Various amoeba-infecting giant viruses have been isolated during the last 20 years [1,2,3,4,5,6]. They are characterized by the large size of their genomes and particles. This group of viruses has been classified as belonging to the phylum Nucleocytoviricota [7]. Nucleo-cytoplasmic virus orthologous groups (NCVOGs) and giant virus orthologous groups (GVOGs) have been widely used for core gene identification and to conduct comprehensive classification of these viruses [8, 9]. As a result, the structure of the phylum Nucleocytoviricota has recently been challenged, and an expansion from seven currently recognized families (Mimiviridae, Phycodnaviridae, Ascoviridae, Iridoviridae, Marseilleviridae, Asfarviridae, and Poxviridae) to 32 families has been proposed [8, 10, 11].

Recently, two viruses were discovered by co-culture with Acanthamoeba castellanii. The first one was isolated from a hot spring in Japan and was named "Acanthamoeba castellanii medusavirus J1" (ACMV-J1) because the host amoebae tend to form cysts upon infection with this virus, and this phenomenon is reminiscent of Medusa in Greek mythology [4]. The second isolate was a close relative of the first virus and was named "medusavirus stheno T3" (MVS-T3) because Stheno is a sister of Medusa [5]. Both viruses show substantial morphological and genomic similarities to members of the phylum Nucleocytoviricota. However, these viruses are not phylogenetically close to any member of the established families within the phylum Nucleocytoviricota. Thus, in order to officially classify these two viruses within the ICTV virus taxonomy framework, we propose to create two species, “Medusavirus medusae” (typified by ACMV-J1) and “Medusavirus sthenus” (with the exemplar isolate MVS-T3), to classify them in the new genus and family, “Medusavirus” and “Mamonoviridae”, respectively, belonging to the class Megaviricetes of the phylum Nucleocytoviricota.

Etymology of taxon nomenclature

The species and genus nomenclature was inspired by the two Gorgon sisters from Greek mythology (Medusa and Stheno), while the family name originates from the Japanese word “mamono” (魔物), meaning “monster”.

Infection cycle

ACMV-J1 has been shown to enter the host cell by endocytosis and then enter the host nucleus at approximately one hour postinfection (hpi). The virus gradually transforms the host nucleus into a viral factory without disrupting the nuclear membrane. At around 10 hpi, the cytoplasm is filled with empty viral capsids, and eventually, viral particles are released outside the cell in a non-lytic way at around 14 hpi [4, 12] (Fig. 1a).

Fig. 1
figure 1

Acanthamoeba castellanii medusavirus J1 (ACMV-J1) replication and the features of its particles [12]. (a) ACMV-J1 replication in an amoeba cell after infection. (b) A cryo-EM image of ACMV-J1. Scale, 200 nm. (c) A 3D reconstruction of an ACMV-J1 virion. Scale, 50 nm

Genomic and proteomic features

The ACMV-J1 virion has an icosahedral shape with a diameter of approximately 260 nm, including surface spikes, as revealed by cryo-electron microscopy (cryo-EM) [12] (Fig. 1b and c). It encapsidates a linear, double-stranded DNA (dsDNA) genome of 381,277 bp with a high G+C content (61.7%) [4]. A total of 461 open reading frames (ORFs) have been predicted in the genome. The ACMV-J1 genome encodes five of the seven core genes of members of the phylum Nucleocytoviricota that are frequently used in phylogeny [8]. These are the major capsid protein (MCP), superfamily II helicase (SFII), DNA polymerase family B (PolB), A32-like packaging ATPase (A32), and virus late transcription factor (VLTF3). However, the virus is unique among amoeba-infecting giant viruses in that it encodes a full set of histone proteins (i.e., linker histone H1, and core histones H2A, H2B, H3, and H4) and lacks two of the core genes, namely, RNA polymerase and DNA topoisomerase II (TopoII).

