DISCUSSION

Previous studies have suggested that the gene predisposing to JP is located in chromosomal band 10q22-24, in the region where the PTEN, gene associated with CS and BRR is located. Jacoby and colleagues (1997) first reported a de novo interstitial deletion of 10q22.3-24.1 in a single JP patient with multiple congenital abnormalities (Jacoby et al., 1997a). To further analyze the relevance of this finding, they studied allelic loss at 10q22 in 47 juvenile polyps from 16 unrelated JP patients. As a result, 83% of these polyps were found to have somatic deletion at 10q22 (Jacoby et al., 1997b). These findings were interpreted as evidence for a tumor suppressor gene on 10q.

In our study, PTEN mutations were analyzed among probands from 14 JP families and 11 sporadic cases and no mutations were detected. The lack of germline PTEN mutations in a total of 25 unrelated JP patients argues that PTEN is not the JP susceptibility gene. Also the possibility of whole gene deletion was excluded among those four cases in which heterozygous sequence variant in intron 8 of PTEN was present.

The possible linkage of JP to chromosomal region 10q22-24, which includes PTEN gene and the putative JP locus (JP1, Jacoby et al., 1997b) was also analyzed in this study. Four microsatellite markers covering the PTEN locus and flanking 20 cM were used to generate haplotypes for members of eight JP families. The linkage analysis excluded linkage to PTEN locus and, hence excluded the possibility of both a PTEN promoter and intron mutations in at least these eight families. Thus, our data suggested that PTEN is not the gene responsible of JP.

Later, another negative study was published by Riggins and colleagues (1997). In their study, 11 JP cases were analyzed for PTEN mutations and no mutations were found (Riggins et al., 1997).

There are however two reports where PTEN mutations have been described in JP patients. Lynch and colleagues (1997) reported one family, which was thought to have both JP and CS, as having nonsense mutation in PTEN (Lynch, et al., 1997).

In the study by Olschwang and colleagues (1998) 14 JP patients were screened for PTEN mutations and three variants were observed in three JP patients. The first

variant was a 1 bp deletion at codon 232. This frameshift mutation leads to a stop codon at position 255. The patient having this variant was a 74 year old man diagnosed with anemia, hypoalbuminaemia, polyps in the GI tract and laryngeal cancer. The second variation was a T to G transversion at codon 35, predicted to substitute arginine for methionine. This patient was ten years old and had a history of rectal bleeding. Juvenile polyps were found throughout the GI tract. The third variant was a T to G transversion at the second position of the consensus splicing sequence of the donor site of exon 6 (intronic variant). This change was predicted to lead to the skipping of at least exon 6 in the mRNA, resulting in a shift of the translational reading frame. The patient with this variant was 14 years old when he underwent colonoscopy that revealed juvenile polyps (Olschwang et al., 1998).

These reports showing that four individuals with juvenile polyposis had PTEN germline mutations would appear to confirm that PTEN is the predisposing gene on 10q in some families with JP (Lynch et al., 1997; Olschwang et al., 1998). However, one of these patients was described as having both CS and JP (Lynch et al., 1997), whereas the other three had no family history of JP (Olschwang et al., 1998). This raises the question whether these patients were truly affected with JP or CS (Eng and Ji, 1998).

To localize the gene predisposing to JP, a whole genome wide search was performed in three Finnish JP families (families 2, 4 and 5, data not shown). Simultaneously and independently, Howe and colleagues succeeded in localizing the gene predisposing to JP on chromosome band 18q21.1, between markers D18S1118 and D18S487 (Howe et al., 1998b). This interval contains two putative tumor suppressor genes, DCC and SMAD4 (Figure 1).

The DCC gene was cloned in 1990, and it encodes a cell-surface protein containing homology with N-CAM (Fearon et al. 1990). There have been many reports on the loss of heterozygosity at the DCC gene locus in human colon cancers (Kikuchi-Yanoshita et al., 1992; Thiagalingam et al., 1996), suggesting that DCC might be a tumor suppressor gene. However, mutations in the coding region of DCC seem to be rare (Cho et al. 1994) and the position of DCC as a candidate tumor suppressor is not clear.

