(2002) 21, 7585 – 7592 ª 2002 Nature Publishing Group All rights reserved 0950 – 9232/02 $25.00 www.nature.com/onc

Germline, and epigenetic events underlying mismatch repair deficiency in colorectal and HNPCC-related

Siu Tsan Yuen*,1,3, Tsun Leung Chan1, Judy WC Ho2,3, Annie SY Chan1, Lap Ping Chung1, Polly WY Lam4, Chun Wah Tse5, Andrew H Wyllie6 and Suet Yi Leung*,1

1Department of Pathology, The University of Hong Kong, Queen Mary Hospital, Hong Kong; 2Department of Surgery, The University of Hong Kong, Queen Mary Hospital, Hong Kong; 3Hereditary Gastrointestinal Registry, The University of Hong Kong, Queen Mary Hospital, Hong Kong; 4Department of Pathology, Queen Elizabeth Hospital, Hong Kong; 5Department of Surgery, Queen Elizabeth Hospital, Hong Kong; 6Department of Pathology, Cambridge University, UK

High-frequency microsatellite instability (MSI-H) results wild-type RNA by as yet undiscovered mechanisms, or from deficiency in nucleotide mismatch repair. It no detectable abnormality in any of these parameters. contributes significantly to in the human Evidence is presented to indicate that methylation in colorectal mucosa. Here we study 41 colorectal and three proximal region of the hMLH1 is a more other HNPCC-related cancers with MSI-H to provide reliable correlate of transcriptional silencing in colorectal comprehensive information on the mechanisms of cancers than methylation in upstream region. These inactivation of the two major proteins involved, hMLH1 observations have significant implications for manage- and hMSH2. Seventeen of the patients had family ment of patients with MSI-H tumours. histories meeting the criteria for Bethesda grades 1, 2 Oncogene (2002) 21, 7585 – 7592. doi:10.1038/sj.onc. or 3. Of these familial cases, 14 (83%) had early-onset 1205968 disease, defined on the basis of diagnosis prior to the age of 50, but in three the disease was of late onset (450 Keywords: microsatellite instability; promoter methyla- years). A second subset of 20 patients had early onset tion; mismatch repair gene ; colorectal cancer; disease without family history. The remaining seven HNPCC patients were selected to allow comparisons with sporadic, late-onset disease, the molecular basis of which has been extensively reported elsewhere. We stratified the tumours initially on the basis of hMLH1 or hMSH2 protein deficiency, detected by immunohistochemistry, High frequency microsatellite instability (MSI-H) is the and then by analysis of and , phenotypic manifestation of mismatch repair defect. In mRNA transcription, loss of heterozygosity (LOH) at colorectal carcinomas most MSI-H is associated with the hMLH1 and hMSH2 loci, and methylation status in inactivation of either hMSH2 or hMLH1, genes that two regions of the hMLH1 promoter. The functional encode proteins involved early in the process of significance of several of these changes in the MSI-H mismatch repair (MMR) (Peltomaki and Vasen, tumours was confirmed by comparisons with 16 tumours 1997). Whilst germline MMR gene mutation is the with low-frequency microsatellite instability and 56 major mechanism responsible for the mismatch repair tumours with stable microsatellites. As anticipated, deficiency in hereditary non-polyposis colorectal cancer patients with family histories usually showed germline (HNPCC) kindreds, the majority of sporadic late-onset mutation of hMSH2 or hMLH1. In many cases the MSI-H colorectal cancer were attributed to hMLH1 residual normal was silenced in their tumours by promoter hypermethylation (Kane et al., 1997; Herman loss of heterozygosity (LOH). The small subset of late- et al., 1998; Cunningham et al., 1998; Kuismanen et onset, sporadic cases confirmed the preponderance in this al., 2000). Cancers with MSI-H constitute a high group of biallelic hMLH1 promoter methylation. In the proportion of those arising in the colon and rectum of early-onset, apparently sporadic subset there were 11 younger patients, and whilst many of these have clear tumours with hMLH1 deficiency, five with hMSH2 family histories of cancer, several do not. Relatively deficiency and four with no detectable abnormality in little is known of the molecular origins of the MMR expression of either protein. These showed a complex deficiency in such apparently sporadic early-onset mixture of lesions, including germline and somatic cancers, although we and others have shown that some , promoter methylation, LOH, suppression of arise on a background of germ-line mutation in hMSH-2 or hMLH1, despite the absence of family history (Liu et al., 1995; Chan et al., 1999a). A comprehensive understanding of the mechanisms *Correspondence: SY Leung; E-mail: [email protected] and ST Yuen; E-mail: [email protected] underlying MSI, in particular distinguishing between Received 1 May 2002; revised 8 August 2002; accepted 8 August germline and somatic or epigenetic events, could 2002 significantly influence both the management of the Basis of MSI in colorectal cancer ST Yuen et al 7586 patients and the level of surveillance required for their Figure 1. Methylation-specific PCR, which assessed relatives. There is good reason to believe that the MSI- methylation status of CpG sites located in bases 7694, H phenotype requires bi-allelic (‘two hit’) inactivation 7688, 7674, 7672, 7600, 7598 and 7588 (referred of the responsible MMR gene. In practice, however, to here as region 1), was successfully performed in 40 only a single inactivating event was identified in up to MSI-H, 15 MSI-L and 26 MSS tumours. hMLH1 65% of currently reported MSI-H colorectal tumours promoter methylation