OPSCC cell lines UD SCC02, UPCI SCC072, UPCI SCC089 and UPCI SCC090 were used in this project: the UPCI cell lines were received under MTA from Prof. S. Gollin, University of Pittsburgh School of Public Health, Pittsburgh, PA. The HPV status was confirmed as previously described (19). Short tandem repeat profiling was undertaken to confirm cell line authenticity. A summary of cell lines used is given in Supplementary Table S1.

Fibroblast cultures were derived from normal tonsil and OPSCC explant cultures respectively, (NTF06, NTF10, NTF322, CAF1, CAF2) as described previously (20), using tissue retrieved from patients during ENT surgery, with written, informed consent, under ethical approval (09/H1308/66 and 13/NS/0120: Supplementary Table S2). The authors are grateful to Dr Helen Colley for access to the NTF cultures). Both fibroblasts and OPSCC cell lines were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal calf serum, with 2 mM L-glutamine and 50 IU penicillin and 50 μg ml⁻¹ streptomycin and incubated under standard conditions (5% CO2, 37°C). All cell cultures were subject to regular mycoplasma testing.

CAF1 and CAF2 were seeded at a density of 5,000 cell/ml in a 6 well-plate with 2 ml growth media for 6 days. At daily intervals, cells from one well of the 6 well-plates were detached and counted giving the total count/ml of cells.

TGF-β1: Cells were plated into 6-well plates (Greiner Bio-one, UK) at a density of 1.5 × 10⁵ per well with 2 ml of growth media and incubated overnight at 37°C. The next day, the media was aspirated, and cells were washed with PBS x2 and serum-starved overnight using DMEM serum-free media. On the third day, 2 ml of DMEM serum-free medium containing 5 ng/ml rh-TGF-β1 (Sigma-Aldrich, UK) was added to each well. The experimental control wells were incubated with 1 ml serum-free medium. Cells were incubated with 5 ng/ml TGF-β1 (R&D Systems, USA) for 24, 48 or 72 h. At the end of the experiment, cells were pelleted for downstream RNA/protein analysis and the media was aspirated, centrifuged, and the supernatant was stored at −20 °C.

H₂O₂: Early-passage NTF322, NTF6 and NTF10 cells were grown in 75 cm² flasks to 70% confluence. Cells were then washed with PBS and incubated with 500 µM of H₂O₂ (Sigma-Aldrich, UK) in serum-free media for 2 h. Negative control cells were incubated with serum-free media alone. Cells were then washed with PBS x3 and incubated with growth media for 15 days. Cells were split on day 8. At the end of the experiment, the media were collected, centrifuged at 3,000 rpm for 5 min and stored at −20 °C. Cells were counted and the culture viability percentage was recorded. 4 × 10⁴ cells were seeded in a 12-well plate, prepared for Senescence-Associated β-galactosidase Assay (SA-β-gal) staining. The rest of the suspension was centrifuged, and the formed pellet was collected and stored at −80 °C for further analysis.

Recombinant Osteopontin (rOPN): Fibroblasts were cultured in a T75 cm² flask in growth media to 70% confluence. Cells were washed with PBS x3 and serum-starved overnight. On the third day, cells were incubated with 10 ml of serum-free medium containing 180 ng/ml of rh-OPN (R&D systems, USA) for 48 h. The experimental control was incubated with 10 ml serum-free media only. At the end of the experiment, the media were aspirated, centrifuged, and stored at −20 °C for IL-6 ELISA analysis. Cells were harvested and counted to normalise ELISA analysis results then cells were collected for further IL-6 gene evaluation.

Anti-CD44v6 antibody: NTF322 and NTF10 were cultured in T75 cm² flask with growth media up to 70% confluence. Cells were incubated for 1 h with serum-free media supplemented with 5 µg/ml of CD44var (v6) Monoclonal Antibody (VFF-18) (eBioscience™ Catalogue: BMS125). The experimental control was incubated with serum-free media. 6.5 ml of cell line conditioned media from SCC89, SCC72 and SCC2 was thawed and supplemented with 1.5 ml of fresh growth media. After 1 h of CD44 blocking, rOPN-containing or control CM was added to NTFs and incubated for 24 h. The next day, media were collected, centrifuged, and stored at −20 °C. Fibroblasts were harvested, counted, and collected as a cell pellet for further analysis.

