Sua5 and TsaC sequences were extracted from UniProt (version 06/2021 or version 11/2022) and UniRef (version 11/2022). TsaC/Sua5 sequences in DPANN archaea were retrieved using BLASTp searches with Pa-Sua5 protein sequence as query against all genome assemblies available (nr database, version 11/2022). The hits were filtered for the presence of KxR/SxN tetrade and one sequence per species was chosen to reduce redundancy. Following this pipeline, 5,889 bacterial, 761 archaeal and 1,318 eukaryotic sequences were recovered. The identified Sua5 or TsaC orthologs were mapped on the RNA polymerase-based universal phylogenetic tree (Da Cunha et al., 2017) or on the previously established phylogeny of DPANN archaea (Moody et al., 2022). The organisms were ranked according to the NCBI taxonomy database (as of 11/2022) or, for DPANN organisms, based on (Moody et al., 2022). The accession numbers for all sequences are available at figshare repository (doi: 10.6084/m9.figshare.22283929).

For computing the universal phylogenetic tree we chose a set of species that has been previously vetted for its ability to produce a robust phylogenetic signal using universal proteins (Da Cunha et al., 2017). The corresponding Sua5/TsaC sequences (Supplementary Tables 2, 3) were extracted from Uniprot database, aligned using MAFFT v7 with auto settings (Yamada et al., 2016). The alignment was trimmed using BMGE (Criscuolo and Gribaldo, 2010) with BLOSUM30 matrix leaving 153 positions for tree construction, corresponding only to the catalytic domain residues (the SUA5 domain was removed by the trimming software). In addition, we generated an alignment of full length Sua5 sequences that contained 470 positions. The maximum likelihood (ML) trees were inferred using IQ-TREE v1.4.3 (Trifinopoulos et al., 2016) with the TESTNEW option for model selection. The branch support was obtained using nonparametric bootstrap (100 replicates), SH-aLRT test and ultrafast bootstrap approximation (1,000 replicates; Minh et al., 2013; Kalyaanamoorthy et al., 2017) or booster (Lemoine et al., 2018).

Archaeoglobi Sua5 sequences were extracted from Uniprot (doi: 10.6084/m9.figshare.22283929) and the Maximum Likelihood tree was inferred using Phylogeny.fr web page (Dereeper et al., 2008).

To screen for putative horizontally transferred genes in Archaea we extracted all sequences annotated as TsaC/Sua5 from Uniprot database (9,553 sequences, database release 2019_06). We filtered this dataset to remove sequences containing less than 100 amino acids or more than 400 amino acids as well as those without the KRSN tetrade. The remaining set of 1,275 archaeal TsaC/Sua5 sequences were classified according to their taxonomic affiliation to 13 different archaeal taxa (doi: 10.6084/m9.figshare.22283929). This allowed to identify candidate proteins in “mixed” groups containing both Sua5 and TsaC users. To reduce the impact of convergent evolution on small proteins, we focused our analyses on transfer of sua5 genes. We used BLASTp searches against the nr protein database to identify the most similar sequences to the candidate query protein. If the best-hit species was only distantly related to the query species, the query protein was selected as candidate for HGT. The HGT candidate were subjected to phylogenetic analysis to determine their origin. The phylogenetic inferences were done as indicated above (phylogenetic analysis section).

To identify Sua5-or TsaC-specific residues we aligned four representative sequences for TsaC and for TsaC-like domain from each domain of life (doi: 10.6084/m9.figshare.22283929) using MAFFT v7 with auto settings (Yamada et al., 2016).

