The Methylomirabilis enrichment culture (~80% Methylomirabilis sp.) was grown in a continuous sequencing batch reactor (Applicon Biotechnology BV, Applisens, Schiedam, the Netherlands) made of stainless steel and glass with a volume of 6 L. The inoculum biomass was derived from a pre-existing enrichment, originally inoculated with sediment samples from a ditch in the Ooijpolder (Ettwig et al., 2009). The culture was kept anoxic by a continuous supply of a gas mixture composed of methane and carbon dioxide (95:5, v/v) and the medium was continuously flushed with a mixture of argon and carbon dioxide (95:5, v/v). The pH was maintained at 7.3 ± 0.1, the temperature was kept stable at 30°C and the system was constantly stirred at 100 RPM. The volume of the enrichment was kept at 4 L by a level sensor controlled pump in sequential cycles of feeding (22.5 h) and rest (30 min to settle, 60 min to pump out excess medium). The HRT (hydraulic retention time) was 4 days and the composition of the medium was (per L) 0.552–2.07 g NaNO₂ (8–30 mM depending on culture performance), 288 μg MgSO₄, 192 μg CaCl₂, 50 μg KH₂PO₄, 2.5 mg FeSO₄*7H₂O. The trace element solution was adapted from Ettwig et al. (2010) and contained per L: 150 μg ZnSO_4·7H₂O, 60 μg CoCl₂6H₂O, 400 μg CuSO_4·, 100μg NiCl₂6H₂O, 10 μg H₃BO₃, 100 μg MnCl₂4H₂O, 10 μg Na₂WO_4·2H₂O, 50 μg Na₂MoO_4·2H₂O, 15 μg SeO₂, 10 μg CeCl₂.

A 20 ml sample was collected from the bioreactor enrichment culture and filtered with 0.45 and 0.2 μm syringe filters, sequentially (Puradisc Cellulose Acetate, Whatman) to remove microorganisms and keep only the viral fraction. The flow-through was concentrated using spin columns (Vivaspin, GE Healthcare, 10,000 MWCO) at 5000 × g for 5 min at 4°C, (Allegra X-15R Centrifuge, Beckman Coulter) to a final volume of about 60 μl. Approximately 3 μl sample was placed on glow-discharged and carbon-Formvar-coated 100 mesh hexagonal copper grids (Veco) and incubated at room temperature for 15 min. The excess liquid was blotted off using filter paper (Ashless, Grade 589/1 Filtration Paper, Whatman). The specimens were stained with 0.5% uranyl acetate for 1 min, blotted dry, washed in Milli-Q, blotted dry with filter paper and let air dry completely. The virioplankton was observed using a JEOL JEM-2100 transmission electron microscope operated at 200 kV.

A 2 ml sample was collected from the bioreactor enrichment culture and centrifuged at 800 × g for 4 min at 30°C (Microcentrifuge 5417R, Eppendorf, Hamburg, Germany). Supernatant was decanted and the pellet was resuspended in remaining supernatant. For high-pressure freezing, a specimen sandwich was assembled in the specimen holder of a Leica EMHPF. The sandwich consisted of an aluminum spacer ring (0.2 × 3 mm) and a membrane carrier (2.6 × 0.1 mm, no. 16707898) in which 0.2 μl sample was loaded, the sample was sealed off with a gold stub (specimen carrier D3/D2, gold, no. 16770131). After closing the holder, the sample was high-pressure frozen in an EMHPF, operating at 2100 bar. Samples were stored in liquid nitrogen.

To obtain a freeze-fracture replica, the sample was placed in a detachable cold table and loaded onto the stage of a Balzers BAF400 freeze-etch machine, pre-cooled to −170°C. Specimens were fractured at −110°C and subsequently etched for 12 min at the same temperature with a vacuum below 10⁻⁷ Bar. The specimen was shadowed with 1.3 nm Pt (angle 45°) and 15 nm C (angle 90°) and a replica of the sample was obtained. The biological material underneath the replica was digested in 70% H₂SO₄ for 48 h. Replicas were washed twice in MilliQ water and picked up with glow discharged 700 mesh hexagonal copper grids (reference number G276OG, Agar Scientific) and investigated by a JEOL JEM-2100 transmission electron microscope operated at 200 kV.

