Medusavirus, a Novel Large DNA Virus Discovered from Hot Spring Water

Medusavirus was isolated using Acanthamoeba castellanii as the laboratory host. Cytopathic effects, such as cell rounding of host cells due to viral infection, were observed 1–2 days postinfection (PI) (Fig. 1a and b). Eventually, the viral infection induced amoebae to undergo morphological changes similar to encystment in a subpopulation of amoeba cells as early as 2 days PI (Fig. 1c). On the other hand, other amoeba cells without encystment were lysed. This encystment-like phenomenon prompted us to name this virus after Medusa, the monster in ancient Greek mythology who turns onlookers to stone.

Single-particle cryo-electron microscopy (cryo-EM) analysis revealed that the medusavirus virion is an icosahedron with T=277 (h=7; k=12) and is approximately 260 nm in diameter (Fig. 2a and d). The single-layered major capsid (approximately 8 nm in thickness) of medusavirus was covered with a number of spherical-headed spikes of approximately 14 nm in length, each of which extended from each capsomer (Fig. 2). The virus particles isolated in the laboratory were either filled with DNA or lacked DNA inside. The resolutions of the cryo-EM maps were estimated at approximately 31.7 Å for the DNA-filled particle and approximately 31.3 Å for the empty particle. The spikes appeared to be rather flexible because they seemed to be blurred in the DNA-filled matured particle or the higher-resolution map with imposed icosahedral symmetry. The viral capsid was backed with a 6-nm thick internal membrane, as commonly found in NCLDVs (Fig. 2a, b, c and e). The membrane extended and directly interacted with the inner surface of the capsid under the 5-fold axis (red arrows in Fig. 2b).

Fluorescent in situ hybridization (FISH) analysis revealed that medusavirus DNA was localized in the nucleus of amoeba cells by 1 h PI. Signals of medusavirus DNA were strong at the periphery of the nucleolus at around 2 to 4 h PI (Fig. 3), suggesting that medusavirus DNA replication is initiated in the host cell nucleus. The amount of viral DNA increased significantly at 8 h PI and spread throughout the nucleus (Fig. 3). The cytoplasm of the host cells was filled with numerous capsids without DNA at 8 to 10 h PI, but the host cells maintained the integrity of the nuclear membrane even after viral infection (Fig. 4). At 14 h PI, signals of putatively newly synthesized medusavirus DNA were observed in the cytoplasm of the host cells and increased until 48 h PI (yellow arrows in Fig. 3). New virions were released from the host cells at 14 h PI, and a number of replicated virions were detected in the culture media around 29 h PI.

The genome of medusavirus was a linear 381,277-bp dsDNA. Its G+C content (61.7%) was the greatest among NCLDVs after pandoraviruses (62.0% on average) (Table 1) and slightly greater than that of the host amoeba genome (58.4%). We identified a total of 461 predicted protein-coding genes (open reading frames [ORFs]) and three tRNA-like sequences, corresponding to a theoretical coding density of 89.5% (Table S1). Among the predicted protein-coding genes, seven harbored putative spliceosomal introns (Table S1). Of all the ORFs, 182 (39%) showed homologs in the public sequence databases, including 115 closest homologs in eukaryotes, 45 in viruses, 18 in prokaryotes, and four in unclassified organisms, whereas 279 (61%) were ORFans (Fig. 5a). Notably, 86 predicted proteins had their closest homologs in A. castellanii strain Neff (Fig. 5a), suggesting that massive gene exchanges occurred between the ancestors of medusavirus and Acanthamoeba. There were 137 predicted proteins with significant sequence similarities to viral proteins in the databases, of which 117 proteins showed the closest viral homologs in other large DNA viruses, including 36 proteins in pandoraviruses, 23 proteins in Phycodnaviridae, and 16 proteins in mollivirus (Fig. 5b).

