Frequent inactivation of p16INK4a in oral premalignant lesions - Nature

Oncogene (1997) 14, 1799 ± 1803  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Frequent inactivation of p16INK4a in oral premalignant lesions Vali Papadimitrakopoulou1, Julie Izzo2, Scott M Lippman1, Jin Soo Lee1, You Hong Fan1, Gary Clayman3, Jay Y Ro4, Walter N Hittelman2, Reuben Lotan5, Waun K Hong1 and Li Mao1 1 5

Department of Thoracic/Head and Neck Medical Oncology, 2Clinical Investigation, 3Head and Neck Surgery, 4Pathology and Tumor Biology, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Boulevard, Houston, Texas 77030, USA

Head and neck carcinogenesis is believed to be a multistep process, whereby genetic events accumulate in the carcinogen-exposed ®eld at risk, resulting in distinct phenotypic premalignant changes that eventually evolve into invasive cancer. Frequent loss of heterozygosity (LOH) at the chromosome 9p21 region and inactivation of p16INK4a by dierent mechanisms have been described in head and neck squamous cell carcinoma (HNSCC). Recently, we reported that loss of 9p21 is also frequent in oral premalignant lesions. To investigate potential inactivation of p16INK4a in these premalignant lesions, we analysed 74 biopsies from 36 patients by immunohistochemistry (IHC) for expression of the p16 protein. Loss of p16 expression was found in 28 (38%) of the lesion biopsies from 17 patients (47%). LOH at the D9s171, a marker in the 9p21 region, was observed in 19 lesion biopsies from 12 cases and correlated with absence of p16 by IHC in 11 (92%) of the 12 comparable cases and 15 (79%) of 19 lesion biopsies. By direct sequencing of ten lesion biopsies from ten individuals with LOH at D9s171 for p16INK4a exon 2, one non-sense mutation at codon 88 (GGA?TGA) was identi®ed. Our data suggest that inactivation of p16INK4a may play an important role in early head and neck cancer development. Keywords: p16INK4a; head and neck cancer; oral premalignant lesions; immunohistochemistry

Introduction Head and neck squamous cell carcinoma (HNSCC) remains a major cause of morbidity and mortality worldwide (Parkin et al., 1988). Despite advances in early diagnosis and treatment, the incidence and mortality rates have not signi®cantly changed over the last two decades (Parker et al., 1996; De Vesa et al., 1995). Oral leukoplakia, harboring histologic features of hyperplasia and/or dysplasia is a frequent precursor of squamous cell carcinoma of the oral cavity. Advanced oral premalignant lesions are associated with a well established rate of malignant transformation; 17% overall and 36% at 8 years for dysplastic lesions in the largest US series (Silverman et al., 1984). The occurrence of multiple premalignant lesions at dierent epithelial sites of the upper aerodigestive tract and multiple primary tumors synchronous or metachronous to the initial HNSCC, are two clinical

Correspondence: L Mao Received 30 September 1996; revised 23 December 1996; accepted 6 January 1997

manifestations of the ®eld cancerization phenomenon (Slaughter et al., 1953). Epithelial carcinogenesis in the head and neck area is believed, as is the case with many other epithelial cancers, to be a multistep process. The de®nition and the sequence of the speci®c genetic events that lead to the development of invasive HNSCC has been the subject of intensive study by many dierent groups. The previously conducted search for a tumor suppressor gene on chromosome 9p21 has identi®ed CDKN2/MTS1 coding for the cyclin dependent kinase (cdk) inhibitor p16 (Serrano et al., 1993; Kamb et al., 1994). Inactivation of p16 was frequently found in many tumor types including HNSCC and is thought to be caused by several dierent mechanisms including point mutations, homozygous deletion and DNA methylation (Okamoto et al., 1994; Merlo et al., 1995; Cairns et al., 1995). In HNSCC, inactivation of p16 by mutation was reported in 10% (Zhang et al., 1994; Cairns et al., 1994), by homozygous deletion in 33% (Cairns et al., 1995), and by methylation in 20% (Merlo et al., 1995) of primary tumors. Another level of complexity in the analysis of p16 inactivation arises from the existence of an alternative gene transcript p16b containing a dierent exon, E1b, in place of the ®rst p16 coding exon and derived from a distinct promoter (Mao et al., 1995; Stone et al., 1995). It was recently shown that clonal events in the form of loss of heterozygosity at the 3p14 and 9p21 chromosomal regions are detectable early in the head and neck carcinogenic process and can predict subsequent development of invasive cancer (Califano et al., 1996; Mao et al., 1996). To de®ne whether inactivation of p16 is an important event in oral premalignant lesions, we analysed a set of leukoplakia biopsies obtained from patients that were enrolled in an NCI chemoprevention trial conducted at MD Anderson Cancer Center (Lippman et al., 1993) by immunohistochemistry (IHC) and direct DNA sequencing analysis.