was detected in 12 of 22 MSI-H (Hemminki et al., 1994; Konishi et al., 1996; cases with hMLH1 protein loss, but also in several Tannergard et al., 1997; Leung et al., 1998; Chan et tumours in which hMLH1 expression was normal: two al., 1999b). This may reflect the restricted number of of 12 MSI-H cases with hMSH2 protein loss, two of inactivation mechanisms that were directly investi- six MSI-H cases with no MMR protein loss, two of 15 gated. Moreover, there are technical limitations to the MSI-L cases, and zero of 26 MSS cases. The incidence detection of mutations, be they germline or somatic, in of promoter methylation was significantly different the MMR genes. These large genes have no mutation between cases with (12/22) and without (6/59) loss of hot spots and hence no simple strategy can detect all hMLH1 protein (P50.0001), but the clear evidence of mutations. The simple detection of hMLH1 promoter region one methylation in six tumours with normal hypermethylation also suffers from several drawbacks. expression of hMLH1 protein prompted search for Firstly, it does not distinguish between mono-allelic or other promoter regions with which the association of bi-allelic inactivation in primary cancer tissue. methylation and protein expression was more specific. Secondly, the specificity of promoter methylation and When bisulphite sequencing was used to identify hence its functional significance varies according to the methylation in a region more proximal to the methods used and the CpG sites examined (Deng et al., transcription start site, covering bases 7178 to 1999). These limitations are particularly important for 7248 (referred to here as region 2), two distinct the group of sporadic early-onset colorectal cancer patterns emerged. DNA extracted from the tumour with MSI, where so far there is no comprehensive was either unmethylated in all eight CpG sites study addressing the relative contributions of germline, (hereafter referred to as the unmethylated pattern), somatic and epigenetic events. Here we characterize the or showed methylation close to or in excess of 50% of basis of MSI-H in a substantial number of MSI-H total DNA at all eight CpG sites (referred to as the colorectal cancers, including familial, sporadic late- methylated pattern) (Figure 2). Of the six tumours onset and sporadic early-onset cases, the latter group with normal expression of hMLH1 protein but with shows an exceptionally high incidence in Hong Kong methylation in region 1 of the hMLH1 promoter, all (Yuen et al., 1997; Chan et al., 1999a). For both showed an unmethylated pattern in region 2. Of the hMSH2 and hMLH1 we sought to identify loss of 12 tumours with loss of hMLH1 expression and protein expression, germline and somatic mutations, methylated region 1, however, seven also showed the mRNA expression and LOH. Methylation was exam- methylated pattern in region 2. Amongst the remain- ined in various regions of the hMLH1 promoter. The ing five tumours, only two (L6, L20) failed to show data show that different mechanisms of MMR clear evidence of bi-allelic inactivation of hMLH1 by deficiency predominate in these three groups. means other than methylation (i.e. somatic or germ- The results of immunohistochemical staining for line mutation together with LOH), and none failed to hMSH2 and hMLH1, and the family or personal show at least one somatic or histories of cancer of 44 individuals with MSI-H (Figure 2B,C). In contrast, of the seven cases with tumours are shown in Table 1 and summarized in region 2 methylation, five showed a predominance of Figure 1. In the majority of these tumours (38/44, methylated DNA in region 2 on bisulphite sequencing 86%) one or other of the hMLH1 or hMSH2 proteins and absence of other germline or somatic inactivation was undetectable. In contrast, both proteins were events. Real time quantitative PCR of these cases readily identified in a nuclear location in all of 16 demonstrated significantly lowered hMLH1 transcript tumours with low levels of microsatellite instability when compared with 10 randomly selected MSS (MSI-L) and 56 tumours with stable microsatellites colorectal cancers (Figure 3). These results strongly (MSS). When the MSI-H tumours were classified on support transcriptional silencing of hMLH1 in these the basis of age and family history into familial cases by bi-allelic promoter methylation in region 2. (satisfying Bethesda criteria 1, 2 or 3), sporadic early- Two further cases showed region 2 methylation and onset (aged 50 years or below at diagnosis) and suppression of hMLH1 expression. In both, approxi- sporadic late-onset (aged above 50 years) groups, all mately 50% of region 2 DNA was methylated, familial cases were found to have lost either hMSH2 suggesting monoallelic inactivation. One of these cases (11/17, 65%) or hMLH1 (6/17, 35%). In the sporadic (L3) also had clear evidence of a germline mutation in early-onset group, 55% (11/20) lost hMLH1 and 25% hMLH1, whilst in the other (L22) a similar mutation (5/20) lost hMSH2. In four cases (20%), however, might be inferred (although we did not detect it nuclear staining for both proteins was retained. In the directly) since there was a family history satisfying sporadic late-onset group, the majority (5/7, 71%) lost Bethesda Criteria 2 and 3. Both these tumours, hMLH1 and none lost hMSH2. therefore, are likely to have arisen through germline The results of hMLH1 promoter methylation mutation of one hMLH1 allele, and transcriptional analysis are listed in Table 1 and summarized in silencing of the other through region 2 promoter

