The study protocol has been described in detail elsewhere 21 . Briefly, 34 underground sewage workers and a control group of 30 office workers from the city of Paris were recruited on a weekly basis and over a 10 months period (July 2008-April 2009). All were male volunteers, current non-smokers since at least six months, aged 20-60 years, employed at the same function for at least six months with no history of chronic or recent illness. Interviews and biological sampling were conducted in the framework of regular occupational medical visits at the offices of occupational and preventive medicine of Paris municipality. The study was approved by the local ethical committees and has been conducted according to the Helsinki Declaration. A signed informed consent was obtained from each participant.

During three consecutive days of work shifts prior to the medical visit, workplaces indoor air concentrations for 13 PAHs and 12 VOCs were measured. Each subject collected 24 hours urine in a sterile plastic bottle in the last day of air sampling. During the medical visit that took place on Thursdays or Fridays, subjects filled in two self-administered questionnaires. One for socio-demographic factors, non occupational exposures (especially possible PAHs and VOCs exposures related to commuting, to area of residence and indoor sources), medical history, lifestyle (smoking history, including passive smoking, alcohol consumption and medications), usage of protection equipments and other putative confounders. The second for diet habits including barbecue usage and daily intake of fruits and vegetables.

Unless otherwise specified, all chemicals and culture media used were purchased from Sigma-Aldrich Chimie S.A.R.L (L'Isle d'Abeau Chesnes, France). For 8-oxodG, the 96-well kits were purchased from CliniSciences SA, Montrouge, France (origin: StressMarq Biosciences Inc., Victoria, BC Canada) 22 .

The sampling equipments were placed in a backpack and handed to one subject per work team (a team being generally composed of two to three subjects). The backpack contained a personal active sampling ChemPass pump (Rupprecht and Patastnick Co., Inc. NY, USA) with a calibrated flow rate of 4 L/minute, to measure PAHs. It was carried during the sewage activities or placed on a desk nearby the office workers. A VOCs passive sampling badge (Radiello code 145, Sigma-Aldrich, France) was also provided and attached to workers' clothes near the breathing zone.

PAHs in particulates and gaseous forms were collected on a quartz filter (Supelco 21038, 32 mm of diameter) and a cartridge containing polyurethane foam (Supelco ORBO 2-0600); respectively. The pump was equipped with a timer and subjects had to turn it on during work shifts and off when finished. The flow rate was measured before and after sampling and the sample was rejected if the difference was greater than 10%. For analysis, the filter and the foam were subjected to extraction by accelerated solvent (hexane-acetone; 50/50, v/v). The extract was concentrated in a water bath at 35°C using automatic evaporator (Turbo Vap II Zymark), then under nitrogen flow evaporator (N-Evap, MA, USA) until obtaining an oily drop. The drop was taken by 1 ml of acetonitrile solvent compatible high performance liquid chromatography by which the extracts were analyzed using a fluorescence emission detector (Waters 2475). The quantification analysis was carried out according to the calibrated response of standard solutions of marked-mixture for the selected PAHs. VOCs were analysed by coupling gas chromatography and mass spectrometry, as described previously 21 .

The workplace collection of the PAHs and VOCs were performed from Monday to Thursday for sewage workers and from Monday to Wednesday for office workers. For comparison with ambient air pollution, results from the Paris air quality monitoring network (AIRPARIF) were retrieved. The average concentrations at the same or at the nearest days of the study period, were obtained from the closest monitoring stations relative to the sampled underground workplaces, with each time two types of monitors: one measuring background air quality and the other close to traffic sources.

Thirteen PAHs were quantified [twelve listed as priority pollutants by the United States Environmental Protection Agency (U.S. EPA)] plus benzo(j)fluoranthene. Of these, four are in gaseous form (phenanthrene, anthracene, fluoranthene and pyrene) and the rest are particles. Twelve VOCs were also quantified.

