NCM 2021: For want of a nail: investigating somatic and germline variations in cancer
Sissel (Senior Director, Genomic Applications, Oxford Nanopore Technologies) presented work on ‘extrachromosomal DNA in cancer’, which has been performed by the Showcasing team, based predominantly in New York and San Francisco. This team aim to showcase nanopore technology ‘in the best possible way’.
Somatic variation: extrachromosomal DNA in cancer
Extrachromosomal DNA (ecDNA) has been detected in 14–40% of solid tumours and its presence is often associated with higher malignancy. These molecules are usually circular, double-stranded DNAs of 1–3 Mbp in length. How they form is not entirely clear, but there are several proposed mechanisms, such as via errors during DNA replication. They can contain full oncogenes and regulatory regions and have been shown to highly impact oncogene activation, with many of the oncogenes they contain being cancer treatment targets, associated with treatment sensitivity and resistance. Sissel explained that, when these ecDNAs are replicated, they are unevenly segregated between daughter cells during cell division, which leads to tumour heterogeneity. As ecDNAs tend to have an open-chromatin structure, the genes they contain tend to be highly expressed.
Current methods to identify and/or reconstruct ecDNAs include manual counting via microscopy, which Sissel stated can be laborious — and as they don’t provide any sequencing information, there is limited value for treatment. Another approach is via short- and medium-read-length sequencing, although this has limited ability to resolve structural variants and repeats, which makes ecDNA assembly challenging; it can also be hard to determine if the sequences originate from the ecDNA or the genome itself; and thirdly it does not provide any methylation information, which could be of value.
Sissel’s team therefore decided to investigate these ecDNAs using nanopore sequence data from three neuroblastoma cell lines. Sergey Aganezov in the team built an analysis pipeline to reconstruct ecDNAs. After de novo assembly of the CHP212 cell line at ~4x depth of coverage, a circular ecDNA of ~230x depth of coverage across the MYCN ongogene locus was obtained. Further analysis demonstrated that the ecDNA MYCN sequences were in a different order to those on the reference genome. Titrating the data down from its original depth of ~4x, Sissel explained that even ~0.25x still provided ~14x depth of the ecDNA, which was sufficient for full ecDNA resolution. Sissel pointed out that, with such low depth requirements, sufficient genomic sequencing for ecDNA resolution could be achieved ‘within an hour, maybe two’, of sequencing a sample on a MinION Flow Cell; and sequencing could also be performed using the low-cost, lower output (1–2 Gb) Flongle Flow Cells.
Sissel’s team also looked at the methylation status of the MYCN locus in the cell lines; in CHP212, the region upstream of the locus had a significantly different methylation profile compared to other cell lines. Although Sissel stated that she couldn’t confirm what that meant, it could be part of the tumour heterogeneity.
Germline variation: hereditary cancer
Phill (Associate Director, Clinical Applications, Oxford Nanopore Technologies) discussed his team’s work on detecting germline variants in hereditary cancer using adaptive sampling. Phill drew our attention to how ~10% of cancers are hereditary; deleterious mutations in cancer risk genes passed on to offspring tend to be heterozygous. However, the value of 10% is likely to be an underestimate, mainly because the impact of structural variants is probably insufficiently appreciated.
Phill outlined that adaptive sampling is a way of performing targeted enrichment of a sequence of interest, without performing additional library prep — targeting is performed on the flow cell during nanopore sequencing. On-target reads are accepted, and off-target reads are rejected by the pores. In Phill’s opinion, this technique could be particularly useful for sequencing large panels of genes – ‘that’s sort of where the hereditary cancer idea…came from’. Phill started looking into Lynch syndrome, associated with genetic aberrations in mismatch repair genes. The first job was to work out what he wanted to target; ‘and like any good scientist I decided that I’d Google it – so I did!’ From this, Phill identified several genes to target; additionally, he included an expanded Lynch syndrome panel. He noted that there were also specific hereditary cancer panels. So ‘why not’ add those too? ‘And then things got a little bit silly’ — as there are also expanded hereditary cancer panels, and even large research panels. That’s when, on some ‘idle Thursday afternoon’, the thought struck him that he had created a massive panel, which would take months or even years to analyse with traditional PCR-based approaches. So, he then needed to go and try this on something.
Luckily Phill had a box of hereditary cancer samples in the freezer ‘as one does’. He chose four of these at random: one ovarian cancer sample, and three breast cancer samples (including a BRCA1/2 negative sample). Phill sequenced these four samples, along with a non-cancer human reference control, for 72 hours on a GridION. Per sample, he obtained over 22.5 Gb under the adaptive sampling conditions. With a ‘moving graph’, Phill showed how they obtained ~57x depth of coverage across the 201 cancer targets. This was ‘quite phenomenal’ as he was only aiming for ~30x coverage. Single nucleotide polymorphisms (SNPs) were called and subsequently intersected with the ClinVar database to identify potentially pathogenic SNPs. In all four of the cancer samples, one pathogenic SNP was found, whereas within the non-cancer control sample, no SNPs were found. The ‘most interesting’ SNP identified was an intronic SNP within a breast cancer sample that formed a cryptic splice site, ultimately resulting in a truncated BRCA1protein. Phill stated that this example nicely demonstrated ‘why you should sequence your introns’.
The SNP in the BRCA1/2-negative sample fell within the gene PALB2, the protein product of which forms a complex with BRCA1 and BRCA2. Disruption to this complex results in defective homologous recombination repair. Although the role of the intronic variant was unclear, PALB2 mutations have been suggested as good targets for BRCA-negative patients, so the mutation was of interest to follow up.
Phill next presented the concept of phasing. He explained that, in heterozygous conditions, it is helpful to identify which haplotype the heterozygous variant is on, as it may have implications for disease (e.g. instances of compound heterozygosity in the CFTR gene in the development of cystic fibrosis). In the case of BRCA1 and BRCA2, the fact that the nanopore sequence data could be phased completely across the genes into two haplotypes, which are reasonably complex and also have ‘nasty’ pseudogenes, suggested that the quality of read mapping was very high.
As native DNA was sequenced, Phill could also directly examine methylation status in the data (using the Megalodon tool). Phill presented ‘a really nice example’ of the MSH2 gene promoter, a core gene associated with Lynch syndrome, where all cancer samples had high levels of methylation in the promoter-enhancer region, but the control didn’t. This wasn’t expected.
In the last part of Phill’s talk, he discussed how the rejected reads from the adaptive sampling process could be ‘reused’ to look for copy number variants (CNVs). Going back to the BRCA1/2 negative samples, ‘I find this bit quite phenomenal’: a CNV deletion was detect in the PIP gene, which is known to be important in development and progression of breast cancer. They also located the breakpoints.
Phill concluded by going back to the talk title: ‘For want of a nail’. He explained that this comes from a proverb that highlights how small aberrations in the system can have large downstream effects. It also shows the requirement for reasonably simple tools to resolve complex problems. So, ‘whatever research you’re doing’, Phill hoped that this work demonstrated how nanopore sequencing can fit into that work, and hopefully in the future, ‘no one will want for a nail’.