Molecular epidemiological studies in 1,3-butadiene exposed Czech workers: Female male comparisons

Chemico-Biological Interactions 166 (2007) 63 77 Molecular epidemiological studies in 1,3-butadiene exposed Czech workers: male comparisons Richard J. Albertini a,, Radim J. Sram b, Pamela M. Vacek c, Jeremiah Lynch d,, Pavel Rossner b, Janice A. Nicklas a, Jake D. McDonald e, Gunnar Boysen f, Nadia Georgieva f, James A. Swenberg f a BioMosaics, Inc., 665 Spear Street, Burlington, VT 05401, United States b Laboratory of Genetic Ecotoxicology, Prague, Czech Republic c University of Vermont, Burlington, VT 05401, United States d Rumson NJ, United States e Lovelace Respiratory Research Institute, Albuquerque, MN, United States f University of North Carolina, Chapel Hill, NC, United States Available online 26 July 2006 The investigators dedicate this report to Jerry Lynch who was responsible for the exposure assessment design for both Czech studies. Jerry passed away during the conduct of this second study; he is greatly missed. Abstract Results of a recent molecular epidemiological study of 1,3-butadiene (BD) exposed Czech workers, conducted to compare female to male responses, have confirmed and extended the findings of a previously reported males only study (HEI Research Report 116, 2003). The initial study found that urine concentrations of the metabolites 1,2-dihydroxy-4-(acetyl) butane (M1) and 1-dihydroxy- 2-(N-acetylcysteinyl)-3-butene (M2) and blood concentrations of the hemoglobin adducts N-[2-hydroxy-3-butenyl] valine (HB-Val) and N-[2,3,4-trihydroxy-butyl] valine (THB-Val) constitute excellent biomarkers of exposure, both being highly correlated with BD exposure levels, and that GST genotypes modulate at least one metabolic pathway, but that irreversible genotoxic effects such as chromosome aberrations and HPRT gene mutations are neither associated with BD exposure levels nor with worker genotypes (GST [glutathione-s-transferase]-m1, GSTT1, CYP2E1 (5 promoter), CYP2E1 (intron 6), EH [epoxide hydrolase] 113, EH139, ADH [alcohol dehydrogenase]2 and ADH3). The no observed adverse effect level (NOAEL) for chromosome aberrations and HPRT mutations was 1.794 mg/m 3 (0.812 ppm) the mean exposure level for the highest exposed worker group in this initial study. The second Czech study, reported here, initiated in 2003, included 26 female control workers, 23 female BD exposed workers, 25 male control workers and 30 male BD exposed workers (some repeats from the first study). Multiple external exposure measurements (10 full 8-h shift measures by personal monitoring per worker) over a 4-month period before biological sample collections showed that BD workplace levels were lower than in the first study. Mean 8-h TWA exposure levels were 0.008 mg/m 3 (0.0035 ppm) and 0.397 mg/m 3 (0.180 ppm) for female controls and exposed, respectively, but with individual single 8-h TWA values up to 9.793 mg/m 3 (4.45 ppm) in the exposed group. Mean male 8-h TWA exposure levels were 0.007 mg/m 3 (0.0032 ppm) and 0.808 mg/m 3 (0.370 ppm) for controls and exposed, respectively; however, the individual single 8-h TWA values up to 12.583 mg/m 3 (5.72 ppm) in the exposed group. While the urine metabolite concentrations for both M1 and M2 were elevated in exposed compared to control females, the Corresponding author. Tel.: +1 802 656 8346; fax: +1 802 656 5446. E-mail addresses: Ralbert315@aol.com, Richard.Albertini@uvm.edu (R.J. Albertini). Deceased. 0009-2797/$ see front matter 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.cbi.2006.07.004

64 R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 differences were not significant, possibly due to the relatively low BD exposure levels. For males, with greater BD exposures, the concentrations of both metabolites were significantly elevated in urine from exposed compared to control workers. As in the first study, urine metabolite excretion patterns in both sexes revealed conjugation to be the minor detoxification pathway (yielding the M2 metabolite) but both M1 and M2 concentration values were lower in males in this second study compared to their concentrations in the first, reflecting the lower external exposures of males in this second study compared to the first. Of note, females showed lower concentrations of both M1 and M2 metabolites in the urine per unit of BD exposure than did males while exhibiting the same M1/(M1 + M2) ratio, reflecting the same relative utilization of the hydrolytic (producing M1) and the conjugation (producing M2) detoxification pathways as males. Assays for the N,N-(2,3-dihydroxy-1,4-butadyl) valine (pyr-val) hemoglobin (Hb) adduct, which is specific for the highly genotoxic 1,2,3,4-diepoxybutane (DEB) metabolite of BD, have been conducted on blood samples from all participants in this second Czech study. Any adduct that may have been present was below the limits of quantitation (LOQ) for this assay for all samples, indicating that production of this important BD metabolite in humans is below levels produced in both mice and rats exposed to as little as 1.0 ppm BD by inhalation (J.A. Swenberg, M.G. Bird, R.J. Lewis, Future directions in butadiene risk assessment, Chem. Biol. Int. (2006), this issue). Results of assays for the HB-Val and THB-Val hemoglobin adducts are pending. HPRT mutations, determined by cloning assays, and multiple measures of chromosome level changes (sister-chromatid exchanges [SCE], aberrations determined by conventional methods and FISH) again showed no associations with BD exposures, confirming the findings of the initial study that these irreversible genotoxic changes do not arise in humans occupationally exposed to low levels of BD. Except for lower production of both urine metabolites in females, no female male differences in response to BD exposures were detected in this study. As in the initial study, there were no significant genotype associations with the irreversible genotoxic endpoints. However, as in the first, differences in the metabolic detoxification of BD as reflected in relative amounts of the M1 and M2 urinary metabolites were associated with genotypes, this time both GST and EH. 