Genetic predisposition to colorectal cancer: Where we stand and future perspectives

Abstract

The development of colorectal cancer (CRC) can be influenced by genetic factors in both familial cases and sporadic cases. Familial CRC has been associated with genetic changes in high-, moderate- and low-penetrance susceptibility genes. However, despite the availability of current gene-identification techniques, the genetic causes of a considerable proportion of hereditary cases remain unknown. Genome-wide association studies of CRC have identified a number of common low-penetrance alleles associated with a slightly increased or decreased risk of CRC. The accumulation of low-risk variants may partly explain the familial risk of CRC, and some of these variants may modify the risk of cancer in patients with mutations in high-penetrance genes. Understanding the predisposition to develop CRC will require investigators to address the following challenges: the identification of genes that cause uncharacterized hereditary cases of CRC such as familial CRC type X and serrated polyposis; the classification of variants of unknown significance in known CRC-predisposing genes; and the identification of additional cancer risk modifiers that can be used to perform risk assessments for individual mutation carriers. We performed a comprehensive review of the genetically characterized and uncharacterized hereditary CRC syndromes and of low- and moderate-penetrance loci and variants identified through genome-wide association studies and candidate-gene approaches. Current challenges and future perspectives in the field of CRC predisposition are also discussed.

Key Words: Hereditary colorectal cancer, Familial colorectal cancer, High penetrance, Low penetrance, Cancer syndromes, Cancer susceptibility, Hereditary cancer genes, Risk variants, Heritability

Core tip: The risk of developing colorectal cancer (CRC) can have genetic influences, especially when there is a family history of the disease. Much of this genetic predisposition to develop cancer is already known, including high-penetrance genes, i.e., those responsible for hereditary cases, and low-penetrance alleles, which are responsible for both sporadic and familial cases. However, despite recent developments in gene identification techniques, the genetic causes of many hereditary cases remain unknown. This review details the hereditary CRC syndromes and their genetic causes, the roles of low- and moderate-risk genetic factors in familial cases and the state-of-the-art in the identification of new causal genes.

BACKGROUND

Colorectal cancer (CRC) is the third most common cancer, accounting for 10% of all cancers and affecting approximately 1 million people worldwide every year[1]. Although most cases of CRC are thought to be sporadic, crude estimates indicate that familial CRC, defined by the presence of two or more first-degree relatives affected with CRC, accounts for more than 20% of all cases[2-4]. All CRC syndromes caused by known high-penetrance CRC genes collectively account for 2%-6% of all cases of CRC. For decades, gene-identification strategies such as genome-wide linkage studies or studies involving high-throughput sequence capture methods and next-generation sequencing technologies have sought to identify new high-penetrance genes that could explain the aggregation of CRC in high-risk families. Despite this technological progress, the genetic etiology of familial cancers such as familial colorectal cancer type X (fCRC-X) or serrated polyposis (SP) remains unknown.

For years, scientists have hypothesized that the heritable nature of CRC might be associated with the co-inheritance of multiple low-risk variants[2,3] that may interact with environmental factors. This hypothesis was supported by the identification of single-nucleotide polymorphisms (SNPs) localizing to different genomic regions that influence the risk of CRC[5]. The risk of CRC associated with each of the variants is individually low; however, the combined effect of these variants could significantly contribute to disease burden, especially given the high prevalence of these variants in the general population. Moreover, the presence of these or other SNPs might modify the risk of cancer in families with mutations in known predisposing genes such as those associated with lynch syndrome (LS)[6-8].

In this review, we present current knowledge on the genetics of inherited CRC syndromes and of moderate- and low-risk variants of CRC; we also describe the approaches currently being used to understand the genetic causes of uncharacterized hereditary cases of CRC. Current challenges and future perspectives are discussed. Clinical issues such as surveillance, prophylactic and preventive measures, treatments and genetic counseling are not reviewed.

GENETICALLY CHARACTERIZED INHERITED COLORECTAL CANCER SYNDROMES

A summary of the main genes associated with hereditary cancer syndromes is provided in Table 1; information on the modes of inheritance of these syndromes, the types of mutations identified in patients with these syndromes and the molecular features of tumors are also presented.

Table 1 Hereditary colorectal cancer genes, major associated syndromes, modes of inheritance, types of mutations identified and specific molecular characteristics of associated tumors.

Gene	Syndrome	Inheritance	Mutations reported	Tumor molecular features
MLH1	Lynch syndrome	Autosomal dominant	Point mutations1	MMR deficiency (MSI)
			Large rearrangements
			CpG island methylation
MSH2	Lynch syndrome	Autosomal dominant	Point mutations	MMR deficiency (MSI)
			Large rearrangements
			CpG island methylation2
MSH6	Lynch syndrome	Autosomal dominant	Point mutations	MMR deficiency (MSI)
			Large rearrangements
PMS2	Lynch syndrome	Autosomal dominant	Point mutations	MMR deficiency (MSI)
			Large rearrangements
EPCAM	Lynch syndrome	Autosomal dominant	Large rearrangements2	MMR deficiency (MSI)
APC	(Attenuated) familial adenomatous polyposis	Autosomal dominant	Point mutations	-
		De novo mutations	Large rearrangements
		Mosaicisms	ASE (deep-intronic and promoter mutations)
MUTYH	MUTYH-associated polyposis	Recessive	Point mutations	Base excision repair deficiency: KRAS c.34G>T
			Large rearrangements
POLE	Polymerase proofreading-associated polyposis	Autosomal dominant	Point mutations (exonuclease domain)	Hypermutated: excess of G:C>T:A transversions
POLD1	Polymerase proofreading-associated polyposis	Autosomal dominant	Point mutations (exonuclease domain)	Hypermutated: excess of G:C>T:A transversions
GREM1	Hereditary mixed polyposis	Autosomal dominant	40-kb upstream duplication3	-
SMAD4	Juvenile polyposis	Autosomal dominant	Point mutations	-
			Large rearrangements
BMPR1A	Juvenile polyposis	Autosomal dominant	Point mutations	-
			Large rearrangements
STK11	Peutz-Jeghers	Autosomal dominant	Point mutations	-
			Large rearrangements
PTEN	PTEN hamartoma tumor4	Autosomal dominant	Point mutations	-
			Large rearrangements
			Promoter

¹Point mutations include missense, non-sense, frameshift and splice-site mutations and small intragenic deletions/insertions;

²MSH2 germline CpG island methylation occurs secondary to EPCAM deletions;

³Founder Ashkenazi mutation;

⁴PTEN hamartoma tumor syndrome includes Cowden, Bannayan-Riley-Ruvalcaba, PTEN-related Proteus and Proteus-like syndromes. ASE: Allele-specific expression; MMR: DNA mismatch repair; MSI: Microsatellite instability.

