Abstract
Aim: To investigate the level of agreement between three non-invasive methods for hrHPV diagnosis in oral and oropharyngeal squamous cell carcinoma (OSCC, OPSCC) and in oral mucosal lesions. Materials and Methods: For hrHPV DNA FTA Elute card™ and Anyplex II HPV28™ were used and for hrHPV mRNA PreTect SEE™ in tumour patients (n=60), non-tumour lesions (n=51), immunosuppression or previous hrHPV-infection (n=32). Results: The level of agreement between the DNA-methods was 82.2% (k=0.54, p=0.001). Pair-wise comparison for the FTA Elute card were close to the reference (AUC=0.83, 95% CI=0.73-0.90). hrHPV mRNA was diagnosed in 50% of the tumours, with an agreement level of 58.3%, compared to Anyplex II (k=0.17, p=0.04). The hrHPV positivity in oral lesions was 3.9% for immunosuppression and for previous HPV infection 9.4%. Conclusion: The FTA card is reliable for hrHPV DNA diagnosis while mRNA gives an insight into viral activity and correlates with severity of the lesion.
- FTA Elute card
- oral- and oropharyngeal cancer
- potentially malignant oral disorders
- human papillomavirus
- hrHPV
- real-time PCR
- DNA
- mRNA
The intention with this study was to evaluate the efficacy of non-invasive brush methods for diagnosis of hrHPV in oral- and oropharyngeal lesions. In 2018 the global burden of oral cancer was 280,000 and that of oropharyngeal cancer 140,000, whereof 5,900 and 42,000 were attributable to hrHPV infection respectively (1). This compared to the incidence of cervical cancer in 2018 which accounted for 570,000 cases, with more than 90% caused by high-risk human papillomavirus (hrHPV) and existing screening programs (2). In Sweden, the number of oral and oropharyngeal cancer, ~1,000 cases yearly, exceeds the incidence of cancer in cervix uteri ~528 cases, in lack of screening program (3-5). Between 2008 to 2017 the incidence of oropharyngeal cancer in Sweden increased by 65% and that of oral cancer by 26% predominantly caused by hrHPV related oropharyngeal carcinomas (6), particularly in the tonsils and at the base of the tongue (7-13). In addition, hrHPV infections have also been argued to be associated with the malignant transformation of the oral mucosa as HPV is recognized as the cause of precancerous and cancerous cervical lesions (14-16). The IARC concluded that in the oral cavity there is enough evidence of the carcinogenicity of HPV16, but only limited evidence for HPV18 (17). One of the key questions is whether hrHPV is the driver of the malignant transformation of potentially malignant oral disease (PMOD). The prevalence of reported hrHPV PMOD varies extensively from 0% to 53% (18-21). Precursor lesions such as the PMOD oral leukoplakia developed to OSCC in 12% during a 7.5 years follow-up period according to Holmstrup et al. (22). The role of hrHPV in malignant transformation of PMOD is still debated. HPV infection of the oral mucosa is reported to be a frequent finding e.g. in youth, similar among males and females (9.3% vs. 9.8%), but higher for females with (17.1%) than without (4.4%) cervical HPV infection (23). Subclinical chronical oral human HPV infection is suspected to be the main factor for development of squamous cell carcinoma of the head and neck. This is a remaining research question to clarify the steps between initial infection and a later possible development of cancer (24). An extensive number of the HPV infections heal as shown by a meta-regression analysis that estimated the 12-month cumulative incidence of oral HPV to be 4.8% and clearance for HPV16 was 43-83% from 7 to 22 months but with extensive variability (25). Since hrHPV infections have been implicated to be a contributing factor for malignant transformation in the oral cavity it is of interest to elucidate methods for oral HPV diagnosis and possible viral genome integration in the host cell. This work reports on the level of agreement between different methods for hrHPV DNA and mRNA diagnosis in hrHPV-positive OSCC and OPSCC cancers in oral mucosal lesions and clinically healthy oral mucosa.
