Updates in the field of hereditary nonpolyposis colorectal cancer

ABSTRACT

Introduction

Up to one third of colorectal cancers show familial clustering and 5% are hereditary single-gene disorders. Hereditary non-polyposis colorectal cancer comprises DNA mismatch repair-deficient and -proficient subsets, represented by Lynch syndrome (LS) and familial colorectal cancer type X (FCCTX), respectively. Accurate knowledge of molecular etiology and genotype-phenotype correlations are critical for tailored cancer prevention and treatment.

Areas covered

The authors highlight advances in the molecular dissection of hereditary non-polyposis colorectal cancer, based on recent literature retrieved from PubMed. Future possibilities for novel gene discoveries are discussed.

Expert commentary

LS is molecularly well established, but new information is accumulating of the associated clinical and tumor phenotypes. FCCTX remains poorly defined, but several promising candidate genes have been discovered and share some preferential biological pathways. Multi-level characterization of specimens from large patient cohorts representing multiple populations, combined with proper bioinformatic and functional analyses, will be necessary to resolve the outstanding questions.

KEYWORDS:

• Hereditary non-polyposis colorectal cancer
• lynch syndrome
• familial colorectal cancer type X
• DNA mismatch repair
• genomic instability
• germline mutation
• constitutional epimutation

1. Introduction: lynch syndrome

1.1. MLH1, MSH2, MSH6, and PMS2 mutations in lynch syndrome predisposition

Lynch syndrome (LS) is an autosomal dominant disorder caused by a defect in one of the DNA mismatch repair (MMR) genes MLH1, MSH2, MSH6, or PMS2 [1]. It represents the MMR-deficient subgroup of hereditary non-polyposis colorectal cancer (HNPCC) (please see section 2.1. below for the evolution of nomenclature). LS is one the most prevalent hereditary cancer syndromes in man, with 1:226 individuals as the highest reported population frequency to date [2]. LS accounts for 1–3% of unselected colorectal or endometrial carcinomas and ~15% of those with high-degree microsatellite instability (MSI) or absent MMR protein in tumor tissues [3].

LS-associated MMR genes are conserved in evolution (Table 1). A specific clinical phenotype or syndrome exists for all known human MMR genes, except for MSH4 and MSH5 whose protein products are necessary for meiotic (and possibly mitotic) recombination but do not participate in MMR [4], and PMS1 whose product forms a complex with MLH1 but no specific function has been identified for this complex (hMutLβ) to date [5]. Of the over 3000 unique germline sequence variants of the four LS-associated MMR genes deposited in the InSiGHT locus-specific database, the relative shares are 40% for MLH1, 34% for MSH2, 18% for MSH6, and 8% for PMS2 [6]. MLH1 and MSH2 proteins are obligatory components in all types of heterodimers that the MMR proteins form with each other [5] which explains the predominant roles of MLH1 and MSH2 in LS predisposition. MSH6 is functionally redundant with MSH3 when these proteins form a complex (hMutSα and hMutSβ, respectively) with MSH2 for mismatch recognition. A MutL homolog is needed to couple mismatch recognition to the downstream events. In humans, the prevalent MutL homolog is a heterodimer of MLH1 and PMS2 (hMutLα), but MLH1 can also heterodimerize with MLH3 (hMutLγ) which is able to partially compensate for the lack of hMutLα in MMR [5].

Table 1. LS-associated MMR genes in evolution and phenotypes resulting from germline mutations.

E. coli	S. cerevisiae	H. sapiens	Phenotype resulting from germline mutation
MutS	Msh2	MSH2	LS (monoallelic), CMMRD (biallelic)
	Msh6	MSH6	LS (monoallelic), CMMRD (biallelic)
	Msh3	MSH3	Attenuated adenomatous polyposis/CMMRD (biallelic)
	Msh1	Not identified	N/A
	Msh4	MSH4	No LS-associated mutations known
	Msh5	MSH5	No LS-associated mutations known
MutL	Mlh1	MLH1	LS (monoallelic), CMMRD (biallelic)
	Pms1	PMS2	LS with reduced penetrance (monoallelic), CMMRD (biallelic)
	Mlh2	PMS1	No LS-associated mutations known
	Mlh3	MLH3	LS? (monoallelic), adenomatous polyposis (biallelic)
MutH	Not identified	Not identified	N/A

N/A, not applicable

In comparison with the average population, heterozygous carriers of MMR gene mutations have significantly increased risks of cancers of the colon and rectum, endometrium, ovary, kidney, and urinary tract, upper gastrointestinal tract, and certain other organs. Compared to previous retrospective and family-based studies, recent prospective investigations have arrived at generally lower risk estimates for cancers occurring in LS, after stratification of cancer risk by the affected MMR gene, age, and gender [7]. Penetrance and expression patterns strongly depend on the MMR gene involved so that germline mutations of MLH1 and MSH2 have the highest and PMS2 the lowest penetrance, whereas MSH6 mutations underlie a sex-limited trait with a high risk of gynecological cancers in females [7]. Evolutionary pathways to cancer may vary (for example, colorectal cancer development in LS may or may not involve a visible polyp precursor), and differential somatic involvement of key cancer-relevant genes (such as APC and CTNNB1) may in part explain differences in (colon) tumor development and penetrance between the four LS-associated MMR genes [8,9].

Germline mutations in MMR genes are detectable in up to 88% of families strongly suspected of having LS based on the fulfillment of the Amsterdam criteria and presence of MSI in tumors [10,11]. Mutation frequencies in smaller or atypical families and families not pre-screened by MSI or immunohistochemical analyses vary between 10% and 40% depending on the method of ascertainment [10,12]. The detection rate of MMR gene mutations in families harboring such mutations is anticipated to increase when deep sequencing by targeted gene panels, or even whole exomes or genomes, will gradually replace single gene-based approaches. On the other hand, these new sensitive and comprehensive sequencing methods are likely to detect increasing numbers of variants of uncertain significance (VUS) which cannot be directly translated to clinical management [13,14]. Furthermore, panels that include predisposition genes for multiple cancer syndromes may broaden genotype-phenotype associations within the syndromes. For example, a retrospective review of clinical histories in patients who underwent multigene panel testing and were diagnosed with a germline mutation in one of the MMR genes, showed the predominance of MSH6 and PMS2 over MSH2 and MLH1 mutations and linked a hereditary breast and ovarian cancer phenotype to MSH6 and PMS2 mutations [14].