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: 1,3-Butadiene; Molecular epidemiology; M1 metabolite; M2 metabolite; pyr-val adducts; HPRT mutations; Chromosome aberrations; Genotypes; Risk assessment 1. Introduction and background Biomarker responses in humans can reflect toxicological mechanisms as well as internal exposures, susceptibility or effects. Several studies describing such responses in 1,3-butadiene (BD) exposed workers have been reported over the past 15 years [reviewed in 1 3]. Study design has evolved over that period. Originally small and focused on one or two biomarkers, these investigations have become more comprehensive. However, most have still suffered from a lack of extensive BD exposure assessments and/or identification of potential confounding agents. Not surprisingly, results have been conflicting. Although urinary metabolites and hemoglobin (Hb) adducts have emerged as potentially useful biomarkers of exposure, it is not clear from the weight of evidence if irreversible genotoxic effects such as chromosome aberrations or somatic gene mutations have been associated with BD exposures. Investigations of susceptibility, as reflected by metabolic genotypes, have also produced mixed results. In an attempt to resolve inconsistencies in past studies, a large-scale international, multi-institution molecular epidemiological study of BD exposed Czech BD workers was initiated in 1998 and reported in 2003 [1,2]. The Czech population investigated was the same as that investigated in several earlier smaller studies; investigators who had previously reported either negative or positive results were collaborators, as were additional investigators who measured the same or additional endpoints. Prior to initiating this first Czech study, an elaborate blinding scheme was developed whereby individual workers were assigned random numbers. Different fractions of the same urine or blood sample also received different numbers, each referenced to the individual worker, with the key to the overall blinding scheme known only to the central study office. Blinding codes were broken for individual investigators only after all samples in any laboratory had been analyzed and the report filed. BD exposures were assessed by 10 independent 8-h personal monitoring measurements (on average) conducted over a 2-month period for each potentially exposed worker. Ambient air BD levels and co-exposures to toluene, styrene and benzene (by personal monitors and air samples) were also measured. Blood and urine samples were collected, the latter on three occasions at the end of the exposure assessment period. Samples were fractionated, blinded, cryo-preserved and subsequently sent to collaborating laboratories. This first Czech study included a total of 83 subjects: 24 BD monomer production workers (mean BD exposure = 0.642 mg/m 3 [0.290 ppm]), 34 polymerization workers (mean BD exposure = 1.794 mg/m 3 [0.812 ppm]) and 25 controls (mean BD exposure = 0.023 mg/m 3 [0.010 ppm]). Mean exposures for individual workers (approximately 10 full 8-h shift

R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 65 measurements per worker over an approximately 2- month period) ranged up to 3.51 mg/m 3 (1.60 ppm) and 9.24 mg/m 3 (4.20 ppm) BD in the monomer production and polymerization facilities, respectively. Compared to controls, group mean 1,2-dihydroxy-4-[N-acetylcysteinyl]-butane (M1) and 1-dihydroxy-2-[N-acetylcysteinyl]-3-butene (M2) urine metabolite concentrations and N-[2-hydroxy-3-butenyl] valine (HB-Val) and N- [2,3,4-trihydroxy-butyl] valine (THB-Val) adduct concentrations were all elevated in the exposed groups and were significantly correlated with workers mean BD exposure levels. The Hb adduct concentrations were the most precise biomarkers of exposure. The ratio of urinary metabolite concentrations [M1/(M1 + M2)] was approximately 0.99 in this initial study. As M1 is a product of the hydrolytic detoxification pathway (mediated by epoxide hydrolase) and M2 derives from conjugation (mediated by glutathione- S-transferase), this finding confirmed that BD detoxification in humans relies almost entirely on the former ([2,4] and references therein). Hb adduct concentrations also reflect internal BD metabolism, with HB-Val being derived from oxidation of BD to 1,2-dihydroxy- 3-butene (EB) and THB-Val being formed from 1,2- dihydroxy-3,4-epoxybutane (EBD), itself a product of the hydrolytic detoxification pathway [2,5]. THB-Val adduct concentrations were 300 400-fold higher than HB-Val adduct concentrations in the first Czech study, also reflecting the dominance of the hydrolytic pathway in producing greater in vivo accumulations of EBD than EB in humans. Despite the well documented external BD exposures, and the urinary metabolites and Hb adducts revealing appreciable internal doses in exposed workers, there was no evidence of irreversible DNA genotoxicity (mutations or chromosome aberrations) related to BD exposure (determined by either the direct measurements or as reflected by the urinary metabolite or adduct levels) in this initial study. Neither sister chromatid exchange (SCE) frequencies nor high frequency cells (HFC) were related to the BD exposures themselves or to their estimates, using the surrogate biomarkers of exposure. Regression analyses showed no relationships between individual BD exposures (means of the multiple measurements per worker over 2 months) and HPRT mutations, chromosome aberrations or SCEs, regardless of assay methods employed. Several metabolic genotypes were determined: GSTM1, GSTT1, CYP2E1 (5 promoter region), CYP2E1 (intron 6), EH113, EH139, ADH (alcohol dehydrogenase)2 and ADH3. GST polymorphisms did influence BD metabolism as reflected by relative M1 and M2 metabolite concentrations in urine, with the null genotypes being associated with lower activity in the conjugation detoxification pathway. However, no other endpoints were affected by genotype. Specifically, combinations of the EH polymorphisms specifying low, intermediate or high enzyme activity were unrelated to any of the endpoints measured. Only male workers were included in this initial Czech study. The no observed adverse level (NOAEL) of 1.794 mg/m 3 (0.812 ppm) for irreversible genotoxic effects (HPRT mutations or chromosome aberrations) was based on the group mean BD exposure level of the highest exposed group, i.e. the polymerization workers. In order to evaluate the possibility that females may be more susceptible to BD genotoxicity than males, a second similar study including females was initiated in 2003. The second study differed from the first in that a longer exposure assessment period was employed (4 