Oncogene (1997) 14, 811 ± 818  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Alterations of the p16 gene in head and neck cancer: frequency and association with p53, PRAD-1 and HPV

Andrew F Olshan1,2, Mark C Weissler2, Hong Pei1, Kathleen Conway1, Steven Anderson3, Daniel B Fried1 and Wendell G Yarbrough2

1Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, North Carolina; 2Division of Otolaryngology/Head and Neck Surgery, School of Medicine, University of North Carolina, Chapel Hill, North Carolina; 3Department of Molecular Biology and Pathology, Laboratory Corporation of America, Research Triangle Park, North Carolina, USA

Alterations, especially homozygous deletions, of the suppressor gene (Brachman, 1994). Recently, the putative tumor suppressor gene, p16 (p16INK4A, MTS1, cyclin-dependent kinase inhibitor, p16 (p16INK4A, CDKN2) have been found in tumor cell lines from a MTS1, CDKN2), has been proposed as the tumor variety of neoplasms. Recent studies have reported suppressor gene at 9p21 (Serrano et al., 1993, 1996; frequent p16 gene deletions in cell lines from squamous Kamb et al., 1994; Nobori et al., 1994; Lukas et al., cell carcinomas of the head and neck (SCCHN), 1995a; Koh et al., 1995; Spillare et al., 1996). Related although the prevalence of alterations was variable in CDK-inhibitors have also been isolated including primary tumors. This study determined the prevalence of p15INK4b, p18INK4C and p19INK4d (Sherr and Roberts, point mutations and deletions of the p16 gene in 33 1995). Further, the p16INK4a locus yields two transcripts SCCHN. In addition, the association of p16 gene one encoding p16INK4a and another encoding p19ARF alterations and abnormalities of p53, PRAD-1 (cyclin (Quelle et al., 1995). All of these genes participate D1), and the presence of human papillomavirus (HPV) either directly or indirectly in the control of the cell was examined. We found an overall prevalence of p16 cycle. alterations of 36% (nine deletions, three single base The transition from G1 to S in the cell cycle relies on substitutions, including one polymorphism). Seven tumors a coordinated interaction of several gene products (of 29, 24%) had an alteration of p16 and p53; ®ve (of including the retinoblastoma protein (Rb), cyclin D1, 33, 15%) had alterations of p16 and PRAD-1; three (of and its catalytic partners CDK4 and CDK6, and the 29, 10%) had alterations of all three genes. In addition, cyclin-dependent kinase inhibitor, p16 (Peter and of the ®ve tumors with human papillomavirus detected, Herskowitz, 1994; Grana and Reddy, 1995; Yeudall only one also had a p16 gene alteration. The results and Jakus, 1995). Loss of control of the G1 to S indicate a potentially important role for the p16 gene in transition is a common defect observed in many head and neck tumorigenesis. In addition, the presence of di€erent tumor types and can occur through a variety tumors with multiple somatic gene alterations suggest a of mechanisms (Hartwell and Kastan, 1994; Kamb, possible interaction in the dysregulation of the cell cycle. 1995). Viruses that are associated with tumors, such as the human papillomavirus (HPV) and SV40, produce Keywords: head and neck neoplasms; mutation; p53; proteins that inactivate Rb and p53. Also, mutations p16INKA4; HPV; PRAD-1 and/or deletions of RB, p53, and p16 and ampli®cation and overexpression of cyclin D1 are relatively common events in a wide range of tumors. Recently, evidence for transcriptional repression of p16 through promotor Introduction methylation has been observed in all major solid tumor types, suggesting another mechanism through which

