Toxicogenomics and Cancer Susceptibility: Advances with Next-Generation Sequencing

Abstract

The aim of this review is to comprehensively summarize the recent achievements in the field of toxicogenomics and cancer research regarding genetic-environmental interactions in carcinogenesis and detection of genetic aberrations in cancer genomes by next-generation sequencing technology. Cancer is primarily a genetic disease in which genetic factors and environmental stimuli interact to cause genetic and epigenetic aberrations in human cells. Mutations in the germline act as either high-penetrance alleles that strongly increase the risk of cancer development, or as low-penetrance alleles that mildly change an individual’s susceptibility to cancer. Somatic mutations, resulting from either DNA damage induced by exposure to environmental mutagens or from spontaneous errors in DNA replication or repair are involved in the development or progression of the cancer. Induced or spontaneous changes in the epigenome may also drive carcinogenesis. Advances in next-generation sequencing technology provide us opportunities to accurately, economically, and rapidly identify genetic variants, somatic mutations, gene expression profiles, and epigenetic alterations with single-base resolution. Whole genome sequencing, whole exome sequencing, and RNA sequencing of paired cancer and adjacent normal tissue present a comprehensive picture of the cancer genome. These new findings should benefit public health by providing insights in understanding cancer biology, and in improving cancer diagnosis and therapy.

INTRODUCTION

All cancers result from genetic and epigenetic aberrations [1]. Genetic aberrations refer to inherited germline mutations and induced somatic mutations, which include single base mutations, indels (insertions and deletions), gene copy number variations, and gene/chromosomal rearrangements [2]. Some inherited germline mutations predispose individuals to particular types of cancer. These inherited variants can be categorized as either (i) rare, high-penetrance alleles and dominant mutations that strongly increase the risk of cancer development or (ii) common and low-penetrance alleles that mildly change an individuals’ susceptibility to cancer [3]. In contrast to germline mutations, which are transmitted passively through inheritance, somatic mutations are acquired among the cells of the body sporadically over the lifetime of the individual. Somatic mutations arise from DNA damage caused by exposure to endogenous or exogenous mutagens or resulting from spontaneous errors in DNA replication or repair. Some of these somatic mutations, termed driver mutations, are involved in the development or progression of the cancer; driver mutations confer growth advantages to cancer cells. Other somatic mutations, referred to as passenger mutations, happen to be present in cancer cells but do not contribute to the cancer phenotype [2]. Epigenetics has been defined as “heritable changes in gene expression that are not due to any alteration in the DNA sequence” [4]. Such epigenetic changes include his-tone modifications, DNA methylation pattern changes, and alterations of non-coding RNA (ncRNA) expression. Examples of epigenetic alterations in cancer include: (1) progressive hypomethylation of total DNA in tumor cells followed by hypermethylation of CpG-islands of tumor suppressor genes [4]; (2) alterations in the patterns of ncRNA expression that play important roles in the regulation of genes related to cell growth, cell proliferation, cell differentiation and cell death [5]; (3) global down-regulation of the expression of microRNA in comparison to counterpart normal tissues, with the ability to classify some poorly differentiated tumors by microRNA signatures [6]; and (4) histone modifications that alter chromatin structure and influence the expression of important tumor suppressor genes [7]. Genetic and epigenetic alterations in the expression of genes that regulate cell growth, cell division, cell differentiation, cell apoptosis, and other biological functions determine the cancer phenotype. In fact, it has been reported that more than 500 gene transcripts were differentially expressed at statistically significant levels between normal cells and cancer cells [8].

Advances in biotechnologies will now allow the detection of more mutations, epigenetic alterations, and gene expression changes in cancer cells. The accurate, economical, and rapid identification of mutations, gene expression profiles, and epigenetic alterations in cancer genome has led to insights in understanding cancer biology and to improved cancer diagnosis and therapy. Over the past several decades germline mutation detection approaches have dramatically improved from low throughput, low resolution methods like restriction fragment length polymorphism (RFLP) analyses [9] to next-generation sequencing (NGS) technology, which provides very high resolution genetic information relevant to carcinogenesis [10]. The power of molecular epidemiological studies of different types of cancer has evolved from single marker association analysis using RFLP to multiple marker association analysis using TaqMan assays [11], and to genome-wide association studies (GWAS) using microarray technologies [12]. Coincident with the announcement in November, 2013 that the Illumina MiSeqDx was the first NGS platform to receive marketing authorization by the FDA, Collins and Hamburg described the vast potential of high-throughput sequencing technology to revolutionize biomedical science and clinical medicine [13]. Currently, NGS methods can provide a much higher level of molecular detail with a relatively low-cost, enabling the comprehensive analysis of human and cancer genomes [14]. Whole genome sequencing of paired cancer and adjacent normal tissue presents a comprehensive picture of the cancer genome. Furthermore, NGS provides the opportunity to study the role of mutational signatures among cancers that can be associated with probable etiologies, such as exposure to genotoxic agents or defects in DNA repair [15]. Comparing somatic and germline mutations at base-pair resolution reveals precisely how the tumor genome is different from the normal genome and sheds light on the mechanisms of carcinogenesis [16]. Several international consortia, such as The Cancer Genome Atlas, the Cancer Genome Project (at the Wellcome Trust Sanger Institute), the International Cancer Genome Consortium, and Catalogue of Somatic Mutations in Cancer (COSMIC) have made tremendous efforts to identify cancer markers and mutations. Mutations, translocations, and potential therapeutic targets have been identified in numerous cancer subtypes by these consortia using NGS technology [17–19]. Microarray technology has made it possible to profile the expression of hundreds to thousands of genes in tumor tissues, allowing classification of cancers (such as breast cancer) into clinic subtypes, and predicting cancer recurrence and response to different treatment protocols [20]. NGS provides much higher resolution and increased depth of cancer gene expression by offering details of transcriptional boundaries, differential expressions, mutations, rare transcripts, and aberrations of alternative splicing [21]. Additional molecular changes associated with cancer can be detected using NGS. Castle and colleagues [22] utilized NGS methods to determine copy number differences for nuclear, mitochondrial, and telomeric DNA sequences with high accuracy. Likewise, NGS is a powerful tool for detection of epigenetic changes that, in turn, alter gene expression. For example, chromatin immunoprecipitation combined with NGS (ChIP-Seq) was used to define epigenome maps, which revealed underlying mechanisms of acquired drug resistance in breast cancer during endocrine therapy [23].

PERSONALIZED ONCOLOGY AND NGS—COLORECTAL CANCER AS AN EXAMPLE

A great wealth of information regarding the biology and treatment of colorectal cancer has been gathered through decades of intensive research. Information gathered from molecular epidemiological studies of colorectal cancer offer valuable perspectives on the need for NGS technologies to enhance future toxicogenomic studies and it also provides rationales for new applications of NGS-based methods in personalized medicine to treat this disease. The capabilities provided by NGS technology to gather and evaluate vast amounts of information on individual patient specimens at the molecular level provides an unprecedented opportunity to leverage what has been learned through past research to increase our understanding of gene-environment interactions in human colorectal cancer and to improve patient care.