1.2. Constitutional epimutations in LS predisposition

In cases suspected of having LS but with no identifiable coding region mutations, constitutional epimutations are worthy of consideration. Constitutional epimutation refers to constitutional hypermethylation at the promoter of one allele of a given (non-imprinted) gene leading to silencing of expression from that allele in all main somatic tissues. Epimutation can be induced by a genetic alteration (usually located in cis), in which case epimutation is secondary, whereas epimutation arising in the absence of any detectable genetic change is considered primary. For MLH1, primary and secondary epimutations are known, whereas constitutional epimutations of MSH2 are secondary to deletions of the 3ʹ end of the EPCAM gene (Table 2). After removal of stop codon, transcription of EPCAM reads into the adjacent, structurally normal MSH2 gene inducing its promoter methylation [15]. Owing to the capability of MSH2 silencing, EPCAM is considered as a LS predisposition gene in addition to the four LS-associated MMR genes.

Table 2. Constitutional epimutations of MLH1 and MSH2 in LS predisposition.

a	MLH1	MSH2
Type of epimutation	Primary or secondary	Secondary
Published gene defects behind	Genomic deletion of MLH1 exons 1–2 [20]	Deletions of EPCAM removing stop codon [15,16,29]
secondary epimutation	Genomic deletion of MLH1 exon 1 [17]
	Duplication of MLH1 and flanking regions [18]
	Duplication of MLH1 exons 1–6 [24]
	c.-167delA in MLH1 promoter [24]
	c.-63_-58delins18 in MLH1 promoter [19]
	c.-27 C > A in MLH1 promoter [36]
	c.27 G > A (synonymous) in MLH1 exon 1 [8,24]
	AluYc insertion in 3ʹ end of MLH1 exon 1 [24]
	c.116 + 106 G > A in MLH1 intron 1 [24]
Pattern of transmission	Variable, from apparent lack of heritability	Autosomal dominant
	(primary epimutation) to autosomal
	dominant (secondary epimutation)
Share of LS^a	1–10%	10–40%

^aEstimated from Lynch-suspected cases with MMR-deficient tumors and negative screens for conventional germline mutations in MMR genes.

Depending on the method of ascertainment, constitutional epimutations of MLH1 (primary or secondary) may explain 1–10% of LS-suspected cases without traditional genetic mutations and with absent MLH1 expression in tumor tissue [20–23]. Distinction of primary epimutations of MLH1 from those that are secondary to genetic changes is challenging, since a comprehensive screening of the MLH1 region [24], and in fact, the entire genome, would be required for a reliable exclusion of a possible inducing alteration. Constitutional epimutations are rare among unselected colorectal cancer patients: in a study on one such series from Spain, no MLH1 constitutional epimutations were detected [25]. On the other hand, low-level methylation may be missed by standard techniques: a sensitive method used by Damaso et al. [26] was able to detect constitutional MLH1 methylation at 1% level in 1/18 (6%) patients with MLH1 hypermethylated colorectal tumors. The share of EPCAM deletion-associated constitutional epimutations of MSH2 among Lynch suspected families without conventional germline mutations and with absent MSH2 protein in tumor tissues varies between 0% and 40% depending on possible founder effects [15,23]. Apart from MLH1 and MSH2, no constitutional epimutations have been reported for the remaining MMR genes [27].

The clinical phenotype of constitutional epimutation carriers is indistinguishable from that of conventional LS mutation carriers. For occasional carriers of MLH1 epimutation, atypical phenotypes, such as colonic polyposis, have been reported [28]. EPCAM deletion carriers have a high risk of colorectal cancer, whereas only deletions extending close to the MSH2 promoter are associated with increased endometrial cancer risk [29]. In analogy to conventional LS mutations, tumor tissues show MSI and absent MMR protein(s) following inactivation of the wild-type allele by a genetic or epigenetic mechanism [15,20,30–32].

Since epigenetic changes are erased on passage through the germline, they are not regularly transmitted from parent to child. Therefore, primary constitutional epimutations segregate in a non-Mendelian fashion and are seldom associated with any remarkable family history of cancer, which makes a clear distinction relative to LS arising from genetic mutations. Mosaic epigenetic inheritance and reversion of the methylated allele to normal active state have been observed in families of primary epimutation carriers, and occasionally apparent heritability is seen [21,33–35]. In contrast, secondary epimutations of MLH1 or MSH2 give rise to classical LS families and co-segregate with their cis-acting genetic changes as dominant Mendelian traits [15,24]. However, there is evidence that the basic mechanism of transmission differs from that of a genetic mutation: a secondary epimutation of MLH1 was found to undergo a complete but transient reversion in the germline, being erased in spermatozoa but reinstated in somatic cells of the next generation [36].

1.3. Lynch-like ‘syndrome’ – a LS mimic

MLH1 promoter methylation as an acquired (somatic) alteration is by far the most frequent cause of MMR deficiency in various carcinomas, attributable to 10–20% of all colorectal carcinomas and 30% of endometrial carcinomas [37,38]. Germline mutations of MMR genes indicative of LS are detected in 1–3% of unselected colorectal or endometrial carcinomas [3]. MMR-deficient cancers not explained by either of the two mechanisms mentioned above represent Lynch-like syndrome (LLS). The share of all colorectal carcinomas that LLS is responsible for (2–3%) [37,39] is comparable to LS. Two somatic mutations of the MMR genes MLH1, MSH2, MSH6, or PMS2 are detectable in more than half of LLS tumors [2,37,40–42]; this subset of LLS tumors is referred to as double somatic cases.

A higher average age at onset compared to LS is a feature of LLS [37,43], but on the level of an individual patient, LLS cannot be reliably distinguished from LS on clinical or histopathological grounds [39,44]. Some LLS cases are young and a (minor) proportion shows a positive family history [39,45], raising the possibility of a hereditary entity (even in the proven double somatic subset). Rare likely pathogenic germline variants in repair genes and other genes that maintain genome integrity were detected in LLS by Vargas-Parra et al. [46] (BUB1, SETD2, FAN1, and heterozygous variant of MUTYH) and Xicola et al. [47] (WRN, MCPH1, BARD1, and REV3 L). Additionally, tumors from variant carriers exhibited a mutational signature different from that seen in tumors from LLS patients without such variants, suggesting a possibly distinct etiology in the former group [47]. On the other hand, Pico et al. [45] found that LLS patients had homogeneous clinical and pathology characteristics regardless of family history of colorectal cancer, arguing against the existence of ‘sporadic’ and ‘hereditary’ subgroups within LLS. Future studies are necessary to resolve this question.

