Comprehensive analyses of archaeal viral genomes reveal genomic characteristics, divergence, and host interactions

Abstract

Background

The ecological significance of bacteriophages has been extensively investigated, while the role of archaeal viruses across different environments remains poorly understood.

Results

Here, we present the Archaeal Viral Genome Database (AVGD), a comprehensive survey of archaeal viruses across eight distinct habitat types, including 3708 archaeal viral genomes, with genome sizes ranging from 3 to 188 kb, identified from 64,521,709 putative viral genomes using 40 public metagenomic datasets, an integrated public viral genome database (IGN), and pig gut viral databases. Our analysis revealed that the majority (92.93%) of archaeal viruses in the AVGD belong to the class Caudoviricetes. Phylogenetic analysis showed that many archaeal viruses diverged with their respective habitats. Using CRISPR spacer matching, we characterized the host composition of these archaeal viruses and uncovered competitive interaction networks between archaeal viruses and other archaeal viruses targeting the same host or different hosts. Furthermore, we identified 129,067 coding genes from 3708 archaeal viral genomes, most of which were associated with essential archaeal viral cellular functions, including replication, assembly, and packaging. Archaeal viruses also encoded a variety of auxiliary metabolic genes, anti-CRISPR (Acr) proteins for evading host immunity, and DNA methyltransferases for escaping host restriction-modification systems.

Conclusions

Together, this study provides a valuable resource and offers new insights into the ecological roles and host interactions of archaeal viruses across diverse environments.

Download video file ^{(42.8MB, mp4)}

Video Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-026-02445-2.

Background

Viruses are one of the most numerous and widely distributed microorganisms in the biosphere, playing important roles in microecosystems, host metabolism, and biogeochemical cycles [1, 2]. The ecological importance of bacteriophages within the extensive microbial niches has been extensively investigated; most of these studies have focused on those targeting bacteria, whether in the human gut or other ecosystems [3–5]. In particular, since the most common groups of archaeal viruses and bacteriophages both belong to the class Caudoviricetes [6], their discrimination remains challenging, especially since large-scale viral genome host prediction is rarely feasible in metagenomic studies [7], leaving viruses that infect archaea comparatively poorly understood. Archaea are widely distributed in various extreme environments as well as in the gastrointestinal tracts of humans and other mammals [8]. The host-associated archaeome of the human gut or ruminant mammals primarily comprises members of the Methanobacteriales, Methanomassiliicoccales, Methanocellales, and Methanosaetaceae [9]. Within host-associated microbial communities, they play roles as vital as those of bacteria, and their activities are closely linked to the health of humans and ruminants [10, 11]. In addition, the archaeal lineage Bathyarchaeia, which is abundant and ecologically versatile in extreme environments such as marine sediments, plays a pivotal role in the global carbon cycle [12]. Meanwhile, viruses infecting archaea are extremely diverse both in terms of genome sequences and viral particle structures [13, 14]. The viral isolates currently available are limited to a few culturable halophilic and thermophilic archaeal lineages [15]. Li et al. constructed the Human Gut Archaeal Viral Database (HGAVD) [16], comprising 1279 species of archaeal viruses. However, an evaluation by Cynthia et al. revealed the presence of some false-positive sequences within the dataset [17]. One major challenge is that, although metagenomic data have contributed to overcoming the lack of reference genomes due to the limited cultivation of archaeal viruses, the number of high-quality archaeal viral genomes remains constrained, possibly due to the limitations of existing bioinformatic approaches [18]. Moreover, unlike RNA viruses, which possess the universally conserved RNA-dependent RNA polymerase (RdRp) [14], the absence of a relatively conserved marker gene among archaeal viruses has hindered their detection, leading to potential false negatives in the identification of archaeal viruses. There is an urgent need for more high-quality genomic data to characterize the genomic features and divergence of archaeal viruses across diverse environments.

Prokaryotes possess CRISPR-Cas adaptive immune systems, which confer enhanced immunity to viruses and other genetic elements [19]. Previous studies have reported that approximately 40–50% of reference bacterial genomes and 85–90% of reference archaeal genomes contain CRISPR-Cas systems [20]. Despite this, the estimated distribution of CRISPR-Cas systems among prokaryotes remains controversial [21, 22]. Nevertheless, the viral sequences retained in the spacer regions of the CRISPR-Cas system can still help uncover the relationship between viruses and their host microbes. Additionally, it has been confirmed to be highly reliable in detecting pairwise viruses and their hosts by matching viral sequences from metagenomic sequencing data to the CRISPR spacers of known organisms [23]. Recently, CRISPR spacer searches have been widely used to identify viruses associated with various archaeal taxa, such as DPANN [24], Asgard [25, 26], anaerobic methanotrophic (ANME) [27], and methanogenic archaea [28], and have overcome the difficulty of characterizing the viruses infecting uncultured archaea. In addition, a growing number of novel defense systems continue to be discovered in both bacteria and archaea [29]. Several distinct CRISPR-Cas subtypes continue to be discovered in archaea, undoubtedly adding complexity to virus-host interactions [24]. The virus-host arms race has driven the evolution of diverse viral counter-defense mechanisms [30]. Bhoobalan-Chitty et al. reported that the archaeal virus Sulfolobus virus SIR2 encodes the anti-CRISPR protein AcrIIIB1 [31]. Additionally, He et al. described four families of archaeal lytic viruses that encode AcrID1 [32]. However, research on immune evasion strategies in archaeal viruses across different habitats remains limited. Until now, few studies have focused on distinguishing between bacteria and archaeal domains of predicted hosts [6], and have limited our understanding of archaeal virus diversity across different environments. Comprehensive data on archaeal viruses are lacking, which hinders our ability to characterize their genomic features, evolutionary diversity, and interactions with their hosts. This gap constrains our understanding of the distinct functional roles that archaeal viruses may play in various ecological contexts [30].

Here, we present the Archaeal Viral Genome Database (AVGD), a comprehensive survey of archaeal viruses across eight distinct habitats. Specifically, AVGD comprises 3708 archaeal viral genomes identified from viral contigs derived from 40 published studies and public databases (Additional file 2: Table S1). Based on this dataset, we further characterized the diversity, lifestyle, genomic architecture, functional potential, and divergence of archaeal viruses. We also conducted in-depth analyses of archaeal virus-host interactions and competitive networks among archaeal viruses. Together, the AVGD represents a valuable resource for exploring the roles of archaeal viruses across diverse ecosystems.

