Patients and DNA isolation

The sample studied consisted of 82 unrelated cases with FAP (>100 colorectal adenomas) or AFAP (5–100 colorectal adenomas). All the patients were included in the study based on colonoscopic findings and/or positive family history. Forty-eight samples were submitted for mutation analysis at the Galician Public Foundation of Genomic Medicine (FPGMX) from health centers across Galicia, and 34 were attended in the at-risk clinic for CRC of the Hospital Clinic in Barcelona. Written informed consent was obtained for each patient before mutation analysis, according to the protocols approved by the ethics review boards of the Hospitals and in compliance with the Helsinki declaration.

All patients were screened for APC germline mutations, and when negative, MUTYH was analysed. MUTYH was also studied in 9 APC-negative families included in a previous article [10].

Genomic DNA from Galician and Catalonian samples was obtained from peripheral blood using the Wizard DNA extraction kit (Promega, Madison, WI), and the QIAamp DNA Blood Mini Kits (Qiagen, Hilden, Germany) respectively. Protocols were performed according to the manufacturer's instructions.

Analysis of the APC gene

Analysis of the MUTYH gene

For every patient without detectable pathogenic mutations in APC, all MUTYH exons and their adjacent intronic splice sites were amplified using primers designed with the Primer3 software [23] [see Additional file 2].

Galician samples were analysed at the FPGMX center by sequencing each amplification fragment, as described above for the APC gene. Real-time PCR using Taqman probes, and SSCP analysis were performed for Catalonian samples at the Hospital Clinic. TaqMan genotyping included the analysis of the two most common mutations found to date in the MUTYH gene: p.Tyr165Cys and p.Gly382Asp, as well as the two rare mutations c.1103delC and c.1186_1187insGG identified in our previous study [26]. This technique is based on allelic discrimination using allele-specific probes resolved on a 7300 Real Time PCR System (Applied Biosystems, Foster City, CA).

Mutation nomenclature

All mutations were described following the guidelines proposed by the Human Genome Sequence Variation (HGSV) site and were referred to the cDNA sequences of APC (NM_000038) and MUTYH (U63329). Furthermore, all mutations were confirmed in two independent DNA extractions.

Variants of Unknown Significance (VUS)

We examined 500 chromosomes from control individuals with no personal or family history of colorectal cancer, in order to estimate the frequency of VUS. Analysis was carried out by direct DNA sequencing (see above).

Polyphen software was used to test the potential role of missense variants. This prediction program is based on observed substitutions of the residues in homologous proteins [27].

Statistical analyses

Ji-squared statistics with Fisher's correction were used to test for differences in APC and MUTYH mutation frequencies between the Galician and the Catalonian populations. Comparisons were also made for mutation frequencies at codons 1061 and 1309 of the APC gene. All statistics were estimated with the SPSS statistical software package (SPSS Inc., Chicago IL).

APCmutations

In this study, germline mutations in APC were found in 39% (32 out of 82) of the Spanish patients with FAP. Frameshift and nonsense mutations were the most frequently identified, and despite the great number of genetic variants described to date in the APC gene, seven new pathogenic mutations and two new VUS were reported. Clinical features displayed by APC-positive patients are shown on Table 1.

Five new frameshift mutations were identified: c.147_150delACAA (p.Lys49AsnfsX20),c.2900delT(p.Val967AlafsX13),c.2934_2935delAA(p.Gln978HisfsX6),c.3531delT(Ile1177MetfsX5) and c.4219_4220delAG (p.Ser1407XfsX1) in patients GAL-07, GAL-17, CAT-02, CAT-08 and GAL-25, respectively. All of these mutations were deletions of few nucleotides, that give rise to premature stop codons (X) which would lead to truncated APC proteins. We also identified two additional nonsense mutations that generate premature stop codons: c.1402 G>T (p.Glu468X) in patient GAL-19, and c.3329C>A (p.Ser1110X) in two unrelated patients (CAT-06 and CAT-07).

The two new VUS: c.3165A>G (p.Ile1055Met) and c.5357G>C (p.Arg1786Thr), were found in GAL-47 and GAL-35. Both patients displayed an attenuated FAP phenotype with an onset at around forty. These variants were not detected in 500 chromosomes from a healthy control population. However, their absence from the control group cannot be taken as prove of a deleterious effect. In silico studies using Polyphen revealed the p.Arg1786Thr as "possibly damaging", while the p.Ile1055Met was reported as "benign". In these families it was not possible to study the co-segregation with the disease, so further functional studies are necessary to consider them deleterious.