To examine the distribution of the signature residue Pro¹⁴³/Thr¹³⁸ we retrieved 26,036 TsaC and Sua5 sequences from the Complete Genome Data Bank of the NCBI (database release 2015_05) using BLASTN, with Sua5 from Pyrococcus abyssi or TsaC from Escherichia coli as query. Using hmmsearch (Finn et al., 2011) we further retrieved 1,088 eukaryotic TsaC and Sua5 sequences from the Uniprot-Proteomes database (database release 2018_11). The hit sequences missing the KRSN catalytic tetrad or partial sequences were eliminated from the analysis. One UniRef 90 sequence per genus was kept for sequence alignment, 654 for prokaryotes and 564 for eukaryotes in total (accession numbers were deposited to doi: 10.6084/m9.figshare.22283929). Sequence alignment depiction was done using ESPript3 with default parameters (Robert and Gouet, 2014).

AlphaFold2 (Jumper et al., 2021) batch colab¹ was used to predict de novo structures of atypical Sua5 proteins from Archaeoglobus archaea: Archaeoglobus profundus (Ap-Sua5, UniProt D2RFV3), A. veneficus (Av-Sua5, UniProt F2KMZ1) and A. fulgidus (Af-Sua5, UniProt O29477). Crystal structures of several TsaC and Sua5 proteins were used for structural comparison. The PDB accession codes of those structures are 1HRU and 2MX1 for Ec-TsaC, 2EQA, 3AJE and 4E1B for St-Sua5, 6F87 for Pa-Sua5, 3VEZ for St-TobZ and 3TTC for Ec-HypF. Structures were visualized and edited with UCSF ChimeraX v. 1.4 (Pettersen et al., 2021).

Previous work established that TsaC/Sua5 family of proteins (COG0009) is ubiquitous however the specific distribution of TsaC or Sua5 orthologs was only examined for a small number of representative isolated species.

Here, we exploited the recent massive increase of (meta)genome sequencing data to perform an up to date comprehensive analysis of the distribution of Sua5 and TsaC proteins among living organisms. To this end, we retrieved 5,889 bacterial, 761 archaeal and 1,318 eukaryotic sequences from public databases and mapped their presence on reference phylogenetic trees (Figure 2; Supplementary Table 1; Supplementary Figure 1).

The data confirm the ubiquitous distribution of the Sua5/TsaC family among organisms. We found that the obligate ectosymbiotic archaea Nanoarchaeum equitans, and Nanobdella aerobiophila, were the only cultured organisms that lack the genes encoding TsaC/Sua5 proteins. In addition, we failed to detect Sua5/TsaC proteins in Undinarchaeota and Huberarchaea lineages from DPANN archaeal superphylum (Supplementary Figure 1) but it is unclear whether they are genuinely missing or were missed during metagenomic assemblies. Notably, in the entire tree we detected few species encoding both TsaC and Sua5: in the Leotiomyceta fungi clade (e.g., Aspergillus bombycis–Uniprot ID A0A1F7ZM40 and A0A1F8A8Q8) and in the Vibrio bacteria (e.g., Vibrio aquaticus – Uniprot A0A3S0QCR7 and A0A3S0V185). The data further show that out of 41 examined taxonomic groups depicted in Figure 2; Supplementary Figure 1, 10 groups contain species that are exclusively Sua5 users, 8 groups contain species that encode the TsaC orthologs only, 3 groups contain at least 98% of species that encode only the TsaC orthologs and the remaining 20 taxons contain species encoding Sua5 or TsaC. Intriguingly, however, the distribution of TsaC and Sua5 users across the tree shows no clear phyletic pattern suggesting a complex evolutionary history for the Sua5/TsaC family of proteins.

To gain insight into the evolutionary history of Sua5/TsaC family we constructed a phylogenetic tree using balanced taxonomic sampling across the three domains of life (Supplementary Tables 2, 3). The tree was inferred from the alignment of TsaC proteins and the catalytic TsaC-like domain of Sua5 proteins. It exhibits bipartite topology whereby TsaC proteins and TsaC-like domains robustly (bootstrap values 100%) segregated into two monophyletic clades (Figure 3), each clade containing sequences from the three domains of life. The branches in the TsaC part of the tree are in general longer suggesting that these proteins evolved at a higher rate. While in the Sua5 part of the tree no clear phyletic pattern can be observed, the TsaC tree shows that eukaryotic sequences cluster together and seem to be more related to archaeal TsaC sequences thus recapitulating the established phylogeny of organisms. Similarly, the tree built from the alignment of full-length Sua5 sequences (Supplementary Figure 2) revealed a clear phyletic pattern whereby the bacterial and archaeal sequences formed two distinct groups while the eukaryotic sequences were split into two clades. The bipartite tree topology suggests that Sua5 and TsaC emerged from an ancient duplication event that occurred before LUCA, followed by either SUA5 domain acquisition or loss to yield the two variants.

