From APC to the genetics of hereditary and familial colon cancer syndromes

Abstract

Hereditary colorectal cancer (CRC) syndromes attributable to high penetrance mutations represent 9–26% of young-onset CRC cases. The clinical significance of many of these mutations is understood well enough to be used in diagnostics and as an aid in patient care. However, despite the advances made in the field, a significant proportion of familial and early-onset cases remains molecularly uncharacterized and extensive work is still needed to fully understand the genetic nature of CRC susceptibility. With the emergence of next-generation sequencing and associated methods, several predisposition loci have been unraveled, but validation is incomplete. Individuals with cancer-predisposing mutations are currently enrolled in life-long surveillance, but with the development of new treatments, such as cancer vaccinations, this might change in the not so distant future for at least some individuals. For individuals without a known cause for their disease susceptibility, prevention and therapy options are less precise. Herein, we review the progress achieved in the last three decades with a focus on how CRC predisposition genes were discovered. Furthermore, we discuss the clinical implications of these discoveries and anticipate what to expect in the next decade.

Introduction

Familial clustering of colorectal cancer (CRC) and gastrointestinal polyposis has been recognized for well over a century (1,2), but little was known about the heritable changes in this heterogeneous group of diseases until the last three decades. Familial clustering is observed in up to 25% of CRC cases (3) and twin studies have yielded heritability estimates of ~ 35% (4).

Inherited predisposition to CRC because of high penetrance germline mutations (Table 1) has been implicated in 9–26% of patients diagnosed before the age of 50 years (5–9). Hereditary CRC syndromes caused by these mutations are phenotypically divided into polyposis and non-polyposis syndromes (Fig. 1). The more common non-polyposis syndromes include Lynch syndrome (OMIM#120435; LS), caused by DNA mismatch repair (MMR) gene mutations and resulting in a hypermutated tumor phenotype, and familial colorectal cancer type X (FCCTX) describing the remaining MMR-proficient cases with a heterogeneous genetic background. Polyposis syndromes are distinguished by adenomatous or non-adenomatous (hamartomatous/serrated) polyp histology (10). Of these, the causative gene for familial adenomatous polyposis (OMIM#175100; FAP), APC, was discovered first (11) (Fig. 2). CRC syndromes also involve predisposition to other tumors, including breast and gastrointestinal cancers in hamartomatous polyposes (12–14), endometrial cancer in LS (15,16), and fundic gland polyps (17) and desmoid tumors (18) in FAP.

Table 1.

Established (in bold) and putative high penetrance genes associated with hereditary CRC and polyposis