1.4. Homozygosity for MMR gene mutations – a shift from non-polypotic colorectal cancer to polyposis

To date, 100–200 patients have been reported to have homozygous or compound heterozygous germline mutations in any one of the four LS-associated MMR genes. These patients represent the constitutional mismatch repair deficiency (CMMRD) syndrome [48–50]. The syndrome is characterized by childhood cancers such as hematological malignancies and brain tumors, signs of neurofibromatosis type 1 (café-au-lait spots), and early-onset colorectal cancer. The distinct tumor spectrum may reflect sensitivity of neural and hematological progenitor cells to deficient MMR through mutational inactivation of certain key genes such as NF1 [51]. Since both alleles of the involved MMR gene are mutant in the germline, the respective protein is absent in normal tissues in addition to tumors. On the other hand, microsatellite length variants induced by defective MMR occur at such low levels in nonneoplastic tissues that their detection requires sensitive techniques [50,52,53].

The predominant genes underlying CMMRD are PMS2 and MSH6 [50]. When heterozygous, PMS2 mutations cause predisposition to LS with low penetrance [7,54], whereas homozygosity or compound heterozygosity for PMS2 mutations gives rise to CMMRD with full penetrance [50]. Adenomatous polyposis ranging from a few polyps to familial adenomatous polyposis-like phenotype is a regular feature of CMMRD [55,56]. In addition to CMMRD due to biallelic germline mutations of the four LS-associated MMR genes (especially PMS2), polyposis is the leading phenotypic feature in the rare patients with biallelic germline mutations of MSH3 [57] and MLH3 [58]. In heterozygous states, MSH3 mutations are not known to be associated with any apparent disease phenotype [57], and MLH3 mutations may act as low-penetrance susceptibility factors for colorectal cancer [59,60]. In CMMRD, germline inactivation of both alleles of the MMR gene in question, together with somatic polymerase exonuclease domain mutations, was shown to result in an ‘ultra-hypermutated’ tumor phenotype (over 100 mutations/Mb) in brain tumors, whereas gastrointestinal polyps revealed no increased mutational load compared with adult polyps [61]. Moreover, no microsatellite instability (at dinucleotide or trinucleotide repeats) was observed in a polyp from a homozygous carrier of a nonsense mutation in PSM2 [62]. Tumor tissues from biallelic MSH3 mutation carriers showed a distinct form of microsatellite instability called EMAST (Elevated Microsatellite Alterations at Selected Tetranucleotide Repeats) [57]. No instability at any type of microsatellite repeats was found in polyps or cancer tissues from biallelic MLH3 mutation carriers [58]. Absent or low-degree MSI may reflect partial redundancy between the different hMutS and hMutL dimers (see above).

At present, no definitive explanation is available for polyp predisposition in biallelic carriers of MMR gene mutations. At least one somatic APC mutation was detected in each of three PMS2-associated polyps examined by Siraj et al. [62], implying that inactivation of APC, a gatekeeper of cellular proliferation, is important for tumor development. In MSH3-associated tumors, the observed inflammatory infiltration and the spectrum of somatic APC mutations (small deletions at repeat sequences) pointed to the significance of MMR deficiency in tumorigenesis [57]. Finally, in MLH3-associated tumors, multiple loci revealed allelic imbalance, suggesting the possible involvement of chromosomal segregation or recombination-related mechanisms [58].

2. Familial colorectal cancer type x

2.1. Division of hereditary non-polyposis colorectal cancer into two main entities

MMR-proficient families meeting the Amsterdam criteria [63,64] are referred to as familial colorectal cancer ‘type X’ (FCCTX); ‘type X’ implies that the genetic basis is largely unknown [65,66]. Prior to the discovery of the LS-associated MMR genes in the 1990 s, the Amsterdam Criteria were used to identify families with a presumably inherited form of colorectal carcinoma termed HNPCC. Molecular analyses have revealed that LS accounts for approximately half of such families, whereas, in the remaining families (FCCTX), no MSI or absent MMR protein in tumor tissue or germline MMR gene defect is detectable [65]. FCCTX families show site-specific (mainly distal) colorectal cancer, as opposed to colonic (mainly proximal) and extracolonic cancers seen in LS (Table 3). The mean age at cancer onset is around 55 years in FCCTX vs. 45 years in LS [67].

Table 3. Two main subgroups of hereditary non-polyposis colorectal cancer (HNPCC).

	LS	FCCTX
Share of HNPCC	~Half	~Half
Age at cancer onset	45 years	50–60 years
Tumor spectrum	Colon cancer & extracolonic cancers	Site-specific colorectal cancer
Precursor lesion of colorectal cancer	Adenomatous polyp	Adenomatous polyp
Transmission pattern	Autosomal dominant	Autosomal dominant
Predisposing genes	MLH1, MSH2, MSH6, PMS2, EPCAM	Mostly unknown, candidate genes exist^a
Tumor characteristics	MMR-deficient and hypermutant,	MMR-proficient, typically non-hypermutant,
	CIMP may be present	global hypomethylation present

^aPlease see Table 4.

Table 4. Proposed novel high-penetrance susceptibility genes for FCCTX.