Results Frequent p16 inactivation in oral leukoplakia We analysed 76 leukoplakia biopsy specimens from 36 patients by immunohistochemistry for the p16 protein. Interpretable staining was obtained in 74 biopsies from 36 patients. Inadequate staining (cytoplasmic only) and partial tissue destruction did not allow evaluation of two biopsies. Paran-embedded buttons of the Hela and MAD886Ln cell lines were used as positive and negative controls for each staining procedure. Western

p16INK4a in oral premalignant lesions V Papadimitrakopoulou et al

1800

blot analysis of these specimens with the p16 monoclonal antibody used for the immunohistochemistry procedure con®rmed the presence of the normal sized protein in the Hela cell line and its absence in the MAD886Ln cell line (data not shown). p16 protein was undetectable in 27 of the 74 evaluable lesions (36%) and in 17 of the 36 evaluable patients (47%). In nine of the total 36 cases there were dierences in p16 expression between dierent biopsies obtained from either distinct anatomic sites or/and at dierent timepoints. Representative immunohistochemical staining of two p16 negative and two positive biopsies is shown in Figure 1. Of the 74 lesions 23 were dysplastic and in these lesions p16 expression was lost in 10 (43%), while it was lost in 18 (35%) of the 51 biopsies without dysplasia. Although in some tissues demonstrating immunoreactivity for p16, the signal localized in both the nucleus and the cytoplasm of individual cells, in most cases there was clear cut nuclear staining only. The signi®cance of cytoplasmic staining is unclear to date and has been observed by others (Geradts et al., 1995). The majority of p16-negative lesions showed complete absence of p16 immunoreactivity, while stromal cells demonstrating positive p16 staining served as internal controls (Figure 1). The percentage of cells demonstrating nuclear staining in p16-positive lesions ranged from 7% to 76% and several of these lesions were focally positive, while other areas of the epithelium were p16-negative. Of the 36 cases with interpretable immunohistochemical staining, 31 were informative for LOH at D9s171 (three non informative and two cases with a novel allele). Of these 31 cases, 12 (39%) demonstrated LOH at D9s171 according to our prior study (Mao et al., 1996), including 19 of 58 (33%) informative lesions. When the results of immunohistochemistry and LOH analysis were compared, LOH at D9s171 was correlated with absence of p16 by IHC in 11 (92%) of the 12 cases with LOH and 15 (79%) of 19 lesion biopsies with LOH. In 15 (79%) of 19 cases and 32 (84%) of 38 biopsies without LOH there was concordance between retention of the

D9s171 locus and presence of p16 protein expression, respectively (Table 1). In four cases (six biopsies) there was retention of D9s171 but no p16 protein expression and in one case (four biopsies) p16 expression was present despite LOH at 9p21. Point mutation of p16 in oral leukoplakia Of the 12 cases that demonstrated LOH at 9p21, DNA for direct sequencing was available in ten. A mutation was detected in only one of these ten cases. It consisted of a codon 88 G to T transversion resulting in a stop codon (Figure 2a) and p16 expression was not detectable by IHC in the same biopsy (Figure 2b). The outcome of patients with loss of p16 expression was examined. After a median follow-up of 63 months, eight of the 36 patients (22%) developed HNSCC. Five of 17 patients (29%) with loss of p16 expression by IHC developed HNSCC while three of 19 (16%) patients with conserved p16 expression developed cancer. Discussion In the present study we attempted to de®ne the role of p16 inactivation in head and neck carcinogenesis by

Table 1 Status of p16 by IHC and correlation with LOH at D9s171 in 36 patients (74 leukoplakia biopsies) Patients Positive Negative LOH No LOH NI ND New allele Total