Methods

Collection of viral genome datasets and the construction of archaeal virus database

We first collected 59,652,008 viral genomes from our previous study (unpublished, Additional file 2: Table S1) based on eight studies of the pig gut microbiome, 38 published virome studies, and an integrated viral genome database (IGN). Among them, the integrated IGN database included the IMG/VR [33], GenBank [34], and the Nucleotide Sequence Database (NT, https://www.ncbi.nlm.nih.gov/nuccore) databases. Briefly, we combined and removed duplicate sequences using MMseqs2 [35] with the “easy-linclust -c 1.0 –min-seq-id 1.0” options to construct the integrated IGN database for contigs from the three databases. Engineered anaerobic systems such as biogas reactors, wastewater treatment plants, and anaerobic digesters are known to be enriched with methanogenic archaea and their associated viruses [36–38]. We downloaded raw metagenomic sequencing data from two previously published studies of engineered anaerobic systems from the NCBI Sequence Read Archive database (https://www.ncbi.nlm.nih.gov/sra) [36, 37], and 4,869,701 contigs were assembled using Megahit (v1.2.9) with the default parameters. These combined contigs encompassed eight defined habitat types, including rumen, pig guts, non-human primate guts, human guts, engineered anaerobic systems, soil and lake sediments, freshwater and meltwater, and ocean (Additional file 1: Fig. S1).

To minimize the likelihood of including false-positive viral genomes, we first used a pipeline developed by Nayfach et al. [39] for detecting viral genomes. Briefly, this pipeline integrates four viral signatures to identify viral genomes: (1) the presence of viral protein families, (2) the absence of microbial protein families, (3) the presence of viral nucleotide signatures, and (4) the occurrence of multiple adjacent genes on the same strand. Gene prediction for this filtering step was performed using Prodigal v2.50 [40] with the default parameters. Furthermore, to identify putative archaeal viral genomes, we developed a pipeline containing three steps to detect archaeal viral genomes. (1) We initially collected a total of 4767 archaeal metagenome-assembled genomes (MAGs), including 185 archaeal metagenome-assembled genomes from the pig gut microbiome, 1170 archaeal MAGs from the UHGG database [41], and 3412 archaeal MAGs from the GTDB database [42]. Subsequently, 90,138 CRISPR spacer sequences were predicted from 4767 archaeal MAGs using MinCED v0.4.2 (https://github.com/ctSkennerton/minced) with default parameters. (2) To further enhance the recovery of archaeal viral genomes from these viral genomes, we incorporated 168,721 CRISPR spacers of methanogenic archaea predicted by Medvedeva et al. [28] and 9312 CRISPR spacers of DPANN archaea predicted using MinCED from 3662 DPANN archaea genomes provided by Wu et al. [24]. We identified putative archaeal viral genomes using blastn from the blast + package [43] with the “-evalue 1e-5 -word_size 8 -task blastn-short” options based on matches to CRISPR spacer sequences [16]. To balance sensitivity with accuracy in viral prediction, we adopted matching thresholds established in previous bacterial and archaeal viral studies [24, 44–46] to minimize false positives while retaining authentic virus-host associations, and retained those putative archaeal viral genomes with at least 95% coverage and a maximum of one mismatch. (3) To remove false-positive archaeal viral genomes, we compiled 8485 archaeal viral hallmark genes predicted by Li et al. [16], and proteins of all putative archaeal viruses were identified using diamond [47] blastp with the “-e 1e-5” options. Those aligned to archaeal viral hallmark genes were included for subsequent analysis as potential archaeal viruses, while those not aligned were excluded. Among these, the 8485 archaeal viral hallmark genes were predicted from archaeal proviruses and reference archaeal viral genomes based on the Pfam [48] and the Virus Orthologous Groups (VOG, release 202, http://vogdb.org) databases. To further remove false-positive archaeal viral genomes, we incorporated 3,912,325 bacterial CRISPR spacers predicted from bacterial genomes in the UHGG, GTDB, and the pig gut, to match all putative archaeal viral genomes using blastn from the blast + package with the “-evalue 1e-5 -word_size 8 -task blastn-short” options, and removed those genomes matched at least three bacterial spacers with at least 95% coverage and a maximum of one mismatch. Furthermore, we manually assessed and removed those genomes exhibiting distinct bacteriophage characteristics, including known bacteriophage taxonomy families and bacteriophage core function. Subsequently, all identified archaeal viral genomes were dereplicated using MMseqs2 [35] with the “easy-linclust -c 1.0 –min-seq-id 1.0” options based on archaeal viral genomes from the same habitat type. We then assessed genome completeness, detected proviruses, and removed host bacterial sequence contamination using CheckV v0.8.1 [49]. To evaluate whether any of the genomes might represent bacterial false positives, we assessed bacterial and viral gene content by quantifying hits to bacterial universal single-copy orthologs (BUSCO) [50] and the curated viral protein family modules (VPFs) [51]. Notably, the BUSCO analysis served strictly for contamination detection and was not intended to assess viral genome completeness. After rigorous filtering and quality assessment, 3708 high-confidence archaeal viral genomes (AVGD) were retained for downstream analysis.

Lifestyle prediction and taxonomy assignment of archaeal viral genomes

The lifestyles of identified archaeal viruses were predicted using two complementary tools: BACPHLIP v0.9.3 [52] and VIBRANT v1.2.1 [53]. Firstly, similar to previous strategies used to infer viral lifestyles in the human gut virome [39], we employed BACPHLIP to predict the lifestyles of archaeal viruses, despite the absence of a dedicated prediction tool for this viral group. This approach was motivated by the observation that temperate bacteriophages and temperate archaeal viruses share several common features. Nevertheless, it should be noted that BACPHLIP was originally trained on characteristics of temperate bacteriophages, which may introduce bias when applied to archaeal viruses. BACPHLIP classifies viruses based on a scoring system derived from integrase and other prophage-associated genes, with scores > 0.9 indicating temperate, 0.5–0.9 uncertain, and < 0.5 virulent. To address potential biases from BACPHLIP’s bacteriophage training data, we additionally ran VIBRANT with default parameters, which leverages curated viral protein families (including those from archaeal viruses) and a machine learning framework. To ensure high-confidence predictions, we retained only those archaeal viral genomes for which the lifestyle assignments from BACPHLIP and VIBRANT were concordant. For taxonomic annotation of all identified archaeal viruses, we used geNomad v1.7.4 [54], which assigns taxonomy based on viral marker genes and encompasses most viral lineages currently recognized by the International Committee on Taxonomy of Viruses (ICTV).

Identifying alternative genetic codes in archaeal viral genomes

Prodigal v2.50 [40] was used to identify open reading frames of protein-coding genes in 3708 archaeal viral genomes (AVGD) under the conditions of the standard genetic code (code 11) and three alternative genetic codes: TAG recoding (code 15), TAA recoding (code 90), and TGA recoding (code 91) as described by Nayfach et al. [39]. Briefly, for an archaeal virus with a genome size < 100 kb, if its protein-coding density with the genetic codes 15, 90, or 91 increased > 10% compared to that with the standard genetic code 11, we considered that this archaeal viral genome tended to use the corresponding alternative genetic code [39, 55]. For those archaeal viruses with a genome size ≥ 100 kb, the threshold for considering the utilization of an alternative genetic code was the increase of protein-coding density > 5%.