SMAD4 (DPC4) was first identified through deletion mapping in sporadic pancreatic cancers (Hahn et al., 1996). It is inactivated in nearly half of sporadic pancreatic carcinomas and in a subset of breast, ovarian, and colorectal tumors (Schutte et al., 1996; Thiagalingam et al., 1996). The SMAD4 gene was first called DPC4 (deleted in pancreatic carcinoma, locus 4), since it was the fourth marker to be investigated for homozygous deletions in pancreatic cancer. Later the name SMAD4 has been used in addition to DPC4, as this gene is a member of the SMAD family of genes related to the Drosophila Mad and C. elecans Sma genes (Sekelsky et al., 1995; Hahn et al., 1996; Savage et al., 1996).

The mutation screening was performed for both of these genes simultaneously. After sequencing 14 DCC exons and all 11 SMAD4 exons, a 4 base pair deletion in exon 9 of SMAD4 was detected. After this finding the mutation screening was limited only to the SMAD4 gene. In total, a set of 12 independent JP cases were analyzed for SMAD4 mutations (studies II and III). Among these, mutations were detected in 5 families and 3 sporadic cases (62%).

Interestingly, the mutation was the same in four out of five familial cases (4 bp deletion in exon 9). This deletion causes a frameshift that creates a new stop codon at codon 434. In the recent study SMAD4 mutations were analyzed from 11 unrelated JP patients and mutations were found in three of them (Friedl et al., 1999). In two of these patients the mutation was the same 4 bp deletion in exon 9, described earlier in four unrelated patients (studies II and III). According to the haplotype analysis, these two patients did not share common alleles for the markers D18S363 and D18S1110 (markers located close to SMAD4) (Friedl et al., 1999). Since this 4 bp deletion has been detected in one kindred originating from Finland, in three kindreds originating from the United States, and in two kindreds originating from Germany, the defect seems to represent a mutational hotspot rather than an ancestral founder mutation.

Other germline SMAD4 mutations detected in JP are a 1 bp insertion in exon 5 (study I), a 2 bp deletion in exon 8 (study I), an A to C transition in exon 8 (study III), a C to G transversion in exon 4 (study III), a 2 bp deletion in exon 6 (Friedl et al., 1999) and a C to T transition in exon 8 (Houlston et al., 1998). Mutations detected in studies II and III, or reported in the literature are summarized in Table 1.

F= familial case, S= sporadic case

Table 1. SMAD4 germline mutations in JP families/patients reported in the literature.

On the basis of structural and functional criteria, the SMAD family can be divided into three subgroups. The first group includes so called receptor regulated SMADs that act as direct substrates for receptors (in human those are SMADs 1, 2, 3, 5 and 8) (Liu et al., 1996; Eppert et al., 1996; Chen et al., 1997; Nishimura et al., 1998). The second group includes SMADs that are not direct receptor substrates, but whose function is essential for signalling by the receptor regulated SMADs. SMAD4 belongs to the second group. The third group includes the inhibitory SMADs (SMADs 6 and 7) (Hata et al., 1998; Hayashi et al., 1997; Nakao et al., 1997). All these SMAD proteins have a domain structure composed of highly conserved N-terminal and C-terminal domains known as MH1 and MH2 (Mad homology domains 1 and 2) which are joined by a linker region (Figure 3). The receptor regulated SMADs are directly phosphorylated at their carboxyl terminus by type I TGFβ superfamily receptors. SMAD4 lacks this carboxy-terminal phosphorylation sequence and thus can not behave as direct substrate for the type I receptors (Macias-Silva et al., 1996; Zhang et al., 1996; Krezschmar et al., 1997). In the inactive state, SMAD4 is located in the cytoplasm as a homo-oligomer and remains in an inactive conformation through an interaction between the MH1 and MH2 domains. After ligand stimulation and phosphorylation of the receptor regulated SMADs, SMAD4 forms hetero-oligomers with receptor regulated SMADs and translocates into the nucleus (Lagna et al., 1996; Liu et al., 1997; Whitman, 1998). In the nucleus this SMAD hetero-oligomer complex may specifically associate with a nuclear transcription factor, which then targets the SMAD transcriptional activation function to a specific promoter (Liu et al., 1997). In addition to targeting promoters via trancription factors, SMADs have a site-specific DNA-binding activity. It has been demonstrated that the MH1 domain of SMAD protein may directly bind to the promoter or enhancer region of the genes, thus regulating the transcription (Whitman, 1998). The third mechanism for SMAD protein signalling in cells is transcriptional activation through MH2 domain (Liu et al., 1997) A summary of the domain structure of SMADs and their interaction is presented in Figure 3.