Squamous cell carcinoma of the head and neck loss of control of the G1-S transition can occur (Merlo (SCCHN, including oral cavity, pharynx and larynx) et al., 1995). It has been suggested that di€erent genes provides an excellent model for understanding the related to the control of the cell cycle represent carcinogenic process due to the de®ned risk factors alternate targets and that functionally inactivating (tobacco and alcohol consumption) and emerging any one component will have the same consequence molecular characteristics. Cytogenetic, allelotyping, of loss of cell cycle control. Little research has been loss of heterozygosity, and gene ampli®cation studies done to evaluate alternate pathways by examining the have implicated 11q as the site of a putative oncogene, interrelationship among abnormalities in cell cycle and 17p, 9p and 3p as sites of putative tumor genes in tumors. The relationship betwen p16 and suppressor genes in SCCHN (Cowan et al., 1992; other genes, including HPV has not been directly Nawroz et al., 1994; Van der Reit et al., 1994). investigated in SCCHN. Molecular studies have identi®ed candidate genes for Examination of primary tumors and cell lines for these regions, namely cyclin D1 (PRAD-1, CCND1) as alterations of the p16 gene has yielded con¯icting the protooncogene and p53 as a putative tumor results. Initial reports indicated a high frequency of homozygous deletions of p16 in a variety of cell lines (Kamb et al., 1994; Nobori et al., 1994). In addition, Correspondence: AF Olshan point mutations were identi®ed in cell lines without Received 14 June 1996; revised 18 October 1996; accepted 18 October deletions. Other studies using primary tumors reported 1996 a much lower prevalence of p16 deletions and p16 gene alterations and head and neck cancer A F Olshan et al 812 mutations, suggesting that tumor cells might acquire a p16 defects as a consequence of forming established 12345678 lines (Carins et al., 1994; Spruck et al., 1994). More GAPDH recent studies suggest that p16 defects are not an p16 artifact of establishing cell lines because they occur in a high percentage of certain types of primary carcinomas (Mori et al., 1994; Zhou et al., 1994; Caldas et al., p16 1994; Schmidt et al., 1994; Carins et al., 1995). Two CPT analyses of primary head and neck tumors have reported a prevalence of point mutations of 7% and 10%, con®rming that the loss of p16 function more commonly occurs through deletion (Carins et al., 1994; Zhang et al., 1994). Two recent reports indicated a b higher prevalence of p16 gene alterations in head and 123456 neck tumors 25% (16/65; all deletions; Carins et al., GAPDH 1995) and 19% (7/36; six deletions; Lydiatt et al., p16 1995). None of the studies of SCCHN evaluated the relationship between p16 gene alterations and the p16 presence of alterations of other cell cycle associated CPT genes. In this study, we determined the prevalence of p16 0 20 40 60 80 100 somatic alterations in 33 SCCHN tumors and have % Normal DNA examined the association of p16 mutation or deletion Figure 1 Representative deletion analysis of p16 gene in head with aberrations of other tumor suppressor genes and and neck tumors. (a), PCR products from six tumors and two oncogenes, and with the presence of human papillo- control samples with two reference genes, GAPDH and CPT. mavirus (HPV). Our results provide additional Lanes 2, 4, 5, deletions of the p16 target gene. Lane 7, sample of normal tonsil tissue. Lane 8, cell line with p16 homozygous evidence for the role of the inactivation of p16 in the deletion (K562). The estimated percentage of contaminating development of head and neck cancer. normal cells and ratio of the intensity of the target gene band to the reference gene (CPT) was: lane 1 (70%, 1.04), 2 (25%, 0.24), 3 (60%, 1.03), 4 (70%, 0.64), 5 (60%, 0.63), 6 (60%, 1.05), 7 (100%, 1.00), 8 (0%, 0.00). (b) Comparison PCR using a mixed Results DNA template from human foreskin ®broblast and a human leukemia cell line (K562) with a known homozygous deletion of Among the 33 tumors examined, a total of 12 (36%) the p16 gene and GAPDH and CPT as reference genes. The had either a deletion (n=9) or single base substitution percentage of normal DNA mixed in the template is shown at the in exon 2 of the p16 gene (n=3). Using di€erential bottom of the lanes. The ratio of the intensity of the target gene band to the reference gene (CPT) band for each sample was: lane PCR, independently coampli®ed each time with a 1 (0.00), 2 (0.31), 3 (0.53), 4 (0.69), 5 (0.78), 6 (1.01) di€erent set of reference genes (PGR/d-IFN, GAPDH/CPT), we detected p16 deletions in 9 (27%) of the tumors (Figure 1 and Table 1). Figure 2 shows an example of selected mutations. tumors had a p16 base change, did not show any p16 Four mutations were identi®ed in three tumors mutations or deletions. Although there were some including an A4T transversion (GAC4GTC, di€erences, no signi®cant association was found Asp4Val) at codon 76 (case #2300), a G4A between the presence of a p16 mutation and SCCHN transition (GCG4ACG, Ala4Thr) at codon 140 site (P=0.74), recurrence status (P=0.43), or history (case #4165), and a tumor (case #3515) with two of preoperative radio- or chemotherapy (P=0.42). G4T transversion mutations, one at codon 79 Additionally, there was no relation between p16 (CGG4CTG, Arg4Leu), the other at codon 80 alterations and Tumor-Node-Metastasis (TNM) stage (GAG4TAG, Glu4Stop). Mutations at codon 76 classi®cation (P=0.47). have been reported in a previous study of head and We were also able to successfully screen 29 of the 33 neck tumors and in a study of esophageal cancer SCCHN tumors for p53 mutations using SSCP and (Spruck et al., 1994; Mori et al., 1994). In both cases, direct sequencing. Twelve of 29 (41%) tumors were the sequence change was a transition from GAC to found to have a mutation of the p53 gene and seven of AAC; in our study the change was a transversion these 12 also had a p16 alteration (®ve deletions and from GAC to GTC. The base substitution two single base substitutions (P=0.15) (Table 1). (GCG4ACG) at codon 140 has been reported in PRAD1 ampli®cation status was evaluated in all 33 head and neck and other cancers and has been tumors (Table 1). Eight (24%) of the tumors had reported to be a population polymorphism (Cairns PRAD1 ampli®cation and ®ve of these also had a p16 et al., 1994; Spruck et al., 1994; Mori et al., 1994). deletion (P=0.10). There was no association between The peripheral blood-derived DNA from this patient PRAD1 ampli®cation and a p53 mutation (P=0.53). contained the same sequence alteration as in the HPV was detected in ®ve of the 29 tumors (17%) for tumor. A recent biochemical analysis found no which a de®nitive HPV determination could be made functional e€ect of this mutation (Ranade et al., (Table 1). All ®ve HPV-positive tumors were of HPV 1995). Because it is suspected that the p16 mutation type 16. The presence of HPV was partially from this tumor (#4165) is a polymorphism it was independent of alterations of the p16 gene, with only excluded from the statistical comparisons. Germline one tumor being both HPV-positive and having a p16 DNA samples for the other two patients, whose gene deletion (#3039). p16 gene alterations and head and neck cancer AFOlshanet al

