GALNT12 is Not a Major Contributor of Familial Colorectal Cancer Type X

Family history is well established to be one of the strongest risk factors for the development of colorectal cancer (CRC) and it is thought to involve approximately 20% of all CRC cases. However, only a minority of all CRC cases (2%–6%) are explained by germline genetic mutations in well-known high penetrance genes [reviewed by Lynch et al., 2009]. The proportion of CRC families that fulfill the most stringent criteria for hereditary CRC, that is, the Amsterdam criteria, and do not show a mismatch repair (MMR) defect is high (∼40%). These individuals are at increased risk of developing CRC and therefore require strict cancer surveillance strategy. The genetic cause of the CRC familial aggregation in these families is unknown and they have been grouped as familial CRC type X (fCRC-X) [reviewed by Ku et al., 2012]. Several dominantly acting predisposition loci mapping to different chromosomal regions, such as 9q22.2-31.2, 5q14-q22, or 3q13.31-q27.1, have been identified through genome-wide linkage studies in CRC families, but so far no causal gene has been reported [Gray-McGuire et al., 2010; Picelli et al., 2008; Tomlinson et al., 1999].

Previous evidence indicates that mutations in the GALNT12 gene (MIM #610290), which codes for the enzyme N-acetylgalactosaminyltransferase-type 12, might explain familial CRC cases of unknown etiology [Clarke et al., 2012; Guda et al., 2009]. This gene, whose protein product is involved in the O-glycosylation of mucin-type glycans, shows high expression levels in the normal colon and is downregulated in a significant proportion of colorectal tumors [Guo et al., 2002; 2004]. GALNT12 is located in 9q21-33, in close proximity to the linkage peak in 9q22-31, recurrently found when studying familial CRC cases [Gray-McGuire et al., 2010; Kemp et al., 2006; Skoglund et al., 2006; Wiesner et al., 2003]. In 2009, Guda et al. identified functionally relevant germline GALNT12 mutations in seven out of 272 (2.6%) colon cancer patients. Of note, no mutations were present in 192 cancer-free controls. Recently, Clarke et al. (2012) studied the occurrence of GALNT12 mutations in a cohort of 118 CRC families of unknown genetic cause. They identified two missense mutations present in four different Bethesda-positive families. However, no mutations were detected in 26 probands that met the Amsterdam I criteria. Based on the previous evidence implicating GALNT12 in CRC predisposition, we wanted to expand the study of the relevance of GALNT12 mutations in the etiology of fCRC-X.

A total of 103 fCRC-X families (all Caucasian) were included in the study: forty-three had been referred to the Genetic Counseling Units of the Catalan Institute of Oncology in the Spanish region of Catalonia between 1998 and 2011; 27 families were recruited by the Hereditary Cancer Unit of the Instituto de Biología y Genética Molecular (IBGM) of the University of Valladolid and the Spanish Research Council (CSIC), between 2006 and 2011; and 33 came from the Hereditary Cancer Program of the Spanish region of Valencia, which belongs to the Valencian Biobank Network, and were collected between 2005 and 2012. All families fulfilled Amsterdam criteria and did not have MMR deficiency (either tumor microsatellite instability and/or lack of expression of the MMR proteins MLH1, MSH2, MSH6, or PMS2). Of all 103 families, 45.6% fulfilled Amsterdam I and 54.4% Amsterdam II. The mean age at cancer diagnosis was 48.4 for the sequenced probands. Informed consent was obtained from all subjects and the study received the approval of the Ethics Committee of IDIBELL (ref. PR073/12). Supp. Table S1 shows the characteristics of the families included in the study.

All exons and exon–intron boundaries were sequenced using a standard protocol for automated direct Sanger sequencing. Primer sequences and PCR conditions are available upon request. Sequencing was performed on an ABI Sequencer 3730 and data analyzed using Mutation Surveyor v. 3.10. Identified variants were submitted to the LOVD database (http://www.lovd.nl/GALNT12).