Clustering archaeal viral genomes into species-, genus-, and family-level viral clusters

The reconstructed 3708 archaeal viral genomes (AVGD) were clustered into species-level viral clusters (sVCs) at the threshold of 95% ANI and ≥ 85% of genome coverage using the greedy and centroid-based algorithm developed by Nayfach et al. [39]. We further clustered archaeal viral genomes into genus-level viral clusters (gVCs) and family-level viral clusters (fVCs) based on average amino acid identity (AAI) and gene sharing information. First, protein-coding genes were predicted from all archaeal viral genomes using Prodigal v2.50 with alternative genetic codes, and protein alignments were performed using Diamond blastp v2.1.9.163. Then, for each pair of archaeal viral genomes, we identified their shared genes and calculated the AAI and the percentage of shared genes. For the genus-level viral clustering analysis, those genome pairs with > 50% AAI or > 20% shared genes were filtered. The MCL v14-137 [56] was used to perform the genus-level viral clustering with an inflation factor of 2.0. The family-level clustering was performed with those genome pairs with < 20% AAI or < 10% shared genes at the threshold of an inflation factor of 1.2 using the MCL v14-137.

Identification of potential novel archaeal viral genomes and divergence analysis of archaeal viral genomes with habitat types

Considering that some sequences in our AVGD represent their first report as archaeal viruses, we were therefore particularly interested in comparing these genomes to those in the integrated IGN database to assess their novelty. Specifically, to identify novel archaeal viral genomes, we performed clustering at the species-level (sVCs, ANI-based), genus-level (gVCs, AAI-based), and family-level (fVCs, AAI-based). Following the ANI/AAI clustering approach, an archaeal viral genome was classified as novel at a given level if it did not cluster with any archaeal viral genome from the integrated IGN database.

For those gVCs that contained at least 4 archaeal viral genomes and existed in at least two habitat types, a phylogenetic tree was constructed for each gVCs [57]. Briefly, protein-coding genes were predicted from archaeal viral genomes. Core genes of each archaeal viral gVCs were identified with Roary v1.7.8 [58] with the default options, and a multi-FASTA alignment was created. Subsequently, phylogenetic trees were built using IQ-TREE2 v2.1.3 [59] with the “-m MFP -B 1000 –bnni” parameters, visualized and annotated using iTOL (https://itol.embl.de/). The branch length between any two archaeal viral genomes was then calculated for each archaeal viral gVCs. The branch lengths between archaeal viral genomes from the same habitat types and those from different habitat types were compared with two single-tailed Wilcoxon rank sum tests. To account for multiple comparisons across viral clusters, we applied Benjamini–Hochberg false discovery rate (FDR) correction. If the branch lengths among archaeal viral genomes in a gVC from different habitat types were significantly longer than those from the same habitat types (P < 0.05 after FDR correction), we considered that this archaeal viral gVC exhibited habitat divergence.

Host prediction for archaeal viral genomes, potential interactions among archaeal viruses, and defense system prediction for hosts

To accurately investigate archaeal viruses with a broad host range, we performed host prediction using CRISPR spacer matching. Specifically, CRISPR spacers were extracted from the 4767 archaeal MAGs using MinCED (https://github.com/ctSkennerton/minced) with default options. All extracted CRISPR spacers were then aligned to archaeal viral genomes using BLASTN with the option “-evalue 1e-5 -word_size 8 -task blastn-short,” and if a CRISPR spacer was aligned to an archaeal viral genome with a maximum of one mismatch and 95% coverage [24], the archaeal taxa identified that the CRISPR spacer was treated as the host of the corresponding archaeal virus. Additionally, to investigate potential interactions among archaeal viruses, all archaeal viruses were processed using MinCED with default options to discover CRISPR spacer arrays. All archaeal viruses were then queried with BLASTN against a database of the CRISPR spacers. Hits were kept if a viral CRISPR spacer was mapped to an archaeal viral genome with a maximum of one mismatch and 95% coverage [24]. Furthermore, antiviral defense systems in archaeal genomes (hosts) were detected using DefenseFinder v1.2.0 [22] (-dbtype gembase) based on MacSyFinder model v1.2.0 [60].

Functional annotation of archaeal viral genomes

A total of 129,067 protein-coding genes were predicted from 3708 archaeal viral genomes using Prodigal v2.50 with alternative genetic codes. The function of these genes was annotated based on HMM searches against the KEGG [61], Pfam-A [48], PHROG [62], and arCOGs [63] protein family databases. All searches were performed using the hmmsearch utility in the HMMER package v.3.1b2 with the “-E 1e-10” option [64], and each gene was annotated by each database according to its top bit score alignment.

Archaeal viruses encoding auxiliary metabolic genes (AMGs) were annotated using the VIBRANT v1.2.1 [53] and DRAM-v v1.3.5 [65]. Briefly, proteins encoded by archaeal viral genomes were first scored by VirSorter2 v2.2.2 [66], and then, the scored proteins were annotated using DRAM-v with default options. AMGs were also annotated using the VIBRANT and assigned to the metabolic pathways using the KEGG database [67]. Only those AMGs annotated by both the VIBRANT and DRAM-v were retained for further analyses.

Construction of phylogenetic trees for archaeal viral genomes of Caudoviricetes.

A large-scale phylogenetic analysis was conducted to evaluate the phylogenetic novelty and contribution of our archaeal viral genomes to Caudoviricetes diversity [39]. The phylogenetic tree of Caudoviricetes genomes was constructed using the method described by Low et al. [68]. This approach allowed us to quantify the phylogenetic distance (PD) between our genomes and reference genomes, thereby assessing the expansion of the known evolutionary landscape. To prevent the archaeal viruses from being over-represented, which can skew the underlying alignment and ultimately the tree, we first clustered combined Caudoviricetes genomes from this study and the RefSeq database. Briefly, we collected 4463 complete Caudoviricetes genomes from the RefSeq database and 3446 Caudoviricetes genomes from this study, and clustered combined Caudoviricetes genomes using CD-HIT (http://cd-hit.org/) with the “-d 0 -c 0.95” option. Then, we identified a set of 77 gene markers of representative Caudoviricetes genomes from the predicted protein sequences based on individually searching against HMM profiles for the 77 markers using HMMER [39, 68]. We trimmed and concatenated individual marker alignments to retain those genome fragments with less than 50% gaps using trimAl v1.4.rev22 [69]. We only kept those viral genomes containing at least three markers and existing in > 5% of alignment columns. Finally, the phylogenetic tree was constructed using IQ-TREE2 (-m MFP -B 1000 -bnni) under the LG + GAMMA model, followed by 1000 nonparametric bootstrap replicates to estimate branch support and visualized using iTOL (https://itol.embl.de/) [70]. Finally, the phylogenetic distances between genomes were computed from the resulting tree branch lengths using the ape v5.7–1 [71] and picante v1.8.2 [72] packages in R.