months) to include the entire life-span of human red blood cells (RBCs). The autographic assay, as a second measure of HPRT mutations, was eliminated and additional genotype polymorphisms were determined. Importantly, a recently developed assay for the N,N-[2,3-dihydroxy-1,4-butadyl] valine (pyr-val) adduct, which is specific for the highly genotoxic 1,2,3,4- dihydroxybutane (DEB) metabolite of BD, was added [6]. Results are reported here. 2. Study design and methods Study design and methods were largely as reported for the first study with the subtractions and additions noted above [1,2]. Exposure assessment measurements, questionnaire administrations and biological sample collections were conducted at the industrial facility in the Czech Republic. Chemical analyses of exposure badges were performed by the EcoChem Company in Prague. Biological samples were processed, fractionated, labeled, cryo-preserved and subsequently sent by Air Express to collaborating laboratories by the Laboratory of Genetic Toxicology in Prague. Cytogenetic analyses were also conducted at this facility. Urine metabolite determinations were made at the Lovelace Respiratory Research Institute in Albuquerque. Hb adduct assays were conducted at the University of North Carolina. HPRT gene mutations were measured in Burlington, as were worker genotype determinations. All data were analyzed by the Central Study Office in Burlington. 2.1. Blinding Random numbers generated in the Central Study Office (Burlington) were sent to Prague for assignment

66 R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 to workers at the beginning of the exposure assessment period and to fractions of biological samples after their processing. 2.2. Study subjects One hundred and four workers from the same industrial facility that took part in the first study volunteered for the current study: 49 females and 55 males. Twentysix of the female subjects were administrative controls and 23 worked in a laboratory setting. Twenty-five of the male subjects were administrative controls and 30 were drawn from a polymerization unit. All participants completed questionnaires, as in the first study, with modifications to include female-specific questions. All participants signed informed consent forms approved by both the Institutional Review Board of the University of Vermont and the Bohemian Institute of Hygiene. The research conformed to the Helsinki Principles. 2.3. Exposure assessment BD exposures were measured by personal monitoring using diffusive solid sorbent tubes for 8-h work shifts on 10 separate occasions over a 4-month interval for each of the 104 study subjects. Ambient air samples were also analyzed for BD concentrations. Co-exposures to toluene, styrene, and benzene were measured by personal monitoring on a single occasion and by ambient air sampling. All tubes were collected at the work facility by representatives of the EcoChem Company of Prague, where chemical analyses were performed. 2.4. Genotyping Genotyping was performed in Burlington using a small portion of samples sent for HPRT mutation studies. The same metabolic genes analyzed in the first study were again analyzed with the addition of a PCR-based method to detect GST heterozygosity. Methods were essentially as described for the first study, i.e. PCR for GST null genotypes and PCR followed by digestion with appropriate restriction enzymes for the others. 2.5. Sample acquisition, processing and blinding Spot urine samples were collected before and after shifts on the last 3 days of the exposure assessment period. Blood samples were collected only on the last day. All samples were obtained in the medical unit of the work facility and transported on ice to the Laboratory of Genetic EcoToxicology in Prague where they were immediately processed. Processing included fractionating the blood samples into components and cryopreserving in the vapor phase of liquid nitrogen. Whole blood cultures for cytogenetic analyses were established at that time. Urine samples were frozen at 70 C. All workers and sample aliquots were labeled with unique code numbers. Coded cryo-preserved aliquots of appropriate samples were later shipped (by air express) in dry shippers (liquid nitrogen vapor) or on ice (urine samples) to participating laboratories performing the various biomarker assays. All assays were conducted blindly, with results electronically transmitted to the Biometry Facility of the University of Vermont for statistical analyses. Codes were broken only after all assays from the collaborating laboratory had been completed. (Codes did not have to be broken for interpretation of the pyr-val adduct determinations as all samples were below the limit of quantitation (LOQ) for the method (see below). Therefore, RBC samples remain blinded for the subsequent HB-Val and THB-Val adduct determinations.) 2.6. Biomarker assays The different biomarkers in blood or urine were assayed using the following methods. The investigators responsible for the different biomarker assays or statistical analyses are indicated. 2.6.1. Urine metabolites (JCM) Urine concentrations of the M1 and M2 metabolite concentrations were determined in coded samples using high-pressure liquid chromatographic separation and triple quadruple mass spectrometry (LC/MS) after isolation by solid phase extraction, as described previously [7]. 2.6.2. pyr-val Hb adducts (JAS) The newly developed immunoaffinity LC-MS/MS assay for the pyr-val adduct was used to analyze 50 mg globin from each of the 104 RBC samples collected in this study exactly as described elsewhere [6]. RBCs have been cryo-preserved for HB-Val and THB-Val assays determinations, but results are not yet available. 2.6.3. HPRT gene mutations (RJA) HPRT mutations in T-cells were measured by cloning assays performed as described elsewhere [8]. After thawing, cells were plated in limiting dilutions with or without thioguanine selection and were scored for colony formation. Cloning efficiencies (CE) were calculated by use of the Poisson relationship, P 0 =e x, where P 0 is the fraction of wells without colony growth and x is the average