MVS-T3 was isolated in 2021 [5]. This virus forms icosahedral particles similar to those of ACMV-J1 and has a G+C-rich (62.64%), 362,811-bp-long dsDNA genome. The average nucleotide identity (ANI) between ACMV-J1 and MVS-T3 is 79.5%. MVS-T3 has the same set of core genes as ACMV-J1 and also encodes a full set of histones, but the genes for H3 and H4 are fused into a single gene.


To clarify the relationship between medusaviruses and other members of the phylum Nucleocytoviricota, we used the seven core genes from GVOGs that have been argued to be the most suitable for phylogenetic analysis of Nucleocytoviricota members (i.e., PolB, SFII, A32, VLTF3, TopoII, TFIIB, and RNAPL) [8]. The two medusaviruses formed a clade in the phylum Nucleocytoviricota with high branch support (SH-aLRT = 100%; ultrafast bootstrap = 100%) (Fig. 2), consistent with a previous study that demonstrated that ACMV-J1 does not belong to any virus group identified so far [4]. In the tree, medusaviruses are close to Feldmannia species virus, Ectocarpus siliculosus virus 1, coccolithoviruses, pandoraviruses, and molliviruses, which were previously suggested to form a proposed new taxonomic order, “Pandoravirales” [8]. However, the branch support for this clade was weak (SH-aLRT = 97.9%; ultrafast bootstrap = 58%). Therefore, we focus here on the position of the two medusaviruses and propose to create two new species in a new genus and a new family.

Fig. 2
figure 2

Maximum-likelihood phylogenetic tree of members of the phylum Nucleocytoviricota. The tree was based on a concatenated amino acid sequence alignment of seven marker genes constructed using MAFFT (v.7.471) and trimAl (v.1.4.1) and was built using IQ-TREE 2 (v.2.1.3) [15,16,17]. The model was LG+F+R8, which was selected by the built-in Modelfinder program of IQ-TREE 2 [18]. Branch support values were computed by 1000 ultrafast bootstrap replicates and SH-aLRT [19]. The tree was visualized using iTOL, and the round labels on the branches represent high-confidence support with ultrafast bootstrap ≥ 95% and SH-aLRT ≥ 80%. The positions of members of the proposed family “Mamonoviridae” are shown with a red background and are indicated by stars.

Relationship between medusaviruses and clandestinovirus

Recently, another giant virus, named clandestinovirus, was isolated by co-culture with another host, Vermamoeba vermiformis, in France [6]. Clandestinovirus has a larger genome, more genes, and a lower G+C content (581 kbp, 617 genes, 43.5% G+C) than medusaviruses. The clandestinovirus genome contains all of the core genes that medusaviruses have and additionally encodes an RNA polymerase and TopoII. Like medusaviruses, clandestinovirus also causes a nucleo-cytoplasmic infection; it enters the nucleus and turns it into a viral factory. A previous study has shown that when comparing the core genes, the closest relative of clandestinovirus is ACMV-J1 [6].

Here, we used a quantitative method to draw a family-level boundary to determine the relationship between clandestinovirus and medusaviruses. We compared these three viruses in terms of nucleotide-level similarity, including ANI and tetranucleotide similarity (TETRA); phylogenomic distance, by calculating the distance between tips on the phylogenomic tree, and the number of shared orthologous genes (OGs). We then compared these metrics to the inter- and intra-family metrics for other virus families. As a result, the degree of similarity between medusaviruses and clandestinovirus lies between the interfamily and intrafamily levels.

Phylogenomic analysis showed that clandestinovirus branched together with the medusaviruses with a high branch support (ultrafast bootstrap = 100%; SH-aLRT = 98.8%) (Fig. 2). However, the tip distances (3.92 to ACMV-J1 and 3.95 MVS-T3) lay between the mean values for intrafamily (2.46) and interfamily distances (7.30) (Fig. 3a).

