Mammalian ACSF4-U26 (Acyl CoA synthetase family member 4), a protein of unknown function, comprises a putative adenylation domain (AMP-binding domain) similar to those of bacterial non-ribosomal peptide synthetases, a putative phosphopantetheine attachment site, and a C-terminal PQQDH (pyrroloquinoline quinone dehydrogenase)-related domain. Orthologues comprising these three domains are present in many eukaryotes including plants. Remarkably, the adenylation domain of plant ACSF4-U26 show greater identity with Ebony, the insect enzyme that ligates β-alanine to several amines, than with vertebrate or insect ACSF4-U26, and prediction of its specificity suggests that it activates β-alanine. In the presence of ATP, purified mouse recombinant ACSF4-U26 progressively formed a covalent bond with radiolabelled β-alanine. The bond was not formed in a point mutant lacking the phosphopantetheine attachment site. Competition experiments with various amino acids indicated that the reaction was almost specific for β-alanine, and a KM of ~ 5 μm was calculated for this reaction. The loaded enzyme was used to study the formation of a potential end product. Among the 20 standard amino acids, only cysteine stimulated unloading of the enzyme. This effect was mimicked by cysteamine and dithiothreitol, and was unaffected by absence of the PQQDH-related domain, suggesting that β-alanine transfer onto thiols is catalysed by the ACSF4-U26 adenylation domain, but is physiologically irrelevant. We conclude that ACSF4-U26 is a β-alanine-activating enzyme, and hypothesize that it is involved in a rare intracellular reaction, possibly an infrequent post-translational or post-transcriptional modification.
Microorganisms are well-known producers of a large number of chemically diverse secondary metabolites, such as antibiotics, immunosuppressants or cytostatic agents. Many of these compounds are peptide-like, and are formed via non-ribosomal pathways, utilizing large multi-functional enzymes called non-ribosomal peptide synthetases (NRPSs) . NRPSs are of modular structures, and each module incorporates one amino acid into the final product. The enzymatic units present in a module form separate domains that catalyse at least three steps of non-ribosomal peptide synthesis, i.e. activation of an amino acid, its covalent binding to the enzyme, and formation of a peptide bond. The activation of a specific, often non-standard, amino acid involves formation of an aminoacyl adenylate in the adenylation domain (also known as the AMP-binding domain), followed by transfer of the aminoacyl residue to a 4′-phosphopantetheinyl group attached to a conserved serine of a nearby thiolation domain. Finally, the condensation domain catalyses formation of a peptide bond between two activated amino acids.
Non-ribosomal synthesis of peptides has long been thought to be restricted to microorganisms. However, the Ebony protein from the fruit fly, which condenses β-alanine to dopamine and other amines, comprises an adenylation domain and a thiolation domain similar to those found in NRPSs . The adenylation domain ligates β-alanine to phosphopantetheine. This reaction is followed by transfer of the aminoacyl moiety onto a free acceptor rather than a protein-bound acceptor as in NRPSs.
Mammalian genomes have also been found to encode a homologue of bacterial NRPSs and the insect Ebony, which was initially named U26  and is now designated acyl CoA synthase family member 4 (ACSF4) . ACSF4-U26 is a protein of ~ 1100 amino acids, containing an adenylation and a thiolation domain, as well as a C-terminal pyrroloquinoline quinone dehydrogenase (PQQDH)-related domain. Distant homology of this protein with the yeast 2-aminoadipate reductase, which also comprises adenylation and thiolation domains, led Kasahara and Kato  to postulate that ACSF4-U26 was 2-aminoadipate semialdehyde dehydrogenase (thus the frequent but incorrect description of ACSF4-U26 as a 2-aminoadipate semialdehyde dehydrogenase; AASDH) and that PQQ was a redox cofactor for the enzyme. These findings have been seriously questioned, as no PQQ-dependent 2-aminoadipate semialdehyde dehydrogenase activity was detected in mammalian tissues either in vivo or in vitro, and ACSF4-U26-dependent oxidation of 2-aminoadipate semialdehyde to 2-aminoadipate has never been experimentally confirmed [6, 7]. The authentic mammalian 2-aminoadipate semialdehyde dehydrogenase has now been molecularly identified, and shown to be an NAD+-dependent enzyme belonging to the family of aldehyde dehydrogenases .
The aim of the present work was to identify the reaction catalysed by ACSF4-U26. We provide evidence that this protein serves as a β-alanine-activating enzyme.