R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 67 number of clonable cells per well. The CE = ln P 0 /N, where N is the average number of cells inoculated per well. The mutant frequency is the ratio of the CE in the presence (selected) and the absence (non-selected) of thioguanine. 2.6.4. Cytogenetic analyses (RJS) Chromosome aberrations and sister-chromatid exchanges (SCE) were assessed using methods as described in detail for the initial Czech study [1,2]. Fluorescent in situ hybridization (FISH) analyses using whole chromosome painting probes in different colors (Biovation, Aberdeen, UK) for chromosomes 1 and 4, scored for chromosome aberrations, translocations, reciprocal translocations (%), co-junctions and stable translocations per 1000 cells. Aberrent cells were classified according to the Protocol for Aberration Identification and Nomenclature Terminology ( PAINT ) [9]. Genomic frequencies of stable translocation frequencies per 100 cells (F G /100) were calculated according to Lucas and associates (1993) using to the equation: F G /100 = F gr /2.05[f r (1 f r )+f g (1 f g ) f r f g ] where F gr is the translocation frequency measured by FISH after two-color painting and f r and f g are the fractions of the genome painted red or green, respectively [10]. 2.7. Statistical analyses (PMV) Questionnaires were mailed to the Biometry Facility of the University of Vermont (CSO) where they were screened for completeness and consistency. Exposure and biomarker data from participating laboratories were transferred electronically to the same facility. Unless otherwise noted, groups were compared using the non-parametric Kruskal Wallis (K W) test. Pairwise differences were assessed by Mann Whitney tests using a Bonferoni adjustment for multiple comparisons. Spearman s rank correlation coefficients were computed to assess bivariate associations between exposure measurements and biomarkers of exposure and effect. Pearson correlation coefficients were also computed to assess the linearity of these relationships after logarithmic transformation of the exposure measurements and biomarkers. Multiple regression was used to examine whether the linear relationships differed between males and females or between metabolic genotypes. These regression analyses included an interaction term to test whether the rate of increase in the biomarker per unit of exposure was modified by sex or genotype. 3. Results 3.1. Study groups Demographic data are presented in Table 1. The four study groups (control male, control female, exposed male and exposed female) are fairly well balanced with respect to age and time in district. As in the first study, there was a statistically significant difference in educational level achieved among males, with the controls having a greater number of years in school. Although fewer control than exposed males are current smokers, the difference is not statistically significant. There were no significant differences between the sexes or between female control and exposed workers as regards to current smoking. Table 1 Demographic information on workers Control (mean ± S.D.) Exposed (mean ± S.D.) (n = 26) (n = 25) (n = 23) (n = 30) Age 42.8 ± 8.2 40.1 ± 10.7 38.7 ± 9.5 40.2 ± 13.0 Years in company 17.6 ± 9.3 13.2 ± 8.7 19.4 ± 9.9 18.8 ± 13.1 Years in district 28.8 ± 14.3 28.4 ± 12.9 35.3 ± 11.1 29.3 ± 15.4 Control (n (%)) Exposed (n (%)) (n = 26) (n = 25) (n = 23) (n = 30) Education a <12 years 2 (8) 1 (4.0) 3 (13) 11 (37) 12 years 13 (50) 15 (60) 14 (60) 14 (47) >12 years 11(42) 9 (36) 6 (26) 5 (17) Current smoking (%) 24.0 31.8 26.1 41.4 No significant difference in current smoking status. a M-exposed differ significantly from M-controls p < 0.05 by χ 2 -test.

68 R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 Table 2 Butadiene exposure assessment (mg/m 3 ) individual 8-h measurements (10 per subject) Table 3 Butadiene exposure assessment (mg/m 3 ) average of ten 8-h measurements/subject control exposed control exposed control exposed control exposed N 260 230 249 300 Mean 0.008 0.397 0.007 0.808 S.D. 0.015 1.094 0.012 1.663 Median 0.004 0.056 0.004 0.241 Minimum 0.004 0.004 0.004 0.004 Maximum 0.219 9.793 0.157 12.583 LOD = 0.0088 mg/m 3 ; significant differences (p < 0.05; K W); M(exp) vs. M(control); F(exp) vs. F(control); M(exp) vs. F(exp). Analysis of questionnaire responses for femalespecific adverse health questions showed no significant differences between controls and exposed for miscarriages, still births, ectopic pregnancies, molar pregnancies, low birth weight (<2500 g) babies, or pre-term births, based on information collected on all pregnancies. 3.2. BD, styrene, toluene and benzene exposure concentrations Results of individual 8-h BD concentration measurements are presented in Table 2. The 260 measurements for the 26 control females showed a mean 8-h exposure of 0.008 mg/m 3 (0.004 ppm), with the maximum single measurement in this group being 0.219 mg/m 3 (0.100 ppm). The 230 measurements for the 23 exposed females showed a mean 8-h exposure of 0.397 mg/m 3 (0.180 ppm), with the maximum single measurement being 9.793 mg/m 3 (4.451 ppm). controls showed a mean 8-h exposure of 0.007 mg/m 3 (0.003 ppm), based on 249 individual measurements, with the single highest measurement being 0.157 mg/m 3 (0.070 ppm), while the male exposed workers had a mean exposure of 0.808 mg/m 3 (0.367 ppm), based on 300 individual measurements, with a maximum single measurement being 12.583 mg/m 3 (5.72 ppm). The limit of detection (LOD) for the assay employed was 0.0088 mg/m 3 (0.004 ppm); measurements below this were assigned half this value N 26 23 25 30 Mean 0.008 0.397 0.007 0.808 S.D. 0.005 0.502 0.005 1.646 Median 0.006 0.229 0.005 0.646 Minimum 0.004 0.006 0.004 0.086 Maximum 0.028 2.199 0.030 3.117 LOD = 0.0088 mg/m 3 ; significant differences (p < 0.05; K W); M(exp) vs. M(control); F(exp) vs. F(control); M(exp) vs. F(exp). for statistical analyses. As shown, the male BD exposed worker concentrations were significantly higher than the male control concentrations; the female BD exposed worker concentrations were significantly higher than the female control concentrations and the male BD exposed worker concentrations were significantly higher than the female BD exposed worker concentrations. Table 3 presents the BD exposure concentrations in terms of averages of the ten 8-h measurements/subject, which is the value assigned to an individual worker over the 4-month interval. Mean exposures are identical when calculated in this fashion because each worker had the same number of measurements. However, the maximum 4-month average for a control female was 0.028 mg/m 3 (0.013 ppm) while, for an exposed