Liquid biopsies — multi-modal cell-free DNA assays using nanopore sequencing for potential cancer detection

The use of liquid biopsies in cancer research involves taking blood samples to examine circulating tumour DNA (ctDNA) or cell-free DNA (cfDNA); they are particularly useful in paediatric cancers, since they are less invasive than tissue biopsies. Interested in the potential for using cfDNA for both longitudinal monitoring of cancer and early cancer detection, Billy highlighted that ‘methylation profile…can be more informative than just looking at the specific cancer mutations, because methylation is reflective of overall transcriptional programs in the cell’. Using short-read sequencing technology, where methylation is inferred using bisulfite treatment, biases such as GC skews, DNA damage, PCR amplification bias, and alignment artefacts are introduced. Nanopore sequencing is PCR-free so that methylation is simultaneously detected alongside canonical bases, with no special library prep required. In his talk, Nanopore sequencing of cell-free DNA for methylation-based breast cancer detection in a case-control research cohort, Billy described how he and his team developed a very high-throughput, nanopore-based single-molecule sequencing approach to accommodate low input amounts of cfDNA. Their multiplexed approach used less than one nanogram of input DNA, equating to ~0.5–1 ml plasma, and generated up to 200 million reads for a single cfDNA sample — improving throughput by over an order of magnitude upon previous techniques.

‘We can sequence very high throughput cfDNA using the Oxford Nanopore platform in a highly multiplex fashion’

The team have used their high-throughput nanopore sequencing approach to create single-molecule methylation profiles of cfDNA from population research cohorts. Billy shared how they have combined both methylation and fragmentation profiles in breast cancer and normal research samples and how their multi-modal data could be used as potential indicators of disease state in early breast cancer detection.

[Methylation] pattern changes… could be reflective of a malignancy…[which] you can detect through liquid biopsy research’

Carolin is also applying liquid biopsy to the field of paediatric cancer research, since one of the key challenges is obtaining sufficient tumour tissue to perform genomic profiling. Explaining that children with cancer, particularly those with solid tumours, have very poor outcomes and relapse rates, Carolin highlighted that paediatric cancers are the leading cause of child death in the western world. Motivated by this, the Cortés-Ciriano group at EMBL-EBI is investigating thanks to CRUK funding the potential utility of nanopore sequencing for a multi-modal liquid biopsy test for cancer screening — covering copy number variants, structural variants, single nucleotide variants, methylation, and fragmentomics — and developing computational tools for data integration and collation for an ‘accurate way to monitor disease progression and treatment response’. Carolin explained that ‘quite a large proportion of paediatric cancer patients have actionable alterations that we might be able to target’.

We are also very interested in looking at the methylation signal...This is something that’s very exciting, in particular for cfDNA

In her talk, The potential application of nanopore sequencing for liquid biopsy analysis in children with cancer, Carolin described how nanopore sequencing can be used to directly detect methylation patterns in native DNA, showing potential to more accurately predict tumour burden and perform disease classification. Showing how the methylation signal strongly correlated with tumour fractions and how cfDNA from different tumour types clustered separately, Carolin highlighted how the methylation signal can differentiate between different tumour types. They would like to further investigate how the methylation pattern correlates with gene expression and how this, in turn, correlates with the nucleosomal footprint. Carolin showed a typical cfDNA fragmentation profile obtained from their nanopore sequencing reads, explaining that mono-, di- and tri-nucleosomal DNA fragments were detected alongside longer cfDNA fragments.

The Cortés-Ciriano and Beggs groups also performed highly accurate copy-number profiling using nanopore sequencing, and Carolin shared an example of five neuroblastoma clinical research samples collected by the CRUK-funded Stratified Medicine Pediatrics (SMPEDS) programme from one individual before surgery, during and after chemotherapy, and at relapse. Their data indicated that the tumour fraction altered over time, with additional copy number events detected.

As the field of cancer detection and monitoring develops, the ability to store the raw nanopore sequence data means it can be re-analysed to investigate additional potential modifications of interest in future. Carolin described how the Cortés-Ciriano group at EMBL-EBI continue to develop computational tools for data analysis, including SAVANA  — a novel SV caller for long sequencing reads — and methylation deconvolution algorithms. Carolin summarised by highlighting that a single nanopore dataset allows simultaneous detection of genomic and epigenomic variants, providing the potential ‘to develop an affordable, easy-to-deploy, multi-modal cfDNA assay with increased sensitivity and specificity for cancer screening.’

‘A really transformative technology for circulating tumour DNA analysis’

Our final showcase panellist, Andrew, began by explaining how his lab like to specialise in doing what others say is impossible. Recollecting in his talk, Nanopore sequencing of ctDNA – better than short read! how he was told you can't detect mutations at low variant allele frequency (VAF) with nanopore, Andrew went on to describe how they did just that. Introducing their ‘homebrew’ nanopore sequencing protocol, Andrew explained how they use a commercial PCR panel and adapt the ctDNA amplicons for direct use with the Oxford Nanopore Ligation Sequencing Kit, before performing duplex sequencing — combining sequence data from both DNA strands, enabling Q30 accuracy, for variant detection at very low VAF. Showcasing nanopore variant detection within codons 12 and 13 of KRAS, one of the most commonly mutated oncogenes in cancer, with a VAF of 15 and 6%, Andrew was keen to point out how ‘noise-free’ the reads were, particularly bearing in mind the very low DNA input commonly associated with liquid biopsy samples. Describing the mutations as ‘easy’ to pick out, the team have shown that their approach works ‘between anything from 5 ng and beyond for both FFPE and ctDNA’. Highlighting nanopore sequencing as ‘a really transformative technology’, Andrew shared how his group detected a 15 bp indel in EGFR, a known genomic aberration in lung cancer at 2% VAF, ‘as clear as day’, and a SNV in oncogene PI3KCA, a common variant detected in lung cancer and ctDNA, present at 1.5% VAF, which Andrew described as ‘very easily pickable outable’. Benchmarking their adapted method with a reference panel, they found that 95% of all comparable mutations were detected. Andrew concluded his talk by explaining they can use their method to detect variants with a VAF down to ~0.1–0.2%.

When asked why use nanopore sequencing given that the regions of interest are small, the answer was simple: ‘we can go from the DNA to the result in eight hours. You can't achieve that with hybridisation, and you can't do that with short-read-based sequences because you need to run the flow cell for 24–36 hours to get data — that's the key advantage to us.’ Andrew also highlighted the cost-efficiency of their protocol, which generates ‘coverage at 10,000x in eight hours’: using the Flongle, an adapter for MinION or GridION sequencing devices, it costs £75 all in, including all the enzymes and reagents.
During the post-session panel, Andrew also touched on fragmentomics and methylation, explaining that they have detected reads up to 1,000 bp long which show tumour-specific methylation patterns, highlighting that ‘long reads are actually quite informative when it comes to classification’ and suggest ‘they'll play a more important role and a unique role’ in liquid biopsy. Generally, cancer detection has relied on genomic mutation information; however, Andrew suggested that the methods we use to detect epigenetic changes are probably going to be more important for early screening assays. Highlighting the need for simple, reliable, and cost-effective methods to detect this methylation, Andrew explained that ‘one of the things that not just us but others… realise is the Nanopore kit works on femtomolar loading, not quantity'. So, for short DNA fragments, ‘150 fm DNA is equivalent to 5–10 ng DNA’, which is easily within the reach of most of the paediatric research samples — ‘that's the key to unlocking a lot of these epigenetics tests’.

Watch all the talks and post-session panel here.