Partner-independent fusion gene detection by multiplexed CRISPR/Cas9 enrichment and long-read sequencing – Christina Stangl

"Cancer genomes are in general very unstable" - 16% of cancers involve a driver fusion gene. Despite their prevalence, fusion genes have long been neglected as it very difficult to identify them. Displaying a fusion gene network of acute myeloid leukaemia for context, Christina stated that there is fusion gene “promiscuity”: a vast number of genes form fusions, often there are no recurrent fusion partners, and breaks can range within 100s of kilobases. In response to the paucity of research into fusion genes, Christina's team aimed to develop a targeted and directional application for their detection.

Christina introduced the question of whether we even need to know both fusion partners involved in a fusion gene? She answered that yes, it is incredibly important - not only for diagnosis, but also for prognosis, and for helping doctors make better treatment choices. Furthermore, fusions can be used as biomarkers for minimal residual disease tracing.

Current diagnostic methods for detecting fusion genes include: fluorescent in situ hybridisation (FISH), a semi-targeted method using probes against one recurrent, specific fusion partner (meaning the other partner will not be detected); real-time PCR, a targeted method to amplify a specific sequence, again missing fusions that don't recur; and RNA-sequencing, an unbiased method capable of identifying "all the fusions present in a sample". However, short-read RNA sequencing has its drawbacks: a deep sequencing depth is needed to identify all fusions and a lot of false positives are detected; if a fusion hasn't been reported in the literature then validations will be needed which take time and effort. Moreover, the process takes an average of 2-4 weeks which is "just too long to be used for clinical practices". In comparison, nanopore long-read sequencing takes only 1-2 days and has a greater ability to identify fusion pairs - long reads give high confidence that the fusion events are true events. However, Christina said that a global sequencing approach does not produce enough output from a single MinION Flow Cell in the clinical setting, and therefore they aimed to develop a more targeted method.

Christina introduced her method of fusion gene detection called FUDGE (FUsion gene Detection from Gene Enrichment), an orientation-specific sequencing method involving single-cut, dsDNA targeting with Cas9. The sequencing data is analysed using the NanoFG (Fusion Gene detection from Nanopore data) algorithm, which outputs the exact genes, breakpoint, and provides primers for breakpoint validation.

As proof-of-principle, FUDGE was tested on a fusion-positive cell line, in which a ESWR1 fusion was known to be present but the partner was undetermined. With such a large increase in depth provided by FUDGE, this method revealed that the fusion was an EWSR1-FLI1 fusion.

Christina also demonstrated that the single-cut method was successful regardless of the fusion partner or breakpoint sequence, and on-target enrichment was up to 1,000x for the RH and AML1 genes which were targeted, including over 200x fold enrichment of reads spanning the fusion. As expected, the number of fusion-spanning reads increased over the 48 hours of sequencing time (the typical length of a sequencing run that they performed).

In a patient with a KMT2A fusion, but unidentified fusion partner, FUDGE identified that MLLT6 was the partner. Christina also demonstrated how tiled guide design across a gene can help identify the partner gene and breakpoint in instances when the fusion has a more variable or "promiscuous" breakpoint, as was the case for a PAX3-FOXO1 fusion. Christina stated that "this is a very nice method to span really long regions", and even though the partner may vary "with FUDGE we can always identify the fusion partners".

Furthermore, after only ~24 hours Christina found that the detection of fusion genes plateaued - suggesting that the length of a run could be reduced from 48 hours to 24 hours.

Next Christina discussed performing FUDGE with multiplexed targets and samples on a single MinION Flow Cell, without barcoding, using NanoFG for fusion calling, followed by breakpoint PCR for validation. NanoFG identified the fusions detected within the multiplexed run, and subsequent PCR performed with the suggested primers identified from which samples the fusions were from. As an example, this successfully identified reciprocal translocations between KMT2A and members of the MLLT gene family. Furthermore, Christina's team were also able to identify reciprocal fusion genes using FUDGE; she stated that prior to using FUDGE "we have never really had an assay to identify reciprocal fusion genes".

In summary, Christina stated that FUDGE can be used for targeted and directional fusion gene detection, which is partner and breakpoint-location independent, fast, and able to span large regions of interest. The method requires little DNA input (the method can be performed from only 10 ng by including a WGA), and targets and samples can be successfully multiplexed.