Extraction of the urinary organic fraction was performed on a Sep Pak Vac C18 cartridge (Waters, Saint-Quentin en Yvelines) 21 . Briefly, the volume of the 24 hours urine was measured and a sample of 150 ml was immediately coded and frozen at -20 C° for each subject. For the assays, samples were then thawed and 50 ml were centrifuged at 3000 t/minute for 5 minutes. The supernatant (40 ml) was collected in a sterile tube. The organic fraction was extracted on a Column Sep Pak^® Vac C18 Cartridges on an aspirated tray (J.T. Baker spe -12G). The cartridge was washed 2 × 3 ml of absolute methanol, then 3 × 3 ml of ultra-pure water. It was then loaded with the 40 ml urine. The column was washed 2 × 3 ml of ultra-pure water and the adsorbed organics were eluted with 3 × 3 ml of absolute acetone (Carlo, CAS 67-64-1). The eluate was evaporated at 45°C under nitrogen stream (Tech Lab Faster-Chemfree, France) until complete dryness. The residue was suspended in 500 μl DMSO and stored at -20°C. All operations were done at room temperature.

HepG2 cells (ATCC, catalog number HB-8065) were routinely cultured in the laboratory in monolayers. They were maintained in EMEM (Eagle's minimum essential medium, 9.6 g/L) supplemented with 10% FBS (Fetal bovine serum) and 1% antibiotics solution (penicillin 10000 U/ml; streptomycin 10 mg/ml), sodium carbonate (2.2 g/L), Hepes (5.96 g/L), and sodium pyruvate (0.11 g/L). The pH of cultured mediums was maintained between (7.2-7.4). Cells were grown at 37 C° in a humidified 5% CO2 and 95% air atmosphere. HepG2 cells were cultivated for 24 hours in 5 ml complete culture medium supplemented with 50 μl of organic urine extract. After harvesting and centrifugation, the cell pellet was diluted by the culture medium to 5 × 10⁵ cell/ml. Trypan blue exclusion test was used to assess toxicity, and cell viability was always ≥ 95%.

Comet assay was performed essentially as in Singh et al. 23 , with some modifications as in Muller-Pillet et al. 24 . Briefly, after layering the cells on conventional microscopic slides previously sprayed by normal melting point agarose, they were lysed at 4°C in the dark for at least 1 hour (2.5 M NaCL, 100 mM Na2EDTA, 10 mM Tris and 1% Triton X-100, Prolabo 28.817.295 and 10% DMSO, pH = 10). The electrophoresis was conducted (20 V, 0.62 V/cm, 300 mA, 20 minutes) in an electrophoresis gel system (EC340, Maxicell^® Primo, Holbrook, New York) filled with electrophoresis buffer (300 mM NaOH, 1 mM Na2EDTA, pH 13). B(a)P 40 μM and DMSO 1% final concentrations (positive and negative controls; respectively) were added with each electrophoretic run. After neutralizing the slides in 0.4 M Tris buffer, they were stained with 40 μl ethidium bromide (2 μg/ml in ultra-pure water). 50 cells (two slides/subject and 25 cells/slide) were examined for DNA migration using an Olympus BX40 fluorescence microscope (Olympus, Tokyo, Japan). We quantified DNA damage by a computerized image analysis system (Komet 4.02 software, Kinetic Imaging, UK) in evaluating the percent DNA tail parameter. The gel was scanned in a systematic way and the comets represented the whole gel. Edges and areas around air bubbles were avoided. Clouds images (indicative of dead cells) were also excluded from acquisition and analysis.

For micronucleus assay, we adapted the assay described by Fenech 25 . Briefly, at time = 0, HepG2 cells were grown in 10 ml complete medium for 24 hours. Then, in a freshly 10 ml diluted medium supplemented with 100 μl of organic urine extract (time = 24 hours). Subsequently, at time = 44 hour, the cells were cultivated in a fresh complete culture medium supplemented with 3 μg/ml cytochalasin-B final concentration. At time = 72 hour, cells were rinsed in cold hypotonic KCL solution (0.075 M, Prolabo 26.764.298) and then fixed in CARNOY solution (methanol, Carlo Erba 525.102: acetic acid, Prolabo 20.104.298, 3:1 v:v). The air-dried slides were stained with 40 μg/ml Acridine orange solution in dark. The slides were examined at 200-fold magnification by an Olympus BX fluorescence microscope (Olympus, Tokyo, Japan). We evaluated the frequency of the micronuclei (MNi) formation in 1000 binucleated cells (BNed)/slide/subject. We respected Fenech 25 criteria in scoring the MNi. Proliferation and cytotoxicity were assessed by calculating the nuclear division index (NDI) on 150 viable cells according to Fenech 25 formula. NDI mean difference between exposed and non-exposed and between positive and negative controls (B(a)P 40 μM and DMSO 1% final concentrations; respectively) was always acceptable and below 25% 26 .