Colorectal cancer affects over one million people globally and 5%–10% of these cases can be associated with the inheritance of high penetrance cancer susceptibility alleles that are typically transmitted as autosomal dominant traits [24]. Genes associated with hereditary nonpolyposis colorectal cancer/Lynch syndrome (MLH1, MSH2, MSH6, PMS2 DNA mismatch repair genes) [25], familial adenomatous polyposis (APC) [26], and hamartomatous polyposis/Peutz-Jeghers syndrome (STK11) [27] are each high penetrance genes associated with colorectal cancer. Variants of DNA polymerase-delta (POLD) and axis inhibitor 2 (AXIN2) genes transmitted through the germline are also associated with colorectal cancer with high penetrance, but these are more rare in the population [28]. Particular inherited variant alleles for APC (APC^*I307K), transforming growth factor beta receptor type 1 (TGFBR1^*6Ala), methylene tetrahydrofolate reductase (MTHFR^*677V), and H-Ras (HRAS1^*VNTR) are examples of alleles that are much more common in the population (5%–7%, 14%, 32%–43%, 1%–6% reported frequencies, respectively) but exhibit lower penetrance for colorectal cancer [relative risk (95% CI) 1.5–2.2 (1.21–2.07); 1.20 (1.01–1.43); 0.76 (0.62–0.92 protective); and 2.50 (1.54–4.05), respectively] (reviewed in [28]). Chromosomal instability is common in colorectal cancer [29]. Associations between telomere shortening, chromosomal instability, and increased susceptibility to colorectal tumors and other types of epithelial cancers are apparent in telomerase-deficient and p53-null mice [30]. The presence of colorectal cancer susceptibility traits and evidence of chromosomal instability or telomere dysfunction can be assessed simultaneously in patient samples using NGS methods.

In addition to the preceding list of genetic traits, dietary and environmental factors impact the risk of developing colorectal cancer. Exogenous risk factors of this type may increase the frequency of sporadic mutations resulting from DNA damage. Heterocyclic amines are well-known colon carcinogens that act in this way [31, 32]. Other exogenous nongenotoxic risk factors can affect cell growth, death, or differentiation processes in ways that lead to the selection of mutations that provide growth and survival advantages to mutant cells. Inflammatory stimuli that activate cyclooxygenase-2 and nuclear factor kappaB signaling are believed to promote colorectal cancer via this mechanism [33]. Or, exogenous factors may induce epigenetic changes, such as DNA methylation, that favor development of a cancerous phenotype. Dietary folate deficiency was associated (p = 0.01) with increased APC promoter methylation among colorectal cancers in a report by Gay and coworkers [34].

The consumption of well-done cooked red meat is considered an exogenous risk factor for colorectal cancer [35] because it is a source of heterocyclic amines including 2-amino-3,4,8-trimethylimidazo[4,5-f]quinoxaline 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) which damages DNA leading to mutations [36]. Protein and DNA adducts were detected in the colons and in the blood of human volunteers exposed to labeled PhiP at levels found in the diet [37]. Human metabolism plays a role in the activation and the detoxification of heterocyclic amines relevant to colorectal cancer. Human cytochrome P450 isoforms CYP1A2, CYP1A1, CYP1B1, and CYP2A6 catalyze N-oxidation of heterocyclic amines and the resulting hydroxylamine derivatives thus produced are activated by acetyltransferase (NAT1, NAT2) or sulfotransferase (SULT1A1) activity to form electrophilic N-acetyloxy or N-sulfonyloxy esters that react with DNA [32, 38]. The inheritance of alleles that provide enhanced catalytic activity for enzymes responsible for the metabolic activation of heterocyclic amines has been associated with increased risks for colorectal cancer [32, 39, 40]. Similarly, inheritance of alleles that result in reduced activity for enzymes involved in detoxification, including glutathione S-transferase (GSTA1^*B) [41] or UDP-glucuronosyltransferase (UDP1A1-3279 GG/TG in Caucasians) [42], are also linked to increased risk for colorectal cancer. Exposure to heterocyclic amines results in a diagnostic mutational signature that involves G:C single base pair deletions, particularly within 5′-GGGA-3′ motifs, along with G:C → T:A transversions [43]. The status of each gene involved in heterocyclic amine metabolism within a tumor sample can be determined using NGS and, by comparison with normal adjacent tissue, the presence of signature mutations associated with exposure to heterocyclic amines can also be identified within colorectal tumors.

In addition to the inherited traits described previously that influence the risk for colorectal cancer, certain genes have been found to be targeted by induced, or acquired, somatic mutations in colorectal cancer. Somatic mutations or epigenetic silencing of the KRAS, DCC, TP53, and CDKN1B genes are important in colorectal cancer (reviewed in [44]), along with epigenetic silencing of MLH1 followed by mutations in TGFBR2, BRAF, and BAX genes and chromosome 18q deletions involving the SMAD4 and DCC tumor suppressor genes (reviewed in [45]). Because NGS techniques can provide a comprehensive analysis of genetic and epigenetic aberrations, induced mutations and epigenetic changes involving all genes associated with colorectal cancer can be accessed directly. If environmental exposures for study populations are also known, molecular fingerprints for colorectal cancer biology may be obtained from enhanced molecular epidemiologic studies using NGS-based strategies.

Important medical decisions regarding patient care choices could be improved by results obtained using NGS methods. Two treatment modalities approved by the U.S. Food and Drug Administration for metastatic colorectal cancer involve the anti-epidermal growth factor receptor (EGFR) antibodies panitumumab, used alone, and cetuximab in combination with irinotecan, 5-fluorouracil, and leucovorin (FOLFIRI). Reported response rates to cetuximab and panitumumab monotherapies for metastatic colorectal cancer are limited to 8%–10% [46, 47] in patient populations that have not been screened for KRAS mutation status. Use of either anti-EGFR antibody requires genetic screening [48] because response rates are much lower for colorectal tumors bearing mutant KRAS genes [49, 50]. Increased detection sensitivity of KRAS mutation status provides enhanced predictive confidence for anti-EGFR antibody therapy [51, 52]. Furthermore, when expression of PTEN and the mutation status of KRAS, BRAF, and PIK3CA were determined simultaneously, up to 70% of responders could be identified [53]. These observations are consistent with a hypothesis presented by Parsons and Myers [54] that most, if not all, colon tumors contain small subpopulations of cells harboring undetected KRAS mutations and that outgrowth of these KRAS mutant tumor cells leads to relapse following anti-EGFR antibody therapy. Although preliminary studies with mutant KRAS colorectal tumor cells showed sensitivity to the combination of the MEK inhibitor AZD6244 and cetuximab [55], acquired resistance to AZD6244 developed in the human colon cancer cell lines HCT116 and COLO205 that was conferred through selective amplification of the KRAS ^G13D and BRAF ^V600E oncogenes, respectively, that they harbored [56]. Thus, future studies are needed to determine whether co-targeting the RAS/RAF/MEK/ERK signaling pathway simultaneously with anti-EGFR antibodies will lead to improved long term survival of metastatic colorectal cancer patients. The combination therapy for colorectal cancer of 5-fluorouracil (5-FU) with leucovorin, a folate derivative, depletes dTTP pools via inhibition of the de novo thymidylate synthesis pathway. In addition, 5-FU acts by also interfering with ribosomal RNA processing via the uridine monophosphate kinase (UMPK)-dependent misincorporation of fluorouridine triphosphate into pre-ribosomal RNA transcripts. Interestingly, variant alleles affecting the 5′- and 3′-untranslated regions of the thymidylate synthase gene (TYMS) were found to be protective for colorectal cancer progression to advanced stages (OR = 0.5, 95% CI: 0.3–0.9) [57], and overexpression of thymidylate synthase has been associated with resistance to fluorouracil plus leucovorin therapy in colorectal cancer [58]. Other markers associated with either response or resistance to fluoropyrimidine therapy for colorectal cancer include the expression of UMPK [59], dihydropyrimidine dehydrogenase, thymidine phosphorylase [60], and orotate phosphoribosyltransferase [61]. Variants forms of the UGT1A1 gene (reviewed in [62]) have been associated with adverse reactions to irinotecan therapy for metastatic colorectal cancer (severe neutropenia and diarrhea), although the Evaluation of Genomic Applications in Practice and Prevention Working Group found that further studies are warranted before recommending routine UGT1A1 genotyping for irinotecan therapy [63]. Powerful NGS technologies allow the expression levels and mutation status to be evaluated simultaneously for each genetic marker associated with sensitivity or resistance to candidate chemotherapeutic strategies for colorectal cancer and for known markers of adverse treatment outcomes.

