NOTE-1234: Kay et al. — Dated gene duplications and the assembly of eukaryotes

Contribution

It puts absolute dates on the order in which eukaryotic features were assembled. Earlier work (Pittis & Gabaldón, LIT-1595, among it) ordered events by the relative lengths of gene-tree branches. This paper dates individual duplication nodes with a relaxed clock. It carries time information from a dated species tree of archaea, bacteria and eukaryotes into each gene tree as priors. Three results follow:

• the eukaryotic stem from the archaeal side (nFECA, 3.05–2.79 Ga) is about twice as long as the one from the alphaproteobacterial side (mFECA, 2.37–2.13 Ga) before LECA (1.80–1.67 Ga);
• most duplications in archaeal-origin families predate the onset of alphaproteobacterial-origin duplications, taken as mitochondrial endosymbiosis at about 2.2 Ga;
• most duplications in non-alphaproteobacterial bacterial families come after it.

Key insight

A gene tree has no node for “the mitochondrion arrived”. It does have nodes where a eukaryotic gene duplicated before LECA, and each of those can be dated. Dating them gives a ceiling for when each paralogue’s function could have begun: “Dating this node provides a maximum estimate for the timing of origin of their paralogue-specific functions”. The authors read the first burst of duplications among genes from the alphaproteobacterial partner as the start of its integration. Against that marker, the host’s own families (actin and its ARP2/3 paralogues, γ-, δ- and ε-tubulin, Golgi and ER trafficking proteins, RNA polymerase subunits, nuclear-targeted paralogues) were already duplicating. Hence “complexified archaeon, late mitochondrion” (CALM).

Assumptions

• The species tree. 62 marker genes, concatenated to 15,761 positions. nFECA branches within Asgard archaea, sister to Heimdallarchaeia. mFECA is sister to Alphaproteobacteria excluding Magnetococcales. Eighteen fossil calibrations, uniform priors, cross-bracing of equivalent nodes (18 calibration instances, 88 congruent splits), LG + F + G5, four MCMCTree runs.
• Sequential Bayesian priors. Pooled species-tree posteriors become skew-T priors and maximum constraints on gene-tree speciation nodes. Duplications in genes of Asgard (or TACK-sister) origin are constrained to lie between nFECA and LECA. Duplications in genes of alphaproteobacterial origin are constrained to lie between mFECA and LECA. The authors say the hierarchy means “the inferred ages of gene duplications are informed by the ages of nodes on the dated species tree”.
• Duplication onset marks endosymbiosis. Gene duplication is much more frequent in eukaryotes than prokaryotes, so the first fixed duplications in symbiont-derived genes are taken to show the symbiont was already inside. This is argued, not tested.
• Function follows duplication. A duplication date is a maximum age for the paralogue’s specific role. For targeting changes, the paper says so: “On the assumption that the change in localization occurred after gene duplication”.
• Present function stands for ancestral function. Localization and function of each LECA group are taken from human reference proteins.
• Sample. Only families of prokaryotic ancestry with dateable pre-LECA duplications and alignments good enough for clock analysis. Genes novel to eukaryotes are excluded, because their duplication node is also the root. That leaves out most of what defines the nucleus, such as nuclear pores.