The urine aliquots (1 ml each) were kept at -20°C. They were thawed at room temperature immediately before analysis. 8-oxodG was measured with a competitive enzymatic immunoassay (EIA) kit. This assay utilizes a specific 8-oxodG monoclonal anti-body (Catalog# SKC-120A), 8-oxodG-acetylcholinesterase (AChE) conjugate and an anti-mouse IgG-coated plate. The manufacturer provided the protocol of analysis 22 . To insure the accuracy and reproducibility of results, each sample was assayed at two dilutions (1/300 and 1/400) and each at duplicates. The 24 hours 8-oxodG excretion was expressed relative to body weight (pmole/kg/24 h) 27 28 . Creatinine in 24 hours urine was determined photometrically as picrate, according to Jaffé method 29 .

PAHs concentrations are presented as mean (range) while VOCs are exhibited as mean ± SE. For analyses of associations between biomarkers and exposure, the workplace air concentrations (26 measurements) were assigned to each subject within a team as his exposure level. Chi square and Fisher exact tests were used to analyse the differences between categorical variables. ANOVA was used to compare the mean differences of quantitative variables, and to evaluate the effect of some socio-demographic and putative confounding factors on comet and micronucleus results. Multiple linear regression was conducted to evaluate the level of 8-oxodG among the two study groups while adjusting for possible confounding factors. Multiple linear regression were also implemented to assess the association between exposure to occupational agents and the two genotoxicity assays while adjusting for confounding variables. Only pollutants significantly associated (P < 0.05) in univariate analysis were tested, while confounders were retained if P < 0.1 in univariate analyses. Effect modification was tested, in particular by age whose distribution was partitioned by the median value into two categories (≤ 39 and < 39 years). SPSS 16 software was used for analysis 30 .

To assess and characterize the carcinogenic potency of PAH mixtures, toxicity equivalent factors (TEFs) were used to convert PAH exposure into an estimated B(a)P equivalent. We used the Nisbet and Lagoy 31 TEFs with the exception for benzo(j)fluoranthene where the TEFs proposed by Collins et al. 32 was used. Cancer risk due to inhalation of PAH mixtures was then calculated using the B(a)P inhalation risk cancer unit of 1.1 × 10^-6 (ng/m³)^-1 proposed by the U.S. EPA 33 . We also used the benzene unit risk range of 2.2 × 10^-6 to 7.8 × 10^-6 (μg/m³)^-1 34 to assess the cancer risk by inhalation of benzene.

General characteristics of the study population are summarized in Table 1. The mean ages were 35.9 (standard deviation; SD = 7.5) and 43.3 (SD = 8.2) years in sewage (n = 34) and office (n = 30) workers respectively (P < 0.001). The mean ± SD number of working years in sewage was 7.05 ± 6.9 years. Never smokers were more frequent in sewage than in office workers (P = 0.03). However, sewage workers were more likely to drink alcohol regularly than the control group (P = 0.01), and less likely to eat vegetables than the office workers (P = 0.01). The 24 hours urinary creatinine differed between the two groups (P = 0.003) but was within the normal human male values for both groups. The level of education also distinguished the study groups (P = 0.04). No other difference was seen regarding factors that might influence study-relevant exposures: environmental tobacco smoke (P = 0.87), type of heating system used at home (individual or collective, P = 0.52), declared proximity of homes to industrial installations (P = 0.82) or consumption of barbecue grilled food.

The mean workplace exposure levels of each PAH compound presented in Table 2 were significantly higher among sewage workers compared to office workers and to ambient air concentrations (23 measurements) (P < 0.01). In general, phenanthrene, fluoranthene and pyrene contribute the largest portion of the total PAHs exposure (e.g. among sewage workers they ranged from 43%, 15%, and 12%; respectively). The other PAHs amount to less than 5%, except anthracene after the traffic monitor ambient air measurements (11.5%). The highest B(a)P concentration was found among sewage workers with a range of 0.5 to 62.1 ng/m³ and a mean value of 7 ng/m³.

