Methylation associated inactivation of RASSF1A from region ... - Nature

Oncogene (2001) 20, 1509 ± 1518 ã 2001 Nature Publishing Group All rights reserved 0950 ± 9232/01 $15.00 www.nature.com/onc

Methylation associated inactivation of RASSF1A from region 3p21.3 in lung, breast and ovarian tumours Angelo Agathanggelou1, So®a Honorio1, Donia P Macartney1, Alonso Martinez1, Ashraf Dallol1, Janet Rader2, Paul Fullwood1, Anita Chauhan1, Rosemary Walker3, Jacqueline A Shaw3, Shigeto Hosoe4, Michael I Lerman5, John D Minna6, Eamonn R Maher1 and Farida Latif*,1 1

Section of Medical and Molecular Genetics, Department of Reproductive and Child Health, University of Birmingham, The Medical School, Edgbaston, Birmingham, B15 2TT, UK; 2Department of Obstetrics and Gynecology, Washington University School of Medicine, St Louis, Missouri, USA; 3Breast Cancer Research Unit, University of Leicester, Clinical Sciences, Glen®eld Hospital, Leicester, UK; 4Department of Internal Medicine, National Kinki-Chuo Hospital, Osaka, Japan; 5Laboratory of Immunobiology, NCI-Frederick Cancer Research and Development Center, Frederick, Maryland, MD, USA; 6Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center at Dallas, Dallas, Texas, TX, USA

Previously we analysed overlapping homozygous deletions in lung and breast tumours/tumour lines and de®ned a small region of 120 kb (part of LCTSGR1) at 3p21.3 that contained putative lung and breast cancer tumour suppressor gene(s) (TSG). Eight genes including RASSF1 were isolated from the minimal region. However, extensive mutation analysis in lung tumours and tumour lines revealed only rare inactivating mutations. Recently, de novo methylation at a CpG island associated with isoform A of RASSF1 (RASSF1A) was reported in lung tumours and tumour lines. To investigate RASSF1A as a candidate TSG for various cancers, we investigated: (a) RASSF1A methylation status in a large series of primary tumour and tumour lines; (b) chromosome 3p allele loss in lung tumours and (c) RASSF1 mutation analysis in breast tumours. RASSF1A promoter region CpG island methylation was detected in 72% of SCLC, 34% of NSCLC, 9% of breast, 10% of ovarian and 0% of primary cervical tumours and in 72% SCLC, 36% NSCLC, 80% of breast and 40% of ovarian tumour lines. In view of the lower frequency of RASSF1 methylation in primary breast cancers we proceeded to RASSF1 mutation analysis in 40 breast cancers. No mutations were detected, but six single nucleotide polymorphisms were identi®ed. Twenty of 26 SCLC tumours with 3p21.3 allelic loss had RASSF1A methylation, while only six out of 22 NSCLC with 3p21.3 allele loss had RASSF1A methylation (P=0.0012), one out of ®ve ovarian and none out of six cervical tumours with 3p21.3 loss had RASSF1A methylation. These results suggest that (a) RASSF1A inactivation by two hits (methylation and loss) is a critical step in SCLC tumourigenesis and (b) RASSF1A inactivation is of lesser importance in NSCLC, breast, ovarian and cervical cancers in which

*Correspondence: F Latif Received 9 November 2000; revised 1 December 2000; accepted 15 December 2000

other genes within LCTSGR1 are likely to be implicated. Oncogene (2001) 20, 1509 ± 1518. Keywords: 3p; tumour suppressor gene; methylation; lung cancer; breast cancer Introduction Deletion of markers on the short arm of human chromosome 3 is one of the most frequent and earliest genetic changes found in lung cancer (Virmani et al., 1998; Wistuba et al., 1999, 2000). Several distinct 3p regions, 3p25 ± 26, 3p24, 3p21.3, 3p14 and 3p12, have been identi®ed as showing frequent allelic losses in lung and other sporadic cancers (Kok et al., 1997). This suggests that multiple tumour suppressor genes (TSGs) may reside on 3p and that their inactivation plays a role in the pathogenesis of a number of common sporadic cancers. We have concentrated our eorts for identi®cation of lung cancer TSG(s) to two lung cancer tumour suppressor gene regions (LCTSGRs) at 3p21.3 (LCTSGR1) and 3p12 (LCTSGR2). These regions have been identi®ed by overlapping homozygous deletions in lung and breast tumour and tumour lines (Rabbitts et al., 1990; Latif et al., 1992; Wei et al., 1996; Sekido et al., 1998; Sundaresan et al., 1998; Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000). The nested deletions at 3p21.3 in three lung and one breast tumour line (Wei et al., 1996; Sekido et al., 1998; Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000) narrowed the region to 120 kb interval within LCTSGR1. A 600 kb cosmid contig containing the overlap and surrounding regions was constructed which has been sequenced jointly by The Washington University and The Sanger Centre. Using a combination of molecular biology and in silico cloning techniques we isolated eight genes from within the 120 kb minimal region,