We were able to assign putative functions to 105 (23%) ORFs. Predicted proteins encoded in the genome of medusavirus included a variety of enzymes for DNA/nucleotide metabolisms, such as a family B DNA polymerase (PolB), a DNA primase, three sliding clump proteins, an RNase HII, a Holliday junction resolvase, ribonucleotide reductase large and small subunits, a nucleoside diphosphate kinase, a thymidylate synthase, a deoxycytidylate deaminase, and a dUTPase (Table S1). In addition, medusavirus was predicted to encode several transcription-related proteins, such as a transcription elongation factor S-II, viral late transcription factors 2 and 3, a poly(A) polymerase regulatory subunit, and four homologs of Rho transcription termination factor (Table S1). Furthermore, we identified genes for the translation initiation factor eIF1 and a tRNA^His guanylyltransferase. Genes for the MCP and three variola virus (VV) A32-like DNA-packaging ATPases were readily identified. However, medusavirus had no genes for a DNA-dependent RNA polymerase, mRNA capping enzyme, or DNA topoisomerase II. All known NCLDVs encode one or more of these enzymes in their genomes, but even fragments or remnants of these genes were not detected in the medusavirus genome. The lack of these enzyme genes in medusavirus may be consistent with its putative dependence on the host nucleus for DNA replication. The genome of medusavirus was found to encode a homolog of GTP-binding Ras-related nuclear protein (Ran), which plays an important role in the directionality of nuclear import and export (55) (Fig. 6).

Medusavirus encoded homologs of all four core histones (i.e., H2A, H2B, H3, and H4) and the linker histone H1 (Fig. 5 and Table 2). In general, the core histone proteins are enriched with basic amino acid residues to facilitate interaction with the negatively charged DNA. Medusavirus core histone homologs were also enriched with basic amino acid residues (Fig. 7). Phylogenetic analysis of these core histone homologs revealed that the branches for medusavirus and other viral homologs were placed at the root of the respective core histone clades (Fig. 8).

The genome of medusavirus harbored other predicted proteins that were atypical for viral proteins. These include a cyclin B homolog, which may regulate the G2-M phase transition of the host amoeba (56, 57); a putative metacaspase; and a homolog of the mitochondrial chaperone BCS1. Putative viral metacaspases have previously been identified in an environmental giant virus single-amplified genome (gvSAG-566-O17) and marine metagenomes (58). As virus infection induces the host’s programmed cell death and the activation of host metacaspase in Emiliania huxyleyi (58–60), the virally encoded medusavirus metacaspase may serve to enhance the efficiency of infection by regulating programmed cell death and/or host stress responses. Our phylogenetic analysis of metacaspases revealed that the medusavirus metacaspase gene forms a monophyletic clade with the environmental sequences from ocean samples, including that of gvSAG-566-O17 (Fig. 9). We also found that a hypothetical protein (GenBank accession no. YP_009507480.1) of Heterosigma akashiwo virus 01 (HaV01) belongs to the same clade. To our knowledge, these medusavirus and HaV01 cases represent the first identification of putative metacaspase homologs encoded in cultured viruses.

Proteomic analysis of medusavirus virions revealed 80 virion proteins (Table 3). Among them, 54 (68%), including 20 ORFans, had unknown functions. Identified proteins included the MCP, two DNA-packaging ATPases, four Rho transcription termination factor homologs, multifunctional redox-active proteins such as glutaredoxin and thioredoxin homologs, and all virally encoded core histones, viz., H2A, H2B, H3, and H4. Although our proteomic analysis was not truly quantitative, the semiquantitative exponentially modified protein abundance index (emPAI) (61) values were relatively high for the core histone proteins (Table 3). Compared with the emPAI value of the MCP showing the highest emPAI value, the emPAI values were 28%, 21%, 9%, and 2.7% for H3, H2B, H4, and H2A, respectively. This suggests that these histones may be sufficiently abundant to package the viral DNA within the capsid.

