Plant material, growth conditions, and stress treatment

Prior to germination, Arabidopsis thaliana seeds were surface sterilized and stratified for 2 days at 4 °C. Before and during stress treatments plants were grown under controlled conditions in a Panasonic MLR-352-PE growth chamber on solid ½ MS medium (1% sucrose, 0.8% agar, pH 5.8) at 21 °C with 12/12 hr (day/night) light cycle. For the heat-stress-induction of ONSEN transposition, a prechilling step was performed to increase the relative activation [10]. Seedlings were first placed at 4° for 24 hr within the growth chamber. This chilling pretreatment was followed by 24 hr with the chamber set to 37° for the heat-stressed (HS) plants; control-stressed (CS) plants were moved to 21 °C. All Arabidopsis mutants used in this study dcl3 [35], rdr6 [36], and ros1 [37], are in the Columbia (Col-0) background. Samples for ddm1 were prepared as part of Fluorescence-activated nuclear sorting (FANS) experiment (described below). All plants were grown to obtain three biological replicates.

Calli induction

Calli tissues were induced from Arabidopsis leaf explant as described before [38] with slight modification. Arabidopsis seeds were grown on ½ MS medium at 21 °C with 12/12 hr (day/night) light cycle for 3 weeks. Mature rosette leaves were harvested under sterile conditions and leaf strips of about 5 mm × 2 mm were cut with midvein going across the width. Similar size leaf strips were transferred to Arabidopsis callus induction medium (MS agar medium with 100 µl 1 mg/L 2,4-D and 100 µl 1 mg/L 6-BA). The plates were shifted back to the growth chamber at 21 °C with 12/12 hr (day/night) light cycle until sufficient callus induction. The calli tissues were harvested in liquid nitrogen for DNA extraction.

Fluorescence-activated nuclear sorting (FANS)

FANS was performed at the Gregor Mendel Institute as previously described [39]. In short, 300 mg of above-ground seedlings (7 d.a.g. ddm1−10 and Col-0, both with the pCLV3::H2BmCherry reporter [40] were collected into 3 mL of nuclei isolation buffer (NIB: 45 mM MgCl₂, 30 mM sodium citrate, 20 mM MOPS, 0.1% triton, 0.5% beta-mercaptoethanol, protease inhibitor cocktail (Roche #6538282001) and RiboLock (Thermo scientific #EO0384)) on ice (3 replicates each). Nuclei were isolated using the TissueRuptor (Qiagen #990890) and 30 μm filters (Sysmex #04-0042-2316). After centrifugation (1500 × g for 10 min at 4 °C), pellets were washed twice with 1 mL of NIB and resuspended in 2 mL of NIB buffer containing 5 μg/mL DAPI. mCherry and DAPI nuclei were sorted on a BD FACSAria III Cell Sorter (70 μm nozzle), and gates were adjusted using Col-0 wild-type samples as a reference. Nuclei were sorted into CTAB buffer (Biochemica #A4150) containing Rnase A (Thermo scientific #R1253). After sorting, the suspension was heated to 60 °C for 10 min, and chloroform/isoamyl alcohol (24:1) was added. Phases were mixed and centrifuged (14000 × g for 1 min). The upper aqueous phase was transferred to a new tube with 0.7 volume of cold isopropanol, 150 mM of NaAc, and 1 μL of glycogen (Thermo scientific #R0561) and incubated at −20 °C overnight. The next day, tubes were centrifuged (14000 × g for 10 min), and the pellets were washed twice with 70% ethanol. After removing ethanol, pellets were resuspended in 10 μL of nuclease-free TE buffer. DNA concentration was measured using the Picogreen dsDNA Assay Kit (Invitrogen #P11496).

Sampling and DNA extraction

For Col-0 WT, CS, HS, and mutant Arabidopsis plants, harvesting was done from MS agar plates in sterile conditions (3 replicates for each condition). Single plantlets were carefully removed from the agar plates and leaf tissues were harvested in liquid nitrogen. DNA extraction was performed using a CTAB (cetyltrimethylammonium bromide) protocol [41] followed by an ethanol precipitation step. DNA quality was measured on 1% agarose gel and with a NanoDrop spectrophotometer. DNA was quantified using a Quantus Fluorometer (Promega). All the following steps involving handling DNA were performed with wide-bore pipet tips.

