Germline variants impact on somatic events during tumorigenesis

Abstract

Cancer is characterized by diverse genetic alterations in both germline and somatic genomes that disrupt normal biology and providea selective advantage to cells during tumorigenesis. Germline and somatic genomes have been extensively studied independently, leading to numerous biological insights. Analyses integrating data from both genomes have identified genetic variants impacting somatic events in tumors, including hotspot driver mutations. Interactions among specific germline variants and somatic eventsinfluence cancer subtypes, treatment response, and clinical outcomes. Investigation of these complex interactions is increasing our understanding of aberrant pathwaysin tumorsthat may uncover novel therapeutic targets. Here, we review the literature describing the role germline genetic variantsplay in promoting selection of and generation of specific mutations during tumorigenesis.

Germline Variants Impact Tumor Characteristics

Using next-generation sequencing, the germline and tumor genomes have been explored independently for mutations that are associated with tumorigenesis, often with the germline genome serving as a reference control for acquired mutations. Analyses integrating data from both genomes are leading to novelassociations between germline variantsand specific somatic events. Recent investigation of the complex interplay between germline variants and cancer-associated mutations is increasingour understanding of aberrant pathways that may be potential therapeutic targets. This review explores the hypothesisthat specific germline variants impact which somatic events and mutations are generated and selected for in cancer cells during tumorigenesis(Figure 1). Here we discuss the body of literature supporting a role for germline polymorphic and pathogenic variants(PVs)in shaping a number of key tumor characteristics. Specifically, this review will address the emerging link between germline variants by somatic mutation (GxM) associations. We focus on tumor characteristics such as: histopathological subtypes, mutational signaturesand microsatellite instability (MSI), allele-specific copy number changesand loss-of-heterozygosity (LOH),somatic mutationsin cisand in transwith the relevant genetic variant, and the tumor immune response

Figure 1: Examples of germline variants by somatic mutation (GxM) association.

Specific germline polymorphicor pathogenic variants impact which somatic events, mutational signatures, allele-specific copy number changes oraltered signaling pathway, are selected for in cancer cellsduring tumor evolution. In the figure, the outer circle shows germline variants and the inner wedges represent the associated somatic events for that specific example of a GxM association. BRCA1PVs are associated with a mutational signature characterized by small tandem duplications and deletions. MDB4PVs are associated with mutational signatures, particularly C>T substitution in CpG dinucleotides, and APOBECgermline variants are associated with the mutational signature C>T or C>G substitution within TCA and TCT motifs. TP53PVs are associated with chromothripsis. More generally, germline PVs and SNVs are associated with preferential gain of the risk/oncogenic allele and/or loss of the non-risk/tumor suppressing allele. Specific HLAhaplotypes have also been linked to higher and lower frequencies of specific pathogenic variants as they result in the immune system being more or less likely to recognize specific somatic mutations.

Germline Variants and Histopathological Subtypes

Tumors from individuals with hereditary cancer syndromesoften exhibit characteristic histopathological featuresat different frequencies than tumors arising in the general population. For example, breast cancers occurring in individuals with BRCA1or BRCA2germline PVs are associated with different frequencies ofcancer subtype, grade, and estrogen receptor (ER) statuscompared to their germline negative counterparts[1]. Specifically, basal-likesubtypeand triple-negative grade 3breast cancers are highly associated with BRCA1PVs. Multiple studies have found associationsof lower-penetrancecommon single nucleotide variants(SNVs) with specific breast cancer subtypes, particularly ER-positive and ER-negativesubtypes[2–7]. The functionsof many of these variants, the majority mapping to intergenic regions, remain unknown. A few germline variants associated with ER status map to the ESR1gene and lead to differential ESR1expression, providing a mechanistic link between the variant and tumor phenotype [8]. Similarly, PVs in renal cell carcinoma(RCC)genes are associated with specific histological subtypes[9]. Histopathological subtypes often correlate with clinical outcomes. Therefore, knowledge of the associated germline SNVs and PVs can contribute to better predictive power for survival and therapeutic response. For example, screening for germline BRCA1/2PVs can predictrisk for ER-negative (triple-negative and basal-like) tumors, which are more aggressive and have poorer prognoses compared to ER-positive (luminal-A and luminal-B) subtypes [10].

Germline Variants and Tumor Mutational Signature

Germline variants are associated with mutation patterns in tumors. Breast tumors from individuals with germline BRCA1and BRCA2PVs havea higher frequency ofsmall tandem duplications at regions of microhomology and deletions <100kb [11](Figure 1). These signatures likelyreflect BRCA1and BRCA2function in homology directedDNA repair and suggest PARP inhibitors will be beneficial to this patient population[12].

