Prevalence, distribution, and viral load of human papillomavirus 16 DNA in tonsillar carcinomas

Authors


Abstract

BACKGROUND

Oncogenic human papillomaviruses (HPV) DNA have repeatedly been observed in many head and neck carcinomas (HNSCCs), and HPV infections are currently considered a possible factor in the etiology of these tumors. However, the reported prevalences of HPV-DNA in HNSCC are variable. In the current study the authors used highly sensitive polymerase chain reactions (PCRs) to analyze the occurrence of viral sequences in 98 carefully stratified HNSCCs. The authors determined the load and localization of HPV DNA in a subset of tonsillar carcinomas and their metastases.

METHODS

Nested PCR and an HPV16 specific single step PCR were used to screen 98 HNSCCs for HPV DNA for genital- and Epidermodysplasia verruciformis (EV)-associated HPVs. Typing was performed by direct sequencing and/or sequencing of cloned amplimers. In two patients HPV16 subtypes in tonsillar carcinomas and their metastases were compared by amplification and sequencing of the long control region of the virus. In a subset of HPV16 positive tonsillar carcinomas and their metastases, localization and viral load were determined using laser assisted microdissection and real time fluorescent PCR, respectively.

RESULTS

Altogether 25 HNSCCs (26%) were found to be HPV positive. Stratified according to the tumor localization, the frequency of HPV positive lesions was 18% in the oral cavity, 45% for oropharynx, 25% for hypopharynx, 8% for nasopharynx, and 7% for larynx. The highest HPV DNA prevalence (58%) was found in tonsillar carcinomas. The high risk HPV type 16 was found in 84% of positive HNSCCs, in 14% of which EV-associated HPVs were detected. Human papillomavirus sequences were detected in 64% of biopsies with normal mucosa from 11 patients with positive carcinomas. As a control group, 14 tumor free tonsils were analyzed. In none of these specimens were HPV sequences detected. Viral long transcriptional control region sequences in homologous metastases were identical with those in primary tumors and the load values in both locations were roughly comparable. Viral loads differed substantially in different areas of one tumor. Statistical evaluation of data related to clinicopathologic parameters showed a significant linkage of HPV with tonsillar carcinomas compared to other locations. Furthermore, a significant correlation of HPV status of tonsillar carcinomas with tumor grading and alcohol consumption was found.

CONCLUSIONS

Our study shows a preferential association of HPV-DNA with tonsillar carcinomas. The data support the view of HPV negative and positive tonsillar carcinomas being different tumor entities and conventional cancer risk factors being of less importance in HPV-infected individuals. The HPV genome is located in the cancer cells, whereas the infection of normal mucosa is a rare event. Data on quantification of HPV16 in tonsillar tumors and their metastases showed mean viral loads comparable to other HPV associated malignancies. Cancer 2001;92:2875–84. © 2001 American Cancer Society.

Human papillomavirus (HPV) infections induce a broad spectrum of epithelial tumors of skin and mucosa, ranging from benign warts, papillomas, and dysplasias to malignant lesions of the cervix uteri and cutaneous cancers of patients with Epidermodysplasia verruciformis.1, 2 The role of HPV infection in carcinogenesis and the involved molecular mechanisms could thus far best be elucidated in the case of the cervical carcinomas. Overwhelming epidemiologic and molecular evidence verifies the existence of several so-called high-risk HPV types (e.g., HPV16 or 18), which consistently associate with genital malignancies and encode proteins which deregulate the control network of cellular proliferation/differentiation processes.3

Apart from anogenital cancer, a number of other neoplasias are currently being examined for possible etiologic associations with HPV infections. Epidemiologic data suggest that the most probable candidates are non-melanoma skin cancer and some of the head and neck squamous cell carcinomas.4, 5