Methylation associated inactivation of RASSF1A A Agathanggelou et al

1510

namely: CACNA2D2- PL6- 101F6- NPRL2- BLURASSF1- FUS1- HYAL2 (Figure 1). Extensive mutation analysis in lung tumours and tumour lines revealed only rare inactivating mutations in these genes, although expression of isoform A of RASSF1 was lost or down regulated in lung tumour lines (Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000). A very recent article reported epigenetic inactivation of RASSF1A in lung cancers (Dammann et al., 2000). RASSF1 has a predicted RAS association domain and was recently identi®ed as a putative phosphorylation target for ATM in in vitro studies (Kim et al., 1999). RASSF1 has several major isoforms due to alternative splicing and promoter usage (Figure 2a,b). RASSF1C is the shorter isoform (32 kDa predicted peptide), and is well expressed in lung tumour lines, whilst RASSF1A the longer isoform (39 kDa predicted peptide) expression is lost or down regulated in many lung tumour lines (Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000). In addition to containing the predicted RAS association domain and a putative substrate for ATM phosphorylation, the long isoform, RASSF1A also contains a predicted diacylglycerol (DAG) binding domain also found in a related gene NORE1 (Figure 2b). RASSF1A has also been shown to dramatically reduce tumorigenicity in vivo (Dammann et al., 2000; Burbee et al., 2001) We analysed RASSF1A promoter region CpG island methylation status in a large series of primary lung,

breast, ovarian and cervical tumours and in lung, breast and ovarian tumour lines, we found that this CpG island was hypermethylated in majority of lung tumours and lung, breast and ovarian tumour lines and to a lesser extent in breast, and ovarian tumours, but not in any of 22 cervical tumours. We also determined 3p allelic losses in the above series of primary lung tumours to determine their extent and frequency and relationship to RASSF1A methylation. We focused on speci®c 3p regions, (3p25, 3p21-22, 3p14.2, 3p12) implicated in tumorigenesis in lung, breast and other cancers. We found 3p loss in a majority of the lung tumours and the two most frequently lost regions on 3p included LCTSGR1 and LCTSGR2. In addition we found SROs within these two regions, hence further evidence of their importance in tumour development.

Results Methylation analysis of a CpG island in promoter region of RASSF1A We directly sequenced 36 lung tumour cell lines DNA from nucleotide 7110 bp to +41 bp after sodium bisul®te modi®cation using primers described previously (Dammann et al., 2000, and see Figure 4a). Eighteen of 25 (72%) SCLC tumour lines and four of 11 (36%) NSCLC (P=0.0665) were methylated for majority of the 16 CpGs contained within this region, of which six SCLC were partially methylated (Figure 3). Seven SCLC and seven NSCLC tumour lines were

Figure 1 Schematic representation of the overlapping homozygous deletions found in the 3p21.3 chromosomal region. The 120 kb minimal deleted segment is included in cosmids LUCA6 to 13; the nine genes that lie in or border this sub-region are shown below Oncogene


1511

Figure 2 (a) The RASSF1 gene structure. The RASSF1 gene has ®ve exons, two promoter-associated CpG islands and encodes several transcripts by alternative promoter usage and RNA splicing mechanisms. All isoforms share exons 2 ± 5 but dier in their 5' regions. RASSF1A (Accession # AF102770) has exons 1A and 1C, RASSF1B (Accession #AF132677) uses exons 1B and 1C and RASSF1C (Accession # AF040703) transcription starts in exon 1. (b) Peptide sequence alignment for RASSF1A (Accession # AAF35127) and RASSF1C (Accession # AAC70910), showing some of the predicted protein domains: the ATM phosphorylation site (in blue) and the Ras association domain (in green), common to both isoforms, and the DAG_PE/C1 domain (in yellow), that is only found in RASSF1A. The rare polymorphisms described in this study are highlighted in red

not methylated for any CpG. We also sequenced corresponding normal DNA from two SCLC lines showing methylation of majority CpGs and found no methylation in the normal samples. Four out of the ®ve breast tumour lines sequenced were found to be completely methylated and the remaining was unmethylated and two out of ®ve ovarian tumour lines were also methylated. From Figure 3 it can be determined that in our set of lung tumour lines nearly all CpG's are methylated whenever methylation occurs. At CpGs 1/2 and 10/11 methylation protected the restriction site for BstUI (GCGC) from modi®cation by the bisul®te reagent. Similarly, at CpGs 6 and 16 bisul®te modi®cation of preceding cytosine residues generates TaqI restriction sites (TCGA) when these CpGs are methylated. Hence, we devised two restriction enzyme digestion protocols in which the bisul®te modi®ed DNA ampli®ed by PCR using the above primers was digested with both BstUI and TaqI separately (Figure 4). In this way 70 primary lung tumours and corresponding normal DNA, 44 primary breast carcinomas and corresponding normal breast tissue, 21 ovarian tumour and 22 cervical tumours were analysed for RASSF1A promoter region methylation at a total of 6/16 CpGs (Figure 4). We found 50% (35/70) lung tumours, 9% (4/44) breast, 10% (2/21) ovarian and 0% (0/22) cervical tumours were methylated either at all four restriction sites or at one or the other of the pairs of restriction sites (P50.0001 lung vs breast; P=0.0008 lung vs ovarian).