A study that systematically evaluatedthe association of germline variants withsomatic mutation signaturesin 2642 individuals across 39cancer typesfound:(1) individuals with PVsshowed enrichment of C to T substitutionsat CpG dinucleotides(Figure 1),(2) BRCA1-deficient tumors wereassociated with complex rearrangements including insertions and tandem duplications, (3)germline L1/LINE elements wereassociated with somatic retrotransposition, and (4) germline TP53PVswereassociated with chromothripsis[13](Figure 1). Thisstudy also highlighted the relevance of the APOBEC family of cytidine deaminases in cancer. Tumors from individuals with germline variants near APOBEC3had reduced levels of APOBEC mutationalsignatures(Figure 1), characterized by substitutions of C with either T or Gwithin TCA or TCT motifs [14–16]. Specifically, rs2395185was associated with APOBECmutation signaturesin lung cancer[14]and rs1014971 in bladder tumors [15](Table 2). APOBEC signatures in breast tumors also associate with rs12628403, a tagging SNV for a polymorphic deletionresultingin a fusion of APOBEC3Awith non-coding regions of APOBEC3B[13, 15, 16]

Table 2:

Germline variants by somatic mutation associations in trans

Cancer type	SNP ID/ aGenomic Location	Gene/Pathway with somatic mutation/ bGene Locus	Impact	Reference
Glioblastoma	rs13222385 chr7:55183900	LANCL2 7p11.2	p=0.037	[42]
Multiple	rs8051518 chr16:7270514	SF3B1 2q33.1	8-fold increase incidence	[49]
	rs25673 chrl9:2109158	PTEN 10q23.31	4-fold increase incidence
Breast	rs252913 chr5:56900019	PIK3CA 3q26.32	p=0.01	[51]
	rs331499 chr5:56915096		p=0.02
Lung	rs36600 chr22:29941597	ARID1A 1p36.11	p=5.78×10−4	[14]
	rs2395185 chr6:32465390	Cell cycle pathway genes	p=3.61×10−4
	rs2395185 chr6:32465390	APOBEC Genome-wide APOBEC mutational signature	p=3.58×10−3
	rs3817963 chr6:32400310	Cell cycle pathway genes	p=4.15×10−4
	rs3817963 chr6:32400310	MAPK pathway genes Ex. MAPK1 22q11.22	p=8.58×10−4
Bladder	Rs1014971 chr22:38936618	APOBEC Genome-wide APOBEC mutational signature	p=1.92×10−5	[15]
Gastric	rs2285947 chr7:21544470	PDGF pathway Ex. PDGFB 22q13.1	p=3.93×10−4	[50]
	Rs1679709 chr6:28260564	RCF4 mismatch repair gene 3q27.3	p=1.25×10−2
Lung (mouse)	rsl3459194 cchrl9:32049010	Kras Q61L cchr6:77.37 cM	p=1.6×10−5	[29]
Melanoma	rs12203592 chr6:396321	BRAF V600E 7q34	OR 0.59; 95% CI=0.43–0.79	[53]
		BRAF V600K 7q34	OR 0.65; 95% CI=0.41–1.03
		BRAF (other in exon 15) 7q34	OR 1.57; 95% CI=0.93–2.65
		NRAS 1p13.2	OR 0.99; 95% CI=0.75–1.30
	rs132985 chr22:38167464	BRAF V600E 7q34	OR 1.32; 95% CI=1.05–1.67
		BRAF (other in exon 15) 7q34	OR 1.82; 95% Cl=1.11–2.98
		BRAF V600K 7q34	OR 1.12; 95% CI=0.78–1.60
		NRAS 1p13.2	OR 0.83; 95% CI=0.64–1.07
Renal cell carcinoma	rs 1093 2384 chr2:211545123	ERBB4 2q34	p=0.003	[46]

^aGenomic locationsobtained from NCBI dbSNP, Human assembly GRCh38/hg38

^bSomatic mutation locations obtained from NIH, Genetic Home Reference

^cMouse genomic location obtained from UCSC Genome Browser and somatic mutation location obtained from NCBI, GRCm38/mm10

OR, odds ratio; CI, Confidence interval; NSCLC, non-small cell lung cancer.

The occurrence of MSIin the tumor genome is also influenced by germline PVs. MSI is observed at a high frequency in colorectal cancer (CRC) with germline or somatic PVs in mismatch repair genes [17–19]. Tumors with high MSI are more likely to be of low pathological stage, more likely to carry activating BRAFmutations, and less likely to carry activating KRASmutations compared to tumors that are microsatellite stable and are more likely of low pathological stage [20]. Compared to microsatellite stable CRC, MSI is predictive of favorable outcome and reduced likelihood of metastases [21]

Collectively, these studies demonstrate that germline variants in genes involved in DNA repair or maintenance ofgenome integrity can result in characteristic mutational patterns reflective of the perturbed function. Additionally, these findingsmay uncover novel strategies for prevention of hypermutation or failed DNA repair, and, asin the case of MSI-high CRC, can inform clinical outcomes.

