Abstract
DNA methylation is one of the epigenetic mechanisms that regulates gene expression, chromosome structure, and stability. Our objective was to determine whether the DNA methylation system could be a target following in utero and postnatal exposure to human blood contaminants. Pregnant rats were dosed daily from gestation day 1 until postnatal day 21 with 2 dose levels of either organochlorine pesticides (OCP; 0.019 or 1.9 mg/kg/day), methylmercury chloride (MeHg; 0.02 or 2 mg/kg/day), polychlorinated biphenyls (PCBs; 0.011 or 1.1 mg/kg/day), or a mixture (Mix; 0.05, or 5 mg/kg/day) including all 3 groups of chemicals. Livers from 1 female offspring per litter were collected at postnatal day 29. Hepatic analysis revealed that the mRNA abundance for DNA methyltransferase (DNMT)-1, -3a, and -3b were significantly reduced by the high dose of PCB, that the high dose of MeHg also reduced mRNA levels for DNMT-1, and -3b, but that OCP had no significant effects compared with control. The high dose of PCB and Mix reduced the abundance of the universal methyl donor S-adenosylmethionine, and Mix also reduced global genome DNA methylation (5-methyl-deoxycytidine/5-methyl-deoxycytidine + deoxycytidine). The latter is consistent with pyrosequencing methylation analysis, revealing that the high-dose groups (except OCP) generally decreased the methylation of CpG sites (position -63 to -29) in the promoter of the tumor suppressor gene p16INK4a. Overall, these hepatic results suggest that the DNA methylation system can be affected by exposure to high doses of blood contaminants, and that OCP is the least potent chemical group from the investigated mixtures.
DNA methylation is a normal epigenetic mechanism involved in controlling DNA structure, chromosome stability, mobility of “parasitic” elements (transposons, retrotransposons, viral elements), gene imprinting, and gene expression. Abnormal DNA methylation is associated with a number of cancers, infertility, and developmental, neurological, immunological, and age-related disorders. 1 There is growing evidence that suggests that early exposure to environmental agents might lead to long-term and transgenerational effects through epigenetic mechanisms involving DNA methylation. 2–4 Global genome methylation, site-specific CpG dinucleotide methylation, and different elements of the DNA methylation system have all been shown to be affected by drugs, 5–7 radiation, 8 environmental estrogens and contaminants, 9–12 disinfection by-products, 13 and metals such as arsenic, 14,15 chromium, 16,17 nickel, 18,19 and cadmium. 20 Abnormal DNA methylation, observed in a number of diseases and cancers, 21,22 may originate from dietary insufficiency (choline, methionine, vitamin B-12, and folic acid) or disorders of the 1-carbon metabolism pathway leading to a reduced availability of the universal methyl donor, S-adenosyl-L-methionine (SAM). 23 The DNA methylation machinery includes numerous methyl-binding proteins, DNA demethylases, 24 and families of DNA methyltransferases (DNMT-1, -2, -3) with multiple variants. 25 The DNMTs catalyze the transfer of a methyl group from SAM mostly to cytosine located next to a guanine (5′-CpG-3′). 2 During mitosis, the methylation pattern of the original DNA strand is copied mostly by the constitutively expressed DNA methyltransferase-1 (DNMT1) onto the replicating DNA. The de novo methyltransferases DNMT3a and 3b catalyze the transfer of a methyl group to previously unmethylated DNA during gameto- and embryogenesis. However, DNMTs often work cooperatively, and the normally low expression of DNMT3b in somatic cells is increased in cancer to cooperate with DNMT1 in silencing tumor suppressor genes. 26 An understanding of the effects of environmental contaminants on the DNA methylation machinery and its biological consequences should open avenues for better differentiating transient effects from adaptive and persistent effects, and for the development of diagnostic tools and screening methods.
