All the data files associated with this work are listed in Table 1. The study design is illustrated in Data file 1. The characteristics of the OSCC cell lines used in this work are described in Data file 2. The characterization of the stem cell-like tumour spheres and the measurements of cisplatin sensitivity of the three cell lines have been reported previously [8]. All the methods provided in the following sections are condensed versions of the methods described in our previous work [8].
MinION sequencing
We performed six MinION sequencing runs for H103, SAS, and SAS tumour spheres using two MinION SpotOn Flow Cells version R9.5 (Oxford Nanopore Technologies (ONT), UK; Data file 3). We first co-extracted supercoiled mtDNA and nuclear DNA of the cells using QIAprep Miniprep Kit (QIAGEN, Germany) and Agencourt AMPure XP (Beckman Coulter Inc., USA) [19]. The sequencing libraries were prepared using the 1D Ligation Sequencing Kit (SQK-LSK108; ONT, UK), loaded onto the flow cells, and sequenced for 48 hours. The flow cells were washed using a Wash Kit (EXP-WSH002; ONT, UK) before they were reused for subsequent sequencing runs.
Raw sequencing signals stored in FAST5 files were acquired by MinKNOW version 1.6 (ONT, UK; Data set 1). The sequencing run performance was assessed using Poretools [20] (Data file 4). During sequencing, live base-calling with a read quality score threshold of 7 was executed by an in-built MinKNOW base-caller. To base-call all the reads, additional post-sequencing base-calling was performed using Albacore version 1.2.6 (ONT, UK). The quality of the base-called reads was assessed using NanoStat [21] (Data file 5). The base-called reads were mapped to the human reference genome assembly GRCh38 using BWA-MEM [22], generating alignment files (Sequence Alignment Map (SAM) format). The mapping statistics are provided in Data file 6 [23]. The SAM files were compressed into the binary format (BAM) using SAMtools [24]. The variants were called by Nanopolish [25], which compared the aligned reads with the revised Cambridge Reference Sequence of mtDNA in the GRCh38 assembly. The accuracy of variant calling was evaluated by a cross-check of the quality-filtered variants with Sanger sequencing, as described in our previous work [8]. The workflow for sequence reads processing and variant-calling analysis is provided in Data file 7.
Microarray analysis
Total RNA was isolated and purified using innuPREP RNA Mini Kit (Analytik Jena, Germany) and RapidOut DNA Removal Kit (Thermo Fisher Scientific Inc., USA). The purified RNA samples were subjected to a whole transcriptomic analysis using the GeneChip Human Clariom S Array (Thermo Fisher Scientific Inc., USA; the analysis outsourced to Research Instruments Sdn. Bhd., Malaysia). The raw data files (CEL files) were obtained from the GeneChip Command Console Software (Thermo Fisher Scientific Inc., USA; Data set 2). The transcriptomic profiles of the cells, as described in Data file 8, were analysed using Transcriptome Analysis Console 4.0 (Affymetric Inc., USA). As reported previously, the findings of the microarray analysis were confirmed by real-time quantitative polymerase chain reactions (qPCR) [8].