1 15 2 ± 1 19

11 4 1 ± 1 17

Total 12 19 3 ± 2 36

Lesions Positive Negative 4 32 4 5 1 46

Total

15 7 1 2 3 28

*ND, not done; NI, non-informative lesions

Figure 1 p16 immunohistochemical staining of oral premalignant lesions. (a) and (b) p16-positive oral premalignant lesions. (c) and (d) Negative p16 immunostaining of oral premalignant lesions with positive stromal elements as internal control

19 39 5 7 4 74


A T G C A

T G C A

Codon 88 GAG ➝ TAG

Epithelium

Stroma

Figure 2 Detection of a p16 mutation in an oral premalignant lesion. (a) Direct sequencing of a biopsy with LOH at D9s171 demonstrating a G to T transversion in codon 88. (b) Negative p16 immunostaining of the same biopsy

analysing oral premalignant lesions in patients that participated in a chemoprevention trial aimed at reversal of these lesions. Cell cycle regulation is crucial in tumorigenesis and depends upon the action of cdks, activities of which are regulated by cyclins and cdk inhibitors (Hunter and Pines, 1994). Cyclin D-cdk4 and/or cdk6 complexes trigger phosphorylation of Rb during mid to late G1 cancelling its growth suppressive function, with subsequent release of the E2F transcription factor and entry into the S phase of the cell cycle. The activity of cyclin D-cdk complexes is positively regulated by mitogenic growth factors and negatively by a group of cdk inhibitors, which include at least four members of the INK4 gene family (p16INK4a, p15INK4b, p18INK4c, and p19INK4d). Cdk inhibitors may be responsible for maintenance of cell cycle and proliferation arrest once the cell's developmental fate has been reached. Therefore, the inactivation of these genes may interfere with terminal dierentiation and lead to unrestricted proliferation and tumorigenesis. Inactivation of the p16INK4a gene by complex mechanisms has been frequently found in many human cancers including HNSCC (Nobori et al., 1994; Cairns et al., 1995; Reed et al., 1996). Although point mutations were not frequent in most of these tumor types (Zhang et al., 1994; Cairns et al., 1994), frequent homozygous deletion in this region (Cairns et al., 1995) and methylation of the gene (Merlo et al., 1995), are considered major inactivation mechanisms in these tumors. Alternatively, another tumor suppressor gene(s), close to the p16INK4a locus such as p15INK4b, p16b, or others may play a role either in cooperating with p16 or acting independently during tumor development. p15INK4b encodes for a cdk inhibitor, p15, thought to play an integral role in transforming growth factor b-induced cell cycle arrest in some cell

types (Hannon and Beach, 1994), acting as an eector of extracellular growth inhibitory signals, while p16 may function in intracellular growth regulatory pathways. p16b is a second transcript of the human p16INK4a gene with distinct coding potential produced from a distinct promoter (Mao et al., 1995; Stone et al., 1995). Both forms of the p16 transcript p16a and p16b have been detected in various tissues (Quelle et al., 1995) in dierent ratios and the balance between the two may play a role in regulation of the cell cycle. The use of immunohistochemistry allows dierential topographic evaluation of the status of protein expression on a cell by cell basis in a tissue. Our immunohistochemical evaluation of the status of p16 protein expression provides further support for the important role of this gene in head and neck tumorigenesis, since 47% of the patients with oral premalignant lesions (and 38% of the oral premalignant lesions) demonstrated lack of p16 protein expression (Table 1). Furthermore, the results of immunohistochemistry were in concordance with LOH analysis in approximately 80% of the cases. Speci®cally, the fact that the percentage of oral premalignant lesions with LOH at 9p21 closely approximates the percentage of lesions with p16 loss by IHC is consistent with the hypothesis that p16 is the major target of inactivation on chromosome 9p. However, in four cases there was a loss of protein expression despite apparent retention of both alleles at the D9s171 locus. One possible explanation for this is that the use of only one marker to characterize the chromosomal region was not sucient to detect a small deletion in the same chromosomal region, which could account for the loss of protein expression. Alternatively, the presence of normal contaminating DNA may contribute to a retention pattern in cases where in fact a homozygous deletion is present. A third possibility that could account for the loss of protein expression is DNA methylation of both alleles of p16INK4a. A single case (and four lesions) that demonstrated LOH at D9s171 with retention of protein expression by IHC, suggest the presence of other putative tumor suppressor gene(s) in the 9p21 region. Direct sequencing of p16INK4a in oral premalignant lesions demonstrating LOH at D9s171 revealed that only one (10%) of ten lesions contained a point mutation, con®rmed by negative immunohistochemical staining for the p16 protein. This codon 88 G to T transversion has been previously described (SmithSùrensen et al., 1996). Alternative mechanisms of p16INK4a inactivation could account for loss of protein expression in the remaining cases. Methylation of the 5' CpG island of p16, resulting in silencing of the p16 promoter and loss of transcription is one such mechanism that has been previously described in approximately 20% of head and neck tumors (Merlo et al., 1995). Another possible type of p16INK4a inactivation consists of small homozygous deletions that eliminate the INK4a locus, which has been described previously in 33% of head and neck tumors (Cairns et al., 1995). Unfortunately, due to limited sample size, homozygous deletions and the methylation status of p16 in the premalignant lesions could not be determined in this study. However, a more extensive mutational, homozygous deletion and methylation