Analysis of the ACSF4-U26 sequence
Figure 1 shows the three domains found in mouse ACSF4-U26: an N-terminal adenylation domain, homologous to similar domains in NRPSs (AMP-binding domain; Pfam00501), a thiolation domain with a typical consensus GXDS sequence, with the serine present in this motif involved in the attachment of phosphopantetheine, and a C-terminal PQQDH-related domain of unknown function that is extremely distant from those of authentic dehydrogenases using pyrroloquinoline quinones as cofactor. Compared to classical NRPS (as found in bacteria and some fungi), ACSF4-U26 is unusual in having only a single adenylation domain and no condensation domain. This is also the case for Ebony, which comprises a C-terminal domain that is hypothetically involved in condensation with amines .
BLAST searches indicated that a protein with a similar architecture was present in many eukaryotes, including mammals, birds, fish, Actiniaria (Nematostella vectensis), Placozoa (Branchiostoma floridae, Ciona instestinalis and Trichoplax adherens), the beetle Tribolium castaneum, insects (Drosophila spp., Anopheles spp. and Culex quinquefasciatus), Hydra magnipapillata, and plants, including Populus trichocarpa, Vitis vinifera, Arabidopsis thaliana, and the moss Physcomitrella patens. No protein with a similar architecture was found in fungi or prokaryotes. However, prokaryotes and fungi do not express NRPSs that contain both an adenylation domain and a PQQDH-related domain.
BLAST searches using the mouse adenylation domain of ACSF4-U26 indicated a significantly higher degree of identity (BLAST scores up to ~ 180) with similar domains found in fungal or bacterial NRPS, but less similarity with mammalian acyl CoA ligases (BLAST scores up to ~ 62). This, together with the presence of a putative phosphopantetheinylation site, suggests a role similar to that of NRPS rather than in formation of an acyl CoA. Pairwise comparisons indicated that, as expected, the adenylation domain of mouse ACSF4-U26 was less similar to plant ACSF4-U26 than to insect ACSF4-U26. Remarkably, the adenylation domain of plant ACSF4-U26s was significantly more similar to the adenylation domain of insect Ebony (~ 30% sequence identity; BLAST scores of ~ 190) than of mammalian or insect ACSF4-U26s (~ 25% sequence identity, BLAST scores of ~ 110). This close relationship between the adenylation domains of plant ACSF4-U26 and insect Ebony is further illustrated by the evolutionary tree (Fig. 2), which shows that the adenylation domain of Ebony clusters with those of plant ACSF4-U26s, rather than those of insect and vertebrate ACSF4-U26s. Additional sequence comparisons indicated that the adenylation domains of plant ACSF4-U26 and insect Ebony cluster with similar domains found in cyanobacteria (data not shown).
The specificity of NRPS is dependent on the residues that line the catalytic pocket of the adenylation domain . Sequence alignments with adenylation domains for which the crystal structure is known allow identification of a set of residues (ten residues in the case of the Stachelhaus motif) that are predicted to line the catalytic pocket and thereby determine the specificity [9, 10]. Specificity prediction for a protein domain may then be made by comparing its Stachelhaus motif to the Stachelhaus motifs of adenylation domains with known specificities. The Stachelhaus motifs of ACSF4-U26 proteins from various species and of Ebony from Drosophila melanogaster were compared to those of adenylation domains (of known specificity) that share the highest degree of identity (Table 1). The Stachelhaus motifs of mouse, chicken and fruit fly ACSF4-U26s were quite similar (seven identical amino acids), but were considerably different (two or fewer identical amino acids) from those of plant ACSF4-U26s. The latter shared seven identical amino acids with the Stachelhaus motif in the adenylation domain from Streptomyces verticillus peptide synthase NRPS2-1, which is specific for β-alanine, and six or seven identical amino acids with D. melanogaster Ebony, which is also specific for β-alanine. In the case of mouse, chicken and fruit fly ACSF4-U26s, the best hit was an adenylation domain specific for glycine; however, this shared only four or five identical amino acids with the queried proteins. This suggests that the specificity prediction was significantly stronger in the case of plant ACSF4-U26s compared to their vertebrate or insect orthologues.
Table 1. Prediction of the substrate specificity of ACSF4-U26 from various species. The indicated proteins were used as queries to identify, among adenylation domains of known specificity, those with a similar Stachelhaus motif. The highlighted residues in the queried proteins are those that match the template proteins. Note that ACSF4-U26 from Drosophila melanogaster was not mentioned in .
ACSF4-U26 is the only known example of a protein that combines an adenylation domain with a PQQDH-related domain. It is therefore likely that it plays a similar role in the organisms in which it is found, i.e. in plants and in animals. As the same amino acid may be activated by adenylation domains with different Stachelhaus motifs , we hypothesized that the specificity of ACSF4-U26 was conserved throughout evolution, and that this protein, whether from animals or plants, activates β-alanine.
