The only known function of S-adenosylmethionine decarboxylase (AdoMetDC) is to supply, with its partner aminopropyltransferase enzymes such as spermidine synthase (SpdSyn), the aminopropyl donor for polyamine biosynthesis. Polyamine spermidine is probably essential for the growth of all eukaryotes, most archaea and many bacteria. Two classes of AdoMetDC exist, the prokaryotic class 1a and 1b forms, and the eukaryotic class 2 enzyme, which is derived from an ancient fusion of two prokaryotic class 1b genes. Herein, we show that ‘eukaryotic' class 2 AdoMetDCs are found in bacteria and are enzymatically functional. However, the bacterial AdoMetDC class 2 genes are phylogenetically limited and were likely acquired from a eukaryotic source via transdomain horizontal gene transfer, consistent with the class 2 form of AdoMetDC being a eukaryotic invention. We found that some class 2 and thousands of class 1b AdoMetDC homologues are present in bacterial genomes that also encode a gene fusion of an N-terminal membrane protein of the Major Facilitator Superfamily (MFS) class of transporters and a C-terminal SpdSyn-like domain. Although these AdoMetDCs are enzymatically functional, spermidine is absent, and an entire fusion protein or its SpdSyn-like domain only, does not biochemically complement a SpdSyn deletion strain of E. coli. This suggests that the fusion protein aminopropylates a substrate other than putrescine, and has a role outside of polyamine biosynthesis. Another integral membrane protein found clustered with these genes is DUF350, which is also found in other gene clusters containing a homologue of the glutathionylspermidine synthetase family and occasionally other polyamine biosynthetic enzymes.
Cellular growth and proliferation in all eukaryotes, most archaea and many bacteria are likely contingent on the presence of the polyamine spermidine . The pyruvoyl-dependent enzyme S-adenosylmethionine decarboxylase (AdoMetDC) is a key enzyme in the biosynthesis of polyamines such as spermidine and spermine . A catalytically essential pyruvoyl cofactor is generated from an internal serine after autocatalytic self-processing of the AdoMetDC proenzyme into two subunits [3,4]. The only known function of the AdoMetDC enzyme is to provide decarboxylated AdoMet that serves as the aminopropyl donor in polyamine biosynthesis (Figure 1). In all known cases in the three domains of life, AdoMetDC must partner with one or more aminopropyltransferases such as spermidine synthase (SpdSyn), to allow transfer of the aminopropyl group of decarboxylated AdoMet to an acceptor such as putrescine (Figure 1). Methylthioadenosine is produced from decarboxylated AdoMet after the aminopropyl group has been transferred  (Figure 1). Structurally related aminopropyltransferases are involved in linear polyamine biosynthesis: SpdSyn transfers an aminopropyl group to putrescine to form spermidine [6,7], spermine synthase forms spermine from spermidine [8–10], thermospermine synthase forms thermospermine from spermidine , agmatine aminopropyltransferase forms aminopropylagmatine from agmatine , and long-chain polyamines are formed by promiscuous long chain polyamine synthases . A different structural class of aminopropyltransferase sequentially transfers the aminopropyl group of decarboxylated AdoMet to the internal N4 position of spermidine or N3 position of norspermidine to form the branched polyamines N4-bis(aminopropyl)spermidine or N3-bis(aminopropyl)norspermidine . To date, there are no known functions of aminopropyltransferases outside of polyamine biosynthesis.
The role of S-adenosylmethionine decarboxylase in spermidine biosynthesis.
Two structural classes of AdoMetDC exist: class 1a is found in bacteria, class 1b in bacteria and archaea, and class 2 in eukaryotes . The class 1a form is typified by the AdoMetDC of Escherichia coli , and is derived from the phylogenetically much more prevalent, smaller class 1b form, typified by the AdoMetDCs of bacterium Thermotoga maritima  and archaeon Methanocaldococcus jannaschii [17,18]. There is little detectable homology between the prokaryotic class 1 forms of AdoMetDC and the eukaryotic class 2 form. However, the eukaryotic AdoMetDC appears to have evolved by head to tail gene fusion of two class 1b AdoMetDC genes, with loss of the processing site of the C-terminal half [2,15,19,20].