Patients and Methods
Study population. A total of 152 patients treated at the Department of Orofacial Medicine at Södersjukhuset, Stockholm, Sweden, between October 2015 and October 2019, were recruited for this study. However, 9 patients with tumours without hrHPV were excluded later. The final 143 consisted of 39.2 % women (n=56) and 60.8 % men (n=87), with a mean age of 57.9 years (SD±13.87) (Table I). Informed consent was obtained from all subjects. The patients were categorized into groups (Figure 1).
Group I. Control group (n=60): hrHPV positive OSCC or OPSCC consecutively referred by the Department of Oncology, Karolinska University Hospital with a mean age of 62.4 (SD±9.43), 73% (n=44) were men and 27% (n=16) women. Tumour diagnosis was tonsillar 66,7% (n=40); base of tongue 26,7% (n=16); and lateral border of tongue or floor of mouth 6,7% (n=4); (Table I).
Group II. Patients with PMOD, ulceration, and follicular epithelial hyperplasia (n=51) referred by general dental practitioners (GDP) with a mean age of 61.1 years (SD±12.88), 41.2% were men (n=21) and 58.8% women (n=30). The clinical diagnoses were oral lichen planus 43.1% (n=22), leukoplakia 31.4% (n=16); dysplasia 19.6% (n=10); follicular epithelial hyperplasia 1,9% (n=1); ulceration in healthy mucosa 2% (n=1); and verruca 2% (n=1), all histologically examined; (Table I).
Group III. Patients previously diagnosed with either HPV positive ano-genital condyloma or immunosuppression due to HIV, or cervix dysplasia, referred by the Department of Infectious Diseases and the Department of Obstetrics and Gynecology, at Södersjukhuset (n=32). The mean age was 44.3 years, (SD±14.07), 69% whereof men (n=22) and 31% women (n=10). Oral lesions were observed in 40.6 % (n=13) of the patients with the following clinical diagnoses: oral leukoplakia 9.4% (n=3); follicular epithelial hyperplasia 9.4% (n=3); papilloma 9.4% (n=3); oral lichen planus 6.2% (n=2); syphilis 3% (n=1); and verruca 3% (n=1), all histologically examined; (Table I).
Prior HPV analysis of Group I. HPV analyses (n=60) conducted prior to inclusion were performed at the Department of Pathology and Cytology, Karolinska University Hospital, with 2 different methods randomly. Positivity for hrHPV was confirmed using p16INK4a immunohistochemistry and with a real-time PCR assay with the Cobas® 4800 (Roche molecular systems, 4300 Hacienda Rd., Pleasanton, CA 94588, USA) that tested for 14 hrHPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68), the most prevalent HPV16 and 18 were detected as single genotypes while the other types were reported as a group denoted by ‘Other HPV types’.
Sample collection. Brush samples were obtained from Groups I, II and III plus healthy contralateral mucosa of all patients and a punch biopsy (5 mm Ø) from each site, with the exception of already diagnosed OSCC or OPSCC, and sites in contralateral non-tumour tonsils or base of tongue.
Liquid base DNA (LBDNA). Cells collected were transferred to a sterile test tube containing 1 ml of sterile physiological saline (9mg/ml) and subsequently analyzed for HPV using a DNA-array based HPV assay (Anyplex 28™ II HPV28 Detection, Seegene) at the Department of Laboratory Medicine, Division of Clinical Microbiology, Karolinska University Hospital, Stockholm, Sweden. Anyplex 28™ assay is based on Tagging Oligonucleotide Cleavage and Extension (TOCE TM) technology (26).