Figure 3. The domain structure of SMADs and their interactions (modified from Hata et al., 1998). In basal state, SMADs form homo-oligomers and remain in an inactive conformation through an interaction between the MH1 and MH2 domains.

The majority of somatic, as well as germline, mutations described in SMAD4 map to the carboxyl-terminus between codons 330-526 (MH2 domain), and half of mutations are located in exons 8 and 9 (Hahn et al., 1996, Hahn et al., 1998; Kim et al., 1996; Schutte et al., 1996; Takagi et al., 1996) (Figure 3). The tumorigenic mutations in SMADs fall in three major categories according to their effect on protein function. The missense mutations in the MH1 domain interrupt DNA binding function, truncating mutations of the MH2 domain remove its ability to activate gene expression, and the missense mutations of the MH2 domain appear to interfere with its nuclear localization (Dai et al., 1998). The location of SMAD4 germline mutations and the hotspots of sporadic SMAD4 mutations are shown in Figure 4.

Figure 4. Summary of the locations of germline SMAD4 mutations in JP patients. The hotspots of somatic mutations in sporadic cancer are also shown.

So far, 12 SMAD4 mutations have been identified in 46 familial and sporadic JP cases (26%) (studies II and III; Friedl et al., 1999; Houlston et al., 1998). In our studies (studies II and III) SMAD4 mutations were detected in eight of 12 JP patients analysed (66%). However, in two recent studies a considerably lower proportion of SMAD4 involvement in JP were reported. Friedl and colleagues analysed 11 unrelated patients with JP and found SMAD4 mutation in three of them (27%). Similarly, Houlston and colleagues analysed 21 JP patients and found a mutation only in one of them (5%). Taken together these results suggest that mutations in SMAD4 predispose to JP, but account only for approximately 25% of cases. Further studies will hopefully clarify the true frequency, though issues such as diagnosis of JP and different mutation detection method utilized (genomic sequencing in studies II, III, and in the study by Friedl and colleagues and conformation-specific gel elctrophoresis in the study by Houlston) may contribute to the discrepancy.

Other explanation for the relatively low SMAD4 mutation frequency among JP patients may be the existence of other gene/genes predisposing to JP. Houlston and colleagues (1998) analyzed eight JP families for linkage to SMAD4 and found no evidence for linkage. It could be excluded in two of the eight kindreds and was unlikely in two others. Of those four families which provided good evidence against linkage to 18q21.1, two were also unlinked to the PTEN locus (Houlston et al., 1998), one was compatible with PTEN linkage but no PTEN mutation was detected (study I) and one has not been tested for PTEN mutations. None of these families had notable features of CS or BRR.

In our studies (studies II and III) SMAD4 mutations were not detected in two families (families 2 and 4; family 2 also excluding linkage to 18q21) and two sporadic cases (sporadic cases 1 and 3), these being negative also in the previous PTEN mutation analysis (study I). Two of these (one family and one sporadic case) display a JP phenotype with heart anomalies; the sporadic case also has a history of Wilms' tumor. The other sporadic case was diagnosed with hereditary hemorrhagic telangiectesia (HHT) and epilepsy.

There is good evidence for the importance of the TGF-β signaling pathway in colorectal carcinogenesis. A large proportion of sporadic colon cancers with MSI have TGF-β type II receptor mutations (Myeroff et al., 1995) and germline variants in the TGF-β type II receptor may predispose to colon cancer (Lu et al., 1998). Also the downstream targets of TGF-β have been implicated in colorectal carcinogenesis. Somatic mutations of SMAD2 and SMAD4 have been found in colorectal cancer in humans and germline Smad3 mutations cause colon cancer in mice (Zhu et al., 1998). It has been hypothesized that the germline mutations in genes (other than SMAD4) encoding different components of the TGF-β signaling pathway may be present in JP.