The genome of medusavirus encoded 18 genes that were classified into 15 of the previously defined 47 NCLDV core genes (25) (Table 4). The number of NCLDV core genes in medusavirus is thus comparable to those found in other NCLDVs with a relatively small genome (e.g., Feldmannia species virus [155 kb, 17 core genes]; Rock bream iridovirus [112 kb, 16 core genes]) (Fig. 10). Phylogenetic analyses of DNA polymerases and MCPs did not support the inclusion of medusavirus in any of the existing groups or families of DNA viruses (Fig. 11). The gene content-based cladistics tree and proteomic tree also indicated that medusavirus represents an independent lineage among known DNA viruses, by branching from the root of the clade comprised of mollivirus and pandoraviruses (Fig. 12).

The reconstructed tree placed medusavirus PolB at the root of, but not inside, the eukaryotic DNA polymerase delta (Pol δ) clade (Fig. 13). Other viral PolB sequences were clearly separated from medusavirus PolB in the reconstructed phylogenetic tree (Fig. 13).

By reciprocal BLAST searches, we identified 57 LGT candidate genes between medusavirus and A. castellanii. 13 of the 57 genes were predicted to be transferred from virus to amoeba (VtoA) and 12 were in the reverse direction (AtoV) (Table 5). The directions for the remaining 32 cases could not be determined (Table 5). AtoV genes included a linker histone H1 gene, while VtoA genes included several viral hallmark genes, such as an MCP gene and a DNA-packaging ATPase gene. LGT candidates with undetermined directions were enriched with hypothetical proteins. We analyzed the transcriptional activities expressed as RPKM (number of reads per kilobase per million reads) of the LGT candidate genes in A. castellanii. The LGT candidate genes showed lower transcriptional activities than other genes conserved among the species of Amoebozoa (i.e., “vertically inherited genes”) (Mann-Whitney U test, P = 2.00 × 10⁻⁶⁰) (Fig. 14). Transcriptional activities of both VtoA genes and LGT candidate genes with undetermined directions were even lower than those of AtoV genes (P = 0.026), suggesting that some of these genes are no longer functional or are silenced in the amoeba genome. This is consistent with the previous observation that approximately half of A. castellanii genes putatively acquired from large DNA viruses are transcriptionally inactive (62).

Recent studies have revealed the presence of NCLDV MCP genes in the genomes of Acanthamoeba spp. (63, 64), suggesting gene transfers from viruses to Acanthamoeba. Phylogenetic analyses indicated that some of these sequences formed unidentified NCLDV clades (62, 64, 65). Therefore, we performed phylogenetic reconstructions of MCPs and DNA-packaging ATPases, including homologs from the genomes of medusavirus and Acanthamoeba spp. The results indicated that medusavirus protein sequences form monophyletic groups with the previously identified homologs in Acanthamoeba (Fig. 15).

Acanthamoeba castellanii (Douglas) strain Neff (ATCC 30010) cells were purchased from the American Type Culture Collection (ATCC, Manassas, VA) and cultured in peptone-yeast-glucose (PYG) medium at 26°C as described previously (14, 81). An outflow water/soil sample (50 ml) was collected from a water sample spilled out from a hot spring in Japan. The water temperature was 43.4°C at the sampling site. After removal of floating bacteria and small viruses by filtration using a 20-μm filter (no. 43; Whatman International, Maidstone, UK), the collected mud and dead leaves were resuspended in 13 ml of sterile phosphate-buffered saline (PBS) and stirred gently for 1 day at room temperature. The sample was again filtered through another 20-µm filter. Then, the filtered sample (10 ml) was further filtered through a sterile 1.2-µm filter (Millex-AA; Merck Millipore, Darmstadt, Germany). The filtrate (9.5 ml) was then mixed with PYG medium (18 ml). Acanthamoeba cell suspension (0.5 ml) was added and incubated with gentle stirring for 1 h at room temperature, followed by incubation at 26°C in a total of 142 wells using two 96-well microplates. After 5 days, amoeba cells with delayed proliferation were screened. Culture supernatant from growth-retarded wells showing phenotypical difference was inoculated into fresh amoeba cells in an individual well of a 12-well microplate. After 3 days, supernatant of all three wells with cell encystment was inoculated into a fresh amoeba cell suspension in three 25-cm² culture flasks and then in three 75-cm² culture flasks. Supernatant from each 75-cm² culture flask was stored at 4°C as an isolated virus solution (named HS-1, HS-2, and HS-3).