Polychlorinated biphenyls (PCBs) and some organochlorine pesticides (OCPs) are legacy chemicals that were banned in the 1970s, and their levels in the environment are declining. Exposure to these chemicals is still of concern to health agencies given that they are still present in human tissues and in the human diet, particularly that of circumpolar populations who consume traditional food. Breast milk and cord blood levels of PCBs, p,p′-dichlorodiphenyldichloroethene (DDE), and other organochlorines are higher in the Inuit of North-Eastern Canada than many other Canadian and non-Canadian populations. 27,28 Recent papers reviewed possible associations between exposure to these chemicals and cancer, 29,30 infertility, 31 developmental anomalies, 32 endocrine disruption, 33,34 immunological function, 35 and neurological impairments. 36,37 Methylmercury (MeHg) is another contaminant of concern in Canada, given that it is substantially higher in the Canadian Arctic marine biota compared to other circumpolar countries, 38 and that fish contaminated with MeHg are components of the Inuit diet. Methylmercury accumulates in the fetus in utero, but in contrast to organochlorines, the levels of MeHg in breast milk are low. 39,40 The fetus is more sensitive than the adult to MeHg, as demonstrated by the fact that high levels of exposure result in severe neurological impairments in babies born from asymptomatic mothers, 41 whereas at low levels of exposure, subtle cognitive and language delays are observed in older children. 42,43
There is little information about the effects on human health of chronic exposure to mixtures of contaminants during the fetal and postnatal periods. The pre- and postnatal phases are critical developmental stages because tissue differentiation, organ growth, the detoxification system, and blood barriers have not yet matured, and abnormal events during these periods can lead to long-term adverse health effects. We have shown that postnatal exposure to a high dose of a reconstituted mixture of arylhydrocarbon receptor (AhR) agonists (non-ortho PCB, polychlorinated-dibenzodioxins [PCDD], and dibenzofurans [PCDF]), prepared based on their proportion found in breast milk, decreased the mRNA expression of DNMT1 in the liver and the brain of female pups at postnatal day (PND) 21. 44 Bowers et al developed a mixture of PCBs, OCPs, and MeHg based on the chemical profiles in Canadian Arctic Inuit maternal blood, and they demonstrated that exposure of rats to this mixture during gestation and lactation produces a blood chemical profile comparable to that of humans in rodent dams and the pups. 45 When this PCB-OCP-MeHg mixture was tested at a dose that produces a serum level in the rat pups 100-fold higher than human levels, Chu et al reported a decrease in the growth rate of rat pups, clinical, biochemical, and histopathological effects. 46 Subsequent work with this mixture is currently studying adverse effects that could be attributed to the PCB, the OCP, or the MeHg component, and investigating developmental, systemic, neurotoxic, and genomic effects (only the latter has been published at this stage). 47 As part of this initiative, the current manuscript reports effects of this mixture and its chemical family components (PCB, OCP, MeHg) on elements of the DNA methylation system in the liver, including the abundance of SAM and DNMT mRNAs and global genome and gene-specific changes in DNA methylation. Genespecific methylation changes will be addressed by bisulfite pyrosequencing analysis of 5 CpG sites in the promoter of the gene cyclin-dependent kinase inhibitor 2A (Cdkn2a) that express the tumor suppressor protein p16INK4a. This gene is a known DNMT methylation target for which the expression is reported to be regulated by DNA methylation during carcinogenesis and following exposure to various environmental agents. 14,17,20,48,49 Finally, the sensitivity of these end points as an indicator of exposure to xenobiotics will be compared with the changes in the abundance of cytochrome P450 1A1 (CYP1A1) mRNA. A phase I metabolism enzyme, CYP1A1 is a sensitive indicator of exposure to AhR agonists. Its expression is regulated through the classical AhR-dioxin response element pathway. 50
Methods
Animal Treatment
Animal treatment and housing conditions were approved by the Animal Care Committee of Health Canada and conformed to the Guidelines of the Canadian Council on Animal Care. Housing, handling and breeding conditions, origin of test chemicals (analytical grade, 99% pure), and methods for the preparation and validation of the mixture were all as previously described. 