabolized to benzene oxide, which either undergoes further metabolism or reacts with macromolecules including proteins. Previously reported levels of benzene oxidealbumin adducts (BOAlb) are analyzed from 30 workers exposed to 0.2302 ppm benzene and 43 controls from Shanghai, China. Although both exposed workers and controls had significant levels of BOAlb in their blood, exposed subjects' adduct levels (GM5378 pmol / g protein) were much greater than those of controls (GM5115 pmol / g protein). When the natural logarithm of the BOAlb level was regressed upon the natural logarithm of exposure among the 30 exposed subjects, a strong effect of benzene exposure was observed (R250.612; p,0.0001). Because the slope of the relationship between BOAlb and benzene exposure was significantly less than one in log-space, we infer that production of benzene oxide was less than proportional to benzene exposure. Since benzene is a substrate for CYP2E1, these results are consistent with saturation of CYP450 metabolism. They indicate that deviations from linear metabolism began at or below benzene exposures of 10 ppm and that pronounced saturation was apparent at 4050 ppm. To our knowledge, this is the first study to investigate the linearity of human metabolism of a carcinogen based upon protein adducts. 2002 Elsevier Science B.V. All rights reserved. Keywords: Benzene oxide; Albumin; Benzene 1. Introduction Benzene is a multi-site carcinogen in rodents that causes hematotoxicity and acute myeloid leukemia in humans [1]. Although the mechanism by which *Corresponding author. CB 7400, School of Public Health, University of North Carolina, Chapel Hill, NC 27599-7400, USA. Tel.: 11-919-9665-017; fax: 11-919-9664-711. E-mail address: stephen]rappaport@unc.edu (S.M. Rappaport). benzene causes these effects is unknown, evidence strongly indicates that metabolism is required [2,3]. As summarized in Fig. 1, benzene is metabolized by CYP450 to benzene oxide (BO) [4,5], which exists in equilibrium with its valence tautomer oxepin. Subsequent enzymatic and non-enzymatic reactions of BOoxepin give rise to other metabolites, notably, phenol, catechol, hydroquinone and the ring-opened muconaldehydes (ultimately transformed to trans, trans-muconic acid). Catechol and hydroquinone can 1570-0232 / 02 / $ see front matter 2002 Elsevier Science B.V. All rights reserved. PII: S0378-4347(01)00457-1 368 S.M. Rappaport et al. / J. Chromatogr. B 778 (2002) 367374 Fig. 1. Simplified scheme for metabolism of benzene in humans. be oxidized to 1,2- and 1,4-benzoquinone, respectively, via the corresponding semiquinones. The uncertain mechanism of benzene's toxicity arises from the possible roles of the many electrophilic metabolites (including BO, the muconaldehydes and the benzoquinones) as well as reactive oxygen species produced by redox cycling of catechol, hydroquinone and the benzoquinones [3]. Thus, even though there appears to be a causal link between human exposure to high levels of benzene (above 10 ppm) and leukemia, the prediction of human risks at lower levels of exposure is fraught with uncertainty. Since airborne benzene is ubiquitous at low levels in the environment, typically at concentrations less than 10 ppb [6], and can reach thousand-fold higher levels in the workplace [7], the shape of the exposureresponse relationship below 10 ppm represents a major quandary for environmental epidemiology. Because the uncertainty surrounding benzene's exposureresponse curve is unlikely to be resolved by conventional epidemiological studies, recent efforts have turned to biomarkers to shed light upon exposurerate effects involving uptake, biotransfor- S.M. Rappaport et al. / J. Chromatogr. B 778 (2002) 367374 369 mation and cell damage in humans exposed to benzene. Here the goal has been to elucidate relationships between benzene