Isolation of an archaeon at the prokaryote–eukaryote interface

The origin of eukaryotes remains unclear1–4. Current data suggest that eukaryotes may have emerged from an archaeal lineage known as ‘Asgard’ archaea5,6. Despite the eukaryote-like genomic features that are found in these archaea, the evolutionary transition from archaea to eukaryotes remains unclear, owing to the lack of cultured representatives and corresponding physiological insights. Here we report the decade-long isolation of an Asgard archaeon related to Lokiarchaeota from deep marine sediment. The archaeon—‘Candidatus Prometheoarchaeum syntrophicum’ strain MK-D1—is an anaerobic, extremely slow-growing, small coccus (around 550 nm in diameter) that degrades amino acids through syntrophy. Although eukaryote-like intracellular complexes have been proposed for Asgard archaea6, the isolate has no visible organelle-like structure. Instead, Ca. P. syntrophicum is morphologically complex and has unique protrusions that are long and often branching. On the basis of the available data obtained from cultivation and genomics, and reasoned interpretations of the existing literature, we propose a hypothetical model for eukaryogenesis, termed the entangle–engulf–endogenize (also known as E3) model.

Reason for using a continuous-flow bioreactor system to enrich deep marine sedimentary microorganisms and deep-sea methane seep sediment as an inoculum source. Culture-independent molecular studies showed that deep marine sediment harbors phylogenetically diverse microorganisms, most of which belong to uncultured taxa and are distinct from those living on the Earth's surface 1-4 . Hence, their physiology and metabolic functions still remain largely unknown 5,6 . To gain insight into deep marine sedimentary microbes, they need to be cultivated, and this has been a significant challenge.
However, only a small fraction of indigenous deep subseafloor microbes has been successfully isolated and characterized 7,8 . It is unclear why the cultivation of deep marine sedimentary microbes is difficult, but the batch-type cultivation techniques commonly used in previous studies may have been inadequate for this purpose. Therefore, the development of a new cultivation technique is needed. We have, therefore, employed a continuous-flow bioreactor technique for the cultivation of deep marine sedimentary microbes since 2006. The bioreactor is called a down-flow hanging sponge (DHS) reactor, which was originally developed to treat municipal sewage at a low cost in developing countries [9][10][11] . Specifically, a polyurethane sponge packed in the DHS reactor column provides a large surface area for microbial colonization and a longer cell residence time.
As such, this type of continuous-flow reactor cultivation can provide substrates at low concentrations, similarly to those found in the natural environment. In addition, continuous-flow bioreactors allow the outflow of metabolic products that may inhibit microbial growth if accumulated. These continuous-flow reactors thereby might increase the culturability of deep marine sedimentary microorganisms in a controlled manner and serve as better sources (incubators) for microbial isolation than the original samples 12 . In fact, using DHS reactors, we have successfully enriched phylogenetically diverse microorganisms from deep marine sediments [12][13][14][15] and isolated and characterized various microorganisms using enriched microbial community from the bioreactors [16][17][18][19][20] .
In this study, we used deep-sea methane seep sediments collected off Kumano area, Japan. In deep-sea methane seep sediments, anaerobic oxidation of methane (AOM) reaction is the major microbial process and is mediated by a syntrophic association of euryarchaeal anaerobic methanotrophs (ANMEs) and deltaproteobacterial sulfate-deducing bacteria (SRB) 21 . In addition to ANMEs and SRB, abundant and diverse microorganisms, most of which are affiliated with uncultured microbial groups of high taxonomic ranks, such as phylum, class, and order, live in methane-seep sediments [22][23][24] .
Therefore, deep-sea methane-seep sediment can be regard as a hot spot for uncultured microorganisms. As such, cultivation and characterization of these uncultured microorganisms can greatly expand our knowledge regarding microbial physiology, genetics, and ecology. This was our rationale for using deep-sea methane seep sediment as an inoculum source for uncultured microorganism cultivation. However, in 2006, when we started the DHS bioreactor cultivation, there was extremely limited information about the metabolism of uncultured microorganisms because the metagenomic approach was not a common technique. We, therefore, could not predict the appropriate carbon and energy sources to culture uncultured microorganisms. However, we were aware that methane-seep microbial communities were sustained by methane released below the sea floor. Thus, we expected that if we provided methane as a major energy source in the DHS reactor system, the uncultured microorganisms could be cultivated from the methane-seep sediment under laboratory conditions, along with ANMEs and SRB. Indeed, using a combination of the DHS bioreactor "pre-enrichment" and subsequent in vitro cultivation, we have successfully cultured and isolated microorganisms representing predominant uncultured taxa. The "Candidatus Prometheoarchaeum syntrophicum" strain MK-D1 reported in this study is an example of cultured microorganism using our deep-sea methane seep-derived bioreactor enrichment.