Head and neck squamous cell carcinomas (HNSCCs) contribute 6% of new cancer cases annually worldwide.6 Well-recognized risk factors for the development of HNSCC are extensive consumption of tobacco and/or alcohol, and betel quid chewing;7 however, a constant proportion of cases have also been observed among non-addicts. During the last decade, a number of studies argued that infections with potentially oncogenic HPVs may present another (independent or complementary) risk factor involved in the etiology of HNSCC (review8). Human papillomavirus sequences have been shown in many of these tumors; however, the reported prevalence values varied drastically between 0 and 100%, apparently biased by the size of the tested collectives, anatomic location of tumors, HPV detection techniques, and the type of tissue material used.8–11 Evaluation of 30 polymerase chain reaction (PCR)-based studies on HNSCC, which included a total of 1205 individual cases, revealed an average of 35% of the examined tumors were positive for HPV DNA.11 Virologic analyses of HPV-positive HNSCCs consistently report almost exclusive occurrence of virus types, known to preferentially infect the genital skin and mucosa. In line with these data is the higher prevalence of anti-HPV16-antibodies in HNSCC patients as compared to control individuals.12

Epidemiologic approaches to HNSCC reveal that HPV infection of the oral cavity is linked to an elevated cancer risk, independent of tobacco and alcohol abuse,13 and that there can be similar risk factors for both HNSCC and cervical cancer.12 The latter findings led to the recently presented hypothesis that some HNSCCs are a sexually transmitted disease.14, 15

In this study we report on prevalence of HPV DNA in a series of carefully stratified HNSCCs using highly sensitive PCR protocols, capable of detecting almost all presently known papillomavirus types. To test for HPV DNA association in either neoplastic or normal epithelial cells, the laser assisted microdissection was used in selected cases, assisted by the quantification of viral load in primary and metastatic tumor.

MATERIALS AND METHODS

Subjects and Material

Only patients with newly diagnosed and histologically confirmed HNSCC were enrolled in the current study. All patients gave their written consent to the Department of Otolaryngology, Head and Neck Surgery of the University of Cologne. Tumor specimens and normal mucosa from sites distant from the tumor (at least 2 cm) were obtained during surgery or diagnostic panendoscopy and stored at −80 °C until further processing. Clinical data, tumor staging, and primary treatment were obtained from medical, radiology, operative, and pathology reports using the 1997 American Joint Committee on Cancer staging criteria.16 Tumor grading was revised for all tonsillar carcinomas by one pathologist. Histologic grading was performed in a blinded manner (V.D.) according to the World Health Organization criteria for squamous cell carcinomas of the oral mucosa.17 All tonsillar carcinoma patients were examined by the head and neck surgeon at each follow-up appointment (every 3 months).

After confirming the integrity of the DNA extracted from tissue samples18 by β-globin gene PCR, the studied population consisted of 98 patients. The patient ages ranged from 26 to 81 years (median 59 years). Seventy-seven (79%) patients were males and 21 (21%) were females. Distribution of the tumor localization was as follows: 22 tumors (23%) in the oral cavity, 33 tumors (34%) in the oropharynx, 13 tumors (13.4%) in the nasopharynx, 16 tumors (16.4%) in the hypopharynx, and 14 tumors (14.4%) in the larynx. The tumors of the oral cavity and oropharynx were assigned to further subsites as shown in Table 1. From four patients with tonsillar carcinomas paraffin blocks of the opposite tonsil were available; however, only two contained adequate DNA for PCR analysis.

Table 1. Occurrence of HPV-DNA in HNSCCs of Different Anatomical Localizations (n = 98)
Tumor subsitenHPV DNA positiveaHPV types/sequences detected
  • HPV: human papillomavirus.

  • a

    Percent numbers correspond to the prevalence in the individual subsites analyzed and among all HPV-positive tumors, respectively.