Gene	Protein product	Mutation type	Share of cases or families (%)	Evidence for pathogenicity	Reference
GALNT12	N-acetyl-galactosaminyl-	Start codon, missense	7/272 CRC patients (3%)	Lost or reduced enzymatic activity	Guda et al. [74]
	transferase type 12	(T491 M, R297 W, R373 H, R382 H, D303 N),
		nonsense (Y395*)
		Missense (D303 N, Y396 C)	Two Bethesda+ families	Cosegregation	Clarke et al. [75]
			(4/118 probands, 3%)	(Function: see ref. [74])
		Missense (D303 N, R297 W, H101Q, I142 T,	10/479 incident CRC cases (2%),	Lost or reduced enzymatic activity	Evans et al. [76]
		E239Q, T286 M, V290 F), splicing variant	including 3 Bethesda+ and 2
		(c.732–8 G > T)	Amsterdam+ mutation positive families
RPS20	Ribosomal protein S20	Frameshift (c.147dupA)	Amsterdam+ family	Cosegregation, defect in pre-	Nieminen et al. [78]
			(1/16 Amsterdam+ or Bethesda+	ribosomal RNA maturation
			families, 6%)
		Frameshift (c.181_182delTT), missense	2/863 familial CRC cases (0.2%)	In silico evaluation for missense	Broderick et al. [79]
		(V54 L)	(missense mutation in Amsterdam+
			family)
		Splicing variant (c.177 + 1 G > A)	Amsterdam+ family	Co-segregation, splicing defect	Thompson et al. [80]
			(1/746 CRC cases, 0.1%)	demonstrated in RNA
BRF1	RNA polymerase III	Splicing variant (c.1459 + 2 T > C), missense	11/503 individuals with familial or	Co-segregation, aberrant mRNA	Bellido et al. [88]
	transcription initiation	(T12 M, V75 M, S81 T, C140 S, P405 R,	early-onset CRC (2%), including 9	splicing, protein stability or expression,
	factor subunit	R572 G, G70 C)	Amsterdam+/Bethesda+ mutation	or yeast growth assay
			positive families
FAN1	FANCD2/FANC1-	Nonsense (C47*, R952*), missense	5/177 Amsterdam+ families (3%)	Co-segregation, defective interstrand	Segui et al. [91]
	associated nuclease 1	(D140Y, P340 S, R591 W)		cross-link repair
FAF1	Fas-associated factor 1	Missense (D371 N, R85P)	One Amsterdam+ family (among 40 CRC	Co-segregation (D371 N), resistance to	Bonjoch et al. [101]
			families, 3%) and one Bethesda+ family	apoptosis, increased NF-κB and Wnt
			(among 473 CRC families, 0.2%)	activity
SEMA4A	Semaphorine A	Missense (V78 M, G484A, S326 F, P682 S)	9/54 Amsterdam+ families (17%)	Co-segregation, increased MAPK/Erk	Schulz et al. [103]
				and PI3 K/Akt signaling (V78 M)

Abbreviations: CRC, colorectal cancer

Apart from MMR status, the epigenomic pattern differs between FCCTX and LS (Table 3). Colorectal carcinomas from FCCTX families display a remarkably low LINE-1 methylation, which is a surrogate marker of global hypomethylation [68]. In contrast, extensive LINE-1 hypomethylation is not characteristic of LS-colorectal cancers; a subset associated with a poor prognosis may constitute an exception [69]. In FCCTX patients, low LINE-1 methylation is a feature of normal mucosae, too [70] and may, therefore, represent a field defect or a constitutional alteration. The mechanistic basis of FCCTX-associated hypomethylation is unknown.

2.2. Candidate susceptibility genes for FCCTX

Genome-wide sequencing and other approaches have resulted in the discovery of novel candidate genes for FCCTX. Selected candidate genes with functional proof available are listed in Table 4. Each gene appears to be involved in only a small fraction of families and no major FCCTX susceptibility genes, analogous to MMR genes in LS, have been identified to date. The genes cluster into some preferential biological pathways to be described in detail below (Table 4, Figure 1).

Figure 1. Pathways affected and interactions between genes involved in FCCTX predisposition. Generalized pathways for the FCCTX-associated genes were constructed according to the WikiPathways database and are indicated with colored circles. Only pathways with at least two affected genes were considered. Gene interactions with at least medium confidence according to the STRING database are indicated with black lines.

2.2.1. Protein glycosylation

2.2.1.1. GALNT12

Genetic linkage studies have provided evidence of the existence of a colorectal cancer susceptibility gene on chromosome 9q22 but the responsible genes remain elusive [71–73]. GALNT12 is one of the genes located in that chromosomal region and codes for the N-acetylgalactosaminyltransferase 12 enzyme involved in O-linked glycosylation of mucin-type glycans. GALNT12 was originally identified as a possible colon cancer susceptibility gene by a candidate gene approach, based on the knowledge of aberrant glycosylation as a hallmark of colon cancers and the high level of expression of GALNT12 in normal colon [74]. Germline mutations of mainly missense type and leading to lost or reduced enzymatic activity of GALNT12 were identified in 2–3% of colorectal cancer patients with variable family histories, ranging from apparently sporadic cases to families referred to a cancer genetics clinic (Table 4; refs [74–76].). No GALNT12 germline mutations were detected in 103 families meeting the stringent Amsterdam criteria [77]. The results argue against GALNT12 as a major high-penetrance susceptibility gene for FCCTX and provide evidence to support a role for GALNT12 as a moderate penetrance gene for colorectal cancer.

2.2.2. Ribosome biosynthesis

2.2.2.1. RPS20

Genome-wide linkage analysis in a multi-generation Finnish FCCTX family fulfilling the Amsterdam criteria identified an area of linkage around the centromeric region of chromosome 8, and the RPS20 gene from that region (8q12.1) was subsequently found to be mutant in four colon cancer-affected siblings by exome sequencing (Table 4; ref. [78]). The mutation (c.147dupA) shared by the siblings was predicted to cause a frameshift and premature truncation. RPS20 encodes a component of the 40 S small ribosomal subunit. RNA analyses revealed a defect in the late steps of ribosome biogenesis affecting the formation of mature 18 S rRNA [78]. Later investigations identified germline mutations (truncating or missense) in three additional probands, two of which represented familial colon cancer cases of European ancestry belonging to the UK National Study of Colorectal Cancer Genetics [79] while the third one represented an Amsterdam criteria-positive MMR-proficient family from Utah [80]. Studies on colorectal cancers from RPS20-associated families [78,80] revealed no loss of the wild-type allele, consistent with observations from zebrafish showing that ribosomal protein genes act as haploinsufficient suppressors of tumorigenesis [81].

Ribosome dysgenesis is connected to cancer in many ways, often involving p53 activation [82,83]. Western blot analysis of lymphoblastoid cells from RPS20 mutation carriers showed increased expression of p53 protein [78]. The same applies to germline mutation carriers of several other ribosomal proteins, such as RPL26 [84]; unlike RPS20 mutations, heterozygous germline mutations of other ribosomal protein genes are not associated with FCCTX but Diamond-Blackfan anemia (DBA), a rare bone marrow failure syndrome with red cell aplasia and congenital anomalies [83]. It is unknown why germline mutations of RPS20 underlie FCCTX while mutations of other ribosomal protein genes cause DBA. Some clinical overlap does exist: no significant anemia appears to accompany RPS20 mutation, but mutations of DBA-associated ribosomal proteins cause significantly increased risk of colon cancer [85]. It is possible that different ribosomal proteins are rate-limiting in different cells, that different types of cells possess different sensitivity to p53 activation, or that p53 pathway activation has different downstream effects in different cells [83,86,87], thereby explaining tissue-specific manifestations of mutant ribosomal protein genes.