Gene symbol	Gene name	Gene function	Cancer syndrome(s)	Mode of inheritance	Method of discovery	Type of genetic instability	PubMed ID	Ref. no.
APC	Adenomatous polyposis coli	Inhibitor of Wnt signaling pathway	FAP	Autosomal dominant	Chromosomal deletion	CIN	3 789 010 1 651 174 1 651 562 1 651 563 1 678 319	(21) (11) (26) (221) (222)
AXIN2	Axis inhibition protein 2	Inhibitor of Wnt signaling pathway	Familial tooth agenesis and attenuated familial polyposis	Autosomal dominant	Linkage and candidate gene analysis	CIN?	15 042 511	(62)
BMPR1A	BMP receptor, type IA	Activates SMAD transcriptional regulators	JPS FCCTX	Autosomal dominant	Linkage analysis	CIN (low)	11 381 269 (JPS) 21 640 116 (FCCTX)	(73) (JPS) (223) (FCCTX)
BRF1	BRF1 RNA polymerase III transcription initiation factor subunit	Activator of RNA polymerase	FCCTX Cerebellofaciodental syndrome (CFDS)	Autosomal dominant	Exome sequencing	CIN	28 912 018 (FCCTX) 25 561 519 (CFDS)	(224) (FCCTX) (225) (CFDS)
EPCAM	Epithelial cell adhesion molecule	Cell adhesion	LS (MSH2 hypermethylation from transcriptional read-through of EPCAM) Congenital tufting enteropathy (CTE)	Autosomal dominant (LS) Autosomal recessive (CTE)	Tumor phenotype	MSI (LS)	16 951 683 (LS) 19 098 912 (LS) 18 572 020 (CTE)	(226) (LS) (227) (LS) (228) (CTE)
ERCC6	ERCC excision repair 6, chromatin remodeling factor	Transcription-coupled nucleotide excision repair	FCCTX Cockayne’s syndrome (CKS)	Autosomal dominant/codominant with WRN? (FCCTX) Autosomal recessive (CKS)	Exome sequencing	CIN	26 344 056 (FCCTX) 1 339 317 (CKS) 9 443 879 (CKS)	(83) (FCCTX) (229) (CKS) (230) (CKS)
FAF1	Fas-associated factor 1	DNA replication	FCCTX	Autosomal dominant	Exome sequencing	CIN	32 179 092	(231)
FAN1	FANCD2- and FANCI-associated nuclease 1	DNA cross-link repair	FCCTX Karyomegalic interstitial nephritis (KMIN)	Autosomal dominant (FCCTX) Autosomal recessive (KMIN)	Exome sequencing	CIN	26 052 075 (FCCTX) 22 772 369 (KMIN)	(232) (FCCTX) (233) (KMIN)
GALNT12	Polypeptide N-acetylgalactosaminyl-transferase 12	Oligosaccharide biosynthesis	FCCTX	Autosomal dominant (possibly dominant negative)	Candidate gene approach	?	19 617 566	(234)
GREM1	Gremlin 1, DAN family BMP antagonist	BMP antagonist	Mixed polyposis syndrome	Autosomal dominant	Candidate gene approach	CIN (low?)	22 561 515	(235)
MBD4	Methyl-CpG binding domain 4, DNA glycosylase	BER	MBD4-associated neoplasia syndrome (MANS)	Autosomal recessive	Whole-genome sequencing	CpG > TpG	30 049 810	(236) (237)
MLH1	MutL homolog 1	MMR	LS (monoallelic) Constitutional MMR deficiency syndrome (CMMRD; biallelic)	Autosomal dominant (LS) Autosomal recessive (CMMRD)	Linkage analysis	MSI-H Hypermutated	7 903 889 8 128 251 8 145 827	(53) (51) (52)
MLH3	MutL homolog 3	MMR recombination	MLH3-associated adenomatous polyposis (biallelic) Low-penetrance CRC (monoallelic)	Autosomal recessive (polyposis)	Exome sequencing (polyposis) Candidate gene approach (CRC)	CIN (polyposis) MSI-low (CRC)	30 573 798 (polyposis) 11 586 295 (CRC)	(91) (polyposis) (238) CRC
MRE11	MRE11 homolog, DSB nuclease	Telomere maintenance	FCCTX Ataxia-telangiectasia-like disorder (ATLD)	Autosomal dominant? Autosomal recessive (ATLD)	Exome sequencing	CIN?	27 329 137 (FCCTX) 10 612 394 (ATLD)	(239) (FCCTX) (240) (ATLD)
MSH2	MutS homolog 2, colon cancer, nonpolyposis type 1 (E. coli)	MMR	LS (monoallelic) CMMRD; biallelic	Autosomal dominant (LS) Autosomal recessive (CMMRD)	Linkage analysis and tumor phenotype	MSI-H Hypermutated	8 484 120 8 252 616 8 261 515 8 484 121	(46) (47) (45) (57)
MSH3	MutS homolog 3	MMR	MSH3-associated adenomatous polyposis (biallelic)	Autosomal recessive (polyposis)	Exome sequencing	MSI (EMAST)	27 476 653	(90)
MSH6	MutS homolog 6	MMR	LS (monoallelic) CMMRD; biallelic	Autosomal dominant (LS) Autosomal recessive (CMMRD)	Candidate gene approach	MSI-H	9 354 786	(81)
MUTYH	MutY DNA glycosylase	BER	MUTYH-associated adenomatous polyposis	Autosomal recessive	Tumor phenotype	G:C > T:A	11 818 965	(38)
NTHL1	Nth-like DNA glycosylase 1	BER	NTHL1-associated adenomatous polyposis	Autosomal recessive	Exome sequencing	C:G > T:A Hypermutated	25 938 944	(89)
OGG1	8-Oxoguanine DNA glycosylase	BER	FCCTX	Autosomal dominant/codominant	Candidate gene approach	G:C > T:A?	21 195 604	(114)
PMS2	PMS2 postmeiotic segregation increased 2	MMR	LS (monoallelic) CMMRD; biallelic	Autosomal dominant (LS) Autosomal recessive (CMMRD)	Candidate gene approach	MSI	8 072 530	(241)
POLD1	DNA polymerase delta 1, catalytic subunit	DNA replication	PPAP	Autosomal dominant	Linkage analysis and genome sequencing	Ultrahypermutated Possible MSI	23 263 490	(87)
POLE	DNA polymerase epsilon, catalytic subunit	DNA replication	PPAP	Autosomal dominant	Linkage analysis and genome sequencing	Ultrahypermutated Possible MSI	23 263 490	(87)
POLE2	DNA polymerase epsilon 2, catalytic subunit	DNA repair and replication	FCCTX Adenomatous polyposis	Autosomal dominant? (27329137) Autosomal recessive? (25529843)	Exome sequencing	Hypermutant?	25 529 843 27 329 137	(242) (239)
POT1	Protection of telomeres 1	Telomere maintenance	FCCTX Familial melanoma (FM) Familial glioma (FG) Li-Fraumeni-like syndrome (LFL)	Autosomal dominant?	Exome sequencing	CIN?	27 329 137 (FCCTX) 24 686 849 (FM) 24 686 846 (FM) 25 482 530 (FG) 26 403 419 (LFL)	(239) (FCCTX) (243) (FM) (244) (FM) (245) (FG) (246) (LFL)
PTEN	Phosphatase and tensin homolog	Phospatase	CS PTEN hamartoma tumor syndrome	Autosomal dominant	Linkage and candidate gene approach	CIN (low?)	8 673 088 9 072 974 9 140 396	(66) (247) (67)
RNF43	Ring finger protein 43	DNA damage response	Serrated polyposis syndrome	Autosomal dominant	Exome sequencing	CIN	24 512 911	(248)
RPS20	Ribosomal protein 20	Ribosome biogenesis	FCCTX DBA	Autosomal dominant	Exome sequencing	CIN?	24 941 021 (FCCTX) 32 790 018 (DBA)	(88) (FCCTX) (98) (DBA)
SEMA4A	Semaphorin 4A	Semaphorin	FCCTX Retinis pigmentosa (RP)	Autosomal dominant (FCCTX) Autosomal recessive (RP)	Linkage analysis and exome sequencing	CIN	25 307 848 (FCCTX) 16 199 541 (RP)	(249) (FCCTX) (250) (RP)
SMAD4	SMAD family member 4	Cytoplasmic mediators of TGF-beta signaling	Juvenile polyposis syndrome	Autosomal dominant	Candidate gene approach	CIN (low?)	9 582 123	(71)
STK11	Serine/threonine kinase 11	AMPK activity regulator	PJS	Autosomal dominant	Chromosomal deletion	CIN (low)	8 988 175 9 428 765	(68) (69)
WRN	WRN RecQ like helicase	DNA DSB repair	FCCTX Werner syndrome	Autosomal dominant/co-dominant with ERCC6 (FCCTX) Autosomal recessive (Werner)	Exome sequencing (FCCTX) Candidate gene approach (Werner)	CIN	26 344 056 (FCCTX) 8 602 509 (Werner)	(83) (FCCTX) (251) (Werner)

