Oncogene (1999) 18, 8044 ± 8047 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc

SHORT REPORT Somatic frameshift in the MBD4 of sporadic colon with mismatch repair de®ciency

Scott Bader*,1, Marion Walker1, Brian Hendrich2, Adrian Bird2, Colin Bird1, Martin Hooper1 and Andrew Wyllie1,3

1Sir Alastair Currie CRC Laboratories, Molecular Medicine Centre, University of Edinburgh, Western General Hospital, Edinburgh EH4 2XU, UK; 2Institute of Cell and Molecular Biology, University of Edinburgh, King's Buildings, May®eld Road, Edinburgh EH9 3JR, UK

Defects of mismatch repair are thought to be responsible sites occurs at elevated rates in mammals and accounts for in hereditary non-polyposis colorectal for a disproportionately high frequency of point and about 15% of sporadic colon cancers. The mutations in human cancers (for example within phenotype is seen as microsatellite instability and is (Rideout et al., 1990)) and genetic diseases (Bird, 1980; known to be caused either by mutations in mismatch Cooper and Youssou®an, 1988; Rideout et al., 1990; repair or by aberrant methylation of these genes Zingg and Jones, 1997). We therefore hypothesized stabilizing their downregulation. Lack of repair of that mutations causing loss of normal MBD4 function microsatellite sequence errors, created during replication, might lead to an increased rate of genome-wide C to T leads to a -prone phenotype. Where mutations transitions, and so allow or promote tumorigenesis. occur within mononucleotide tracts within exons they MBD4 could have a tumour suppressor role similar to cause translation frameshifts, premature cessation of that for mismatch repair (MMR) genes (including translation and abnormal expression. Such MSH2, MSH3, MSH6, MLH1, PMS1, PMS2) that mutations have been observed in the TGFbRII, BAX, are mutated by nucleotide alterations or downregulated IGFIIR, MSH3 and MSH6 genes in colon and other by aberrant methylation (Liu et al., 1995, 1996; cancers. We describe here frameshift mutations a€ecting Herman et al., 1998). The presence of a translated the gene for the methyl-CpG binding thymine glycosy- poly- tract of ten consecutive adenine nucleo- lase, MBD4, in over 40% of microsatellite unstable tides (A10) within exon 3 suggested a region sensitive to sporadic colon cancers. The mutations all appear mutation in microsatellite unstable (RER+) tumours heterozygous but their location would ensure truncation (Aaltonen et al., 1993; Ionov et al., 1993; Thibodeau et of the protein between the methyl-CpG binding and al., 1993), similar to the unstable mononucleotide glycosylase domains, thus potentially generating a tracts within a number of other genes (Markowitz et dominant negative e€ect. It is thus possible that such al., 1995; Rampino et al., 1997; Souza et al., 1996; mutations enhance mutation frequency at other sites in Malkhosyan et al., 1996). We have therefore in this these tumours. A suggestion has been made that MBD4 study screened the MBD4 gene for mutations in (MED1) mutations may lead to an increased rate of colorectal carcinomas, focusing in particular on the microsatellite instability but this mechanism appears A10 tract of RER+ tumours. unlikely due to the nature of mutations we have found. Fresh tumour and matching normal colon tissue samples were collected from patients with colorectal Keywords: MBD4; ; RER; mismatch carcinoma undergoing surgery at the Royal In®rmary repair of Edinburgh NHS Trust. All patients had no known family history of cancer and all but two (43 and 49 years old respectively) were 60 years of age or older. MBD4 was identi®ed as a methyl-CpG binding protein Samples were stored frozen at 7708C and genomic related to the transcriptional repressor MeCP2 and DNA extracted from them later by standard techni- containing a C-terminal domain homologous to DNA ques. RER+ tumours were identi®ed on the basis of repair (Hendrich and Bird, 1998). It has been microsatellite instability demonstrated at least two of demonstrated that the protein can bind to methyl- ®ve interrogated sites while RER7 tumours were CpG/TpG and to symmetrically methylated CpG, and selected on the basis of absence of microsatellite also has the ability to remove Thymine (T) or Uracil instability at any of these sites. The tumours exhibited (U) from mismatched CpG sites (Hendrich and Bird, the histopathological features characteristic of RER+ 1998; Hendrich et al., 1999a). Transition mutation of and RER7 types as described by ourselves and others methyl- (methyl-C) to T at CpG dinucleotide (Bubb et al., 1996; Aaltonen et al., 1993). We screened 23 RER+ sporadic colon cancers (20 primary samples