GENETIC VARIANTS AND ENVIRONMENTAL EXPOSURE

Most human illnesses are outcomes of the interactions between inherited genetic factors and modifiable environmental factors [64]. Similarly, most cancer cases are the consequence of the interaction of genetic variants and environmental factors. It is believed that the genetic factors by themselves contribute to the development of approximately 5% of all cancers [3]. For example, familial breast cancer results from a predisposition due to genetic variants. However, only 5%–10% of all breast cancers run in families. Even in familial breast cancer, well-known mutations, such as mutations in BRCA1 and BRCA2 genes, account for only approximately 20% of the familial risk. The development of sporadic breast cancers, accounting for the rest of all breast cancers, might be related to the interaction of genetic variation and environmental exposures, such as radiation, chemical carcinogens, and xenoestrogens [65]. An individual may inherit a genetic make-up that predisposes them to the development of cancers, yet the onset of the cancer and the severity of the cancer are usually modified by exposure to environmental factors. In the biological DNA-RNA-protein-biofunction dogma, many studies have shown that environmental factors, such as nutrients, pharmaceuticals, air pollutants, carcinogens, and pesticides, have impacts on genomic events, such as gene mutations, gene transcription, histone modifications, DNA methylation, and DNA repair, which ultimately affect disease phenotypes such as cancer, diabetes, allergies, and infertility [66].

The risk of developing cancer from environmental exposures may depend on particular genetic variants carried by individuals. A clear example is the greater risk of cutaneous melanoma as a result of ultraviolet (UV) exposure for individuals that harbor xeroderma pigmentosum variant (XP-V) alleles. Compared to normal individuals, XP-V carriers have an increased risk (1000-fold higher) of developing cutaneous melanoma after exposure to UV light (or sun-exposure) [67] because the defective form of human DNA polymerase-eta is unable to perform damage-bypass replication of UV-damaged DNA [68]. In terms of affecting the risk of cancer development, a less dramatic gene-environment interaction in affecting the risk of cancer development is illustrated by variants in the Cockayne syndrome B protein (ERCC6). ERCC6 is a base-excision repair enzyme playing an important role in DNA repair. The association between a common variant (rs3793784: C > G) in ERCC6 and lung cancer risk was examined in a case-control study with 1000 cases and 1000 controls. A 1.76-fold (p < 1.0×10⁻⁷) excess risk of developing lung cancer was found for the -6530CC carriers compared with noncarriers. Through gene-environment interaction analysis, it was found that the −6530CC allele interacts with smoking to heighten lung cancer risk. An odds ratio of 8.87 (95% CI 5.74–13.71) for developing lung cancer was found among heavy smokers, suggesting the importance of environmental factors in the development of genetic variant-related diseases [69]. Colorectal cancer provides another example of the importance of gene-environment interaction. The risk/protective factors for colorectal cancer include family history, consumption of red meat, fat, vegetables, fruits, fiber, alcohol, cigarettes, micronutrients, and anti-inflammatory drugs. Genetic mutations in genes involved in the APC (adenomatous polyposis coli)-β-catenin-Tcf (T-cell factor) pathway, DNA mismatch repair pathway, and xenobiotic metabolizing pathway are significantly associated with the incidence of colorectal cancer, and these genetic factors interacting with environmental factors intensify the risk [70].

Molecular epidemiological studies have revealed that many environmental factors may interact with genetic variants to affect the risk of cancer development. Humans are continually exposed to harmful environmental factors or hazardous life styles and have evolved metabolic pathways and cellular controls that can minimize the biological impact introduced by hazardous environmental agents. Such pathways and controls involve the metabolic activation or detoxification of carcinogenic chemicals, controlling the cell cycle, cell differentiation and cell death, and DNA repair [71]. All the genes involved in these environmental response pathways are subject to genetic variability, which in turn can change the efficiency of these biological functions. Most cancers seem to be polygenic, arising from complex interactions among several genes or genetic variants, and gene-environment interactions. With the exception of some high penetrance, dominant alleles (such as BRCA1 discussed above, and others discussed next), susceptibility genes or genetic variants are usually insufficient to cause disease. Although these low penetrance genes or genetic variants only have relatively small impacts on increasing disease risk (e.g., only increase the risk by several fold), they could contribute a significant effect on the incidence of cancer development in a human population owing to their relative high allele frequency [3]. In general, a specific individual’s risk for developing a cancer is dependent on the interaction of environmental exposures to carcinogenic agents within the context of that person’s inherited set of cancer susceptibility or resistance traits. Table 1 lists some examples of known genotypic markers associated with particular types of cancers and environmental risk factors.

Table 1.

Examples of Genes-Environment Interaction in Cancer Development

Cancer	Gene	Environment Factor	Reference
Melanoma	XP-V, NER, CS	Ultraviolet Radiation	Van Steeg 1999 [67]; Masutani 1999 [68]
Lung Cancer	CYP1A1, ERCC6	Cigarette Smoking	McLemore 1990 [167]; Lin 2008 [69]
Liver Cancer	TERT, MAPK1, p53	Viral Infection AflatoxinB1	Murakami 2005 [168]; Bressac 1991 [169]
Bladder Cancer	GSTM1 NAT2	PAH Cigarette Smoking	Bell 1993 [170] Marcus 2000 [171]
Colorectal Cancer	MTHFR NAT1, NAT2, CYP1A2	Folate, VitB12, Red Meat	Chen 1996 [172]; Ma 1997 [173] Chen 1998 [174]; Kadlubar 1992 [175]

Although some of the more abundant genetic variants may pose lower individual risk or lower penetrance than others, it has been shown that these low penetrance genetic traits contribute to substantial differences in cancer susceptibility among populations. Moreover, gene–environment interactions may intensify the risks of cancer development among carriers of these low penetrance traits, thus they are important for the determination of population risk.

On the other hand, numerous cancer genes have been identified as germline-inherited tumor predisposition genes with very high penetrance. Well-known examples of inherited cancer susceptibility traits include RB1 for retinoblastoma [72], TP53 for colorectal cancer [73], and for the familial syndrome of breast cancer, sarcomas, and other neoplasms [74], BRCA1 [75] and BRCA2 [76] for breast cancer and ovarian cancer, RET for familial medullary thyroid carcinoma [77], MET for hereditary papillary renal carcinoma [78], APC for colorectal cancer [79], CDKN2A for melanoma [80], and KRAS for bladder, lung [81], colorectal, and other cancers [82]. Mutations in these genes confer high risks of cancer development among carriers. Compared to low penetrance genetic variants, the high penetrance germline mutations have the following characteristics: they cause familial cancers; their allele frequencies are generally rare in the population; their risks are sufficient to increase the incidence of a cancer in the affected families; and, because powerful activating mutations are already present in the germline, genotoxic environmental factors play a smaller role in carcinogenesis among these families [83].