1801


1802

status analysis in specimens from patients with oral premalignant lesions enrolled in our ongoing chemoprevention study is currently underway. The ®ndings of the present study though raise a number of important issues. We found that 43% of the dysplastic lesions had lost p16 expression, whereas 18 of the remaining nondysplastic lesions (35%) demonstrated such a loss. These ®ndings support inactivation of p16 at an early stage in the multistep process of head and neck tumorigenesis and are consistent with previous findings that showed LOH at 9p21 in 20% of hyperplastic lesions, with a sharply increasing percentage in dysplastic lesions (57%) and further increase in the carcinoma in situ and invasive cancer stage (Califano et al., 1996). Our lower percentage of 9p21 loss and p16 loss by IHC in dysplastic lesions may be due to the fact that the majority of dysplastic lesions in our study demonstrated only mild dysplasia, with only two of 23 lesions showing moderate to severe dysplasia. Our ®ndings are also consistent with those of investigators that found 9p loss early during lung tumorigenesis (in hyperplastic lesions) by studying non-small cell lung cancer specimens containing premalignant lesions (Kushimoto et al., 1995). Another interesting ®nding in our study is that despite our ability to sequentially biopsy the same lesion from many of these patients we found dierences regarding p16 expression between biopsies from the same patient obtained at dierent timepoints in nine of 36 cases. This re¯ected the fact that a dierent anatomical site was biopsied in four cases and the possibility that clonal populations of cells can exist even in histopathologically homogeneous lesion that could be detectable in one biopsy and not another. Of the eight individuals with oral premalignant lesions who subsequently developed invasive squamous cell carcinoma, ®ve (63%) had lost p16 expression in at least one of the biopsies. Since the total number of patients that eventually developed cancer was small, no de®nitive statement can be made about the predictive value of loss of p16 for subsequent malignant transformation. Our prior study, however, demonstrated a signi®cantly shorter time to cancer development in the group with LOH at 9p21 and/or 3p14 than in the group without LOH (Mao et al., 1996), suggesting that the combination of multiple genetic events is necessary for and a better predictor of malignant transformation. These ®ndings also lend support to the concept of multistep head and neck tumorigenesis, where p16 alterations could cooperate with other genetic changes, such as cyclin D1 overexpression and ampli®cation (Lukas et al., 1995; Izzo et al., 1996) and loss of 3p, in the progression to invasive cancer. The disruption of G1-S transition control, through alteration of the key regulators of this transition, namely cyclin D1 and p16 could result in genomic instability and carcinogenic promotion. Further investigation of these and other genetic events implicated in deregulation of cell cycle control and tumorigenesis, in a larger number of individuals with oral premalignant lesions from our current large-scale NCI chemoprevention trial is ongoing. These studies provide unique opportunities to gain insight into the genetic events that drive the carcinogenesis process and the eects of chemopreventive interventions on possible modi®cation of the changes that de®ne the malignant phenotype.