and three cell lines) across the A10 tract by two *Correspondence: S Bader methods. The region was polymerase chain reaction 3 Current address: Department of Pathology, University of (PCR) ampli®ed and then (a) run on a 10% non- Cambridge, Tennis Court Road, Cambridge CB2 1QP Received 17 May 1999; revised 14 September 1999; accepted denaturing acrylamide gel under conditions suitable to 14 September 1999 detect a single base-pair di€erence in length, and (b) Frameshift mutations in MBD4 in sporadic colon cancer SBaderet al 8045 cycle sequenced. We found that a total of ten (nine is statistically signi®cant (P50.001, w2 test with the primary and one cell line) cases were mutated (43%), Yates continuity correction). of which eight had lost one A and two had gained one We screened the entire coding region of the gene in A (Figure 1), with concordance between the detection the 23 RER+ tumours, and in 52 RER7 tumours, by methods used. The cell line HCT116 (MLH1 mutant) single-stranded conformation polymorphism (SSCP) was clearly heterozygous for the mutation, apparently assay looking for additional mutations. Genomic still retaining a normal A10 sequence in addition to an DNA was PCR ampli®ed for 30 cycles, each cycle

A9 allele (Figure 1c). Both normal and mutant alleles being (948 30", 588 30"), in the presence of Taq are expressed at the mRNA level in roughly equivalent polymerase (Life Technologies UK) and 33P adATP amounts (data not shown). The status of mutations in (ICN). PCR products were run on 0.56 SequaGel MD primary tumours is more dicult to determine, as the gels (National Diagnostics) containing 10% glycerol coincidence of normal and mutated sequence might and 0.66 TBE for 18 h at 6W at room temperature. re¯ect either heterozygosity in the tumour or contam- Primers for PCR were derived from genomic sequence ination by normal stroma of tumour with no residual (Hendrich et al., 1999b) and were located just inside normal alleles. We do, however, suspect that the introns ¯anking splice acceptor and donor sites of each primary tumours with mutations were heterozygous: of the seven smaller exons, or in overlapping segments the tissues analysed comprised mainly tumour yet the along exon 3, a total of 14 primer pairs (primer band intensities of mutant alleles varied from weaker sequences available from the authors upon request). than, to closely similar to, those of the wild-type We found no other mutations in any of the RER+ alleles. Indeed, their relative intensities were often very primary tumours or cell lines. We found only one similar to those derived from heterozygous HCT116 mutation in an RER7 tumour, in exon 2 encoding the cells, where no stromal contamination was possible. methyl-CpG binding domain (G to C, causing a glycine The matching normal tissues for the mutated primary to alanine mis-sense change). The nucleotide di€erence tumours did not show any alterations in the exon 3 A10 was also seen in normal tissue from the same patient, tract, as determined by either non-denaturing gel or suggesting either a polymorphism or an inherited cycle sequencing analysis. The A10 tract was also mutation. If it is a polymorphism, it is extremely rare normal in 36 microsatellite stable (RER7) primary as it was not seen in any of the other 74 colon patients, tumours and three RER7 colon cancer cell lines nor blood DNA from 110 independent normal

(Figure 1b,c). This association of MBD4 A10 tract individuals. Two authentic exonic polymorphisms mutations with RER+ status in sporadic colon cancers were also identi®ed, one silent (C to T, cDNA nt

Figure 1 MBD4 is mutated in some RER+ colon tumours but not in matched normal tissue nor RER7 colon tumours. (a) Matched pairs of primary normal (N) and tumour (T) tissues from three RER+ cases. All three tumours carry a mutation, two with a deletion of one A, one with an insertion of one A. A 95 basepair section of exon 3 incorporating the A10 tract was PCR ampli®ed for 30 cycles, each cycle being (948 30", 588 30"), using primers MBD4.13a (GATGCTGGAGCATGTGGT) and MBD4.12a (CAGAACAAAAATTTGATCCTGAACTC) in the presence of Pfu polymerase (Stratagene). The primer MBD4.13a was then end labelled with 33P gdATP (ICN) using T4 polynucleotide kinase (Life Technologies UK), and used to label the original PCR products by performing two rounds of PCR using the conditions described above. Radiolabelled PCR products were diluted with non-denaturing dye and loaded onto a non-denaturing 10% acrylamide 16TBE gel, run overnight for 16 h at 10W. The gels were dried and exposed to autoradiographic ®lm. (b) Primary tumours from six RER7 cases, without any changes in the A10 tract. Experimental protocol as in (a). (c) Colon cancer cell lines. RER7: 1, HT29, 2, SW480, 3, COLO-320; RER+: 4, HCT15, 5, HCT116, 6, LOVO. Experimental protocol as in (a). (d) Sequences including the A10 tract from examples of a mutated 71A RER+ tumour, normal A10 tissue and a mutated +1A RER+ tumour. PCR products were made using primers MBD4.11 (GATAGCAAAAGAGAATCTGTGTGTA) and MBD4.14 (CTTTCCTTTCCACAACTTCTACT), then directly sequenced with primer MBD4.36 (GTCAGCTTGATAGAACTGTCT) using Thermosequenase radio-labelled terminator cycle sequencing kits (Amersham-USB Life Sciences). Sequencing products were run on 6% acrylamide 7 M urea gels for 2 h at 80W and exposed to autoradiographic ®lm Frameshift mutations in MBD4 in sporadic colon cancer SBaderet al 8046 1589) and one coding (G to A, cDNA nt 993, changing MBD1 and MBD2. Thus, not only would G/T Alanine to Threonine). Both were con®rmed as natural mismatch repair be impaired, but also normal gene polymorphisms, with allelic frequencies 89 and 11% in regulation, through interference with repression of each case, upon screening blood from 110 from methylated regions. independent, normal individuals. The MBD4 gene is A gene with sequence identical to MBD4 has been therefore mutated, with one possible exception, only in cloned recently by another group using the yeast sporadic colorectal cancers with the RER+ phenotype, interaction trap system in which its protein product,

