Hide and seek: de novo identification in sugar beet reveals impact of non-autonomous LTR retrotransposons

Plant genomes are filled with retrotransposons and their derivatives, subject to constant sequence turnover. As short, non-autonomous retrotransposons do not encode a protein product, they experience reduced selective constraints on their DNA sequence, leading to diversification into multiple families, usually limited to only a few species. This absence of any coding capacity and their tendency to form subfamilies are the reasons for the incomplete description of non-autonomous LTR retrotransposons in most to all genomic repeat annotations. Here, we focus on non-autonomous LTR retrotransposon identification. Are all of these sequences derivatives of easier-to-identify full-length elements? Or is there more variability, which is currently overlooked? For this, we capitalize on our comprehensive understanding of the TE landscape in sugar beet to assess the extent of the blind spot on non-autonomous LTR retrotransposons Here, we present a workflow to identify non-autonomous LTR retrotransposons without prior sequence information, retrieving more than 100 families within the sugar beet genome. We only include TEs without the ability for complete self mobilization. Spanning up to 15,000 bp, these non-autonomous families are often longer than expected and characterized by reshuffling and modular evolution. Most strikingly, only a few of these families are directly derived from autonomous partners, showing that there is a large, undiscovered TE variety in the non-autonomous TE fraction. We highlight that a large fraction of non-autonomous TEs wont be retrieved with the current TE identification workflows, even if the output is well-curated and condensed into TE libraries and suggest procedures to remedy this gap. This study is the first insight into the non-autonomous LTR retrotransposon landscape within a single genome and serves as an example to estimate the error in non-autonomous TE detection.