Low Doses of the Carcinogen Furan Alter Cell Cycle and Apoptosis Gene Expression in Rat Liver Independent of DNA Methylation

Background Evidence of potent rodent carcinogenicity via an unclear mechanism suggests that furan in various foods [leading to an intake of up to 3.5 μg/kg body weight (bw)/day] may present a potential risk to human health. Objectives We tested the hypothesis that altered expression of genes related to cell cycle control, apoptosis, and DNA damage may contribute to the carcinogenicity of furan in rodents. In addition, we investigated the reversibility of such changes and the potential role of epigenetic mechanisms in response to furan doses that approach the maximum estimated dietary intake in humans. Methods The mRNA expression profiles of genes related to cell cycle, apoptosis, and DNA damage in rat liver treated with furan concentrations of 0.1 and 2 mg/kg bw were measured by quantitative polymerase chain reaction (PCR) arrays. We assessed epigenetic changes by analysis of global and gene-specific DNA methylation [methylation-specific PCR, combined bisulfite restriction analysis (COBRA), and methylated DNA immunoprecipitation chip] and microRNA (miRNA) analyses. Results The expression profiles of apoptosis-related and cell-cycle–related genes were unchanged after 5 days of treatment, although we observed a statistically significant change in the expression of genes related to cell cycle control and apoptosis, but not DNA damage, after 4 weeks of treatment. These changes were reversed after an off-dose period of 2 weeks. None of the gene expression changes was associated with a change in DNA methylation, although we detected minor changes in the miRNA expression profile (5 miRNA alterations out of 349 measured) that may have contributed to modification of gene expression in some cases. Conclusion Nongenotoxic changes in gene expression may contribute to the carcinogenicity of furan in rodents. These findings highlight the need for a more comprehensive risk assessment of furan exposure in humans.


Research
Furan is an important industrial compound that was recently found in a number of heated food items, including baby food (Food and Drug Administration 2004). Furan has been reported to cause liver tumors characteristic of hepato cellular carcinoma and cholangio carcinoma (CC) in both male and female rats at doses of 2, 4, or 8 mg/kg body weight (bw), 5 days a week for 2 years [National Toxicology Program (NTP) 1993]. Male rats treated with 30 mg/kg bw furan for 3 months develop cholangiofibrosis, which in some cases progressed to tumors after 9 or 15 months without further treatment (Maronpot et al. 1991). There are no tumorigenicity data at doses < 2 mg/kg bw. Bearing in mind that CC is the next most common primary hepatic malignancy in humans after hepato cellular carcinoma and that it is associated with high mortality, an assessment of the effects of furan relevant to carcinogenicity at doses < 2 mg/kg is essential. A recent evaluation by the Joint Food and Agriculture Organization/World Health Organization Expert Committee on Food Additives (JECFA) also indicated a human health concern (JECFA 2010).
The mechanisms of furan carcinogenicity are still not well understood. Furan has shown negative results in a number of in vitro geno toxicity assays (Mortelmans et al. 1986), and an in vivo DNA binding study provided no evidence for covalent binding of furan to DNA (Burka et al. 1991). Wilson et al. (1992) found no evidence of increased DNA synthesis in an in vivo DNA repair assay, and furan was not clastogenic or aneugenic in either in vivo or in vitro studies (Durling et al. 2007). Kellert et al. (2008) demon strated that furan was not genotoxic when added directly to mouse lymphoma cells, although its metabo lite cis2butene1,4dial was genotoxic. Thus, the potential contribution of both genotoxic and non genotoxic mechanisms to the carcino genicity of furan must be taken into account.
Carcinogenesis is a complex multistage process during which the tightly controlled balance between cell proliferation and cell death is disrupted. Furaninduced cell prolifera tion and apoptosis have been found in mouse and rat liver (FranssonSteen et al. 1997;Mugford et al. 1997;Wilson et al. 1992). Cell proliferation, as a neces sary component for tumor development, may influence tumor formation by increasing spontaneous mutations or providing a pro motional influence to spontaneously initiated cells. Additionally, secondary oxidative DNA damage has been associated with chronic inflammation after furan exposure (Hickling et al. 2010).