46 Female Sprague Dawley rats (200–230 g) were mated with males (320–350 g) purchased from Charles River Laboratories (St. Constant, Québec). The day of detection of a vaginal plug was denoted as gestation day 0 (GD 0). All pregnant dams were dosed daily from GD 1 to postnatal day (PND) 21 by providing them with a Teddy Graham cookie (Nabisco Ltd., Toronto, ON) laced with a measured dose (adjusted for body weight every day) of the mixture dissolved in corn oil. The mixtures are described in Table 1. Pregnant dams were exposed to either corn oil as control (CTRL), a low dose (L) or high dose (H) of OCP (0.019 or 1.9 mg/kg/day), methylmercury chloride (MeHg, 0.02 or 2 mg/kg/day), PCB (0.011 or 1.1 mg/kg/day), or a complete mixture including all these chemicals (Mix: 0.05, or 5 mg/kg/day). The dams were monitored for parturition beginning on GD 18, and litters were culled to 4 males and 4 females on PND4. One female pup from each litter was sacrificed by decapitation at PND29, providing the following number of investigated pups per treatment group: 11 CTRL, 10 OCP-L, 10 OCPH, 10 PCB-L, 11 PCB-H, 9 MeHg-L, 10 MeHg-H, 11 Mix-L, and 9 Mix-H. At PND99, liver samples were also collected from older siblings treated with the high doses (11 CTRL, 10 OCP-H, 10 PCB-H, 10 MeHg-H, and 9 Mix-H). Male pups were not investigated, given that PCBs and AhR-agonists are classical hepatocarcinogens mostly in females. 51
Measurement of SAM
S-Adenosyl methionine is a universal methyl donor involved in various methylation reactions, including DNA methylation. The concentration of SAM was quantified from liver samples using a commercially available fluorescence assay kit (Mediomics LLC, St. Louis, MO, USA). Samples (0.5 to 1 g) were homogenized in 2 volumes (v/g) of 50 mM NH4HCO3 and clarified by centrifugation (11 750 × g, 30 min, 4°C). The supernatant was diluted with cold absolute (100%) ethanol to a final concentration of 70% and maintained in an ice bath for 30 minutes prior to being clarified by centrifugation (11 750 × g, 30 min, 4°C). The supernatant was lyophilized and resuspended in cold buffer S (20 mM Hepes, pH 8.0, 100 mM NaCl and 0.5 mM 2-mercaptoethanol) prior to a final clarification by centrifugation (11 750 × g, 15 min, 4°C). S-Adenosyl methionine concentrations were measured in this supernatant at 3 dilution levels using a fluorescence microplate reader (excitation 485 nm, emission 665 nm) and a 9-point standard curve (0–50 μM SAM), as described by the manufacturer’s protocol (Mediomics LLC, St. Louis, MO, USA). All samples were measured in a single assay. The intra-assay coefficient of variation calculated over the range of the standards was 10.4%.
Measurement of 5-Methyl-deoxycytidine as an Indicator of Global Genome DNA Methylation
5-Methyl-deoxycytidine (5mdC) is measured as an indicator of the level of DNA methylation in the genome. It is expressed as a percentage ratio of 5mdC/(5mdC + dC), in which the measurement of dC also provides a correction for the amount of DNA in the sample.
Total DNA was prepared from liver samples using the Qiagen DNAeasy kit (Qiagen, Mississauga, Ontario, Canada). The concentration of purified DNA was measured for each sample using a spectrophotometer at 260 nm. For each sample, 10 μg DNA (usually in 50 μL) was first denatured for 3 minutes in a 100°C water bath and then chilled immediately in ice water. After adding a 1/10 volume of ammonium acetate buffer (0.1 M, pH 5.3) containing 20 units of nuclease P1 (Sigma Chemical Co., Canada), the mixture was incubated at 45°C for 2 hours in an autoshaker (Thermomixer-R, Eppendorf AG). A 1/5 volume of ammonium bicarbonate buffer (0.5 M, pH 8.0) including 0.02 unit of snake venom phosphodiesterase I (Sigma Chemical Co.) was added and incubated at 37°C for 2 hours. Then, 5 units of shrimp alkaline phosphatase (1.23 units/μL glycerol solution from Sigma Chemical Co.) was added, and the mixture was incubated at 37°C for 1 hour. Finally, the mixture was incubated at 65°C to inactivate the shrimp alkaline phosphatase, then the samples were spun at 16 000 × g for 10 minutes, and the supernatant was stored at −20°C until analysis by liquid chromatography–ultraviolet electrospray ionization–mass spectrometry (LC-UV/ESI-MS). To control for DNA hydrolysis and analytical variability, aliquots of pooled rat genomic DNA were prepared and included in each batch of samples.
Ten nucleotides (cytosine [C], deoxycytidine [dC], 5-methyl-cytosine [5mC], 5-methyl-deoxycytidine [5mdC], uracil [U], adenine [A], deoxyadenosine [dA], guanine [G], deoxyguanosine [dG], and thymine [T]) were quantified using a LC-UV/ESI-MS Agilent 1100 system (Agilent Technologies, Palo Alto, CA, USA), with a Waters Atlantis dC 18 column (3 μm, 150 × 2.1 mm; Waters, Milford, MA, USA) at 30°C. Liquid chromatography separation was performed at a flow rate of 220 μL/min using a solution of 0.1% (v/v) formic acid and a linear gradient of methanol increasing from 2% to 29% at a rate of 1.5% per minute. Methanol was further increased to 80% in 2 minutes and maintained for 11 minutes. The injection volume was 10 μL. Optimal ESI was conducted using nitrogen as the nebulizer (45 psi, 8 L/min) and drying gas (350°C). The capillary voltage of MS was maintained at + 3000 V, and the sampling orifice was set for the positive mode. The MS was operating in both scan and SIM mode, and spectra were acquired over the range of 100 to 650 Da. The diode array detector wavelength was set at 280 nm. Quantification was accomplished by analyzing the UV readings and confirmed by MS scan using single ion monitoring for pairs of ions: m/z 244.1/112.2, 228.2/112.2, 258.3/126.1, 242.1/126.3, 245.1/112.9, 268.1/136.1, 252.3/135.9, 284.1/152.3, 268.1/152.3, and 243.3/127.2 for C, dC, 5-mC, 5-mdC, U, A, dA, G, dG, and T, respectively.