Large genomic deletions were found in 11% (6/56) of the families that tested APC mutation-negative by conventional techniques. This frequency is consistent with published data that comprise a range between 8–12% for such rearrangements [11, 12, 28]. Three different deletions were detected in 6 unrelated families: two exon 4 deletions (CAT-13 and GAL-15), an exon 1–15 deletion (GAL-14), and 3 whole-gene deletions (including the promoter) (CAT-12, GAL-27 and GAL-16). All of them were further confirmed by either cDNA studies (Figure 1) or FISH analysis (Figure 2).

Carriers of whole allelic deletions generally displayed a severe polyposis phenotype with an early onset of symptoms, as previously described [11]. A correlation between site of mutation and clinical phenotype was also observed for six of the seven new mutations identified. Mutations occurring at the beginning and middle of exon 15 were generally associated with a more severe phenotype than those located at the 5' and 3' ends of the gene, which is consistent with other studies [29]. Although mutations in exon 10 would then be associated with FAP, p.Glu468X was found in a patient classified as AFAP based on number of polyps (GAL-19) (Table 1).

Phenotypic differences about number of adenomas and extracolonic disease were observed in unrelated probands carrying the p.Lys1061LysfsX2 mutation (CAT-03 and CAT-04) and the whole gene deletions (CAT-12 and GAL-16) (Table 1). This phenotypic heterogeneity suggests that either modifier genes, epigenetic mechanisms or environmental factors could modulate the FAP phenotype. There is good evidence from humans, and particularly from mouse models, of the involvement of modifier genes that influence the severity of FAP. It is known that same sex siblings in their early twenties often show phenotypic differences which cannot be easily explained except by the action of modifier genes [30]. Despite those findings, further clinical information and an accurate follow-up of patients is necessary to confirm our results.

The two hotspots at codons 1061 and 1309 of the APC gene accounted in this study for 9.4% of the APC-positive families. However, they were reported mainly in Catalonian families. Mutation at codon 1061 was detected in 3 (CAT-03, CAT-04 and CAT-05) out of 13 Catalonian positive families (23%), but was not found in 19 APC-positiveGalician families. A similar trend was observed for the deletion at codon 1309: 15% (2 out of 13, CAT-10 and CAT-11) for Catalonians vs 5,2% (1 out of 19, GAL-24) for Galician patients. Therefore, we performed statistical analyses in order to test whether the Galician population had a significantly different mutation frequency at these codons. Ji-squared tests yielded significance for the 1061 mutation (p = 0,058) but not for the 1309 variant (p = 0,356).

Although different mutation screening methods were used to study these two populations, the possibility that this fact could have caused the different spectra observed would be very small. Firstly, the mutation frequency in both populations was similar (40% for Galicia and 38% for Catalonia, p = 0.888). Besides, Galician samples were directly sequenced, which was the most sensitive of the techniques used.

Similar results had already been observed in a previous study where 15 unrelated Galician patients were analysed, but in that study it was not possible to establish whether the inability to detect the recurrent mutations at codons 1061 and 1309 actually reflected an underrepresentation of these genetic variations in this population, or was simply due to a sampling bias [10]. Furthermore, when we considered all the data available from the Galician FAP patients altogether (the 19 APC-positive families from this study plus the 6 from our previous one [10]), the frequencies observed were 4% (1/25 APC-positive) for the 1309 deletion, and still 0% for the deletion at codon 1061, which is certainly quite remarkable. Thus, we recalculated the Ji-squared test for the 1309 mutation and a trend towards the underrepresentation of this variant was indeed observed (p = 0,265).

Although there is plenty of evidence that the mutational spectrum of the APC gene varies in different populations, these two hotspots are thoroughly reported worldwide. They represent around 8% and 20% of the APC-positive families, and the 5 bp deletion at codon 1309 is reported as the most common germline mutation [31, 32]. Interestingly, a frequency range for this deletion in different countries has also been described: a high rate in Japan (14%), a moderate frequency in most European populations (5–6%), and no 1309 deletions detected in Australia (0% out of 27 APC-positive families) [33]. It has also been recently reported that the frequency of the deletion at codon 1061 in 46 Czech and Slovak APC-positive families was lower than the expected (3%) [34]. The variation in the distribution of these hotspots could presumably be caused by an ascertainment bias, but in isolated populations it could as well be explained by a founder effect. For instance, in the Balearic Islands, where the hotspot at codon 1061 is overrepresented (50%), the haplotype analysis of the families sharing this deletion was consistent with the presence of a founder effect [35]. The recurrence of these two mutations has been linked to the molecular properties of the DNA region around codons 1061 and 1309, rather than with specific haplotypes. These two are located within a short hypermutable polyA repeat that may associate with an increased probability of DNA polymerase slippage during DNA replication, leading to an overrepresentation of deletions. The high incidence of the 1309 deletion among de novo cases, and the fact that this alteration was found to segregate with different haplotypes associated with the disease supports this hypothesis [33].