LS

LS (MIM No. 120435) is an autosomal dominantly inherited disorder caused by germline mutations or epimutations in a DNA mismatch repair (MMR) gene (MLH1, MSH2, MSH6 or PMS2).

Carriers of a heterozygous mutation (or epimutation) in a MMR gene are at high risk of developing CRC and at increased risk of developing malignancies at extracolonic sites such as the endometrium, ovary, stomach, small bowel, hepatobiliary tract, urinary tract, brain and skin[9]. A detailed description of the clinical and pathological features of LS is provided in Table 2. Of note, biallelic deleterious germline mutations in MMR genes lead to a constitutional mismatch repair-deficiency, a syndrome characterized by a broad spectrum of early-onset malignancies such as hematologic neoplasms and brain and LS-associated tumors and a phenotype that resembles the phenotype associated with neurofibromatosis type 1[10].

Table 2 Clinico-pathological characteristics of Lynch syndrome.

Clinico-pathological characteristics
The onset of colorectal cancer (CRC) occurs at an early age (average 45 yr)
Predilection to develop proximal (right-sided) colon cancer
High risk of multiple primary colorectal tumors (synchronous or metachronous)
Specific pathological features of lynch syndrome-related colorectal tumors:
Poorly differentiated
Mucinous
Signet-cell features
Crohn’s-like lymphocytic reaction
Excess of tumor-infiltrating lymphocytes
Increased survival (in patients with CRC)
Accelerated carcinogenesis
Increased risk of cancer at extracolonic sites:
Endometrium
Ovary
Stomach
Small bowel
Hepatobiliary tract
Pancreas
Upper uroepithelial tract
Brain (Turcot’s syndrome)
Sebaceous adenomas, carcinomas and keratoacanthomas (Muir-Torre syndrome)

Mismatch repair genes behave like tumor suppressors; cancer arises when a second hit (mutation, deletion or CpG island methylation) somatically inactivates the wild-type allele in a target cell (e.g., a cell of the colonic epithelium)[11-14]. The complete inactivation of the corresponding MMR gene in the tumor causes a marked reduction in MMR function, which results in microsatellite instability (MSI)[15,16].

The identification of MMR gene mutation carriers is critical for improving cancer surveillance and the effectiveness of preventive measures[17,18]. Before MMR genes and their causal role in hereditary CRC cancer were identified, the International Collaborative Group on hereditary non-polyposis colorectal cancer established the Amsterdam criteria in 1990. These criteria, the first clinical criteria used to define hereditary non-polyposis colorectal cancer, were used to identify families for research studies[19] and subsequently modified (Amsterdam II) to include extracolonic LS-related cancers[20]. However, the Amsterdam criteria failed to identify a large portion of MMR gene mutation carriers[21,22]. The Bethesda guidelines, which were less restrictive and had a sensitivity greater than 90% but a specificity of only 25%, were later defined[23,24].

Tumor testing is used to enhance the predictive power of clinical selection features and to identify the genes most likely to have a causative germline mutation. Standard tumor testing for LS involves the study of MSI and/or immunohistochemistry to detect the protein products expressed by MMR genes. However, because 10%-15% of sporadic CRCs also exhibit MSI[21,25-28], the detection of somatic MLH1 promoter methylation and somatic BRAF V600E mutations in patients with a MMR deficiency could help identify tumors that are more likely to be sporadic[29]. If the results of these tests suggest a diagnosis of LS, then germline molecular genetic testing of MMR genes is performed. The National Comprehensive Cancer Network has established unified CRC screening strategy guidelines (http://www.nccn.org).

In recent years, the concept of population-based universal screening for LS has gained strength among researchers and clinicians. The identification of individuals who are at increased risk of hereditary cancer allows for the possibility of specialized surveillance and early cancer detection, potentially resulting in decreased disease-specific mortality[30]. Because the prevalence of LS in the population is relatively high (approximately 3% of all diagnosed cases of CRC) and because surveillance strategies aimed at cancer prevention and early detection in LS patients have proven benefits[17], there is a clear rationale for exploring universal LS screening at the population level. Moreover, universal screening for LS is feasible, as LS tumors exhibit MMR deficiencies that can easily be identified with a simple PCR-based assay for MSI or by immunohistochemistry to identify the loss of expression of a MMR protein. Several studies have demonstrated the feasibility of this approach from a research and a clinical perspective[31-34]. In fact, in 2009, the Evaluation of Genomic Applications in Practice and Prevention recommended that all patients newly diagnosed with CRC be screened for LS through PCR-based MSI testing or immunohistochemistry[35]. However, at the population level, significant challenges and barriers to the successful implementation of this screening process exist[36].

MLH1 and MSH2 germline mutations, MSH6 mutations and PMS2 mutations account for approximately 90%, 7%-10% and less than 5% of mutations in families with LS, respectively[37-40]. Germline deletions in EPCAM that inactivate MSH2 (via the methylation of CpG islands) occur in approximately 1% of LS cases[41,42]. Finally, the constitutional inactivation of MLH1 by CpG island hypermethylation also causes Lynch syndrome; for this reason, MLH1 promoter methylation screening could be useful in individuals who have experienced a loss of MLH1 expression in their tumors and who have a negative germline sequence screen[43-45]. Large deletions and genetic rearrangements account for 20%, 5%, 20%, 7% and 100% of mutations in MSH2, MLH1, PMS2, MSH6 and EPCAM, respectively[40,46-49].

In some populations, recurrent mutations (i.e., those occurring repeatedly de novo) or ancestral (founder) mutations can change the aforementioned proportions; preliminary screening for these mutations can facilitate the molecular diagnosis of LS[50-54].