EPIGENETIC ALTERATIONS AND ENVIRONMENTAL EXPOSURE

While cancer is considered a genetic disease as discussed, it also can be considered as an epigenetic disease. Epigenetics refers to stably maintained or altered gene expression patterns, without changing the underlying DNA sequence, via DNA methylation [84], chromatin modification [85], and ncRNA modulation [86]. The epigenetic mechanisms that control gene expression are based on covalent modifications of DNA and/or protein molecules, or modulation of protein translation. These mechanisms are susceptible to environmental challenges such as chemicals or metabolites derived from the diet, nutritional deficiencies, and exposure to other xenobiotics [87]. The dynamics of DNA methylation/de-methylation, chromatin status, and modulation of protein translation by ncRNA each function in a precisely controlled manner to maintain cellular homeostasis by fine-tuning gene expression patterns in normal cells. However, dramatic changes of epigenetic status may result in the activation of oncogenes, inactivation of tumor suppressor genes, induction of chromosomal instability and mutations, abnormal expression and translation of genes, and production of aneuploidy, all being pivotal biological events in the carcinogenic process [88, 89].

DNA methylation is achieved by addition of a methyl group to the cytosine in CpG dinucleotides, catalyzed by DNA methyltransferases. The genome of the cancer cells is usually characterized by aberrations in DNA methylation: hypermethylation in the CpG island of tumor suppressor genes and hypomethylation in the global genome [88]. In many cancers, hypermethylation at CpG islands in the promoter regions of tumor suppressor genes is usually a major cancer-predisposing event. Hypermethylation provides a mechanism to inactivate a variety of genes involved in normal biological functions that also function to repress cancer. Hypermethylated targets associated with cancer include tumor suppressor genes (e.g., VHL, p15, p16, p73 and BRCA1), DNA-repair genes (e.g., hMLH1 and MGMT), drug metabolizing genes (e.g., GSTP1), cytokine receptor genes (e.g., SOCS1), and apoptotic signal genes (e.g.,TMS1) [4]. The hypermethylation pattern is cancer-specific, with different types of cancer utilizing different sets of hypermethylated genes [4].

In contrast to hypermethylation of CpG islands in tumor suppressor genes, the overall genome of cancer cells is globally hypomethylated. It has been found that cancer cell genomes have only 20%–60% of the methylation level as their normal counterparts [90, 91]. Hypomethylation commonly affects the exonic and intronic regions of genes, and has biological impacts on carcinogenesis through mechanisms which include the increase of chromosomal instability, reactivation of transposable elements, loss of imprinting, recombination and chromosome rearrangement, and the promotion of aneuploidy [88]. However, a comparison of methylation patterns in the colon cancer cell line SW48 to those of primary colon cells revealed that the methylation patterns were markedly similar between the primary cells and the transformed cells, “suggesting that aberrant methylation of CpG island promoters in malignancy might be less frequent than previously hypothesized” [92].

In eukaryotes, DNA molecules are associated with small proteins including histones, which are the fundamental components of chromatin. Chromatin remodeling controlled by chemical modifications of the histones is important for many biological processes [93]. Dynamic structural changes of chromatin affect the “packing” of DNA chains. Highly condensed chromatin can exclude the transcriptional machinery, thus influencing the expression of genes, DNA replication and repair, chromosome condensation and segregation, and apoptosis [94, 95]. Highly regulated modifications of histones, including histone acetylation/deacetylation, phosphorylation/dephosphorylation, and methylation/demethylation are involved in pivotal cellular events, such as proliferation and differentiation, and are thus modifiers of the carcinogenesis process [96]. For example, the dynamics of histone acetylation, which results in a more open, accessible chromatin state, is precisely regulated by histone acetyltransferases (HAT) and histone deacetylases (HDAC). Impaired enzymatic activities of HAT and/or HDAC are associated with the development of cancers: mutated/truncated HATs have been found in colorectal, gastric tumors and other epithelial cancers; translocation-generated HDACs have been identified in leukemia and lymphoma [97].

ncRNA is another epigenetic modulator for gene expression that can operate through two mechanisms: translational repression by microRNA (miRNA) and mRNA degradation by small interfering RNA (siRNA). These ncRNAs are important players in gene regulatory networks. Consequently, mutations and or genetic variants residing in ncRNA sequences, as well as differential expression of ncRNA molecules among individuals, could lead to a substantial change in phenotypes. Such changes may have a significant influence on quantitative traits, including the development of cancer and other diseases [98].

As a complex process, carcinogenesis is a result of aberrant cellular processes caused by both genetic and epigenetic events. Epigenetic machinery, including DNA methylation, histone modification, and ncRNA expression, can be modulated by environmental exposure such as dietary chemicals and pollutants. Dietary factors, such as vitamins and micronutrients, are critical for the dynamic balance of epigenetic patterns that keep cells healthy. For example, the one-carbon metabolic pathway is utilized to generate S-adenosylmethionine, the methyl donor necessary for DNA methyltransferase activity. Choline, methionine, folate, vitamin B₁₂, vitamin B₆, and riboflavin are strategic dietary factors involved in the efficient maintenance of methyl donor pools and healthy DNA methylation status [99]. Deficiencies of diet-provided substrates and cofactors in one-carbon metabolism may therefore result in impaired DNA methylation which is linked to increased risks for neural tube defects, cardiovascular diseases and cancers [100]. Xenobiotic stressors, including heavy metals, water disinfection byproducts, air pollutants, endocrine-disrupting toxicants, and cigarette smoke, can also affect epigenetic components. Table 2 summarizes impact of environmental chemicals in terms of their effects on epigenetic alterations.

Table 2.

Examples of Interaction between Environmental Exposure and Epigenetic Alteration

	DNA Methylation	Histone Modification	ncRNA Expression
Food Constituents
Folate	Affected [176]	Affected [182]	Affected [186]
Retinoic Acid	Affected[177]	Affected [177]	Affected [187]
Vitamin B-12	Affected [178]	Affected [100]	Affected [188]
Vitamin D	Affected [179]	Affected [183]	Affected [189]
Selenium	Affected [180]	Affected [184]	Affected [190]
Polyphenols	Affected [181]	Affected [185]	Affected [191]
Pollutants
Cigarette smoke	Affected [192]	Reduced [202]	Affected [210]
Heavy metals
Cadmium	Decreased [193]	Reduced Acetylation [194] Increased Phosphorylation [203]	Affected [211]
Nickel	Increased [194]	Decreased Acetylation [204] Increased Demethylation [204]	Affected [212]
Mercury	Globally Decreased [195]	Reduced Methylation [205]	Affected [213]
Lead	Decreased [196]	Reduced [196]	Affected by Pb Mix. [214]
Arsenic	Globally Increased [197]	Globally Affected [206]	Affected [215]
Disinfection byproducts (TCE, DCA, TCA)	Increased [198]	Unknown	Unknown
Air pollution	Globally Decreased [199]	Affected [207]	Affected [211]
Other Chemicals
Benzene	Affected [200]	Increased Phosphorylation [208]	Affected [216]
Bisphenol A	Globally Decreased [201]	Increased Trimethylation [209]	Affected [217]

GENE EXPRESSION AND TOXICOGENOMICS

The merging of conventional toxicological research and functional genomics resulted in the emergence of toxicogenomics after the introduction of DNA microarray (i.e., transcriptomics) technology and, more recently, NGS [101]. Toxicogenomics has now incorporated other high throughput omic technologies such as proteomics and metabonomics. Using bioinformatics software and databases, the raw data are analyzed and meaningful biological information such as gene expression changes and alterations in biological pathways/functions are obtained. Therefore, toxicogenomics has become an important sub-discipline in the field of toxicology, with its primary goals to understand the relationship between environmental stress and human disease susceptibility, such as cancer susceptibility; exploring the molecular mechanisms of environmental mutagens and/or carcinogens; and identifying potential biomarkers of disease and toxicity, including mutagenicity and carcinogenicity [102].

