NOTE-1239: Santana-Molina, Williams, Snel & Spang — Chimeric origins and dynamic evolution of central carbon metabolism in eukaryotes

Contribution

It traces, enzyme by enzyme, where the eukaryotic central carbon metabolism (CCM) came from. The textbook view was that archaea gave eukaryotes their informational genes and bacteria, chiefly the mitochondrial ancestor, gave them metabolism. The paper finds archaeal, and specifically Asgard, contributions in glycolysis and the pentose phosphate pathway (PPP). It finds alphaproteobacterial contributions concentrated in the TCA cycle, and further contributions from other bacteria. It then maps how those enzymes were lost, replaced and retargeted across 207 eukaryotes after LECA. It is a pathway-scale study: 64 gene families, each tree inspected by hand.

Key insight

Syntrophy models say the archaeal and bacterial partners fed each other. If so, the eukaryotic metabolism they left behind should carry genes from both, not only from the bacterium. It does. The archaeal enzymes sit in the cytosol and in the oxidation of sugars, and the alphaproteobacterial ones sit in the mitochondrion and the TCA cycle. That matches a picture in which the archaeon fermented sugars and passed reducing equivalents to a bacterium that burned them. In the authors’ words, modern metabolism “preserves information about the nature of the original asgardarchaeal–alphaproteobacterial interactions”. The trace is faint, because most of the Asgard enzymes have since been replaced in many lineages.

Assumptions

• A LECA clade is a clade with two of three supergroups. Eukaryotic clades with members from at least two of Excavata, Amorphea and Diaphoretickes are “potential LECA clades”. The species tree is unrooted; “our interpretations of gene family origins do not assume a particular root position”.
• Donor by sister group. The prokaryotic donor of a LECA clade is its sister group in a maximum-likelihood tree (IQ-TREE, LG + C20 + G + F, ultrafast bootstrap). Trees were “manually rooted”, by outgroup where possible, otherwise arbitrarily “to ease visualization”. When sisters are mixed, the taxa in them are tallied (Fig. 3e).
• KEGG defines the pathways. Gene families come from KEGG orthology annotations, with homologous KO families merged by HMM comparison.
• Targeting predicted, not measured. Subcellular localization is TargetP prediction.
• Sampling. A core set of 483 archaeal, 487 bacterial and 224 eukaryotic proteomes (2021). Potential LECA clades were then searched against 187,681 prokaryotic and 44,889 viral proteomes, adding the top 15 hits per class up to 150 sequences.

Key results

• LECA had the full CCM. Nine of ten glycolytic (Embden–Meyerhof–Parnas) steps, seven of eight PPP steps and all ten TCA steps have putative LECA clades. The Entner–Doudoroff enzymes came later, in photosynthetic lineages. Pyruvate dehydrogenase, acetyl-CoA synthetase and lactate dehydrogenase were in LECA; the anaerobic alternatives (pyruvate formate lyase, POR, ADP-forming acetyl-CoA synthetase) are mostly post-LECA, though “others (such as POR) may have been present in LECA”. LECA had redundant isofunctional enzymes for several steps. The authors: “these phylogenies unambiguously show that LECA had the complete set of enzymes needed for the CCM”.
• Asgard contributions (Fig. 3a,d).
  • Clear: ADP-dependent glucokinase (ADPGK, 76% bootstrap), 2,3-bisphosphoglycerate-independent mutase APGM (89%), enolase (ENO, 39%), ribose 5-phosphate isomerase A (RPIA, 97%).
  • Ambiguous: GPMI_1 (a few Asgard sequences sister to eukaryotes) and pyruvate kinase PK_5 (mainly Amoebozoa nested in Asgard). For PK_5 and for a Lokiarchaeia–Fornicata group of ACDA (100%), the authors give two readings: present in LECA and later lost, or a post-LECA transfer.
  • Tentative and not Asgard: ATP-citrate lyase (sisters in DPANN and Thermoplasmatota) and malate dehydrogenase MDH1/2 (TACK and Baldarchaeota sisters, “long branches”). The Discussion still calls these “even more salient” possible “archaeal host contributions” to the mitochondrial TCA cycle.
  • The Discussion counts six Asgard contributions: ADPGK, GPMI, APGM, ENO, PK, RPIA. This contradicts earlier claims of none (Eme et al. 2023) or of enolase alone. Except for enolase, they are “patchily distributed in extant eukaryotes, apparently because of independent horizontal replacement events”.
• Alphaproteobacterial contributions (Fig. 3b). Seventeen: ALDO and TPI (glycolysis), PRPS (PPP), PDHA/B/C/D (pyruvate), and ten in the TCA cycle (IDH1, SUCA/B, LSC1/2, SDH1–4, FUMC); counts checked here. Most are mitochondrial; ALDO, TPI and PRPS are not. Signals are “not always unequivocal”: TPI and PRPS are sister to Alpha- and Gammaproteobacteria together, and other taxa intrude in LSC2 and ACS.
• Cyanobacterial contributions to plastid-bearing eukaryotes in glycolysis, the PPP and pyruvate conversion, and PDH trees that separate the green and red secondary plastid events.
• Other donors (Fig. 3c,e). Some with good support (> 95%): Chlamydia (TAL), Planctomycetota–Verrucomicrobiota (PGK, G6PD, RPE), Fusobacteriota (PK), Dependentiae (two LDH acquisitions), Chloroflexota (CS). Sister-group tallies show recurrent Myxococcota–Desulfobacterota, Bacteroidota and Acidobacteriota. “no additional dominant source of gene donations is apparent.”
• Unresolved (Fig. 2). At least 12 phylogenies give no clear LECA clade. For GPI, conserved intron positions suggest it was in LECA anyway, and a recombination between nuclear and plastid paralogues may explain its tree.
• Remodelling after LECA (Fig. 4).
  • Asgard-derived ADPGK, APGM and RPIA were replaced in some lineages by bacterial HK, PGMA/B and RPIB.
  • Enzyme repertoires correlate partly with lifestyle (|phi| ≥ 0.5): plastid status with glycolysis and the PPP, anaerobiosis with pyruvate and TCA enzymes.
  • Targeting does not follow origin. ALDO and TPI are alphaproteobacterial but not mitochondrial. CS, ACNB and IDH1/2 are not alphaproteobacterial but are mitochondrial. “The targeting of proteins is not always in agreement with their origins.”
  • In Archaeplastida, plastid glycolysis and PPP come mostly from duplicated nuclear genes relocated to the plastid, “which might have promoted the genome reduction of the endosymbiont”. The mitochondrial glycolysis of some SAR lineages is retargeting, not an endosymbiotic origin of glycolysis.
• The scenario (Fig. 5). Host and symbiont contributions plus continuous gene transfer. The compartment split “is consistent with symbiogenetic models”, and “Specifically, our results support the view of syntrophic interactions between host and endosymbiont, in which the archaeal partner produced reducing equivalents by the degradation of organic substrates via glycolysis which, in the absence of a suitable electron acceptor, were shuttled to a bacterial symbiont which contributed a TCA cycle and an electron transport chain”. Other donors “may be due to transient interactions with other prokaryotes before and during eukaryogenesis”.

