Perspectives

Precancer Atlas to Drive Precision Prevention TrialsAvrum Spira1,2, Matthew B. Yurgelun3, Ludmil Alexandrov4, Anjana Rao5,Rafael Bejar6, Kornelia Polyak3, Marios Giannakis3, Ali Shilatifard7, Olivera J. Finn8,Madhav Dhodapkar9, Neil E. Kay10, Esteban Braggio11, Eduardo Vilar12,Sarah A. Mazzilli1,2, Timothy R. Rebbeck13, Judy E. Garber3,Victor E.Velculescu14,15,Mary L. Disis16, Douglas C.Wallace17,18, and Scott M. Lippman6

ABSTRACT Cancer development is a complex process drivenby inherited and acquiredmolecularand cellular alterations. Prevention is the holy grail of cancer elimination, butmaking

this a reality will take a fundamental rethinking and deep understanding of premalignant biology. In thisPerspective, we propose a large-scale concerted effort to create a Precancer Atlas (PCA), integrating multi-omics and immunity – basic tenets of the neoplastic process. Study of the biology of neoplasia caused bygermline mutations has led to new paradigm-changing therapy and precision prevention efforts, includingimmune interception and early detection research. Breakthrough advances include: studies of luminalprogenitor biology in BRCA1-mutation carriers leading to an international breast cancer prevention trial,combinatorial chemoprevention efficacy in familial adenomatous polyposis (FAP), signal of benefit fromimaging-based early detection research in individuals at high-germline risk for pancreatic neoplasia, andinsights into the complex germline-somatic-immunity interaction landscape, e.g., germline APOBEC3A/3B-deletion allele confers predisposition to breast cancer, while surprisingly increasing somatic APOBEC-mutational burden and innate immunity to prevent infection. Cancer vaccine preventive activity in earlymouse model and clinical precancer trials are building on the success of HPV vaccines. The increasinglycomplex cellular microenvironment (e.g., now including adipocytes, myocytes, neutrophils, fibroblasts,vascular and neuronal cells, microbiota, and even normal epithelial cells), interplay between defective DNArepair, mitophagy, mitochondrial dynamics, and ROS-dependent and TLR-induced inflammasome acti-vation, and relationships between somatic copy number alterations (SCNA), mutations and immunity areall recent findings with major implications on malignant transformation and prevention. Next generationsequencing (NGS) studies discovered premalignantmutational accumulation in aging, established in clonalhematopoiesis, and cancer mutational signatures reflecting exogenous exposures or endogenous processesimprinted over time in precursors. Single-cell technologies are beginning to reveal vast molecular hetero-geneity within precancers (e.g., in ductal carcinoma in situ [DCIS]and Barrett's esophagus) and will beneeded to dissect the detailed cellular interactions within the neoplastic microenvironment. Preclinicalmodels will allow controlled study of many emerging complexities in this field, including lineage-specificregulation of stem/progenitor cell fate. Accelerating the prevention of cancer will require a large-scale,longitudinal effort, leveraging diverse disciplines (from genetics and biochemistry to mathematicsand engineering), initiatives, technologies, and models in developing an integrated multi-omics andimmunity PCA – an immense national resource to interrogate, target and intercept events that driveoncogenesis. Cancer Res; 77(7); 2–28. �2017 AACR.

Perspective

Cancer Res; 77(7) April 1, 20172

IntroductionIn early 2016, President Obama announced the creation of the

National Cancer Moonshot initiative – a commitment to eradi-cate cancer by investing in greatly accelerating progress withindiverse fields of cancer research, including prevention and earlydetection. Funding for this effort (and other cancer-related inno-vative research initiatives, e.g., stem cell research) was signed intolaw by the President inDecember 2016 as part of the 21st CenturyCures Act (1). Historically, the rate-limiting step for developingand implementing precision prevention approaches has been ourlimited understanding of precancer biology in contrast to theextensive study of advanced disease. For example, The CancerGenome Atlas (TCGA), which includes volumes of omics data

from>11,000patients across 33 tumor types (including colorectalcancer [CRC]), has transformed our understanding of cancerbiology, identified hundreds of driver mutations that alter hall-mark pathways, immense heterogeneity within and betweentumors, and become a tremendous national resource for discov-ery, catalyzing the development of novel computational tools. Incontrast, although the seminal multi-step, genetic model ofcarcinogenesis was defined in the colorectal adenoma-carcinomasequence nearly 30 years ago (2), limited numbers of adenomashave undergone deep sequencing (3). The interaction betweenhost immunity andneoplasia is nowestablished as a fundamentalprinciple of cancer development and progression. A diverse arrayof engineered models, single-cell technologies, and broad dis-ciplines are beginning to be leveraged to probe early oncogenesis

1Department of Medicine, Boston University School of Medicine, Boston,Massachusetts. 2Department of Pathology and Bioinformatics, Boston Uni-versity School of Medicine, Boston, Massachusetts. 3Department of MedicalOncology, Dana-Farber Cancer Institute, Boston, Massachusetts. 4Theoret-ical Division, Center for Nonlinear Studies, Los Alamos National Laboratory,Los Alamos, New Mexico. 5Division of Signaling and Gene Expression, LaJolla Institute for Allergy and Immunology, La Jolla, California. 6Departmentof Medicine, Moores Cancer Center, University of California San Diego, LaJolla, California. 7Department of Biochemistry and Molecular Genetics,Northwestern University Feinberg School of Medicine, Chicago, Illinois.8Department of Immunology, University of Pittsburgh, Pittsburgh, Penn-sylvania. 9Department of Hematology and Immunology, Yale Cancer Center,New Haven, Connecticut. 10Department of Hematology, Mayo Clinic Hos-pital, Rochester, Minnesota. 11Department of Hematology, Mayo ClinicHospital, Phoenix, Arizona. 12Department of Clinical Cancer Prevention,The University of Texas MD Anderson Cancer Center, Houston, Texas.13Division of Hematology and Oncology, Dana-Farber Cancer Institute andHarvard T.H. Chan School of Public Health, Boston, Massachusetts.

14Department of Oncology, Sidney Kimmel Comprehensive Cancer Centerat Johns Hopkins, Baltimore, Maryland. 15Department of Pathology, SidneyKimmel Comprehensive Cancer Center at Johns Hopkins, Baltimore, Mary-land. 16Department of Medicine, Center for Translational Medicine inWomen's Health, University of Washington, Seattle, Washington. 17Centerfor Mitochondrial and Epigenomic Medicine, Children's Hospital of Phila-delphia, University of Pennsylvania, Philadelphia, Pennsylvania. 18Depart-ment of Pathology and Laboratory Medicine, Perelman School of Medicine,University of Pennsylvania, Philadelphia, Pennsylvania.

Note: A. Spira and M.B. Yurgelun contributed equally to this article.

Corresponding Author: Scott M. Lippman, UC San Diego Moores CancerCenter, La Jolla, CA 92093. Phone: 858-822-1222; Fax: 858-822-0207; E-mail:slippman@ucsd.edu

doi: 10.1158/0008-5472.CAN-16-2346

�2017 American Association for Cancer Research.

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017

and malignant transformation. Large-scale longitudinal and sys-tematic mapping is critical to develop an integrated omics andimmune PCA, allowing dissection of the sequential molecularand cellular events that promote oncogenesis, to drive novelprevention and interception (4–6) (Fig. 1).

Translating Inherited Risk into PrecisionPreventionGermline cancer susceptibility

The genetics of various hereditary forms of cancer risk havebeen studied extensively and long been used to aid our under-standing of sporadic neoplasia. Historically, most cancer suscep-tibility genes were identified through linkage studies in cancer

families (7). Furthermore, the biology of germline mutations incertain cancer genes (e.g., BRCA1 carriers) is much better under-stood than in the somatic setting and is now providing tremen-dous insight into personalized therapeutic and preventive strat-egies by leveraging the biology of the mutated gene itself. Forexample, study of the biology of tumors that develop in BRCA1/2-mutation carriers has led to paradigm-changing precision therapywith PARP inhibitors (8), which have demonstrated preventiveactivity in BRCA1-deficient mice (9). Similarly Lynch syndrome isserving as a model of immune oncology for sporadic MSI-Htumors (10, 11). Furthermore, understanding the convergenceof Wnt and EGFR signaling in FAP biology, characterized bygermline APC mutations, led to a breakthrough trial of combi-nation chemoprevention with erlotinib and sulindac, which

Immune Surveillance

VEGF, Cytokines(e.g., IL-6, IL-8, IL-17)

Innate and Adaptive Immune Cells:

Mutations and Neoantigens

Viral Antigens

Post-Translational

• Glycosylation

• Phosphorylation

• Splice Variants

LEGENDNaive lymphocytes

Activated CD8+ Cytotoxic T- cells

Exhausted CD8+ Cytotoxic T-cells

Regulatory T-cells

M1 Macrophages

M2 Macrophages

Dendritic Cells

Myeloid Derived Suppressor Cells

Natural Killer Cells

Adipocytes

Neutrophils

Immune Suppression/ EscapeImmune Suppressing Cells:

• Tumor Associated Macrophages (TAMs/M2)

• Myeloid-Derived Suppressor Cells (MDSCs)

• Regulatory T-cells (Tregs)

• Adipocytes, Neutrophils, Fibroblasts

Immune Evasion Markers

• Checkpoint Inhibition of T-cells

• Somatic Copy Number Alterations

INFECTION

ImmunityAtlas

HPV, HBV, EBV

NORMAL PRECANCER

Multi-omics Atlas

CANCER

GERMLINEBRCA1, MMRD, APC

SOMATICKRAS, TET2, ND5

Figure 1.

An integrated multi-omics and immunity precancer atlas (PCA). Inherited and acquired molecular alterations and their interaction with the local microenvironmentand immune system influence oncogenic progression to invasive carcinoma. Normal cells (light orange, far left) that have nuclear or mitochondrial germlinemutations (green nuclei) acquire somatic alterations (dark orange), including due to viral infection (purple). These all can potentially alter the oncogenic state withloss of cell growth control, immune evasion and other hallmarks of cancer (7), which can then result in the development of advanced precancers (multicoloredcell mass with subclonal diversity/heterogeneity) that immediately precede invasive cancer (red). Molecular alterations, depicted by symbols in the nucleus,include mutations, SNPs, or epigenetic alterations. The accumulation of mutations (e.g., UV, ABOBEC) during life creates signatures (shown by the chromosomalinsets and colored dots in gradient from cancer to normal). These mutational signatures, often identified in cancers, may predict early events, now extending tothe precancer state (depicted by the orange/red cells with bold, yellow border). Multi-omic alterations interact (bidirectionally) with the complex tissuemicroenvironment, including the well-established immune cells (tumor-associated macrophages, MDSCs, NK and cytotoxic/regulatory T-cells) and more recentlyidentified and less well understood cells in this context (e.g., adipocytes, myocytes, neutrophils, fibroblasts; and vascular, endothelial, neuronal, B- and othercells), microbiota and other cells/events to influence oncogenesis (see text). The continuum between immune state modulated by cytokines and growth factorsincludes immune surveillance, composed of the antigenic repertoire, adaptive and immune cells (upper left box), and immune suppression/escape (upper right box)along with the cells and markers that can lead to immune escape. Elucidation of the integrated multi-omics and immunity PCA will continue to evolve in complexity(e.g., recent SCNA/immune suppression finding), requiring continuous updating in this fast moving field.

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research4

successfully reduces duodenal neoplasia, the leading cause ofcancer death following standard prophylactic colectomy for thisdevastating syndrome (12).

Multigene NGS is broadening the spectrum and complexityof cancer risk linked to various hereditary syndromes, fre-quently identifying individuals with high-penetrance germlinemutations that are unexpected based on clinical history (e.g., CRCin BRCA1/2- and TP53-mutation carriers and breast cancerpatients with Lynch syndrome) (13–15). Furthermore, broadNGS identified germline mutations in cancer-predisposing genesin �10% of >1100 pediatric cancer patients, most of whom lacksuggestive family histories of cancer (16).

A fundamental question regarding inherited cancer suscepti-bility is why germline mutations in certain genes predispose to aparticular spectrum of cancers. Although some genes have organ-specific effects (e.g., mutations leading to hepatic overload whichpredispose to liver cancer),most cancer susceptibility genes have abroad range of functions (e.g., cell-cycle regulation). It is unclear,therefore, why different inherited DNA repair defects predisposeto specific cancers (e.g., DNAmismatch repair [MMR] gene defectsin Lynch syndrome to CRC and endometrial cancer; BRCA1/2mutations [e.g., homologous recombination deficiency] to breastand ovarian cancer) rather than generalized systemic cancer risk,or why specific MMR and BRCA2 mutations are associated withdifferent cancer patterns (17, 18). For example, BRCA2 loss-of-function mutations confer increased risks of breast and ovariancancer (but only small increases in prostate and pancreatic can-cer),mutations in the central part of the gene confer higher risks ofovarian (vs. breast) cancer, and risks differ by context (e.g., familyhistory, other modifying factors). Mouse models of Lynch syn-drome have begun to unravel key mechanisms of the CRCpredisposition by finding that butyrate, generated by gut micro-biota from dietary carbohydrates, can act as an oncometabolite(19). Interestingly, butyrate can have the opposite effects indifferent CRC models, likely reflecting different germline back-grounds, e.g., reduced dietary butyrate markedly decreased CRCdevelopment in MMR-deficient, but not MMR-proficient mice.Studies in APCMin/þ mice have found that celecoxib (a COX-2inhibitor known to reduce intestinal adenoma burden) increasedgut Coriobacteriaceae, which suppressed production of oncogenicmetabolites (e.g., glycine and serine) (20).

The mechanisms underlying the breast-specific predominanceseen in BRCA1/2-mutation carriers are not clear but may behormonally driven, with impaired local steroid-hormone regu-lation potentiating the BRCA1/2mutagenic effect (21). Defectiveovarian hormone regulation and signaling, DNA repair, andBRCA cross talk in normal breast tissue from BRCA1-mutationcarriers may also contribute to neoplasia at this site (22). BRCA1-haploinsufficient non-cancer cells have differences in gene dos-age, DNA damage repair, and lineage commitment that havebegun to shed light on tissue specificity (23); mutations in asingle BRCA1 allele are sufficient to alter non-neoplastic breastepithelial and mammary progenitor cells. Direct cell-type com-parisons showed thatDNAdamage response, genomic instability,telomere erosion, and premature senescence are all increased inprimary breast epithelial (but not fibroblast) cells with BRCA1-haploinsufficiency. BRCA mutations have been shown to induceperturbations in telomere-telomerase system,which contribute toneoplasia, based mainly on data from cancer cells. Recent studiesextended this work to BRCA1-silenced nonmalignant breast epi-thelial cells and revealed profound alterations in telomere

homeostasis, involving decreased DNMT1 levels and TIN2 telo-mere binding and increased telomere trimming, shortening ofsingle-stranded overhang, telomere length heterogeneity, and losson chromosomal ends (24). Exposure to estrogen is central tobreast cancer development; therefore, the identification of ER-a asa putative BRCA1/BARD1 ubiquitination substrate reveals apotential link between loss of this ubiquitin ligase activity andbreast cancer (25). We have only recently begun to uncover theearliest steps of BRCA1-mutant breast neoplasia. Regulators ofmammary epithelial cell identity are crucial to cell fate determi-nation and switching, and manipulation of mammary lineageregulators can affect progenitor cell fate decisions, placing pre-malignant cells on a path towards specific cancer phenotypes. Inthe case of BRCA1-mutation carriers, aberrant expression of thetranscription factor Slug was shown to be a cause for cell fateswitching. Identifying additional epigenetic and transcriptionalregulators of cell fate and linage transitioning will provide deeperinsight into this complex process and interception targets (26).

Fanconi Anemia (FA), a rare hereditary syndrome related toBRCA1/2-mutation carriers, involves 19 genes (including BRCA2/FANCD1 and BRCA1/FANCS) and is characterized by a geneticdefect in double-strand DNA repair pathway. FANCC and certainother FA pathway genes are linked to host innate anti-viralimmunity. Heterogeneous immune defects, including B-cell dys-function, may account for susceptibility to HPV (and other viral)infections andvariable immune responses toHPVvaccines,whichhas implications for preventive cancer vaccine development.Recent study elucidated the mechanistic interplay of defectivemitophagy, mitochondrial dynamics, and ROS-dependent andTLR-induced inflammasome activationmediating IL-1b secretion(27). Somatic mutations or epigenetic silencing of FA genes iscommon in sporadic breast, ovary, and pancreatic cancer. Failureto remove damaged mitochondria can lead to increased mt-ROS,genotoxic stress and tumorigenesis in FA, and sporadic settingwith somatic FA driver mutations (28).

While the majority of breast cancers arising in BRCA2-muta-tion carriers are ER positive, most BRCA1 breast cancers are ERnegative at diagnosis. Large randomized controlled trials oftamoxifen and other anti-estrogens consistently show preventivebenefit confined to ER-positive breast cancer, although obser-vational studies suggest that tamoxifen can reduce breast cancerrisk, especially contralateral breast cancer, in BRCA1-mutationcarriers (29). Potential mechanisms for the latter include: BRCA1carriers have augmented estrogenic signals (21), breast cancerrisk associated with SNPs located near ESR1 (which encodesERa) (30) and estrogen and anti-estrogen (including tamoxifen)effects on proliferation in normal breast epithelial tissue fromBRCA1 carriers, similar to non-carriers at increased risk andpopulation risk of breast cancer (31). Brca1 loss with estrogensignaling promotes tumor development in mice (32). Estrogencontrols the survival of BRCA1-deficient mammary epithelialcells via a PI3K-NRF2-regulated pathway (33). Estrogen deple-tion also augments innate and adaptive immunity (34), e.g., bydecreasing mobilization and immunosuppressive activity ofmyeloid-derived suppressor cells (MDSC) (35, 36). These andother studies indicate important estrogen/hormonal effects inearly ontogeny. Indeed, BRCA1-mutant mammary stem andprogenitor cells (which lack ERa and PR) appear to be indirectlyactivated by steroid hormones via paracrine signaling mediatedby RANK ligand (RANKL), which is reduced by tamoxifen (36),see below.

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 5

Elegant studies of luminal progenitor biology (37, 38) have ledto a transformative potential to prevent/delay BRCA1-associatedbreast cancer, a disease for which the best current preventiveoption is prophylactic surgery. A highly proliferative subset ofluminal progenitor cells that give rise to basal-like breast cancer,constitutively express RANK and are hyper-responsive to RANK-L,a key mediator of progestin-driven mammary carcinogenesis.Furthermore, RANKþ BRCA1-deficient/mutant progenitor cellsare more susceptible to DNA damage and aberrant downstreamNFkB activation than RANK- mammary progenitor cells (39).Pharmacologic RANK-L inhibition or RANK deletion (in mam-mary epithelium) in several GEMMs markedly inhibits Brca1-driven mammary tumorigenesis (37, 38). High-level RANK-Ldrives epithelial-mesenchymal transition with invasion and cor-relates with high-grade ER/PR negative disease and mammarystem cell number (in young breast cancer patients) (40). RANK-L/RANK signaling (41) also can influence innate and adaptiveimmunity (42, 43). RANK-L produced by regulatory T-cells (Treg)promotes mammary cancer and T-cell tolerance to intestinalbacteria (43). Finally, NFkB hyperactivation in BRCA-mutantbreast cancer cells appears to be associated with an immunesignature (36). Recently, a second RANK-L receptor, LGR4, hasbeen implicated in the regulation of multiple developmentalpathways (44). Serum levels of osteoprotegerin, the endogenousRANK-L inhibitor, are significantly lower in BRCA1 carriers (vs.controls), premenopausal women, and are inversely associatedwith breast cancer risk (45, 46). Certain SNPs in the TNFRSF11Alocus have been associated with increased RANK expression andbreast cancer risk in BRCA1 carriers (38). Denosumab (a RANK-LmAb inhibitor FDA-approved in 2010 with a well-establishedsafety record for treating osteoporosis and bone metastases,espousing both efficacy and toxicity superior to biphosphonates(47)) blocked progesterone-induced proliferation in BRCA1-mutant organoids and reduced breast epithelial cell proliferationand progenitor cell clonogenic potential in small pilot windowstudies of BRCA1 carriers (37, 38). Theoretically, RANK-L inhi-bition would work best to prevent/delay tumor onset for pre-menopausal BRCA1-mutation carriers based on the above science

and recent data finding risk-reducing salpingo-oophorectomy tobe ineffective in this setting (48). Based on these data, denosumabis in late-stage development for a large-scale international breastcancer prevention trial in BRCA-mutation carriers (37, 38).

The first real benefit of early detection research in pancreaticcancer, imaging people with high-penetrance mutations, wasrecently reported (49). Unselected patients may have a very high(>15% in one clinic-based cohort) prevalence of germline muta-tions (particularly of BRCA1/2 and among Ashkenazi Jewishindividuals) (50). Precursor lesions in people with high-pene-trance germline mutations have a higher malignant potential(than other pancreas high-risk groups) (51, 52) – leading somecenters to recommend germline testing for all new pancreaticcancer patients. Precedent for such an approach already exists inNCCN guidelines for ovarian cancer, where a substantial fractionof unselected patients will carry BRCA1/2 mutations (most with-out consistent clinical/family histories) and screening has limitedbenefit (53). The development of molecular imaging techniquesto detect high-grade lesions (PanIN-3) may further improveprevention and early detection of this fatal disease (54).

Immunologic mechanisms, including cancer vaccines, mayalso be key to realizing precision prevention in certain types ofneoplasia characterized by immunogenic antigen production, asseen in various inherited settings (Table 1). Cancers that arise inLynch syndrome with inherited DNA MMR gene defects displaya high-level of microsatellite instability (MSI-H) and widespreadaccumulation of somatic frameshift mutations/neoantigens(55), thought to underlie the success of immune checkpointinhibitors in this setting (10, 11). These breakthrough advancesin immunotherapy along with early immunosurveillance (T-cellsspecific to MSI-related neoantigens (56)) in "healthy" Lynchsyndrome carriers have generated interest in developing preven-tive cancer vaccines targeting predictable frame-shift mutation-derived peptides.

Universal CRC tumor testing for MMR deficiency to screen forLynch syndrome, a paradigm-changing approach for identifyinginherited cancer risk, has become standard practice for all newlydiagnosed CRCs and is now being extended to endometrial

Table 1. Immunologic features of three distinct germline neoplastic pathways

Germline mechanism ofhypermutability

Characteristics ofcancers

Somatic mutations,signatures in cancers

Evidence of immunogenicphenotype Precancer immune biologyy

DNA MMR mutations: MLH1,MSH2, MSH6, PMS2; BMMR-D

CRC, endometrial, andothers with MSI-H;pediatric CRC/adenomas,gliomas, lymphomas inBMMR-D

Hotspot microsatellites indriver genes, e.g.,TGFBR2, BAX in LS; POLE,POLD1 in BMMR-D;biallelic MSH3

MSI-induced FSPs found inassociated cancers; PD-1inhibitors active in LS,BMMR-D, sporadic MSI-H

MSI-H in precancers, e.g.,adenomas/crypt foci,IPMNs, DCIS; microbiota;FSP T-cell immunity

BRCA1/2 mutations�

(HR deficiency)Breast, ovarian, pancreatic,prostate, and othercancers

Specific signatures oftandem duplications and/or deletions

Markers of immunity inBRCA1/2: breast, ovary,and pancreatic cancers

LPs, RANKL/RANK, NFkB,immunity; wild-type loss insome precancers

APOBEC3A/B chimeric deletionpolymorphism��

Modest increased breastcancer risk, somatichypermutation

APOBEC-mutationalburden; hotspots withindriver genes, e.g., PIK3CA

Upregulation of cytokineresponse, immunitygenes; penetrance ofimmune effects appearshigh

APOBEC mutagenesis insporadic pre-invasivebladder carcinoma and CIN;innate immunity to infection

NOTE: Immune biology of precancers in three distinct inherited neoplastic pathways: high-penetrance MMR deficiency and BRCA1/2mutations, and low penetrancepolymorphisms of APOBEC3A/B. Cancers associated with all three pathways are associated with distinct, predictable forms of somatic hypermutability and havevarious features suggesting an immunogenic phenotype. The PCA will be key to devising immune interception (see text).Abbreviations: BMMR-D, biallelic mismatch repair deficiency; CIN, cervical intraepithelial neoplasia; CRC, colorectal cancer; FSP, frameshift peptide; HR, homologousrecombination; IPMN, intraductal papillary mucinous neoplasms; LP, luminal progenitor; LS, Lynch syndrome; MMR, mismatch repair; (high level) MSI, microsatelliteinstability.�Fanconi anemia pathway (including BRCA1/2) defects associated with immunity. Estrogen depletion augments innate, adaptive immunity.��Increased stability of chimeric APOBEC3A/B mRNA leads to increased APOBEC3A activity correlating with germline copy number.