Oral cavity
Tongue12 (12%)2 (17%; 8%)16, 19
Floor of the mouth4 (4%)1 (25%; 4%)16
Other6 (6%)1 (17%; 4%)16
Total22 (23%)4 (18%; 16%)
Oropharynx
Tonsil24 (25%)14 (58%; 56%)16, 5b, 33, ADX1
Base of the tongue6 (6%)1 (17%; 4%)16
Other3 (3%)0
Total33 (34%)15 (45%; 60%)
Nasopharynx13 (13%)1 (8%; 4%)IA09
Hypopharynx16 (16%)4 (25%; 16%)16, 19
Larynx14 (14%)1 (7%; 4%)16
Total98 (100%)25 (26%)

Control specimens of tonsillar tissue were taken from eight patients who underwent tonsillectomy due to chronic tonsillitis and from six patients who underwent tonsillectomy due to primary tumor search (patients with a carcinoma of unknown primary). The patient ages in the control group ranged from 20 to 65 years (median 52 years). Ten (71%) patients were male and four (29%) were female.

Sample Preparation, Polymerase Chain Reaction, and HPV Typing

Tissues were processed with the QIAamp Tissue Kit (Qiagen, Hilden, Germany), and 10 μL of purified total cellular DNA were employed in each PCR reaction. Negative controls (water or human placental DNA) were included in each PCR run. Human papillomavirus sequences were detected by highly sensitive nested PCR protocols with degenerate primers A10/A5-A6/A8 for HPV group A (genital/mucosal) HPV19 and CP62/70-CP65/69a for group B1 (cutaneous/EV) HPV20. Human papillomavirus 16 specific PCR was performed as described by van den Brule et al.21 The HPV 16 long control region (LCR, nucleotides [nt] 7009-129) was amplified according to Dong et al.22 Polymerase chain reaction products (10 μL) were separated in 2% agarose gels and visualized by ethidium bromide staining.

Human papillomavirus typing was performed by sequencing of PCR products and comparison of the obtained sequences with an HPV database.23 Direct sequence analysis of purified PCR products (QIAquick PCR purification kit, Qiagen) was carried out with an ABI Prism 377 DNA sequencer using the Taq FS BigDye-Terminator cycle sequencing method (PE Applied Biosystems, Weiterstadt, Germany). Additionally, A6/A8 and CP65/CP69a PCR products (270 base pairs and 422 base pairs, respectively) were cloned into the vector pCR-Blunt II-Topo using the ZeroBlunt-Topo-PCR-Cloning Kit (Invitrogen, Leek, Netherlands). Clones that carried an EcoRI insert of the expected size were sequenced as mentioned above. Human papillomavirus 16-LCR was directly sequenced from both sides as described by Dong et al.22

Laser Assisted Microdissection

Histologically defined tissue areas of tumors, normal mucosa, lymph node metastases, and connective tissue from patients with tonsillar carcinoma were microdissected and ejected with laser pressure directly into the cap of the sample tubes in an entirely “non-contact” manner (Fig. 1).23

Figure 1.

Laser-assisted microdissection of carcinoma (A-C) and normal epithelium tissue (D-F). After cutting the area of interest with a focused laser beam (A, D), the specimens were ejected from the slides (B, E) and catapulted directly into the microfuge cap (C, F).

Sections (5 μm) of formalin-fixed and paraffin-embedded tissue samples from patients with HPV positive tonsillar carcinoma were mounted on glass slides covered with a 1.35 μm thin polyethylene membrane (PALM, Wolfratshausen, Germany) coated with 1% poly-L-lysin. Slides were deparaffinized and stained with hematoxylin. The Robot-MicroBeam system (PALM, Wolfratshausen, Germany) was used, consisting of a pulsed, low-energy 337 nm nitrogen laser, an inverted microscope (Axiovert 135, Carl Zeiss, Oberkochen, Germany), and a motorized, computer-controlled micromanipulator.24 Microdissection was performed under microscopic control, and the tissue was then catapulted into the cap (Fig. 1). Successful catapulting was microscopically controlled. Samples were incubated for at least 3 hours at 55 °C in 15 μL PCR buffer (Life Technologies, Karlsruhe, Germany) containing proteinase K (400 μg/mL). Proteinase K activity was inactivated at 95 °C for 10 minutes. DNAs from microdissected specimens were analyzed either with conventional nested HPV PCR or quantitative HPV PCR as described below.

