NOTE-1238: Imachi et al. — Isolation of an Asgard archaeon at the prokaryote–eukaryote interface

Contribution

Before it, Asgard archaea were known only from metagenomes. Their physiology and cell biology were inferred from gene content. This paper grows one. It shows what the cell looks like, what it eats and that it cannot grow without partners that remove its hydrogen and formate. It also gives a closed genome, which removes the contamination worries that attend metagenome-assembled genomes. From these it proposes a model of eukaryogenesis.

Key insight

The organism most closely related to eukaryotes that anyone has grown is a small, internally simple cell whose way of life is dependence on other microbes. It ferments amino acids and can do so only while a partner consumes the hydrogen and formate it makes. Its biosynthesis is also incomplete. The authors’ model makes this dependence the starting point of eukaryogenesis: a host already “entangled” with partners by metabolism and by its protrusions, which later engulfs one of them and then “endogenizes” it, handing over its own energy metabolism.

Assumptions

• A modern relative stands in for an ancient host. The authors stress that “roughly 2 billion years” separate MK-D1 from the lineage that led to LECA. The model rests on “reasoned assumptions” and “extrapolations”.
• The Asgard ancestor resembled MK-D1. The model assumes it was “syntrophic, internally simple (that is, similar to MK-D1)” and lived in anaerobic marine sediment.
• Gene presence stands for capacity. The ancestral reconstruction rests on which catabolic genes are present across metagenome-assembled Asgard genomes and form monophyletic clusters.
• Lipids by subtraction. Lipids were measured on the co-culture, and MK-D1’s composition is inferred using a reference Methanogenium isolate.
• Phylogeny among cultured organisms. The ribosomal-protein trees were built from cultured organisms only. Metagenome-derived sequences were used for the alignment and then removed “to avoid any influence of contamination”, except for Ca. Korarchaeum.

Key results

• Cultivation (Fig. 1, Extended Data Fig. 1, Extended Data Table 1).
  • A methane-fed bioreactor ran for more than 2,000 days on Nankai Trough seep sediment. Subcultures on casamino acids with four antibiotics gave faint turbidity after about a year.
  • Growth is slow: a lag of 30–60 days, more than 3 months to full growth, about 10⁵ 16S rRNA gene copies per ml, and a doubling time of about 14–25 days. Temperature and substrate changes did not improve it.
  • After six transfers MK-D1 was 13% of a tri-culture with Halodesulfovibrio (85%) and Methanogenium (2%). Later transfers removed Halodesulfovibrio, giving “a pure co-culture” with Methanogenium.
• Syntrophy (Fig. 2, Extended Data Tables 2–3).
  • MK-D1 degrades ten amino acids. Inhibitors of its partners (molybdate for sulfate reducers, 2-bromoethanesulfonate for methanogens) “significantly impaired” its growth.
  • High hydrogen or formate “completely suppressed” growth, as thermodynamic dependence on a partner predicts.
  • ¹³C labelling of methane and CO₂ varied with the partner, so MK-D1 makes both H₂ and formate. The genome has one hydrogenase (MvhADG–HdrABC) and one formate dehydrogenase (FdhA).
  • The partner is replaceable: MK-D1 also grows with Methanobacterium sp. MO-MB1.
  • NanoSIMS shows uptake of amino-acid nitrogen into MK-D1 aggregates, “performed without replicates”, on two kinds of culture.
  • Fourteen culture conditions did not raise yield.
  • The metabolic reconstruction (Fig. 2a) marks vitamin B12, biotin, lipoate and thiamine pyrophosphate among compounds predicted to come from the partners.
• Cells (Fig. 3, Extended Data Fig. 3).
  • Cocci 300–750 nm across (average 550 nm), in aggregates with extracellular polymer.
  • Cryo-EM and TEM show “no visible organelle-like inclusions, in contrast to previous suggestions”. The cryo-EM rests on 14 images from two observations, TEM on 131 from six. DNA and ribosomes could not be told apart.
  • Membrane vesicles (50–280 nm), chains of blebs, and protrusions 80–100 nm wide, of various lengths and often branching. They form neither networks nor connections between cells.
  • Archaeal ether lipids (C20-phytane, C40-biphytanes with 0–2 rings), and no genes for ester lipids.
• Genome and expression (Fig. 4a, Extended Data Figs 2, 4).
  • A closed genome of 4.46 Mb.
  • Ribosomal-protein trees (31 proteins, maximum likelihood and Bayesian) show MK-D1 as sister to eukaryotes among the cultured taxa. “Thus, MK-D1 represents the closest cultured archaeal relative of eukaryotes.”
  • 80 eukaryotic signature proteins. 23 are among the 500 most highly expressed genes, including actin-, gelsolin-, ubiquitin- and ESCRT-III-related proteins. Three possible division systems (FtsZ, actin, ESCRT) are expressed together.
• The Asgard ancestor (Fig. 4b, Extended Data Figs 5–7). Most Asgard genomes encode amino-acid catabolism and NiFe hydrogenases and have restricted biosynthesis. Catabolic genes for histidine, serine and threonine degradation and for butyrate and propionate fermentation are conserved across phyla including Heimdallarchaeota. The authors propose that the last Asgard common ancestor “was an amino-acid-degrading anaerobe that produced H2 and fatty acids as by-products … and depended on metabolic partners, although we do not reject the possibility of other additional lifestyles”.
• The E3 model (Fig. 5). “a hypothetical model”, “not definitive”.
  • Entangle: the host degrades amino acids with H₂-scavenging sulfate reducers, then interacts with a facultatively aerobic partner that scavenges O₂ (the future mitochondrion), as oxygen rises.
  • Engulf: protrusions wrap the partner, and a membrane around the chromosome forms at the same time. Phagocytosis is judged unlikely, given the host’s small size and lack of machinery and energy. “further evidence is required to support this conjecture”.
  • Endogenize: the host shares 2-oxoacids with the endosymbiont. It then loses its own anaerobic 2-oxoacid catabolism, which eukaryotes lack, resolving “catabolic redundancy” and O₂ sensitivity. An ADP/ATP carrier makes the symbiont the cell’s ATP supply.
  • The authors depart from hydrogen-transfer models: “Although previous models propose that H2 transfer was a key interaction that drove endosymbiosis …, we believe that current data favours the above interaction” (Supplementary Note 8, not read).

