Identification of multidrug chemoresistant genes in head and neck squamous cell carcinoma cells

Multidrug resistance renders treatment failure in a large proportion of head and neck squamous cell carcinoma (HNSCC) patients that require multimodal therapy involving chemotherapy in conjunction with surgery and/or radiotherapy. Molecular events conferring chemoresistance remain unclear. Through transcriptome datamining, 28 genes were subjected to pharmacological and siRNA rescue functional assays on 12 strains of chemoresistant cell lines each against cisplatin, 5-fluorouracil (5FU), paclitaxel (PTX) and docetaxel (DTX). Ten multidrug chemoresistance genes (TOP2A, DNMT1, INHBA, CXCL8, NEK2, FOXO6, VIM, FOXM1B, NR3C1 and BIRC5) were identified. Of these, four genes (TOP2A, DNMT1, INHBA and NEK2) were upregulated in an HNSCC patient cohort (n = 221). Silencing NEK2 abrogated chemoresistance in all drug-resistant cell strains. INHBA and TOP2A were found to confer chemoresistance in majority of the drug-resistant cell strains whereas DNMT1 showed heterogeneous results. Pan-cancer Kaplan-Meier survival analysis on 21 human cancer types revealed significant prognostic values for INHBA and NEK2 in at least 16 cancer types. Drug library screens identified two compounds (Sirodesmin A and Carfilzomib) targeting both INHBA and NEK2 and re-sensitised cisplatin-resistant cells. We have provided the first evidence for NEK2 and INHBA in conferring chemoresistance in HNSCC cells and siRNA gene silencing of either gene abrogated multidrug chemoresistance. The two existing compounds could be repurposed to counteract cisplatin chemoresistance in HNSCC. This finding may lead to novel personalised biomarker-linked therapeutics that can prevent and/or abrogate chemoresistance in HNSCC and other tumour types with elevated NEK2 and INHBA expression. Further investigation is necessary to delineate their signalling mechanisms in tumour chemoresistance. Supplementary Information The online version contains supplementary material available at 10.1186/s12943-023-01846-3.


Clinical Samples
The use of fresh clinical specimens collected in the UK was approved by the NHS Research Ethics Committee (06/MRE03/69).All tissue samples were previously collected according to local ethical committee-approved protocols and informed patient consent was obtained from all participants [1][2][3].Fresh tissue biopsies were preserved in RNALater (#AM7022, Ambion, Applied Biosystems, Warrington, UK) and stored short-term at 4°C (1-7 days) prior to transportation and subsequent storage at -20°C until used.All frozen samples were digested with nuclease-free proteinase K at 60°C prior to mRNA extraction (Dynabeads mRNA Direct kit, Invitrogen).

Crystal Violet Cell Viability Assay
Crystal violet cell viability assay was performed in 96-well plates (Figure S1).Growth medium was aspirated and 30 µL/well of crystal violet solution (0.5% crystal violet in 30% ethanol) was added and incubated for 10 minutes at room temperature (RT).Cells were then washed with 200 µL/well distilled water prior to the addition of 100 µL/well of 1% SDS and incubated for 30 minutes at RT prior to absorbance measurement at 595 nm using a CLARIOstar microplate reader.

Figure S1. Chemosensitivity assays for wildtype (WT) and drugresistant cells measured by crystal violet cell viability assay following 72 hr drug incubation. An example shown here using WT and PTXresistant (PTX-R) CaLH2 cell lines. IC50 values (shown here in nM)
representing the degree of chemosensitivity for WT and PTX-R cells were determined using sigmoid-curve fitting algorithm on respective data points plotted as logarithmic drug concentrations on the X-axis and survival fraction (Absorbance at 595 nm) on the Y axis.IC50-fold difference was calculated between WT and PTX-R cells as indicated within the graph.