813

„—˜le I hes™ription of study ™—ses —nd mole™ul—r results

— ˜ ™ d e f g h

ƒex ƒite „xw ‚—dG‚e™ pSQ r€† €‚ehI pIT sh ege ‚—™e

PPSS TS f w or—l n— ± ± ± B C ±

PQHH RR ‡ w or—l „PxHwH ± ± ± B ± mut

PQIT US ‡ p or—l „PxHwH C C C ± ± ±

PQRR SW f w or—l „QxHwH ± ± B C ± ±

PQWP RS f w or—l „QxIwH ± ± C ± C del

PRIS TV ‡ w or—l „QxPwH ± ± ± B ± ±

QHII UH f w l—rynx „RxHwH ± C C ± ± del

QHQW ST ‡ w ph—rynx „RxHwH C C ± C ± del

QHRW TP f w ph—rynx „RxHwH C C C C ± ±

QPRH UH ‡ p or—l „QxQwH ± ± ± B ± ±

QPRS RQ f w or—l „QxIwH ± ± C ± ± del

QPSI PS ‡ p or—l „IxHwH ± ± B ± ± ±

QQRH TI ‡ w or—l „RxQwH ± ± ± C ± ±

QSHI SV ‡ p or—l „RxHwH C C ± C ± ±

QSIH UP ‡ w l—rynx „PxHwH ± ± ± ± ± ±

QSIS UI f w l—rynx „IxHwH C C C ± ± mut

QSQH UT ‡ w l—rynx „RxHwH ± ± ± ± ± ±

QSQU VQ ‡ p or—l „QxHwH ± ± C ± ± ±

QTHT SR f w l—rynx „RxQwH ± ± C ± C del

QTIV SH ‡ w l—rynx „PxQwH ± ± ± ± ± ±

QUTI SS f w or—l „PxHwH ± ± ± ± C ±

RHHQ SV ‡ w l—rynx n— ± ± ± ± C del

RITS TT f w l—rynx „QxHwH ± ± C ± ± ply

RIUV SW ‡ w or—l „PxHwH ± ± ± ± ± del

RIWP SI f w or—l „PxHwH C C ± ± C del

RIWU RW ‡ w or—l „IxHwH ± C C ± ± ±

RQHV SW ‡ w or—l „RxPwH ± ± ± ± ± ±

RQIS TW ‡ p or—l „IxHwH ± ± B ± ± ±

RQPU RV f w or—l „RxHwH C C” C ± C del

RQQI TS f w l—rynx „IxHwH ± ± ± ± C ±

RQQU TS f p l—rynx „xxPwH ± ± B ± ± ±

RQRW RR f w l—rynx „IxHwH C C C ± ± ±

RSHI SV ‡ p ph—rynx „PxHwH ± ± ± ± ± ±

— ˜ ™ d

fafl—™kD ‡a‡hiteF wam—le pafem—leF „umorExodeEwet—st—sis gl—ssi®™—tionD n—aunknownF €reoper—tive r—di—tion tre—tmentG

e f

pSQ gene mut—tionD C a yesD ± a noD B a st—tus indetermin—teF r€† dete™tedD C a yesD ± a noD Ba st—tus ‚e™urren™eD Ca yesD ± a noF

g h

€‚ehEI gene —mpli®™—tionD C a yesD ± a noF pIT gene—lter—tionD mut a mut—tionD del a deletion ply a suspe™ted indetermin—teF

polymorphismD ± a no —lter—tionF ” €reoper—tive ™hemother—py

Discussion If there is sucient degradation of the DNA from the tumor samples one would expect the appearance of The overall prevalence of p16 alterations found in this false negatives. In addition, some deletions may be study was 33% (excluding the one p16 polymorphism), heterozygous rather than homozygous, although of which, nine (27%) were deletions and two (6%) previous studies have not supported this possibility. were single base mutations. The prevalence of point As described in the Materials and methods section, we mutations is very similar to Cairns et al. (1994) who dissected ®xed tumor specimens allowing us to obtain found one of 15 primary head and neck tumors tumor samples enriched for neoplastic cells. We also contained a single base mutation of p16 and margin- used a variation of multiplex PCR with two sets of ally lower than the 10% ®gure of Zhang et al. (1994). reference genes, each set including one gene fragment With respect to deletions, Zhang et al. did not ®nd any larger and one smaller than the target gene. In deletions among the 68 primary tumors they evaluated addition, we performed a mixing experiment to assess and Lydiatt et al. found six p16 gene deletions in 36 our limits of detection, as well as demanding agreement (17%) tumors. However, Cairns et al. (1995) recently among three independent reviewers. Nevertheless, this reported that among 65 SCCHN tumors, 16 (25%) had methodology will lead to some degree of under- a homozygous deletion. estimation of p16 deletions. It is encouraging that A common method of detecting p16 gene deletions is Cairns et al. (1995) reported a very similar estimate multiplex PCR which can be unreliable due to factors using other methods. They detected a homozygous such as variable amounts of contaminating normal deletion in 25% of their head and neck tumors cells. In general it would be expected that this method (compared with 27% in this study) using analysis of would be conservative, that is, false negatives due to microsatellite markers ¯anking the p16 locus and the appearance of bands resulting from the amplifica- ¯uorescent in situ hybridization (FISH). tion of normal DNA would lead to an underestimate The high proportion of deletions relative to point of the true prevalence of homozygous deletions. This mutations has lead some to suggest that the target of may help explain the results of Zhang et al. whose deletion may be p16 and/or a neighboring gene, p15 samples were probably fresh tumor samples that most (Hannon and Beach, 1994; Jen et al., 1994). Fine likely contained an unde®ned mixture of tumor and mapping of deletions in bladder tumors has identi®ed a normal cells. In addition, they utilized a single and minimal region that includes p16, but not p15 (Cairns rather large reference gene (GAPDH, 474 bp) in et al., 1995). In addition, an analysis of p16 and p15 comparison with the p16 target gene (around 200 bp). expression in oral tumor cell lines has indicated that p16 gene alterations and head and neck cancer A F Olshan et al 814 necessary to provide further growth advantage for a a EXON 2 A tumor (Serrano et al., 1996). The retinoblastoma gene N123456 (RB) appears to play a central role in the transition