exposure and the corresponding levels of urinary metabolites, protein adducts, cytogenetic changes and blood abnormalities (e.g., see Refs. [811]). Since the toxic effects of benzene are so clearly linked to metabolism, one promising avenue of inquiry involves the relationship between benzene exposure and levels of BO, from which all other metabolites arise (Fig. 1). However, direct measurement of BO in humans is impractical due to the reactivity of this compound in blood and by the limited sensitivity of assays [4]. Thus, recent efforts have employed adducts of BO with hemoglobin (Hb) and serum albumin (Alb) among benzene-exposed workers to make inferences about levels of BO in blood. (These adducts will henceforth be designated as BOHb and BOAlb, respectively). As shown in Fig. 2, cysteinyl adducts are presumed to be formed from reactions of BO with free cysteinyl residues in proteins to give S-(2-hydroxycylohexa-3,5-dien-1-yl) cysteine derivatives, which may persist or spontaneously dehydrate to S-phenylcysteine. Since BO Alb and BOHb are stable in vivo [12], they accumulate over periods of 1 month (the mean residence time of human serum albumin, corresponding to a half life of 21 days) or 4 months (the life span of human erythrocytes), respectively. Hence, levels of these protein adducts reflect average concentrations of BO produced by exposure to benzene over 1 to 4 months [13]. Exposure-related increases in both BOHb and BOAlb have been reported among workers exposed to benzene [11,14,15]. Because Alb adducts tend to be much more abundant than Hb adducts at a given exposure to benzene, BOAlb is preferable for human biomonitoring [11,14]. We previously reported relationships between benzene exposures and levels of Fig. 2. Reaction of benzene oxide with a free cysteinyl residue of a protein (R-SH) to produce a protein adduct. BOAlb among 86 workers (43 currently exposed to benzene and 43 controls) in Shanghai China, in factories where benzene was used as a solvent [11,14]. Due to heteroscedasticity, we used weightedleast-squares regression to predict linear relationships between subject-specific benzene exposures (1.65328 ppm) and the corresponding levels of BOAlb. It was also necessary to adjust regression coefficients for respirator use by some workers. These factors, coupled with a high background level of adducts among controls (arising primarily from non-benzene sources), obscured the relationship between BOAlb and benzene exposure. In the current study, we employ alternative statistical methods to investigate the linearity of the relationship between BOAlb and benzene exposure among workers who did not wear respirators (n530). The results indicate that the rate of production of BOAlb begins to diminish at or below 10 ppm of benzene and approaches saturation at 4050 ppm. 2. Experimental The data used in this study were obtained from a previous investigation of protein adducts among 86 workers [11,14]. The methods are briefly summarized as follows. 2.1. Subjects Forty-four benzene-exposed workers were recruited from three factories in Shanghai, China where benzene was used to solubilize natural rubber (factory 1), to manufacture adhesive tape (factory 2) and to paint wooden toys and boxes (factory 3); 44 control workers, who had been frequency matched with exposed subjects by gender and age, were recruited from a sewing machine manufacturing plant (factory 4) and an administrative facility (factory 5) in the same geographic region [8]. Since blood was not available for one of the exposed workers, complete sets of data were available for 87 of these frequency-matched subjects (43 exposed and 44 controls). Of these, 16 exposed subjects in factory 1 wore respirators and were excluded from analyses in the current study. Two additional workers were excluded from the present study