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research6

cancers (57). This benefits both the patient and at-risk familymembers for intensive and early screening and aspirin or poten-tially other NSAID prevention (58, 59). The profound activity ofimmune checkpoint blockade in MMR deficient tumors (11) hasadded to the enthusiasm of this universal screening approach.This tumor testing approach is under study in lung cancer patients:the �1% with tumor EGFR T790M mutation at diagnosis have ahigh risk of carrying germline T790Mmutations (60). Large-scalepan-TCGA analysis of tumor sequencing data (>4,000 tumors, 12cancer types) revealed rare germline mutations (e.g., BRCA1/2,FANCM, MSH6) in 4–19% of cancer types, unselected for familyhistory and associated with increased somatic mutation frequen-cies (61).Germline variants and somatic events are also intricatelylinked, including specific haplotypes with JAK2 V617F in mye-loproliferative neoplasms (62) and EGFR exon 19microdeletionsin non-small cell lung cancer (63). Germline variation was alsofound to influence gene expression in breast cancer risk loci (64).

Recent large-scale approach (22 tumor types, 5954 tumors)integrated common genotypes (412 germline loci) with somaticchanges to obtain a pan-cancer view of how inherited variationcan influence oncogenesis. Translating genome-wide associationstudies (GWAS) to the clinic has been challenged by the generallysmall effect sizes of risk variants (see below). This genome-wideanalysis of germline-somatic interactions among cancer patientsidentified common germline alleles that had very large effects onsomatic events. Novel associations included a 15q22.2 allele with>10-fold increased somatic alteration of GNAQ and an intronicSNP in RBFOX1 with >8-fold increase in mutation rate of SF3B1affecting RNA splicing. Interestingly, some frequently alteredhigh-penetrance cancer genes (e.g., TP53 and PIK3CA) were notfound to be influenced by common germline variants. For genesnear the median mutation frequency, such as BRCA1, which isaltered in �3% of tumors, the study was powered to detectchanges in mutation rate of �3.0-fold. This study found thatgenetic background can influence the somatic evolution of atumor by directing where (organ site) and how (which genes areimpacted transcriptionally) cancer develops –making it possibleto anticipate and intercept key early events during tumor devel-opment (65). As broad NGS of blood and neoplasias becomemore widely used and germline-somatic relationships compre-hensively mapped, shared and distinct germline and/or somaticevents can be integrated into the PCA and exploited as preventiontargets (17, 66–68).

GWAS have contributed to expanding catalogs of implicatedgenes and pathways for many complex human diseases and arebeginning to shed light on shared and unique etiological andpathological disease components. Combining large-scale GWASfindings across cancer types (breast, ovarian, and prostate) andusing fine-mapping pathway analysis and polygenic risk scores(PRS) discovered cross-cancer risk loci has the potential to shedlight on shared biology underlying these hormone-related cancers(69). Only a few SNPs are within known cancer predispositiongenes (e.g., ATM and TP53 carriers), and none are associated withcancers that occur in carriers of rare, penetrant mutations. Whilethe low penetrance of most GWAS loci has limited clinicaltranslation, the combined effects of such SNPs to create robustPRS (70)may bemore useful clinically, especially, for example, asmodifiers of high-risk BRCA1 and BRCA2-mutation carriers giventhat even small relative risk changes translate into large absoluteeffects. Most general population breast cancer SNPs are associatedwith ER-negative or -positive disease and, respectively, in BRCA1-

or BRCA2-mutation carriers. Similarly, only general populationloci (e.g., 19p13.11) associated with high-grade serous ovariancancer modify risk of ovarian cancer in BRCA1- and BRCA2-mutation carriers, where tumors are of this high-grade subtype(71). The importance of post-GWAS functional studies in thiscontext is illustrated by finding a SNP in the OGG1 glycosidasegene of the Base Excision Repair pathway, associated with increas-ed ovarian cancer risk inBRCA1-mutation carriers, possibly due toOGG1 transcriptional down regulation, telomere shortening, andgenome instability. These results provide insight into BRCA1mutation mechanism and prevention (72). The new OncoArrayapproach to study genetic architecture of common cancers, whichincludes dense mapping of loci associated with single or multiplecancers, pleotropic effects and cancer related traits, will also buildon first-generation GWAS findings (73). Finally, the ability ofGWAS to discover novel cancer genes/pathways underlying theobserved risk is now being exploited for future drug developmentor repositioning, aided by studies in relevant preclinical models.These GWAS-initiated efforts have already led to prevention-relevant drug targets, including IL-17 and ESR1 (70, 74, 75).

A key challenge is that many GWAS-identified loci are not nearknown coding or regulatory regions, leaving associated mechan-isms and functions unclear. Linking susceptibility variants to theirrespective causative genes and cell-specific regulatory elementsthus remains a high priority in order to realize the full potential ofassociation studies. A rigorous integrated approach to this issuerecently confirmed the widespread genetic regulation of immuneand inflammatory pathways, including in CRC precursors (76).High-resolution genetic, epigenetic, transcriptomic, and RNAsplicing studies, provided powerful approaches to annotate theputative consequence of disease associations. These integratedinvestigations were extended to primary human cells in disease-relevant contexts to unravel the cell- and context-specific regula-tory effects of complex disease variants. This deep characterizationof transcriptional regulation of molecular events will greatlyadvance functional studies of human disease variants, identifyingnovel disease mechanisms and prevention opportunities. CertainSNPs underlying cancer risk are linked to hypermutability andimmune activation (77–79) and intestinal barrier function (dis-cussed below). For example, a germline APOBEC3A/3B-deletionallele confers breast cancer risk via increased APOBEC-dependentmutational processes. Great variation in population frequencyof this deletion allele suggests selection for it. Some APOBECsare involved in innate immunity to infection possibly conferredby this deletion allele, suggesting strikingly disparate effects ofAPOBEC mutagenesis on infection and cancer 80).

Mitochondrial biology and genetics in cancer predispositionMitochondrial DNA (mtDNA) with its mutations and poly-

morphisms is a relatively underappreciated field in cancerresearch. Human mtDNA is maternally inherited (and mappedinitially fromAfrica) (81) and encodes 37 genes: 22 transfer RNAs,2 ribosomal RNAs and 13 protein subunits of the electrontransport chain (ETC) complexes and ATP synthase (mtOXPHOSproteins), essential for respiration. There are several mtDNAcopies per mitochondrion and hundreds of mitochondria percell (82). Generally, neoplastic cells possess functionalmitochon-dria and retain the ability to conduct oxidative phosphorylation.In fact, targeted depletion of mitochondrial DNA can reducetumorigenic potential in vivo (83). While it has long been knownthat somatic mtDNA alterations are frequently acquired during

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 7

oncogenesis (84), recent intriguing data indicate that inheritedmtDNA variants influence multiple innate mitochondrial func-tions, including reactive oxygen species (ROS) production andredox control, signal transduction and epigenome systems, autop-hagy, apoptosis, and immunity (81). mtDNA has a very highmutation rate, greater than an order of magnitude higher than thenuclear genome. Furthermore, mtDNA genes are intimatelylinked with�2000 nuclear genes encoding proteins that functionwithin mitochondria, which can produce Mendelian patterns ofinheritance. Inherited deleterious missense alterations inmtDNAgenes, such as ND6 and COI (cytochrome oxidase subunit I),which code for subunits of OXPHOS complexes I and IV, havebeen associated with risks of various cancers. Inheritedmutationsin the COI gene (associated with increased ROS production,which contributes to tumorigenesis) are associated with prostatecancer risk (somatic COI mutations are also found in prostatecancer), andmay explain increased risks inAfricanAmericanmen;certain African mtDNA lineages harbor COI gene variants thatmay contribute to risk of other cancers among African Americans(85, 86). mtDNA variants have been associated with risks ofovarian (87), bladder (88), breast (89), endometrial (90), andHPV-infection and cervical cancers (91), among others. Theprogression of nonalcoholic fatty liver disease to nonalcoholicsteatohepatitis (NASH), which predisposes to hepatocellular car-cinoma, has been shown to be associated with a mtDNA SNP inthe mt-Atp8 gene (a subunit of OXPHOS complex V, ATPsynthase). This variant has profound effects on hepatic lipid andacylcarnitine metabolism and susceptibility to high-fat diet-induced NASH (92). Generating animal models of these andother inherited mtDNA mutations (93, 94) will be critical toprobing the contribution of mitochondrial biology to inheritedcancer risk.

Deleterious alterations inmtDNA are inherently heteroplasmic(harboring amixture ofmutant and wild-typemtDNA) with highlevels of these severe mutations being lethal. Since mtDNAtransmission duringmitosis is the result of stochastic distributioninto daughter cells, milder mtDNA polymorphisms can shift tobecome predominantly enriched within individual cells andcloser to pure mutant (homoplasmy), potentially contributingto neoplastic transformation. The importance of this phenome-non for cancer predisposition has been demonstrated in a ped-igree in which a mtDNA complex IND5m.12425delA frameshiftmutation is present in homoplasmy in a nasopharyngeal onco-cytic tumor, which was inherited as a germline mutation, trans-mitted at lower heteroplasmy levels (5–10% mutant), and thusmasked by wild-type mtDNAs. Shift to homoplasmicND5muta-tion occurred exclusively in tumor cells and correlatedwith lack ofthe ND6 subunit, indicating that complex Imutationsmay have aselective advantage. Thus, while the transmission of the mutantmtDNA in this pedigree was phenotypically silent, the chanceincrease of the mutant mtDNA in somatic cells caused oncocytictransformation (95). In contrast to Mendelian genetics, the het-eroplasmic mtDNA genotype is continuously changing duringsuccessive cytokinesis to generate cells with varying oncogenicpotential (86, 96).Widespreadheterogeneity has been reported inthe mtDNA of normal human cells and heteroplasmic variantsamong different tissues within the same individual, highlightingthe existence of complexmixtures of related genotypes rather thana single genotype. Mechanistic study of the regulation of mtDNAheteroplasmy may yield novel prevention insights. Because of itshigh mutation rate, mtDNA is highly polymorphic, harboring

functional variants that canbebeneficial or deleterious dependingon context. A subset of mtDNA variants cause subtle changes inOXPHOS, which in turn could modulate a wide range of context-dependent cellular functions of adaptive relevance, includinginflammation, stress, autophagy, and oncogenic responses to dietand other factors (97, 98). Those functional mtDNA variants thatwere beneficial (adaptive) in a particular environment increasedin number and gave rise to descendant mtDNAs, which sharethe founders' beneficial variant. Migration (or other majorchanges in diet, stress, etc.) canmake previously adaptive variantsbecomemaladaptive andpredisposed to awide range of commondiseases (81).

A recentmurine study highlighted the importance ofmtDNA intumorigenesis by examining PyMT transgenic mice (which areinherently predisposed to developing mammary tumors) withidentical nuclear genomes and found that varying the mtDNAgenetic background of these mice influenced both mammarytumor latency and progression (99). In normal mice, the mito-chondrial-targeted catalase transgene (mtCAT) reduces somaticmtDNAmutations (100) and, when crossed with PyMTmice, theincidence of mammary cancer is greatly reduced (101). Mecha-nistic studies have demonstrated the profound influence of subtlechanges in mtDNA haplotype/variation on obesity and aging,both ofwhich are key cancer risk factors (102). Emerging data alsosuggest a complex communication between the nucleus-cytosoland mitochondria (103). Murine models with germline muta-tions in the nuclear gene SUV3, which encodes for a mtRNAhelicase, are characterized bymarked somatic mtDNA instability,hypermutability, shortened lifespan, and various cancers – aunique model to study mitochondrial genomic instability incancer predisposition. Clinical relevance was shown by reducedSUV3 expression in two independent cohorts of human breastcancer (104). Mutations in nuclear DNA genes influencing trans-formation involve some of the same targets/mechanisms affect-ed by mtDNA, including TETs, succinate, fumarate, NRF2, anda-ketoglutarate dioxygenases (105, 106) – all important in cancerrisk.

Mitochondria may also be intimately involved with T-cellimmune surveillance, since T-effector cells are more glycolyticwhile Treg cells are more oxidative. Within neoplastic cells,glucose is converted to lactate (which promotes inflammation,angiogenesis and tumorigenesis), thus inhibiting T-effector func-tion. Treg function is enhanced, which further inhibits the T-effector cells, and suggests that immune rejection of neoplasiamight be enhanced in this setting by mild complex I inhibitorssuch as metformin, whose effectiveness should be increased inneoplastic cells with partial OXPHOS dysfunction (107, 108).Mitochondria can also influence the inflammasome, innateimmunity, IL-1b and NFkB inflammatory pathways (109, 110).BRCA1 has been observed within mitochondria, suggesting a rolein repair of mtDNA and raising speculation that germline BRCA1mutations reprogram breast metabolism towards mitochondrial-dependent biosynthesis and oncogenesis. Metformin can alsoprevent BRCA1 haploinsufficiency-driven oncogenesis by: restric-tion ("starvation") of mitochondrial-dependent biosyntheticintermediates (111); regulating mitochondrial-nuclear commu-nication; and modulating epigenetics (targeting histone acetyla-tion) in genomically unstable precancers (112) to guide meta-bolomic-epigenetic prevention strategies. As with nuclear GWAS,certainmtDNA alterationsmodify (lower) risks of breast cancer ingermline BRCA2mutations (113). Future GWAS integrating both

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research8

nuclear and mitochondrial studies will provide a more compre-hensive germline landscape, including somatic interactions.

Big Genomics Data of PremalignantSomatic Tissues

The collection and analyses of NGS big data are beginningto provide biological insights into cancer prevention and earlydetection in the context of studies characterizing somatic geno-mic alterations. It is worth noting that, in addition to compre-hensive analyses of "big genomics data," several recent studiesthat have also examined cancer microbiomes (reviewed in(97, 114–116)), transcriptomes (117–119) and epigenomes(reviewed in (120, 121)). In addition to big data generatedfrom somatic sequencing efforts, GWAS has involved over onemillion cancer patients worldwide across most organ sitesidentifying �3000 cancer-related genetic associations (122,123) and has been studied in some precancers such as Barrett'sesophagus (124–126), colorectal adenomas (127), DCIS (128)and hematologic premalignancies (below).

The genome of a malignancy can be examined as an arche-ological record bearing the cumulative imprints of all mutationalprocesses that have been operative throughout the cellular lineagebetween the fertilized egg and cancer (129). Each mutationalprocess leaves a characteristic imprint, termed mutational signa-ture, which can change over time, and almost all mutationalsignatures detected in a cancer genome have been imprintedduring the precursor phase of a cancer cell (129). Examinationof the cancer genomes from >12,000 patients has revealed morethan 30 distinctmutational signatures (129,130), including thoserelated to environmental exposures, such as UV-light, aflatoxin,and tobacco (cancer.sanger.ac.uk/cosmic/signatures). Some ofthese signatures have already been used for identifying the pres-enceof specific carcinogens, including aristolochic acid (131), oneof the most potent known human carcinogens – a chemicalpresent in certain plants still in use even today especially in China– with global public health risks of urologic and hepatic cancers(132–134). It will be important for the PCA to study bothpremalignant andnormal-aged tissueswith the goal of identifyingrelevant mutational signatures that may provide important cluesinto premalignant oncogenesis, prevention, and interception.However, it is important to note that approximately half of thecurrently known cancer signatures have unknown etiology andongoing efforts have started exposing experimental systems toknown carcinogens in an attempt to reproduce/identify them(135,136).

Another set of widespread and extensively studied endogenousmutational signatures are those attributed to ectopic activity of theAPOBEC family of deaminases (130, 137). Recent examinationacross more than 10,000 specimens from 36 distinct cancer typesrevealed that these signatures are found in more than 30% ofcancer samples and account for approximately 15%of all somaticmutations across these cancers. The activity of these APOBECmutational signatures is especially strong in bladder and cervicalcancer where they account for more than 75% of all somaticmutations in each of these cancer types (137). In cancers of thecervix and oropharynx, these APOBEC mutational signaturesare predominately triggered early by HPV infection (137, 138),and it has been speculated that APOBEC mutational signaturesare imprinted during the preinvasive phases of these cancers(137, 138). In contrast, APOBEC signatures are believed to be

late events in lung and many other cancer types, and are foundin subclonal expansions and intratumor heterogeneity (139).Germline variants can affect endogenous mutational signaturessuch as the APOBEC3 germline copy number polymorphismdiscussed above associated with APOBEC-dependent mutationalburden in breast cancer (80). Germline variants can also influenceexogenous mutational signatures due to environmental expo-sures. For example, disruptive germline polymorphisms inMC1R,contributor to phosphorylation of DNA repair proteins, havebeen associated with increased somatic mutation burden due toa UV light–related mutational signature and a higher risk fordeveloping skin cancer (140).

Multiple mutational signatures reflect failure of different DNArepair pathways. A specific mutational signature associated withboth germline and somatic BRCA1 or 2mutations (130) has beenobserved in breast, ovarian, pancreatic, gastric, and esophagealcancers (even those without BRCA1 or 2 mutations) (141–146),associated with markers of immunity in subsets of the formerthree cancers (147), suggesting a potential role for immune-basedprevention against such cancers. Amutational signature reflectingthe accumulation of unrepaired reactive oxygen species, mainly8-oxoguanine, has been identified in CRC and adenomas arisingin individuals with pathogenic germline MUTYH mutations,which cause base excision repair (BER) defects (148). Addition-ally, a failure of transcription-coupled nucleotide excision repair(NER) due to somatic mutations in ERCC2 in bladder (includingpreinvasive) neoplasia has been shown to exhibit a specificmutational signature (149). Germline defects in other NER genescan cause Xeroderma pigmentosum, a rare autosomal-recessivegenetic disorder associated with high risk of UV-associated skincancer due to faster accumulation of UV-associated DNA damageand mutational signatures (150). UV-induced nonmelanomalesions can be reduced using bacterial DNA repair enzymes(151) or nicotinamide, which can prevent UV-induced immunesuppression and enhance DNA repair (152).

In addition to mutational signatures related to the failure ofDNA repair mechanisms and ones due to endogenous/exoge-nous exposures, large-scale genomics studies have also identifiedmutational signatures responsible for the unavoidable back-ground mutation rate in somatic cells. Notably, two unrelatedmutational signatures have been found to act as endogenousmutational "clocks," characterized by accumulating somaticmutations within all normal somatic cells of the human bodywith the progression of age (153, 154). One of these mutationalsignatures has been attributed to spontaneous deamination of5-methylcytosine in the context of CpG (its rate of "ticking"appears to be influenced by cellular division), while the etiologyof the second clock-like signature remains unknown. Interest-ingly, a recent study demonstrated the increased rate of onemutational clock to be mechanistically linked to tobacco smok-ing (149, 155). The somatic mutation loads in single-celllineages provide information about an individual's lifetimehistory of mutagenic exposure and the impact of intrinsic factorson mutagenesis. Expanding this study to precancers, more celltypes and larger populations in the PCA would further refineestimates of the rates of somatic changes in human genomes.Understanding the contributions of environmental and endog-enous mutagenic processes to somatic mutation loads is fun-damental to develop preventive strategies (156).

Analyses of omics data from precancers are beginning toemerge. Despite the relatively small sample sizes within such

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 9

analyses, cross-sectional, precancer/cancer pair studies suggestthat many precancers share genomic alterations with their respec-tive invasive cancers, including ductal and lobular breast cancer(157, 158), pancreas (159), non-melanoma skin (160, 161),melanoma (162), lung adenocarcinoma (163) and colorectalneoplasia (148). From an NGS, genomics, transcriptomics (seebelow) and big data perspective, Barrett's esophagus is the best-studied epithelial precancer (141, 164–167), including threeGWAS (125, 126, 168) and a post-GWAS analysis reporting someCDKN2A SNPs associated with reduced esophageal cancer risk(169). Somatic tissue studies of Barrett's esophagus/esophagealcancer pairs revealed that the overall mutation rate (single-nucle-otide variations) was lower in Barrett's esophagus than esoph-ageal cancer but higher than that of several invasive cancers (e.g.,breast cancer, multiple myeloma [MM]) and most recurrentlymutated genes in esophageal cancer were remarkably similar tothe matched Barrett's esophagus. Only mutations in the tumorsuppressors TP53 and SMAD4 were associated with advanced/invasive neoplasia, low or absent from Barrett's esophagus (166),and that intra-tumor genomic heterogeneity, with some contri-bution of aberrant methylation, seemed to drive neoplastic pro-gression. See longitudinal section below for Barrett's esophagusSCNA studies.

Two recent large-scaleNGS studies ofmtDNA in cancer (>2,000human cancer specimens, 30 tumor types) identified amutationalsignature with unique heavy strand-specific C > T transition.Moreimportantly, this missense mutational signature was consideredneutral (analogous to passenger mutations in nuclear DNA),not compromising mitochondrial function (170). One of thesestudies further refined the mtDNA mutational map with match-ed transcriptome sequencing (RNA-seq) data (171). AlthoughDNA/RNA allelic ratios generally were consistent, some muta-tions in mt-tRNAs displayed strong allelic imbalances caused byaccumulation of unprocessed tRNA precursors, indicative ofimpaired tRNA folding and maturation, which underlie a rangeof diseases. Both studies found a selective pressure against dele-terious coding mutations affecting oxidative phosphorylation,indicating that tumors require functional mitochondria. Unex-pectedly, known dominant mutagens, such as cigarette smoke orUV light, had a negligible effect on mtDNA mutations. Anothernew study has reported significant correlations between mtRNA-Seq and mtDNA copy number, with some important exceptions(e.g. MT-ND5 and MT-ND6) (172). mtDNA mutations are fre-quent in Barrett's esophagus without dysplasia (173) and havebeen used to characterize genetic lineages that assess clonalrelationships, e.g., between Barrett's esophagus and esophagealsquamous cells (174). Clonal expansion of mtDNA mutationscan result in mitochondrial dysfunction, such as decreased ETCenzyme activity and impaired cellular respiration. NGSofmtDNAof oncocytomas, which are rare benign tumors of epithelial cellsdefined by excessive amounts of mitochondria, has identified apathogenic mutation signature that compromises the overallfunction of the mitochondria, proposed to serve as a metabolicbarrier for these benign tumors, and perhaps precancers, progres-sing to malignancy (175).

In addition to the studies above, a systematic approach toclassify cancers using transcription profiles at both bulk tumorsand single-cell resolution has been described (176–179). Theseprofiles not only provide molecular bases for classifying cancerswith shared transcriptional programs across different cancer typesbut also characterize heterogeneity that exists within individual

neoplasias (180). Whole transcriptome profiling using RNAseq,pathway enrichment, and functional assays of BE found novelcell-cell communication, with normal epithelial cells sensing anddramatically suppress dysplastic cell behavior (e.g., motility,signaling pathways). These effects are distinct from the stromaland immune cell microenvironment effects. DownregulatedTGFB, EGF, and Wnt pathways associated with the differentialtranscriptional profiles observed in the dysplastic cells in co-culture (normal and dysplastic Barrett's epithelial cells, whichoften co-exist in vivo) (167). Single-cell approaches would allowanalysis of different subpopulations of cells, including highlyvariable epithelial cell motility and immune, stromal, and othermicroenvironment cellular components surrounding the neopla-sias (Figure 1) (167, 179). Furthermore, oncogenic pathways anddevelopmental- and immune-based gene expression signaturescan be used for "pathway/phenotype"-based molecular charac-terization (181–183).