Real-time Reverse Transcriptase Polymerase Chain Reaction
Total RNA was isolated using RNeasy mini-kits and the RNase-free DNase set (Qiagen). The Qiagen Omniscript Reverse Transcriptase kit was used for reverse transcription (RT). For negative control samples to test for residual DNA contamination, RT enzyme was omitted. The resulting cDNA was amplified by quantitative real-time polymerase chain reaction (PCR) analysis using QuantiTect SYBR Green PCR Master Mix (BioRad, Mississauga, ON, Canada), RT products, and primers (forward and reverse). The temperature cycles were 15 minutes at 95°C, then 40 cycles of 30 seconds at 95°C, 30 seconds at 55°C, and 1 minute at 72°C. The PCR primers (forward and reverse), designed using Beacon Designer 2.0 (BioRad) were: 5′-CAA TGA GCT GCG TGT GGC-3′ and 5′-GGG TCA TCT TTT CAC GGT TGG-3′ for β-actin (NM031144), 5′-AAC GGA ACA CTC TCT CTC ACT CA-3′ and 5′-TCA CTG TCC GAC TTG CTC CTC-3′ for dnmt-1 (AF116344), 5′-ATA GAA GTC AAG CGG GAA GTG-3′ and 5′-GGA GGC TGA GGT ATT GAT TGG-3′ for dnmt-3a (XM220319), 5′-CGT GGT AGG AGA TGG AGA TGG-3′ and 5′-TGG AGA TAC TGT TGC TGT TTC G-3′ for dnmt-3b (BN000397), 5′-CTT CAC ACT TAT CGC TAA TGG-3′ and 5′-TTG GGT CTG AGG CTA TGG-3′ for Cyp1A1 (NM 012540), and 5′-CTT CAC CAA ACG CCC CGA ACA C-3′ and 5′-CGG GAG AGG GTG GTG GGG TC-3′ for p16ink4a (NM031550). The results were normalized for the abundance of β-actin mRNA.
CpG Methylation Analyses of the Gene Cdkn2a by Pyrosequencing
Total DNA was extracted from liver samples using the DNAeasy kit (Qiagen, Valencia, CA), and the samples were shipped to an independent laboratory for blind analysis and further processing of the samples (EpigenDx, Inc., Worcester, MA, USA). The nucleotide sequence 1369–1427 (accession number AB081658) of the rat gene cyclin-dependent kinase inhibitor 2A (Cdkn2a) was investigated. This sequence included 7 CpG dinucleotides located upstream of the start codon (A in ATG is position 0). The pyrogram was disrupted at position -16 likely because of a point mutation A→G, which prevented the quantification of the percentage methylation of the 2 CpG dinucleotides closest to the start codon (positions -14 and -9). Therefore, the methylation analysis focused on the CpG dinucleotides located at positions -63, -55, -40, -29, and -23.
Sodium bisulfite conversion was performed using the EZDNA Methylation kit (Zymo Research, Orange, CA, USA). The bisulfite-treated DNA (1 μL, approximately 100 ng) was amplified by PCR reaction (95°C 15 min; 45 × [95°C, 30 s; 59°C, 30s; 72°C, 30 s]; 72°C, 5 min, 4°C) in 30 μL buffer including 3 mM MgCl2, 200 μM of each dNTP, 6 pM of forward (5′-GGAAGGAGGGTTTTATTGGTT-3′) and reverse (5′-biotin-CTACAAAAAACTCCATACTACTCC-3′) primers, and 0.75 U of HotStar Taq Polymerase. The PCR product, with 1 strand 5′ end biotin labeled, was captured on streptavidin-coated beads (GE Health Care Life Sciences, Baie d’Urfé, Québec, Canada). The biotinylated strand of DNA was separated and purified by alkali treatment and subjected to pyrosequencing using the sequencing primer, 5′-GGGTTTTATTGGTTATATTA -3′. The pyrosequencing analysis was performed using the PyroMarkMD system (Biotage, Uppsala, Sweden), and the degree of methylation was analyzed as an artificial “C/T SNP” using the AQ mode. The percentage methylated C was calculated as: C peak height × 100 / (C peak height + T peak height). As quality control, the completeness of the sodium bisulfite reaction was confirmed by the analysis of the conversion of Cs→Ts at sites that were not part of CpG dinucleotides. The absence of PCR bias (ie, disproportional amplification of 1 allele over another) and the capability for an absolute quantification of methylation was verified by testing the linearity of stepwise increments in methylation from samples prepared by mixing unmethylated and artificially methylated DNA. The 0% methylated DNA was derived from normal DNA extracted from rat blood. The 100% methylated DNA was artificially fully methylated in vitro using the SssI enzyme system according to the manufacturer’s protocol (New England Biolabs, Ipswich, MA, USA). These 2 DNA controls were mixed at an increasing proportion of methylated DNA (0%, 2.5%, 5%, 10%, 20%) before bisulfite modification. The pyrosequencing analysis revealed a strong linear correlation (r 2 = 0.999) between the expected and measured methylation levels.