Because of the uncertainty about the rela tive importance of genetic and epigenetic mechanisms, we investigated potential effects of furan on gene expression in relation to cell cycle, apoptosis, and DNA damage, and we assessed potential changes in DNA methyla tion and microRNA (miRNA) expression. DNA methylation is a potentially reversible chemical modification of cytosine residues that occurs predominantly in CpG dinucleo tides in mammals. Cancer cells often display global hypo methylation, which can cause genomic instability and activation of onco genes, whereas the promoter region of some specific genes, especially tumor suppressor genes, can be methylated, resulting in tran scriptional repression. DNA methylation is somewhat dynamic and responsive to envi ronmental exposures (Szyf 2007). Thus, sev eral nongenotoxic chemicals have been found to affect gene function through changes in DNA methylation (Bombail et al. 2004). For example, the non genotoxic rodent carcino gen pheno barbital induces global hypo methylation and regional hyper methylation in rodent livers (Bachman et al. 2006;Phillips and Goodman 2009). Aberrant DNA hypermethylation of tumor suppressor genes in human CC has also been frequently reported (Sandhu et al. 2008;Yang et al. 2005). miRNAs are a group of regulatory RNAs of 19-22 nucleo tides involved in post transcriptional gene regu la tion (Bartel 2004). miRNAs are abundant in the liver and play very important roles in liver functions (Chen 2009), and modulation of miRNA expression profiles is closely linked to the biological and clinical behavior of human intra hepatic CC ).
The aim of this study was to test the hypothesis that furan can modulate the expres sion of genes relevant to tumor induction, possibly through effects on DNA methyla tion Background: Evidence of potent rodent carcinogenicity via an unclear mechanism suggests that furan in various foods [leading to an intake of up to 3.5 μg/kg body weight (bw)/day] may present a potential risk to human health. oBjectives: We tested the hypothesis that altered expression of genes related to cell cycle control, apoptosis, and DNA damage may contribute to the carcinogenicity of furan in rodents. In addition, we investigated the reversibility of such changes and the potential role of epigenetic mechanisms in response to furan doses that approach the maximum estimated dietary intake in humans. Methods: The mRNA expression profiles of genes related to cell cycle, apoptosis, and DNA damage in rat liver treated with furan concentrations of 0.1 and 2 mg/kg bw were meas ured by quantitative polymerase chain reaction (PCR) arrays. We assessed epigenetic changes by analysis of global and gene-specific DNA methylation [methylation-specific PCR, combined bisulfite restriction analysis (COBRA), and methylated DNA immuno precipitation chip] and microRNA (miRNA) analyses. results: The expression profiles of apoptosis-related and cell-cycle-related genes were unchanged after 5 days of treatment, although we observed a statistically significant change in the expression of genes related to cell cycle control and apoptosis, but not DNA damage, after 4 weeks of treatment. These changes were reversed after an off-dose period of 2 weeks. None of the gene expression changes was associated with a change in DNA methyla tion, although we detected minor changes in the miRNA expression profile (5 miRNA alterations out of 349 measured) that may have contributed to modification of gene expression in some cases. conclusion: Nongenotoxic changes in gene expression may contribute to the carcinogenicity of furan in rodents. These findings highlight the need for a more comprehensive risk assessment of furan exposure in humans. or miRNA expression in rat liver. Importantly, we treated rats with 0.1 mg/kg bw furan, a dose closer to the estimated highest level of human exposure (3.5 μg/kg bw/day) (European Food Safety Authority 2004), and with 2 mg/kg bw furan, the lowest dose asso ciated with rat carcinogenicity (Maronpot et al. 1991). We used furaninduced CC and paired non tumor tissues as reference samples for DNA methylation assays.

Materials and Methods
Chemicals. We obtained furan (CAS no. 110009, ≥ 99% pure) from SigmaAldrich (Munich, Germany), NovaTaq Hot Start DNA Polymerase (catalog no. 71091) from Merck (Darmstadt, Germany), and all other chemicals and enzymes from New England Biolabs (Ipswich, MA, USA) or Sigma Aldrich, if not mentioned otherwise.
Animals. Male F344/N rats at 6-7 weeks old were purchased from HarlanWinkelmann (Borchen, Germany). Animals were treated humanely and with regard for alleviation of suffering. All procedures involving ani mals were performed according to national animal welfare regulations after authoriza tion by the local authorities (Regierung von Unterfranken). All animals were given free access to pelleted standard rat maintenance diet and tap water, and were housed in groups of five in Makrolon cages (TECNIPLAST, Hohenpeißenberg, Germany) at 21 ± 2°C and a 12/12 hr day/night cycle. Rats were allowed to acclimatize for 5-7 days before furan treatment. Furan was prepared in corn oil vehicle immediately before use and was administered orally via gavage at doses of 0, 0.1, and 2 mg/kg bw for 5 days, 28 days, and 28 days plus a 14day recovery period, for a total of nine treatment groups; animals were treated 5 days/week. Sections of left lobes of the liver, including subcapsular proliferative regions [which have previously been shown to be susceptible to furaninduced tumor for mation Maronpot et al. 1991)] were removed, frozen in liquid nitro gen, and stored at -80°C.