The patchy distribution of Sua5 and TsaC proteins could, at least in part, be explained by horizontal gene transfers (HGT) of sua5 and tsaC genes. To identify such events, we performed an initial screen using BLASTp searches whereby we focused on “mixed” taxonomic groups containing Sua5 users and TsaC users. For example, within Chlorophyta phylum (eukaryotes, green algae, E6 in Figure 2), the Chlorophyceae family (e.g., Volvox carteri f. nagariensis – Uniprot D8TKZ3) is the only one carrying Sua5 orthologs (Supplementary Table 1). This protein is most similar to the Sua5 from Apophysomyces ossiformis (Uniprot A0A162WET7, 57% sequence identity) which is a Mucoromycota fungus. Similarly, Rhodophyta (E3), are Sua5 users except for one species (Rhodosorus marinus – Uniprot A0A7S0G1B8) that encodes TsaC. This protein is most similar (49% identity) to the TsaC from Firmicutes bacteria (Brochothrix campestris). Leotiomyceta fungi such as Aspergillus are a peculiar case because several encode both Sua5 and TsaC proteins with a complete KxR/SxN motif suggesting that both orthologs are functional. However, the latter protein shows highest sequence similarity with Actinobacteria (Amycolatopsis bartoniae, 47.5% identity) and Proteobacteria (Verticiella sediminum, 41.9% identity) suggesting that this may be a case of inter-domain HGT. This initial screen identified several other candidate proteins in Bacteria and Archaea. Among those, we observed that all Methanothermococcus species encoded TsaC ortholog except M. thermolithotrophicus which carries a Sua5 ortholog (WP_018153339.1). The sequence of this ortholog is identical to Sua5 from Methanococcus maripaludis KA1 and the gene is encoded on a ~ 14 kbp fragment that is 99% identical to M. maripaludis KA1 genome (Supplementary Figure 3). This suggests a recent acquisition of the whole fragment by horizontal gene transfer.

We next tested the robustness of the BLASTp results using phylogenetic analysis. As an example, we selected outlier Sua5 proteins from groups which are predominantly TsaC-users (the Methanococci archaea (A7), Thaumarchaea (A1) and Thermotogae bacteria (B2); Supplementary Table 1) and aligned those with representative Sua5 sequences from Bacteria and Archaea. The tree was resolved into two distinct clades containing bacterial or archaeal sequences allowing to identify the closest homologs for the outlier Sua5 proteins (Figure 4A). The tree topology showed that (i) the Sua5 proteins from Methanococcus species and Thaumarchaeota branched within Euryarchaeota; (ii) the Sua5 protein from Methanocaldococcus archaeon branched within bacterial Sua5 orthologs; (iii) Sua5 proteins from Thermotoga bacteria branched within Crenarchaeal Sua5 orthologs. This suggested that the outlier Sua5 proteins were acquired by HGT from closely related (Methanococcus) or distantly related (Thaumarchaeota, Methanocaldococcus, and Thermotoga) organisms (Figure 4B).

Together, these analyses show that TsaC and Sua5 encoding genes have been submitted to horizontal transfers across different phylogenetic distances including at the highest taxonomic level.