Identification of ACSF4-U26 as a β-alanine-activating enzyme
To verify our hypothesis that ACSF4-U26 catalyses the activation of β-alanine, we produced the mouse enzyme in HEK293T cells as a fusion protein with a 6 x His tag at its N-terminus. We also produced a mutant lacking the C-terminal PQQDH-related domain (truncated ACSF4-U26), as well as a protein in which the serine that putatively binds the phosphopantetheine group (S591) was replaced by an alanine (S591A-ACSF4-U26). As shown in Fig. 3A, recombinant proteins of the expected molecular masses (~ 125 and 78 kDa for the full-length and truncated ACSF4-U26, respectively) were over-expressed in HEK293T cells. We next tested the activity of the recombinant proteins in cell lysates by measuring [3H]β-alanine incorporation into cell protein. [3H]-radiolabelled protein was only found in cell lysates containing either full-size ACSF4-U26 or its truncated version, and no incorporation was detected with the S591A-ACSF4-U26 mutant (Fig. 3B), indicating the requirement for the phosphopantetheine group, which presumably serves as the acceptor. The higher rate of β-alanine incorporation with truncated ACSF4-U26 than with the full-length protein is probably due to a higher level of expression of the former protein (see Fig. 3A).
To verify whether β-alanine was specifically incorporated into ACSF4-U26, fluorographic visualization of [3H]β-alanine-labelled proteins present in the cell extract was performed in lysates of HEK293T cells expressing ACSF4-U26 or not. Fig. 3C shows exclusive ATP-dependent radiolabelling of a single protein band of the expected molecular mass in lysates containing recombinant ACSF4-U26.
This observation was confirmed by results showing that β-alanine was incorporated into recombinant ACSF4-U26 that had been purified to homogeneity by Ni-Sepharose affinity chromatography (Fig. 4). These experiments also confirmed that incorporation of [3H]β-alanine was totally dependent on the presence of a nucleotide triphosphate in the reaction mixture. ATP was preferentially utilized compared to other ‘standard' nucleotides. KM values of 0.05, 1.25, 1.7 and 0.61 mm were obtained with ATP, CTP, GTP and UTP, respectively (Fig. S1). The relative Vmax values were 0.062, 0.026, 0.042 and 0.043 pmol·min−1·μg−1 protein, respectively. The reaction rate was low at 1 μm β-alanine (0.038 ± 0.002 pmol·min−1·μg−1 protein) but was clearly accelerated in the presence of a tenfold increase in the concentration of β-alanine (0.16 ± 0.02 pmol·min−1·μg−1), although the enzyme was still not fully β-alanylated after 120 min of incubation (Fig. 4A). According to our calculations, ~ 75% of enzyme molecules were labelled at this time point. The β-alanine saturation curve was hyperbolic (Fig. 4B), and ACSF4-U26 displayed high affinity towards this substrate (KM = 5.9 ± 1.4 μm).
Substrate specificity of the loading reaction
To investigate the possibility that ACSF4-U26 uses substrates other than β-alanine, we tested the ability of several compounds, particularly β-alanine analogues, to inhibit loading of ACSF4-U26 with radiolabelled β-alanine. As expected, non-radiolabelled β-alanine, at 1 mm, strongly inhibited incorporation of [3H]β-alanine into the enzyme (Fig. 5). d,l-2-methyl-β-alanine, a close structural analogue of β-alanine, also caused inhibition, but its effect was almost 100-fold less potent than that of β-alanine, as the inhibition reached only ~ 60% at 1 mm. Interestingly, glycine, which was weakly predicted to be a potential substrate of mouse ACSF4-U26, did not significantly inhibit the reaction. Except for β-alanine and its methyl derivative, the only amino acid that was found to cause apparent inhibition of β-alanine incorporation was cysteine. However, as shown below, this is probably due to the fact that cysteine promotes unloading of ACSF4-U26.
Search for an acceptor
By analogy with the reaction catalysed by the fruit fly protein Ebony, which not only loads β-alanine onto its phosphopantetheine group, but also transfers it onto a free amine acceptor , we investigated the possibility that ACSF4-U26 catalyses the formation of an amide bond in the presence of a suitable amine acceptor. This was tested by incubating [3H]β-alanine-loaded ACSF4-U26 in the presence of several amino compounds (tested at 1 mm). The reaction was stopped by protein precipitation with trichloroacetic acid to separate free and ACSF4-U26-bound radioactivity. These experiments were performed with the 20 standard amino acids and several amines, such as histamine, dopamine, tyramine, tryptamine, octopamine, spermine, spermidine, cadaverine and putrescine (Fig. S2). Unloading of ACSF4-U26 was only observed with cysteine (Fig. 6 and Fig. S2).