Methylation of the promoter region of RASSF1A was higher in SCLC tumours (72%, 21/29) compared to NSCLC tumours (34%, 14/41) (P50.0033). The majority of tumour samples contained the unmethylated allele (204 bp, Figure 4b), this is not surprising since none of the tumours had been micro-dissected and hence would be contaminated with normal cells. Complete digestion of the PCR products with TaqI and BstUI should have produced fragments of 92, 81, 31 and 89, 83 32 bp respectively. However, in most of the cases which were methylated the full spectrum of restriction enzyme digestion products were observed suggesting that there is a heterogeneity in the methylation pattern within the tumour cell population. None of the matching normal lung or breast tissue demonstrated methylation, and none of the matching DNA from blood for ovarian and cervical tumours showed any methylation. The results of all methylation experiments are in Table 1. LOH analysis of primary lung tumours To determine the frequency and extent of 3p deletions in primary lung tumours and to clarify the role of 3p allele loss at LCTSGR1 and LCTSGR2 we undertook LOH analysis using 15 microsatellite markers targeted at regions 3p12 (LCTSGR2), 3p14.2, 3p21.3 (LCTSGR1) and 3p25 ± 26 (Figure 5). We analysed 84 primary lung tumours of which 36 were SCLC, and 48 NSCLC including 11 squamous, 21 adenocarcinoOncogene


1512

Figure 3 This bar chart summarizes the sequencing data from the SCLC and NSCLC cell lines that show methylation (n=22). The black ®lled bars show the proportion of cell lines that are completely methylated at all 16 CpGs, the grey ®lled bars show partially methylated CpGs and the un®lled bars show the proportion of unmethylated CpGs

b

a

Figure 4 (a) Schematic representation of the reverse sequence of the sodium bisul®te-treated RASSF1A promoter region CpG island fragment ampli®ed with the primers described in Material and methods. The 16 CpG's pairs (in this sequence they are all shown as methylated) and the restriction sites for TaqI and BstUI are shown. (b) Representative results of bisul®te-PCR methylation analysis in SCLC and NSCLC tumours of the 5' region of RASSF1A. Tumour and normal DNA was treated with bisul®te, ampli®ed using speci®c primers described in Dammann et al. (2000) and digested using restriction enzymes that recognize methylated alleles only, bands separated by 3% agarose gel stained with ethidium bromide. Methylated alleles cleaved by restriction enzyme (TaqI, top panel and BstUI bottom panel). The sizes of the digested products are against the brackets. The 204 bp undigested product is seen in each lane Oncogene


mas, two large cell, one adeno-squamous and 13 of unknown histopathology. 3p LOH was very frequent with 86% (31/36) of SCLC, 91% (10/11) of squamous cell cancers, and 71% (15/21) of adenocarcinomas showing loss at one or more loci. LOH was observed for all informative markers in 53% (19/36) of SCLC, 36% (4/11) of squamous cell cancers, and 33% (7/21) of adenocarcinomas, consistent with the complete loss

Table 1 Showing results of all tumour/tumour lines analysed for RASSF1A methylation status Sample SCLC cell lines NSCLC cell lines Breast tumour lines Ovarian tumour lines Primary SCLC Primary NSCLC Primary breast tumours Primary ovarian tumours Primary cervical tumours

Methylated Unmethylated % Methylated 17 4 4 2 21 14 4 2 0

7 11 1 3 8 27 40 19 22

68 36 80 40 72 34 9 10 0

of chromosome 3p. A subset of tumours with partial losses (n=31, Figure 5) allowed us to identify three candidate minimal regions of non-overlapping deletion (Figure 6 shows typical LOH results for SRO1 and breakpoints for SRO2 and SRO3). Region 1, between the markers D3S4597 and D3S1621 within LCTSGR1, a distance 5600 kb that includes RASSF1. Region 2 within LCTSGR2 between markers D3S1274 and D3S3049 (tumour 83 showed loss at D3S1274 and retention at D3S3049 while tumour 53 showed retention at D3S1274 and loss at D3S3049) a genetic distance of 51 cM. Region 3, also within LCTSGR2 but more centromeric, between markers D3S1577 and D3S1254 (see tumours 3, 13 and 67) a genetic distance of 51.0 cM. We also observed that LCTSGR1 was lost more frequently in SCLC and squamous cell carcinomas compared to adenocarcinomas (80% [28/35], 82% [9/ 11], and 60% [12/20] respectively); but this dierence was not statistically signi®cant. When considering all tumours with any loss at 3p, LCTSGR1 was always seen to be lost in combination with other loci. Markers