Germline Variants and Somatic Copy Number Changes or Mutations in cis

In addition to global influences on tumor subtype and mutational signatures, germline variants are also associated with specific somatic events “in cis”; theseincludesuch phenomena as LOH, allele-specific imbalance (ASI), and somatic mutation occurring at the same locus as the germline variant. A classic example of the connection between germline variants and somatic events in tumor development is illustrated by Alfred Knudson’s 1971 observation, today known as the “two-hit hypothesis” [22](Box 1). Knudson showed that retinoblastoma occurs much earlier in individuals who inherited a germline PV in the RBgenecompared to individuals in the general population. Individuals who inherit a PV in every cell in the retina only need one cell to acquire a “hit” in the remaining wild-type allele for selective growth advantage leading to cancer, whereas retinal cells from individuals in the general population must acquire two independent events within the same cell to become at-risk of developing a tumor. Tumors arising in the context of hereditary cancer syndromes, like retinoblastoma, frequently show evidence of LOH of the wild-type allele because of the selective advantage that these tumor cells have over cells retaining one wild-type allele.

Box 1.

Knudson’s two-hit hypothesis, also referred to as the two-mutation hypothesis, was proposed in 1971 by Alfred Knudson from his work on the genetic mechanism behind retinoblastoma. In hisstudy of retinoblastoma patients, data such as age at diagnosis, gender, family history, unilateral or bilateral disease occurrence, and the approximate number of tumors in each eye werecollected and analyzed. At the time of Knudson’s study, it was believed that retinoblastoma could becaused by either somatic or germline mutations. Knudson’s findingsshowedthat retinoblastoma was caused by a germline mutation in a subset of cases, but that in order for retinoblastoma to occur, a subsequent inactivating mutation in the wild-type allele was required. Hereditary cases, in which individuals inherited one copy of a mutated RB gene, were frequently bilateral and occurred at a younger age compared to non-hereditary cases. Knudson concluded that retinoblastoma was caused by two mutations (one in each copy of the gene), hence the two-hit hypothesis. Individuals with no inherited mutation have to somatically accumulate a loss-of-function mutationin each allele for the disease to manifest, and this process takeslonger compared to individuals who already haveoneinherited mutation. Knudson’s work illustrates the concept of allele-specific somatic events: in order for an individual with an inherited pathogenic variant in the RBgene to develop retinoblastoma a somatic loss-of-function mutation in the wild-type allele is required.

In tumors, loci harboring oncogenesoften show copy number gains.Tumorsuppressor geneloci are more typically lostthrough somatic mutations, chromosomal rearrangements or deletions. Copy neutral alteration(in which the DNA is neither lost nor gained)and LOH of the wild-type (non-mutated) alleleare frequent mutations in tumor suppressor genes (Figure 2A). This phenomenon isreferred to asASI(Box 2 and Figure 1). In contrast, activating mutant alleles of oncogenessuch as EGFR, KRAS, PIK3CAand BRAFarepreferentially amplified over non-activated alleles(Figure 2A), thereby increasing the selective advantage of these cellsby promoting proliferation and survival [23]

Figure 2: Examples of germline alleles conferring selective advantage to specific somatic alterations.

(A) ASI incis(affecting the same gene), can arise in at least three ways (from left to right): by amplification or gain of the risk allele (“G”), by copy neutral loss of the resistance allele (“T”), or by LOH due to loss of the resistance allele (“T”). ASI can increase the selective advantage of these cells (in pink) via the risk allele promoting proliferation and survival pathways leading to preferential tumor growth of cells with gain of the risk allele over cells that have gain of the protective allele (in blue).(B) A germline genetic variant (“a”) may associate with a mutation in an unlinked gene (“b”). However, when a somatic mutation occurs in cells with the non-associated allele (“A”) there is no selective advantage to the cells. When a somatic mutation occurs in cells with the associated allele (“a”), there is dysregulation of normal pathways and/or functions, resulting in selective advantage and expansion of those cells (in pink).

Box 2.

Allele-Specific Imbalance (ASI).