Gene-sharing network analysis for all archaeal viruses in the AVGD

A gene-sharing network was constructed using vCONTACT v2.0 [73], containing the genomes of archaeal viruses identified in this study, previously reported 22 DPANN viruses [24], and prokaryotic viruses from the NCBI RefSeq database (release 211). Cytoscape v3.9.0 [74] was used for visualizing the network with an edge-weighted spring-embedded model.

Statistical analysis

All statistical analyses and data visualization were performed using the packages in R v4.2.1.

Results

Construction of an archaeal viral genome database using viral genomes from different habitats

To comprehensively uncover the composition and diversity of archaeal viruses in the gut microbial community of pigs, we identified genomes or genome fragments of archaeal viruses from the pig gut viral genome database (PGV) constructed in our laboratory, which included 131,673 medium and high-quality viral genomes or genome fragments, as evaluated by CheckV v0.8.1. Among these, these 131,673 genomes comprised viral genomes with medium quality (n = 66,474, 50–90% complete), viral genomes with high quality (n = 15,245, > 90% complete), and complete viral genomes (n = 49,954, 100% complete). We also recovered genome fragments of archaeal viruses from metagenomic sequencing data of pig feces samples from eight countries (Australia, Canada, China, Denmark, France, Ghana, Ireland, and the USA) in the repository databases (Additional file 3: Table S2). Furthermore, to enhance the comprehensiveness of the archaeal viral genome database, viral genomes from 40 public datasets spanning eight different habitats (rumen, pig guts, non-human primate guts, human guts, engineered anaerobic systems, soil and lake sediments, freshwater and meltwater, and ocean) and from an integrated viral database (IGN) including the IMG/VR [33], GenBank [34], and the Nucleotide Sequence Database (NT, https://www.ncbi.nlm.nih.gov/nuccore) databases were also collected. We obtained a total of 9,671,535 putative viral genomes through a custom virus identification process [39] based on previously assembled viral genomes or contigs using the Megahit (Fig. 1a). An approach integrating the CRISPR-Cas spacer sequence matching and archaeal viral hallmark gene identification was used to identify archaeal viral genomes from viral genomes [16, 28]. After rigorous screening and the removal of false-positive viral contigs, a total of 3708 archaeal viral genomes (defined as the archaeal viral genome database, AVGD) were obtained and used for further analysis. Archaeal viral genomes were identified in viral genome datasets from six different habitat types (Fig. 1b), but no archaeal viral genomes were identified from the habitats “Freshwater and meltwater” and “Non-human primate guts.” The proportion of archaeal viral genomes in total viral genomes identified in datasets among eight habitat types varied greatly, with a higher proportion in the engineered anaerobic systems (0.269%, 173/64,300), rumen (0.098%, 232/235,909), human gut (0.055%, 293/535,705), and pig gut (0.046%, 751/1,637,933) (Additional file 1: Fig. S1), highlighting the widespread prevalence of archaeal viruses in diverse environments. This supports our investigation into the diversity of archaeal viruses across diverse habitats.

Fig. 1.

Identification, characterization, and annotation of archaeal viral genomes from different habitats. a An overview of the workflow for the identification of archaeal viral genomes. b The distribution of archaeal viral genomes in the AVGD database among different habitats. c The lifestyles of 3708 archaeal viral genomes in the AVGD database. The different colors represent different lifestyles in the stacked bar for 3708 archaeal viruses. d The quality and length distribution of archaeal viral genomes in the AVGD database. e The taxonomic distribution of 3708 archaeal viral genomes from different habitats in the AVGD database at different taxonomic levels. The pie charts show the proportion of known taxa at each taxonomic level for 3708 archaeal viral genomes. The Sankey diagram represents the number and detailed taxonomies of 3708 archaeal viruses at different taxonomic levels. The gray lines connect the archaeal viruses across taxonomic levels. Line width corresponds to the number of viruses in each taxonomic level

We used both BACPHLIP v0.9.3 and VIBRANT v1.2.1 [53] to predict the lifestyle of 3708 archaeal viruses. The archaeal viruses in the AVGD had a higher proportion of virulent archaeal viruses (n = 1800, 58.54%) than temperate archaeal viruses (Fig. 1c). Interestingly, proviruses (n = 53, 1.32%) belonging to archaeal viruses were identified across distinct habitat types. This pattern could suggest an extensive history of virus-archaea interaction. However, a more parsimonious explanation might be that these proviruses were acquired prior to the geographic and lineage divergence of their hosts.

We then evaluated the completeness of all archaeal viral genomes (n = 3708) using CheckV v0.8.1 and found that the AVGD contained 566 (15.26%) genomes with medium to high quality (Fig. 1d). To further evaluate the taxonomic composition of archaeal viral genomes in the AVGD, we annotated 3708 archaeal viral genomes using the geNomad v1.7.4 classification software. A total of 3499 (94.12%) archaeal viral genomes could be taxonomically annotated (Fig. 1e and Additional file 4: Table S3), and the taxonomic composition of archaeal viruses varied across habitat types, suggesting a large heterogeneity of archaeal viral compositions in different habitats. We found that only 3.72% archaeal viral genomes could be classified into known families, such as Pleolipoviridae, Hafunaviridae, and Vertoviridae, while 92.93% genomes were known to belong to Caudoviricetes. These findings indicated the challenges of taxonomic annotations of archaeal viral genomes at the current stage. Furthermore, despite our systematic screening, no archaeal viral genome was found to utilize a genetic code other than the standard genetic code 11 (Additional file 4: Table S3).

Expanded diversity of archaeal viruses and divergence with habitat types using the AVGD

To evaluate the genomic diversity of archaeal viruses in the AVGD, a total of 3708 archaeal viral genomes were classified into 1806 species-level viral clusters using a threshold of 95% average nucleotide identity (ANI) and 85% coverage [39]. Notably, 328 (18.16%), 213 (11.79%), 80 (4.43%), and 40 (2.21%) species-level viral clusters were specifically identified in the pig gut, rumen, engineered anaerobic system, and human gut samples, respectively. We found that 1150 archaeal viral genomes (31.01%) belonging to 731 species-level viral clusters (40.28%) in the AVGD were not clustered with any archaeal viral genomes in the integrated IGN database, and these genomes were considered novel archaeal viral sequences in AVGD (Fig. 2a). We further clustered archaeal viral genomes in the AVGD into genus-level viral clusters using a threshold of 50% average amino acid identity (AAI) or > 20% of shared genes and used an inflation factor of 2.0, and into family-level viral clusters using a threshold of > 20% AAI or > 10% shared genes with an inflation factor of 1.2 [39]. A total of 850 genus-level viral clusters and 383 family-level archaeal viral clusters were obtained, and we described the range of archaeal viral genome numbers within singleton genera (1–160 genomes) and families (1–424 genomes). Among them, 75 genus-level (8.82%) viral clusters and 25 family-level (6.53%) viral clusters of viruses from the pig gut samples, 9 genus-level (2.35%) viral clusters and 4 family-level (1.04%) viral clusters from the human gut samples, exhibited no overlap with archaeal viral genomes in the integrated IGN database and the datasets from other habitat types. This absence of shared genomic signatures suggests the discovery of novel archaeal viral genomes, including habitat-specific genus- and family-level clusters associated with pig gut and human gut ecosystems (Additional file 1: Figs. S2a and S2b). Archaeal viruses identified in the pig gut and human gut made up a significant portion of the AVGD, and we focused on the genomic clusters between archaeal viruses from human and pig intestinal samples. We further found that some habitat-specific archaeal viruses were found between the human and pig gut samples at the cluster levels, including species-, genus, and family-level clusters (Additional file 1: Fig. S3a). Furthermore, accumulation curve analysis revealed that the numbers of family-level, genus-level, and species-level viral genome clusters did not reach saturation using the AVGD (Fig. 2a, Additional file 1: Figs. S2a and S2b), highlighting the great diversity of archaeal viruses and suggesting that additional samples from diverse sources were needed to capture a more comprehensive range of archaeal viral taxa.