2.2.2.2. BRF1

An exome sequencing study on three colon cancer affected relatives from an Amsterdam-positive FCCTX family from Spain identified a shared heterozygous splice site alteration (c.1459 + 2 T > C) that led to the skipping of exon 13 of BRF1 (Table 4; ref. [88]). The protein product of BRF1 is part of the TFIIB complex that initiates transcription by the RNA Pol III. Among other RNAs, RNA Pol III synthesizes 5 S ribosomal RNA, a limiting molecule required for the formation of the large 60 S ribosomal subunit [89]. A validation series of 502 individuals with familial or early-onset colorectal cancer sequenced for the coding region and exon-intron boundaries of BRF1 revealed 10 independent families with rare heterozygous germline mutations that affected protein expression and/or function of BRF1 [88]. Thus, BRF1 was mutant in a total of 2% (11/503) of all cases investigated, most of which met at least the revised Bethesda criteria [90] for hereditary non-polyposis colorectal cancer. In analogy to RPS20, colorectal tumors from BRF1-associated families showed no loss of the wild-type allele, supporting haploinsufficiency [88]. Taken together, cosegregation and functional evidence derived from several families having mutant RPS20 or BRF1 with converging functions, link ribosomal biosynthesis to colorectal cancer predisposition, and RPS20 and BRF1 can be considered as high-penetrance susceptibility genes for FCCTX.

2.2.3. DNA repair and damage response

2.2.3.1. FAN1

Sequencing of the exomes of three colon cancer affected members from an Amsterdam-positive FCCTX family from Spain led to the discovery of a nonsense mutation (c.141 C > A, p.C47*) in FAN1 which was shared by all affected family members and was absent in control individuals (Table 4; ref. [91]). FAN1 encodes FANCD2/FANCI-associated nuclease 1 involved in interstrand cross-link repair. Lymphoblastoid cells from a heterozygous mutation carrier showed increased sensitivity to mitomycin C which is characteristic of defects in Fanconi anemia-related genes and indicates impaired interstrand cross-link repair [91]. Homozygous FAN1 germline mutations underlie karyomegalic interstitial nephritis, where chronic kidney failure is a predominating feature and possible cancer-associations are unknown due to the rarity of the syndrome [92]. A subsequent screen of 176 MMR-proficient Amsterdam-positive families identified 4 additional families with FAN1 germline mutations co-segregating with colorectal cancer in the families and having low mutant allele frequencies (<0.1%) in the control individuals [91]. Thus, FAN1 germline mutations accounted for a total of 5/177 (3%) of Spanish FCCTX families. Colorectal tumors from FAN1 mutation carriers showed no evidence of somatic second hits, suggesting haploinsufficiency. The overall somatic mutational load in tumor tissue was in the non-hypermutant range [91].

Interpretations of the role of FAN1 mutations in FCCTX susceptibility are complicated by the fact that some population-based control series, too, have been found to show disruptive FAN1 germline variants with frequencies less than 1% [79,91]. Analysis of the exomes of 863 familial colorectal cancer cases and 1604 controls from the UK National Study of Colorectal Cancer Genetics detected no significant increase of (likely) damaging FAN1 variants in cases (15/863, 1.7%) vs. controls (17/1604, 1.1%, p = 0.19). Therefore, additional investigations seem necessary to settle the role of defective FAN1 in FCCTX predisposition.

2.2.3.2. WRN and ERCC6

Arora et al. [93] pursued the hypothesis that constitutional defects in double-strand break (DSB) repair may underlie familial colorectal carcinomas of an undefined genetic basis. In addition to colorectal cancer, 25 patients studied by Arora et al. [93] had cancers of several other organs and some also had (mild) polyposis. Following the observation that peripheral blood cells from familial colorectal cancer patients vs. controls showed higher levels of DSB-associated γ(phospho)-H2AX foci after treatment with DNA-damaging agents, the patients were subjected to exome sequencing. Compared to controls, a significantly higher frequency of disruptive variants in genes relevant to DSB repair, nucleotide excision repair (NER), or other repair mechanisms were found. These included the WRN (RecQ-helicase-like) gene involved in DSB and the ERCC6 (excision repair cross-complementation group 6) gene involved in transcription-coupled NER. The variants occurred in a heterozygous state in the colon cancer patients and their consequences were validated by functional experiments. Homozygous null mutations in WRN are known to cause Werner’s syndrome, while homozygous ERCC6 mutations cause Cockayne’s syndrome; Werner syndrome is associated with increased risks of various tumors, including gastrointestinal cancers [94], whereas in Cockayne’ syndrome, cells die prematurely which may offer protection against cancer development [95].

2.2.3.3. POT1, POLE2, and MRE11

Chubb et al. [96] hypothesized that rare high-impact alleles might contribute to the ‘missing heritability’ of colorectal cancer and conducted an exome sequencing-based epidemiological investigation on the occurrence of such alleles in 1006 early-onset familial colorectal cancer cases vs. 1609 healthy controls from the UK population. Highly penetrant rare mutations were identified in 16% of familial colorectal cancer cases. Excluding established colon cancer susceptibility genes (such as MMR genes and APC), POT1 (Protection of Telomeres 1), POLE2 (DNA Polymerase Epsilon 2, Accessory Subunit), and MRE11 (MRE11 Homolog, Double-Strand Break Repair Nuclease) were identified as new candidate colon cancer susceptibility genes, by virtue of truncating germline mutations that were significantly more frequent in cases than controls. The protein product of POT1 is part of the shelterin complex required for telomere maintenance. Heterozygous germline mutations have been associated with an autosomal dominant cancer family syndrome with diverse malignancies, such as glioma, angiosarcoma, and melanoma [97]. POLE2 is part of the accessory (non-catalytic) subunit of the polymerase epsilon enzyme complex. In a previous study, targeted sequencing of polymerase genes resulted in the identification of rare POLE2 variants (one truncating and one missense) in two patients with colorectal polyposis [98]. The MRE11 protein is involved in DSB repair, homologous recombination, and telomere length maintenance. Heterozygous germline mutations of MRE11 may cause increased breast cancer risk [99], whereas homozygous mutations underlie ataxia-telangiectasia-like disorder [100].