Supplementary Note 2
Reason for use of the four antibiotics to isolate the Lokiarchaota. In parallel to the attempted the cultivation of microorganisms from methane-seep sediment, we tried to isolate anaerobic microorganisms from the enriched methanogenic microbial community in another DHS reactor, which was established from deep marine sediments collected off the Shimokita Peninsula, Japan 15 . During the isolation attempt, we detected few DSAG (i.e., Lokiarchaeota) sequences in a propionate-fed culture (Supplementary Table S8 in Imachi et al. [2011] 15 ). However, the Lokiarchaeota sequences became undetectable after five successive transfers. After this, we could not revive the culture any longer. However, we detected some Lokiarchaeota sequences in several anaerobic enrichment cultures supplemented with four antibiotics (i.e., ampicillin, vancomycin, kanamycin, and streptomycin, each at a final concentration of 50 µg/ml), via archaeal 16S rRNA genebased clone analysis (data not shown). This finding suggested that Lokiarchaeota members can tolerate these antibiotics. Therefore, we added these four antibiotics into the media to serve as selective agents for the isolation of Lokiarchaeota members from the enriched AOM microbial community in the DHS reactor.

Reason for provisional Candidatus status.
Although the strain was extensively characterized and other organisms have been isolated and published in co-cultures as well (e.g., Pelotomaculum schinkii strain HH 25 ), we opted to use the provisional Candidatus status due to challenges associated with (i) deposition and maintenance of the strain in culture collections (a requirement for validly naming a strain) stemming from the extremely low growth rate/yield and need for qPCR track growth and (ii) growing enough cell mass for accomplishing experiments that provide sufficient biological data for meeting the current standards for validly proposing a name for a strain.

Supplementary Note 4
Fate of 13 C-AAs in syntrophic degradation. MK-D1 and Methanobacterium cocultures fed with 13 C-AAs generated 13 C-enriched CH4 (Extended Data Table 2). On the other hand, 13 C-AA-fed tri-cultures of MK-D1 with Halodesulfovibrio and Methanogenium generated 13 C-enriched CO2. Given that the medium is buffered with 12 C-HCO3 and 12 CO2, any 13 CO2 produced by MK-D1 AA degradation would be diluted by the exogenous 12 C-carbonate pool and not directly reach the partner methanogen (i.e., very little 13 CH4 generation); thus, syntrophy in the tri-culture is primarily mediated by the oxidative path. Syntrophy in the co-culture is likely mediated by the hydrolytic formate-transferring path as 13 C-AA-derived 13 C-formate would directly reach the symbiotic partner, enrich the intracellular carbonate pool through oxidation to 13 CO2, and supply 13 CO2 for 13 CH4 generation. With ocean oxygenation, sulfate levels rose 26 , which may have afforded an opportunity for thermodynamically enhanced syntrophy via interaction with SRB 27 (an interaction evidenced by MK-D1 cultures). In addition, SRB are aerotolerant 28 and produce a reductant that can passively consume O2 (H2S). By contrast, the alternative syntrophic partner, methanogenic archaea, are sensitive to O2 and produce an inert byproduct (CH4).
We suggest that interaction with SRB may have conferred a low-level of aerotolerance to enable the archaea to inhabit more oxygenated environments, where excess organic matter is predicted to have been available from increased cyanobacterial activity during the Great Oxidation Event 29,30 . Growth in oxygenated environments may have further selected for a more refined aerotolerance.

Supplementary Note 7
Transition towards aerobiosis may have benefited from multiple symbiosis.
To further adapt to higher O2 concentrations and compete with facultatively aerobic organotrophs, interaction with an O2-consuming partner might have been beneficial in gaining aerotolerance 31,32 . The alternative hypothesis would require the gain of genes for aerobic respiration, but this may have significantly weakened the selection for (endo)symbiosis between the host archaeon and pre-mitochondrial alphaproteobacterium symbiont (PA), assuming that the host-PA symbiosis that led to endosymbiosis was driven by the host's sensitivity to O2 and dependency on an O2-scavenging partner as we posit.
Given that, with the current data, eukaryotes and Ca. Heimdallarchaeota represent the most recent branchpoint of archaea and eukaryotes, the lineages of Ca.
Heimdallarchaeota that possess aerobic respiration genes 33 [40][41][42] . If AAC was originally used for parasitism, the above loss of 2-oxoacid catabolism has the potential to resolve parasitism by reducing host ATP production and reversing the ATP transport direction of AAC (direction dependent on ATP concentration 39 ).

