Article pubs.acs.org/jpr
Disruption of DNA Methylation via S‑Adenosylhomocysteine Is a Key Process in High Incidence Liver Carcinogenesis in Fish Leda Mirbahai,†,‡ Andrew D. Southam,†,‡ Ulf Sommer,§ Tim D. Williams,† John P. Bignell,∥ Brett P. Lyons,∥ Mark R. Viant,†,§,⊥ and James K. Chipman*,†,⊥ †
School of Biosciences, University of Birmingham, Edgbaston, Birmingham, United Kingdom NERC Biomolecular Analysis Facility − Metabolomics Node (NBAF-B), School of Biosciences, University of Birmingham, Edgbaston, Birmingham, United Kingdom ∥ Cefas, Weymouth Laboratory, Weymouth, Dorset, United Kingdom §
S Supporting Information *
ABSTRACT: Interactions between epigenome and the environment in biology and in disease are of fundamental importance. The incidence of hepatocellular adenomas in flatfish exceeds 20% in some environments forming a unique opportunity to study environmental tumorigenesis of general relevance to cancer in humans. We report the novel finding of marked DNA methylation and metabolite concentration changes in histopathologically normal tissue distal to tumors in fish liver. A multi-“omics” discovery approach led to targeted and quantitative gene transcription analyses and metabolite analyses of hepatocellular adenomas and histologically normal liver tissue in the same fish. We discovered a remarkable and consistent global DNA hypomethylation, modification of DNA methylation and gene transcription, and disruption of one-carbon metabolism in distal tissue compared to livers of non-tumor-bearing fish. The mechanism of this disruption is linked not to depletion of S-adenosylmethionine, as is often a feature of mammalian tumors, but to a decrease in choline and elevated S-adenosylhomocysteine, a potent inhibitor of DNA methyltransferase. This novel feature of normal-appearing tissue of tumor-bearing fish helps to understand the unprecedentedly high incidence of tumors in fish sampled from the field and adds weight to the controversial epigenetic progenitor model of tumorigenesis. With further studies, the modifications may offer opportunities as biomarkers of exposure to environmental factors influencing disease. KEYWORDS: epigenetics, DNA methylation, metabolomics, liver cancer, methylation index, choline, chemical pollutants, SAH
■
INTRODUCTION Cancer is considered as much an epigenetic disease as it is a genetic disease.1,2 The highly regulated epigenetic profile of cells, controlled by a range of enzymes and metabolites,3 suffers a dramatic transformation during tumorigenesis.1 This is partly triggered by the susceptibility of the epigenome to environmental signals,4 which contributes to the increased risk of developing a cancer in response to environmental factors.1,5 In higher eukaryotes, the immediate environment influencing cells is determined by external exposure and by the metabolism and physiology of both neighboring cells and organs.3 Therefore, the epigenome can be altered not only by environmental factors, such as exposure to exogenous chemicals,6 but also by changes in the levels of endogenous cofactors and metabolites.2,3 Recently we have demonstrated that chemically induced liver tumors in zebrafish (Danio rerio), commonly used as a human disease model, have a similar epigenetic signature to human liver tumors,7 as well as similarities in the transcriptional profile.8 These findings imply a degree of conservation of © XXXX American Chemical Society