2.2.3.4. FAF1

Exome sequencing on 40 FCCTX families identified a missense variant (c.1111 G > A, p.Asp371Asn) in the FAF1 (Fas-associated factor 1) gene in an Amsterdam-positive family of Spanish origin [101]. The variant was present in all four affected family members tested, including three with colorectal cancer and one with advanced adenoma. Targeted sequencing of FAF1 was subsequently conducted on 473 additional families mostly fulfilling the revised Bethesda criteria, and another presumably pathogenic missense variant (c.254 G > C, p.Arg85Pro) was discovered. No co-segregation studies were possible in the latter family. The two missense variants together explained a 0.4% fraction (2/513) of colorectal cancer families investigated [101].

The FAF1 gene product is a tumor suppressor with diverse functions; among others, it is a negative regulator of NF-κB and Wnt signaling [101]. Fas-associated factor 1 also interacts with proteins involved, for example, in TFGβ signaling, Fas-mediated apoptosis, and ubiquitin-related functions [102]. The two missense variants detected by Bonjoch et al. [101] rendered the resulting proteins unstable. When FAF1 was knocked out in DLD1 colorectal cancer cells, cellular proliferation increased, and the cells became resistant to programmed cell death. Transient expression of the missense variants in DLD1 knockout cells recapitulated the features mentioned above and revealed NF-κB and Wnt signaling activation, as judged from increased nuclear translocation of NF-κB, and impaired degradation and cytoplasmic accumulation of β-catenin, respectively. As pointed out by Bonjoch et al. [101], investigations of additional cohorts are necessary to confirm and further evaluate the significance of FAF1 as a FCCTX susceptibility gene.

2.2.4. Other pathways

2.2.4.1. SEMA4A

Combined linkage analysis and exome sequencing in an Amsterdam-positive FCCTX family from Austria identified a missense variant (c.232 G > A, p.Val78 Met) in SEMA4A as the only variant shared by all four colorectal cancer-affected members tested [103]. The gene is located on chromosome 1q22 and encodes semaphorine 4A, a membrane protein with various functions (e.g., in neuronal axon extension, immunomodulation, and cell proliferation). The Val78 Met mutation was associated with increased MAPK/Erk and PI3 K/Akt signaling. Semaphorins have been found to have pro- and anti-carcinogenic functions; here, the evidence as a whole suggested loss of function, which was further supported by homozygosity of the variant in colorectal tumor tissues from two heterozygous mutation carriers [103]. A targeted screen of 53 additional Amsterdam-positive FCCTX families from Austria, Germany, and the United States identified three further missense changes in SEMA4A; among these, the rare G484A and S326 F variants were predicted to be deleterious by in silico tools, whereas a case–control association study suggested a significant enrichment of the single nucleotide polymorphism p.Pro682 Ser in FCCTX cases (6%) vs. controls (1%, p = 0.0008). However, an attempt to validate the proposed cancer-predisposing role of the recurrent G484A and P682S variants in an independent case–control study involving ~7,000 colorectal cancer cases and 10,000 controls from six European populations failed to do so [104]. Furthermore, the entire coding region of SEMA4A was sequenced in 158 FCCTX families, which resulted in the detection of protein changing variants in 16% of FCCTX families, compared with 14% of controls, suggesting no significantly increased frequency of SEMA4A variants in FCCTX families [104]. In conclusion, additional investigations are needed to resolve the role of SEMA4A variants in FCCTX predisposition.

2.2.4.2. BMPR1A

Genome-wide linkage scan conducted on a large Amsterdam-positive FCCTX family from Finland resulted in the observation of tentative linkage on 10q23 [105]. Sequencing of genes from the linked region revealed an in-frame germline mutation (c.264_266del, p.Glu88del) in exon 3 of BMPR1A which co-segregated with colon carcinoma and/or adenoma in the family [105]. The protein product of the BMPR1A gene is a type I bone morphogenetic protein receptor that belongs to a family of serine/threonine kinases. BMPR1A was subsequently sequenced in 17 additional Finnish families fulfilling the Amsterdam or Bethesda criteria and having MMR-proficient tumors. The proband of an Amsterdam-positive family revealed a truncating splicing mutation; no other affected members from that family were available for co-segregation studies [105]. Neither of the two predisposing mutations, each occurring in one family [105], have been found in healthy reference populations.

Loss of function mutations of BMPR1A are known to underlie juvenile polyposis syndrome, characterized by hamartomatous polyps and increased risk of gastrointestinal cancer [106]. Interplay between epithelial and stromal cells is believed to be important for the development of hamartomatous polyps, and loss of BMP signaling in the epithelial compartment alone may be insufficient [107,108]. For comparison, polyps from the two BMPR1A-linked FCCTX families described by Nieminen et al. [105] were of adenomatous histology and the highest combined number of adenomatous and hyperplastic polyps per colonoscopy screen per person was 4; thus, the clinical picture was that of a non-polypotic colorectal cancer rather than polyposis. When bulk tumor tissues were investigated, the FCCTX-associated BMPR1A germline mutations were accompanied by loss of heterozygosity in colorectal tumors: the splicing mutation was associated with wild-type allele loss, whereas wild-type and mutant allele losses were equally common in tumors from the in-frame deletion carriers [105].

Additional phenotypes that BMPR1A germline mutations have been linked to include early-onset MMR-proficient colorectal cancer [109] and unexplained adenomatous polyposis [110]. The basis of the reported genotype-phenotype associations is unknown. In the BMP signaling pathway, multiple ligand, receptor, and downstream effector combinations are possible and might allow for tissue or cell-specific responses. Furthermore, the seemingly different phenotypes may overlap. For example, Lieberman et al. [111] evaluated seven unrelated families sharing a genomic deletion of the entire coding region of BMPR1A and found extremely variable phenotypes despite the same predisposing mutation and the same genetic background (Bukharin Jewish ancestry). Importantly, adenomatous polyps predominated over juvenile polyps in the families [111]. Lastly, Evans et al. [112] conducted a targeted investigation on 22 unrelated FCCTX families from Newfoundland to determine the frequency of BMPR1A germline mutations: no rare (mutant allele frequency < 1%) or novel variants were found. Therefore, studies on additional FCCTX families from different populations are warranted to define the relationship between BMPR1A mutations and FCCTX.

