An Atlas of Plant Transposable Elements

Advances in genomic sequencing have recently offered vast opportunities for biological exploration, unraveling the evolution and improving our understanding of Earth biodiversity. Due to distinct plant species characteristics in terms of genome size, ploidy and heterozygosity, transposable elements (TEs) are common characteristics of many genomes. TEs are ubiquitous and dispersed repetitive DNA sequences that frequently impact the evolution and composition of the genome, mainly due to their redundancy and rearrangements. For this study, we provided an atlas of TE data by employing an easy-to-use portal ( APTE website ). To our knowledge, this is the most extensive and standardized analysis of TEs in plant genomes. We evaluated 67 plant genomes assembled at chromosome scale, recovering a total of 49,802,023 TE records, representing a total of 47,992,091,043 (~47,62%) base pairs (bp) of the total genomic space. We observed that new types of TEs were identified and annotated compared to other data repositories. By establishing a standardized catalog of TE annotation on 67 genomes, new hypotheses, exploration of TE data and their influences on the genomes may allow a better understanding of their function and processes. All original code and an example of how we developed the TE annotation strategy is available on GitHub ( Extended data).


Introduction
The growing number of sequenced plant genomes is providing unprecedented opportunities for biological studies, evolution, and growing of many algal and Viridiplantae species. We estimate more than 13k plant genomes have been released (NCBI), revealing that plant genomes are faintly explored. High diversity in terms of ploidy, heterozygosity, and genome size, probably due to a dynamic set of old and recent bursts of transposable elements (TEs), are common hallmarks of many plant genomes. 1 TEs can comprise between 32% to 56% (Utricularia genomes), 2,3 to up to 90% in many plant genomes, 4-6 e.g., maize 7 and wheat. 5,8,9 TEs can be organized into two main classes. Each class is hierarchically organized into orders, superfamilies, families, and subfamilies. This terminology is primarily associated with the type of their transposition mechanisms. 3,10 They are classified into: (i) retrotransposons (Class I), which are propagated by a "copy-and-paste" mobility mechanism and are the most redundant TE class in plant genomes; and (ii) DNA transposons (Class II), which are known for the "cut-andpaste" mechanism that allows them to move to a completely different position. Moreover, both Classes may contain autonomous members 10,11 for which the transposition mechanism depends on an autonomous and cognate type of TE. 11,12 However, despite their significance, in-depth identification and analysis of TEs content in the sequenced plant genomes remains barely explored. 13,14 The lack of concise data available may prevent the enrichment of in silico, functional genomics research and compromises the appearance of new strategies to investigate TEs. Recently, many computational models and entire wet-lab efforts have increasingly been helping to understand these sequences. [15][16][17][18] For example, Ensembl Plants 19 provides high-quality, primary genomic information for 67 plant (in the broad sense, including green plants, green and red algae) genomes, assembled near or at chromosome scale; however, mobile sequences are poorly systematized and have a humble coverage.
These observations prompted us to standardize tools and methods aiming to improve TE detection, annotation and standardization. In this work, we developed a new method for systematic annotation of plant TEs, using the 67 genomes available at Ensembl Plants assembled at chromosome scale as a starting point. Our identification was standardized, applying the same methodologies to all genomes and delivering a concise Atlas of TEs annotation in plant genomes. We also provided an updated analysis of non-coding RNAs (ncRNAs) overlapping TEs. This annotation is accessible on the Atlas website for exploration and download, which might be relevant to any type of research involving mobile sequences.

Data source
All genomes (Supplementary Material 1, Extended data) were downloaded from the Ensembl Plants 19 database, version 41 (57 genomes) and 45 (plus 10 new genomes).