Viral Load Determination by Real Time Quantitative PCR

Viral load of HPV16 in biopsies was determined by real time fluorescence PCR with the LightCycler System (Roche Molecular Biochemicals, Mannheim, Germany) and expressed as HPV16 DNA copies per β-globin-gene copy. Gene copy numbers of β-globin were determined using the LightCycler-Control Kit DNA (Roche Molecular Biochemicals) in 20 μL reactions containing 4.0 mM MgCl2, 20 mM TrisHCl (pH 8.4), 50 mM KCl, 500 μg/μL bovine serum albumin (Sigma, Deisenhofen, Germany), and 0.5% Tween 20 (Sigma), 200μM dNTP each, 0.8 U Platinum Taq DNA polymerase (Life Technologies, Karlsruhe, Germany), β-globin primers (0.5 μM), and fluorescein- (0.2 μM)- and LC Red 640 (0.4 μM)-labeled probes according to the manufacturer′s instructions. Quantification of HPV16 and β-globin sequences in DNA extractions from whole samples was done in duplicate, and a mean value was calculated. Quantification of HPV16-DNA and β-globin-gene in microdissected material could only be done once. Amplification of the PCR-products was monitored using type-specific hybridization probes labeled with fluorescein and LightCycler (LC) Red 640 (TIB MolBiol, Berlin, Germany), respectively.25 Calculation of initial copy numbers in samples was performed by the LightCycler software (Version 3.1.102) using a standard curve generated with exactly quantified HPV DNA standards (ten fold dilution series of full length HPV16 plasmid26 that were amplified in the same PCR run). Input copy numbers of the HPV standards were estimated by ethidium bromide fluorescent quantification.27 Ten copies of HPV16 standard DNA were regularly detectable, and five copies were frequently detectable by real time PCR. The slope of the generated standard curves indicated amplification efficiencies from 0.89 to 0.97 (efficiency of 1.00 = 100% turnover). Negative controls (human placental DNA) were included in duplicate in each run and never yielded fluorescence signals above the background. Hot start amplification of HPV16 DNA was performed in LightCycler capillaries in 20μL containing 4.25 mM MgCl2, 20 mM TrisHCl (pH 8.4), 50 mM KCl, 500 μg/μL bovine serum albumin (Sigma, Deisenhofen, Germany), 0.5% Tween 20 (Sigma), 200μM dNTP each, and 0.8 U Platinum Taq DNA polymerase (Life Technologies, Karlsruhe, Germany). Cycling conditions were: 60 seconds at 95 °C, followed by 45 cycles of 1 second at 95 °C, 10 seconds at 56 °C, and 8 seconds at 72 °C. Type-specific primers and probes (So) for HPV16 quantification were: A6-16.3 5′GCACAGGGCCACAATAAT 3′ (HPV16 nt 6584-6601; 0.25μM), A8-16.3 5′GACCAAAATTCCAGTCCTC 3′ (HPV16 nt 6854-6836; 0.25μM), So16-1 5′TCAACTGTGCAAAATAACCTTAACTGC-fluorescein 3′ (HPV16 nt 6766-6792; 0.15μM), So16-2 5′LC Red 640-ACGTTATGACATACATACATTCTATGAATTCCACT-ph3′ (HPV16 nt 6795-6829; 0.15μM).

Sequence Analysis

Sequence analysis was performed using the BLAST 2.028 and MacVector 7.0 (Oxford Molecular Group PLC, U.K.) program packages. GenBank, EMBL, DDJB, and PDB served as sequence databases.

Statistical Analysis

Factors associated with HPV status and HPV load were selected on cross-tabulations, which were analyzed by the use of the chi-square test or t-test using the SPSS Base System, version 6.1 (SPSS, Chicago, IL). An initial significance level of α- 0.05 was chosen.