It has been shown that SpdSyn was likely present in the Last Universal Common Ancestor (LUCA), however, AdoMetDC was not identified as part of the LUCA proteome . This may be due to phylogenetic occlusion caused by greater horizontal gene transfer of AdoMetDC or to the fact that the class 1b AdoMetDC is less than half the size of SpdSyn and therefore has a weaker phylogenetic signal. It is more likely that the class 2 AdoMetDC was present in the Last Eukaryotic Common Ancestor (LECA), since it is found throughout eukaryotic supergroups but it is unclear whether it is a eukaryotic invention. Previously, we noted that a homologue of a eukaryotic-type class 2 AdoMetDC was present in the bacterium Shewanella oneidensis . We sought to determine whether eukaryotic-type class 2 AdoMetDCs encoded in bacteria are functional. This was not obvious since class 2 AdoMetDC homologues in kinetoplastid genomes are catalytically dead but function as allosteric regulators of the paralogous functional AdoMetDC [23,24]. Furthermore, class 1b AdoMetDC homologues in Crenarchaeota are in fact arginine decarboxylases . For the current manuscript, we sought to determine whether the class 2 AdoMetDC is a eukaryotic invention or whether it arose in bacteria. Addressing these questions led us to discover a new class of probable membrane-anchored spermidine synthase homologues that do not appear to function in polyamine biosynthesis even though they colocalize in gene clusters with functional AdoMetDCs from the class 2 or class 1b, suggesting a novel nonpolyamine substrate for aminopropylation. In addition, we uncovered a link between these gene clusters and others encoding homologues of the glutathionylspermidine synthetase family.
Materials and methods
Growth of bacterial strains
Cultures of the γ-proteobacterium Shewanella oneidensis MR-1 (a kind gift from Prof. Bin Cao, Nanyang Technological University, Singapore) were initiated from frozen stocks and grown in MM1 chemically defined minimal medium  at 30°C. The β-proteobacterium Ralstonia pickettii ATCC 27511 was obtained from the American Type Culture Collection and was grown aerobically at 30°C in chemically defined M199 medium (Sigma–Aldrich, cat. no. M4530). Heterologous expression of chemically-synthesized genes was performed in E. coli BL21ΔspeD or BL21ΔspeE, and derived strains were grown aerobically in chemically defined M9 medium at 37°C.
Heterologous gene expression in E. coli BL21-derived strains
All heterologous genes were synthesized by GenScript with E. coli-optimized codons. Class 2 and 1b AdoMetDC ORFs, cloned as 5′-NcoI-3′-HindIII were expressed from pETDuet-1 (Novagen) in E. coli BL21ΔspeD. To coexpress the E. coli speD (encoding AdoMetDC) and R. pickettii 12D membrane protein-speE (SpdSyn) gene fusion (mpSpdSyn: WP_012430203) in pETDuet-1, the E. coli speD ORF was cloned into the 5′-NcoI and 3′ BamHIII sites, and the synthesized R. pickettii membrane protein-speE fusion gene ORF (mpSpdSyn) was cloned into the 5′-NdeI and 3′-XhoIII sites of pETDuet-1 to form pSK823. To express the SpdSyn domain only of the R. pickettii mpSpdSyn fusion gene, the C-terminal domain of the fusion gene starting at amino acid position A229 was used to replace the whole fusion protein in pSK823 to form pSK831. The control plasmids were pETDuet-1 with the E. coli speD alone (pSK815), or pETDuet-1 with E. coli speD and E. coli speE (pSK817), with the same cloning sites as pSK823 and pSK831.
Generation of an AdoMetDC (speD) deletion mutant in E. coli BL21 (DE3)
Construction of a speE (SpdSyn) gene deletion mutant in E. coli BL21 was described previously . To generate a gene deletion of the E. coli speD (AdoMetDC) locus, strain JW0116-1 where the speD ORF is replaced by a kanamycin resistance cassette (ΔspeD::kanFRT), was obtained from the Keio Collection . The ΔspeD::kanFRT sequence of JW0116-1 was transduced from JW0116-1 into BL21(DE3) (F−ompT gal dcm lon hsdSB (rB− mB−) λ(DE3 [lacI lacUV5-T7 gene 1 ind1 sam7 nin5]) using a cleared lysate of phage P1-infected JW0116-1 to infect BL21(DE3) followed by selection for kanamycin resistance. Successful replacement of the original BL21(DE3) AdoMetDC ORF by ΔspeD::kanFRT was assessed by PCR amplification of replicated colonies using the upstream primer 5′-catacggcagcttttgcctt-3′ and downstream primer 5′-cgttgaggattttctgcagg-3′. The original speD+ product produced a PCR product of 968 bp, whereas the ΔspeD::kanFRT replacement generated a product of 1473 bp.