from the G1 to S and its inactivation may completely eliminate this checkpoint (Weinberg, 1995). It has been found that the loss of p16 in tumors is inversely related to the presence of a functional RB protein (Rb) (Okamoto et al., 1994; Shapiro et al., 1995; Aagaard et al., 1995; Lukas et al., 1995). In addition, previous studies have shown that inactivation or Rb and overexpression of PRAD-1 are nonoverlapping in esophageal tumors (Jiang et al., 1993). Lukas et al. (1995b) using cell lines, including SCCHN lines, have shown that concurrent abnormalities of cyclin D1 and p16 can cooperate in Rb-positive cells, but not in Rb- negative cells. Thus, one might expect to ®nd coexistent alterations of upstream e€ectors of Rb in some tumors. Investigation of the relationship among these genes may be of particular interest in SCCHN, where mutational inactivation of Rb is infrequent (Yoo et al., 1994). In our sample, only two tumors (3340, 3530) were found to lack expression of Rb as determined by immunostaining methods and both did not have any abnormalities of p53, p16, or PRAD-1. Although a b statistically powerful analysis was not possible in this CTAG CTAG study we examined the potential interrelationship between p16 and other genes involved either directly or indirectly in regulation of the cell cycle including p53 and PRAD-1. Previous studies of SCCHN have not fully examined this complement of genes. Overall, we found of the 29 tumors that could be evaluated for alterations of p53, p16 and PRAD-1, three (10%) had an alteration of all three genes, six (21%) had two genes altered, and 11 (38%) had only one gene a€ected. Nine tumors (of 29, 31%) had an alteration of p16 and either p53 or PRAD-1. Of the eight tumors G G that had ampli®cation of PRAD-1, we found a p16 A Val T Asp deletion in ®ve (62%) providing further evidence for a C C possible collaborative role of p16 and PRAD-1 in some tumors. An analysis of p16 and p53 point mutations in a Normal Mutant variety of cell lines has suggested that the two genes may act in independent pathways (Okamoto et al., Codon 76 1994). Another study of melanoma and bladder cancer Figure 2 Representative SSCP and sequencing analysis. (a), cell lines and primary tumors did not ®nd that the SSCP analysis of p16 exon 2A gene fragment (see Table 1 for coexistence of mutations in p53 and p16 was primer sequence used). Lanes 4 and 5 show mobility shifts. (b), Codon 76 A to T mutation in case #2300. Mutation results in statistically signi®cant, although the study did not amino acid change from Valine (Val) to Aspartic acid (Asp) examine homozygous deletions of the p16 gene (Gruis et al., 1995). In our study, alteration of p16 was not statistically signi®cantly associated with p53 mutation, the inactivation of p16 may be of more signi®cance although seven of 12 (58%) tumors with a p53 (Yeudall et al., 1994). Deletions of p16 will also result mutation also had a p16 mutation. The Zhang et al. in deletion of p19ARF, a gene that shares exon 2 with study found that all the cell lines with a deleted or p16INK4a and also appears to be involved in cell cycle mutated p16 gene also had a mutated p53 gene; three control (Quelle et al., 1995). Further work is needed to of the eleven primary tumors with a mutation of p16 more precisely de®ne the roles of p15, p16 and p19ARF also had a p53 mutation. Neither this study nor the in SCCHN. study of Zhang et al. examined p16 promoter Recent research has provided new data on the role methylation as a mechanism of p16 inactivation. of various gene products in the regulation of the cell Further studies examining all mechanisms of p16 cycle, especially the biochemical pathway controlling inactivation may clarify the relationship between p16 and p53 in head and neck tumors. progression through the G1 phase. Abnormalities of these genes can lead to loss of control and tumorigen- Our study is the ®rst to evaluate the relationship esis (Hartwell and Kastan, 1994). It is generally between p16 and HPV in tumors. The lack of a assumed that genes such as p16, PRAD-1 (cyclin positive association between the presence of HPV and D1), CDK4 and RB are alternate targets in a common mutations or deletions of p16 and ampli®cation of pathway and alteration of multiple genes is not PRAD-1 extends the previously reported independence p16 gene alterations and head and neck cancer AFOlshanet al 815 of HPV and p53 gene mutations. Our study sample did (10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl2 and include two (of ®ve) tumors with HPV and either a p53 0.001% gelatin), 200 mM each deoxyribonucleotide and 2.5 or p16 gene deletion. It is thought that the HPV type units AmpliTaq DNA polymerase (Perkin Elmer.) Amplifica- 16 oncoproteins (E6 and E7) are related to carcinogen- tion was performed using 32 cycles of: 948C for 30 s, 608C esis through binding and inactivation of p53 (E6) and for 30 s and 728C for 30 s with a ®nal extension of 728C for 10 min. Exon 2 of the p16 gene was ampli®ed under similar Rb (E7) (Munger et al., 1989; Sche€ner et al., 1990). conditions but with 5% dimethylsulfoxide (DMSO), an Binding with these proteins appears to result in loss of annealing temperature of 558C and 34 cycles of amplifica- cell cycle control and genomic instability (White et al., tion. We have not evaluated exon 1 of the p16 gene owing to 1994; Xiong et al., 1996). One would not expect to ®nd the fact that there are a paucity of mutations reported for many p16 gene alterations in tumors in which Rb this exon and all missense mutations that have been proven inactivation had occurred by means of binding with to functionally inactivate the p16 protein reside in exon 2. HPV oncoprotein, analogous to the inverse correlation SSCP-PCR was performed on p53 exons 5, 6, 7 and 8, seen between p16 and Rb in tumors with inactivating using ®rst PCR products as template. Because of its size, mutations of RB. A recent study of p16 and HPV in exon 4 was divided into two segments (4-1 and 4-2) for SSCP. cervical carcinoma cell lines provides additional Exon 2 of p16 was analysed directly from genomic DNA as three separate fragments. SSCP ± PCR reactions consisted of support for this hypothesis (Kelley et al., 1995). a total volume of 20 ml containing 500 nM primers, 16 Cetus In summary, using primary tumors, we have bu€er, 200 mM each dNTP, 2.5 units AmpliTaq polymerase provided additional support for the importance of and 100 mCi/ml [32Pa]dCTP. Cycle conditions for p53 were as p16 alterations in SCCHN. Our ®ndings also suggest described for ®rst PCR. The p16 SSCP ± PCR reaction that a single tumor can harbor alterations of multiple conditions were the same as for p53 except that 5% DMSO genes involved in cell cycle control. Further work is was added and an annealing temperature of 558C was used. needed to clarify the relative importance and functional SSCP ± PCR products were diluted 30-fold with 1% SDS/ consequences of alterations of p16 and other genes 10 mM EDTA, the diluted product (3 ml) was mixed with an involved in the regulation of the cell cycle. Future work equal volume of stop bu€er (95% formamide, 20 mM EDTA, should also take into account the new ®ndings that 0.05% bromophenol blue, 0.05% xylene cyanol FF) and denatured at 948C for 5 min. SSCP reaction products were inactivation of the p16 gene by methylation may subjected to electrophoresis in 6% polyacrylamide non- represent an important pathway in the genesis of denaturing gels, both at room temperature (these gels head and neck tumors (Merlo et al., 1995). Finally, included 5% glycerol) and at 48C. Negative control samples molecular epidemiologic studies are also needed to containing only wild type p53 or p16 genes were included on evaluate whether SCCHN risk factors such as tobacco each gel. In addition, all experiments were run using a blank and alcohol are related to p16 gene mutation and control without template DNA and none showed ampli®ed deletion. products due to outside contamination. Tumors showing altered SSCP bands were further analysed as follows. The ®rst PCR product or gel-puri®ed abnormal SSCP bands were used as template for asymmetric PCR to generate single forward and reverse DNA strands. Asymmetric PCR was carried out under PCR conditions Materials and methods previously described but with primer concentrations of either 500 nM forward primer with 10 nM reverse primer or 10 nM Tumour samples forward primer with 500 nM reverse primer. Both forward A total of 33 tumors from patients diagnosed with either and reverse asymmetric PCR products were puri®ed and used primary (n=23) or recurrent (n=10) squamous cell as template in sequencing reactions. Samples were determined carcinoma of the head and neck at the University of to be positive only if both strands exhibited the same North Carolina Hospitals was obtained. Informed consent mutation. Mutations were veri®ed by repeating the entire was obtained from all patients for the use of tissue samples sequencing analyses on suspected positive samples. in research.