based upon regres- 370 S.M. Rappaport et al. / J. Chromatogr. B 778 (2002) 367374 sion diagnostics, which revealed great influence upon coefficients and model fit. hOne was a control worker for which laboratory notes indicated poor chromatographic resolution during gas chromatographymass spectrometry (GCMS) analysis and the other was a worker with very high exposure (328 ppm) who had unrealistically low levels of adducts as well as urinary benzene metabolites [11,14]j. In addition to these frequency-matched subjects, four additional workers with low exposures to benzene were included in the present study. Thus, data from 73 subjects were analyzed in the current investigation (30 exposed workers and 43 controls). 2.2. Exposure measurements Passive monitors had been used to measure personal benzene exposures of each exposed subject during the full work shift on 5 consecutive days (two workers had six measurements) just prior to blood collection [8]. The geometric mean of these air measurements was used to estimate each subject's exposure to benzene; for the 30 exposed subjects the overall (geometric mean) exposure to benzene was 11.0 ppm benzene. Control subjects from factory 4 had been monitored for benzene exposure on a single day (n530); only three of these measurements (0.047, 0.052 and 0.110 ppm) were above the limit of detection of 0.016 ppm. Control subjects in factory 5 had not been monitored and were assumed to have no exposure to benzene (n513). 2.3. Adduct measurements and methanesulfonic acid to produce phenyltrifluorothioacetate (PTTA), which was measured by GCMS in the negative chemical ionization (NCI) mode [16,17]. [The initial adduct shown in Fig. 2, an S-(2-hydroxycylohexa-3,5-dien-1-yl) cysteine derivative, dehydrated to form an S-phenylcysteine residue either spontaneously or under the acidic conditions of the assay]. Following isolation, 45 mg Alb were combined with 310 pmol [2H5]Sphenylcysteine (internal standard) and 200 ml of 100 mM ascorbic acid (to reduce adducts of the benzoquinones which were measured simultaneously as the corresponding hydroquinones). An 800-ml volume of trifluoroacetic anhydride and 20 ml of methanesulfonic acid were then added and the reaction mixture was heated to 1008C for 40 min. The remaining trifluoroacetic anhydride was removed under a stream of nitrogen. Following extraction with 1 ml hexane, the hexane layer was washed with 1 ml of 0.1 M Tris buffer (pH 7.5) and 231 ml of deionized water. After concentrating the hexane layer to |200 ml, 23 ml of each sample were analyzed by GC MS in the NCI mode using selected ion monitoring of PTAA (m /z 206) and the derivatized internal standard [2H5]PTTA (m /z 211). A representative chromatogram is shown in Fig. 4 for Alb obtained from a worker exposed to 30 ppm benzene. The relative standard deviation of this assay had been estimated to be 36% based upon duplicate analyses of 48 specimens of Alb from these subjects [14]. Average levels of BOAlb were assigned to those subjects with duplicate analyses. BOAlb had been measured in a single sample of venous blood obtained from each subject, as described previously [11,14]. The assay, illustrated in Fig. 3, is based upon reaction of an S-phenylcysteine residue (in serum Alb) with trifluoroacetic anhydride Fig. 3. Reaction of BOAlb with trifluoroacetic anhydride and methanesulfonic acid to produce phenyltrifluorothioacetate (PTTA). 