The five HPV-negative cell lines Cal27, Cal33, UPCI:SCC040, UPCI:SCC131, UPCI:SCC099 (ATCC, Wesel and DSMZ, Braunschweig, Germany) were cultured in DMEM (Life Technologies, Darmstadt, Germany) supplemented with 10% fetal bovine serum (Capricorn, Ebsdorfergrund, Germany) and maintained at 37°C in a humidified 5% CO₂ atmosphere. The GenePrint 10 kit (Promega, Walldorf, Germany) and GeneMapper 5.0 software were used for short-tandem repeat analysis. To confirm the authenticity of all cell lines, the data were compared to Expasy and DSMZ databases (30). Cell lines are routinely tested free from mycoplasma contamination using a PCR-based assay (31).

Cell lines and samples were irradiated at the Marburg Ion-Beam Therapy Centre (MIT) with a vertical beam of 114.5–129.5 MeV/n ¹²C-ions and positioned in a spread-out Bragg peak (SOBP) of 10–20 mm as described previously (32). Fields were applied using active scanning with a field size of 18 × 18 cm². Doses of 0.25, 0.5, 1, 2, 3, and 4 Gy were applied as indicated.

An X-ray biological irradiator Precision X-RAD 320ix (Precision X-Ray, North Branford, CT, USA) was used to irradiate cell lines and samples at 320 kV and 8 mA, dose rate of 1.1 Gy/min, and Thoräus filter 0.5 mm Cu+ 0.5 mm Al. Absolute dose measurements confirmed the applied doses. Doses of 1, 2, 4, 6, and 8 Gy were applied as indicated.

For the preplating assay (33), exponentially growing single cells were seeded in a range of 1 × 10¹–10⁴ cells/cm², depending on the cell line and irradiation dose. Cells were irradiated after adherence with the indicated irradiation quality and dose.

For the delayed plating assay, exponentially growing cells were seeded in a range of 5 × 10⁴—10⁵ cells/cm², depending on the cell line. Cells were irradiated after adherence with the indicated irradiation quality and dose. 16 hours after irradiation, growing single cells were seeded in a range of 1 × 10¹–10⁴ cells/cm².

After incubation for 10–14 days, the grown colonies were fixed and stained (10% formaldehyde, 0.1% crystal violet) and colonies >50 cells were counted. The clonogenic survival was calculated by normalization to the plating efficiency of untreated cells. Survival curves were fitted to the linear–quadratic model (SF = exp - [a × D + b × D^2]) according to a least square fit (GraphPad Prism 8.1.1 software, San Diego, CA, USA). D10 was extrapolated from these data and used for calculation of RBE. Each experiment was done at least in three biological triplicates with a minimum of three independent technical repetitions.

Senescence was measured by detection of senescence-associated β-galactosidase (SA-βgal) activity with a flow cytometer (34). Cells were seeded at 2–4 × 10⁴ cells/cm², irradiated after adherence at the indicated doses, and incubated for 1–6 days. Staining was done with 5-dodecanoylaminofluorescein-di-β-galactopyranoside (C12-FDG, Thermo Scientific, Waltham, MA, USA) after lysosomal alkalinization with bafilomycin A1, which ensures the lysosomal origin of SA-βgal activity. The experiment was described previously (9). In deviation from this study, all cells with high C12-FDG and high SSC signal were considered senescent and the setup of flow cytometric parameters in terms of photodiode settings varied, explaining the difference in total intensities of senescence, but showing similar amounts of relative senescence. All measurements were performed on a LSRII cytometer (BD Biosciences, Heidelberg, Germany), and data were analyzed with FlowJo v10 Software (Tree Star Inc., Ashland, OR, USA). Each experiment was done in biological triplicates.

DSB repair foci were analyzed using co-staining of γH2AX and 53BP1, as described previously (32). Cells were fixed and stained at 0, 4, and 24 h after irradiation with a primary antibody solution as follows: mouse monoclonal anti-phospho-S139-H2AX antibody (1:500, clone JBW301, Millipore, Darmstadt, Germany) and rabbit polyclonal 53BP1 antibody (1:500, Novus Biologicals, Wiesbaden, Germany). Secondary antibody solutions were mouse Alexa-Fluor 594 (1:1,000) and rabbit Alexa-Fluor 488 (1:1,000, both Invitrogen, Karlsruhe, Germany). Cells were mounted in ProLong Gold Antifade Reagent with DAPI (Invitrogen, Karlsruhe, Germany). Immunofluorescence was analyzed using the Leica DM5500 wide-field microscope and LAS-AF software (Leica, Wetzlar, Germany). All experiments were performed at least twice and with 100 counted nuclei per experiment.