PRAD-1 ampli®cation DNA extraction The method of binary di€erential PCR (Liu et al., 1993) Molecular analyses were carried out on DNA puri®ed from was used to identify tumors with ampli®cation of the 10 mM sections of formalin-®xed paran-embedded tu- PRAD-1 (cyclin D1) gene. Gene ampli®cation was mors. These tumor sections were subjected to pathologic evaluated using both a 127 bp fragment of PGR review in which tumor areas were identi®ed, tumor- (progesterone receptor gene) and an 85 bp fragment of d- enriched tissue isolated from slides, and DNA extracted IFN (d interferon gene) as references for the 101 bp using a standardized protocol (Wright and Manos, 1990). fragment of PRAD-1. Each di€erential PCR reaction contained PRAD-1 as the target gene and one reference gene. We used additional reference loci (GAPDH, CPT) to p53 and p16 mutation screening verify ampli®cation in possible positive samples, and Tumors were screened for mutations in p53 exons 4 ± 8 and ampli®cation was performed at least twice for each p16 exon 2. Analyses were performed on DNA isolated reference gene on apparent positives. In addition, we used DNA lysates prepared from tumor-enriched formalin-®xed the FaDu cell line, which carries an approximately 10-fold paran-embedded tumor tissue and were veri®ed using level of PRAD-1 ampli®cation, as a positive control. DNA extracted from snap frozen tumors. All PCR, SSCP Typically, our positive samples showed a 5 ± 10-fold level and sequencing primers are given in Table 2. of ampli®cation. Primer sequences for di€erential PCR are The following segments of the p53 gene were ampli®ed in given in Table 2. Reactions were carried out in a 50 ml separate PCR reactions to generate template for both SSCP ± reaction mixture containing 100 ng genomic DNA or 2 ± PCR and sequencing: exon 4, exons 5 and 6, and exons 7 and 4 ml DNA lysate, 16 Cetus bu€er, 200 mM dNTPs, 2.5 8. PCR was carried out in a 50 ml reaction mixture containing units AmpliTaq. Primer concentrations were determined 100 ng genomic DNA, 500 nM each primer, 16 Cetus bu€er empirically and depended upon both the combination of p16 gene alterations and head and neck cancer A F Olshan et al