Genomic DNA and RNA purification. Samples of frozen rat liver (30 mg; n = 5) were processed using Precellys kit 03961CK14 (Bertin Technologies, Montignyle Bretonneux, France) containing 0.6 mL RLT buffer (Qiagen, Hilden, Germany) with 6 μL βmercaptoethanol. After loading on the Precellys 24 system (Bertin Technologies), samples were homogenized and centrifuged two times at 2,400 rpm for 10 sec. Genomic DNA and RNA were purified at the same time using the AllPrep DNA/RNA Mini Kit (Qiagen) according to the manufacturer's instructions. Genomic DNA contamination was removed from purified RNA samples using the Turbo DNAfree kit (Applied Biosystems, Foster City, CA, USA). The concentrations of DNA and RNA were measured by ultra violet (UV) absorbance using a NanoDrop 1000 Spectrophotometer (Thermo Scientific from Fisher Scientific, Loughborough, UK). The purity of DNA and RNA were deter mined by UV scanning between 200 and 300 nm and by the 260:280 nm ratio. For DNA and RNA, the ratios were 1.8-1.9 and 1.9-2.0, respectively. RNA quality was assessed by RNA gel electrophoresis. mRNA polymerase chain reaction (PCR) array. One microgram of purified RNA was used to synthesize complementary DNA using the RT 2 Profiler PCR Array kit (PARN012, 020, and 029; SABiosciences, Frederick, MD, USA). We performed realtime PCR using an ABI7000 PCR system and PowerSYBR reagents (Applied Biosystems) following the standard twostep cycling and dissociation program. Results from three different rat liver samples in each group were analyzed using the Excelbased data analysis template provided by SABioscience. Relative gene expression was cal culated using the comparative threshold cycle (Ct) method (2 −ΔΔCt ). miRNA PCR array. We extracted total RNA containing miRNA from three differ ent rat liver samples in each group using a mirVana miRNA isolation kit (Applied Biosystems). Purified RNA (2 μg) was reverse transcribed into complementary DNA using a QuantiMir Kit (RA680A1; System Biosciences, Mountain View, CA, USA). Realtime PCR was performed on a 384well plate using an ABI7900 PCR system (Applied Biosystems) and a QuantiMir kit (System Biosciences) and PowerSYBR reagents (Applied Biosystems) following standard thermo cycling conditions. We calculated rela tive miRNA expression (Ct) using the 2 −ΔΔCt method. We used miRGen software (Megraw et al. 2007) and MicroCosm Targets soft ware (version 5; European Molecular Biology Laboratory, European Bioinformatics Institute 2010) to predict miRNA targets.

Methylation-specific PCR (MSP) and combined bisulfite restriction analysis (COBRA).
To generate positive methylated DNA con trols, we incubated rat genomic DNA with 4 U M.SssI (CpG methylase) in the presence of Sadenosylmethionine. Bisulfite conversion of genomic DNA was performed using an EZ DNA MethylationGold Kit (Zymo Research, Orange, CA, USA) following the manufac turer's protocol. Genomic DNAs (~ 400 ng) from five different samples in each group were used for bisulfite treatment. Elution volume was 40 μL, and 2 μL of the eluted DNA was used as a PCR template. PCR reactions were performed on a Mastercycler (Eppendorf, Hamburg, Germany); primers are listed in Supplemental Material, Table 1 (doi:10.1289/ ehp.1002153). For MSP, the PCR products were directly run on a 2% agarose gel. For COBRA, the PCR products were digested by the corresponding restriction enzymes and then run on a 2% agarose gel.
