Mucosotropic high-risk human papillomaviruses (HPVs), known to cause cervical and other anogenital cancers, have been proposed to play a role in the etiology of head and neck squamous cell carcinomas (HNSCCs).1 The presence of high-risk HPV DNA in a subgroup of HNSCCs has supported this hypothesis.2–5 Molecular studies have provided important data on the role and oncogenic mechanism of high-risk HPV in carcinogenesis.6–8 By expression of the viral oncoproteins E6 and E7, the virus dysregulates crucial cellular mechanisms such as the cell cycle and the apoptotic pathway. The E6 oncoprotein specifically inactivates wild-type p53, and the E7 oncoprotein inactivates Rb. In this way the high-risk HPV E6-mediated degradation of the p53 protein should be considered an alternative pathway for “classical” mutation to knock-out the p53 regulated pathways; it provides the biological basis to expect that tumors originating from HPV infection will show wild-type p53. Indeed, this general biological mechanism is supported by the finding that p53 mutations hardly occur in cervical carcinomas.9, 10 However, in most studies on head and neck cancer, HPV DNA presence and p53 mutations were overlapping,2, 11, 12 an observation that gave rise to a long debate as to whether HPV is causally related to the development of a subset of these tumors.
On the other hand, the inconsistent observations might be explained by the methods and criteria that were used for HPV assessment and mutational analysis. HPV DNA detection by PCR is extremely sensitive, up to a level of a few DNA copies13 and might lead to the detection of a few viral genomes that may not be clonally associated with the tumor. Moreover, the discrepancies in the data might be caused by the source of the tissue material, i.e., purified DNA or crude extracts of cryosections or paraffin sections. Furthermore, p53 mutation frequencies were not determined in all cases by sequencing but were often based on immunohistochemical or single-strand conformation polymorphism (SSCP) analysis. Finally, not all sequencing methods are equally reliable in detecting mutations in tumor DNA.14
In the present study, we investigated HPV involvement in HNSCC, making use of the known biological properties of the virus in cervical carcinogenesis. We hypothesized that if the virus plays an important role in the genesis and progression of HNSCC: then 1) HPV DNA should be present in the tumor, 2) the E6 viral oncogene should be expressed in the tumor, 3) the p53 gene should be wild-type and 4) HPV DNA and E6 mRNA should be present in corresponding lymph node metastases.
MATERIAL AND METHODS
Patients and tumor specimens
In total 84 patients who underwent surgical treatment for squamous cell carcinoma of the upper aerodigestive tract were included. The study was approved by the Institutional Review Board of the Vrije Universiteit Medical Center, and informed consent was obtained from all patients. From each patient a fresh primary tumor sample was obtained, and, if present, a sample of each macroscopic lymph node metastasis. The tumor sample was directly snap-frozen into liquid nitrogen and stored at −80°C until further processing. Selection criteria used to include patients were tumor site (mainly oropharynx, oral cavity and hypopharynx) and size. The distribution of the tumors by anatomical site was as follows: 45 tumors were located in the oral cavity, 30 in the oropharynx, of which 18 were assigned as tonsil, 4 in the hypopharynx and 5 in the larynx. The age of the patients ranged from 40 to 77 years, with a mean of 58 years. In total, 55 patients were male, and 29 were female.
Routine hematoxylin-and-eosin staining was performed on 10 μm cryosections to confirm the presence of squamous cell carcinoma in the specimen sampled from the primary tumor and to guide microdissection. Neoplastic areas were microdissected and tumor DNA isolated. A 1.8 kb fragment of the p53 gene, encompassing exons 5 to 9 was amplified by PCR as described by Sidransky et al.15 Purified PCR products were sequenced directly by exon-specific primers using the radioactive dideoxynucleotide method (AmpliCycle Sequencing Kit, Perkin Elmer, Norwalk, CT),15 a very reliable sequencing method to detect p53 mutations.14 For a subset of tumors (20) without a mutation in exons 5 to 9, the remaining exons 2, 3, 4, 10 and 11 were sequenced in addition. Primer sequences and reaction conditions are available on request. To check the reliability of the sequencing method, a plaque assay15 was performed on 35/44 tumors. In all cases the sequenced mutation corresponded to the results with the plaque assay.