816

„—˜le P yligonu™leotide primers used for pSQD €‚ehEI —nd r€† —n—lyses

ƒequen™e

x—me @S9 to Q9A x—me ƒequen™e ƒize @f€A

€r—dEI emplifi™—tion or pIT heletion en—lyses

„—rget genes

pr—dEIEIHI e IHI €r—dEIEIHI ƒ ATGCTGAAGGCGGAGGAGA ATCCAGGTGGCGACGATCT

€ITEIHI e IHI pITEIHI ƒ GCCCTAAGCGCACATTCAT CCCTGAGCTTCCCTAGTTCA

‚eferen™e genes

dEspxEVS e VS EspxEVS ƒ d AGTGATGGCTGAACTGTCGC CTGGGATGCTCTTCGACCTC

€q‚EIPU e IPU €q‚EIPU ƒ TACGGAGCCAGCAGAAGTCC TCGTGCATGCTGTGAAGCTC

qe€hrEIPS e IPS qe€hrEIPS ƒ CCGGGAAACTGTGGCGTGAT GAAGGCCATGCCAGTGAGCT

g€„IEVS e VS g€„IEVS ƒ GCCTACCTGGTCAATGCGTA CCGACTGGGTTTGGGTAAAC

pSQ wut—tion ƒ™reening

pSQ ixon R@IAX

€R@IAEQ9 IVV 9 €R@IAES GACCTGGTCCTCTGACTGCT CGGTGTAGGAGCTGCTGGTG

9sx €R@IAES CCTCTGACTGCTCTTTTCAC

pSQ ixon R@PAX

€R@PAEQ9 PRW 9 €R@PAES TCCAGATGAAGCTCCCAGAA ACGGCCAGGCATTGAAGTCT

€R@PAEQ9 sx 9 sx €R@PAES AAGCTCCCAGAATGCCAGAG TCTCATGGAAGCCAGCCCCT

pSQ ixon SX

€SEQ9 PWR 9 €SES GCTGCCGTGTTCCAGTTGCT CCAGCCCTGTCGTCTCTCCA

€SEQ 9sx 9 sx €SES CCAGTTGCTTTATCTGTTCA TGTCGTCTCTCCAGCCCCAG

pSQ ixon TX

€TEQ9 IWW 9 €TES GGCCTCTGATTCCTCACTGA GCCACTGACAACCACCCTTA

€TEQ9 sx 9 sx €TES CCTCTGATTCCTACATGATT ACCACCCTTAACCCCTCCTC

pSQ ixon UX

€UEQ9 IWT 9 €UES TGCCACAGGTCTCCCCAAGG AGTGTGCAGGGTGGCAAGTG

€UEQ9 sx 9 sx €UES GCGCAGTGGCCTCATCTTGG TGTGCAGGGTGGCAAGTGGC

pSQ ixon VX

€VEQ9 PPS 9 €VES CCTTACTGCCTCTTGCTTCT ATAACTGCACCCTTGGTCTC

€VEQ9 sx 9 sx €VES CCTCTTGCTTCTCTTTTCCT TCTCCTCCACCGCTTCTTGT

pIT wut—tion ƒ™reeningB

pIT ixon PeX

PVT‚ PHR ˆPFTPp AGCTTCCTTTCCGTCATGC GCAGCACCACCAGCGTG

pIT ixon PfX

QRT‚ IRU PHHp AGCCCAACTGCGCCGAC CCAGGTCCACGGGCAGA

pIT ixon PgX

ˆPFRP‚ IVW QHSp TGGACGTGCGCGATGC GGAAGCTCTCAGGGTACAAATTC

r€† exh ˜Eglo˜in €rimersBB

r€† €rimersX

w‰HW RSH w‰II GCMCAGGGWCATAAYAATGG CGTCCMAARGGAWACTGATC

˜Eglo˜in €rimersX

€gHR PTV qrPH GAAGAGCCAAGGACAGGTAC CCACTTCATCCACGTTCACC

‚ƒRP SQT uwPW GGTTGGCCAATCTACTCCCAGG GTCCACTCAGTGTGGCAAG

sxY intern—l primers used for ƒƒg€D —symmetri™ €g‚ —nd sequen™ingF ƒD ƒense primer eD entisense primerF Bprimer sequen™es fromX russussi—n