This is the term used to describe non-random genetic copy number alterations of one allele relative to another. Typically, this is associated with a relative gain of a highly penetrant oncogenic allele or relative loss of an allele associated with suppression of tumorigenesis. Studies have also found ASI for lower-penetrance alleles associated with increased cancer risk. Quantitative genotyping and copy number analysis are used to identify these events. Normal and tumor DNA from individuals heterozygous for a particular SNV or PV of interest are sequenced, and an ASI score is determined by comparing the allele peak heights between normal and tumor DNA sequencing graphs. In retinoblastoma, the classic tumor suppressor gene RB shows ASI, asthe wild-typeor resistanceallele ofatumor suppressor geneis lost and the activating mutated or susceptibility allele is retained and selected for gain or amplification. These allele-specific copy number alterations have an impact on tumorigenesis, progression, and metastasis. Activating mutations in oncogenes increase the selective advantage of cells by affecting proliferation and survival. While activating somatic mutation in one allele of an oncogene is sufficient for selective growth advantage, malignancies can be promoted by copy number gain, copy neutral alteration and loss of heterozygosity events enhance malignancies. With advances in genome-wide analyses of the tumor genome, ASI is being identified in regions of the genome showing copy number gain, copy neutral alteration, and loss of heterozygosity

Observation of either preferential gain of the mutant allele of anoncogene or LOH ofa tumor suppressor gene provides strong evidence for the association of inherited germline variants with specific somatic events in tumors. For example, allele-specificgain of the KRASallele containing a G12D mutation has been observed in 13–55% of CRCcases [24]and is associated with poor overall survival [25]. However, ASI is not restricted to somatic events with large functional impact. Earlyevidence suggesting that loci harboring low-penetrantcancer susceptibility(risk)and resistance(protective)allelescanalso undergo selective copy number changes came from mouse modelsin which cancer-associatedloci were identified by linkage[26–29]. In a mouse model for chemically-induced skin cancer[26]about 40% of previously identified skin tumor-susceptibility (Skts) loci[30]showed evidence of ASIin F1 backcross mice withrelative preferential gain of the allele from the susceptible strain orrelative preferential loss of the allele from the resistant strain. Notably, the oncogeneHrashad activating mutations in a majority of the skin tumors, and 65% ofheterozygoustumors showed preferential gain of the chromosome carrying themutant allelefrom the susceptible strain[26].

Genome-wide association studies (GWAS)have identified a number of haplotype-tagging SNVs thatassociate with cancer risk andsomatic mutation frequency (Box 3). One of the first GWASvariantstested forASIwas rs6983267, whichmaps to an intergenic region on8q24.21and is associated with increased risk of colorectal and prostate cancers[31–33]. Whenrs6983267 was genotyped in colon tumor DNA from 466patients heterozygousfor thisSNV, 101 tumor DNAs exhibited ASI, with66% showing relativegainof the risk (G)allele and 34% showing relative gainof the non-risk (T)allele[34](Table 1). Similar findingswere observed for rs6983267in84 heterozygous cell lines from various malignancies[35]andin an expanded ASI studyof16 CRC GWAS variantsin 490totalpaired normal and tumor samples[28](Table 1). However, ASIdoesnot occur at every risk-associated locus, which isconsistent with previous mouse studies[26]. Functional studies of rs6983267 found thatgain ofthe risk allele associates with higher MYC expressionandactivity, thereby promotingCRC development [35]. Importantly, many variants identified by GWAS are intergenic and can be difficult to functionally assess. Evaluatingvariants for ASI offers one strategy for prioritizing risk alleles for functional evaluation.

Box 3.

Genome-wide Association Studies(GWAS) are association studies investigating the correlation between genetic variants and phenotype or disease. When an allele is observed more frequently in the genomes of individuals with a disease or phenotype (cases)compared to individuals without the disease or phenotype (controls), the variant is classified as associated with that disease. Hundreds of GWAS for cancer have been conducted and have identified variants that are associated with increased risk of cancer, as well as other phenotypes such as tumor subtype, tumor aggressiveness, propensity to metastasize, and response to therapy. Because the effect size of most of these variants is small, individually they do not have strong predictive value for the phenotype. However, the predictive value increases when variants are combined into a risk score of multiple variants, or a polygenic risk score. Data from GWAS are leading to a better understanding of the biology and mechanisms of disease that may lead to prevention or more effective personalized risk assessments and precision treatment options. Germline variant by somatic mutation (GxM) associationstudies, which identify associations betweenparticular germline variantsandan increased likelihood of a specific somatic mutation in an individual’s tumor, area new area of genomic exploration. GxM associations may lead to the identification of the context of the cells that is important for selection during tumor development and uncover novel pathways. Such associations may be utilized to predictprognosis, to predict therapeutic response and clinical outcome, and to identify specific pathways that can be targeted for therapeutic intervention.