Polyamine extraction, HPLC and LC–MS analysis
Polyamines were extracted from exponentially growing cells using trichloroacetic acid as previously described [29–31]. For analysis of polyamines by HPLC, polyamines were derivatized using an AccQ-Tag Ultra Derivatization kit (Waters) as previously described [29,30] and detected by fluorescence, using pure polyamine standards obtained from Sigma–Aldrich. Analysis of polyamines from E. coli by LC–MS was performed on benzoylated cell extracts as described [27,32], employing an Agilent 1290 Infinity HPLC system using an Eclipse XDB-C18 column (4.6 × 150 mm, 5 µm; Agilent) that was coupled to an Agilent 6130 quadrupole ESI mass spectrometer run in the positive mode with a scan range of 100 to 1100 m/z. Liquid chromatography was carried out at a flow rate of 0.5 ml/min at 20 °C with a 5 µl injection volume, using a gradient elution with aqueous acetonitrile containing 0.1% formic acid.
Class 2 AdoMetDC sequences in bacteria were detected using BLASTP searches against the nonredundant protein database at the National Center for Biotechnology Information (NCBI), GenBank™, employing class 2 AdoMetDC sequences from Saccharomyces cerevisiae and Arabidopsis thaliana. Identification of bacterial class 2 sequences was based on a combination of cut-off E value (2e × 10−4), ORF completeness and presence of conserved motifs. Sequence alignments were performed with ClustalX , after sequences from the N- and C-termini were trimmed to facilitate the alignment. Aligned sequences were edited and formatted for display with GeneDoc . A Neighbour Joining phylogenetic tree was constructed using PAUP* , node support was determined using 1000 bootstrap replicates, and the tree was imported into TreeView . All taxonomic descriptions were obtained from the NCBI. Hidden Markov Models (TMHMM) were used to predict transmembrane sequences . Other protein domains were identified using InterPro .
Eukaryotic-type AdoMetDCs are present in bacteria
The prokaryotic forms of AdoMetDC, class 1a and Ib had been found previously to possess regions of homology, based on comparison of the E. coli class 1a and Thermotoga maritima class 1b AdoMetDC amino acid sequences . By aligning more sequences from phylogenetically diverse species, it is clear that class 1a evolved from the class 1b protein through two internal peptide insertions and a short N-terminal and long C-terminal extension (Supplementary Figure S1). As mentioned above, the eukaryotic class 2 AdoMetDC was formed from an ancient gene fusion of two prokaryotic class 1b AdoMetDCs. Although the class 2 AdoMetDC is regarded as a specifically eukaryotic form of the enzyme, we had previously noted that a class 2 AdoMetDC homologue was encoded in the genome of the γ-proteobacterium Shewanella oneidensis . With the current availability of many more sequenced bacterial genomes since our original observation, we now detect by BLASTP 257 bacterial genomes encoding class 2 AdoMetDC homologues. Of these, 237 are encoded by Proteobacteria, with the remaining sequences found in metagenome-assembled genomes from Elusimicrobia, Acidobacteria, Planctomycetes and Chloroflexi. Within the Proteobacteria, 192 homologues are encoded by γ-Proteobacteria, 24 by Oligoflexia and 18 by δ-Proteobacteria. Of the 192 γ-proteobacterial homologues, 176 are found in the Alteromonadales order. Using BLASTP to search genomes of the Alteromonadales family Shewanellaceae, only one genome was found to encode a class 1b AdoMetDC, none encoded class 1a but 111 encoded complete or partial class 2 AdoMetDCs. The bacterial class 2 AdoMetDCs exhibit sequence similarity to eukaryotic homologues across the entire protein (Supplementary Figure S2). A Neighbour Joining tree of class 2 AdoMetDC homologues indicates that the bacterial sequences form a clade within the eukaryotic tree, suggesting that the bacterial genes were likely derived from a single eukaryotic source (Figure 2). No class 2 AdoMetDC homologues are found encoded in the archaea or α-Proteobacteria, which are thought to be the progenitors of eukaryotes. The class 2 AdoMetDC is found in all eukaryotic supergroups, indicating that this form of AdoMetDC is a eukaryotic invention.