Toxicology has traditionally relied on in vivo and in vitro models to study the adverse effects of chemicals, including cellular, biochemical, and molecular mechanisms of action. Toxicogenomics has been used to understand toxicant-induced effects in vivo and in vitro and to address challenges that are difficult to overcome by conventional toxicology methods. Among the omic technologies, microarray-based toxicogenomics methods have been exploited most extensively for mechanistic studies because they can evaluate the relative expression of thousands of genes conveniently and cheaply to identify early biomarkers of toxicity and disease [103]. Recently, next-generation sequencing (i.e., RNA-Seq), a newly developed technology, has been used in toxicogenomics studies and it has significantly accelerated genomic research and discovery. Compared to DNA microarrays, RNA-Seq may provide more sensitivity in detecting genes with low expression levels [104, 105].

Genotoxic carcinogens cause genetic errors via molecular modifications of DNA that lead to mutations that become fixed following DNA replication. Because of the association between DNA damage and cancer development, genetic toxicology data have been used for hazard identification and cancer risk assessment. A multitude of toxicogenomics studies have been performed to evaluate a number of environmental mutagens and/or carcinogens with both traditional toxicity endpoints and genomics changes. Generally, there are two ways to perform toxicogenomics studies: (1) starting with conventional toxicological research and then focusing on omics approaches to detect systematic biological effects, and (2) starting with the omics study followed by conventional toxicological research to interrogate molecular mechanisms. NGS technologies have been used in both approaches [105–108].

Approaches for investigating simple chemically-induced toxicity or carcinogenicity are well established. For example, aristolochic acid is a potent human nephrotoxin and carcinogen, and the International Agency for Research on Cancer classified it as a Group 1 human carcinogen [109]. DNA adduct formation and mutagenicity by aristolochic acid were detected in the kidney, spleen, and liver of Big Blue transgenic rats gavaged with 0, 0.1, 1.0, and 10.0 mg/kg body weight aristolochic acid for 3 months [110, 111]. Treatments with aristolochic acid resulted in three major DNA adducts in the tissues tested and the induced mutations in the cII gene exhibited linear dose-responses to aristolochic acid. A:T → T:A transversion was identified as the predominant mutation in aristolochic acid-treated rats [110, 112]. A dose-dependent induction of H-Ras mutation fraction at codon 61 having CAA → CTA mutation in liver and kidney was also determined [113]. Using microarray analysis, it was revealed that gene expression alterations in cancer-related pathways were more significant in kidney than in liver. The genes involved in the biological processes related to defense response, apoptosis, and immune response were significantly altered by aristolochic acid exposure in kidney, but not in liver [114]. Later, the same set of RNA samples were analyzed by NGS using an Illumina Genome Analyzer II [105]. When gene ranking was compared between NGS and microarray, the overlap of the differentially expressed genes was about 40%, because the dynamic range of the NGS platform is greater than that of the microarray technology. However, when the common gene ontology terms generated from the differentially expressed genes were compared for the two methods, the biological interpretation was largely consistent between the NGS and microarray data [105]. NGS technologies are capable of generating more data and providing additional insight into the mechanisms involved in toxicity and carcinogenesis associated with environmental mutagen/carcinogen exposure.

It is a challenge to determine the mechanisms of toxicity induced by mixtures containing many chemical components, such as diesel exhaust, herbal dietary supplements, and cigarette smoke condensates. Tobacco smoke contains more than 4000 chemicals, of which >200 chemicals are toxic to humans and >50 of them are recognized as known or probable human carcinogens [115, 116]. Tobacco smoking is a major public health problem that threatens the lives of one billion people and tobacco use is estimated to kill more than five-million people worldwide each year [117]. Using the NGS technology (i.e., the SOLiD platform), Pleasance and colleagues [107] sequenced a small cell lung cancer cell line NCI-H209 to investigate the mutational burden associated with tobacco smoking. Overall, 22,910 somatic substitutions were observed across the NCI-H209 genome, illustrating the power of NGS to detect the many complex mutational signatures associated with tobacco smoke-induced cancers [107]. Beane and associates [106] evaluated total RNA from bronchial airway epithelial cells of current smokers using both microarray and RNA-Seq methods and observed a significant correlation between the RNA-Seq gene expression data and Affymetrix microarray data generated from the same samples. The RNA-Seq technology detected more smoking- and cancer-related gene expression differences, and NGS data has the potential to provide greater detail for information concerning mutations that may provide insights into the biological changes associated with smoking and lung cancer [106].

ADVANCE OF NGS TECHNOLOGIES

NGS technologies and platforms have rapidly evolved over the past eight years with a number of innovations and developments in sequencing chemistries, microfabrication and imaging to increase speed, throughput, and accuracy, and decrease costs of sequencing [118, 119]. There are currently four companies including Illumina, Life Technologies, Roche/454, and Pacific Biosciences offering a variety of NGS platforms in the NGS market. Also, many new technologies with promising features, such as nanopore sequencing, are under development and it is likely that commercialization will take place in a few years [120]. Different NGS platforms may rely on dissimilar sequencing chemistries and techniques, but they all share the technical strategy of miniaturization and parallelization of individual sequencing chemical reactions to boost sequencing speed and throughput [121]. Roche/454 was the first to commercialize their NGS platforms, but Illumina, which developed the first short read sequencer, is currently dominating the NGS market. Each NGS technology and platform has its own pros and cons, and may fit different analysis applications. Table 3 provides a summary of the characteristics of commercially available and near-to-release NGS platforms. In this review, we will not discuss each NGS technology and related platforms in detail as most of them have been well-reviewed elsewhere [121, 122]. We will focus on recently developed technologies and small-scale, low-cost platforms including Pacific Biosciences PacBio RS II, Illumina MiSeq, Life Technologies IonTorrent, and Oxford nanopore sequencing.

Table 3.