Statistical Analysis
The normality of the data was verified by goodness-of-fit Shapiro test. Data failing normality were log transformed. If normality was reached, the data were analyzed by analysis of variance (ANOVA) followed by control to treatment means comparison using the Tukey-Kramer HSD test for comparison of all pairs. If the log-transformed data failed the normality test, the data were analyzed using the nonparametric Wilcoxon test when 2 groups were compared, and the Kruskal-Wallis rank sum test when differences among more than 2 groups were tested.
Three assays were performed to measure the mRNA abundance for DNMT1, 3a, 3b, and CYP1A1 at PND29 (Figure 3). The analysis of all control rats was repeated in each assay, explaining the presence of 3 control bars in Figure 3. Assay 1 included all samples from the control and the PCB-H groups, Assay 2 included controls and the other high-dose groups (OCP-H, MeHg-H, Mix-H), and Assay 3 included controls with all low-dose groups. Levels of p16 mRNA were analyzed in 2 assays, 1 for the low doses and a second assay for the high-dose groups. Levels of CYP1A1 mRNA at PND99 were analyzed in a single assay. In the text, comparisons to the control group refer to the control of the respective assay. The data for DNMT3a and p16INK4a were normally distributed, and comparisons among treatment groups (excluding the control) were analyzed using a 2-way ANOVA, testing for effects of treatment (OCP, PCB, MeHg, Mix), dose (low, high), and their interactions. This step was possible given that all the data are expressed as a percentage relative to the mean of the respective control group, and the variance of the control groups obtained in the different assays were not significantly different (O’Brien, Brown-Forsyth, and Levene tests). Further mean comparisons for these 2 end points were performed as described in the previous paragraph for the other data sets.
Regressions between 5mdC and dC were analyzed using the simultaneous linear regression models of SAS, version 9.1 TS Level 1M3 XP_PRO Platform (SAS Institute, Inc., Cary, NC, 2002), whereas other analyses were performed using the software JMP, version 3 (SAS Institute Inc., Cary, NC, 1998), all considering P ≤ .05 as the statistically significant level.
Results
The hepatic concentrations of the universal methyl donor SAM were significantly reduced in the high-dose PCB and Mix groups compared with control (Table 2). To assess a possible effect of a reduced availability of SAM on global genome DNA methylation, an LC-UV/ESI-MS method was developed to quantify nucleosides and to calculate the percentage ratio of 5mdC (100 × 5mdC/5mdC + dC) as an indicator of global genome DNA methylation. This LCUV/ESI-MS method separates and quantifies 10 nucleosides (Figure 1) with confirmed identity using distinct molecular ions. The limit of detection was less than 0.02 μg. Table 3 shows that 3.49% ± 0.06% of the deoxycytidine is methylated in the DNA of the control group. This percentage of methylation is lower in the group treated with a low dose of PCB (P = .02) than in the control, whereas the high dose of mixture induced a tendency for reducing global genome DNA methylation (P = .053). It is only in the high-dose Mix group that a statistically significant linear relationship could be obtained between the availability of SAM and global genome DNA methylation (r 2 = 0.5, P = .03). As expected, given the large number of deoxycytidine in the genome, the differences in global genome DNA methylation (Table 3) are small among treatment groups, therefore the relationship between dC and 5mdC was further investigated by a simultaneous linear regression analysis (Figure 2). Briefly, simultaneous linear regression models were fitted (5mdC = a + b × dC) to the full data set for each treatment by dose combination. The overall F test was significant (F = 2.85, P < .008), indicating that the slope for at least 1 of the treatment dose combinations was different. When the Mix-H group was removed from the data and the simultaneous regression model re-run, there was no statistical difference between the slopes of the remaining treatment by dose combination groups (F = 0.58, P = .77). Therefore, only the group Mix-H had a statistically significantly different slope compared to the other groups (Figure 2), which supports abnormal DNA methylation in that group. The data for the Mix-H group are located more to the upper right area of the graph only because more genomic DNA was analyzed in some samples of this group. The quantity of total nucleotides (5mdC + dC + dA + dT + dG) measured in the 4 rats to the right of the graph range from 11.47 to 11.60 μg, whereas the average for all samples was 8.57 ± 1.63 μg (mean ± SD). The amount of genomic DNA analyzed has no impact on the slope.