For years, researchers have sought to identify genetic modifiers that could affect the risk of cancer in MMR gene mutation carriers to explain the high variability in individual cancer risk among carriers. Identifying these modifying factors can enable an efficient stratification of mutation carriers based on their predicted risk and thereby offer a more appropriate clinical management strategy based on personalized surveillance programs. Initial attempts to identify cancer risk modifiers in patients with LS were based on the study of candidate genes; however, few of these studies were validated in larger-sized populations[55]. With the arrival of genome-wide association studies (GWAS), researchers hypothesized that the common variants associated with the risk of CRC in the general population could modify cancer risk in LS families. This hypothesis was verified in the case of rs16892766 (8q23.3) and rs3802842 (11q23.1) in MLH1 mutation carriers[6-8]. Similarly, we recently identified an association between the presence of a common variant in the telomerase gene (hTERT rs2075786) that causes shorter telomeres and an increased risk of developing LS-related tumors at a young age (< 45 years) in two independent series of patients with LS[56]. The identification of additional modifying factors will enable the estimation of individualized cancer risks that can be used to deliver tailored clinical surveillance protocols to mutation carriers.

Familial adenomatous polyposis

Familial adenomatous polyposis (FAP; MIM No. 175000) is the second most common inherited CRC syndrome. In its classic form, FAP is an autosomal dominantly inherited disease characterized by the development of hundreds to thousands of colorectal adenomatous polyps after the first decade of life. FAP is estimated to have a prevalence of 2-3 per 100000 individuals and to account for 0.2%-1% of all CRCs[57-60]. If left untreated, the classic form of FAP results in nearly complete penetrance of CRC by the age of 50 years. FAP is usually classified into classic and attenuated FAP (AFAP) depending on the number of polyps detected. A summary of the clinical characteristics of FAP is shown in Table 3[61,62].

Table 3 Clinical characteristics of familial adenomatous polyposis.

Clinical characteristics
Hundreds to thousands of colonic adenomatous polyps (on average beginning at age 16 yr)1
Colorectal cancer (100% penetrance if not treated; average age 39 yr)1
Other gastrointestinal polyps and malignant lesions:
Fundic gland polyps in the stomach
Adenomatous polyps in the stomach and small bowel
Periampullary carcinoma
Duodenal cancer
Congenital hypertrophy of the retinal pigmented epithelium (CHRPE)
Other less common manifestations:
Embryonal tumors (hepatoblastoma and medulloblastoma)
Pancreatobiliary carcinoma
Papillary thyroid carcinoma (especially cribriform-morular variant)
Adrenal cortical tumors
Gardner syndrome subtype (specific characteristics):
Colonic adenomatous polyposis
Desmoid tumors
Epithelial inclusion cysts
Osteoid osteomas
Supernumerary and/or impacted teeth
CHRPE
Turcot syndrome subtype (specific characteristics):
Colonic adenomatous polyposis
Tumors of the central nervous system (medulloblastoma)

¹AFAP: Patients have 10-100 colorectal adenomas. Polyps develop preferentially in the proximal colon, and the onset of colorectal cancer (CRC) occurs 10-15 years later than in patients with classic familial adenomatous polyposis. The cumulative risk of CRC by age 80 years is estimated to be approximately 70%.

Most classic FAP cases arise as a consequence of a germline heterozygous mutation in adenomatous polyposis coli (APC), a gene located on chromosome 5q21. All individuals who carry a germline pathogenic mutation in the APC gene (the first hit according to Knudson’s two-hit hypothesis) eventually develop FAP. As is the case for other tumor suppressor genes, tumor development requires the somatic inactivation of the wild-type allele. Given that thousands of adenomas can form within 15-40 years, it is likely that only two hits are necessary for the initiation of tumorigenesis; however, given that only one or a few of these adenomas progress to cancer, it is likely that several additional mutations are needed[63,64].

In most cases of FAP, the APC mutation is inherited in an autosomal dominant manner; however, in 15%-20% of cases, the APC mutation appears to arise de novo (i.e., spontaneously). Patients with these types of mutations therefore do not present with a family history of the disease[65]. However, approximately 20% of individuals with an apparent de novo APC mutation appear to have somatic mosaicism[66].

A truncating germline APC mutation that constitutively activates the Wnt pathway can be detected in approximately 80% of classic FAP cases[67-69], whereas fewer than 30% of individuals with attenuated phenotypes carry an identifiable APC mutation[70]. Approximately 90% of mutations are detected by sequence analysis (small intragenic deletions/insertions and missense, nonsense or splice-site mutations); the remaining 8%-12% consist of whole or partial gene deletions[67-69,71,72]. Moreover, interstitial deletions of chromosome 5q22 that also delete APC have been reported in individuals with the classic and attenuated forms of FAP. These individuals often have dysmorphic features and mild-to-moderate cognitive impairments[73,74]. No germline epimutations (CpG island methylation) have been identified in the APC gene[75].

It is widely accepted that the methods used to identify mutations fail to detect certain mutations because of factors such as polymorphisms in the sequences to which PCR primers bind that lead to allele dropout, or due to somatic mosaicism or because the mutations occur in regions not targeted by the currently used methods. Castellsagué et al[76] and Spier et al[77] reported the occurrence of imbalanced allele-specific expression of APC in 8%-9% of APC/MUTYH mutation-negative polyposis cases, indicating that the underlying mutations were not detected by standard mutation detection techniques. Some of these cases carried pathogenic deep intronic variants predicted to activate cryptic splice sites[77], whereas others carried mutations in the promoter region of APC[78].

If no disease-causing APC mutation is found, molecular genetic testing of MUTYH, POLE and POLD1 (exons coding for the exonuclease domain) should be considered (more information on MUTYH- and polymerase proofreading-associated polyposes can be found in the corresponding sections of this review).

In FAP, mutations in certain codons or regions are associated with specific phenotypic features. For example, profuse polyposis (which corresponds to an average of 5000 polyps) has been reported to be associated with mutations in codons 1250-1464[79]; AFAP is associated with mutations in the 5’ region of the gene (codons 1-177), in exon 9 and in the 3’ region of the gene[80-84]. AFAP has also been associated with interstitial deletions of chromosome 5q22 that also delete APC[73] and with somatic mosaicism for APC mutations that are generally associated with classic FAP[66,85,86]. APC mutations that cause Gardner’s syndrome typically occur in the region between codons 1403 and 1578. Additionally, certain genotypes have been found to be associated with extracolonic manifestations of the disease[87].

MUTYH-associated polyposis

Bi-allelic (homozygous or compound heterozygous) mutations in the MUTYH gene, which encodes a base excision repair protein, are responsible for certain cases of adenomatous polyposis. MUTYH-associated-polyposis (MAP; MIM No. 608456) represents the first known polyposis syndrome with a recessive pattern of inheritance; therefore, the disease is theoretically restricted to one generation.