1513

Figure 5 Idiogram of human chromosome 3p indicating regions of interest and the location and order of microsatellite markers and the 31 lung tumours with partial losses. LOH was carried out using standard PCR conditions and primers as described in The GDB and in Fullwood et al. (1999). Alleles were visualized either by autoradiography or silver staining and LOH scored based on absences in tumour derived DNA or a dierence in the relative intensities of alleles in the tumour derived DNA upon comparison to the normal tissue derived DNA. Markers D3S4597, 4604, 1568 (equivalent to D3S4615), 1621 (equivalent to D3S4623) represent LCTSGR1; markers D3S1274, 3049, 1604, 1577, 1511 and 1254 represent LCTSGR2. Black ovals represent LOH; White ovals represent retention. I represents microsatellite instability and no symbols represent uninformative loci Oncogene


1514

Figure 6 (a) Delineation of SRO1 by LOH in tumour pairs 90 and 11. (b) Demonstration of LOH breakpoints in LCTSGR2 for tumour pairs 83 and 3. RET : retention of both alleles in tumour DNA, arrows indicate alleles lost in tumour DNA

from LCTSGR2 were again lost in a high proportion of SCLC (83%, 30/36) but to a lesser extent in squamous cell carcinoma (54%, 6/11) and adenocarcinoma (55%, 11/20) (SCLC vs adenocarcinomas, P=0.03). Furthermore, four tumours showed 3p LOH for 3p12 markers alone (two adenocarcinomas, one large cell carcinoma and one SCLC). Clinical-pathological parameters and 3p LOH in lung tumours We analysed the relationship between tumour stage, and frequency and extent of 3p LOH. The majority of SCLC were stage IV (n=33) and showed loss of most 3p markers. There were four stage IV tumours (4/33) that did not show loss for any informative 3p marker. It is not clear if this was due to contamination from normal DNA in the tumour sample, which can lead to diculty in the determination of loss, or if these tumours underwent allelic losses at chromosome(s) other then 3p. However, the two stage I and one stage II tumour showed extensive 3p loss. Stage I tumours 18 and 45 gave LOH for 62 and 89% of informative loci respectively, whilst stage II tumour 51 underwent loss Oncogene

at all informative loci. All but one (10/11, 91%) squamous cell carcinoma had loss of one or more 3p markers. Stage III and IV tumours (n=5) had more extensive 3p loss compared to stage I and II tumours (n=5). All stage III and IV tumours had loss of over 50% of informative loci, whilst only one tumour amongst stage I and II had loss of over 50% of the informative loci. There was also a signi®cant dierence between the level of loss seen in stage III and IV adenocarcinomas compared to stage I and II tumours. Eight out of 14 stage III and IV adenocarcinomas had loss of over 50% of informative loci, whilst only one out of seven stage I and II adenocarcinomas underwent loss of over 50% of informative loci. 3p allele loss and RASSFIA promoter region methylation in lung, ovarian and cervical tumours We investigated correlations between 3p allelic loss with RASSF1A promoter region methylation. Twenty of the 26 SCLC tumours with 3p21.3 (and other 3p markers) allelic loss demonstrated RASSFIA methylation compared to six of 22 NSCLC (P=0.0012). Of 27 NSCLC (27/41) with no RASSFIA methylation, 16 had


1515

allele loss at 3p21.3 (and other 3p markers), three tumours had loss at only 3p12 markers (LCTSGR2) and retention at 3p21.3, indicating inactivation of gene(s) at 3p12 in a subset of NSCLCs, six tumours had no loss at any 3p loci, for the remaining two tumours we did not have LOH data. We had analysed the same series of ovarian tumours for LOH at 3p in an earlier study (Fullwood et al., 1999), and for this study we also analysed cervical matched normal/tumour pairs for markers on 3p21.3 (D3S1621 and D3S4597) (data not shown). Only one of the ®ve ovarian tumours and zero of six cervical tumours with 3p21.3 loss had RASSF1A methylation (P=0.0274 SCLC vs ovarian; P=0.001 SCLC vs cervical). We did not ®nd any correlation between RASSF1A promoter region hypermethylation and stage or histology of our set of NSCLC tumours. Of the two ovarian tumours showing RASSF1A methylation one was serous (stage IA) and the other endometrioid (stage IIIB), whilst three grade II and one grade I breast ductal carcinomas underwent RASSF1A promoter region hypermethylation. Homozygous deletions in lung tumours We also analysed our series of primary lung tumours to search for homozygous deletions, Gene Scan analysis was used to determine whether markers showing retention ¯anked by closely related markers showing loss were genuinely retained (discontinuous LOH) or homozygously deleted (apparent retention due to contaminating normal DNA in the tumour sample) (Cairns et al., 1995; Reed et al., 1996). This approach enabled us to con®rm retention where tumour samples were relatively clean (520% normal DNA contamination); for instance, tumour 3 genuinely retained marker D3S1577 and lost marker D3S1604 (Figure 7), tumour 14 genuinely retained D3S1568, tumour 42 genuinely retained D3S4597 etc. In this way we were able to rule out the occurrence of homozygous deletions in most of the lung tumours. Mutation analysis of RASSF1 In view of the low frequency of RASSF1A methylation in breast cancer we performed mutation analysis of RASSFIA, RASSF1C and RASSF1B in 36 breast carcinomas and four breast tumour lines by SSCP analysis. No inactivating mutations were detected, but we did ®nd six single nucleotide polymorphisms, four of which resulted in amino acid substitutions. One amino acid substitution was in the candidate ATM target sequence, two in the predicted RAS association domain, two in the promoter region of isoform A and one sequence variant speci®c for isoform B (Table 2). Discussion We analysed a large series of primary lung tumours for 3p allele loss at speci®c loci and determined the