Oncogene Table 1 Summary of promoter methylation, LOH, germline and somatic mutations of MMR genes; and family history in MSI-H colorectal cancers with loss of hMLH1 protein(a), loss of hMSH2 protein(b) and preserved expression of both proteins(c) (a.) Protein expression hMLH1 promoter methylation Case Sex Age Sitea hMSH2 hMLH1 Region 1 Region 2 LOHb 3p/hMLH1 Germline hMLH1 mutationc Somatic hMLH1 mutationc Family historyd

L1 M 46 R + 77 7 L c.793C?T(del exon 10)d Allele freq 0/218 None – L2 M 38 R + 7 + 7 L 1.8-kb genomic del exon 11 (del exon 10-11)* None – L3 M 38 R + 7 + + R c.1151T?A(V384D)# – Allele frequency 0/216 L4 M 46 R + 7 + 7 L None c.1988-1989delAG(del exon 17)# – L5 M 47 R + 77 7 R None c.738-750delATCCAATGCAAAC# – Wild type mRNA loss L6 M 34 R + 7 + 7 L None None – L7 F 44 L + 7 + + MI None None – L8 M 39 R + 77 7 MI None None – L9 M 30 L + 77 7 MI None – L10 F 34 R + 77 7 RNone – L11 M 36 R + 77 7 RNone – L12 F 81 R + 77 7 R c.199G?A(G67R) None 7 L13 F 76 R + 7 + + MI None None – L14 F 79 R + 7 + + R None None – L15 F 74 R + 7 + + R None None – L16 M 75 R + 7 + + R None None – L17 M 44 R + 77 7 L c.1731G?A(del exon 5)^ Allele freq 0/218 None BG3 L18 F 53 R + 7 + 7 L c.1731G?A(del exon 15)^ Allele freq 0/218 BG2+3 L19 F 45 R + 77 7 L c.793C?T(del exon 10)d Allele freq 0/218 BG3 L20 M 39 L + 7 + 7 R 1.8-kb genomic del exon 11 (del exon 10-11)* Wild type mRNA loss BG1 cancer Yuen colorectal ST in MSI of Basis L21 M 30 R + 77 7 R [c.350C?T(T117M); c.790C?T(H264Y)]* BG3 L22 M 51 R + 7 + + MI None BG2+3 tal et

(b.) Protein expression hMLH1 promoter methylation Case Sex Age Sitea hMSH2 hMLH1 Region 1 Region 2 LOHb 2p/hMSH2 Germline hMSH2 mutationc Somatic hMSH2 mutationc Family historyd

S1 M 33 L 7 + 7 L c.1452-1455delAATG – S2 M 39 R 7 ++ 7 R c.1452-1455delAATG Deletion ex2-8 – S3 M 25 R 7 ++ 7 R None c.1759G?C (del exon 11)# – Wild type mRNA loss – S4 F 33 R 7 + 7 R None c.1807G?T(D603Y)# – c.2479G?A(G827R)# – S5 M 42 R 7 + 7 MI None – S6 M 28 L 7 + L c.595T?C (C199R)* BG3 S7 M 29 R 7 + 7 L c.1760delG* BG1 S8 F 60 B 7 + 7 L IVS5+3A?T (del exon 5) BG3 S9 F 40 E 7 + 7 R c.1144insC# BG3 S10 M 42 L 7 + 7 R c.513delG* BG1 S11 M 25 L 7 + 7 R c.1738G?T (E580X)* BG2§ S12 F 42 O 7 + R IVS15+IG?T(del exon 15) BG2 S13 F 39 R 7 + R c.1452-1455delAATG BG1 S14 M 32 R 7 + 7 R c.1452-1455delAATG* BG3 S15 F 34 R 7 + c.1452-1455delAATG* BG3