and then only within its A10 tract. via the C-terminal region, was found to bind MLH1 Our data clearly show that the gene encoding the protein (Bellacosa et al., 1999). Although in that report recently described methyl-CpG binding domain pro- it was proposed that MBD4 has endonuclease activity, tein, MBD4, is a target for frequent mutation in this has not been found by others (Hendrich et al., RER+ colorectal tumours. The mutations in our 1999a). It was reported by Bellacosa et al. (1999) that a colon samples invariably involve addition or deletion deletion mutant lacking the N-terminal methyl-C

of a single adenine nucleotide within the exon 3 A10 binding domain caused an increase in microsatellite tract, creating frameshifts that encode a premature instability due to inappropriate binding to MLH1 and translation stop after 12 or three altered amino acids an inhibition of that protein's normal repair function. respectively. These truncated proteins, if expressed, It was therefore suggested that MBD4 mutations might would lack the glycosylase domain but retain the N- occur in RER+ tumours that lack primary de®ciencies terminal half of the protein including the methyl-CpG in MMR genes. These MBD4 mutations might binding domain. A similar but even shorter truncation inactivate the methyl-C binding function but still of wild-type MBD4 comprising only the methyl-CpG allow complex formation with MLH1 in a dominant binding domain and lacking the glycosylase domain, negative fashion. Our data show a di€erent phenom- has been shown still to bind methylated DNA in vitro enon: mutations in MBD4 do occur in authentic,

(Hendrich et al., 1999a). Thus, the truncated proteins naturally occurring RER+ cases but only in the A10 of RER+ tumours would lack mismatch repair tract, truncating the protein and so removing the function but are expected to retain the ability to bind glycosylase and MLH1-binding domains but retaining methylated DNA sites, whether normal or mismatched, a normal methyl-C binding domain. Mutation of the and thus could compete with full-length MBD4 protein methyl-C binding domain itself as suggested does not, in a dominant negative manner. therefore, contribute signi®cantly to microsatellite

The absence of mutations outwith the A10 tract instability in vivo because of the rarity of the necessary suggests that all the observed mutations in RER+ methyl-C binding domain mutations in RER+ tumours are likely to be the result rather than the cause tumours. It is possible that truncation of MBD4, of defective mismatch repair. The overall incidence in losing the MLH1-binding domain, might further our series (43%) is similar to that observed in the most enhance the accumulation of mutations at microsatel- labile microsatellite sites used for identi®cation of the lite sequences via loss of a potential recruitment RER+ phenotype, consistent with the view that the function. It is perhaps of note that the one other changes in MBD4 might merely re¯ect the presence of sequence abnormality we found in a colon tumour was microsatellite instability. The incidence of mutations in a mis-sense mutation within the methyl-CpG binding

MBD4 is less than that in the A10 tract in the domain. Were this change to impair normal methyl- transforming growth factor b type II receptor CpG binding function of MBD4, it would mimic the (TGFbRII) gene (80%) (Markowitz et al., 1995), but mutant form created by Bellacosa et al. (1999), and

close to that observed in a similar G8 tract in the BAX would in theory still be able to complex with MLH1

gene (51%) (Rampino et al., 1997) and MSH6 (C8, and so increase microsatellite instability. However, this 30%) (Malkhosyan et al., 1996) and substantially does not appear to be the case since the tumour is higher than in other sites including insulin-like growth RER7. The biological signi®cance of both the