To study this hypothesis further, we focused on SMAD genes, which are involved in TGF-β signaling via TGF-β type II and I receptors (TGFβR-II and TGFβR-I). SMAD4 is known to be the co-SMAD and its function is essential for signaling by the receptor-regulated SMADs. This group includes SMADs 2 and 3, which are substrates of the TGF-β and activin type I receptors (TGFβR-I and ActR-IB) and mediators of TGF-β and activin signals (Whitman, 1998; Hata et al., 1998). The same pathway also includes SMAD7, which is an inhibitory SMAD. Expression of SMAD7 inhibits the receptor-regulated phosphorylation of SMAD2 or SMAD3 (Hayashi et al., 1997; Nakao et al., 1997). In this study a cDNA based mutation screening was performed for SMAD2, SMAD3 and SMAD7 genes and no mutations were detected in these genes in any of our patients (study III). Recently, similar results were reported by Bevan and colleagues (1999). In their study, 30 unrelated JP patients, who were negative in previous SMAD4 mutation analyses (Houlston et al., 1998) were screened for germline mutations at SMAD1, SMAD2, SMAD3 and SMAD5. No mutations were found in any of the patients screened (Bevan et al., 1999).

Previous results together with the occurrence of other genetic syndromes with juvenile polyposis, suggest genetic heterogeneity in the etiology of JP. BRR, Gorlin syndrome and HHT are genetic syndromes, which commonly are associated with JP. The genes responsible for these conditions may also be considered as candidate genes for JP. The gene predisposing to BRR (and CS) is PTEN, and as discussed in previous chapters it is unlikely that it would also be the gene predisposing to JP. Gorlin syndrome is characterized by multiple basal cell carcinomas of the skin. The gene for Gorlin syndrome has been localized to chromosome 9 and shown to be the human homolog of the Drosophila developmental gene patched (Johnson et al., 1996). The third syndrome, which is occasionally described among JP patients, is HHT (also known as Osler's disease or Osler-Rendu-Weber disease). HHT is an autosomal dominant disorder characterized by multisystemic vascular dysplasia and recurrent haemorrhage (Guttmacher et al., 1995). Two predisposing genes have been identified for HHT. Endoglin, which maps to chromosome 9q33-34 has been identified as the gene for HHT1 (HHT type 1), and ALK1 (activin receptor-like kinase 1 gene), which maps to chromosomal locus 12q11-14, is the gene for HHT2 (HHT type 2) (McDonald et al., 1994; Shovlin et al., 1994; Johnson et al., 1995; Vincent et al., 1995).

One of our JP patients was diagnosed with HHT (study III, sporadic case 3). Similar cases have previously been reported in some other studies. In 1980 two patients (mother and daughter) with HHT and polyposis were reported (Cox et al., 1980). Two years later an autosomal dominant syndrome with juvenile gastrointestinal polyposis, cutaneous telangiectesia, and pulmonary arteriovenous malformations in a father and two children were described. The father died from colon cancer at the age of 36 (Conte et al., 1982). Recently, Desai and colleagues (1998) described one JP patient with HHT and two patients with some features of HHT (Desai et al., 1998).

The association of HHT with juvenile polyposis in a subset of patients raises the possibility that JP could occur as part of a 9q or 12q microdeletion syndrome in some cases, or JP with HHT might be a unique syndrome due to mutation in a single gene. To further study this hypothesis, endoglin and ALK1 mutation screening was performed among those JP patient, where SMAD2, SMAD3, SMAD4, SMAD7 or PTEN mutations were not detected. Both endoglin and ALK1 are members of the TGF-β receptor superfamily and are therefore cosidered as good candidates for JP gene. No mutations were detected in either of these genes (unpublished data). Our data suggest that there are still unidentified genes predisposing to JP, perhaps associated with developmental defects such as cardiovascular anomalies.