Following the protocol of the collagen gel contraction kit (Cell Bio-Lab Inc, CBA-201), NTFs were harvested and resuspended in growth medium at 2 × 10⁴ cells/ml. A collagen lattice was prepared by mixing 2 parts of cell suspension and 8 parts of cold collagen gel working solution. From this mixture, 0.5 ml was pipetted in a well in a 24-well plate and incubated for an hour at 37 °C. After collagen polymerization, 1 ml of growth medium was added, and the plate incubated for 24 h at 37 °C. The next day, the cells were serum-starved for a further 24 h, then were treated with 5 ng/ml TGF-β1 or 180 ng/ml OPN before releasing the stressed gel matrix from the plate using a sterile spatula. The gels were then incubated for 72 h (TGF-β1 stimulation experiment) or 48 h (rOPN stimulation experiment). Photographs of the gel disc were taken at 24, 48 and 72 h. The contractility of the gels was calculated by measuring the distance between the collage disc border and the plate well using ImageJ (NIH).

Cell lines were grown to 70% confluence in T75 flasks, washed in PBS and incubated with fresh normal media for 24 h. Conditioned medium (CM) was then retrieved, centrifuged at 3,000 rpm for 5 min to remove cell debris and stored at −20 °C. The remaining cells were counted, and CM normalized to a concentration of 3 × 10⁶cells/ml.

Tonsil fibroblast cultures (normal and CAF) were grown to confluence in T75 flasks, washed in PBS and incubated with either OPSCC cell line CM or normal media control for 24 h. This media was collected (M1) and flasks were washed in PBS and incubated with fresh normal media for a further 24 h to collect a “stimulated” fibroblast CM (M2). CM was retrieved, centrifuged at 3,000 rpm for 5 min and stored at −20 °C. The remaining cells were counted to confirm an equal number of cells in each experiment post-stimulation.

OPSCC cell lines were exposed to unstimulated fibroblast CM collected in a similar manner to that described above.

Cancer cells (3 × 10⁵) were seeded in 1.5 ml of growth media on cell culture inserts containing a 0.45 μm plastic membrane filter and placed on a 6-well plate filled with 2.6 ml of growth media. Cells were allowed to adhere overnight. The filters were conditioned before cell seeding to enhance the cellular adhesion by preincubation for a minimum of 1 h at 37 °C. 50,000 cells of NTF 3, NTF 10, CAF1 or CAF2 were plated in 6-well plate in 2 ml growth media and incubated for 24 h The experimental controls were plated in the same way but omitting the fibroblasts or cancer cells. The next day, 1.5 ml fresh growth media was added to the inserts and 2.6 fresh growth media was added to the fibroblasts in 6-well plate, then the cell culture inserts were inserted into the wells carefully to avoid air bubble formation. The co-culture system was maintained for 48 h at 37 °C. At the end of the experiment, the media were aspirated and centrifuged at 3,000 rpm for 5 min and stored immediately at −20 °C. Cancer cells and fibroblasts were detached from the culture system, counted, and stored separately as pellets at −80 °C.

Serum-free conditioned media were collected from cell lines and stimulated fibroblasts as described above and soluble factors analysed by cytokine array Proteome Profiler Human XL Oncology Array kit (R&D systems: ARY026) following the manufacturer's protocol. This kit provides relative expression levels for 84 human cancer-related proteins. Developed films were scanned at high resolution and analysed using ImageJ Studio software (Version 5.2; LI-COR, Inc, USA). Densitometry data were then normalised to the positive control spot (reference spots).

Total RNA was extracted from cell pellets using an Isolate II RNA Mini Kit (Bioline), following the manufacturer's instructions. RNA was quantified using a Nanodrop 1,000 Spectrophotometer (Thermo Fisher Scientific). Five hundred nanograms of isolated RNA was reverse transcribed using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), following the manufacturer's protocol using a Peltier thermal cycler (MJ Research). cDNA was then stored at −20 °C.

Gene expression was quantified using a Rotor-gene Q real-time PCR cycler (Qiagen) with SYBR green or TaqMan chemistry. Quantification was achieved using delta CT values normalized to either U6 or B2M. Each reaction was performed in triplicate. The standard thermal cycle settings for a reaction consisted of 40 cycles (each 95 °C for 10 s, 60 °C for 15 s and 72 °C for 20 s), including a melt curve analysis (when using SYBR green).

Taqman (inventoried, Thermo Fisher Scientific, UK) and SYBR green primers (designed in-house, purchased from Sigma-Aldrich) were used, as seen in Supplementary Table S3.