High-risk HPV DNA detection and typing
Detection of high-risk HPV DNA in HNSCC samples from all 84 patients was performed by general primer GP5+/GP6+-mediated PCR enzyme immunoassay (PCR-EIA), essentially as described previously.13 This method allows the group-specific detection of 14 high-risk HPV genotypes by hybridization of HPV GP5+/6+ PCR products in an EIA format with an oligonucleotide-probe cocktail specific for HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68. GP5+/6+ PCR products of positive cases (using a cutoff of 3× background) were subsequently hybridized with the type-specific probes individually as described elsewhere.16 The HPV test was only considered positive when both the EIA OD value reached at least 3× background and a fragment was visible after Southern blot hybridization. All L1 negative tumor DNA was subsequently tested for HPV16 E7 amplification using type-specific primers, again in an EIA format. High-quality purified DNA from microdissected tumor was used as starting material in all assays. All positive cases were confirmed on crude extracts of cryosections by the same assays. As a control of the integrity of the target DNA, all samples were subjected to a p53 gene or β-globin gene PCR. To prevent crossing-over contamination, all samples were handled carefully and cut on different blades; DNA isolations as well as pre-PCR pipetting were performed in laboratories separate from post-PCR processing. In a number of cases negative tissues were cut between positive samples to check for crossing-over contamination in the cryotome. In all experiments a serial dilution of SiHa DNA was added as well as a number of non-template controls to check both the sensitivity and specificity of the assays adequately. In all experiments the controls showed the expected results.
HPV16 E6 RT-PCR
To assess the integrity of the RNA to be used for HPV16 transcript analysis, RNA samples were first pre-screened by RT-PCR specific for the U1 small nuclear ribonucleoprotein specific A protein (snRNP U1A), as described by Snijders et al.17 All samples showed the presence of snRNP U1A mRNA and were subsequently subjected to RT-PCR specific for the HPV16 E6 region. RT-PCR was performed as described previously,17 except that HPV16 E6-specific primers were used spanning nucleotides 204 to 525 of the HPV16 genome. These primers allow the detection of both full-length E6 transcripts and spliced E6*I mRNA, the latter being the major E7-encoding mRNA species. Moreover, RNAs were preincubated with 0.5 U RQ1 DNase (Promega, Leiden, The Netherlands) in RT reaction mix for 15 min at 37°C prior to reverse transcription to remove traces of genomic DNA that might be present. Reactions without RT were included as negative controls during cDNA synthesis. RT-PCR products were run on 1.5% agarose gels and blotted onto nylon membranes (GeneScreen Plus, NEN, Hoofddorp, The Netherlands). Hybridization was performed using an oligonucleotide-probe (nucleotide position 386 to 415) that specifically detects full-length E6 but not E6*I mRNA.
In total 84 HNSCC cases were examined for the presence of HPV DNA by L1 GP5+/6+ general primer-mediated PCR-EIA13 on purified DNA from microdissected tumor, the most appropriate template. In total 12/84 specimens (14%) were positive by this assay, all typed as HPV16. Subsequently, all positive cases were reanalyzed on crude extracts of cryosections. Only 10/12 cases could be confirmed using these crude extracts as template. Because single integration events involving the L1 region might reveal false-negative results with this primer set, all negative cases were re-tested by PCR-EIA amplification of the HPV16 E7 region, again using purified DNA as template. Eight additional tumors scored positive for HPV DNA, but none could be confirmed using crude extracts of cryosections as template. Revision of the L1 PCR-EIA data indicated that a number of the E7-positive cases also showed increased L1 values, but not on a level to exceed the threshold for a positive test (3× background).13 In summary, in 20/84 cases one or more HPV DNA assays were positive, but only 10/20 showed consistent HPV DNA-positivity with all assays and templates used.