Because of the variability in clinical features observed among mutation carriers, the diagnosis of MAP based on clinical findings alone remains difficult. Two thirds of MAP patients have CRC at the time of diagnosis, and up to one third of patients have CRC but no polyps. Most MAP patients have < 100 adenomas at diagnosis and a mean age of 45 years; these patients tend to develop CRC at a mean age of 50 years[88-92]. Other features variably present in MAP include: duodenal polyps and cancer; gastric fundic gland polyps; gastric, ovarian, bladder, breast or endometrial tumors; benign and malignant tumors of the skin and thyroid gland; dental abnormalities (jaw-bone cysts); and CHPRE[93-97]. Because the phenotypes associated with MAP are highly variable, a wide spectrum of clinical characteristics should be considered in patients with suspected MAP (Table 4)[98,99].

Table 4 Clinical characteristics of individuals with suspected MUYTH-associated polyposis.

Clinical characteristics
One to ten colonic adenomas before 40 yr of age
Tens to hundreds of colonic adenomas and/or hyperplastic polyps
Colonic polyposis (i.e., > 100 colonic polyps) in the absence of a germline APC mutation
Colorectal cancer with the somatic KRAS mutation c.34G>T in codon 12
Family history of colon cancer (with or without polyps) consistent with autosomal recessive inheritance

The definitive diagnosis is confirmed by the presence of a biallelic MUTYH mutation. APC: Adenomatous polyposis coli.

The MUTYH protein is a base excision repair glycosylase involved in repairing one of the most frequent and stable forms of oxidative damage, namely the oxidation of a guanine leading to the formation of 8-oxo-7, 8-dihydro-2’-deoxyguanosine (8-oxoG). When an oxoG:A mismatch is present in the DNA-template, a G:C to T:A transversion occurs in the subsequent round of replication[100]. For this reason, G:C to T:A transversions frequently occur in MUTYH-associated adenomas and tumors. One such transversion in the KRAS gene (c.34G>T in codon 12) is frequently encountered (64%) in patients with MAP CRC. Therefore, the analysis of somatic KRAS has been recommended as a pre-screening test to identify CRC patients eligible for MUTYH germline molecular genetic testing[98,101,102].

Colorectal tumors that develop in the context of a MUTYH mutation have specific molecular and histological features that differentiate these tumors from sporadic tumors and that overlap with features of hereditary (LS) and sporadic MSI tumors. These features include a preferential proximal location, a mucinous component and the increased presence of tumor infiltrating lymphocytes. However, only a minority (range: 0%-18%) of MUTYH-associated tumors exhibit MSI. All of these features raise the suspicion of a MAP etiology for the CRC, especially when the disease is diagnosed at a young age and when polyps and/or a recessive inheritance pattern are detected[103].

Approximately 30% of APC mutation-negative cases of polyposis harbor bi-allelic mutations in the MUTYH gene. At least one of two MUTYH missense mutations found in 1%-2% of the general population [c.536A>G (p.Tyr179Cys) and c.1187G>A (p.Gly396Asp), annotated according to the longest (hypothetical) coding sequence NM 001128425.1] is present in approximately 90% of MAP patients in the western part of the world; a biallelic status for one and/or the other variant was found in up to 70% of Caucasian patients with MAP[98]. Additional common mutations that were most likely founder mutations have been reported in different populations: the c.1147delC (p.Ala385Profs*23) mutation was reported in northern European MAP patients, the c.1214C>T (p.Pro405Leu) mutation was reported in Dutch MAP patients, the c.1437_1439del (p.Glu480del) mutation was reported in Italian MAP patients, the c.1438G>T (p.Glu480*) mutation was reported in British Indian MAP patients, the p.Tyr104* mutation was reported in Pakistani MAP patients, the c.1227_1228dup (p.Glu410Glyfs*43) mutation was reported in Spanish, Portuguese and Tunisian MAP patients and the p.Ala359Val mutation was reported in Japanese and Korean MAP patients[90,104-111]. The presence of recurrent mutations facilitates genetic testing for MUTYH, thus allowing for an initial screening of the common mutations found in the corresponding population.

Polymerase proofreading-associated polyposis

DNA polymerase ε (POLE) and δ (POLD1) mutations have recently been identified in patients with familial CRC, many of whom have multiple adenomas[112].

The two germline mutations POLE p.Leu424Val and POLD1 p.Ser478Asn were detected in individuals with multiple colorectal adenomas and CRC. An additional variant of POLD1, p.Pro327Leu, the pathogenicity of which has not yet been determined, was also identified in a multiple adenoma patient[112]. The two pathogenic mutations are characterized by a dominant pattern of inheritance and associated with a high risk of multiple colorectal adenomas, large adenomas, early-onset CRC and multiple CRCs. POLD1 mutations are also associated with an increased risk of endometrial cancer in female carriers[112,113]. A recent study performed by our group identified a de novo POLE p.L424V mutation in patient with adenomatous polyposis and early onset CRC, and a novel pathogenic mutation in POLD1, p.L474P, in a non-polyposis Amsterdam II family without MMR defects[113]. Based on these findings, the term “polymerase proofreading-associated polyposis” may be misleading and should be carefully used, at least until more POLE/POLD1 families are described and the full phenotypic spectrum of this syndrome is defined.

All germline mutations identified thus far in POLE and POLD1 are located within the proofreading (exonuclease) domain of the respective polymerase, suggesting a deficient ability to proofread and repair errors during DNA replication[112,114-116]. Non-exonuclease domain POLE and POLD1 mutations do not appear to be associated with familial CRC. Mutations in non-exonuclease domain regions have been identified in colorectal and endometrial tumors that are mostly MSI-positive; however, these mutations appear to be passenger mutations[116].

Tumors that develop in patients with polymerase proofreading-associated polyposis and sporadic colorectal and endometrial tumors with POLE mutations (somatic POLD1 mutations are rare) are hypermutant and microsatellite-stable. These hypermutated tumors have approximately 5000 somatic base substitutions in their coding regions and an altered mutation spectrum characterized mostly by increased proportions of G:C→T:A and A:T→C:G transversions[116]. As is the case with microsatellite instability in LS, a feasible molecular approach for identifying hypermutated tumors in patients with polyposis should be developed to facilitate the selection of cases suspected of carrying germline polymerase proofreading mutations.