Cell lines were grown to approximately 70% confluence, washed in PBS. On completion of each respective incubation period (as above), flasks were washed in cold TBS and detached using cell dissociation solution at 4 °C on a rocking machine for 10 min. Cells were centrifuged at 1,000 rpm for 5 min. Cell pellets were lysed on ice in 50 mM Tris–HCL pH 7.4, 250 mM NaCl, 5 mM EDTA, 0.3% Triton X-100 and EDTA-free protease inhibitor cocktail (Roche, Germany).

Lysates were boiled for 5 min in an equal volume of sample buffer and separated using 4%–15% polyacrylamide pre-cast gels (Mini-Protean TGX, Bio-Rad, CA). Separated samples were transferred to a nitrocellulose membrane (Amersham Hybond ECL, GE Healthcare, Chicago), then blocked in 5% skimmed milk in Tris-buffered saline (pH 7.4) with 0.1% Tween-20. This was followed by incubation with primary antibody (Supplementary Table S4) or anti-β actin control (1:3,000) in 5% bovine serum albumin/TBST overnight at 4 °C. Membranes were then incubated with horseradish peroxidase conjugated secondary antibody at a dilution of 1:10,000 for 45 min and imaged using ECL reagent (GE Healthcare).

An 8-well slide chamber (Starstedt, GER) was used to culture the cells for immunofluorescence analysis. 20,000 cells were seeded in each chamber with 800 μl of growth media. Chambers were incubated for 24 h at 37 °C in a humidified incubator with 5% CO2. After treatment with TGF-β1 or OPN, cells were fixed using the recommended reagent according to the manufacturer's instructions This was followed by another incubation with the permeabilization reagent if needed. After that, cells were blocked by the appropriate blocking on a shaker for 15 mins. 200 μl of the diluted primary antibody was added, including the experiment negative control, and incubated for 1.5 h (Supplementary Table S4). Cells were then washed twice with PBS and incubated with the secondary antibody for 1 h. The antibody control chamber was incubated with 200 μl of secondary antibody only. The chambers were washed x3 with PBS and mounted with a coverslip containing a drop of Prolong diamond antifade mountant (DAPI, Thermo Fisher P36966). The slide was placed in a slide box at 4 °C for 24 h. Images were captured using a Ziess Axioplan2 fluorescence light microscope (Carl Zeiss, UK). Staining intensity was quantified using Image J software quantifying the green colour only in fields with equal numbers of cells.

Secreted IL-6 was quantified using the BD OptEIA (BD Biosciences, San Jose, USA) kit for human IL-6 detection. The procedure was performed following the manufacturer's recommendation using 96-well plates. Absorbance was read at 450 nm within 30 min of adding the stop solution. Wavelength correction was calculated by subtracting the absorbance at 570 nm from the absorbance at 450 nm.

Twenty thousand cells per well were seeded in a 12-well plate and left to adhere overnight. Senescence associated β-galactosidase activity was assessed using a senescence detection kit (ab65351, Abcam) following the manufacturer's recommendations.

This study utilised patient samples and associated clinical data within the period of primary diagnosis 2002–2012. Patients were selected using the Oncology databases and cross-referenced against the histopathology archive at Sheffield Teaching Hospitals NHS Trust. Patients were included in the study if there was sufficient tissue remaining after diagnosis for TMA generation and if there was a reasonably complete dataset of associated clinicopathological information. All biopsy material was obtained before commencement of anti-cancer treatment. Paraffin-wax embedded tissue samples were retrieved and collated in a blinded fashion with respect to HPV status. Clinical data regarding tumour recurrence, clinical status and overall survival were retrieved from the patient's medical files at the Sheffield Teaching Hospital NHS Foundation Trust, UK. Overall survival was determined by the difference between the date of treatment and either the date of death due to the tumour or last follow-up. The study was conducted with National Research Ethical Committee approval (UK 12/LO/2018).

TMAs were constructed by selecting tumour regions displaying more than 70% cellularity with minimal necrosis and marked on haematoxylin and eosin-stained sections. The TMAs were constructed using a Beecher tissue arrayer with 3 × 1.0 mm diameter cores from each tissue block arranged at mapped locations into recipient paraffin blocks. TMAs were sectioned (5 μm) by microtome onto Superfrost™ adhesive glass slides (ThermoFisher Scientific).

An additional cohort of 96 cases was added, which had formed part of the PREDICTR study cohort (21). These were received from the University of Birmingham under MTA. An overall summary of the whole cohort is presented in Table 1.