Removing large linear DNA fragments

To remove large genomic linear DNA fragments and small degradation products, 5 μg of genomic DNA was purified using a Geneclean kit (MPBio) according to the manufacturer’s instructions with slight modifications. 5 μg DNA was dissolved in 20 μl water and mixed with 3 volumes (60 μl) NaI 6 M solution. Glassmilk suspension of silica matrix (MPBio), was thoroughly resuspended by vortexing for 1 min and 10 μl glassmilk was added to the DNA-NaI solution. The mixture was gently inverted and incubated at room temperature for 5 min, mixing every 1–2 min. DNA bound to glassmilk was pelleted by centrifugation and the pellet was washed twice with NEW wash solution (MPBio). After washing, the DNA was resuspended in 10 μl water and quantified again on a Quantus Fluorometer (Promega).

Extrachromosomal circular DNA enrichment

Rolling circle amplification (RCA) was performed to amplify eccDNA as previously described [42] with some modifications. A 20 μl reaction was set up using 20 ng of size-selected template DNA, 1 mM dNTPs, 10 U Phi29 DNA polymerase (Thermo Fisher), 50 μM Exo-resistant random primer (Thermo Fisher), 0.02 U inorganic pyrophosphatase (Thermo Fisher) and 1× Phi29 DNA polymerase buffer (Thermo Fisher). The reaction was run at 30 °C for 18 hr and stopped by heating to 65 °C for 2 min. DNA was precipitated by adding 0.1 volume of 3 M sodium acetate (pH 5.2), 2.5 volumes of ethanol, and 1 μl of glycogen (Thermo Fisher) and incubating overnight at −20 °C.

Resolving hyperbranched RCA amplicons

A series of DNA debranching, branch-release, and DNA repair reactions were used to resolve the hyperbranched RCA amplicons as described before [43] with some modifications. An amount of 10 μg of RCA amplified DNA was used in a debranching reaction with 5 U of Phi29 DNA polymerase (Thermo Fisher) without primers at 30 °C for 2 h and stopped by heating at 65 °C for 2 min. The product was precipitated with sodium acetate/ethanol as described above. The purified product was treated with 50 U S1 nuclease (Thermo Fisher) in a 20 μl reaction at 37 °C for 30 min and stopped by adding 3.3 μl of 0.5 M EDTA and heating at 70 °C for 10 min. DNA was purified by sodium acetate/ethanol precipitation. Purified DNA was treated with 3 U T4 DNA polymerase (New England Biolabs) and 10 U Escherichia coli DNA polymerase I (New England Biolabs) with 1× NEBuffer 2 and 1 mM dNTPs in a 50 μl reaction. The reaction was incubated at 25 °C for 1 h and stopped by heating at 75 °C for 20 min. Dephosphorylation was conducted by adding 5 U of Alkaline Phosphatase (Thermo Fisher) and incubating at 37 °C for 10 min, followed by stopping the reaction by heating at 75 °C for 5 min. The repaired DNA was purified using Agencourt AMPure XP Beads (Beckman Coulter) at a 1.5× volumetric ratio.

Libraries preparation and PacBio sequencing

A Bioanalyzer 2,100 12K DNA Chip assay (5067-1508, Agilent) was used to assess the fragment size distribution of the enriched DNA samples. The sequencing libraries were produced using the SMRTBell Barcoded Adapter Complete Prep Kit-96, following the manufacturer’s instructions (100-514-900. Pacific Biosciences). Approximately 200 ng of each DNA sample was end-repaired using T4 DNA Polymerase and T4 Polynucleotide Kinase according to the protocol supplied by Pacific Biosciences. A PacBio barcoded adapter was added to each sample via a blunt end ligation reaction. The samples were then pooled together and treated with exonucleases in order to create a SMRT bell template. Library fractions were quality inspected and quantified on the Agilent Bioanalyzer 12 kb DNA Chip and on a Qubit Fluorimeter respectively. A ready-to-sequence SMRTBell-Polymerase Complex was created using the P6 DNA/Polymerase binding kit 2.0 (100-236-500, Pacific Biosciences) according to the manufacturer’s instructions. The Pacific Biosciences RS2 instrument was programmed to load and sequence the samples on SMRT cell v3.0 (100-171-800, Pacific Biosciences), taking one movie of 360 min. A MagBead loading (100-133-600, Pacific Biosciences) method was chosen to improve the enrichment of longer DNA fragments. After the run, a sequencing report was generated for every cell via the SMRT portal to assess the adapter dimer contamination, sample loading efficiency, the obtained average read length, and the number of filtered sub-reads.

