Wide spectrum and high frequency of genomic structural variation, including transposable elements, in large double stranded DNA viruses

Our knowledge of the diversity and frequency of genomic structural
variation segregating in populations of large double stranded (ds) DNA
viruses is limited. Here we sequenced the genome of a baculovirus (AcMNPV)
purified from beet armyworm (Spodoptera exigua) larvae at depths
>195,000X using both short-read (Illumina) and long-read (PacBio)
technologies. Using a pipeline relying on hierarchical clustering of
structural variants (SVs) detected in individual short- and long-reads by
six variant callers, we identified a total of 1,141 SVs in AcMNPV,
including 464 deletions, 443 inversions, 160 duplications and 74
insertions. These variants are considered robust and unlikely to result
from technical artifacts because they were independently detected in at
least three long reads as well as at least three short reads. SVs are
distributed along the entire AcMNPV genome and may involve large genomic
regions (30,496 bp on average). We show that no less than 39.9% of genomes
carry at least one SV in AcMNPV populations, that the vast majority of SVs
(75%) segregate at very low frequency (<0.01%) and that very few
SVs persist after 10 replication cycles, consistent with a negative impact
of most SVs on AcMNPV fitness. Using short-read sequencing datasets, we
then show that populations of two iridoviruses and one herpesvirus are
also full of SVs, as they contain between 426 and 1102 SVs carried by 52.4
to 80.1% of genomes. Finally, AcMNPV long reads allowed us to identify
1,757 transposable elements (TEs) insertions, 895 of which are truncated
and occur at one extremity of the reads. This further supports the role of
baculoviruses as possible vectors of horizontal transfer of TEs.
Altogether, we found that SVs, which evolve mostly under rapid dynamics of
gain and loss in viral populations, represent an important feature in the
biology of large dsDNA viruses.