Accuracy of High-risk HPV DNA PCR, p16INK4a Immunohistochemistry or the Combination of Both to Diagnose HPV-driven Oropharyngeal Cancer

The incidence of high-risk human papillomavirus (hrHPV)-driven head and neck squamous cell carcinoma, in particular oropharyngeal cancers (OPC), is increasing in high-resource countries. Patients with HPV-induced cancer respond better to treatment and consequently have lower case-fatality rates than patients with HPV-unrelated OPC. These considerations highlight the importance of reliable and accurate markers to diagnose truly HPV-induced OPC. of p16 INK4a in the HPV-AHEAD study. Presence of HPV E6*I was HPV etiology. All methods carried out in with relevant guidelines and regulations. The overall study included patients with oropharyngeal, oral cavity, laryngeal, hypopharyngeal, and unspecied head and neck squamous cell carcinoma. For a detailed description of the patient characteristics, we refer to the previous publication of the Belgian HPV-AHEAD study data (Simoens, 2021, Cancer Epidemiology). For the analyses in this study, the focus was on the oropharyngeal cancers (ICD-O-3 codes: C01 (base of tongue); C02.4 (lingual tonsil); C05.1 (soft palate); C05.2 (uvula); C09 (tonsil): C09.0, C09.1, C09.8 & C09.9; C10 (oropharynx): C10.0-10.4, C10.8 & C10.9). FFPE tumour blocks and clinical information were collected from OPC patients treated in two Belgian hospitals (GZA and UZA), diagnosed between 1980 and 2010. Ethical was obtained Ethical


Background
In high-resource countries, human papillomavirus (HPV)-driven head and neck squamous cell carcinoma represents an increasing health problem, particularly in cancers of the oropharyngeal region, including base of tongue and tonsils (1)(2)(3)(4). An estimated 20-40% of oropharyngeal cancers (OPC) is believed to be caused by HPV infection, and the large majority of them (>80%) are due to HPV16 (2,5). Patients with HPV-induced cancers respond better to treatment and consequently have better survival than patients with HPV-unrelated OPC (6)(7)(8)(9)(10)(11). Therefore, therapy may be de-escalated with reduced toxicity and less sequelae. However, treatment de-escalation may negatively impact patient survival if HPV was not the carcinogenic driver of the respective tumour. These considerations highlight the importance of reliable and accurate markers, or marker combinations, to diagnose truly HPV-induced OPC.
The most widely applied detection methods thus far are based on PCR ampli cation of viral DNA to determine HPV-positivity. However, several independent studies have highlighted that PCR-based assays for the detection of HPV DNA are not su ciently accurate to establish the viral causality (12)(13)(14)(15)(16). These PCR-based methods are highly sensitive and can detect even a few DNA copies per sample, which might yield false-positive results mainly re ecting transient infections (17)(18)(19). Additional markers, such as the presence of viral E6/E7 mRNA transcripts and p16 INK4a expression as surrogates for HPV-induced transformation, allow a more accurate classi cation of HPV-driven head and neck cancers (HNC) (14,(20)(21)(22)(23)(24).
The detection of HPV E6/E7 oncogene transcripts is currently regarded as gold standard in this context, as HPV-driven carcinomas critically depend on the carcinogenic action of the HPV E6 and E7 oncoproteins (22,25). However, the detection of viral transcripts is laborious and may not be routinely feasible in all laboratories. This is particularly true for mRNA transcript detection from formalin-xed, para n-embedded (FFPE) tissue specimens, which are frequently used during routine diagnostic work up; however, RNA integrity may be affected by the xation protocol.
Apart from single p16 INK4a IHC, combined detection of p16 INK4a IHC and HPV DNA detection by PCR is frequently used.

HPV-AHEAD (FP7 funded network)
The HPV-AHEAD study ("Role of human papillomavirus infection and other co-factors in the aetiology of head and neck cancer in Europe and India") group comprised partners from six European countries and from India (website HPV-AHEAD: https://hpv-ahead.iarc.fr/). The main goal of the study was to perform a comprehensive analysis on a large number of HNC cases to clarify pathogenic pathways in HNC carcinogenesis, and to identify clinically useful biomarkers.
In a previous publication (33), the results of the histological and molecular assessment of 1039 archived HNC specimens from Belgian patients were described, primarily using the detection of HPV DNA, mRNA, and p16 INK4a IHC. In this study, we focus on the accuracy of these tests -individually and in combinationto diagnose hrHPV-driven oropharyngeal cancer.