Oncogene S16 F 39 L 7 + 7 R None c.1738insA# BG3

Continued on next page 7587 Basis of MSI in colorectal cancer ST Yuen et al 7588 d

y methylation. Collectively, therefore, these results suggest that methylation in region 2 of the hMLH1 history promoter is more specific for transcriptional silencing

c than in region 1. Table 1a and Figure 1 summarized the alterations of hMLH1 in colorectal cancers with hMLH1 protein

mutation loss. Overall, mono- or bi-allelic inactivation of

this mutation, recorded in hMLH1 was identified in 83% (18/22) of tumours that d lacked hMLH1 protein expression. As expected from inoma; O=ovarian carcinoma; Germline and somatic Family the published results of others, all but one of the five sporadic, late-onset tumours that lacked hMLH1 *Germline mutation reported in our c osed at younger than 40 years); – =no

b expression showed bi-allelic promoter methylation; in including endometrial, ovarian, gastric, every case this was evident in region 2 as well as region 1 of the promoter. 2p/hMSH2 hMSH2 Also concordant with previous reports were the results from the six familial cases. Most (5 of 6, 83%) c had germline mutations. The ‘second hit’ leading to complete gene inactivation was identified in four of the

mutation tumours (three due to LOH, one with loss of wild type transcript of undetermined cause). In contrast to these two groups, the sporadic, early-

hMLH1 onset patients showed remarkable plurality of mechan- ism for hMLH1 suppression. Three (L1, 2, 3) had c germline mutations, two with LOH of the residual normal allele in their tumours, and region 2 promoter

mutation methylation in the third. Two had normal germline sequence but somatic mutations in their tumours (L4, Germline Somatic LOH 5). In one of these this was accompanied by LOH involving the residual normal allele, but in the other we did not identify a second lesion in DNA, although wild BG1=individuals with family histories of colorectal cancer that meet Amsterdam criteria; d glioma); BG3=individuals with colorectal cancer and a first-degree relative with colorectal cancer b § type RNA was lost. One showed LOH together with L None None L None – R None c.1852-1854delAAG MI None – RRRR None None None None None None None None MI R U MIpromoter None None None methylation None – – – + (L6), although this was found only in region 1. Only one of the 11 hMLH1 deficient 3p/hMLH1 hMLH1 these two patients were from the same family, all other patients were unrelated;

^ tumours in the sporadic early-onset group showed bi- allelic methylation in region 2 (L7), a significantly lower incidence compared with the sporadic late-onset 7 7 group, (P=0.01). Most of the hMLH1 germline mutations identified in this study presented marked diagnostic challenge. The mutations, either in form of single nucleotide substitu- tion that causes splicing defect (patients L17, L18, L1 promoter methylation LOH and L19) or large genomic deletion (patients L2 and 7 7 7 7 L20) (Chan et al., 2001), all generate abnormal

hMLH1 transcripts that are inherently unstable and mimic alternative splicing. The pathogenic significance of these mutations was revealed by examination of the L=loss of heterozygosity; R=retention of heterozygosity; MI=microsatellite instability; U=uninformative.

b microdissected tumour tissue, in which the wild type mRNA was absent, leaving only the truncated this patient had a first degree relative with colorectal cancer, and both were of late-onset. Blank denotes not studied

y transcripts. Table 1b and Figure 1 summarized the mechanisms

Protein expression of hMSH2 protein loss. Germline hMSH2 mutation hMSH2 hMLH1 Region 1 Region 2 was detected in the majority of tumours deficient in ., 1999a; 2001); #novel mutation not previously reported; a hMSH2 expression (12/16, 75%), including 91% (10/ et al 11) of familial and 40% (2/5) of sporadic early-onset cases. The majority of mutations resulted in frameshifts or generation of truncated protein products. Interest- ingly, a 4-bp deletion involving nucleotides 1452-1455 of exon 9, reported by us previously (Chan et al., 1999a) and so far not identified in other ethnic groups, was found in five apparently unrelated patients (three R=cancers of the colon arising at/or proximal to the splenic flexure; L=cancers of the colon or rectum distal to the splenic flexure; E=endometrial carc a the ICG-HNPCC database by others, has been misinterpreted as a missense mutation. previous studies (Chan Case Sex Age Site and/or HNPCC-related extracolonic cancerpersonal and/or a or colorectal family adenoma history (one of of the cancer; cancers diagnosed at age younger than 45 years and the adenoma diagn BG2=individuals with twohepatobiliary, HNPCC-related small bowel cancers, cancer, including transitional synchronous cell carcinomas and of metachronous the colorectal renal pelvis cancers or or ureter, associated or extracolonic cancers ( B=transitional cell carcinoma of bladder. (c.) N1 F 42 R + + + N2 M 36 R + + + N3N4N5 MN6 M 35 M 42 M 66 R 55 L L + L + + + + + + + familial and two early-onset sporadic) and constitutes