We next tested a series of other compounds that also contained a thiol group, and found that all tested sulfhydryl compounds promoted unloading of radiolabelled ACSF4-U26 (Fig. 6). Furthermore, the rate of unloading was the same whether full-length or truncated ACSF4-U26 was used, indicating that the PQQDH-related domain is not involved in the transfer of active β-alanine onto sulfhydryl compounds. The thiol-induced unloading of [3H]β-alanine-labelled ACSF4-U26 was most likely due to a transthioesterification reaction, which involves nucleophilic attack by a thiol anion on the carbonyl carbon of the β-alanyl moiety. l-cysteine thioesters are known to get rearranged to amides by an intramolecular S,N-acyl shift . To determine the identity of the unloaded products, we took advantage of the alkali lability of the thioester bonds compared to amide bonds, and of the large difference in isoelectric points of the end products depending on whether the bond was a thioester or an amide (Table S1).
The radiolabelled product formed in the presence of dithiothreitol clearly bound to a strong cation exchanger at close to neutral pH (pH 7.5) but not after being subjected to alkaline hydrolysis (Fig. S3), indicating that the compound was β-alanyl-dithiothreitol, rather than free β-alanine. Similar analysis performed with the radioactive products liberated with cysteamine, l-cysteine and l-cysteinyl-glycine indicated that these product contained an amide bond rather than an ester bond. Finally, as expected from the previous results, there was no difference in the chemistry of the end products whether the full-length or the truncated form of ACSF4-U26 was used, confirming that the PQQDH-related domain was not involved in their formation (data not shown).
ACSF4-U26 is a β-alanine-activating enzyme
The present work shows that ACSF4-U26, defined as a protein containing an adenylation domain, a phosphopantetheine domain and a PQQDH-related domain, is a widespread protein in eukaryotes, being found in animals and plants but not in fungi. Its adenylation domain is closer to similar domains found in bacterial and fungal NRPSs than those in acyl CoA ligases. This finding, together with the presence of a putative phosphopantetheine domain, indicated that ACSF4-U26 may ligate an amino acid to the thiol group of a phosphopantetheine prosthetic group. Prediction of the putative enzyme substrate using the Stachelhaus motif was inconclusive in the case of vertebrate and insect ACSF4-U26, but strongly suggested that β-alanine was the activated amino acid in the case of the plant enzymes. As the same amino acid may be activated by adenylation domains that differ in their Stachelhaus motifs , we hypothesized that β-alanine was the substrate for all forms of ACSF4-U26, including the mammalian enzyme.
This results was verified by showing that mouse ACSF4-U26 incorporates radioactivity when incubated in the presence of ATP and radiolabelled β-alanine. No incorporation was observed in an ACSF4-U26 mutant in which the serine that is predicted to bind phosphopantetheine was replaced by an alanine (S591A-ACSF4-U26). By analogy with NRPSs , these findings indicated that ACSF4-U26 covalently binds β-alanine to form a thioester with its phosphopantetheine group. The specificity for β-alanine is indicated by the low KM for this substrate and the finding that, among many compounds tested, including the 20 classical amino acids and β-alanine analogues, only 2-methyl-β-alanine behaved as a competitor. Remarkably, competition required ~ 100-fold higher concentrations of this close structural analogue of β-alanine than of non-radiolabelled β-alanine. Cysteine was also found to decrease the incorporation of β-alanine, but, as discussed below, this was due to the fact that it promoted unloading.
The specificity of ACSF4-U26 for β-alanine has not been investigated in the case of plant ACSF4-U26, but the high degree of conservation of its Stachelhaus motif compared with insect Ebony, which synthesizes β-alanyl-dopamine and related β-alanine derivatives , suggests that it activates β-alanine.
Search for the unloading reaction
Bacterial and fungal NRPSs have condensation domains that allow transfer of an activated acyl onto an acceptor bound to the phosphopantetheine group of a neighboring unit. No condensation domain was found in ACSF4-U26, which, furthermore, comprises only one amino acid activation domain and one phosphopantetheine-binding site. In this respect, ACSF4-U26 resembles Ebony. However, the latter enzyme is able to transfer the activated β-alanine onto free dopamine and other biogenic amines. This acyltransferase activity is presumably catalysed by the putative amine-selecting C-terminal domain of Ebony . No homology with acyltransferases was found in the case of the C-terminal PQQDH-related domain of ACSF4-U26.
We nonetheless tested a series of potential acceptors, including amino acids, several amines (particularly those known to be substrates for Ebony) and related compounds, and found unloading activity only with cysteine, cysteamine and other thiols. Analysis of the compounds formed in the presence of cysteine and cysteamine indicated that an amide bond rather than a thioester bond was formed. The fact that this formation occurred at similar rates with ACSF4-U26 and the truncated ACSF4-U26 indicated that the transfer reaction was not catalysed by the PQQDH-related domain.