Fig. 3
figure 3

Boxplots for (a) tip distance, (b) ANI, (c) TETRA, and (d) normalized OG sharing level. The horizontal black line represents the value obtained by comparing clandestinovirus and Acanthamoeba castellanii medusavirus J1 (ACMV-J1), a member of proposed species “Medusavirus medusae

In terms of genome-level nucleotide sequence similarity, ANI and TETRA were calculated using the Python package pyani [13]. The ANI between clandestinovirus and each of the two medusaviruses was 0, whereas the average intra- and interfamily values were 0.36 and 0.01, respectively. In addition, only kaumoebavirus had a non-zero ANI (0.68) when compared to clandestinovirus. The TETRA values obtained when comparing the two medusaviruses with clandestinovirus were both 0.32, which is lower than the average interfamily TETRA value (0.38). In addition, the TETRA values obtained from the comparison of clandestinovirus and the medusaviruses ranked only 134th and 139th among 220 comparisons between clandestinovirus and other viruses (Fig. 3b and c).

We then used Orthofinder v.2.5.2 to identify OGs and calculated the gene-sharing level \({S}_{ij}\) based on the number of shared OGs between viral genomes [14]. The number of shared OGs was normalized to the total number of OGs of each virus under comparison using the following formula:

$${S}_{ij} = \frac{{OG}_{ij}}{\sqrt{{OG}_{i} \times {OG}_{j}}}$$

Here, \({OG}_{ij}\) is the number of shared OGs between virus \(i\) and virus \(j\), and \({OG}_{i}\) is the total number of OGs in virus \(i\). The gene-sharing level between clandestinovirus and the medusaviruses (0.16 to ACMV-J1, ranked 30th among all comparisons between clandestinovirus and other viruses; 0.17 to MVS-T3, 25th) lay between the mean values for intra- and interfamily levels (0.47 and 0.07, respectively) (Fig. 3d).

Among the known viruses, clandestinovirus is the closest relative of medusaviruses. However, they show large divergence that places their phylogenetic relationships between the intra- and interfamily levels. We therefore do not currently recommend including the clandestinovirus into the proposed new family “Mamonoviridae”.

Finally, we propose the following simple and ready-made criteria for species, genus, and family demarcations within the family “Mamonoviridae”. If a virus shares >95% ANI, similar morphology, and comparable genome size to the members of two proposed species (e.g., “Medusavirus medusae” and “Medusavirus sthenus”) in the genus “Medusavirus”, it should be classified as a member of one of these two taxa. The average intra-genus ANI is 70% within five families of the phylum Nucleocytoviricota (i.e., Mimiviridae, Ascoviridae, Phycodnaviridae, Poxviridae, and Iridoviridae). By taking this statistic into consideration, we propose that if a virus shares >70% ANI, similar morphology, and comparable compositions of core genes with members of the proposed genus “Medusavirus”, it should be classified as a member of this genus. For a virus distantly related to the members of the proposed family “Mamonoviridae”, its inclusion in or exclusion from the family should be considered based on phylogenomic analyses like those presented here. We acknowledge that these criteria might need to be updated according to the progress of analytical methods and discoveries of new traits in viruses.


Medusaviruses are amoeba-infecting giant viruses that undergo a nucleo-cytoplasmic infection cycle and are unique among known viruses because they encode a full set of histone genes. Currently, there are two well-characterized but not yet officially classified medusaviruses (ACMV-J1 and MVS-T3). Our phylogenomic analysis revealed that this group of viruses does not branch within any groups of viruses. Thus, based on the overall characteristics of the two currently known medusaviruses, in particular genome features and phylogenomics, we here propose the creation of two species, “Medusavirus medusae” and “Medusavirus sthenus”, in a new genus, “Medusavirus”, and a new family, “Mamonoviridae”, to classify ACMV-J1 and MVS-T3, respectively. We propose that the new family be included in the class Megaviricetes of the phylum Nucleocytoviricota. This article is related to a taxonomic proposal that was recently submitted to the ICTV for consideration but had not yet been approved or ratified at the time of publication. Therefore, the taxa proposed in this paper are not part of the official ICTV taxonomy.