Establishment of Drug-resistant Cell Strains
Crystal violet cell viability dose-response curve (kill-curve) assays were first performed to determine the half maximal growth inhibition (IC50) concentrations of cisplatin, 5-fluorouracil (5FU), paclitaxel (PTX) and docetaxel (DTX) on SVpgC2a, SVFN8 and CaLH2 cell lines.Each cell type was then cultured in growth media containing IC50 concentrations of each drug with regular changes of growth media containing freshly diluted drugs.When cells were proliferating, a 3-fold higher concentration above the IC50 of each drug were then added to the growth medium.This process was repeated until cells were able to proliferate in the highest drug concentrations over a period of ~6 months.Drug-resistant cells were then expanded in drug-free growth medium to create aliquots for cryopreservation until used for experiments.We have established a total of 12 drug resistant strains: four drug-resistant (Cisplatin, 5FU, PTX and DTX) strains for each of the three cell types (SVpgC2a, SVFN8 and CaLH2), with a minimum of 3-fold higher IC50 values than their corresponding parental wildtype cells Additional File 2: Figure S14-S16).

Drug-dependent Chemoresistant Gene Expression Assay
To identify genes that are differentially expressed when challenged with increasing doses of chemotherapeutic drugs in WT and its corresponding drug-resistant strains, cells were seeded into 96-well plate (8,000 cells/well) one day prior to drug treatment.Cells were then treated with serial dilutions of each drug for 24h prior to harvest for RT-qPCR to quantify the relative mRNA expression levels of 28 target genes (Figure S2).

Reverse transcription quantitative PCR (RT-qPCR)
RT-qPCR assays were performed as described previously [1][2][3] with minor modifications.Briefly, mRNA purified directly from cells using the Dynabeads™ mRNA DIRECT™ Purification Kit (61012; ThermoFisher Scientific, UK) were used directly in RT-qPCR reaction containing qPCRBIO SyGrene 1-Step Go Lo-ROX (PB25.31-12;PCRBiosystems, UK) and gene-specific primers for onestep reverse transcription and qPCR to quantify gene expression in the LightCycler 480 qPCR system (Roche, UK) based on our previously published protocols [1,3,11] which are MIQE compliant [12].Briefly, thermocycling begins with 45ºC for 10 mins (for reverse transcription) followed by 95ºC for 30s prior to 45 cycles of amplification at 95ºC for 1s, 60ºC for 1s, 72ºC for 1s, 78ºC for 1s (data acquisition).A 'touch-down' annealing temperature intervention (66ºC starting temperature with a stepwise reduction of 0.6ºC/cycle; 8 cycles) was introduced prior to the amplification step to maximise primer specificity.Melting analysis (95ºC for 30s, 75ºC for 30s, 75-99ºC at a ramp rate of 0.57ºC/s) was performed at the end of qPCR amplification to validate single product amplification in each well.Relative quantification of mRNA transcripts was calculated based on the second derivative maximum algorithm [13] (Roche).Primer sequences are provided in Additional File 2: Table S2.All target genes were normalised to two stable reference genes validated previously [4] to be amongst the most stable reference genes across a wide variety of primary human epithelial cells, dysplastic and squamous carcinoma cell lines, using the GeNorm algorithm [14].No template controls (NTCs) were prepared by omitting cells/tissue sample during RNA purification and eluates were used as NTCs for qPCR assays to monitor contamination.

Determination of Chemosensitivity (IC50)
Chemosensitivity or drug potency on cell viability or gene expression was determined by performing curve-fitting on dose-response data points to calculate the concentration of drug which induced 50% inhibition (IC50) using the four-parameter logistic Hill equation [15]: where Y is the percentage of cell death or gene downregulation, X is the drug concentration, A is the maximal cell density or gene expression, B is the slope factor, C is the IC50 and D is the minimal cell density or gene level.Cell viability or gene expression datapoints of dose-response assays were curve-fitted based on the above algorithm using the Quest Graph™ IC50 Calculator (AAT Bioquest, Inc, 04 Jul.2019, https://www.aatbio.com/tools/ic50-calculator).