Summary of Next-Generation Sequencing Platforms

Company	Sequencing reaction	Amplification	Sequencer	Read length (base)	Number of reads (M/run)	Run time (day)	Throughput (Gb/day)	Error type	Comments
Illumina	SBS* with RT‡	Bridge PCR±	HiSeq 1500/1000 high output	36–100	1500–3000	2–8.5	23–27	Substitution	Currently the most widely used platforms with uitra high throughput, without scalability as SOLiD 5500×1
			HiSeq 2500/2000 high output	36–100	3000–6000	2–11	49–55
			HiSeq 1500 rapid run	36–150	300–600	0.3–1.7	30–54
			HiSeq 2500 rapid run	36–150	600–1200	0.3–1.7	60–108
			HiScanSQ	35–100	750–1500	1.5–8.5	15–18		Dual platform: microarray and NGS Fast, long reads, automated workflow, highly scalable
			MiSeq	25–300	22–50	0.17–2.7	3.3–5.5
			GA-llx	35–150	320–640	2–14	5–6.8		Legacy platform
Life Technologies	SBL†	Emulsion PCR	SOLID	75/2×50	3200	1	20–45	Substitution	ECC-based error detection and correction, scalable runs for each lane on flow cell
			5500×1 W	MP/50×50 PE
	SBS with H + detection		Ion PGM 314 Chip	200/400	0.4–0.55	0.1–0.15	0.31–0.65	Indel	Label-free chemistry, cheap, run fast, highly scalable, long reads.
			Ion PGM 316 Chip	200/400	2–3	0.12–0.2	2.4–4.9
			Ion PGM 318 Chip	200/400	4–5	0.18–0.3	3.3–6.6
			Ion Proton	<200	60–80	0.08–0.17	60–120
Roche/454	Pyrosequencing	Emulsion PCR	GS Junior	~400	0.07/0.1	0.42	0.035		Long reads and short run time, high reagent cost and high error rate in homopolymer repeats
			GS FLX	~700	1	0.96	0.7	Indel
			Titanium XL+ GS FLX	~450	1	0.42	1
Pacific Bioscience	SBS, SMRTλ	None	Titanium XLR70 PacBio RS II	~5000	~0.05	<0.1	2.2		Long reads, single-molecular sequencing, less PCR bids
Oxford Nanopore Technologies Ltd.	Strand sequencing	None	MinlON					Deletion	Long reads, single-molecular DNA sequencing. RNA and protein sequencing are under development
			GridlON

^*SBS, sequencing by synthesis;

^‡RT, reversible terminators;

^±PCR, polymarase chain reaction;

^†SBL, sequencing by ligation;

^λSMRT, single molecule real time sequencing.

The Pacific Biosciences PacBio RS II is the latest upgrade of its first commercial NGS sequencer PacBio RS which was released in 2011. The new system produces longer reads and offers higher sequence throughput than the original instrument. But both platforms utilize the company’s zero-mode waveguide (ZMW) technology [123] to conduct single molecule real-time sequencing (SMRT) [124]. The sequencing by synthesis is conducted on a small plastic SMRT cell. Each SMRT cell contains thousands of ZMWs. A ZWM is a nanophotonic confinement structure that can be used to detect fluorescence signals from single nucleotide incorporation. In each ZMW, a single active DNA polymerase enzyme is immobilized with a single molecule of single-stranded DNA template. During sequencing by synthesis, four nucleotides labeled with different fluorescent dyes are supplied to the SMRT cell. When a nucleotide is incorporated, the fluorescent dye is cleaved off as a part of a natural DNA synthesis process. The fluorescent signal is detected and the base call is made in real time. The cleaved fluorescent dye molecule then diffuses out of the observation area of the ZMW. The SMRT sequencing can generate very long reads (average read length >4 kilobases [125] with relatively low throughput and has been used successfully for de novo genome sequencing [125], transcriptome sequencing [126], targeted resequencing [127, 128], methylation detection [129], and in vitro diagnostics [130].

The Illumina MiSeq employed the same technologies for sequencing as those used in the earlier HiSeq systems but is aimed at smaller laboratories and clinical diagnostic applications. In contrast to the HiSeq series of platforms, the MiSeq is a lower throughput but fast-turnaround instrument in which cluster generation, bridge PCR amplification, sequencing by synthesis with fluorescently labeled reversible-terminator nucleotides, and data analysis were integrated and streamlined without user intervention needed [131]. Such a design not only provides rapid, cost-effective NGS analysis but also offers much more flexibility for applications such as clone checking, amplicon sequencing, targeted transcript sequencing, small genome resequencing, de novo sequencing, and small RNA sequencing [132–134].

The Life Technologies Ion Torrent semiconductor sequencing is based on the detection of protons that are released when nucleotides are incorporated into growing DNA strands during sequencing-by-synthesis reactions. Libraries are constructed by the fragmentation of DNA or cDNA, and fragments are then linked to specific adapter sequences to form sequencing templates which are clonally amplified with emulsion PCR. The amplified products are subsequently loaded onto an Ion Torrent chip which contains millions of ion-sensitive field-effect transistor (ISFET) sensors that allow parallel detection of multiple sequencing reactions [135, 136]. Sequencing is primed from a specific position in adapter sequences and each of the four nucleotides is supplied sequentially. If a specific type of nucleotide is incorporated, protons are released and pH is altered and measured with an ISFET sensor. Although pH signal is proportional to the number of bases incorporated, the determination of homopolymer bases introduces major sequence errors as false insertions or deletions (Indels). Without the needs for modified nucleotides and optical signal detection, the Ion Torrent sequencing process is much faster and its instruments, chips, and reagents are less expensive than the other platforms. Recently, Life Technologies has released the Ion PGM sequencer with Ion 314/316/318 chips and the Ion Proton sequencer with Ion PI chips to provide cheap and rapid sequencing.

The Oxford Nanopore strand sequencing is based on the detection of electric current changes as a single DNA sequence strand passes through a nanopore, which is a nanoscale hole formed by proteins and set in an electrically resistant membrane bilayer [137]. A constant voltage is set across the bilayer membrane which creates an ionic current passing through nanopores. If a DNA strand passes through a nanopore, each of the four nucleotides induces a different type of electric current change which can be translated into sequence data. By monitoring electric current changes in real time, the base call can be made for a single DNA molecule. Oxford Nanopore developed two sequencing systems, that is, MiniION and GridION, and planned to commercialize at the end of 2013. The MiniION is a disposable self-contained device for real-time single molecule sensing sequencing and can be run directly with a laptop or desktop computer through a USB port. The GridION is a scalable system consisting of single or multiple nodes. Each GridION node works with a disposable, self-contained cartridge for sequencing experiments. Currently, the Nanopore DNA sequencing platform can be used to produce very long read-lengths with average lengths of 4200 to 8500 bases [138]. It also has the potential to be adapted for RNA, protein, and other polymer sequencing.

Advancements in NGS technologies have enabled a number of sequencing-based analytical approaches that were not affordable previously or even technically feasible. Over the past few years, NGS-based approaches have been widely investigated and implemented in a variety of biological and biomedical research settings including de novo genome sequencing [139], detection of genetic variation [140], quantification of gene expression [105], discovery and annotation of genes and transcripts [141], and analysis of gene regulation [142]. Recently, the release of low cost, smaller-scale, and easy-to-use NGS platforms such as MiSeq, Ion Torrent, and Nanopore sequencing, make NGS technologies and NGS-based approaches more accessible and will speed up the application of NGS technologies in clinical settings [120]. Table 4 provides a summary of NGS-based approaches and their potential applications.

Table 4.