Figure 7 Con®rmation of retention of D3S1577 in tumour 3 by Gene Scan analysis. Gene Scan analysis was carried out using multiplex reactions containing 40 ng DNA, and standard cycling for 23 cycles. Forward primers were labelled with chromophores Hex, Fam or Ned and run on an ABI 377. Data was analysed using Gene Scan 3.1 software

methylation status of the 5'CpG island (in lung, breast, ovarian and cervical tumours) in the promoter region of isoform A of RASSF1 contained within a 120 kb minimal region at 3p21.3 de®ned by overlapping homozygous deletion in lung and breast tumour lines. We found: (a) RASSF1A is epigenetically inactivated in 72% of SCLC and 34% NSCLC primary tumours, in 9% breast tumours, 10% of ovarian and 0% of cervical tumours and in 72, 36, 80 and 40% of SCLC, NSCLC, breast and ovarian tumour lines respectively; (b) high allelic loss at 3p21.3 and 3p12 loci in primary lung tumours; (c) minimal regions of non overlapping deletions in lung tumours within regions LCTSGR1 and LCTSGR2; (d) two hits (methylation and 3p loss) are consistent with the role of RASSF1A as a classical TSG (Knudson's model). We have concentrated our eorts on identi®cation of 3p TSGs on two speci®c regions on 3p, LCTSGR1 at 3p21.3 and LCTSGR2 at 3p12. Both regions have been narrowed by ®nding of overlapping homozygous deletions in lung and breast tumour and tumour lines (Rabbitts et al., 1990; Latif et al., 1992; Wei et al., Oncogene


1516

Table 2 Exon 1A

Polymorphisms in the RASSF1 gene found in breast cancer Polymorphisma

Codon altered

Domain affected

c.61A?C

K21Qb (Lys21Gln) R28R (Arg28Arg)

RASSF1A promoter region

c.84T?A 1B

c.-237insG

±

RASSF1B 5' untranslated region (5'UTR)

2

c.433G?T

A133S (Ala133Ser)

ATM phosphorylation site

3

c.772G?A

4

c.806G?A

E246K (Glu246Lys) R257Gb (Arg257Gln)

Ras association domain

a

All the polymorphisms described in this table are numbered using RASSF1A cDNA sequence as a reference, except the one found in exon 1B, that is speci®c to RASSF1 isoform B. bThese polymorphisms are also described by Dammann et al. (2000)

1996; Sekido et al., 1998; Sundaresan et al., 1998; Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000). We isolated eight genes from the 120 kb minimal overlap from LCTSGR1 and so far one gene has been isolated from LCTSGR2. But thus far none of the genes show frequent inactivating mutations in lung or breast tumours. But RASSF1A residing in the 120 kb minimal region at 3p21.3 showed loss of expression in lung tumour lines (Dammann et al., 2000; Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000). We identi®ed RASSF1C by screening gridded cDNA libraries using LUCA12 cosmid probe, whilst the long form (RASSF1A) was identi®ed by using RT ± PCR (The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000). RASSF1A and RASSF1C were found to be well expressed in many tissues analysed including lung, heart, brain, liver, kidney, pancreas, placenta etc. Whilst RASSF1C was well expressed in lung tumour lines, expression of RASSF1A was absent in many lung tumour lines (Dammann et al., 2000; Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000; Burbee et al., 2001). RASSF1 contains a RAS association domain (RAD) by SMART and PFAM analysis, and was recently shown to be an eector of RAS both in vitro and in vivo studies (Vos et al., 2000). Kim et al. (1999) recently reported several putative in vitro phosphorylation targets for ATM, including RASSF1 (PTS in Kim et al., 1999), in addition RASSF1A contains a predicted diacylglycerol (DAG) binding domain also found in a related mouse gene NORE1, interestingly NORE1 was recently identi®ed as a potential new Ras eector in a yeast two-hybrid screen (Vavvas et al., 1998). Oncogene

