From an established retrospective database of patients with primary OPSCC diagnosed between January 2000 and January 2012 in the department of Otolaryngology, Head & Neck surgery & Audiology at Rigshospitalet (Copenhagen, Denmark), a balanced cohort of 48 (52%) HPV-positive and 46 (48%) HPV-negative patients were randomly assembled. The inclusion criteria were primary OPSCC with preserved large biopsies or diagnostic resection specimens being available for IHC expression analysis. Exclusion criteria were insufficient tumor tissue for IHC analysis or incomplete dataset in the database. All specimens were collected prior to non-surgical treatment if this was indicated. Because HPV is a well-known powerful prognostic marker of OPSCC outcome, a balanced selection based on HPV status was performed. Both p16 and HPV-DNA status was available. The dataset in this current study has been part of previous publications and assays for the analysis of p16 and HPV-DNA have previously been described (25). Briefly, for p16 detection by IHC, the anti-P16 clone E6HA OptiView DAB IHC Detection Kit (Roche Diagnostics); for HPV-DNA analysis, DNA was isolated using the QIAamp DNA FFPE Tissue Kit (Qiagen, Inc.) and PCR was performed using General Primers GP5⁺/GP6⁺. The 8th version of the TNM Union For International Cancer Control (UICC8) staging manual was applied (26). The recorded clinicopathological characteristics and treatment modalities are provided in Table I. The present study was approved by the Ethical Committee of the Capital Region of Denmark (protocol H-15016322; Copenhagen, Denmark) and was conducted in accordance with The Declaration of Helsinki.

At the time of collection, fresh tissue was fixed in 10% buffered formalin at room temperature for 24 h and then embedded in paraffin. The paraffin-embedded tissue blocks were sliced into 4-µm thick serial sections. Deparaffinization was performed with xylene for 15 min followed by dehydration using an ethanol gradient (from 99 to 70%, then demineralized water). IHC staining was performed on a Ventana Benchmark Ultra semi-automated autostainer (Roche Diagnostics). All antibody protocols have been previously optimized on positive and negative control tissues. The following antibodies were used: anti-cytokeratin (CK) AE1/AE3+8/18 (Pan Cytokeratin Plus; cat. no. 162; 1:200; Biocare Medical, LLC), mouse anti-human uPAR R2 (1:20,000; in-house antibody; Finsen Laboratory), anti-EGFR (Confirm anti-EGFR 5B7; cat. no. 790-4347; ready-to-use; Ventana Medical Systems, Inc.; Roche Diagnostics) and anti-TF (anti-TF; cat. no. 4509; 1:150; American Diagnostica, Inc.). Antigen retrieval for uPAR was performed with proteinase K (1:50 in Tris-HCl; cat. no. S3004; DAKO, Agilent Technologies, Inc.) for 8 min at room temperature, followed by 100°C heating with Cell Conditioning 1 buffer (CC1; Ventana Medical Systems, Inc.; Roche Diagnostics) for 16 min. For TF, EGFR and CK, heat-induced epitope retrieval was performed at 100°C in CC1 buffer for 32 min. The blocking step was performed using Peroxidase Blocking Solution (cat. no. S2023; DAKO, Agilent Technologies, Inc.) for 8 min at room temperature, followed by 2% BSA (cat. no. A7906; Sigma-Aldrich; Merck KGaA) blocking for 10 min at room temperature. For uPAR and TF, EnVision mouse HRP-conjugated EnVision FLEX+ secondary antibodies were used at room temperature for 40 min (cat. no. K8002; ready-to-use; DAKO; Agilent Technologies, Inc.), followed by detection with DAB and DAB+ Chromogen Solution (cat. no. K4065; diluted according to manufacturer's instructions; DAKO, Agilent Technologies, Inc.). For EGFR and CK, ultraView Universal DAB Detection Kit combined with ultraView Kit+ amplification (cat. nos. 760-500 and 760-080, respectively; Roche Diagnostics) and OptiView DAB IHC Detection Kit (cat. no. 760-700; Roche Diagnostics), respectively, were applied for detection.

In addition, separate sections were also stained with H&E for a 60 min protocol including deparaffinization, dehydration as described above, then stained in hematoxylin (cat. no. 854183; Capital Region, Hospital Farmacy) for five minutes, rinsed and stained with eosin (cat. no. 854653; Capital Region, Hospital Farmacy) for three minutes at room temperature. Digital images of all IHC slides were acquired using a Axio Scan.Z1 (Carl Zeiss AG) brightfield microscope slide scanner.

Five slides from each individual tumor were reviewed and scored by two experienced head and neck pathologists (KK and GL) blinded to the clinical dataset. In case of any discrepant assessment of IHC target expression, individual slides would be reviewed together to produce a consensus score. For each tumor, the extent of tumor compartmentalization and differentiation was evaluated using H&E and CK staining images. For uPAR, TF and EGFR respectively, IHC staining was semi-quantitatively assessed for intensity (I-score) and proportional expression (P-score; that is, the proportion of cells with positive biomarker expression) within the tumor compartment by assessment of the whole tumor section at lowest magnification. The I-score was stratified as 0–3 (none, weak, moderate and strong intensity, respectively) whereas the P-score was stratified as 0–3 (0–10, 11–50, 51–75 and 76–100%, respectively). The P-score was estimated based on the observed target expression on tumor cells, stromal cells and inflammatory cells within the tumor compartment. By adding the I-score and P-score, a composite semi-quantitative score (PI-score) was calculated, as proposed by Allred et al (27). The PI-score is a 7-point system stratified from 0 to 6. If present, IHC expression in adjacent non-cancerous tissues would also be qualitatively characterized. To estimate the expression rate of uPAR, TF and EGFR within the cohort, expression would be considered positive if the PI-score was >1. To assess the possible prognostic value of the three targets, the PI-score was dichotomized into high and low expression based on the mean value for each target for association analyses of associations. PI-scores <3 was used for the dichotomization of uPAR and TF, whereas PI-scores <4 was used for EGFR.

