Precancer Atlas to Drive Precision Prevention Trials

Abstract

Cancer development is a complex process influenced by inherited and acquired molecular and cellular alterations. Prevention is the holy grail of cancer eradication, but making this a reality will take a fundamental rethinking based on deep understanding of premalignant biology. Although the seminal multi-step, genetic model of human tumorigenesis was defined in the colorectal adenoma-carcinoma sequence three decades ago, only a handful of sporadic adenomas analyzed by next-generation sequencing (NGS) have been reported, in striking contrast to the wealth of NGS data in cancers. Recent precancer advances include: germline mutation biology driving precision prevention – RANK-L effects on luminal progenitors leading to international trial in BRCA1 carriers, first real signal of potential benefit of early detection research in pancreatic neoplasia, breakthrough trial of combination chemoprevention in familial adenomatous polyposis, and germline/microbiota interactions in intestinal oncogenesis, including in mouse model of Lynch syndrome. Other developments include: novel (e.g., copy number alterations) regulators and imaging of immunosuppressive microenvironments; immune prevention (from HPV to cancer vaccines); mutational signatures; premalignant mutational accumulation in aging, established in clonal hematopoiesis, with recent deep NGS suggesting a more generalized phenomenon; and single-cell analyses of DCIS and Barrett’s esophagus identifying importance of genetic heterogeneity. The challenges are substantial, including hefty computational and data prices for unprecedented whole-genome, single-cell resolution, with population scale data estimates in the exabyte range (>10¹⁸ bytes), requiring new computational frameworks. To accelerate the prevention of cancer, this field needs a large-scale, longitudinal national effort, leveraging diverse disciplines, technologies, and models to develop an integrated multi-omics and immunity precancer atlas (PCA), to interrogate and target events that drive oncogenesis.

Introduction

In early 2016, President Obama announced the creation of the National Cancer Moonshot initiative – a commitment to eradicate cancer by investing in dramatically advancing progress within diverse fields of cancer research, including prevention and early detection (1). Historically, the rate-limiting step for developing and implementing precision approaches has been our relatively limited understanding of precancer biology in contrast to the extensive study of advanced disease. For example, The Cancer Genome Atlas (TCGA), which includes volumes of omics data from >11,000 patients across 33 tumor types, has transformed our understanding of cancer biology, identified hundreds of driver mutations that alter hallmark pathways, and become a tremendous national resource for continuing discovery, including catalyzing the development of novel computational tools to analyze mutational signatures and single-cell NGS. Recent studies are redefining the spectrum and biology of neoplasias linked to various hereditary cancer predisposition syndromes (2,3). The influence of mitochondrial genetics and biology on cancer predisposition and development is a critical, and until recently, understudied field (4–7). Finally, a diverse array of engineered preclinical models, technologies, and disciplines are now being leveraged to probe early oncogenesis. Large-scale longitudinal and systematic mapping is critical to develop an integrated omics and immunity precancer atlas (PCA), allowing dissection of the sequential molecular and cellular events that promote oncogenesis leading to novel prevention and interception strategies (8–10) (Figure 1).

Figure 1.

An integrated multi-omics and immunity precancer atlas (PCA). Inherited and acquired molecular alterations and their interaction with the local microenvironment/immune system influence oncogenic progression to invasive carcinoma. Normal cells (light orange, far left) that have a germline mutation (green nuclei) acquire somatic mutations (dark orange) or molecular alterations due to viral infection (purple). These all can potentially have an altered cell state with dysregulated molecular pathways that result in the loss of cell growth control and other hallmarks of cancer (228), which can then result in the development of advanced precancers (multicolored, subclonal diversity/heterogeneity) that immediately precede invasive cancer (red). Molecular alterations, depicted by symbols in the nucleus, represent mutations, SNPs, or epigenetic modifications. The accumulation of mutations (e.g., UV, ABOBEC) during life creates signatures (shown by the chromosomal insets and colored dots in gradient from cancer to normal). Somatic mutations in nuclear and mitochondrial genes, mutational signatures, other omic events, and germline changes/interactions are key drivers of oncogenesis. The mutational signatures, often identified in cancers, may predict early events, now extending to the precancer state (depicted by the orange/red cells with bold, yellow border). These omic alterations interact (bidirectionally) with the complex tissue microenvironment, including immune cells, stromal cells (adipocytes and fibroblasts), and other events, which promote oncogenesis and invasion/escape; microbiota can prevent or promote oncogenesis (see text). The continuum between the immune state modulated by cytokines and growth factors includes immune surveillance, composed of the antigenic repertoire, adaptive and immune cells (upper left box), and immune suppression/escape (upper right box) along with vascular and endothelial cells (not shown) that can lead to immune escape – another hallmark of cancer. The elucidation of the integrated multi-omics and immunity PCA will continue to evolve in complexity (e.g., recent SCNA finding), requiring continuous updating in this fast moving field.

Translating Inherited Cancer Risk into Precision Prevention

Germline nuclear genetics

The genetics of various hereditary forms of cancer risk have been studied extensively and long been used to aid our understanding of sporadic neoplasia. Our knowledge about the biology of germline mutations in certain cancer genes (e.g., BRCA1) is much better understood than in the somatic setting and has historically provided tremendous insight into fundamental neoplastic processes since the mechanisms of the neoplasia in these mutation carriers directly reflect the biology of the mutated gene itself. For example, study of the biology of tumors that develop in individuals carrying high-penetrance, predisposition mutations has led to paradigm-changing therapy and is beginning to drive prevention and early detection research, as evidenced by PARP inhibitors, one of the most compelling forms of precision cancer therapy in various forms of BRCA1/2-associated cancers (11), shown to delay mammary tumor development in Brca1-deficient mice (12). Understanding familial adenomatous polyposis (FAP) biology, characterized by germline APC mutations and chromosomal instability, led to a recently reported breakthrough trial of combination chemoprevention targeting the convergence of Wnt and EGFR signaling and COX-2 (13) in this devastating syndrome. In contrast, although invasive lobular cancers occur excessively in women who carry germline mutations in BRCA2 or CDH1, there is no reported excess of lobular carcinoma in situ (LCIS) in these rare families (14), illustrating the complexity of precancer types, biology, and patterns in carriers of different high-penetrance germline mutations.

Germline genetic heterogeneity, single phenotype/genetic disorders caused by mutations in several inherited genes, is relatively common. A striking example is pheochromocytoma, which can be caused by a number of germline mutations, including in RET, VHL, FH, IDH1, NF1, and certain SDH nuclear genes (SDHA, SDHB, SDHC, and SDHD), which encode mitochondrial complex II genes (15). The SDH complex functions in mitochondrial energy generation, links the Krebs cycle and the electron transport chain (ETC), and plays a key role in oxygen sensing and tumor suppression. Germline SDHD variants alter PTEN function, a novel mechanism of thyroid pathogenesis in Cowden syndrome (16). Recent large series of pediatric cancer have identified germline mutations in predisposition genes in only about 10% of pediatric cancer patients (17), comparable to the prevalence in adult cancer patients (18). The fundamental question of why certain germline mutations in cancer susceptibility genes predispose to a particular spectrum and magnitude of neoplasias is largely unknown. Although some genes have organ-specific functions (e.g., mutations leading to hepatic overload cause liver cancer) (19), most mutations have a broad range of functions (e.g., mismatch repair (MMR)). It is unclear why MMR inactivation predisposes more to colorectal cancer (CRC) rather than generalized systemic cancer risk or why specific MMR gene mutations are associated with different cancer patterns (20).

Mouse models have begun to unravel some of the key mechanisms of intestinal neoplastic transformation in the germline mutation setting. The first major study involved germline/microbiota interaction in a mouse model of Lynch syndrome (MSH2-deficient intestinal carcinogenesis) and demonstrated that butyrate, generated by gut microbiota from dietary carbohydrates, can act as an oncometabolite (21). Interestingly, butyrate can have opposing effects in different colon models functioning as a tumor-suppressive metabolite with energetic and epigenetic functions, which likely reflect the different germline backgrounds, e.g., reduced dietary butyrate markedly decreased CRC development in MMR-deficient, but not MMR-proficient mice. Building on this germline microbiome/CRC prevention work, a recent study demonstrated that celecoxib (a COX-2 inhibitor known to reduce intestinal adenoma burden) induces alterations in the gut microbiome and metabolome in APC^Min/+ mice, a model of FAP (22). Specifically, celecoxib increased gut Coriobacteriaceae, which suppressed production of oncogenic metabolites (e.g., glycine and serine). In this Min mouse model, innate and adaptive sources of IL-17 drive colon tumorigenic response to enterotoxigenic Bacteroides fragilis (ETBF); and anti-IL-17 monoclonal antibody (mAb) and T-regulatory (Treg) cell depletion suppressed tumorigenesis at the micro-adenoma stage (23). Emerging data suggest a link between ETBF, inflammatory bowel disease (IBD), and CRC. ETBF toxin-triggered colon tumorigenesis is characterized by a specific immune signature (combining IL-17-driven colitis and altered myeloid differentiation into MDSC) (24). Germline influence on bacterial translocation also involves GWAS-identified laminin nuclear and mtDNA variants (21,25–28).

Studies in BRCA1/2 carriers are providing novel insights into the high penetrance for breast and ovarian cancer. Data suggest that BRCA1/2 germline mutations are driving oncogenesis via combined effects of compromised DNA repair capability and changes in the endocrine system in the organ at risk, leading to breast and ovary cancers (29). Somatic loss of the BRCA wild-type allele is required to provoke genomic instability and tumorigenesis. Other studies have suggested this tissue specificity may be linked to inhibition of estrogen receptor (ER)-α transcription activation by BRCA1 ubiquitin ligase (30). Although most breast cancers arising in BRCA1-mutation carriers are ER-negative, tamoxifen use appears to be associated with a reduction in risk, particularly of contralateral breast cancer (31), likely due to female hormone effects in the early ontogeny or stromal-/ myeloid-derived suppressor cell (MDSC) (32,33) estrogen signaling. These insights suggest new prevention strategies that exploit hormonal dysregulation in BRCA1/2 carriers.