Materials and methods Cell lines For immunohistochemistry we used the HeLa cell line with a known positive p16 status as a positive control and the MAD886Ln cell line as a negative control (Serrano et al., 1993; Lydiatt et al., 1995). The cultured cells were harvested, pelleted and ®xed in 10% buered formalin, processed and paran embedded using standard procedures. Premalignant lesions Leukoplakia biopsies were obtained from patients who participated in a previously reported oral premalignancy chemoprevention study (Lippman et al., 1993). Paranembedded tissue blocks from a total of 76 biopsy samples from 36 patients were available for this study. For the purposes of microdissection for LOH analysis and sequencing, four to ®ve sections (4 mm) from each samples were stained with hematoxylin and eosin, and one section from each specimen was reviewed by a pathologist (JYR) for histologic features. Microdissection and DNA extraction The epithelial part of each lesion was microdissected from three to four serial sections of each biopsy by using a 25 gauge steel needle and microdissected stroma cells were used as normal control for each lesion. The samples were digested in 100 ml 50 mM Tris-HCl (pH 8.0) containing 1% dodecyl sulfate-proteinase K and incubated at 428C for 12 ± 24 h. Digested products were boiled for 10 min to inactivate enzymes and puri®ed by using phenol chloroform. DNAs were precipitated by using the ethanol precipitation method in the presence of glycogen (Boehringer-Mannheim, Indianapolis IN). LOH analysis Between 500 and 1000 nuclei were dissected from both epithelium and stroma of each sample and DNA from at least 150 nuclei was used for each PCR ampli®cation. The marker used was D9s171 at 9p21 (Research Genetics, Huntsville, Alabama). PCR ampli®cation was performed as described previously (Mao et al., 1996). LOH was de®ned as more than 50% reduction of the intensity by visual inspection in one of the two alleles as compared with normal control. The patient was considered to exhibit LOH if at least one sample showed LOH. A novel allele was de®ned as a clear band that did not appear on the normal panel. PCR and sequence analysis The exon speci®c primers for p16 exon 1 were described previously (Washimi et al., 1995). The PCR product was run on 1.5% agarose gel and visualized by ethidium bromide staining. After ethanol precipitation and recovery in the appropriate amount of distilled water, an aliquot of ampli®ed DNA was used for each direct sequencing reaction. Brie¯y, puri®ed DNA and sequencing primers labeled with [g-32P] or [g-33P]ATP as described were subjected to PCR ampli®cation using the AmpliCycle sequencing kit (Perkin-Elmer, Branchburg, NJ). Each ampli®ed product (3 ml) was run on 6% Long-range gel (FMC BioProducts, Rockland, ME). Autoradiography with ®lm exposure for 12 to 48 h was performed. Each mutation identi®ed was con®rmed by repeat sequence analysis.


Immunohistochemistry Mouse monoclonal IgG1 antibody, against the full-length p16 was obtained from PharMingen (San Diego, CA). For detection the Elite ABC kit (Vector, Burlingame, CA) was used. The immunohistochemical assay for p16 in paranembedded tissues was performed following a microwave oven antigen retrieval step using hot citrate buer for 10 min (Cattoretti et al., 1993). Five mm paran sections of oral premalignant lesions were mounted on polysilanecoated slides, dewaxed, rehydrated and incubated with p16 antibody (normal preimmune mouse serum was used as negative control) at a 1 : 400 dilution at 48C overnight. Biotin-conjugated antimouse antibody was added at a 1 : 200 dilution for 45 min at 378C. For color development we used 3,3'-diaminobenzidine tetrahydrochloride and H2O2. Hematoxylin was used as counterstain. A cell was considered positive if there was a visibly detectable signal within the nucleus or within the nucleus and cytoplasm. A lesion was considered positive if more than 5% of the cells demonstrated nuclear staining. For this purpose at least

250 epithelial cells were counted in each lesion. A case was considered to be negative if at least one of the biopsied lesions was negative. The slides were reviewed by two independent observers (VP, JI) under light microscopy. Additional controls for immunohistochemical analysis included formalin-®xed, paran-embedded HeLa cell line which shows p16 expression and the MAD886Ln cell line which does not express p16. One of each (negative and positive control) was included in each staining procedure along with the patient slides. In addition nuclear staining in admixed normal stromal cells was used as internal positive control for each stained lesion.

Acknowledgements We thank Susan Cweren for technical support. Waun K Hong is an American Cancer Society Clinical Research Professor. This work is supported in part by NIH-NCI Grant CA-52051 (WKH) and MD Anderson PRS grant (LM).