Figure 3 summarizes the treatment effects on DNMTs, CYP1A1, and p16INK4a mRNA abundance. Only the high-dose PCB and MeHg treatments significantly reduced the DNMT1 mRNA abundance below the control level (statistically significant comparisons to the control group are indicated by asterisks). Additional analysis revealed that DNMT1 mRNA abundance was significantly less in the Mix than in the OCP groups, and a significant difference between the low-dose and high-dose PCB treatments. Only the high-dose PCB treatment significantly reduced the DNMT3a mRNA abundance below the control level. In addition, DNMT3a mRNA was significantly less abundant in the Mix than in the OCP group (treatment group effects are indicated by capital letters). Abundance of DNMT3b mRNA was significantly reduced by all high-dose treatments compared with low-dose treatments (brackets), but only the high-dose PCB and high-dose MeHg treatments reduced mRNA abundance below the control level (asterisks). Over all the DNMT data, only the high-dose PCB and MeHg treatments changed the mRNA abundance relative with controls (asterisks): PCB affected all 3 DNMTs, decreasing hepatic mRNA abundance for DNMT1, 3A, and 3B to 4%, 54%, and 17% of the control values, respectively, whereas MeHg reduced DNMT1 and DNMT3B mRNA levels to 51%, and 18%, of the control value, respectively. Levels of mRNA in the OCP and the Mix groups were not significantly different from control.
Except for the low-dose OCP treatment group, CYP1A1 mRNA was significantly more abundant in all treatment groups compared with control (Figure 3, see asterisks). In all high-dose groups, mRNA was significantly more abundant than in the low-dose groups, except for the MeHg groups, where the abundance in the low-dose and high-dose groups was not significantly different. A synergic effect is apparent in the Mix high-dose group (5-, 64-, 4-, and 158-fold increase above control in the OCP, PCB, MeHg, and Mix group, respectively; note the log scale), but not among the low-dose groups (1.8-, 4.7-, 2.6-, and 6-fold increase above control in the OCP, PCB, MeHg, and Mix group, respectively).
For the cyclin-dependent kinase inhibitor p16INK4a, a known methylation target for DNMT3b, mRNA abundance was significantly reduced below the control level only by the high-dose Mix group (28% of control, asterisk). In the OCP group, mRNA was significantly more abundant than in the other groups (capital letters). Abundance of mRNA was significantly reduced in the high dose compared with the low dose only in the MeHg- or Mix-treated groups.
The mRNA expression of these 5 genes was also measured at PND99 in liver samples from sibling females of control and high-dose groups. Significant effects (P < .05) could be detected only for CYP1A1 mRNA, which was more abundant in the PCB and the Mix groups compared with other groups (Control, 100 ± 33; OCP, 258 ± 76; PCB, 3570 ± 1075; MeHg, 91 ± 30; Mix, 1950 ± 445).
The reduction in mRNA expression of p16INK4a, the decrease in availability of SAM, and the reduction in global genome DNA methylation, all in the high-dose Mix group, led us to investigate at PND 29 the methylation level of the CpG dinucleotides close to the initiation sites in the promoter region of the gene for p16INK4a. Figure 4 demonstrates a consistent pattern of methylation with significantly different levels of methylation over all CpG sites (capital letters) and significant treatment effects (asterisks). Although OCP treatment had no significant effects at any of the CpG dinucleotides, other treatments altered methylation of CpG at all sites except position -40. Generally, methylation levels were decreased by PCB, MeHg, and Mix treatments. The CpG dinucleotide located at position -23 appears to be different from the others since methylation was increased by most treatments (except OCP). In all cases, Mix appears to produce the greatest change in methylation levels, whereas OCP treatment never induced significant effects. Accurate measurements of the methylation level for the 2 CpG dinucleotides closest to the start codon (location -14 and -9) could not be obtained due to a disruption of the pyrogram, which suggests single-nucleotide polymorphism (SNP) attributable to a point mutation A→G (location -16) that occurred at a frequency that followed a mendelian inheritance pattern (+/+ 31%, +/− 48%, and −/− 21% of rats).