Elegant studies of luminal progenitor biology in BRCA1-mutation carriers (34,35) have led to a transformative potential to prevent/delay BRCA1-associated breast cancer, a disease for which the best current option is prophylactic mastectomy. A highly proliferative subset of luminal 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, RANK+ BRCA1-deficient/mutant progenitor cells are more susceptible to DNA damage and aberrant downstream NFkB activation than RANK- mammary progenitor cells (36). Blocking RANK-L signaling in several Brca1-deficient/mutated mouse models markedly inhibits mammary tumorigenesis (34). RANK-L/RANK signaling (37) also can influence immune surveillance/evasion via innate and adaptive immune responses (38,39). RANK-L produced by Tregs promotes mammary cancer and T-cell tolerance to intestinal bacteria (39). Recently, a second RANK-L receptor, LGR4, has been implicated in the regulation of multiple developmental pathways (40). Serum levels of osteoprotegerin (OPG), the endogenous RANK-L inhibitor, are significantly lower in BRCA1 carriers (vs controls), premenopausal women, and are inversely associated with breast cancer risk (41,42). Certain germline RANK SNPs (TNFRSF11A) have been associated with increased RANK expression and breast cancer risk in BRCA1 carriers (35). Denosumab (a RANK-L mAb inhibitor FDA-approved for the treatment of both osteoporosis and bone metastases) blocked progesterone-induced proliferation in BRCA1-mutant organoids. In small pilot window studies of BRCA1 carriers, denosumab reduced breast epithelial cell proliferation and progenitor cells clonogenic potential (34,35). Furthermore, denosumab has a well-established safety record and was recently shown to significantly delay (by 50%) the time to first fracture in postmenopausal ER+ breast cancer (43). Theoretically, RANK-L inhibition would work best to prevent/delay tumor onset for premenopausal BRCA1-mutation carriers since risk-reducing salpingo-oophorectomy has recently been shown to be ineffective in this setting (44) and RANK-L is a progesterone-responsive gene. Progesterone receptor modulators prevent mammary tumorigenesis in Brca-mutant mice but are limited by toxicity for clinical prevention. Based on these data, denosumab is being developed for a large-scale international breast cancer prevention trial in BRCA-mutation carriers (34,35).

Pancreatic cancer early detection research with magnetic resonance and endoscopic ultrasound in germline high-risk individuals produced the first real signal of potential benefit in this setting (45). Unselected pancreatic cancer patients have a very high (>15% in a clinic-based cohort) prevalence of germline mutations (mostly BRCA1/2), even higher among Ashkenazi Jewish individuals (46). Pancreatic precursor lesions in people with high-penetrance germline mutations have a higher malignant potential (than other pancreas high-risk groups) (47,48). These data recently led some centers to recommend germline testing for all new pancreatic cancer patients. Precedent for such an approach already exists in ovarian cancer, where NCCN guidelines recommend germline analysis of BRCA1/2 in all new patients (49). The development of new molecular imaging techniques to detect high-grade pancreatic intraepithelial neoplasia (PanIN-3) may further improve prevention and early detection of this fatal disease (50–52).

Universal tumor testing for MMR deficiency, a paradigm-changing approach for identifying inherited cancer risk, has become standard practice for all newly diagnosed CRCs, recently extended to endometrial cancers (53,54) to identify Lynch syndrome probands. This benefits both the patient and at-risk family members for intensive and early screening and potentially aspirin and/or other NSAID prevention (54–56). The recent profound activity of immune checkpoint blockade to treat MMR-deficient cancers (57) has added to the enthusiasm for this approach (see below). A similar approach is under study in lung cancer patients: the ~1% with tumor EGFR T790M mutation at diagnosis have a high risk of carrying germline T790M mutations (58). These families appear to have a different lung tumor biology, supporting precision prevention (with T790M inhibitors) and early detection research (59). This highlights the substantial overlap between somatic driver mutations and germline predisposing mutations (19). Recent data reveal that pathogenic genetic variants identified within many cancer types are of germline origin in 10–15% of unselected childhood and adult cancers. As NGS becomes more widely used and germline-somatic relationships comprehensively mapped, shared and distinct oncogenic events can be integrated into the PCA and assessed for preventive targeting (18,19,60,61).

Immunologic mechanisms, including cancer vaccines, may also be key to realizing precision prevention in certain types of neoplasia characterized by immunogenic antigen production, as seen in various inherited settings (Table 1). Cancers that arise in Lynch syndrome with inherited DNA MMR gene defects display a high-level of microsatellite instability (MSI-H) and widespread accumulation of somatic frameshift mutations. This results in very large numbers of neoantigens/mutations (62) that make the tumor appear more foreign to the host immune system, underlying the success of immune checkpoint inhibitors and serving as a model of sporadic MSI-H tumors (57,63). These breakthrough advances using immunotherapy to treat MMR-deficient/MSI-H cancers along with early immunosurveillance (T-cells specific to MSI-related neoantigens (64)) in “healthy” Lynch syndrome carriers have generated interest in developing cancer vaccines as immune prevention targeting predictable frame-shift mutation-derived peptides. DNA damage response plays an important role in innate immunity, activates inflammatory cytokines, and induces the expression of immune-receptor ligands on damaged cells. As such, inhibitors of DNA damage response signaling may, in fact, attenuate the immune response following DNA damage (65). A specific mutational signature associated with both germline and somatic BRCA1 or 2 mutations (66) has been observed in breast, ovarian, pancreatic, gastric, and esophageal cancers (even those without BRCA1 or 2 mutations) (67–72) in association with markers of immunity in subsets of the former three cancers (73), suggesting a potential role for immune-based prevention against such cancers. In contrast, a recent seminal study revealed an increasingly complex interplay between chromosomal abnormalities and immunity. High-level arm- and chromosome-somatic copy number alterations (SCNAs), which can drive precancer/cancer progression, produced immune evasion. Consistent with gene-dosage imbalance rather than a specific gene, SCNAs drove an immunosuppressive microenvironment with immune depletion that opposed immune response (74), ), even in the setting of high mutational/ neoantigen burden as seen in mismatch repair-deficient tumors (57,63) revealing an increasingly complex interplay between chromosomal abnormalities and immunity.

Table 1:

Immunologic features of three distinct neoplastic pathways.

Germline mechanism of hypermutability	Characteristics of associated cancers	Loci of somatic hypermutations in invasive cancers	Evidence of immunogenic phenotype	Immunologic considerations in premalignant biology
DNA mismatch repair gene mutations: Lynch syndrome (LS); biallelic mismatch repair deficiency (BMMR-D)	Gl, gyn, and other cancers with high-level microsatellite instability (MSI); pediatric-onset Gl cancers, gliomas, and lymphomas in BMMR-D	Mono- and dinucleotide repeat microsatellites in non-coding and coding DNA; certain “hotspot” microsatellites within driver genes, e.g., TGFBR2 and BAX in LS; POLE or POLD1 in BMMR-D	MSI-induced frameshift neopeptides/neoantigens found in associated cancers; early success of PD-1 antibodies	Not systematically studied; MSI-H (presumably a surrogate for neoantigen load) seen in various premalignant states (adenomas, intestinal crypt foci, IPMNs, DCIS, etc.)
BRCA1/2 mutations	Breast, ovarian, pancreatic, prostate, and other cancers with homologous recombination deficiency	Specific signatures of tandem duplications and/or deletions	BRCA1/2-associated ovarian cancers demonstrate infiltrating lymphocytes and increased expression of genes related to immune-mediated cytotoxicity	Not well studied; loss of wild-type BRCA1/2 allele is seen in some premalignant lesions, but BRCA1/2 haploinsufficiency leads to increased expression of EGFR and genomic instability in non-neoplastic breast epithelium cells prior to loss of wild-type BRCA function
APOBEC3A/B chimeric deletion polymorphism	Modestly increased risk of breast (and possibly other) cancers with somatic hypermutation; increased stability of chimeric APOBEC3A/B mRNA leads to increased APOBEC3A activity correlating with germline copy number	TCA and TCG trinucleotide sequences, including certain “hotspots” within driver genes, e.g., PIK3CA	Cancers show upregulation of genes associated with immune activation, cytokine response, and lymphocytic infiltration; penetrance of immune-activating effects appears to be high	Not well studied; APOBEC3A/B-related mutational patterns seen in sporadic (i.e., not associated with germline polymorphism) pre-invasive bladder carcinoma and cervical intraepithelial neoplasia

This schematic outlines the rationale for investigating the immunobiology of premalignancy in three distinct inherited neoplastic pathways: mismatch repair deficiency, BRCA1/2 mutations, and APOBEC3A/B-mediated neoplasia. Cancers associated with all three pathways are associated with distinct, predictable forms of somatic hypermutability and have various features suggesting an immunogenic phenotype. In each case, however, an in-depth understanding of premalignant biology will be key to devising strategies for primary cancer prevention.

Genome-wide association studies (GWAS) have contributed to expanding catalogs of implicated genes and pathways for many complex human diseases and are beginning to shed light on shared and unique etiological and pathological disease components. A key challenge is that many GWAS-identified loci are not near known coding or regulatory regions, making determining the underlying mechanisms and functions related to the associations difficult. Linking susceptibility variants to their respective causative genes and cell-specific regulatory elements thus remains a high priority in order to realize the potential of association studies to advance understanding of disease biology, etiology, and prevention. Their ability to identify novel cancer genes/pathways (through functional follow-up studies) underlying the observed risk is now being exploited for future drug development or repositioning (75). While the low penetrance of most GWAS-identified risk loci has limited translation to clinical trials, the combined effects of such SNPs are being used to create robust polygenic risk scores (PRS) (76), which may be useful in developing personalized risk estimates. Combining large-scale GWAS findings across cancer types (breast, ovarian, and prostate) and using fine-mapping pathway analysis and PRS discovered cross-cancer risk loci with the potential to shed new light on the shared biology underlying these hormone-related cancers (77). A recent GWAS led to the identification of miR-3662 at the 6q23.3 locus shown in mechanistic studies (when overexpressed) to inhibit NFkB signaling and leukemogenesis (78). Certain SNPs (e.g., involving APOBEC3), underlying cancer (breast and bladder) risk are linked to hypermutability and immune activation (79–81) and intestinal barrier function (discussed above). Future possibilities of drug development or repositioning, possibly aided by studies in relevant animal models, include GWAS-identified loci at 1p31.3 (82) and 6q25 (83), which helped identify cancer prevention relevant drug targets IL-17 (key immune/microbiome target) and ESR1 (tamoxifen target). A GWAS-identified locus at 6p23 (84), associated with breast cancer risk in BRCA2 carriers, is located near the SIRT5 gene, a drug target under active study (85).

Recent findings confirm the occurrence of widespread genetic regulation of immune and host defense pathways overlapping disease loci and involving not only gene expression but also splicing and epigenetic modifications, including CRC precursors (86). The results suggest the convergence of independent regulatory layers for cell-specific function, and used independent measurements to yield robust biological validity to mapped traits for associations between genetic variants and intermediate phenotypes, providing powerful approaches to annotate the putative consequence of disease associations. The biological resolution of this approach was further increased by assessing by evaluating context-specific events in studies probing multiple primary cell types, in different conditions, and extending analyses beyond gene expression to histone modification or methylation greatly enhancing the functional and mechanistic interpretation of genetic associations.