PacBio data analysis

Barcode-separated PacBio subreads were generated as the result of SMRT sequencing. The sequencing data was processed on SMRT Link v8.0 (Pacific Biosciences) for demultiplexing and extraction of circular consensus sequencing (ccs) reads using the following filtering criteria: minimum predicted accuracy = 99.9, and minimum read length of insert (in bases) = 3,000. Processing PacBio data on SMRT Link ccs enables the selection of reads with user-defined accuracy [44]; thus, the same analysis was repeated by changing the minimum predicted accuracy to 99.5 and 99.0 and only the reads with 99.9% accuracy post-ccs were selected for downstream analyses. Post-99.9-ccs PacBio reads (ccs reads) were processed on deconcat (https://github.com/devang-mehta/ciderseq2) as described before [30] and only the reads with more than one round of concatenation were retained, referred in this paper as eccDNA reads. Clustering analysis was performed using cd-hit [45] at default identity and tolerance to generate a non-redundant database.

EccDNA mapping and annotation

For eccDNA mapping on the Arabidopsis genome, Col-0 genome assembly was obtained from TAIR version 10 (https://www.arabidopsis.org/). Mapping was done using BLAST [46] with following parameters; -evalue = 1e-50, -num_threads = 20, -max_hsps = 1, -max_target_seqs = 1,000, -best_hit_overhang = 0.1, -best_hit_score_edge = 0.05 (S10 Table). A BLAST database was generated with all annotated Arabidopsis TEs obtained from TAIR version 9 (https://www.arabidopsis.org/) and eccDNAs were mapped using BLAST on this database to obtain percentages of partial and full-length TEs in eccDNAs; where full TEs represent complete TE within eccDNA and partial TE means the eccDNA does not cover the whole TE (eccDNA length < TE length; S9 Table). TE family and superfamily classification was obtained from TAIR (https://www.arabidopsis.org/browse/transposon_families). Mapping on the CDS database was done the same way by utilizing the tair9 cds dataset (S4 Table). Gene ontology analysis was done using ShinyGO 0.77 (http://bioinformatics.sdstate.edu/go/).

EccDNA quantification and downstream statistical analyses

For eccDNA quantification, chromosomes from Arabidopsis genome assembly tair10 were divided into 100 kb windows, and eccDNA reads were mapped to each window using BLAST [46], to obtain eccDNA counts per 100 kb. For downstream comparative analyses, the eccDNA counts were normalized using the formula ‘eccDNA counts per 100 kb window/total eccDNA reads from respective biological replicate × 100’.

Normalization of TE-derived eccDNA was performed using the formula: Normalized eccDNA abundance = (Number of TE-derived eccDNA reads/Total number of CCS reads in the sample) × 100. This normalization approach scales the eccDNA reads to a percentage of the total CCS reads in each sample, which effectively standardizes the data across different sequencing depths and biological replicates. This standardized approach mitigates bias due to the differing total number of reads obtained per sample, and allows for comparisons of eccDNA abundance across samples. Statistical analyses were then performed on these normalized values and this information was used in the downstream comparative analyses such as to calculate standard deviation, indicated by error bars in respective figures. Statistical analyses were conducted using ANOVA or t-tests, for the comparison of multiple groups simultaneously or to analyze data involving comparisons between two groups, respectively. Statistical significance was only accepted when the p-value < 0.05.

DNA blot analysis

Southern blots were performed as described previously [47] using an ONSEN-specific probe (S2 Table). For eccDNA chr3_14202454 and chr5_3253118, inverse primers were used to amplify the region encompassing the junction of eccDNA, and respective amplicons were used as probes (S2 Table). Ten micrograms of total gDNA and RCA-amplified DNA was digested by restriction with 20 U of PsiI and EcoRI (Thermo Fisher), respectively, in an overnight reaction. DNA was separated on a 1.5% agarose-TAE gel and transferred overnight onto a positively charged nylon membrane (Roche). This was followed by UV-crosslinking and hybridization with the ONSEN IR-DIG-labeled probe. DNA bands were visualized using alkaline phosphatase-conjugated anti-DIG (1:10,000) and CPD chemiluminescent substrate.