Key results

• The timeline (Fig. 1). nFECA 3.05–2.79 Ga, mFECA 2.37–2.13 Ga, LECA 1.80–1.67 Ga. The authors give median stem durations of about 1.1 and 0.6 Gyr. Differences of the interval midpoints give about 1.19 and 0.52 Gyr, as computed here. The posterior medians are not in the text, so the 0.6 could not be checked exactly. The sensitivity analyses (Supplementary Note 3, not read) are reported to leave nFECA and mFECA and the gap between them robust.
• The sample. 135 time-resolved gene trees: 95 of archaeal origin and 40 of bacterial origin. 9 of the 95 lack support for Asgard origin. 19 of the 40 lack strong support for alphaproteobacterial origin.
• Dating the mitochondrion (Figs 1d, 4a).
  • Alphaproteobacterial-origin duplications begin soon after mFECA, “within 200 Myr”, which the authors align to “around 2.2 Ga”.
  • Among the earliest are paralogues that work outside the mitochondrion (METTL4, 2.24–2.11 Ga; type Y polymerases, 2.24–2.12 Ga; mismatch repair, 2.18–2.08 Ga), and parts of the protein import system (TIM14 and PAM16, 2.24–1.99 Ga; TIM44, 2.22–2.00 Ga; PMPCA and PMPCB, 2.16–1.94 Ga).
  • “85% of the pre-LECA duplicates of archaeal origin were estimated to have been fixed prior to mitochondrial endosymbiosis.” The text does not say whether the base is 95 or 86 families.
  • 16 of 19 duplications of bacterial but not alphaproteobacterial origin have average ages after the inferred endosymbiosis.
• Cytoskeleton (Fig. 2a). Actin duplications from about 2.8 Ga. ARP3 2.90–2.47 Ga and ARP2 2.72–2.06 Ga, a maximum age for branched actin and so for phagocytosis “as carried out by extant eukaryotes”. γ-tubulin 2.84–2.41 Ga, δ 2.75–2.25, ε 2.61–2.07, and the filament-forming α and β later, 2.47–1.92 Ga. The authors conclude “an organized microtubule network before mitochondrial endosymbiosis, probably by 2.7 Ga”.
• Endomembranes (Figs 3, 4b,c). The oldest trafficking duplications are in Golgi–ER–plasma-membrane families (TRAPPC3, 3.00–2.54 Ga; YIPF1/2, 2.91–2.55 Ga; COPD, 2.90–2.61 Ga). Endosome, lysosome and autophagosome families begin around 2.4 Ga (for example STX7/STX12, 2.63–2.04 Ga; ARF6, 2.55–1.84 Ga). Bacterial-origin lipid enzymes duplicate late (SPTLC1–3, 2.16–1.99 Ga; ACSL, 2.12–1.87 Ga), some not from Alphaproteobacteria, which the authors read as “late complexification” of the ER by bacterial genes from non-mitochondrial sources. Endolysosomal transporters (CLC, SLC, ABCB) date with endolysosomal trafficking, and the CLC and SLC families are not of alphaproteobacterial origin.
• Nucleus (Fig. 2b,c). Sm–LSm duplications span 2.91–1.87 Ga. The nuclear localization signals in two subunits arose before the final duplication, dated 2.87–1.87 Ga. RNA polymerase I/II/III duplications fall in 2.91–2.34 Ga, pol III’s at 2.73–2.32 Ga. Nuclear-targeted paralogues fall in 2.86–2.08 Ga. The authors note that localization signals “might have evolved before the nuclear compartment”.
• Scenarios (Fig. 5, Discussion).
  • Mitochondria-early (the hydrogen hypothesis, LIT-1597) is rejected.
  • Serial endosymbiosis is argued against, because most non-alphaproteobacterial genes duplicate after the mitochondrion.
  • Phagocytosis is left open: “our results do not allow us to discriminate between the origin of phagocytosis and the timing of mitochondrial endosymbiosis”.
  • The results are “not fully compatible with any of the competing evolutionary scenarios”, and fit CALM models best.
• Environment. Mitochondrial endosymbiosis is “almost coincident” with the Great Oxidation Event (2.43–2.22 Ga), in oceans that stayed largely anoxic. The authors say this is “compatible with” syntrophy models, while their timeline “precludes a syntrophic association between nFECA and mFECA prior to the GOE”.
• The abstract’s dates against the body. The abstract puts the cytoskeleton, trafficking, endomembranes, phagocytotic machinery and a nucleus “all between 3.0 and 2.25 billion years ago”. The body’s intervals for these features run lower: ARP2 to 2.06, α/β-tubulin to 1.92, the Sm–LSm signal duplication to 1.87, nuclear-targeted paralogues to 2.08. Many overlap the 2.2 Ga marker. The 3.0–2.25 Ga range is the span of the older estimates, not of every interval.

Claims

id	claim	strength	support
C1	The archaeal stem (nFECA to LECA) is about twice as long as the alphaproteobacterial stem (mFECA to LECA)	moderate	species-tree clock with 18 calibrations; sensitivity analyses reported robust (Supplementary Note 3, not read)
C2	Mitochondrial endosymbiosis was under way by about 2.2 Ga	moderate to weak	the onset of alphaproteobacterial-origin duplications, which the priors confine to after mFECA; rests on the assumption that duplication onset marks endosymbiosis
C3	Most archaeal-origin duplications (85%) were fixed before mitochondrial endosymbiosis	moderate	95 gene trees; partly set by the priors, which let archaeal duplications range back to nFECA and keep alphaproteobacterial ones after mFECA
C4	The host had an elaborated cytoskeleton, endomembrane trafficking and nuclear localization before the mitochondrion	moderate for gene families, weak for cellular features	dates are maximum ages for paralogue functions; many intervals overlap 2.2 Ga; features inferred from human orthologues’ present roles
C5	Mitochondria-early scenarios are rejected	moderate	follows from C3 and C4, granted C2
C6	Serial endosymbiosis is not supported, since most non-alphaproteobacterial duplications postdate the mitochondrion	weak	16 of 19 families; duplication dates, not acquisition dates, which the authors grant (“Irrespective of whether these genes were already present …”)
C7	Phagocytosis and mitochondrial endosymbiosis cannot be ordered	moderate	stated by the authors; endolysosomal duplications date with the earliest alphaproteobacterial ones

Concepts

• nFECA and mFECA: the points where the eukaryotic lineage branches from its closest sampled archaea and alphaproteobacteria. They move if closer relatives are found.
• Pre-LECA duplication: a node uniting two LECA clades in a gene tree. Its age is a maximum for the paralogues’ distinct functions.
• Cross-bracing: fixing equivalent nodes (the same speciation in the nuclear and mitochondrial copies of the tree) to the same age.
• Mitochondria-early, -intermediate, -late: the authors’ grouping of scenarios. Early: the hydrogen hypothesis. Late: the phagocytosing archaeon, syntrophy and serial endosymbiosis. Intermediate: reverse flow, inside-out and E3 (LIT-1591’s model).
• CALM: “complexified archaeon, late mitochondrion”, the class of models consistent with the results.