Table1:

Cancer susceptibility variants showing allele-specific imbalance in cis

SNV ID/ aGenomic Location	Cancer	Number of heterozygotes with evidence of copy number imbalance	Risk allele (n) percentage	Non-risk allele (n) percentage	bp-value	Reference
rs6983267 chr8:127401060	CRC	101	G, Loss (34/101) 34%	T, Loss (67/101) 66%	p=0.0007	[34]
rs6983267 chr8:127401060	CRC (cell lines)	84	G, Loss (33/84) 39%	T, Loss (51/84) 61%	p=0.05	[35]
rs6983267 chr8:127401060	CRC	48	G, Loss (15/48) 31%	T, Loss (33/48) 69%	p=0.03	[28]
rs2273535 chr20:56386485	CRC	23	A, Gain (19/23) 83%	T, Gain (4/23) 17%	p=0.018	[36]
rs2273535 chr20:56386485	CRC	54	A, Gain (38/54) 70%	T, Gain (16/54) 30%	p=0.03	[41]
rs13281615 chr8:127343372	cSCC	35	A, Gain (28/35) 80%	G, Gain (7/35) 20%	p=0.012	[43]
rs6959338 chr7:112077957	Glioblastoma	41	T, Gain (33/41) 80%	C, Gain (8/41) 20%	p=1.1×10−4	[42]
rs13222385 chr7:55183900		45	G, Gain (35/45) 78%	A, Gain (10/45) 22%	p=2.5×10−4
rs4367471 chr7:104742426		24	Minorc, Gain (20/24) 83%	Majorc, Gain (4/24) 17%	p=0.0015
rs4132013 chr7:104762262		32	Minorc, Gain (27/32) 84%	Majorc Gain (5/32) 16%	p=0.00011
rs12343867 chr9:5074189	Myeloproliferative neoplasm	109	C, Gain (93/109) 85%	T, Gain (16/109) 15%	p=5.7×10−6	[44]

^aGenomic locationsobtained from NCBI dbSNP, Human assembly GRCh38/hg38

^bIn all cases, the risk allele showed relative preferential gain or the non-risk allele showed relative preferential loss.

^cMinor and Major refer to the minor and major allele frequency in Caucasian populations, respectively.

Whilers6983267maps to a putative long-range enhancer for the MYC oncogene, variants altering protein-coding sequence also exhibit ASI. One example is the variant rs2273535(Phe31Ile)in AURKA, whichwas previouslyidentifiedas a candidate tumor-susceptibility allele for multiple cancer types using data from mouse models, human tumors, and case-control association data[36–40]. In one study, among 48 heterozygous CRC samples with copy number changes at 20q13.2, gainof therisk (A/Ile)allele was favoredover the non-risk (T/Phe)allele (Table 1) and was associated with higher likelihood of aneuploidy [36]. An independent study found ASI of the rs2273535(A/Ile) allele using125 familial and 110 sporadic CRCcases, with familial heterozygous tumors exhibiting preferential gain of the A allele relative to the Tallele[41](Table 1).

Genomic analyses of glioblastoma have also identified selective amplificationof germline variants associated with increased cancer risk. Researchers analyzed 178 glioblastoma tumors from The Cancer Genome Atlas (TCGA) and foundselectively amplifiedSNVsin kinase-encoding genes including AGK, DGKB, EGFR, INSR, KIT andRELN[42]. This groupalso compared their list of 139 amplified SNVs with 406 SNVs identified in a glioblastoma GWAS, with the rationale that a selectively amplified SNV in glioblastoma may predispose the carrier to tumor initiation and therefore occur at higher frequency in cases versus controls. Two SNVs, rs4367471and rs4132013, demonstrated amplification of the risk alleles over the non-risk alleles in the glioblastoma tumors of germline heterozygotes (Table 1). Two SNVs, rs6959338in DOCK4at 7q31.1 and rs13222385in EGFRat 7p11.2, also showed preferential amplification of the risk over the non-risk allele in the tumor DNA of germline heterozygotes (Table 1). Samples with amplification of the selected-for allele had higher gene expression of DOCK4and EGFR.

Interestingly, LANCL2, located near EGFR, also exhibited higher expression in rs13222385heterozygotes with amplified risk allele (Table 2). Thus, for some variants preferential allelic gain in tumors may reflect selection based on their role in regulation of gene expression. This study illustrates the power of having multiple data sets (germline SNV, tumor exomes, and tumor gene expression) for making biological connections.

While the aforementioned studies were inherently limited to testing single tumors from individuals, researchers have utilized solid organ transplant recipients to better address the role of germline DNA on genetic alterations. Solid organ transplant recipients are distinctive in that some will develop multiple independent cutaneous squamous cell carcinomas(cSCCs)in the context of an unchanging constitutional genome. This provides a uniquely powerfulopportunity to test the hypothesis that an individual’s germline DNA influences the types of genetic alterations that promote cancer initiation and growth. Copy number alterations from individuals with multiple cSCCs revealed higher concordance of chromosomal aberrations within an individual compared to across unrelated individuals, andrs13281615 showed evidence of ASI in thiscohort [43](Table 1). Thus, genetic background affects the pattern of specific somatic alterations that predisposes individuals to cSCC initiation or progression.