RESULTS

Occurrence of HPV-Sequences in HNSCC Biopsies

DNA preparations from biopsies of 98 HNSCC-patients were examined for the presence of HPV sequences by means of the PCR analyses employing both group- and type-specific primers. The analyzed tumors represented carcinomas of the oral cavity, oropharynx, nasopharynx, hypopharynx, and larynx. As shown in Table 1, only 26% of tested samples were found to be HPV-positive. However, significant differences in HPV prevalence could be observed depending on the anatomic location of the HPV-positive tumors. Generally, HPV sequences were most frequently detected in the oropharyngeal lesions (45%), especially among tumors arising from the palatine tonsils (58%). In contrast, other tested HNSCCs showed comparatively low HPV infection rates, ranging from 7 to 25%.

HPV Types Infecting HNSCC

To identify HPV types present in the analyzed tumors, PCR-amplified viral DNA-fragments of the open reading frame (ORF) L1 were subjected to sequence analysis. Eighty-four % of tested DNA samples (21 out of 25 HPV-positive biopsies) carried sequences of HPV type 16 (Tables 1 and 2). Examination of amplified sequences revealed 99-100% homology with the HPV16 prototype described by Seedorf et al.26 The only other genital HPV type identified was HPV33, detected in a biopsy of tonsillar carcinoma. Five of the analyzed tumors contained sequences of typical cutaneous, EV-associated papillomaviruses: HPV19 (one hypopharyngeal and one tongue cancer), IA09 (nasopharyngeal tumor; accession number AFO42003) and ADX1 (tonsillar carcinoma, accession number AJ001483). In two tonsillar cancers, double infection was identified by HPV5/HPV16 and ADX1/HPV16 was identified.

Table 2. HPV Detection, Clinicopathologic Parameters, Therapy Modalities, and Outcome of the Tonsillar Carcinomas
Patient no.HPVClinicopathologyaRisk factorsTherapy modalitiesLocal controlFollow up month
GradingTNStageSmokingAlcohol consumption
  • HPV: human papillomavirus; S: surgery; R: radiation; RC: combined radiotherapy.

  • Smoking: cigarettes per day, alcohol consumption: g alcohol per day.

  • a

    Tumor not more than 2 cm in greatest dimension [T1]; tumor more than 2 cm but not more than 4 cm [T2]; tumor more than 4 cm [T3]; tumor invades adjacent structures [T4]; no regional lymph node metastasis [N0]; single ipsilateral lymph node metastasis, 3 cm or less in greatest dimension [N1]; single ipsilateral lymph node metastasis, more than 3 cm but not more than 6 cm in greatest dimension [N2a]; metastasis in multiple ipsilateral lymph nodes, none more than 6 cm in greatest dimension [N2b]; metastasis in bilateral or contralateral lymph nodes, none more than 6 cm in greatest dimension [N2c]; metastasis in a lymph node more than 6 cm in greatest dimension [N3].

  • b

    Died from distant metastasis.

  • c

    Lymph node metastasis.

116312aIVnonoS + Ryes45
216342cIV3054Ryes19b
316220II60540Sno30
433311III30noS + Ryes19
516212bIVformernoS + Ryes18
616, 5b221III4036S + Rno13
716222aIV40360S + Ryes28
816312bIVNonoS + Ryes33
916242cIV20noRyes21
1016322bIVno18S + RCyes7
1116322bIVnonoS + RCyes16
1216, ADX1230IIIformer36Sno10c
1316220II3054Syes3
1416340IV25noS + RCyes9
15negative243IV40288RCno21
16negative240IV6090S + Ryes22
17negative241IV6072S + RCyes6
18negative232bIV2036S + Rno7
19negative221III2072S + RCyes5
20negative230III2036Syes2
21negative231III20126S + RCyes9
22negative222cIVformer36S + RCyes9
23negative240IV20180Rno17
24negative222bIVformer18S + Ryes4

HPV Sequences in Non-Neoplastic Head and Neck Tissue

In order to control for the occurrence of viral DNA head and neck tissue samples in from nononcologic cases from macroscopically uninvolved locations have been analyzed. These included: 11 biopsies obtained during primary surgery or panendoscopy of 26 patients with HPV-positive HNSCC, taken at least 2 cm away from the tumor location, and 28 tonsils of 14 patients removed due to non-oncologic conditions. Human papillomavirus sequences were detected in 7 of 11 (64%) of the healthy tissue samples taken during surgeries undertaken as part of the diagnosis or treatment of HNSCC. In contrast, none of the biopsies of the normal tonsil samples analyzed contained HPV-sequences. From two tonsillar carcinoma patients the routine paraffin material from the opposite tonsil contained adequate DNA for PCR-analysis: however, both were found to be HPV DNA negative.

