Somatic alterations of the SMAD-2 gene in human colorectal cancers.

The SMAD-2 gene, which is located at 18q21, has been identified as a candidate tumour-suppressor gene from work on colorectal cancers. The aim of the present study was to determine the clinical alterations and the significance of its mutations in a series of colorectal cancers previously examined for SMAD-4/DPC-4 gene. Mutation analyses of the SMAD-2 gene were carried out on cDNA samples from 36 primary colorectal cancer specimens using a combination of the polymerase chain reaction (PCR), single-strand conformation polymorphism (SSCP) and DNA sequencing. Only one missense mutation (2.8%), producing an amino acid substitution in the highly conserved region, and two homozygous deletions (5.5%) of the total coding region of the SMAD-2 gene were detected in the 36 cancers. The SMAD-2 gene may play a role as a candidate tumour-suppressor gene in a small fraction of colorectal cancers. However, allelic loss at 18q21 is very often seen in this type of tumour. Even in combination with changes in SMAD-4, the observed frequency was not sufficient to account for all 18q21 deletions in colorectal cancers. Thus, another tumour-suppressor gene, such as DCC, discovered as the first tumour-suppressor candidate in the region may also exist in this chromosome region. ImagesFigure 1Figure 2

Tumour-suppressor genes are charactenrzed by alterations that inactivate both alleles in cancers (Knudson et al. 1985). this often being accomplished by intragenic mutations in one allele accompanied by loss of a chromosomal region containing the other allele termed loss of heterozygosity (LOH). LOH at 18q21 has been reported to be frequent in various types of human tumours such as pancreatic and colon cancers (Vogelstein et al. 1989: Cliby et al. 1993: Brewster et al. 1994: Shibagaki et al. 1994: Hahn et al. 1995: Vogelstein et al. 1988). Recently. we showed that the SMAD-4 gene. which was initially isolated from 1 8q2 1 as a candidate tumour-suppressor gene for pancreatic cancers (Hahn et al. 1996). is somatically mutated in a proportion of colorectal cancers in vivo (Takagi et al. 1996). However. the observed frequency of SMAD-4 mutations was significantly lower than expected from the numbers of 18q21 deletions in colorectal cancers. indicating that another tumour-suppressor gene might be present in this chromosome region. For instance. the DCC was the first candidate tumour suppressor in the region. although data on DCC mutations in the same set of tumour samples presented here are not available at present because of its length and complexity.
More recently. a novel gene. termed SMAD-2. has been isolated and revealed to be closely related to the SMAD-4 gene (Riggins et al. 1996: Eppert et al. 1996. Of interest. it w-as mapped to the 18q21 region within a short distance. 3 Mb. from the SMAD4 gene. These observations led us to examine whether the SMAD-2 gene might also be mutated in the same set of colorectal cancers already examined for SMAD-4 gene changes. We herein report our results for SMAD-2 mutations from 36 colorectal cancer specimens. taken at surgery from Japanese patients. together with comparatis e data for the molecular status of the SMAD4 gene. Somatic in vivo SMAD-2 mutations were found. although at very low frequency. providing further evidence that alterations of MAD-related genes at 1 8q2 1 are indeed involved in the pathogenesis of colon cancers.

MATERIALS AND METHODS Patients and tissue samples
Samples of tumours and matched control normal colorectal tissue located far from the tumour site were collected at surgery from 36 patients diagnosed histologically as havsing colorectal cancers. All tissues were quickly frozen in liquid nitrogen and stored at -80°C until analysed.
Reverse transcriptase polymerase chain reaction (RT-PCR) Two micrograms of total RNA. isolated and purified by standard procedures. were reverse transcribed usin-0.3 ,g of random primers. 20 U of RNase inhibitor. 10 mmol 1-' of deoxynucleoside triphosphates and 20 U of reverse transcriptase (Takara Biomedicals. Kyoto. Japan) using, the manufacturer's suggested reaction conditions. The polymerase chain reaction (PCR) was performed in 25-pg reaction mixtures containing, the followinga: 1 pi of the complementarv DNA mix. 2.5 jl of 10 x PCR buffer  Detection of single-strand conformation polymorphism (SSCP) The SSCP analysis was performed essentially as described by Orita et al (1989) Figure 3 Schematic diagram of the SMAD-2 protein as well as locations of the predicted amino acid alterations resulting from the identified mutations in case 31. The amino acid sequence of SMAD-2 from amino acids 431 to 454 of the MH2 domain are aligned with the corresponding regions in SMAD-1 and SMAD-4. Conserved sequences are highlighted (black box) mixture wvas applied to 6%c non-denaturing polyacrvlamide gels (acrylamide and N-N'-bisacrvlamide. 28:2. and 10%7c glvcerol) and electrophoreticaly separated at 30 W constant power for 4 h. The resultingy bands were visualized after autoradioggraphy with Kodak XAR films for day. Abnormal samples were repeatedly tested in independent PCR reactions and with separate gel loadings to ensure reproducibility.