In addition to SNVs being linked to copy number alterations in cis, associations between SNVs and somatic mutations within the genes harboring the variants have been detected. A GWAS for myeloproliferative neoplasms associated rs12343867in JAK2 with JAK2^V617F somatic mutations. Among individuals heterozygous for rs12343867, a significant proportion carried the risk (C) allele for JAK2^V617F mutation [44](Table 1). Similar findings in anon-small cell lung cancer (NSCLC) study demonstrated that germline variants in EGFRwere associated with somatic EGFRmutations, particularly exon 19 deletions, and that these showed evidence of ASI [45]. In a case-control study of 141 NSCLC patients, two SNVsinEGFR, rs45559542and rs712829, wereobserved at a higher frequency in patients with exon 19 deletions. A similar observation was seen inrs712829heterozygouscell lines. These data demonstrate that variants can contribute to ASI for oncogenes such as EGFR. As another example, analysis ofTCGA data for127 significantly mutated genes across major cancers determined that SNVsmay be predictive of RCCrisk and predictors of clinical outcomes. Using data from over 650 individuals with RCC, an association was observed between the risk (C) allele of the ERBB4intronic variant rs10932384and ERBB4mutation [46](Table 2). The same variant was also associated with recurrence and overall survival, suggesting that GxMassociationsmaypredict recurrence and clinical outcomes.

Germline Variant by Somatic Mutation (GxM) Associationin trans

While GWAS have identified hundreds of cancer risk variants[47], fewstudies have identifiedGxM associationslinking germline variants with distal somatic mutations [48] (Figure 2B). Studies conducting exome sequencing of tumor tissue and genome-wide SNV genotyping of the germline DNA from the same individual can be used to identify some GxM associations but miss intergenic SNVs where the majority ofrisk alleles are located. These GxM associations refer to trans effects of germline variants, inwhich the variants influence somatic mutations at distant, and even seemingly unrelated, loci. Significantly, GxM studies have great potential to illuminate unknown biological connections tethering the germline genome to the tumor genome.

One of the firstGWAS to systematically test forassociations between germline variants and somatic events in human tumors used TCGA data from nearly 6000cancer cases[49]. Thehypothesiswas that germline background establishes a context in which a loss-or gain-of-function event in a particular gene may confer a selective growth advantage. GxM analysesof 138 frequently mutated cancer genes identified 62 associations. Validation studies revealed 28 germline loci that were associated with increased somatic alterations in 20 cancer-related genes. For example, the intronic variant rs8051518 in RBFOX1(mapping to 16p13.3)was associated with eight-fold increased incidence of somatic mutations in SF3B1(2q33.1)(Table 2). SF3B1and RBFOX1 encode for RNA-binding proteins involved in splicing, suggesting a biological rationale for the observed association. Another GxM interactionwas observed between rs25673(19p13.13)and PTEN(10q23.31); individuals with the risk allele of rs25673were four-fold more likely to have PTENmutations (Table 2). Two genes at the 19p13.13locus, GNA11and STK11, are involved in the PIK3CA/mTOR pathway for which PTEN is a negative regulator.

Therefore,rs25673could increase the selective advantage of inactivating PTENmutations in cancer progression.

Other studies have identified GxM interactionsthat provide evidence for risk SNVs influencing somatic events in tumorigenic signaling pathways. In lung cancer, SNVsassociatewith somatic driver gene mutations or copy number alterations in ARID1A, CDKN2Aand genes linked to the cell cycle and MAPK pathways[14](Table 2). In gastric cancer, several germline SNVs associatewith somatic alterations in SOS1and other genes in the PDGF and DNA mismatch repair pathways [50](Table 2). In a cohort of ER-positive breast cancer patients, GxM analysis found associations of two SNVs, rs252913and rs331499(5q11.2)withsomatic PIK3CAvariants(3q26.32)[51](Table 2), which results in MAP3K1gene overexpression. While PIK3CA mutations are frequent in breast cancer, clinical trials of agents targeting this mutationhave produced disappointing results[52]. Future studies may capitalize on new insights highlighted by GxM interactions to identify altered signaling pathways for which precision therapieswill be effective

While much of the literaturehas described how germline variants are associated with specificgenes in the mutation landscape of cancer, some variants are associated with very specific mutations in these genes. Amouse lung cancer susceptibility locus, Pas1, which maps nearKras, showed ASIof the chromosome from the tumor-susceptible strain[27]. A recent studytested the associationof germline variants with the frequency and type of Krasmutation in mouse lung tumorigenesisusing two chemically-inducible lung tumor mouse intercross models. In both models, Krascodon 61 was the most frequently mutatedcodon, and somatic mutations wereassociatedwithgermline markers atthesame chromosome 19 locus. However, two Q61Rwere associated with different SNVs different Krascodon 61 mutations:Kras^Q61L withrs3655407and Kras^Q61L with rs13483612. Fine-mapping studies found that Kras^Q61L was associated with rs13459194in mice homozygous for the allele from the susceptible parent, but this was not observed for Kras^Q61R mutation [29](Table 2). Thesestudiesillustrate that different somatic mutationswithin the same gene and evenaffecting the same codoncan occur in the context ofspecific germline variants.