Figure 1.

CRC syndromes. Division between syndromes is traditionally based on the number of and histopathology of intestinal polyps and mode of inheritance. The most common germline mutant genes for each syndrome are given. Ratios indicate estimates of prevalence of each syndrome based on literature (201–213).

Figure 2.

Timeline of CRC and polyposis susceptibility gene discoveries in the Sanger sequencing and NGS eras. Timeline assignment is based on the discovery of the gene as a CRC or polyposis susceptibility gene (the gene itself may have been identified before). Established susceptibility genes are colored and putative are indicated with a gray box.

In this anniversary review, we give an overview on genetics of hereditary CRC syndromes with a focus on Mendelian conditions and developments in genetic analysis techniques over the past three decades.

Clues From Molecular Tumor Phenotype

Chromosomal abnormalities have been attributed to the genesis of cancer for over a century (19), and deletions of the long arm of chromosome 5 (5q) have been recognized as the most frequent chromosome anomaly in secondary leukemia and in some solid cancers for some 40 years (20). As 5q was already a prominent factor in cancer development and familial clustering of colon polyposis was acknowledged, a clinical report of a Gardner syndrome patient with a de novo interstitial deletion within the 5q arm (21) provided the ‘missing link’ connecting 5q to polyposis predisposition. Soon, several research groups confirmed frequent allelic loss in 5q in CRC (22) and linked the adenomatous polyposis phenotype to the locus (23–25), leading to the detailed characterization of the APC gene (26) and initiating the hunt for gene defects responsible for inherited CRC and polyposes. Today, germline APC mutations are recognized as the underlying cause of FAP (Table 1); of these, 10–25% arise de novo. Typical of disorders with high de novo mutation rates (27), somatic APC mutations are sometimes restricted to certain tissues or organs, reflecting the timing of the mutation occurrence in embryogenesis. In these cases multiple genotypes are present in the patient, collectively referred to as genetic mosaicism. Patients with somatic APC mosaicism generally manifest as sporadic cases, as a disease phenotype only occurs when a substantial number of cells with the mutation are present in the target tissue. The detection of somatic APC mosaicism with conventional methods often proves difficult; the mutations might not be observable in samples derived from peripheral blood (28–30). Somatic APC mutations are commonplace in colorectal tumors and key players initiating colonic tumorigenesis (31–33).