Liquid base RNA (LBRNA). Samples were collected using an Isohelix SK-3S brush (Cell Projects Ltd, Harrietsham, Kent, ME17 1AB, UK) and preserved in a standard commercially available methanol-buffer solution (PreTect™), and sent to PreTect AS, Klokkarstua, Norway. PreTect® SEE is a real-time nucleic acid amplification and detection assay used for qualitative determination of E6/E7 viral messenger RNA (mRNA) from the 3 most prevalent carcinogenic HPV types; 16, 18 and 45 in a single analysis (27).
Dry base DNA (DBDNA). Indicating FTA Elute micro cards (art. No WB129308, GE Healthcare, Cardiff CF14 7YT, UK) were used for dry collection, transportation and storage of brush samples. Indicating FTA elute cards contain chemicals that lyse cells, denature proteins and protect, immobilizes and stabilizes nucleic acids, thereby making the samples non-infectious. FTA cards were processed using a dedicated automated laboratory system (easyPunch STARlet; Hamilton Robotics, Bonaduz, Switzerland). The system collects each card, acquires a photographic image of the sample collection area and uses machine learning software to identify parts of the sampling deposition area containing the highest amount of cellular material. Using a 3-mm Ø knife, punches were taken from the area containing most material and deposits and placed in a single well of a 96-well microtiter plate. DNA extraction from the punches was performed as described earlier (28). HPV testing was performed using a real-time PCR-based hpVIR assay (28,29). Briefly, this test detects and quantifies the following HPV types: 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, and 59. It also measures a human single copy gene (HMBS), which serves as a control. For this purpose, the samples must contain sufficient amounts of cellular material so that the test would be informative. The limit of detection for both HMBS and HPV was set to 10 copies per PCR.
Fresh frozen tissue sample for PCR analysis. Punch biopsies (5 mm Ø) of Group II and contralateral healthy oral mucosa were immediately submerged in RNAlater™ stabilization solution (Invitrogen, Thermo Fischer Scientific, Stockholm, Sweden). Thereafter, the samples were frozen at –20°C and analyzed at the Department of Clinical Virology, Sahlgrenska University Hospital, Gothenburg, Sweden. A real-time TaqMan PCR assay targeting 12 hrHPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58 and 59) and 2 low-risk HPV types (6 and 11) was used for the identification of HPV DNA. Real-time PCR for the human betaglobin gene was performed in parallel for sample and amplification quality control; only samples with Ct-values below 35 were included (30).
Statistical analysis. Categorical data were expressed as numbers and percentages, while continuous data were expressed as mean±standard deviation. Descriptive values and differences were calculated using the Kruskal-Wallis rank test. Cohen’s κ (kappa) was used to evaluate agreement between sample sites for pairwise comparisons. A k <0.40 signifies poor agreement, a k from 0.40-0.75 signifies fair to good agreement and a k>0.75 signifies excellent agreement. Receiver Operator Characteristic (ROC) curve and area under the curve (AUC) analyses were performed to assess the diagnostic value of DBDNA. The statistical package StataCorp. 2017, Stata Statistical. Software, Release 15 (College Station, TX, USA) and StataCorp LLC® for Windows was used. Statistical significance was set at <0.05.
Ethical approval. The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethical Review Board Stockholm Sweden (2015-1213-31).
Results
Prevalence of hrHPV in group I. The hrHPV16 type was found in 90% (n=54) of tumour sample sites, whereof 10% (n=6) of cases in combination with other hrHPV, followed by single representations of hrHPV33 (n=3), hrHPV 35 (n=1), hrHPV53 (n=1) and hrHPV54 (n=1); (Table II). Positives detected for hrHPV among tumour sites was highest for LBDNA at 85% (n=51), followed by 66.7% (n=30) for DBDNA and 50% (n=30) for LBRNA. The rate of hrHPV positives detected for contralateral healthy mucosa was highest for LBDNA with 41.7% (n=25) followed by 25.6% for DBDNA (n=10) and 5.2% (lowest) for LBRNA (n=3); (Table III).