Sequence analysis for mutations
After digestion w ith EcoRI-SacI. the RT-PCR products of colorectal cancer specimens shosing abnormal PCR/SSCP pattems were resolved on 0.8% agarose gels and isolated using Geneclean II (Bio 101. La Jolla. CA. USA). After cloning into EcoRI-SacI sites of pBluescript SKII(-) (Stratagene. La Jolla. CA. USA). plasmid DNAs prepared from pooled clones were sequenced by the dideoxv chain termination method (Suzuki et al. 1992). Identified mutations were confirmed by separate cDNA/PCR amplification and subsequent sequencing. RT-PCR products of the corresponding normal colorectal RNAs vk-ere also subjected to PCR/SSCP and sequencing analyses.
Clinicopathological data Data including gender. age. stage of disease and histopathological findings were available from the clinical and pathological records. A clinicopathological staging derised from the oniginal Dukes' classification was applied. Stage A tumours were defined as being confined to the bowel wall. stage B tumours as extending into the British Joumal of Cancer (1998) 78(9). 1152-1155 Tumour G A T C .. I pericolic/perirectal fat or beyond the serosa or both, stage C tumours were either stage A or stage B with lymph node metastases, and stage D tumours were those with distant metastases. The differentiation grade was assessed.

RESULTS
The present examination of 36 colorectal cancer specimens using the RT-PCR and analysis of single-strand conformation polymorphisms (RT-PCR/SSCP), yielded positive data for three cases. In case 31, a distinct mobility shift was present in the colorectal cancer specimen but not in the corresponding normal tissue, indicating a somatic nature for the change. Both mutant and wild-type alleles were expressed at similar RNA levels indicating retention of heterozygosity ( Figure 1B). In two colorectal specimens, cases 4 and 5, the total coding region of both alleles were absent, suggesting possible homozygous deletions at 1 8q2 1 (data not shown).
Sequence analysis of the normal and colorectal cancer specimens for case 31 showed the presence of a somatic missense mutation. A change of CAT to CGT in a highly conserved residue at codon 441 ( Figure 2) within the MAD homology 2 (MH2) region of the SMAD-2 protein was found, leading to an amino acid substitution (change from histidine to arginine) (Figure 3).
Correlations between mutations and clinicopathological status are shown in Table 1. The SMAD-2 mutation was present in an early stage lesion, whereas the SMAD-4 mutations were primarily in cases with liver and lymph node metastasis.

DISCUSSION
The vast majority of human epithelial and lymphoid malignant tumour cell lines demonstrate escape from transforming growth factor (TGF-P)-mediated growth control, which may represent an important step in tumour progression (Fynan et al, 1993). Recently, several novel human genes related to the Drosophila gene called MAD, thought to transduce signals from TGF-, family members, have been identified, with SMAD-2 being one of these (Sekelsky et al, 1995;Graff et al, 1996;Hoodless et al, 1996;Savage et al, 1996). We here identified one somatic mutation in the highly conserved MH2 domain of the SMAD-2 gene in 36 colorectal cancer specimens. This alteration would be expected to disrupt regulation of phosphorylation by the TGF-f signalling pathway.
Previously, we reported detection of alterations of the SMAD-4 gene in a subset of the present series of colorectal cancers (Takagi et al, 1996). In total, one SMAD-2 and five SMAD-4 mutations and two homozygous deletions of the two genes were observed ( Table 1). The SMAD-2 mutation was present in an early stage lesion, whereas the SMAD-4 mutations were primarily in cases with liver and lymph node metastasis (P = 0.0026 and P = 0.0023 respectively by the chi-squared test). Further studies should be carried out to clarify the relation between mutations of MAD-related genes and disease stage of colorectal cancers.
The discrepancy between the frequencies of SMAD-2 and SMAD-4 mutations and that of allelic loss at 1 8q2 I still remains to be explained. Using RNAs from microdissected tumour specimens, RT-PCR/SSCP analysis might yield higher mutation frequencies of MAD-related genes. However, another possibility is that there is yet another putative tumour-suppressor gene at 1 8q2 1 playing a role in colorectal carcinogenesis. The other alternative is that MAD-related genes might be inactivated by other molecular aPreviously published in Takagi et al (1996). bAll cases with liver metastasis. mechanisms such as aberrant hypermethylation leading to transcriptional repression (Baylin et al, 1991), although it might not play a major role.
In conclusion, together with our previous demonstration of SMAD-4 alterations in colorectal cancers, the present findings indicate that the biological and biochemical functions of the MADrelated genes, SMAD-2 and SMAD-4, warrant further investigation to gain insights into the molecular pathogenesis of neoplasia. Future studies may lead to identification of another, yet unidentified, tumour-suppressor gene(s) linked with the reported frequent l 8q21 deletions in colorectal cancers.