Although the phenotypes associated with POLE and POLD1 mutations vary among carriers, the evidence gathered so far supports the recommendation of the sequencing of the exonuclease domains of POLE and POLD1 for genetic testing purposes.

Hereditary mixed polyposis syndrome

Hereditary mixed polyposis syndrome (HMPS; MIM No. 601228) is an unusual disease characterized by the apparent autosomal dominant inheritance of multiple types of colorectal lesions (including Peutz-Jeghers polyps, juvenile polyps, serrated lesions, conventional adenomas and CRC) and a lack of extracolonic manifestations[117].

Linkage studies conducted in large families identified CRAC1 on chromosome 15q13.3 as the candidate region that causes HMPS[118,119]. Moreover, families of Ashkenazi descent with hereditary mixed polyposis syndrome shared a disease haplotype in the CRAC1 region[120]. The sequencing of the shared region did not yield useful results; however, the study of copy number alterations revealed the presence of a heterozygous single-copy duplication of a region approximately 40 kb in length that co-segregated with the disease. The duplication extended from intron 2 of SCG5 to a site immediately upstream of the GREM1 CpG island. The SCG5-GREM1 duplication increased the transcription of GREM1, a gene that encodes the secreted BMP antagonist[121]. No non-Ashkenazi affected individuals with duplications in the region implicated in HMPS have yet been identified.

Hamartomatous polyposis syndromes

Juvenile polyposis: Juvenile polyposis syndrome (JPS; MIM No. 174900) is the most common hamartomatous syndrome, with an estimated incidence of one per 100000[122].

The diagnosis of JPS is made when any of the following three criteria is met[123-125]: (1) the patient has multiple (3-10) juvenile polyps of the colorectum; (2) a patient with a familial history of JPS has any number of juvenile polyps; or (3) the patient has extracolonic (e.g., in the stomach or small intestine) juvenile polyps.

Juvenile polyps are hamartomas that have a normal epithelium with a dense stroma, an inflammatory infiltrate and a smooth surface with dilated, mucus-filled cystic glands in the lamina propria. Most juvenile polyps are benign; however, malignant transformation can occur. Members of families with JPS have an estimated lifetime risk of developing gastrointestinal cancer of 9% to 50%[126]. Most of these cancers consist of colon cancer; however, cancers of the stomach, upper gastrointestinal tract and pancreas have also been reported[122].

Germline mutations in SMAD4 or in BMPR1A have been identified in approximately 40% of JPS patients[127,128]. Both genes encode proteins involved in the TGF-beta signaling pathway, an important modulator of many cellular processes.

JPS patients with mutations in the SMAD4 gene are predisposed to developing massive gastric polyps and usually have a family history of upper gastrointestinal polyposis[127,129]. A large proportion of JPS patients with SMAD4 mutations have a juvenile polyposis/hereditary hemorrhagic telangiectasia overlap syndrome (MIM No. 175050). Hereditary hemorrhagic telangiectasia is a dominant disorder characterized by epistaxis, visceral arteriovenous malformations and telangiectasias[130].

Sweet et al[131] found two rare germline variants of ENG, a gene associated with a predisposition to hereditary hemorrhagic telangiectasia, in two JPS patients with no symptoms of hemorrhagic telangiectasia. Subsequent studies in other JPS patients did not identify deleterious ENG mutations in genetically uncharacterized JPS patients[132-134]. PTEN mutations have been identified in JPS patients[134]; however, it has been suggested that these patients were clinically misclassified and most likely belonged to the PTEN hamartoma tumor group[135]. Microdeletions at 10q22-q23, a region that includes both PTEN and BMPR1A, have also been reported[136].

Peutz-Jeghers syndrome: Peutz-Jeghers syndrome (PJS; MIM No. 175200) is an autosomal-dominant condition caused by germline mutations in STK11 (formerly known as LKB1), which encodes a serine-threonine kinase. PJS is clinically characterized by the occurrence of gastrointestinal polyposis and mucocutaneous pigmentation and a predisposition to develop cancer.

The presence of PJS-type hamartomatous intestinal polyps is required for a clinical diagnosis of PJS, even though these patients also develop other types of polyps, including colonic adenomatous polyps or gastric PJS polyps that resemble hyperplastic polyps. PJS-type polyps have characteristic histological features, including a notable frond-like, elongated epithelial component, cystic gland dilatation extending into the submucosa or muscularis propria and arborizing smooth muscle extending into the polyp fronds. These polyps are found throughout the gastrointestinal tract but occur predominantly in the small intestine and colon[137].

Cutaneous lesions found in patients with PJS include small melanocytic macules on the labial mucosa, lips, palate and tongue, around the eyes and nostrils and in the perianal region. Hyperpigmented macules on the fingers are also common. Mucocutaneous pigmented lesions, which usually develop in childhood, are found in 95% of patients with PJS[138,139].

A clinical diagnosis can be made when any of the following criteria are fulfilled[137]: (1) the patient has two or more histologically confirmed PJS-type hamartomatous polyps; (2) a patient with a family history of PJS has any number of PJS-type polyps; (3) a patient with a family history of PJS has characteristic mucocutaneous pigmentation; (4) a patient with characteristic mucocutaneous pigmentation has any number of PJS-type polyps; and (5) the patient has a pathogenic mutation in STK11.

The risk of gastrointestinal and extraintestinal malignancies, including duodenal, colon, breast, pancreas, stomach, small bowel, cervix, uterus, ovary, testes, and thyroid tumors, is significantly increased in patients with JPS[140-143]. Benign and malignant gonadal and gynecologic tumors, including ovarian sex cord tumors with annular tubules, mucinous tumors of the ovaries and fallopian tubes and large-cell calcifying Sertoli cell tumors of the testes, can also be observed in these patients[144].

The clinical manifestations of JPS can vary; however, there are no reports of STK11 mutation carriers lacking clinical manifestations of the disease.

Aretz et al[145] reported that 100% of individuals with familial PJS have detectable STK11 mutations, whereas 91% of simplex cases (i.e., a single occurrence in a family) who met the relevant diagnostic criteria had a detectable mutation. Clinical misdiagnoses of PJS could account for the decreased rate of detection of mutations in simplex cases.