Based on these data, single PAHs concentrations were converted into total B(a)P equivalent concentration ([B(a)P]eq) using the TEFs (see statistical analysis) and translated into lifetime cancer risk estimates. The average sewage workers' total [B(a)P]eq exposure value (13.66 ng/m³) is more than 10 times greater than those encountered for office workers, traffic and urban ambient air levels (respectively 1.15 ng/m³, 1.08 ng/m³, and 0.73 ng/m³). Table 2 also shows the associated lifetime cancer risks. The PAHs cancer risk level for sewage workers is 1.5 × 10^-5 (0.13 × 10^-5 for office workers).

Figure 1 presents the mean of each VOC concentrations in the workplace air of sewage and office workers, and in the air of urban and traffic environments nearest to the study locations. A high heterogeneity was observed between the different locations with the highest values found in the sewage workplace (P < 0.01), followed by indoor air of office workers, traffic and urban background ambient air, respectively. Benzene (mean ± SE) concentrations were 19.1 ± 2.9 and 4.1 ± 0.53 μg/m³ among sewage and office workers, respectively; corresponding values were 3.7 ± 0.13 and 1.0 ± 0.09 μg/m³ in traffic and urban background, respectively. Using the Integrated Risk Information System (IRIS) unit risk estimate range 34 , the benzene associated lifetime excess cancer risk for sewage workers ranged from 4.2 × 10^-5 to 14.9 × 10^-5 (0.9 × 10^-5 to 3.2 × 10^-5 for office workers).

The mean percent DNA tail and MNi/1000 BNed among sewage workers was statistically higher than in office workers [mean ± SD of percent DNA tail = 8.07 ± 3.12 and 2.70 ± 0.58, and mean ± SD of MNi/1000 BNed = 38.02 ± 7.16 and 28.30 ± 3.74, respectively in the two populations [P < 0.001 in both tests] (Figure 2A and 2B).

In multivariate linear regression models, we tested the differences in percent DNA tail and MNi/1000 BNed between the two study occupation groups while adjusting for possible confounders including 24 hours urinary creatinine. The differences between sewage and office worker are, (point estimate and [95% confidence interval]), 5.01 [3.01-7.00] for percent DNA tail and 9.41 [4.47-14.36] for MNi/1000 BNed, respectively. No interaction was observed according to subjects' characteristics.

The percent DNA tail was positively associated with age and educational level with a borderline positive effect of alcohol consumption (P = 0.09) and a protective effect of vegetable intake (P = 0.05). MNi/1000 BNed was positively associated with age and alcohol consumption while inversely with vegetable intake. Table 3 also presents the results of the association of percent DNA tail and MNi/1000 BNed with each individual VOC and PAH.

Results of the multiple linear regression models are presented in Table 4 that exhibits the association between each of the in vitro genotoxicity assays and the assigned personal atmospheric exposure variables, controlling for covariates. Effect modification is also accounted for, with age being the sole factor influencing this association. A significantly positive association with the comet test response was seen among older workers only (> 39 years; n = 34) for nine PAHs (the four gaseous PAHs and five out of the nine particulate PAHs). This association with PAHs was not found for the MNi/1000 BNed. All VOCs were significantly associated with MNi/1000 BNed among older workers, while percent DNA tail was only influenced by exposure to benzene, ethylbenzene, m+p-xylene, o-xylene, decane, tri and tetra-chloroethylene. Noteworthy is that no difference in 24 hours urinary volume or creatinine levels between the two age groups was observed (P = 0.91 and 0.20; respectively).

Figure 3 shows the box and whisker plots of 24 hours urinary 8-oxodG in sewage and office workers. There was a slightly (but not significantly) higher mean level in sewage compared to office workers (mean ± SD, 8.26 ± 4.26 pmole/kg 24 h and 7.22 ± 3.32 pmole/kg 24 h respectively, P = 0.28). There was no significant difference in 8-oxodG level regarding other exposure factors mentioned in Table 1. No clear association could be found between the workplace concentrations of the measured pollutants or of total [B(a)P]eq and the level of 24 hours urinary 8-oxodG (data not shown).