Connections

The authors place their method against branch-length studies that “used gene tree branch lengths, expressed in number of substitutions per site and normalized by relative evolutionary rates post-LECA”. That is the method of Pittis & Gabaldón (LIT-1595, read in NOTE-1240). The paper reaches the same order for the mitochondrion. It disagrees on what came before. Pittis & Gabaldón infer a host already chimaeric with non-alphaproteobacterial genes, and list serial symbioses among the explanations. Kay et al. find most non-alphaproteobacterial duplications after the mitochondrion and argue against serial endosymbiosis. The two measure different things (duplication against acquisition), so the disagreement is partial. The comparison is mine.

On Asgard cell biology the paper cites the cultivated isolate of LIT-1591 (read in NOTE-1238): cytoskeleton-supported protrusions, but “in terms of overall size and complexity, the Asgard archaea cultivated so far resemble typical prokaryotes”. It classes E3, that paper’s model, as mitochondria-intermediate and consistent with early cytoskeletal development. It counts the hydrogen hypothesis (LIT-1597) as mitochondria-early and rejected. Tobiasson et al. (LIT-1558), Bernabeu et al. (LIT-1561) and Santana-Molina et al. (LIT-1564) are not cited in the text read here.

Bearing on the record

• QUESTION-277, eukaryogenesis as a test. The question set this paper as the “late mitochondrion after a complex host” programme. The reading bears that out, with two refinements.
  • It speaks to one endosymbiotic partner, not a sequence. The paper argues against serial endosymbiosis. Its non-alphaproteobacterial genes are explained as transfer, mostly after the mitochondrion. On the question’s choice between “one two-partner merger or a sequence of associations”, it takes the first, preceded by a long phase of host elaboration with no partner.
  • What it dates is G and, partly, C. The host’s complexity before the mitochondrion is its own gene duplication, not dependence on anyone. After the mitochondrion, the earliest dated symbiont-gene duplications include the protein import system (TIM14/PAM16, TIM44, PMPCA/PMPCB). If the onset marker is right, the machinery for importing proteins into the symbiont was being elaborated within about 200 Myr. That is the gene-transfer-and-import route of the question, early in this association. The inference is mine.
  • Nothing on D, R or S. The paper does not address when partners became dependent, when they reproduced together, or conflict between them.
• NOTE-159. O’Malley and Parke report inferences placing the mitochondrion “late” and question whether eukaryogenesis was “driven primarily by a transformation in individuality”. This paper is one of those inferences, dated. It supports the premise. It is silent on individuality.
• THEORY-141. No bearing on selection levels. A host that complexified for about half a billion years before the endosymbiosis is a case where much of the new cell’s machinery owes nothing to the collective. The inference is mine.
• Filed as THEORY-1099. No instruction for machine-learning practice.

Limitations

• Maximum ages read as presence. A duplication date bounds a paralogue’s function from above. To say the host “had” a feature before the mitochondrion needs the function to have followed the duplication soon after. The paper states this for nuclear targeting and leaves it implicit elsewhere.
• The priors shape the order. Alphaproteobacterial duplications cannot be older than mFECA; archaeal ones can reach back to nFECA. The authors grant this and report robustness in Supplementary Note 3, which was not read.
• The marker is an assumption. Equating the onset of symbiont-gene duplication with endosymbiosis is argued from the rarity of duplication in prokaryotes.
• Overlapping intervals. Many host-feature intervals reach below 2.2 Ga. The abstract’s “all between 3.0 and 2.25” summarizes the older ends.
• A biased sample. Only well-conserved, dateable families of prokaryotic origin. Archaeal families were easier to analyse, which the authors note “may simply reflect” more duplications on the archaeal stem, or may not. Eukaryote-specific genes, including most of the nucleus, are excluded.
• Function from human orthologues. Present localization in human cells stands for the role a paralogue had at its origin.
• Not read here: the supplementary notes, the gene-tree reports and the code. The robustness claims rest on them.

Open questions

• Does the onset of duplication in symbiont-derived genes coincide with endosymbiosis, or lag it? A dated case with a known entry time (a younger endosymbiosis) would calibrate the marker.
• When were the non-alphaproteobacterial genes acquired, as against duplicated? This decides between transfer to the host before the mitochondrion and transfer after it.
• Can the endolysosomal system and the mitochondrion be ordered? The authors say their data cannot.
• How would placing nFECA closer (a deeper-sampled Heimdallarchaeia relative) move the archaeal-stem dates?