Treatment of cell lines with BaP.
Jurkat T cells were maintained in RPMI media (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS). Two lung adenocarcinoma cell lines, A549 and H1793 were maintained in F12K (Invitrogen) and Dulbecco’s modified Eagle medium (DMEM)/F12, respectively. Both media were supplemented with 10% FBS (Hyclone, Logan, UT). For H1793 cultures, the DMEM/F12 medium was supplemented with 10 nM hydrocortisone, estradiol-17β (Sigma, St. Louis, MO), 1X insulin/transferrin/selenite, and 4.5 mM l
-glutamine (Invitrogen). For BaP treatment, cells were exposed to 0.1, 1.0, 10, or 100 nM BaP (Sigma) or with DMSO alone as control, every other day for 4 days. Total RNA and DNA were isolated as previously described (Perera et al. 2009
Real-time reverse-transcriptase-PCR (RT-PCR).
Total RNA was reverse transcribed and the transcript levels quantified by SYBR Green-based RT-PCR as previously described (Perera et al. 2009
). The following primers were used for the quantification of transcripts of β-actin
(X00351.1): Fwd-5´-GCGGGAAATCGTGCGTGACATT-3´ and Rev-5´-GATGGAGTTGAAGGTAGTTTCGTG-3´; IFN
γ (NM_000619.2): Fwd-5´-TTTGGGTTCTCTTGGCTGTT-3´ and Rev-5´-CTGTCACTCTCCTCTTTCCAA-3´); and IL4
(NM_000589.2 and NM_172348.1): Fwd-5´-TGAACAGCCTCACAGAGCAG-3´ and Rev-5´-CTCTGGTTGGCTTCCTTCAC-3´. Mean transcript levels in cell cultures were obtained from at least three separate experiments. The 2-ΔΔCt method was used to calculate the relative expression levels of a transcript by normalization to the level of β-actin mRNA. Values in vehicle-treated (BaP: 0 nM) cells were assigned arbitrarily an abundance value of 1.00 for each gene and values from other treatment groups were compared with values in the vehicle-treated cultures.
Bisulfite genomic sequencing.
Bisulfite genomic sequencing was conducted according to previously described protocols (Perera et al. 2009
). Genomic DNA was bisulfite-modified before PCR. In silico
analyses were used to predict the 5´-CGI(s) of IFN
γ (NC_000012:c66843688-66838788) () and IL4
[NC_000005.9:132007373-132010373; see Supplemental Material, Figure S1
)]. Primers were designed to amplify a 365 bp (–2082 to –1718) fragment upstream of TSS encompassing the predicted 5´-CGI(s) of IFN
γ (Region 1: BS-IFNg-F1: 5´-GAGGGGAAAAAGAATTTAAGATTAAAG-3´ and BS-IFNg-R1:5´-ATCACCCAAACTAAAATACAATAAC-3´) and a 551 bp (–329 to +222) fragment including 6 CpG dinucleotides flanking the TSS of the gene (Region 2: BS-IFNg-F2: 5´-TGTGAATGAAGAGTTAATATTTTATTA-3´ and BS-IFNg-R2: 5´-TTCTACTTCTTTTACATATAAATCCTAACA-3´) from bisulfite-modified DNA. Similarly, a pair of primers was used to amplify a 337 bp (–1388 to –1052) fragment of the predicted 5´-CGI of IL4
(BS-IL4-F:5´-GGGAAGTGGAATAGAGGTAAAATTT-3´ and BS-IL4-R:5´-ATCACCCAAACTAAAATACAATAAC-3´). Six clones were picked from individual UCWBC samples, and a total of 12 clones were picked from three individual sets of cell line samples and sequenced (Macrogen, Rockville, MD).
Figure 1 CG content (%) in the 5’ flanking region of IFNγ and the location of primers used in this study. The CGI(s) (shaded in gray) were identified in silico based on a CG content of 60%, with an observed:expected ratio of 0.6 according to instructions (more ...)
Cell cytotoxicity assay. Jurkat T cells, A549, and H1793 were treated with BaP (0, 0.1, 1, 10, or 100 nM), or with DMSO as control, every 2 days for a total of 4 days. Triplicate experiments were performed. Percentages of viable lung cancer cells following BaP exposure were determined by cell proliferation assay using MTS solution (Promega, Madison, WI). For Jurkat T cells, the numbers of viable cells present in treated or untreated samples were counted after staining cells with Tryptan Blue (Invitrogen).