PTEN hamartoma tumor syndrome: The PTEN hamartoma tumor syndrome (PHTS, MIM No. 601728) comprises Cowden syndrome, Bannayan-Riley-Ruvalcaba syndrome, PTEN-related Proteus syndrome and Proteus-like syndrome. A presumptive diagnosis of PHTS is based on clinical signs; the definitive diagnosis of PHTS is, by definition, made only when a PTEN mutation is identified. The clinical characteristics of PHTS are shown in Table 5[146]. The specific clinical features of the four PHTS syndromes were reviewed by Orloff and Eng[147] and are not further described in this review.

Table 5 Clinical characteristics of the PTEN hamartoma tumor syndrome.

Clinical characteristics
Benign neoplasia
Dermatologic
Palmoplantar keratoses
Trichilemmomas
Lipomas
Fibromas
Freckling of the glans penis
Vascular anomalies/hemangiomas
Lhermitte-Duclos (dysplastic gangliocytoma of the cerebellum)
Genitourinary tumors/malformations
Colorectal polyposis
Mucosal lesions
Thyroid goiter/nodules
Proliferative breast changes
Malignant neoplasia
Breast cancer
Non-medullary thyroid cancer
Renal cancer
Endometrial cancer
Colorectal cancer
Melanoma
Central nervous system
Macrocephaly
Autism/developmental delay
Dysmorphic characteristics
Dolichocephaly
Postaxial polydactyly

The clinical phenotypes of PTEN mutation carriers are highly variable and range from macrocephaly and developmental delays (reported in a two-year-old patient) to a history of multiple primary neoplasias (reported in a 60-year-old patient)[146]. However, to date, no strong genotype-phenotype correlations have been reported[148]. Moreover, significant intra-familial phenotypic variability and overlapping mutation spectra have been observed[149].

The lifetime risks for a variety of cancers are increased in patients with PHTS; more specifically, these patients have an estimated lifetime risk of breast, thyroid, endometrial, renal cell and colon tumors and melanoma of 85%, 35%, 28% 34%, 9% and 6%, respectively[150,151].

Germline PTEN mutations have been identified in patients with autism/ pervasive developmental disorder and macrocephaly and to a particularly significant extent in patients with a personal or family history of Cowden or Bannayan-Riley-Ruvalcaba syndromes[152-156].

To determine whether germline methylation is found in patients with Cowden syndrome and in patients with a Cowden-like syndrome who lack germline PTEN mutations, Bennett et al[157] observed that germline methylation upstream of PTEN occurred in 42% and 33% of mutation-negative patients with Cowden syndrome and Cowden-like syndromes, respectively. This hypermethylation did not silence PTEN. However, a newly characterized tumor suppressor gene, KILLIN, the promoter of which overlaps with the PTEN 5’UTR and the 5’ end of its coding region, was silenced by this hypermethylation. This finding must be validated in other groups of patients before it can be used for diagnostic purposes.

Current challenges and future perspectives in genetically characterized inherited syndromes

One of the most significant challenges for researchers, clinicians and genetic counselors in treating or investigating hereditary CRC (and any other cancer syndrome) involves the assessment of the pathogenicity of variants of unknown or uncertain significance (VUS). Enormous efforts are currently being undertaken to establish gene-specific interpretation guidelines that can be made available to diagnostic laboratories, research laboratories, genetic counselors and clinicians worldwide. Some of these efforts have been conducted with the support and coordination of international societies or consortiums; for example, the International Society for Gastrointestinal Hereditary Tumors (InSiGHT), which is under the umbrella of the Human Variome Project, has attempted to classify MMR VUS (http://www.insight-group.org/variants/classifications/)[158].

When a pathogenic germline mutation is identified in a family, carriers of the mutation can benefit from increased surveillance and a more informed decision about preventive measures; at the same time, non-carriers do not have to undergo intensive (and, in the case of CRC, invasive) surveillance. In the absence of an identified pathogenic mutation, these individuals may decide to undergo preventive surgery based on family history alone. The effects of germline variants in many of the main cancer-related genes on protein function are unknown (VUS); as a result, it is difficult to make any inferences on the risk of cancer in patients with these variants. It has been estimated that up to 10% of Caucasians undergoing genetic testing have variants that are designated VUS, which leads to important issues in genetic counseling. Current attempts to classify these variants involve the use of data from co-segregation studies, in silico functional predictions, personal and family cancer history, the co-occurrence of these variants and pathogenic mutations, the frequency of these variants in the general population (controls), molecular characteristics of the tumors, effects on RNA (splicing, allele-specific expression) and in vitro functional consequences, which, in the case of MMR genes, include impairment of MMR activity and the abnormal subcellular localization and abrogation of the formation of physiological dimers[158-164].

The coupling of next-generation sequencing technologies with genomic sequence enrichment methods has made the sequencing of comprehensive panels of cancer-predisposing genes technically feasible; consequently, this approach has become cost-effective for diagnostic applications and can be used to overcome the issue of syndrome-overlapping genes and gene-overlapping syndromes. However, the more frequent use of this approach for genetic diagnostic purposes will result in an exponentially increased number of identified VUS and a more urgent need to classify these VUS to their highest level.

Another challenge for the coming years involves the identification of additional cancer risk modifiers, including environmental and genetic factors. Identifying these factors for syndromes with incomplete cancer penetrance will facilitate an accurate individual risk assessment that will enable the application of personalized surveillance protocols and preventive measures. For example, the importance of individual risk assessments is supported by the extreme heterogeneity in CRC risk in carriers of MMR gene mutations. Dowty et al[165] studied 17500 family members of 166 MLH1-mutated and 224 MSH2-mutated families and showed that the cumulative risk of CRC by age 70 follows a U-shaped distribution. These authors also observed that 17% of male MSH2 mutation carriers have estimated lifetime risks of CRC of 0%-10% and that 18% of these carriers have lifetime risks of 90%-100%. If carriers who are at low risk of developing CRC can be distinguished from patients who will definitely develop a tumor, subsequent cancer surveillance strategies could be applied accordingly.