3. Prospects to identify novel cancer-predisposing genes – challenges and promises

A single biological pathway, MMR, with the major players already known (Table 1) defines LS. The diagnosis of FCCTX on the other hand is based on the exclusion of LS and the demonstration of MMR-proficient tumors, until the responsible genes are identified. The available experience suggests that FCCTX is genetically heterogeneous and private genes and mutations may be the norm (Table 4). A burning question is how to best find the genes underlying colon cancer susceptibility in the majority of FCCTX families that currently lack molecular definition.

3.1. Methodological considerations

Over the years, the discovery of new cancer predisposing genes has been based on three main strategies: candidate gene analysis, genetic linkage analysis, and most recently, genome-wide sequencing [113]. Large families with multiple cases of preferably early-onset colorectal cancer affecting several generations would provide convincing evidence of a hereditary nature of cancer predisposition and would obviously be ideal for searches for novel susceptibility genes. This setting was utilized in previous linkage-based studies that led to the identification of many currently known colorectal cancer predisposition genes [114]. Genetic linkage analysis was also part of the strategies used to identify many of the FCCTX candidate genes listed in Table 4; however, genome-wide sequencing, usually of the whole exomes, is nowadays the main approach. While the available methods are constantly improving, some challenges remain and may guide searches for predisposing mutations that have possibly escaped detection. First, not all relevant regions of the genome may be covered. High-penetrance mutations are assumed to mainly reside in the coding portions of the genome, which is why exome sequencing is the preferred method. Exome sequencing interrogates the exons and their immediate flanking regions (~ 2% of the human genome). However, some high-penetrance mutations may involve promoter regions, introns, or non-coding RNA. For example, mutations residing deep in the introns can activate sequences (pseudoexons) flanked by splice sites, which can result in the insertion of intronic sequences in the mature mRNA [115]. In such cases, whole-genome sequencing and RNA-sequencing would be required to catch the alteration [115]. Likewise, for the detection of constitutional epimutations, a combination of techniques (genetic and epigenetic) would be needed. Second, gaps may remain in data analysis. While single nucleotide variants (SNVs) are effectively detected by the available variant calling pipelines, the ability to accurately call insertion-deletion type changes (indels) remains imperfect [116]. The same applies to copy number variant (CNV) analysis, although CNV detection algorithms have been developed for whole genome, whole exome, and targeted deep sequencing data [117].

3.2. Data interpretation

Protein truncating variants, which predominate among known disease-causing germline mutations of cancer genes [118], are considered deleterious by nature. However, panel sequencing studies suggest that up to 40% of patients suspected to have a hereditary cancer syndrome may carry a variant of uncertain significance, VUS [119]. VUSes typically consist of single amino acid substitution (missense) changes which can be classified neither pathogenic nor harmless a priori. Even synonymous nucleotide changes (i.e., nucleotide substitutions which do not result in a coding change) can be pathogenic if they, for example, disrupt motifs for proper splicing [120]. Thus, if a novel candidate susceptibility gene is affected by non-truncating alterations alone, the cancer predisposing role of the gene may remain unknown in cases where confirmatory studies (e.g., co-segregation or functional investigations) are not possible. Occasionally, two or more variants (of the same gene or different genes) may co-exist and even co-segregate with disease phenotype in a single family, which could be compatible with several alternative interpretations, including a true predisposing change coexisting with spurious variants, or a possibly multigenic basis of colon cancer susceptibility (see below).

3.3. Genotypic and phenotypic heterogeneity

A plausible reason why searches for new predisposition genes for FCCTX often fail is that FCCTX families do not have any pathognomonic features, either clinical (such as multiplicity and/or distinct histology of polyps that may aid the dissection of polyposis syndromes) or molecular (such as deficient MMR characteristic of LS), which could guide gene discovery efforts and distinguish true gene carriers from phenocopies (non-carriers). The possibility to investigate multiple (affected and unaffected) family members or utilize other hallmarks of hereditary cancer besides family history (such as multiple tumors in a single individual or the combination of specific tumor types) can be important advantages. Genome-wide sequencing studies have demonstrated that the inclusion of multiple affected family members who can reasonably be assumed to share the same high-penetrance susceptibility variant can significantly reduce the number of candidate variants worth considering [78]. Tumor genotype and phenotype can likewise be informative, including the type of genomic instability and epigenomic pattern (Table 3) or mutational signature [121]. Since each mutational process leaves a specific imprint in the genome of a cell, somatic mutational signatures can offer clues to the predisposing genes involved. Diaz-Kay et al. [122] performed an integrated whole-exome analysis of germline and tumor DNA to detect novel tumor suppressor genes underlying familial colorectal cancer. They required that genes of interest should be altered in both normal and tumor DNA according to Knudson’s two-hit hypothesis. The strategy pinpointed DNA repair genes, including BRCA2, BLM, ERCC2, RECQL, REV3 L, and RIF1, and mutational signatures additionally provided evidence of the existence of novel predisposition genes to be defined by further investigations [122].

3.4. Polygenic basis of colon cancer susceptibility

Rare high-penetrance variants in known susceptibility genes account for only ~5% of the total colorectal cancer burden [123]. Genome-wide association studies of common colorectal cancers have identified over 40 independent loci of low-penetrance risk variants frequent in the average population [124]. Each low-penetrance locus is estimated to increase colorectal cancer risk no more than up to 1.5-fold [123]. Although mostly residing outside coding regions, low-penetrance variants can influence the expression of genes from pathways associated with high-penetrance susceptibility genes (including, for example, the TGFβ/BMP pathway [125]). Thus, epistatic (synergistic non-additive) interactions between two or more variants from different loci may account for some of the currently unexplained colorectal cancer susceptibility. Rare variants with moderate effects are another worthwhile possibility in searches for novel colon cancer predisposition genes. This is because such variants are easily overlooked in family-based studies due to incomplete co-segregation, and the same could happen with genome-wide association studies because large sample sizes would be required to reach statistical significance [123].