Next-Generation Sequencing-Based Approaches and Their Applications

Sequencing Approaches	Strategy	Feature	Case Study
Whole genome sequencing (WGS)	Sequencing DNA fragments from a whole genome	SNVs, Indels, structural rearrangement, and copy number variations s can be obtained in a single assay with single-base resolution; high cost for data generation, storage and analysis, low coverage.	Whole genome resequencing [140], de novo genome sequencing, and assembly [139]
Targeted resequencing	Sequencing DNA fragments from captured genome regions	Low cost, easy to get deep coverage for exomes, specific genes, or other genomic regions of interest; Need isolation of genomic regions of interest, only targeted regions can be seen	Whole exome sequencing (WES) [218], a targeted panel of selected genes [219], amplicon sequencing [220]
RNA-Seq	Sequencing RNA/cDNA fragments	High dynamic range, enable to identify novel genes and transcripts, gene fusion events; High cost, lack of standards for data analysis	Gene expression quantification [105], transcriptome annotation [126], gene fusion detection [221], allele specific expression [222]
ChIP-Seq	Combination of chromatin immunoprecipitation (ChIP) with NGS	High resolution and sensitivity, but the cost is correlated with sensitivity and resolution	Exploring interactions between protein, DNA and RNA [135]
MethyIC-Seq (BS-Seq)	Sequencing DNA fragments from bisulphite treated genome	Genome-wide survey	Genome-wide methylation characterization [129]
MNase-Seq	Sequencing DNA fragments from micrococcal nuclease digestion	Genome-wide survey	Detection of nucleosome localization [223]
DNase I-Seq	Sequencing DNA fragments from DNase I digestion	Genome-wide survey	Detection of chromatin accessibility [224]

LIMITATIONS AND CHALLENGES IN NGS

Although NGS technologies are very powerful when applied appropriately to certain research and clinical applications, many investigations have revealed that there are significant technical drawbacks that limit the utility of NGS technologies in certain situations. Beyond the high cost of NGS instruments and reagents, the storage, management, analysis, and interpretation of NGS data are also posing challenges. For example, a study [143] in which NGS was tailored to detect RNA-editing events reported that an astonishingly large number of exonic mismatches were discovered between mRNA sequences and corresponding DNA sequences from the same individuals and showed that many of these RNA sequence changes were not produced by any known RNA regulatory mechanisms. This report questioned the basic assumption in molecular biology that DNA and RNA of the same individuals are nearly identical. However, this report has been criticized by several groups who reanalyzed the data and found that the majority of those mismatches found in the report can be explained either by systematic technical errors in NGS technology or by analytical artifacts [144–147]. Thus, it is crucial to analyze NGS raw sequencing data properly, such as FASTQ files, and translate them into final variant calls. In addition, the single-base resolution capability of NGS-based approaches in the detection of sequence variation can be compromised by the biases, artifacts, or errors from NGS-related protocols, platforms, sequencing depth, data analysis pipelines, and reference genomes or transcriptomes, resulting in both high rates of false positive and false negative results. Bearing this fact in mind will help avoid pitfalls in identifying true variants from various artifacts generated by NGS technologies.

APPLICATION OF NGS IN CANCER GENOMICS AND EPIGENOMICS

Advances in NGS technology made possible new approaches to cancer research. With much faster speed, higher throughput, and lower cost, the NGS platforms provide tools to identify the genetic mutations, epigenomic alterations and transcriptome changes occurring in cancer cells and to establish a framework for understanding the complexity and heterogeneity of cancers. Cancer genome sequencing is an endeavor to sequence homogeneous or heterogeneous groups of cancer cells to characterize DNA or RNA molecules for discovery of their mutations and gene expression variations. A careful cancer genome sequencing strategy not only performs sequencing analysis on primary cancer cells/tissues directly but also analyzes paired adjacent or distal normal tissue, as well as tumor-related fibroblast/stromal cells and metastatic tumor tissues. The strategy is not limited to DNA sequencing of the whole genome, but also includes exome sequencing or target gene sequencing, transcriptome sequencing or RNA-Seq, and sequencing analysis of the epigenome. Single base-resolution sequencing allows for the characterization of DNA or RNA molecules for the discovery of cellular aberrations including sequence variants, structural rearrangements (such as chromosomal translocation and gene fusion), and variations in gene expression, gene copy numbers, and epigenetic status [148].

Various databases are being developed to organize the massive amount of cancer genome sequence data being generated. The Cancer Gene Census (http://cancer.sanger.ac.uk/cancergenome/projects/census/) has documented 513 genes (as of December 2013) with a wide variety of mutations that are associated with cancers. These mutations include chromosomal translocations, gene fusions, gene amplifications, large deletions, frameshift mutations, and missense mutations. Approximately 90% of these are dominant mutations occurring in oncogenes, thus one allele is sufficient to activate an encoded protein (oncogene) to cause cancer; whereas 10% of these mutations occur in tumor suppressor genes and act in a recessive manner [2]. Launched in 2004, the COSMIC (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/) is an online database that curates somatic mutations found in human cancers from the scientific literature and large-scale experimental screens from the Cancer Genome Project (http://www.sanger.ac.uk/research/projects/cancergenome/). The COSMIC database presents complex phenotype-specific mutation data in a statistical manner. As of October 2013 (version 67), COSMIC has documented over 1,592,109 mutations, 9190 gene fusions, 7584 genomic rearrangements, and 422,314 copy number aberrations that affected 25,606 genes from 947,213 tumor samples. These findings were identified from 17,731 scientific papers. COSMIC has the capability to annotate somatic mutations across the whole genome. With data from large-scale systematic candidate gene screening, whole genome sequencing, and exome sequencing, COSMIC has curated genomic information from 7954 whole genomes. For example, by sequence analysis of coding exons of 21,416 protein coding genes and 1664 miRNAs in 100 primary breast cancer samples, somatic copy number changes and mutations in the coding exons of protein-coding genes were documented. To exclude inherited sequence variation, paired normal DNAs from the same individuals were also sequenced. In such a way, multiple mutational signatures and driver mutations were identified in at least 40 cancer genes. Novel cancer genes were also identified and included AKT2, ARID1B, CASP8, CDKN1B, MAP3K1, MAP3K13, NCOR1, SMARCD1, and TBX3, demonstrating the genetic diversity, heterogeneity, and complexity of primary breast cancer [149].

Sponsored by the National Cancer Institute and the National Human Genome Research Institute, The Cancer Genome Atlas is a project aimed at cataloguing genetic mutations in 20–25 major cancer types, including glioblastoma, breast, colorectal, stomach, ovarian, prostate, and lung, using high-throughput genome sequencing technology and bioinformatics approaches (http://cancergenome.nih.gov/). The definition of mutations in the various cancers will allow a better understanding of the biology and pathology of cancer, leading to improved cancer diagnoses, treatments, and prevention. Identification of mutations in glioblastoma, ovarian, colorectal, lung, head and neck, multiple myeloma, chronic lymphocytic leukemia, large B-cell lymphoma, and many other cancers have been accomplished. For instance, Lawrence and colleagues reported a great heterogeneity of mutations in cancer-associated genes [150]. Among 27 cancer types, with a sequencing data set of 3083 tumor–normal pairs, including 2957 pairs by whole-exome sequencing and 126 pairs by whole-genome sequencing, a total of 373,909 non-silent coding mutations were found. The average mutation frequency in the exome was approximately 4.0/Mb per sample, and the median mutation frequency was 44 nonsilent coding mutations per sample, or 1.5/Mb per sample. Furthermore, across different cancer types and among different patients with the same cancer type, the exome mutation frequency and spectrum were extraordinarily variable. The lowest mutation frequencies, as low as 0.1/Mb (approximately one change across the entire exome) were found in pediatric cancers. In contrast, the frequency of mutations in the exomes of melanoma and lung cancer were more than 100/Mb, more than 1000-fold higher than in pediatric cancers. Notably, the highest mutation frequencies appear to be associated with extensive exposure to environmental carcinogens, such as ultraviolet radiation in the melanoma patients, and tobacco smoke in lung cancer patients [150]. In addition, among lung cancer patients, smokers have 10 times more somatic mutations than nonsmokers, indicating the involvement of environmental carcinogens in the pathogenesis of lung cancer [151].