Sometimes the landscape of somatic mutations can provide direct clues to the underlying germline mutation. In a British polyposis family negative for APC germline mutations, 11 tumors revealed somatic inactivation of APC because of G:C → T:A transversions (34). Previous efforts had identified spontaneous G:C → T:A transversions in OGG1 (35), MUTYH (36,37), and MTH mutant Escherichia coli and Saccharomyces cerevisiae, which gave researchers an incentive to test the aforementioned genes for mutations. This resulted in the discovery of two MUTYH missense mutations in the family, p.Tyr165Cys and p.Gly382Asp, segregating with the disease in an autosomal recessive mode. Nowadays both variants are recognized as common causes for MUTYH-associated polyposis (OMIM#608456) when biallelic (34,38,39). Later mutational signature analyses have attributed the G:C → T:A transversions in MUTYH-deficient tumors to the mutational signatures 18 (40) and 36 (41). Tumors deficient of the base-excision repair protein NTHL1 also have a conserved mutational signature characterized by C:G → T:A transitions (mutational signature 30) (42,43). Similarly, small insertions and deletions at repetitive sequences were consistently observed in tumors belonging to the spectrum of hereditary non-polyposis CRC (44), simultaneously attributed to genes in charge of MMR (45–47). This tumor phenotype is widely in use for diagnostic purposes as we will later discuss.

Gene-Hunting With Linkage Analysis—Narrowing It Down

Genetic linkage analysis remains a powerful tool for rare variant detection when combined with whole-genome sequencing (48,49) and has played an important role in gene discoveries made in CRC syndromes. LS is caused by germline mutations in MMR genes MLH1 (42% of unique variants), MSH2 (33%), MSH6 (18%), and PMS2 (7.5%) (50) (Table 1, Fig. 3). MSH2 and MLH1 were the first LS genes discovered in 1993–1994 (45,47,51,52), fueled by mapping of the respective predisposing loci by genetic linkage analysis (46,53,54). For MSH2, two large LS kindreds were studied using 345 microsatellite markers across the genome, revealing highly significant linkage with marker D2S123 at 2p15-16 (46). For MLH1, restriction fragment length polymorphisms and microsatellite markers studied in Swedish LS families revealed a predisposing locus at 3p21.3-23 (53,54). Subsequent gene identification by positional cloning was aided by the evolving understanding of MMR proteins in bacteria and yeast, providing a connection with the instability of DNA microsatellite markers observed in LS and other human tumors (55–61).

Figure 3.

Key biological pathways associated with hereditary CRC syndromes. Genes with known germline mutations are shown in purple font. (A) Fidelity of DNA replication depends primarily on DNA polymerases ε and δ for correct base pairing and proofreading exonuclease activity. Any persisting base–base mispairs or insertion–deletion loops both on the leading and lagging strands are subsequently targeted by MMR proteins. MSH2:MSH3/6 dimers recognize the error. Interaction with MLH1:PMS2 triggers downstream repair events initiated by removal of the erroneous DNA by exonuclease 1 (Exo1). (B) Oxidation, alkylation, and deamination of DNA bases are repaired by base excision repair (BER) pathway. Eleven DNA glycosylases recognize and remove the damaged/mispaired bases; these are divided into monofunctional (e.g. MUTYH) and bifunctional (e.g. NTHL1, OGG1) glycosylases based on the presence of additional endonuclease activity. Their function is exemplified by repair of oxidative 8-oxoguanine (oxoG) and 5-hydroxycytosine (5-OHC) lesions. (C) Wnt signaling regulates cell development and stemness and is commonly hyperactivated in cancer. In the absence of secreted Wnt ligands, a cytoplasmic destruction complex directs β-catenin for proteasomal degradation and RNF43/ZNRF3 ubiquitin ligases downregulate Frizzled receptors. Upon receptor binding of Wnts, β-catenin is released and translocates to the nucleus, where it binds TCF/LEF transcription factors and displaces the Groucho repressor to activate target gene transcription. Binding of R-spondins (Rspo) to LGR5 inhibits RNF43/ZNRF3, enhancing Wnt signaling. (D) TGF-β signaling restricts proliferation of colonic epithelial cells. Bone morphogenetic proteins (BMPs) are TGF-β superfamily ligands that trigger receptor dimerization and activation via trans-phosphorylation, resulting in SMAD-dependent target gene transcription. Also, non-SMAD signaling pathways, including the PI3K–AKT–mTOR pathway, can be activated. Growth factor signaling via receptor tyrosine kinases (RTK) activates PI3K to generate phosphoinositide-3,4,5-triphosphate (PIP₃), resulting in AKT-mediated derepression of mTOR complex 1 (mTORC1). mTOR integrates nutrient and growth factor signals, promoting cell growth upon activation. PTEN lipid phosphatase antagonizes this pathway by converting PIP₃ back to PIP₂. Conversely, low glucose conditions trigger the STK11-dependent activation of AMPK, suppressing mTORC1. See references (214–220) for further details.