EccDNAs are predominantly derived from centromeric and pericentromeric regions and are enriched in TEs

A comprehensive map of the eccDNA population in Arabidopsis was generated using the CIDER-Seq pipeline (Fig 1a) [30] on DNA purified from Arabidopsis leaf tissues. In this de-novo assembly of high-accuracy circular DNA molecules, we identified on average 4,675 eccDNAs per sample (S1 Table; S1 Fig) including molecules as large as 22,431 nt (Fig 1b), with an average length of 3,681 nt (median 3,347; Fig 1c). We validated our approach by verifying a previously reported LTR/Copia retrotransposon, ATCOPIA78/ONSEN, that is activated upon heat stress (S2 Fig) [10,48]. Comparative analysis of eccDNA copies in Arabidopsis under heat stress (HS; 37 °C for 24 hr), control stress (CS; 4 °C for 24 hr), and no stress (NS; no temperature shift) indicated the accumulation of ONSEN eccDNAs upon heat stress (S3a Fig). We further verified a 5-fold increase of ONSEN copies in HS compared to CS (S3b Fig), as observed in other studies using independent molecular techniques [10,48] indicating that our eccDNA datasets can be used for quantitative analyses. Comparing the copies of younger and older ONSEN eccDNAs indicated that most eccDNAs (73%) are derived from younger ONSEN genomic loci (S3c Fig). Thus, the long-read eccDNA datasets allowed us to quantitatively study the population diversity of eccDNA-making TE copies and revealed the association between the evolutionary timescale, mutation accumulation, and TE mobilization. Finally, for additional independent validation of eccDNAs, we selected two previously uncharacterized eccDNA, chr3_14202454; and VANDAL18NA (chr5_3253118). By using a probe reverse-complementary to the junction region in chr3_14202454 and chr5_3253118 (S2 Table) in a DNA blot, we detected their presence in Arabidopsis leaf tissue (S4 Fig), providing an independent validation of TE mobilization via eccDNA intermediates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Dynamics and distribution of eccDNA in Arabidopsis.

(a) Schematic representation of eccDNA amplification and sequencing via CIDER-Seq pipeline. (b) EccDNA size distribution in Arabidopsis Col-0 leaf tissues (n = 4,158). The x-axis shows eccDNA size in kb and the y-axis shows the percentage of eccDNAs belonging to each window. (c) Size ranges of eccDNAs with marked median and quantiles (n = 4,190). (d) Percentages of eccDNA reads mapping to TAIR CDS, TE, or intergenic databases (values represent the average of three replicates). (e) Percentages of partial and full-length TEs in eccDNAs (values represent the average of three replicates); where full TEs represent complete TE within eccDNA and partial TE means the eccDNA does not cover the whole TE, i.e., eccDNA length < TE length. (f) Percentages of TE-derived eccDNA reads mapping to different Arabidopsis TE superfamilies (values represent the average of three replicates). (g) Cd-hit clustering analysis indicates that the majority of clusters (83.84%) are single-event clusters (n = 9,801). (h) EccDNAs mapped to Arabidopsis chromosomes. The x-axis indicates chromosome size in Mb and the y-axis indicates normalized eccDNA reads per 100 kb window. Heat stress (hs), control (cs), and non-stressed (ns) Arabidopsis samples (3 replicates for each condition). The raw data supporting all figures can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003275.g001