Briefly, 1–1.5 × 10⁵ cells/10 cm² were grown in 2 ml medium and irradiated after a 16 h adherence with photons or ¹²C-ions as indicated. Supernatants were collected on indicated days after irradiation, immediately frozen in liquid nitrogen, and kept at -80°C until further analysis. Cytokine concentrations in cell culture supernatants were measured with the human IL1B DuoSet ELISA System (R&D Systems, Minneapolis, MN, USA) according to the manufacturer’s protocol. All experiments were performed twice in technical duplicates.

Transient knockdown of the transcripts was performed using Lipofectamine 2000 (Life Technologies, Carlsbad, CA) as described previously (35). Cells were seeded in a density of 1.5 × 10⁴ cells/cm² and transfected after a 16 h adherence with 50 nM siRNA, according to the manufacturer’s protocol. Human IL1A or IL1B and control siRNA oligonucleotides (ON-TARGETplus, SMARTpool) were purchased from Dharmacon (Horizon Discovery Group, Cambridge, UK). Sequences are listed in Supplementary Table 1 . Knockdown efficiency was verified by qRT-PCR 1 day after transfection.

Cells were seeded at 5 × 10⁴ cells/cm² and treated and RNA extracted as indicated (dose and time point) using NucleoSpin RNA II Kit (Macherey-Nagel, Dueren, Germany) according to the manufacturer’s instructions. cDNA was reverse transcribed by incubation of 500 ng RNA with 200 U RevertAid Reverse Transcriptase in the presence of 5 μM random hexamers, 5 μM Oligo(dT)₁₈, 500 μM dNTPs, and 1 U/μl RiboLock RNase inhibitor (all from Thermo Scientific). qRT-PCR was carried out on a QuantStudio5 Real-Time PCR System (Thermo Scientific, Germany). Relative quantification was calculated by the ddCT method. For normalization, a housekeeper reference gene was used. Primer sequences are stated in Supplementary Table 2 .

Transcriptome generation and data preprocessing were conducted as described in Hirschberger et al. (36). In brief, sequencing libraries were prepared using the QuantSeq 3′ mRNA-Seq Library Prep Kit FWD for Illumina (Lexogen GmbH, Vienna, Austria), while the optimal number of amplification cycles was determined using the PCR Add-on kit for Illumina (Lexogen GmbH, Austria). Prior to sequencing on an Illumina HiSeq 4000 machine, the quantity and quality of libraries were assessed using the Quanti‐iT PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA, USA) and the Bioanalyzer High Sensitivity DNA Analysis Kit (Agilent Technologies, Inc., Santa Clara, CA, USA). An equimolar pool of libraries was prepared and sequenced in 150 bp paired-end mode. Preprocessing of data included adapter trimming, followed by alignment to the human genome (GRCh38) and counting of reads per gene. Only genes with a total read count greater than five times the sample size were kept in the data set. Generation of variance-stabilized expression values was performed using the DESeq2 R Bioconductor package (37).

Graphs: Data were visualized using mean values and standard deviations over all individual experiments. GraphPad Prism version 8.1 (GraphPad Software, San Diego, CA, USA) was used for statistical analysis. Student’s t-test statistics were applied with significant p-values < 0.05.

Area under curve (AUC): for correlation analysis, AUC was determined for each individual cell line on linear survival curves for the clonogenic survival data sets, and on linear time kinetics for the senescence data sets. The total area and SD values are given in Supplementary Tables 3, 4 .

RNA seq: Qlucore Omics Explorer v3.7 (Qlucore, Lund, Sweden) was used to analyze RNA sequencing data based on vst expressions. Principal component analysis (PCA) was performed on z-scaled data. Hierarchical clustering was used to visualize genes differentially expressed between irradiated and unirradiated samples (absolute log2-fold change > 1.5, Benjamini–Hochberg false discovery rate (FDR) < 0.05).

To understand the relevance of senescence and SASP in HNSCC cells, radioresistance was determined after two types of radiation with different efficacies in cytotoxic cell killing. Cells were exposed to 2, 4, 6, or 8 Gy photon or 0.25, 0.5, 1, 2, 3, or 4 Gy ¹²C-ion irradiation, and survival was scored via colony formation assay ( Figure 1A ). For the five cell lines used, there was a considerable variation in radioresistance, which was more pronounced after photon irradiation ( Supplementary Figure 1A ). To express the specific radioresistance by applying the entire dose–response curve, area under curve (AUC) values were calculated for the linear presentation of the data ( Supplementary Figure 1B , Supplementary Table 3 ). Cal33 and UPCI:SCC040 are the two most radioresistant cell lines after both radiation qualities, while Cal27 is the most sensitive cell line after photon but of intermediate radioresistance after ¹²C-ion irradiation, and the opposite is seen for UPCI:SCC131; UPCI:SCC099 is the second sensitive cell line for both photons and ¹²C-ions ( Figure 1B ).