Among the three isolated virus solutions, virus cloning of HS-1 was performed according to a cloning method used for Mollivirus sibericum (32) with several modifications as described below. Briefly, HS-1-infected amoeba cells in 75-cm² culture flask were harvested and washed with an excess amount of fresh PYG medium to remove surplus viruses. Amoeba cells were then resuspended in 16 ml of fresh PYG medium. Eight serial 3-fold dilutions were performed in a 96-well microplate by mixing 50 μl of the solution from the previous well with 100 μl of fresh PYG. Each last eighth dilution was examined under a light microscope to verify the existence of fewer than two amoeba cells per well. Only one amoeba cell was observed in each well. Several hundred fresh amoeba cells were added to the wells containing only one cell and cultured for 3 days until most cells exhibited encystment. The obtained viral clone was designated “Acanthamoeba castellanii medusavirus (Medusavirus),” and amplified and stored at 4°C for further use.

To routinely culture medusavirus, amoeba cells were initially cultured using eight 25-cm² culture flasks, each containing 25 ml of PYG medium. The cells were inoculated with medusavirus (multiplicity of infection [MOI], ∼1 to 2), and then the culture media containing medusaviruses were harvested 1–4 days postinfection (PI). Amoeba cells and cell debris were removed by centrifugation (800 × g, 5 min, 24°C), and the medusavirus particles were collected by centrifugation (8,000 × g, 35 min, 4°C). The collected medusavirus particles were resuspended in 5 ml of PBS, and filtered through a 0.45-μm filter (Millex-AA, Merck Millipore, Darmstadt, Germany), centrifuged (8,000 × g, 35 min, 4°C), and resuspended in 10 μl of PBS. This purification protocol was performed 5 to 10 times to obtain high numbers of medusavirus particles.

A suspension of 2.5 ml of purified medusavirus particles was applied to an R1.2/1.3 Mo grid (Quantifoil Micro Tools GmbH, Germany), which was previously glow-discharged, and snap-frozen in liquid ethane using a Vitrobot Mark IV unit (FEI Company, USA) at a condition of 95% humidity at 4°C. Frozen grids were imaged using a JEM-2200FS electron microscope operated at 200 kV accelerating voltage and equipped with an omega-type energy filter and field emission electron source (JEOL Ltd., Japan). The images were recorded on a DE20 direct detector (Direct Electron LP, USA) at a nominal magnification of ×25,000 for a 3-s exposure time with 75 movie frames. The total electron dose was below 20 electrons (e⁻)/Ų for each image. The numerical pixel size corresponds to 2.3 Å on the specimen. The movie frames were motion corrected using a manufacture-provided script, DE_process_frames.py, and summed. The resulting images were subjected to single-particle analysis.

For single-particle analysis, a total of 5,406 medusavirus particles were selected from 1,198 motion-corrected images and binned by two using RELION software (82). The extracted particles were classified by reference-free alignment, where the class averages were simultaneously separated into DNA-filled, partially DNA-filled, filled with non-DNA, and empty particles classes. For structural analysis of the viral capsid, a total of 2,288 DNA-filled particles and a total of 1,397 empty particles were selected from well-aligned two-dimensional (2D) classes, respectively, and used for three-dimensional (3D) reconstruction by imposing the icosahedral symmetry. The handedness of the 3D map was determined by independent subtomogram averaging. The final map resolutions were estimated using the gold-standard Fourier shell correlation (GS-FSC) criterion of 0.143. The cryo-EM maps were visualized and annotated by UCSF Chimera (83). The icosahedral T-number was determined by manually counting the surface spike-like short fibers that extended from each capsomer.