It is known that gene diversity in the Galician population is generally lower than in other European populations, as a result of its relative isolation from the rest of the Iberian Peninsula and the high emigration rates during the last two centuries [36, 37]. These genetic features would have possibly caused the selection of not yet identified allelic variants in DNA repair genes. Hypothetically, those variants would repair more efficiently the DNA polymerase slippage caused by the repetitive sequences around codons 1061 and 1309 during replication. Therefore, lower frequencies for these two APC hotspots should be observed. Such founder effects have already been observed in this population for some genetic diseases, including BRCA1 in familial breast cancer [38].

MUTYH mutations

Biallelic germline mutations in MUTYH were found in 24% of the APC-negative patients, i.e., 14/59 (fifty from this study plus 9 from the previous one [10]). This data is consistent with previous results [39, 40]. Differences between these populations were not significant (p = 0.517).

Table 2 shows that the two most frequent mutations reported to date (p.Tyr165Cys and p.Gly382Asp) were detected in quite a number of cases, the frequency of these alleles being 71%. This observation is comparable to what has been described in the literature [9, 41]. Among the other mutations found, the c.1186_1187insGG accounted for 21% of the mutant alleles reported. This mutation was previously reported in Portuguese families with a similar frequency [22].

Biallelic MUTYH carriers displayed an attenuated polyposis phenotype without extracolonic manifestations, with the exception of patient CAT-19 who showed breast cancer at 59 years and CRC at 69 (Table 2). It is noteworthy that the BRCA1 and BRCA2 tumor suppressor proteins participate in the base excision repair of 8-oxo-7,8-dihydroguanine (8-oxoG) lesions [42]. Accordingly, loss of BER function due to biallelic MUTYH mutations may underlie breast cancer risk. In Dutch MAP patients, breast cancer occurred in 18% of females, significantly more than the expected from national statistics. This observed increased breast cancer risk should be thoroughly investigated [21].

The median age at diagnosis of CRC in MAP families was 51,5 years (ranging from 43 to 69) (Table 2). In contrast, classical FAP patients showed a CRC onset 10 years earlier (median 41, ranging from 20 to 46) (Table 1). As previously reported [8], it appears that disease symptoms in MAP are not as severe as those observed in APC- driven FAP, and that they resemble an attenuated polyposis phenotype. However, patients GAL-08, CAT-15, GAL-03 and CAT-18 presented ambiguous clinical manifestations, with a display of CRC at around their forties, which is more likely a feature of FAP, but a number of polyps and an onset typical of the attenuated phenotype. Hence, we have thought it appropriate to classify them as AFAP. We realise that classification of such patients is difficult, since it is well-known there is a lack of agreement concerning the exact diagnostic criteria that should be used to classify attenuated polyposis [43].

As expected, most of the bilallelic MUTYH carriers were found in families with an autosomal recessive model of inheritance, or in cases with apparent sporadic presentation. However, we identified three patients (CAT-14, GAL-22 and CAT-16) with a family history of vertical transmission of CRC; similar results had already been described [44].

Biallelic MUTYH mutations have been consistently linked to higher CRC susceptibility. However, the risk for monoallelic MUTYH carriers remains controversial. Balaguer et al. [26] used a meta-analysis of published case-control studies and concluded that monoallelic MUTYH carriers were not at increased risk for CRC, although an effect of bordeline statistical significance was observed for p.Tyr165Cys. In the present study, monoallelic changes with predicted functional relevance (p.Tyr165Cys, p.Gly382Asp, p.Val232Phe) were found in 3/45 patients, and accounted for 6.7% of cases. Nevertheless, they were not included as positive within the overall data, even though the p.Val232Phe was recently shown to reduce glycosylase activity [45].

Five new VUS were identified: c.-56 G>C, c.39 C>T (p.Ala13Ala), c.269A>G (p.Tyr90Cys), c.508C>T (p.Arg170Trp) and c.762 G>A (p.Gln254Gln). We also found p.Arg412Cys, previously described by Aceto et al. [46] and predicted by Polyphen as "possibly damaging". Neither of these variants were found in healthy controls when genotyping 500 chromosomes. It is quite remarkable though, that two of these previously not reported variants (p.Tyr90Cys and p.Arg170Trp), predicted as "probably damaging", were both found in CAT-22. This patient displayed multiple adenomas (50–60) and CRC at 53 years, but had no family history of polyposis. However, further studies are necessary to assess if these two variants are indeed deleterious.