A total of 727 nonsmoking African-American and Dominican women and their children, who lived in a high-traffic area of Manhattan and South Bronx were enrolled in the CCCEH cohort study as previously described (Miller et al. 2004
; Perera et al. 2003
; Perzanowski et al. 2006
). Written informed consent was obtained from all subjects following procedures approved by the Institutional Review Board of the New York–Presbyterian Medical Center. Completion of prenatal air monitoring for the mothers and collection of umbilical cord blood sample at delivery were required. We used the same study sample described previously (Perera et al. 2009
). Briefly, 53 cohort participants were dichotomized on the median PAH levels of the CCCEH cohort (2.27 ng/m3
), including 27 selected from those above the median (high PAH exposure) and 26 selected from those below the median (low exposure). This sample was representative of the cohort population regarding maternal age, child sex, median maternal PAH exposure, and percentage of children with probable asthma, but differed with respect to ethnicity (full cohort, n
= 606, 63.5% Dominican and 36.5% African-American; study sample, n
= 53, 49% Dominican and 51% African-American).
Monitoring and sample collection.
Maternal PAH exposure was assessed from personal prenatal air monitoring during the third trimester of pregnancy as previously described (Perera et al. 2003
). In brief, total PAH exposure levels for each mother were calculated as the sum of eight nonvolatile (molecular weight 228–278: benzo[a
]anthracene, chrysene/iso-chrysene, benzo[b
]fluoranthene, BaP, indeno[1,2,3-c,d
]perylene) carcinogenic PAHs collected as part of total dust on the filters of the personal monitors. Samples with total PAH < 0.125 ng/m3
were considered below the limit of detection. Umbilical cord blood (30–60 mL) was collected at delivery. The buffy coat, packed red blood cells, and plasma samples were separated and stored at –80°C. DNA (100–500 ng) was extracted from the buffy coat. All samples were de-identified.
Statistical methods. The distributions of characteristics of 53 participants were compared after stratification at the median level of maternal PAH exposure levels, 2.27 ng/m3. Pearson’s chi-square test of independence was used for comparing exposure levels with respect to ethnicity (percent African American vs. Dominican), sex (percent male), maternal exposure to tobacco smoke during pregnancy (ETS) (percent yes), and receipt of public assistance (percent yes). Differences between medians of age at delivery (years) were compared by quantile regression. Based on the median as an approximation to the center of subsets of distributions of percent methylation, quantile regressions were performed to estimate differences between the medians of percent methylation for the specified categories of each characteristic. We included receipt of public assistance as an indicator of family socioeconomic status (SES) because it was more complete (no missing data) and considered to be more objective than some of the other available measures of SES. The statistical analyses focused on estimating the association between maternal PAH exposure and DNA percent methylation using multiple regression analyses, adjusted for participant characteristics. Because of the skewness of the distribution of PAH, the loge-transformation (ln-PAH) was applied to improve symmetry.
Generalized additive models (GAMs) were analyzed in which smooth plots of region-specific curves relating percent methylation of the IFNγ promoter to ln-PAH and age at delivery were drawn. General regressions were analyzed where ln-PAH and age at delivery were modeled as restricted cubic spline functions with turning points (knots), identified by the GAMs. A (restricted) cubic spline function consists of piecewise cubic polynomial functions relating a dependent and independent variable. The parameters of the polynomial functions depend on the number and location of specified knots. In addition to an intercept term, the number of parameters required is one less than the number of knots, because the polynomials are constrained to be linear in the tails. Thus, a restricted cubic spline function with three knots will require two parameters. The first parameter describes the effect of the linear component of the independent variable; the second parameter describes the effect of the nonlinear component. The linear and nonlinear components of the function should not be separated in interpreting the trend that is graphically displayed.
Regressions were adjusted for participant characteristics, ethnicity, sex, ETS exposure, and receipt of public assistance as dichotomous variables modeled dichotomously, and loge
-percent (ln-percent) methylation was the dependent variable (Harrell 2001
). This transformation improved the fit of the full and reduced models, compared with the untransformed analyses, of each region, as determined by the Aikaike Information Criterion. Restricted cubic splines for ln-PAH were constructed with knots at the 10th and 90th percentiles and a third knot at 3.5 ng/m3
(ln-PAH = 1.3). Splines for age at delivery varied according to IFNγ region (at the 10th and 90th percentiles and at third knot at 25 for region 1; at the 5th and 95th percentiles for region 2.) Reduced models included covariates that predicted the outcome with p
< 0.20. The significance of the linear and nonlinear terms of the restricted cubic spline functions were evaluated by the likelihood ratio test statistic (LRT), where subsequent models were reduced by testing nonlinear terms of the spline function first. Spline terms were retained if p
< 0.20. The analyses were performed using SAS for Windows, version 9.2 (SAS Institute Inc., Cary, NC). Graphs were generated using S-Plus software, S-Plus 2000 (TIBCO Software Inc., Palo Alto, CA). An alpha level of 0.05 indicates statistical significance.