The segregation of TsaC and TsaC-like sequences into two distinct phylogenetic clades indicated that each variant must contain a specific set of residues in addition to the residues conserved across the whole TsaC/Sua5 family. In search for such residues, we aligned TsaC or Sua5 sequences representative for each domain of life (Supplementary Figure 4). This showed that almost half out of approximately 200 positions in the alignment are similar, in line with a strong conservation of structure and function of the TsaC domain across the tree of life. In addition, we identified 15 conserved residues specific for Sua5 proteins (Supplementary Figure 4), 14 of which are variable among TsaC proteins. Notably, three of them, Pro⁵⁹, Asn⁶² and His⁶⁷ (Pa-Sua5 numbering) are found in the loop that forms the deep part of the catalytic cavity surface, with their side chains pointing towards the cavity (Figures 5A,C). His⁶⁷ side chain forms a H-bond (2.7 Å) with the hydroxyl function of L-threonine in the structures of St-Sua5 (Supplementary Figure 5A). In the structure of Pa-Sua5 the side chain nitrogen of Asn⁶² acts as a donor to form a H-bond (2.4 Å) with PPi (Figure 5C) while its side chain oxygen acts as an acceptor to form a H-bond with the 3’-OH function of the ribosyl moiety of TC-AMP in the structure of St-Sua5 (2.6 Å; Supplementary Figure 5A). In several TsaC variants, Asn⁶² is replaced by a lysine (Supplementary Figure 4). However, unlike asparagine, lysine can only act as a H-bond donor suggesting their respective role in TC-AMP synthesis may not be identical.

Most of the other Sua5 specific conserved residues form two motifs at the inter-domain interface: the KPSPT motif is composed of (Lys/Arg)¹⁴⁹-Pro¹⁵⁰-Ser¹⁵¹-Pro¹⁵²-Thr¹⁵³ and the GVE motif is composed of Gly¹⁷⁸-Hyd¹⁷⁹-Glu¹⁸⁰ (where Hyd corresponds to the hydrophobic residue Val, Ile or Leu; Figure 5D). The motifs interact with one another via a H-bond (2.7 Å) formed by the side chains of Ser¹⁵¹ and Glu¹⁸⁰. The backbone chain of the KPSPT motif forms several H-bonds with inter alia the side chain of His¹⁵⁷, another Sua5-specific residue, and the conserved Arg³⁰¹ present in the SUA5 domain. In addition, the Sua5-specific amino acids Asp¹⁶¹ and Glu¹⁸⁰ are forming salt bridges with Arg ³⁰¹ and Arg³²⁸, respectively (Figure 5D), thus likely playing a key role in the stabilization of the inter-domain interaction.

Remarkably, only one TsaC-specific conserved residue, Thr¹³⁸ (Ec-TsaC numbering), could be clearly identified (Supplementary Figure 4). This residue is found in the catalytic cavity of TsaC proteins and could potentially interact with substrate molecules (Figure 5B). To investigate this further we analyzed the structures of TsaC-like domains of TobZ and HypF proteins that co-crystalized with substrate molecules. The corresponding residue Thr⁵²⁹ in TsaC-like domain of TobZ protein forms a H-bond (2.8 Å) with the phosphor-moiety of the carbamoyl-phosphate a precursor for the formation of carbamoyl-adenylate (Figure 5F). This phosphor-moiety occupies the same position as the phosphor-moiety of TC-AMP and as the β phosphate of AMPPNP in the structures of St-Sua5 (Supplementary Figure 5B). In the TsaC-like domain of HypF from E. coli, the equivalent residue Thr³²¹ interacts via a H-bond (2.6 Å) with the β phosphates of the bound ADP (Supplementary Figure 5C).

Overall, the comparative sequence and structural analyses identified variant-specific conserved residues with putative key roles in substrate binding thus highlighting previously unrecognized functional differences between TsaC and Sua5 proteins.