The existence of genetic anticipation in cancer syndromes and the mechanisms that might explain this phenomenon have been studied and discussed for years. In LS, despite the numerous reports and clinical observations identifying anticipation in the age of cancer onset in successive generations, it is still unclear whether true genetic anticipation contributes to the early diagnosis of LS. More recently, methods that correct for random effects, that isolate the confounding effect of changes in secular trends, screening and medical practices and that adjust for changes in age-specific incidence across birth cohorts, appear to confirm the presence of this phenomenon in families with LS[166-168]. However, the molecular mechanism underlying this phenomenon has not yet been identified. Telomere shortening, the accumulation of mismatch repair slippage events in subsequent generations and environmental factors have been suggested as causative mechanisms of anticipation[169,170]. Our group recently ruled out telomere length attrition as the cause of anticipation in patients with LS[171]. Anticipation has also been observed in patients with hereditary non-polyposis CRC without MMR deficiency[172].

LOW-PENETRANCE LOCI IDENTIFIED BY GWAS IN HEREDITARY CRC

GWAS conducted since 2007 using samples from the general population and common genetic markers (SNPs) have successfully identified low-penetrance loci associated with CRC. To date, at least 21 independent loci have been conclusively associated with the risk of CRC in Caucasians (P < 5.0 × 10^-8, and these associations were confirmed in independent case-control series) (source: http://www.genome.gov/gwastudies). These loci include 1q25 (LAMC1)[173], 1q41 (DUSP10)[174], 2q32.3 (NABP1)[173], 3q26.2(MYNN)[174], 5q21[175], 6p21.2 (CDKN1A)[176], 8q23.3 (EIF3H)[177], 8q24.21 (c-MYC)[173,177-181], 10p14[177], 11q13.4 (POLD3)[176], 11q23.1[180], 12p13.3 (CCND2)[173], 12q13.13 (DIP2B, ATF1)[174], 14q22.2 (BMP4)[182], 15q13.3[181], 16q22.1 (CDH1)[182], 18q21.1 (SMAD7)[173,183], 19q13.11 (RHPN2)[182], 20p12.3[182], 20q13.33 (LAMA5)[174] and Xp22.2 (SHROOM2)[176].

There is almost no evidence of interactive effects among these loci. However, the distribution of alleles associated with a risk of CRC follows a normal distribution in both cases and controls, with a shift towards higher numbers of these alleles in cases, which is consistent with a polygenic model of disease predisposition. It has been estimated that individuals carrying a large number of these alleles have an approximately threefold higher risk of developing CRC than those with a median number of these alleles[184]. Data suggest that only a small proportion (at most 10%) of the heritability associated with CRC can be explained by the identified loci[174,182,184-188].

OTHER RISK VARIANTS

In addition to the low-risk variants identified by GWAS, numerous genetic variants (> 3500 variants in > 1300 genes) that are associated with a low-moderate risk of CRC have been identified through candidate-gene approaches. The results from these candidate-gene association studies are usually inconsistent and difficult to interpret. In an effort to comprehensively evaluate candidate-gene association studies for CRC, Ma et al[189] recently performed meta-analyses for variants included in at least three independent datasets (267 variants in 150 genes) and used Venice criteria and false-positive report probability tests to assess the evidence for true associations. A total of 67 variants in 50 genes were found to be significantly associated with a risk of CRC. The cumulative epidemiological evidence for a risk of CRC was strong, moderate and weak in eight, two and 52 of the variants, respectively. Table 6 shows the 10 variants with strong and moderate evidence of association and their estimated risks. The authors of this study suggested that these variants may explain approximately 5% of the familial cancer risk in Caucasians.

Table 6 Genetic variants identified using candidate-gene association studies.

Gene	Variant	Frequency in controls	OR (95%CI)	Cumulative evidence of association	Ethnicity
MUTYH	Biallelic mutation	0.01%	10.19 (5.0-22.0)	Strong	Caucasian
MUTYH	G382D (rs36053993)	0.00%	6.49 (2.6-10.4)	Strong	Caucasian
MUTYH	Y165C (rs34612342)	0.01%	3.32 (1.1-9.8)	Strong	Caucasian
APC	I1307K (rs1801155)	6.80%	1.96 (1.4-2.8)	Strong	Ashkenazi
CHEK2	1100delC	0.71%	1.88 (1.3-2.7)	Strong	Caucasian
CHEK2	I157T (rs17879961)	3.91%	1.56 (1.3-1.8)	Strong	Caucasian
MLH1	rs1800734 (promoter)	21.11%	1.51 (1.3-1.7)	Strong	Caucasian
DNMT3B	rs1569686 (promoter)	16.99%	0.57 (0.5-0.7)	Strong	All
GSTM1	Present/null	50.64%	1.10 (1.0-1.2)	Moderate	All
TERT	rs2736100 (intron 2)	49.34%	1.07 (1.0-1.1)	Moderate	Caucasian

Genetic variants identified using candidate-gene association studies, significantly associated with a risk of colorectal cancer in meta-analyses and showing strong and moderate cumulative evidence of association according to Venice criteria and false-positive report probability tests[189].

HEREDITARY CRC OF UNKNOWN ETIOLOGY

Despite recent developments in genotyping and sequencing technologies, the genetic etiology of several familial CRCs, including serrated polyposis (formerly known as “hyperplastic polyposis”) and hereditary non-polyposis CRC without a MMR defect (also known as familial CRC type X) remains unknown.

Serrated polyposis

SP is a rare condition characterized by multiple and/or large serrated colonic polyps and an increased risk of CRC. The diagnosis of serrated polyposis is made based on established clinical criteria[190] (Table 7). Patients with SP most likely consist of a heterogeneous group of patients with a variety of SP phenotypes that are most likely caused by different genetic alterations[191]. At least three different subgroups have been described: (1) a right-sided phenotype with large sessile serrated adenomas associated with early-onset CRC characterized by the presence of a BRAF mutation; (2) a left-sided phenotype with large amounts of small polyps characterized by the presence of a KRAS mutation; and (3) a mixed phenotype with features of phenotypes 1 and 2[191,192]. Conventional colonic adenomas have been identified in up to 80% of individuals with SP and are more frequently present in CRC-affected individuals with SP[193,194].

Table 7 Clinical criteria established for the identification of serrated polyposis.

Clinical criteria
At least five serrated polyps proximal to the sigmoid colon, two of which are larger than 10 mm in diameter
Any number of serrated polyps occurring proximally to the sigmoid colon in an individual who has a first-degree relative with serrated polyposis
More than 20 serrated polyps of any size distributed throughout the colon

Diagnosis is made when one of the criteria is fulfilled.