et —lF @IWWRAF BBwaeCgD ‡aeC„D ‰agC„D ‚aeCq

target and reference gene as well as the oligonucleotide lot the smaller of the two reference fragments of the pair is used. PCR cycle conditions were 34 cycles of 948Cfor30s, loaded ®rst and run into the gel for 10 min. Next, the 608C for 30 s and 728C for 30 s and a ®nal extension of product with the larger reference is loaded on top of the 728C for 10 min. PCR products were separated in 6% smaller. The gel is run for 60 min, then stained with polyacrylamide minigels at 100 volts for 70 min, then were ethidium bromide and photographed. A deletion was stained with ethidium bromide and photographed. considered to be present if both sets of reference genes independently showed a reduction in the intensity of the p16 band as determined by the consensus of three p16 deletion independent reviewers. The method of layered di€erential PCR was used to To examine the potential problem of failing to detect a identify tumors with deletion of the p16 gene. Layered homozygous deletion because of the contamination of tumor PCR involves performing two separate binary PCR samples by normal cells we conducted an additional mixing reactions with each binary reaction containing the p16 experiment. Cells with a known homozygous deletion of p16 target primers and one reference primer set. Deletions were (K562) were mixed in varying proportions (0%, 20%, 40%, evaluated using each of two pairs of reference genes (Table 60% and 80%) with untransformed human foreskin 2). One pair consisted of a 127 bp fragment of PGR ®broblast cells. The cell mixture was then paran- (Progesterone receptor gene) and an 85 bp fragment of d- embedded and DNA was extracted from sections as IFN as references for the 101 bp p16 fragment. The second described above. In addition, gels were photographed and pair consisted of a 125 bp fragment of GAPDH scanned by a laser densitometer. The intensity of the p16 (glyceraldehyde phosphate dehydrogenase gene) and an band and the reference gene band was determained and an 85 bp fragment of CPT1 (carnitine palmitoyl-transferase) intensity ratio (P16/reference) derived. Comparison of the as references for the p16 target. To minimize variability, p16/CPT ratios and estimated percentage of normal cells for the reactions for both reference genes in the pair are run the representative tumors in Figure 1a with the cell mixture simultaneously and in the same PCR machine, using the experiment data in Figure 1b shows that the qualitative same reagents. PCR products were separated in 6% threshold we used to classify samples corresponds quite well polyacrylamide minigels at 100 volts. The product with with the ratio expected with approximately 60% normal cell p16 gene alterations and head and neck cancer AFOlshanet al 817 contamination. For example, lanes 4 and 5, which were 6, 11, 16, 18, 31, 33, 35, 39 and 45). Blots were hybridized classi®ed as deletion positive, had band intensity ratios (p16/ as described and detected by binding with streptavidin- CPT) of 0.64 and 0.63, respectively. A ratio of 0.69 was horseradish peroxidase (Vector Laboratories), followed by obtained in the mixing experiment for the sample containing incubation with ECL chemiluminescent reagents (Amer- cells with a known homozygous deletion and 60% sham Corp.). Blots were exposed to X-ray ®lm (Kodak contaminating normal cells. Similar results were obtained XAR-5) for 5 min and 2 h. with ratios based on GAPDH, PGR, and d-IFN reference genes. Statistical analysis To evaluate the relationship between p16 alterations and HPV methods somatic mutations in p53, ampli®cation of PRAD-1, or the Ten micron tissue sections were cut for the analysis of HPV presence of HPV, we calculated standard chi-square DNAsequencesbyPCR.Asectionofappendixtissueswas statistics for contingency tables employing exact testing cut at the beginning, end and after every ®ve specimens to methods. serve as a negative processing control. Tissues were extracted and digested as previously described (Wright and Manos, 1990). Co-ampli®cations of a 268 bp b-globin product and a 450 bp HPV L1 gene product were performed (Bauer et al., 1992). Separate ampli®cation reactions for a 536 bp b-globin fragment were also performed and served as a measure of DNA integrity and Acknowledgements sample suciency (Table 2). DNA extracted from the SiHa The authors thank Dr Harold Pillsbury and the surgeons cervical carcinoma cell line served as a positive control and and sta€ of the Division of Otolaryngology/Head and a negative control without DNA was included. Amplifica- Neck Surgery for their support. We also want to thank tion products were initially analysed on 1% agarose gels, Jeanette Wedsworth for her assistance with tissue procure- then spotted onto Biodyne B nylon membranes for dot-blot ment and Dr Joseph Geradts for the Rb immunostaining hybridization. Replicate blots were probed with biotin- data. This work was supported by a grant from the labeled probes; a b-globin oligonucleotide; a generic HPV Lineberger Comprehensive Cancer Center, University of probe; and HPV type-speci®c oligonucleotides (HPV types North Carolina.