Methylated DNA immunoprecipitation (MeDIP) and microarray analysis. We used this method as an "open" unbiased system to detect methylation changes. Methylated DNA was immuno precipitated using the MagMeDIP kit (mcmagme048; Diagenode, Liège, Belgium). Briefly, genomic DNA puri fied from rat liver was sheared to 200-1,000 bp at 30% power for ten 15sec pulses using VCX130 (Sonics, Newtown, CT, USA) and then incubated with anti-5methylcytosine antibody at 4°C overnight. After washing four times, the methylated DNAenriched genomic DNA faction was eluted in provided elution buffer. For micro array analysis, the immuno precipitated DNA and input DNA were amplified with a whole genome ampli fication kit (WGA250RXN; SigmaAldrich) and then were sent to NimbleGen's service laboratory to perform the array experiments using Rat CpG Island Promoter micro array (C71200001; Roche NimbleGen, Madison, WI, USA). Data were extracted from scanned images using NimbleScan 2.0 extraction soft ware (Roche NimbleGen) and were entered into GeneSpring GX software (version 7.3; Agilent Technologies, Santa Clara, CA, USA) for analysis.

Liquid chromatography/tandem mass spectrometry (LC-MS/MS).
Global DNA methyla tion was detected in at least five different rat liver samples in each group. RNA contamina tion of isolated DNA was removed by treating with 100 μg/mL RNase A and 2,000 U/mL RNase T in a final volume of 100 μL at 37°C for 2 hr; DNA was then purified by phenol/ chloroform extraction followed by ethanol precipitation. Genomic DNA (5 μg) was then denatured at 100°C for 3 min and chilled on ice. After treating with 5 U nuclease P1 (Calbiochem, Darmstadt, Germany) for 1 hr at 37°C, 1 U alkaline phosphatase (Sigma Aldrich, St. Louis, MO, USA) was added to the sample and incubated for 30 min at 37°C. The hydrolyzed DNA was transferred onto a cutoff filter (Millipore, Billerica, MA, USA) and centrifuged for 20 min at 4°C and 14,000 rpm. DNA hydrolysate (5 μL) was diluted with 495 μL doubledistilled H 2 O in a vial. LCMS/MS analysis was performed using an Agilent 1100 series LC coupled to an API 3000 triple quadrupole mass spec trometer equipped with a turbo ion spray source (Applied Biosystems). Separation was carried out on a Reprosil Pur ODS 3 column (150 × 2 mm, 5 μm) by gradient elution with 0.1% formic acid (solvent A) and methanol (solvent B) using the following conditions: 90% A and 10% B (starting conditions) fol lowed by an increase to 40% in 3 min and a linear increase to 100% B in 2.5 min, at a flow rate of 0.3 mL/min. Analytes were detected in the positive ion mode at a vapor izer temperature of 400°C. Data acquisition was performed by multiple reaction monitor ing of mass transitions m/z 268.2 to 152.1 for 2deoxyguanosine, and m/z 242.17 to 126.0 and m/z 242.17 to 108.95 for 5methyldeoxy cytidine. Quantitation of 2deoxyguanosine and 5methyldeoxycytidine was performed using external standards.
Statistical analyses. We used the Student ttest to analyze PCR arrays for both mRNA and miRNA expression profiles; only expres sion changes that were statistically significant and at least two times higher or lower than expression in controls were considered. mRNAs and miRNAs with average Ct values > 32 in both control and treated groups were not investigated further. Global DNA methyla tion levels were compared using oneway analysis of variance followed by Dunnett's and Tukey's multiple comparison test. A p value < 0.05 was considered statistically significant.

mRNA expression profiles of genes related to apoptosis, cell cycle, and DNA damage.
Each PCR array contained 84 genes related to apoptosis, cell cycle, and DNA damage. Three genes (casp3, Bcl2, and Trp63) are important to both apoptosis and cell cycle and were included in both arrays. Twelve other genes were present in both cell cycle and DNA damage arrays. Moreover, Tp53 and Gadd45a existed in all three arrays. Thus, in total, we examined 233 genes in this study using PCR arrays [see Supplemental Material (doi:10.1289/ehp.1002153)]. The expression of 18 genes, including Cdnk2a, was barely detectable in both control and treated sam ples (with an average Ct value > 32; data not shown); expression of these genes was not considered further.
Analysis of DNA methylation of selected genes. To determine if there was any methyla tion change in furantreated samples at the 2 mg/kg bw dose, we first examined the methylation status of five genes rele vant to carcinogenesis: p16 INK4a , p15 INK4b (Cdkn2b), Bid3 (BH3 interacting domain), Myc (myelocytomatosis oncogene), and Sfn (stratifin, also known as 1433 sigma). Two of these genes (p15 INK4b and Bid3) were up regulated in furantreated rat liver samples based on mRNA PCR arrays, consistent with the hypothesis that methylation may, in part, contribute to the mechanism of action. Myc is an oncogene, whereas Sfn, p16 INK4a , and p15 INK4b are tumor suppressor genes. MSP results shown in Figure 2A and B demonstrate that both p16 INK4a and p15 INK4b promoter regions were unmethylated in all control and treated samples, and only a very small propor tion of Myc promoter region was methylated. COBRA results ( Figure 2C) further proved the low methylation status of Myc promoter in both control and treated rat liver. Lack of methylation of Bid3 and methyla tion of Sfn in all control and treated samples were also demonstrated in COBRA results ( Figure 2D). Overall, we found no furaninduced DNA methylation changes in the selected genes.
