Oncogenic driver mutations predict outcome in a cohort of head and neck squamous cell carcinoma (HNSCC) patients within a clinical trial

234 diagnostic formalin-fixed paraffin-embedded (FFPE) blocks from homogeneously treated patients with locally advanced head and neck squamous cell carcinoma (HNSCC) within a multicentre phase III clinical trial were characterised. The mutational spectrum was examined by next generation sequencing in the 26 most frequent oncogenic drivers in cancer and correlated with treatment response and survival. Human papillomavirus (HPV) status was measured by p16INK4a immunohistochemistry in oropharyngeal tumours. Clinicopathological features and response to treatment were measured and compared with the sequencing results. The results indicated TP53 as the most mutated gene in locally advanced HNSCC. HPV-positive oropharyngeal tumours were less mutated than HPV-negative tumours in TP53 (p < 0.01). Mutational and HPV status influences patient survival, being mutated or HPV-negative tumours associated with poor overall survival (p < 0.05). No association was found between mutations and clinicopathological features. This study confirmed and expanded previously published genomic characterization data in HNSCC. Survival analysis showed that non-mutated HNSCC tumours associated with better prognosis and lack of mutations can be identified as an important biomarker in HNSCC. Frequent alterations in PI3K pathway in HPV-positive HNSCC could define a promising pathway for pharmacological intervention in this group of tumours.


Results
cohort characteristics. 234 FFPE blocks with diagnostic biopsies from HNSCC patients within a multicentre phase III clinical trial were incorporated in this study (Fig. S1). Clinical demographic factors such as age, gender, disease site and tumour stage are consistent between the whole cohort within the clinical trial and the subsequent random selection due to FFPE block availability in this study. Overall, most were from men (89.7%), with pharyngeal carcinoma (65.4%) and diagnosed in tumour stage IV-A (71.4%) with an average of 57 years old (Table 1). Clinicopathologic features by locations are shown in Table 1.

Association of mutations with clinical variables. General comparison of the mutational status and
tumour characteristics such as location, grade and histology, did not show any significant difference (p > 0.05) ( Table 3). However, considering variants of uncertain significance, women were associated with a lower percentage of mutation than men in our cohort (p = 0.002) ( Table 3).

Mutational profile and HPV presence in oropharyngeal tumours. HPV mutational profile in oro-
pharyngeal tumours is shown in Fig. 2. HPV-positive samples presented slightly more pathogenic mutations than HPV-negative (76.2% versus 69.2%, p = 0.762) ( Table 2). Despite the fact that TP53 was the most frequently mutated gene in both groups, these mutations were more recurrent in HPV-negative tumours (71.4% in HPVnegative and 30.8% in HPV-positive), difference statistically significant (p = 0.009). Conversely, the second most mutated gene, PIK3CA, although more represented in HPV-positive tumours (9.5% in HPV-negative versus 23.1% in HPV-positive), did not show any statistically significant difference (p = 0.178). While HPV-negative tumours did not present pathogenic mutations in other genes, PTEN was the third most commonly mutated in HPV-positive tumours (15.4%), followed by FBXW7 (7.7%).
Mutational status and response to treatment. After  www.nature.com/scientificreports/ Preliminary data indicated that the two regimens showed similar survival, response rates, toxicity and locoregional control 22 . For that reason, both arms were evaluated within the same group as final response (or response after randomization). Evaluation of the two time-point responses according to the mutational profile did not show any statistical difference (Table 4). There was, however, a tendency between mutated tumours and complete response at the end of the treatment taking VUS into consideration, p = 0.096 (Table S3). Considering only HPV profile in oropharyngeal tumours, there were no differences between HPV-positive and HPV-negative individuals either after induction chemotherapy (p = 0.396) or randomization (p = 0.914) ( Table 5).
Finally, an exploratory analysis was performed using the two most mutated genes in the study: TP53 and PIK3CA (Table S4). Analysing patients with mutations in those genes alone or within other genes and the clinical response, indicated that none of the TP53 subgroups were associated in any of the clinical trial treatment timepoints (p > 0.05). By contrast, considering only PIK3CA mutations, a statistically positive association was found in the complete response group after induction chemotherapy (p = 0.024). However, this finding was not corroborated in final response group (p = 0.235) (Table S4) what could suggest that this could be a false positive result taken into consideration multiple testing and sample size bias.
Poeta's 23 and Neskey's 24 classification in patients harbouring TP53 mutations in relation with clinical response before and after randomization did not show any statistically significant association (p > 0.05, Table S5).
A correlation with the number of mutations also showed that tumours with one mutation had lower OS (p = 0.038, HR = 1.544 (1.025-2.327)) than non-mutated patients, with the exception of PFS (p = 0.259, HR = 1.264 (0.842-1.898)) ( Fig. 3E,F). Equally, tumours with more than one mutation showed lower OS (p = 0.001, Table 1. Clinicopathologic characteristic of the 234 HNSCC patients included in the study: overall and by subtypes. TNM classification system stands for tumour, node and metastasis. SD standard deviation, Unk unknown.  www.nature.com/scientificreports/ HR = 2.524 (1.441-4.422)) and PFS (p = 0.036, HR = 1.824 (1.039-3.203)) than non-mutated samples. Conversely, the differences between tumours with one or more mutations were not statistically significant (p > 0.05). Finally, we compared TP53 mutations based on Poeta's 23 and Neskey's 24 stratification models with OS and PFS (Fig. S2). No association was observed between low-risk/high risk mutations or non-disruptive/disruptive mutations and survival in these patients (Fig. S2).

