Asgard archaea shed light on the evolutionary origins of the eukaryotic ubiquitin-ESCRT machinery

Abstract

The ESCRT machinery, comprising of multiple proteins and subcomplexes, is crucial for membrane remodelling in eukaryotic cells, in processes that include ubiquitin-mediated multivesicular body formation, membrane repair, cytokinetic abscission, and virus exit from host cells. This ESCRT system appears to have simpler, ancient origins, since many archaeal species possess homologues of ESCRT-III and Vps4, the components that execute the final membrane scission reaction, where they have been shown to play roles in cytokinesis, extracellular vesicle formation and viral egress. Remarkably, metagenome assemblies of Asgard archaea, the closest known living relatives of eukaryotes, were recently shown to encode homologues of the entire cascade involved in ubiquitin-mediated membrane remodelling, including ubiquitin itself, components of the ESCRT-I and ESCRT-II subcomplexes, and ESCRT-III and Vps4. Here, we explore the phylogeny, structure, and biochemistry of Asgard homologues of the ESCRT machinery and the associated ubiquitylation system. We provide evidence for the ESCRT-I and ESCRT-II subcomplexes being involved in ubiquitin-directed recruitment of ESCRT-III, as it is in eukaryotes. Taken together, our analyses suggest a pre-eukaryotic origin for the ubiquitin-coupled ESCRT system and a likely path of ESCRT evolution via a series of gene duplication and diversification events.

Fig. 1. The genomes of Asgard archaea, Lokiarchaeota, Heimdallarchaeota, Helarchaeota, and Odinachaeota possess homologues of the ubiquitin-ESCRT pathway.

A List of proteins in the Asgard archaea and eukaryotic Ubiquitin-ESCRT (Ub-ESCRT) pathway. B Co-location of Ub/ESCRT protein-encoding genes in Heimdall- (i; 22 genomes), Hel- (ii; 9 genomes), Loki- (iii; 29 genomes) and Thorarchaeota (iv; 30 genomes). A colour gradient indicates the fraction of genomes in which a pair of genes was found to co-locate within a region of <10 kb. White cells indicate gene pairs not found co-existing in any Asgard genome analysed. C Synteny plot of selected genomes. Arrows represent genes and are coloured if their products are annotated as containing diagnostic domains for Ub/ESCRT proteins (see Methods). Homologues of Vps23/37 (as determined via alignment (Supplementary Fig. 6)) in the vicinity of the ESCRT gene cluster in Helarchaeote Hel_GB_B, Heimdallarchaeote B3_Heim and Ca. Odinarchaeum yellowstonii LCB_4 only possess E2/UEV domains. A gene encoding a fusion of Vps23/37 and Vps28 is coloured as both orange and red. Genome regions are plotted up to a distance of 2 kb from ubiquitin or ESCRT protein-encoding genes (coloured), or until a contig boundary (thicker vertical lines). Similarity lines indicate best-reciprocal BLAST-p hits with an e-value lower than 1e-5. The names of the organisms used for experimental analyses in later sections are marked in orange. D Phylogenetic reconstruction of Vps22 and Vps36. Unrooted maximum likelihood phylogenetic tree of Vps22 (blue), Vps36 (purple) and Vps25 (orange) and outgroup (black) sequences. The tree was reconstructed using IQ-Tree under the LG + C60 + R4 + F + PMSF model. Support values are only shown for the deeper branches connecting gene homologues and represent standard Felsenstein bootstrap proportions (upper left) or Transfer-Bootstrap Expectation (TBE) (lower right) values based on 100 bootstrap pseudoreplicates. The full tree with all leaf and support labels is shown in Supplementary Fig. 4. E Phylogenetic reconstruction of UEV domain-containing proteins and E2 ubiquitin-conjugating enzymes. Unrooted maximum likelihood phylogenetic tree of the UEV domain-containing proteins (gold [Asgard], red [Eukarya]) and E2 ubiquitin-conjugating proteins (grey [Asgard], black [Eukarya]) in Eukarya and Asgard archaea, respectively. The tree was reconstructed using IQ-Tree under the Q.pfam+C20 + G4 + F + PMSF model. Support values are only shown for the deeper branches, following the same pattern as in D.