Mitochondrial Biology and Genetics in Cancer Predisposition

Mitochondrial DNA (mtDNA) with its mutations and polymorphisms is a relatively underappreciated field in cancer research. Most mitochondrial proteins are nuclear encoded. Human mtDNA is maternally inherited and exists as a circular, double-stranded genome encoding for 37 mitochondrial genes: 22 transfer RNAs, 2 mitochondrial ribosomal RNAs and 13 protein subunits of the ETC complexes (with the exception of Complex II, which is nuclear encoded) and ATP synthase (mtOXPHOS proteins), essential for respiration. There are several mtDNA copies per mitochondrion and hundreds of mitochondria per cell (87). Generally, neoplastic cells possess functional mitochondria and retain the ability to conduct oxidative phosphorylation. In fact, targeted depletion of mitochondrial DNA generally reduces tumorigenic potential in vivo (88). While it has long been known that somatic mtDNA alterations are frequently acquired during oncogenesis (89), recent intriguing germline data indicate that mtDNA variants influence multiple innate mitochondrial functions, including reactive oxygen species (ROS) production and redox control, signal transduction and epigenome systems, autophagy, apoptosis, and immunity (90). mtDNA has a very high mutation rate, over an order of magnitude higher than the somatic nuclear genes. Furthermore, mtDNA genes are intimately linked with ~2000 nuclear genes encoding proteins that function within mitochondria, which can produce nuclear inheritance of mitochondrial disease. The most common germline mtDNA mutations in neoplasia occur in the non-gene encoding region. Inherited high-penetrance of deleterious missense alterations in mtDNA genes, such as ND6 and COI, which code for subunits of OXPHOS complexes I and IV, have been associated with risks of various cancers. For example, oncocytomas tend to have mutations in one of the seven mtDNA coded polypeptides of respiratory complex I (91), while missense mutations in the mtDNA complex IV (cytochrome c oxidase) subunit COI gene are commonly found in prostate cancer, and certain African mtDNA lineages harbor COI gene variants that may contribute to cancer risk among African Americans (4,92). MtDNA variants have been associated with ovarian (93), bladder (94), breast (95), endometrial (96), and HPV-infection and cervical cancers (97), among multiple other cancers. Mitochondrial mutations are frequent in Barrett’s metaplasia without dysplasia (98). The progression of nonalcoholic fatty liver disease to the liver precancer nonalcoholic steatohepatitis (NASH) has been shown to be associated with a mtDNA SNP in the Mtatp8 gene (a subunit of OXPHOS complex V, ATP synthase). This variant has profound effects on hepatic lipid and acylcarnitine metabolism and susceptibility to diet-induced (e.g., high-fat Western diet) NASH (99). Generating mouse models of these and other deleterious missense mt germline mutations, e.g., COI nt 6589T>C V421A (100) and ND6 nucleotide G13997A P25L (101), will be critical to probe mitochondrial biology.

Deleterious alterations in mtDNA are inherently heteroplasmic (harboring a mixture of mutant and wild-type mtDNA) with high levels of these severe mutations being lethal. Since mtDNA transmission during mitosis is the result of stochastic distribution into daughter cells, milder mtDNA polymorphisms can shift to become predominantly enriched within individual cells and closer to pure mutant (homoplasmy), resulting in neoplastic transformation. The importance of this phenomenon for cancer predisposition has been demonstrated in a pedigree in which a mtDNA complex I ND5 m.12425delA frameshift mutation was transmitted through the maternal germline at lower heteroplasmy levels (5–10% mutant) and was thus masked due to the preponderance of WT mtDNAs. Thus, while the transmission of the mutant mtDNA in this pedigree was phenotypically silent, the chance increase of the mutant mtDNA in somatic cells caused neoplastic transformation and seemingly sporadic cancer (102). In contrast to Mendelian genetics, mtDNA heteroplasmic genotype is continuously changing during successive cytokinesis to generate cells with varying oncogenic potential (92,103). Widespread heterogeneity has been reported in the mtDNA of normal human cells. Furthermore, the frequency of heteroplasmic variants among different tissues of the same individual vary considerably indicating that individuals (and perhaps even a single cell) are characterized by a complex mixture of related genotypes rather than a single genotype. Mechanistic study of heteroplasmy (shifting percentages of WT and mutant mtDNAs) regulation will yield novel prevention insights. Because of its high mutation rate, human mtDNA is highly polymorphic, harboring functional variants that can be beneficial or deleterious depending on environmental context. Because of the strict maternal inheritance of the mtDNA, sequencing mtDNAs from global indigenous populations permitted reconstructing the origins and ancient migrations of women (initially from Africa) (90). A subset of the mtDNA variants cause subtle changes in OXPHOS, which in turn could modulate a wide range of context-dependent cellular functions of adaptive relevance, including inflammatory, stress, autophagy, and oncogenic responses to diet and other factors (104,105). Those functional mtDNA variants that were beneficial (adaptive) in a particular environment increased in number and gave rise to descendant mtDNAs, which share the founders’ beneficial variant creating a regional group of related mtDNA haplotypes known as a haplogroup. A result of migration has been that previously adaptive mtDNA haplogroups have become maladaptive and predisposed to a wide range of common diseases, including diabetes, obesity, aging and cancer (90). Dietary or caloric restriction slows the development of many age related diseases, including cancers, although the mechanisms involved are complex and context-dependent, such as autophagy (106) and translational regulation involving mitochondrial (and ribosomal) genes/proteins/components (107). Autophagy is essential for BRAF V600E-driven melanoma and lung tumor development in GEMMs by overcoming senescence; deletion of Atg7 inhibits tumorigenesis, likely via a mitochondrial mechanism (108).

A recent murine study highlighted the importance of mtDNA genetic background in tumorigenesis by examining PyMT transgenic mice (inherently predisposed to developing mammary tumors) with identical nuclear genomes but varying mtDNA backgrounds. The mtDNA background influenced both mammary tumor latency and progression (109). In normal mice, the mitochondrial-targeted catalase transgene (mtCAT) reduces somatic mtDNA mutations (110) and when crossed with PyMT, the incidence of mammary cancer was greatly reduced (111). Consistent with cancer predisposition, mechanistic studies demonstrated the profound influence of subtle changes in mtDNA haplotype/variation on obesity and aging – two common cancer risk factors (112). Mitochondria may also be intimately involved with T-cell tumor surveillance. T-effector cells are more glycolytic while Treg cells are more oxidative. Within neoplasia, glucose is converted to lactate (which promotes inflammation, angiogenesis and tumorigenesis), thus inhibiting T-effector function. By contrast, the Treg cells can metabolize lactate by OXPHOS (113), further inhibiting T-effector cell immune rejection (114). Germline mtDNA ND6 P25L-mutant mice harboring a mild mtDNA complex I gene mutation have reduced Treg cells (113,114), suggesting that tumor immune rejection might be enhanced by mild complex I inhibitors such as metformin, whose effectiveness should be increased in neoplastic cells with partial OXPHOS dysfunction (115,116). Mitochondria can also influence the inflammasome, innate immunity, IL-1β, and NFkB inflammatory pathways (7,117).

Adding to the tissue- and geographic-specific context of mitochondrial effects (118), emerging data also suggest a complex interplay between the nucleus-cytosol and mitochondria. Murine models with germline mutations in the nuclear gene SUV3, which encodes for a mtRNA helicase, have marked somatic mtDNA instability, hypermutability, shortened lifespan, and various cancers – a unique model to study mitochondrial genomic instability in cancer predisposition. Clinical relevance was shown by reduced SUV3 expression in two independent cohorts of human breast cancer (5). Mutations in nuclear DNA genes influencing transformation include some of the same targets/mechanisms affected by mtDNA, including TETs, succinate, fumarate, NRF2, and α-ketoglutarate dioxygenases (119,120) – all important in cancer risk. Nuclear BRCA1 has been found in the mitochondria, where its function is unclear (121,122). Germline BRCA1 mutation reprograms breast epithelial cell metabolism towards mitochondrial-dependent biosynthesis and increased risk of oncogenesis. Metformin studies in BRCA1 haploinsufficiency-driven oncogenesis support potential prevention approaches in BRCA1 carriers: inhibition of complex I and restriction of mitochondrial-dependent biosynthetic intermediates (123) may open a new avenue for “starvation” strategies; and regulating mitochondrial-nuclear communication and modulating the epigenetic landscape (targeting histone acetylation) in genomically unstable precancerous cells (124), might guide the development of new metabolomic-epigenetic strategies. As with nuclear GWAS, certain mtDNA alterations modify (lower) risks of breast cancer in germline BRCA2 mutations (125). Future GWAS integrating nuclear and mitochondrial studies will provide a more full germline landscape.

Big Genomics Data of Premalignant Somatic Tissues

The collection and analyses of NGS big data are beginning to provide biological insights into prevention/early detection in the context of studies characterizing somatic genomic alterations. It is worth noting that, in addition to comprehensive analyses of “big genomics data,” there are few recent studies that have also examined cancer microbiomes (reviewed in (104,126–128)), transcriptomes (129–131) and epigenomes (reviewed in (132,133)). In addition to big data generated from somatic sequencing efforts, GWAS has involved hundreds of thousands of cancer patients across most organ sites identifying ~3000 cancer-related genetic associations (recently reviewed (134,135)) and has been studied in some precancers – Barrett’s esophagus (BE) (136,137), colorectal adenomas (26), ductal carcinoma in situ (DCIS) (138) and hematologic premalignancies (below).

The genome of a malignancy can be examined as an archeological record bearing the cumulative imprints of all mutational processes that have been operative throughout the cellular lineage between the fertilized egg and cancer (139). Each mutational process leaves a characteristic imprint, termed mutational signature, which can change over time, and almost all mutational signatures detected in a cancer genome have been imprinted during the precursor phase of a cancer cell (139). Examination of the cancer genomes from >12,000 patients has revealed more than 30 known distinct mutational signatures (66,139), including those related to environmental exposures, such as UV-light, aflatoxin, and tobacco (cancer.sanger.ac.uk/cosmic/signatures). Some of these signatures have already been used for identifying the presence of specific carcinogens, including aristolochic acid, one of the most potent known human carcinogens – a chemical present in certain plants still in use even today – to global risks of urologic and hepatic cancers (140). It will be important for the PCA to study mutational signatures of premalignant and normal aged tissues, perhaps guided by the many cancer signatures, to provide a lens into mutational precancer patterns. However, it is important to note that approximately half of the currently known cancer signatures have unknown etiology and ongoing efforts have started exposing experimental systems to known carcinogens in an attempt to reproduce/identify them (141,142).

Another set of widespread and extensively studied endogenous mutational signatures are the ones attributed to ectopic activity of the APOBEC family of deaminases (66,143). Recent examination across more than 10,000 specimens from 36 distinct cancer types revealed that these signatures are found in more than 30% of cancer samples and account for approximately 15% of all somatic mutations across these cancers. The activity of these APOBEC mutational signatures is especially strong in bladder and cervical cancer where they account for more than 75% of all somatic mutations in each of these cancer types (143). In cancers of the cervix and oropharynx, these APOBEC mutational signatures are predominately triggered early by HPV infection (143,144). It has been speculated that the APOBEC mutational signatures have been imprinted during the precancer phases of these cancers (143,144). In lung and most other cancer types, APOBEC signatures are believed to be late events and are found in subclonal expansions and intra-tumor heterogeneity (145). Recent studies using mutational signatures have also demonstrated that the presence of certain germline variants can affect the accumulation of somatic mutations due to environmental exposures. For example, disruptive germline polymorphisms in MC1R, contributor to phosphorylation of DNA repair proteins, have been associated with increased number of somatic mutations due to a UV-light-related mutational signature and a higher risk for developing skin cancer (146).