Since there is no formally accepted classification system for plant eccDNAs, we investigated the different types by mapping eccDNA reads on different TAIR datasets, such as coding sequences (CDS), intergenic regions, and TE (Fig 1d). These analyses revealed several interesting insights. Firstly, we identified the enrichment of gene ontologies related to the RNA metabolic processes (S3 and S4 Tables). Secondly, we observed that on average 53% of the TE-mapped eccDNA reads encompassed full-length TEs (Fig 1e, S5 Table). Furthermore, the majority of TE-containing eccDNAs belonged to the RC/Helitron (38%), followed by the DNA/MuDR, as observed in the TE superfamilies-based eccDNA mapping (Figs 1f; S5; S6 Table). Individual TEs with the most abundant eccDNA reads were also belonging to RC/Helitron (S7 Table). Further analysis of RC/Helitron indicated that, within the RC/Helitron superfamily, the most abundant TE families were ATREP3, ATREP4, ATREP10, and HELITRONY3 (S8 Table). Because helitrons are also the most abundant TEs inserted in the Arabidopsis genome [49], we checked if the TE families found in eccDNA are enriched more than expected to their respective composition in the Arabidopsis genome. In these analyses, we observed that while helitrons conformed to their expected genomic proportions in the Arabidopsis genome, other TE superfamilies, e.g., LTR/Gypsy TEs were significantly overrepresented (p-value < 0.05). Moreover, the ‘DNA’ TE superfamily, the fourth most abundant TE superfamily in the Arabidopsis genome, was not enriched in the eccDNA datasets (Fig 1f; S9 Table). These results, alongside the observations on the varied abundance of TE superfamilies in Arabidopsis mutants, suggest that TE distribution in eccDNA datasets does not directly mirror the overall TE composition of the Arabidopsis genome.

The clustering analyses of Arabidopsis eccDNAs identified 84% of single-event eccDNAs (Fig 1g), where single-event is defined as described by [29]; ‘single continuous genomic loci (continuous eccDNAs, self-circularization of a single genomic fragment)’. Our observation is comparable to the 89% of single-event eccDNAs observed in the mammalian cell lines [29]. Interestingly, the Arabidopsis eccDNAs were enriched in the centromeric and pericentromeric regions, as revealed by mapping the eccDNA datasets on Arabidopsis genome and evaluating eccDNA abundance per 100 kb bins on Arabidopsis chromosomes (Fig 1h, S10 Table). To further investigate the genomic context of eccDNA biogenesis, we specifically analyzed ATHILA family elements based on their chromosomal location. Our analysis revealed that a significantly higher number of eccDNAs originated from ATHILA elements located within the centromeric regions, compared to those from non-centromeric regions (S6a Fig). This enrichment was particularly striking for ATHILA6A and ATHILA6B elements, which are highly abundant in the centromeric region of chromosome 5 [50]. Consistent with this, we observed an overrepresentation of these families in eccDNAs derived from chromosome 5 (S6b Fig), indicating that centromeric ATHILA elements might be active contributors to eccDNA formation. Together, the generated eccDNA datasets revealed centromeric/pericentromeric regions as hotspots of Arabidopsis eccDNAs with several eccDNAs containing full-length TEs.

DNA methylation selectively regulates the production of EccDNAs from different TE classes

DNA methylation is one of the major mechanisms used by plants to control transposon activity [21]. Since we identified eccDNA-making TEs, we asked if there is a correlation between eccDNA loci identified in this study, DNA methylation (methyl C) [51], and active histone H3 lysine 4 methylation (H3K4me) sites [52]. This analysis indicated no significant correlation between eccDNA activation and methyl C or active H3K4me3 (S7 Fig). Correlation analysis with DNA methylation levels specifically in euchromatin, pericentromeric, and centromeric regions corroborated these results (S8 Fig). Since there can be multiple cellular eccDNA biogenesis mechanisms operating in a methylation-independent manner (e.g., in the case of apoptosis [29]), or methylation-dependent manner (e.g., in the case of TE-derived eccDNAs), we analyzed eccDNAs from independent TE superfamilies/families in representative Arabidopsis methylation mutants. Specifically, we performed eccDNA sequencing of rdr6, a knock-out of an RNA-dependent RNA polymerase involved in TE silencing [53]; dcl3, a mutant lacking 24nt siRNAs-generating dicer involved in RNA-directed DNA methylation (RdDM) [54]; and ros1, a knock-out of the Arabidopsis demethylase that limits the spread of DNA methylation at TE/repeat boundaries [37]. In these analyses we discovered several eccDNAs affected by RdDM, non-canonical DNA methylation or DNA hypermethylation from Arabidopsis TEs belonging to RC/Helitron (S9 Fig), LTR/Copia (S10 Fig), LTR/Gypsy (S11 Fig), DNA/MuDR (S12 Fig), DNA/En-Spm (S13 Fig) and LINE/L1 (S14 Fig) superfamilies. Because long inverted repeats (LIR) have been linked to TE activation, we analyzed if LIRs could promote the formation of eccDNAs by examining LIRBase [55]. We observed a significantly high number of LIRs in eccDNAs from methylation mutants compared to the wild-type (Fig 2a). We found that eccDNAs corresponding to LIR structures were more frequently detected in methylation mutants dcl3, ros1, and rdr6, but not in ddm1 (S15 Fig). This suggests that eccDNA formation from LIR fragments may rely more on non-CG methylation pathways rather than on the global heterochromatin decompaction observed in ddm1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Impaired DNA methylation shifts eccDNA profile.