Fig. 2.

Expanded diversity and divergence of archaeal viruses with habitat types. a The number of known and novel archaeal viral genomes identified in this study, and the distributions of archaeal viruses among different habitats at the species-level viral genome clusters. The pie chart represents the proportions of novel and known archaeal viral genomes in the AVGD at the species-level viral genome clusters. The bar graph shows the number of archaeal viruses that are unique to or shared among different habitat types. The accumulation curve indicates the relationship between the number of archaeal viral genomes and the corresponding number of species-level viral genome clusters. b Divergence of archaeal viral genomes with habitat types based on genus-level clusters. The dot plot shows the likelihood (P values) distributions of divergence relationships between 10 genus-level viral clusters and habitat types. The pie chart shows the proportion of divergent and non-divergent genus-level viral clusters. c Representative examples of genus-level viral genome clusters that exhibit divergence with habitat types (clusters 1 and 2), and those that do not diverge with habitat types (cluster 3)

Given that the AVGD database contains archaeal viruses from diverse habitats, it provides an opportunity to investigate whether the divergence of archaeal viral genomes is influenced by their associated environments. Considering the heterogeneity and diversity of archaeal viruses, we selected genus-level viral genome clusters with relatively broad and even distribution for subsequent analyses. Specifically, 10 genus-level viral genome clusters, each comprising at least four archaeal viral genomes and present in at least two distinct habitat types, were included in the analysis. For each genus-level viral genome cluster, we evaluated whether archaeal viral genomes originating from the same habitat type were more closely related than those from different habitat types by constructing a phylogenetic tree [57]. We observed that, in 70.00% of genus-level viral genomics clusters analyzed (n = 7) (Additional file 5: Table S4), archaeal viral genomes from the same habitat were clustered together in the phylogenetic tree and exhibited significantly closer evolutionary relationships compared to those from different habitat types. This finding suggests that a substantial proportion of archaeal viruses may have diverged with their respective habitat types (Fig. 2b). The examples of genus-level viral genome clusters showing habitat divergence (genus-level viral clusters 1 and 2) and the clusters not exhibiting habitat divergence (genus-level viral cluster 3) were presented (Fig. 2c). These results indicated the heterogeneity of habitat divergence among archaeal viral taxa.

Identification of potential hosts of archaeal viruses and insights into virus-host interactions

The distribution and composition of phages mainly depend on their host microbes. Conversely, phage communities can also significantly influence the structure and function of bacterial and archaeal communities [16, 45]. Matching CRISPR spacers has been used to link viral genomes with their putative archaeal hosts [16, 24]. In this study, 2150 archaeal viral genomes (57.98% of the total 3708) were successfully matched to one or multiple known archaea hosts with detailed host taxonomy (Additional file 6: Table S5). Among these, 276 viral genomes (7.44%) were targeted with multiple archaeal genera (Fig. 3a). The other 1874 archaeal viral genomes (50.54%) only targeted one archaeal genus. Hosts for these archaeal viruses targeted one archaeal genus mainly belonged to Methanobrevibacter A, Methanobrevibacter, and Halolamina, which are keystone archaea in the gut [10] microbiome and hypersaline ecosystems [75]. In more detail, the 2150 matched archaeal viral genomes were linked to 796 archaeal MAGs spanning 11 archaeal phyla, suggesting broad interactions between viruses and host archaea. The phyla Methanobacteriota, Halobacteriota, and Thermoplasmatota harbored the highest number of viral hosts identified in this study (Additional file 1: Fig. S2c). Furthermore, the Methanobrevibacter A genus was the most common archaeal host genus for archaeal viruses from the pig gut and human gut samples. Comparing hosts of archaeal viruses from human guts, archaeal viruses from pig guts commonly infected the genera Methanobrevibacter, Methanobrevibacter B, and MX-02 (Additional file 1: Fig. S3b).

Fig. 3.

Identification of potential hosts and virus-host interactions for archaeal viruses. a The proportion of archaeal viruses infecting ≥ 2 archaeal genera and infecting 1 archaeal genus among 3708 archaeal viral genomes (pie chart), distribution of the number of putative host genera targeted by archaeal viruses infecting ≥ 2 archaeal genera (box plot), and the distribution of putative archaeal hosts for archaeal viruses infecting 1 archaeal genus (cycle diagram). b The network diagram of archaeal virus-methanogen interaction for archaeal viruses in the AVGD. c Genomic distribution and 3D structure of anti-CRISPR (Acrs) proteins and DNA methyltransferases encoded by archaeal viral genomes

Methanogenic archaea (methanogens) are key producers of methane, a potent greenhouse gas and biofuel, and are widely distributed across diverse environments [28]. Although viruses are likely to influence the ecophysiology of methanogens, they remain largely uncharacterized. In this study, we identified 1035 archaeal viruses in the AVGD that target methanogenic archaea, including Methanobrevibacter_A, Methanobrevibacter, and Methanothermobacter based on the methanogenic-derived CRISPR spacer matches (Fig. 3b and Additional file 7: Table S6). Interestingly, 422 of these viruses were capable of targeting multiple methanogenic archaea species. The comprehensive characterization of archaeal viruses within the AVGD provides important insights into their potential roles in shaping methanogen communities and lays the foundation for developing novel tools for methanogen research and management.