Cutaneous melanomas are categorized byfourgenomicsubtypes: mutant BRAF, mutant RAS(mainly NRAS), mutant NF1, and triple-wild-type. In one study, germline DNAfrom 1,223 participantswas genotyped for 47 SNVs previously associated with melanoma risk[53]. Melanomas were also interrogated for BRAFand NRASmutations. GxM analyses found two SNVssignificantlyassociated with BRAFmutations. The rs12203592(T)alleleof IRF4was associated with decreased risk of BRAF^V600E and BRAF^V600K mutations, but an increased likelihood ofother BRAFexon 15 mutations (Table 2). The rs132985(T) allele of PLA2G6was associated with increased likelihood of BRAF^V600E and other BRAFexon 15 mutations(Table 2)[53]

Germline MC1Rvariantsalso influence the somaticmutational landscape of melanoma. These variants are associated with red hair, freckling, and sun sensitivity. Melanomas from individuals with MC1Rvariants have a higher somatic mutational burden than individuals without MC1Rvariants.[54] Furthermore, BRAF^V600K somatic variants wereless likely to occur in individuals withMC1R pigment-related germlinevariants(D84E, R142H, R151C, R160W and D294H). These same MC1Rvariants were associated with the presence of BRAF^V600E PVs, but only in individuals with darker eye and hair color[55]. Collectively these data suggest that pigment phenotypes, or genesdetermining these phenotypes, are associatednot only with what genes are somatically mutated but with precise mutation (BRAF^V600K the versus BRAF^V600E).

Germline Variant by Somatic Mutation Associations and Tumor Immune Response

There are emerging connections between germline variants in immune-related genes and somatic mutations in tumors. Correlations among MHC-I(HLA)genotypes and a subset of 1000 common driver mutations for cancer were testedin 9000 cancer patients from TCGA. This studyfound that patient-specific HLAvariants are associated with which oncogenic mutationspass undetected by the immune system and allow for clonal selection. This powerful studyrevealed that an individual’s HLAgenotype may predict which drivermutations are likely to occur in the tumors of specific patients [56]. In general, across individuals with assorted HLAvariations, BRAF^V600E mutation is associated with weak immune presentation and higher population mutation frequency while IDH1^R132C mutation is associated with strong immune presentation and lower population mutation frequency. Other. somatic mutations such as PIK3CA^E545K,PIK3CA^H1047R, KRAS^G12D and KRAS^G12V and are more frequent in individuals with specific MHC-Igenotypes(Figure 1). Patient-specific MHC-IIgenotypesinfluence the somatic mutational landscape and anti-tumor responsesimilar to MHC-I [57]. In another GxM study, rs351855nearFGFR4was associated with increased STAT3 signaling, poor prognosis, and increased progression in multiple cancers [58]. Knock-in mice with thers351855risk allele showed suppressed CD8/CD4 regulatory T-cell ratio, decreased tumor infiltration of CD8 T-cells, and increased STAT3 signalinginvivo. With a better understanding of the role of the innate immune system in recognition of specific driver mutations, we may be able to predict which somatic mutations an individual’s tumor is more likely to contain. Collectively, these data suggest a link between germline variantsandimmune evasionorsuppressed anti-tumor response during cancer progression.

Concluding Remarksand Future Directions

Integrating germlineand somatic tumor genome data providesinsight into pathways and molecular mechanisms important for tumorigenesis. Just as tumors develop resistance mutations or have clonal expansions of rare cell populations during therapeutic intervention, an individual’s germline genetic background can lead to particular mutational profilesand/or may impart an earlyselective pressure fostering environments in which the particular somatic events may be more likely to expand or escape normal cell controls. Due to the paucity of studies integrating the germline and somatic genomes, and considering that this area of research is still in its infancy, many such associationsand mechanisms of these associationsremain undiscovered. It also remainsundetermined if findings from GxM studies will inform individual response to therapiesbeyond PVssuch as BRCA1and MLH2. However, given the connections described here, this research direction shows promise. Future studies evaluatingGxM associationsmayalsolead topredictionof tumor subtype or prognosis even before an individualis diagnosed with cancer, which could additionally inform prevention strategies(Figure 3)

Figure 3: Summary of GxM association studies and clinical application.