Claims

id	claim	strength	support
C1	MK-D1, an Asgard archaeon, can be grown in stable co-culture, and is the closest cultured archaeal relative of eukaryotes	strong for the culture; moderate for the placement, which holds among cultured taxa	12 years of enrichment; closed genome; ribosomal-protein trees without MAGs
C2	MK-D1 degrades amino acids only in syntrophy, transferring H₂ and formate to replaceable partners	strong	partner inhibitors, suppression by H₂ and formate, isotope data, growth with two different methanogens
C3	MK-D1 has no visible organelle-like internal structure, and has long branching protrusions	moderate to strong	cryo-EM (14 images), TEM (131), SEM (122); for one organism in culture
C4	The last Asgard common ancestor was a partner-dependent, H₂-producing amino-acid degrader	weak to moderate	gene presence across incomplete MAGs and conserved clusters; one cultured species
C5	Eukaryogenesis proceeded by entanglement, engulfment by protrusions, and endogenization of an O₂-scavenging partner	weak	a hypothesis by the authors’ own description; built on C2–C4 and the literature
C6	Phagocytosis was not the mechanism of engulfment	weak	argued from the host’s small size and assumed lack of machinery and energy

Concepts

• Syntrophy: growth that depends on a partner consuming one’s products, here H₂ and formate, so that degradation stays thermodynamically favourable.
• Eukaryotic signature proteins: proteins otherwise found only in eukaryotes. MK-D1 has 80.
• Pure co-culture: MK-D1 with one partner and nothing else. It is not an axenic culture, which is why the name keeps Candidatus (Supplementary Note 3, not read).
• E3 (entangle–engulf–endogenize): the authors’ three-stage model, described above.
• Endogenization: the host delegating catabolism and ATP generation to the endosymbiont, and losing its own.