In this study we have shown RASSF1A to be methylated in majority of lung cancers and to a lesser extent in breast, and ovarian tumours. We have also demonstrated that RASSF1A acts as a classical TSG, requiring inactivation of both alleles. Under Knudson's two hit model (Knudson, 1971, 1996) the ®rst `hit' is often a point mutation, small deletion or an epigenetic event followed by a second inactivating mutation which is often a large chromosomal loss. In the case of RASSFIA it is likely that the ®rst hit is epigenetic silencing followed by 3p loss. For SCLC there was strong evidence that RASSFIA is the critical 3p21.3 TSG thus 20/26, (77%) SCLCs with 3p21.3 loss demonstrated RASSF1A hypermethylation but only six out of 22 (27%) NSCLC, one out of ®ve (20%) ovarian and none out of six cervical tumours with 3p21.3 loss had RASSF1A promoter region methylation. This dierence between SCLC and NSCLC, ovarian and cervical tumours may indicate diering mechanisms of tumourigenesis in the four tumour types. RASSF1A mutations are rare in lung cancers (Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000) so although there is strong evidence for RASSF1A inactivation being a critical step in most SCLC, RASSF1A inactivation makes a lesser contribution to NSCLC. Moreover, the high frequency of NSCLC with 3p21.3 loss without RASSF1A inactivation suggests that another, as yet unidenti®ed, 3p21.3 TSG gene is important in NSCLC (two other candidate TSGs from 3p21.3 region show loss/down regulation of expression in NSCLC cell lines, Lerman and Minna for The International Lung Cancer Chromosome 3p21.3 Tumour Suppressor Gene Consortium, 2000). In addition the high frequency of LCTSGR2 allele loss in both SCLC and NSCLC irrespective of RASSF1 methylation status suggests that inactivation of TSGs in both regions is advantageous to lung cancer cells. We note that three NSCLC tumours displayed 3p12 loss with retention of markers at 3p21.3 and no RASSF1A methylation. Identi®cation of the 3p21.3 TSG(s) implicated in NSCLC without RASSF1A inactivation will reveal whether such tumours have developed via a LCTSGR1-independent pathway. LOH analysis de®ned a small critical region in LCTSGR1 which contained RASSF1 and the other two genes whose expression is either lost or down regulated in NSCLC. LOH analysis also de®ned two small critical regions within LCTSGR2 suggesting there are two TSGs at 3p12 or one large gene. The low frequency of RASSF1A inactivation in breast, ovarian and cervical cancers despite a relatively high incidence of 3p21.3 allele loss also suggests the presence of other breast, ovarian and cervical cancer TSGs in LCTSGR1. The higher percentage of RASSF1A methylation in breast and ovarian tumour lines compared to primary tumours may represent dynamic changes of methylation during cell culture or indicate that tumours with RASSF1A methylation are more oncogenic and are likely to be established as cell lines.


The identi®cation of polymorphic RASSF1 variants with amino acid substitutions in proposed functional domains will enable the role of these variants as candidate low pentrance susceptibility alleles to be investigated. VHL TSG at 3p25 is inactivated in patients with the familial cancer syndrome, von Hippel Lindau disease and in a high percentage of clear cell renal cell carcinomas (Latif et al., 1993; Crossey et al., 1994; Gnarra et al., 1994; Cliord et al., 1998; Kaelin and Maher, 1998), but inactivation of VHL gene is rare in other tumours showing 3p loss (Gnarra et al., 1994; Sekido et al., 1994; Foster et al., 1995; Patridge et al., 1999). The hunt for other major TSGs on 3p has intensi®ed in recent years and resulted in identi®cation of RASSF1A. Our results suggest that RASSF1A is not the only 3p21.3 TSG and further studies of RASSF1A methylation status in other tumours that show frequent 3p allele loss such as kidney, head and neck etc, and in early lesions are required to de®ne the role of RASSF1A and other candidate 3p21.3 TSGs in a range of common cancers.

Materials and methods Patients and samples A total of 36 SCLC and 48 NSCLC (11 squamous, 21 adenocarcinomas, two large cell and one adeno-squamous and 13 of unknown histopathology) tumour and corresponding normal tissue and 25 SCLC and 11 NSCLC tumour lines DNA was provided by Dr Shigeto Hosoe. The primary tumours and corresponding normal tissue were either obtained at autopsy or by operation (Hosoe et al., 1994). All cases were categorized according to the criteria of the Japan Lung Cancer Society. Forty-four invasive breast carcinomas (in®ltrating ductal carcinomas) (eight grade 1, 13 grade 2 and 23 grade 3) and corresponding normal tissue were provided by Dr Rosemary Walker, they were either detected by mammographic screening or had presented symptomatically. Breast carcinomas were excised at Glen®eld Hospital NHS Trust between July 1995 and July 1997. None of the tumours were from women with a known family history of breast or other cancers. DNA extraction and microdisection from paran embedded sections has been described previously (Shaw et al., 1996). The breast carcinomas were graded using the modi®ed Bloom and Richardson system (Elston and Ellis, 1991), all tissue histological assessment was performed by RAW. Epithelial ovarian normal/normal pairs were described in an earlier study (Fullwood et al., 1999), whilst DNA from cervical normal/normal pairs was provided by Dr Janet Rader and consisted of 15 squamous and seven adenocarcinomas. Bisulfite modification, direct sequencing and restriction enzyme digestion Bisul®te DNA sequencing was done as described (Herman et al., 1996). Brie¯y, 0.5 ± 1.0 mg of genomic DNA was denatured in 0.3 M NaOH for 15 min at 378C and then unmethylated cytosine residues were sulphonated by incuba-