Neighbour Joining Tree of class 2 S-adenosylmethionine decarboxylases (AdoMetDCs) from eukaryotes and bacteria.
Bacterial eukaryotic-type AdoMetDCs are functional
To determine whether bacterial class 2 AdoMetDC homologues possess AdoMetDC enzymatic activity, we expressed several homologues, three from the γ class and one from the Oligoflexia class of the Proteobacteria (indicated in red text in Figure 2) in an AdoMetDC (speD) gene deletion of E. coli BL21 (BL21ΔspeD). This strain is completely deficient in spermidine when grown in chemically defined medium. As a positive control, we expressed the class 1b AdoMetDC-encoding speD gene from Bacillus subtilis in E. coli BL21ΔspeD. Polyamines were extracted from the supernatants of cell extracts expressing the AdoMetDC genes, subjected to benzoylation, and spermidine was detected by LC–MS using the Extracted Ion Chromatogram (EIC) for the mass of tribenzoylated spermidine (457.7 : 458.8). Spermidine was absent from E. coli BL21ΔspeD containing only the empty pETDuet-1 vector (Figure 3). Heterologous expression of the class 2 AdoMetDCs from Shewanella algae, S. oneidensis, S. denitrificans and Bacteriovorax sp. BSW11-IV, and the class 1b AdoMetDC from B. subtilis resulted in spermidine production that was confirmed by MS analysis of the corresponding EIC peak (Figure 3). Production of spermidine confirms that the bacterial AdoMetDCs have complemented the loss of the endogenous AdoMetDC in the E. coli BL21ΔspeD strain.
LC–MS analysis of heterologous spermidine production by diverse bacterial class 2 and class 1b AdoMetDC (speD) genes.
Comparison of class 2 (eukaryotic-type) AdoMetDC-encoding bacterial genomes reveals non-canonical role
Many bacterial genomes contain gene clusters of functionally related genes that may or may not be co-regulated . In some Shewanella species such as S. algae, the class 2 AdoMetDC (speD) gene is clustered with genes encoding other polyamine biosynthetic enzymes: PLP-dependent arginine decarboxylase, agmatine ureohydrolase and SpdSyn, constituting a complete spermidine biosynthetic pathway (Figure 4). The S. algae polyamine biosynthetic gene cluster also contains a pdxH gene encoding pyridoxamine 5′-phosphate oxidase catalyzing the terminal step in PLP production, which in principle could affect arginine decarboxylase activity. However, the majority of Shewanella species do not encode a canonical SpdSyn, instead, a SpdSyn homologue is found in an unlinked gene cluster elsewhere in the genome. It is fused at its N-terminus to an integral membrane protein of the Major Facilitator Superfamily (MFS) group of transport proteins, encompassing ∼200 amino acids, producing a fusion protein of ∼500 amino acids. The membrane protein-SpdSyn (mpSpdSyn) fusion protein is clustered with genes encoding other membrane-associated proteins such as the small integral membrane protein DUF350 (Figure 4). To assess whether the mpSpdSyn fusion protein possesses spermidine synthase activity, we grew S. oneidensis to mid-log phase in polyamine-free, chemically defined medium and determined the polyamine content of the S. oneidensis cells by HPLC (Figure 5A). We found that S. oneidensis lacks spermidine but contains putrescine. This is concordant with the findings of Hamana  that other Shewanella species lack any spermidine, and suggests that the functional class 2 AdoMetDC is not involved in polyamine production, and that if mpSpdSyn is a functional aminopropyltransferase, it is not aminopropylating putrescine.
Gene clusters from Shewanella algae and S. oneidensis containing class 2 AdoMetDCs.
HPLC analysis of polyamine content in (A) Shewanella oneidensis and (B) Ralstonia pickettii cells.