These virus-host interactions occur within the context of a constant evolutionary arms race [76]. Our analysis confirmed that the linked archaeal hosts possess diverse antiviral defense systems, including CRISPR-Cas and restriction-modification systems (Additional file 1: Fig. S2c). Among CRISPR-Cas systems, the complex multi-protein Class 1 type was far more common than the single-effector Class 2 systems [24]. Additionally, a diverse suite of other defense systems was prevalent. Restriction-modification (RM) systems were particularly widespread, present in 85.93% of host archaeal genomes. Alongside these, toxin-antitoxin systems like ShosTA, which can trigger programmed cell death to halt viral spread, were also identified in 15.95% of host genomes. In turn, interrogation of the AVGD revealed that archaeal viruses encoded potential counter-defense proteins, such as anti-CRISPR proteins (Acrs) and DNA methyltransferases (Fig. 3c), which were identified in 292 and 295 viral genomes, respectively. The presence of these counter-defense genes within viruses linked to specific hosts offers molecular clues to the dynamics of their interaction.

In summary, the AVGD enables a detailed, genome-resolved perspective on archaeal virus-host interactions. By establishing potential host links and highlighting specific, ecologically significant relationships, such as those with methanogens, this resource lays a critical foundation for future experimental and ecological studies aimed at understanding the role of viruses in archaeal-driven processes.

CRISPR spacer analysis reveals competitive networks among archaeal viruses and their hosts

CRISPR-Cas systems are adaptive immune systems widespread in archaea. Previous studies have reported that some phages encode their own CRISPR arrays, and some phages have intact and functional CRISPR-Cas systems; however, our current understanding of CRISPR-Cas systems encoded with archaeal viral genomes remains limited [24, 77]. To address this gap, we analyzed archaeal viral genomes and identified 38 viral genomes that encoded CRISPR systems. It is known that phage-encoded CRISPR-Cas systems are typically associated with mechanisms that eliminate competing phages or mobile genetic elements within the host cell [78]. Based on sequence matches between virus-encoded spacer regions and other viral genomes, it appears that these CRISPR-Cas systems may be involved in competing with other phages for the same archaeal hosts [79]. Interestingly, we identified a relatively complete CRISPR-Cas system in one archaeal viral genome, while the majority of viral genomes encoded only orphan Cas proteins (Fig. 4a). These orphan systems lacked components of the spacer acquisition module, such as Cas1, Cas2, and Cas4. The absence of these genes may be the result of partial horizontal gene transfer events among viruses [24].

Fig. 4.

Revealing potential interactions and competitive networks among archaeal viruses. a An overview of the process of identifying CRISPR-Cas systems in archaeal viral genomes and potential interactions among archaeal viruses. b The network diagram illustrating archaeal virus-archaeal virus interactions. Each node represents an individual archaeal virus, with node colors indicating different viral taxonomic groups. Edge colors distinguish between competitive and non-competitive interactions, while edge styles differentiate one-target pairs and two-target pairs

Viruses can carry a wide variety of CRISPR spacers [79–81]; however, competitive networks among archaeal viruses have rarely been characterized. In this study, a total of 191 CRISPR spacers were detected across 38 archaeal viral genomes in the AVGD. Among these, 72 spacers (37.70%) were matched to 36 archaeal viral genomes, suggesting the existence of complex networks of archaeal virus-archaeal virus interaction (Fig. 4b and Additional file 8: Table S7). When comparing the “one target pairs” (defined as spacers encoded by one archaeal virus could match another archaeal viral genome, while the other archaeal viral genome either did not encode spacers or encoded spacers that did not match this archaeal viral genome, 33.33%, n = 38), the “two target pairs” (defined as two archaeal viruses encoding spacers that could alternatively match another archaeal viral genome) was found to have a high proportion (66.67%, n = 76). Furthermore, 66.67% (n = 76) of these archaeal virus-archaeal virus pairs shared the same archaeal host and were classified as competitive archaeal viruses.

Functional potentials of archaeal viral genomes

Although the functional potentials of gut phages have been extensively reported [39, 82], the functional capacities of archaeal viruses from diverse habitats remain largely unknown. In this study, we identified 129,067 coding genes from 3708 archaeal viral genomes and annotated them using the PHROG, KEGG, Pfam-A, and arCOGs database (Fig. 5a). Overall, 60.58% of archaeal viral genes had no significant matches in any databases above, 9.19% were matched to the hits in the databases but could not be assigned to any known biological function categories, and 30.23% of archaeal viral genes could be functionally annotated (Fig. 5b), implying the largely unexplored functional diversity of archaeal viruses.

Fig. 5.

Functional annotations of archaeal viral genomes in the AVGD. a The proportion of archaeal viral genes matched each of the four databases. b Overview of the functional annotation for 129,067 archaeal viral genes. c The top 50 functional categories based on the number of archaeal viral genes; different colors represent different functional categories. d The distribution of AMGs encoded by archaeal viral genomes in the AVGD. The numbers in brackets denote the number of archaeal viral genomes with AMGs

We further classified these genes into functional categories, and the items with the number of annotated genes in the top 50 were listed. Among them, categories related to the structure, assembly, and packaging, DNA replication and transcription, and lysis, all of which were typical functional capacities of phages, were enriched by annotated genes (Fig. 5c). Notably, most of the annotated genes belonged to DNA methyltransferase and HNH endonuclease. As a common class of epigenetic modification in phages and eukaryotic viruses, DNA methylation confers viruses an escape from the host restriction-modification systems [83, 84]. Similarly, we found that some archaeal viruses could encode restriction endonucleases that could help to degrade the host DNA for the recycling of nucleotides and facilitate the switching of transcription from host to virus [84]. Intriguingly, some archaeal viruses harbored the HNH homing endonucleases that could help to cleave specific DNA sequences from competitors [85, 86]. Glycosylation has also been extensively identified in bacteriophages and giant viruses as a class of post-translational modification that can protect viruses from the host defense system and inhibit cell superinfection [87]. Interestingly, some archaeal viruses harbored various glycosyltransferases such as acetylgalactosaminyltransferases, mannosyltransferases, and glucosyltransferases, which may be involved in catalyzing the transfer of hexose or octose to DNA, proteins, or lipids, and could assist viruses in evading host defense systems [87]. Thus, these findings revealed the extensive and distinct molecular strategies employed by archaeal viruses to interact with and manipulate their archaeal hosts.

Viruses can directly influence the metabolism of bacterial or archaeal hosts through their auxiliary metabolic genes (AMGs) [88]. In this study, AMGs within archaeal viral genomes were identified using the VIBRANT v1.2.1 [53] and DRAMv v1.4.4 [65]. A total of 233 potential AMGs were detected in 171 archaeal viral genomes. Among these viruses, 36 archaeal viral genomes contained more than two AMGs (Additional file 9: Table S8). To further investigate the auxiliary metabolic functions of archaeal viral AMGs, we analyzed the distribution of archaeal viral AMGs in the genomes of their corresponding archaeal hosts by focusing on AMGs whose numbers were ranked in the top 10 (Fig. 5d). Furthermore, comparing archaeal viral genomes from the human gut, archaeal viral genomes from the pig gut carried more diverse AMGs (Additional file 1: Fig. S3c). These results provided a comprehensive characterization of AMGs encoded by archaeal viruses and underscore their significant roles in modulating host metabolism and shaping microbially driven biogeochemical processes.