Analysis of HPV16 Sequences in Primary and Metastatic Tumors

In order to test if the HPV DNA found in lymph node metastases was identical to the HPV DNA found in the homologous primary tumors, we amplified and compared sequences of the HPV16 LCR from primary and metastatic tumors in two patients. Comparison with a sequence database identified HPV16 variant HO59 (accession number AF67025) with one substitution over 854 nt in patient nine and 16W12E (accession number AF125673) with 100% identity over 853 nt in patient 11 (Table 2). Human papillomavirus16 sequences from primary tumors and homologous lymph node metastasis were identical over 973 nt and 970 nt in patients 9 and 11, respectively.

HPV Analysis of Microdissected Tumor Specimens

We used laser assisted microdissection to acquire specimens containing exclusively tumor cells or histologically normal mucosa immediately next to the tumor for analysis with nested PCR. We tested 27 tumor specimens from 5 HPV positive tonsillar carcinoma cases. We were able to detected HPV sequences in 74% of microdissected tumor specimens. In two patients HPV sequences were still detectable after collecting less than 5 cells by microdissection. In none of the 26 specimens from histologically normal mucosa were HPV sequences identified.

Quantification of HPV16

Viral load determination was performed for seven patients with HPV16 positive tonsillar carcinomas. Viral loads were determined in DNA extracted from whole tissue and three to eight microdissected samples of primary tumor, lymph node metastasis, normal mucosa, and control tissue, respectively. The mean HPV16 loads for each patient and the different tissues are given in Table 3. Human papillomavirus 16 loads in whole sample DNA extractions from primary tumors ranged from 6.6 ×10math image to 172 copies per β-globin-gene copy. The mean HPV16 loads of the microdissected tumor samples were one-and-a-half to three times higher than in the corresponding whole tumor samples in four cases. In contrast, in microdissected tumor material of patient 6 (Table 3), no HPV16-DNA was detected. The levels of the mean HPV16 loads of lymph node metastases were similar to corresponding primary tumors. Two microdissected samples from the metastasis of patient 6 revealed a mean HPV16 load of 77.8, whereas six samples from other tumor regions were HPV16 negative. Except for patient 11, where an HPV16 load of 0.3 was measured in one out of four samples, all the microdissected samples of histologically normal mucosa from tumor margins were HPV negative. Figure 2 shows the mean HPV16 load in whole samples and the minimum, maximum, and mean HPV16 load in microdissected samples of the primary tumor from patients 10 and 12 and in the lymph node metastasis from patient 12. The HPV16 minimum and maximum loads in these cases ranged from 0.2 to 40.4, 0.1 to 58, and 4.4 to 19 HPV16 copies per β-globin copy for a single patient in different tumor areas.

Table 3. Mean HPV16 Loada in Primary Tumor, Metastasis, and Normal Mucosa
Patient no.Primary tumorMetastasisMucosaControl tissueb
Whole samplesMicrodissected samplesWhole samplesMicrodissected samplesMicrodissected samplesMicrodissected samples
  • HPV: human papillomavirus; ND: not done.

  • a

    HPV16 copies/β-globin gene copy.

  • b

    Lymphocytes, connective tissue or acinic cells of small salivary glands.

  • c

    No material available.

  • d

    Mean out of two positive samples, six other samples were negative.

  • e

    Value from one sample, three other samples were negative.

  • f

    Patient had no lymph node metastasis.

1172.4NDcNDcNDcnegativenegative
2ND19,1NDcNDcnegativenegative
66.6 × 10−4negativeND77.8dnegativenegative
109.515.51.622.4negativenegative
111.15.8ND16.30.3enegative
128.220.34.49.2negativenegative
1449.5152.6ffnegativenegative
Figure 2.

Maximum, minimum, and mean human papillomavirus 16 load in microdissected specimens and in whole biopsies of primary and metastatic tonsillar carcinomas.

Correlations between HPV Status and Clinicopathologic Parameters

Tonsillar carcinomas were significantly likelier to be HPV positive than tumors from other examined sites (p < 0,001). In the tonsillar subset a significant correlation of the HPV status with primary tumor grading, alcohol consumption, and cigarette smoking was found (Tables 2 and 4). Poorly differentiated tumors were likelier to be HPV positive (p = 0.008). Patients with HPV positive tumors had a significantly lower alcohol intake per day than patients with HPV negative tumors (p = 0.029). Further, there was a trend for patients with HPV positive tumors to be nonsmokers or light smokers (p = 0.037). Statistical testing revealed no significant correlation of the HPV status to TMN-stage or local tumor control. No significant correlations between HPV positive and negative carcinomas to median age (p = 0.97, t-test) or gender (p = 0.44, Chi-Quadrat test) were seen.

Table 4. Correlation of HPV Status and Clinicopathologic Parameters in the Tonsillar Carcinoma SubsetThumbnail image of

DISCUSSION

Currently it is generally recognized that HPV infections are responsible for induction of several benign head and neck tumors. Best known examples are juvenile laryngeal papillomas, most likely caused by infections with HPV type 6 or 11.29 Although cigarette smoking and alcohol consumption have been accepted as major causative agents for HNSCC, there is now increasing evidence that oncogenic HPVs may present an independent or complementary risk factor for HNSCC development.5, 8, 12 The current study focused on that evidence to evaluate the association of HPV with HNSCC using a combination of newly developed techniques to detect and determine the distribution of HPV genomes in lesional and healthy epithelium of the upper aerodigestive tract.

Using PCR methods capable of detecting over 90% of presently known HPV–types, papillomavirus sequences were detectable in 26% of the analyzed tumors. In view of the number of tested tumors (n = 98) and the high sensitivity of the detection methods employed, we believe this estimate reflects a truer prevalence of HPV DNA in a majority of HNSCCs. This value corresponds to those reported by the recent studies of Gillison et al.5 and Schwartz et al.12 indicating that roughly one quarter of HNSCC cases are HPV-associated.

The analyzed collective was selected to include individuals suffering from primary HNSCC of different localizations, except paranasal sinus and salivary gland tumors. The highest prevalence of HPV sequences was found in oropharyngeal cancer (45%), particularly tonsillar carcinomas (58%). This corresponds with several previous reports30–33 which in most cases did not consider anatomic stratification of the lesions within the oropharynx.5 Our results thus indicate that within the oropharynx HPV infection of the palatine tonsils imply a particular risk. Conversely, only 17% of the tumors from the base of the tongue and none from other oropharyngeal sites were found to be HPV positive.

Tumors from the other localizations (oral cavity, nasopharynx, hypopharynx and larynx) showed low or moderate prevalences of HPV sequences, varying between 7 and 25% (Table 1), again in line with previously reported data.8

It is not clear why tonsils are more susceptible to HPV infection than other localizations of the upper aerodigestive tract. However, it is tempting to speculate that for at least two reasons a viral infection can be facilitated by the specific anatomic structure of these organs. Tonsils contain deep invaginations of the mucosal surface (tonsillar crypts) which might trap the infectious agents simply by lowering their mechanical clearance. Furthermore, the crypts are lined with a monolayer epithelium. This kind of tissue may presumably be more sensitive to HPV infections than stratified squamous epithelium, known to be readily infectable only within their basal keratinocyte layer.

Analysis of HPV sequences isolated from positive HNSCC samples revealed a clear predominance of infections by HPV16 (84%), a virus type believed to act as the primary cause of most of the cervical carcinomas.1, 2 With the exception of one case of HPV33, no other genital HPV types were detected in the tested biopsies, including HPV6 and 11, common causes of laryngeal papillomas. In contrast, 20% of analyzed HPV-positive tumors contained sequences of typical (HPV5 and 19) or putative (ADX1 and IA09) EV-associated virus types. This was unexpected, since EV-HPVs have thus far been regarded as viruses strictly specific for non-mucosal, keratinized epithelia.