AXIN2 locus underlying oligodontia-CRC syndrome (OMIM#608615) (Table 1) was discovered in a genome-wide search utilizing microsatellite markers in an affected Finnish family (62). Linkage was excluded for APC and candidate loci associated with tooth development, pointing to a region in chromosome 17. AXIN2 (17q24.1) became the strongest candidate owing to its role in WNT signaling (Fig. 3) and somatic mutations in CRC (63,64). Sequencing revealed an AXIN2 nonsense mutation segregating with the phenotype (62) as later confirmed in unrelated cases (65).

The genes for the main hamartomatous polyposis syndromes—Cowden syndrome (OMIM#158350; CS), Peutz-Jeghers syndrome (OMIM#175200; PJS), and juvenile polyposis syndrome (OMIM#174900; JPS)—were all identified through linkage studies. For CS, a traditional linkage and candidate gene analysis identified PTEN as the culprit gene (66,67) (Table 1, Fig. 3). In case of PJS, a targeted linkage analysis based on allelic loss detected in PJS polyps by comparative genomic hybridization revealed the disease locus, followed by fine mapping and systematic sequencing of transcripts in the region to identify the gene, LKB1/STK11 (68,69). For JPS, targeted linkage analysis excluded loci previously associated with CRC and closely related hamartomatous polyposis syndromes, revealing linkage to 18q21.1 (70). This led to the discovery of SMAD4 (18q21.2) mutations (71), but these were found to explain only a minority of the cases (72). Genome-wide linkage analysis in four JPS families without SMAD4 or PTEN mutations resulted in the discovery of germline BMPR1A (10q23.2) nonsense mutations in all families (73). Together, mutations in SMAD4 and BMPR1A, both involved in transforming growth factor beta (TGF-β) signaling, account for 50–60% of JPS cases (74) (Table 1, Fig. 3).

Candidate Genes Pinpointed by Biological Function

Existing knowledge of biological functions of candidate genes can greatly facilitate predisposition gene discoveries. Human MSH6 was identified as a 160 kDa protein dimerizing with MSH2 in a mismatch-binding factor purified from HeLa cells (75–77) (Fig. 3). Both proteins were required for restoring MMR in extracts of hypermutable CRC cell lines (76–78). Sequence analysis of proteolytic peptides from p160 allowed cDNA identification, revealing homology to human and yeast MSH2 and E. coli MutS (75–77). An Msh6-deficient mouse model gave the first evidence for a role in cancer susceptibility (79). While initial attempts to find germline mutations of MSH6 in 20 LS kindreds failed (78), mutations were soon found in atypical LS cases with weak family history (80,81). In contrast to MLH1 or MSH2 mutant tumors, microsatellite instability (MSI) is not always detected in MSH6 mutant tumors (79,82), in agreement with its preferential role in the repair of single-nucleotide mismatches (76–78).

In a study of 25 patients with familial CRC of undefined genetic basis (83), mostly falling into the category of FCCTX (Fig. 1), the authors hypothesized to find mutations predisposing to DNA double-strand breaks (DSBs) based on earlier reports on replication stress as the underlying mechanism of chromosomal instability in CRC (84). Compared with controls, CRC patient T cells had increased susceptibility to DSBs under ultraviolet radiation or aphidicolin-induced replication stress, and exome sequencing showed significantly higher frequency of rare disruptive germline mutations in DSB repair-related genes, found in 25 genes in 17/20 patients (83). Heterozygous mutations in WRN (genome stability during DNA repair) and ERCC6 (transcription-coupled nucleotide excision repair) found in one of the patients were functionally validated, further supporting a role for constitutional defects in suppression of DSBs as a new molecular class of hereditary CRC. Homozygous mutations cause Werner (OMIM#277700) and Cockayne (OMIM#133540) syndromes, respectively, the former associated with various neoplasms and the latter with heightened sensitivity to ultraviolet. Interestingly, the WRN-encoded DNA helicase was suggested to have a cell/tissue type specific role in MMR (85), and a related RecQ family DNA helicase BLM directly interacts with MSH6 (86).

From Single Genes to the Exome and Genome

The emergence of new sequencing techniques has enabled scientists to make advances in oncogenomics, especially in the study of somatic changes in tumors. Nevertheless, success in research of CRC predisposition has been limited. In conjunction with family information, be it in the form of linkage or other methods, a handful of novel causative genes have been identified in the last decade (Fig. 2, Table 1) (87–91). One of the successful results is the characterization of polymerase proofreading-associated polyposis (PPAP) by discovery of the associated genes: POLE and POLD1 (OMIM#615083 and OMIM#612591, respectively). Whole exome sequencing (WES) analysis of 15 mutation negative index cases with polyposis and family history of CRC, as well as some of their family members, initially left the researchers empty-handed: after excluding common and benign variants across multiple families, no susceptibility genes were identified (87). However, previous linkage data from a few families in this study (92,93) pinpointed the analysis to POLE and POLD1 genes, resulting in the identification of three families wherein the mutations segregated with the disease. Subsequent research has further defined the roles of POLE and POLD1 in familial CRC and polyposis (94).