(a) Long inverted repeats (LIR) identified in the eccDNAs of Col-0 and mutants dcl3, ros1, and rdr6. A higher number of LIRs are observed in eccDNAs from Arabidopsis methylation mutants, compared to the Col-0. (b) EccDNA from TE superfamilies in Col-0 WT and dcl3, rdr6, and ros1 mutants. y-axis indicates the normalized eccDNAs abundance belonging to each TE superfamily. (c) LTR/Gypsy-derived eccDNAs from Col-0 WT and dcl3, rdr6, and ros1 mutants mapped on Arabidopsis Chr 4 (3 replicates). (d) RC/Helitron eccDNA from Col-0 WT and dcl3, rdr6, and ros1 mutants mapped on Arabidopsis Chr 4 (3 replicates). (e) Mapping eccDNA-producing and non-eccDNA-producing LTR/Gypsy elements identified in ddm1 and mapped on Arabidopsis chromosomes. Gray dashed line indicated the centromere on respective chromosomes and the x-axis indicates chromosome size in Mb. No particular enrichment of eccDNA-making LTR/Gypsy is observed within or near the protein coding regions. The raw data supporting all figures can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003275.g002

Previous studies have shown that ROS1 removes methylation from TEs and their neighboring genes, thus, a ros1 mutant has a hypermethylated genome [56,57]. Our eccDNA profiling of ros1 indicated an overall increase in LTR/Gypsy (Fig 2b and 2c) and a decrease in RC/Helitron transposons (Fig 2d), while these TEs remained relatively stable in wt and dcl3 plants (Fig 2c and 2d). Hyperactivation of LTR/Gypsy-eccDNAs in ros1 mutant was surprising because previous studies have shown that ROS1 preferentially targets TEs closer to protein-coding genes [57] while the LTR/Gypsy TEs are enriched in the pericentromeric regions [50]. One explanation could be that the LTR/Gypsy TEs identified in ros1 eccDNA datasets, such as ATHILA elements (S6 Fig), belong to a subset of these TEs activated in the absence of ROS1 demethylase and thus could be concentrated in the neighboring regions of protein-coding genes away from the centromeres. However, comparing eccDNA-producing LTR/Gypsy elements identified in ros1 eccDNA datasets with all LTR/Gypsy TEs in the Arabidopsis genome indicated no particular eccDNA-making LTR/Gypsy enrichment in the neighboring regions of protein-coding genes (Fig 2e). These findings suggest a noncanonical TE control by DNA methylation pathways, potentially involving eccDNA biogenesis, warranting further investigation into the specific mechanisms by which ROS1 and other demethylases influence TE eccDNA formation and their potential roles in plant genome regulation.

Transcriptomic and eccDNA correlations in actively dividing Arabidopsis cells

To check if differential methylation associated with active cell division could alter eccDNA profiles, we tested the correlation between eccDNA abundance and gene expression in actively replicating pluripotent cells of Arabidopsis [58]. First, we generated the eccDNA profile of Arabidopsis calli and performed a comparative analysis of the relative eccDNA abundance in calli versus leaf tissues (S16 Fig). We then used a published dataset of Arabidopsis transcript fold change in calli versus leaf tissues [59] to investigate a potential correlation with eccDNA induction in calli. Despite a weak correlation between transcript levels and eccDNA in calli (Pearson’s R = 0.25; S17 Fig), the analysis revealed calli-enriched eccDNAs, and among them, several eccDNAs originated from Arabidopsis TE superfamily LTR/Copia (Fig 3a). Of particular interest was an LTR/Copia element ATCOPIA53, that has been linked previously with the retrotransposon-mediated control of self-incompatibility in polymorphic A. thaliana lines [60]. Likewise, ATCOPIA66’s noncanonical structure with short internal tandem repeats has been proposed to increase RNAi-mediated chromatin modifications [61]. The eccDNA enrichment and active transcription of those TEs particularly in calli call for further investigation of their potential role in cell differentiation/dedifferentiation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. EccDNA in Arabidopsis meristem and callus.