Connections

The paper builds on the metagenomic discovery of Lokiarchaeota and the Asgard superphylum, and on metagenomic proposals of Asgard hydrogen metabolism, some of which it contradicts in detail. Its engulfment step takes up a “previously proposed alternative route” in which external structures enclose a partner. It sets aside hydrogen-transfer models, the hydrogen hypothesis among them (LIT-1597, unavailable). Kay et al. (LIT-1566, read in NOTE-1234) cite this isolate’s protrusions and its prokaryote-like size, and class E3 as “mitochondria intermediate”. Kay et al. place nFECA as sister to Heimdallarchaeia, not to the Loki lineage that MK-D1 belongs to, which agrees with this paper’s narrower claim of “closest cultured” relative. The paper dates the TACK–Asgard lineage to “approximately 2.1–2.4 billion years ago”, citing other work. Kay et al. date nFECA, a younger node, to 3.05–2.79 Ga. The two timescales do not fit together. The comparison is mine.

Bearing on the record

• QUESTION-277. The question registered this work as “the empirical case for” syntrophy. The reading supports that for the host lineage’s way of life, with corrections and one new point.
  • Dependence before endosymbiosis (D). The host lineage, on this evidence, was already obligately dependent on other microbes, for removing H₂ and formate and probably for some vitamins. But the dependence is on a service, not on a partner. Methanogenium, Methanobacterium and Halodesulfovibrio each serve. That is high D with no partner specificity, no cohesion (R) and no control (C). It is a case of the question’s “interdependence that is not individuality”, at the start of the most famous transition. The inference is mine.
  • A loss route inside the endosymbiosis (G). E3’s endogenization is a host-side loss of a redundant function, its own 2-oxoacid catabolism, made affordable because the symbiont supplies the product. That resembles the Black Queen route (LIT-1568, NOTE-1212, THEORY-1048), but inside a partnership and with a non-leaky product passed through a dedicated ATP carrier. The authors propose it; it is not shown. The comparison is mine.
  • Not the hydrogen hypothesis. The question’s later note groups this work with Martin & Müller (LIT-1597) under syntrophy. Imachi et al. make the host the H₂ producer and say current data favour an interaction other than H₂ transfer as the driver of endosymbiosis. The two works share the word “syntrophy” and differ on who feeds whom and why the partner entered.
  • Weak bearing on host complexity. The LIT says the isolate “bears on” whether host complexity preceded the mitochondrion, since it “shows no internal complexity”. The authors themselves set 2 billion years between MK-D1 and the host lineage, and Kay et al. say the same of cultivated Asgard archaea. One modern relative without internal membranes neither shows nor refutes an ancient complex host.
• NOTE-159. O’Malley and Parke cite Asgard inferences to doubt that eukaryogenesis was primarily a transition in individuality. A host whose ancestors lived by dependence on a guild of partners is consistent with that doubt. The inference is mine.
• Filed as THEORY-1108. No instruction for machine-learning practice.

Limitations

• One organism, not axenic. All characterization is of MK-D1 with partners. Lipids were inferred by subtracting a reference strain.
• “Closest relative” is among cultured taxa. The trees omit metagenome-derived Asgard genomes, including the Heimdallarchaeota that other work places closer to eukaryotes.
• The ancestral state is inferred from gene presence in incomplete metagenome-assembled genomes. The authors keep other lifestyles open.
• Small imaging samples. The cryo-EM rests on 14 images, and NanoSIMS had no replicates, as the authors say.
• The E3 model is a hypothesis. Its steps (an SRB stage, engulfment by protrusions, a protonuclear membrane formed during engulfment) are not tested, and the supporting arguments sit in supplementary notes not read here.
• Inconsistent timescale. The cited 2.1–2.4 Ga for the TACK–Asgard lineage conflicts with later dating of nFECA (LIT-1566).

Open questions

• Do cultured Heimdallarchaeia, or other lineages closer to eukaryotes, share MK-D1’s obligate syntrophy and internal simplicity?
• Is the partner dependence obligate in principle, or does MK-D1 grow alone under H₂-removing conditions (a gas-sparged culture, an electrode)? That would separate thermodynamic dependence on a service from dependence on organisms.
• What are the protrusions for? The authors suggest physical interaction with partners; whether they wrap or exchange with partners is not shown.
• Can a symbiont-to-host ATP supply evolve before the host loses its own catabolism, as E3 needs?