Multiple mutational signatures reflect failure of different DNA repair pathways. A mutational signature reflecting the accumulation of unrepaired ROS, mainly 8-oxoguanine, has been attributed to failure of base excision repair (BER) due to defects in MUTYH, and has been identified in colorectal cancers and adenomas arising in individuals with pathogenic germline MUTYH mutations (147). Additionally, a failure of transcription-coupled nucleotide excision repair (NER) due to somatic mutations in ERCC2 in bladder (including preinvasive) neoplasia has been shown to exhibit a specific mutational signature (148). Germline defects in other NER genes can cause Xeroderma pigmentosum (XP), a rare autosomal recessive genetic disorder associated with high risk of UV-associated skin cancer due to faster accumulation of UV associated DNA damage and mutational signatures (149). UV-induced non-melanoma lesions can be reduced using bacterial DNA repair enzymes (150) or nicotinamide, which can prevent UV-induced immune suppression and enhance DNA repair (151).

In addition to mutational signatures related to failure of DNA repair mechanisms and ones due to endogenous/exogenous exposures, large-scale genomics studies have also identified mutational signatures responsible for the unavoidable background mutation rate in somatic cells. Notably, two mutational signatures (which are not correlated and have different frequencies in different tissues) have been found to act as endogenous mutational clocks, characterized by accumulating somatic mutations within all normal somatic cells of the human body with the progression of age (152,153). One of these mutational signatures has been attributed to spontaneous deamination of 5-methylcytosine in the context of CpG (its rate of “ticking” appears to be influenced by cellular division), while the etiology of the second clock-like signature remains unknown. Interestingly, a recent study demonstrated the increased rate of one mutational clock to be mechanistically linked to tobacco smoking (148,154). The somatic mutation loads in single-cell lineages provide information about an individual’s lifetime history of mutagenic exposure and the impact of intrinsic factors on mutagenesis. Expanding this study to precancers, more cell types, and larger populations would further refine estimates of the rates of somatic changes in human genomes. Understanding the contributions of environmental and endogenous mutagenic processes to somatic mutation loads is fundamental to develop preventive strategies (155).

Analyses of omics data from precursors are beginning to emerge. Despite their cross-sectional, precancer/cancer pair designs, and relatively small sample sizes, these emerging data suggest that many precancers share genomic alterations with their respective invasive cancers, including ductal and lobular breast cancer (156,157), pancreas (158), non-melanoma skin (159,160), melanoma (161), lung adenocarcinoma (162) and colorectal neoplasia (147). From an NGS, genomics, transcriptomics (see below) and big data perspective, BE is the best-studied epithelial precancer (67,163–166), including three recent GWAS (136,137,167) and a post-GWAS analysis reporting some CDKN2A SNPs associated with reduced EAC risk (168). Somatic tissue studies of Barrett’s/esophageal adenocarcinoma (EAC) pairs revealed that most recurrently mutated genes in EAC were remarkably similar to the matched precancer, only TP53 and SMAD4 were associated with advanced neoplasia (165). Intra-tumor genomic heterogeneity, with some contribution of aberrant methylation, drives neoplastic progression in this setting (see longitudinal section below) and most cancers (169).

Two recent large-scale NGS of mtDNA of cancer (total > 2,000 human cancers, 30 tumor types) identified a unique heavy strand-specific C > T transitions and mutational signature. More importantly, this cancer mitochondrial missense mutational signature was considered neutral (analogous to passenger mutations in nuclear DNA) and did not compromise the function of the mitochondria (170). One of these studies further refined the mtDNA mutational map by requiring that mutations also be detectable in matched transcriptome sequencing (RNA-seq) data from the same tumors (171). Although DNA/RNA allelic ratios generally were consistent, some mutations in mt-tRNAs displayed strong allelic imbalances caused by accumulation of unprocessed tRNA precursors, indicative of impaired tRNA folding and maturation, which underlie a range of diseases. Both studies found a selective pressure against deleterious coding mutations affecting oxidative phosphorylation, indicating that tumors require functional mitochondria. Unexpectedly, known dominant mutagens, such as cigarette smoke or UV light, had a negligible effect on mtDNA mutations. Another recent study has reported significant correlations between mtRNA-Seq and mtDNA copy number, with some important exceptions (e.g. MT-ND5 and MT-ND6) (172). Clonal expansion of mtDNA mutations can result in mitochondrial dysfunction, such as decreased ETC enzyme activity and impaired cellular respiration. NGS of mtDNA of oncocytomas, which are rare benign tumors of epithelial cells defined by excessive amount of mitochondria, has identified a pathogenic mutation signature that compromises the overall function of the mitochondria, proposed to serve as a metabolic barrier for these benign tumors, and perhaps precancers, to progress to more malignant tumors (173).

In addition to the studies above, a systematic approach to classify cancers using transcription profiles at both bulk tumors and single-cell resolution have been well described (174–177). These profiles not only provide molecular basis for classifying cancers with shared transcriptional programs across different cancers but also characterize heterogeneity that exists within individual precancers (178) and tumors. Whole transcriptome profiling using RNAseq, pathway enrichment, and functional assays of BE found novel cell-cell interactions between dysplastic and normal epithelial cells (which often coexist in vivo) in the microenvironment that can dramatically suppress dysplastic cell behavior. These effects are distinct from the stromal and immune cell microenvironment effects on precancer. Differential gene expression revealed TGFβ, EGF, and Wnt as key pathways associated with the differential transcriptional profiles observed in co- vs. mono-culture (166). Single-cell approaches will allow analysis of different subpopulations of cells, including the highly variable epithelial cell motility as well as enumerating the immune cell infiltrates, stromal cells, and other microenvironment components surrounding the neoplasias (166,177). Comparing gene expression-based subtypes defined in tumors with those in their precursor lesion will provide insights that can inform which precursors may progress into malignant disease. Further, oncogenic pathways and developmental- and immune-based gene expression signatures can be used for “pathway/phenotype”-based molecular characterization (179–181). More recently, a novel analytic approach to define oncogenic states and produce functional maps of cancer has been established. This serves as a framework for combining experimental and computational strategies to deconvolve oncogenic pathways/signatures derived from oncogene activation into transcriptional components that can be used to determine oncogenic states. By mapping precancers and tumors onto distinct oncogenic states, the resulting functional map can be used to characterize how these states relate to various omics features, including NGS mutations, copy number alterations, gene and protein expression, gene dependencies, and biological phenotypes; and to predict which interventions are more likely to have a significant effect (182). This approach was used to effectively map cancers with altered KRAS/MAPK pathways into divergent functional states. Studies in pancreatic oncogenesis highlight the need for big data approaches to interpret neoplastic complexity (including KRAS mutation subtypes and Hippo pathway interactions), profound effects on cell metabolism, DNA repair, immunity, mitochondrial biology, and distinct precursor pathways (183). Mutant Kras in pancreatic acinar cells induces expression of ICAM-1 to attract macrophages and drive PanIN development: direct early cooperative mechanism between a driver mutation and inflammatory environment (184). Even B-cells can initiate pancreatic tumorigenesis (185). These maps can be generalized to consider gene networks and interactions (186), including the contributions of the germline and to study the close interplay with the immune microenvironment. Integration of the results from functional genomics studies described above to the functional maps of oncogenic states will provide insight into the cellular contexts in which genomic alterations contribute to malignant transformation (187).

Epigenetics

Previous work has yielded only a limited big data perspective of the neoplastic epigenome, primarily in hematologic neoplasia, where chromatin modifiers are among the most frequently mutated in cancer in general (133,188). Most studies have focused on performing functional analysis on a few genes in a limited number of samples and are reviewed below. Widespread epigenetic field defects have been observed in apparently normal breast tissue located adjacent to breast cancer (189) and also associated with inflammation-related cancers, such as H. pylori-induced neoplasia, in which NGS has revealed more cancer pathway-related genes affected by DNA methylation than by genetic alterations (190). The ten eleven translocation (TET) enzymes oxidize 5-methylcytosines (5mCs) and promote locus-specific reversal of DNA methylation (191).

An epigenetic mitotic-like clock was developed using a novel approach based on an underlying mathematical model. A key feature underlying the construction of this clock is the focus on Polycomb group target promoter CpGs, which are unmethylated in many different fetal tissue types, thus allowing defining a proper ground state from which to then assess deviations in aged tissue. By correlating the tick rate predictions of this model to the rate of stem cell divisions in normal tissue, as well as to an mRNA expression-based mitotic index in cancer tissue, this model approximates a mitotic-like clock. The epigenetic mitotic clock-like signature exhibits a consistent, universal pattern of acceleration in cancer in normal epithelial cells exposed to a major carcinogen. The epigenetic clonal mosaicism is maximal before cancer emerges. Unlike the recently identified mutational clock-like signatures discussed above, this epigenetic clock is based on clinical DCIS and lung CIS progression to cancer and normal at-risk tissue; a concrete example of a molecular mitotic-like clock that predicts universal acceleration in precancer (192). Smoking was associated with an increased rate of this mutational clock. Another approach to ITH analyzed DNA methylation patterns at two genomic loci that were assumed to have no role in gene regulation, in contrast to driver methylation changes. Methylation at such neutral loci were unlikely to be under selective pressure and therefore, could serve as a “molecular clock” to measure mitotic divisions based on the higher error rate of DNA methylation maintenance relative to the error rate of DNA polymerase (193). This molecular clock analysis revealed highly heterogeneous tumors, suggesting that the tumors had not undergone any recent clonal expansion.

Aberrations of the epigenetic modulator TET2 are one of the first alterations in several hematologic premalignancies (TET1 and TET3 are rare in hematologic neoplasias). TET2 mutations are found in premalignant hematopoietic stem cells (HSCs), including of myeloproliferative neoplasms (MPNs) and myelodysplastic syndrome (MDS), and are frequently observed in aged healthy individuals (194) with propensity to transform (see clonal hematopoiesis below). Disruption of TET2 in mouse models increases HSC proliferation and clonal expansion, prone to additional oncogenic events, which are generally required for malignant transformation (191). Mouse models have addressed the functional relevance of co-occurring alterations and found that Tet2 disruption with Asxl, Ezh2, or Jak2 V617F results in MDS or MPN phenotypes. Recurrent dominant point mutations in IDH1 and IDH2 appear to be early events in glioblastoma (affecting a common glial precursor cell population) and hematologic neoplasias that lead to loss of TET activity and other epigenetic changes (195,196). In addition, TET2, IDH1, and IDH2 mutations are frequently observed in lymphoma precursors (197–199), and the frequency of TET loss-of-function (which drives hematologic transformation) in these settings supports testing IDH inhibitors and/or TET activators. For instance, ascorbic acid acts as a cofactor for the α-ketoglutarate-dependent dioxygenases and can affect DNA methylation in embryonic stem cells and mouse embryonic fibroblasts in a manner that is dependent on TET2. TET modulators (200), can enhance antigen presentation and increase IL-6 production by macrophages (201), affect regulatory T-cells (Tregs) (202), and alter expression of endogenous retroviruses, cancer testis antigens, and stem cell antigens in premalignant lesions resulting in enhanced immunogenicity (203).