Supplementary Methods
Culturing. The purity of Ca. P. syntrophicum strain MK-D1 was routinely examined by microscopy and iTAG analysis. In addition, the purity was verified by the whole genome shotgun sequencing, which only detected the sequences of MK-D1 and Methanogenium genomes. We also confirmed the culture purity based on failure to amplify the bacterial 16S rRNA gene through PCR using the bacterial primer pairs 27F/907R 24 and EUB338F*/1492R [43][44][45] . Moreover, we evaluated the culture purity based on the failure of microbial growth in the following media at 10°C, 20°C, 30°C, 37°C, and 55°C: (i) thioglycolate medium (Difco); (ii) basal medium supplemented with 1 mM sucrose, 1 mM glucose, 1 mM fructose, 1 mM xylose, and 0.01% (w/v) yeast extract; and (iii) basal medium containing 5 mM lactate, 10 mM sulfate, 0.05% (w/v) CA, and 0.01% (w/v) yeast extract.
To confirm Halodesulfovibrio has ability to use H2 and formate, we isolated the bacterium from the MK-D1 enrichment culture using a roll-tube technique, with lactate (10 mM) and sulfate (10 mM), acting as an electron donor and acceptor, respectively.
After isolation, we confirmed that the Halodesulfovibrio, designated strain MK-HDV, could grow on a hydrogen-or formate-fed medium supplemented with sulfate at 20°C.
Methanobacterium sp. strain MO-MB1 was previously isolated from subseafloor sediment in our laboratory as a hydrogen-and formate-utilizing methanogenic archaea 15 .

Supplementary Figure 7 | Hypothetical protein with profilin-like domain. a,
Maximum likelihood tree. Homologs were collected through BLASTp analysis of the Asgard archaea sequences against the UniProt database (release 2019_09). Of homologs with sequence similarity ≥20%, representative sequences were selected using CD-HIT (c 0.8). Additional homologs with verified biochemical activity, sequence similarity ≥20% were collected through BLASTp analysis of the Asgard archaea sequences against the UniProt/SwissProt database (2019_09). Sequences were aligned using MAFFT v7. Only sequences with ≥70% overlap with the corresponding MK-D1 sequence were retained for further analyses. The phylogenetic tree was constructed using RAxML-NG (--model LG+G4+F) and 100 bootstrap replicates. 221 sites of the alignment were used for tree construction. b, Domain analysis based on sequences registered in NCBI CDD (v3.17) profilin domain pfam00235. A sequence logo is shown for the region of the CDD alignment overlapping with the consensus sequence registered in CDD. Amino acid colors are based on their chemistry (polar, neutral, basic, acidic, and hydrophobic). For each position, the letter width corresponds to the percentage of non-gap sequences. MK-D1 sequences were aligned to the CDD reference alignment (mafft --add x --keeplength). Residues aligned with any positions with bit values ≥1.5 are bolded and colored based on their chemistry. Those with chemistry consistent with at least one residue in the CDD alignment are shown with a yellow background (otherwise orange background). Of homologs with sequence similarity ≥20%, representative sequences were selected using CD-HIT with a clustering cutoff of 80% similarity (default settings otherwise). Additional homologs with verified biochemical activity, sequence similarity ≥20% were collected through BLASTp analysis of the Asgard archaea sequences against the UniProt/SwissProt database (2019_09). Sequences were aligned using MAFFT v7 with default settings. Only sequences with ≥70% overlap with the corresponding MK-D1 sequence were retained for further analyses. The phylogenetic tree was constructed using RAxML-NG using fixed empirical substitution matrix (LG), 4 discrete GAMMA categories, empirical amino acid frequencies from the alignment, and 100 bootstrap replicates. 225 sites of the alignment were used for tree construction. b, Domain analysis based on sequences registered in NCBI CDD (v3.17) ribosomal protein L22e pfam1776. A sequence logo is shown for the region of the CDD alignment overlapping with the consensus sequence registered in CDD. Amino acid colors are based on their hydrophobicity (hydrophilic, neutral, and hydrophobic). For each position, the letter width corresponds to the percentage of non-gap sequences. MK-D1 sequences were aligned to the CDD reference alignment (mafft --add x --keeplength). Residues aligned with any positions with bit values ≥2 are bolded and colored based on their chemistry. Those with chemistry consistent with at least one residue in the CDD alignment are shown with a yellow background (otherwise orange background). c, Comparison of secondary structure between eukaryotic ribosomal protein L22e and MK-D1 homolog (HHpredpredicted) (blue = beta sheet; red = alpha helix).  Figure 6 caption for details. MUSCLE v.3.8.31 was used for alignment instead of MAFFT and CD-HIT clustering was performed with a 60% cutoff rather than 70%. Positions with gaps in more than 10% of the sequences were excluded from the alignment using trimAl v1.2 (gt 0.9; and default settings otherwise). 553 sites of the alignment were used for tree construction.
Supplementary Figure 17 | Maximum likelihood tree of Asgard archaea biotin ligase (BirA). BirA homologs were collected through BLASTp analysis of the Asgard archaea sequences against the UniProt database (release 2019_06). Of homologs with sequence similarity ≥40% and overlap ≥70%, representative sequences were selected using CD-HIT with a clustering cutoff of 70% similarity (default settings otherwise). Additional homologs with verified biochemical activity, sequence similarity ≥30%, and overlap ≥70% were collected through BLASTp analysis of the Asgard archaea sequences against the UniProt/SwissProt database. Sequences were aligned using MAFFT v7 with default settings and trimmed using trimAl with default settings. The phylogenetic tree was constructed using FastTree using fixed empirical substitution matrix (LG) and 1000 bootstrap replicates. 332 sites of the alignment were used for tree construction.