Non-canonical role for class 1b (bacterial-type) AdoMetDC
The protein sequence of the S. oneidensis mpSpdSyn fusion protein was used to search for other homologues in bacteria. As of 09 May 2019, there are over 7000 bacterial genomes encoding such homologues (COG4262). We selected seven genomes encoding gene clusters containing an mpSpdSyn fusion protein gene and a typical class 1b AdoMetDC (Figure 6), and which did not encode any typical SpdSyn or any other aminopropyltransferase, or another AdoMetDC. As with the class 2 AdoMetDCs, we determined whether the bacterial class 1b AdoMetDCs, from genomes encoding the mpSpSyn fusion protein were enzymatically functional by expressing each gene in spermidine-deficient E. coli BL21ΔspeD, and verifying whether each gene could complement the spermidine deficiency. Tribenzoylated spermidine (m/z 458.2, Na+ adduct 480.2) was again detected by LC–MS, and the analysis indicated that spermidine was produced with all tested class 1b AdoMetDC-encoding speD genes derived from: Ralstonia pickettii, R. solanacearum GMI1000 (β-proteobacteria), Paraglaciecola psychrophila (γ-proteobacteria), Magnetococcus marinus (α-proteobacteria), Gemmatimonas aurantica (Gemmatimonadetes), and Clostridium hiranonis and Oceanibacillus iheyensis (Firmicutes) (Figures 3 and 7). Although R. solanacearum encodes a functional AdoMetDC, we had previously shown that this species accumulates putrescine and 2-hydroxyputrescine but spermidine is absent . Another Ralstonia species, R. pickettii also encodes a functional AdoMetDC (Figure 7), and is clustered with an mpSpdSyn fusion protein gene (Figure 6). We grew R. pickettii in chemically defined polyamine-free medium to mid-log phase, and by HPLC detected putrescine but spermidine was not detected (Figure 5B). The growth culture supernatant was also analyzed and only a small amount of putrescine but no spermidine was detected (Supplementary Figure S3).
Gene clusters containing a class 1b AdoMetDC and membrane protein-SpdSyn fusion protein.
LC–MS analysis of heterologous spermidine production in E. coli by diverse bacterial class 1b S-adenosylmethionine decarboxylase (speD) genes.
It is formally possible that the mpSpdSyn fusion protein is expressed only under very specific conditions, possibly explaining why we did not detect spermidine in either S. oneidensis, R. solanacearum or R. pickettii cultures. The whole R. pickettii mpSpdSyn fusion protein gene or the SpdSyn-like domain only were expressed from pETDuet-1 in a SpdSyn (speE) gene deletion strain of E. coli (BL21ΔspeE). To avoid any polar effects of the speE deletion, which is immediately upstream of the speD gene on the E. coli chromosome, the native E. coli speD gene was coexpressed with the mpSpdSyn gene in pETDuet-1. Cell extracts were analyzed by HPLC (Figure 8). Unsurprisingly, the BL21ΔspeE strain contained putrescine but no spermidine (Figure 8A), and the BL21ΔspeE strain complemented with the native E. coli speE gene expressed from pETDuet-1 contained a substantial spermidine peak (Figure 8B). Neither strain expressing either the full length mpSpdSyn fusion protein (Figure 8C) or the stand-alone C-terminal SpdSyn-like domain (Figure 8D) contained any spermidine. It is possible that the genes were not expressed, or that the encoded proteins misfolded. Nevertheless, the lack of spermidine production by the stand-alone aminopropyltransferase domain is consistent with the lack of spermidine in the native R. pickettii cells, and would suggest that the mpSpdSyn fusion protein, if a functional aminopropyltransferase, is not aminopropylating putrescine to form spermidine.
The SpdSyn-like domain of the Ralstonia pickettii mpSpdSyn fusion protein does not complement an E. coli SpdSyn gene deletion mutant.