Annotation of transposable elements
We used similarity-based methods and de novo techniques to build a collection of putative transposable elements, based on the SPTEdb pipeline. 21 We refined, extended and increased steps in order to produce a novel annotation ( Figure 1). Our reformulated steps (details in Supplementary Material 2, Extended data) guarantee a comprehensive knowledgebase of these TEs. RepeatScout was performed separately; the output was unified in a library to be labeled by PASTEClassifier 32 and later combined into a final annotation. To automate the pipeline, an in-house framework in Perl language was developed for each software output to be uniformized, described in steps 1 to 4. A main script in Bash starts the process of automatization using Perl scripts. All steps were supervised by researchers, carefully checked, and the output was manually verified at each step for each genome. Records classified as low complexity, simple repeat and other nomenclature not related to Class I or Class II TEs were discarded.
Due to the extensive genome sizes of Triticum aestivum (14,5 Gb), Triticum dicoccum (10,4 Gb) and Triticum turgidum (10,4 Gb), we adapted our pipeline for their analysis, based on the approach of Jamilloux et al. 24 For these species, we applied our pipeline on chromosome 1 (which is the longest pseudomolecule), as the large genomes were eventually duplicated into new copies, increasing the number of these same repeats in the genome, and did not significantly impact discoveries related to new or different TEs families. 24

TE evidence score
To test the reliability of our TE annotation pipeline, we scored sequences that had duplicated annotation in the same loci ( Figure 2). We developed a statistical metric (labeled as TE-Score, shown in each record as the ninth column for each genome annotation file) that identify and ponder sequences types that have been identified by the programs. The TE-Score is a metric (0 to 1) that is given by Steps in transposable elements identification. Dataset: Genome assemblies were downloaded from Ensembl Plants. Identification: 1A) RepeatScout was used to search for putative repetitive sequences and further classification by PASTEClassifier, resulting in a library. 1B) RepeatModeler was also used to find a consensus of TEs sequences. 2) RepeatMasker was run with Repbase library and libraries from RepeatModeler and RepeatScout.
3) For Class II -Subclass 2 TEs, we also used HelitronScanner and MITE-Hunter. 4) In order to find LTR and Non-LTR retrotransposons, we used LTR_retriever and MGEScan-non-LTR, respectively. Filter: A cut-off filter was applied to remove low complexities, simple repeats and other nomenclatures that were not classified into TEs. Annotation: In result of the pipeline, we have a Transposable Element annotation for each genome analyzed. where QIP = Quantity of identification by program and QP = Quantity of programs. To illustrate an average of the amount sequences annotation by programs, see Figure 2.

Correlation analysis
To test for correlations between genome size and transposable elements percentage by genome in base pairs, we first normalized using log10, and then we applied the Pearson Correlation Coefficient in SPSS version 25.
Web implementation APTE is hosted at the Universidade Tecnológica Federal do Paraná (Cornélio Procópio, PR, Brazil). It uses Debian 11 as operating system, Apache 2 as web server, PHP 5.6 as web programming language. We also used Zend Framework 2, which implements model, view, controller (MVC), a methodology for web development that can be expanded for any future additional functionality. On the front-end, we used HyperText Markup Language 5 (HTML5), Cascading Style Sheet 3 (CSS3) and JavaScript to perform dynamic functions that provide user-friendly navigation. A built-in genome browser (JBrowse, version 1.14.1) is available to visualize and download the data as well.

Computational resources
To run the pipeline described in Figure 1, we used three platforms: (i) to runRepeatModeler, 25  class, order and superfamily. In addition,~42,64% (21,236,989) elements could not be assigned to any type of known TE and they were labelled as unknown. They likely represent chimeric and/or partial elements for which we were not able to perform the full classification. For known TEs, we identified that~62,85% were retrotransposons, and~37,15% were DNA transposons. All assigned classifications of TEs identified along the 67 genomes are shown in Figure 3. The distribution of TEs in the analyzed genomes are somewhat similar (Figure 4), especially in genomes that have a shorter phylogenetic distance (e.g., Oryza spp, Triticum spp). However, even close-related genomes exhibit uneven TEs distribution (e.g., Arabidopsis spp). Two main hypotheses might explain the variation of TE content: (a) different evolutionary stories, since the two major genome duplication events are shared by all seed plants (epsilon) and flowering plants ( gamma), followed by the lineage-specific duplication events, 20 and (b) specific pressures to maintain, expand and purge TEs in each lineage.
We have noted that our approach permitted better TE annotations in genomes assembled at chromosome scale, and we also observed that the amount of TEs is generally related to the genome size, since larger genomes have higher occurrences of TEs ( Figure 5). However, for incompletely and draft-assembled genomes, it tends to decrease the number of TEs, once the assembly into small parts (scaffolds or contigs) may impact the genome assembly quality, collapsing repeated contigs (mostly TE-derived) and interfering with the proper identification of these TEs.