Discussion
The study of epigenetic effects of environmental contaminants, including mechanisms of DNA methylation, is a relatively new area of investigation in toxicology. By investigating the rat liver, this project demonstrates that in utero and lactational exposure to a mixture of persistent organochlorine pollutants and MeHg decreases the availability of SAM, DNMT mRNA abundance, and DNA methylation as supported by small decreases in global genome DNA methylation and in CpG methylation in a segment of the promoter of the p16INK4a gene. These changes occur concomitantly with hepatic inflammation, vacuolation, and hypertrophy, but no hyperplasia, necrosis, or changes in liver weight, as reported in other rats treated with the same mixture. 46 The current investigation suggests that the OCP component of the mixture is the least potent chemical group in affecting the DNA methylation system, as it had no obvious effects on DNMT mRNA abundance, SAM concentration, and DNA methylation. Interestingly, consistent with the current reduction in DNA methylation in the rat, a negative correlation was recently observed between serum level of persistent organic pollutants and global genome DNA methylation in Greenlandic Inuit. 52 Similar associations have also been reported in other human populations exposed to chemicals. 53,54
The changes in the abundance of the different DNMT mRNAs induced by the chemical groups suggest chemical- and mixture-specific effects. Compared with controls, PCB treatment reduced mRNA expression of the 3 DNMTs and MeHg treatment reduced the expression of 2 DNMTs, whereas OCP treatment did not affect any. The observation that the DNMT mRNA levels in the Mix group are statistically similar to the controls suggests antagonism among chemical components, an effect particularly evident for DNMT1. Explaining the origin of these chemical-specific imbalances in DNMT mRNA expression would require the investigation of end points known to be relevant to the DNMT promoters, including occupation of dioxin response elements, the methylation of CpG dinucleotides, and the role of Sp regulatory sites, 55,56 as well as microRNAs known to interfere with DNMT expression. 57,58 The general pattern of reduction in DNMT mRNA expression is consistent with the reduction in DNA methylation observed by both techniques (LC-UV/ESI-MS and sodium bisulfite pyrosequencing), but at this stage no specific relationship can be established between mRNA expression and DNA methylation. The literature reports that in some instances DNMTs work independently, but in others they work cooperatively. 26,59 The expression of many genes is often regulated beyond the mRNA level by translational or post-translational mechanisms, and others found no correlation between DNMT mRNA expression and DNA methylation in cancerous or noncancerous tissues. 60,61
In contrast to DNMT mRNA abundance, mRNA levels of CYP1A1 appear to exhibit a synergistic increase. However, using other rats treated with the same mixtures, ethoxyresorufin-o-deethylase (the classical enzyme reaction for assessing CYP1A1 activity) measurements did not reveal a synergistic increase; the PCB and Mix treatments induced enzyme activities that were similar and approximately 136-fold above control level. 62 This increase in enzyme activity is close to the 158-fold increase in CYP1A1 mRNA abundance reported here. These changes in CYP1A1 mRNA abundance and activity suggest significant and complex detoxification response, which was likely associated with increases in the amount of proteins that are known to sequester contaminants, including CYP proteins, metallothionein, 63,64 and glutathione involved in oxidative stress modulation. 65 As an indicator of exposure and effects, changes in the expression of DNMT mRNAs are not as sensitive as changes in CYP1A1 mRNA expression. However, whereas increases in CYP1A1 expression are normal detoxification responses, the biological consequences of the treatment-specific reductions in DNMT mRNA levels require further investigation.