factor II receptor (IGFIIR)(G8, 9%) (Souza et al, observed binding of MBD4 to MLH1 in vitro and 1996). the experimental e€ects of N-terminal truncated MBD4 Our data, however, do not exclude the possibility of protein on microsatellite instability thus remain to be

positive selection in favour of the MBD4 A10 mutation determined. during tumour development. If, as discussed above, such mutations exert a dominant negative e€ect, this would result in accumulation of mutations at methyl- CpG sites throughout the genome compounding the Acknowledgements e€ects of MMR gene-related microsatellite instability. The work of this study was supported by grants from the Furthermore, the truncated protein might exert a gain Cancer Research Campaign (S Bader, C Bird, M Hooper of function e€ect by inhibiting the binding not only of and A Wyllie), the Scottish Hospitals Endowment Re- the normal MBD4 protein, but also other methyl-C search Trust (M Walker) and the Wellcome Trust (B binding domain-containing proteins such as MeCP2, Hendrich and A Bird).

References

Aaltonen LA, Peltomaki P, Leach FS, Sistonen P, Pylkkanen Bellacosa A, Cicchillitti L, Schepis F, Riccio A, Yeung AT, L, Mecklin J-P, Jarvinen H, Powell SM, Jen J, Hamilton MatsumotoY,GolemisEA,GenuardiMandNeriG. SR, Petersen GM, Kinzler KW, Vogelstein B and de la (1999). Proc. Natl. Acad. Sci. USA, 96, 3969 ± 3974. Chapelle A. (1993). Science, 260, 812 ± 816. Frameshift mutations in MBD4 in sporadic colon cancer SBaderet al 8047 Bird AP. (1980). Nucl. Acids Res., 8, 1499 ± 1504. Liu B, Parsons R, Papadopoulos N, Nicolaides NC, Lynch Bubb VJ, Curtis LJ, Cunningham C, Dunlop MG, Carothers HT,WatsonP,JassJR,DunlopM,WyllieA,Peltomaki AD, Morris RG, White S, Bird CC and Wyllie AH. (1996). P, de la Chapelle A, Hamilton SR, Vogelstein B and Oncogene, 12, 2641 ± 2649. Kinzler KW. (1996). Nature Med., 2, 169 ± 174. Cooper DN and Youssou®an H. (1988). Hum. Genet., 78, Malkhosyan S, Rampino N, Yamamoto H and Perucho M. 151 ± 155. (1996). Nature, 382, 499 ± 500. Hendrich B, Abbott C, McQueen H, Chambers D, Cross S Markowitz S, Wang J, Myero€ L, Parsons R, Sun LZ, and Bird A. (1999b). Mammalian Genome, 10, 906 7912. Lutterbaugh J, Fan RS, Zborowska E, Kinzler KW, Hendrich B and Bird A. (1998). Mol. Cell. Biol., 18, 6538 ± Vogelstein B, Brattain M and Willson JKV. (1995). 6547. Science, 268, 1336 ± 1338. Hendrich B, Hardeland U, Ng H-H, Jiricny J and Bird A. RampinoN,YamamotoH,IonovY,LiY,SawaiH,Reed (1999a). Nature, 401, 3017304. JC and Perucho M. (1997). Science, 275, 967 ± 969. Herman JG, Umar A, Polyak K, Gra€ JR, Ahuja N, Issa J- Rideout WM, Coetzee GA, Olumi AF and Jones PA. (1990). PJ, Markowitz S, Willson JKV, Hamilton SR, Kinzler Science, 249, 1288 ± 1290. KW,KaneMF,KolodnerRD,VogelsteinB,KunkelTA Souza RF, Appel R, Yin J, Wang S, Smolinski KN, and Baylin SB. (1998). Proc. Natl. Acad. Sci. USA, 95, Abraham JM, Zou T-T, Shi Y-Q, Lei J, Cottrel J, Cymes 6870 ± 6875. K,BidenK,SimmsL,LeggettB,LynchPM,FrazierM, Ionov Y, Peinado MA, Malkhosyan S, Shibata D and Powell SM, Harpaz N, Sugimura H, Young J and Meltzer Perucho M. (1993). Nature, 363, 558 ± 561. SJ. (1996). Nat. Genet., 14, 255 ± 257. Liu B, Nicolaides NC, Markowitz S, Willson JKV, Parsons Thibodeau SN, Bren G and Schaid D. (1993). Science, 260, RE, Jen J, Papadopoulos N, Peltomaki P, de la Chapelle 816 ± 819. A, Hamilton SR, Kinzler K and Vogelstein B. (1995). Zingg J-M and Jones PA. (1997). Carcinogenesis, 18, 869 ± Nature Genet., 9, 48 ± 55. 882.