Harvested cells infected by medusaviruses (8 h PI) or purified medusavirus particles were subjected to regular transmission electron microscopic observation as described previously (81). Plastic-embedded virus-infected amoeba cells were sectioned at 70-nm thickness using an ultramicrotome (EM-UC7; Leica Microsystems, Austria). The thin sections were mounted on a Formvar-coated slot mesh and stained with 2% uranyl acetate and 1% lead citrate for 5 min each. Transmission electron microscopy observation was done using a JEM1010 microscope (JEOL Ltd., Japan) at 80 kV accelerating voltage. The images were recorded in a 2k × 2k side-mount Veleta charge-coupled device (CCD) camera (Olympus, Japan).

For tracing medusavirus DNA in host cells after infection, FISH analysis was performed as described below. Briefly, purified medusavirus DNA (4.68 μg) was labeled with Cy3 using the nick translation method. Amoeba cells cultured in a 12-well plate were infected with medusaviruses and harvested at 10 min, 30 min, 1 h, 2 h, 4 h, 8 h, 14 h, 24 h, and 48 h PI from each individual well of the 12-well plate. Cells were washed twice with PBS and fixed with methanol:acetate (3:1) solution. One drop of the fixed cell suspension was placed on a glass slide and air dried completely. Cy3-labeled medusavirus DNA probe was placed on the glass slide and incubated at 67°C for 5 min, followed by hybridization at 37°C for 2 h, and stringent washing with 50% formamide in 2× and 1× SSC buffer (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate). Cells on the glass slide were also stained with 4′,6-diamidino-2-phenylindole (DAPI). The detection of FISH and DAPI signals were performed using the Leica CW-4000 cytogenetic workstation (Leica Microsystems K.K., Tokyo, Japan).

After virus cloning and purification, the genomic DNA of medusavirus (1.2 μg) was prepared using NucleoSpin tissue XS (Macherey-Nagel, Germany), following the manufacturer’s protocol, and further purified using AMPure XP (Beckman Coulter). The DNA library for sequencing was prepared using a g-Tube (Covaris) and an SMRTBell template prep kit 1.0 (Pacific Biosciences), and sequencing was performed on a PacBio RS II sequencer (Pacific Biosciences). The total number of subreads was 304,607, and the total number of sequenced nucleotides was 1,325,027,506. Canu v1.5 (84) was used to assemble the reads to generate a final single contig of 381,277 bp.

Gene prediction was performed using GeneMarkS v4.32 (85). Putative introns were predicted using GeneWise v2.4.1 (86) with visual inspection of the alignments. Amino acid sequence similarity searches (E value < 1E−5) were performed against the UniRef90 database, Virus-Host Database (Virus-Host DB) (87), RefSeq database, and the Conserved Domains Database (CDD) using BLASTP and RPS-BLAST of BLAST+ (88) (v2.6.0). tRNA genes were identified using tRNAscan-SE (89) (v1.3.1). NCLDV core genes were assigned to viral genomes through homology searches against the Nucleo-Cytoplasmic Virus Orthologous Groups (NCVOG) database (90).

The hidden Markov model (HMM) profiles for DNA polymerases, MCPs, and DNA-packaging ATPases were constructed using sequences in NCVOG (90). Homologs of each protein were identified using HMMsearch (91) against the Virus-Host DB (87). Sequences were aligned using Multiple Alignment using Fast Fourier Transform (MAFFT) v7.220 (92) with default parameters. Tree reconstruction was performed using RaxML v8.2.4 (93) with the selected LG+F model and PROTGAMMA parameter with 100 bootstrap replicates. The HMM profiles of histone families were constructed using sequences in the Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthology (KO) database (94) (H2A, KO no. K11251; H2B, no. K11252; H3, no. K11253; and H4, no. K11254). The homologs of each protein were recruited using HMMsearch (91) against the Virus-Host DB (87). Eukaryotic and archaeal histone and eukaryotic DNA polymerase sequences were manually collected. Tree reconstruction was performed using PhyloBayesMPI (95) with four chains for at least 4,000 cycles. The cladistic tree was computed using the neighbor-joining method based on the presence/absence matrix of gene clusters derived from OrthoFinder (96) clustering with a previously proposed similarity score (97). Branch support values were estimated using 100 times of bootstrap resampling. The proteomic tree was computed using ViPTreeGen (98) (v1.1.0).