A total of 20 variants were identified in 90 patients. Of the 20 variants, eight were located in protein-coding regions, seven in introns, and five in the 3′-UTR. Three of the eight changes in protein-coding regions, c.136G>A (p.G46R), c.356A>T (p.E119V), and c.781G>A (p.D261N), were nonsynonymous. The change p.E119V was identified in 18 fCRC-X patients, p.D261N in 2, and p.G46R in 1 patient. The three variants had been previously described in public databases with population minor allele frequencies of 0.06, 0.008, and 0.09, respectively (Table 1). Also, previous studies identified these three variants both in cancer cases and controls [Clarke et al., 2012; Guda et al., 2009]. No relevant effects on the protein were predicted by the in silico algorithms SIFT and Condel, while PolyPhen-2 predicted functional effects for p.D261N (score ∼1) and p.E119V (score 0.8, when applying HumanDiv) (Table 2). Nevertheless, Guda et al. (2009) reported that none of these three variants altered the enzymatic activity of the protein.

The remaining five variants in protein-coding regions predicted to translate into synonymous amino-acid changes, affecting codons 79, 133, 464, 499, and 569. All but two, c.237C>T and c.399C>T, had been previously described in public databases (dbSNP or ESP). The population allelic frequencies reported for other two, c.1392C>T and c.1497C>T, were below 2% (Table 1). The in silico algorithm NNSplice [Reese et al., 1997] only predicted a possible effect on the splicing process for c.237C>T, where a new donor splice would be created (score 0.52) (Table 2). However, the generation of this new donor splice site was not predicted by other algorithms, such as NetGene2 [Hebsgaard et al., 1996] or SoftBerry [Burset et al., 2001]. Moreover, this variant did not segregate with the disease in the family: the father of the proband, diagnosed with rectal cancer at age 64, did not carry the variant, being therefore absent in the CRC-affected family branch.

Of the intronic variants identified, only one, c.1036–4G>A, had not been reported as a polymorphism. Likewise, the change c.1605+4G>A (rs79574929) had a population MAF of 0.006–0.012 (sources: dbSNP and ESP, respectively). Despite their proximity to the corresponding consensus splice sites, no relevant effect on the splicing process was predicted (Table 2). Regarding c.1036–4G>A, it was not present in the mother of the proband, who was diagnosed with two metachronous colon tumors at the age of 57 and 59 years, discarding it as the genetic cause of the CRC familial aggregation. Unfortunately, no cosegregation analysis could be performed for c.1605+4G>A. In summary, no functionally relevant variants were identified in any of the 103 fCRC-X families evaluated.

The results obtained in our series (n = 103 fCRC-X families), in line with those of Clarke et al., who found no mutations in 26 Amsterdam-positive families, support the notion that GALNT12 is not a major high-penetrance gene for CRC predisposition. The lack of information on the familial cancer history of the carriers of GALNT12 germline mutations in Guda et al. (2009) does not allow further investigation in this regard.

Considering the presence of GALNT12 mutations in non-Amsterdam familial CRC cases, Clarke et al. identified two missense variants, c.907G>A (p.D303N) and c.1187A>G (p.Y396C) in 4 families that met the revised Bethesda criteria [Clarke et al., 2012]. The c.907G>A change, rs145236923, identified in three different families, partially inactivates the enzymatic activity of the protein, in contrast to the other mutations, germline and somatic, identified by Guda et al. (2009) which cause almost total inactivation of the enzyme. Moreover, the evidence of cosegregation of c.907G>A with CRC in one of the two families studied was weak. The second missense variant identified, c.1187A>G (p.Y396C), is not reported in public databases, is located in the catalytic domain of the protein, and in silico algorithms predict it to be functionally relevant [Clarke et al., 2012]. All reported patients carrying inactivating germline mutations developed CRC later in life (median age: 71), and half of them were diagnosed with multiple primary epithelial tumors, including breast and colon cancers [Guda et al., 2009]. Further studies are required to provide a definitive answer about the role of GALNT12 mutations in this subset of familial cancer cases. Either germline mutations in GALNT12, as a moderate-risk gene, or in other genes located within the 9q22-31 linkage peak might still explain some familial CRC cases. Indeed, the strength of the studies that replicate the linkage signal in 9q22-31, which suggest that the disease locus housed on 9q is specific to a familial syndrome with a phenotype of young age of onset and/or severity of colon neoplasia [Gray-McGuire et al., 2010], supports the idea that probably the other genes under the linkage peak, such as ZNF367, HABP4, and GABBR2, might be relevant in CRC susceptibility.

In conclusion, our findings indicate that GALNT12 is not a major high penetrance gene for CRC. Further comprehensive studies in Amsterdam-negative CRC families are required to clarify the gene's role as moderate/low susceptibility gene in CRC familial aggregation.