HNPCC is an autosomal dominant cancer susceptibility syndrome. The genes responsible for the disease are MSH2, MLH1, PMS1, PMS2 and MSH6 (Leach et al., 1993; Bronner et al., 1994; Papadopoulos et al., 1994; Nicolaides et al., 1994, Miyaki et al., 1997). Inactivation of both alleles of one of these genes results in microsatellite instability (MSI) that is a hallmark of HNPCC tumors. The genes responsible for microsatellite stable (MSS) HNPCC are still largely unknown (Peltomäki et al., 1997; Aaltonen et al., 1993, Nyström-Lahti et al., 1996).

Loss of growth inhibition by transforming growth factor β(TGF-β) is an important step in colorectal tumorigenesis. In HNPCC tumors with MSI, this is mainly due to frameshift mutations within a polyadenine sequence repeat in the TGFβRII gene. It has been proposed that germline mutations in TGFβ RII underlie a subset of MSS HNPCC cases (Lu et al., 1998). Other genes involved in the TGF-β pathway could also be candidates for MSS HNPCC (Grady et al., 1999).

SMAD proteins play a key role in intracellular TGF-β signaling and mutations in some of the SMAD genes disrupt this pathway. To investigate whether germline mutations in SMAD2, SMAD3 or SMAD4 could predispose to MSS HNPCC, we analyzed mutations in these genes in 14 HNPCC kindreds in which germline mutations in MSH2 and MLH1 had been ruled out (Nyström-Lahti et al., 1996; Holmberg et al., 1997). The only changes detected were three polymorphisms in SMAD3. One of these was an A to G transition at nucleotide 545, which changes isoleucine to valine at amino acid 170, in the so called linker-region of SMAD3 protein. This variant has not been reported earlier. We analyzed 110 Finnish controls and 132 additional colon cancer patients by restriction enzyme digestion to clarify the frequency of this polymorphism. In total, seven control individuals (6.4%) and 13 colon cancer patients (8.9 %, including two HNPCC cases) displayed the variant, suggesting that this polymorphism is rather neutral than related to colorectal cancer. Even SMAD2 and SMAD4 genes are mutated in a subset of human colorectal tumors, our results suggest that these genes are not involved in the MSS HNPCC. To date, SMAD3 mutations have not been reported in human cancers. In a recent study SMAD3 mutations were analysed among 35 sporadic colorectal and 15 HNPCC cancers and no mutations were found (Arai et al., 1998). Further work is necessary to unravel the molecular background of MSS HNPCC.

SMAD4 seems to be mutated in only less than 10% of cancers analysed. The only exception is pancreatic cancer, where the mutation frequency is approximately 20% (Takagi et al., 1996; Thiagalingham et al., 1996; Schutte et al., 1996; Hahn et al., 1996; Nagatake et al., 1996; Rozenblum et al., 1997).

Knudson's hypothesis that two hits are required for the full inactivation of a tumor-suppressor gene has been shown to be correct in most human cancers. Traditionally, it has been thought that tumor suppressor genes are inactivated by intragenic mutations and loss of chromosomal material. However, the fact that methylation of CpG islands located in the promoters of genes can cause transcriptional silencing, has led to the suggestion that hypermethylation of tumor-suppressor gene promoters may be one of the mechanisms leading to malignant transformation (Jones and Laird, 1999). Several tumor suppressor genes contain CpG islands in their promoters and some of them, like RB1 and VHL also show evidence of methylation silencing (Herman et al. 1994; Sakai et al., 1991). On the other hand, promoters of some tumor suppressor genes are very rarely hypermethylated. For example, TP53 is mostly inactivated by point mutations, whereas p16 is generally inactivated by homozygous deletion (Hussain and Harris, 1998; Kamb et al. 1994).

There has been two studies on the possible SMAD4 promoter region. The first candidate for the SMAD4 promoter was reported by Minami and colleagues (1998), when they cloned a 1.4 kb fragment of the SMAD4 5'-flanking region from a phage library by using the first coding exon's sequence as a primer. This SMAD4 promoter lacks typical TATA boxes and CpG islands, but contains some TATA-like structures (TAAAAT) as well as some binding sites for transcription factors (Minami et al., 1998). Later, another candidate for the SMAD4 promoter was published by Hagiwara and colleagues (submitted). First, they identified a new non-coding exon (exon 1) by using the 5'-rapid amplification of the cDNA ends (5'-RACE) (Hagiwara et al., submitted). Then SMAD4 promoter was cloned using the exon 1 sequence as a probe. The SMAD4 promoter published by Hagiwara and colleagues is a typical CpG island having a high G+C content (Hagiwara et al., submitted).

In study V, we examined whether hypermethylation of the promoter could be an alternative mechanism for SMAD4 inactivation. The CpG island near non-coding exon 1 was selected for this analysis, since it is well documented that methylation has important regulatory effects especially when involving these CpG rich areas (Bird, 1986). In total, 26 MSI and 16 MSS tumors and corresponding normal DNAs were selected for the analysis and no evidence of hypermethylation was found.

In the recent study SMAD4 mutations were analysed from 176 colorectal tumors and SMAD4 mutation frequencies were found to be 0% in adenoma, 10% in intramucosal carcinoma, 7% in primary invasive carcinoma without metastasis, 35% in primary invasive carcinoma with metastasis and 31% in distant metastasis. So, the frequency of SMAD4 mutations was strongly correlating with the stage of colorectal cancer (Miyaki et al., 1999). This correlation was further confirmed by Koyma and colleagues (1999). They analysed SMAD4 mutations among 64 advanced colorectal cancers and found them in seven tumors. They also demonstrated that in all these seven tumors carrying intragenic mutations in one allele, LOH analysis showed the loss of the other allele. Their results suggested that inactivation of both alleles of the SMAD4 gene occurs in a substantial portion of advanced colorectal cancers (Koyama et al., 1999).

In our study (study V), 22 primary CRCs were taken into SMAD4 mutation screening and the only change detected was a polymorphism in exon 2. The tumors included in our mutation screening study were mainly classified as Dukes' stages A and B (15/22,) and one explanation for the lack of SMAD4 mutations could be that the tumors were mainly limited to mucosa and submucosa and were not metastasizing.

There are many studies reporting SMAD4 mutation analyses for different tumors. In most of them, the mutation screening is limited to the coding exons and small fragments of the introns around these exons. In one recent study, Zhou and colleagues (1999) analysed six endometrial tumors that do not express normal SMAD4, and found mutation in the 5'-untranslated region in two of them (Zhou et al., 1999). The region included in mutation screening was a 331 bp long fragment, which spanned nucleotides -262 to +69 from the transcription start site of SMAD4. This fragment contains most of the important transcription factor binding sites (Minami et al., 1998). To further study whether this 5'-untranslated region is also mutated in colorectal carcinomas, we sequenced the same fragment from 15 MSS and 7 MSI tumor samples. We however found no changes.

It has generally been considered that the genes involved in hereditary cancer syndromes are also mutated in the same type of sporadic tumors (Fearon, 1998). For example APC, the gene responsible for FAP, is involved in up to 80% of all sporadic colon cancers (Beroud and Soussi, 1996). There is good evidence that the hamartomas in PJS, JP and CS can progress to colorectal carcinoma. However, mutations in PTEN, LKB1 or SMAD4 genes in the sporadic colon cancers occur at very low frequencies (Kong et al., 1997; Avizienyte, et al., 1998; Wang et al., 1998b; MacGrogan et al., 1997; Takagi et al., 1996).

There are also examples of cancer syndromes, where a gene, which is mutated in germline is not frequently mutated in the corresponding somatic cancers. The familial breast/ovarian cancer genes, BRCA1 and BRCA2, cause cancer when mutated in the germline, but are only seldom mutated in sporadic cancers (Futreal et al., 1994; Miki et al., 1996; Teng et al., 1996). There is, however, evidence that BRCA1 is inactivated by promoter methylation in some sporadic breast cancers (Dobrovic and Simpfendorfer, 1997). TP53 gene is an another example of discrepancy in somatic and germline mutations. The TP53 mutations were first described in sporadic colorectal carcinomas (Baker et al., 1989), in which they are frequent, however families with germline TP53 mutations (Li-Fraumeni syndrome) are not predisposed to colorectal carcinomas (Knudson, 1993).

Our results together with previously published data strongly suggests that the SMAD4 is not frequently mutated in primary colorectal carcinoma and that the hypermethylation of SMAD4 promoter region is not a crucial mechanism in colorectal tumorigenesis.