Cells from the bioreactor enrichment culture were harvested and cryofixed by high-pressure freezing (Leica HPM 100; Leica Microsystems, Vienna, Austria). Samples were placed into a 100 μm cavity of a type A platelet (3 mm diameter; 0.1–0.2 mm depth, Leica Microsystems) and closed with the flat side of a lecithin-coated type B platelet (3 mm diameter, 0.3 mm depth). The platelets were stored in liquid nitrogen.

For epon embedding, frozen samples were freeze-substituted in acetone containing 2% OsO₄, 0.2% uranyl acetate, and 1% H₂O (Walther and Ziegler, 2002). The substitution followed several intervals: cells were kept at −90°C for 47 h; brought to −60°C at 2°C per hour and kept at −60°C for 8 h; brought to −30°C at 2°C per hour and kept at −30°C for 8 h in a freeze-substitution unit (AFS; Leica Microsystems, Vienna, Austria). To remove uranyl acetate, the samples were washed four times for 30 min in the AFS device at −30°C with acetone containing 2% OsO₄and 1% H₂O. Next, fixation was continued on ice for 1 h. OsO₄ and H₂O were removed by washing the sample twice in anhydrous acetone for 30 min. Samples were gradually infiltrated with epon resin and polymerized for 72 h at 60°C (Mollenhauer, 1964). Ultrathin sections of 60 nm were cut using a Leica UCT ultramicrotome (Leica Microsystems, Vienna, Austria) and collected on carbon-Formvar-coated 100 mesh hexagonal copper grids (Veco). The sections were then post-stained with 2% uranyl acetate for 20 min and lead citrate for 2 min. After each of the two steps, grids were washed in MilliQ water. Sections were investigated using a JEOL JEM1010 transmission electron microscope operated at 60 kV.

Semi-thin sections (200–300 μm) were cut using a diamond knife (Diatome, Biel, Switzerland) and an ultramicrotome (UCT, Leica microsystems) and collected on 50 mesh copper grids with a carbon coated formvar support film. After air drying, grids were post-stained with 4% uranyl acetate in MilliQ for 30 min and Reynolds lead citrate for 2 min. Ten nanometers proteinA gold (CMC, UMC Utrecht, The Netherlands) was applied to the sections to act as fiducial marker during tilt-series acquisition and reconstruction.

Virus infected cells and isolated virus particles were chosen as the region of interest. Dual axis tilt series (−60 to +60) were recorded on the JEOL JEM-2100 microscope operating at 200 kV, using SerialEM for automated image acquisition (Mastronarde, 2005).

Recorded tilt series were reconstructed using the IMOD package (Kremer et al., 1996) and tomograms were generated using both the weighted back-projection and SIRT algorithms. Reconstructed tomograms were segmented by hand using 3DMOD. Iso-surface model was generated from a sub-tomogram employing a circular mask. Summed virtual slices (36 slices) were visualized using the UCSF Chimera package (Pettersen et al., 2004), the histogram was adjusted to fit the density associated with the virus particle.

Over a period of ~3 months bioreactor material and effluent (13 L) were collected from the bioreactor enrichment culture and stored at 4°C until further analyses. Since the majority of the microbial population in the bioreactor grows in aggregates, the total sample was split in two: the bioreactor supernatant containing free bacteriophages and the bacterial biomass containing bacteriophages in the aggregates and within the cells. The two samples were obtained by centrifuging the sample at 20,000 × g for 1 h in 350 ml centrifuge tubes (Thermo Scientific, SorvallTM LYNX 4000 centrifuge). The resulting biomass pellets were resuspended in a total volume of 50 mL of the supernatant. Free-floating bacteriophages were precipitated by iron chloride flocculation, whereas bacterial biomass-associated bacteriophages were collected by PEG 8000.