Recently, a novel analytic approach to define oncogenic statesand produce functional maps of cancer has been established.This serves as a framework for combining experimental andcomputational strategies to deconvolve oncogenic pathways/signatures derived from oncogene activation into transcriptionalcomponents that can be used to determine oncogenic states. Bymapping precancers and cancers onto distinct oncogenic states,the resulting functional map can be used to characterize howthese states relate to various omics features, including NGSmutations, copy number alterations, gene and protein expres-sion, gene dependencies, and biological phenotypes; and topredict which interventions are more likely to have a significanteffect (184). This approach has been used to effectively mapcancers with altered KRAS/MAPK pathways into divergent func-tional states. Studies in pancreatic oncogenesis highlight theneed for big data approaches to interpret neoplastic complexity,including KRAS mutation subtypes and Hippo pathway inter-actions, profound effects on cell metabolism, DNA repair,immunity, mitochondrial biology, and distinct precursor path-ways (185). Mutant Kras in pancreatic acinar cells inducesexpression of ICAM-1 to attract macrophages and drive PanINdevelopment: direct early cooperative mechanisms betweendriver mutations and inflammatory environments (186). EvenB-cells can initiate and promote pancreatic tumorigenesis. Forexample, primary human and mouse models found B-cell infil-tration in proximity of PanIN lesions, due to stromal secretion ofthe B-cell chemoattractant CXCL13 (187). Another study foundhighly expressed HIF1a in PanINs; deletion of HIF1a increasedsecretion of CXCL13 and related attractants and acceleratedmalignant transformation. Depletion of B-cells reduced PanINprogression (188). These dynamic maps can continually inte-grate new data, be generalized to consider gene networks andinteractions (189), including the germline, and study the closeinterchange with the immune microenvironment. Integratingfunctional genomics data, such as those described above, to themaps of oncogenic states will provide further insight intothe cellular contexts in which genomic alterations contribute tomalignant transformation (190).

EpigeneticsPreviouswork has yielded only a limited big data perspective of

the neoplastic epigenome, primarily in hematologic neoplasia,where chromatin modifiers are in general among the most

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research10

frequently mutated in cancer (121, 191). Most studies havefocused on performing functional analysis on a few genes in alimited number of samples as reviewed below. Widespread epi-genetic field defects have been observed in apparently normalbreast tissue located adjacent to breast cancer (192) and alsoassociated with inflammation-related cancers, such as H. pylori-induced neoplasia, in which NGS has revealed more cancerpathway-related genes affected by DNA methylation than bygenetic alterations (193). The ten eleven translocation (TET)enzymes oxidize 5-methylcytosines (5mCs) and promotelocus-specific reversal of DNA methylation (194).

An epigenetic mitotic clock was developed using a novelmathematical approach. A key feature underlying the construc-tion of this clock is the focus on Polycomb group target promoterCpGs, which are unmethylated in many different fetal tissuetypes, thus allowing defining a proper ground state from whichto then assess deviations in aged tissue. By correlating the tickrate predictions of this model to the rate of stem cell divisions innormal tissue, as well as to an mRNA expression-based mitoticindex in cancer tissue, this model approximates a mitotic-likeclock. The epigenetic mitotic clock-like signature exhibits a con-sistent, universal pattern of acceleration in cancer in normalepithelial cells exposed to a major carcinogen. Epigenetic clonalmosaicism is maximal before cancer emerges. Unlike the muta-tional clock-like signatures discussed above, this epigeneticclock is based on clinical DCIS and lung CIS progression to cancerand normal at-risk tissue; a concrete example of a molecularmitotic-like clock that predicts universal acceleration in precancer(195). Smoking was associated with an increased rate of thismutational clock. Another approach to intratumoral heterogene-ity analyzed DNA methylation patterns at two genomic loci thatwere assumed to have no role in gene regulation, in contrast todriver methylation changes. Methylation at such neutral loci wereunlikely to be under selective pressure and therefore, couldserve as a ``molecular clock'' to measure mitotic divisions basedon the higher error rate of DNAmethylationmaintenance relativeto the error rate of DNA polymerase (196).

Aberrations of the epigenetic modulator TET2 are among thefirst alterations identified in several hematologic premalignan-cies. The TET family of proteins was originally named due to theTET1 fusion partner in mixed-lineage leukemia (MLL)-rear-ranged AML (see "biochemistry" subsection below), however,this translocation is rare, and its significance in leukemogenesisis unclear. In contrast, TET2 mutations are found in premalig-nant hematopoietic stem cells (HSCs), including of myelopro-liferative neoplasms (MPNs) and MDS, and are frequentlyobserved in aged healthy individuals (197) with propensity totransform (see clonal hematopoiesis below). Disruption ofTET2 in mouse models increases HSC proliferation and clonalexpansion, prone to additional oncogenic events generallyrequired for malignant transformation (194). Mouse modelsfound that co-occurring Tet2 disruption with Asxl, Ezh2, or Jak2V617F results in MDS or MPN. Recurrent dominant pointmutations in IDH1 and IDH2 are early events in some hema-tologic neoplasias that lead to loss of TET activity and otherepigenetic changes (198, 199). In addition, TET2, IDH1, andIDH2 mutations are frequently observed in lymphoma precur-sors (200–202), and the frequency of TET loss-of-function(which drives hematologic transformation) in these settingssupports testing TET activators. TET modulators (203), canenhance antigen presentation, increase IL-6 production by

macrophages (204), affect Tregs (205), and alter expressionof endogenous retroviruses, cancer testis antigens, and stem cellantigens in premalignant lesions resulting in enhanced immu-nogenicity (206).

Another epigenetic mechanism found to be important inpremalignant biology involves RNA editing by ADAR enzymes,which results in adenosine-to-inosine conversion of RNA therebyinducing virtual adenosine-to-guanine mutations (since inosinebears molecular resemblance to guanine) (207). Depending onwhether the editing events occur in coding regions or 30 UTRs,ADAR-mediated editing of mRNAs can result in post-transcrip-tional protein codingmutations or altered susceptibility tomicro-RNAs (208). Recent data suggest that germline variation of ADARgenes may influence ovarian cancer susceptibility (209). ADAR1editase activity has been implicated in the oncogenic transforma-tion of premalignant progenitors that harbor clonal self-renewal,survival, and cell cycle-altering mutations (210, 211), such as inhepatocellular carcinoma precursors, where aberrant RNA editingof AZIN1 has been found to be a key oncogenic driver (210, 212).Transgene expression of APOBEC-1 causes dysplasia and cancer inmouse and rabbit livers likely due to RNA editing ofNAT1 (213).Study of ADAR1 regulation of APOBEC3 in neoplasia will becritical, potentially suppressing hypermutation and immunity(214). Finally, inflammatory cytokine networks and JAK2/STATsignaling activate ADAR1 during relapse/progression in leukemiastem cell renewal, linking RNA editing to the development ofinnate immunity and potential preventive activity (215).

Emerging data also suggest that some premalignant lesionsmay progress to cancer via fundamental epigenetic reprogram-ming. Indeed, reprogramming of the epigenome to a progenitor-like state may be required for driver mutations to induce tumor-igenesis (216). The role of BRAF mutations in benign nevi is amajor historical conundrum (217). In the BRAF V600E zebrafishmodel of melanoma, deletion of p53 promotes the nevus-to-melanoma transformation, but melanomas remain surprisinglyinfrequent considering that all of the cells bear both the oncogeneand tumor suppressor loss (218) – a feature that replicates thephenomenon of "field defect" in human tumors. Two recentstudies using preclinical models addressed this issue. Work withBRAF V600E/p53-null zebrafish now suggests that initiation ofmalignant transformation within such a "cancerized field"requires fundamental epigenetic reprogramming of these prema-lignant cells into an embryonic state via transcription factor-mediated reactivation of genes typically expressed only in neuralcrest progenitor cells (216). This reprogramming involves bindingof multiple transcription factors and generation of "superenhan-cer" regions. Engineered models, including epigenetic mechanis-tic studies, suggest key roles of p15 loss and autophagy (over-coming senescence) in promoting BRAF V600E-mutant transfor-mation to melanoma (219). Deletion of Atg7 inhibits tumori-genesis, likely via a mitochondrial mechanism (220). Similarly,mouse model research recently demonstrated that basal cellcarcinomas, known to be driven by oncogenic signaling in thehedgehog pathway, only originate from stem cells located in veryspecific areas of the murine epidermis, rather than from morecommitted progenitor cells (221). Like the zebrafish model, thisstudy provides evidence that the earliest stages of tumorigenesisare characterized by reprogramming to a more embryonic cellstate. Such data suggest that tumor-initiating cells can be identi-fied– andpotentially targeted for early destruction– through theirability to reactivate an "embryonic" epigenetic state, highlighting

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 11

one of the PCS's key missions: studying premalignant cells andmodel systems to better understand when epigenomic changesarise and how stable they are over time.

The Power of Immunology andBiochemistry to Facilitate CancerPreventionImmune oncology

The integration of multiple omics analysis platforms withimmune-informatics analysis can be the foundation of a moreeffective framework for precision prevention (222). There is nowawealth of evidence from both animal models and cancer patientsof how the immune system can survey and recognize peptidesencoded by certain genetic mutations when such peptides arepresented on the surface of the cancer cell bound to MHC-Class Iand Class II molecules. For example, RAS mutations, which arekey oncogenic drivers in a wide array of cancers, may also betargets of immunosurveillance since T-cells specific to mutatedRAS peptides have been found in cancer patients (223) and maybe a viable target for immune approaches to treatment and evenprevention (224). Proof-of-principle studies of vaccine targetingmutant Kras (with Treg depletion) in a pancreas mouse modelinduced CD8þ T-cells specific for the Kras mutation and showedpreventive efficacy in the early PanIN setting (225). In addition topredicted mutations in known oncogenes, cancer cells and theirprecursors can harbor tens to hundreds of random mutationsthroughout their genome.

Vaccine-based approaches hold particular promise since theyare a form of precision prevention with few side effects. Viralvaccines (e.g., to HPV) given before exposure can provide long-term protection from cancer development after only one or twotreatments (5, 226). HPV vaccines havemodest activity, however,in treating chronic viral infections and related precancers, and farless activity in treating established cancers. The mechanisms ofviral immune evasion vary by pathogen (seven viruses have beenlinked with human neoplasia), but are strikingly similar to thoseused by non-viral neoplasias, supporting the inclusion viraloncogenesis in this PCA (227). For example, HLA genes havebeen linked to development of certain cancers. The minor MHCclass I gene HLA-G plays a key role in immune evasion and HPVinfection and transformation of CIN III to invasive cancer (228).Evidence for immune surveillance has been reported in healthypeople and associated with lowered lifetime cancer risk. Child-hood febrile viral infections have been associated with reducedcancer risk, consistent with an influenza mouse model in whichthe virus infection elicited protective antibodies and T-cells spe-cific for host and some tumor-associated antigens (229). Thesedata suggest that infection-induced immunity and immunemem-ory could provide long-term immune surveillance of cancer,important properties for vaccine targets. T-cells are likely themaineffector cells in preventing most forms of cancer. The immunesystem has the ability to recognize precancers and generateimmune responses to potentially intercept and prevent cancer(51, 58, 230) and avoiding immune elimination is a hallmark ofcancer (7, 231). We must learn how to both strengthen T-cellimmunity– either through immunization, drugs, or engineering–and concurrently overcome a dynamic hostile tumor microenvi-ronment that prevents T-cell activation and infiltration into earlyneoplasia. The latter involves multiple factors, for example, met-abolic reprogramming of the microenvironment by the high

utilization of extracellular glucose and glutamine results in extra-cellular lactate, which attenuates dendritic and T-cell activation,stimulatesmacrophage polarization to anM2 state, induces VEGFsecretion by stromal cells, and activates NF-kB. The microenvi-ronment can in turn have profound effects on the metabolism ofneoplastic cells (83). Emerging data suggest that microenviron-ment barriers develop early in precursor lesions but are likelyqualitatively different from more established cancer-associatedbarriers. The progressive accumulation of somatic changes thatleads to neoplasia also co-opts neighboring vascular, neuronal,and other normal cells to support/promote oncogenesis.

Remarkable new data revealed that high-level arm- and whole-chromosome-SCNAs drive immune evasion (232). This unantic-ipated effect is in striking contrast mutational burden effects onimmunity, and can override immune response even in highmutational/neoantigen burden settings (e.g., MSI-H/MMR-defi-cient tumors), revealing an increasingly complex interplaybetween SCNA levels, gene (including driver mutations), andimmunity. Critical to vaccine development, therefore, is theidentification of potent immune enhancers/adjuvants (e.g., met-formin, STING) that can specifically target one or more innatepathways (233) and alter the developing inflammation thatpromotes immune suppression (234). Experience with therapeu-tic cancer vaccines shows that targeting a single antigen or a singlemutated peptide invariably leads to outgrowth of cancer cells thathave lost that mutation. This may happen in the precancer settingas well, requiring a vaccine that elicits a polyclonal and poly-specific immune response. Trial endpoints could include T-cellreceptor sequencing to look at clonality and clone expansion andpotentially liquid biopsy detection of low levels of specific muta-tions or mutational load, SCNAs and imaging of the microenvi-ronment, immune response (235), and high-grade precancer(e.g., PanIN-3) (56).

In a clinical feasibility trial of advanced adenoma patients, lackof immune response to a MUC1 cancer vaccine correlated withincreased levels of circulating MDSCs responsible for inhibitingadaptive immunity (236), suggesting that these may be usefulbiomarkers to identify individuals unlikely to benefit from pre-ventive cancer vaccines. As above, research into drugs that couldhelp overcome such immune resistance will be critical. Metfor-min, for example, has many key properties in this context and hasbeen shown to enhance T-cell immunity and immune memory,and also influence the microbiome, by various mechanisms,including involving mitochondrial biology and RANK-L inhibi-tion (see above) (41, 237–239). Furthermore, prospective cohortdata suggest that aspirin prevention of CRC is related to its effectson T-cell immunity (240). Large study of adenomas reported ahighly inflammatory microenvironment (241), which varied byhistology and location (242). Two recent NGS studies of IBD(colitis)-associated CRC suggested that, compared to their spo-radic counterparts, colitis-associated CRCs have a distinct muta-tional profile associated with cell-to-cell signaling, cell adhesion,and epigenetic regulators/chromatin modifiers, all linked to IBDinflammatory mediators (243, 244). IDH1 mutation (extremelyrare in sporadic CRC) was found only in Crohn's associated CRC.Future IBD NGS should include epigenomic and microbiomeprofiles and dysplasia.

Microbiota-immune interactions are an increasingly importantand challenging field. Each human body contains at least 40trillion microbial cells (microbiota), which have exponentiallymore genes (the microbiome) than do human cells; 99% of

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research12

microbiota reside in the gut microbiome and certain bacteriacan influence oncogenesis, inflammatory microenvironments,and immune interventions. Low-grade gut inflammation dueto microbiota weakens epithelial tight junctions and may causecancer risk factors – obesity and Type 2 diabetes (245). Intesti-nal barrier function (and bacterial translocation) is regulatedby GWAS-identified laminin nuclear and mtDNA variants(19, 127, 246–248) and inflammatory cytokines such as IL-1Band IL-18 (249), autophagy (250), and carbohydrates (whichaffect gut mucous layer and microbiota spatial organization)(251). Thymic stromal lymphopoietin (TSLP) is a cytokineexpressed mainly by epithelial cells at barrier surfaces (skin, gut,and lung). Short-course calcipotriol, a topical specific TSLP induc-er FDA approved for psoriasis, durably suppressed skin cancerdevelopment in genetically engineered mouse models (GEMMs)consistent with an immune memory response, markedly reducedactinic keratosis (mediated T-cell adaptive immunity) in a ran-domized clinical trial (252). In theMinmouse model, innate andadaptive sources of IL-17 promote pathologic myeloid inflam-matory signature and colon tumorigenic response to enterotoxi-genic Bacteroides fragilis (ETBF); and anti-IL-17 monoclonal anti-body (mAb) and Treg cell depletion suppressed tumorigenesis atthe micro-adenoma stage (23). Emerging data suggest a linkbetween ETBF, inflammatory bowel disease (IBD), and CRC.ETBF toxin-triggered colon tumorigenesis is characterized by aspecific immune signature (combining IL-17-driven colitis andaltered myeloid differentiation into MDSC) (253). Potentialother immune and/or microbiota prevention factors includelifestyle (97), antibiotics, diet, and microbial reprogramming(254, 255). It has been shown that gut microbes modulate wholehost immune and hormonal factors influencing the fate of distantprecancers (e.g., breast), for example, by stimulatinghost immunecells to prevent dietary and genetic predisposition to mammarycancer in mice. This raises the possibility that the tumor micro-environment interacts with broader systemic microbial-immunenetworks (256). Caloric restriction is the most consistently effec-tive cancer preventive approach in virtually every mouse model/tumor type tested. Recent data, including from the mutant Kraslungmousemodel, indicate that caloric restriction or itsmimetics(e.g., over-the counter hydroxycitrate) elicit autophagy, whichimproves immunosurveillance via Treg depletion and preventsmalignant transformation (257).

Analyses of NGS genomic data are critical to develop vaccinesthat target specific epitopes derived frommutations, copy numberalterations, or other variants common to precancers. However,this direct strategy is especially challenging given the large numberof alterations, low penetrance of driver mutations, so-called"long tail" problem (low frequency mutations) (258), and corre-sponding mutant peptides, which do not lead to effective antigenpresentation and response (259). Mass spectrometry-based anal-ysis and immunogenic assays in mice (260), coupled with pre-cursor NGS genomic data, will help select better precancer vaccinetargets. Computational methodologies using existing resourcesand databases that catalog potential antigens (261) andmachine-learning classification approaches to predict peptide bindingaffinity to HLA I and II have also been developed (262, 263).Computational studies of neoantigen-epitope prediction algo-rithms have shown that only a very small proportion of predictedneo-epitopes are actually presented on MHC class I as targets ofendogenous T-cell responses (11, 264, 265). Using the NGSgenomic data from precancers could use two strategies to nom-

inate candidate neoantigens: 1) mutation-calling algorithms toidentify frequently occurring antigens and 2) for the low frequen-cy events, utilize functional maps described above to identifycomplementary antigens that associate with oncogenic states(184). These filtered lists of neoantigens could predict strongepitope candidates in silico using algorithms and analytic pipe-lines based on stabilized peptide p–MHC-I binding affinity(264, 266). Novel approaches are being developed to identifyneoantigens that elicit T-cell responses and could help prioritizepeptide candidates for vaccination (267). Leveraging the PCA todevelop high-throughput approaches and improved predictivealgorithms to identify (and quantify) immunogenicity of theprecancer antigenic repertoire will be critical to identifying poten-tial immunosurveillance and vaccine targets (268–270).

As above, the MHC/HLA complex (located on 6p21) is criticalto immune oncology and vaccine development. HLA loci arehighly polymorphic and implicated in cancer risk by multiplestudies in a variety of tumor types, including via search of theNHGRIGWAS catalog (228, 271).Highlighting the importance ofbiochemical studies in this context, a recent analysis foundsomatic mutations that disrupt b-2 microglobulin (a componentof the class I MHC complex, located on chromosome 15) protein-protein interactions, with a striking enrichment for mutations atprotein interfaces involving b-2 microglobulin's binding partners(272). It has been shown that disruption of b-2microglobulin canminimize immunogenicity of human embryonic stemcells (273).It is conceivable that such mechanisms may be employed byprecancers and cancers to escape immune surveillance (274, 275).

Biochemistry: understanding the molecular basis of neoplasiaTCGA and related studies have demonstrated that a large

number of genetic and epigenetic factors, such as chromatinmodifiers and remodelers, are highly mutated in a large numberof solid tumors and in hematologic malignancies (276). Recur-rent mutations in genes that encode regulators of chromatinstructure and function highlight the central role that aberrantepigenetic regulation plays in the pathogenesis of these neo-plasms. Deciphering the molecular mechanisms for how altera-tions in epigenetic modifiers, specifically histone and DNAmethylases and demethylases, drive hematopoietic transforma-tion could provide avenues for developing novel targeted epige-netic prevention for hematologic neoplasia and could also informfuture studies in solid tumors. Many such protein complexes –including the MLL family (191), the polycomb complexes PRC1and PRC2, which contain EZH2, ASXL1, and BAP1 (277), and theSWI/SNF chromatin–remodeling complex (278) – contain genesthat are frequently mutated in human cancers (276), but wereinitially identified in simple model systems, such as Drosophilaand yeast, emphasizing the importance of including studies ofmodel organisms in any large-scale efforts in cancer prevention.While genomic deletions and nonsense, frameshift, and splice sitemutations that introduce a premature stop codon or alter proteinstructure can be obvious loss-of-function events, missense muta-tions can be hard to classify unless they alter enzymatic functionor disrupt protein–protein interfaces within large functionalprotein complexes.

For example, a large number of hematologic malignanciesharbor translocations of the N-terminal region ofMLL1 to diversefusion partners that share very little sequence or functionalsimilarity. To understand how these diverse MLL translocationsresult in leukemogenesis, biochemical and enzymological studies

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 13

were essential. First, MLL and its yeast homolog SET1 have beenshown to be present in a complex named COMPASS (Complex ofProteins Associated with Set1) and to function as histone H3K4methylases (279). Second, AFF4, itself a fusion partner of MLL inleukemia, has been found to be a common factor among allpurifiedMLL translocations (280). Third, ELL, one of the frequenttranslocation partners of MLL in leukemia, has been found tofunction as an RNA Pol II elongation factor that increased thecatalytic rate of transcription elongation by RNA Pol II by sup-pressing transient pausing (281). Finally, it has been shown thatmany MLL translocation partners are found in association withELL and the positive transcription elongation factor (P-TEFb),within a complex named the Super Elongation Complex (SEC)(278, 282, 283). The translocation of MLL into SEC is involved inthe misrecruitment of SEC to MLL target genes, perturbing tran-scription elongation checkpoints at these loci and resulting inleukemia (283).MLL-induced leukemogenesis highlights the roleof deregulated histone methylation in tumorigenesis (284).

Another example of the importance of biochemistry is deci-phering themolecular role of an observed genetic link of EZH2 incancer. EZH2 encodes the catalytic subunit of the polycombrepressive complex 2 (PRC2), which is responsible for methylat-ing lysine 27 of histone 3 (H3K27). Trimethylation at this site isassociated with closed chromatin and silencing of neighboringgene expression. In neoplasia, EZH2 can influence T-cell biology(285) and function as either an oncogene or a tumor suppressorgene depending on the cellular context, for example, EZH issufficient to transform lung cells in transgenic mouse modelsoverexpressing EZH (286), and loss-of-function EZH2mutationsoccur in MDS and chronic myelomonocytic leukemia (CMML)(287). In germinal center diffuse large B-cell lymphomas, recur-rent mutations essentially of only one codon (Y641) create aprotein with reduced affinity for unmethylated H3K27 but highlyincreased affinity formonomethylatedH3K27, resulting in higherlevels of H3K27 trimethylation overall. In contrast, pre-AMLsyndromes like MDS and CMML do not develop Y641mutationsbut instead recurrently develop nonsense, frameshift, and otherloss-of-function mutations in EZH2 resulting in low levels ofH3K27 trimethylation (288). Ezh2 loss synergizes with JAK2-V617F in hematopoietic cells to contribute to the developmentand progression of MPN. The MPN phenotype induced by JAK2-V617Fwas accentuated in JAK2-V617F;Ezh2(�/�)mice, resultingin expansion of the stem cell and progenitor cell compartmentsand severe disease progression, including more advanced mye-lofibrosis and reduced survival (289). These results, which sup-port tumor suppressor function of EZH2 in patients with MPN,have important clinical implications for EZH2 inhibitor devel-opment in this setting. It is possible that EZH2 inhibition willmimic malignancy-associated, loss-of-function EZH2 mutationsin normal myeloid cells leading to dysregulated growth or dif-ferentiation in these cells, highlighting the need for future context-dependent studies.

SWI/SNF, also known as the BAF complex, is also a criticalregulator of nucleosome remodeling conserved from yeast tohumans. Biochemical investigation combinedwith bioinformaticassessments have demonstrated widespread genomic alterationsthat occur across the members of the complex in 19.6% of allhuman tumors reported in 44 studies (290). In synovial sarcoma,SS18–SSXoncogenic fusion that results froma fusionof 78 aminoacids of SSX to the SS18 subunit of BAF complex was shown todisrupt binding of BAF47, tumor-suppressive member of the

complex, leading to reversible dysregulated growth (291). Inliver, genetic suppression of SWI/SNF complex member ARID1Bwas shown to overcome oncogene-induced senescence and leadto liver neoplastic progression (292). While these studies suggestan emerging role for SWI/SNF in tumorigenesis, better delineat-ing the role of SWI/SNF complexes in precancers will also beimportant. Furthermore, prior studies demonstrate antagonisticrelationship between the SWI/SNF and PRC2 complex in medi-ating oncogenic transformation (293).