Fig. 2. Assembly of Asgard ESCRT-I complexes with ubiquitin binding to the UEV domain of Vps23(TSG101).

A Schematic diagram of the domain structure of HeimAB125 Vps28 and the truncation design used in the experiments. The N-terminal ubiquitin E2 variant domain (“UEV”), the steadiness box (SB), and the core domain of Vps28 (“Vps28”) identified previously are highlighted. B A model of the three-dimensional structure of HeimAB125 UEV. The three-dimensional structure model was created using AlphaFold 2 by templating the structure of the budding yeast Vps23 bound to ubiquitin. Vps23 (PDB: 1UZX [https://www.rcsb.org/structure/1UZX], chain A, light green) and HeimAB125 UEV (model structure, purple) are superimposed. Ubiquitin bound to Vps23 (PDB: 1UZX [https://www.rcsb.org/structure/1UZX], chain B) is shown in yellow. C The structural model of HeimAB125 ubiquitin generated using AlphaFold 2 was superimposed with the structure of ubiquitin in complex with Vps23 (PDB: 1UZX [https://www.rcsb.org/structure/1UZX], chain B). Ubiquitin (PDB: 1UZX [https://www.rcsb.org/structure/1UZX], chain B, yellow), model structure of HeimAB125 ubiquitin (light blue), Vps23 (PDB: 1UZX [https://www.rcsb.org/structure/1UZX], chain A, light green) were shown. An amino acid residue (Val45) on the model structure located in the ubiquitin hydrophobic patch, which is important for ubiquitin-UEV interactions, is highlighted in magenta. The structure of HeimAB125 ubiquitin is illustrated in the ribbon diagram (i) and the surface model (ii). D HeimAB125 UEV binds ubiquitin dependent on a hydrophobic patch. The interaction between HeimAB125 ubiquitin (wild-type or V45D mutant) and UEV-Vps28 (full-length, top panel) or UEV domain (bottom panel) was tested by BS3-mediated chemical crosslinking, followed by SDS-PAGE to detect the increase in their molecular weight. Three experimental repeats performed with representative experiment displayed. Source data are provided as a Source Data file. E Size-exclusion chromatography analysis of the thermophilic Odinarchaeota ESCRT-I subcomplex assembly. All proteins and complexes were incubated at 60 °C for 10 min before analysis. From top to bottom: Vps28 protein only (top); Odin Vps23 protein only; ubiquitin only; Vps28 pre-incubated with ubiquitin (no interaction); Odin Vps23 pre-incubated with Vps28 (stable complex formation); Odin Vps23 pre-incubated with Vps28 and ubiquitin (bottom—ubiquitin binds to the Odin Vps23 / Vps28 complex, via the UEV domain of Odin Vps23. For additional controls see Supplementary Fig. 10). All proteins were separated on a Superdex S200 HR 10/300 size-exclusion chromatography column. The relative elution volumes of the size standards β-amylase (200 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (BSA) (66 kDa) and carbonic anhydrase (29 kDa) and cytochrome-c (12.4 kDa) are also indicated in grey. Eluted fractions were resolved by SDS-PAGE and visualized by Coomassie stain. Left: chromatography UV traces (at 280 nm) for the respective elution profiles. Three experimental repeats performed with representative experiment displayed. Source data are provided as a Source Data file.

Fig. 3. Heimdall Vps22 forms stable dimers.

