Sponsored content brought to you by

‘It was amazing to discover that AML evolution is mainly driven by methylation alterations that affect low density CpGs regions in the gene body of master transcription factors. These findings completely changed our point of view on epigenomic alterations of AML.’ — Alberto Magi, Department of Information Engineering, University of Florence, Italy

 

The haematological malignancy acute myeloid leukaemia (AML) accounts for around 2% of all cancer deaths each year. Alberto Magi, Associate Professor of Bioengineering at the Department of Information Engineering, University of Florence, Italy, shared the story of the last two years of his team’s research exploring the genetic basis of chemoresistance and relapse in individuals that had been treated for AML*.

Exploring the genetic basis of chemoresistant AML

Samples from two females of 22 and 20 years of age, and one male of 67 years of age, were taken for their research. All three subjects had been diagnosed with AML, had undergone standard chemotherapy, and had gone into remission for around a year, but had then relapsed exhibiting a chemoresistant phenotype. As whole-exome sequencing (WES) data had been obtained from each subject at diagnosis, remission, and relapse, the team compared the WES data from diagnosis and relapse to uncover the relapse-specific genetic changes. This revealed a few relapse-specific variants that affected key cancer genes, but none of those genes had a documented role in chemoresistance activity. Alberto’s team therefore decided to perform high-coverage nanopore whole-genome sequencing (WGS) for each subject to gain a more comprehensive picture of the genomic and epigenomic changes that may be responsible for the chemoresistance phenotypes observed in these three subjects.

To achieve this, the team created two novel computational tools: GASOLINE and PoreMeth. Alberto explained that GASOLINE is one of the first tools for detecting somatic structural variants (SVs) in the genome, by comparing two WGS datasets, whilst PoreMeth is the first tool to identify differentially methylated regions (DMRs), by comparing nanopore methylation profiles of two samples. These analyses were ‘a big challenge’: the team wrote 211,000 lines of code for somatic SV detection and 215,000 lines of code to call methylation in the samples.

They applied their new analysis workflow GASOLINE to their research samples obtained at diagnosis and relapse to explore relapse-specific somatic SVs. Similar to their WES analyses, they detected a few relapse-specific variants that affected important cancer genes, including those occasionally mutated in AML or myelodysplastic syndromes (e.g. FLT3 and SRGAP2). However, none of the genes affected by SVs had a documented role in drug resistance, arguing against their involvement in reprogramming the leukaemic cells to their chemoresistant phenotype. They therefore shifted their focus to exploring the epigenome, benefitting from the ability to call variants and methylation from the same dataset with nanopore sequencing technology.

The importance of gene body methylation: a ‘paradigm shift’

Alberto explained the importance of DNA methylation in genome regulation; DNA methylation involves methyl groups being added to a DNA molecule, and this can change the activity of the DNA without changing the nucleotide sequence. In mammals, methylation is almost exclusively found in CpG dinucleotides. Around 75% of CpG dinucleotides are methylated in somatic cells, except for CpG-rich sequences termed CpG islands (CGIs) that are generally unmethylated. The methylation state of CGIs in the promoter regions of a gene, at the transcription start site, strongly influences the transcriptional activity of a gene. Usually, a gene is transcribed when CpGs are not methylated.

Alberto explained that bisulfite sequencing and arrays have traditionally been used to explore DNA methylation. However, bisulfite conversion methods have limited resolution for methylation analysis: arrays interrogate less than 1% of CpGs in the human genome; enhanced reduced representation bisulfite sequencing (ERRBS) approaches target around 25% of genomic CpGs; and even whole-genome bisulfite sequencing (WGBS) can only interrogate around 50-60% of CpGs.

In comparison, using their high-coverage nanopore WGS dataset to interrogate methylation, ‘we were capable [of calling] the methylation state of around 90-99% of all CpGs of the human genome…with at least one read’, with over 90% of CpGs covered by at least five reads. Alberto stated that this is probably ‘the highest resolution map of CpGs’ ever produced for an AML genome.

Using the PoreMeth pipeline, the team investigated DMRs at relapse compared to diagnosis, as captured in the nanopore sequence data. Their approach detected DMRs with unprecedented resolution – 30-40% of DMRs were found to be below the detection limit of bisulfite conversion methods. This high-resolution map of differential methylation states of CpG dinucleotides allowed them to ‘explore low-density CpG regions that were never explored before’. The vast majority of regions were differentially hypermethylated, suggesting the acquisition of methylation during cancer evolution. The DMRs were associated with around 1,000 genes in each sample, many of which were affected by DMRs that fell outside of CGIs, in low-density CpG regions.

To understand the functional impact of these DMRs, they looked for enrichment in cancer-specific signalling pathways, such as those involved in cell proliferation, including genes known to have a role in chemoresistance. Surprisingly, genes within these pathways were affected by DMRs located outside of CGIs – so not at high-density CpG regions. In particular, DMRs seemed to be associated with gene bodies (i.e. introns and exons), where CpGs are found at low density. Alberto emphasised that this is a ‘paradigm shift’ – we are moving away from the classical paradigm that high-density CpGs around the promoter are of central importance, and towards the idea that low-density CpGs in the gene body, which couldn’t be seen before with other technologies, are key.

Regarding the genes affected by differential methylation at low-density CpGs, there was an enrichment for transcription factor genes in all three subjects, as opposed to chemoresistance genes. So, what might the hypothesis be here? Alberto stated that differential methylation is probably being selected for during AML progression, and this differential methylation was affecting transcription factor genes that regulate the expression of key genes, such as those imparting chemoresistance. In other words, differential methylation at low-density CpG regions within transcription factor genes was being selected, and this was indirectly inducing the chemoresistant phenotype.

A regulatory cascade to resistance

To test this hypothesis further, they investigated gene expression using RNA obtained from their AML research samples. Thousands of genes exhibited differential expression between diagnosis and relapse, for each of the three subjects. Putting this all together, all three subjects displayed the same methylation perturbations – extensive and recurrent perturbations in gene regulatory networks, with hypermethylation at gene bodies, outside of CGIs, and decreased expression of cancer genes of interest, including tumour suppressor genes, oncogenes, and chemoresistance genes.

Overall, as Alberto had hypothesised, their findings suggested an indirect role for DMRs in selection of a chemoresistant phenotype: during cancer evolution, differential methylation at low-density CpG regions within transcription factor gene bodies impacted transcription factor function and, consequently, regulatory cascades; target genes in such cascades were therefore deregulated. Ultimately, this induced the observed chemoresistant phenotype.

With the ‘unprecedented resolution’ for identifying differential methylation provided by high-coverage nanopore sequence data, the team were therefore able to study the subtle but crucial methylation changes involved in the development of chemoresistance in these AML samples. Their two-year story had started with exploring genomic variants, and ‘we finished with finding that AML evolution is mainly driven by epigenomic alterations’.

 

*London Calling 2022 hybrid conference, hosted by Oxford Nanopore Technologies; May 18–20, 2022

 

Oxford Nanopore Technologies and the Wheel icon are registered trademarks of Oxford Nanopore Technologies plc in various countries. All other brands and names contained are the property of their respective owners. © 2022 Oxford Nanopore Technologies plc. All rights reserved. Oxford Nanopore Technologies products are not intended for use for health assessment or to diagnose, treat, mitigate, cure, or prevent any disease or condition.