References

Aagaard L, Likas J, Bartkova J, Kjerul€ AA, Strauss M and Kamb A, Gruis N, Feldhaus J, Liu Q, Harshman K, Bartek J. (1995). Int. J. Cancer, 61, 115 ± 120. Tavtigian S, Stockert E, Day R, Johnson B and Skolnick Bauer HM, Greer CE and Manos MM. (1992). Diagnostic M. (1994). Science (Washington DC), 264, 436 ± 439. Molecular Pathology, Volume II. CS Herrington and JO'D Kelley MJ, Otterson GA, Kaye FJ, Popescu NC, Johnson BE McGee. (eds). Oxford University Press: Oxford UK, pp and Dipaolo JA. (1995). Int. J. Cancer, 63, 226 ± 230. 131 ± 152. Koh J, Enders GH, Dynlacht BD and Harlow E. (1995). Brachman DG. (1994). Seminars in Oncology 21, 320 ± 329. Nature 375, 506 ± 510. Cairns P, Mao L, Merlo A, Lee D, Schwab D, Eby Y, Tokino Liu ET, He M and Rajgopal U. (1993). Semin. Cancer Biol., K, Van der Riet P, Blaugrund J and Sidransky D. (1994). 4, 47 ± 58. Science (Washington DC), 165, 415 ± 416. Lukas J, Aagaard L, Strauss M and Bartek J. (1995a). Cairns P, Polascik TJ, Eby Y, Tokino K, Califano J, Merlo Cancer Res., 55, 4818 ± 4823. A, Mao L, Herath J, Jenkins R, Westra W, Rutter JL, Lukas J, Parry D, Aagaard L, Mann DJ, Bartkova J, Strauss Buckler A, Gabrielson E, Tockman M, Cho KR, Hendrick M, Peters G and Bartek J. (1995b) Nature, 375, 503 ± 506. L, Bova GS, Isaacs W, Koch W, Schwab D and Sidransky Lydiatt WM, Murty VVVS, Davidson BJ, Xu L, Dyomina D. (1995). Nat. Genet., 11, 210 ± 212. K, Sacks PG, Schantz SP and Chaganti RSK. (1995). Caldas C, Hahn SA, da Costa LT, Redston MS, Schutte M, Genes, Chromosomes & Cancer 13, 94 ± 98. Seymour AB, Weinstein CL, Hruban RH, Yeo CJ and MerloA,HermanJG,MaoL,LeeDJ,GabrielsonE,Burger Kern SE. (1994). Nat. Genet., 8, 27 ± 32. PC, Baylin SB and Sidransky D. (1995). Nature Medicine, Cowan JM, Beckett MA, Ahmed-Swan S and Weichselbaum 1, 686 ± 692. RR. (1992). J. Natl. Cancer Inst., 84, 793 ± 797. Mori T, Miura K, Aoki T, Nishihira T, Mori S and Grana X and Reddy EP. (1995). Oncogene, 11, 211 ± 219. Nakamura Y. (1994). Cancer Res., 54, 3396 ± 3397. Gruis NA, Weaver-Feldhaus J, Liu Q, Frye C, Eeles R, Munger KB, Werness A, Dyson N, Phelps WC, Harlow E Orlow I, Lacombe L, Ponce-Castaneda V, Lianes P, and Howley PM. (1989). EMBO J., 8, 4099 ± 4105. Latres E, Skilnick M, Cordon-Cardo C and Kamb A. Nawroz H, Van der Riet P, Hruban RH, Koch W, Ruppert (1995). Am. J. Pathology, 146, 1199 ± 1206. JM and Sidransky D. (1994). Cancer Res., 54, 1152 ± 1155. Hannon GJ and Beach D. (1994). Nature (Lond.), 371, 257 ± Nobori T, Miura K, Wu D, Louis A, Takabayashi K and 261. Carson D. (1994). Nature (Lond.) 368: 753 ± 756. Hartwell LH and Kastan MB. (1994). Science, 266, 1821 ± Okamoto A, Demetrick DJ, Spillare EA, Hagiwara K, 1828. Hussain SP, Bennett WP, Forrester K, Gerwin B, Serrano Hussussian CJ, Struewing JP, Golstein AM, Higgins PAT, M, Beach DH and Harris CC. (1994). Proc. Natl. Acad. Ally DS, Sheahan MD, Clark Jr, WH, Tucker MA and Sci., 91, 11045 ± 11049. Dracopoli NC. (1994). Nature Genetics, 8, 15 ± 21. Peter M and Hershkowitz I. (1994). Cell, 79, 181 ± 184. Jen J, Harper JW, Bigner SH, Bigner DD, Papadopoulos N, Quelle DE, Zindy F, Ashmun RA and Sherr CJ. (1995). Cell, Markowitz S, Willson JK, Kinzler KW and Vogelstein B. 83, 993 ± 1000. (1994). Cancer Res., 54, 6353 ± 6358. Ranade K, Hussussian CJ, Sikorski RS, Varmus HE, Jiang W, Zhang YJ, Karn SM, Hollstein MC, Santella RM, Goldstein AM, Tucker MA, Serrano M, Hannon GJ, Lu SH, Harris CC, Montesano R and Weinstein IB. Beach D and Dracopoli NC. (1995). Nature Genetics, 10, (1993). Proc. Natl. Acad. Sci., 90, 9026 ± 9030. 114 ± 116. Kamb A. (1995). Trends in Genetics 11, 136 ± 140. p16 gene alterations and head and neck cancer A F Olshan et al 818 Sche€ner M, Werness BA, Huibregtse JM, Levine AJ and Wright DK and Manos MM. (1990). PCRProtocols:A Howley PM. (1990). Cell, 63, 1129 ± 1136. Guide to Methods and Applications.MMInnis,DH Schmidt EE, Ichimura K, Reifenberger G and Collins VP. Gelfand, JJ Sninsky and TJ White (eds). Academic (1994). Cancer Res., 54, 6321 ± 6324. Press, Inc.: San Diego, pp 153 ± 158. Serrano M, Hannon G and Beach D. (1993). Nature (Lond.), XiongY,KuppuswanyD,LiY,LivanosEM,HixonM, 366, 704 ± 707. White A, Beach D and Tlsty TD. (1996). J. Virol., 70, Serrano M, Lee HW, Chin L, Cordon-Cardo C, Beach D and 999 ± 1008. Dephinho RA. (1996). Cell, 85, 27 ± 37. Yeudall WA, Crawford RY, Ensley JF and Robbins KC. Shapiro GI, Edwaards CD, Kobzik L, et al. (1995). Cancer (1994). Carcinogenesis, 15, 2683 ± 2686. Res., 55, 505 ± 509. Yeudall WA and Jakus J. (1995). Oral Oncol. Eur. J. Cancer, Sherr CJ and Roberts JM (1995). Genes Dev., 9, 1149 ± 1163. 31B, 291 ± 298. Spillare EA, Okamoto A, Hagiwara K, Demetrick DJ, Yoo GH, Xu JH, Brennan JA, Westra W, Hruban RH, Koch Serrano M, Beach D and Harris CC. (1996). Mol. W, Benedict WF and Sidransky D. (1994). Cancer Res., Carcinogenesis 16, 53 ± 90. 54, 4603 ± 4606. Spruck C, Gonzalez-Zuleta M, Shibata A, Simoneau A, Lin Zhang SY, Klein-Szanto AJP, Sauter ER, Shafarenko M, M, Gonzales F, Tsai Y and Jones P. (1994). Nature Mitsunaga S, Nobori T, Carson DA, Ridge JA and (Lond.), 370, 183 ± 194. Goodrow TL. (1994). Cancer Res. 54, 5050 ± 5053. Van der Riet P, Nawroz H, Hruban RH, Corio R, Tokino K, Zhou X, Tarmin L, Yin J, Jiang HY, Suzuki H, Rhyu MG, Koch W and Sidransky D. (1994). Cancer Res., 54, 1156 ± Abraham JM and Meltzer SJ. (1994). Oncogene, 9, 3737 ± 1158. 3741. Weinberg RA. (1995). Cell, 81, 323 ± 330. White A, Livanos EM, and Tisty TD. (1994). Genes Dev., 8, 666 ± 677.