The genetic background of FCCTX has been a particularly difficult one to crack: the underlying genes have eluded geneticists worldwide despite being under investigation for almost 20 years (95). In a similar fashion to the discovery of PPAP, a combination of linkage analysis and WES identified a truncating variant of RPS20 in association with MMR-proficient CRC in a four-generation family (88,95). Subsequent studies have since identified three RPS20 mutations in familial CRC cases (96,97) and two in Diamond-Blackfan anemia (DBA) patients (98). Available experience suggests that only a small fraction of FCCTX susceptibility genes have been discovered so far, and that private genes and mutations are common, strengthening the idea of the heterogeneous genetic background of FCCTX (94,99).

WES has become the main approach to detect novel cancer susceptibility genes and has enabled much of the recent success in the discovery of novel CRC and polyposis susceptibility genes. In addition to POLE, POLD1, and RPS20, recent successful reports of novel CRC and polyposis susceptibility variant discoveries include NTHL1 (89), MSH3 (90), and MLH3 (91) (Table 1). While WES is able to detect (or exclude) mutations in exons of protein-coding genes, a disadvantage of the method is that only ~2% of the genome is covered and many of the regulatory elements controlling the expression of these genes remain unrevealed. For example, intronic variants resulting in a splicing acceptor or donor domains can create pseudoexons, intronic sequences inserted in the mature mRNA of a gene. Detection of such variants is impossible with just WES and requires simultaneous transcriptome and whole genome-based approaches. Few cases of APC pseudoexons have been reported (100,101).

Toward Understanding ‘the Missing Heritability’—Oligogenic and Polygenic Etiology

Though much has been learned from hereditary CRC, mutations in known susceptibility genes explain only a fraction of familial CRC risk, and multiple explanations for this ‘missing heritability’ have been proposed (102–105). Genome-wide association studies (GWAS) continue to uncover novel susceptibility loci and common low-risk variants associated with CRC at the population level, further shaping our view of the genetic landscape of CRC (105,106). A recent effort combining whole-genome sequencing and imputation into GWAS data almost doubled the number of known association signals for sporadic CRC into ~ 100 independent signals at > 80 CRC risk loci (107), implicating novel pathways and gene families in CRC risk. These GWAS findings may provide further clues into etiology of hereditary CRC syndromes, as genes previously linked with hereditary syndromes have been found to harbor also common CRC risk variants with weaker effects (105). Common variants may modify CRC risk in patients with hereditary CRC syndromes, possibly explaining some of the observed differences in the ages of cancer diagnosis and disease presentation among individuals carrying the same high-penetrance mutation (108–112). Moreover, combinations of multiple common low-penetrance variants may work synergistically to substantially contribute to overall disease heritability. A recent example is from serrated polyposis syndrome, a hereditary CRC predisposition with a poorly understood genetic basis, which was analyzed in a case–control study for contribution of 65 common low-penetrance CRC risk variants, reporting significant associations of the condition with seven of these (113). These findings, together with emerging reports on di- or oligogenic inheritance via rare moderately penetrant variants (114–117), highlight the complex genetic make-up of hereditary CRC.

Clinical Implications

Molecular diagnostics of hereditary CRC has advanced significantly in the past decades, enabling today efficient screening strategies to identify individuals at high risk for referral to genetic testing and specialized cancer surveillance programs for each hereditary CRC predisposition (118). The genetic testing of affected individuals and predictive testing of their at-risk relatives, combined with intensive cancer surveillance, has an enormous cancer-preventive potential in these families, while also reducing unnecessary procedures and anxiety in family members who are found mutation negative.

Compared with polyposis syndromes, LS lacks distinctive endoscopic features and presents a particular challenge for diagnostics. Diagnosis was initially based on so-called Amsterdam criteria formulated in 1991 (119,120); these focused on family history and age of onset and were highly useful for selection of families for scientific gene identification studies. More practical and less stringent Bethesda guidelines (121,122) followed, utilizing tumor MSI for the selection of patients for further genetic testing. Both criteria became suboptimal for identifying LS cases as tools for genetic diagnostics advanced, prompting recommendations to offer genetic testing to all individuals with newly diagnosed CRC (123,124). The discovery of a high frequency of DNA replication errors in LS tumors compared with normal tissue paved the way for universal screening based on MSI (57,125–127). MSI testing involves polymerase chain reaction (PCR) amplification of a set of short repetitive DNA sequences (microsatellites) to determine changes in their length in tumors (128). Apart from LS, MSI occurs in ~ 12% of sporadic CRCs because of defective MMR caused by MLH1 promoter methylation or, rarely, bi-allelic somatic mutations of MMR genes (129,130). The former can be excluded by testing the tumors for the BRAF-V600E mutation and/or MLH1 promoter hypermethylation (131). Immunohistochemistry for detecting loss of expression of MMR proteins in tumor tissue is another prevalent LS screening approach (132–134). MSI testing and immunohistochemistry identify LS cases with an estimated clinical sensitivity of 85/83% and specificity of 90/89%, respectively (124). Subsequent detection of germline mutation in an MMR gene by sequencing then establishes the diagnosis of LS.