Reported case series indicate that 25%-70% of SP patients had CRC at the time of diagnosis or during follow-up. Additionally, 10%-50% of SP patients had a family history of CRC[191,195-199]. In fact, studies have reported a fivefold increase in the risk of CRC and a 3.5-fold increase in the risk of pancreatic cancer in first-degree relatives of individuals with SP[194,199,200].

Serrated polyps are the precursors of CRC tumors developed through the serrated neoplasia pathway, which is characterized by BRAF mutations and the CpG island methylator phenotype with or without MSI depending on whether MLH1 is methylated. Several subtypes of serrated polyps have been defined: hyperplastic polyps, sessile serrated adenomas and traditional serrated adenomas[201]. However, the majority of CRC tumors arising in patients with serrated polyps exhibit a diverse range of molecular profiles and generally do not harbor molecular hallmarks of tumors developed via the serrated pathway[202].

Although the genetic basis of SP is unknown, both recessive and dominant transmission patterns have been proposed[197,203,204]. Because serrated polyps were reported in individuals with biallelic mutations in PTEN, BMPR1A, SMAD4 and MUTYH or with a duplication in the GREM1 gene[205-208], Clendenning et al[209] hypothesized that these genes might be altered in individuals with SP and might account for some of the cases with this condition. However, no deleterious germline mutations were identified in a case series of 65 patients with SP.

It has been suggested that lifestyle factors such as smoking, obesity and diet, which have been associated with the presence of serrated polyps[210-212], could be responsible for SP or for the modification of the risk of disease in the presence of predisposing genetic mutations or risk variants.

If the cause of SP is genetic, current sequencing and genotyping technologies or methods that identify copy number alterations, structural variants or epigenetic modifications can be used to understand the etiology of this disease in the near future.

Familial CRC type X

Approximately 40% of the families meeting the Amsterdam criteria for a diagnosis of hereditary non-polyposis CRC lack evidence of heritable defects in the MMR system; more specifically, these patients have no germline mutations in the MMR genes, no tumor microsatellite instability and no loss of immunohistochemical staining of the MMR proteins. Because the genetic etiology of this disease is unknown, these families are said to have fCRC-X. As has been the case for other familial cancer syndromes, the identification of the genes associated with fCRC-X will facilitate the molecular diagnosis of the disease and the development of appropriate surveillance guidelines and clinical management protocols for these patients.

Familial CRC-X is clearly clinically different from Lynch syndrome; in particular, patients with familial CRC-X have a lower incidence of CRC and a lower risk of extracolonic tumors and tend to develop cancer at a later age[213-217]. Familial CRC-X tumors are characterized by the presence of microsatellite stability and chromosomal instability and the absence of high CpG methylator phenotypes; these characteristics overlap with some of the characteristics of sporadic MMR-proficient tumors. However, some molecular features specific to familial CRC-X tumors have been reported[218-221].

Significant but mostly unsuccessful efforts have been made to understand the genetic cause(s) of fCRC-X. Several dominant predisposition loci that have been mapped to different chromosomal regions such as 3q13.31-q27.1, 3q22, 4q21.1, 5q14-q22, 7q31, 8q13.2, 9q22.2-31.2, 10p15.3-p15.1, 12q24.32 and 13q22.1-13q31.3 have been identified using genome-wide linkage studies in families with CRC; however, no causal genes have yet been identified[222-230].

Despite the ability of whole-exome and whole-genome sequencing to uncover numerous new causal mutations and genes in Mendelian disorders, few such genes and mutations have been identified in hereditary cancer syndromes[112,231,232] and none have been identified in fCRC-X.

Current evidence indicates that families with fCRC-X constitute a very heterogeneous group. Because the Amsterdam criteria indicate that this disease is characterized by strong familial aggregation, it is likely that certain cases of fCRC-X are caused by high-penetrance mutations (i.e., that have a monogenic component). If this is the case, reports on new hereditary CRC genes identified by whole-exome or whole-genome strategies will likely be published in the near future. However, because no such genes have yet been identified, it is likely that any genes identified in the future would explain only a small number of fCRC-X cases.

In contrast, candidate-gene approaches have identified several high-penetrance genes that might be involved in the etiology of uncharacterized familial CRC[233-237].

It is likely that most of the familial aggregation observed in fCRC-X is associated with non-genetic factors. Lifestyle and environmental factors could interact with multiple genetic risk factors to increase the risk of CRC in these families. This scenario is consistent with a multifactorial disease model associated with polygenic diseases and supported by the less aggressive clinical characteristics of fCRC-X (e.g., the late onset of the disease, the lower risk of CRC and the almost complete absence of multiple primary tumors).

Because some families appear to fit the monogenic model and others the polygenic model, finding the optimal approach for exploring the genetic basis of fCRC-X remains challenging[238]. The selection criteria used to identify patients with fCRC-X will need to include the Amsterdam criteria, a very early onset of cancer and severe clinical manifestations; these criteria and insightful data analyses will play a key role in determining the ability of exome sequencing to identify rare and deleterious mutations within gene-coding regions. Furthermore, it is possible that other mechanisms of gene silencing such as germline epigenetic or copy number alterations or the deregulation of tumor suppressor genes via regulatory non-coding RNAs such as microRNAs could be associated with the hereditary forms of the disease.

Elucidating the polygenic component of the disease will also remain challenging. GWAS of patients with fCRC-X would be very useful; however, collecting an adequate number of samples (thousands of samples) is almost impossible given the rarity of fCRC-X, even if samples were to be collected worldwide. A closely related approach based on the hypothesis that variants associated with the risk of CRC in the general population are also associated with the risk of CRC in fCRC-X involves genotyping the population-based GWAS CRC risk variants in a large fCRC-X cohort. GWAS have already provided evidence suggesting that low-penetrance alleles may explain the risk of cancer in familial cases[187,188,239]. Sequencing the loci identified by GWAS to identify common and rare variants in patients with fCRC-X therefore represents an alternative approach. This approach has been successfully used in other diseases such as hypertriglyceridemia, diabetes and inflammatory bowel disease[240-243].

In summary, the genetic basis of fCRC-X will become clearer when all of the approaches mentioned above are applied in practice. However, alternative mechanisms involving gene-gene and gene-environment interactions, epigenetic and structural alterations and other non-classic gene silencing mechanisms might explain fCRC-X cases that are not detected by current risk variant or mutation-identification techniques.

ACKNOWLEDGMENTS

I thank Marta Pineda for critical review of the paper.