Immunohistochemistry for p16 and αSMA was performed on 4 µm formalin-fixed paraffin-embedded TMA sections. Human skin sections were used as positive controls. Antigen retrieval was performed with heat-induced epitope retrieval in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0). The primary antibody used are noted in Supplementary Table S4, and samples were incubated overnight at 4 °C. After secondary antibody incubation, staining was visualized using a Vectastain ABC Kit (Vector Laboratories) with 3,3′-diaminobenzidine substrate and a haematoxylin counterstain. Staining was assessed as follows: •

p16 (E6H4 CINTECH, using standard diagnostic p16 protocol). Expression was assessed semi-quantitatively, using H-score, to a maximum score of 300. Staining intensity (0–3) and % of cells at each stating level is recorded.

HR-HPV16, 18 E6 and E7 RNA ISH was performed with the RNAscope assay (Advanced Cell Diagnostics, Newark, CA, USA), according to the manufacturer's instructions. The following scoring system was followed for HPV RNA ISH slide assessment: 0: No staining or <1 dot every 10 cells; ×40 magnification. 1: 1–3 dots/cell; ×20–×40 magnification. 2: 4–10 dots/cell; very few dot clusters at ×20–×40 magnification. 3: >10 dots/cell; <10% have dot clusters at ×20 magnification. 4: >10 dots/cell; >10% positive cells have dot clusters at ×20 magnification. Score 0 was counted as a negative result and scores 1–4 as positive.

All experimental data are expressed as mean ± SD of at least three independent experiments performed in triplicate, unless otherwise stated. Data are expressed as mean ± standard error (unless stated). A p-value of <0.05 was considered to be statistically significant.

The study was approved after ethical reviews: 09/H1308/66 (generation of NOFs) and 13/NS/0120 (generation of CAFs) and 12/Lo/2018 (patient cohort). All the clinical cases were pseudo-anonymised before analysis.

The demographic details of the clinical cohort are presented in Table 1. Assessment of HPV status in the clinical cohort (n = 143) by both p16 IHC and HPV 16/18 RNA ISH demonstrated poorer overall survival related to the expression/presence of either biomarker (Figure 1A: p16 p = 0.0002, HPV ISH p < 0.001). In the assessment of αSMA expression (assessed as low or high: see Figures 1B1,B2), high αSMA expression was related to survival in only in HPV-positive cases (Figure 1B: HPV+ p = 0.02, HPV− p = 0.77).

We previously published a comprehensive characterisation of normal oral fibroblasts (NOF) in a overall similar experimental protocol (19), but it is possible that anatomical variations in mesenchyme may result in differences in fibroblast reactivity. Thus, we undertook characterisation of 3 new normal tonsil fibroblast (NTF322, NTF6 and NTF10) cultures, which allowed comparison with NOF cultures. We have demonstrated that all the NTF cultures respond to treatment with TGFβ1, with maximal effect seem at 48 h at transcript level (Figure 2A1), and 48–72 h at protein level (Figure 2A2). The development of αSMA stress fibres mirrored these changes (Figure 2A3). The functional consequences of this are seem in the increase in collagen gel contraction under the same treatment and timescale (Figure 2B), and in an increase in the production of IL-6, at protein level, but this was variable between the cultures and timepoints, with no increase noted in NTF322 (Figure 2C).

Treatment with H₂O₂ resulted in a significant increase in the percentage of SA-β-Galactosidase activity (widely used as an indicator of senescence) in all cultures (Figure 3A), with associated increases in the expression of p16 and p21 in NTF332 and NTF6, assessed by western blot (Figure 3B). H₂O₂ treatment also resulted in an increase in the production of IL-6 in NTF6 and NTF10 (Figure 3C).

These results are very similar to those previously demonstrated using NOFs and a similar treatment schedule, with the induction of aSMA stress fibres on treatment with TGFβ1, and the induction of SA-β-Galactosidase activity on treatment with H₂O₂ (16).

A similar process of characterisation for the two new tonsil CAF cultures (CAF1 and CAF2) was undertaken, after confirmation of lack of contamination of these cultures by other cell types. qPCR assessment of cytokeratin 6, HLA-DR and CD31 demonstrated no significant expression in the cultures, indicating lack of epithelial, lymphocyte and endothelial cell contamination (Supplementary Figure S1). The two CAF cultures, which were both derived from HPV-positive tumours in patients who were both current smokers varied in their phenotype (Figure 4). Whilst both CAF1 and CAF2 increased their expression of αSMA (mRNA transcript and stress fibre formation; Figure 3A), the baseline expression of αSMA was higher in CAF2 than in CAF1. We also identified higher expression of FAP-α and FSP1 in the CAFs when compared to NTF6 (Figure 4B), and similar patterns of expression of PDGRA (data not shown). On assessment of the proportion of senescent cells at baseline, there was a higher proportion of senescent cells in CAF1 cultures (Figure 4C). Treatment of CAF1 and CAF2 with H₂O₂ resulted in an increase in the proportion of SA-β-Gal expressing cells, with the dynamic change greater in CAF2 (Supplementary Figure S2).