Prevalence of hrHPV for group II. Overall hrHPV findings were low in group II with only 7.7% (n=2) positive cases for lesion sites detected via DBDNA, of which one patient with dysplasia presented a positive value for the healthy sample site tested with LBDNA); (Table III).
Prevalence of hrHPV for group III. LBDNA detected hrHPV positives in the healthy contralateral mucosa of 9.4% (n=3) of Group III patients of which one patient with oral leukoplakia was positive for both lesions and healthy sites (Table III).
Results from fresh frozen invasive biopsies for PCR analysis. PCR analysis did not detect any clinically relevant levels of hrHPV among the incisional biopsies of healthy oral mucosa, lesions or PMOD (Table III).
HPV detection agreement for the different sampling methods. All pairwise agreements between the sampling methods for all participants were calculated, using the kappa coefficient. For tumour sites, the kappa based level of agreement between DBDNA and LBDNA was 82.22% (k=0.54, p=0.001) followed by that between LBRNA and LBDNA, which was 58.33% (k=0.17, p=0.04). The agreement between DBDNA and LBDNA was characterized as “fair” whereas the agreement between LBRNA and LBDNA was characterized as “poor”. Both agreements were statistically significant. The level of agreement between DBDNA and LBRNA was 66.67% (k=0.38, p=0.001) and characterized as “poor” (Table IV).
Similar results were obtained for contralateral healthy sites. Kappa for the level of agreement between DBDNA and LBDNA was 69.23% (k=0.42, p=0.001) followed by that between LBRNA and LBDNA which was 63.79% (k=0.14, p=0.02). The agreement between DBDNA and LBDNA was characterized as “fair” and the agreement between LBRNA and LBDNA was characterized as “poor”. Both agreements were statistically significant. The agreement between DBDNA and LBRNA was the highest for these sites with 78.38% (k=0.27, p=0.001). However, it was characterized as “poor” (Table IV).
Diagnostic value of sampling methods using ROC analysis. ROC analysis was used to test the accuracy and predictability of tumours verified as hrHPV-positive. The first indicative HPV-test was performed on tissues excised via a biopsy prior to oral brush sampling. LBDNA showed the highest accuracy for detecting HPV (AUC=0.93, 95% CI=0.83-0.98), with a sensitivity of 86.44%. The DBDNA method showed good predictability (AUC=0.84, 95% CI=0.71-0.94), with a sensitivity of 68.18%. The LBRNA method resulted in a lower detection rate of HPV-mRNA with a sensitivity of 50.85%, indicating that the accuracy of LBRNA was lower compared to that of DNA-methods (AUC=0.75, 95% CI=0.62-0.85). All 3 methods showed 100% specificity. The difference between the 2 ROC curves was statistically tested and showed a p-value of 0.03, which proved that LBDNA was the most predictive for HPV (Table V).
Discussion
The FTA card has been reported to contain the amount of genomic DNA (>10 copies of a single copy gene) that is sufficient for HPV typing (27) and we found a sensitivity of the FTA card to be comparable to that of the LBDNA method. This result is substantiated by previous reports that used the FTA card for detecting high-grade cervical intraepithelial neoplasia (29, 31). The current prospective study had limited options to compare the different methods, because the prevalence of HPV in the non-tumour lesions was low. However, with optimization of workflow, FTA card-based sampling may be used for hrHPV diagnosis and continuous clinical control of oral cavity lesions (32, 33). We surmised that self-sampling of the oral cavity can be introduced using FTA technology, as shown by repeated sampling of patients with potentially malignant cervical lesions, which has been proven to be practical and well accepted (34-36). Furthermore, dry FTA cell sampling technology is an easy system to use in primary care, an effective, reliable and sensitive method for transporting and storing of specimens (37).