The high dose of Mix and PCBs reduced the availability of SAM, the universal methyl donor involved in methylation reactions. It is well known that changes in the availability of SAM are associated with DNA hypomethylation and hepatocarcinogenesis. 23,66 Pyrosequencing provides robust quantitative methylation analysis for sequential series of CpG dinucleotides, 67 and our results (Figure 4) appear to follow an additive pattern among chemical treatments leading to a consistent reduction in methylation in the Mix group, which is in line with the reduction in availability of SAM (Table 2) and global genome methylation (Table 3). Global genome hypomethylation and site-specific hypermethylation are usually landmarks in the progression of carcinogenic events. Hypomethylation is involved with the activation of prometastatic genes and other tumor-promoting genes, 68–70 and it precedes genomic damage and cancer. 71,72 Perhaps exposure to these contaminants decreases global genome methylation and induces a predisposition to hepatocarcinogenesis. This hypothesis would be in line with the observations that PCBs and AhR agonists are classical hepatocarcinogens, mostly in females. 51
The high-dose Mix treatment decreased the abundance of p16INK4a mRNA, and this might be another observation supporting a predisposition to hepatocarcinogenesis. This protein originates from the INK4b-ARF-INK4a locus; its elevated expression causes a G1-phase cell-cycle arrest that is dependent on functional retinoblastoma protein. 73 p16INK4a has an important role in oncogene-induced senescence, a tumorigenesis barrier that slows or inhibits the progression of preneoplastic lesions to neoplasia. 74 A decrease in p16INK4a expression is common in various cancers and suggests increased mitotic activity with less time for DNA repair. Reduction in p16INK4a mRNA expression can be due to mechanisms unrelated to DNA methylation (eg, polycomb repressor proteins and mRNA stability, reviewed in Jacobs and de Lange 75 ), but it is often associated with DNA methylation of its promoter. Our investigation could not demonstrate a link between mRNA expression and DNA methylation. Others were able to associate hypermethylation of the p16INK4a promoter region with a reduced mRNA expression in chromium-, 17 arsenic-, 14,76 radiation-, 49 or 2-acetylaminofluorene (2-AAF)-treated rats. 77 Comparisons with the current investigation are hazardous, given the differences in chemical treatments, in the various techniques used to assess the methylation levels, in sex differences in chemical sensitivity and effects on DNA methylation, 77 and most importantly, in the DNA segment investigated. Our results are consistent with other pyrosequencing analysis of this promoter region, showing low methylation level of the CpG sites upstream of the transcriptional start site, 78 as well as with bisulfite genomic sequencing results demonstrating very little methylation of the CpG sites in this promoter region in normal lung tissue and in preneoplasia-containing tissues in the rat. 79 In fact, Blanco et al showed increases in methylation of CpG sites occurring only downstream of the transcriptional start site in lung tumors (a DNA segment not investigated here). 79 Perhaps the small increase in methylation at position 23 (the position closest to the transcriptional start site in Figure 4) is part of a progressive process, given that silencing of p16INK4a precedes the DNA methylation changes that were suggested to maintain rather than initiate gene silencing. 80 Investigating human livers and hepatocarcinomas using pyrosequencing, and reporting only the average methylation level over multiple CpG sites of the p16INK4a promoter, Kondo et al obtained, in normal tissues, a mean DNA methylation level of 3%, which is very close to our observations in the rat (Figure 4). 61 This percentage rises early during carcinogenesis to reach 37% in cancer. 61 These comparisons with human tissues may suggest that the changes observed here have not reached the carcinogenic process involving hypermethylation of the p16INK4a promoter.
Finally, the current pyrosequencing analysis supports the existence of an SNP in the p16INK4a gene (A→G, location -16) that followed a mendelian inheritance pattern. Single-nucleotide polymorphisms are more frequent in outbred strains like the Sprague Dawley rat than in inbred strains and complicate the interpretation of methylation analysis. This SNP, however, may not have a biological impact on protein activity given that it is not part of an exon.
In summary, this project assessed the impact on components of the DNA methylation system of in utero and lactational exposure to a mixture of environmental contaminants based on the chemicals measured in Inuit blood, and that generates in the high-dose rat pups a serum level of organochlorines 100 times the human level. 46 The decreases in abundance of DNMT1, 3a, and 3b mRNA induced by some treatment groups and the decrease in the availability of SAM and in DNA methylation as observed by 2 independent techniques (pyrosequencing of some CpG sites in the promoter of the gene p16INK4a and global genome DNA hypomethylation by LCUV/ESI-MS in the Mix group) support that the DNA methylation system could be affected following exposure to environmental contaminants. Compared with the PCB and MeHg component of the complex mixture, the OCP group was the least potent in affecting the current series of end points. Methylation of DNA is of fundamental importance in cellular and molecular biology, and the identification of the specific DNA loci where these changes have biological impact is slowly emerging. 81 Given the biological importance of changes in DNA methylation, it is plausible that exposure to contaminants induces site-specific changes in methylation that could be predictive of adverse health effects, and this possibility warrants further investigation.
Footnotes
Figures and Tables
Acknowledgements
This work was supported by the Northern Contaminants Program, Department of Indian Affairs and Northern Development (DIAND) and by Health Canada.
The authors are grateful to Dr. Nanqin Li for the preparation and analysis of the dosing solutions; to Leonora Marro for statistical advices; to Karen Leingartner, Bruce Martin, Susan Kelly, and Kelly Brennan for their technical assistance; to Liying Yan (EpigenDx Inc, Worcester, MA) for the pyrosequencing analysis; and to Dr. Tim Schrader and Dr. Craig Parfett for carefully reviewing this manuscript.