Methods
All methods were carried out in accordance with relevant guidelines and regulations.

Patient selection, clinical information and tissue specimen collection for the Belgian cohort
The overall study included patients with oropharyngeal, oral cavity, laryngeal, hypopharyngeal, and unspeci ed head and neck squamous cell carcinoma. For a detailed description of the patient characteristics, we refer to the previous publication of the Belgian HPV-AHEAD study data (Simoens, 2021, Cancer Epidemiology). For the analyses in this study, the focus was on the oropharyngeal cancers Preparation of tissue sections (University of Antwerp, Antwerp, Belgium) FFPE-tissue blocks were all processed at the UA laboratory of cell biology and histology, following the optimized HPV-AHEAD sectioning protocol (33,34). Brie y, minimal ten sections (S) were prepared from each FFPE block. The rst (S1) and the last (S10) 5µm sections were hematoxylin and eosin (H&E) stained for morphologic histology interpretation and used to check for the presence of tumour. S2 and S9 (5µm) were used for p16 INK4a IHC staining, while the 10µm sections S3-5 and S6-8 were used for the extraction of RNA and DNA, respectively.

Histological review
All sections were re-evaluated by the HPV-AHEAD pathology review panel. The review was blinded with respect to the original local diagnosis. Only FFPE blocks where S1 and S10 H&E sections re ected squamous tumour tissue were included in the analysis.
HPV E6*I mRNA analysis (DKFZ, Heidelberg, Germany) All OPC cases were analysed for the presence of: (i) HPV16 E6*I mRNA, and (ii) ubiquitin C (ubC) mRNA as a cellular mRNA positive control (housekeeping gene used for RNA quality control). OPC cases positive for DNA of a non-HPV16 genotype were additionally analysed for E6*I mRNA of the respective genotype. Specimens that were HPV E6*I and/or ubC mRNA-positive (RNA+) were considered RNA valid. Reverse Transcription-PCR was carried out using the QuantiTect Virus Kit (Qiagen, Hilden, Germany), as described in full in former HPV-AHEAD publications (33,35). The type-speci c E6*I mRNA assays identifying transcripts of fourteen high-risk and six possible/probable high-risk HPV genotypes (36) were applied.
HPV DNA genotyping (IARC, Lyon, France) HPV DNA genotypes were detected by a E7 type-speci c multiplex genotyping (E7-MPG) assay, which combines multiplex PCR and bead-based Luminex technology (Luminex Corporation, Austin, TX), as previously described (37,38). Type speci c-MPG uses HPV type-speci c primers targeting the E7 region of thirteen high-risk and six possible/probable high-risk HPV genotypes (HPV 16,18,26,31,33,35,39,45,51, 52, 53, 56, 58, 59, 66, 68a and b, 70, 73 and 82), as well as two low-risk HPV genotypes (HPV 6 and 11). Two primers for ampli cation of the beta-globin (β-globin) gene were also included to control for the DNA quality of each specimen. A detailed description of the DNA extraction and further HPV genotyping test characteristics can be found in the previous publication on the Belgian HPV-AHEAD study (33).
Genotyping controls and DNA preparation were blindly analysed, and no sign of contamination of negative controls was detected during the laboratory work.
p16 INK4a expression (Roche mtm laboratories, Mannheim, Germany) Expression of p16 INK4a was evaluated by IHC, with a dual-immunostaining protocol for the simultaneous immunostaining of both p16 INK4a and Ki-67 biomarkers (CINtec PLUS kit, Roche mtm laboratories AG, Mannheim, Germany), as previously described in the HPV-AHEAD publications (33,35). As in all HPV-AHEAD studies (35,39), a continuous, diffuse staining for p16 INK4a within the cancer area of the tissue sections was considered as positive, while a focal staining or no staining was considered negative. IHC slides were analysed without knowledge of any other clinical information (including HPV DNA and RNA status) by the scientists RR or DH and reviewed by one of the European members of the HPV-AHEAD pathology review panel (JPB, BLR, or FM). Discrepant cases were re-evaluated by a pathologist outside the review panel (AC), and nal classi cation of the staining was based on majority consensus results.
Corresponding to the former Belgian HPV-AHEAD publication (33), the p16 slides with technical issues were restained and re-evaluated, to minimize the number of missing results. Relative sensitivity and speci city of each test compared to the other tests were also assessed. Binomial 95% con dence intervals (95% CIs) were computed for proportions. Given the matched testing of the same patient specimens, McNemar 95% CIs were computed for relative accuracy parameters. Continuous variables were summarized by their mean and 95% CIs. ANOVA was used to compare mean age by gender.