female, it was 2.199 mg/m 3 (1.000 ppm). For a control male, the maximum 4-month exposure average was 0.030 mg/m 3 (0.136 ppm), while, for an exposed male, it was 3.117 mg/m 3 (1.414 ppm). Four-month average concentrations are significantly higher for exposed than for controls for both sexes, while these concentrations are significantly higher in exposed males than in exposed females. The BD exposure concentrations for male polymerization workers in this study were lower than those found in this same facility in the initial study. Co-exposures to styrene, toluene and benzene are shown in Table 4. Many values were below the LOD Table 4 VOC coexposures (mg/m 3 ) average of ten 8-h measurements/subject Styrene a Toluene b Benzene c Control Exposed Control Exposed Control Exposed 0.024 ± 0.000 0.478 ± 0.672 0.011 ± 0.000 0.636 ± 0.746 0.013 ± 0.000 0.029 ± 0.047 0.041 ± 0.059 1.354 ± 0892 0.024 ± 0.036 0.018 ± 0.019 0.013 ± 0.000 0.013 ± 0.000 a Significant difference: F(exp) vs. F(control), M(exp) vs. M(control), M(exp) vs. F(exp). b Significant difference: F(exp) vs. F(control), F(exp) vs. M(exp). c Significant difference: F(exp) vs. M(exp).

R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 69 and were assigned half the LOD for statistical purposes. Forty-six workers had styrene levels above the LOD; 22 had toluene levels above the LOD while only three (all exposed females) had benzene levels above the LOD. All female controls had levels below the LOD for all three volatile organic chemicals (VOCs) measured. There were some significant differences in these exposure levels between study groups: male BD exposed worker styrene concentrations were higher than male controls; female BD exposed worker concentrations of styrene and toluene were higher than female controls; and male BD exposed worker styrene concentrations were higher than female BD exposed worker concentrations, while female BD exposed workers had higher concentrations of toluene and benzene than male BD exposed workers. 3.3. Urine M1 and M2 metabolites M1 and M2 metabolite concentrations in urine were determined for all study subjects before and after work shifts on a day of BD exposure measurement. Urine samples were obtained from all workers on the last three days of the BD exposure assessment period. To date, urine metabolite concentrations have been determined only for the first of these samples from each worker. Before shift concentrations, after shift concentrations, and their difference, were analyzed, and all analyses led to the same overall conclusions. Analyses were also done both with and without correction for creatinine concentration. However, adjustment for creatinine did not reduce the variability of the metabolite measurements and is problematic for comparison of males and females because of sex differences in creatinine excretion (see below). Therefore, only after work shift values uncorrected for creatinine are presented here. Table 5 shows after work M1 concentrations by exposure group and sex. The mean concentration for control Table 5 M1 after work urine concentrations by group and sex ( g/l) N Mean S.D. Control 26 331.6 284.9 Exposed 23 508.1 597.4 Control 25 512.8 272.1 Exposed 30 854.1 567.0 exposed significantly greater than male control (K W test; p < 0.05). exposed vs. female control: NS. control and exposed significantly different from female control and exposed, respectively (p < 0.05). Table 6 M2 after work urine concentrations by group and sex ( g/l) N Mean S.D. Control 26 8.3 10.1 Exposed 23 19.2 27.5 Control 25 14.9 10.3 Exposed 30 47.9 44.3 exposed significantly greater than male control (K W test; p < 0.05). control and exposed significantly greater than female control and exposed, respectively (p < 0.05). females was 331.6 ± 284.9 g/l while, for BD exposed females, it was 508.1 ± 597.4 g/l. While the mean value was elevated in the exposed group, the difference is not statistically significant. Control males had a group mean M1 concentration of 512.8 ± 272.1 g/l while BD exposed males had a group mean concentration of 854.1 ± 567.0 g/l. This mean difference is statistically significant, as are the mean higher values for male controls and BD exposed workers as compared to female controls and BD exposed workers, respectively. After work M2 concentrations for these groups are shown in Table 6. These values are much lower than the M1 values, with female controls having a mean of 8.3 ± 10.1 g/l compared to female BD exposed workers having a mean of 19.2 ± 27.5 g/l. While elevated in the exposed group, this difference is not statistically significant. controls had a mean M2 concentration of 14.9 ± 10.3 g/l compared to male BD exposed workers with a mean concentration of 47.9 ± 44.3 g/l. The higher mean concentration in the exposed, compared to the control group, is significantly different, as are the male control and BD exposed worker values compared to the female control and BD exposed worker values, respectively. Associations between BD exposure concentrations and individual worker urine metabolite concentrations were analyzed for each metabolite for the female and male workers separately as well as in combination, the latter to determine if the lower excretions in females as a group were the result of their lower external BD exposures or because they excrete less as individuals per unit of BD exposure (Figs. 1 and 2). The BD exposure levels presented in the figures are subjects 4-month means (10 measurements). There are significant associations between BD exposure and M1 metabolite concentration for males (Fig. 1) and between BD exposure and M2 for both sexes (Fig. 2). This difference between female and male excretions per unit of BD exposure is statistically