4. Expert opinion

Hereditary non-polyposis colorectal cancer offers an illustrative example of the remarkable progress that, along with technical developments, has taken place in understanding the genetic basis of human disorders in recent years. LS is well-defined in terms of predisposing genes, and genetic challenges that remain mainly concern optimal detection methods of predisposing variants and interpretation of clinical significance of the variants. It is currently recommended that all new patients with colorectal or endometrial cancer should undergo systematic screening for LS, consisting of MSI and MMR protein tests on tumor tissue followed by further analyses (for MLH1 methylation and additionally, BRAF-V600E mutation for colorectal cancers), and ultimately, germline mutation testing when appropriate [126]. This ‘universal’ screening, combined with high-throughput sequencing that is replacing gene-based methods in germline mutation detection, is likely to multiply the numbers of MMR gene variants that need to be annotated for pathogenic significance. While various in silico tools [127] may be helpful for a preliminary assessment, there is an urgent need for functional tests robust enough that could unequivocally distinguish between normal and abnormal MMR gene function. LS carrier tests based on the effect of MMR protein dosage on MMR capability [128,129] are under development and/or already available for clinical diagnostics (DiagMMR®).

Increased knowledge of genotype-phenotype correlations in LS is expected to improve the possibilities of tailoring cancer prevention and treatment according to the predisposing gene and variant [7,9] as well as the genetic, epigenetic, or immunological type of tumors [130,131]. Clinical trials conducted to date have already established the responsiveness of LS and other hypermutant/MSI tumors to immune checkpoint inhibitors, motivating the use of such drugs to treat metastatic/advanced disease [132]. At the same time, immunovaccination [133] and chemoprevention [134] represent promising new approaches for cancer prevention in LS.

Despite recent progress, the underlying genes remain to be identified for most cases of FCCTX. Even for those few families with plausible candidate genes discovered, the true predisposing nature of the genes and variants is typically awaiting further proof. The decreasing costs for genome-wide sequencing make investigations of larger patient cohorts affordable, thereby facilitating new gene discoveries on the one hand and providing accurate information of the molecular epidemiology of existing candidate genes and variants on the other hand. Experience from pan-cancer analysis of whole genomes shows that common and rare germline variants influence the patterns of somatic mutations [135]; hence, mutational signatures of tumor tissues are becoming an integral part of future searches for FCCTX susceptibility genes. Moreover, a closer scrutiny of the non-coding genome (comprising 98% of the whole genome) is anticipated to pinpoint new genes and mechanisms that may underlie colon cancer susceptibility or modify the disease phenotype.

Based on available knowledge (Table 3), a modified cancer prevention protocol for FCCTX is currently recommended [136], consisting of colonoscopy once in every 5 years (vs. once in every 1–2 years in LS [126]) and starting from the age of 45 years (vs. 20–35 years in LS [126]). Cancer surveillance in FCCTX focuses on colon cancer alone since extracolonic cancers are considered rare. A reappraisal may be warranted since FCCTX families from a national (Danish) HNPCC register revealed an increased risk of five types of extracolonic cancers (cancers of the urinary tract, breast, stomach, and pancreas, as well as ocular melanoma) compared to the average population [137]. In the study mentioned above [137], information of possible predisposing genes was not available, and FCCTX families were ascertained through the Amsterdam criteria and MMR-proficient tumors A better understanding of the individual nature of the susceptibility genes, their biological function and associated clinical features (including tumor spectra), is likely to refine protocols for cancer surveillance and enable targeted therapy for subsets of FCCTX in the future.

Article highlights

• Hereditary non-polyposis colorectal cancer refers to cancer predisposition caused by single genes affected with high-penetrance mutations. Transmission is usually autosomal dominant. While colorectal tumorigenesis may involve polyp precursors, no significant polyposis is present.

• DNA mismatch repair deficiency vs. proficiency divides hereditary non-polyposis colorectal cancer into two genetically and clinically distinct categories, Lynch syndrome and familial colorectal cancer type X, respectively.

• Germline mutations in MLH1, MSH2, MSH6, PMS2, and EPCAM cause predisposition to Lynch syndrome. Over 3,000 different germline mutations are known.

• Following somatic inactivation of the wild-type allele, tumor tissues from Lynch syndrome patients show absent mismatch repair protein(s) and high-degree microsatellite instability. Defective mismatch repair results in increased numbers of somatic mutations in other genes (hypermutability).

• Clinical phenotype of Lynch syndrome depends on the predisposing gene and typically involves colorectal and extracolonic cancers. Constitutional homozygosity for mismatch repair gene mutations gives rise to distinct clinical features and frequent polyposis.

• No established major predisposition genes are known for familial colorectal cancer type X, but several candidate genes have been proposed, each affected in a few families or a single family. The genes may participate in protein glycosylation (GALNT12), ribosome biosynthesis (RPS20 and BRF1), DNA repair (FAN1), or various other biological pathways.

• Tumors from patients with familial colorectal cancer type X are mismatch repair-proficient and microsatellite-stable. Somatic mutational signatures may provide clues to the predisposing genes involved.

• While the low numbers of mutation carriers make assessments of tumor spectra and other clinical features unreliable, site-specific colorectal cancer diagnosed at somewhat later ages compared to Lynch syndrome appears true for most families with colorectal cancer type X.

• Genome-wide profiling of constitutional and tumor tissues, combined with functional investigations, are likely to define genotype-phenotype correlations in the existing syndromes and identify novel high-penetrance susceptibility genes for familial colorectal cancer type X.

Declaration of interest

P Peltomäki receives research grant support from Jane and Aatos Erkko Foundation, the Academy of Finland (grant number 294,643 to P.P.); the Finnish Cancer Foundation; the Sigrid Juselius Foundation; the HiLIFE Fellows 2017-2020; T Nieminen is the recipient of research grants from Jane and Aatos Erkko Foundation and the Päivikki and Sakari Sohlberg Foundation. A Olkinuora is supported by the Doctoral Programme of Biomedicine of the University of Helsinki. The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose

Additional information

Funding

This work was funded in part by grants from the Jane and Aatos Erkko Foundation, the Academy of Finland (grant number 294,643 to P.P.); the Finnish Cancer Foundation; the Sigrid Juselius Foundation; the HiLIFE Fellows 2017-2020; and the Päivikki and Sakari Sohlberg Foundation and the Doctoral Programme of Biomedicine of the University of Helsinki.