epigenetic abnormalities important in tumorigenesis among vertebrates. Tumours in fish collected from their natural habitat are mainly observed in demersal species, such as the common flatfish dab (Limanda limanda). This has been linked with their high levels of exposure to sediment-associated chemical carcinogens,9 making them an ideal species for biomonitoring and environmental carcinogenesis studies. An unusually high incidence of liver tumors, exceeding 20% of fish sampled at some sites, has been reported in dab captured from UK waters as part of the Clean Seas Environmental Monitoring Programme (CSEMP).10,11 This program not only provides information on the health status of the UK marine environment, it also serves as a unique resource for dedicated research into the subject of carcinogenesis.12−15 Our previous nontargeted 1H NMR metabolomic study of the non-cancerous distal tissue (DT) versus hepatocellular adenoma (HCA) in the same fish revealed disruption of the choline cycle, indicating the Received: March 4, 2013
A
dx.doi.org/10.1021/pr400195u | J. Proteome Res. XXXX, XXX, XXX−XXX
Journal of Proteome Research
Article
possibility of altered DNA methylation in dab tumors.16 This hypothesis was investigated in our recent study using DNA methylation-specific approaches.12 Remarkably, global DNA methylation levels in both hepatocellular adenoma (HCA) and non-cancerous DT of the same fish were statistically significantly lower (1.8 fold) compared to global DNA methylation levels in livers of non-cancer bearing dab. Furthermore, it was hypothesized that chronic exposure to a mixture of pollutants, such as estrogenic non-planar polychlorinated biphenyls (PCBs) and heavy metals that are known to influence DNA methylation,6 might contribute to this global DNA hypomethylation. In combination, these studies showed that alterations in DNA methylation are key tumor markers and led us to infer that they contribute to the process of tumorigenesis in dab. The one-carbon cycle is the principal biochemical pathway regulating DNA methylation, since it can alter production of the immediate methyl donor, S-adenosylmethionine (SAM).2 Chronic imbalance in the concentrations of one-carbon cycle metabolites can influence DNA methylation and underlies the pathogenesis of many diseases.17 For example, several studies in rodents have demonstrated that diets deficient in primary methyl donors such as choline, folate and methionine can lead to the development of tumors, especially in the liver. These tumors are induced through alteration of several biological pathways including the one-carbon cycle and consequently DNA methylation levels.18−20 Although the significance of the one-carbon cycle in the development of tumors in mammals is well established,18,19,21 this area is under-studied in relation to regulation of responses to chemical contaminants and diseases in fish. Here we present the discovery of the mechanism leading to disruption of DNA methylation in histologically normal tissue in tumor-bearing livers. Upon the basis of the successful use of “omic” studies in tumor characterization, discovery of disease mechanisms and diagnosis,22−27 we applied a multi-omics study, integrating observations at the levels of the epigenome, transcriptome and metabolome. First, an unbiased investigation into the metabolome was conducted using direct-infusion mass spectrometry, which enabled considerably deeper detection of the metabolome than our previous 1H NMR study, to identify key methylation-related metabolites that were altered in late stages of tumorigenesis between HCA and DT samples of the same fish. This non-targeted investigation discovered that the pathways related to the one-carbon cycle were among the top metabolic pathways altered in HCA. Consequently we developed a liquid chromatography−tandem mass spectrometry (LC−MS/MS) targeted analysis of key one-carbon metabolites, and quantified their levels in HCA and DT samples compared to within livers of non-cancer bearing fish. Furthermore, DNA methylation and transcriptomics measurements related to the one-carbon cycle were integrated with the metabolomics data sets to achieve a better understanding of the role of the one-carbon cycle in the development of liver tumors.