70 R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 Table 7 M2/(M1 + M2) ratio by group and sex N Mean S.D. Control 26 0.0239 0.0193 Exposed 23 0.0373 a 0.0232 Control 25 0.0291 0.0157 Exposed 30 0.0519 a,b 0.0385 a Exposed significantly greater than control of same sex (p < 0.05; K W test). b exposed significantly greater than female exposed (p < 0.05; K W test). Fig. 1. Association between individual urine M1 concentrations (ln[ g/l]) and 4-month average BD exposure (ln[mg/m 3 ]) by sex. Open circles and dashed line indicate females; closed circles and solid line indicate males. Association significant (p < 0.05) for males; intercepts for males and females differ significantly (p < 0.05). significant. It should be noted that this difference is no longer apparent when urine metabolite concentrations are corrected for creatinine concentrations. This can be attributed to the fact that female creatinine concentrations in the present study were approximately 60 80% of male concentrations, a ratio that is similar to normal differential values for creatinine between the sexes. These differences are thought to reflect differences in muscle mass between the sexes so correcting for them when making inter-sex comparisons of urinary excretion concentrations is not appropriate [11 13]. The male BD exposed worker group mean values for both M1 and M2 metabolite concentrations are lower than those determined in the first Czech study for male polymerization workers, as expected, because of the lower external BD exposure levels found in the current study. Urinary M2/(M1 + M2) ratios reflecting the proportion of detoxification occurring via the GST mediated conjugation pathway are presented in Table 7. (Although this ratio has been presented as M1/(M1 + M2) thus far in this report, the complementary proportion is given here and elsewhere when attention is directed to differences in the minor conjugation detoxification pathway.) As shown, in female controls approximately 2% of the detoxification leading to urine metabolites utilized this pathway, compared to close to 4% that utilized it in the female BD exposed worker group. This difference is significant. Similarly, for males, the control males showed approximately 3% utilization of this GST mediated conjugation pathway compared to BD exposed male workers who showed approximately 5% utilization, a significant difference. The M2/(M1 + M2) ratio is also significantly greater in male BD exposed workers compared to female BD exposed workers. This sex difference however, reflects the greater BD exposures in males because, in contrast to the M1 and M2 metabolite concentrations alone (Figs. 1 and 2), this ratio, while significantly associated with individual BD exposure levels, does not show a sex difference per unit of BD exposure. 3.4. Hb adducts Fig. 2. Association between individual urine M2 concentrations (ln[ g/l]) and 4-month average BD exposure (ln[mg/m 3 ]) by sex. Open circles and dashed line indicate females; closed circles and solid line indicate males. Associations significant (p < 0.05) for both sexes; intercepts for males and females differ significantly (p < 0.05). Assays for pyr-val adducts have been completed for all blood samples in the study. Adduct concentrations for all 104 workers (females and males, controls and BD exposed) were below the LOQ when 50 mg globin samples were analyzed. This is the same assay that had previously detected and quantified pyr-val adducts in

R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 71 Table 8 HPRT mutations (group means) a Control ( 10 6 ) Exposed ( 10 6 ) 16.5 ± 7.5 (n = 26) 14.5 ± 7.4 (n = 23) 14.7 ± 10.1 (n = 25) 13.7 ± 8.3 (n = 30) No significant differences between control and exposed, either sex. Overall female significantly > overall male MFs when corrected for cloning efficiency. a Uncorrected for non-selected cloning efficiencies. blood from rats and mice exposed to BD by inhalation at levels as low as 1.0 ppm [14,15]. HB-Val and THB-Val HB adduct concentrations have not yet been determined. 3.5. HPRT mutations HPRT mutations arising in vivo in T-lymphocytes were measured by cloning assays for all 104 subjects. Assays for all workers were technically acceptable with mean non-selected cloning efficiencies (CE s) (measure of cell growth) of 0.29 ± 0.11 and 0.28 ± 0.09 for control and BD exposed female workers, respectively, and 0.28 ± 0.17 and 0.23 ± 0.11 for control and BD exposed male workers, respectively. There were no significant differences in CE among the four groups. Mutant frequency (MF) values are presented in Table 8. The group mean for control females was 16.5 ± 7.5 10 6 as compared to 14.5 ± 7.4 10 6 for BD exposed females. Control males had a group mean value of 14.7 ± 10.1 10 6 compared to a group mean value of 13.7 ± 8.3 10 6 for exposed males. These values are uncorrected. However, analyses correcting for CE, age and smoking status gave similar results. Smoking had no effect on MF. Similar to the lack of association between group mean HPRT MF values and BD exposures, there was no association between workers mean BD exposure levels (average of 10 measurements) and individual MF values (Fig. 3). 3.6. Cytogenetic changes Cytogenetic studies included assays for SCEs, chromosome aberrations determined by traditional methods, and chromosome changes determined by fluorescence in situ hybridization (FISH). SCE results are shown in Table 9. As can be seen, the mean values for SCEs/cell are almost identical for all four groups. The same is true for chromosome aberrations assayed by conventional staining methods. The percent of cells showing aberrations was 2.4 ± 1.7 in female controls compared with Fig. 3. Association between individual HPRT mutant frequencies (ln MF) and 4-month average BD exposure (ln[mg/m 3 ]) for all workers. No significant association. 2.8 ± 1.6 for female BD exposed workers (Table 10). The comparable values were 2.6 for both male controls and BD exposed workers. The mean value for breaks/cell was 0.03 in all four groups. Table 11 presents the multiple endpoints measured in the FISH analysis. As for the SCE studies and the conventional analysis for chromosome aberrations, there were no significant differences between groups. Neither SCEs nor chromosome aberrations determined by conventional methods, nor chromosome aberrations determined by FISH studies, showed significant differences between groups in these multiple analyses. Associations between these many cytogenetic endpoints and workers mean BD exposure levels (10 measurements) were also tested and none showed associations with BD exposure (results not shown). 3.7. Genotype effects The first Czech study (male workers only) showed that individuals with either the GSTM1 or GSTT1 null genotype had a lower rate of rise in excretion of the M2 urinary metabolite associated with increasing BD exposure levels than did GST positive individuals. This Table 9 Sister chromatid exchanges control exposed control exposed N 26 23 23 a 30 SCEs/cell 5.3 ± 0.5 5.0 ± 0.7 5.3 ± 0.8 5.3 ± 0.8 No significant differences. a Two observations missing.

