Study subjects. We recruited 63 male healthy workers, free of cancer and cardiopulmonary disease, who had been working in a steel production plant in Brescia, Italy, for at least 1 year. Individual written informed consent and approval from the local institutional review board were obtained before the study. All of the study subjects had a rotating weekly schedule based on 4 consecutive working days of 8 hr each, followed by 2 days of rest. The study subjects worked in 11 different areas within the plant, which were selected to provide a wide contrast of exposures between the study subjects. The exposure of each of the study subjects in the plant was monitored for the first 3 working days of a work week. On the fourth day, each subject donated a 20-mL peripheral blood sample. We used ethylene diamine tetraacetic acid (EDTA) tubes to collect 7 mL whole blood that was promptly (within 30 min from the blood drawing) centrifuged on site at 2,500 rpm for 15 min. The buffy coat (400 μL) was separated and transferred in a cryovial, immediately snap frozen in vapor phase of liquid nitrogen, shipped in nitrogen dry shippers to the laboratory, and kept stored in vapor phase of liquid nitrogen until histone extraction. A self-administered questionnaire was used to collect detailed information on lifestyle, drug use, medical conditions, body mass index (BMI), education, and residential history. Records from the factory administrative files were used to extract information on occupational history.
Exposure assessment. PM metal components (aluminum, manganese, nickel, zinc, arsenic, lead, iron) and PM mass [PM with aerodynamic diameters ≤ 10 μm and ≤ 1 μm (PM10 and PM1, respectively)] were measured in each of the 11 work areas of the steel production plant. We measured air concentrations of individual metal PM components in PM10, through multielement analysis performed by means of inductively coupled-plasma mass spectrometer (ELAN DRC II; PerkinElmer, Waltham, MA, USA) using the total quant method. External calibration was performed using calibration standard 3, stock multielement (10 μg/mL; PerkinElmer). PM10 and PM1 were measured using a Grimm 1100 light-scattering dust analyzer (Grimm Technologies, Inc., Douglasville, GA, USA).
Study subjects recorded in a personal log the time they spent in each of the work areas. Personal exposure was calculated as the average of work area levels weighted by the time spent in each area. All metal and PM levels were expressed in micrograms per cubic meter. In the plant, exposure levels have shown very little variability over time, as measures repeated over 3 years in a subset of the study population showed very high correlations (r2
> 0.90). Therefore, the time-weighted levels of metals and PM represented, in addition to the exposure during the week of the study, also a measure of the usual exposure of the study subjects (Tarantini et al. 2009
). We estimated cumulative exposures as the product of the time-weighted levels of metals and PM during the study by the years of employment in the plant.
Total histone extraction and histone modification analysis.
We obtained buffy coat from peripheral blood collected in EDTA tubes centrifuged at room temperature (22–25°C) for 15 min at 1,500 × g
. Red blood cell lysis solution (s.r.l. cod.A7933; Promega, Madison, WI, USA) was added to the buffy coat to wash out red blood cells. After 10 min at room temperature, the mixture was centrifuged at 2,500 × g
for 15 min, and the supernatant was discarded. Remaining monolayer cells were processed according to protocol used by Chen et al. (2006)
. Briefly, cells were lysed in 1 mL ice-cold radioimmunoprecipitation assay buffer (Santa Cruz Biotechnology, Santa Cruz, CA, USA) supplemented with a protease inhibitor mixture (Roche Applied Sciences, Indianapolis, IN, USA) for 10 min. The sample was then collected and centrifuged at 10,000 × g
for 10 min. After discarding supernatant, remaining pellet was resuspended in 0.4 N H2
. After incubation on ice for 90 min, the sample was centrifuged at 14,000 × g
for 15 min. The supernatant was mixed with cold acetone and kept at –20°C overnight. The histones were collected by centrifugation at 14,000 × g
for 15 min. After one wash with acetone, the histones were air dried and resuspended in 500 μL water. We measured total proteins in each sample by means of the Bradford assay according to manufacturer’s instructions (protein assay kit 500-0002; Bio-Rad Laboratories, Milan, Italy). We used equal amounts of proteins (4 μg) to normalize results of the subsequent analysis on histones.
We used a solid-phase sandwich enzyme-linked immunosorbent assay (ELISA), using monoclonal antibodies, to detect endogenous levels of H3K4me2 and H3K9ac (PathScan Sandwich ELISA Kits; Cell Signaling Technology, Beverly, MA, USA) according to the manufacturer’s protocol. The assays we used in our study are the PathScan dimethyl-histone H3 (Lys4) Sandwich ELISA Kit 7124 and PathScan acetyl-histone H3 (Lys9) Sandwich ELISA Kit 7121. The assays use dimethyl-histone H3 (Lys4) (C64G9) rabbit monoclonal antibody 9725 and acetyl-histone H3 (Lys9) antibody 9671, respectively, which have been shown by the manufacturer to be specific for the histone modifications of concern. Sample measurements were performed in duplicate. We used a Synergy HT-BioTek spectrophotometer (Winooski, VT, USA) to read 450 nm absorbance. The absorbance values at 450 nm directly reflected the concentration of modified histones (Deligezer et al. 2010
). According to the Beer–Lambert law, optical density (OD; absorbance) is used for colorimetric analysis so that readings relate directly to concentration. The coefficient of variation in replicate samples of the assays was 0.30% for H3K4me2 and 0.42% for H3K9ac.
Statistical analysis. We performed trend tests using the continuous variable in the regression model and presenting the corresponding two-sided p-values. We evaluated the association of PM metal components and PM mass levels with histone modifications using simple linear regression models, as well as multivariable models adjusting for age, BMI, pack-years, and percent granulocytes in the differential blood count as continuous variables and education as categorical variable. The independent variables used in multivariable models were selected a priori and included general characteristics potentially associated with cancer risks or other carcinogenic exposures. In addition, we adjusted for percent granulocytes to account for possible shifts in the proportion of leukocytes subtypes associated with the exposures. As sensitivity analyses, we fitted in the models as independent variables data from differential white blood counts (i.e., percent lymphocytes, monocytes, eosinophils, or basophils) or duration of sample storage and found no major changes in the results. Regression diagnostics were performed separately for each model. We examined whether the exposure–response relationships were linear through graphical inspection. Furthermore, we fitted a polynomial regression by including a quadratic term for exposure and compared these models with the linear model using the likelihood ratio test. Neither graphical inspection nor the likelihood ratio tests suggested any departure from linearity.
Outliers were excluded from regression analysis by dropping observations with studentized residuals that exceeded +3 or –3. The number of outliers removed ranged from a minimum of 0 to a maximum of 3.
Regression coefficients were computed with ordinary least squares estimators. To compare the magnitude of the associations of H3K4me2 and H3K9ac with different exposures, we calculated standardized regression coefficients and 95% confidence intervals (CIs) that express the change in histone modifications associated with an increase in exposure equal to the difference between the 90th and 10th percentile of the exposure distribution. We checked regression assumptions by performing diagnostic tests for each model, which included the Shapiro–Wilk test to verify normality of residuals and the White test to verify the homogeneity of variance of the residuals. A two-sided p-value < 0.05 was considered statistically significant. Statistical analyses were performed in SAS (version 9.1.3; SAS Institute Inc., Cary, NC, USA) and R (R Foundation for Statistical Computing, Vienna, Austria).