Structural and sequence analyses identified Thr¹³⁸ (Ec-TsaC numbering) as the only conserved residues specific for TsaC proteins. Intriguingly, the corresponding residue in Sua5 proteins is a conserved Proline (Pro¹⁴³ in Pa-Sua5; Supplementary Figure 4; Figure 5C) suggesting that this single and evolutionary conserved residue is a key determinant of specific mechanisms by which Sua5 and TsaC proteins catalyze the synthesis of TC-AMP. To test the relevance of this observation, we extracted 653 prokaryotic and 566 eukaryotic sequences from complete bacterial and archaeal genomes (one sequence per genus) and plotted the nature of the signature residue as a function of the size of the protein (Figure 6; Supplementary Figure 6). This showed that Sua5 proteins have exclusively Pro¹⁴³ at this position. In TsaC sequences, Thr¹³⁸ is predominant with up to 99% of occurrences in eukaryotes but is replaced by Ser¹³⁸ in 16% of bacterial sequences. The only exception to this rule are TsaC sequences from the bacterial genera Brachyspira, Lactococcus, Erysipelothrix as well as the entire Aquificae phylum where Thr¹³⁸ is replaced, intriguingly, by a Sua5-typical proline. These TsaC proteins lack the Sua5 specific motifs important for the interdomain interaction (Supplementary Figure 7) suggesting that these are genuine exceptions.

Together, these data confirm that the highly conserved Thr¹³⁸/Ser¹³⁸ and the corresponding Pro¹⁴³ accurately identify TsaC and Sua5 proteins, respectively, in the vast majority of cases.

The identification of TsaC and Sua5 specific residues allowed us to spot atypical sequences such as those of Pro¹⁴³-containing bacterial TsaC proteins. In addition to these outliers, our attention was drawn to the Sua5 sequence from the archaeon Archaeoglobus profundus (Ap-Sua5) that carried Thr¹³⁸ instead of Pro¹⁴³ signature residue and was significantly shorter (289 AA) than a classical Sua5 protein (average size 331 AA). Extending the analysis to six other isolated Archaeoglobi species showed that, in addition to Ap-Sua5, the Sua5 proteins from A. veneficus (Av-Sua5) and A. fulgidus (Af-Sua5) carry deletions in the interdomain loop and in the SUA5 domain (Supplementary Figure 8). Around 50% of the interdomain loop residues are missing in these proteins and among those the highly conserved and functionally important motif Pro²²⁸-Gly²²⁹-Met²³⁰. Moreover, when comparing with the archetypal Sua5 from A. sulfaticallidus these three proteins lack 12%, (Av-Sua5), 13% (Af-Sua5), and 36% (Ap-Sua5) of residues in the SUA5 domain.

To get insight into the history of the acquisition of the detected mutations and their potential consequences for the function of these atypical Sua5 proteins, we performed phylogenetic analyses and AlphaFold2 modelling. The sequences that accumulated the highest number of mutations clustered within a single clade (Figure 7A). Of note, they all carry a deletion in the interdomain loop leading to the loss of the PGM motif and, concomitantly, the substitution of Asn⁶² to Lys and His⁶⁷ to Gly/Val, the residues that we have tentatively identified as being involved in interaction with substrates (Figures 5C, 7A). We next modelled the structure of Ap-Sua5 and compared it to the crystal structure of Pa-Sua5. This showed that the TsaC-like domain adopts a typical fold whereby Lys and Val residues superpose well with Asn⁶² and His⁶⁷, respectively, suggesting that these mutations affect the binding of substrates (Figure 7B). Despite a significant shortening, the SUA5 domain of Ap-Sua5 was modelled as a SUA5-like globular fold with the two arginines (Arg²⁵⁷ and Arg²⁷⁹ in Ap-Sua5) being correctly positioned to build salt bridges with the catalytic domain (Figure 7B). However, the interdomain loop seem to adopt a much more linear conformation leaving the entrance to the catalytic cavity open to the solvent (Figure 7C).

Together, the data suggest a progressive accumulation of mutations in interdomain loop and SUA5 domain of Archaeoglobi Sua5 with Ap-Sua5 being the extreme case. Consequently, these atypical Sua5 proteins seem to have acquired compensatory mutations in the catalytic domain that probably affect the binding to substrates and/or TC-AMP molecules.