72 R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 Table 10 Chromosome aberrations (conventional analysis) control exposed control exposed N 26 23 25 30 % cells with aberrations 2.4 ± 1.7 2.8 ± 1.6 2.6 ± 1.8 2.6 ± 2.0 Breaks/cell 0.03 ± 0.03 0.03 ± 0.02 0.03 ± 0.02 0.03 ± 0.03 No significant differences. Fig. 4. Association between individual urinary M2/(M1 + M2) ratios and 4-month average BD exposure (ln[mg/m 3 ]) by GSTT1 genotype. Open circles and dashed line indicate null genotype individuals; closed circles and solid line indicate individuals with GSTT1+/+ or ± genotypes. Slopes are significantly different (p < 0.05) when a common intercept is assumed. lower rate of rise was statistically significant only for the GSTM1 null individuals. The same phenomenon was observed in the current study, although here it is the GSTT1 null genotype that showed statistical significance. Fig. 4 shows the relationship between after work M2/(M1 + M2) as a function of GSTT1 genotype. There is a much steeper rise in this ratio with increasing BD exposure in the GSTT1 positive individuals. Analysis of EH effects using combined genotypes (different combinations of EH113 and EH139) specifying high, intermediate, and low activity phenotypes showed no associations with any endpoint in the first Czech study. In the current study however, an effect on urinary metabolite excretion was observed. In this case, the rate of rise in M2 excretion was significantly higher for individuals (males and females combined, with M2 values for females adjusted for the gender difference) with low activity genotype combinations compared to individuals with high activity genotype combinations (Fig. 5). This is reflected in the rate of rise in the M2/(M1 + M2) ratio with increasing BD exposure levels, being significantly higher for individuals with the genotype combinations specifying low activity compared with individuals with genotype combinations specifying high or intermediate activity (Fig. 6). These observations are the mirrors of what occurs with GST null genotypes and are consistent with the activities of the different detoxification pathways for BD [reviewed in 4]. Small differences in M1 excretion would be difficult to see because of the greater baseline activity in the hydrolytic detoxification pathway. Table 11 Chromosome aberrations (FISH) control exposed control exposed N 26 23 25 28 a Aberrant chromosomes b 2.7 ± 1.6 2.9 ± 2.3 28. ± 2.8 2.5 ± 2.1 Translocations b 4.2 ± 3.0 4.0 ± 4.0 4.0 ± 3.9 4.0 ± 3.2 Reciprocal translocations b 1.4 ± 1.3 1.3 ± 1.3 1.5 ± 1.6 1.6 ± 1.4 Insertions b 0.04 ± 0.20 0.26 ± 0.54 0.12 ± 0.33 0.07 ± 0.26 Dicentrics b 0.08 ± 0.27 0.00 ± 0.00 0.04 ± 0.20 0.00 ± 0.00 Acentrics b 0.19 ± 0.49 0.43 ± 0.79 0.56 ± 0.96 0.29 ± 0.60 Co-junctions b 4.3 ± 3.0 4.5 ± 4.7 4.3 ± 4.5 4.1 ± 3.4 F G t/100 1.6 ± 1.1 1.5 ± 1.5 1.5 ± 1.4 1.5 ± 1.2 F G t/100 = F gr /2.5[f r (1 f r )+f g (1 f g ) f r f g ] where F gr is the translocation frequency measured by FISH after two color painting and f r and f g are the fractions of the genome painted red and green, respectively. a Two missing observations. b 0/00 denotes per 1000 metaphases.

R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 73 Fig. 5. Association between individual urine M2 concentrations (ln[ g/l]) and BD exposures (ln[mg/m 3 ]) by combination EH genotypes indicating high (triangles and long dashes), low (filled circle and solid line) and intermediate (open circle and short dashes) activity. M2 values for females have been adjusted for the sex difference. Slopes for low and high activity genotypes are significantly different (p < 0.05) when a common intercept is assumed. Fig. 6. Association between individual urine M2/M1 + M2 ratios and BD exposures (ln[mg/m 3 ]) by combination EH genotypes indicating high (triangles and long dashes), low (filled circle and solid line) and intermediate (open circle and short dashes) activity. Slope for low activity genotype is significantly different (p < 0.05) from the other two genotypes when a common intercept is assumed. 4. Discussion This second comprehensive molecular epidemiological study of BD exposed workers, undertaken to determine if there were any differences in biomarker responses between BD exposed females and males, revealed that there might be gender differences, but in an unanticipated direction. s excrete significantly lower concentrations of both measured urine metabolites (M1 and M2 together reflect both detoxification pathways) per unit of BD exposure. These results are from measurements made in after-work urine samples uncorrected for creatinine concentration. This correction, however, would have been inappropriate as the sexes have intrinsically different urine creatinine concentrations related to body mass rather than urine dilution or concentration [11 13]. The BD exposures used in these analyses is the 4-month (10 measurements) TWA mean for each subject rather than the individual same day measurement because, as seen in both the first Czech study and in the current one, results are similar regardless of which BD exposure value is used. The 4-month mean, however, contains less variability. We have concluded that, although BD exposures show day-to-day variability, the mean TWA concentration can be used for all analyses because it has the strongest correlation with the urinary metabolites as well as the hemoglobin adducts. Furthermore, our initial Czech study showed a greater than expected carry-over in urine metabolite concentrations from day to day, indicating that washout intervals are not as rapid as previously thought [2]. Using each worker s overall mean TWA BD exposure value also minimizes this source of variability. Our tentative conclusion from these female male differences in urine metabolite excretion levels is that females absorb LESS BD per unit of exposure than do males. The pending HB-Val and THB-Val Hb adduct determinations will either refute or confirm this conclusion. Overall, the BD exposure levels measured in this second Czech study are lower than those measured in the first. The male workers in this second study were drawn from the same polymerization population as in the first study, so there truly might have been a decline in BD exposure levels with time. However, BD measurement badges were analyzed by a different laboratory in the second study (EcoChem Company in Prague) than in the first (Health and Safety Laboratory, UK), so the difference may, in part, be technical as LODs appeared to be higher in the second study than in the first. Also, the BD exposure assessment period in the current study took place in mid-winter, compared to assessment during the spring in the first study. Time of year could conceivably be a factor in the lower measured BD exposure levels. Again, the results of THB-Val adduct concentrations, being determined by the same technique in the same laboratory in this study as in the first, with the addition of repeat determinations in cryo-preserved red blood cells maintained from the first study, should resolve this issue. In this second study, as in the first, the M1 and M2 urine metabolite concentrations reflected external BD