In recent years, the sequencing of each established hereditary CRC gene individually has become increasingly replaced by next-generation sequencing (NGS)-based diagnostic gene panels. These allow simultaneous germline mutation testing of a wide selection of CRC predisposition genes in a cost-effective manner in suspected CRC/polyposis syndrome cases, supporting utility as a first line test (135). The overlapping clinical presentation and genetic heterogeneity in CRC syndromes may also be better addressed by panel testing. The diagnostic yield and clinical actionability of results from CRC panel testing vary depending on the cohort selected for testing, genes included in the panel, and variant interpretation (6–8,136–147). In cohorts of unselected CRC patients representing different populations and ethnicities, panel testing of 25, 18, 73, 83, and 27 cancer-predisposing genes identified pathogenic or likely pathogenic germline mutations in 9.9%, 2%, 18.1%, 15.5%, and 3.3% of the patients, respectively (136,137,143,144,147). A large variety of commercial CRC gene panels are available (148). These show inconsistent criteria for inclusion of genes, including both high and low penetrance genes with highly variable levels of supporting evidence (148). To address these and other concerns, the Collaborative Group of the Americas on Inherited Gastrointestinal Cancer recently published a position statement proposing a minimal set of 11 genes to be included on a multigene panel for evaluation of hereditary CRC/polyposis; MLH1, MSH2, MSH6, PMS2, EPCAM, APC, BMPR1A, MUTYH, PTEN, STK11, and SMAD4 (149). The statement also recommends an additional set of 16 genes to be considered for panel testing (149), including genes with low-to-moderate CRC risk, preliminary but limited data on CRC risk, or actionable mutations without proven causation in CRC. Notably, this set includes several genes with germline mutations not traditionally associated with CRC/polyposis, including TP53, CHEK2, and BRCA1/2, highlighting the utility of a broader analysis of hereditary cancer genes in diagnostic panels.

The hypermutable state in MSI tumors causes accumulation of frameshift mutations that generate neoantigens, immunogenic peptides recognized as non-self by the immune system (150–156). Neoantigen burden is significantly increased in MSI versus microsatellite stable (MSS) CRCs and in LS versus sporadic MSI CRCs (154,155). The neoantigens can be presented at the tumor cell surface in human leukocyte antigen type I (HLA I) molecules, eliciting antitumor cytotoxic T cell responses (150–152,154,157,158). These observations likely explain the high level of lymphocyte infiltration long known as a hallmark of MSI CRCs (159,160) and more recently shown to be a predictor of better prognosis (154,161) similar to MSI (162). The highly immunogenic tumor microenvironment in MSI CRCs is balanced by elevated expression of several immune checkpoint molecules that block antitumor T cell responses, making this subgroup an excellent candidate for treatment using immune checkpoint inhibitors (163,164). Clinical studies have consistently demonstrated efficacy of immune checkpoint inhibition with anti-PD-1 antibodies (pembrolizumab, nivolumab) in MSI tumors from colorectum and many other tissues (165–171). Both drugs have received US Food and Drug Administration approval for treatment of patients with MSI-high or MMR-deficient metastatic CRC, followed by approval of nivolumab combination immunotherapy with anti-CTLA-4 antibody (ipilimumab) for the same indication (169). Similar to MSI, also the presence of POLE and POLD1 mutations, associated with an ultramutator phenotype in tumors, may predict survival benefit from immune checkpoint therapy in CRC and other cancers (172,173).