Oncogene Basis of MSI in colorectal cancer ST Yuen et al 7589

Figure 1 Graphic presentation of the genetic/epigenetic events in relation to mismatch repair protein expression in 44 MSI-H color- ectal and HNPCC-related tumours. The 41 high-level microsatellite instability (MSI-H) colorectal cancers in this study were identi- fied in a systematic ongoing study of colorectal cancer in Hong Kong examined in the period 1991 to 1998. Three MSI-H cancers arising from other organs were also identified in the same study period from putative HNPCC families. Some of the colorectal can- cer cases have been included in two previous reports of the incidence of MSI and germline MMR gene mutation in our early-onset population (Chan et al., 1999a; Ho et al., 2000). Sixteen low-level microsatellite instability (MSI-L) and 56 microsatellite stable (MSS) colorectal cancers were also randomly selected from these series for comparison. These cases were either from patients who had undergone colectomy in our hospital, and thus there was frozen tumour tissue stored at 7708C available for study, or from patients who were referred to the Hereditary Gastrointestinal Cancer Registry for further work up because of young age at presentation or a positive family or personal history of cancer. The family history of cancer for all patients aged 50 years or below was ascertained using stringent criteria and method as described previously (Chan et al., 1999a; Ho et al., 2000). For those aged above 50, a similar approach was used whenever possible, but otherwise the family history was obtained from the medical records. Cases were designated ‘familial’ if they had a family or personal history of cancer satisfying the Bethesda Criteria 1, 2 or 3 (Rodriguez-Bigas et al., 1997); otherwise, they were considered ‘sporadic’. Because of the methods of case ascertainment and collection, the familial and early-onset MSI-H colorectal tumours are likely to be representative of tumours of this type in our community. The small number of late-onset MSI-H tumours, however, were drawn from a much larger population of sporadic tumours. Venous blood was taken, with informed consent, from the patients for extraction of DNA and RNA. This study was ap- proved by the Ethics Committees of the University of Hong Kong and Queen Mary Hospital. DNA and/or RNA extraction from tumour, normal mucosa or blood leucocytes, and MSI analysis were performed as described (Chan et al., 1999a). Sections were also microdissected in the analysis for LOH, bisulphite sequencing, quantitative RT – PCR and loss of wild type mRNA. Immunostain- ing for hMSH2 and hMLH1 was performed as described (Leung et al., 1999). In vitro synthesized protein assays (IVSP) were used to screen for germline or somatic mutation of hMSH2 (Chan et al., 1999a), and the site and nature of the underlying mutations were confirmed by direct DNA and cDNA sequencing. For cases showing altered hMSH2 protein by immunohistochemistry but without detectable truncated protein products by IVSP, direct sequencing of all exons of hMSH2 was performed. For the detection of hMLH1 mutation, direct cDNA sequencing of the whole gene was performed (Chan et al., 1999a). In the majority of cases, fro- zen tumour tissue was available and mutation detection was performed in the tumour first. Confirmed tumour mutations were sub- sequently searched for in the germline. For those without frozen tumour/normal colorectal tissue, sequencing was performed in cDNA derived from the blood leukocytes. Missense mutations not previously reported to be pathogenic were further analysed by screening for allele frequency in the control healthy Chinese population using single-strand conformation polymorphism. For cases with detectable germline or somatic mutation, loss of heterozygosity (LOH) was determined by analysing the presence of the wild type allele by direct sequencing of DNA extracted from the microdissected tumour. For those without demonstrable muta- tion, or if the mutations were of large genomic deletions, LOH analysis was performed using microsatellites including D2S123, D3S1100, D3S1448, D3S3564 and D3S3593 as described by Wheeler et al. (1999). RNA from microdissected tumour was used for the detection of loss of wild type mRNA in cases with mutation whenever frozen tumour samples were available. Data was cross-referenced with other data including promoter methylation and loss of heterozygosity (LOH) information of the respective gene. Only those cases without identifiable mechanism in the DNA level are highlighted in the results