A similar effect of cysteine on the unloading of aminoacyl tRNAs bound to their cognate aminoacyl tRNA synthase was described previously . This reaction leads to formation of dipeptides such as Arg-Cys and Val-Cys. Studies with various thiols and cysteine analogues indicated that the reaction involves a transient S-aminoacyl-cysteine intermediate, which is rearranged to a peptide bond. The reaction is orders of magnitude faster than the non-enzymatic unloading of aminoacyl tRNAs by thiol derivatives, and therefore appears to be catalysed by aminoacyl tRNA synthases. The unloading reaction of ACSF4-U26 is presumably catalysed by the adenylation domain of this protein. Although structurally distinct, the adenylation domain of NRPSs (including ACSF4-U26) are similar to aminoacyl tRNA ligases in that they catalyse a ligation reaction involving formation of an aminoacyl adenylate followed by transfer of the acyl group to form a (thio)ester. Although these enzymes are structurally unrelated, common functional groups in the two types of catalytic sites may participate in catalysis of the thiol-mediated breakdown reaction. The fact that the cysteine-dependent unloading reaction is rather slow, requires elevated concentrations of this amino acid and does not involve the PQQDH-related domain led us to conclude that it most likely does not represent the physiological activity of ACSF4-U26, but instead represents a minor side activity without significance.
Potential physiological role of ACSF4-U26
The PQQDH-related domain of ACSF4-U26 belongs to the family of β-propeller proteins, which also include WD40 domains. β-propeller domains frequently serve as scaffolds for association with other proteins, although some of them do have catalytic activity . The PQQDH-related domain of ACSF4-U26 shows very low sequence identity with genuine dehydrogenases that use PQQ as a cofactor. Furthermore, the residues that bind PQQ in these enzymes are not conserved in ACSF4-U26, indicating that it is unlikely to have dehydrogenase activity. BLAST searches indicated that the closest bacterial homologues of ACSF4-U26 (showing up to 30% sequence identity) form a bidomain protein with a protein kinase (e.g. in Chloroflexus aurantiacus) or participate in assembly of outer membrane lipids (e.g. YfgL, also known as BamB ), suggesting that the PQQDH-related domains are probably involved in protein–protein interactions. By analogy, we hypothesize that the PQQDH-related domain of ACSF4-U26 is involved in a scaffolding function. Thus, this domain may interact with an acyltransferase that catalyses transfer of β-alanine onto its physiological acceptor.
The pathway for synthesis of the most abundant physiological β-alanine derivatives is known, and their synthesis is therefore unlikely to involve ACSF4-U26. The most important of these derivatives are carnosine (β-alanyl-histidine), which is formed by an ADP-forming ligase that was recently molecularly identified , and the methylated forms of carnosine (anserine and balenine), which are formed by methylation of carnosine by various enzymes . Much less abundant derivatives of β-alanine are β-alanyl-lysine and β-alanyl-hypusine [17-19]. β-alanyl-lysine is most likely formed by a side activity of carnosine synthase, which is known to be rather non-specific [15, 20-23] as it ligates β-alanine to several basic amino acids. The hypusine derivative is most likely formed from hypusine in a similar manner. Hypusine is a form of lysine modified with a 2-hydroxy-4-aminobutyl extension on its epsilon amino group. It is exclusively found in eiF5A (eukaryotic initiation factor 5A) eukaryotic translation initiation factor 5A (eIF5A) , and is presumably released in free form by proteolysis.
Our present hypothesis regarding the function of ACSF4-U26 is that it is required for an as yet unknown post-translational modification of a protein or for post-transcriptional modification of an RNA. In addition to post-translational modifications that are very common (such as phosphorylation, acetylation and methylation), proteins may undergo modifications that are infrequent and even in some cases restricted to a single protein species. An example of this is the modification of a lysine residue to hypusine in eIF5A (see above), conversion of a histidine residue to diphthamide in eukaryotic elongation factor 2 , and addition of an ethanolamine phosphoglycerol group to a glutamate in eukaryotic elongation factor 1A (eEF1A) . Post-translational modification by β-alanine may have been overlooked not only because of its scarcity, but also because of technical difficulties encountered in its detection. The presence of a β-alanyl moiety on the N-terminus of a protein is likely to block its chemical sequencing. Its attachment to the side chain of a lysine may prevent cleavage with trypsin, a frequent strategy used to sequence proteins. In addition, β-alanine is an isomer of α-alanine, and therefore requires specific attention for detection by mass spectrometry .
Similarly to proteins, RNAs undergo multiple modifications after their assembly, and some of these modifications are restricted to only one type of RNA, such as the mannosylqueuosine and galactosylqueuosine modification found in mammalian tRNA(Asp) and tRNA(His), respectively [28, 29]. Modification of RNA by β-alanine may have been overlooked because β-alanine is produced when dihydrouridine, a constituent of tRNAs, is degraded in alkali .