Discussion
As most of the head and neck cancers are diagnosed at a locally advanced stage the identification of biomarkers of response is a main goal to optimize treatment and reduce side effects. In recent years, induction chemotherapy has been shown to produce a benefit in organ preservation without a clear improvement in survival. In addition, this approach led to a high toxicity, particularly when concurrent radiotherapy was given with high doses of cisplatin. At present, very few predictive biomarkers of response have been described. For this reason, we proposed a study of the mutational status in 26 of the most common altered genes in cancer with next-generation sequencing in a homogeneously treated representative Spanish cohort of HNSCC from the phase III clinical trial TTCC-2007-01 22 .
The epidemiology characteristics of the HNSCC patients included in our study were similar to other series reported from the same region: the ratio between sexes is 9:1 in detriment of men, and most of the patients were diagnosed at stage IV 25 . p16 IHC, a surrogate of HPV infection in oropharyngeal tumours, showed that HPV was present in 17.1% of samples, a lower percentage than previously reported in Europe 26 but with similar location to other Southern European countries in oropharynx 27 .
Globally, the most mutated gene in our series was TP53 (61.1%). We observed a statistically significant lower percentage of mutated TP53 in HPV-positive oropharyngeal tumours (71.4%) than in HPV-negative (30.8%) as has been previously reported in HNSCC 28,29 . These results could be explained if TP53 sequestration by the viral oncoprotein E6 prevents gaining mutations in this gene under selective pressure of 30,31 . Comparing to other series, there was a higher percentage of TP53 mutations in HPV-positive tumours 29 . This fact could be explained by the coexistence of viral infection and other aetiological factors such as tobacco smoking and alcohol consumption Table 3. Association between mutational status and clinicopathological characteristics. TNM classification system stands for tumour, node and metastasis. SD standard deviation, Unk unknown. a Initially, statistical analysis was done comparing normal, pathogenic mutation and variants of uncertain clinical significance (VUS), p-value. b To avoid bias with VUS, a direct comparison only between normal and pathogenic mutation was done, p-value. www.nature.com/scientificreports/   www.nature.com/scientificreports/ during tumourigenesis 32 ; these data were not collected in this study. TCGA data described 85% of TP53 mutation in HPV-negative tumours and only 3% in HPV-positive ones 18 . However, the sample population was very different with a high predominance of oral cavity tumours (62%) and mainly heavy smokers. PI3K/AKT/mTOR has been reported as the most mutated pathway in HNSCC (13% to 56%), regardless of the HPV status 18 . PIK3CA gene, that encodes the catalytic subunit of the family, has been reported with an average mutational rate of 10.53% in HNSCC 33 , similar to the 10.25% found in this cohort, and with a higher frequency in laryngeal tumours 34 . Mutations in this gene have also been related to HPV-positive tumours 4 . Our results corroborate this fact, being PIK3CA more frequently mutated in HPV-positive tumours (23.1% versus 9.5% in HPV-negative oropharyngeal tumours), similar to previously described data 35 . We did not, however, see an increased percentage in laryngeal carcinoma. 73% of the mutations in PIK3CA are commonly located in 3 hotspots (E542K, E545K and H1047R/L) 36 , result also found in 76% of PIK3CA mutated samples in our study, emphasising the accuracy of using the targeted panel in HNSCC.
Mutations in FBXW7: an E3 ubiquitin ligase member of the F-box protein family, have been previously observed in HNSCC 19 . This tumour suppressor gene targets NOTCH1, being an important protein in cell proliferation control. Previous studies found FBXW7 mutated in 5% of HNSCC 37,38 and a higher percentage of mutations was previously considered as a prevalent event in HPV-positive tumours 39 . Our cohort confirmed these results in FBXW7 with a similar percentage only found in HPV-positive tumours (7.7%).
PTEN was the third most mutated gene in 15.4% of the HPV-positive oropharyngeal tumours while not mutations were found in HPV-negative ones. Contrary to our results, TCGA study showed PTEN mutated in 12% of HPV-negative tumours and 6% of HPV-positive 18 . Apart from PTEN, there were other genes which mutated at a lower percentage in our series, such as CKIT or CTNNB1 (both mutated at less than 1% and only in nonoropharyngeal HPV-negative tumours), have been reported in HNSCC in varied percentages 30,38,40 . Together with PIK3CA, our result enhances the hypothesis of higher prevalence of PI3K pathway activated mutations in HPV-positive tumours 41 .
Overall, excluding TP53 mutations, recurrent alterations in PIK3CA, PTEN and FBXW7 genes, all belonging to the PI3K/AKT/mTOR pathway, could define a potential new target for pharmacological intervention in HNSCC, as it has been suggested in other publications 42 .
In terms of survival, HPV-positive oropharyngeal tumours were associated with better prognosis, showing an increased OS and PFS compared to HPV-negative tumours as it was previously defined 26,[43][44][45][46][47] . Secondly, the presence of mutation in the targeted genes was associated with inferior outcome demonstrated by the presence of detrimental OS. These results could be an indirect measure of tumour aggressiveness, as has been reported in other series 43,47 . Moreover, the fact that carriers of tumours with more than one mutation have lower OS than those with non-mutated tumours reinforces this concept.
Lastly, there was a lack of association between mutational status and response after treatment. This can indicate that, excluding genetic-driven druggable targets, HNSCC mutational profile is not related to any clinical response but is a matter of mutational burden as is shown in the survival analyses. Similarly, there was no association between TP53 mutations stratified by Poeta's 23 and Neskey's models 24 and response to treatment or survival. These classification systems can serve as an important tool in individualizing and improving treatment for high TP53 mutated tumours, as it was previously identified in a subset of high-risk patients with a decreased response to platinum-based therapies 48 . Nevertheless, these classification models did not have any implication on outcome in our cohort.
Overall, our data strongly support and expand previously published studies exploring the presence and prognosis of mutations in this population. We have characterized the mutational profile of HPV-positive/HPVnegative oropharyngeal HNSCC in a representative cohort of patients. In this context apart from TP53 mutations, frequent alterations in PIK3CA, PTEN and FBXW7 genes, define possible pathways for pharmacological intervention. Finally, survival analysis showed that mutational status in the tumour could define patient prognosis, and may potentially be used as biomarkers to stratify patients for more intensive treatment. However, larger studies should be performed to confirm these results aiming at stratifying patients to different therapeutic interventions.

Methods
Samples. 234 FFPE blocks with diagnostic biopsies from HNSCC patients were included in this study. A consort diagram reporting the dropout is shown in Fig. S1. All samples belong to the clinical trial TTCC-2007-01 entitled: "Open label randomized, multi-centre phase III trial of TPF plus concomitant treatment with cisplatin and radiotherapy versus concomitant cetuximab and radiotherapy in locally advanced, unresectable head and neck cancer", ClinicalTrials.gov identifier: NCT00716391 22 .

TTCC-2007-01 trial design and data collection.
It was a non-inferiority, randomized and controlled study with a parallel assignment intervention model and an endpoint of safety/efficacy, carried out between 2008 and 2013. The follow-up of the clinical trial finished in November 2016. According to protocol, written informed consent was obtained from living subjects and the protocol was approved by the University Hospital of Salamanca and the ethical committees of each hospital in accordance with the 1964 Helsinki declaration and its later amendments.
Eligible patients: histologically or cytologically confirmed, previously untreated unresectable locally advanced (Stage III-IV) tumours (from oral cavity, oropharynx, larynx, hypopharynx), ECOG performance status 0-1. www.nature.com/scientificreports/ RT up to 70 Gy + cetuximab 400/250 mg/m 2 weekly until the completion of RT, and they were stratified by primary tumour site. Surgery after RT (neck dissection) was allowed. The primary endpoint was non-inferiority of cetuximab-radiotherapy versus cisplatin-radiotherapy in terms of overall survival. Response rate, loco-regional control and toxicity in both arms were considered secondary objectives. Preliminary data of this trial did not show any difference in terms of survival or response rates, toxicity and loco-regional control as secondary end points in the two regimens 22 . Clinical data were compiled in a case report form by medical oncologists involved in the clinical trial. All data were treated with the security measures established in compliance with the Protection of Personal Data Organic Law 15/1999, 13th December, and safe-keeping at the University Hospital of Salamanca in its specific server.
DNA extraction. Percentage of tumour cells was measured in haematoxylin-eosin tissue sections by central pathologist. Between four and ten 10 µm FFPE sections from diagnosis blocks were treated with deparaffinization solution (Qiagen, Hilden, Germany) and DNA extraction was done using QIAamp DNA FFPE Tissue kit (Qiagen, Hilden, Germany).
DNA quality evaluation and targeted NGS. Following TruSight Tumor 26 Reference Guide (Illumina, San Diego, USA), DNA quality was measured by qPCR. Comparing FFPE-gDNA amplification potential with a reference non-FFPE gDNA (QCT), delta Cq value was used to predict the dilution required for each sample.
TruSight Tumor 26 panel includes a set of 174 amplicons in complete exons of 26 cancer-associated genes (Table S1). This panel was selected due to its exceptional success rate using minimal DNA input even from FFPE samples where genetic material is often degraded. Following steps of hybridization with the oligo pool, removing unbound oligos and extension and ligation with bound oligos, an amplification of the libraries were performed. PCR products were checked on a 4% TBE agarose gel and finally the libraries were cleaned up by AMPure XP magnetic beads (Beckman Coulter, Brea, CA, USA). PCR products were quantified using Qubit Fluorometer (Invitrogen, Carlsbad, CA, USA) and libraries were normalized at 4 nM in a final pool. Sequencing was performed in a NextSeq 500 System (Illumina, San Diego, USA).