74 R.J. Albertini et al. / Chemico-Biological Interactions 166 (2007) 63 77 exposures. Metabolite concentrations were lower in this second study than in the first mirroring the lower BD exposure levels. Although both M1 and M2 concentrations were higher in the BD exposed females than in the control females, the differences were not significant. The exposed females had relatively low BD exposures. The higher M1 and M2 concentrations in exposed males compared to control males, however, were significant, as were the higher metabolite concentrations in males compared to females. Two factors appear to be responsible for this latter observation. Exposed male workers had significantly greater BD exposures than did the exposed female workers and females excreted less of both urine metabolites per unit of BD exposure than did males. The dominance of hydrolytic detoxification of BD in humans was again clearly demonstrated by the M1/(M1 + M2) ratios for both females and males. As in the first study, however, utilization of the conjugation pathway appeared to be stimulated by BD exposure. The complementary proportion, M2/(M1 + M2) (used to emphasize differences in the minor conjugation detoxification pathway) is greater in BD exposed males than in BD exposed females, again reflecting the greater BD exposures of the males. It is important to note that there was no difference in this ratio between females and males per unit of BD exposure, indicating that the relative use of the two detoxification pathways is similar in the two sexes. An extremely important finding in the current study was the lack of quantifiable pyr-val adduct in any of the 104 samples tested. The method used was sufficiently sensitive, however, to detect and quantify this adduct in mice and rats exposed to BD concentrations as low as 1.0 ppm [14,15]. As noted by Swenberg (in this issue), although the TWA BD exposures in the current study are below 1.0 ppm for both females and males, the duration of exposure was much greater for the humans than for the rodents, making the cumulative exposures quite comparable [15]. The pyr-val adduct is specific for the highly genotoxic (and carcinogenic) DEB metabolite of BD. Comparing the results of the current study with those of the rodent studies suggests that humans form at least 3 and 10 times less pyr-val than similarly exposed rats or mice, respectively [15]. Current efforts are being directed at developing a more sensitive assay for this critical adduct. When developed, it will be used to restudy all blood samples from this second Czech study, as well as all samples from the first. This second study again found absolutely no associations between HPRT gene mutations or chromosome aberrations (or SCEs) and BD exposures in any of the exposed workers. This finding is quite consistent with and to be expected from the non-detectable pyr-val adducts. The NOAEL of 1.794 mg/m 3 for these irreversible genotoxic effects, determined in the first study, remains the no effect level for these genotoxic endpoints. s were NOT shown to be more susceptible to these genotoxic effects than males although, as noted, their BD exposures were less. Our new results for males agree with those of the first Czech study, and are consistent, for both sexes, with those from a large, comprehensive Chinese molecular epidemiological study that found no association between BD exposures and mutations or chromosome changes in either females or males [16]. This current study also included a comprehensive evaluation of genotype effects in BD exposed workers. Genotypes did not significantly affect irreversible genotoxicity, as was shown in the first study. However, the influence of GST genotypes on metabolic detoxification was again seen in the present study where GSTT1 null individuals showed a significantly slower rise in the rate of M2 excretion (and presumably production) than did individuals with the other genotypes. In the first study, the significant association was with the GSTM1 null genotype, but both showed the rate alteration. In contrast to the first study, EH genotype combinations specifying high, intermediate, and low activity in the hydrolytic detoxification pathway also influenced urine metabolite excretion patterns in this second study. For EH, individuals with the genotype combinations specifying low activity showed a significantly faster rise in M2 excretion, reflected in a faster rise in M2/(M1 + M2) ratios with increasing BD exposure levels, than did individuals with the genotype combinations specifying high activity. This finding mirrors what was seen for the GST null individuals and presumably reflects greater detoxification via the conjugation pathway in individuals who have slower hydrolytic pathways. Again, the hemoglobin adduct concentrations will be very informative. All biomarkers studied in experimental systems using BD exposed animals have now been measured in human observational studies. It is, therefore, possible to compare the BD metabolic and genotoxicity profiles for rodents (mice and rats) with those for humans as a basis for developing a mechanistically based cancer risk assessment procedure. Such comparisons were not previously possible because of a paucity of human data primarily for exposure that could be related to key events in carcinogenesis, as reflected by biomarkers. The exposure, metabolic and genotoxicity results of the human Czech studies have been described. Analogous data sets are available from studies of BD exposed mice and rats [14,17 21, reviewed in 22]. While vastly