The involvement of ACSF4-U26 in a rare modification may account for the slowness of the reaction it catalyses: full loading of the enzymes takes several tens of minutes compared to ~ 0.15 min and 1 min for the two adenylation domains studied by Stachelhaus et al. . Furthermore, the scarcity of the modification involving β-alanine would also account for the fact that it has been overlooked until now.
In conclusion, our studies indicate that ACSF4-U26 is responsible for an uncommon reaction leading to activation of β-alanine by its fixation to a phosphopantetheine group. The name ACSF4-U26 is misleading, and we propose that it be replaced by ‘β-alanine activating enzyme'. Now that the substrate specificity of this enzyme is known, further large-scale studies on protein or RNA modifications, or on the interactome of the β-alanine activating enzyme are likely to lead to identification of the other substrate, i.e. the β-alanine acceptor, and the putative acyltransferase involved in forming the final product.
[3H]β-alanine was purchased from Moravek Biochemicals (Brea, CA, USA). HisTrap HP (Ni2+ form, 1 mL) and PD-10 columns were obtained from GE Healthcare Bio-Sciences (Uppsala, Sweden). Dowex AG50W-X4 resin (200 mesh) was obtained from Sigma-Aldrich (St Louis, MO, USA) and Vivaspin 20 centrifugal concentrators were obtained from Sartorius Stedim (Kostrzyn, Poland). All other enzymes and DNA-modifying enzymes, as well as the TurboFect transfection reagent, were obtained from Fermentas/Thermo Scientific (Waltham, MA, USA).
ACSF4-U26 and Ebony protein sequences were identified by protein BLAST searches using mouse ACSF4-U26 and fruit fly Ebony protein sequences (accession CAA11962.1). The amino acid sequences of adenylation domain were obtained from identified proteins based on data provided by the National Center for Biotechnology Information Conserved Domain Database (http://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml). A phylogenetic analysis was performed on the Phylogeny.fr platform (www.phylogeny.fr) . Amino acid sequences were aligned using T-Coffee version 6.85 . After alignment, ambiguous regions were removed using Gblocks version 0.91b . Phylogenetic trees were generated using PhyML  (maximum-likelihood method) with the WAG model for amino acid substitution . The final tree was customized using the editing interface TreeDyn . The confidence level was assessed using the aLRT test (minimum of SH-like and χ2-based parametric) . Stachelhaus motifs were searched for using the SBSPKS web server .
Cloning, over-expression and purification of mouse recombinant ACSF4-U26s
Mouse brain cDNA was used to PCR amplify the open reading frame encoding ACSF4-U26 (GenBank accession number NM_173765.3) using Pfu DNA polymerase in the presence of 1 m betaine. Due to its length (3303 bp), the open reading frame was amplified as three separate and overlapping fragments using the primers listed in Table S2. The lengths of the PCR-amplified fragments of ACSF4-U26 were 1662 bp (fragment 1, primers #1 and #2), 1083 bp (fragment 2, primers #3 and #4) and 731 bp (fragment 3, primers #5 and #6). These were cloned into pBluescript II SK (Stratagene, La Jolla, CA, USA), sequenced and digested using AvrII (fragment 1), AvrII and NdeI (fragment 2), or NdeI and NotI (fragment 3) and assembled by ligation into pBluescript to create a full-length open reading frame encoding ACSF4-U26 protein (Table S1). An expression plasmid for ACSF4-U26 fused to an N-terminal 6 x His tag (pEF6/ACSF4-U26) was constructed by sub-cloning the assembled open reading frame into the pEF6/His-B vector (Invitrogen, Carlsbad, CA, USA), using KpnI and NotI enzymes, and verified by sequencing (see Table S1).
N-terminal 6 x His-tagged truncated ACSF4-U26 was obtained by PCR amplification of nucleotides 1-2004 from the mouse ACSF4-U26 open reading frame using Pfu DNA polymerase and primers #1 and #7 (see Table S1). The PCR-amplified fragment was digested using KpnI and NotI restriction enzymes, and cloned into the pEF6/His-B vector. The S591A-ACSF4-U26 mutant was generated by site-directed mutagenesis using a QuikChange II XL kit (Stratagene, La Jolla, CA, USA), with pEF6/ACSF4-U26 as the template and mutagenic primers #8 and #9 (see Table S1). The sequences of both truncated ACSF4-U26 and S591A-ACSF4-U26 were verified by DNA sequencing.
HEK293T cells were transfected (6 μg of expression plasmid per plate), and cell extracts were prepared as described previously . After removing the medium and washing the plates once with 5 ml phosphate buffered saline for 30 sec., cells were harvested by scraping in 1 mL of 50 mm HEPES, pH 7.5, containing 10 mm KCl, 1 mm MgCl2, 1 mm tris(2-carboxyethyl)phosphine (TCEP), 5 μg·mL−1 leupeptin and 5 μg·mL−1 antipain.
