Detection and identification of fungal pathogens using MinION metabarcoding

High-throughput sequencing (HTS) using Illumina sequencing has proven its huge potential to detect plant pathogens in various types of samples, but the high turnaround time (especially when sequencing is outsourced), limited read lengths and the need to pool many samples with the aim to reduce costs make it less attractive as a fast, low-throughput diagnostic tool. Nanopore sequencing offers a potential solution for this problem, since this can be performed in-house within a limited timeframe and with a low number of samples when using the Flongle flowcell.
Here, we evaluated metabarcoding for detection and identification of fungal pathogens in seed samples and spore traps using nanopore sequencing. We compared different barcodes (ITS2 and a long ITS fragment) and a 5’ and 3’ region of tef1-α. The comparisons were done with 1) different mock communities consisting of DNA of various fungal pathogens in different relative concentrations; 2) different spiked communities, i.e. a background community of DNA derived from either a seed or a spore trap sample spiked with DNA from three fungi in different concentrations and 3) real spore trap and seed samples.
We observed that the mock, spike and real communities can be profiled well using MinION and Flongle metabarcoding. An important observation is that the background of the sample (i.e. the absolute amount of fungal DNA present as well as the composition of the fungal community), plays a huge role in the success of detecting and identifying a pathogen. For example, the background community determines the sensitivity; for spore trap samples this was found to be around 1 pg, while for seed samples it was 0.01 pg (more sensitive due to less background fungi). The specificity is acceptable for most investigated species, but it depends on the barcode used and the completeness of the reference database. The long ITS fragment was determined as being a very good general barcode with both good sensitivity and specificity. The 5’ end tef-1α barcode is however preferred when detecting and identifying Fusarium sp.
Based on these results, we can state that the developed Flongle metabarcoding approach is definitely useful for a first (more general) pathogen screening in a fast way, but that presence or absence of specific pathogens still needs to be confirmed by targeted tests such as qPCR since we observed that sensitivity and specificity depend on the context of the sample. We also recommend setting up and following a scheme that evaluates controls (no template, positive and alien control) to be able to detect, for example cross-contamination and interpret the results correctly.