AVGD greatly expanded the diversity of Caudoviricetes and current archaeal viruses

Caudoviricetes is a class of tailed double-stranded DNA (dsDNA) phages found in various environments in the AVGD. In recent years, an increasing number of novel phages belonging to the class Caudoviricetes have been discovered, particularly those that infect bacterial hosts [44]; however; archaeal viruses of Caudoviricetes have been rarely characterized. To explore the evolutionary relationships of archaeal viral genomes belonging to Caudoviricetes across different habitat types, we first clustered Caudoviricetes archaeal viral genomes from the AVGD database and the NCBI RefSeq database at the threshold of 95% ANI and constructed a phylogenetic tree using representative Caudoviricetes archaeal viral genomes based on tandem comparisons of 77 protein-coding marker genes [39]. After removing those genomes with less than three marker genes or < 5% of aligned sequence in the multiFASTA alignment, ultimately, a total of 4695 archaeal viral genomes of Caudoviricetes (AVGD 595 and Refseq 4100) were retained for phylogenetic analysis. Based on the cumulative branch lengths, all archaeal viral genomes of Caudoviricetes from the AVGD covered 26.08% (427/1810 PD units) of the total phylogenetic diversity, while novel genomes alone contributed 8.78% (159/1810 PD units), and these genomes represented major lineages in the tree (Fig. 6a).

Fig. 6.

Phylogeny analysis of Caudoviricetes and gene-sharing network analyses among archaeal viral genomes. a The phylogenetic tree of representative Caudoviricetes genomes constructed based on 77 conserved marker proteins. The tree was visualized using iTOL. b Gene-sharing network analysis of archaeal viruses in the AVGD and previously described prokaryotic viruses. Each node represents a viral genome, and each edge denotes the existence of shared protein clusters. Viral sources are distinguished by different colors

All 3708 archaeal viral genomes in the AVGD were further compared to previously reported archaeal viruses and bacteriophages from the NCBI RefSeq database (n = 4646) using gene-sharing network analysis (Fig. 6b) [24]. All archaeal viruses rarely had connections with any bacterial reference viruses in the network. Notably, many archaeal viral clusters or singletons (n = 3640 genomes, 98.17%) had no bacterial or archaeal viruses in the NCBI RefSeq database, suggesting that they were previously uncharacterized taxa of the archaeal virosphere.

Overall, the combined evidence from the ANI-based species-level approach, phylogeny-based analysis of Caudoviricetes, and gene-sharing network analysis robustly supports the substantial novelty and diversity of archaeal viral genomes represented in AVGD. Specifically, at the species level, ANI clustering identified 1150 novel viral genomes in AVGD after excluding clusters containing sequences from the previously identified virus (IGN database), as described above. Among these novel genomes, 361 out of 1150 were included in the phylogeny-based analysis of Caudoviricetes and collectively accounted for 8.78% of the total Caudoviricetes diversity (Additional file 1: Fig. S4a). In addition, 1122 of the 1150 novel genomes formed distinct clusters or singletons in the gene-sharing network, indicating a high degree of genomic uniqueness (Additional file 1: Fig. S4b).

Discussion

Archaeal viruses, like bacteriophages, are consistent members of the microbiome across diverse habitats, including the gastrointestinal tract of humans and mammals, as well as water, sediment, and soil environments [44, 82, 89]. The cultivation of archaeal viruses remains limited, often leading to an underestimation of their diversity and ecological roles within the microbiome [14]. For instance, in human gut microbiome studies, archaeal viruses are typically less abundant than bacteriophages [16, 39]. Despite advances in metagenomic research, which have revealed a wide range of archaeal viruses across various environments [88, 90], the majority of these viruses belong to the Caudoviricetes class [6]. Additionally, accurately predicting virus-host interactions remains a significant challenge [7], further complicating the distinction between bacterial and archaeal viral hosts [6]. This gap in knowledge hinders our understanding of the diversity and evolutionary dynamics of archaeal viruses within different ecosystems. Consequently, the primary goal of our research was to investigate the diversity of archaeal viruses across diverse environments and their interactions with host organisms. In this study, we constructed an archaeal viral genome database containing 3708 archaeal viral genomes, using publicly available metagenomic sequencing datasets from pig guts and a comprehensive viral database spanning diverse habitat types. Based on the data, we systematically characterized the archaeal viral genomic characteristics, host interactions, and divergence across different habitat types. This study provides a valuable foundation for future investigations into the roles of archaeal viruses within microbial ecosystems.

Although metaviromic and metagenomic sequencing data have accelerated the discovery of phages [39, 91], identifying archaeal viruses from complex sequencing datasets remains challenging. For instance, Li et al. established a database of archaeal viruses from the human gut [16], but Cynthia et al. later re-evaluated the database and identified several false-positive viral sequences [17]. Their critical analysis revealed two main issues: (i) the reliance on viral signature genes that lacked specificity and on an insufficient set of reference genomes for host sequence subtraction, leading to the inclusion of chromosomal contaminants; and (ii) the over-interpretation of archaeal CRISPR spacer matches as a definitive marker for viral origin. In light of these methodological lessons, our study employed a more rigorous, multi-step strategy. After an initial identification based on a combination of viral features, we applied stringent filtering against comprehensive reference databases to remove potential host sequences. Furthermore, while we utilized CRISPR-Cas spacer matches for ecological inference, we explicitly did not use them as a primary or standalone criterion for viral classification. This conservative approach was designed to ensure the authenticity of the viral genomes in our final dataset. Based on our analysis, archaeal viruses accounted for only 0.038% of the total viral genomic sequences we identified. In engineered anaerobic systems, we observed a relatively higher proportion (0.269%) of archaeal viruses, which may be attributed to the presence of a greater number of archaeal hosts in this environment. Overall, however, the proportion of archaeal viruses we detected remained low. This could be due to our stringent archaeal virus discovery process. The true prevalence of archaeal viruses across different environments warrants further investigation. Nevertheless, we constructed a database containing 3708 archaeal viral genomes, which, to our knowledge, represents the most comprehensive collection of archaeal viral genomes to date. Interestingly, the majority of these viruses were classified as virulent archaeal viruses (58.54%), based on combined predictions from BACPHLIP [52] and VIBRANT [53]. While BACPHLIP was originally designed for bacteriophage lifestyle prediction, its application to archaeal viruses in this study was informed by the shared features observed between certain temperate archaeal viruses and temperate bacteriophages [39]. Consequently, this approach should be considered a preliminary inference. The absence of tools specifically trained on archaeal viruses is a recognized limitation of our study, and definitive determination of their lifestyles will ultimately require experimental validation. This resource significantly expands the known genomic diversity of archaeal viruses, as reflected at both the family and genus levels. A previous study has shown that certain phages from different rumen systems exhibit closer evolutionary relationships within the same host species [57]. Similarly, we observed that members of some archaeal viral genera exhibited closer evolutionary relationships within the same habitat, even though not all lineages exhibited strict phylogenetic monophyly, further supporting that divergence of archaeal viruses was potentially linked to their environmental context. This was consistent with reports that the type of habitat or host may influence the distribution and diversity of archaeal viruses, further emphasizing the importance of these viruses in maintaining the stability of microbial ecosystems and their coevolution with host organisms [90, 92]. Investigating the divergence patterns of archaeal viruses across different environments inevitably requires pooling viral genomes from diverse habitats. However, combining data from heterogeneous environments may introduce noise in viral detection, novelty estimation, and divergence analyses, which represents an important limitation of this study. Undoubtedly, the inclusion of more environmental samples, particularly those with deeper sequencing coverage, would likely lead to the discovery of additional novel archaeal viruses. Nonetheless, based on our novelty assessments using both nucleotide and protein-level features, we are confident that the AGVD dataset captures a substantial proportion of archaeal viral diversity across the environments surveyed. In addition, we acknowledge that heterogeneity in archaeal viral communities within specific habitats may influence our divergence analyses. Future efforts incorporating a broader collection of archaeal viral sequences will help to further elucidate these patterns. We further investigated the interactions between archaeal viruses and their hosts. Firstly, we found that a proportion of archaeal viruses present in AVGD tended to target multiple archaeal genera, and their archaeal hosts spanned multiple phyla. At the genus level, Methanobrevibacter A, Halolamina, and Methanobrevibacter were the most commonly targeted, and the majority of these viruses appeared to be lytic archaeal viruses, suggesting their potential role in regulating archaeal community dynamics. In addition, we examined the distribution of archaeal antiviral defense systems and corresponding anti-defense mechanisms encoded by archaeal viruses. Similar to bacteriophages, we identified incomplete CRISPR-Cas systems in some archaeal viral genomes, a finding consistent with a previous study [24]. Similar to bacteriophages [79], one of the archaeal viruses identified in this study carried a complete functional CRISPR-Cas system. These CRISPR-Cas systems might respond to competition between archaeal viruses for the same host [24, 79]. Additionally, these systems can encode newly identified or divergent variants of Cas effector proteins, which, like the extensively engineered Cas9, may possess features making them suitable for development as next-generation gene-editing tools. Furthermore, we constructed interaction networks between archaeal viruses and their hosts, as well as archaeal virus-archaeal virus networks, based on CRISPR spacer matching. These networks provided a more comprehensive view of the complex interactions between archaeal viruses and their archaeal hosts; for example, when two archaeal viruses target the same host, one archaeal virus can also carry spacer regions that match the other archaeal virus. This is similar to the recent findings by Brenes et al., who reported that Escherichia coli prophages encode an arsenal of defense systems to protect against temperate phages [93].

We further investigated the functional potential encoded by archaeal viral genomes in AVGD. Notably, DNA methyltransferases were one of the top three annotated functional categories (0.80%) of annotated gene functions, which are likely related to the ability of archaeal viruses to evade host restriction-modification (RM) systems. In addition, we identified a range of genes associated with immune evasion functions, including acetylgalactosaminyltransferases and mannosyltransferases. Similar to bacteriophages [89, 94, 95], archaeal viral genomes also carry a variety of auxiliary metabolic genes (AMGs). Across all examined environments, the dcm gene was consistently the most prevalent AMG in archaeal viral genomes. This gene is closely linked to the host’s cysteine and methionine metabolism. Of course, it is important to note that the dcm gene may have contributed to AMG functions in a small proportion of instances, with the majority being annotated as methyltransferases, often linked to anti-RM systems. Martin et al. addressed the limitations of current semi-automated tools for AMG annotation [96], underscoring the necessity for experimental validation of the auxiliary metabolic gene functions identified in our microbiome analysis.

The class Caudoviricetes comprises tailed double-stranded DNA (dsDNA) phages and represents one of the most diverse taxa in the virosphere [97]. Since 2015, when the GenBank database included only 16,232 genomes classified under Caudoviricetes [98], the known genomic diversity within this class has been greatly expanded. Research from diverse environments has significantly accelerated the exploration and diversification of Caudoviricetes. However, most studies have focused primarily on their potential bacterial hosts [44, 68, 82], often overlooking the fact that these viruses may instead target archaeal hosts [6]. The development of the AVGD database has greatly expanded the known diversity of Caudoviricetes that specifically infect archaea, thereby significantly enriching our understanding of the archaeal virosphere. Based on phylogenetic analysis, the newly identified archaeal viral genomes in AVGD may contribute approximately 8.78% of the overall phylogenetic diversity. Furthermore, we performed a gene-sharing network analysis to examine the relationships between archaeal viral genomes in AVGD and bacteriophages. The results revealed that archaeal viruses formed distinct and separate clusters, indicating a clear genomic segregation between archaeal and bacterial viruses within Caudoviricetes. Collectively, the convergent evidence derived from the ANI-based species-level approach, the phylogeny-based analysis of Caudoviricetes, and the gene-sharing network analysis robustly confirms the substantial novelty and broad diversity of the archaeal viral genomes represented in AVGD. Admittedly, the number of samples collected from certain habitats remains limited, and progress in the isolation and cultivation of archaeal viruses has been constrained, representing the primary limitations of this study. Nonetheless, this unique separation warrants further investigation to understand its biological significance. Overall, leveraging viral data from diverse habitats, we constructed a comprehensive archaeal viral genome database.

Abbreviations

AVGD

Archaeal Viral Genome Database

IGN

An integrated public viral database

Acrs

Anti-CRISPR proteins

HGAVD

Human Gut Archaeal Viral Database

RdRp

RNA-dependent RNA polymerase

CRISPR

Clustered regularly interspaced short palindromic repeat

ANME

Anaerobic methanotrophic

PGV

Pig gut viral genome database

NT

Nucleotide Sequence Database

ANI

Average nucleotide identity

AAI

Average amino acid identity

RM

Restriction-modification

MTases

DNA methyltransferases

AMGs

Auxiliary metabolic genes

dsDNA

Double-stranded DNA

MAGs

Metagenome-assembled genomes

BUSCO

Bacterial universal single-copy orthologs

sVCs

Species-level viral clusters

gVCs

Genus-level viral clusters

fVCs

Family-level viral clusters

Authors’ contributions

Z. C. interpreted the results, analyzed the data and revised the manuscript; C. W. performed the experiments, analyzed the data and wrote the manuscript; Y. W. analyzed the data.

Funding

This study received no funding.

Data availability

The codes for detecting archaeal virus analysis are available at https://doi.org/10.5281/zenodo.15274252. All other datasets are available in the supplementary materials.

Declarations

Ethics approval and consent to participate

All animal work was conducted according to the guidelines for the care and use of experimental animals established by the Ministry of Agriculture of China. The project was also approved by Animal Care and Use Committee (ACUC) in Jiangxi Agricultural University.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.