Data were also used to calculate the relative biological effect at 10% survival (RBE10). The values obtained varied from 2.22 to 3.16 showing that on average ¹²C-ions are two to three times more effective than photons ( Figure 1C ), which is well in line with previous data (32).

The results obtained for photons are in good congruence with a previous study (9), separating a radioresistant (Cal33, UPCI:SCC040) from a rather radiosensitive cluster (other cell lines). Nonetheless, some differences in the detailed ranking are seen, which might be due to the fact that delayed plating was used instead of preplating and cellular cooperation is a strong determinant of assay performance (38, 39) but also that after photon irradiation the overall variation in radioresistance was rather small. To overcome this deficit, now also ¹²C-ions were used, enhancing the overall variation in radiation response.

DNA-DSBs are the most critical damage induced by irradiation when left un- or misrepaired. Therefore, DSB repair is considered to be a main terminator of radioresistance. DSB repair was studied after the isoeffective doses of 2 Gy photon and 1 Gy ¹²C-ion. DSBs were detected 4 and 24 h after irradiation via γH2AX/53BP1 foci colocalization ( Figure 2A ). For both radiation qualities, all cell lines exhibited an efficient repair as indicated by the significant reduction in foci with increasing repair incubation ( Figure 2B and Supplementary Figure 1C ). On average, the number of initial foci measured 4 h after irradiation was slightly lower for ¹²C-ion irradiation ( Figure 2C ), which is due to the lower physical dose. However, despite this difference in dose significantly more residual foci are found 24 h after ¹²C-ion irradiation, indicating a clearly less efficient DSB repair when compared to photons. This lower DSB repair efficiency is considered to contribute to the higher effect of ¹²C-ions on cell survival. For both radiation qualities, there was a moderate correlation of residual foci with radioresistance AUC ( Figure 2D ). The correlation did not reach significance but is in good agreement with previous data from our lab (32). In conclusion, residual foci are an indicator of radiosensitivity but are not sufficient to explain the variations in radiation response of the HNSCC cell-line panel.

Cellular senescence is a frequent event after irradiation, to prevent propagation of damaged cells. Due to metabolic changes, senescent cells switch their expression profiles toward inflammatory factors summarized as SASP, which might also induce radioresistance (40). To evaluate senescence, cells were irradiated with 2–6 Gy photon or 1–3 Gy ¹²C-ions and senescence-associated lysosomal β-galactosidase activity (34) was determined by flow cytometry over a period of 6 days after irradiation ( Supplementary Figure 2 ). The fraction of senescent cells increased from day 2 after irradiation over time and with dose but showed large variations between cell lines and radiation qualities ( Figure 3A ).

Senescence was quantified by calculating AUC ( Figure 3B ). Cal33 and UPCI:SCC040 showed a strong induction after both photon and ¹²C-ion, while UPCI:SCC099 presented intermediate senescence levels and UPCI:SCC131 and Cal27 were characterized by very low senescence ( Figures 3B, C ). Overall, there was a strong correlation for the senescence induced by these two types of irradiation ( Figure 3D ), with a much stronger effect after ¹²C-ion irradiation. For an identical dose of 2 Gy, senescence was about 7 times higher after ¹²C-ions when compared to photons ( Figure 3E ). The correlation of senescence AUC and radioresistance AUC revealed a strong and significant association of the two parameters for ¹²C-ions and similar results for photons ( Figure 3F ). It is noticeable that the slope of the association was much steeper for ¹²C-ion irradiation. Albeit the strong increase in senescence, radiosensitivity did not increase to the same extent. This suggests that overall senescence appeared to have a lower impact on radioresistance after ¹²C-ions than it did after photons.

