Repressive Cytosine Methylation is a Marker of Viral Gene Transfer Across Divergent Eukaryotes

Abstract

Cytosine DNA methylation patterns vary widely across eukaryotes, with its ancestral roles being understood to have included both transposable element (TE) silencing and host gene regulation. To further explore these claims, in this study, we reevaluate the evolutionary origins of DNA methyltransferases and characterize the roles of cytosine methylation on underexplored lineages, including the amoebozoan Acanthamoeba castellanii, the glaucophyte Cyanophora paradoxa, and the heterolobosean Naegleria gruberi. Our analysis of DNA methyltransferase evolution reveals a rich ancestral eukaryotic repertoire, with several eukaryotic lineages likely subsequently acquiring enzymes through lateral gene transfer (LGT). In the three species examined, DNA methylation is enriched on young TEs and silenced genes, suggesting an ancestral repressive function, without the transcription-linked gene body methylation of plants and animals. Consistent with this link with silencing, methylated genomic regions co-localize with heterochromatin marks, including H3K9me3 and H3K27me3. Notably, the closest homologs of many of the silenced, methylated genes in diverse eukaryotes belong to viruses, including giant viruses. Given the widespread occurrence of this pattern across diverse eukaryotic groups, we propose that cytosine methylation was a silencing mechanism originally acquired from bacterial donors, which was used to mitigate the expression of both transposable and viral elements, and that this function may persist in creating a permissive atmosphere for LGT in diverse eukaryotic lineages. These findings further highlight the importance of epigenetic information to annotate eukaryotic genomes, as it helps delimit potentially adaptive LGTs from silenced parasitic elements.

Keywords: DNA methylation; epigenetics; eukaryotes; heterochromatin; lateral gene transfer; viral endogenization.

Fig. 1.

Evolutionary dynamics of DNMTs in the eukaryotes. a) Maximum likelihood phylogenetic tree of DNMTs, including eukaryotes, asgard archaea, viruses and bacteria. Each major eukaryotic DNMT family is highlighted with a color. Eukaryote orphan are eukaryotic sequences that do not cluster with any previously defined family. Gray branches depict viral sequences, brown branches depict Asgardarchaeota sequences, and black sequences depict bacteria. b) Distribution of eukaryotic DNMTs across the eukaryotes, based on phylogenetic profiling of panel a) and Dollo parsimony reconstruction of the ancestral state in each clade. Green cells depict presence, whereas yellow cells represent potential presence. Potential presence implies distinct domain architectures from “archetype” and low nodal branch supports. Black triangles highlight the two competing hypotheses regarding the root of extant eukaryotes. Raw data with per species DNMT repertoires can be found in supplementary fig. S1 and table S1, Supplementary Material online. An asterisk in animals indicates that few species are included in this analysis, therefore orphans are not ruled out. c) Protein domain architecture of the main eukaryotic DNMT families, as defined by PFAM domains. Each dot implies presence of the domain, but colored dots indicate that the domain is restricted to an eukaryotic lineage. d) Domain architectures of selected DNMT1 orthologues across major eukaryotic supergroups.

Fig. 2.

Cytosine methylome characteristics of three divergent eukaryotes. a) Global methylation levels at the four dinucleotide contexts as quantified with EM-seq in A. castellanii, C. paradoxa and N. gruberi. Dashed line indicates the cytosine methylation nonconversion rate (lambda genome control) in each EM-seq experiment. b) Average methylation levels on genes separated by their methylation status, with a threshold of ≥10% in N. gruberi, ≥ 20% in A. castellanii, and ≥50% in C. paradoxa. c) Distribution of transcriptional levels among methylated and unmethylated genes in each species, shown as the Transcript per Million+1, with a log10 transformed y axis. d) Average methylation levels on the three major classes of TEs (>500 bp) defined by RepeatModeler2, as well as unclassified repeats (Unknown), colored by legend below. Only methylated repeats (as per previous thresholds) with at least 20 methylated insertions are included. Heatmaps display methylation levels across hypermethylated regions of the A. castellanii e) and N. gruberi f) genomes, as measured by three independent technologies, alongside the average enrichment of histone posttranslational modifications, based on data from (Navarrete et al. 2025). 5mC levels in A. castellanii are for CG dinucleotides whereas in N. gruberi to all Cs.

Fig. 3.

Viral endogenization is marked by cytosine methylation across distantly related eukaryotes. a) Genome browser examples of hypermethylated genomic regions belonging to Giant Viral Endogenous Elements in three distantly related eukaryotes. Methylation levels range from 0% to 100%, red genes have best NCBI nr hits against viral sequences, whereas blue genes have other ancestries. GEVE potential origin defined by gathering the taxonomy of all the viral hits in the region. b) Potential taxonomic origin of genes in eight distantly related eukaryotes, separated by their methylation status. Asterisks indicate significant enrichment of a given taxonomic category in the methylated or unmethylated state (fisher exact's test P < 0.001 two sided).

Fig. 4.

Diagram of 5mC pattern evolution across the eukaryotes. Phylogenetic tree of eukaryotes is based on current consensus branching patterns, and main DNMT families are shown as numbered colored squares, showing gains in the branches where they appear. Only lineages for which direct whole genome 5mC data is available are shown. TE and GBM are subdivided as different cases, specified in the legend below. Methylated viral elements are shown for confirmed cases, in the case of bilaterians, it is restricted (so far) to endogenous retrovirus in vertebrates. Predominant 5mC sequence context in the lineage, while some exceptions exist (e.g. CH methylation in the vertebrate neural system [de Mendoza et al. 2021]). Main DNMTs found in each lineage, although secondary loss occurs in many instances (e.g. many animals have lost DNMT1/3). Co-localization with H3K9me2/3 is highlighted for lineages in which both 5mC and H3K9me2/3 have been empirically measured. Among bilaterians, only vertebrates have been shown to exhibit co-localization of 5mC with H3K9me3.