MeDIP and DNA methylation micro array results. To screen for alternative genes with potential methylation alteration, we conducted MeDIP and micro array experiments.We focused on the 2 mg/kg bw furan treatment group because this dose can induce CC after 2 years of treatment (Maronpot et al. 1991). R. Maronpot kindly provided a furaninduced CC sample from a female SpragueDawley rat treated with 2 mg/kg bw furan for 500 days that we used as a reference sample and in which we found modulation of methylation.
After immunoprecipitation, we examined the enrichment of four singlecopy genes as well as the positive and negative controls pro vided in the Diagenode MeDIP kit to confirm the enrichment of methylated DNA by anti-5 methylcytosine antibody. The hypo methyla tion status of Myc gene promoter region was independently shown by MSP and COBRA (Figure 2A,C), whereas the 5´upstream region of the H19 gene has been found to be highly methylated in the adult rat by bisulfite sequencing (Manoharan et al. 2004). Thus, Myc and H19 could be employed as negative and positive controls, respectively. We also investigated the enrichment rate of LINE1 (long interspersed nucleotide element type 1) and ID (identifier) elements because repet itive elements represent a large part of the genomes and can reflect global methylation level to some extent. ID elements are mem bers of a family of SINEs (short interspersed nucleotide elements) in rodents. We speculate   TI T2  C2  CI  TI T2  C2  CI  TI T2  C2  CI  TI T2   UM   UM  M  M   UM   M   C2  CI  TI T2  C2  CI  TI T2  C2   CI  TI  T2  C2  CI  TI  T2  Me  Me  Mv  M100  C2  CI  TI  T2  C2 that cytosines lie within the CpG islands of the transposons, including LINE and SINE. Only the H19 gene and the positive internal control were enriched in immuno precipitated DNA [see Supplemental Material, Figure 1 (doi:10.1289/ehp.1002153)]. LINE1 showed little enrichment, which is consistent with a previous report that LINE1 was hypomethy lated in rat liver (Asada et al. 2006). Although the methylation level of one CpG of ID ele ments has been reported at > 60% (Kim et al. 2007), considering the high mutation level and the possible low methylation status of other CpG sites, it is not surprising that few ID elements were enriched in immuno precipitated DNA.
We found no evident methylation change in samples treated with furan for 4 weeks compared with control samples by the micro array method. However, when validating the assay, we noted methylation changes in pro moter regions of the CC sample relative to paired normal liver samples (data not shown).
Global DNA methylation. We observed no significant global methylation change in either furan treatment group (Figure 3). However, CC samples showed slight but sig nificant hypomethylation levels compared with non tumor samples (p < 0.01; Figure 3). Global methylation in mammals usually ranges from 3% to 5% (Bombail et al. 2004;Özden et al. 2008), in agreement with our findings.

Discussion
We assessed gene expression changes and epi genetic parameters relevant to carcinogenesis at furan doses lower than the minimum dose that has been shown to cause tumors, and closer to estimated exposure levels in humans. Although treatment with 2 mg/kg bw furan for 5 days did not appear to influence gene expression, we observed statistically signifi cant expression changes in the expression of a number of apoptosisrelated and cellcy cle-related genes after 4 weeks of treatment with 0.1 mg/kg bw furan. We also found that changes in gene expression following 4 weeks of treatment with 2 mg/kg bw furan were reversed 2 weeks after furan treatment was discontinued, which suggests that sustained exposure to furan is required to elicit effects at relatively low doses. Consistent with this, Hamadeh et al. (2004) reported a significant gene expression difference between 2week exposure and 3 or 7day exposure in rats treated with 4 mg/kg furan.