Senescent cells are characterized by an altered expression profile summarized as SASP. Members of the SASP include chemokines and cytokines, but also growth factors, which are secreted and enable communication with the microenvironment (41). We used RNA sequencing data to examine the gene expression profiles of SASP family members in the cell line panel. The cell lines were irradiated with 8 Gy photons or 4 Gy ¹²C-ions or mock irradiated, and RNA was extracted 48 and 72 h after irradiation ( Figure 4A ). 3′ sequencing of the RNA (RNA-seq) was performed, data were preprocessed, and log2 values were extracted. After QA, from the RNA-seq dataset, 49 SASP genes could be extracted for further analysis ( Supplementary Table 4 ). PCA on z-scaled data revealed a cell-line-specific SASP gene expression profile ( Figure 4B ). It can also be observed that unirradiated samples with low SASP can be distinguished from irradiated samples with high SASP. Separation is not completely achieved for UPCI:SCC131 and UPCI:SCC099 cells. UPCI:SCC131 unirradiated cells express low but noticeable SASP levels and cannot be clearly distinguished from irradiated samples, while for UPCI:SCC099 SASP is not induced after irradiation and, as a consequence, all samples appear in the unirradiated cluster ( Figures 4B–D ). 8 Gy photons and 4 Gy ¹²C-ions induce highly similar expression patterns for almost all factors, and hence samples cannot be separated according to radiation quality. Using hierarchical clustering, irradiated samples can be separated from unirradiated samples in case of photons (q = 0.001; Figure 4C ) as well as ¹²C-ions (q = 0.001; Figure 4D ) by an 11-gene identifier. These genes are expressed significantly differently between irradiated and unirradiated samples with a fold change >1.5 ( Figure 4E ) and include members of the IL1 pathway, various chemokines, and growth factors. Correlation studies for IL1A ( Figure 4F ) and IL1B ( Figure 4G ) show strong associations for these genes with radioresistance and senescence, where the associations are always stronger for ¹²C-ions. This analysis demonstrates that IL1A and IL1B gene expression are strong indicators for the radioresistance of HPV-negative HNSCC cells. The result was confirmed using qRT-PCR analysis of independent samples ( Supplementary Figure 3A–C ).

Since both proteins are known to activate the identical pathway, we were interested to examine other relevant genes of the IL1 pathway. 25 genes could be extracted from the RNA-seq data set ( Supplementary Table 5 ), and similar to the analysis with the SASP genes, hierarchical clustering was able to separate irradiated from unirradiated samples for photons (q = 0.006; Supplementary Figure 4A ) and ¹²C-ions (q = 0.001; Supplementary Figure 4B ). Separation was excellent for Cal27, Cal33, and UPCI:SCC040. Also, the IL1 pathway gene expression was enhanced in unirradiated samples of UPCI:SCC131 and was low and not induced in irradiated UPCI:SCC099 cells. The 9-gene identifier showed a fold change above 1.5 and highly significant q values ( Supplementary Figure 4C ). For association analysis with survival and senescence, the mean over log2 values of all IL1 members was calibrated to the unirradiated control and used as a representative value for IL1 pathway activity. The associations obtained were strong and significant for both radiation qualities ( Figure 4H ).

We also observed that after photon irradiation with 8 Gy, IL1B protein secretion as detected by ELISA nicely coincided with gene expression data ( Figures 4G, I ). In all three cell lines, secretion of IL 1B increased with time after irradiation. In UPCI:SCC131 cells showing the lowest gene expression, also the lowest protein secretion was seen, and protein secretion was detectable in unirradiated samples, which was not the case for the two other cell lines. For Cal33 cells, where gene expression was strong, also protein secretion was strong and UPCI:SCC040 exhibited intermediate levels.

In conclusion, a strong and cell line-specific increase in expression of SASP factors can be detected after irradiation with photons and ¹²C-ions. Especially the IL1 pathway is highly associated with radioresistance and senescence of HPV-negative HNSCC cells.

Results up to now disclosed a role of IL1 as a strong indicator of radioresistance in HNSCC cells. It was tested whether the strong correlation with survival and senescence also implies that IL1 is functionally involved in these processes. To this end, specific siRNA oligonucleotides were used to perform a single or double knockdown of IL1A and IL1B in the HNSCC cell lines. Efficiency of the knockdown was validated by qRT-PCR and demonstrated a decrease in expression levels from 34% down to 5% ( Supplementary Figure 5A ).

Surprisingly, neither the IL1A and IL1B single nor double knockdown was found to have an effect on the cellular radioresistance, as determined by colony formation assay ( Figure 5A ). The result was identical for all cell lines examined, and for both radiation qualities. There was also no effect of the knockdown on the number of residual foci as found after irradiation with 2 Gy photon or 1 Gy ¹²C-ion ( Figure 5B ). Likewise, there was no effect of IL1B KD on irradiation-induced senescence ( Figure 5C ), whereby this knockdown was characterized by stable depression of both gene expression and protein secretion over the whole observation period of 5 days ( Supplementary Figure 5B, C ).

Overall, these data demonstrate for the first time that IL1A and IL1B, despite being strong indicators of radioresistance and senescence in HNSCC cells, are not functionally involved in these processes.