Subsequently, all carcinomas that were HPV DNA positive (20/84) by either of these assays were examined for the presence of HPV16 E6 transcripts. In only 9/20 cases could we demonstrate expression of E6 mRNA (Table I). These nine cases were consistently positive with all the DNA PCR assays. All carcinomas showing HPV E6 mRNA expression were located either in the oropharynx (7/9, 4/7 being tonsillar carcinomas) or in the oral cavity (2/9) (Table II). It has been suggested previously that HPV-positive HNSCC tumors have a basaloid morphology.18 All 84 cases were revised. After histopathological review, indeed 6/9 HPV E6-positive cases showed a basaloid morphology. In contrast, only 1/75 HPV E6-negative cases showed a basaloid morphology, a highly significant observation (p < 0.0001, Fisher's exact test).
Table I. Results of HPV DNA/RNA Assays and p53 Sequencing
Used DNA assays: a) GP5+/6+ PCR-enzyme immunoassay (EIA) and subsequent Southern blot analysis of amplimers, both with purified DNA of microdissected tumor samples and crude extracts of cryosections as template; b) HPV16 E7 PCR-EIA and subsequent Southern blot analysis of amplimers with both purified DNA of microdissected tumor samples and crude extracts of cryosections as template. +, consistently positive in all DNA assays; +/−, positive in one or more DNA assays, but not consistently in all DNA assays; −, negative in all assays with purified DNA as template. HPV, human papillomavirus.–*p-values were calculated by Fisher's exact test using the mutation frequency in the various groups against the frequency in the HPV DNA-negative group. A p-value < 0.05 was considered significant.
Table II. Distribution of HPV Status as Assessed by E6 RT-PCR and p53 Mutation by Primary Tumor Site
The tumors of these two patients were located in the anterior floor of the mouth and the lateral floor of the mouth/gingiva, respectively.
To establish the role of the virus in tumor progression further, we analyzed the DNA and RNA of seven lymph node metastases (LNMs) from six patients with HPV16 E6 RT-PCR positive carcinomas. In all six cases the LNMs were shown to contain both HPV16 DNA as well as E6 transcripts. In Figure 1, representative results of HPV DNA and RNA assays are presented for two cases. Case 98-39 was unequivocally positive for HPV DNA by all assays but negative for E6 transcripts, and case 98-8 was unequivocally positive for HPV DNA and positive for E6 mRNA.
Subsequently all 84 tumor samples were analyzed for the presence of p53 mutations. Sequencing of exons 5 to 9 revealed a mutation in 41 samples (49%). Additional sequencing of exons 2, 3, 4, 10 and 11 in the tumors without a mutation in exons 5 to 9 (20 cases) revealed only three mutations: one in exon 3 (oral cavity) and two in exon 4 (oropharynx and oral cavity), increasing the total mutation frequency to 52% (44/84) (Table II). Interestingly, all HPV E6 mRNA-positive cases lacked a p53 mutation, a highly significant observation (χ2 Fisher's exact test, p < 0.001). Strikingly, 4/11 tumors that were positive for HPV DNA-PCR but negative for E6 RT-PCR showed a p53 mutation (Table I), a frequency not significantly different from the HPV DNA-PCR negative group.
Subsequently, tumor and patient data were compared between HPV-E6-positive tumors lacking a p53 mutation and HPV-E6-negative tumors displaying p53 mutations. Parameters analyzed were smoking history, sex, age, differentiation grade of the tumors and survival. We could not demonstrate any correlation of HPV status and these variables. Kaplan Meier analysis failed to show any difference in survival between the two groups (p = 0.82). It should, however, be noted that the number of cases is small.
In this study we have clearly demonstrated an absolutely inverse correlation of HPV E6 mRNA expression and mutations in the p53 gene. The presence of p53 mutations in HPV DNA-positive HNSCC tumors has long overshadowed a clear etiological role of the virus in a subset of these tumors, although it cannot be exluded that the virus more often plays a role in the initial phase of carcinogenesis. The percentage of HPV positivity as assessed by E6 mRNA expression is, however, relatively low in comparison with many other studies.18, 19 This discrepancy might be explained by the fact that previous studies mainly considered the presence of HPV by DNA PCR only. The number of tumors containing solely HPV DNA as detected by PCR exceeds that of carcinomas that also show expression of HPV E6 mRNA and are likely to be clonally related to HPV.4 Importantly, we have also shown that viral DNA and E6 mRNA is maintained in LNMs of the HPV E6 RT-PCR positive cases, providing strong evidence for a key role of the virus in these particular tumors.