Although we found no DNAdamagerelated gene expression changes in samples from animals treated with 2mg/kg bw furan, Brca2 and Chek1-two cellcycle-related genes that respond to DNA doublestrand breaks (DSBs)-showed a slight over expression in samples from animals treated with 0.1 mg/kg bw furan. cis2Butene1,4 dial, the major reactive metabolite of furan, has been shown to form covalent adducts with nucleosides in vitro (Byrns et al. 2006), and DSBs have been found in mitogenstimulated spleno cytes of furantreated mice (Leopardi et al. 2010). Highdose (30 mg/kg bw) and longterm (3 months) treatment with furan has been found to induce oxidative stress and associated DNA damage (Hickling et al. 2010). However, in the present study, which was shorter term and used a lower dose, we found no change in oxidative DNAdamagerelated gene expression, nor did we find a gene expression profile charac teristic of a range of geno toxic carcinogens, such as described by EllingerZiegelbauer et al. (2005). In another study , we observed only a reversible enhanced proliferation in the subcapsular region of the left liver lobe after treatment with 2 mg/kg bw furan for 4 weeks, with little or no DNA oxidation. Considering that oxidative DNA damage is related to pro moting activity more than initiating activity (EllingerZiegelbauer et al. 2005;Umemura et al. 1999), the oxidative DNA damage may be a secondary mechanism associated with a persistent inflammatory response (Hickling et al. 2010).
Recent studies indicate that miRNAs are involved in regulation of cell proliferation and apoptosis (Ambros 2004). Moreover, miRNA expression profiles have been closely associ ated with the biological and clinical behav ior of human intra hepatic CC ). Thus, altered miRNA expression might contribute to the deregulation of certain cell cycle-related and apoptosisrelated genes. Here we found that 13 of the 349 miRNAs examined were aberrantly expressed in sam ples from the 2mg/kg-treated animals, and two downregulated miRNAs (rnolet7e* and rnomiR489) were consistent with the up regulation of mRNA expression of their pre dicted target genes, Bcl10 and Ccna2. The let-7 family of miRNAs, which is functionally con served from worms to humans, is important to normal development and differentiation and has been reported to be deregulated in various cancers (reviewed by Boyerinas et al. 2010). For example, let-7e* was downregulated in malignant mesothelioma (Guled et al. 2009), and let-7a was upregulated in lung, lym phoma, and ovarian cancers (Boyerinas et al. 2010). Caspase-3, Dicer, and Myc have been confirmed to be let-7a targets (Boyerinas et al. 2010). In the present study, rnomiR296 showed the greatest change in expression (~ 8fold downregulation). Human miR-296 has been reported to be downregulated in breast cancer and para thyroid cancer (Barh et al. 2008;Corbetta et al. 2010). Furthermore, inhibition of miR296 in Hela cells has been reported to decrease cell growth and increase the level of apoptosis (Cheng et al. 2005).
Changes in DNA methylation consti tute a mechanism for altering gene expression and are important non genotoxic mechanisms contributing to cancer. In the present study, we investigated the methylation status of five genes (p16 INK4a , p15 INK4b , Bid3, Myc, and Sfn) in samples from rats treated with furan at 2 mg/kg bw. Promoter methylation changes of p16 INK4a , p15 INK4b , and Sfn have been reported in CC (Lee et al. 2002;Yang et al. 2005), and hypomethylationinduced Myc over expression has been found in various types of tumors (Del Senno et al. 1989;Tao et al. 2002;Tsujiuchi et al. 1999). However, in the present study, we found no DNA methylation change in the five genes evaluated by MSP and/or COBRA. Furthermore, we found no methylation changes using the "open" unbiased method of MeDIP and micro array assay using NimbleGen's Rat CpG Island Promoter micro array, which covers 15,809 islands and 14,490 promoters. In addition, we did not observe evidence of a global methyla tion change based on LCMS/MS. Taken together, these results suggest that the reversible gene expression changes we observed were not caused by meth ylation changes. The findings contrast with the observation of irreversible gene expression changes, a DNA damage response, and altered methyla tion of at least one gene in liver sam ples from rats treated with a much higher furan dose of 30 mg/kg bw for 3 months (Hickling et al. 2010; Chen T, Williams T, Mally A, Hamberger C, Hickling K, Chipman JK, unpublished data).

Conclusion
Shortterm treatment with furan at a dose approaching the maximum estimated human dietary intake resulted in reversible changes in the expression of genes that control the cell cycle and cell death but did not appear to influence the expression of genes involved in responses to DNA damage. These changes did not appear to result from changes in DNA methylation, although modulation of miRNA may have a minor role in these responses. Bearing in mind the demonstrated differen tial profile of gene expression of genotoxic and non genotoxic carcinogens in rodent liver