tion in 3.12 M sodium bisul®te (pH 5.0) (Sigma)/5 mM hydroquinone (Sigma) in a thermocycler (Hybaid) for 30 s at 998C/15 min at 508C for 20 cycles. The bisul®te modi®ed DNA was recovered using the Wizard DNA cleanup system (Promega) in accordance with the manufacturers instructions. The conversion reaction was completed by desulphonating in 0.3 M NaOH for 10 min at room temperature. The DNA was ethanol precipitated and resuspended in water. DNA sequences speci®c for the RASSF1A promoter region were ampli®ed using primers and conditions already described (Dammann et al., 2000). Methylated cytosine residues were identi®ed either by direct sequencing of PCR products using a dRhodamine sequence cycling kit (Perkin Elmer) or by restriction enzyme digestion. Brie¯y, 16 ml of the 204 bp PCR product was incubated with 20 units of TaqI (Roche) or BstUI (New England BioLabs) in separate reactions for 2 h at 65 or 608C, respectively. To ensure that restriction enzyme was present in all incubations a master mix containing TaqT or BstUI was prepared and aliquoted into sample and control PCR products. The size of the TaqT restriction enzyme digestion products possible are 173, 112, 92, 81 and 31 bp; and the size of the BstUI digestion products are 172, 121, 89, 83 and 32 bp, see Figure 3 for restriction map. The restriction enzyme digestion products were then visualized by separation in a 3% agarose gel or an 8% polyacrylamide gel stained with ethidium bromide.

1517

Mutation analysis Intron-exon boundaries were determined by matching the cDNA sequence for each isoform with the GenBank deposited genomic sequences of cosmids LUCA12 (Accession number AC002481) and LUCA13 (Accession number AC002455). Mutation screening was performed on 36 primary breast carcinomas and three breast tumour lines by the PCR ± SSCP (Single Strand Conformation Polymorphism) method using intronic primers derived as described above. Aberrantly migrating bands were sequenced on an ABI 377 automated sequencer. Microsatellite repeat analysis PCR ampli®cation of di, tri and tetranucleotide microsatellite sequences were utilized in this study. Up to 15 markers were selected spanning the regions of interest on 3p. All are available through GDB with the exception of new primers for the D3S1621 locus (Forward Primer: 5'CCTCACTACTCCTGGAATTG-3' Reverse Primer 5'CCAAGGAAGGGTTTTACTTA-3', PCR product size 140 bp, annealing temperature 558C). The microsatellite analysis has been described previously by us (Fullwood et al., 1999). Gene scan analysis Gene Scan analysis was carried out using multiplex reactions containing 40 ng DNA, and standard cycling for 23 cycles. Forward primers were labelled with chromophores Hex, Fam or Ned and run on an ABI 377. Data was analysed using Gene Scan 3.1 software. Statistical analysis Comparisons were made by Fisher's exact test and Chisquare test as appropriate. P values of 50.05 were taken as statistically signi®cant. Oncogene


1518

Acknowledgments This work is supported in part by Association for International Cancer Research, Cancer Research Campaign, Breast Cancer Campaign, Wellbeing and The Royal Society. S Honorio is supported by Fundacao para a Cientia e a Technologia. A Martinez is supported by the

University of Antioquia, Medellin, Colombia, and a fellowship from COLCIENCIAS. J Minna was supported by NCI grants Lung Cancer SPORE P50 CA70907 and CA71618. J Rader was supported by grant ACS #RPG-96088-03-CCE