Based on the large international efforts described, as well as others, it has become possible to start defining a comprehensive cancer genome landscape that will aid our understanding of this complex disease process and help in the diagnosis, treatment, and prevention of cancer. This cancer genome landscape for common types of human cancer has revealed previously unanticipated complexities in the patterns of somatic alterations in cancers [152, 153]. The landscape of common human cancer suggests that approximately 140 genes are frequently altered in many types of cancer. The altered genes conferring a selective growth advantage are defined as driver genes and the mutations in these genes are driver mutations. Driver mutations contribute greatly to individuals’ cancer susceptibility. Owing to the knowledge provided by NGS technology, it is realized that a cancer can be initiated by 2–8 driver mutations in driver genes [152, 153].

While whole-exome sequencing has been widely used in cancer genome analysis, this strategy is unable to efficiently detect gene fusions, which are important drivers of several types of blood cancers. Sequencing of the transcriptome (RNA-Seq), however, is a powerful tool to detect fusion genes, in addition to somatic mutations, novel transcripts and alternative splicing variants expressed in tumor cells. Furthermore, it can be used to analyze gene expression profiles at a low cost and high sensitivity [154]. RNA-Seq has been used to identify gene fusion events in many cancer types, such as breast cancer [155], lung cancer [156], prostate cancer [157], and glioblastoma [158, 159]. For example, using RNA-Seq data, Shah et al [159] were able to identify gene fusion events in approximately 30%–50% of 185 glioblastoma multiforme tumors. The majority of fusions were located at the chromosomal 7p11 and 12q14–15 regions. Further analysis demonstrated complex genomic rearrangements in these affected chromosomal regions that result in aberrant EGFR gene amplification of a tyrosine kinase domain, suggesting the importance of gene fusion in glioblastoma. On the other hand, the use of RNA-Seq to profile gene expression patterns in tumor-normal pairs may be able to provide a more comprehensive view that comprises the overall outcome resulting from the interaction between an individual’s variable genetic makeups, the epigenetic regulation, environmental exposures, and mRNA splicing patterns. For example, by comparing of RNA-Seq profiles between of 17 breast cancer samples and matched normal breast tissues, Horvath and colleagues [160] identified a total of 4847 genetic and functional novel variations, including mutations, mRNA splicing changes, and alterations of gene expression. Their results indicated that the most prevalent novel variants associated with cancer related genes included ESRP2, GBP1, TPP1, MAD2L1BP, GLUD2, and SLC30A8, which are pivotal in breast cancer tumorigenesis [160].

NGS technology is also being used for surveying DNA methylation, mapping of transcription factor binding sites, and measuring histone binding sites [161, 162]. NGS can be applied to obtain the whole genome DNA methylation status at base-resolution level through bisulphite sequencing [163]. Genome-wide transcription factor binding sites can be analyzed using chromatin immunoprecipitation-mediated massively parallel DNA sequencing (ChIP-seq) [164]. Similarly, ChIP-seq can be used to reveal genome-wide histone modifications. These methods are advanced in comparison to array-based interrogation for discovery, since they make no a priori assumptions of where to interrogate the genome. As a high-throughput tool, NGS has added a new avenue to revolutionize our appreciation of how chromatin and DNA methylation are involved in the pathology of tumorigenesis.

Gene mutations that may drive cancer-relevant changes in the epigenome have been found by NGS technology. For example, in a comprehensive study, using whole-genome sequencing or whole-exome sequencing, combined with RNA-Seq, miRNA sequencing and DNA-methylation profiling, the Cancer Genome Atlas Research Network analyzed genomes from 200 cases of de novo acute myeloid leukemia (AML) [165]. DNA methylation changes were identified at 160,519 CpG loci across the AML genomes, which is 42% of sites tested. Among these methylation-altered loci, 67% of them had increased methylation, and 33% of them had decreased methylation. Further analysis indicated that some genetic mutations were highly associated with DNA methylation status. A significant decrease of DNA methylation was associated with MLL fusions or co-occurring NPM1, DNMT3A, and FLT3 mutations. In a small portion of samples (7/200), genetic variants were identified in some miRNA genes. The miR-142 gene was highly expressed in AML samples, indicating significant epigenomic alterations in AML [165]. In another study, genome-wide methylated CpG islands were profiled by NGS. It was found that the regulatory regions of many genes are hypermethylated and several repeated elements are hypomethylated in melanoma cell lines. By integrative analysis of DNA methylation data with RNA-Seq data, a gene co-expression network significantly related to melanoma was determined, which could partially explain the molecular pathogenesis of the disease [166].

FUTURE PERSPECTIVES

Multiple factors such as genetic variants, environmental exposure, and lifestyle are all involved in cancer etiology. Genome-wide association studies (GWAS) based on the hypothesis “common disease-common variants” have identified many susceptibility loci in the human genome for a variety of cancers using genotyping microarray technologies. The first wave of GWAS was successful in understanding cancer etiology. But the findings have not fulfilled the expectations of the scientific community because the cancer-associated genetic variants only explained a very small portion of cancer risk. As described in this article, research efforts to understand cancer susceptibility have been shifting to ascertaining genetic-environmental interactions and to identifying rare genetic variants with high penetrance as well as somatic variants. These new efforts have been made possible by comprehensive, rapid and relatively low-cost next-generation sequencing technologies. Significant progress in the following areas is expected to be achieved in the next few years.

Additional rare genetic variants with high penetrance will be identified that confer a portion of total cancer susceptibility, maybe larger than that explained by the common genetic variants that have been identified from GWAS. NGS technologies, unlike genotyping microarrays, enable interrogating all genetic variants (known and unknown, common and rare) by sequencing the whole genomes of cancer patients. However, challenges to distinguish the causal variants from the large number of apparently novel genetic variants present by chance in any human genome will need to be solved before true genetic variants can be identified.

Somatic mutations resulting from the interaction of genetic and environmental factors may be involved in the development of cancers by conferring growth advantages to cancer cells. Genotyping microarray technologies are hypothesis based, that is the genetic mutations that can be detected are known before they can be designed in the microarrays. Some, if not most, somatic mutations that confer cancer susceptibility may not be presently known so they will not be used in the microarray design. However, detection of genetic variants by NGS is hypothesis-free and, in principle, all somatic mutations in cancer patient genomes can be detected by NGS. Therefore, more somatic mutations that contribute to cancer risk are expected to be identified in future cancer genetic studies using NGS.

Understanding interactions between genetic and environmental factors will provide insight into the etiology of cancers, as well as susceptibilities, possibly allowing the prevention of cancers. Environmental factors can affect the expression of the genome through alterations in the epigenome. Application of NGS technology to the analysis of DNA methylation, chromatin modification, and ncRNA will allow better definition of the role of the environment in cancer development. Coupling the genomic and epigenomic information may ultimately give the insights needed to reduce the human costs of this prominent disease.

To better understand cancer etiology, not only NGS data but also clinical data, lifestyle records, healthcare records, genealogy records and other types of information need to be analyzed and integrated. Analysis and integration of these data, as well as interpretation of the results, are the key steps for applying NGS technologies in ascertaining cancer susceptibility and translating this knowledge into public health actions. In the near future, powerful infrastructures will be developed to support the storage, access, analysis and management of the huge amount and diversity of data. Powerful bioinformatics tools will be developed for identification of genetic variants and genetic-environmental interactions that contribute to cancer risk.

With the expected progress described, we can look forward to a better understanding of cancer susceptibility that will lead to better diagnosis, treatments, and ultimately prevention.