Purification of mouse recombinant ACSF4-U26, truncated ACSF4-U26 and S591A-ACSF4-U26 was performed using HisTrap HP columns (1 mL) as described previously , except that the soluble fraction of HEK293T lysates (7–12 mL) was diluted threefold in buffer containing 50 mM Hepes, pH 7.5, 300 mM NaCl, 20 mM imidazole, 10 mM KCl2, 1 mM MgCl2, 5 μg/ml leupeptin and 5 μg/ml antipain (buffer A). After loading and washing the column (15 mL) with buffer A, the retained protein was eluted using a stepwise gradient of imidazole (30, 60 and 300 mm) in buffer A. The recombinant proteins were eluted from the column using either 60 mm (ACSF4-U26 and S591A-ACSF4-U26) or 300 mm imidazole (truncated ACSF4-U26). In all cases, the homogeneity of the preparations was > 95% as confirmed by SDS/PAGE (data not shown). The elution fractions (4–8 mL) containing recombinant proteins were concentrated to ~ 2.5 mL using a Vivaspin 20 filter, and supplemented with BSA (1.4 mg·mL−1). The enzyme preparations were further desalted on PD-10 columns equilibrated with 50 mm HEPES, pH 7.5, 10 mm KCl, 5 mm TCEP and 1 mm MgCl2. Protein concentration was determined before addition of BSA as described previously  using bovine γ-globulin as the standard. Between 0.2 and 0.7 mg of pure recombinant enzyme was obtained from 70 mg of soluble HEK293T cell protein. The purified enzymes were stored at −70 °C.
When appropriate, the N-terminal 6 x His-tagged recombinant proteins were detected by western blot analysis as described previously .
Fluorography of [3H]β-alanine-labelled proteins
Fluorography was performed using Amplify fluorographic reagent (GE Healthcare, Buckinghamshire, UK) according to the manufacturer's instructions. Briefly, cell-free HEK293T lysates (0.1 mg protein) collected 48 h post-transfection were incubated in the presence or absence of 3 mm ATP-Mg in a reaction mixture (110 μL) containing 50 mm HEPES, pH 7.5, 10 mm KCl, 1 mm MgCl2, 1 mm dithiothreitol and [3H]β-alanine (~ 3 × 106 cpm; 1 Ci mmol−1) for 2 or 10 min at 37 °C. The incubation was stopped by transferring 100 μL to 4 x concentrated Laemmli sample buffer containing 40 mm dithiothreitol and heating at 90 °C for 5 min. Denatured samples (38 μg protein) were electrophoresed on a 10% Tris gel using Tris/glycine/SDS running buffer, stained with Coomassie Brilliant Blue, soaked in Amplify fluorographic reagent for 30 min, dried onto Whatman blotting paper (GE Healthcare, Buckinghamshire, UK) under vacuum at 60 °C, and placed in close contact with Hyperfilm ECL (GE Healthcare, Buckinghamshire, UK), enclosed in a hypercassette (GE Healthcare, Buckinghamshire, UK) for 2 weeks at −70 °C, and finally developed.
Assay of ACSF4-U26 activity (loading assay)
ACSF4-U26-dependent activation of β-alanine was determined by measuring the incorporation of [3H]β-alanine into the enzyme protein due to formation of a thioester bond between the amino acid and a prosthetic group of the enzyme (loading assay). The standard incubation mixture (110 μL) contained 50 mm HEPES, pH 7.5, 10 mm KCl, 1 mm MgCl2, 5 mm TCEP, 3 mm ATP-Mg and 1 μm [1H+3H]β-alanine (~ 0.75 × 106 cpm). The reaction was started by addition of the enzyme preparation and performed at 37 °C for the time intervals indicated in figures. The incubation was stopped by addition of 100 μL of the reaction mixture to 25 μL BSA (40 mg·mL−1) and 800 μL of ice-cold 10% w/v trichloroacetic acid. After 30 min on ice, the precipitate was pelleted (13 000 g for 10 min at 4 °C) and washed twice with 1 mL of chilled 10% trichloroacetic acid. The pellet was finally dissolved in 500 μL of formic acid. Incorporated radioactivity was counted in 12 mL Ultima Gold scintillation fluid (Perkin-Elmer, Waltham, MA, USA) using a Beckman LS6000 IC liquid scintillation counter (Beckman Instrument, Fullerton, CA, USA).