References Burbee D, Forgacs E, ZoÈchbauer-MuÈller S, Shivakuma L, Fong K, Gao B, Randle D, Virmani A, Bader S, Sekido Y, Latif F, Milchgrub S, Gazdar AF, Lerman MI, Zabarovsky E, White M and Minna JD. (2001). J. Natl. Cancer Inst., In press. Cairns P, Polascik TJ, Eby Y, Tokino K, Califano J, Merlo A, Mao L, Herath J, Jenkins R, Westra W, Rutter JL, Buckler A, Gabrielson E, Tockman M, Cho ER, Hedrick L, Bova GS, Isaac W, Koch W, Scwab D and Sidransky D. (1995). Nature Genetics, 11, 210 ± 212. Cliord SC, Prowse AH, Aara NA, Buys CH and Maher ER. (1998). Genes Chromo. Cancer, 22, 200 ± 209. Crossey PA, Richards FM, Foster K, Green JS, Prowse A, Latif F, Lerman MI, Zbar B, Aara NA, Ferguson-Smith MA and Maher ER. (1994). Hum. Mol. Genet., 3, 1303 ± 1308. Dammann R, Li C, Yoon JH, Chin PL, Bates S and Pfeifer GP. (2000). Nat Genet., 25, 315 ± 319. Elston CW and Ellis IO. (1991). Histopathology, 19, 403 ± 410. Foster K, Osborne RJ, Huddart RA, Aara NA, FergusonSmith MA and Maher ER. (1995). Eur. J. Cancer, 31A, 2392 ± 2395. Fullwood P, Marchini S, Rader JS, Martinez A, Macartney D, Broggini M, Morelli C, Barbanti-Brodano G, Maher ER and Latif F. (1999). Cancer Res., 59, 4662 ± 4667. Gnarra JR, Tory K, Weng Y, Schmidt L, Wei MH, Li H, Latif F, Liu S, Chen F, Duh FM, Lubensky I, Duan D, Florence C, Pazzatti R, Walther MM, Bander NH, Grossman HB, Brauch H, Pomer S, Brooks JD, Isaac WB, Lerman MI, Zbar B and Linehan WM. (1994). Nat Genet., 7, 85 ± 90. Herman JG, Gra JR, Myohanen S, Nelkin BD and Baylin SB. (1996). Proc. Natl Acad. Sci. USA, 93, 9821 ± 9826. Hosoe S, Shigedo Y, Ueno K, Tachibana I, Osaki T, Tanio Y, Kawase I, Yamakawa K, Nakamura Y and Kishimoto T. (1994). Lung Cancer, 10, 297 ± 305. Kaelin Jr WG and Maher ER. (1998). Trends Genet., 14, 423 ± 426. Kim ST, Lim DS, Canman CE and Kastan MB. (1999). Biol Chem., 274, 37538 ± 37543. Knudson Jr AG. (1971). Proc. Natl Acad. Sci. USA, 68, 820 ± 823. Knudson AG. (1996). J. Cancer Res. Clin. Oncol., 122, 135 ± 140. Kok K, Naylor SL and Buys CH. (1997). Adv. Cancer Res., 71, 27 ± 92. Latif F, Tory K, Gnarra J, Yao M, Duh FM, Orcutt ML, Stackhouse T, Kuzmin I, Modi W, Geil L, Schmidt L, Zhou F, Li H, Wei M, Chen F, Glenn G, Choyke P, Walther MM, Weng Y, Duan D-SR, Dean M, Glavac D, Richards FM, Crossey PA, Ferguson-Smith MA, Le Paslier D, Chumakov I, Cohen D, Chinault AC, Maher ER, Linehan WM, Zbar B and Lerman MI (1993). Science, 260, 1317 ± 1320.

Oncogene

Latif F, Tory K, Modi WS, Graziano SL, Gamble G, Douglas J, Heppell-Parton AC, Rabbitts PH., Zbar B and Lerman MI. (1992). Genes Chrom. Cancer, 5, 119 ± 127. Lerman MI, Minna JD for the International Lung Cancer Chromosome 3p21.3 Tumor Suppressor Gene Consortium. (2000). Cancer Res., 60, 6116 ± 6133. Partridge M, Emilion G, Pateromichelakis S, Phillips E and Langdon J. (1999). Int. J. Cancer., 83, 318 ± 325. Rabbitts P, Bergh J, Douglas J, Collins F and Waters J. (1990). Genes Chrom. Cancer, 2, 231 ± 238. Reed AL, Califano J, Cairns P, Westra WH, Jones RM, Koch W, Ahrendt S, Eby Y, Sewell D, Nawroz H, Bartek J and Sidransky D. (1996). Cancer Res., 56, 3630 ± 3633. Sekido Y, Ahmadian M, Wistuba II, Latif F, Bader S, Wei MH, Duh FM, Gazdar AF, Lerman MI and Minna JD. (1998). Oncogene, 16, 3151 ± 3157. Sekido Y, Bader S, Latif F, Gnarra JR, Gazdar AF, Linehan WM, Zbar B, Lerman MI and Minna JD. (1994). Oncogene, 9, 1599 ± 1604. Shaw JA, Walsh T, Chappell SA, Carey N, Johnson K and Walker RA. (1996). Br. J. Cancer, 73, 1393 ± 1397. Sundaresan V, Chung G, Heppell-Parton A, Xiong J, Grundy C, Roberts I, James L, Cahn A, Bench A, Douglas J, Minna MI, Sekido Y, Lerman MI, Latif F, Bergh J, Li H, Lowe N, Ogilvie D and Rabbitts P. (1998). Oncogene, 17, 1723 ± 1729. Vavvas D, Li X, Avruch J and Zhang XF. (1998). J. Biol. Chem., 273, 5439 ± 5442. Virmani AK, Fong KM, Kodagoda D, McIntire D, Hung J, Tonk V, Minna JD and Gazdar AF. (1998). Genes Chrom. Cancer, 21, 308 ± 319. Vos M.D, Ellis CA, Bell A, Birrer MJ and Clark GJ. (2000). J. Biol. Chem., in press. Wei MH, Latif F, Bader S, Kashuba V, Chen JY, Duh FM, Sekido Y, Lee CC, Geil L, Kuzmin I, Zabarovsky E, Klein G, Zbar B, Minna JD and Lerman MI. (1996). Cancer Res., 56, 1487 ± 1492. Wistuba II, Behrens C, Milchgrub S, Bryant D, Hung J, Minna JD and Gazdar AF. (1999). Oncogene, 18, 643 ± 650. Wistuba II, Behrens C, Virmani AK, Mele G, Milchgrub S, Girard L, Fondon III JW, Garner HR, McKay B, Latif F, Lerman MI, Lam S, Gazdar AF and Minna JD. (2000). Cancer Res., 60, 1949 ± 1960.