We aligned the C-terminal SpdSyn-like domains from the mpSpdSyn fusion proteins of R. pickettii and S. oneidensis with the bona fide SpdSyn proteins from the human and hyperthermophilic bacterium Thermotoga maritima, for which X-ray crystal structures with ligands have been solved [6,7] (Supplementary Figure S4). By reference to the human SpdSyn amino acid positions, the presence of an aspartate at D104 in both the R. pickettii and S.oneidensis fusion proteins indicates that decarboxylated AdoMet rather than AdoMet is the substrate. Almost all of the main amino acids for binding decarboxylated AdoMet and putrescine are conserved in each sequence (Supplementary Figure S4). They include D104, D173 and Q80 that form a negatively charged binding pocket for binding decarboxylated AdoMet. Interaction with the ribose hydroxyls is achieved by E124 and Q49, although an aspartate in present at the E124 position of the R. pickettii and S. oneidensis fusion proteins. The 6-NH2 group of adenosine is bound by D155, conserved in human and the fusion proteins, and a putrescine binding pocket consists of D173, S174, Y79 and Y214, although the tyrosine at Y214 is replaced by a phenylalanine in the fusion proteins. Residues 171–180 form the gate-keeping loop, which is not particularly conserved between the human and T. maritima proteins. Potential equivalent tryptophan positions corresponding to human W28 that closes off the substrate-binding pocket, preventing binding of longer polyamine substrates are also present near the W28 position in the fusion proteins. There are three insertions in the S. oneidensis sequence but only one in the R. pickettii sequence, and one two-amino acid deletion. There is no clear indication from the sequence alignment why the fusion proteins do not aminopropylate putrescine.
An analysis was made of the mpSpdSyn fusion protein amino acid sequences for the presence of transmembrane domains using a TMHMM, from species we had shown to possess a functional AdoMetDC. The analysis revealed that each of the six mpSpdSyn fusion proteins possesses seven transmembane domains in the N-terminal domain, and the C-terminal aminopropyltransferase domain is predicted to be extracellular (Supplementary Figure S5).
Commonly occurring genes in the AdoMetDC/mpSpdSyn gene clusters
In bacterial gene clusters that contain the class 1b AdoMetDC and mpSpdSyn fusion protein genes, the most commonly co-occurring gene encodes DUF350, a small integral membrane protein containing two transmembrane segments (Figure 6). Genes encoding DUF350 domains may contain one, two or three DUF350 domains that encode two, four or six transmembrane segments, and these open reading frames encode ∼70, 140 and 280 amino acids. Another frequently co-localized gene (DUF4178) possesses an N-terminal signal peptide, and some variants possess a second DUF4178 domain and two transmembrane segments in the C-terminus. Some clusters encode a flotillin-like homologue, a putrescine oxidase-like homologue from the FAD/NAD(P)-binding oxidoreductase family that contains a twin-arginine translocation pathway signal sequence, and a phage shock protein A-like homologue (Figure 6).
DUF350 is a nexus to other branches of polyamine metabolism
The small integral membrane protein DUF350 is also found in gene clusters containing what is annotated as a glutathionylspermidine (Gsp) synthetase without an amidase domain. The E. coli Gsp synthetase/amidase is an ATP-dependent bifunctional enzyme producing Gsp from glutathione and spermidine, and hydrolyzing the same product through its amidase domain. There are two other Gsp synthetase paralogues in E. coli without an amidase domain, YgiC and YjfC, and neither produces Gsp but they do hydrolyze ATP . The bacterial gene clusters, although annotated as containing Gsp synthetase, in fact encode proteins homologous to YgiC or YjfC across their entire lengths, and the ygiC-containing clusters also include genes related to other branches of polyamine metabolism (Figure 9). Some gene clusters encode DUF350, a fusion protein of a class 1b AdoMetDC and SpdSyn or a fusion protein of ornithine decarboxylase and SpdSyn, together with a homologue of YgiC. Other clusters encode DUF350, homospermidine synthase and YgiC, indicating that these clusters may ligate spermidine or homospermidine to an amino acid or small peptide but not glutathione. A number of bacterial genomes encode clusters containing DUF350, YgiC and some or all of DUF1190, DUF2491, and DUF2170, and being Domains of Unknown Function, these proteins do not have a known function. Some clusters encode DUF350 and YjfC homologues (Bacillus genomes) and a fusion of two YjfC genes is found in Clostridium botulinum. The N-terminal YjfC domain in this fusion protein is not detected by BLASTP but is identified by InterProScan. In addition to DUF350, another common feature of some of the YgiC gene clusters and AdoMetDC/mpSpdSyn clusters is the presence of DUF2170 immediately upstream of phage shock protein A, eg., in the Caulobacter crescentus YgiC cluster in Figure 9 and the S. oneidensis AdoMetDC/mpSpdSyn cluster in Figure 4.
Gene clusters containing DUF350 and glutathionylspermidine synthetase-family homologues YgiC or YjfC.
There are two prominent forms of AdoMetDC in prokaryotes, constituting class 1a and 1b. The class 1a form typified by the E. coli AdoMetDC has evolved from the smaller class 1b enzyme, found widely distributed in bacteria and archaea, through two internal insertions and a short N-terminal and long C-terminal extension. It has been presumed that the class 2 enzyme was specific to eukaryotes . Structural analysis revealed that the class 2 AdoMetDC is composed of two fused class 1b proteins, with the C-terminal copy having lost the ability to self-process and produce a pyruvoyl cofactor . Often, after gene duplication, one of the copies develops a new function, a process known as neofunctionalization . Sometimes two copies of a gene become mutually dependent for function, a process known as subfunctionalization. The class 2 AdoMetDC, consisting of a fusion of two class 1b genes, is an unusual phenomenon because there has been some neofunctionalization in the C-terminal copy as it no longer undergoes catalytic self-processing but it is still essential for AdoMetDC activity, which is reminiscent of subfunctionalization. It is not clear whether the original two copies of the class 1b AdoMetDC that formed the class 2 enzyme came from gene duplication or acquisition of a second copy by horizontal gene transfer. In our study, we have shown that functional class 2 AdoMetDCs are found in bacteria, albeit in a phylogenetically limited group, and importantly, that group of sequences is found within the eukaryotic tree of AdoMetDCs. This indicates that the bacterial class 2 genes were acquired from eukaryotes, most likely from a single eukaryotic source. The class 2 gene then likely disseminated in bacteria by vertical descent and some horizontal gene transfer. Horizontal acquisition of a eukaryotic gene in bacteria would be initially problematic for regulated expression due to the lack of bacterial promoter and other regulatory signals. However, it has been shown recently that eukaryotic genes in a bacterial genome can be rapidly activated for function by promoter capture .
The last universal common ancestor of all bacteria, archaea and eukaryotes probably encoded SpdSyn . Yet, the phylogenetic analysis used to trace SpdSyn back to LUCA did not find convincing phylogenetic signals for the presence of AdoMetDC, the obligate partner enzyme of SpdSyn. It is possible that AdoMetDC might be more susceptible to horizontal gene transfer, which would have confounded the phylogenetic signal. Or, it may be that the class 1b AdoMetDC, which must have been the original form of this enzyme since both the class 1a and 2 forms are derived from it, being less than half the size of SpdSyn, has a much weaker phylogenetic signal that may obscure its detection in LUCA. Despite the fact that the class 2 AdoMetDC is derived from two bacterial class 1b genes, it is likely that the class 2 AdoMetDC is a eukaryotic invention. It is found in all major eukaryotic supergroups, and horizontal gene transfer between eukaryotes is relatively rare , suggesting that the class 2 form was present in LECA. Recent evolutionary thought proposes that eukaryotes arose from the merger of an archaeal host with an α-proteobacterial endosymbiont that became the mitochondrion . The archaeal host may have resembled extant members of the Asgard superphylum . However, the class 2 AdoMetDC is not found in any archaea or α-proteobacteria, so it may have emerged after the archaeal/α-proteobacterial merger but before LECA.
It has been generally assumed that the sole function of AdoMetDC and its partner aminopropyltransferases is to produce polyamines. Our discovery of functional class 1b and class 2 AdoMetDCs in bacteria that do not produce spermidine suggests that many, possibly thousands of bacterial strains employ decarboxylated AdoMet with a non-canonical aminopropyltransferase to aminopropylate novel, unknown substrates. AdoMetDC is often co-localized with a SpdSyn homologue containing an MFS-type transporter domain at the N-terminus (the mpSpdSyn protein), although in many bacteria they are not necessarily physically clustered. In those bacteria containing mpSpdSyn homologues, it is usually the only aminopropyltransferase encoded in the genome. The AdoMetDCs encoded in those genomes appear to be typical AdoMetDCs, with the proviso that some of those AdoMetDCs in bacteria are the eukaryotic-type class 2 form. All the AdoMetDCs that we tested were functional, whether class 1b or class 2, so they have not evolved into a non-enzymatic regulatory role such as the AdoMetDC homologue ‘prozyme’ in kinetoplastidia , or switched substrate such as the AdoMetDC-like arginine decarboxylases found in Crenarchaeota . Sequence-based analysis of mpSpdSyn proteins using a TMHMM suggests that the aminopropyltransferase domain is outside of the cellular membrane, which needs to be experimentally confirmed.
The frequently co-occurring genes encode proteins with either integral membrane domains (DUF350), transmembrane spans (DUF4178), or that are associated with membrane microdomains (band 7/flotillin), or associated with membrane stress (PspA), or are exported as folded proteins through the plasma membrane using the twin-arginine pathway for export to the periplasm (putrescine oxidase-like). The predicted nature of these proteins suggests that they may form a co-localized protein complex in/at the membrane. Associated AdoMetDCs, whether class 1b or class 2 do not contain any obvious membrane localization signal. The aminopropyltransferase domain of the mpSpdSyn proteins contains almost all the conserved, critical amino acids required for spermidine synthase activity. If the mpSpdSyn protein is a functional aminopropyltransferase and located outside of the cellular membrane, decarboxylated AdoMet would need to be exported out of the cell to provide the aminopropyl groups needed for aminopropylation. It is intriguing therefore that the N-terminal domain of the mpSpdSyn fusion protein is homologous to an MFS transporter.
The Gsp synthetase of E. coli is part of a bifunctional synthetase/amidase fusion protein that can synthesize Gsp from glutathione and spermidine, and also hydrolyze the same product . There are two paralogues of the Gsp synthetase encoded in the E. coli genome that lack the amidase domain, YgiC and YjfC, and these two proteins are much more similar to themselves than to Gsp synthetase . Gsp synthetase belongs to the ATP-grasp structural group and uses ATP to activate a carboxylate group. Although YgiC and YjfC bind ATP and hydrolyze it, they do not synthesize Gsp and neither spermidine nor glutathione increase the rate of ATP hydrolysis . The small integral membrane protein DUF350 found in the AdoMetDC/mpSpdSyn gene clusters is also found in completely different gene clusters containing a gene annotated as a Gsp synthetase but which is actually more homologous to ygiC or yjfC. Furthermore, diverse bacteria contain polyamine biosynthetic genes in the ygiC-containing clusters. The presence of the polyamine biosynthetic genes suggests that the YgiC homologues may conjugate either spermidine or homospermidine to an amino acid or peptide but not glutathione. As homospermidine synthase, gene fusions of AdoMetDC and SpdSyn, or of ornithine decarboxylase and SpdSyn are found together with YgiC, this indicates that there is selective pressure for polyamine biosynthetic genes to be physically clustered with ygiC, with likely functional significance. YgiC and YjfC homologues do not contain any signal peptide or transmembrane domains, so it maybe that they interact with the cytoplasmic segments of DUF350. Genomic evidence suggests that DUF350 has evolved to be associated with a non-canonical polyamine or polyamine biosynthesis-derived metabolism.
domain of unknown function
Extracted Ion Chromatogram
last eukaryotic common ancestor
last universal common ancestor
major facilitator superfamily
membrane protein-spermidine synthase-like
Hidden Markov Model
Biochemical, molecular biological, analytical chemistry, and microbiological experimentation and data acquisition: B.L., S.H.K., S.K.; Mass Spectrometry: J.L.; Study concept, bioinformatics and general ruminations: A.J.M.; writing of manuscript: A.J.M.
A.J.M. is supported by UT Southwestern Medical Center.
The Authors declare that there are no competing interests associated with the manuscript.
Current address: Division of Applied Life Science, Graduate School of Bioresources and Environmental Sciences, Ishikawa Prefectural University, Nonichi, Ishikawa, 921-8836, Japan
Current address: Realtox Labs, 200 Business Center Dr., Reistertown, MD 21136, U.S.A.