Claims

id	claim	strength	support
C1	LECA had a complete central carbon metabolism, with redundancy at several steps	strong	LECA clades for nearly all steps across 64 trees, with intron evidence for unresolved cases
C2	Asgard archaea contributed several glycolytic and PPP enzymes to LECA, contrary to earlier claims of none or one	moderate	four trees with eukaryotes nested in Asgard (39–97% bootstrap); two more ambiguous between inheritance and later transfer
C3	Alphaproteobacterial contributions concentrate in the TCA cycle and pyruvate conversion and mostly act in the mitochondrion	moderate	17 trees, 10 in the TCA cycle; some sister groups mixed with Gammaproteobacteria
C4	Other prokaryotes contributed CCM enzymes, with no third dominant donor	moderate for the first part, weak for the second	individual well-supported trees; sister-group tallies; 64 families is a small sample for detecting a donor
C5	Since LECA, CCM was remodelled by loss, horizontal replacement and retargeting, so that origin no longer predicts compartment	strong	phylogenetic profiles over 207 eukaryotes; TargetP predictions
C6	The archaeal-cytosolic, bacterial-mitochondrial split supports syntrophy between an archaeal host and a bacterial endosymbiont	weak	the split is predicted by symbiogenetic models generally; no test against alternatives is made
C7	Archaeal TCA enzymes (MDH, ATP-citrate lyase) may be host contributions integrated into the mitochondrion	weak	non-Asgard archaeal sisters on long branches; called “tentative” in the Results

Concepts

• Central carbon metabolism (CCM): here, glycolysis (Embden–Meyerhof–Parnas and Entner–Doudoroff), the pentose phosphate pathway, pyruvate and acetate conversion to acetyl-CoA, and the TCA cycle.
• Potential LECA clade: a eukaryotic clade with members from at least two of Excavata, Amorphea and Diaphoretickes.
• EGT: endosymbiotic gene transfer, from an endosymbiont to the host nucleus.
• Syntrophy: metabolic cross-feeding in which one partner depends on the other, here “as an external electron sink”.
• Isofunctional enzymes: unrelated enzymes for the same reaction, such as the four phosphoglycerate mutases.