The two viral samples obtained by iron chloride and PEG 8000 precipitation were pooled together and bacteriophages were further concentrated by ultracentrifugation (Optima XE90, Beckman-Coulter, rotor type 90 Ti, Beckman-Coulter) at 77,000 rcf for 1 h at 4°C (Szpara et al., 2011). A P1 reference bacteriophage NC_005856.1 was used as a positive control for DNA extraction. The pellet was resuspended in 1 ml of supernatant and the total DNA was extracted according to the protocol published by Thurber et al. (2009). Using the Qubit dsDNA HS assay kit (Life), the extracted DNA was quantified at 0.2 ng DNA.

To prevent amplification biases, yet obtain enough DNA for Ion Torrent sequencing, 43.2 ng of 16S ribosomal DNA of “Candidatus K. stuttgartiensis” was added to 0.1 ng of viral DNA (Cremers et al., in preparation). The total DNA was sheared for 6 cycles using the Bioruptor (Diagenode, Liege, Belgium; 1 min on, 1 min off) and prepared according to manufactures protocol (IonXpress Plus gDNA fragment library preparation Rev C.0, Life). The sample was sequenced using an Ion Torrent Personal Genome Machine (Life) on a v318 chip following manufacturer's protocol.

Computational phage-host prediction signals (Edwards et al., 2015) were calculated to identify which of the 2094 assembled contigs belonged to the bacteriophage infecting Methylomirabilis. First, genetic homology between the Methylomirabilis genomes (Methylomirabilis oxyfera; Ettwig et al., 2010 and a new Methylomirabilis species; Guerrero et al., in preparation) and the viral contigs was determined by using blastn and tblastx 2.2.30+ with default settings and E ≤ 10⁻⁵ (Camacho et al., 2009), and the total bitscore for each viral contig was recorded. Second, both Methylomirabilis genomes were checked for CRISPR sequences using CRISPRfinder (http://crispr.u-psud.fr/Server/) and CRASS (Skennerton et al., 2013). The resulting spacer sequences were subsequently cross-referenced with the viral genomes and contigs from the SPAdes assembly using CRISPRtarget (Biswas et al., 2013). Third, similarity in oligonucleotide usage was calculated as 1—E, where E is the Euclidean distance between the k-mer profiles of each viral contig and the Methylomirabilis genomes (k = 2, 4, 6).

We investigated the bacteriophage population in a bioreactor containing an enrichment culture of the anaerobic methanotroph Methylomirabilis sp. (Table 1). Several free bacteriophages with different morphologies were observed using negative staining. Among these, four morphotypes were identified. All bacteriophages had putative icosahedral capsid symmetry and three of them possessed a tail. Capsids and tails varied in diameter and length, thereby allowing a clear distinction between the morphotypes (Figure 1).

In addition to free bacteriophages, the microbial population was investigated for phage infection using thin sections, freeze-etching, and electron tomography. Thin sections of high-pressure frozen, freeze-substituted, and epon embedded biomass samples showed a lytic phage infection in Methylomirabilis cells (Figure 2). The infection rate (amount of infected Methylomirabilis cells in the total Methylomirabilis population) was calculated to be 2.3%. The infection did not affect bioreactor performance with respect to activity (methane and nitrite consumption). The intracellular bacteriophages had a hexagonal capsid (ca. 55 nm wide, vertex-to-vertex), indicating an icosahedral 3D symmetry. The capsid contained an electron dense and round central core (ca. 20 nm diameter), probably enclosing the genetic material. The bacteriophages were present both outside and inside the Methylomirabilis cells. Infected Methylomirabilis cells displayed different stages of bacteriophage assembly. In addition to completely assembled bacteriophages, entities were observed inside the cells that contained only the central core, i.e., the capsid was not yet assembled. As more and more bacteriophages were assembled, they were organized in a highly packed formation within the cell (Figures 2C,D). This caused the infected cell to swell approximately 1.9x in size compared to a not infected cell [area measured on TEM sections based on 10 infected (0.44 ± 0.2 μm²) and 10 not infected cells (0.23 ± 0.06 μm²)]. Afterwards the infected cell lysed thereby releasing the bacteriophages (Figures 2E,F). Comparing the intracellular Methylomirabilis bacteriophage to the free bacteriophages observed in the bioreactor enrichment culture using negative staining, its morphology resembled that of the bacteriophage depicted in Figure 1A based on size and symmetry. Freeze-etching of infected Methylomirabilis cells revealed the surface of the bacteriophage capsid (Figure 3). The capsomeres, the building subunits of the capsid, were arranged in triangular faces that constitute the icosahedral symmetry of the proteinaceous capsid.