A Elution chromatogram of Vps22 (27.9 kDa) using a Superdex 200 16/600 size-exclusion column. The inset shows the column calibration curve established with standard proteins (see Methods section). Grey lines indicate the Ve/Vo and predicted molar mass (85 kDa) of Vps22. This assay suggests that this protein forms a trimer or an elongated dimer. Source data are provided as a Source Data file. B SEC-MALS analysis of Heimdall Vps22 using a Superdex 200 increase 10/300 analytical column. The chromatograms display the calculated molar mass of the peaks (kDa) and refractive indexes (A.U.) as dots and lines, respectively, for loaded sample concentrations of 2.0 (blue) and 0.5 (red) mg/ml. The estimated masses are 54.4 and 54.2 kDa for the two protein concentrations, indicating stable formation of a Vps22 dimer, as the theoretical dimer mass is 55.9 kDa. Source data are provided as a Source Data file. C Purified HeimAB125 Vps22 showed slower migration on SDS-PAGE gel after chemical crosslinking, whose mobility is consistent with that of a cross-linked dimer. The left panel shows Vps22 treated with or without BS3. The right panel shows Vps22 treated with or without cross-linker EDC. Note that following crosslinking, Vps22 showed a reduced mobility by SDS-PAGE with an estimated molecular weight double of that predicted for monomeric Vps22. Three experimental repeats performed with representative experiment displayed. Source data are provided as a Source Data file. D A model structure of HeimAB125 Vps22 superimposed on human Vps22 in the structure of ESCRT-II complex (3CUQ). E Vps22-Vps22 interacting regions (i) aa 41-47 (ii) aa 160-166 on the enlarged Vps22 model structure are shown. The crosslink positions on the peptide sequence are highlighted in red in the model structure. The bottom panel shows the cross-linked peptide sequences and the location of the crosslink.

Fig. 4. Crystal structure of the Odinarchaeota Vps25ΔN tandem winged helix (WH) domain.

A Asgard Odinarchaeota Vps25ΔN tandem WH domain structure coloured from blue to red (N-terminus to C-terminus) shown in ribbon form, with secondary structural sequence elements indicated. Refinement and model statistics are shown in Supplementary Table 1. B Superposition of the N- (blue) and C-terminal (wheat) Asgard Vps25 WH domains demonstrating their similarity. C Structural alignment of the Odinarchaeota Vps25ΔN crystal structure (wheat) with Vps25 from S. cerevisiae (grey) (PDB: 1XB4, chain A [https://www.rcsb.org/structure/1XB4]) and H. sapiens (salmon) (PDB: 3CUQ, chain C [https://www.rcsb.org/structure/3CUQ]). The alignment shown superimposes the N-terminal WH domains, only, and yields all-atom RMSD values against the Odin N-terminal WH domain of 0.7 Å and 1.1 Å for the yeast and human N-terminal WH domains, respectively. A similar alignment of the C-terminal WH domains yielded all-atom RMSD values of 1.5 Å and 1.1 Å for the yeast and human domains against the Odin C-terminal WH domain. Because of changes in the relative orientations of the WH domains with respect to each other in the tandem arrangement of the three Vps25 homologues, overall alignment gives much higher all-atom RMSD values of 3.5 Å and 5.1 for the yeast and human proteins against Odin, respectively.

Fig. 5. Systematic reciprocal Yeast two-hybrid assays between Asgard ESCRT proteins and insights gained from investigating ESCRT from Asgard.

A Summary of Y2H interactions. Molecules related to the Ub-ESCRT pathway found in Lokiarchaeota (Lokiarch), Heimdallarchaeota (HeimAB125), Thorarchaeota, and Ca. Odinarchaeum yellowstonii LCB_4 (OdinLCB_4) were examined comprehensively using Y2H, and the detected interactions are illustrated. See Figs. 3, 4 and S9 for further biochemical validations. B Top panel: Schematic representation of the organization of the Asgard ESCRT pathway based on this work. In the Heimdallarcheota and Lokiarchaeota ESCRT-II complexes, the Vps22 subunit forms a homodimer like the eukaryotic Vps22/Vps36 ESCRT-II heterodimeric stalk. In Odinarcheota, however, the Vps22 homologue does not appear to dimerise, and yet undetermined factor(s) therefore likely bridge the interaction between the ESCRT-I and -II subcomplexes. The Odinarcheota Vps23 ESCRT-II homologue forms a dimer, thereby presenting two ubiquitin-binding UEV domains. The Vps23 dimer interacts with a single Vps28 protein thus forming a tripartite complex, reminiscent of the eukaryotic Vps37/Vps23/Vps28 complex. In Heimdallarcheota, the Vps23 and Vps28 functions are fused in a single protein that also dimerises. Compare with the eukaryotic arrangement as shown in Supplementary Fig. 1. Bottom panel: a schematic representation of a hypothetical Asgard archaeal cell using ESCRT-III polymers to facilitate extracellular vesicle formation and potentially in virus release.