A transformative example of biochemistry's importance inpremalignant biology involves the discovery of recurrent muta-tions in IDH1 and IDH2 in glioblastoma, AML, and their precur-sor cells. Such mutations were found through broad sequencingefforts (294) although their role at the molecular level was notclear until the advent of modern metabolomics profiling (295),which found that mutant IDH enzymes convert the normalintracellular metabolite alpha-ketoglutarate into 2-hydroxygluta-rate. A competitive inhibitor of a large class of dioxygenaseenzymes that utilize alpha-ketoglutarate, 2-hydroxyglutarateaccumulates to very high levels in IDH-mutated cancers, potentlyinhibiting many important intracellular dioxygenases, includingthe TET family, prolyl hydroxylases, and several histone demethy-lases (296–298). Thus, biochemistry and metabolomics haveillustrated how 2-hydroxyglutarate contributes to carcinogenesisin a hitherto unprecedented way by acting as a novel "oncome-tabolite" generated by somatic IDH1/IDH2 mutations that canpotentially serve as targets for both cancer prevention and therapy,including vaccines (299).

Biochemical approaches have also focused on the significanceof metabolism and its link to epigenetic factors, such as the TETfamily in the regulation of cell-lineage specification and thedevelopment of cancer (191). These discoveries are only a fewexamples among a large number of biochemical approaches inneoplastic cancer studies and are the testimony to the power ofbiochemistry in understanding neoplasia and the design of itstargeted prevention, for example, by highlighting the importanceof epigenetic regulation. High information content mass spec-trometry to profile global histonemodifications inhuman cancers(300), when combined with the DNA-sequencing data, can beused to identify novel variants that can drive epigenetic changesthat can lead to oncogenic transformation. Chromatin-immuno-precipitation technology combined with NGS sequencing (ChIP-seq) can provide systematic information regarding the architec-ture of the chromatin cell states of cancers. New technologicaladvances have demonstrated that CHIP-seq can be carried out inhuman tissues including tumors (301). Interestingly, examina-tion of the chromatin landscape was able to fully distinguishnormal versus cancer samples. These results suggest the possibilityof gaining additional insights into precancers by systematicassessment of chromatin states using key histone acetylation andmethylation patterns, super-enhancers, as well as TET, SWI/SNF,and PRC2 complex, all of which are critical for chromatin regu-lation (301).

It is essential that the PCA incorporate detailed biochemicaland enzymological studies on purified protein complexes todecipher the precise, context-dependent function of chromatinand other epigenetic modifiers and somatic mutations in pre-cancer development and progression (280). This will also allowthe profiles to be cross-referenced with the landscapes in primarytumors, as well as of the corresponding transcriptomic datato identify critical epigenetic changes that are necessary for

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research14

malignant transformation. The context-dependent, complex rolesof EZH2 mutations and PRC2 and SWI/SNF complexes in chro-matin regulation in normal development and neoplasia requirefurther study, especially in precancers. Finally, metabolic altera-tions in neoplasia continue to uncover novel connectionsbetween nutrient utilization and oncogenic state, critical to pre-cancer progression.

Single-Cell AnalysesThe natural history of precancers is heavily influenced by the

heterogeneity (including genetic and epigenetic (302) of neo-plastic cells and tissue microenvironment. Single-cell RNA- orDNA-sequencing technologies can be specifically leveraged tounravel the complex cellular relationships within these lesionsthat cannot be addressed by assaying bulk tissue (303, 304). Inthe case of mRNA profiles, downstream analyses can character-ize known populations and novel subpopulations of cells andassess how these populations change in abundance as diseaseprogresses or regresses. These data also can be used to moreaccurately infer important disease-associated gene regulatoryand immune cell networks (305) because the gene expressionvariability has not been averaged across all sampled cells as inbulk tissue. In addition, single-cell sequencing can reveal andmonitor lesion heterogeneity in somatic alterations and dissectcomplex clonal dynamics among epithelial cells sampled atdifferent geographic locations and over time to complementexisting multiregion bulk sequencing approaches (306). Thesedata will provide a high-resolution picture of cell types presentin precancers and their surrounding microenvironment and thetranscriptional programs active within each cell type that drivedisease progression.

However, several technical limitations need to be overcometo realize the full potential of single-cell sequencing of precan-cers. These lesions are relatively small and frequently onlydiagnosed in formalin-fixed, paraffin-embedded (FFPE) tissuespreviously precluding comprehensive genome-sequencing stud-ies using current methods (307). Furthermore, informationregarding the location of neoplastic cells with particular muta-tions within a given lesion is especially important for earlylesions, as this often defines the boundary between preinvasiveand invasive neoplasia. Therefore, the development and appli-cation of methods that allow assessing the genetic and pheno-typic features in situ using intact FFPE tissue samples is especiallycritical for the improved understanding of preinvasive lesions.Several technologies enable copy number alteration and geneexpression analyses at the single-cell level from FFPE slides.These include FISH and immuno-FISH (combination of FISHwith immunofluorescence; refs. 308, 309), mRNA in situ hybrid-ization, in situ PCR, and STAR-FISH (310–312). The applicationof immuno-FISH for the analysis of cellular phenotypic hetero-geneity and genetic features revealed extensive intratumor diver-sity in DCIS and clear expansion of minor subclones in DCIS todominant clones in invasive ductal carcinoma (309). A short-coming of these methods is the limited set of markers that canbe assessed on a single slide and the need for a priori knowledgeof the changes to be analyzed.

Single-cell methods are beginning to be applied to prema-lignancy, including sequencing on both fresh/frozen and FFPEhas been applied to an epithelial precancer site, DCIS, andassociated invasive breast cancers, and included massively par-

allel single-cell sequencing for copy number analysis (180). Thisproof-of-principle analysis established technical feasibility anddemonstrated intralesion genetic heterogeneity at DCIS, sug-gesting complex and distinct evolutionary processes involved inearly DCIS to subclonal selection in invasive disease. MulticolorFISH to study clonal evolution at single-cell resolution inBarrett's esophagus found extensive genetic diversity in pro-gressors (313). A whole-exome single-cell sequencing methodwas developed to assess genetic heterogeneity and tested on apremalignant JAK2-negative myeloproliferative neoplasm(essential thrombocythemia) patient (314). Such profiling ofa different myeloproliferative neoplasm myelofibrosis revealedsubstantial heterogeneity in cytokine production (315). Recentsingle-cell genomic study of childhood ALL provided a high-resolution view of the preleukemic sequence of events leadingto ALL, including early ETV6-RUNX1 translocations in hemato-poietic precursors to later APOBEC-mediated transformation toleukemia (316). Importantly, however, current single-cellsequencing approaches have important technological limita-tions, especially in epithelial precancers. First, the methodsavailable are labor and cost intensive. Second, it is currentlynot possible to obtain accurate detailed copy number andmutational data from the same cell, given that some whole-genome amplification methods yield templates optimal forcopy number analysis, whereas others are optimal for mutationprofiling. Therefore, efforts are required for the development ofless labor-intensive and more cost-effective methods forsequencing approaches and clonal lineage tracing, which areessential for a detailed analysis of the evolutionary paths of insitu disease and its progression to invasive cancer. Two moretechnologies, FISSEQ (fluorescent in situ sequencing; ref. 304)and "spatial transcriptomics" (317), allow for complete tran-scriptome analysis of single cells in intact tissue sections.Recently, single-cell techniques have been developed to studychromatin maps/signatures and epigenetic heterogeneity inneoplasia (318, 319) and the microbiome (320), includingimaging of host–microbiota interchanges (251, 321).

Cost reduction and advances in cell sequencing (and cfDNAtechnology) could theoretically allow temporal monitoring ofblood and epithelial premalignancies on a population scalewithin the PCA (322–324). Periodic single-cell DNA sequencingof multiple cells from an individual will be invaluable for cancerprevention as it will allow one to assess the overall baselineaccumulation of somatic mutations over time in a person, tosurvey andmonitor multiple different endogenous processes andexogenous exposures through the use of mutational signatures,and to reveal the existence of premalignant clones and clonalevolution over time (325–327). The unprecedented resolution ofsequencing single cells comes with a hefty computational anddata price. Monitoring even a single individual will requiremultiple sequencing of one's genome every year resulting inseveral terabytes of data per person. As such, population-scaleexaminations will generate millions of whole-genome sequencesresulting in exabyte scale data (>1018 bytes) requiring next gen-eration of computational infrastructure (328) and novel compu-tational frameworks (e.g., to take into account their relatively lowsignal resolution when compared with traditional bulk tissuesequencing). Challenges for whole-genome analyses with sin-gle-cell resolution will be highly amplified with multi-omicsequencing (329). More than ever, the rate-limiting step will bedata analysis.

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 15

Liquid Biopsies for Early Detection andIntervention

By virtue of the clonal nature of tumor cells, somatic changesare present inmany copies that are continuously released and canbe detected in the blood as cell-free (cf) circulating tumor DNA(330). In cancer patients, alterations in cfDNA can be detectedusing sequencing and bioinformatic approaches, although suchalterations can be difficult to detect as they often represent aminute fraction (<1%) of cfDNA. A variety of both targeted andwhole-genome approaches have been developed to detectsuch alterations in cfDNA (331, 332). These have been used forearly detection of recurrence in colorectal cancer, pancreaticcancer, neuroblastoma, hematopoietic malignancies, and others(333–335). Importantly, cfDNA detection of pancreatic cancerrecurrence appears to precede radiographic detection of recur-rence by over sixmonths in some cases, providing a largerwindowfor potential intervention in this challenging disease (334). A newphylogenetic multiplex-PCR NGS platform approach to ctDNAprofiling provides sufficient sensitivity to both identify lungcancer patients destined to relapse within a year subclonal detec-tion. Of importance to early detection research, cfDNA wasrecently detected in plasma in patients with premalignant lungand bladder disease (336, 337). Differentially methylated loci inmultiple HPV and host genes were detected by NGS in urinecfDNAcould lead toprecision screening in cervical premalignancy(338). Somatic mutations in cfDNA among individuals withoutany precancer diagnosis, however, pose even more serious chal-lenges for the development of ctDNA screening tests (339).

Achieving acceptable levels of sensitivity and specificity forearly cancer detection will require further technical advances toidentify possible combinations of cancer-specific mutations anddefine potential quantitative thresholds to avoid overdiagnosis.Enrichment steps can be based on biological properties (e.g.,surface expressionmarkers) or physics (e.g., size, density, deform-ability). Based on findings to date, it can already be envisaged thatDNA sequencing needs to be broad, to encompass considerabletumor heterogeneity, and deep, to detect minute amounts ofctDNA fragments in the milieu of extensive genetically normalcfDNA. Nonmalignant conditions that can lead to the death ofnormal cells may also lead to a further dilution of ctDNA mole-cules and hamper quantitative evaluations. Very sensitive tech-nologies that allow the detection of less than 0.1% of ctDNA inblood plasma (e.g., digital droplet PCR or cancer personalizedprofiling [CAPP] by deep sequencing methods) have been devel-oped and applied to cfDNA analyses in patients with variousforms of cancer (340), but the key biological limitation might bethenumber of genomeequivalents present in blood samples fromearly-stage cancer patients (341).

Much work remains to be done in improving methodologiesfor detection of circulating tumor DNA. In addition to the needfor higher sensitivity, the specificity of cfDNA measurementsalso faces serious challenges. Several emerging reports haveshown that cancer-associated mutations are not restricted tocancer patients. For example, very low levels of TP53-mutatedcfDNA were observed in plasma of 11.4% of 123 matched non-cancer controls (339) and in the peritoneal fluid and periph-eral blood of women with benign ovarian lesions (327).Likewise, a potential confounding issue is the detection ofclonal alterations that arise in blood cells of healthy indivi-duals that may be associated with aging and clonal hemato-

poiesis (see below). Distinguishing between alterations ingenes associated with MDS/AML versus solid tumors may beone way to overcome this issue. Additionally, even whenmolecular alterations are identified, determining the tissue oforigin of the incipient neoplastic lesion can be extremelydifficult and complex. Combining sequencing of cfDNA withepigenetics markers (342, 343), mutational signatures, imag-ing and mathematical modeling can be used for pinpointingthe most likely tissue in which the clone originated. Evengreater issues challenge the promise of ctDNA analysis forearly cancer detection. Genomic alterations, such as those inthe BRAF, RAS, EGFR, HER2, FGFR3, PIK3CA, TP53, CDKN2A,and NF1/2 genes, all considered hallmark drivers of specificcancers, can also be identified in benign and premalignantconditions, occasionally at frequencies higher than in theirmalignant counterparts (214).

Analysis of other blood/fluid components, such as exosomes,platelets (344), urine, peritoneal fluid, and circulating tumorcells (CTCs), may help increase the sensitivity of detection(330, 336–338). CTCs are released early during tumor devel-opment (330) and have been found in patients with smallprimary tumors. Thus, detection of CTCs might be an alterna-tive approach to early cancer identification. Circulating tumorcells have been detected in 3% of patients with chronic obstruc-tive pulmonary disease (COPD) who have an elevated risk ofdeveloping lung cancer 1–4 years before annual CT-screeningdetected lung nodules (345), possibly complementing a CLIA-approved bronchial airway genomic classifier (FDA LaboratoryDeveloped Test [LDT]) validated for lung cancer detection(346), but were not detected in control individuals (bothsmokers and nonsmokers) with normal lung function. Tech-nical advances have established the feasibility of detecting andNGS of CTCs collected from blood at the very early steps oftumor invasion (347).

Focusing on population groups at an elevated risk of devel-oping cancer (e.g., COPD patients or individuals with inheritedcancer susceptibility) is a good strategy to speed up the processof testing and validation of emerging approaches and technol-ogies before considering large population-based screeningefforts as indicated above. CTCs have been detected in �10%of CRC precancers (adenomas) possibly stimulated by cytokine-driven epithelial migration (348). Enormous resources arerequired to identify genomic aberrations specific and sensitiveenough to detect early malignant lesions in large cohort studies.Finally, verification of the findings of approaches involvingliquid biopsy will call for acquisition of information on theputative location of the occult lesion to select the appropriateimaging modalities or other diagnostic means prior to planningappropriate therapies. In this context, tissue-specific multiplextranscriptome profiling of single CTCs (349) might be a prom-ising approach. With these approaches, it is possible to imaginea time when individuals at high risk of developing cancer due toeither genetic or environmental risk factors could be seriallymonitored using a blood-based test. A potential advantage ofthis approach would be the relative ease of compliance com-pared with other more invasive screening technologies. Ulti-mately, the specific methodology would be determined bypractical considerations such as cost, sensitivity, specificity, androbustness of the assays, but these approaches may foreverchange screening for cancers that are currently incurable unlessdiagnosed at an early stage.

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research16

Longitudinal Analysis of Premalignancies

It is likely that synchronous precancer/cancer pair studies willnot always accurately reflect the temporal clonal evolution under-lying neoplastic transformation, i.e., genetic alterations found inprecancer adjacent to cancer are not comparable to those foundin precancer in a patient who never progressed (as well shown inBarrett's esophagus). To fully appreciate such mechanisms, sys-tematic longitudinal analyses of malignant cells will be essential.To date, such analyses of epithelial premalignancies, with theexception of Barrett's esophagus, have been extremely limited,with reports of a relatively small number of patients with squa-mous lung premalignancy (350, 351). DNA- and RNA-seq ofthese lesions have identified molecular alterations in both epi-thelial cell signaling pathways and immune cell pathways thatassociate with progression of these premalignant lesions overtime. Surprising relationships between SCNA types and levels,mutation burden (including cancer drivers), cell cycle markers,and immunity have major implications on malignant transfor-mation. Large longitudinal studies of Barrett's esophagus estab-lished SCNAs as essential oncogenic drivers (352, 353). Non-progressors largely maintained stable genomes, in contrast topatients with high levels of SCNAs, genome doubling, geneticdiversity, and chromosomal instability/catastrophe associatedwith rapid (<2 years) progression to esophageal cancer (232,353).

Another recent Barrett's study of clonal evolution at single-cellresolution using multicolor FISH found that baseline geneticdiversity predicts progression (to cancer) and remains in stabledynamic equilibrium over time, suggesting that clonal make upand evolutionary trajectory of the lesion is predetermined fromthe outset (313). Clonal expansions were rare, often involvingp16. Importantly, this work has established the feasibility andmodel of longitudinal study in epithelial premalignancy.

Focusing on premalignancies of the blood has several advan-tages, including the ease of repeatedly acquiring neoplastic cells tostudy their clonal evolution, discoveries that inform single-cellsequencing studies in epithelial neoplasia. For example, study ofmyeloproliferative syndromes (201) provides the only direct datathat somatic mutation order (JAK2 and TET2) can greatly influ-ence disease features. The overall malignant transformation ratefor clonal hematopoiesis, monoclonal B-cell lymphocytosis(MBL), and monoclonal gammopathy of undetermined signifi-cance (MGUS) is about 1%–2% per year, but individual risk ishighly variable. Comprehensive single-cell and cfDNA omicsstudies will play a key role in improving our understanding ofdisease pathogenesis. We now have the ability to monitor hun-dreds of individual cells, thus overcoming bulk-cell/tissue limita-tions and allowing precise study of intraclonal and microenvi-ronment architecture and crosstalk in the process and timing oftransformation. The complexity and importance of follow-upis highlighted by findings in pancreas precancers, which canexhibit significant clonal heterogeneity/diversity that surprisinglydecreases during transformation (354).

Clonal hematopoiesis was initially discovered (355, 356) byNGS identification of somatic mutations in genes (mutantclones of mostly single driver mutations) similar to the muta-tional spectrum seen in MDS (notably DNMT3A, JAK2, TET2,and ASXL1) that increased markedly with age in the generalpopulation (357, 358). Although the mechanism is unclear,most people with clonal hematopoiesis of indeterminate poten-tial (CHIP) can stably harbor small hematopoietic clones for

long periods of time. These clones must have initially expandedto the point of being detectable but then were held in check(358). One possibility is that the somatically mutated clonesare able to persist in a finite number of pseudo-niches notavailable to normal stem cells. Once these pseudo-niches arefilled, the clone cannot expand further until additional muta-tions are acquired or the microenvironment is altered to createadditional area hospitable to these cells. This is analogousto why BRAF-mutant moles only grow so much before theybecome stable, until epigenetic reprogramming or some otherevent promotes transformation (see epigenetic section). Recentstudies have demonstrated how cells with a wide range of soma-tic mutations typical of myeloid malignancies (e.g., in U2AF1,SF3B1, SRSF2, ASXL1, or TET2) can induce inflammatory con-ditions and innate immune responses that then favor thegrowth of these mutant cells. Drugs targeting the inflammasomeand innate immune responses implicated in remodeling ofthe microenvironment are potential preventive approachesunder investigation (359, 360).

Whether adaptive immune surveillance plays a role in limitingthe expansion of pre-malignant clones is less clear. However,T-cellmediated autoimmunity is well documented in some formsof MDS and downregulation of MHC-II is common suggestingthat interactions with the adaptive immune system likely play arole in the development of thesemyeloidmalignancies. It appearsthat the aging bone marrow microenvironment promotes out-growth of clones driven bymutations in splicing factor genes (e.g.SF3B1, SRSF2) and is associated with innate and adaptiveimmune attrition (361). Murine findings suggest that aberrantsplicing may also produce neoantigens, capable of eliciting animmune response that would need to be suppressed or evaded forprogression to occur.

Aplastic anemia is a slightly different scenario, with two distinctprocesses occurring – immune mediated destruction of hemato-poietic cells and a constant stimulus for the regeneration of thesecells. Two kinds of clonal hematopoietic cells may be selected forin that environment – those that can evade the immune system(reverting to amore normal stem cell state) and those that acquiremutations promoting clonal expansion. Clonal hematopoiesis ishighly prevalent among patients with aplastic anemia (�50%), inwhich two broad types of genetic alterations are identified: (1)age-related mutations and CNAs commonly seen in myeloidmalignancies (e.g., DNMT3A, RUNX1, ASXL1) and (2) those notage related and highly specific to aplastic anemia – PIGA andBCOR/BCORL1 mutations and uniparental disomy (UPD) in 6p(6pUPD) identified SNP array karyotyping, which in contrast tothe first type, is associatedwith stable or decreasing clone size overtime and much lower rates progression to MDS/AML (362).6pUPD functionally results in deletion of half of the class I HLAlocus, providing a mechanism for evasion of an immuneresponse. Outside of this context, this event would not be pre-dicted to confer a clonal advantage. In aplastic anemia, this formof immune evasion would allow stem cells to function as theymight were the immune response not present. Clones with drivermutations in MDS-related genes, on the other hand, may gainproliferative or niche-independent growth capabilities that aremore oncogenic. This is born out by observations that 6pUPD ismore common in children with aplastic anemia, an age-groupwhere CHIP mutations are rare and that 6pUPD is exceedinglyrare in elderly MDS and AML. This suggests that 6pUPD is not anoncogenic event but only allows stem cells to persist in the setting

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 17

of autoimmunity. Nearly 40% of clonal hematopoiesis indivi-duals with unexplained cytopenias harbor detectable mutations(200, 358), many with clones having more than one drivermutation and higher risk of transformation to MDS/AML andall-cause mortality (200, 358, 363). Cooperating mutations alsocan be identified during periods of clonally skewed hematopoi-esis in sporadic and hereditary settings that precede myeloidtransformation. DNMT3A haploinsufficiency transformed FLT-mutant myeloproliferative disease into AML in mouse model(364). The frequent co-occurrence of germlineGATA2 and somat-icASXL1 events (365) and germline SNPs associatedwith somaticmutations of JAK2 have uncovered several targetable cooperativemutations (366) driving premalignant progression (367). Exam-ples of hereditarymutated transcription factors that predispose tohematologic neoplasia include mutations in CEBPA, RUNX1,ETV6, and PAX5 (367). A recent GWAS identified germlinevariants that predispose toboth JAK2V617F clonal hematopoiesisand myeloproliferative neoplasms (366). Four genes (JAK2,SH2B3, CHEK2, and TET2) altered in both inherited and somaticsettings, contribute to V617F clonal hematopoeisis and/or MPNdevelopment. This study’s identification of a predisposition alleleassociated with TET2 is intriguing since somatic TET2 mutationsare common early events in myeloid precursors, including clonalhematopoiesis and myeloproliferative neoplasms, and can beidentified in hematopoietic stem cells, either preceding or fol-lowing the acquisition of V617F, and the mutational order ofthese two genes can influence the clinical and biologic behavior ofthese neoplasms (201).

MBL is an asymptomatic expansion of clonal B cells in theperipheral blood present in roughly 4% of all U.S. individualsover the age of 40 years (368). Genetic predisposition to MBL issuggested by the finding that the incidence ofMBL is 3-fold higherfor individuals within familial chronic lymphocytic leukemia(CLL) kindreds (defined as families with at least two first-degreerelatives with CLL). A large GWAS of CLL families (including60 relatives with MBLs) found significant germline variant asso-ciations in two out of eight regions tested (369). NGS has shownthat most mutations and intraclonal heterogeneity found in CLLare already present in MBL years before progression (370, 371).Furthermore, longitudinal MBL studies, including those frompatients who progressed to CLL, have begun to elucidate thesequence, timing, and impact of subclonal expansion and T-cellexhaustion on malignant transformation (369, 372–374). Risksof serious bacterial infections in individuals with MBL are similarto those with CLL (375–377) and linked to MBL transformation(378, 379) and solid tumor risk is 3- to 4-fold higher in MBL andCLL (versus healthy controls), all thought to be due to defectsin immune surveillance (376, 377). Antibody responses to pri-mary and secondary antigen challenges are typically inefficientamong patients with early-stage CLL. Preliminary data haveshown that immunologic T-cell synapse is defective in individualswith MBL as well (368). Efforts to generate efficient vaccineresponses will, therefore, be challenging. Enhancement of cyto-lytic T-cell function in MBL via vaccine therapy should be a long-term challenge as graft-versus-leukemia is highly effective ineradicating leukemic B-cells. Both in vitro and in vivo data suggestthat lenalidomide can repair defects in the T-cell immune synapseand reduce Tregs in CLL patients (380, 381), an approach cur-rently being tested clinically in MBL. Of note, new mouse modeldata suggest that age and inflammatory status of the host micro-environment promotes selection for adaptive oncogenic events

in B-cell progenitors (382), analogous to clonal hematopoesis.Related work involves a hereditary syndrome of susceptibility topre–B-cell neoplasia caused by inherited mutations of PAX5.Recent mechanistic data indicated that inherited susceptibilityand aberrant immune responses to postnatal infections drivesB-cell clonal evolution of premalignant B-cells and transforma-tion to leukemia and lymphoma, by showing that pre-B-ALLwas initiated in Pax5 heterozygous mice only when exposed tocommon infections (383).

Studies of MGUS have provided some of the best evidencethat the immune system has the capacity to recognize precursorstates (384). Search for shared targets of immune responseled to the finding that T cells against stemness antigens (suchas SOX2) are particularly enriched in MGUS (versus MM;ref. 385). Prospective data demonstrate that SOX2 immunitycorrelates with risk of transformation (230). Prevention strat-egies include boosting preexisting T-cell immunity, for exam-ple, with SOX2 vaccines and immune-modulatory drugs (386).NGS has highlighted clonal evolution and heterogeneity inlongitudinal studies (387, 388). Similar to clonal hematopoi-esis above, MGUS cells demonstrate clinical dormancy despiteNGS suggesting that the majority of genomic alterations foundin MM are found in precursor gammopathies (387, 388).Interestingly, new humanized models developed to grow pre-cursor cells in vivo indicate that MGUS cells have the capacity forprogressive growth, suggesting that the clinical stability/dorman-cy of these cells is in part mediated by features extrinsic to tumorcells, such as the immune system or bone marrow niche, wheresignals derived from osteoblasts may be important for mediatingdormancy of MGUS cells (389). These humanized models togrow premalignant cells in vivo should greatly advance the studyof clonal evolution and malignant transformation in this set-ting (390). Inherited genetic variation in specific SNPs increasesMGUS predisposition and risk of transforming toMM (391). Lociidentified also points to a role for chronic antigen-driven stimu-lation in driving clonal origins and evolution in MM and otherB-cell tumors. The risk of MM is particularly increased in somepopulations such as those with inherited lipid storage disorders,such as germline GBA mutations in Gaucher disease (GD), duein part to lysolipid-induced chronic inflammation and genomicinstability. GD mouse models, for example, lysolipid substratereduction in Gba1-deficient mice decreased the risk of gammopa-thies, studies already show marked preventive efficacy of tar-geting the underlying trigger of genomic instability (392, 393).This led to the recent discovery that in nearly 25% of all cases ofMGUS/MM, the underlying clonemay be driven by lipid antigenssuch as inflammation-associated bioactive lipids (392). Studiesto characterize the precise nature of microbial/dietary/endo-genous lipid antigens in the setting of lipid-reactive gammopathywill facilitate targeted prevention with pharmacologic/lifestylechanges.

SummaryWhile a number of interventions are already FDA approved for

cancer prevention (detailed in ref. 394), this is only the tip of theiceberg. New precision prevention approaches will be needed,including novel trial designs (e.g., involving molecular selection;refs. 395, 396) and agents (e.g., denosumab for BRCA1 carriers;ref. 37)). Developing cancer vaccines as potent as polio, diphthe-ria, and rubella vaccines would protect future generations from

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research18

developing cancer. As in cervical cancer, where vaccination againstHPV has virtually eliminated the disease, the development ofcancer vaccines to stimulate T-cells to recognize precancer anti-gens as foreign may prevent cancer. The communication betweenthe immune system and neoplasia reflects a fundamental princi-ple that is applicable to all organ/cell types and continues tobecome more involved (e.g., the complex microenvironmentmilieu now includes adipocytes, myocytes, pericytes, chondro-cytes, osteoblasts, fibroblasts, neutrophils; vascular, neuronal,and B-cells; extracellular matrix and secreted inflammatory med-iators; mutations and SCNAs; and systemic microbial-immunenetworks; ref. 397). The human genome, microbiome, and dietshare a complex set of interdependencies requiring further bio-chemical study, including metabolomics. Recent systematic,genome-wide transcriptional and epigenetic studies have begunto uncover the role of heterogeneity and variability across mul-tiple immune cell types (398).

Analogous to clonal hematopoiesis, albeit in much smallernumbers of individuals studied to date, ultra-deepNGS of somat-ic cancermutations reported a distinct mutational profile (mostlyNOTCH1) in normal skin (161), potentially MBL (382), andliquid-biopsy technology is detecting very low levels of TP53-mutated cfDNA were observed in plasma of matched non-cancercontrols (339) and in peritoneal fluid and peripheral blood fromwomen with benign ovarian lesions (399), likely representing apremalignant mutational background events that accumulates incancer and aging. Similar to BRAF-mutant moles, CHIP clonesseem to expand to the point of being detectable but then are heldin check (358), until an epigenetic, immune, or other transform-ing event. The aging bone marrow microenvironment promotesclonal outgrowth driven by splicing factor mutations with innateand adaptive immune attrition (361). Somatic myeloid muta-tions can induce inflammation and innate immunity favoringclonal growth. T-cell immunity and downregulation ofMHC-II inMDS suggests a role for adaptive immunity in myeloid dysplasia.Probing germline (raremutations and commonvariants)-somaticinteractions is leading to novel prevention targets (61, 65). Single-cell tools/approaches are beginning to reveal immense heteroge-neity of certain precancers and are needed to dissect the detailedinteractions of precancer cells with immune cells, and othercomponents of the tumormicroenvironment. Establishing a PCAthat integrates multi-omics and immunity (Figure 1) will becritical to better interrogate and disrupt the disabling immuno-suppressive properties in the tumor microenvironment and toidentify immunogenic antigens to design vaccines to activate/reprogram the immune response to detect, prevent, and rejectprecancers.

As indicated in this Perspective, the complexities of this pro-ject are daunting, but the disruptive tools to make the PCAfeasible are coming on line and the potential public healthbenefits of preventing cancer are immeasurable. To accelerate

the prevention of cancer, this field needs a large-scale, longi-tudinal effort, leveraging major initiatives and infrastructure,diverse disciplines, technologies, and models (e.g., involvingstem/progenitor cells) (400–402) to develop a multi-omics andimmunity PCA. This unified Atlas will provide a vast nationalresource for discovery to interrogate, target, and interceptevents that drive oncogenesis.

In summary, to fully achieve cancer prevention, we mustbuild teams with multiple areas of expertise from the NIH,academia, FDA, private foundations, philanthropic partners,and industry. The best analogy of assembling such a multi-disciplinary team is the Manhattan Project—one goal, multipleexperts.

Disclosure of Potential Conflicts of InterestA.E. Spira reports receiving a commercial research grant from Janssen

Pharmaceuticals and is a consultant/advisory board member for JanssenPharmaceuticals and Veracyte Inc. M.B. Yurgelun reports receiving a com-mercial research grant from Myriad Genetic Laboratories, Inc. R. Bejar is aconsultant/advisory board member for Genoptix, Celgene, FoundationMedicine, and Alexion. J.E. Garber reports receiving other commercialresearch support from Novartis and Ambry Genetics and is a consultant/advisory board member for Biogen, Novartis, and GTx. V.E. Velculescu hasownership interest (including patents) in Personal Genome Diagnostics andis a consultant/advisory board member for Personal Genome Diagnostics.M.L. Disis reports receiving a commercial research grant from VentiRx,Celgene, Jannsen, Seattle Genetics, and EMD Sorono and has ownershipinterest (including patents) in VentiRx and Epithany. No potential conflictsof interest were disclosed by the other authors.

Authors' ContributionsConception and design: A.E. Spira, M.B. Yurgelun, R. Bejar, E. Vilar, T.R.Rebbeck, V.E. Velculescu, S.M. LippmanAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): S.M. LippmanAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computationalanalysis):M.B.Yurgelun,T.R.Rebbeck,D.Wallace,S.M.LippmanWriting, review, and/or revision of the manuscript: A.E. Spira, M.B. Yurgelun,L. Alexandrov, A. Rao, R. Bejar, K. Polyak, M. Giannakis, A. Shilatifard, O.J. Finn,M. Dhodapkar, N.E. Kay, E. Braggio, E. Vilar, S.A. Mazzilli, T.R. Rebbeck, J.E.Garber, V.E. Velculescu, M.L. Disis, D. Wallace, S.M. LippmanStudy supervision: S.M. Lippman

AcknowledgmentsThe authors thank Jennifer Beane, PhD, for her input on the single-cell se-

quencing section and Leona Flores, PhD, for editorial assistancewith this article.

Grant SupportA.E. Spirawas supported byNIH/NCI 1U01CA214182 and5U01CA196408.

S.M. Lippman was supported for this work by NCIP30-CA023100-29. ASand SML co-chair the PremalignanT Cancer Genome Atlas (PreTCGA) Demon-stration Project, NCI Blue Ribbon Panel.

Received August 24, 2016; revised January 20, 2017; accepted January 20,2017; published online April 3, 2017.

References1. Hudson KL, Collins FS. The 21st Century Cures Act – A View from the

NIH. N Engl J Med 2017;376:111–3.2. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA Jr, Kinzler

KW. Cancer genome landscapes. Science 2013;339:1546–58.3. Borras E, San Lucas FA, Chang K, Zhou R, Masand G, Fowler J, et al.

Genomic landscape of colorectal mucosa and adenomas. Cancer Prev Res2016;9:417–27.

4. Blackburn EH. Cancer interception. Cancer Prev Res 2011;4:787–92.5. Finn OJ. Vaccines for cancer prevention: a practical and feasible

approach to the cancer epidemic. Cancer Immunol Res 2014;2:708–13.

6. Spira A, Disis ML, Schiller JT, Vilar E, Rebbeck TR, Bejar R, et al.Leveraging premalignant biology for immune-based cancer prevention.Proc Natl Acad Sci U S A 2016;113:10750–8.

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 19

7. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell2011;144:646–74.

8. Livraghi L, Garber JE. PARP inhibitors in the management of breastcancer: current data and future prospects. BMC Med 2015;13:188.

9. ToC, KimEH, RoyceDB,WilliamsCR,Collins RM, Risingsong R, et al. ThePARP inhibitors, veliparib and olaparib, are effective chemopreventiveagents for delaying mammary tumor development in BRCA1-deficientmice. Cancer Prev Res 2014;7:698–707.

10. Dudley JC, Lin MT, Le DT, Eshleman JR. Microsatellite instability as abiomarker for PD-1 blockade. Clin Cancer Res 2016;22:813–20.

11. Le DT, Uram JN,WangH, Bartlett BR, KemberlingH, Eyring AD, et al. PD-1 blockade in tumors with mismatch-repair deficiency. N Engl J Med2015;372:2509–20.

12. Samadder NJ, Neklason DW, Boucher KM, Byrne KR, Kanth P, SamowitzW, et al. Effect of sulindac and erlotinib vs placebo on duodenal neoplasiain familial adenomatous polyposis: A Randomized Clinical Trial. JAMA2016;315:1266–75.

13. Kamihara J, Rana HQ, Garber JE. Germline TP53 mutations andthe changing landscape of Li-Fraumeni syndrome. Hum Mutat 2014;35:654–62.

14. TungN, LinNU,Kidd J, AllenBA, SinghN,WenstrupRJ, et al. Frequency ofgermlinemutations in 25 cancer susceptibility genes in a sequential seriesof patients with breast cancer. J Clin Oncol 2016;34:1460–8.

15. Yurgelun MB, Kulke MH, Fuchs CS, Allen BA, Uno H, Hornick JL, et al.Cancer susceptibility gene mutations in individuals with colorectalcancer. J Clin Oncol 2017:JCO2016710012.

16. Zhang J, Walsh MF, Wu G, Edmonson MN, Gruber TA, Easton J, et al.Germline mutations in predisposition genes in pediatric cancer. N Engl JMed 2015;373:2336–46.

17. Rahman N. Realizing the promise of cancer predisposition genes. Nature2014;505:302–8.

18. Sijmons RH, Hofstra RM. Review: clinical aspects of hereditary DNAMismatch repair gene mutations. DNA Repair (Amst) 2016;38:155–62.

19. Belcheva A, Irrazabal T, Robertson SJ, Streutker C, Maughan H, Rubino S,et al. Gut microbial metabolism drives transformation ofMSH2-deficientcolon epithelial cells. Cell 2014;158:288–99.

20. Montrose DC, Zhou XK, McNally EM, Sue E, Yantiss RK, Gross SS, et al.Celecoxib alters the intestinal microbiota andmetabolome in associationwith reducing polyp burden. Cancer Prev Res 2016;9:721–31.

21. WidschwendterM, Rosenthal AN, Philpott S, Rizzuto I, Fraser L, HaywardJ, et al. The sex hormone system in carriers of BRCA1/2 mutations: a case-control study. Lancet Oncol 2013;14:1226–32.

22. Communal L, VilascoM,Hugon-Rodin J, Courtin A,MourraN, LahlouN,et al. Proliferation and ovarian hormone signaling are impaired in normalbreast tissues from women with BRCA1 mutations: benefit of a proges-terone receptormodulator treatment as a breast cancer preventive strategyin women with inherited BRCA1 mutations. Oncotarget 2016;7:45317–30.

23. Sedic M, Kuperwasser C. BRCA1-hapoinsufficiency: Unraveling themolecular and cellular basis for tissue-specific cancer. Cell Cycle 2016;15:621–7.

24. Uziel O, Yerushalmi R, Zuriano L, Naser S, Beery E, Nordenberg J, et al.BRCA1/2 mutations perturb telomere biology: characterization of struc-tural and functional abnormalities in vitro and in vivo. Oncotarget 2016;7:2433–54.

25. Eakin CM, Maccoss MJ, Finney GL, Klevit RE. Estrogen receptor alpha is aputative substrate for the BRCA1ubiquitin ligase. ProcNatl Acad Sci U S A2007;104:5794–9.

26. Gross K, Wronski A, Skibinski A, Phillips S, Kuperwasser C. Cell fatedecisions during breast cancer development. J Dev Biol 2016;4:4.

27. CainoMC, Seo JH, Aguinaldo A,Wait E, Bryant KG, Kossenkov AV, et al. Aneuronal network of mitochondrial dynamics regulates metastasis.Nat Commun 2016;7:13730.

28. Sumpter RJr, Sirasanagandla S, Fernandez AF, Wei Y, Dong X, Franco L,et al. Fanconi anemia proteins function in mitophagy and immunity.Cell 2016;165:867–81.

29. Xu L, Zhao Y, Chen Z, Wang Y, Chen L, Wang S. Tamoxifen and risk ofcontralateral breast cancer among women with inherited mutations inBRCA1 and BRCA2: a meta-analysis. Breast Cancer 2015;22:327–34.

30. Antoniou AC, Kartsonaki C, Sinilnikova OM, Soucy P, McGuffog L,Healey S, et al. Common alleles at 6q25.1 and 1p11.2 are associated

with breast cancer risk for BRCA1 andBRCA2mutation carriers.HumMolGenet 2011;20:3304–21.

31. Bramley M, Clarke RB, Howell A, Evans DG, Armer T, Baildam AD, et al.Effects of oestrogens and anti-oestrogens on normal breast tissue fromwomen bearing BRCA1 and BRCA2 mutations. Br J Cancer 2006;94:1021–8.

32. Jones LP, Tilli MT, Assefnia S, Torre K, Halama ED, Parrish A, et al.Activation of estrogen signaling pathways collaborates with loss of Brca1to promote development of ERalpha-negative and ERalpha-positivemammary preneoplasia and cancer. Oncogene 2008;27:794–802.

33. Gorrini C, Gang BP, Bassi C, Wakeham A, Baniasadi SP, Hao Z, et al.Estrogen controls the survival of BRCA1-deficient cells via a PI3K-NRF2-regulated pathway. Proc Natl Acad Sci U S A 2014;111:4472–7.

34. Corriden R, Hollands A, Olson J, Derieux J, Lopez J, Chang JT, et al.Tamoxifen augments the innate immune function of neutrophils throughmodulation of intracellular ceramide. Nat Commun 2015;6:8369.

35. Welte T, Zhang XH, Rosen JM. Repurposing antiestrogens for tumorimmunotherapy. Cancer Discov 2017;7:17–9.

36. Nolan E, Lindeman GJ, Visvader JE. Out-RANKing BRCA1 in mutationcarriers. Cancer Res 2017;77:595–600.

37. Nolan E, Vaillant F, BranstetterD, Pal B,GinerG,Whitehead L, et al. RANKligand as a potential target for breast cancer prevention in BRCA1-mutation carriers. Nat Med 2016;22:933–9.

38. Sigl V, Owusu-Boaitey K, Joshi PA, Kavirayani A,Wirnsberger G, Novatch-kova M, et al. RANKL/RANK control Brca1 mutation-driven mammarytumors. Cell Res 2016;26:761–74.

39. Sau A, Lau R, CabritaMA,Nolan E, Crooks PA, Visvader JE, et al. Persistentactivation ofNF-kappaB in BRCA1-deficientmammary progenitors drivesaberrant proliferation and accumulation of DNA damage. Cell Stem Cell2016;19:52–65.

40. PalafoxM, Ferrer I, Pellegrini P, Vila S,Hernandez-Ortega S,Urruticoechea A,et al. RANK induces epithelial-mesenchymal transition and stemness inhuman mammary epithelial cells and promotes tumorigenesis andmetastasis. Cancer Res 2012;72:2879–88.

41. Mai QG, Zhang ZM, Xu S, Lu M, Zhou RP, Zhao L, et al. Metforminstimulates osteoprotegerin and reduces RANKL expression in osteoblastsand ovariectomized rats. J Cell Biochem 2011;112:2902–9.

42. Galluzzi L, Buque A, Kroemer G. Prevention of breast cancer by RANKL/-RANK blockade. Cell Res 2016;26:751–2.

43. TanW, ZhangW, Strasner A,Grivennikov S, Cheng JQ,Hoffman RM, et al.Tumour-infiltrating regulatory T cells stimulate mammary cancer metas-tasis through RANKL-RANK signalling. Nature 2011;470:548–53.

44. Luo J, Yang Z,MaY, Yue Z, LinH,QuG, et al. LGR4 is a receptor for RANKLand negatively regulates osteoclast differentiation and bone resorption.Nat Med 2016;22:539–46.

45. Rebbeck TR,Mitra N,Wan F, SinilnikovaOM,Healey S,McGuffog L, et al.Association of type and location of BRCA1 and BRCA2 mutations withrisk of breast and ovarian cancer. JAMA 2015;313:1347–61.

46. WidschwendterM, BurnellM, Fraser L, Rosenthal AN, Philpott S, ReiselD,et al. Osteoprotegerin (OPG), the endogenous inhibitor of receptoractivator of NF-kappaB ligand (RANKL), is dysregulated in BRCA muta-tion carriers. EBioMedicine 2015;2:1331–9.

47. Gnant M, Pfeiler G, Dubsky PC, Hubalek M, Greil R, Jakesz R, et al.Adjuvant denosumab in breast cancer (ABCSG-18): a multicentre,randomised, double-blind, placebo-controlled trial. Lancet 2015;386:433–43.

48. Kotsopoulos J, Huzarski T, Gronwald J, Singer CF, Moller P, Lynch HT,et al. Bilateral oophorectomy and breast cancer risk in BRCA1 and BRCA2mutation carriers. J Natl Cancer Inst 2017;109.

49. VasenH, Ibrahim I, PonceCG, Slater EP,Matthai E, CarratoA, et al. Benefitof surveillance for pancreatic cancer in high-risk individuals: outcome oflong-term prospective follow-up studies from three European ExpertCenters. J Clin Oncol 2016;34:2010–9.

50. Salo-Mullen EE, O'Reilly EM, Kelsen DP, Ashraf AM, Lowery MA, Yu KH,et al. Identification of germline genetic mutations in patients withpancreatic cancer. Cancer 2015;121:4382–8.

51. Beatty PL, van der Geest R, Hashash JG, Kimura T, Gutkin D, Brand RE,et al. Immunobiology and immunosurveillance in patients with intra-ductal papillary mucinous neoplasms (IPMNs), premalignant precursorsof pancreatic adenocarcinomas. Cancer Immunol Immunother 2016;65:771–8.

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research20

52. Potjer TP, Schot I, Langer P, Heverhagen JT, Wasser MN, Slater EP, et al.Variation in precursor lesions of pancreatic cancer among high-riskgroups. Clin Cancer Res 2013;19:442–9.

53. Alsop K, Fereday S, Meldrum C, deFazio A, Emmanuel C, George J, et al.BRCA mutation frequency and patterns of treatment response in BRCAmutation-positive women with ovarian cancer: a report from the Austra-lian Ovarian Cancer Study Group. J Clin Oncol 2012;30:2654–63.

54. Serrao EM, KettunenMI, Rodrigues TB,Dzien P,Wright AJ, GopinathanA,et al. MRI with hyperpolarised [1–13C]pyruvate detects advanced pan-creatic preneoplasia prior to invasive disease in a mouse model. Gut2016;65:465–75.

55. Saeterdal I, Bjorheim J, Lislerud K, Gjertsen MK, Bukholm IK, Olsen OC,et al. Frameshift-mutation-derived peptides as tumor-specific antigens ininherited and spontaneous colorectal cancer. Proc Natl Acad Sci U S A2001;98:13255–60.

56. Schwitalle Y, Kloor M, Eiermann S, LinnebacherM, Kienle P, Knaebel HP,et al. Immune response against frameshift-induced neopeptides inHNPCC patients and healthy HNPCC mutation carriers. Gastroenterol-ogy 2008;134:988–97.

57. Goodfellow PJ, Billingsley CC, Lankes HA, Ali S, Cohn DE, Broaddus RJ,et al. Combined microsatellite instability, MLH1 methylation analysis,and immunohistochemistry for lynch syndrome screening in endometrialcancers from GOG210: An NRG Oncology and Gynecologic OncologyGroup Study. J Clin Oncol 2015;33:4301–8.

58. Ait OuakrimD, Dashti SG, Chau R, BuchananDD, ClendenningM, RostyC, et al. Aspirin, ibuprofen, and the risk of colorectal cancer in lynchsyndrome. J Natl Cancer Inst 2015;107.

59. Burn J, Gerdes AM, Macrae F, Mecklin JP, Moeslein G, Olschwang S, et al.Long-term effect of aspirin on cancer risk in carriers of hereditary colo-rectal cancer: an analysis from the CAPP2 randomised controlled trial.Lancet 2011;378:2081–7.

60. Oxnard GR, Heng JC, Root EJ, Rainville IR, Sable-Hunt AL, Shane-CarsonKP, et al. Initial results of a prospective,multicenter trial to study inheritedlung cancer risk associated with germline EGFR T790M: INHERIT EGFR. JClin Oncol 2015.

61. Lu C, Xie M, Wendl MC, Wang J, McLellan MD, Leiserson MD, et al.Patterns and functional implications of rare germline variants across 12cancer types. Nat Commun 2015;6:10086.

62. Campbell PJ. Somatic and germline genetics at the JAK2 locus. Nat Genet2009;41:385–6.

63. Liu W, He L, Ramirez J, Krishnaswamy S, Kanteti R, Wang YC, et al.Functional EGFR germline polymorphisms may confer risk for EGFRsomatic mutations in non-small cell lung cancer, with a predominanteffect on exon 19 microdeletions. Cancer Res 2011;71:2423–7.

64. Li Q, Seo JH, Stranger B, McKenna A, Pe'er I, Laframboise T, et al.Integrative eQTL-based analyses reveal the biology of breast cancer riskloci. Cell 2013;152:633–41.

65. Carter H,Marty R, HofreeM, Gross A, Jensen J, Fisch KM, et al. Interactionlandscape of inherited polymorphismswith somatic events in cancer. NatCommun in press.

66. Parsons DW, Roy A, Yang Y, Wang T, Scollon S, Bergstrom K, et al.Diagnostic yield of clinical tumor and germline whole-exome sequencingfor children with solid tumors. JAMA Oncol 2016.

67. Schrader KA, Cheng DT, Joseph V, Prasad M, Walsh M, Zehir A, et al.Germline variants in targeted tumor sequencing using matched normalDNA. JAMA Oncol 2016;2:104–11.

68. Stoffel EM, Fearon ER. Germline sequence variants and ovarian cancer:known-knowns and known-unknowns. JAMA Oncol 2016;2:491–2.

69. Kar SP, Beesley J, Amin Al Olama A, Michailidou K, Tyrer J, Kote-Jarai Z,et al. Genome-wide meta-analyses of breast, ovarian, and prostate cancerassociation studies identify multiple new susceptibility loci shared by atleast two cancer types. Cancer Discov 2016;6:1052–67.

70. Yurgelun MB, Chenevix-Trench G, Lippman SM. Translating germlinecancer risk into precision prevention. Cell in press.

71. Milne RL, Antoniou AC. Modifiers of breast and ovarian cancer risks forBRCA1 and BRCA2 mutation carriers. Endocr Relat Cancer 2016;23:T69–84.

72. Benitez-Buelga C, Vaclova T, Ferreira S, Urioste M, Inglada-Perez L,Soberon N, et al. Molecular insights into the OGG1 gene, a cancer riskmodifier in BRCA1 and BRCA2 mutations carriers. Oncotarget 2016;7:25815–25.

73. Amos CI, Dennis J, Wang Z, Byun J, Schumacher FR, Gayther SA, et al. TheOncoArray Consortium: a network for understanding the genetic archi-tecture of common cancers. Cancer Epidemiol Biomarkers Prev2017;26:126–35.

74. Cargill M, Schrodi SJ, Chang M, Garcia VE, Brandon R, Callis KP, et al. Alarge-scale genetic association study confirms IL12B and leads to theidentification of IL23R as psoriasis-risk genes. Am J HumGenet 2007;80:273–90.

75. Dunning AM, Michailidou K, Kuchenbaecker KB, Thompson D, FrenchJD, Beesley J, et al. Breast cancer risk variants at 6q25 display differentphenotype associations and regulate ESR1, RMND1 and CCDC170.Nat Genet 2016;48:374–86.

76. Chen L, Ge B, Casale FP, Vasquez L, Kwan T, Garrido-Martin D, et al.Genetic drivers of epigenetic and transcriptional variation in humanimmune cells. Cell 2016;167:1398–414e24.

77. Cescon DW, Haibe-Kains B, Mak TW. APOBEC3B expression in breastcancer reflects cellular proliferation, while a deletion polymorphism isassociated with immune activation. Proc Natl Acad Sci U S A 2015;112:2841–6.

78. Middlebrooks CD, Banday AR, Matsuda K, Udquim KI, Onabajo OO,Paquin A, et al. Association of germline variants in the APOBEC3 regionwith cancer risk and enrichment with APOBEC-signature mutations intumors. Nat Genet 2016;48:1330–8.

79. WenWX, Soo JS, Kwan PY, Hong E, Khang TF,Mariapun S, et al. GermlineAPOBEC3B deletion is associated with breast cancer risk in an Asianmulti-ethnic cohort andwith immune cell presentation. Breast Cancer Res2016;18:56.

80. Nik-Zainal S, Wedge DC, Alexandrov LB, Petljak M, Butler AP, Bolli N,et al. Association of a germline copy number polymorphism of APO-BEC3A and APOBEC3B with burden of putative APOBEC-dependentmutations in breast cancer. Nat Genet 2014;46:487–91.

81. Wallace DC. Mitochondrial DNA variation in human radiation anddisease. Cell 2015;163:33–8.

82. DeBalsi KL, Hoff KE, Copeland WC. Role of the mitochondrial DNAreplication machinery in mitochondrial DNA mutagenesis, aging andage-related diseases. Ageing Res Rev 2017;33:89–104.

83. Pavlova NN, Thompson CB. The emerging hallmarks of cancer metab-olism. Cell Metab 2016;23:27–47.

84. Brandon M, Baldi P, Wallace DC. Mitochondrial mutations in cancer.Oncogene 2006;25:4647–62.

85. Arnold RS, Sun Q, Sun CQ, Richards JC, O'Hearn S, Osunkoya AO, et al.An inherited heteroplasmic mutation in mitochondrial gene COI in apatient with prostate cancer alters reactive oxygen, reactive nitrogen andproliferation. Biomed Res Int 2013;2013:239257.

86. Petros JA, Baumann AK, Ruiz-Pesini E, Amin MB, Sun CQ, Hall J, et al.mtDNA mutations increase tumorigenicity in prostate cancer. ProcNatl Acad Sci U S A 2005;102:719–24.

87. Permuth-Wey J, Chen YA, Tsai YY, Chen Z, Qu X, Lancaster JM, et al.Inherited variants in mitochondrial biogenesis genes may influenceepithelial ovarian cancer risk. Cancer Epidemiol Biomarkers Prev 2011;20:1131–45.

88. Williams SB, Ye Y, Huang M, Chang DW, Kamat AM, Pu X, et al.Mitochondrial DNA content as risk factor for bladder cancer and itsassociation with mitochondrial DNA polymorphisms. Cancer Prev Res2015;8:607–13.

89. Canter JA, Kallianpur AR, Parl FF, Millikan RC. Mitochondrial DNAG10398A polymorphism and invasive breast cancer in African-Americanwomen. Cancer Res 2005;65:8028–33.

90. Liu VW,Wang Y, YangHJ, Tsang PC,Ng TY,Wong LC, et al.MitochondrialDNA variant 16189T>C is associated with susceptibility to endometrialcancer. Hum Mutat 2003;22:173–4.

91. Zhai K, Chang L, Zhang Q, Liu B, Wu Y. Mitochondrial C150T polymor-phism increases the risk of cervical cancer and HPV infection. Mitochon-drion 2011;11:559–63.

92. Schroder T, Kucharczyk D, Bar F, Pagel R, Derer S, Jendrek ST, et al.Mitochondrial gene polymorphisms alter hepatic cellular energy metab-olism and aggravate diet-induced non-alcoholic steatohepatitis. MolMetab 2016;5:283–95.

93. FanW, Waymire KG, Narula N, Li P, Rocher C, Coskun PE, et al. A mousemodel of mitochondrial disease reveals germline selection against severemtDNA mutations. Science 2008;319:958–62.

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 21

94. Lin CS, Sharpley MS, Fan W, Waymire KG, Sadun AA, Carelli V, et al.Mouse mtDNAmutant model of Leber hereditary optic neuropathy. ProcNatl Acad Sci U S A 2012;109:20065–70.

95. Gasparre G, Iommarini L, Porcelli AM, Lang M, Ferri GG, Kurelac I,et al. An inherited mitochondrial DNA disruptive mutationshifts to homoplasmy in oncocytic tumor cells. Hum Mutat 2009;30:391–6.

96. Ishikawa K, Takenaga K, Akimoto M, Koshikawa N, Yamaguchi A, Ima-nishiH, et al. ROS-generatingmitochondrial DNAmutations can regulatetumor cell metastasis. Science 2008;320:661–4.

97. O'Keefe SJ. Diet,microorganisms and theirmetabolites, and colon cancer.Nat Rev Gastroenterol Hepatol 2016;13:691–706.

98. O'Keefe SJ, Li JV, Lahti L, Ou J, Carbonero F,MohammedK, et al. Fat, fibreand cancer risk in African Americans and rural Africans. Nat Commun2015;6:6342.

99. Feeley KP, Bray AW,WestbrookDG, Johnson LW, Kesterson RA, BallingerSW, et al. Mitochondrial genetics regulate breast cancer tumorigenicityand metastatic potential. Cancer Res 2015;75:4429–36.

100. Schriner SE, Linford NJ, Martin GM, Treuting P, Ogburn CE, Emond M,et al. Extension of murine life span by overexpression of catalase targetedto mitochondria. Science 2005;308:1909–11.

101. Goh J, Enns L, Fatemie S, Hopkins H, Morton J, Pettan-Brewer C, et al.Mitochondrial targeted catalase suppresses invasive breast cancer inmice.BMC Cancer 2011;11:191.

102. Latorre-Pellicer A, Moreno-Loshuertos R, Lechuga-Vieco AV, Sanchez-Cabo F, Torroja C, Acin-Perez R, et al. Mitochondrial and nuclear DNAmatching shapes metabolism and healthy ageing. Nature 2016;535:561–5.

103. Picard M, Zhang J, Hancock S, Derbeneva O, Golhar R, Golik P, et al.Progressive increase in mtDNA 3243A>G heteroplasmy causes abrupttranscriptional reprogramming. Proc Natl Acad Sci U S A 2014;111:E4033–42.

104. Chen PL, Chen CF, Chen Y, Guo XE, Huang CK, Shew JY, et al. Mito-chondrial genome instability resulting from SUV3 haploinsufficiencyleads to tumorigenesis and shortened lifespan. Oncogene 2013;32:1193–201.

105. Frezza C, Zheng L, Folger O, Rajagopalan KN, MacKenzie ED, Jerby L,et al. Haem oxygenase is synthetically lethal with the tumour sup-pressor fumarate hydratase. Nature 2011;477:225–8.

106. Xiao M, Yang H, XuW, Ma S, Lin H, Zhu H, et al. Inhibition of alpha-KG-dependent histone and DNA demethylases by fumarate and succinatethat are accumulated in mutations of FH and SDH tumor suppressors.Genes Dev 2012;26:1326–38.

107. Birsoy K, Possemato R, Lorbeer FK, Bayraktar EC, Thiru P, Yucel B, et al.Metabolic determinants of cancer cell sensitivity to glucose limitation andbiguanides. Nature 2014;508:108–12.

108. Vernieri C, Casola S, Foiani M, Pietrantonio F, de Braud F, Longo V.Targeting cancer metabolism: dietary and pharmacologic interventions.Cancer Discov 2016;6:1315–33.

109. Zong WX, Rabinowitz JD, White E. Mitochondria and Cancer. Mol Cell2016;61:667–76.

110. Dan Dunn J, Alvarez LA, Zhang X, Soldati T. Reactive oxygen speciesand mitochondria: a nexus of cellular homeostasis. Redox Biol 2015;6:472–85.

111. Cuyas E, Fernandez-Arroyo S, Alarcon T, Lupu R, Joven J, Menendez JA.Germline BRCA1 mutation reprograms breast epithelial cell meta-bolism towards mitochondrial-dependent biosynthesis: evidence formetformin-based "starvation" strategies in BRCA1 carriers. Oncotarget2016;7:52974–92.

112. Cuyas E, Fernandez-Arroyo S, Joven J, Menendez JA. Metformin targetshistone acetylation in cancer-prone epithelial cells. Cell Cycle 2016;15:3355–61.

113. Blein S, Bardel C, Danjean V, McGuffog L, Healey S, Barrowdale D, et al.An original phylogenetic approach identified mitochondrial haplogroupT1a1 as inversely associated with breast cancer risk in BRCA2 mutationcarriers. Breast Cancer Res 2015;17:61.

114. Garrett WS. Cancer and the microbiota. Science 2015;348:80–6.115. Thaiss CA, Zmora N, Levy M, Elinav E. The microbiome and innate

immunity. Nature 2016;535:65–74.116. Tilg H, Cani PD, Mayer EA. Gut microbiome and liver diseases. Gut

2016;65:2035–44.

117. Brunner AL, Li J, Guo X, Sweeney RT, Varma S, Zhu SX, et al. A sharedtranscriptional program in early breast neoplasias despite genetic andclinical distinctions. Genome Biol 2014;15:R71.

118. Conway C, Graham JL, Chengot P, Daly C, Chalkley R, Ross L, et al.Elucidating drivers of oral epithelial dysplasia formation and malig-nant transformation to cancer using RNAseq. Oncotarget 2015;6:40186–201.

119. Marquardt JU, Seo D, Andersen JB, Gillen MC, KimMS, Conner EA, et al.Sequential transcriptome analysis of human liver cancer indicates latestage acquisition of malignant traits. J Hepatol 2014;60:346–53.

120. Schuyler RP, Merkel A, Raineri E, Altucci L, Vellenga E, Martens JH, et al.Distinct trends of DNAmethylation patterning in the innate and adaptiveimmune systems. Cell Rep 2016;17:2101–11.

121. Stahl M, Kohrman N, Gore SD, Kim TK, Zeidan AM, Prebet T. Epigeneticsin cancer: a hematological perspective. PLoS Genet 2016;12:e1006193.

122. Stadler ZK, Vijai J, Thom P, Kirchhoff T, Hansen NA, Kauff ND, et al.Genome-wide association studies of cancer predisposition. HematolOncol Clin North Am 2010;24:973–96.

123. Rebbeck TR, Sellers TA. The fruits of the genomic revolution. CancerEpidemiol Biomarkers Prev in press.

124. Gregson EM, Bornschein J, Fitzgerald RC. Genetic progression ofBarrett's oesophagus to oesophageal adenocarcinoma. Br J Cancer2016;115:403–10.

125. Levine DM, Ek WE, Zhang R, Liu X, Onstad L, Sather C, et al. A genome-wide association study identifies new susceptibility loci for esophagealadenocarcinoma and Barrett's esophagus. Nat Genet 2013;45:1487–93.

126. Su Z, Gay LJ, Strange A, Palles C, Band G, Whiteman DC, et al. Commonvariants at the MHC locus and at chromosome 16q24.1 predispose toBarrett's esophagus. Nat Genet 2012;44:1131–6.

127. Peters U, Jiao S, Schumacher FR, Hutter CM, Aragaki AK, Baron JA, et al.Identification of genetic susceptibility loci for colorectal tumors in agenome-wide meta-analysis. Gastroenterology 2013;144:799–807e24.

128. Petridis C, Brook MN, Shah V, Kohut K, Gorman P, Caneppele M, et al.Genetic predisposition to ductal carcinoma in situ of the breast. BreastCancer Res 2016;18:22.

129. Petljak M, Alexandrov LB. Understanding mutagenesis through delinea-tion of mutational signatures in human cancer. Carcinogenesis 2016;37:531–40.

130. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SA, Behjati S, BiankinAV, et al. Signatures of mutational processes in human cancer. Nature2013;500:415–21.

131. Rosenquist TA, Grollman AP. Mutational signature of aristolochic acid:clue to the recognition of a global disease. DNA Repair (Amst) 2016;44:205–11.

132. Jin K, Su KK, Li T, Zhu XQ, Wang Q, Ge RS, et al. Hepatic premalignantalterations triggered by human nephrotoxin aristolochic acid I in canines.Cancer Prev Res 2016;9:324–34.

133. Li T, Jin K, Zhu DY, Li L, Mao ZR, Wu BW, et al. Premalignant alterationassessment in liver-like tissue derived from embryonic stem cells byaristolochic acid I exposure. Oncotarget 2016;7:78872–82.

134. PoonSL, Pang ST,McPherson JR, YuW,HuangKK,GuanP, et al.Genome-wide mutational signatures of aristolochic acid and its application as ascreening tool. Sci Transl Med 2013;5:197ra01.

135. Meier B, Cooke SL, Weiss J, Bailly AP, Alexandrov LB, Marshall J, et al.C. elegans whole-genome sequencing reveals mutational signaturesrelated to carcinogens and DNA repair deficiency. Genome Res2014;24:1624–36.

136. Nik-Zainal S, Kucab JE, Morganella S, Glodzik D, Alexandrov LB, ArltVM, et al. The genome as a record of environmental exposure. Muta-genesis 2015;30:763–70.

137. Roberts SA, Lawrence MS, Klimczak LJ, Grimm SA, Fargo D, Stojanov P,et al. An APOBEC cytidine deaminase mutagenesis pattern is widespreadin human cancers. Nat Genet 2013;45:970–6.

138. Henderson S, Chakravarthy A, Su X, Boshoff C, Fenton TR. APOBEC-mediated cytosine deamination links PIK3CA helical domain mutationsto human papillomavirus-driven tumor development. Cell Rep 2014;7:1833–41.

139. Swanton C, McGranahan N, Starrett GJ, Harris RS. APOBEC enzymes:mutagenic fuel for cancer evolution and heterogeneity. Cancer Discov2015;5:704–12.

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research22

140. Robles-Espinoza CD, Roberts ND, Chen S, Leacy FP, Alexandrov LB,Pornputtapong N, et al. Germline MC1R status influences somaticmutation burden in melanoma. Nat Commun 2016;7:12064.

141. Agrawal N, Jiao Y, Bettegowda C, Hutfless SM, Wang Y, David S, et al.Comparative genomic analysis of esophageal adenocarcinoma andsquamous cell carcinoma. Cancer Discov 2012;2:899–905.

142. Alexandrov LB, Nik-Zainal S, Siu HC, Leung SY, Stratton MR. Amutational signature in gastric cancer suggests therapeutic strategies.Nat Commun 2015;6:8683.

143. Connor AA, Denroche RE, Jang GH, Timms L, Kalimuthu SN, Selander I,et al. Association of distinct mutational signatures with correlates ofincreased immune activity in pancreatic ductal adenocarcinoma. JAMAOncol 2016.

144. Secrier M, Li X, de Silva N, Eldridge MD, Contino G, Bornschein J, et al.Mutational signatures in esophageal adenocarcinoma define etio-logically distinct subgroups with therapeutic relevance. Nat Genet2016;48:1131–41.

145. Smid M, Rodriguez-Gonzalez FG, Sieuwerts AM, Salgado R, Prager-Vander Smissen WJ, Vlugt-Daane MV, et al. Breast cancer genome and tran-scriptome integration implicates specific mutational signatures withimmune cell infiltration. Nat Commun 2016;7:12910.

146. Strickland KC, Howitt BE, Shukla SA, Rodig S, Ritterhouse LL, Liu JF,et al. Association and prognostic significance of BRCA1/2-mutationstatus with neoantigen load, number of tumor-infiltrating lympho-cytes and expression of PD-1/PD-L1 in high grade serous ovariancancer. Oncotarget 2016;7:13587–98.

147. Bailey P, Chang DK, Nones K, Johns AL, Patch AM, Gingras MC, et al.Genomic analyses identify molecular subtypes of pancreatic cancer.Nature 2016;531:47–52.

148. Rashid M, Fischer A, Wilson CH, Tiffen J, Rust AG, Stevens P, et al.Adenoma development in familial adenomatous polyposis andMUTYH-associated polyposis: somatic landscape and driver genes. J Pathol 2016;238:98–108.

149. Kim J, Mouw KW, Polak P, Braunstein LZ, Kamburov A, Tiao G, et al.Somatic ERCC2 mutations are associated with a distinct genomic signa-ture in urothelial tumors. Nat Genet 2016;48:600–6.

150. DiGiovanna JJ, Kraemer KH. Shining a light on xeroderma pigmentosum.J Invest Dermatol 2012;132:785–96.

151. Yarosh D, Klein J, O'Connor A, Hawk J, Rafal E, Wolf P. Effect of topicallyapplied T4 endonuclease V in liposomes on skin cancer in xerodermapigmentosum: a randomised study. Xeroderma Pigmentosum StudyGroup. Lancet 2001;357:926–9.

152. Chen AC, Martin AJ, Choy B, Fernandez-Penas P, Dalziell RA, McKenzieCA, et al. A phase 3 randomized trial of nicotinamide for skin-cancerchemoprevention. N Engl J Med 2015;373:1618–26.

153. Alexandrov LB, Jones PH, Wedge DC, Sale JE, Campbell PJ, Nik-Zainal S,et al. Clock-like mutational processes in human somatic cells. Nat Genet2015;47:1402–7.

154. PodolskiyDI, LobanovAV,KryukovGV,GladyshevVN.Analysis of cancergenomes reveals basic features of human aging and its role in cancerdevelopment. Nat Commun 2016;7:12157.

155. Alexandrov LB, Ju YS, Haase K, Van Loo P, Martincorena I, Nik-Zainal S,et al. Mutational signatures associated with tobacco smoking in humancancer. Science 2016;354:618–22.

156. Saini N, Roberts SA, Klimczak LJ, Chan K, Grimm SA, Dai S, et al. Theimpact of environmental and endogenous damage on somatic mutationload in human skin fibroblasts. PLoS Genet 2016;12:e1006385.

157. Kim SY, Jung SH, Kim MS, Baek IP, Lee SH, Kim TM, et al. Genomicdifferences between pure ductal carcinoma in situ and synchronousductal carcinoma in situ with invasive breast cancer. Oncotarget 2015;6:7597–607.

158. Sakr RA, Schizas M, Carniello JV, Ng CK, Piscuoglio S, Giri D, et al.Targeted capture massively parallel sequencing analysis of LCIS andinvasive lobular cancer: Repertoire of somatic genetic alterations andclonal relationships. Mol Oncol 2016;10:360–70.

159. Murphy SJ, Hart SN, Lima JF, Kipp BR, Klebig M, Winters JL, et al. Geneticalterations associated with progression from pancreatic intraepithelialneoplasia to invasive pancreatic tumor. Gastroenterology 2013;145:1098–109e1.

160. Chitsazzadeh V, Coarfa C, Drummond JA, Nguyen T, Joseph A, ChilukuriS, et al. Cross-species identification of genomic drivers of squamous cell

carcinoma development across preneoplastic intermediates. Nat Com-mun 2016;7:12601.

161. Martincorena I, Roshan A, Gerstung M, Ellis P, Van Loo P, McLaren S,et al. Tumor evolution. High burden and pervasive positive selectionof somatic mutations in normal human skin. Science 2015;348:880–6.

162. Shain AH, Yeh I, Kovalyshyn I, Sriharan A, Talevich E, Gagnon A, et al. Thegenetic evolution of melanoma from precursor lesions. N Engl J Med2015;373:1926–36.

163. Vinayanuwattikun C, Le Calvez-Kelm F, Abedi-Ardekani B, Zaridze D,Mukeria A, Voegele C, et al. Elucidating genomic characteristics of lungcancer progression from in situ to invasive adenocarcinoma. Sci Rep2016;6:31628.

164. Ross-Innes CS, Becq J, Warren A, Cheetham RK, Northen H, O'DonovanM, et al. Whole-genome sequencing provides new insights into the clonalarchitecture of Barrett's esophagus and esophageal adenocarcinoma.Nat Genet 2015;47:1038–46.

165. StachlerMD, Taylor-Weiner A, Peng S,McKennaA, Agoston AT,Odze RD,et al. Paired exome analysis of Barrett's esophagus and adenocarcinoma.Nat Genet 2015;47:1047–55.

166. Weaver JM, Ross-Innes CS, Shannon N, Lynch AG, Forshew T, Barbera M,et al. Ordering of mutations in preinvasive disease stages of esophagealcarcinogenesis. Nat Genet 2014;46:837–43.

167. Zeng J, Kelbauskas L, Rezaie A, Lee K, Ueberroth B, Gao W, et al.Transcriptional regulation by normal epithelium of premalignant tomalignant progression in Barrett's esophagus. Sci Rep 2016;6:35227.

168. Gharahkhani P, Fitzgerald RC, Vaughan TL, Palles C, Gockel I, TomlinsonI, et al. Genome-wide association studies inoesophageal adenocarcinomaand Barrett's oesophagus: a large-scale meta-analysis. Lancet Oncol2016;17:1363–73.

169. BuasMF, LevineDM,Makar KW,UtsugiH,Onstad L, Li X, et al. Integrativepost-genome-wide association analysis of CDKN2A and TP53 SNPs andrisk of esophageal adenocarcinoma. Carcinogenesis 2014;35:2740–7.

170. Ju YS, Alexandrov LB, Gerstung M, Martincorena I, Nik-Zainal S, Ramak-rishna M, et al. Origins and functional consequences of somatic mito-chondrial DNA mutations in human cancer. Elife 2014;3:1–28.

171. Stewart JB, Alaei-Mahabadi B, Sabarinathan R, Samuelsson T, Gorodkin J,Gustafsson CM, et al. Simultaneous DNA and RNA mapping of somaticmitochondrial mutations across diverse human cancers. PLoS Genet2015;11:e1005333.

172. Reznik E, Wang Q, La K, Schultz N, Sander C. Mitochondrial respiratorygene expression is suppressed in many cancers. Elife 2017;6:1–16.

173. Lee S, Han MJ, Lee KS, Back SC, Hwang D, Kim HY, et al. Frequentoccurrence of mitochondrial DNA mutations in Barrett's metaplasiawithout the presence of dysplasia. PLoS One 2012;7:e37571.

174. Nicholson AM, Graham TA, Simpson A, Humphries A, Burch N, Rodri-guez-Justo M, et al. Barrett's metaplasia glands are clonal, contain mul-tiple stem cells and share a common squamous progenitor. Gut2012;61:1380–9.

175. Joshi S, Tolkunov D, Aviv H, Hakimi AA, Yao M, Hsieh JJ, et al. Thegenomic landscape of renal oncocytoma identifies a metabolic barrierto tumorigenesis. Cell Rep 2015;13:1895–908.

176. Dienstmann R, Vermeulen L, Guinney J, Kopetz S, Tejpar S, Tabernero J.Consensus molecular subtypes and the evolution of precision medicinein colorectal cancer. Nat Rev Cancer 2017.

177. Sinicrope FA, Shi Q, Smyrk TC, Thibodeau SN, Dienstmann R, Guinney J,et al. Molecular markers identify subtypes of stage III colon cancerassociated with patient outcomes. Gastroenterology 2015;148:88–99.

178. SturmD,Orr BA, ToprakUH,Hovestadt V, JonesDT, CapperD, et al. Newbrain tumor entities emerge frommolecular classification of CNS-PNETs.Cell 2016;164:1060–72.

179. Tirosh I, Izar B, Prakadan S, Wadsworth MII, Treacy D, Trombetta J, et al.Dissecting the multicellular ecosystem of metastatic melanoma bysingle-cell RNA-seq. Science 2016;352:189–96.

180. Martelotto LG, Baslan T, Kendall J, Geyer FC, Burke KA, Spraggon L, et al.Whole-genome single-cell copy number profiling from formalin-fixedparaffin-embedded samples. Nat Med 2017; [Epub ahead of print].

181. Bild AH, Yao G, Chang JT, Wang Q, Potti A, Chasse D, et al. Oncogenicpathway signatures in human cancers as a guide to targeted therapies.Nature 2006;439:353–7.

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 23

182. Godec J, Tan Y, Liberzon A, Tamayo P, Bhattacharya S, Butte AJ, et al.Compendium of immune signatures identifies conserved and species-specific biology in response to inflammation. Immunity 2016;44:194–206.

183. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, GilletteMA, et al. Gene set enrichment analysis: a knowledge-based approach forinterpreting genome-wide expression profiles. Proc Natl Acad Sci U S A2005;102:15545–50.

184. Kim JW, Botvinnik OB, Abudayyeh O, Birger C, Rosenbluh J, Shrestha Y,et al. Characterizing genomic alterations in cancer by complementaryfunctional associations. Nat Biotechnol 2016;34:539–46.

185. Ying H, Dey P, Yao W, Kimmelman AC, Draetta GF, Maitra A, et al.Genetics and biology of pancreatic ductal adenocarcinoma. Genes Dev2016;30:355–85.

186. Liou GY, Doppler H, Necela B, Edenfield B, Zhang L, Dawson DW, et al.Mutant KRAS-induced expression of ICAM-1 in pancreatic acinar cellscauses attraction of macrophages to expedite the formation of precan-cerous lesions. Cancer Discov 2015;5:52–63.

187. Pylayeva-Gupta Y, Das S, Handler JS, Hajdu CH, Coffre M, Koralov SB,et al. IL35-producing B cells promote the development of pancreaticneoplasia. Cancer Discov 2016;6:247–55.

188. Lee KE, Spata M, Bayne LJ, Buza EL, Durham AC, Allman D, et al. Hif1adeletion reveals Pro-neoplastic function of B cells in pancreatic neoplasia.Cancer Discov 2016;6:256–69.

189. Krogan NJ, Lippman S, Agard DA, Ashworth A, Ideker T. The cancer cellmap initiative: defining the hallmark networks of cancer. Mol Cell 2015;58:690–8.

190. Hyman DM, Puzanov I, Subbiah V, Faris JE, Chau I, Blay JY, et al.Vemurafenib in multiple nonmelanoma cancers with BRAF V600 muta-tions. N Engl J Med 2015;373:726–36.

191. Hu D, Shilatifard A. Epigenetics of hematopoiesis and hematologicalmalignancies. Genes Dev 2016;30:2021–41.

192. Teschendorff AE, Gao Y, Jones A, Ruebner M, Beckmann MW, WachterDL, et al. DNA methylation outliers in normal breast tissue identi-fy field defects that are enriched in cancer. Nat Commun 2016;7:10478.

193. Yoda Y, Takeshima H, Niwa T, Kim JG, Ando T, Kushima R, et al.Integrated analysis of cancer-related pathways affected by genetic andepigenetic alterations in gastric cancer. Gastric Cancer 2015;18:65–76.

194. Rasmussen KD, Helin K. Role of TET enzymes in DNA methylation,development, and cancer. Genes Dev 2016;30:733–50.

195. Yang Z, Wong A, Kuh D, Paul DS, Rakyan VK, Leslie RD, et al.Correlation of an epigenetic mitotic clock with cancer risk. GenomeBiol 2016;17:205.

196. Siegmund KD, Marjoram P, Woo YJ, Tavare S, Shibata D. Inferringclonal expansion and cancer stem cell dynamics from DNA methyl-ation patterns in colorectal cancers. Proc Natl Acad Sci U S A 2009;106:4828–33.

197. Busque L, Patel JP, FigueroaME, Vasanthakumar A, Provost S, Hamilou Z,et al. Recurrent somatic TET2 mutations in normal elderly individualswith clonal hematopoiesis. Nat Genet 2012;44:1179–81.

198. Lu C,Ward PS, Kapoor GS, Rohle D, Turcan S, Abdel-WahabO, et al. IDHmutation impairs histone demethylation and results in a block tocell differentiation. Nature 2012;483:474–8.

199. Watanabe T, Nobusawa S, Kleihues P, Ohgaki H. IDH1 mutationsare early events in the development of astrocytomas and oligodendro-gliomas. Am J Pathol 2009;174:1149–53.

200. Kwok B, Hall JM, Witte JS, Xu Y, Reddy P, Lin K, et al. MDS-associatedsomatic mutations and clonal hematopoiesis are common in idio-pathic cytopenias of undetermined significance. Blood 2015;126:2355–61.

201. Ortmann CA, Kent DG, Nangalia J, Silber Y, Wedge DC, Grinfeld J, et al.Effect of mutation order on myeloproliferative neoplasms. N Engl J Med2015;372:601–12.

202. Quivoron C, Couronne L, Della Valle V, Lopez CK, Plo I, Wagner-BallonO, et al. TET2 inactivation results in pleiotropic hematopoietic abnor-malities in mouse and is a recurrent event during human lymphomagen-esis. Cancer Cell 2011;20:25–38.

203. Blaschke K, Ebata KT, Karimi MM, Zepeda-Martinez JA, Goyal P, Maha-patra S, et al. VitaminC induces Tet-dependentDNAdemethylation and ablastocyst-like state in ES cells. Nature 2013;500:222–6.

204. Zhang Q, Zhao K, Shen Q, Han Y, Gu Y, Li X, et al. Tet2 is required toresolve inflammation by recruiting Hdac2 to specifically repress IL-6.Nature 2015;525:389–93.

205. Yue X, Trifari S, Aijo T, Tsagaratou A, PastorWA, Zepeda-Martinez JA, et al.Control of Foxp3 stability throughmodulation of TET activity. J ExpMed2016;213:377–97.

206. Ali MA, Matboli M, Tarek M, Reda M, Kamal KM, Nouh M, et al.Epigenetic regulation of immune checkpoints: another target for cancerimmunotherapy? Immunotherapy 2017;9:99–108.

207. Han L, Diao L, Yu S, Xu X, Li J, Zhang R, et al. The genomic landscape andclinical relevance of A-to-I RNA editing in human cancers. Cancer Cell2015;28:515–28.

208. Wulff BE,Nishikura K.Modulation ofmicroRNA expression and functionby ADARs. Curr Top Microbiol Immunol 2012;353:91–109.

209. Permuth JB, Reid B, Earp M, Chen YA, Monteiro AN, Chen Z, et al.Inherited variants affecting RNA editingmay contribute to ovarian cancersusceptibility: results froma large-scale collaboration.Oncotarget 2016;7:72381–94.

210. Chen L, Li Y, Lin CH, Chan TH, Chow RK, Song Y, et al. Recoding RNAediting of AZIN1 predisposes to hepatocellular carcinoma. Nat Med2013;19:209–16.

211. Jiang Q, Crews LA, Barrett CL, Chun HJ, Court AC, Isquith JM, et al.ADAR1 promotes malignant progenitor reprogramming in chronic mye-loid leukemia. Proc Natl Acad Sci U S A 2013;110:1041–6.

212. Qi L, Chan TH, Tenen DG, Chen L. RNA editome imbalance in hepato-cellular carcinoma. Cancer Res 2014;74:1301–6.

213. Yamanaka S, Poksay KS, Arnold KS, Innerarity TL. A novel translationalrepressor mRNA is edited extensively in livers containing tumors causedby the transgene expression of the apoB mRNA-editing enzyme. GenesDev 1997;11:321–33.

214. ZipetoMA, JiangQ,Melese E, JamiesonCH. RNA rewriting, recoding, andrewiring in human disease. Trends Mol Med 2015;21:549–59.

215. O'Connell MA, Mannion NM, Keegan LP. The Epitranscriptome andInnate Immunity. PLoS Genet 2015;11:e1005687.

216. Kaufman CK, Mosimann C, Fan ZP, Yang S, Thomas AJ, Ablain J, et al. Azebrafish melanoma model reveals emergence of neural crest identityduring melanoma initiation. Science 2016;351:aad2197.

217. Kato S, Lippman SM, Flaherty KT, Kurzrock R. The Conundrum of genetic"Drivers" in benign conditions. J Natl Cancer Inst 2016;108:1–13.

218. Patton EE, Widlund HR, Kutok JL, Kopani KR, Amatruda JF, MurpheyRD, et al. BRAF mutations are sufficient to promote nevi formation andcooperate with p53 in the genesis of melanoma. Curr Biol2005;15:249–54.

219. McNeal AS, Liu K, Nakhate V, Natale CA, Duperret EK, Capell BC, et al.CDKN2B loss promotes progression from benign melanocytic nevus tomelanoma. Cancer Discov 2015;5:1072–85.

220. Xie X, Koh JY, Price S, White E, Mehnert JM. Atg7 overcomes senescenceand promotes growth of BrafV600E-driven melanoma. Cancer Discov2015;5:410–23.

221. Sanchez-Danes A, Hannezo E, Larsimont JC, Liagre M, Youssef KK,Simons BD, et al. Defining the clonal dynamics leading to mouse skintumour initiation. Nature 2016;536:298–303.

222. Martin SD, Coukos G, Holt RA, Nelson BH. Targeting the undruggable:immunotherapy meets personalized oncology in the genomic era. AnnOncol 2015;26:2367–74.

223. Gjertsen MK, Gaudernack G. Mutated Ras peptides as vaccines in immu-notherapy of cancer. Vox Sang 1998;74Suppl 2:489–95.

224. June CH. Drugging the undruggable Ras - immunotherapy to the rescue?N Engl J Med 2016;375:2286–9.

225. Keenan BP, Saenger Y, Kafrouni MI, Leubner A, Lauer P, Maitra A, et al. AListeria vaccine and depletion of T-regulatory cells activate immunityagainst early stage pancreatic intraepithelial neoplasms and prolongsurvival of mice. Gastroenterology 2014;146:1784–94e6.

226. Finn OJ, Beatty PL. Cancer immunoprevention. Curr Opin Immunol2016;39:52–8.

227. Santana-Davila R, Bhatia S, Chow LQ. Harnessing the immune systemas a therapeutic tool in virus-associated cancers. JAMA Oncol2017;3:106–12.

228. Li XJ, Zhang X, Lin A, Ruan YY, Yan WH. Human leukocyte antigen-G(HLA-G) expression in cervical cancer lesions is associated with diseaseprogression. Hum Immunol 2012;73:946–9.

Perspective

Cancer Res; 77(7) April 1, 2017 Cancer Research24

229. Iheagwara UK, Beatty PL, Van PT, Ross TM, Minden JS, Finn OJ.Influenza virus infection elicits protective antibodies and T cellsspecific for host cell antigens also expressed as tumor-associatedantigens: a new view of cancer immunosurveillance. Cancer ImmunolRes 2014;2:263–73.

230. Dhodapkar MV, Sexton R, Das R, Dhodapkar KM, Zhang L, Sundaram R,et al. Prospective analysis of antigen-specific immunity, stem-cell anti-gens, and immune checkpoints in monoclonal gammopathy. Blood2015;126:2475–8.

231. Chu NJ, Armstrong TD, Jaffee EM. Nonviral oncogenic antigens and theinflammatory signals driving early cancer development as targets forcancer immunoprevention. Clin Cancer Res 2015;21:1549–57.

232. Davoli T, Uno H, Wooten EC, Elledge SJ. Tumor aneuploidy correlateswith markers of immune evasion and with reduced response to immu-notherapy. Science 2017;355.

233. Riteau N, Sher A. Chitosan: an adjuvant with an unanticipated STING.Immunity 2016;44:522–4.

234. Balkwill F, Mantovani A. Cancer and inflammation: implications forpharmacology and therapeutics. Clin Pharmacol Ther 2010;87:401–6.

235. Rashidian M, Keliher EJ, Bilate AM, Duarte JN, Wojtkiewicz GR, JacobsenJT, et al. Noninvasive imaging of immune responses. Proc Natl Acad Sci US A 2015;112:6146–51.

236. Kimura T, McKolanis JR, Dzubinski LA, Islam K, Potter DM, SalazarAM, et al. MUC1 vaccine for individuals with advanced adenoma ofthe colon: a cancer immunoprevention feasibility study. Cancer PrevRes 2013;6:18–26.

237. Eikawa S, Nishida M, Mizukami S, Yamazaki C, Nakayama E, Udono H.Immune-mediated antitumor effect by type 2 diabetes drug, metformin.Proc Natl Acad Sci U S A 2015;112:1809–14.

238. Forslund K, Hildebrand F, Nielsen T, Falony G, Le Chatelier E, SunagawaS, et al.Disentangling type 2diabetes andmetformin treatment signaturesin the human gut microbiota. Nature 2015;528:262–6.

239. Incio J, Suboj P, Chin SM, Vardam-Kaur T, Liu H, Hato T, et al.Metformin reduces desmoplasia in pancreatic cancer by reprogram-ming stellate cells and tumor-associated macrophages. PLoS One2015;10:e0141392.

240. Cao Y, Nishihara R, Qian ZR, Song M, Mima K, Inamura K, et al. Regularaspirin use associates with lower risk of colorectal cancers with lownumbers of tumor-infiltrating lymphocytes. Gastroenterology 2016;151:879–92.e4.

241. McLean MH, Murray GI, Stewart KN, Norrie G, Mayer C, Hold GL, et al.The inflammatory microenvironment in colorectal neoplasia. PLoS One2011;6:e15366.

242. Peters BA, Dominianni C, Shapiro JA, Church TR,Wu J,Miller G, et al. Thegut microbiota in conventional and serrated precursors of colorectalcancer. Microbiome 2016;4:69.

243. Robles AI, Traverso G, Zhang M, Roberts NJ, Khan MA, Joseph C, et al.Whole-exome sequencing analyses of inflammatory bowel disease-asso-ciated colorectal cancers. Gastroenterology 2016;150:931–43.

244. Yaeger R, Shah MA, Miller VA, Kelsen JR, Wang K, Heins ZJ, et al.Genomic alterations observed in colitis-associated cancers are dis-tinct from those found in sporadic colorectal cancers and vary bytype of inflammatory bowel disease. Gastroenterology 2016;151:278–87e6.

245. Komaroff AL. The microbiome and risk for obesity and diabetes. JAMA2017;317:355–6.

246. Bar F, Bochmann W, Widok A, von Medem K, Pagel R, Hirose M, et al.Mitochondrial gene polymorphisms that protect mice from colitis.Gastroenterology 2013;145:1055–63e3.

247. Schwabe RF, Jobin C. The microbiome and cancer. Nat Rev Cancer2013;13:800–12.

248. WhiffinN,Hosking FJ, Farrington SM, Palles C,Dobbins SE, Zgaga L, et al.Identification of susceptibility loci for colorectal cancer in a genome-widemeta-analysis. Hum Mol Genet 2014;23:4729–37.

249. Nowarski R, Jackson R,Gagliani N, de ZoeteMR, PalmNW, BailisW, et al.Epithelial IL-18 equilibrium controls barrier function in colitis. Cell2015;163:1444–56.

250. Levy J, Cacheux W, Bara MA, L'Hermitte A, Lepage P, Fraudeau M, et al.Intestinal inhibition of Atg7 prevents tumour initiation through a micro-biome-influenced immune response and suppresses tumour growth.Nat Cell Biol 2015;17:1062–73.

251. Earle KA, Billings G, Sigal M, Lichtman JS, Hansson GC, Elias JE, et al.Quantitative imaging of gut microbiota spatial organization. Cell HostMicrobe 2015;18:478–88.

252. Cunningham TJ, Tabacchi M, Eliane JP, Tuchayi SM, Manivasagam S,Mirzaalian H, et al. Randomized trial of calcipotriol combined with5-fluorouracil for skin cancer precursor immunotherapy. J Clin Invest2017;127:106–16.

253. Thiele Orberg E, Fan H, Tam AJ, Dejea CM, Destefano Shields CE, Wu S,et al. The myeloid immune signature of enterotoxigenic Bacteroidesfragilis-induced murine colon tumorigenesis. Mucosal Immunol 2016.

254. Lynch SV, Pedersen O. The human intestinal microbiome in health anddisease. N Engl J Med 2016;375:2369–79.

255. Poutahidis T, Varian BJ, Levkovich T, Lakritz JR, Mirabal S, Kwok C, et al.Dietary microbes modulate transgenerational cancer risk. Cancer Res2015;75:1197–204.

256. Poutahidis T, Erdman SE. Commensal bacteria modulate the tumormicroenvironment. Cancer Lett 2016;380:356–8.

257. Pietrocola F, Pol J, Vacchelli E, Rao S, Enot DP, Baracco EE, et al. Caloricrestriction mimetics enhance anticancer immunosurveillance. CancerCell 2016;30:147–60.

258. Garraway LA, Lander ES. Lessons from the cancer genome. Cell 2013;153:17–37.

259. Neefjes J, Ovaa H. A peptide's perspective on antigen presentation tothe immune system. Nat Chem Biol 2013;9:769–75.

260. Bassani-Sternberg M, Braunlein E, Klar R, Engleitner T, Sinitcyn P,Audehm S, et al. Direct identification of clinically relevant neoepitopespresented on native human melanoma tissue by mass spectrometry.Nat Commun 2016;7:13404.

261. Olsen L, Johan Kudahl U, Winther O, Brusic V. Literature classificationfor semi-automated updating of biological knowledgebases. BMCGenomics 2013;14Suppl 5:S14.

262. Backert L, Kohlbacher O. Immunoinformatics and epitope predictionin the age of genomic medicine. Genome Med 2015;7:119.

263. LarsenMV, Lundegaard C, Lamberth K, Buus S, LundO,NielsenM. Large-scale validation of methods for cytotoxic T-lymphocyte epitope predic-tion. BMC Bioinformatics 2007;8:424.

264. GubinMM,ArtyomovMN,Mardis ER, Schreiber RD. Tumor neoantigens:building a framework for personalized cancer immunotherapy. J ClinInvest 2015;125:3413–21.

265. RajasagiM, Shukla SA, Fritsch EF, Keskin DB, DeLuca D, Carmona E, et al.Systematic identification of personal tumor-specific neoantigens inchronic lymphocytic leukemia. Blood 2014;124:453–62.

266. Gubin MM, Zhang X, Schuster H, Caron E, Ward JP, Noguchi T, et al.Checkpoint blockade cancer immunotherapy targets tumour-specificmutant antigens. Nature 2014;515:577–81.

267. Stronen E, ToebesM, Kelderman S, van BuurenMM, YangW, van Rooij N,et al. Targeting of cancer neoantigens with donor-derived T cell receptorrepertoires. Science 2016;352:1337–41.

268. Duan F, Duitama J, Al Seesi S, Ayres CM, Corcelli SA, Pawashe AP, et al.Genomic and bioinformatic profiling of mutational neoepitopes revealsnew rules to predict anticancer immunogenicity. J Exp Med 2014;211:2231–48.

269. Srivastava PK. Neoepitopes of cancers: looking back, looking ahead.Cancer Immunol Res 2015;3:969–77.

270. Zhang X, Sharma PK, Peter Goedegebuure S, GillandersWE. Personalizedcancer vaccines: targeting the cancer mutanome. Vaccine 2017;35:1094–100.

271. Welter D, MacArthur J, Morales J, Burdett T, Hall P, Junkins H, et al. TheNHGRI GWAS Catalog, a curated resource of SNP-trait associations.Nucleic Acids Res 2014;42:D1001–6.

272. Engin HB, Kreisberg JF, Carter H. Structure-based analysis reveals cancermissense mutations target protein interaction interfaces. PLoS One2016;11:e0152929.

273. WangD,QuanY, YanQ,Morales JE,Wetsel RA. Targeted disruptionof thebeta2-microglobulin gene minimizes the immunogenicity of humanembryonic stem cells. Stem Cells Transl Med 2015;4:1234–45.

274. Echterdiek F, Janikovits J, Staffa L, Muller M, Lahrmann B, FruhschutzM, et al. Low density of FOXP3-positive T cells in normal colonicmucosa is related to the presence of beta2-microglobulin mutations inLynch syndrome-associated colorectal cancer. Oncoimmunology 2016;5:e1075692.

The Precancer Atlas

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 25

275. Zaretsky JM, Garcia-Diaz A, Shin DS, Escuin-Ordinas H, Hugo W, Hu-Lieskovan S, et al. Mutations associated with acquired resistance to PD-1blockade in melanoma. N Engl J Med 2016;375:819–29.

276. Lawrence MS, Stojanov P, Mermel CH, Robinson JT, Garraway LA, GolubTR, et al. Discovery and saturation analysis of cancer genes across 21tumour types. Nature 2014;505:495–501.

277. Piunti A, Shilatifard A. Epigenetic balance of gene expression by Poly-comb and COMPASS families. Science 2016;352:aad9780.

278. Lin C, Garruss AS, Luo Z, Guo F, Shilatifard A. The RNA Pol II elongationfactor Ell3 marks enhancers in ES cells and primes future gene activation.Cell 2013;152:144–56.

279. Shilatifard A. The COMPASS family of histone H3K4 methylases:mechanisms of regulation in development and disease pathogenesis.Annu Rev Biochem 2012;81:65–95.

280. Smith E, Lin C, Shilatifard A. The super elongation complex (SEC) andMLL in development and disease. Genes Dev 2011;25:661–72.

281. Shilatifard A, Lane WS, Jackson KW, Conaway RC, Conaway JW. An RNApolymerase II elongation factor encoded by the human ELL gene. Science1996;271:1873–6.

282. Lin C, Smith ER, Takahashi H, Lai KC, Martin-Brown S, Florens L, et al.AFF4, a component of the ELL/P-TEFb elongation complex and a sharedsubunit of MLL chimeras, can link transcription elongation to leukemia.Mol Cell 2010;37:429–37.

283. MohanM, Lin C, Guest E, Shilatifard A. Licensed to elongate: a molecularmechanism for MLL-based leukaemogenesis. Nat Rev Cancer2010;10:721–8.

284. Zhu L, Li Q, Wong SH, Huang M, Klein BJ, Shen J, et al. ASH1L linkshistone H3 lysine 36 dimethylation to MLL leukemia. Cancer Discov2016;6:770–83.

285. Christofides A, Karantanos T, Bardhan K, Boussiotis VA. Epigeneticregulation of cancer biology and anti-tumor immunity by EZH2. Onco-target 2016.

286. ZhangH,Qi J, Reyes JM, Li L, RaoPK, Li F, et al.Oncogenic deregulation ofEZH2as anopportunity for targeted therapy in lung cancer. CancerDiscov2016;6:1006–21.

287. Kim KH, Roberts CW. Targeting EZH2 in cancer. Nat Med 2016;22:128–34.

288. Lund K, Adams PD, Copland M. EZH2 in normal and malignant hema-topoiesis. Leukemia 2014;28:44–9.

289. Shimizu T, Kubovcakova L, Nienhold R, Zmajkovic J, Meyer SC, Hao-Shen H, et al. Loss of Ezh2 synergizes with JAK2-V617F in initiatingmyeloproliferative neoplasms and promoting myelofibrosis. J Exp Med2016;213:1479–96.

290. Kadoch C, Hargreaves DC, Hodges C, Elias L, Ho L, Ranish J, et al.Proteomic and bioinformatic analysis of mammalian SWI/SNF com-plexes identifies extensive roles in human malignancy. Nat Genet 2013;45:592–601.

291. Kadoch C, Crabtree GR. Reversible disruption of mSWI/SNF (BAF)complexes by the SS18-SSX oncogenic fusion in synovial sarcoma. Cell2013;153:71–85.

292. Tordella L, Khan S,Hohmeyer A, BanitoA, Klotz S, Raguz S, et al. SWI/SNFregulates a transcriptional program that induces senescence to preventliver cancer. Genes Dev 2016;30:2187–98.

293. Wilson BG, Wang X, Shen X, McKenna ES, Lemieux ME, Cho YJ,et al. Epigenetic antagonism between polycomb and SWI/SNFcomplexes during oncogenic transformation. Cancer Cell 2010;18:316–28.

294. Parsons DW, Jones S, Zhang X, Lin JC, Leary RJ, Angenendt P, et al. Anintegrated genomic analysis of human glioblastomamultiforme. Science2008;321:1807–12.

295. Dang L, White DW, Gross S, Bennett BD, Bittinger MA, Driggers EM, et al.Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature2009;462:739–44.

296. Koivunen P, Lee S, Duncan CG, Lopez G, Lu G, Ramkissoon S, et al.Transformation by the (R)-enantiomer of 2-hydroxyglutarate linked toEGLN activation. Nature 2012;483:484–8.

297. Losman JA, Kaelin WGJr. What a difference a hydroxyl makes: mutantIDH, (R)-2-hydroxyglutarate, and cancer. Genes Dev 2013;27:836–52.

298. Xu W, Yang H, Liu Y, Yang Y, Wang P, Kim SH, et al. Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of alpha-ketoglutarate-depen-dent dioxygenases. Cancer Cell 2011;19:17–30.

299. Dang L, Yen K, Attar EC. IDH mutations in cancer and progress towarddevelopment of targeted therapeutics. Ann Oncol 2016;27:599–608.

300. PomerantzMM, Li F, TakedaDY, Lenci R, Chonkar A, ChabotM, et al. Theandrogen receptor cistrome is extensively reprogrammed in humanprostate tumorigenesis. Nat Genet 2015;47:1346–51.

301. Dimitrova E, Turberfield AH, Klose RJ. Histone demethylases in chro-matin biology and beyond. EMBO Rep 2015;16:1620–39.

302. Mazor T, PankovA, Song JS, Costello JF. Intratumoral heterogeneity of theepigenome. Cancer Cell 2016;29:440–51.

303. Heath JR, Ribas A,Mischel PS. Single-cell analysis tools for drug discoveryand development. Nat Rev Drug Discov 2016;15:204–16.

304. Stegle O, Teichmann SA, Marioni JC. Computational and analyticalchallenges in single-cell transcriptomics. Nat Rev Genet 2015;16:133–45.

305. Mahata B, Zhang X, Kolodziejczyk AA, Proserpio V, Haim-Vilmovsky L,Taylor AE, et al. Single-cell RNA sequencing reveals T helper cells syn-thesizing steroids denovo to contribute to immunehomeostasis. Cell Rep2014;7:1130–42.

306. Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E,et al. Intratumor heterogeneity and branched evolution revealed bymultiregion sequencing. N Engl J Med 2012;366:883–92.

307. Navin NE. Cancer genomics: one cell at a time. Genome Biol 2014;15:452.

308. Almendro V, Kim HJ, Cheng YK, Gonen M, Itzkovitz S, Argani P, et al.Genetic and phenotypic diversity in breast tumor metastases. Cancer Res2014;74:1338–48.

309. Park SY, Gonen M, Kim HJ, Michor F, Polyak K. Cellular and geneticdiversity in the progression of in situ human breast carcinomas to aninvasive phenotype. J Clin Invest 2010;120:636–44.

310. Grundberg I, Kiflemariam S, Mignardi M, Imgenberg-Kreuz J, Edlund K,Micke P, et al. In situ mutation detection and visualization of intra-tumor heterogeneity for cancer research and diagnostics. Oncotarget2013;4:2407–18.

311. Ikeda S, Takabe K, Inagaki M, Funakoshi N, Suzuki K. Detection ofgene point mutation in paraffin sections using in situ loop-mediatedisothermal amplification. Pathol Int 2007;57:594–9.

312. Janiszewska M, Liu L, Almendro V, Kuang Y, Paweletz C, Sakr RA, et al. Insitu single-cell analysis identifies heterogeneity for PIK3CAmutation andHER2 amplification in HER2-positive breast cancer. Nat Genet 2015;47:1212–9.

313. Martinez P, TimmerMR, Lau CT, Calpe S, Sancho-SerraMdel C, StraubD,et al. Dynamic clonal equilibrium and predetermined cancer risk inBarrett's oesophagus. Nat Commun 2016;7:12158.

314. Hou Y, Song L, Zhu P, Zhang B, Tao Y, Xu X, et al. Single-cell exomesequencing and monoclonal evolution of a JAK2-negative myeloprolif-erative neoplasm. Cell 2012;148:873–85.

315. Kleppe M, Kwak M, Koppikar P, Riester M, Keller M, Bastian L, et al. JAK-STAT pathway activation in malignant and nonmalignant cells contri-butes to MPN pathogenesis and therapeutic response. Cancer Discov2015;5:316–31.

316. Gawad C, Koh W, Quake SR. Dissecting the clonal origins of childhoodacute lymphoblastic leukemia by single-cell genomics. Proc Natl Acad SciU S A 2014;111:17947–52.

317. Stahl PL, Salmen F, Vickovic S, Lundmark A, Navarro JF, Magnusson J,et al. Visualization and analysis of gene expression in tissue sections byspatial transcriptomics. Science 2016;353:78–82.

318. Cusanovich DA, Daza R, Adey A, Pliner HA, Christiansen L, GundersonKL, et al. Multiplex single cell profiling of chromatin accessibility bycombinatorial cellular indexing. Science 2015;348:910–4.

319. RotemA,RamO, ShoreshN, SperlingRA,GorenA,WeitzDA, et al. Single-cell ChIP-seq reveals cell subpopulations defined by chromatin state. NatBiotechnol 2015;33:1165–72.

320. Cleary B, Brito IL, Huang K, Gevers D, Shea T, Young S, et al. Detection oflow-abundance bacterial strains in metagenomic datasets by eigengen-ome partitioning. Nat Biotechnol 2015;33:1053–60.

321. Geva-Zatorsky N, Alvarez D, Hudak JE, Reading NC, Erturk-Hasdemir D,Dasgupta S, et al. In vivo imaging and tracking of host-microbiotainteractions via metabolic labeling of gut anaerobic bacteria. Nat Med2015;21:1091–100.

322. Diaz LAJr, Bardelli A. Liquid biopsies: genotyping circulating tumorDNA.J Clin Oncol 2014;32:579–86.

Cancer Res; 77(7) April 1, 2017 Cancer Research26

Perspective

323. Grun D, van Oudenaarden A. Design and analysis of single-cell sequenc-ing experiments. Cell 2015;163:799–810.

324. Macaulay IC, Voet T. Single cell genomics: advances and future perspec-tives. PLoS Genet 2014;10:e1004126.

325. Chen X, Love JC, Navin NE, Pachter L, Stubbington MJ, Svensson V,et al. Single-cell analysis at the threshold. Nat Biotechnol 2016;34:1111–8.

326. Navin NE. The first five years of single-cell cancer genomics and beyond.Genome Res 2015;25:1499–507.

327. Pestrin M, Salvianti F, Galardi F, De Luca F, Turner N, Malorni L, et al.Heterogeneity of PIK3CA mutational status at the single cell level incirculating tumor cells frommetastatic breast cancer patients. Mol Oncol2015;9:749–57.

328. Stephens ZD, Lee SY, Faghri F, Campbell RH, Zhai C, Efron MJ, et al. Bigdata: astronomical or genomical? PLoS Biol 2015;13:e1002195.

329. Bock C, Farlik M, Sheffield NC. Multi-omics of single cells: strategiesand applications. Trends Biotechnol 2016;34:605–8.

330. Haber DA, Velculescu VE. Blood-based analyses of cancer: circulatingtumor cells and circulating tumor DNA. Cancer Discov 2014;4:650–61.

331. Forshew T, Murtaza M, Parkinson C, Gale D, Tsui DW, Kaper F, et al.Noninvasive identification and monitoring of cancer mutations by tar-geted deep sequencing of plasma DNA. Sci Transl Med 2012;4:136ra68.

332. Leary RJ, Sausen M, Kinde I, Papadopoulos N, Carpten JD, Craig D, et al.Detectionof chromosomal alterations in the circulationof cancer patientswith whole-genome sequencing. Sci Transl Med 2012;4:162ra54.

333. Sausen M, Leary RJ, Jones S, Wu J, Reynolds CP, Liu X, et al. Integratedgenomic analyses identify ARID1A and ARID1B alterations in the child-hood cancer neuroblastoma. Nat Genet 2013;45:12–7.

334. Sausen M, Phallen J, Adleff V, Jones S, Leary RJ, Barrett MT, et al. Clinicalimplications of genomic alterations in the tumour and circulation ofpancreatic cancer patients. Nat Commun 2015;6:7686.

335. Tie J, Wang Y, Tomasetti C, Li L, Springer S, Kinde I, et al. Circulatingtumor DNA analysis detects minimal residual disease and predictsrecurrence in patients with stage II colon cancer. Sci Transl Med 2016;8:346ra92.

336. Birkenkamp-Demtroder K, Nordentoft I, Christensen E, Hoyer S, ReinertT, Vang S, et al. Genomic alterations in liquid biopsies from patients withbladder cancer. Eur Urol 2016;70:75–82.

337. Izumchenko E, Chang X, Brait M, Fertig E, Kagohara LT, Bedi A, et al.Targeted sequencing reveals clonal genetic changes in the progression ofearly lung neoplasms and paired circulating DNA. Nat Commun2015;6:8258.

338. Guerrero-Preston R, Valle BL, Jedlicka A, Turaga N, Folawiyo O, Pirini F,et al. Molecular triage of premalignant lesions in liquid-based cervicalcytology and circulating cell-free DNA from urine, using a panel ofmethylated human papilloma virus and host genes. Cancer Prev Res2016;9:915–24.

339. Fernandez-Cuesta L, Perdomo S, Avogbe PH, Leblay N, Delhomme TM,Gaborieau V, et al. Identification of circulating tumor DNA for the earlydetection of small-cell lung cancer. EBioMedicine 2016;10:117–23.

340. NewmanAM, Lovejoy AF, KlassDM,KurtzDM,Chabon JJ, Scherer F, et al.Integrated digital error suppression for improved detection of circulatingtumor DNA. Nat Biotechnol 2016;34:547–55.

341. Pantel K. Blood-based analysis of circulating cell-free DNA and tumorcells for early cancer detection. PLoS Med 2016;13:e1002205.

342. Warton K, Samimi G. Methylation of cell-free circulating DNA in thediagnosis of cancer. Front Mol Biosci 2015;2:13.

343. Wen L, Li J, GuoH, Liu X, Zheng S, ZhangD, et al. Genome-scale detectionof hypermethylated CpG islands in circulating cell-free DNA of hepato-cellular carcinoma patients. Cell Res 2015;25:1250–64.

344. Best MG, Sol N, Kooi I, Tannous J, Westerman BA, Rustenburg F, et al.RNA-Seq of tumor-educated platelets enables blood-based pan-cancer,multiclass, and molecular pathway cancer diagnostics. Cancer Cell2015;28:666–76.

345. Ilie M, Hofman V, Long-Mira E, Selva E, Vignaud JM, Padovani B, et al."Sentinel" circulating tumor cells allow early diagnosis of lung cancer inpatients with chronic obstructive pulmonary disease. PLoS One 2014;9:e111597.

346. Silvestri GA, Vachani A, Whitney D, Elashoff M, Porta Smith K, FergusonJS, et al. A bronchial genomic classifier for the diagnostic evaluation oflung cancer. N Engl J Med 2015;373:243–51.

347. Laget S, Broncy L, Hormigos K, Dhingra DM, BenMohamed F, Capiod T,et al. Technical insights into highly sensitive isolation and molecularcharacterization offixed and live circulating tumor cells for early detectionof tumor invasion. PLoS One 2017;12:e0169427.

348. Pantel K, Deneve E, Nocca D, Coffy A, Vendrell JP, Maudelonde T, et al.Circulating epithelial cells in patients with benign colon diseases. ClinChem 2012;58:936–40.

349. Gorges TM, Kuske A, Rock K, Mauermann O, Muller V, Peine S,et al. Accession of tumor heterogeneity by multiplex transcriptomeprofiling of single circulating tumor cells. Clin Chem 2016;62:1504–15.

350. Beane JE, Campbell J, Moy C, Perdomo C, Schaffer M, Mazzilli S, et al.Abstract 2878: Development of the pre-cancer genome atlas (PCGA) forsquamous cell lung carcinoma. Cancer Res 2015;75:2878–78.

351. Pipinikas CP, Kiropoulos TS, Teixeira VH, Brown JM, Varanou A, FalzonM, et al. Cell migration leads to spatially distinct but clonally relatedairway cancer precursors. Thorax 2014;69:548–57.

352. Andor N, Graham TA, Jansen M, Xia LC, Aktipis CA, Petritsch C, et al.Pan-cancer analysis of the extent and consequences of intratumorheterogeneity. Nat Med 2016;22:105–13.

353. Li X, Paulson TG, Galipeau PC, Sanchez CA, Liu K, Kuhner MK, et al.Assessment of esophageal adenocarcinoma risk using somatic chromo-some alterations in longitudinal samples in barrett's esophagus. CancerPrev Res 2015;8:845–56.

354. Maddipati R, Stanger BZ. Pancreatic cancer metastases harbor evidenceof polyclonality. Cancer Discov 2015;5:1086–97.

355. Jaiswal S, Fontanillas P, Flannick J, Manning A, Grauman PV, Mar BG,et al. Age-related clonal hematopoiesis associatedwith adverse outcomes.N Engl J Med 2014;371:2488–98.

356. Xie M, Lu C, Wang J, McLellan MD, Johnson KJ, Wendl MC, et al. Age-related mutations associated with clonal hematopoietic expansion andmalignancies. Nat Med 2014;20:1472–8.

357. Bejar R, Stevenson K, Abdel-WahabO,Galili N, Nilsson B,Garcia-ManeroG, et al. Clinical effect of point mutations in myelodysplastic syndromes.N Engl J Med 2011;364:2496–506.

358. Steensma DP, Bejar R, Jaiswal S, Lindsley RC, Sekeres MA, Hasserjian RP,et al. Clonal hematopoiesis of indeterminate potential and its distinctionfrom myelodysplastic syndromes. Blood 2015;126:9–16.

359. Basiorka AA, McGraw KL, Eksioglu EA, Chen X, Johnson J, Zhang L, et al.The NLRP3 inflammasome functions as a driver of the myelodysplasticsyndrome phenotype. Blood 2016;128:2960–75.

360. Schneider RK, Schenone M, Ferreira MV, Kramann R, Joyce CE, HartiganC, et al. Rps14 haploinsufficiency causes a block in erythroid differenti-ation mediated by S100A8 and S100A9. Nat Med 2016;22:288–97.

361. McKerrell T, Vassiliou GS. Aging as a driver of leukemogenesis. SciTransl Med 2015;7:306fs38.

362. Ogawa S. Clonal hematopoiesis in acquired aplastic anemia. Blood2016;128:337–47.

363. Cargo CA, Rowbotham N, Evans PA, Barrans SL, Bowen DT, Crouch S,et al. Targeted sequencing identifies patients with preclinical MDS at highrisk of disease progression. Blood 2015;126:2362–5.

364. Meyer SE, Qin T, Muench DE, Masuda K, Venkatasubramanian M, Orr E,et al. DNMT3A haploinsufficiency transforms FLT3ITD myeloprolifera-tive disease into a rapid, spontaneous, and fully penetrant acute myeloidleukemia. Cancer Discov 2016;6:501–15.

365. Lubking A, Vosberg S, Konstandin NP, Dufour A, Graf A, Krebs S, et al.Young woman with mild bone marrow dysplasia, GATA2 and ASXL1mutation treated with allogeneic hematopoietic stem cell transplanta-tion. Leuk Res Rep 2015;4:72–5.

366. Hinds DA, Barnholt KE, Mesa RA, Kiefer AK, Do CB, Eriksson N, et al.Germ line variants predispose to both JAK2 V617F clonal hematopoiesisand myeloproliferative neoplasms. Blood 2016;128:1121–8.

367. Churpek JE, Pyrtel K, Kanchi KL, Shao J, Koboldt D, Miller CA, et al.Genomic analysis of germ line and somatic variants in familial myelo-dysplasia/acute myeloid leukemia. Blood 2015;126:2484–90.

368. Parikh SA, Kay NE, Shanafelt TD. Monoclonal B-cell lymphocytosis:update ondiagnosis, clinical outcome, and counseling. ClinAdvHematolOncol 2013;11:720–9.

369. Slager SL, Rabe KG, Achenbach SJ, VachonCM,Goldin LR, Strom SS, et al.Genome-wide association study identifies a novel susceptibility locus at6p21.3 among familial CLL. Blood 2011;117:1911–6.

www.aacrjournals.org Cancer Res; 77(7) April 1, 2017 27

The Precancer Atlas

370. Ojha J, Secreto C, Rabe K, Ayres-Silva J, Tschumper R, Dyke DV, et al.Monoclonal B-cell lymphocytosis is characterized by mutations in CLLputative driver genes and clonal heterogeneity many years before diseaseprogression. Leukemia 2014;28:2395–8.

371. Puente XS, Bea S, Valdes-Mas R, Villamor N, Gutierrez-Abril J, Martin-Subero JI, et al. Non-coding recurrent mutations in chronic lymphocyticleukaemia. Nature 2015;526:519–24.

372. Barrio S, Shanafelt TD, Ojha J, Chaffee KG, Secreto C, Kortum KM, et al.Genomic characterization of high-count MBL cases indicates that earlydetection of driver mutations and subclonal expansion are predictors ofadverse clinical outcome. Leukemia 2017;31:170–6.

373. Goldin LR, Lanasa MC, Slager SL, Cerhan JR, Vachon CM, Strom SS, et al.Common occurrence of monoclonal B-cell lymphocytosis among mem-bers of high-risk CLL families. Br J Haematol 2010;151:152–8.

374. Oddsson A, Kristinsson SY, Helgason H, Gudbjartsson DF, Masson G,Sigurdsson A, et al. The germline sequence variant rs2736100_C inTERT associates with myeloproliferative neoplasms. Leukemia 2014;28:1371–4.

375. Landau DA, Tausch E, Taylor-Weiner AN, Stewart C, Reiter JG, Bahlo J,et al. Mutations driving CLL and their evolution in progression andrelapse. Nature 2015;526:525–30.

376. Moreira J, Rabe KG, Cerhan JR, Kay NE, Wilson JW, Call TG, et al.Infectious complications among individuals with clinical monoclonalB-cell lymphocytosis (MBL): a cohort study of newly diagnosed casescompared to controls. Leukemia 2013;27:136–41.

377. Solomon BM, Chaffee KG, Moreira J, Schwager SM, Cerhan JR, Call TG,et al. Risk of non-hematologic cancer in individuals with high-countmonoclonal B-cell lymphocytosis. Leukemia 2016;30:331–6.

378. Anderson LA, Landgren O, Engels EA. Common community acquiredinfections and subsequent risk of chronic lymphocytic leukaemia.Br J Haematol 2009;147:444–9.

379. Slager SL. Genome-wide association analysis implicates dysregulation ofimmunity genes in chronic lymphocytic leukemia. Nat Commun 2017;141:14175.

380. Ramsay AG, Johnson AJ, Lee AM, Gorgun G, Le Dieu R, Blum W, et al.Chronic lymphocytic leukemia T cells show impaired immunologicalsynapse formation that can be reversed with an immunomodulatingdrug. J Clin Invest 2008;118:2427–37.

381. Shanafelt TD, Ramsay AG, Zent CS, Leis JF, Tun HW, Call TG, et al.Long-term repair of T-cell synapse activity in a phase II trial ofchemoimmunotherapy followed by lenalidomide consolidation inpreviously untreated chronic lymphocytic leukemia (CLL). Blood2013;121:4137–41.

382. Henry CJ, Casas-Selves M, Kim J, Zaberezhnyy V, Aghili L, Daniel AE,et al. Aging-associated inflammation promotes selection for adaptiveoncogenic events in B cell progenitors. J Clin Invest 2015;125:4666–80.

383. Martin-Lorenzo A, Hauer J, Vicente-Duenas C, Auer F, Gonzalez-HerreroI, Garcia-Ramirez I, et al. Infection exposure is a causal factor in B-cellprecursor acute lymphoblastic leukemia as a result of Pax5-inheritedsusceptibility. Cancer Discov 2015;5:1328–43.

384. Dhodapkar MV, Krasovsky J, Osman K, Geller MD. Vigorous prema-lignancy-specific effector T cell response in the bone marrow of patientswith monoclonal gammopathy. J Exp Med 2003;198:1753–7.

385. Spisek R, Kukreja A, Chen LC, Matthews P, Mazumder A, Vesole D, et al.Frequent and specific immunity to the embryonal stem cell-associatedantigen SOX2 in patients with monoclonal gammopathy. J Exp Med2007;204:831–40.

386. MateosMV,HernandezMT,GiraldoP,de laRubia J, deArriba F,Corral LL,et al. Lenalidomide plus dexamethasone versus observation in patientswith high-risk smouldering multiple myeloma (QuiRedex): long-termfollow-up of a randomised, controlled, phase 3 trial. Lancet Oncol2016;17:1127–36.

387. Walker BA, Wardell CP, Melchor L, Brioli A, Johnson DC, Kaiser MF, et al.Intraclonal heterogeneity is a critical early event in the development ofmyeloma and precedes the development of clinical symptoms. Leukemia2014;28:384–90.

388. Zhao S, ChoiM,HeuckC,Mane S, Barlogie B, LiftonRP, et al. Serial exomeanalysis of disease progression in premalignant gammopathies. Leuke-mia 2014;28:1548–52.

389. Lawson MA, McDonald MM, Kovacic N, Hua Khoo W, Terry RL, Down J,et al. Osteoclasts control reactivation of dormant myeloma cells byremodelling the endosteal niche. Nat Commun 2015;6:8983.

390. Das R, Strowig T, Verma R, Koduru S, Hafemann A, Hopf S, et al.Microenvironment-dependent growth of pre-neoplastic and malignantplasma cells in humanized mice. Nat Med 2016;(in press).

391. Weinhold N, Johnson DC, Rawstron AC, Forsti A, Doughty C,Vijayakrishnan J, et al. Inherited genetic susceptibility to mono-clonal gammopathy of unknown significance. Blood 2014;123:2513–7; quiz 93.

392. Nair S, Branagan AR, Liu J, Boddupalli CS, Mistry PK, Dhodapkar MV.Clonal immunoglobulin against lysolipids in the origin of myeloma.N Engl J Med 2016;374:555–61.

393. Pavlova EV, Archer J, Wang S, Dekker N, Aerts JM, Karlsson S, et al.Inhibition of UDP-glucosylceramide synthase in mice prevents Gaucherdisease-associated B-cell malignancy. J Pathol 2015;235:113–24.

394. MaressoKC, Tsai KY, BrownPH, Szabo E, Lippman S,HawkET.Molecularcancer prevention: current status and future directions. CA Cancer J Clin2015;65:345–83.

395. Bauman JE, Grandis J. Oral cancer chemoprevention–the end of EPOC,the beginning of an epoch of molecular selection. JAMA Oncol 2016;2:178–9.

396. WilliamWNJr, Papadimitrakopoulou V, Lee JJ, Mao L, Cohen EE, Lin HY,et al. Erlotinib and the risk of oral cancer: The Erlotinib Prevention ofOralCancer (EPOC) Randomized Clinical Trial. JAMA Oncol 2016;2:209–16.

397. Turley SJ, Cremasco V, Astarita JL. Immunological hallmarks of stro-mal cells in the tumour microenvironment. Nat Rev Immunol 2015;15:669–82.

398. Ecker S, Chen L, Pancaldi V, Bagger FO, Fernandez JM, Carrillo de SantaPau E, et al. Genome-wide analysis of differential transcriptional andepigenetic variability across human immune cell types. Genome Biol2017;18:18.

399. Krimmel JD, Schmitt MW,Harrell MI, Agnew KJ, Kennedy SR, EmondMJ,et al. Ultra-deep sequencing detects ovarian cancer cells in peritonealfluidand reveals somatic TP53 mutations in noncancerous tissues. Proc NatlAcad Sci U S A 2016;113:6005–10.

400. Drost J, van Jaarsveld RH, Ponsioen B, Zimberlin C, van Boxtel R, Buijs A,et al. Sequential cancermutations in cultured human intestinal stem cells.Nature 2015;521:43–7.

401. Keysselt K, Kreutzmann T, Rother K, Kerner C, Krohn K, Przybilla J, et al.Different in vivo and in vitro transformation of intestinal stem cells inmismatch repair deficiency. Oncogene 2016.

402. Zhang Z, Christin JR, Wang C, Ge K, Oktay MH, GuoW. Mammary-stem-cell-based somatic mousemodels reveal breast cancer drivers causing cellfate dysregulation. Cell Rep 2016;16:3146–56.

Cancer Res; 77(7) April 1, 2017 Cancer Research28

Perspective