Conclusions and Future Perspectives

Advances in the dissection of hereditary CRC syndromes offer a remarkable example of how the understanding of human disease has improved in the past 30 years. Interestingly, with the new knowledge, the borders between established CRC syndromes have become blurrier: for example, evidence shows that some LS patients can mimic attenuated FAP (91,174) and few families exhibiting digenic inheritance of CRC have been reported (114,117). Thus, genotype–phenotype diversity in inherited CRC will likely continue to be a source of difficulty for diagnostics (140) and treatment (175). Much further work will be required to better understand the key components of CRC susceptibility and optimal management strategies for most syndromes, though effective targeted treatments already exist for some of the affected signaling pathways (Fig. 3). For example, patients with APC germline mutations respond well to treatment with the nonsteroidal anti-inflammatory drug sulindac and COX inhibitor celecoxib. In these patients, adenoma burden was significantly reduced (176), likely because of the suppression of canonical WNT/β-catenin signaling (177,178). Very recent breakthroughs have been made in LS vaccination by exploiting the recurrent frameshift-induced neoantigens consequent of the defective MMR machinery (179). Dense lymphocyte infiltration accompanies the MSI tumors of LS patients (159,160,180) and specific T cell reactivity to frameshift peptides is detectable already in healthy LS family members without a history of cancer (158). These observations, combined with the fact that LS carcinomas often lose HLA I expression likely because of a selection pressure (181,182), suggest that the LS patient’s immune system effectively recognizes MMR-deficient cells and may control tumor outgrowth. Therefore, vaccination is an attractive option for cancer prevention in the future. Moreover, vaccination could direct the development of MSI tumors toward a surgically curable phenotype by increasing the immunoselective pressure and likelihood of HLA I loss via mutations in beta-2-microglobulin (182); these mutations have been associated with an excellent prognosis with a low risk of relapse (183) and metastases (184,185). The first clinical trials for frameshift peptide vaccines are promising, with low toxicity and high immune response (179).

Similar to the use of MSI-induced frameshift products in LS-diagnosis, tumor-specific mutation spectra or signatures have laid the groundwork for connections between the molecular profiles of cancers and the tumor development (186). One notable advance in the last decade is the detailed characterization of mutational signatures across different cancer types (187,188). As each signature is a reflection of the mutational processes in cancer, especially prominent in samples whose DNA repair machinery is defective (35–37,40,41,43,189), they can serve as a hint toward the underlying cancer-causing mechanisms and serve as potential diagnostic or prognostic biomarkers (190). Furthermore, mutational signatures may provide evidence for accurate classification of variants of unknown significance as pathogenic or benign (191). For example, owing to the discovery of the mutational signature 30 in a breast cancer sample, a discovery of a germline mutation of NTHL1 was made (42). Finding characteristic features, such as MSI or specific mutational signatures, in tumor samples could reflect the germline defect in molecularly unexplained familial CRC or polyposis cases, acting as a starting point for finding the underlying cause of disease. Unfortunately, availability of tumor DNA is often a limiting factor for NGS-based studies. Liquid biopsies could be a reliable source of tumor DNA for such purposes, as significant numbers of clinically actionable variants have been present in cell-free DNA samples when tumor samples were either unavailable or had not yielded any results (192). Similar conclusions can be drawn from studies utilizing mutation-specific PCR (193) and methylation analysis (194,195) of cell-free DNA from patients with metastatic disease. MSI status and mutational signatures observed in cell-free DNA seem to reflect those of the primary tumor as well (196,197).

Despite major advances in the field, CRC remains one of the leading causes of cancer-related death and a major economic burden (198,199) that personalized medicine is thought to alleviate (200). The decreasing costs of genomic research technologies bring the onset of precision medicine closer to reality as more research is carried out. Our current knowledge of germline and somatic changes in CRC predisposition has given us meaningful clinical predictors for the best-characterized syndromes; however, extending this to the rarer or lesser-known syndromes and understanding the genotype–phenotype correlations present in recognized syndromes is still awaiting further research. This includes the analysis of the noncoding genome and disease-modifying variants, which are anticipated to shed light onto the susceptibility to cancer as well as its progression and treatment. The work of cancer genomics projects, such as The Cancer Genome Atlas, The International Cancer Genome Consortium, or the Pan-Cancer Analysis of Whole Genomes Consortium, will likely deepen our understanding of CRC through the development of research tools and techniques able to identify generalized patterns and characteristics of cancer, such as mutational signatures, contributing to the missing factors prohibiting detailed molecular profiling of patient-matched cancer and interpretation of variants of unknown significance.

Currently, many of the syndromes are not surgically curable, and patients need life-long monitoring. With the rate of new technological advances, the next decade will be exciting and will likely yield a better understanding of the genetic and environmental determinants of CRC susceptibility, the phenotype–genotype correlations of the already established susceptibility genes, and optimal management of heritable CRC in its various forms.

Conflicts of Interest statement. None declared.

Funding

The Biomedicum Helsinki Foundation (to A.P.O.); The Finnish Center of Excellence in Tumor Genetics (to L.A.A.); the Jane and Aatos Erkko Foundation (to P.T.P.); the Academy of Finland (grant number 294643 to P.T.P.); the Academy of Finland’s professorship (L.A.A.); the Cancer Foundation Finland (to P.T.P. and L.A.A.); iCAN Digital Precision Cancer Medicine Flagship (to L.A.A.); the Sigrid Juselius Foundation (to P.T.P. and L.A.A.); The Doctoral Programme of Biomedicine of Helsinki University offered a paid doctoral student position to A.P.O.