Our data suggest that the role of HPV in HNSCC can easily be overestimated when using DNA assays only (20/84 DNA-positive versus 9/84 E6 mRNA-positive cases). The sensitivity of the HPV-DNA PCR is often so high that various types of misinterpretations might be the result of technical artifacts (false-positive findings) or positive findings far below the minimal level of 1 HPV genome copy per tumor cell.20 We thereby assume that when HPV indeed plays a role in tumor formation and maintenance, the presence of virus should at least be 1 copy per tumor cell. This assumption is consistent with the findings of Snijders et al.,17 who showed that only carcinomas containing relatively high copy numbers of viral DNA express the E6 mRNA. Possibly quantitative or semi-quantitative PCR assays might allow discrimination of E6 mRNA-positive versus E6 mRNA-negative cases. This is supported by the fact that the HPV PCR-EIA OD values read in this study after overnight incubation were markedly higher in the HPV DNA-positive samples with E6 mRNA expression than in those samples without detectable E6 mRNA (Table I).
In the recent study of Gillison et al.,18 evidence was presented for a causal association between HPV infection and HNSCC. Although the technical procedures appeared to be highly consistent and solid, their data again did not establish a clear biological link between HPV and HNSCC. Various HPV-positive tumors appeared to have mutations in the p53 gene, and the overall mutation frequencies were not significantly different from those of the HPV-negative group. Moreover, it was shown that in 43% of the cases the presence of HPV DNA as assessed by PCR could not be confirmed by Southern blotting, an alternative and less sensitive technique. Possibly these authors also detected a number of putatively “biologically irrelevant” cases due the sensitivity of their DNA-PCR assays. Expression of the E6 oncogene, which confirms the etiological role of the virus in these tumors, might well be used to check for possibly confounding findings of HPV DNA-PCR assays.
Still, we cannot exclude the possibility that in some E6 mRNA-negative cases the virus played a role in the initial phase of tumorigenesis according to a “hit and run” principle. This may imply that HPV-mediated degradation of p53 is substituted by mutational events during malignant progression. As an example, in one case we could confirm HPV DNA presence with all the different assays on all material, but we failed to demonstrate E6 mRNA expression. This particular case also showed a p53 mutation. Since this patient did not show LNMs, we were not able to investigate the putative association in more detail. Besides this case, there were a number of other E6 mRNA-negative cases that were HPV DNA positive. However, in these cases the various DNA analyses were not unequivocally positive, which might have resulted from the sensitivity of particular primer sets and PCR strategies. Transient infection far below the level of 1 copy per cell might explain these observations.
The implications of our findings are that the large expectations with respect to vaccination programs for the treatment of HPV-induced head and neck cancers, as suggested by McNeil et al.,21 should be interpreted with caution. Particularly in HPV DNA-positive cases in which E6 expression is absent, the etiological role of the virus is questionable, and the rationale for immunotherapy in our view is lacking, although in these cases of DNA positivity only, vaccination might still play a prophylactic role. For immunotherapy trials, it seems a more appropriate approach to select patients by E6 RT-PCR or semiquantitative DNA PCR assays at a level of at least 1 copy per cell. In those cases viral factors are apparently indispensable for the malignant state, and vaccination strategies might be successful.
Our study provides the “missing” biological link, which clearly demonstrates that high-risk HPVs are indeed responsible for a number of squamous carcinomas in the head and neck. There is some epidemiological evidence that this viral etiology could be related to sexual behavior.22, 23 Further research in these patient groups, combined with molecular data, should demonstrate whether specific sexual contacts with HPV-positive partners might cause the disease, possibly in concert with other risk factors such as smoking or drinking.
We thank Ing. H. Schrijnemakers (Vrije Universiteit Medical Center, Amsterdam, The Netherlands) for excellent technical assistance.