2.4. Data analysis Linear regression was performed using Proc REG of SAS system software (SAS Institute, Cary, NC, USA). To eliminate heteroscedasticity and to satisfy normality assumptions, regression employed (natural) logarithmic transformation of levels of BOAlb and benzene exposure. Goodness of fit and influence were evaluated with standard residual diagnostics and normality plots. Regression analysis was performed with data only from exposed subjects (n5 30). The regression was repeated after subtracting the GM level of 115 pmol / g Alb observed in controls (n526 subjects with adjusted levels greater S.M. Rappaport et al. / J. Chromatogr. B 778 (2002) 367374 371 Fig. 4. GCMS (negative-ion chemical ionization) selected-ion monitoring chromatogram of phenyltrifluorothioacetate (PTTA) obtained by reaction of 4 mg of albumin, from a worker exposed to 30 ppm benzene, with trifluoroacetic anhydride and methanesulfonic acid. [2H5]S-Phenylcysteine (3 pmol) was added as the internal standard. than zero). Controls were not included in the regression analysis because incomplete exposure data were available from unexposed subjects; i.e., while each exposed subject had five full-shift measurements some controls had only one measurement and others had no measurements. Within- and between-subject components of variance of benzene exposure were determined via one-way analysis of variance (ANOVA) of the logged exposure levels using Proc NESTED of SAS. 3. Results Summary statistics, showing median values and ranges of benzene exposures and BOAlb levels, are presented in Table 1. The level of adducts was much greater for exposed subjects [geometric mean (GM)5378 pmol / g protein, range: 52.72240 pmol / g Alb] than for controls (GM5115 pmol / g protein, range: 44.7248 pmol / g Alb) ( p,0.0001, Wilcoxon rank sum test). Levels of BOAlb are depicted with individual exposures in Fig. 5A as a loglog plot. The figure shows that logarithmic transformation normalized the variance of adduct levels and that the fit of the straight-line model was reasonable. Simple linear regression revealed a strong effect of benzene exposure according to the following relationship: BOAlb sln[BOAlb, pmol / g protein]d 5 4.89 1 0.434*sln[benzene, ppm]d (R250.612; p,0.0001). After adjustment for background adducts the corresponding relationship was: Bo-Alb54.3810.497*(ln [benzene, ppm]) (R25 Table 1 Benzene exposures and adduct levels among exposed and control subjects Variable Controls (n 5 43) Exposed (n 5 30) Exposure GMa Exposure GSDa Exposure rangea ,0.0160.110 11.0 5.76 0.160302 BOAlb GMb BOAlb GSDb BOAlb rangeb 115 1.53 44.7248 378 2.58 52.72240 GM5Geometric mean; GSD5geometric standard deviation. a Benzene exposures (in ppm) based upon five daily measure- ments per exposed subject and 01 measurement per control subject; the GM and GSD shown for exposed subjects are based upon the mean values of individual workers logged exposures. b Benzene oxideAlb adducts in pmol / g protein based upon one blood specimen per subject. 372 S.M. Rappaport et al. / J. Chromatogr. B 778 (2002) 367374 0.359, p50.0012, n526). Since the estimated slope of 0.497 was less than one in log-space (95% C.I.: 0.217, 0.777), we infer that the rate of BOAlb production diminished at higher benzene exposures (forming a plateau at high levels). The same data plotted in Fig. 5B show this plateau-shape for the corresponding relationship between adducts and benzene exposure in natural space. The dashed line in Fig. 5B represents the straight-line relationship that would be expected if adduct levels were strictly proportional to benzene exposure; here a slope of 39.5 pmol BOAlb / g protein / ppm benzene was predicted. Fig. 6 presents the same relationship as Fig. 5B with the scale reduced to focus upon adduct production below 25 ppm of benzene; the GM adduct level among controls is designated by a triangle on the y-axis. Fig. 5. Scatterplots of levels of BOAlb versus benzene exposure for 30 subjects. (A) Log-space plot; the line represents the leastsquares regression. (B) Natural-space plot; the solid curve represents the adduct level predicted from (A) and the dashed line represents a linear relationship between the adduct level and benzene exposure. Fig. 6. Scatterplot of levels of BOAlb versus benzene exposure for 30 subjects. (Curves are from Fig. 5B with axes modified to focus upon benzene exposures below 25 ppm). The solid curve represents the adduct level predicted from Fig. 5A; the dashed line represents a linear relationship between adduct level and benzene exposure; the triangle represents the geometric mean adduct level measured in control subjects. 