(a) TE-derived eccDNAs differentially regulated in Arabidopsis callus compared to leaf tissue with their respective p-values. (b) TE-derived eccDNAs in ddm1 mapped to representative TE superfamilies. EVD LTR/Copia with high transcript and eccDNA copy number is highlighted in red. (c) High copy number TE-derived eccDNAs with their respective p-values and transcript log₂ fold change in ddm1. ATCOPIA93 (EVD) and VANDAL21 have been previously identified in ddm1 datasets; the newly identified TEs are highlighted. (d) TE-derived eccDNAs differentially regulated in shoot apical meristems (mCherry +) compared to surrounding cells (mCherry −) obtained by fluorescence-activated nuclear sorting. The x-axis in panels a, c, and d indicates normalized eccDNA abundance. The raw data supporting all figures can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003275.g003

In the shoot apical meristem (SAM) of Arabidopsis, the dynamic transcription and DNA methylation patterns, particularly the transient increase in transposon expression and changes in CHG and CHH methylation, suggest an epigenetic reprogramming that may protect genome integrity in stem cells as they transition into the reproductive lineage, potentially involving eccDNA [40]. We therefore aimed at profiling eccDNAs in Arabidopsis SAM-enriched cells obtained using FANS [39,40] from the Arabidopsis mutant ddm1 [62] in which the altered access of DNA methyltransferases to condensed heterochromatin results in TEs activation [63,64]. We generated the eccDNA profiles of SAM and surrounding cells with respective controls (S18 Fig). Next, we utilized the published transcriptomic dataset of ddm1 [65] and performed correlation analyses between TE eccDNAs and transcripts in TE superfamilies (Fig 3b). As seen in the calli, while no overall correlation was found, we observed several TEs with both high eccDNA copy number and transcript expression in ddm1 (Figs 3b, 3c, and S19). Among these high copy number eccDNAs was a Ty1/Copia element ATCOPIA93 Évadé (EVD) (Fig 3b) that has been reported to be reactivated and mobilized upon release of epigenetic silencing in ddm1 mutant background [63,64]. Of note, these analyses also led us to identify new TEs that have not been characterized in previous ddm1 studies [19], such as VANDAL2 that belong to a subset of TEs defined as variable TEs (Fig 3c). ‘Variable TEs’, as defined by Dubin and colleagues, refer to a specific subset of TEs identified by their dynamic methylation patterns in response to environmental temperature fluctuations. These elements exhibit elevated CHH methylation at higher temperatures and are characterized by their euchromatic localization, high expression levels, and recent insertion into the genome, suggesting a role in rapid genomic adaptation to environmental changes [66]. Identification of variable TEs, therefore, provides a unique insight into the interaction between eccDNA, epigenetic modifications and environmental response in plants.

Finally, by focusing the analysis on TEs displaying significant eccDNA activation and transcriptional up-regulation, we could identify SAM-enriched TE eccDNAs (Figs 3d and S19). In these analyses, we identified several TEs detected as eccDNAs specifically in SAM such as ATREP19 and ATCOPIA58. Interestingly, like VANDAL2, ATREP19 has also been characterized as variable TE, indicating that the several properties associated with the variable TEs, particularly differential CHH methylation, may be linked to the eccDNA biogenesis. Likewise, ATCOPIA58 insertion and methylation state was detected in MethylC-seq data by the epiTEome program previously designed to analyze DNA methylation specifically in the context of TEs [67]. Because the Arabidopsis genome contains a small population of ATCOPIA58 elements, with only one putatively autonomous copy and four deletion derivatives [68], its specific detection in SAM might reflect altered epigenetic states or chromatin accessibility associated with the unique epigenetic landscape of stem cell-enriched tissues. Together, these findings suggest that eccDNA biogenesis in actively dividing and stem cell-enriched tissues reflects an interplay between TE structure, transcriptional activity, and dynamic epigenetic regulation.