Statistical analysis
Statistical analyses were performed using STATA 16 (StataCorp, College Station, TX). p-values were two sided and statistical signi cance was set at p equal or less than 0.05.

Study population characteristics
Ninety-nine of 116 OPC patients were included in the analysis, after exclusions based on the pathology review (for reasons of not being a squamous cell carcinoma or not re ecting invasive cancer, N=13) and beta-globin PCR negativity (N=4). 71.7% (71/99) of the patients were male, 26.3% (26/99) were female, and for the remaining two samples the gender of the patients was not speci ed. The mean age at diagnosis was 67.3 years (95%CI: 64.9-69.7) and was not different between males and females (p=0.47).  Table 2, for all OPC cases, and for the tonsils and the base of tongue in speci c. hrHPV DNA PCR testing was slightly more sensitive compared to p16 INK4a IHC (ratio: 1.04, 95% CI: 0.97-1.11), but equally speci c (ratio: 1.00, 95% CI not computable) (see Table 3b and the Supplementary   Table). The combination of hrHPV DNA and p16 INK4a was slightly more speci c (ratio: 1.05, 95% CI: 0.97-1.21) compared to both single tests.

Discussion
In these Belgian OPC patients, PCR testing for HPV DNA showed the highest proportion of positive cases (36%). Due to its high sensitivity, HPV DNA PCR-based assays can detect low viral copy numbers, which may not trigger carcinogenesis. These assays cannot distinguish between transcriptionally-active and passenger HPV infections. Therefore, other assays (HPV E6*I mRNA, and p16 INK4a staining) were evaluated to assess whether combined testing could improve the speci city. Only 28% of the samples with a valid HPV E6*I mRNA result were positive for HPV DNA and p16 INK4a , which might be the fraction of oropharyngeal cancers caused by HPV. After all, the detection of HPV E6 and/or E7 mRNA is seen as the gold standard in this context, because HPV-driven carcinomas critically depend on the continuous expression of E6/E7 oncogenes of hrHPVs (22,25,40). However, the detection of viral transcripts is challenging, not only because the RNA extraction step is laborious and time consuming, but speci c infrastructures and equipment are needed. Therefore, HPV RNA assays may not be routinely feasible in all laboratories or on all tissue samples.
HPV-driven OPC represents an increasing health problem, and therefore, reliable and accurate diagnosis becomes essential. In our Belgian HPV-AHEAD study, HPV DNA PCR testing alone was 100% sensitive, but less speci c (92.5%) compared to mRNA testing. The inherent strength of the PCR-based methodology lies in its capacity to detect very small amounts of HPV DNA. At the same time, strict laboratory procedures and controls are critical in reducing contamination-related false-positive ndings (41). Immunohistochemistry for p16 ink4a is most widely used as a surrogate marker for hrHPV infection in FFPE tissues, also for oropharyngeal squamous cell carcinoma. p16 IHC testing alone was a bit less sensitive (96.4%), but co-presence of hrHPV DNA and p16 INK4a positivity was similarly sensitive (96.4%) and more speci c (97.0%) compared to each test separately. In literature, several authors have advocated against the use of either p16 INK4a or hrHPV DNA alone as indicators of HPV-induced etiology in cancers, but recommended their combined use as a reliable and practical approach to differentiate HPV-induced from HPV-unrelated tumours (25,(42)(43)(44). The meta-analysis performed by Prigge et al. (45) also showed a similarly high (pooled) sensitivity of the combined testing (93%) and either p16 (94%) or HPV DNA (98%) alone, and the (pooled) speci city (96%) was signi cantly higher than either testing method alone (83% and 84%, respectively). These accuracy data were con rmed by these Belgian results.