Another epigenetic mechanism found to be important in premalignant biology involves RNA editing by ADAR enzymes, which results in adenosine-to-inosine conversion of RNA thereby inducing virtual adenosine-to-guanine mutations since inosine bears molecular resemblance to guanine (204). Depending on whether the editing events occur in coding regions or 3’ UTRs, ADAR-mediated editing of mRNAs can result in post-transcriptional protein coding mutations or altered susceptibility to microRNAs (205). Recent data suggest that germline variation involving RNA editing ADAR genes may influence cancer (ovarian) susceptibility (206). ADAR1 editase activity has been implicated in the oncogenic transformation of premalignant progenitors that harbor clonal self-renewal, survival, and cell cycle-altering mutations (207,208), such as in hepatocellular carcinoma precursors, where aberrant RNA editing of AZIN1 has been found to be a key oncogenic driver (207,209). Transgene expression of APOBEC-1 causes dysplasia and cancer in mouse and rabbit livers likely due to RNA editing of NAT1 (210). Study of ADAR1 regulation of APOBEC3 in neoplasia will be critical, potentially suppressing hypermutation and immunity (211). Finally, inflammatory cytokine networks and JAK2/STAT signaling activate ADAR1 during relapse/progression in leukemia stem cell renewal, linking RNA editing to the development of innate immunity and potential preventive activity (212).

Emerging data also suggest that some premalignant lesions may progress to cancer via a fundamental epigenetic reprogramming. Epigenetic defects may be a common mechanism linking genetic mutations to cancer phenotypes, although the details on how they are linked are just beginning to be explored. Indeed, reprogramming of the epigenome to a progenitor-like state may be required for driver mutations to induce tumorigenesis (213). The role of BRAF mutations in benign nevi is a major historical conundrum in the melanoma field (214). In the BRAF V600E zebrafish model of melanoma, deletion of p53 promotes the nevus-to-melanoma transformation, but melanomas remain surprisingly infrequent considering that all of the cells bear both the oncogene and tumor suppressor loss (215) – a feature that replicates the phenomenon of “field defect” in human tumors. Two recent studies using preclinical models addressed this issue. Work with BRAF V600E/p53-null zebrafish now suggests that initiation of malignant transformation within such a “cancerized field” requires fundamental epigenetic reprogramming of these premalignant cells into an embryonic state via transcription factor-mediated reactivation of genes typically expressed only in neural crest progenitor cells (213). This reprogramming involves binding of multiple transcription factors and generation of “superenhancer” regions. New engineered human models, including epigenetic mechanistic studies, suggest a key role of p15 loss in promoting BRAF V600E-mutant benign melanocytic nevi transformation to melanoma (216). Similarly, mouse model research recently demonstrated that basal cell carcinomas, known to be driven by oncogenic signaling in the hedgehog pathway, only originate from stem cells located in very specific areas of the murine epidermis, rather than from more committed progenitor cells (217). Like the zebrafish model, this study provides evidence that the earliest stages of tumorigenesis are characterized by reprogramming to a more embryonic cell state. Such data suggest that tumor-initiating cells can be identified – and potentially targeted for early destruction – through their ability to reactivate an “embryonic” epigenetic state and highlights the importance of studying premalignant cells and model systems to better understand when epigenomic changes arise and how stable they are over time.

The Power of Immunology and Biochemistry to Facilitate Cancer Prevention

Immune oncology

The integration of multiple omics analysis platforms with immune-informatics analysis can be the foundation of a more effective framework for precision prevention (218). There is now a wealth of evidence from both animal models and cancer patients of how the immune system can survey and recognize peptides encoded by certain genetic mutations when such peptides are presented on the surface of the cancer cell bound to MHC-Class I and Class II molecules. For example, RAS mutations, which are key oncogenic drivers in a wide array of cancers, may also be targets of immunosurveillance since T-cells specific to mutated RAS peptides have been found in cancer patients (219) and may be a viable target for immune approaches to treatment and even prevention (220). Proof-of-principle studies of vaccine targeting mutant Kras (with Treg depletion) in a pancreas mouse model induced CD8⁺ T-cells specific for the Kras mutation and showed preventive efficacy in the early PanIN setting (221). In addition to predicted mutations in well-known oncogenes, cancer cells and their precursors can harbor tens to hundreds of random mutations throughout their genome. Elucidation of the mutated precancer repertoire will allow for efforts aimed at determining which mutated genes produce peptides that can bind MHC molecules and be presented to the immune system (222–224) as potential targets for immunosurveillance and vaccination.

Vaccine-based approaches hold particular promise since they are a form of precision prevention with few side effects. Furthermore, vaccines (e.g., to HPV) could provide long-term protection from cancer development after only one or two treatments unlike prevention drugs that must be taken for many years, challenging an individual’s compliance and/or will to endure accompanying toxicities (9,225). Evidence for immune surveillance has been reported in healthy people and associated with lowered lifetime cancer risk. Childhood febrile viral infections have been associated with reduced cancer risk. Recent mouse model data found that influenza virus infection elicited protective antibodies and T-cells specific for host antigens also expressed on some tumors (226). These data suggest that infection-induced immunity and immune memory could provide long-term immune surveillance of cancer and have important implications for vaccine targets. T-cells are likely the main effector cells in preventing all forms of cancer. The immune system has the ability to recognize precancers and generate immune responses to potentially intercept and prevent cancer (47,64,227) and avoiding immune elimination is a hallmark of cancer (228). We must learn how to both strengthen T-cell immunity – either through immunization, drugs, or engineering – and concurrently overcome a hostile dynamic tumor microenvironment that prevents T-cell activation and infiltration into early neoplasia. The latter involves multiple factors, for example, metabolic reprogramming of the microenvironment by the high utilization of extracellular glucose and glutamine results in extracellular lactate which attenuate dendritic and T-cell activation, stimulate macrophage polarization to an M2 state, induces VEGF secretion by stromal cells and activates NF-kB. The microenvironment can in turn have profound effects on the metabolism of neoplastic cells (88). Emerging data suggest that the microenvironment barriers develop early in precursor lesions but are likely qualitatively different from more established cancer-associated barriers. The progressive accumulation of somatic changes that lead to neoplasia also co-opt neighboring vascular, neuronal, and other normal cells to support/promote oncogenesis. Critical to vaccine development, therefore, is the identification of potent immune enhancers/adjuvants that can specifically target one or more innate pathways (229) and alter the developing inflammation that promotes immune suppression in favor of a neoplastic response (230). Experience with therapeutic cancer vaccines shows that targeting a single antigen or a single mutated peptide invariably leads to outgrowth of cancer cells that have lost that mutation. This may happen in the precancer setting as well, requiring a vaccine that elicits a polyclonal and polyspecific immune response. Trial endpoints could include T-cell receptor sequencing to look at clonality and clone expansion, liquid biopsy approaches to detect low levels of the identified mutations or mutational load, and novel techniques to image immune response (231) and high-grade pre-invasive neoplasia (PanIN-3) (52), or depending on risk, even cancer incidence.

In a clinical feasibility trial in advanced adenoma patients, lack of immune response to a MUC1 cancer vaccine correlated with increased levels of circulating MDSCs responsible for inhibiting adaptive immunity (232), suggesting that these may be useful biomarkers to identify individuals unlikely to benefit from preventive cancer vaccines. For those deemed unlikely to respond to the vaccine alone, research into other immunomodulatory drugs that could help overcome such immune resistance will be critical. Metformin, for example, has been shown to enhance T-cell immunity and immune memory, influence the microbiome in mouse models, by various mechanisms including involving mitochondrial biology and RANK-L inhibition (see above) (37,233–235). Furthermore, recent prospective cohort data suggest that aspirin prevention of CRC is related to its effects on T-cell immunity (236). Adenomas have been reported to have a highly inflammatory microenvironment (237), which varied by histology and location in a recent large microbiota/adenoma study (238). Two recent NGS of tumor from patients with IBD (colitis)-associated CRC were compared with sporadic CRC. The comparisons suggested that colitis-associated CRCs have a distinct mutational profile associated with cell-to-cell signaling, cell adhesion, and epigenetic regulators/chromatin modifiers, all of which may be linked with the inflammatory mediators of IBD (239,240). IDH1 mutation (extremely rare in sporadic CRC) was found only in Crohn’s disease. Extension of NGS to include epigenomic and microbiome profiles in IBD dysplasia has great potential for cancer prevention and early detection research.

Microbiota-immune interactions are an increasingly important and challenging field. Each human body contains at least 40 trillion microorganisms that populate complex ecosystems called microbiomes; 99% of microbiota reside in the gut microbiome and certain bacteria can influence oncogenesis and immune interventions via complex microenvironment interplay. Germline/microbiota interactions effects are discussed above. Intestinal barrier function is regulated by inflammatory cytokines such as IL-1β and IL-18 (241), autophagy (242), and microbiota-accessible carbohydrates (which affect gut mucous layer and microbiota special organization) (243). Thymic stromal lymphopoietin (TSLP) is a cytokine expressed mainly by epithelial cells at barrier surfaces (skin, gut, and lung). Short-course calcipotriol, a topical TSLP inducer FDA approved for psoriasis, suppressed skin cancer development in genetically engineered mouse models (GEMMs) in a TSLP-dependent, long-lasting manner consistent with an immune memory response. A randomized clinical trial of this agent showed a marked reduction in actinic keratosis number mediated by specific induction CD4+ T-cell adaptive immunity (244). Non-specific innate immune activation by imiquimod is also active in actinic keratosis patients (245). Potential immune and/or microbiota prevention factors include lifestyle (104), metformin (234), antibiotics, diet, and microbial reprogramming (246,247). It has been recently shown that gut microbes modulate whole host immune and hormonal factors impacting the fate of distant precancers toward malignancy or regression, for example, by stimulating host immune cells to prevent dietary and genetic predisposition to mammary cancer in mice. This raises the possibility that the tumor microenvironment interacts with broader systemic microbial-immune networks (248). Caloric restriction is the most consistently effective cancer preventive approach in virtually every mouse model/tumor type tested. Recent data, including from the mutant Kras lung mouse model, indicate that caloric restriction or its mimetics (e.g., over-the counter hydroxycitrate) elicits autophagy, which improves immunosurveillance via Treg depletion and prevents malignant transformation (249).

Analyses of NGS genomic data are critical to develop vaccines that target specific epitopes derived from mutations, copy number alterations or other variants common to precancers. However, this direct strategy is especially challenging given the large number of alterations, the low penetrance of driver mutations, the so-called “long tail” problem (low frequency mutations) (250), and the fact that the corresponding mutant peptides do not always lead to effective antigen presentation and response (251). Recent work demonstrates the ability to utilized mass spectrometry (MS)-based analysis to identify attractive target antigen candidates from a native human melanoma tissue, which were subsequently narrowed down using somatic mutation information and subsequent immunogenic assays in mice (252). This approach, coupled with NGS genomic data in precursor lesions, will help better nominate strong candidate antigens for vaccinations. Computational methodologies can also help to identify suitable antigens from a large number of candidates, for example, by using existing resources and databases that catalog potential antigens (253). Prediction of peptide binding affinity to HLA I and II has also been developed using machine-learning classification approaches (254,255). Computational studies have focused on neoantigen-epitope prediction algorithms and have shown that only a very small proportion of predicted neo-epitopes are actually presented on MHC class I as targets of endogenous T-cell responses (57,256,257). Whole-exome sequencing has identified higher antigen load was predictive of overall lymphocytic infiltration, tumor-infiltrating lymphocytes (TILs), memory T-cells, survival in colorectal cancers (258). Using the NGS genomic data from precancers, we will use two strategies to nominate candidate antigens: 1) use the mutation calling algorithms to identify the most frequently occurring neoantigens and 2) for the low frequency events, utilize functional maps described above to identify complementary neoantigens that associate with oncogenic states (182). These filtered lists of neoantigens will then be used to predict strong epitope candidates in silico using algorithms that employ either Artificial Neural Networks (ANN) or Support Vector Machines (SVM). These analytic pipelines can be used generate lists of the most likely vaccine candidates based on stabilized peptide p–MHC-I binding affinity (256), an approach that has already produced promising recent results in mice (259). Attractive candidate neoantigens from this analysis will be used for systematic testing in vivo and for vaccine generation and characterization.

Developing cancer prevention vaccines will require in-depth insight into the molecular events that drive premalignant biology, building off of groundbreaking biochemical studies. A recent analysis found somatic mutations that disrupt beta-2 microglobulin (B2M; a component of the class I MHC complex) protein-protein interactions, with a striking enrichment for mutations at protein interaction interfaces involving B2M’s binding partners (260). It has been shown that disruption of B2M can minimize immunogenicity of human embryonic stem cells (i.e., foreign human cells with possible regenerative benefit but that cause immune reactions) (261). It is conceivable that such mechanisms may be employed by precancers and cancers to escape immune surveillance (262,263).

Biochemistry: Understanding the molecular basis of neoplasia

Recent TCGA and related studies have demonstrated that a large number of genetic and epigenetic factors, such as chromatin modifiers and remodelers, are highly mutated in a large number of solid tumors and in hematological malignancies (264). Recurrent mutations in genes that encode regulators of chromatin structure and function highlight the central role that aberrant epigenetic regulation plays in the pathogenesis of these neoplasms. Deciphering the molecular mechanisms for how alterations in epigenetic modifiers, specifically histone and DNA methylases and demethylases, drive hematopoietic transformation could provide new avenues for developing novel targeted epigenetic prevention for hematological neoplasia and could also inform future studies in solid tumors. Many such protein complexes – including the mixed-lineage leukemia (MLL) family (188), the polycomb complexes PRC1 and PRC2, which contain EZH2, ASXL1 and BAP1 (265), and the SWI/SNF chromatin remodeling complex (266) – contain genes that are frequently mutated in human cancers (264) but were initially identified in simple model systems, such as Drosophila and yeast, emphasizing the importance of including studies of model organisms in any large-scale efforts in cancer prevention. While genomic deletions and nonsense, frameshift and splice site mutations that introduce a premature stop codon or alter protein structure can be obvious loss of function events, missense mutations can be hard to classify unless they alter enzymatic function or disrupt protein-protein interactions within large functional protein complexes.

For example, a large number of hematological malignancies harbor translocations of the N-terminal region of MLL1 to diverse fusion partners that share very little sequence or functional similarity. To understand how these diverse MLL translocations result in leukemogenesis, biochemical and enzymological studies were essential. First, MLL and its yeast homologue SET1 were shown to be present in a complex named COMPASS (Complex of Proteins Associated with Set1) and to function as histone H3K4 methylases (267). Second, AFF4, itself a fusion partner of MLL in leukemia, was found to be a common factor among all purified MLL translocations (268). Third, ELL, one of the frequent translocation partners of MLL in leukemia, was found to function as an RNA Pol II elongation factor that increased the catalytic rate of transcription elongation by RNA Pol II by suppressing transient pausing (269). Finally, it was discovered that many MLL translocation partners are found in association with ELL and the positive transcription elongation factor (P-TEFβ), within a complex named the Super Elongation Complex (SEC) (266,270,271). The translocation of MLL into SEC is involved in the misrecruitment of SEC to MLL target genes, perturbing transcription elongation checkpoints at these loci and resulting in leukemia (271). Recent study of MLL-induced leukemogenesis highlights the role of deregulated histone methylation in tumorigenesis (272).

Another example of the importance of biochemistry is deciphering the molecular role of an observed genetic link of EZH2 in cancer. EZH2 encodes the catalytic subunit of the polycomb repressive complex 2 (PRC2) responsible for methylating lysine 27 of histone 3 (H3K27). Trimethylation at this site is associated with closed chromatin and silencing of neighboring gene expression. In neoplasia, EZH2 can influence T-cell biology (273) and function as either an oncogene or a tumor suppressor gene depending on the cellular context, e.g., EZH is sufficient to transform lung cells in transgenic mouse models overexpressing EZH (274), and loss-of-function EZH2 mutations occur in MDS and chronic myelomonocytic leukemia (CMML) (275). In germinal center diffuse large B-cell lymphomas, recurrent mutations essentially of only one codon (Y641) create a protein with reduced affinity for unmethylated H3K27 but highly increased affinity for mono-methylated H3K27, resulting in higher levels of H3K27 trimethylation overall. In contrast, pre-AML syndromes like MDS and CMML do not develop Y641 mutations but instead recurrently develop nonsense, frameshift, and other loss-of-function mutations in EZH2 resulting in low levels of H3K27 trimethylation (276). These distinctions have important clinical implications for EZH2 inhibitor development. It is possible that EZH2 inhibition will mimic malignancy-associated, loss-of-function EZH2 mutations in normal myeloid cells leading to dysregulated growth or differentiation in these cells, highlighting the need for future context-dependent studies.

SWI/SNF also known as the BAF complex is also a critical regulator of nucleosome remodeling conserved from yeast to humans. Recent biochemical investigation, combined with bioinformatic assessments have demonstrated widespread genomic alterations that occur across the members of the complex in 19.6% of all human tumors reported in 44 studies (277). In synovial sarcoma, SS18-SSX oncogenic fusion that results from a fusion of 78 amino acids of SSX to the SS18 subunit of BAF complex was shown to distrupts binding of BAF47, tumor suppressive member of the complex, leading to reversible dysregulated growth (278). In liver, genetic suppression of SWI/SNF complex member ARID1B was shown to overcome oncogene-induced scenscence and lead to liver neoplastic progression (279). While these studies suggest a newly emerging role for SWI/SNF in tumorigenesis, better delineating the role of SWI/SNF complexes in precancers will also be important. Further, prior studies demonstrate antagonistic relationship between the SWI/SNF and PRC2 complex in mediating oncogenic transformation (280).

A transformative example of biochemistry’s importance in premalignant biology involves the discovery of recurrent mutations in IDH1 and IDH2 in glioblastoma, AML, and their precursor cells. Such mutations were found through broad sequencing efforts (281) although their role at the molecular level was not clear until the advent of modern metabolomics profiling (282), which found that mutant IDH enzymes convert the normal intracellular metabolite alpha-ketoglutarate into 2-hydroxyglutarate (2-HG). 2-HG is a competitive inhibitor of a large class of dioxygenase enzymes that utilize alpha-ketoglutarate, and accumulates to very high levels in IDH-mutated cancers, potently inhibiting many important intracellular dioxygenases, including the TET family, prolyl hydroxylases, and several histone demethylases (283–285). Thus, biochemistry and metabolomics have illustrated how 2-HG contributes to carcinogenesis in a hitherto unprecedented way by acting as a novel “oncometabolite” generated by somatic IDH1/IDH2 mutations that can potentially serve as targets for both cancer prevention and therapy, including vaccines (286).

Recent biochemical approaches have also focused on the significance of metabolism and its link to epigenetic factors, such as the TET family in the regulation of cell-lineage specification and the development of cancer (188). These discoveries are only a few examples among a large number of biochemical approaches in neoplastic cancer studies and are the testimony to the power of biochemistry in understanding neoplasia and the design of its targeted prevention, for example, by highlighting the importance of epigenetic regulation. High-information-content mass spectrometry to profile global histone modifications in human cancers (287), when combined with the DNA sequencing data, can be used to identify novel variants that can drive epigenetic changes that can lead to oncogenic transformation. Chromatin-IP technology combined with NGS sequencing (CHIP-seq) can provide systematic information regarding the architecture of the chromatin cell states of cancers. Recent technological development has demonstrated that CHIP-seq can be carried out in human tissues including tumors (288). Interestingly, examination of the chromatin landscape was able to fully distinguish the normal vs. cancers. These results suggest the possibility of gaining additional insights into precancers by systematic assessment of chromatin states using key histone acetylation and methylation patterns, superenhancers, as well as TET, SWI/SNF, and PRC2 complex, all of which are critical for chromatin regulation.

The study of cancer metabolism is not only shedding light on tumorigenesis carcinogenesis but is also revealing new principles of how the biochemistry of anabolic metabolism is orchestrated to support normal cell growth and function. While most of the studies in this context have been focused on alterations in the metabolism of glucose and glutamine (see above Immune Oncology), neoplastic cells utilize many other nutrients, including sulfur–containing amino acids cysteine and methionine, essential fatty acids, choline, trace metals, and vitamins. We are only beginning to understand the extent to which these nutrients contribute to tumorigenesis. Finally, the contribution of a broad spectrum of metabolites produced by the body’s microbiota, to tumor initiation and progression is only beginning to be elucidated.

It is essential to incorporate detailed biochemical and enzymological studies on purified protein complexes to decipher the precise, context-dependent function of chromatin and other epigenetic modifiers and somatic mutations in cancer development and progression (268). This will also allow the profiles to be cross-referenced with the landscapes in primary tumors, as well as of the corresponding transcriptomic data to identify critical epigenetic changes that are necessary for malignant transformation. The context-dependent, complex roles of EZH2 mutations and PRC2 and SWI/SNF complexes in chromatin regulation in normal development and neoplasia require further study, especially in precancers. Finally, biochemical field of metabolic alterations in neoplasia continues to uncover new connections between nutrient utilization and tumorigenic state, critical to precancer progression.

Single-Cell Analyses

The natural history of precancers is heavily influenced by the heterogeneity of neoplastic cells and tissue microenvironment. Single-cell RNA or DNA sequencing technologies can be specifically leveraged to unravel the complex cellular interactions within these lesions that cannot be addressed by assaying bulk tissue (289,290). In the case of mRNA profiles, downstream analyses can characterize known populations and novel subpopulations of cells and assess how these populations change in abundance as disease progresses or regresses. These data also can be used to more accurately infer important disease-associated gene regulatory and immune cell networks (291) because the gene expression variability has not been averaged across all sampled cells as in bulk tissue. In addition, single-cell sequencing can reveal and monitor lesion heterogeneity in somatic alterations and dissect complex clonal dynamics among epithelial cells sampled at different geographic locations and over time to complement existing multi-region bulk sequencing approaches (292). These data will provide a high-resolution picture of cell types present in precancers and their surrounding microenvironment and the transcriptional programs active within each cell type that drive disease progression.

However, several technical limitations need to be overcome to realize the full potential of single-cell sequencing of precancers. These lesions are relatively small and frequently only diagnosed in formalin-fixed, paraffin-embedded (FFPE) tissues previously precluding comprehensive genome sequencing studies using current methods (293). Furthermore, information regarding the location of neoplastic cells with particular mutations within a given lesion is especially important for early lesions, as this often defines the boundary between preinvasive and invasive neoplasia. Therefore, the development and application of methods that allow assessing the genetic and phenotypic features in situ using intact FFPE tissue samples is especially critical for the improved understanding of preinvasive lesions. Several technologies enable copy number alteration and gene expression analyses at the single-cell level from FFPE slides. These include FISH and immuno-FISH (combination of FISH with immunofluorescence) (294,295), mRNA in situ hybridization, in situ PCR, and STAR-FISH (296–298). The application of immuno-FISH for the analysis of cellular phenotypic heterogeneity and genetic features revealed extensive intratumor diversity in DCIS and clear expansion of minor subclones in DCIS to dominant clones in invasive ductal carcinoma (295). A shortcoming of these methods is the limited set of markers that can be assessed on a single slide and the need for a priori knowledge of the changes to be analyzed.

Single-cell methods are beginning to be applied to premalignancy, including sequencing on both fresh/frozen and FFPE has been applied to an epithelial precancer site, DCIS and associated invasive breast cancers, and included massively parallel single-cell sequencing for copy number analysis (178). This proof-of-principle analysis established technical feasibility and demonstrated intra-lesion genetic heterogeneity at DCIS suggesting complex and distinct evolutionary processes involved in early DCIS to subclonal selection in invasive disease. Multicolor FISH to study clonal evolution at single-cell resolution in BE (see below) found extensive genetic diversity in progressors (299). A whole-exome single-cell sequencing method was developed to assess genetic heterogeneity and tested on a premalignant JAK2-negative myeloproliferative neoplasm (essential thrombocythemia) patient (300). Such profiling of a different myeloproliferative neoplasm myelofibrosis revealed substantial heterogeneity in cytokine production (301). Importantly, however, current single-cell sequencing approaches have important technological limitations. First, the methods available are labor- and cost-intensive. Second, it is currently not possible to obtain accurate detailed copy number and mutational data from the same cell, given that some whole genome amplification methods yield templates optimal for copy number analysis, whereas others are optimal for mutation profiling. Therefore, efforts are required for the development of less labor-intensive and more cost-effective methods for sequencing approaches and clonal lineage tracing, which are essential for a detailed analysis of the evolutionary paths of in situ disease and its progression to invasive cancer. Two more technologies, FISSEQ (fluorescent in situ sequencing) (290,302) and “spatial transcriptomics” (303), allow for complete transcriptome analysis of single cells in intact tissue sections. Recently, single-cell techniques have been developed to study chromatin maps/signatures and epigenetic heterogeneity in neoplasia (304,305) and the microbiome (306), including imaging of host-microbiota interactions (243,307).

Cost reduction and advances in cell sequencing (and cfDNA technology) could theoretically allow temporal monitoring of blood and epithelial premalignancies on a population scale (308–310). Periodic single-cell DNA sequencing of multiple cells from an individual will be invaluable for cancer prevention as it will allow one to assess the overall baseline accumulation of somatic mutations over time in a person, to survey and monitor multiple different endogenous processes and exogenous exposures through the use of mutational signatures, and to reveal the existence of premalignant clones and clonal evolution over time (311–313). The unprecedented resolution of sequencing single cells comes with a hefty computational and data price. Monitoring even a single individual will require multiple sequencing of one’s genome every year resulting in several terabytes of data per person. As such, population scale examinations will generate millions of whole-genome sequences resulting in exabyte scale data (>10¹⁸ bytes) that will require a new generation of computational infrastructure (314) and novel computational frameworks (e.g., to take into account their relatively low signal resolution when compared with traditional bulk tissue sequencing). More than ever, the rate-limiting step will be data analysis.

Liquid Biopsies for Early Detection and Intervention

Alterations in precancers and localized tumors may have their greatest impact in early detection of cancer. By virtue of the clonal nature of tumor cells, somatic changes are present in many copies that are continuously released and can be detected in the blood as cell-free (cf) circulating tumor DNA (315). In cancer patients, alterations in cfDNA can be detected using new sequencing and bioinformatic approaches. Such alterations can be difficult to detect as they often represent a minute fraction (<1%) of cfDNA. A variety of both targeted and whole-genome approaches have been developed to detect such alterations in cfDNA (316,317). These have been used for early detection of recurrence in colorectal cancer, pancreatic cancer, neuroblastoma, hematopoietic malignancies, and others (318–320). Of importance to early detection research, cfDNA was recently detected in plasma and urine in patients with premalignant lung and bladder disease (321,322), however, somatic mutations in cfDNA among individuals without any cancer diagnosis poses serious challenges for the development of ctDNA screening tests (323). Importantly, cfDNA detection of pancreatic cancer recurrence appears to precede radiographic detection of recurrence by over six months in some cases, providing a larger window for potential intervention in this challenging disease (319).

Much work remains to be done in improving methodologies for detection of circulating tumor DNA, both in terms of increasing the sensitivity and preventing false positives. A potential confounding issue is the detection of clonal alterations that arise in blood cells of healthy individuals that may be associated with aging and clonal hematopoiesis (see below). Distinguishing between alterations in genes associated with MDS/AML versus solid tumors may be one way to overcome this issue. Additionally, even when molecular alterations are identified, determining the tissue of origin of the incipient neoplastic lesion can be extremely difficult and complex. Combining sequencing of cfDNA with epigenetics markers (324,325), mutational signatures, imaging and mathematical modeling can be used for pinpointing the most likely tissue in which the clone originated. Analysis of other blood/fluid components, such as exosomes, platelets (326), and circulating tumor cells (315), may help increase the sensitivity of detection (327). With these approaches, it is possible to imagine a time when individuals at high risk of developing cancer due to either genetic or environmental risk factors (e.g., individuals with inherited gene mutations at risk for breast, ovary, pancreas, colorectal or other cancers, or heavy smokers or obesity at risk for several cancers) could be serially monitored using a blood-based test. A potential advantage of this approach would be the relative ease of compliance compared with other more invasive screening technologies. Ultimately, the specific methodology would be determined by practical considerations such as cost, sensitivity, specificity, and robustness of the assays, but these approaches may forever change screening for cancers that are currently incurable unless diagnosed at an early stage.

Longitudinal Analysis of Premalignancies

It is likely that synchronous precancer/cancer pair studies will not always accurately reflect the temporal clonal evolution underlying neoplastic transformation, although new analytical methods clonal/subclonal structures from NGS of a single sample. To fully appreciate such mechanisms, systematic longitudinal analyses of malignant cells will be essential. To date, such analyses of epithelial premalignancies, with the exception of BE, have been extremely limited, with reports of a relatively small number of patients with squamous lung premalignancy (328,329). DNA- and RNA-seq of these lesions have identified molecular alterations in both epithelial cell signaling pathways and immune cell pathways that associate with progression of these premalignant lesions over time. The largest longitudinal study of BE assessed genome-wide somatic chromosomal alterations (SCA) using SNP arrays over time and space. Non-progressors largely maintained stable genomes, in contrast to high levels of SCA, increased diversity, and chromosomal instability, were associated with progression to EAC (330). For some patients the transition to malignancy evolves rapidly through a genome-doubling event or chromosomal catastrophe termed chromothripsis; recently shown drive profound immune suppressivon (74,330). Another recent Barrett’s study of clonal evolution at single-cell resolution using multicolor FISH found that baseline genetic diversity predicts progression (to cancer) and remains in stable dynamic equilibrium over time, suggesting that clonal make up and evolutionary trajectory of the lesion is predetermined from the outset (299). Clonal expansions were rare, often involving p16. Importantly, this work has established the feasibility and model of such study in epithelial premalignancy.

Focusing on premalignancies of the blood has several advantages, including the ease of repeatedly acquiring neoplastic cells to study their clonal evolution, discoveries that inform single cell sequencing studies in epithelial neoplasia. For example, study of myeloproliferative syndromes (198) provides the only direct data that somatic mutation order (JAK2 and TET2) can greatly influence disease features. The overall malignant transformation rate for clonal hematopoiesis, monoclonal B-cell lymphocytosis (MBL), and monoclonal gammopathy of undetermined significance (MGUS) is about 1–2% per year, but individual risk is highly variable. Comprehensive single-cell and cfDNA omics studies will play a key role in improving our understanding of disease pathogenesis. We now have the ability to monitor hundreds of individual cells, thus overcoming bulk-cell/tissue limitations and allowing precise study of intraclonal and microenvironment architecture and crosstalk in the process and timing of transformation. The complexity and importance of follow-up is highlighted by recent findings in pancreas precancers, which can exhibit significant clonal heterogeneity/diversity that surprisingly decreases during transformation (331).