The results of our LBRNA assay showing 50% HPV in OPSCC were in accordance with those of a previous study by Koneva et al. (15). Previous in situ hybridization results for the detection of hrHPV E6/E7 mRNA in OPSCC were found to be highly specific, sensitive, and clinically relevant to the de-escalated treatment approach that was used for retesting p16-positive tumours (38). The prevailing consensus on cervical cancer is that upregulated expression of E6/E7 allows insight into viral activity and correlates well with severity of the lesion (39). Detection of mRNA transcripts may, therefore, be a better indicator of the increased risk for progression to neoplasia than HPV DNA (40-42). However, in HPV-positive OPSCC, the incidence of viral integration is reportedly lower, and many tumours contain either extrachromosomal or mixed extrachromosomal and integrated viral DNA (43-46). Whether HPV is a driver of malignant transformation of oral cavity lesions such as PMOD is debatable. The number of infected oral lesions and in clinical healthy oral mucosa in our study was 3.9%. Our records are substantiated by the earlier findings by Sundberg et al., who reported that none of the 74 oral leukoplakia cases were found to be infected (21). This contradicts the results of several previous studies that reported the presence of a significant number of HPV infections in OSCC and in PMOD (14, 19). Pierangeli et al. reported that 53.2% of patients with PMOD were HPV-infected, with HPV16 and 18 being the most frequently detected subtypes (20). The significant levels of variance in prevalence are most probably due to the quality of samples and the geographical region from which the samples were obtained (32, 47). In clinical practice p16INK4a is frequently used as a biomarker of HPV. However, a major issue arises since overexpression of p16 does not correlate with HPV infection in PMOD, OSCC or cervical carcinoma, the latter being one of the main causes for variation in HPV prevalence (14, 21). The large variation in age of patients between studies, as well as the methodology used for sampling and analysis, may also account for these differences (48). In the case of large tumours, cells may contaminate healthy sample sites that account for HPV positivity in healthy contralateral mucosa of the control group. The fact that brush samples from the oral cavity mucosa came from lesions obtained from various areas with different surface layers, some of which were keratinized, may explain the difficulties encountered in acquiring representative samples. Della Vella et al. (49) found the HPV prevalence in incisional biopsies of oral leukoplakia to be 17% compared to 5% in brush biopsies analyzed with the same method as in our study. Their result differs significantly from ours as we found a level of agreement between the two DNA-methods to be 82.2% (k=0.54, p=0.001). This indicates that the brush sampling technique may need to be adjusted to fit both the region of interest and the tentative tumour diagnosis.
Conclusion
The FTA card is a robust and reliable method for detecting hrHPV DNA in potentially malignant oral disorders, as well as post treatment follow-up of OSCC and OPSCC. However, since the prevailing consensus is that upregulated expression of E6/E7 allows insight into viral activity and correlates with severity of the lesion, detection of mRNA transcripts is suggested to be a better indicator of the increased risk for progression to neoplasia.
Acknowledgements
This study was supported by Region Uppsala; the Thuréus Foundation (Grant number 463349110), Region Stockholm (SOF 4-640/2016) and Folktandvården Stockholms län AB (Grant number 7072) for the conduct of the research.
Footnotes
Authors’ Contributions
Conceptualization: CRS, IG, PH, UG, JMH. Data curation: CRS, IG, PH, MK, UG, JMH. Formal analysis: CRS, IG, PH, MK, UG, JMH. Funding acquisition: CRS, IG, PH, UG, JMH. Investigation: CRS, IG, PH, UG, JMH. Methodology: CRS, IG, PH, UG, JMH. Project administration: CRS, JMH. Resources: CRS, IG, PH, UG, JMH. Supervision: CRS, JMH. Validation: CRS, IG, PH, MK, UG, JMH. Writing of draft: CRS, IG, PH, MK, UG, JMH. Writing and editing: CRS, IG, PH, MK, UG, JMH.
This article is freely accessible online.
Conflicts of Interest
The Authors declare no conflicts of interest.
- Received October 12, 2020.
- Revision received November 16, 2020.
- Accepted November 23, 2020.
- Copyright© 2021, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved.