AKR1C3 is a biomarker and druggable target for oropharyngeal tumors

Abstract

Purpose

Oropharynx squamous cell carcinoma (OPSCC) is a subtype of head and neck squamous cell carcinoma (HNSCC) arising from the base of the tongue, lingual tonsils, tonsils, oropharynx or pharynx. The majority of HPV-positive OPSCCs has a good prognosis, but a fraction of them has a poor prognosis, similar to HPV-negative OPSCCs. An in-depth understanding of the molecular mechanisms underlying OPSCC is mandatory for the identification of novel prognostic biomarkers and/or novel therapeutic targets.

Methods

14 HPV-positive and 15 HPV-negative OPSCCs with 5-year follow-up information were subjected to gene expression profiling and, subsequently, compared to three extensive published OPSCC cohorts to define robust biomarkers for HPV-negative lesions. Validation of Aldo-keto-reductases 1C3 (AKR1C3) by qRT-PCR was carried out on an independent cohort (n = 111) of OPSCC cases. In addition, OPSCC cell lines Fadu and Cal-27 were treated with Cisplatin and/or specific AKR1C3 inhibitors to assess their (combined) therapeutic effects.

Results

Gene set enrichment analysis (GSEA) on the four datasets revealed that the genes down-regulated in HPV-negative samples were mainly involved in immune system, whereas those up-regulated mainly in glutathione derivative biosynthetic and xenobiotic metabolic processes. A panel of 30 robust HPV-associated transcripts was identified, with AKR1C3 as top-overexpressed transcript in HPV-negative samples. AKR1C3 expression in 111 independent OPSCC cases positively correlated with a worse survival, both in the entire cohort and in HPV-positive samples. Pretreatment with a selective AKR1C3 inhibitor potentiated the effect of Cisplatin in OPSCC cells exhibiting higher basal AKR1C3 expression levels.

Conclusions

We identified AKR1C3 as a potential prognostic biomarker in OPSCC and as a potential drug target whose inhibition can potentiate the effect of Cisplatin.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13402-020-00571-z.

Introduction

Head and neck squamous cell carcinomas (HNSCCs) arise in all anatomical areas of the upper aero-digestive tract, i.e., pharynx, salivary glands, oral cavity, larynx, nasal cavity and paranasal sinuses [1]. They represent the seventh most common cancer in the USA, accounting for more than half a million new patients each year [2]. The 5-year survival rate from diagnosis is ~50%, despite improvements that have been made in both surgical techniques and other therapeutic options. In localized tumors, surgery is the preferred therapy, being the only potentially curative approach. It is frequently combined with chemo-radiotherapy, which remains the standard of care for locally advanced tumors [3]. Combination of anti-EGFR cetuximab with standard therapeutic compounds (i.e., platinum compounds, taxanes and 5-FU) is highly recommended for the treatment of recurrent or metastatic tumors. It is also administered as maintenance therapy, showing good results in terms of overall response rate (ORR), disease-free survival (DFS), progression-free survival (PFS) and overall survival (OS) [4–6], even though these approaches do not have durable effects in most patients. Recently, the use of checkpoint inhibitors as single agents has been tested in this tumor setting. In particular, the administration of nivolumab instead of standard chemotherapy considerably improved the quality of life and the OS [7, 8]. It has been found, however, that only subgroups of patients with specific molecular characteristics and with particular clinical features may benefit from this therapeutic option.

Oropharyngeal squamous cell carcinomas (OPSCCs) arise from the base of the tongue, lingual tonsils, tonsils, oropharynx, pharynx and the Waldeyer ring [9]. Besides the traditional risk factors (alcohol and tobacco consumption, dietary regimens poor of fresh fruit and vegetables, advanced age), Human Papilloma Virus (HPV) infection has been found to affect a subgroup of OPSCC patients, usually non-smokers and non-alcohol consumers [10, 11]. The HPV16 and 18 genotypes are mainly involved in the development of these tumors, as their integration induces malignant transformation of infected cells through the action of E6 and E7 oncoproteins, located in the early region of the viral genome, which promote tumor growth by inactivating the p53 and Rb tumor suppressor pathways [12, 13]. On these bases, the actual trend is to consider HPV-related and non-related OPSCCs as two different entities, with distinct molecular aberrations and clinical prognoses, supported by evidence indicating that HPV-positive tumors show increased local control as well as better DFS and OS rates compared to HPV-negative tumors [14]. Moreover, deep sequencing has revealed that HPV-positive tumors are enriched in mutations in the PIK3CA, FGFR3, TRAF3 and E2F1 genes, amplifications of the PIK3CA and FGFR3 genes, and aberrations in DNA repair-related genes (BRCA1, BRCA2, ATM, FANCG, FANCA, FANCD2 and RAD51B) [15]. By contrast, HPV-negative tumors are characterized by TP53 mutations, CCND1 amplifications, and TERT promoter and NOTCH pathway gene alterations [16]. These molecular differences have been substantiated by gene expression studies [17], showing that transcription factors and genes involved in cell cycle regulation and DNA repair are generally deregulated in HPV-positive tumors. Recurrent fusions of protein kinases have also been reported. In particular FGFR3-TACC3 fusions, which have a direct impact on the Akt/mTOR/PTEN pathway, have been found in HPV-positive samples, while rearrangements in the NTRK2 and NTRK3 genes have been encountered in HPV-negative samples only [15, 18].

Recently, attention has been focused on the identification of biomarkers associated with the clinical outcome of distinct molecular HNSCC subgroups [19], but only a few studies have been aimed at the stratification of OPSCC patients in terms of survival and sensitivity to therapy taking into account HPV infection. Keck and collaborators described two different subtypes within HPV-positive cases (one of them partially overlapping HPV-negative cases) and three subgroups within HPV-negative cases [20]. It has recently been shown that OPSCCs with integrated HPV16 may exhibit upregulation of both aldo-keto-reductases 1C1 and 1C3 (AKR1C1 and AKR1C3) expression, strongly correlating with poor survival rates [21, 22].

In this work, we analyze four independent OPSCC datasets to derive a panel of robust markers in relation to HPV status. Furthermore, we investigate the role of AKR1C3 as putative target in OPSCCs that may develop resistance to chemo/radiotherapy, and test the efficacy of specific inhibitors of AKR1C3, alone or in combination with the standard chemotherapeutic agent Cisplatin.

Materials and methods

Patients and samples

The study includes 29 cases of OPSCC (cohort A), collected from three centres of the Piedmont region (Novara, Vercelli and Biella Hospitals, Italy) in the period 2006–2015. The research was approved by the Ethical Committee of Novara University Hospital (Protocol No. CE42/2011). Clinical, pathological, HPV infection, alcohol and smoke data, as well as follow-up information, are reported in Table 1. The median patient age was 60 years (range 44–86 years).

Table 1.

Clinical and pathological characteristics of the Piedmont cohort

Patient ID	Gender	Age at diagnosis	HPV	Follow-up	Therapy	Alcohol	Smoke	TNM V8	Site	Time to death/last FU
BI100	M	60	Neg	DCD	S-RT	Yes	Yes	T2 N2aM0	tonsil	2892
BI102	M	50	Neg	DCD	S-RT/CT	Yes	Yes	T2 N2cM0	tonsil	482
BI103	M	62	Neg	REL	CT-RT	NA	NA	T1 N1M0	soft palate	1955
NO300	M	86	Neg	DCD	palliative	cYaeres	Yes	cT4bN2cM1	right tonsil	545
NO301	M	52	Neg	DCD	S-CT	Yes	Yes	pT1N0	right tonsil	984
NO302	M	55	Pos	NED	CT-RT	Yes	No	cT3 N2a	right tonsil	1081
NO303	M	51	Pos	NED	CT-RT	No	Yes	cT2N2c	left tonsil	1795
NO304	M	58	Neg	DCD	palliative	cYaeres	Yes	cT4bN3M1	right tonsil	168
NO305	M	76	Pos	NED	S/CT-RT	Yes	No	pT1 N2b	left tonsil	608
NO306	M	73	Pos	NED	CT-RT	Yes	No	cT4aN2b	right tonsil	1646
NO307	F	54	Pos	NED	CT-RT	No	Yes	cT2N0	tonsil/soft palate	1720
NO310	M	62	Neg	DCD	S/CT	No	Yes	cT1N0M2	right tonsil	673
NO312	M	80	Pos	NED	CT-RT	Yes	No	pT2N1Mx	right tonsil	1530
NO313	M	60	Neg	NA	NA	NA	Yes	cT4N0Mx	right tonsil	NA
VC200	M	77	Pos	NED	S-RT	Yes	No	cT3N2b	left tonsil	180
VC201	M	44	Neg	NED	RT	Yes	No	T2N0M0	soft/hard palate	317
VC202	M	61	Neg	REL	RT	Yes	Yes	T3N0M0	soft/hard palate	60
VC203	F	51	Neg	DCD	CT-RT	Yes	Yes	T2N12M0	left tonsil	579
VC204	M	51	Neg	NED	CT-RT	No	Yes	T3N0	right tonsil	450
VC205	M	54	Pos	NED	RT	Yes	Yes	T2N0	left tonsil/soft palate	604
VC206	M	59	Neg	DCD	palliative	cYaeres	Yes	T3N3M0	left tonsil/soft palate	51
VC207	M	45	Pos	NED	S-RT/2S-IT	Yes	Yes	pT2N2b	right tonsil	720
VC208	F	62	Pos	NED	S/RT	Yes	Yes	pT2N1	right tonsil	1007
VC209	M	69	Pos	NA	NA	Yes	No	T3N2a	right tonsil	NA
VC210	M	55	Pos	NED	CT-RT	Yes	Yes	cT4N2b	left tonsil	932
VC211	M	63	Neg	NED	RT	Yes	Yes	T3N0 + T2N0	soft palate	908
VC212	F	62	Pos	NED	CT-RT	No	Yes	T3N1	soft palate	828
VC213	F	60	Pos	DCD	CT-IT	Yes	Yes	T4bN2b	base tongue/left tonsil	177
VC214	M	66	Neg	DCD	RT-CT	Yes	Yes	pT2N2bM0	right tonsil	355

M male, F female, NED non evidence of disease, DCD deceased, REL relapser, S surgery, CT chemotherapy, RT radiotherapy, IT Immunotherapy, Pos positive, Neg negative, NA Not available

An independent cohort of OPSCC patients (cohort B, n = 111) was collected from Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. Clinical pathological characteristics of these patients are summarized in Online Resource 1-ESM_1. The patients were treated from 2008 to 2017 at Fondazione IRCCS Istituto Nazionale dei Tumori (Milan, Italy) and include the BD2Decide observational study cohort (ClinicalTrials.gov identifier NCT02832102). The research was approved by the Ethical Committee of Fondazione IRCCS Istituto Nazionale dei Tumori of Milan (Protocol No. INT65–16/INT66–16). All research was performed in accordance with the declaration of Helsinki and all patients included in the study signed an informed consent.

RNA extraction

Fresh tissues from cohort A obtained immediately after resection were preserved in RNA later and stored at −20 °C. Briefly, tissues were fragmented through mechanical homogenization by TissueLyser LT (Qiagen) and the addition of TRI Reagent (Sigma Aldrich). Total RNA isolation was performed using an Absolutely RNA miRNA kit (Agilent Technologies) following the manufacturer’s instructions. The quality of the RNA was assessed by capillary electrophoresis using a Bioanalyzer 2100 (Agilent Technologies). The purity and quantity of the RNA were assessed using a NanoDrop spectrophotometer (Thermo Scientific).

Gene expression profiling

Five hundred ng total RNA of each patient sample and Human Universal Reference Total RNA (Clontech, Mountain View, CA, USA) were amplified using an Amino Allyl MessageAmp I aRNA Kit (Ambion, Austin, TX, USA), as previously described [23]. Labeling was performed using cyanine 3 (IRIS 3) and cyanine 5 (IRIS 5) (Molecular Targeting Technologies, Inc). One μg purified cRNA from the patient samples and reference were hybridized to oligo glass two-color arrays using Agilent Technologies SurePrint G3 Human Gene Expression Microarrays 4X44K v2. A dye-swap replication procedure was applied. Images were scanned using an Agilent C dual-laser microarray scanner, and analysed using Agilent Feature Extraction (FE) Software (Agilent Technologies). Microarray data were analysed using the R limma package.

Functional enrichment analysis

Gene Ontology (GO) biological processes level 5 enrichment within selected transcript lists was assessed using the DAVID functional annotation tool (https://david.ncifcrf.gov). MetaCore software (Clarivate Analytics) was used to carry out process network and pathway map enrichment analyses.

External datasets

Our initial results were compared with those obtained from independent datasets available on-line or from previously published meta-analyses. Specifically, the OPSCC cohorts contained in the TCGA-HNSCC, GSE40774, [20] and GSE65858 [24] datasets were re-analyzed and the results were compared to ours.

qRT-PCR validation in an external dataset

The TaqMan gene primer sets used were purchased from ThermoFisher (ThermoFisher, Carlsbad, CA, USA). RT-qPCR data for AKR1C3 (ID: Hs00366267_m1) were normalized using BTF3 (ID: Hs00852566_g1), ACTB (Hs01060665_g1) and MTCH1 (Hs01077107_g1) as housekeeping genes. Reverse transcription and PCR amplification were performed using a High Capacity cDNA Reverse Transcription kit and a Taqman Universal PCR Master Mix no Amperase UNG, respectively, following the manufacturer’s instructions (ThermoFisher, Carlsbad, CA, USA). Ten μl cDNA reaction volumes were assembled in a 384-well plate using a Janus Varispan 8-tip Liquid Handler (PerkinElmer, Waltham, MA, USA) and amplified as follows: 95 °C for 10 min, followed by 45 cycles at 95 °C for 15 s and 60 °C for 60 s. Gene expression was quantified using a detection system (QuantStudio 12 K Flex Real Time PCR System; ThermoFisher, Carlsbad, CA, USA), after which the threshold cycle (C_T) for each sample was determined. QuantStudio 12 K Flex software v1.2.3 (ThermoFisher, Carlsbad, CA, USA) was used to recover the data. Data analysis was conducted using the comparative -ΔC_T method.

Differences in -ΔC_T values between HPV-negative and HPV-positive patients were assessed using a Wilcoxon rank-sum test. For survival analysis, we stratified the validation cohort into three groups: HPV-negative patients and HPV-positive patients with high (higher than the median) or low (lower than the median) AKR1C3 expression. Disease-free and overall survival curves were established using the Kaplan-Meier method and their differences were assessed using a log-rank test. α = 0.05 was set as level of significance for the statistical tests.

Cell lines, AKR1C3 inhibitors and Cisplatin

Two OPSCC HPV-negative cell lines (Fadu, Cal-27) were kindly gifted by Prof. Trusolino. The Fadu cells, derived from a patient with pharynx squamous cell carcinoma, were cultured in MEM supplemented with penicillin/streptomycin (P/S) 0.1%, 10% fetal bovine serum (FBS), 1 mM Na-pyruvate and 1x non-essential amino acid solution. The Cal-27 cells, derived from a patient with squamous cell carcinoma of the tongue, were cultured in DMEM high glucose supplemented with 0.1% P/S and 10% FBS.

The AKR1C3 inhibitors used were synthetized by the Department of Drug Science and Technology, University of Turin (Italy). The compounds, named MEDS461 and MEDS436, have previously been described as compound 3 and 6 in [25]. The respective AKR1C3 inhibitors were dissolved in DMSO after which stock solutions at 20 mM were prepared and stored at −20 °C. Cisplatin (CIS) was purchased from Pfizer and stored at −20 °C at a concentration of 1 mM.

Cell treatment and viability assays

Cells were seeded at a density of 3000 cells/well in 96 multi-well plates in media containing 10% FBS. After 24 h, various dilutions of compounds MEDS461 and MEDS436 (from 100 μM to 6.25 μM in appropriate complete medium) and of CIS (from 10 μM to 0.625 μM in appropriate complete medium) were prepared, the culture media were withdrawn and the drugs in different dilutions were added. For co-treatment experiments, various dilutions containing both compound MEDS461 and CIS at the doses indicated above (from the highest to the lowest) were prepared in complete medium and, after medium withdrawal, added to the cells and incubated for 72 h at 37 °C. For sequential treatment experiments, cells were pre-incubated for 24 h with MEDS461 (at 40 μM or 70 μM concentrations, corresponding to the IC50 values for Fadu and Cal-27 cells, respectively) and, after medium withdrawal, incubated for another 48 h with CIS at the dilutions indicated above (from 10 μM to 0.625 μM). For each experiment, control cells (untreated, NT, in triplicate) were incubated with the same final concentration of DMSO (maximum concentration 0.5% v/v).

Cell viabilities were assessed using a Cell TiterGlo assay (Promega). Briefly, media were withdrawn and 100 μl/well Cell TiterGlo solution was added to the cells. After 10 min of incubation, chemiluminescence was read and optical densities (OD) representing ATP production were evaluated. OD value ratios of treated and non-treated (NT) cells for each dilution were calculated, after which the % of viable cells was estimated. For statistical analysis, GraphPad-Prism software was used. One-way ANOVA with Tukey test was used to compare responses at different doses of drugs within the same cell line, with a p value < 0.05 considered as statistically significant (confidence interval 95%). Data are expressed as mean ± standard deviation (SD) of values obtained from at least triplicate assays. Asterisks indicate significant p values (* p < 0.05 **, p < 0.01, *** p < 0.001, **** p < 0.0001). IC50 values were determined using the Chou Talalay method within the Calcusyn software tool.

Western blot analysis

Cells were incubated with MEDS461 (40 μM or 70 μM, in case of Fadu or Cal-27 cells, respectively) for 1, 3 and 24 h. Treated and untreated cells were lysed on ice using RIPA Buffer 1x (ThermoFisher Scientific) supplemented with protease/phosphatase inhibitor cocktails. After clarification, protein concentrations were evaluated using the BCA method (Euroclone). Thirty μg of the proteins were loaded on polyacrylamide mini Protean precast gels (10%; Bio-Rad) and transferred to 0.45 μm nitrocellulose membranes (Bio-Rad) using Trans Turbo Blot (Bio-Rad). The membranes were saturated using Pierce Clear Milk Blocking Buffer 10X (ThermoFisher Scientific), diluted 1x in TBS-Tween for 1 h at room temperature, and next incubated overnight at 4 °C with specific primary antibodies. The following primary antibodies were used: mouse monoclonal anti-AKR1C3 (1:1000, Sigma) and mouse monoclonal anti-actin (1:3000, Sigma). The secondary antibodies used were HRP anti-rabbit (1:5000, Sigma) and HRP anti-mouse (1:5000, Sigma). Next, the protein bands were detected using Liteablot Turbo Chemiluminescent Substrate (Euroclone) and acquired using a BioRad Chemidoc Image System (Bio-Rad). Densitometric analysis of the bands was performed using Quantity-One software (BioRad). The intensities of the actin bands were normalized to the highest intensity value to obtain the normalization factor for each lane. This factor was used to calculate the intensity of each AKR1C3 band with the following formula: AKR1C3 intensity value/normalization factor.

AKR1C3 silencing

Fadu cells were seeded at a concentration of 350,000/well in 6 well-plates in the appropriate media. After 24 h, transient silencing was performed using oligofectamine (Life Technologies, ThermoFisher) with the addition of AKR1C3 siRNA (Ambion, ThermoFisher), or negative control siRNA (Life Technologies, ThermoFisher) following the manufacturer’s instructions. Untreated cells (only OPTIMEM medium) were used as reference. Effective RNA and protein expression knockdown was checked after 24 h by qRT-PCR and Western blotting, respectively. To this end, cells were washed twice with PBS and lysed in 1 ml TRIZOL (Qiagen). Total RNA was extracted from cells at all conditions using an Absolutely RNA miRNA kit (Agilent Technologies). Five hundred ng total RNA was reverse transcribed using a High Capacity cDNA reverse transcription kit (Applied Biosystems). Subsequent qRT-PCR was performed using SybrGreen (Promega) to analyze the expression levels of AKR1C3 and the housekeeping gene BTF3 using a CFX96 Biorad qPCR machine. The analyses were performed from three independent experiments. The housekeeping gene BTF3 was chosen from the microarray-based expression data as one of the most stable genes across the cohort of patients analyzed. Quantitative analyses were performed using the comparative -ΔC_T method. To assess AKR1C3 protein knockdown, the cells were lysed after 24 h and processed for Western blotting as described above.

Results

Gene expression profiling of OPSCC and association with HPV infection

We first analyzed the gene expression profile of our cohort of OPSCC samples and applied unsupervised two-dimensional hierarchical clustering to group the OPSCC samples based on similarities in their expression pattern (Online Resource 2-ESM_2). Remarkably, probably due to the small number of samples included in the study, this analysis did not highlight any group of genes able to stratify the samples according to specific characteristics such as HPV infection, outcome, gender, age, and alcohol or tobacco consumption. The R limma package, that combines linear models with empirical Bayes analysis, was next used to compare expression profiles of the HPV-negative and HPV-positive samples. After filtering the results for |fold-change| > 1.5 and adjusted p value < 0.01, we identified 848 differentially expressed genes, of which 348 were down- and 500 were up-regulated in the HPV-negative OPSCCs (Online Resource 3-ESM_3). Subsequent gene ontology (GO) analysis revealed several biological processes with significant enrichment in the list of downregulated transcripts, of which the first 20 are shown in Fig. 1a. They can be grouped into a few classes: immune system, muscle-related processes, response to stimuli, actin organization, tissue development and adhesion (Fig. 1b). Transcripts involved in enriched process networks and pathway maps are listed in Tables 2 and 3, respectively. The first 20 biological processes overrepresented in the list of upregulated genes (Fig. 2a) can also be grouped into few classes: response/transport, migration/proliferation, biosynthetic and metabolic processes, and Wnt pathway (Fig. 2b). Transcripts involved in enriched process networks and pathway maps are listed in Tables 4 and 5, respectively.

Fig. 1.

a The first 20 GO biological processes (level 5) overrepresented within the list of downregulated transcripts. b Distribution of all significantly enriched biological processes (p value < 0.01)

Table 2.

Networks overrepresented within down-regulated genes in HPV-negative vs HPV-positive samples

Networks	Genes	p value
Muscle contraction	beta-MHC, MYBPC1, MYBPC2, MELC, Galpha(q)-specific peptide GPCRs, Myosin II, MRLC, MYBPH, Myotilin, ACTC, Galpha(i)-specific peptide GPCRs, MLC2, Telethonin, Actin muscle, MyHC, PKC, MuRF1, Titin, ACTA1, Actin	2.18 × 10−10
Development_Skeletal muscle development	beta-MHC, MYBPC1, MYBPC2, MELC, MYF6, CRP3 (MLP), Myosin II, MRLC, MYBPH, Myotilin, MLC2, Actin muscle, MKL1, MyHC, MuRF1, Titin, ACTA1, Actin	3.26 × 10−8
Cytoskeleton_Regulation of cytoskeleton rearrangement	Galpha(i)-specific amine GPCRs, MELC, RhoGAP4, SDF-1, VAV-3, Myosin II, MRLC, ACTC, Actin muscle, MyHC, PKC, Advillin (p92), ACTA1, Actin	1.97 × 10−4
Cell adhesion_Leucocyte chemotaxis	RASSF5, CCL19, CXCR5, VCAM1, ICAM1, Galpha(q)-specific peptide GPCRs, SDF-1, ZAP70, CCL21, Galpha(i)-specific peptide GPCRs, CCR7, MHC class II	2.95 × 10−4
Cell adhesion_Platelet aggregation	GP-IB alpha, PLA2, MELC, ZAP70, VAV-3, Myosin II, MRLC, MyHC, SLC21A2, Alpha-2A adrenergic receptor, cPLA2	3.84 × 10−4
Cell cycle_G1-S Growth factor regulation	IGF-2, p15, TGF-beta 3, PKC-theta, p16INK4, FGFR1, RelB (NF-kB subunit), PKC, Tcf(Lef), TGF-beta, p14ARF	2.02 × 10−3
Immune response_Antigen presentation	CD8 alpha, CD8, ICAM1, MHC class II beta chain, CD1b, CD1d, HLA-DQA1, RelB (NF-kB subunit), HLA-DRB1, CD1a, MHC class II	2.19 × 10−3
Immune response_TCR signaling	CD8 alpha, SIT1, CD8, ICAM1, ZAP70, PKC-theta, TRAF1, Bcl-2, MHC class II, Actin	2.81 × 10−3
Cytoskeleton_Actin filaments	MELC, SDF-1, Myosin II, MRLC, ACTC, MLC2, Actin muscle, MyHC, ACTA1, Actin	3.05 × 10−3
Cell adhesion_Platelet-endothelium-leucocyte interactions	LEKTI, GP-IB alpha, VCAM1, ICAM1, Multimerin, PAI2, TGF-beta 3, VLDLR, TGF-beta	8.93 × 10−3
Cardiac development_BMP_TGF_beta_signaling	beta-MHC, TGF-beta 3, CRP3 (MLP), ACTC, MLC2, MyHC, PDLIM3	9.67 × 10−3

Table 3.

Maps overrepresented within down-regulated genes in HPV-negative vs HPV-positive samples

Maps	Genes	p value
T cell generation in COPD	Langerin, CCL19, CD8, ICAM1, CCL21, CCR7, CD1a, CD1c, MHC class II	2.18 × 10−11
Role of iNKT and B cells in T cell recruitment in allergic contact dermatitis	AID, CXCR5, VCAM1, ICAM1, CD1d, CD5, CD21, CD79 complex, CD19	1.92 × 10−9
Maturation and migration of dendritic cells in skin sensitization	MHC class II alpha chain, CCL19, ICAM1, MHC class II beta chain, HLA-DRB, HLA-DRB5, CCR7, HLA-DRB1, MHC class II	3.01 × 10−9
Skeletal muscle atrophy in COPD	beta-MHC, MYBPC1, MELC, Myosin-IIA, MRLC, MLC1F, MLC2, MyHC, MuRF1	9.03 × 10−8
Differences between Langerhans cells and dermal dendritic cells in allergic contact dermatitis	CCL19, CD8, SDF-1, CCL21, CCR7, MHC class II	3.67 × 10−7
Cell adhesion_Integrin inside-out signaling in T cells	RASSF5, CCL19, VCAM1, ICAM1, SDF-1, ZAP70, CCL21, CCR7	4.49 × 10−7
Immune response_Antigen presentation by MHC class II	MHC class II alpha chain, Langerin, MHC class II beta chain, ARL14, CD79B, Bcl-2, PKC, CD79A, Cathepsin V, MHC class II, CD79 complex	5.08 × 10−7
CHDI_Correlations from Replication data_Cytoskeleton and adhesion module	MHC class II alpha chain, Langerin, MHC class II beta chain, ARL14, CD79B, Bcl-2, PKC, CD79A, Cathepsin V, MHC class II, CD79 complex	2.3 × 10−6
Dysregulation of germinal center response in SLE	AID, CXCR5, VCAM1, ICAM1, CD79B, RGS13, CD21, CD79 complex	2.6 × 10−6
Populations of skin dendritic cells involved in contact hypersensitivity	Langerin, CD1d, CCR7, CD1a, MHC class II	3.27 × 10−6

Fig. 2.

a The first 20 GO biological processes (level 5) overrepresented within the list of upregulated transcripts. b Distribution of all significantly enriched biological processes (p value < 0.01)

Table 4.

Networks overrepresented within up-regulated genes in HPV-negative vs HPV-positive samples

Networks	Genes	p value
Response to hypoxia and oxidative stress	GSTM1, HSP90 beta, GPX2, GSTA3, GSTM2, GSTM3, GSTA1, GSTA4, GSTA2, GSTT2, GSTO2, GSTA5, GSTM5	1.62 × 10−4
Cell adhesion_Platelet-endothelium-leucocyte interactions	EGFR, PAI1, OLR1, IL-6, JAM1, ITGA2, IL-8, EGF, ENA-78, Collagen IV	0.01
DNA damage_Checkpoint	14–3-3 zeta/delta, ATF-3, Cyclin A1, Cyclin A, RIF1, Cyclin D, Cyclin D1, 14–3-3	0. 01
Development_Regulation of angiogenesis	EGFR, PAI1, HB-EGF, TrkB, IL-6, Galpha(q)-specific peptide GPCRs, Pitx2, GLI-1, IP3 receptor, IL-8, Clusterin	0.02
Immune response_Th17-derived cytokines	Mucin 5B, IL-6, IL-8, ENA-78, CCL7, GCP2	0.03
Reproduction_Male sex differentiation	EGFR, Histone H2, HB-EGF, BMP7, TCP1-zeta-2, GLI-1, OCA2, Cyclin A1, Cyclin A, EGF, Clusterin	0.03
Cytoskeleton_Regulation of cytoskeleton rearrangement	Galpha(i)-specific amine GPCRs, BPAG1, 14–3-3 zeta/delta, Profilin, MLCK, MRCK, MyHC, Profilin II, 14–3-3	0.03
Proliferation_Negative regulation of cell proliferation	PTHrP, IBP2, TFF1, IL-6, GPNMB (Osteoactivin), IBP, Pitx2, IL-8, Cyclin D1	0.03
Reproduction_Progesterone signaling	EGFR, TFF1, AKR1C3, Galpha(q)-specific peptide GPCRs, Pitx2, Dynein 1, cytoplasmic, intermediate chains, DYNC1I1, IP3 receptor, Cyclin A, EGF	0.04
Signal transduction_ESR2 pathway	EGFR, TFF1, AKR1C3, EGF, ERAP140	0.04

Table 5.

Maps overrepresented within up-regulated genes in HPV-negative vs HPV-positive samples

Maps	Genes	p value
Glutathione metabolism	GSTM1, GSTA3, GSTM2, GSTM3, GSTA1, GSTT2, GSTA4, GSTA5, GSTM5, GSTA2	3.01 × 10−9
Androstenedione and testosterone biosynthesis and metabolism p.2	UGT1A9, UGT1A8, UGT1A10, AKR1C3, UGT1A4, AKR1C4, AKR1C1	1.06 × 10−6
TGF-beta signaling via SMADs in breast cancer	PTHrP, HMGA2, PAI1, JAM1, ATF-3, Cyclin A1, IL-8, CUX1	1.35 × 10−6
G protein-coupled receptors signaling in lung cancer	EGFR, NT, HB-EGF, Galanin, Galpha(q)-specific peptide GPCRs, IL-8, EGF, HB-EGF(mature), Cyclin D1	6.76 × 10−6
Calcium-dependent regulation of normal and asthmatic smooth muscle contraction	PAR2, TrkB, FKBP1B, MLCK, IP3 receptor, MYLK1, ACM3, ACM2	2.59 × 10−5
NRF2 regulation of oxidative stress response	GPX2, GSTA3, GSTM3, GSTA1, GSTA2, SLC7A11, NQO1	4.04 × 10−5
EGFR signaling pathway in colorectal cancer	PAR2, EGFR, NT, HB-EGF, ACM3, EGF, Cyclin D1	8.67 × 10−5
Androgen receptor activation and downstream signaling in Prostate cancer	EGFR, IL-6, FOLH1 (GCP2), STEAP1, ER81, IL-8, EGF, Cyclin D1, Clusterin	1.31 × 10−4
Cigarette smoke-mediated regulation of NRF2-antioxidant pathway in airway epithelial cells	ME1, GPX2, GSTA1, UGT1A4, NQO1	1.31 × 10−4
Neuroendocrine transdifferentiation in Prostate Cancer	PTHrP, EGFR, NT, HB-EGF, IL-6, IL-8	1.72 × 10−4

Next, we carried out the same analysis on the TCGA RNA-seq dataset, selecting only patients affected by OPSCC and for whom HPV status information was available. This cohort (OPSCC-TCGA) consists of 30 HPV-negative and 59 HPV-positive samples. GO functional enrichment analysis of the list of downregulated transcripts showed an overlap of biological processes involved in immune response activation and regulation (Online Resource 4-ESM_4). Migration, metabolic processes, Wnt pathway, differentiation, secretion and signaling were confirmed among the biological processes overrepresented within the upregulated transcripts (Fig. 3).

Fig. 3.

Significantly overrepresented biological processes within transcripts upregulated in HPV-negative OPSCC, in common between our data and the TCGA OPSCC dataset

We subsequently analyzed the OPSCC subgroup of the GSE40774 dataset used by Keck and collaborators [20] to generate an 821-gene signature able to classify HNSCCs in five distinct subtypes, using 76 OPSCCs (22 classified as HPV-negative and 54 as HPV-positive). Significant enrichment of immune system-related processes for the downregulated genes and for metabolic/xenobiotic processes, as well as response to stimuli for the upregulated genes, were again confirmed (Online Resource 5-ESM_5, Panel A and panel B, respectively).

Comparison with differentially expressed genes in the 204 samples of the OPSCC-GSE65858 dataset used by Wichmann and collaborators [24], subdivided in HPV-negative (n = 147) and HPV-positive (n = 57), revealed de-regulation of 64 genes, of which 16 were down- and 48 were up-regulated in the HPV-negative compared to the HPV-positive cases. Due to the small number, up- and downregulated genes were considered together, revealing an enrichment of metabolic, response to stimuli, and cell differentiation processes (Online Resource 6-ESM_6).

Next, common up- or downregulated genes from the 4 independent datasets were ordered based on their sum of the statistical ranks within each individual list. Table 6 reports the 30 transcripts with lower rank sums. Interestingly, half of them are also differentially expressed in the samples of the OPSCC-GSE65858 dataset that are positive both for HPV16 DNA and for HPV16 RNA. Within this list, AKR1C3 is the gene with the highest expression in HPV-negative patients, thus suggesting that it could be a promising biomarker and a potential target for therapy in these subgroups.

Table 6.

Thirty gene signature obtained from the intersection of common genes from 4 independent datasets

Gene Name	our cohort	OPSC-GSE40774	OPSC-TCGA	GSE65858	GSE65858 DNA+RNA+
CDKN2A	-3.57	-2.83	-2.46	-0.74	-1.51
GABRP	-1.42	-1.51	-2.87	-0.68
CLCA4	-1.18	-1.25	-1.92	-0.52
STAG3	-1.16	-2.73	-5.16	-0.49	-0.55
UBD	-1.08	-2.65	-2.36	-0.71
VCAM1	-0.90	-1.45	-3.25	-0.63
LTB	-0.89	-1.54	-1.52	-0.56
CCL19	-0.84	-1.77	-2.12	-0.54
CLDN10	-0.80	-1.10	-2.41	-0.66
TCAM1P	-0.74	-2.56	-3.98	-0.48	-0.76
KLHL35	-0.73	-1.55	-2.88	-0.55	-0.83
FCGBP	-0.71	-1.90	-1.52	-0.45
SYCP2	-0.70	-1.89	-4.64	-0.70	-1.04
MUC4	-0.69	-1.17	-2.74	-0.43	-0.70
MEI1	-0.64	-3.71	-3.58	-0.42	-0.15
PSIP1	-0.60	-1.15	-1.35	-0.45	-0.40
SPOCK2	-0.59	-1.74	-2.10	-0.50
LMO4	-0.58	-1.45	-1.23	-0.47	-0.41
FAM3B	-0.58	-2.94	-2.41	-0.67	-0.09
AKR1C3	1.81	2.26	2.58	0.61
GPX2	1.08	1.71	1.93	0.53
MYH10	0.86	0.71	1.22	0.45	0.41
SERPINE1	0.86	0.93	2.58	0.67
IRX2	0.85	2.02	1.94	0.39	0.44
GJA1	0.76	0.94	1.81	0.51
F2RL1	0.70	1.65	2.02	0.50
COL4A6	0.65	2.21	3.85	0.53	0.46
KRTDAP	0.64	3.38	2.49	1.23	1.01
PTHLH	0.64	3.17	2.51	0.84
CCND1	0.60	1.27	1.90	0.58	1.02

AKR1C3 serves as a prognostic biomarker in OPSCC

In order to confirm our findings, we evaluated AKR1C3 expression in an independent series of 111 OPSCC samples (cohort B), of which 82 were HPV-positive and 29 were HPV-negative. We found that AKR1C3 expression normalized to BTF3 was significantly higher in the HPV-negative samples (p = 0.00038, Wilcoxon rank sum test; Fig. 4a), as previously found in our and the independent cohorts. Moreover, Kaplan-Meier curves showed that not only the HPV-negative cases, but also the subgroup of HPV-positive cases with a high AKR1C3 expression (HPV-pos _high) have a worse prognosis compared to those with a low expression (HPV-pos _low), both for disease-free survival (DFS; p = 0.00057) and overall survival (OS; p = 0.0057) (Fig. 4b, c, respectively). Similar statistically significant results were obtained by normalizing data on MTCH1 (HPV-negative vs HPV-positive: p = 2.8 × 10⁻⁵; DFS: p = 0.00067; OS: p = 0.0036) and ACTB (HPV-negative vs HPV-positive: DFS: p = 8 × 10⁻⁴; p = 2.2 × 10⁻⁵; OS: p = 0.01) (data not shown).

Fig. 4.

a Differential AKR1C3 expression determined using the -ΔC_T method in an independent cohort of OPSCC samples (n = 111) according to HPV status. Kaplan-Meier curves for disease-free survival b and for overall survival c according to HPV-negative cases (green line), HPV-positive cases with a high AKR1C3 expression (red line), and HPV-positive cases with a low AKR1C3 expression (orange line). P values were obtained from Wilcoxon rank-sum test (a) and log-rank tests (b and c)

AKR1C3 inhibition decreases cell proliferation and potentiates the effect of Cisplatin

Next, we evaluated the potential role of AKR1C3 as a target for therapy, selecting two OPSCC HPV-negative cell lines (Fadu and Cal-27) as suitable in vitro models. Both Fadu and Cal-27 cells expressed the enzyme, but in Fadu cells we detected a higher level of expression (Online Resource 7-ESM_7 panel A-B). The effect of two selective AKR1C3 inhibitors (MEDS461 and MED436) on cell proliferation was tested, and we found that Fadu cells were more responsive than Cal-27 cells. The highest effect was obtained with MEDS461, with an IC50 of 40 μM for Fadu (Fig. 5a and Online Resource 8-ESM_8A) and 70 μM for Cal-27 (Fig. 6a and Online Resource 8-ESM_8B). MEDS436 compound had IC50 values of 50 μM and 80 μM for Fadu and Cal-27, respectively (Online Resource 9-ESM_9A-B). Considering only the compound with the highest efficacy on both cell lines, i.e., MEDS461, we tested whether treatment with IC50 values was able to reduce the expression of AKR1C3 at the protein level. We found that AKR1C3 expression after 1, 3 and 24 h of treatment was not significantly affected (data not shown). To assess the effect of AKR1C3 on cell proliferation, we transiently silenced it in Fadu cells using siRNA, since these cells exhibited higher levels of the enzyme. We found that the transfection resulted in a ~ 50% reduction in mRNA and protein levels, respectively (Online Resource 10-ESM_10A-B). We also found that the silencing led to a reduction in cell proliferation of about 30% (Online Resource 10-ESM_10C). Moreover, a significant reduction in cell viability was observed when the silenced cells were treated with MEDS461 for another 72 h. In fact, we found that the IC50 value of MEDS461 was reduced from 40 μM for inhibitor alone to 7.33 μM for siRNA+MEDS461, with an enhancement factor (EF) of 5.4.

Fig. 5.

Viability assay performed in Fadu cells after 72 h of treatment with escalating concentrations of MEDS461 inhibitor (a), CIS (b) and the combination of these two compounds (c). Panel d represents the % of viable cells after 24 h of pre-treatment with the IC50 value of MEDS461 followed by 48 h of CIS treatment. *: p = 0.05; **: p = 0.01; ***: p = 0.001; ****: p < 0.0001. M461: MEDS461; PT: pre-treatment

Fig. 6.