The assessment of growth curves, with NTF6 for comparison, also demonstrated differences between CAF1 and CAF2, with the rate of increase in cell numbers less in CAF1, likely due to the higher proportion of senescent cells in these cultures (Figure 4D). Notably, both CAF cultures produced a much higher baseline level of IL-6 production when compared to NTF6, in keeping with the higher baseline SA-β-Gal and αSMA expression (Figure 4E).

In our previous OPSCC paper, we demonstrated the baseline cytokine production of the OPSCC cells (19). The current assessment, which includes a small number of additional cytokines not assessed previously, was undertaken on untreated CAF1, CAF2 and NTF322 cultures (Supplementary Figure S3). This demonstrated a limited baseline repertoire of chemokine production: In CAF1 and CAF2, only CCL2 and MMP2 were detected (with weak expression of SPARC in CAF2), whilst SPARC, Serpin B5 and DCN were detected in NTF322 conditioned media.

After 24 h of treatment of the fibroblast cultures with OPSCC CM, two different collections of conditioned media were made: firstly, the initial 24 h CM, which contains cytokines from both the SCC cells and fibroblasts, and after washing and incubation for a further 24 h in fresh media, the second CM was collected (fibroblast produced cytokines only). The initial conditioned media (CM1) was subjected to cytokine array analysis as shown in Figures 5A–F. The chemokines which were detected are presented in Supplementary Table S5. Several cytokines were detected in the CM which was originally derived from HPV+ and HPV− cells, some of which were also detected in the baseline fibroblast media (Supplementary Figure S3), including CCL2. Increased production was detected solely related to HPV+ cell CM. The cytokine profiles are much reduced in the later CM2 (Supplementary Figure S4), but with a notably increased repertoire still evident in fibroblast CM which has been stimulated by HPV− SCC CM, when compared to HPV+.

We identified OPN (SPP1) as a consistently produced cytokine present in the CM1 from all fibroblast cultures in response to SCC89 CM (HPV negative). Validation of the findings from the cytokine array by ELISA in CM1 demonstrated similar findings (Figure 5H), but OPN is not detected in CM2 (Supplementary Figure S4), which indicates that the source is not the fibroblast cells. qPCR assessment of the OPSCC cells indicated variable baseline transcription of SPP1, higher in HPV-negative cells (SCC 72, SCC89) than in HPV positive (SCC2, SCC90) cells (Figure 5J). We also validated the production of CCL2 in CM1 by ELISA, with no difference in SCC89 and SCC2 CCL2 production in response to CAF1 and CAF2 CM, but variable when stimulated by NTF media (data not shown).

We used a transwell co-culture method to allow the cells to communicate in a paracrine manner but with no direct contact, allowing for separate transcript analysis of the different cells used. Assessment of OPN transcript levels by qPCR indicated that these were only elevated in the HPV-negative cells when co-cultured with either CAFs or NTFs (Figure 6A). No expression was seen either in HPV-positive SCC cells or in CAFs or NTFs. Conversely, IL-6 transcript was increased only in CAFs and NTFs when cultured with conditioned media from HPV-negative SCCs (Figure 6B). no increase was seen in the OPSCC cells of in CAFs/NTFs cultured with HPV-positive SCC conditioned media.

Having established that OPN was being produced by the HPV-negative SCC cells in response to fibroblasts, we treated CAF2 and NTF cultures with recombinant OPN (rOPN). This resulted in increased contraction of collagen gels for NTF10 and CAF2, with increased production of IL-6 (transcript and protein) also detected (Figures 7A,B). The increase in IL-6 production could be reduced, but not abolished, by pre-treatment with anti-CD44V6 antibody (Figure 7C), in keeping with data which demonstrates OPN binding to proteins other than its specific receptor, such as multiple integrins (22). Pre-treatment of fibroblasts with anti-CD44V6 before culture with OPSCC conditioned media showed a reduction in IL-6 production in fibroblasts which were treated with HPV-negative SCC CM, whilst no significant difference was noted in fibroblasts which had been cultured in HPV-positive SCC CM (Figure 7D).