To confirm that incorporation of [3H]β-alanine into the enzyme protein was due to formation of a thioester bond between β-alanine and the prosthetic group of the enzyme, the loaded enzyme was subjected to alkaline hydrolysis. Briefly, the enzyme was first loaded with [3H]β-alanine by incubating purified recombinant ACSF4-U26 protein (113 μg) or truncated ACSF4-U26 (118 μg) for 25 min at 37 °C in the reaction mixture (900 μL final volume) containing 50 mm HEPES, pH 7.5, 10 mm KCl, 1 mm MgCl2, 5 mm TCEP, 3 mm ATP-Mg, 1 mg·mL−1 BSA and 1 μm [1H+3H]β-alanine (~ 75 × 106 cpm). Non-incorporated β-alanine was removed by gel filtration of the reaction mixtures on PD-10 columns equilibrated with 50 mm HEPES, pH 7.0, 10 mm KCl, 5 mm TCEP and 1 mm MgCl2. The PD-10 eluates (3 mL) containing exclusively radiolabelled ACSF4-U26 were subsequently concentrated to 1.3 mL in Vivaspin 20 ultrafiltration devices.
β-alanine-loaded ACSF4-U26 or truncated ACSF4-U26 (~ 4.4 μg) was then incubated for 60 min at 55 °C in the reaction mixture (125 μL) containing 50 mm HEPES, pH 7.0, 10 mm KCl, 1 mm MgCl2, 5 mm TCEP in the absence or presence of 1 m KOH. To neutralize the hydrolysis reaction mixture, 17 μL of 6 m HCl were added and the incubation was stopped by addition of 25 μL BSA (40 mg·mL−1) and 800 μL of ice-cold 10% w/v trichloroacetic acid. After 30 min on ice, the radioactivity that remained bound to ACSF4-U26 was recovered by centrifugation (13 000 g for 10 min) of the mixture followed by washing the pellet twice with 1 mL of chilled 10% trichloroacetic acid and dissolving it in 500 μL of formic acid. The radioactivity released from the loaded enzyme (present in the supernatant of the centrifugation and washing steps) was also analysed. The radioactivity in ACSF4-U26 or released from ACSF4-U26 was counted after addition of scintillation liquid to both samples.
Assay of ACSF4-U26 activity (unloading assay)
ACSF4-U26-dependent transfer of the activated β-alanine onto its potential acceptor was determined by measuring acceptor-induced release of the radioactivity from [3H]β-alanine-loaded enzyme (unloading assay). β-alanine-loaded ACSF4-U26 or truncated ACSF4-U26 (~ 4.9 μg) prepared as described above and free from contaminating radioactivity was incubated for 30 min at 37 °C in a reaction mixture (100 μL) containing 50 mm HEPES, pH 7.0, 10 mm KCl, 1 mm MgCl2, 5 mm TCEP in the absence or presence of tested compounds at 1 mm concentration. The incubation was stopped by addition of 25 μL BSA (40 mg·mL−1) and 200 μL of ice-cold 10% w/v perchloric acid. After 30 min on ice, the precipitate was pelleted by centrifugation (13 000 g for 10 min), and washed twice with 1 mL of chilled 10% perchloric acid. The pellet was dissolved in formic acid and its radioactivity was determined as described above.
Cation-exchange chromatography of products of the unloading reaction and their hydrolysates
To determine the chemical nature of the products of the unloading reaction, the chromatographic properties and sensitivity to alkali of these compounds were investigated. Briefly, the deproteinized reaction mixtures obtained following the unloading assay (325 μL) were diluted with water (100 μL). After neutralization with 3 m K2CO3, the salts were removed by centrifugation (13 000 g for 15 min), and the clear supernatant was split into two equal halves (200 μL). One half was acidified with 30% perchloric acid (49 μL), stored on ice for ~ 10 min, and neutralized with 5 m KOH. The other half was hydrolysed in the presence of 1 m KOH for 60 min at 55 °C, and then neutralized by addition of 30% perchloric acid (49 μL). The salts were removed by centrifugation (13 000 g, at 20 °C for 10 min), and the clear supernatants were diluted five times with 20 mm HEPES, pH 7.5, and 2 mL aliquots were applied to Dowex 50WX4-200 mesh columns (1 mL, Na+ form), equilibrated with 20 mm HEPES, pH 7.5. Radiolabelled products with no positive charge were eluted by washing the columns with 10 mL of the same buffer. To elute products that were positively charged, the columns were washed with 10 mL of 1 m NH4OH (see Table S1). In all cases, the samples were mixed with seven volumes of scintillation liquid, and their radioactivity was determined as described above.
This investigation was co-financed by the European Regional Development Fund under the Operational Programme Innovative Economy via a grant from the Foundation for Polish Science to J.D. (HOMING PLUS/2010-2/2). Work in the laboratories of M.V.D.C. and E.V.S. was supported by a grant from the Fonds National de la Recherche Scientifique (FNRS) and the Center of Excellence des Désordres Inflammatoires dans les Affections Neurologiques (DIANE) programme of the Région Wallonne. M.V.D.C. is a Chercheur Qualifié of the Belgian Fonds National de la Recherche Scientifique.