Viability assay performed in Cal-27 cells after 72 h of treatment with escalating concentrations of MEDS461 inhibitor (a), CIS (b) and the combination of these two compounds (c). Panel d represents the % of viable cells after 24 h of pre-treatment with the IC50 value of MEDS461 followed by 48 h of CIS treatment. *: p = 0.05; **: p = 0.01; ***: p = 0.001; ****: p < 0.0001. M461: MEDS461; PT: pre-treatment

Both cell lines are sensitive to CIS (Fig. 5b and Online Resource 8-ESM_8C; Fig. 6b and Online Resource 8-ESM_8D), the standard chemotherapeutic treatment, with IC50 values of 3.4 μM for Fadu and 3.9 μM for Cal-27 cells. In a search for potential synergistic effects, we next tested combination effects of MEDS461 and CIS. To this end, we evaluated two different associations: co-treatment with scalar dilutions of the compounds for 72 h, and sequential treatment with MEDS461 for 24 h and subsequently CIS for 48 h. In Fadu cells, the combined administration of the two compounds revealed a statistically significant effect, except for the highest dilution (6.25 μM), compared with the non-treated (NT) cells (Fig. 5c). Unfortunately, even with a significant increased activity at low dilutions, the combination did not provide an improvement of the effect compared to CIS alone, with an enhancement factor (EF) of 1.26 (Online Resource 11-ESM_11, Panel A).

In the sequential treatment, a significant reduction in Fadu cell viability was observed at all the concentrations tested compared with NT cells (Fig. 5d). Interestingly, preconditioning with MEDS461 improved the effect of CIS alone at all doses, especially at the lowest dose, with an EF of 3.65 (Online Resource 11-ESM_11, Panel A). Also in Cal-27 cells, co-treatment was effective (Fig. 6c). However, when compared with CIS alone, an antagonist effect was noted, with an EF of 0.78 (Online Resource 11-ESM_11, Panel B). This phenomenon was more evident in the pretreated cells. Compared with NT, a significant reduction of viable cells was observed, but compared with CIS, the effect was lower, with an EF of 0.34 (Fig. 6d; Online Resource 11-ESM_11, Panel B).

Taken together, these results show that the effect of AKR1C3 inhibitors alone is more effective in cells that have a higher baseline expression of the target, and that combination therapy with CIS is synergistic when pretreatment with AKR1C3 inhibitors is carried out before the administration of CIS.

Discussion

Two series of 29 and 111 OPSCCs, derived from patients recruited in four Northern Italy hospitals, were analyzed using microarray-based gene expression profiling and qRT-PCR, respectively. Additional gene expression data were retrieved from three independent published OPSCC cohorts, to derive biological processes and transcripts with a robust association to HPV infection status. HPV-positive tumors, strongly associated with a better prognosis in all the datasets, showed an enrichment of processes, as already described in [26], involved in muscle contraction and development, growth factor regulation at the G1-S cell cycle phase and immune responses, mainly characterized by TCR signaling, antigen presentation by MHC class II molecules and leukocyte chemotaxis. Interestingly, enrichment of immune related processes has already been described as a characteristic feature of a subgroup of patients [27–29], suggesting that immunotherapy may be an option for patients experiencing disease progression. The recurrence rate of HPV-positive patients is about 20% and tends to be delayed, in general after three years from the end of therapy. Some clinical trials are ongoing, in particular for recurrent/metastatic HNSCC patients, testing the efficacy of immunotherapeutic agents, such as nivolumab (Clinical Trial ID: NCT03349710, NCT02105636) or pembrolizumab (Clinical Trial ID NCT03480672- NCT03813836), alone or in combination with standard chemo-radiotherapy [30]. Until now, however, HPV infection status has not influenced the decision to treat HNSCC patients with immunotherapy. Additionally, as demonstrated by Keck and colleagues [20], a subgroup of HPV-negative tumors belongs to a subclass enriched for immune-related genes, which partially overlaps with a subgroup of HPV-positive tumors, suggesting that HPV status should not be the only discriminant in the therapy regimen choice. Radiotherapy, alone or in combination with chemo/targeted therapies, is the standard of care for the treatment of OPSCC. The first chemotherapeutic approach comprises platinum compounds, alone or in combination with anti-EGFR monoclonal antibody, depending on EGFR expression status. Unfortunately, intrinsic resistance development occurs frequently [31], especially in tumors with an increased expression of genes involved in drug metabolism and oxidative stress [32], which are associated with apoptotic escape. HPV-negative tumors are part of this setting. In fact, they are characterized by an enrichment of genes involved in proliferation, leukocyte migration, homeostasis, cell cycle regulation, glutathione derivative biosynthetic and xenobiotic metabolic processes and response to stimuli, in particular hypoxia and oxidative stress. Together, these features are related to a more aggressive phenotype. In particular, alterations in glutathione, cytochrome P450 and hypoxia pathways have already been described as being involved in disease progression and drug resistance [33]. These alterations lead to increased ROS production, which is responsible for oxidative damage of proteins, lipids and DNA, resulting in increased mutation rates, as well as uncontrolled cellular growth and apoptosis [34]. On the other hand, it has been found that high levels of cytoplasmatic ROS may promote the expression of antioxidant scaffold genes (SOD, CAT, NRF2) able to detoxify cells from excessive ROS accumulation and cell death induction [35]. Nevertheless, tumor cells may escape from this checkpoint by activating ROS adaption mechanisms, which are particularly evident in cancer stem cell (CSC) compartments, in which apoptosis resistance, enhanced drug-efflux pump activity and DNA repair are induced. These characteristics concur with the development of drug resistance and increased survival [36]. Of interest, we observed overexpression of Wnt pathway-related genes in HPV-negative patients. This pathway is known to sustain cancer stem cell compartments, thereby promoting drug resistance and aggressiveness [37].

By combining the HPV-related signatures of four independent datasets, we ended up with 30 genes strongly associated with HPV status, 15 of which were also deregulated in HPV DNA- and RNA-positive OPSCCs. Next to CDKN2A, several transcripts have previously been described as prognostic biomarkers in this tumor type: Cyclin D1 (CCND1), whose amplification has a strong correlation with OS [38] and a poor radiation response in OPSCC, typical of HPV-negative tumors [39], FCGBP (Fc Fragment IgG Binding Protein), which is highly associated with a longer survival and a HPV-positive status in HNSCC patients [40], SYCP2 (Synaptonemal Complex Protein 2), an early tumor biomarker in OPSCC [41] frequently overexpressed in HPV-positive tumors, thereby promoting genomic instability and consequent molecular aberrations [42], GPX2 (Glutathione Peroxidase 2), deregulated in HNSCC and belonging to a prognostic three-gene signature able to predict clinical outcome in HNSCC [43], KRTDAP (Keratinocyte Differentiation Associated Protein), a predictor of resistance to chemo-radiotherapy in HPV-positive HNSCC, up-regulated in patients who failed first line therapy [44], and FAM3B (Family With Sequence Similarity 3 Member B), which is overexpressed in HPV-positive tumors and in esophageal squamous carcinoma cells compared to their normal counterparts, promoting epithelial to mesenchymal transition [45], as well as in colon cancer cells where it promotes invasion and metastasis formation [46]. We found an upregulation of PTHLH in HPV-negative patients. Literature data indicate that high expression of this gene is associated with a poor prognosis of HNSCC patients [47, 48]. Puzzo and collaborators showed that JGA1 overexpression is able to identify a subset of HNSCC patients characterized by a poor prognosis and a reduced responsiveness to radio/chemotherapy [49]. For COL4A6 (Collagen Type IV Alpha 6 Chain) and CLDN10 (Claudin 10), no information on their role in OPSCC is currently available. COL4A6 has been found to be down-regulated in prostate cancer and to negatively correlate with its prognosis [50]. In colorectal cancer, loss of COL4A6 has been found to be associated with promoter hypermethylation [51]. CLDN10 has been reported to be highly expressed in many cancer types, including papillary thyroid cancer and hepatocellular carcinoma, and to be associated with a worse overall survival [52, 53].