Clonal hematopoiesis was only recently characterized (332,333) by somatic mutations in genes (mutant clones of mostly single driver mutations) similar to the mutational spectrum seen in MDS (notably DNMT3A, JAK2, TET2, and ASXL1) that increased markedly with age in the general population (334–336). Nearly 40% of clonal hematopoiesis individuals with unexplained cytopenias harbor detectable mutations (197,336,337), many with clones having more than one driver mutation and higher risk of transformation to MDS/AML and all-cause mortality (197,336–338). Cooperating mutations also can be identified during periods of clonally skewed hematopoiesis in sporadic and hereditary settings that precede myeloid transformation. A recent mouse study showed that DNMT3A haploinsufficiency transforms FLT-mutant myeloproliferative disease into AML (339). Examples of hereditary mutated transcription factors that predispose to hematologic neoplasia include mutations in CEBPA, RUNX1, ETV6, and PAX5 (340). The frequent co-occurrence of germline GATA2 and somatic ASXL1 events (341) and germline SNPs associated with somatic mutations of JAK2 have uncovered several targetable cooperative mutations (342) driving premalignant progression (340). A recent GWAS identified germline variants, which predispose to both JAK2 V617F clonal hematopoiesis and myeloproliferative neoplasms (342). Four genes (JAK2, SH2B3, CHEK2, and TET2) altered in both inherited and somatic settings, contribute to V617F clonal hematopoeisis and/or MPN development. This study’s identification of a predisposition allele associated with TET2 is intriguing since somatic TET2 mutations are one of the common early events in myeloid precursors, including clonal hematopoiesis and myeloproliferative neoplasms, and can be identified in hematopoietic stem cells, either preceding or following the acquisition of V617F, and the mutational order of these two genes can influence the clinical and biologic behavior of these neoplasms (198). Most patients with clonal hematopoiesis can stably harbor small hematopoietic clones for long periods of time that expand to the point of detectability but are then held in check, possibly involving immune mechanisms (336). Clonal hematopoiesis in aplastic anemia characterized a distinct mutational pattern, including 6pUPD and PIGA, which are thought to be involved in immune escape, possibly involving loss of relevant HLA class I (343). Further, the aging microenvironment promotes outgrowth of clones driven by spliceosome mutations (SF3B1, SRSF2) in clonal hematopoiesis, associated with innate and adaptive immune attrition (344). Murine model findings suggest that aberrant splicing could produce neoantigens, which elicit an immune response. Drugs targeting the inflammasome and innate immune responses implicated in remodeling the microenvironment are potential preventive approaches.

MBL is an asymptomatic expansion of clonal B-cells in the peripheral blood present in roughly 4% of all U.S. individuals over the age of 40 (345). Genetic predisposition to MBL is suggested by the finding that the incidence of MBL is three-fold higher for individuals within familial chronic lymphocytic leukemia (CLL; at least two first-degree relatives with CLL). A large GWAS of CLL families (including 60 relatives with MBLs) as part of the replication study samples found significant germline variant associations in two out of eight regions tested (346). NGS has shown that most mutations and intraclonal heterogeneity found in CLL are already present in MBL years before progression (347,348). Furthermore, longitudinal MBL studies, including those from patients who progressed to CLL, have begun to elucidate the sequence, timing, and impact of subclonal expansion and T-cell exhaustion on malignant transformation (346,349–351). Risks of serious bacterial infections in individuals with MBL are similar to those with CLL (352–354) and linked to MBL transformation (355,356) and solid tumor risk is 3–4 fold higher in MBL and CLL (versus healthy controls), all thought to be due to defects in immune surveillance (353,354). Related work involves a hereditary syndrome of susceptibility to pre–B-cell neoplasia caused by inherited mutations of PAX5. Recent mechanistic data indicated that inherited susceptibility and aberrant immune responses to postnatal infections drives B-cell clonal evolution of premalignant B-cells and transformation to leukemia and lymphoma, by showing that pre-B-ALL was initiated in Pax5 heterozygous mice only when exposed to common infections (357). Antibody responses to primary and secondary antigen challenges are typically inefficient among patients with early stage CLL. Preliminary data have shown that immunological T-cell synapse is defective in individuals with MBL as well (345). Efforts to generate efficient vaccine responses will, therefore, be challenging. Enhancement of cytolytic T-cell function in MBL via vaccine therapy should be a long-term challenge since GVL is highly effective in eradicating leukemic B-cells. Both in vitro and in vivo data suggest that lenalidomide can repair defects in the T-cell immune synapse and reduce Tregs in CLL patients (358,359), an approach currently being tested clinically in MBL. Of note, recent mouse model data suggest that age and inflammatory status of the host microenvironment promotes selection for adaptive oncogenic events in B-cell progenitors (360), analogous to clonal hematopoesis.

Studies of MGUS have provided some of the best evidence that the immune system has the capacity to recognize precursor states (361). Search for shared targets of immune response led to the discovery that T-cells against stem-ness antigens (such as SOX2) are particularly enriched in MGUS (versus MM) (362). Prospective data demonstrate that SOX2 immunity correlate with risk of transformation (227). Prevention strategies include boosting pre-existing T-cell immunity, e.g., with SOX2 vaccines and immune modulatory drugs (363). NGS has highlighted clonal evolution and heterogeneity in longitudinal studies (364,365). Similar to clonal hematopoiesis above, MGUS cells demonstrate clinical dormancy despite NGS suggesting that the majority of genomic alterations found in MM are found in precursor gammopathies (364,365). Interestingly, new humanized models developed to grow precursor cells in vivo indicate that MGUS cells have the capacity for progressive growth, suggesting that the clinical stability/dormancy of these cells is in part mediated by features extrinsic to tumor cells, such as the immune system or bone marrow niche where signals derived from osteoblasts may be important for mediating dormancy of MGUS cells (366). The recent development of new humanized models to grow premalignant cells in vivo should greatly advance the study of clonal evolution and malignant transformation in this setting (367). Inherited genetic variation in specific SNPs increases MGUS predisposition and risk of transforming to multiple myeloma (368). Loci identified also points to a role for chronic antigen-driven stimulation in driving clonal origins and evolution in MM and other B-cell tumors. The risk of MM is particularly increased in some populations such as those with inherited lipid storage disorders such as germline GBA mutations in Gaucher disease (GD), due in part to lysolipid-induced chronic inflammation and genomic instability. GD mouse models, e.g., lysolipid substrate reduction in Gba1-deficient mice decreased the risk of gammopathies, studies already show marked preventive efficacy of targeting the underlying trigger of genomic instability (369,370). This led to the recent discovery that in nearly 25% of all cases of MGUS/MM, the underlying clone may be driven by lipid antigens such as inflammation-associated bioactive lipids (369). Studies to characterize the precise nature of microbial/dietary/endogenous lipid antigens in the setting of lipid-reactive gammopathy will facilitate targeted prevention with pharmacologic/lifestyle changes.

Summary

While a number of interventions are already approved for cancer prevention (245), this is only the tip of the iceberg. Recent paradigm changing advances include early detection research (universal tumor testing, pancreatic imaging), vaccine prevention of cervical neoplasia and combination chemoprevention in FAP. New precision prevention approaches will be needed, including novel trial designs (e.g., involving molecular selection (371,372)) and agents (e.g., denosumab for BRCA1 carriers; (34)). Developing cancer vaccines as potent as polio, diphtheria, and rubella vaccines would protect future generations from developing cancer. The elimination of cancers “before” they develop is within our grasp. As in cervical cancer, where vaccination against HPV has virtually eliminated the disease, the development of cancer vaccines to stimulate T-cells to recognize precancer antigens as foreign will prevent cancer. The interaction between the immune system and neoplasia reflects a fundamental principle that is applicable to all organ/cell types and continues to become more involved. For example, recent study revealed a striking complex interplay between chromosomal abnormalities and immunity, with mutational burden and arm- and chromosome-level SCNAs having opposing immune surveillance by showing that high level somatic copy-number alterations, consistent with a mechanism related to general gene dosage imbalance rather than the action of specific gene, creates an immune suppressive microenvironment with immune depletion/evasion that can oppose an immune response (74), even in the setting of high mutational/ neoantigen burden (e.g., MSI-H/mismatch-repair deficient tumors) (57,63). The observation that precursor conditions can exhibit similar genomic complexity to their malignant counterparts, shown in longitudinal studies of hematologic premalignancies and BE to exist years before progression, highlights the importance of microenvironment crosstalk in the process of malignant transformation. Therefore, further such studies using new technologies (e.g., single-cell approaches) are needed to better interrogate and dissect the detailed interactions of precancer cells with immune cells, and other components of the tumor microenvironment. The collaborative, focused study of nuclear, mitochondrial, epigenomic, proteomic, transcriptomic, metabolomic, and multi-omic signatures/ landscapes can be bioinformatically timed to distinguish cancer initiation and clonal diversity/heterogeneity. Organelle-targeted omics and studies of translational activity (e.g., by ribosome profiling) will also yield valuable biologic insights into mitochondrial function and dysfunction influence on oncogenesis. Establishing a PCA that integrates multi-omics and immunity (Figure 1) will be critical to develop approaches to disrupt the disabling immunosuppressive properties in the tumor microenvironment and to identify immunogenic antigens to design vaccines to activate/reprogram the immune response to detect, prevent, and reject precancers.

Specifically, we need a large-scale, systematic effort to longitudinally map the biology of the premalignancies that will bring together leaders from key disciplines (from biology and biochemistry to mathematics and engineering); leverage the NIH/NCI TCGA, GWAS, and other programs; develop/enhance statewide registries and disease specific cohorts, such as web-based registries for hereditary cancers (e.g., the PROMPT registry, www.promptstudy.info) and the recent NHLBI/ NCI National study of clonal hematopoiesis. Analogous to clonal hematopoiesis, albeit in much smaller subsets of older individuals, ultra-deep NGS of somatic cancer mutations reported a distinct mutational profile (mostly NOTCH1) in normal skin (160), and low-frequency p53 mutations in normal peritoneal fluid and peripheral blood from women without ovarian cancer (327), both likely representing a premalignant mutational background events that accumulates in cancer and aging. The NCI PCA demonstration project, which came from the Cancer Moonshot recommendations, is designed to evaluate current state of the science and provide feedback to NCI leadership for a concerted effort to comprehensively profile these premalignant lesions and provide a blueprint, including application of existing genomic, proteomic, immune, bioinformatics methods to delineate the signals and interactions within and around these lesions to create a three dimensional analysis of neoplastic progression; and development of a national database analogous to the Pan-TCGA (373,374) and pan-GWAS (77) shared databases with links to clinical annotations (1). This would provide an immense national resource for discovery and shed new light on shared precancer biology and new targets for interception.

The PCA project will require expansion of a diverse array of companion technologies and preclinical models in which to study (88,375), develop, and implement strategies for cancer prevention and interception. New technologies (e.g., single-cell sequencing, liquid biopsy) and preclinical models, e.g., GEMMs such as autochthonous mouse models of cancer to dissect the in vivo contribution of stroma-specific factors to tumor progression; CRISPR/Cas9-engineered, immunosuppressive mouse strains (376,377); methods to isolate and inducibly engineer primary premalignant/benign human cells to study very early oncogenesis (216); clinically relevant organoids (to study T-cell immune surveillance) and sequential, inducible activation of cancer-causing driver genes (and associated neoantigens) in a cell lineage-specific manner, e.g., in germline (MSH2-deficient) and sporadic intestinal neoplasia (378–382), will allow study of complex interactions between genetically susceptible host, microenvironment, and microbiota (246,383).

In summary, to fully achieve cancer prevention, we must build teams with multiple areas of expertise from the NIH, academia, Food and Drug Administration, private foundations, philanthropic partners, and industry. The best analogy of assembling such a multi-disciplinary team is the Manhattan Project – one goal, multiple experts.