4. Discussion This analysis indicates that levels of BOAlb were less than proportional to benzene exposures over the range of 0.2302 ppm. Since BOAlb should reflect the average level of BO in the blood over a period of about 1 month prior to blood collection [13], this suggests that the rate of BO production diminished with increasing benzene exposure among these subjects. In fact, Fig. 6 indicates that the rate of adduct production began to decline noticeably in the range of 10 ppm benzene. Since benzene is a substrate for human and rodent CYP2E1 [1821], saturable metabolism would be expected at some level of exposure. However, given the paucity of human data, rates of benzene metabolism have been predicted from physiologically based pharmacokinetic models [2224], which pointed to linear kinetics below 10 ppm [23] and to saturation above 100 ppm [22]. Our results, indicating nearly linear production of adducts below 10 ppm (Fig. 6) and substantial saturation at 4050 ppm of benzene (Fig. 5B), are generally consistent with these pharmacokinetic models. The estimated regression coefficient (i.e., 0.477 (ln[pmol BOAlb / g protein / ppm benzene]) for background-adjusted adduct levels assumes that benzene exposure was known without error. Since this was certainly not the case, the observed regression S.M. Rappaport et al. / J. Chromatogr. B 778 (2002) 367374 373 coefficient was probably attenuated. However, assuming that errors in estimation of BOAlb were uncorrelated with those of benzene exposure (which is reasonable given a 1-month residence time of BOAlb and daily measurement of exposure), it is possible to adjust the regression coefficient based upon the variance components of benzene exposure within and between subjects [25]. From one-way ANOVA of the (logged) exposure data, the estimated ratio of within- to between-person variance components was 0.332 for the 30 subjects in our sample (with five measurements each). Based upon Ref. [25], we predict an expected value for the ``true'' regression coefficient of 0.477?(11[0.332 / 5])5 0.509?(ln[pmol BOAlb / g protein / ppm benzene]), an attenuation of only 6.7%. Thus, if the individual air measurements were representative of exposures over the month preceding blood collection (the residence time of BOAlb), the straight-line relationship shown in Fig. 5A should be reasonable. By extension, the non-linear relationships shown in Figs. 5B and 6 should be relatively unbiased over the observed ranges of benzene exposure. Production of total urinary metabolites was also observed to be non-linear with benzene exposure among these same workers [9,26]. However, in order to estimate total urinary metabolites it was necessary to add the levels of all major urinary products of benzene, each of which had been measured independently with error. Thus, the variability of total urinary metabolites was probably greater than that of BOAlb. Also, because the residence times of urinary metabolites are less than 1 day, the errors in these estimated (urinary) levels would be correlated with those of benzene exposure from day to day. In this case, the effect of measurement error on the estimated regression coefficient would be more profound than that for stable protein adducts. For these reasons, we regard the inferences derived from BOAlb, about the overall rate of benzene metabolism, to be more accurate and precise than those from urinary data. This inherent strength of protein adducts