Interestingly, we found that the most highly overexpressed gene in HPV-negative OPSCCs was AKR1C3. The encoded enzyme is involved in the regulation of prostaglandin, steroid hormone, retinoid metabolism and detoxification, and has been associated with resistance to therapy. By stratifying HPV-positive samples according to AKR1C3 expression (HPV-pos _high and HPV-pos _low) we found that HPV-pos _high patients had a worse prognosis, similar to HPV-negative patients. Similar results were obtained by Huebbers et al. [21]. They found that this gene was upregulated, both at the mRNA and the protein level, not only in HPV-negative cases, but also in a subset of HPV-positive cases with a poor prognosis.

In addition to its association with a poor prognosis, AKR1C3 has been found to be involved in the development of CIS resistance in colon cancers, and to be associated with the acquisition of castration resistance and resistance to radiotherapy in prostate cancer [54–56]. Thus, we hypothesized that it could also be a candidate target for alternative personalized or combined therapies. Selective inhibitors of AKR1C3, developed at the Department of Drug Science and Technology of Turin University, have already been tested in colorectal cancer [55] and prostate cancer [55, 57] cells. In fact, it has been reported that MEDS461 and MEDS436, the most promising compounds of the studied series, are highly selective for AKR1C3 over the 1C2 isoform with minimal COX1 and COX2 off-target effects. In cell-based assays, the compounds reduced proliferation and prostate specific antigen and testosterone production in AKR1C3-expressing 22RV1 prostate cancer cells, and showed synergistic effects when used in combination with abiraterone and enzalutamide. We, therefore, decided to treat oropharyngeal cancer cells with these AKR1C3 inhibitors to test their potency and selectivity [25, 57]. To this end, we focused our attention on HPV-negative cell lines, since HPV-negative patients have a worse prognosis and limited therapeutic options. We are well aware, however, that the utility of AKR1C3 as a therapeutic target should also be explored in HPV-positive cells. A relatively higher efficacy of MEDS461 was observed in Fadu cells, which express more AKR1C3 protein than Cal-27 cells. A similar finding has already been made in prostate cancer cells treated with the selective AKR1C3 inhibitor KV-37, in which the drug response was directly related to AKR1C3 expression [58]. Moreover, we found that the combination of AKR1C3 silencing with MEDS461 treatment enhanced its effect on proliferation, significantly reducing the IC50 dose 5.4 fold, thereby underlining the key role of AKR1C3 in HNSCC cell proliferation.

Next, we verified whether AKR1C3 inhibition may affect the response of HNSCC cells to CIS. For this reason, we tested the efficacy of CIS, consecutive and simultaneous with a selective AKR1C3 inhibitor synthetized by our group, in two HPV-negative OPSCC cell lines expressing AKR1C3 at different levels. Pretreatment with the selective inhibitor enhanced the effect of CIS alone by 3.65 fold, only in the cell line (Fadu) expressing a higher level of AKR1C3. This observation is in line with literature data, in particular on colon cancer cells [54], where upregulation of AKR1C family members, in particular C1 and C3, was found to be associated with a lower response to CIS. According to our results, no synergistic effect of the two drugs was obtained by simultaneous treatment, but prior AKR1C3 inhibition by pretreatment with the most active compound yielded a strongly enhanced CIS response. In Cal-27 cells, exhibiting a lower expression of AKR1C3, no significant improvement of a CIS alone response was noted. Further studies are warranted, in particular in CIS- and radio-resistant cells [59, 60], to substantiate the putative role of AKR1C3 in HNSCC therapy resistance.

Taken together, we uncovered a robust gene expression signature associated with HPV status and identified AKR1C3 as a strong prognostic biomarker and interesting druggable target for oropharyngeal tumors.

Abbreviations

HNSCC

head and neck squamous cell carcinoma

ORR

overall response rate

PFS

progression free survival

OS

overall survival

DFS

disease free survival

OPSCC

oropharyngeal squamous cell carcinoma

HPV

human papilloma virus

AKR1C1

aldo-keto-reductases 1C1 and 1C3

AKR1C3

aldo-keto-reductases 1C3

CIS

cisplatin

IRIS 3

cyanine 3

IRIS5

cyanine 5

FE

feature extraction

GO

gene ontology

MEM

minimal essential medium eagle

DMEM

dulbecco’s modified essential medium

FBS

fetal bovine serum

PS

penicillin streptomycin

DMSO

dimethyl sulfoxide

NT

untreated

ATP

adenosine triphosphate

OD

optical density

BCA

Bicinchoninic Acid

TBS

Tris-buffered saline

HRP

horseradish peroxidase

DCD

deceased

REL

relapsed

NED

not evidence of disease

ECM

extracellular matrix

EF

enhancement factor

TCR

T cell receptor

MHC

Major histocompatibility complex

ROS

Reactive oxygen species

Authors’ contributions

G.C. conceived and designed the work; C.P.-N., M.M.-G., I.G., A.M. acquired data and performed experimental work; P.O., F.G and A.C. analyzed and interpreted the data; D.B., S.O-B., L.D.-C. revised the work; A.C.P. synthetized the compounds; F.P., S.C., G.V., P.A.-V., G.A., L.M., R.D., P.B. and M.K. informed and recruited patients and supplied follow-up information.

All authors have approved the submitted version of the manuscript and have agreed both to be personally accountable for the author’s own contributions and to ensure that questions related to the accuracy or integrity of any part of the work, even ones in which the author was not personally involved, are appropriately investigated, resolved, and the resolution documented in the literature.

Funding

This research was financially supported by Bando POR-FESR Asse I misura I.1.3 of Piedmont Region; Fondazione Boschetto; Fondazione Cassa di Risparmio di Torino (Grant BOSD-CRT-17-2). The validation analysis received funding from AIRC (ID 18519 and ID 23573 projects -P.I. De Cecco Loris).

Data availability

Raw and processed data have been deposited to the GEO Omnibus database (GSE142583).

Compliance with ethical standards

Conflict of interest

The authors declare that they have no competing interests.

Ethical approval

All procedures performed involving human participants were in accordance with the ethical standards of the Ethical Committee of Novara University Hospital (Protocol No. CE42/2011) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Consent for publication

Not applicable.