A correlation of HPV-positive tonsillar carcinomas with poorly keratinized tumor histology33 and a decreased expression of the retinoblastoma tumor suppressor protein has been reported.34 In line with this, our data show a significant correlation of the HPV status with tumor grading, suggesting an influence of the viral oncogenes on tumor cell differentiation. Further evidence for HPV positive tonsillar carcinomas representing a different tumor entity is suggested by the finding that patients with HPV-positive carcinomas had a significantly lower exposure to the known risk factors for HNSCC (Table 4). In addition, an improved prognosis in patients with HPV-positive HNSCC has been previously reported.5, 34 Other groups35 as well as the current study were unable to confirm a better prognosis for the HPV-positive cases, most likely due to insufficient follow-up. It has been suggested that the reason for the better survival of the HPV positive tumor patients might be a higher susceptibility to radiation or chemotherapy,34 but this hypothesis could not be substantiated. Speculations that the HPV positive HNSCC might be associated with a younger age of patients are refuted by the findings of the current study and earlier investigations.5, 36

Existing data about the prevalence of HPV in normal mucosa of the oral cavity, pharynx, and normal tonsils in healthy persons are conflicting (overview11). Human papillomavirus DNA was not detected in the tonsils of the control group using highly sensitive PCR-techniques. Conversely, 67% of the biopsies with normal mucosa from the HPV-positive tumor cases tested HPV positive with nested PCR. As microdissected mucosa samples remained predominantly HPV-negative, these findings suggest that even in these patients very few of the normal epithelial cells are infected with oncogenic HPVs. Furthermore, two histologically normal opposite tonsils were HPV-negative. As it has been suggested that HPV positive tonsillar carcinomas are a sexually transmitted disease,12, 14 further epidemiologic studies should determine HPV prevalence in tonsils of healthy populations and the relation to sexual behavior.

In the current study we present, to our knowledge, the first data on the viral load in HPV-positive head and neck carcinomas. In cases investigated by quantitative real time PCR, lower levels of HPV16 were found in whole sample DNA preparations than in microdissected tumor tissue. This observation is in accordance with the assumption that only the tumor cells contain HPV DNA, and that the viral load will be diluted by HPV negative cells (e.g., connective tissue, lymphocytes) in DNA preparations from whole samples. This agrees with HPV-negative microdissected control tissue (Table 3). The mean HPV16 loads in microdissected tumor sections ranged from 6 to 150 HPV16 copies per β-globin gene copy in both primary tumors and metastases. This is comparable to the HPV16 load measured in other HPV-associated malignancies, as in Erythroplasia of Queyrat37 and cervical swabs (S. J. Weissenborn et al., unpublished data).

Human papillomavirus 16 loads differed substantially between different tumor areas in both primary tumors and metastases. These data may be explained by a large degree of episomal persistence of viral DNA and unequal distribution to daughter cells during mitosis. This contrasts with the majority of cervical cancers, which show comparatively homogeneous distribution of viral DNA within the tumor due to stable association of integrated viral DNA with the cellular genome. However, sharply demarcated compartments with highly different levels of viral DNA have been shown by in situ hybridization in a cervical carcinoma.38

In a recent comment39 it was suggested that the prevalence of biologically relevant HPVs in tonsillar carcinomas might be lower than the incidence found by PCR because of a low viral load. In our laser assisted microdissection experiments, the HPV DNA was clearly located in the tumor cells, while the normal mucosa at the tumor margin either was HPV negative or contained very low HPV levels. In five of six studied cases measured, viral loads in the tonsillar carcinomas were comparable to cervical carcinomas, which are fully accepted HPV-induced malignancies.

The current study's data are accord with former evidence that HPV positive tonsillar carcinomas are a newly recognized HPV-associated malignancy distinct from other HNSCCs.5, 14, 15, 30

Ancillary