Integration of data from exome sequencing of tumor DNA and genome-wide SNV genotyping of germline DNAis revealing associations between specific germline variants and somatic events that influence mutational profilesand cancer subtypes. This information may predict treatment response, clinical outcomes, and identification of novel therapeutic pathways. For example, BRCA1PVs are associated with more aggressive breast cancers (triple-negative and basal-like) and poorer prognosis. Tumors with PVs in BRCA1are more likely to contain small tandem duplications and deletions reflecting perturbation of homology directed DNA repair function. Tumors with disrupted DNA repair are more responsive to chemotherapeutic agents that block other means to repair DNAdamage, such as PARP inhibitors.

Outstanding Questions.

Are germline variant by somatic mutation interactions cancer type specific or common across multiple tumor types?

Can identification of GxM interactions aid in distinguishing driver somatic mutations from passenger mutations?

Will characterization of the mechanisms leading to specificGxM interactions provide insight into newtherapeutic strategies?

Are histological and pathological subtypes of tumors largely driven by the genetic background of an individual?

EGFRmutations are muchmore common in lung tumors from individuals of Asian ancestry than in lung tumors from individuals of European ancestry. Are differences in the frequency of somatic mutations observed between racial and ethnic groups associated with alleles that show differences in germline variant frequency between these groups?

Highlights.

The genetic context in which a somatic mutation occurs can impact whether it is likely to be selected for during tumor development.

Germline pathogenic variants in highly-penetrant cancer susceptibility genes are associated with specific tumorsubtypes as well as somatic mutations in specific genes and pathways.

A subset of cancer susceptibility alleles identified through genome-wide association studies showsallele-specific copy number gains and lossesin tumors.

Genome-wide association studies have identified genetic variants associated with specific somatic events in cancer, highlighting new biological connections.

Germline variants in immune system genessuch as MHC class 1 genesenable cells withsomatic mutations at specific amino acid residues to evade the immune system.

Glossary

Allele

Alternative forms of DNA located at the same genetic locus on a chromosome, also known as a variant. Diploid organisms have two alleles at each genetic locus, with one allele being inherited from each parent.

Allele-Specific Imbalance (ASI)

Somatic DNA copy number alterations in which one allele shows preferential copy number changes (loss or gain) compared to the other allele.

Germline Variants

Variationsin DNA sequence transmitted from parent to offspring via the sperm or egg. These variants are in all cells of the offspring and can be transmitted to future generations.

Genome-Wide Association Studies (GWAS)

Association studies for a disease (or phenotype)disease in which genetic variants across the entire genome are tested for association with disease in individuals with the disease (cases) and individuals without the disease (controls).

Germline Variant by Somatic Mutation (GxM) Association:

When a germline variant is associated with an increased likelihood that a specific somatic mutation will be present in a tumor.

Linkage

The tendency for two or more genes located close together on the same chromosome to be inherited together. Linkage study refers to a family-based method used to map a trait to a genomic location by demonstrating co-segregation of the disease with genetic markers of known chromosomal location.

Loss-of-Heterozygosity (LOH)

A genetic event whereby one of two different alleles at a locus is lost. When LOH occurs in tumors, the genome is homozygous at that locus but is heterozygous in the corresponding germline DNA.

Microsatellite Instability (MSI)

The condition of genetic hypermutability resulting from defective DNA mismatch repair.

Oncogene

A gene or a mutant variant of a gene that is associated with tumorigenesis.

Pathogenic Variant (PV)

A genetic alteration that is associated with or increases an individual’s predisposition to a particular disease, also known as a mutation.

Penetrance

The proportion of individuals in a population carrying a phenotype or disease-associated genetic variant who manifest the trait. When individuals carry a particular phenotype-associated allele and do not exhibit the phenotype, the gene is said to have reduced or lower-penetrance.

Resistance Allele

Also known as anon-risk or protective allele. Depending on the context, these alleles are associated with no change in disease risk or are associated with a decreased risk.

Single Nucleotide Variant (SNV)

Also known as a single nucleotide polymorphism(SNP). A SNVis a sequence variation in a single nucleotide occurringat a specific genetic location. The term SNP traditionally refers tovariation present at a frequency >1% in a population, whereasaSNV can occur at any frequency in a population.

Somatic Mutation

A change in DNA that occurs in non-germ cellsafter conception.

Susceptibility Allele

Also known as arisk-allele. This allele is associated with increased likelihood of developing the disease.

Tumor Suppressor Genes

Genes that suppress tumor development by regulating cell growth and division, stimulating cell death and/or DNA repair.

Two-hit Hypothesis

A hypothesis in which two mutations, one in each copy of a tumor suppressor gene, are required for a cell to give rise to a tumor.