For associations between biomarker expression and each of the clinicopathological variables, Pearson's χ² test or Fisher's exact test (for low number of events) was applied; for age comparison, an unpaired t-test was performed. For survival statistics, overall survival (OS) was defined as time from the date of diagnosis until death of any cause. Recurrence-free survival (RFS) was defined as time from the date of diagnosis until the date of recurrence confirmed by biopsy, death by any cause. Kaplan-Meier curves were applied for survival plots and 5-year survival estimates, where log-rank test was used for comparison between the groups. Using the Cox proportional hazards model, univariate and multivariate estimates of hazard ratios (HRs) were calculated for OS and RFS following adjustment for sex, age, smoking status, HPV status, tumor location, T-site, T-stage, N-stage, UICC8 TNM-stage, tumor differentiation and biomarker expression. P<0.05 was considered to indicate a statistically significant difference. Data analysis was performed in the SAS software package, version 6.1 (SAS Institute, Inc.) and R statistics (version 3.6.1).

In this cohort of 93 patients, the median age at diagnosis was 60 years (range, 40–84 years), where 69 of the patients (74%) were men. The median follow-up calculated from the day of diagnosis was 4.8 years (range, 0.03-14.2 years). The tumor subsites were palatine tonsils (75%), base of the tongue (13%), pharyngeal wall (9%) and soft palate (3%). At follow up, January 2020, 51 patients were alive, and 42 patients were deceased, 25 (27%) of whom succumbed to OPSCC. During follow-up, 19 (20%) patients had recurrence confirmed by biopsy after primary treatment. The median time to recurrence was 9 months.

The positive expression rates in tumors based on the PI-score (PI>1) for uPAR, TF and EGFR were calculated to be 98.9, 76.3 and 98.9%, respectively. For the I-score, positive staining (I>0) for uPAR, TF and EFGR was present in 100, 89.3 and 100% of the specimens. The exact extension of the tumor compartment, and the margin demarcation from adjacent non-cancerous tissues were evaluated on the H&E and CK stainings and then compared to the topographic distribution of the IHC stainings for each of the three biomarkers. For uPAR a tumor-specific staining pattern was observed, with enhanced expression within the tumor compartment but limited or absent expression in the adjacent non-cancerous tissues (Fig. 1). In the tumor compartment, membranous uPAR staining was seen on tumor cells as well as on stromal cells. Tumor-associated stromal uPAR expression was also observed on macrophages, neutrophils, fibroblasts. At the mucosal margin, there was a sharp demarcation between carcinoma and adjacent epithelium, with the absence of uPAR expression in the unaffected epithelial lining. At the deep margins, uPAR staining delineated the tumor compartment. In addition, in the majority of cases, a narrow rim of expression was present in the stroma at the invasive border. Intense uPAR staining was frequently observed at the invasive front of the tumor and in the microscopic tumor buds separated from the primary tumor. In 10 (11%) of the specimens, uPAR positivity was also found on neutrophilic granulocytes, where pronounced inflammatory infiltration or abscess formation was discovered outside the tumor compartment.

TF demonstrated a tumor-specific expression pattern when positive, though in some cases adjacent normal epithelium also demonstrated weak positivity. As a general pattern, more intense TF staining was typically seen at the invasive front at the deep margin (Fig. 1).

For EGFR, strong and homogeneous expression was observed in the tumor compartment, but high expression could also be observed in the normal mucosal epithelium (Fig. 1). EGFR expression in the salivary tissue was consistently positive in the ductal epithelium (Fig. 1).

A Cox proportional hazards univariate and multivariate analysis was performed for RFS and OS (Table II). Age, HPV status and UICC8 TNM stage were identified to be independent prognostic factors for OS according to the univariate analysis, but only age and HPV status maintained significance in the multivariate analysis (Table II). Only increase in TNM staging for more advanced stage 3 and 4 disease was associated with RFS according to univariate analysis. For the three biomarkers, high uPAR PI-scores had a non-significant trend towards reduced OS according to univariate analysis only (Table II). In addition, Kaplan-Meier survival curves were generated with univariate log-rank tests performed to assess the association of the PI-scores of the three biomarkers with the OS and RFS (Fig. 2). For the entire cohort, the 5-year OS and RFS was 59 and 76%, respectively. For HPV-positive and HPV-negative patients, the respective 5-year OS was 77 and 41%, whereas the respective 5-year RFS was 81 and 67%. For patients with a high and low PI-scores for uPAR, the 5-year OS was 57 and 68%, respectively, whereas the RFS was 74 and 79% for patients with high and low PI-scores, respectively (Fig. 2). In addition, separate tests were performed for the P-score and I-score but no statistically significant association with OS or RFS could be detected (data not shown).

Possible associations between clinicopathological parameters and high or low PI-scores of uPAR, EGFR and TF were next investigated by univariate analysis (Table III). No statistically significant relationships were be detected.