for elucidating rates of human metabolism has not, to our knowledge, been exploited heretofore. Finally, we recognize that our finding about the non-linearity of human benzene metabolism is based upon only 30 subjects from a single investigation. We are currently conducting a similar study among a larger population of workers, to more thoroughly explore human metabolism of benzene over the critical exposure range of 110 ppm. Acknowledgements This work was supported by grants P42ES05948 and P42ES04705 from the National Institute of Environmental Health Sciences. The authors are grateful to Lawrence Kupper for helpful discussions regarding attenuation of exposureadduct relationships. References [1] IARC, Monographs On the Evaluation of the Carcinogenic Risk of Chemicals To Humans, Benzene, Vol. 45, International Agency for Research on Cancer, Lyon, 1989. [2] R. Snyder, C.C. Hedli, Environ. Health Perspect. 104 (Suppl. 6) (1996) 1165. [3] M.T. Smith, Environ. Health Perspect. 104 (Suppl. 6) (1996) 1219. [4] A.B. Lindstrom, K. Yeowell-O'Connell, S. Waidyanatha, B.T. Golding, R. Tornero-Velez, S.M. Rappaport, Carcinogenesis 18 (1997) 1637. [5] M.R. Lovern, M.J. Turner, M. Myer, G.L. Kedderis, W.E. Bechtold, P.M. Schlosser, Carcinogenesis 18 (1997) 1695. [6] L. Wallace, Environ. Health Perspect. 104 (Suppl. 6) (1996) 1129. [7] OSHA, Fed. Reg. 52 (1987) 34460. [8] N. Rothman, G.L. Li, M. Dosemeci, W.E. Bechtold, G.E. Marti, Y.Z. Wang, M. Linet, L.Q. Xi, W. Lu, M.T. Smith, N. Titenko-Holland, L.P. Zhang, W. Blot, S.N. Yin, R.B. Hayes, Am. J. Ind. Med. 29 (1996) 236. [9] N. Rothman, W.E. Bechtold, S.N. Yin, M. Dosemeci, G.L. Li, Y.Z. Wang, W.C. Griffith, M.T. Smith, R.B. Hayes, Occup. Environ. Med. 55 (1998) 705. [10] M.T. Smith, L. Zhang, Y. Wang, R.B. Hayes, G. Li, J. Wiemels, M. Dosemeci, N. Titenko-Holland, L. Xi, P. Kolachana, S. Yin, N. Rothman, Cancer Res. 58 (1998) 2176. [11] K. Yeowell-O'Connell, N. Rothman, M.T. Smith, R.B. Hayes, G. Li, S. Waidyanatha, M. Dosemeci, L. Zhang, S. Yin, N. Titenko-Holland, S.M. Rappaport, Carcinogenesis 19 (1998) 1565. [12] M. Troester, A.B. Lindstrom, S. Waidyanatha, L.L. Kupper, S.M. Rappaport, Toxicol. Sci. 54 (2000) 88. [13] F. Granath, L. Ehrenberg, M. Tornqvist, Mutat. Res. 284 (1992) 297. [14] K. Yeowell-O'Connell, N. Rothman, S. Waidyanatha, M.T. Smith, R.B. Hayes, W.E. Bechtold, G. Li, M. Dosemeci, L. 374 S.M. Rappaport et al. / J. Chromatogr. B 778 (2002) 367374 Zhang, S. Yin, S.M. Rappaport, Cancer Epidemiol. Biomarkers Prev. 10 (2001) 831. [15] W.E. Bechtold, J.K. Willis, J.D. Sun, W.C. Griffith, T.V. Reddy, Carcinogenesis 13 (1992) 1217. [16] K. Yeowell-O'Connell, T.A. McDonald, S.M. Rappaport, Anal. Biochem. 237 (1996) 49. [17] A.B. Lindstrom, K. Yeowell-O'Connell, S. Waidyanatha, T.A. McDonald, B.T. Golding, S.M. Rappaport, Chem. Res. Toxicol. 11 (1998) 302. [18] D.R. Koop, C.L. Laethem, G.G. Schnier, Toxicol. Appl. Pharmacol. 98 (1989) 278. [19] F.P. Guengerich, D.H. Kim, M. Iwasaki, Chem. Res. Toxicol. 4 (1991) 168. [20] M.J. Seaton, P.M. Schlosser, J.A. Bond, M.A. Medinsky, Carcinogenesis 15 (1994) 1799. [21] J.L. Valentine, S.S.T. Lee, M.J. Seaton, B. Asgharian, G.M. Farris, J.C. Corton, F.J. Gonzalez, M.A. Medinsky, Toxicol. Appl. Pharmacol. 141 (1996) 205. [22] K.H. Watanabe, F.Y. Bois, J.M. Daisey, D.M. Auslander, R.C. Spear, Occup. Environ. Med. 51 (1994) 414. [23] F.Y. Bois, E.T. Jackson, K. Pekari, M.T. Smith, Environ. Health Perspect. 104 (Suppl. 6) (1996) 1405. [24] M.A. Medinsky, E.M. Kenyon, M.J. Seaton, P.M. Schlosser, Environ. Health Perspect. 104 (Suppl. 6) (1996) 1399. [25] S.M. Rappaport, E. Symanski, J.W. Yager, L.L. Kupper, Environ. Health Perspect. 103 (1995) 49. [26] S. Waidyanatha, N. Rothman, S. Fustinoni, M.T. Smith, R.B. Hayes, W. Bechtold, M. Dosemeci, G. Li, S. Yin, S.M. Rappaport, Carcinogenesis 22 (2001) 279.