Pan-cancer and HNSCC transcriptome data mining
Pan-cancer transcriptome datasets were queried in the Oncomine (www.oncomine.org)[16] and Kaplan-Meier Plotter (KM-Plotter.com)[17] databases.The initial differentially expressed gene selection study was performed in Oncomine with the main inclusion criterion that the studies must involve comparison between HNSCC tumour samples with normal tissues.Studies using HNSCC cell lines were excluded.At the time of analysis, there were eight studies eligible for analysis (Additional File 2: Table S1).Differentially expressed genes were ranked according to their median P-values for over-expression and under-expression.Candidate genes were selected based on their top-ranking positions across the eight studies.For differential gene expression of selected candidate genes in HNSCC tumour tissues and matching normal margins, we have used the pan-cancer GEPIA 'Box Plot' tool (GEPIA, http://gepia.cancer-pku.cn/)[18] which is based on transcriptomic data of The Cancer Genome Atlas (TCGA)/The Genotype-Tissue Expression (GTEx) [19]).We have also used GEPIA to survey candidate gene expression profile across 33 human cancer types.For pan-cancer biomarker prognostic survival analysis, hazard ratio (HR) and logrank P values were extracted from each corresponding Kaplan-Meier plots for each cancer type with either single marker or different combinations of 2, 3 or 4 markers (there were a total of 15 unique combinations of 1 to 4 markers studied).The main exclusion criterion was when HR values were associated with logrank P values of <0.05 (Additional File 3).

Drug Library Screen
A total of 537 compounds were obtained from the Drug Synthesis & Chemistry Branch, Developmental Therapeutics Program (DTP) at the National Cancer Institute (National Institute of Health, USA), consisting of 147 approved oncology drugs (AOD IX) and 390 known natural products (Set V) selected from the DTP Open Repository collection of 140,000 compounds.Factors in selection were origin, purity (>90% by ELSD, major peak has correct mass ion), structural diversity and availability of compound.Drug library compounds were each given individual NSC ID number searchable at DTP Chemical database (https://dtp.cancer.gov/dtpstandard/ChemData/index.jsp).Original drug stocks (10 mM in DMSO) were diluted to 0.1 mM (in DMSO) as working stocks arrayed in 384-well reservoir plates for downstream screening using alamarBlue™ Cell Viability Reagent (DAL1025/DAL1100; ThermoFisher Scientific, Paisley, UK) in 384-well format.Cells (4000 cells/well in 384-well plates) were seeded one day before the addition of drugs (final drug concentration at 1 µM) which were incubated for 72h before addition of alamarBlue™ for 24h incubation before measuring fluorescence (excitation 540 nm and emission 590 nm) using a CLARIOstar microplate reader.Candidate drugs were selected based on anti-proliferative effects between wildtype and drug-resistant cells.For drug-gene dose-dependent interaction study, cells (8000 cells/well in 96well plates) were incubated with candidate drugs (5-fold dilution containing 6 concentrations from 0.32 nM to 1 µM) for 24h prior to harvest for RT-qPCR to investigate their dose-response effects on gene expression.Assay format and detail protocol are shown in Figure S4 below.

Statistical Analysis
Statistical t-tests P values were used for differential analysis between two groups of data.Linear and non-linear regression analyses were used to quantify the relationship between serial drug concentrations and gene expression levels (Additional Data Figure S5-S16).Beeswarm Boxplots were created in R (version 2.13.1;The R Foundation for Statistical Computing) [20].

Figure S4 .
Figure S4.Drug-gene dose-dependent interaction screening protocol and plate setups.A. Cells (8x10 3 cells/well) were treated for 24h with serial dilution series of nine different drugs as indicated in the diagram.Wells along the edges were not used to eliminate non-specific edge-associated effects.Control cells were treated with equal volume of vehicle (distilled water).B. Direct mRNA extraction method using Dynabeads were performed using a 96-well PCR plate and DynaMag™-96 Side Magnet (12331D) with protocol as illustrated.A total of 60 mRNA samples harvested from panel A (using protocol in panel B) were simultaneously extracted for RT-qPCR in panel C. C. One-step RT-qPCR 384-well plate map for target and reference gene expression quantification in the 60 mRNA samples performed in duplicates.