Connections

• Syntrophy models. It cites the reverse-flow model (Spang et al.), the revisited syntrophy hypothesis (López-García & Moreira) and Imachi et al.’s cultured Asgard archaeon (LIT-1591, unread here) among the models it supports. It sets itself against the view that “the archaeal host metabolism was replaced by counterparts from the endosymbiont”, for which it cites Martin & Müller’s hydrogen hypothesis (LIT-1597, unavailable here).
• Tobiasson et al. (LIT-1558, read in NOTE-1242) cite this paper for alphaproteobacterial associations with mitochondrial carbon metabolism. The two agree on that. They are harder to square on Asgard’s share. In this pathway set, alphaproteobacterial contributions (17) outnumber the Asgard ones (six, two ambiguous), while Tobiasson et al. report Asgard leading across “large parts of the core metabolic network”. The measures differ (counts of hand-inspected trees here, likelihood weights there), so this is a tension, not a contradiction. The inference is mine.
• Bernabeu et al. (LIT-1561, read in NOTE-1237). The same two phyla that paper names as major donors turn up here: Planctomycetota– Verrucomicrobiota as sister to three LECA enzymes, and Myxococcota– Desulfobacterota recurring in sister groups. This paper reads them as minor, “no additional dominant source”. The two also differ on LECA’s anaerobic enzymes. Bernabeu et al.’s reconstruction lacks pyruvate:ferredoxin oxidoreductase (PFO, the same enzyme as POR here); this paper allows that POR “may have been present in LECA”.
• The Fe–S parallel. The authors compare their mosaic to Fe–S cluster synthesis, with alphaproteobacterial mitochondrial steps and cytosolic steps of mixed origin, the pattern Tobiasson et al. also report.

Bearing on the record

• QUESTION-277. The record cited this work for “the chimeric metabolism”, and for discriminating syntrophy-first from complexity-first scenarios (LIT-1564’s “Why this is here”). The reading supports the first and not the second.
  • On discrimination. The paper does not address timing: nothing in it says whether the host was complex before the mitochondrion. Its syntrophy support rests on a compartment split that a host taking up a TCA-bearing bacterium by any route would also leave. What it does count against is the view that the host’s metabolism was simply replaced by the symbiont’s.
  • G, the distribution of essential functions across genomes. This is the paper’s dimension. In LECA, carbon metabolism was split by origin between the partners, and all of it is now nuclear-encoded. Origin and compartment have since come apart through retargeting.
  • C, host control through import. Mitochondrial TCA enzymes of non-alphaproteobacterial origin (CS, ACNB, IDH1/2) are a case of the host supplying the symbiont’s compartment with its own proteins. That is QUESTION-277’s gene transfer and import route, and its host compensation route, at the scale of one pathway.
  • Host compensation as a cause of symbiont reduction. In Archaeplastida, host genes were duplicated and sent to the plastid, which “might have promoted the genome reduction of the endosymbiont”. That is a mechanism the question’s routes do not name: the host’s supply makes the symbiont’s genes redundant. The authors offer it as a possibility only.
  • Replacement. Asgard-origin enzymes replaced by bacterial ones in many lineages show the theme of CASE-171 and LIT-1572 (constituents replaced, function kept) at the level of single genes.
  • D. The syntrophy reading implies an initial metabolic dependence of the archaeon on a bacterial electron sink. That is inferred from present gene origins, not observed.
• The live disagreement. This paper sits with the two-partner reading: “we could not identify a third dominant donor lineage”. Its other donors are “transient interactions”.
• Monocercomonoides. The metamonads’ mitochondrial PDHD and FUMC branch within LECA clades, consistent with secondary loss of mitochondria. That bears on CASE-168’s lineage, though not on Monocercomonoides itself.
• Filed as THEORY-1109. No instruction for machine-learning practice.

Limitations

• Single-gene trees over deep time. The authors call resolution “sometimes limited”. Enolase, the one Asgard contribution earlier work agreed on, has 39% bootstrap here. The trees themselves are in the unread supplement.

• Hand curation. LECA clades and orthogroups were “manually selected”, and trees manually rooted. That is careful but not reproducible as an algorithm, and the paper gives no estimate of curator error.

• A small sample for donor questions. 64 families can establish a donor’s presence but not its absence or its rank.

• The syntrophy inference outruns the data, as above. The authors list the distinct predictions syntrophy models make (“types of metabolites exchanged, the origin of eukaryotic cell membranes, the timing of mitochondrial acquisition, the mechanism of mitochondrial uptake and the origin of the nucleus”) and test none of them directly.

• The species tree is uncertain at points. Ancyromonadida and the Anaeramoebae move between analyses, and the PhyloBayes chains “did not converge”. The authors say gene-origin calls do not depend on the root.

• Slips.

  • “four enzymes of the EMP” introduces a list of five (GPI, ALDO class II, FBA, GAPDH, PGK).
  • The six cyanobacterial PPP contributions include EDD and EDA, which the paper elsewhere assigns to the Entner–Doudoroff pathway.
  • “Amoprhea” in the Results; “Mailin” for Malin in one place.

  None changes a result.

Open questions

• Would other pathways show the same archaeal-cytosol, bacterial-mitochondrion split? The authors expect so.
• Which of the syntrophy models’ distinct predictions can gene phylogenies test? The exchanged metabolites might be one, since the donors’ hydrogenases and electron carriers have trees of their own.
• Were the patchy Asgard enzymes in LECA and then replaced, or are some later transfers from Asgard archaea? PK_5 and the Lokiarchaeia–Fornicata ACDA are the test cases.
• Does host retargeting of its own enzymes into an endosymbiont generally precede and promote the symbiont’s gene loss? The plastid case is offered, not shown.