TE database comparison
To compare the results of the identification performed and to ensure the reliability (details in Supplementary Material 1, Extended data) of our approach, we used SPTEdb 21 annotation data of the genome Populus trichocarpa (black cottonwood), which is explored in Table 1. The second comparison of TE annotations was performed for the Glycine max (soybean) genome, in which we used SoyTEdb 22 to compare the amount vs. type of TEs, shown in Table 1. A third comparison used data from GrTEdb 23 to explore the amount of TEs in Gossypium raimondii (cotton), available in Table 1.

Conclusion
Our analysis brought an exhaustive, systematic and comprehensive genome identification in plant genomes, using seven programs to annotate TEs in plant genomes. In both TE classes, several orders and superfamilies were found ubiquitously in all genomes. Additionally, 21,236,989 out of 49,802,023 mapped TE sequences could not be classified into any of the nomenclatures known for TEs, and were labeled as "Unknown" in GFF3, a standard file format for gene annotation.
For plant species whose TE complement may be quite well-annotated, i.e., Arabidopsis thaliana, we yielded an increased number of identified TEs. In species with less curated annotation in Ensembl, we were able to deliver a more detailed identification of TEs. For example, in three particular genomes, i.e., Populus trichocarpa (black cottonwood), Glycine max (soybean) and Gossypium raimondii (cotton), we increased the TE identification levels by 2,295%, 900% and 2,643%, respectively. We observe that for several other genomes, new types of TEs were identified and annotated; this ensures that our pipeline delivers not only the same TE identification, but also new ones, making the annotation process possible to use for any species.
In this study, we contributed to expand the knowledge on TEs, by providing a large-scale, organized and standardized TE Atlas. We integrated all annotations to make it available to download in each genome separately from the Atlas of Plant Transposable Elements (APTE) website. An example how our pipeline works using the A. thaliana genome, software dependencies, and in-house scripts developed, which can be downloaded, used and changed freely, are available from https://github.com/alerpaschoal/apte_pipeline/.  The full TE annotation is provided in the APTE database that is informative and straightforward. It is a very good source of data for other researchers that might be interested in proceeding deepest analysis of a specific group of elements. The authors have identified a great number of repetitive sequences in each genome. My biggest concern is related to the sequences that fail to be classified. These "Unknown" sequences make up a huge proportion of all the repetitive sequences found. Even with the filtering of low complexity and simple repeat sequences, they could still be false positive. The authors could mention this possibility. Moreover, the difference in the amount of TE identification between the two databases shown in Table 1 is impressive. My doubt is whether there is a possibility that part of this difference is caused by the artificial "breaks" of TE copies into a greater number of fragments. Finally, it is not clear if the authors have filtered short sequences that could also be false positive.

Data availability
Minor suggestions/comments: Page 6: "evolutionary stories". I believe "evolutionary histories" would be more appropriate. MITE is not an order. They are non-autonomous elements from order TIR. In the same way, LARD and TRIM are not superfamilies, but non-autonomous elements related to some LTR superfamily. I suggest a change to "Superfamily/Group" since some groups specified are not independent established superfamilies, but rather groups/clades of a superfamily. The same occurs in Table 1. Moreover, in Table 1, Crypton should be placed as an order. 2.