Relative sensitivities and speci cities were not signi cantly different from unity in our study, nor in the meta-analysis of Prigge et al. (45), where signi cance was only reached for the relative speci city of the combination of HPV DNA PCR and p16 INK4a tests versus p16 INK4a IHC (rel. spec.: 1.13 (95% CI: 1.04-1.23)), by pooling from multiple studies. Furthermore, better clinical outcomes have been reported for patients with HPV-induced compared to HPV-unrelated OPC (6)(7)(8)(9)(10)(11). Given the ongoing discussions about the value of therapy de-escalation in patients with HPV-induced OPC, the reliable identi cation of viral etiology in the tumour and its impact on patient prognosis gain particular momentum. Improved prognostication by combined p16 INK4a and hrHPV DNA detection compared to single marker analysis has been demonstrated in a large metaanalysis on tumours in the head and neck region (46). In our study, 97% (92/95) of the patients would have been correctly diagnosed with the combined testing approach. However, one 58-year old male (current smoker and alcohol consumer) with a T1 (N0 M0) microinvasive tonsillar tumour might have been treated too aggressively by applying this combined test strategy, as his tumour would have been classi ed as a HPV-unrelated case, being p16 negative while actually both DNA and mRNA HPV16 positive. More than six years after diagnosis, the patient was alive without any evidence of disease after having had radical wide excision surgery.
On the other hand, treatment de-escalation may negatively impact disease-speci c patient survival rates if HPV in the tumour was not the cause of carcinogenesis. In our Belgian OPC patients, the number of false positive cancers would be reduced from ve to two with the combined p16+DNA detection, who may otherwise receive a less aggressive treatment regimen. These two HPV-unrelated cancer patients were male and diagnosed with a tonsillar cancer. The rst patient, with a Tx N2 M1 tonsil NOS tumour, was a heavy drinker and current smoker who only underwent surgery and had a recurrence within six months. He died two years after being diagnosed. The other patient, with a T3 N2 tonsil NOS tumour, received surgery followed by radiotherapy and concurrent chemotherapy for two months. At the last follow-up date, this patient was alive, but with evidence of residual disease. Of note, these two HPVunrelated patients clearly had a much worse disease status at diagnosis and thereby reduced survival chances compared to the above described HPV-driven tonsillar cancer patient.
Ninety-seven percent of the OPC patients in our study would have been correctly diagnosed as patients with a HPV-driven cancer by combined p16 INK4a and hrHPV DNA (p16+DNA) detection. Sequential testing algorithms (p16->DNA and DNA->p16) resulted in equally accurate results, however, with a 60% reduction in the number of tests needed to be performed. This will cause a substantial reduction in costs and laboratory time, while providing the same clinical value. Especially the algorithm of p16 INK4a on all samples followed by HPV DNA PCR on p16 INK4a -positive samples only would be a practical strategy.
After all, p16 INK4a IHC can easily be combined with standard histology when a H&E-stained tissue section is prepared for examination by a pathologist. It is a routine diagnostic procedure, at a relatively low cost, available as a validated in-vitro diagnostic reagent and a fully automated protocol, which generates results within several hours after the procedure has been requested. HPV DNA PCR is also a standard laboratory procedure with high throughput and quick results, although against a higher cost. Preselection by p16 INK4a -staining therefore reduces the workload and associated costs, and the combination of HPV DNA and p16 INK4a testing leads to an important reduction of the number of false-positive observations versus the use of either of these assays alone.

Conclusions
In conclusion, combined testing for hrHPV DNA and p16 INK4a enhances speci city up to 97%, while maintaining high sensitivity (96%), compared to single testing. The diagnostic test combination represents an accurate, accessible and high-throughput testing strategy for the diagnosis of HPV-induced OPC in the clinical setting and for disease prognostication.