Oncogene Basis of MSI in colorectal cancer ST Yuen et al 7590

Figure 2 Analysis of colorectal cancer for hMLH1 promoter methylation in region 2 using bisulphite DNA sequencing. CpG sites located in bases 7193 to 7231 relative to the transcription start site of hMLH1 were shown and underlined in (A) and (B). (A)a sporadic MSI-H colorectal cancer (patient L7) with loss of hMLH1 protein but have no detectable germline or somatic hMLH1 mutation. All CpG sites are methylated. Only small amount of unmethylated cytosines are present (likely from contaminating nor- mal cells), indicating biallelic methylation. (B) and (C) a sporadic MSI-H colorectal cancer (patient L4) with hMLH1 promoter hy- permethylation involving region 1 as detected by MS – PCR. All CpG sites are unmethylated in region 2 (B). Direct sequencing of hMLH1 in this tumour shows two hit inactivation in the DNA level, with delAG at the 3’ end of exon 17 and loss of wild type allele (C). The mutations result in skipping of exon 17 and loss of the wild type mRNA. Genomic DNA was prepared from microdis- sected tumour tissue followed by bisulphite modification. Analysis for hMLH1 promoter methylation was performed using the fol- lowing two methods: (1) Methylation-specific PCR using primers and methods similar to those described by Herman et al. (1998). This examined methylation status of the CpG sites located in region 1. (2) Cases with methylation in region 1 and all cases with hMLH1 protein loss, were also subjected to bisulphite-sequencing procedure to determine the methylation status of 8 CpG sites located in region 2. The sequences of the primers were: F: TGTAAAACGACGGCCAGTTATTTTTGTTTTTATTGGTTGGATA, R: AATACCAATCAAATTTCTCAACTCT. The forward primer was tagged with M13. Bisulphite-modified DNA was amplified and the PCR products were sequenced using the DYEnamic ET Terminator Cycle Sequencing Kits (Amersham Pharmacia) with M13 as sequencing primer. The ratio of peak height of C to T at a CpG site indicates the ratio of methylated to unmethylated cytosine

the most common germline hMSH2 mutation in this that virtually all patients with a family history Hong Kong Chinese population. Two sporadic early- satisfying not only Bethesda Criterion 1 (Amsterdam), onset tumours had somatic hMSH2 mutations only. In but also the more relaxed Criteria 2 and 3, had patient S3, though there was retention of hetero- detectable germline mutation in hMLH1 or hMSH2 zygosity at the DNA level, the wild type mRNA was (15 of 17, 83%). not expressed in the cancer tissue. The genetic In 21 of the 38 tumours showing loss of expression mechanisms underlying this loss of wild type mRNA, of hMLH1 or hMSH2 our methods accounted be they germline or somatic events, remained obscure. satisfactorily for suppression of both . The Overall, either one or two hit inactivation of hMSH2 commonest lesion in familial cases was LOH (6/17, was identified in 94% of cases with absence of hMSH2 35%). In the small group of sporadic, late-onset protein expression. tumours, as expected from the results of others, most One prominent feature of these data is the ease with showed bi-allelic promoter methylation in hMLH1 (5/ which the genetic basis of MSI-H tumours may be 6, 83%). The apparently sporadic early onset tumours, established, in the different patient groups, using however, showed a complex mixture of germline and immunohistochemistry for hMLH1 and hMLH2 as a somatic mutation, LOH, promoter methylation or loss primary screen. All of the tumours arising in patients of wild type transcript for undetermined causes. with a family history (17/17) showed loss of one or In three cases (L5, L20, S3) we showed unequivocal other protein. This is also the case in 16 of 20 sporadic loss of wild type mRNA, but with convincing evidence early onset tumours (80%) and five of seven sporadic for inactivation of only one of the alleles of the late-onset cases (approximately 70%). In contrast, in relevant MMR gene. It is not clear whether the failure tumours other than those with MSI-H, both proteins to account for the loss of wild type mRNA in our cases were invariably expressed. Further, these data show is attributable to promoter mutation (Shin et al., 2002)

Oncogene Basis of MSI in colorectal cancer ST Yuen et al 7591 fact that, of the several CpG rich regions within the promoter, methylation in some may correlate better than others with transcriptional silencing and loss of the protein. Deng et al. (1999) examined different regions of the hMLH1 promoter, and found that methylation of CpG sites in region 2 correlated best with MSI and hMLH1 protein loss in colorectal cancer cell lines. Moreover, this group has shown by functional assay that this proximal region of the hMLH1 promoter is capable of driving transcription, and methylation of this region is more important than the upstream regions in inhibiting transcription (Deng et al., 2001). Subsequently, others have demonstrated methylation in the upstream regions both in normal mucosa and a small number of colorectal cancers without protein loss or MSI phenotype (Deng et al., 2002). Our present study using primary cancer tissue, with exhaustive and systematic analysis of hMLH1, Figure 3 Box-and-whisker plot of the hMLH1 mRNA level as provides additional evidence for the relative signifi- measured by Real-time quantitative RT – PCR in MSI-H colorec- tal cancers with promoter methylation and microsatellite-stable cance of methylation in these two regions. Using colorectal cancers. Me+ MSI CRC, MSI-H colorectal cancers bisulphite sequencing performed on microdissected with hypermethylation involving regions 1 and 2 of hMLH1 pro- cancer tissue, we have proven that biallelic methylation moter and without detectable mutation in the gene. Line inside in region 2 is invariably associated with transcriptional box, the median; upper and lower limits of box, hinges – 75th silencing of the gene, loss of the protein, and absence and 25th percentiles, respectively; vertical bars above and below the box, the upper and lower quartiles, respectively. *The differ- of DNA mutation. Methylation of region 2 in mono- ence in hMLH1 level between the two groups is statistically signif- allelic form may also act as the second hit event. In icant (Mann-Whitney U test, P50.001). 2 – 3 mg of total RNA contrast, the functional significance of methylation in extracted from microdissected tumour tissue was reverse tran- region 1, especially when assessed by MSP in which scribed into first-strand cDNAs. Real-time quantitative RT – PCR was performed using LightCycler – FastStart DNA Master only a few CpG sites are examined and the extent of SYBR Green I (Roche Diagnostics Ltd.) according to the manu- methylated alleles are not known, should be interpreted facturer’s protocol. Primers for the detection of hMLH1 were for- with caution. It can occur in tumours with no hMLH1 ward primer 5’-AAGAAGATCTGGATATTGTATGTG-3’ and protein loss, and in tumours that already harbour two- reverse primer 5’-CCTAACATCAGCTACTGTCTCTCC-3’. The hit inactivation of hMLH1. Our data thus support the human b-actin was used for standardization of the initial RNA content of a sample. The primers for human b-actin were forward functional significance of hMLH1 promoter methyla- primer 5’-ACTCTTCCAGCCTTCCTTCC-3’ and reverse primer tion involving region 2 in primary colorectal cancer 5’-CGTCATACTCCTGCTTGCTG-3’. For hMLH1, 6 pmol of tissue and the use of bisulphite sequencing to assess each primer, 5 mM of MgCl2 and an annealing temperature of methylation in this region. 628C were used. For b-actin, 3 pmol of each primer, 3 mM of The management of patients with MSI-H colorectal MgCl2 and an annealing temperature of 608C were used. The PCR was run at 958C for 10 min for antibody inactivation, fol- cancer can be greatly influenced by information on its lowed by 45 to 55 cycles of amplification at 958C for 15 s, specific genetic basis. This is particularly true when the annealing temperature for 5 s and 728C for 12 s, with a tempera- tumours arise in young subjects, and when family ture slope of 208C per s. Real-time PCR monitoring was achieved history is unclear. We show here that a combination of by measuring the SYBR Green I signal at 828C and 868C after the extension period for hMLH1 and b-actin respectively. Melting germline and somatic mutations, LOH, monoallelic curves were analysed in each run to ensure that the products were hMLH1 promoter hypermethylation and loss of wild homogeneous. To quantitate the amount of specific mRNA in the type mRNA accounts for the loss of hMSH2 or samples, a standard curve was generated for each run using serial hMLH1 proteins in a substantial proportion of these dilution of a reference cDNA generated from a lymphoblastoid cell line of a normal individual. This standard curve was used sporadic early-onset tumours. In contrast, bi-allelic to calculate the concentration of the template copies in the un- promoter methylation, the predominant cause of MSI- known samples. All the experiments were performed in duplicate H in late-onset cancers, is rare in the early-onset group. and the result for individual sample was expressed as the mean However, the data presented here also show that in a level of hMLH1/b-actin relative to the reference lymphoblastoid small proportion of sporadic early-onset tumours cell line where deficiency of mismatch repair protein expression was clearly established, no satisfactory explanation for suppression of the relevant mismatch repair gene was or to other events that adversely affect the transcrip- demonstrated. Moreover, alterations in these two genes tion, processing or stability of mRNA. are rarely seen in sporadic MSI-H tumours when the In the course of this study, we compared the corresponding proteins are normally expressed. It association of methylation in two regions of the remains to be seen whether germline or somatic events hMLH1 promoter with suppression of the protein affecting other MMR genes, in particular hMSH6, product. One of the major problems in the analysis of hPMS1 or hPMS2, are responsible for the mismatch hMLH1 promoter hypermethylation derives from the repair deficiency in these cases.

Oncogene Basis of MSI in colorectal cancer ST Yuen et al 7592 In conclusion, our study has revealed different Acknowledgements pathways of carcinogenesis in MSI-H colorectal We thank Mr ISH Ho and Miss KPS Cheung for excellent tumours, related to age of onset of the disease and technical assistance, and Miss Kedo Kwan for coordina- the presence or absence of family history. The tion work of the Hereditary Gastrointestinal Cancer observations impact on management, since relatives Registry. We also thank Professors Jonathan Sham and TM Wong for their support of this project. This work was of patients with germline mutations require stringent supported by grants from the Research Grants Council of prophylactic screening, whilst those of patients with the Hong Kong Special Administrative Region (HKU two-hit somatic inactivations or bi-allelic promoter 7330/00M, HKU 7006/99M), a donation from the Honk methylation without mutation should be relieved of the Kong Cancer Fund and a donation from the Hong Kong psychological burden of suspected cancer susceptibility. Society of Gastroenterology.

References

Chan TL, Yuen ST, Chung LP, Ho JW, Kwan KY, Chan AS, Konishi M, Kikuchi-Yanoshita R, Tanaka K, Muraoka M, Ho JC, Leung SY and Wyllie AH. (1999a). J. Natl. Cancer Onda A, Okumura Y, Kishi N, Iwama T, Mori T, Koike Inst., 91, 1221 – 1226. M, Ushio K, Chiba M, Nomizu S, Konishi F, Utsunomiya Chan TL, Yuen ST, Chung LP, Ho JW, Kwan K, Fan YW, J and Miyaki M. (1996). Gastroenterology, 111, 307 – 317. Chan AS and Leung SY. (1999b). Genes Kuismanen SA, Holmberg MT, Salovaara R, de la Chapelle Cancer, 25, 75 – 81. A and Peltomaki P. (2000). Am. J. Pathol., 156, 1773 – Chan TL, Yuen ST, Ho JW, Chan AS, Kwan K, Chung LP, 1779. Lam PW, Tse CW and Leung SY. (2001). Oncogene, 20, Leung SY, Chan TL, Chung LP, Chan AS, Fan YW, Hung 2976 – 2981. KN, Kwong WK, Ho JW and Yuen ST. (1998). Am. J. Cunningham JM, Christensen ER, Tester DJ, Kim CY, Pathol., 153, 1181 – 1188. Roche PC, Burgart LJ and Thibodeau SN. (1998). Cancer Leung SY, Yuen ST, Chung LP, Chu KM, Chan AS and Ho Res., 58, 3455 – 3460. JC. (1999). Cancer Res., 59, 159 – 164. Deng G, Chen A, Hong J, Chae HS and Kim YS. (1999). Liu B, Farrington SM, Petersen GM, Hamilton SR, Parsons Cancer Res., 59, 2029 – 2033. R, Papadopoulos N, Fujiwara T, Jen J, Kinzler KW, Deng G, Chen A, Pong E and Kim YS. (2001). Oncogene, 20, Wyllie AH, Vogelstein B and Dunlop MG. (1995). Nat. 7120 – 7127. Med., 1, 348 – 352. Deng G, Peng E, Gum J, Terdiman J, Sleisenger M and Kim Peltomaki P and Vasen HF. (1997). Gastroenterology, 113, YS. (2002). Br. J. Cancer, 86, 574 – 579. 1146 – 1158. Hemminki A, Peltomaki P, Mecklin JP, Jarvinen H, Rodriguez-Bigas MA, Boland CR, Hamilton SR, Henson Salovaara R, Nystrom-Lahti M, de la Chapelle A and DE, Jass JR, Khan PM, Lynch H, Perucho M, Smyrk T, Aaltonen LA. (1994). Nat. Genet., 8, 405 – 410. Sobin L and Srivastava S. (1997). J. Natl. Cancer Inst., 89, Herman JG, Umar A, Polyak K, Graff JR, Ahuja N, Issa JJ, 1758 – 1762. Markowitz S, Willson JV, Hamilton SR, Kinzler KW, Shin KH, Shin JH, Kim JH and Park JG. (2002). Cancer Kane MF, Kolodner RD, Vogelstein B, Kunkel TA and Res., 62, 38 – 42. Baylin SB. (1998). Proc. Natl. Acad. Sci. USA, 95, 6870 – Tannergard P, Liu T, Weger A, Nordenskjold M and 6875. Lindblom A. (1997). Hum. Genet., 101, 51 – 55. Ho JW, Yuen ST, Chung LP, Kwan KY, Chan TL, Leung Wheeler JM, Beck NE, Kim HC, Tomlinson IP, Mortensen SY, Chan AS, Tse C, Lam PW and Luk IS. (2000). Int. J. NJ and Bodmer WF. (1999). Proc. Natl. Acad. Sci. USA, Cancer, 89, 356 – 360. 96, 10296 – 10301. KaneMF,LodaM,GaidaGM,LipmanJ,MishraR, Yuen ST, Chung LP, Leung SY, Luk IS, Chan SY, Ho JC, Goldman H, Jessup J and Kolodner R. (1997). Cancer Ho JW and Wyllie AH. (1997). Br. J. Cancer, 76, 1610 – Res., 57, 808 – 811. 1616.

Oncogene