To identify LGT candidates between A. castellanii and medusavirus, bidirectional BLASTP searches were performed by including sequences from UniRef90 but excluding the query genome. UniRef90 contains the proteome sequences of A. castellanii strain Neff but does not contain most of the protein sequences from draft genome sequences of other A. castellanii strains. When a gene of A. castellanii got a best hit for a gene of medusavirus and the same medusavirus gene got a best hit for the same A. castellanii gene, the pair of genes were considered a candidate for LGT. For inference of the directions of the LGT candidates, the most similar homologs of the bidirectional BLASTP searches were examined after excluding the hits against A. castellanii or medusavirus genes. In the BLASTP result with a query of a medusavirus gene, the best-hit gene after excluding hits to A. castellanii genes was determined and considered to be the closest gene. In the same way, in the BLASTP result with a query of an A. castellanii gene, the best-hit gene after excluding hits to medusavirus genes was considered to be another closest gene. In this way, we defined the two closest genes for a pair of LGT candidates. If at least one of the closest genes was a viral gene, it was inferred that LGT occurred from virus to amoeba (VtoA). Conversely, if at least one of the closest genes was a eukaryote gene, it was inferred that LGT occurred from amoeba to virus (AtoV). In other cases, we did not determine the direction of LGT. The transcriptional activity of the candidate LGT genes was determined using the transcriptome sequencing (RNA-seq) data sets of A. castellanii in the GenBank Sequence Read Archive (SRA), namely accession no. SRR611709, SRR611787, SRR611788, SRR611790, SRR611791, SRR611792, SRR611793, SRR611795, SRR611796, SRR611797, SRR629488, SRR957287, SRR957291, and SRR957297. For selected genes (i.e., MCP and DNA-packaging ATPase sequences), we confirmed their LGT directions with the use of phylogenetic tree reconstruction.

Following virus collection, medusavirus was further purified using 10% to 60% sucrose density gradient centrifugation (2,300 × g, 86 min, 4°C). A white-colored virus fraction with a sucrose gradient of approximately 10% to 20% was resuspended in PBS and washed twice with PBS with subsequent centrifugation (8,000 × g, 35 min, 4°C). Medusavirus particles were resuspended in PBS containing 0.5% SDS and protease inhibitor cocktail (product no. 25955-11; Nacalai Tesque), and incubated for 1 h at 65°C. Samples were subjected to trichloroacetic acid (TCA) precipitation followed by resuspension in 250 mM Tris-HCl (pH 8.5) containing 2 mM EDTA, and protein was quantified by the bicinchoninic acid (BCA) method. Proteins were reduced for 2 h at 37°C with 0.67 M dithiothreitol (DTT) in 250 mM Tris-HCl (pH 8.5) containing 2 mM EDTA, subsequently alkylated with 1.4 M iodoacetamide in 250 mM Tris-HCl (pH 8.5) containing 2 mM EDTA for 30 min at room temperature, and treated with trypsin for 20 h at 37°C. After desalination and concentration, the treated proteins were subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis using the East-nLC 1200 system (Thermo Fisher Scientific Inc., USA) and a Q Exactive Plus spectrometer (Thermo Fisher Scientific Inc., USA). All spectra data were then subjected to NCBI homology search using the Mascot server (http://www.matrixscience.com/server.html).

The electron microscopy (EM) density maps of the DNA-filled and empty medusavirus particles reported in this paper have been deposited in the EMDatabank (http://emdatabank.org) under accession no. EMD-9619 and EMD-9620, respectively. The genome sequence of medusavirus has been submitted to DDBJ under accession no. AP018495. The mass spectrometry proteomics data have been deposited to the Japan ProteOme STandard Repository (jPOSTrepo) under jPOST identifier JPST000467 (https://repository.jpostdb.org/entry/JPST000467) and PXID identifier PXD010830 (http://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD010830).