■
Aquaculture Science (Cefas, Weymouth, UK; Sample details are presented in Table S1, Supporting Information). Following euthanasia, fish were assessed for external diseases in accordance to Bucke et al.,28 and livers were visually assessed for the presence of macroscopic lesions. For fish exhibiting macroscopic evidence of large putative HCAs, a standardized 3−4 mm liver section was sampled and preserved in 10% neutral buffered formalin (NBF) for 24 h. Livers were subsampled further with suspected neoplastic tissues meticulously dissected and snap frozen in liquid nitrogen for downstream analysis. Corresponding distal tissue (DT) was also subsampled in the same manner. Only histologically normal liver tissue distal to tumors was used in this study. All histological analyses and interpretation was undertaken within the pathology section at Cefas, UK according to guidelines described by members of that team (Feist et al.29). Histopathology images of HCA and distal tissue are presented in our previous publication.12 The polar fraction of cellular metabolites was extracted as previously described.16 DNA Methylation and Transcriptomics Analysis
DNA and RNA samples were extracted from HCA and the matching DT of 10 fish and 12 non-cancerous livers. DNA Methylation and transcriptomics analysis of these fish livers has been described in our previous publication.12 Changes in DNA methylation and expression of genes involved in the one-carbon cycle were investigated using methylated DNA immunoprecipitation coupled to de novo high throughput sequencing (Data available at NCBI database with accession numbers GSE31124 and GSM770685) and gene expression microarray (MIAMEcompliant raw microarray data were submitted to ArrayExpress with accession number A-MEXP-2084.), respectively as described in our previous publication.12 Direct Infusion Mass Spectrometry-based Metabolomics
Direct infusion Fourier transform ion cyclotron resonance (FTICR) mass spectrometry (Thermo Fisher Scientific) was applied to the HCA and corresponding DT livers from twelve fish, providing an overview of the differences between the metabolome of HCA samples compared to DT. For this, dried polar tissue extracts were resuspended in three times their original volume of 4:1 methanol:water, with 0.25% formic acid for positive ionization analysis or 20 mM ammonium acetate for negative ionization analysis. Samples were directly infused using a chip-based nanoelectrospray ionization assembly (TriVersa) and analyzed by FT-ICR mass spectrometry in positive and negative ionization modes. Data were collected as transient data using the “SIM-stitching” method30 using 22 × 30 m/z overlapping SIM windows covering a mass range of 70−515 m/z. Transient data were processed, stitched and peaks were identified as described in detail in Protocol S1, Supporting Information. Projection of Putatively Annotated, Significantly Changing Metabolites onto KEGG Pathways
Of the significantly changing (FDR < 10%) metabolites between DT and HCA in the direct infusion mass spectrometry negative ion data set, putative annotations were projected onto pathways in the KEGG database (Protocol S2, Supporting Information). Pathways with high numbers of metabolites significantly altered between DT and HCA were hypothesized to reflect tumor-specific processes. Only negative ion data sets were analyzed with KEGG database as negative ion data set had
METHODS
Collection and Histology of Liver Samples and Extraction of Metabolites
As part of the UK Clean Seas Environmental Monitoring Programme (CSEMP), several hundred dab flatfish were sampled from the waters around the UK during June and July 2006−2009 by the Centre for Environment, Fisheries and B
dx.doi.org/10.1021/pr400195u | J. Proteome Res. XXXX, XXX, XXX−XXX
Journal of Proteome Research
Article
Table 1. Selected Putatively Annotated Peaks (from Tables S4 and S5, Supporting Information) Involved in Methylation and the One-carbon Cycle that Significantly Changed between DT and HCA (FDR < 10%)a putative metabolite identification betaine choline proline-betaine
S-adenosyl-homocysteine a
ionization mode
observed peak (m/z)
empirical formulas
Negative Negative Positive Negative Negative Positive Positive Negative Negative
152.0484 176.0928 104.107 178.064 202.1085 144.1019 166.0839 383.1144 419.091
C5H11NO2 C5H11NO2 C5H14NO C7H13NO2 C7H13NO2 C7H13NO2 C7H13NO2 C14H20N6O5S C14H20N6O5S
ion form [M [M [M [M [M [M [M [M [M
+ Cl]− + Ac]− − e]+ + Cl]− + Ac]− + H]+ + Na]+ − H]− + Cl]−
m/z error (ppm)
q-value
fold change from linear data (HCA/DT)
0.07 0.042 0.002 0.004 0.063 0.033 0.002 0.278 0.02
0.05 0.023 0.037 0.027 0.021 0.068 0.071 0.047 0.030
−1.34 −1.54 −1.95 −1.68 −1.63 −1.3 −1.24 −1.27 −1.32
p-values have been corrected for multiple hypothesis testing (q-value).
Data that were normally distributed with equal variance were analyzed by 2-tailed Student’s t test. Data that did not meet the requirements for a normal distribution or homogeneity of variance were analyzed by non-parametric statistics, using a Mann−Whitney U test. P-values