References Cairns P, Mao L, Merlo A, Lee DJ, Schwab D, Eby Y, Tokino K, van der Riet P, Blaugrund JE and Sidransky D. (1994). Science, 265, 415 ± 416. 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 KR, Hedrick L, Bova GS, Isaacs W, Koch W, Schwab D and Sidransky D. (1995). Nat. Genet., 11, 210 ± 212. Califano J, van der Riet P, Westra W, Nawroz H, Clayman G, Piantadosi S, Corio R, Lee D, Greenberg B, Koch W and Sidransky D. (1996). Cancer Res., 56, 2488 ± 2492. Cattoretti G, Pileri S, Parravacini C, Becker MHG, Poggi S, Bifulco C, Key G, D'Amato L, Sabattini E, Feudale E, Reynolds F, Gerdes J and Rilke F. (1993). J. Pathol., 171, 83 ± 98. De Vesa SS, Blot WJ, Stone BJ, Miller BA, Tarone RE and Fraumeni JF. (1995). J. Natl. Cancer Inst., 87, 175 ± 182. Geradts J, Kratzke RA, Niehans GA and Lincoln CE. (1995). Cancer Res., 55, 6006 ± 6011. Hannon GJ, Beach D. (1994). Nature, 371, 257 ± 261. Hunter T and Pines J. (1994). Cell, 79, 573 ± 582. Izzo J, Papadimitrakopoulou V, Lee JS, Ro JY, Li XQ, Hong WK, Hittelman WN. (1996). Proc. Am. Assoc. Cancer Res., 37, 3687. (abstract). Kamb A, Gruis NA, Weaver-Feldhaus J, Lie Q, Harshman K, Tatvigian SV, Stockert E, Day RS, Johnson BE and Skolnick MH. (1994). Science, 264, 436 ± 440. Kushimoto Y, Sugio K, Hung JY, Virmani AK, McIntire DD, Minna JD and Gazdar AF. (1995). J. Natl. Cancer Inst., 87, 1224 ± 1229. Lippman SM, Batsakis JG, Toth BB, Weber RS, Lee JJ, Martin JW, Hays GL, Goepfert H and Hong WK. (1993). N. Engl. J. Med., 328, 15 ± 20. Lukas J, Aagard L, Strauss M and Bartek J. (1995). Cancer Res., 55, 4818 ± 4823. Lydiatt WM, Murty VVVS, Davidson BJ, Xu L, Dyomina K, Sacks PG, Schantz SP and Chaganti RSK. (1995). Genes Chromosom. Cancer, 13, 94 ± 98. Mao L, Merlo A, Bedi G, Shapiro GI, Edwards CD, Rollins BJ and Sidransky D. (1995). Cancer Res., 55, 2995 ± 2997.

Mao L, Lee JS, Fan YH, Ro JY, Batsakis JG, Lippman SM, Hittelman WN, Hong WK. (1996). Nature Med., 2, 682 ± 685. Merlo A, Herman JG, Mao L, Lee D, Gabrielson E, Burger PC, Baylin SB and Sidransky D. (1995). Nature Med., 1, 686 ± 692. Nobori T, Miura K, Wu DJ, Lois A, Takabayashi K and Carson DA. (1994). Nature, 368, 735 ± 756. Okamoto A, Demetrick DJ, Spillare EA, Hagiwara K, Hussain SP, Bennett WP, Forrester K, Gerwin B, Serrano M, Beach DH and Harris CC. (1994). Proc. Natl. Acad. Sci. USA, 91, 11045 ± 11049. Parker SL, Tong T, Bolden S, Wingo PA. (1996). CA Cancer J. Clin., 46, 8 ± 9. Parkin DM, Laara E, Muir CS. (1988). Int. J. Cancer, 41, 184 ± 197. Quelle DE, Zindy F, Ashmun RA and Sherr CJ. (1995). Cell, 83, 993 ± 1000. 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. Serrano M, Hannon GJ and Beach D. (1993). Nature, 366, 704 ± 707. Silverman S Jr, Gorsky M and Lozada F. (1984). Cancer, 53, 563 ± 568. Slaughter DP, Southwick HW and Smejkal W. (1953). Cancer, 6, 963 ± 968. Smith-Sùrensen B, Hovig E. (1996). Human Mutation, 7, 294 ± 303. Stone S, Jiang P, Dayananth P, Tatvigian SV, Katcher H, Parry D, Peters G, Kamb A. (1995). Cancer Res., 55, 2988 ± 2994. Washimi O, Nagatake M, Osada H, Ueda R, Koshikawa T, Seki T, Takahashi T and Takahashi T. (1995). Cancer Res., 55, 514 ± 517. Zhang SY, Klein-Szanto AJP, Sauter ER, Shafarenko M, Mitsunaga S, Nobori T, Carson DA, Ridge JA, Goodrow TL. (1994). Cancer Res., 54, 5050 ± 5053.

1803