Discerning the origin of Epstein-Barr virus in patients using nanopore-derived DNA methylation signatures

Christopher Oakes (The Ohio State University) began his plenary talk by highlighting the potential of tumour viruses to improve patient care. He then introduced Epstein-Barr Virus (EBV), the "prototypical tumour virus". A very common virus, EBV infects ~90-95% of humans: after entering the body and infecting cells, cells proliferate and the virus replicates. The human immune system is typically very effective in launching a response: NK cells limit this proliferation, and T cells learn to recognise the viral antigens and kill infected cells. In response, EBV can get around these defences by moving into a latent state to hide from the immune system. In rare cases, the virus is able to survive and proliferate through a number of complex processes, leading to the growth of tumours. Christopher noted that one of the most significant difficulties in treating cancers is that tumours are "self": how can killing of normal cells be avoided when targeting tumours? This is where EBV's properties as a tumour virus can be exploited: EBV is "not self" and provides a potential target.

Christopher explained that there are three ways in which EBV can be useful in detecting and treating cancer. Firstly, the presence of elevated levels of EBV in the blood is used to screen for cancer, enabling its identification in early, asymptomatic stages. Christopher displayed the results of a study by Lo et al. (New England Journal of Medicine, 2017) into the use of EBV DNA detection from plasma for nasopharyngeal cancer screening. Initial identification of elevated levels of EBV lead to the identification of 34 undiagnosed cases of nasopharyngeal cancer, with one undiagnosed after not proceeding to further screening. Whilst this demonstrated its efficacy, Christopher also noted the lack of specificity of the test, given the initial ~1,000 people displaying elevated EBV levels. Secondly, EBV can be used to diagnose and classify the disease present; Christopher listed cancers that his team investigate - several forms of lymphoma, and gastric adenocarcinoma. For example, for extranodal NK/T cell lymphoma (ENKTL), EBV titre is a diagnostic. Lastly, Christopher describes how targeting the virus specifically can provide a "trojan horse way" of entering and killing tumour cells. This could be achieved via antiviral therapy or by training T/NK cells to attack EBV-infected tumour cells. However, it is here important to determine the state of the virus when it is detected in the blood; Christopher explains that in the clinic, a PCR test from plasma DNA is used to detect EBV levels, but elevated levels may represent an active infection of mononucleosis, or may be a result of immunosuppresion, or may be due to tumour cells. In order to determine the correct course of treatment, it is important to determine which of these that the test represents. How could this be achieved?

Christopher highlighted here that there is "more information in DNA than just the sequence itself." Methylation of EBV DNA, he explained, is intrinsic to the life cycle of the virus. When the virus first infects a host, its DNA is in an unmethylated state; when it passes into the latent phase, it becomes highly methylated, silencing most genes. When the virus is reactivated and passes into its lytic phase, this methylation is again lost. This differential methylation could hold the key to identifying the stage of infection represented by elevated EBV levels. Christopher displayed the results of an investigation of methylation in ~30 regions of the EBV genome in cell lines, using sodium bisulphite conversion and PCR. This showed that methylation was high in tumour cell lines, but low in in-vitro infected samples. Differential methylation was also observed across different regions of the genome. Latent virus can be activated, inducing lytic cells; analysis indicated that methylation decreased when lytic activation was induced. However, Christopher noted that diagnosis using this method is challenging, and cannot determine whether methylation levels represent a heterogeneous population or separate populations with either high or low levels of methylation.

Christopher then introduced his team's work with nanopore sequencing in detecting methylation in EBV. Firstly, samples with known proportions of methylated and unmethylated EBV DNA were produced. Tumour virus DNA was amplified via WGA, removing all methylation. Some of this methylation-free DNA was reserved; the rest was methylated using SssI (CpG) methylase. The resulting samples were mixed in known proportions of methylated and unmethylated sample, for use in producing a standard curve for calculating the percentage of methylation present. Samples were prepared for sequencing with the Ligation Sequencing Kit (SQK-LSK109), sequenced on the MinION device and the data basecalled using Guppy. Reads were mapped to the EBV genome with MiniMap2. Reads were also re-basecalled with the high-accuracy Flip-Flop basecaller. Nanopolish was used to index and call methylated CpG sites: Christopher noted the robust, qualitative distinction between methylated and non-methylated DNA, and good concordance between the expected and preserved proportions.

Tumour cell EBV DNA was then sequenced on the MinION to assess and compare methylation levels, revealing different levels of methylation from different samples; it was possible to detect different populations in some samples. Detection was also possible from primary cells - for example, in a sample representing 100% tumour cells in the latent phase. Christopher demonstrated two examples in which the proportion of methylation clearly dropped after induction of the lytic phase.

Christopher then asked: can we detect methylation levels of EBV from cell-free (plasma) DNA, and if so, is cell-free DNA methylation representative of tumour DNA methylation? The answer is: yes - it is detectable, and good correlation was seen between methylation in cell-free DNA and tumour DNA. Christopher showed an example in which much higher methylation was detected in a cancer patient than that seen in a mononucleosis patient. In another example, a sample from a patient prior to treatment showed high methylation; analysis of another sample from during cancer treatment displayed a new, low-methylation population indicative of a change in the life cycle of the virus. Christopher pointed out that mapping data demonstrates that tumours feature very complex methylation patterns. He then showed a comparison of gene expression between an ENKTL sample and a Burkin's Lymphome (BL) sample, investigating the differential expression of three genes (EBNA1, EBNA2 and BART) between the two: methylation analysis across these genes correlated well with this expression. This suggests that it could be possible to use methylation patterns to determine gene expression, without even having to look at the expression itself. In future, the group plan to further investigate the combination of genetic and genomic information, and also intend to investigate how to enrich for EBV DNA, as levels are often still below the threshold of detection.