We have identi®ed parameters which de®ne a causal role of HPV16 in head and neck cancer. Twenty-eight tumours which were typed positive for HPV16 DNA, were comprehensively analysed for expression of the viral oncogenes E6 and E7, the status of the p53 gene, and the protein status of pRb and p16 INK4a. In a subset of cases, we have searched for integrated viral DNA, and have determined the genomic status of the E6 gene. Expression of E6/E7 was found in 12 tumours most of which were derived from the oropharynx, whereas p53 mutations were present in 13 tumours from various sites. The tumours either carried p53 mutations but did not express E6/E7, or they did express E6/E7 but were p53-wild-type. Coexistence of E6/E7 expression with a mutated p53 was found in only one case. Strikingly, in most p53-mutated tumours without E6/E7 expression, we found the E6 gene to be disrupted. E6/E7 expression was associated with reduced pRb and overexpressed p16INK4a . Viral-cellular fusion transcripts were found in two cases. Our data demonstrate that HPV16 DNA-positivity in head and neck cancers is not indicative of a causal role. A causal role of HPV16 in head and neck cancer is de®ned by: E6/E7 expression, viral integration with an intact E6 gene, and perturbation of pRb cell cycle control. Mostly, the p53 gene is wild-type.

Recent studies have suggested that di erent mutation types within the core domain of the tumour suppressor protein p53, i.e. DNA contact mutations and structural mutations, confer di erent biological properties. We have analysed in 86 head and neck squamous cell carcinomas (HNSCC), whether these p53 mutation types have a di erential clinical impact. Thirty-seven missense mutations were identi®ed. Thirteen of these (36%) were DNA contact mutations, occurring in the L3 loop, in the H2 loop sheet helix motif, in the S10 b strand and in Zinc binding residues. Microsatellite marker analysis revealed a selective association between these mutations and the loss of wild-type alleles (100% LOH vs 50% LOH in tumours with structural mutations; P=0.0034, Fisher's exact, 2-tailed). In comparison to structural mutations or to the absence of mutations in the core domain, DNA contact mutations were associated with higher tumour stages (84.6% vs 62%), a higher incidence of lymph node metastasis (91.7% vs 56%; P=0.014, Fisher's exact, 2-tailed), a shortened recurrence-free survival (8.1 months vs 23.7 months, P=0.047, log rank test) and overall survival (11 months vs 29.2 months; P=0.003, log rank test). The latter was also the case when only stage IV tumours were analysed (P=0.0055, log rank test). These data indicate that in HNSCC, TP53 DNA contact mutations confer a strong selection pressure to eliminate wild-type alleles, and that they result in an accelerated tumour progression and reduced therapeutic responsiveness.

p21 CIP1/WAF1 is an inhibitor of cyclin-dependent kinases and, in normal tissues including squamous epithelia, has been associated with cell-cycle exit and differentiation. As shown in this pilot study, however, the majority of head-and-neck squamous-cell carcinomas (HNSCC) display aberrant p21 CIP1/WAF1 expression: of 42 tumors analyzed by immunohistochemical staining, 28 (67%) over-expressed the p21 CIP1/WAF1 protein. Accumulation of p21 CIP1/WAF1 was independent of the histological grade of the tumors as well as the genetic status of the p53 gene. In many cases, most notably in poorly differentiated or undifferentiated HNSCC, p21 CIP1/WAF1 -positive cells were actively proliferating tumor cells, since they also expressed proliferating-cell nuclear antigen (PCNA) and Ki-67. Accumulation of p21 CIP1/WAF1 occurred through a posttranscriptional mechanism since, in contrast to immunohistochemical analysis of the p21 CIP1/WAF1 protein, in situ hybridization showed no increase of mRNA levels as compared with cells in normal mucosa (n 5 25). Clinically, among the patients with p21 CIP1/WAF1 -over-expressing tumors, there was increased recurring disease (p 5 0.03; x 2 -test), shortened disease-free survival (p 5 0.0019; log-rank test) and shortened overall survival (p 5 0.0071; log-rank test). These in vivo data indicate that in many HNSCC, accumulated p21 CIP1/WAF1 is compatible with increased tumor-cell proliferation, and they provide preliminary evidence that p21 CIP1/WAF1 may be of prognostic and predictive significance. Int. J. Cancer 74:383-389, 1997.r 1997 Wiley-Liss, Inc. p21 CIP1/WAF1 has been shown to inhibit the activity of several cyclin/cyclin-dependent kinase complexes and to block cell-cycle progression (Harper et al., 1993;Xiong et al., 1993;Gu et al., 1993). It was identified as a gene whose product was transcriptionally activated by wild-type but not mutant p53 protein, serving as a mediator of the cell-cycle-arrest function of p53 (El-Deiry et al., 1993). Simultaneously, it was found to be associated with senescence (Noda et al., 1994) and terminal differentiation (Jiang et al., 1994;Steinman et al., 1994; Halevy et al., 1995). These properties made the p21 CIP1/WAF1 gene a prominent candidate tumorsuppressor gene. However, extensive searches failed to reveal genetic alterations of the gene in human tumors, except for the rare occurrence of polymorphisms (Shiohara et al., 1994).The precise role of p21 CIP1/WAF1 in the control of proliferation and differentiation is not clear. During mouse embryonic development, p21 CIP1/WAF1 mRNA expression was found in the post-mitotic cells of many tissues, and this expression occurred independently of p53. This suggested that p21 CIP1/WAF1 provides a tight link between cell-cycle arrest and exit and differentiation (Parker et al., 1995). However, knockout mice lacking a functional p21 CIP1/WAF1 developed fairly normally, indicating that the presumed function of p21 CIP1/WAF1 in differentiation was redundant (Deng et al., 1995). In addition, in contrast to p53 kno...