Low input, long-read viromics reveals previously hidden abundant viruses and their genomic islands in soil and sea samples

Abstract:

Metagenomic long-read sequencing through the Nanopore platform holds the promise of extracting deeper and more complete biological signals compared to short-read data. However, maximizing the benefits of long-reads from low-input samples such as natural viral communities remains challenging. Here we introduce VirION 2, which builds on our previous long-read metagenomic wet-lab and informatics pipeline (VirION) by reducing DNA input 100-fold (now 1 nanogram), generating longer reads (now ~6kb median lengths), and improving accuracy (now 99.97% with short-read correction) – all as benchmarked against short-read data from a seawater sample. In seawater samples, 29% of the most abundant viruses were uniquely found by adding long-read sequencing. As proof of concept, VirION 2 was applied to soil viromes, previously inaccessible to long-read sequencing. These data revealed ~3-fold increase in community-level microdiversity patterns in soil phage communities, as well as ~100-fold increased detection of viral genomic islands, thus highlighting heavy selection pressures at play in soil viruses. Some of these genes were auxiliary metabolic genes, thought to enhance viral reproduction, some of which specific to the soil environment, like sporulation genes. Overall, the application of long-read sequencing to low-input viromes should make more samples accessible, produce more ‘complete’ virus catalogs, while revealing previously missed genomic- and biological-signals.