Electron tomography was used to study the 3D structure of the bacteriophage and the infected Methylomirabilis cells (Figure 4). 3D reconstruction and modeling showed that the cytoplasmic membrane of infected cells was broken at many places while the cell wall was often still intact (Figures 4A,B and Supplementary Movie 1). Inside infected cells, both completely assembled bacteriophages were present as well as incompletely assembled bacteriophages (with only the electron dense core visible—no capsid yet). 3D reconstruction and modeling of free bacteriophages suggested the presence of a putative internal membrane surrounding the electron dense core (Figures 4C–F). In addition, no tail was observed on the free bacteriophages.

To characterize the bacteriophage population present in the Methylomirabilis bioreactor enrichment culture, the total viral DNA was extracted from the bioreactor and the effluent and sequenced. After assembly, 2094 contigs were obtained (Supplementary Files 2A,B). The five longest and putative complete viral genomes (Table 2) were selected for manual annotation (Figure 5 and Supplementary Files 5A–E). Four of these contigs (197, 85, 71, 16 kb) contained identical sequences at both ends, indicating a full genome and a circular genomic arrangement. The 41 kb contig did not have overlapping end sequences, possibly representing a linear genome.

A main focus of the viral metagenomics in this study was to genomically identify the bacteriophage infecting Methylomirabilis cells (Figures 2–4). Based on the five viral genomes and over two thousand contig sequences obtained above, it is not directly clear which cellular hosts the recovered bacteriophages infect. Thus, to address this question we applied several computational approaches to predict this phage-host relationship from sequence signals (Edwards et al., 2015) that linked the assembled viral contigs to both the original genome and that of the new Methylomirabilis species. As described in Edwards et al. (2015), we calculated three main categories of phage-host prediction signals: genetic homology, similarity in CRISPR spacers, and similarity in oligonucleotide usage (Supplementary File 6). Genetic homology is expected to occur between the viral and bacterial genome sequences due to processes including transduction. This was identified at the nucleotide level (blastn, E ≤ 10⁻⁵) and at the protein level (tblastx, E ≤ 10⁻⁵), the former detecting only five short fragments, the latter detecting many more contigs including the long genomes 1–4. However, in all cases the matches between the bacteriophages and the Methylomirabilis genomes were based on relatively short, scattered regions of similarity with low sequence identity. As it was also showed by Edwards et al. (2015), the similarities in oligonucleotide usage profiles (k = 2, 4, 6) are not strong as phage-host predictors.

In the previously described M. oxyfera genome (Ettwig et al., 2010) one CRISPR array was present, which contained 23 spacers (Supplementary File 7). In the genome of a new Methylomirabilis species (Guerrero et al., in preparation) derived from the reactor from which the virome was sequenced, two CRISPR arrays were observed; one of which contained 26 spacers and the other one six spacers (Supplementary File 7). No significant nucleotide similarity was detected between these spacers and any of the 2094 possible viral contig sequences with fewer than 11/37 mismatches, which means these hits are indistinguishable from random noise (Edwards et al., 2015).

We also looked for signs of lysogeny in the M. oxyfera metagenome (Ettwig et al., 2010) and the new Methylomirabilis sp. metagenome (Guerrero et al., in preparation). The viral contigs were compared using BLASTn searching for contigs that were partly viral and partly bacterial. However, no such contigs were present in neither of the two metagenomes.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.