Data were transformed in BaseSpace platform and the VCF file format were read in the Variant Studio Software (Illumina, San Diego, USA). Following Illumina recommendations, somatic variants over 5% of frequency, with yields at least 1000 × cumulative coverage between the 2 strands and considered from the software of PASS filter were reported. Those variants of uncertain significance were considered pathogenic if at least two in silico prediction tools (SIFT and PolyPhen) classified them as deleterious/probably damaging 49 , and they were defined as likely pathogenic in the Catalogue Of Somatic Mutations in Cancer (COSMIC; https ://cance r.sange r.ac.uk/ cosmi c) or the National Center for Biotechnology Information (NCBI; https ://www.ncbi.nlm.nih.gov/clinv ar) databases.

Assessment of HpV status.
In the original study protocol, the assessment of HPV status was carried out by p16 immunohistochemistry (IHC), a surrogate marker for HPV infection 50 as the gold-standard technique. FFPE sections were deparaffinized and exposed to 10 mM citrate buffer antigen retrieval at 92 °C for 30 min and then they were stained using a p16 INK4a mouse monoclonal antibody (Cell Marque, Rocklin, CA, USA). Percentage of p16 staining was measured and only those tumours > 70% nuclear and cytoplasmic p16+ were considered positive. 33 samples were considered HPV-positive following this methodology: 13 oropharyngeal, 4 hypopharyngeal, 2 laryngeal and 9 oral cavity tumours. However, after the publication of the guidelines from the college of American pathologists, p16 IHC is only recommended in oropharyngeal tumours but other locations, where DNA/RNA viral determination should be performed as a confirmatory test 51 . Since there was not more DNA from all the samples after the library preparation, only oropharyngeal tumours with > 70% p16 positive staining were considered HPV-positive.

Statistical analyses. Statistical analysis compared categorical parameters and mutational status by the
Chi-square or Fisher's exact tests; while in continuous nonparametric variables, the Mann-Whitney U or Kruskal-Wallis H tests were used. p-values were calculated excluding missing values and they were considered statistically significant when p < 0.05. Significant variables were included in the logistic regression analysis and size effects were indicated by odds ratio (OR) with their 95% confidence interval (95% CI). Mutational status was classified as presence or absence of mutations, number of mutations (none, one or more than one) and the status of TP53 and PIK3CA (mutant or wild-type). Response was divided in two groups of treatment: after induction chemotherapy and after chemo/cetuximab plus radiotherapy (final response) due to the similar outcome in both arms 22 . Response was classified in both groups as complete response versus partial response/stabilization. No progressions were shown in the cohort.
Survival analysis was done according to the overall survival (OS) and progression-free survival (PFS) by Kaplan-Meier plots and log-rank test p-values were calculated in all the curves. Median was indicated in those plots in which it was achieved. Hazard-ratio was calculated to measure the risk of the event with its 95% confidence interval (95% CI) by Cox regression. Median follow-up in OS was 32.23 months while in PFS it was 15.31 months.
Due to high prevalence in TP53 mutations, we applied Poeta's and Neskey's classifications stratifying the mutations according to its change and functional effect, allowing a better comprehensive understanding on their relevance in clinical outcome. Following Poeta's classification 23 , TP53 mutations were divided in two categories: disruptive and non-disruptive according to their functional effects on the p53 protein. Additionally, according to Neskey's model 24 , also named as Evolutionary Action score of TP53-coding variants (EAp53), missense mutations Scientific RepoRtS | (2020) 10:16634 | https://doi.org/10.1038/s41598-020-72927-2 www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.