A novel way to synthesize pantothenate in bacteria involves β‐alanine synthase present in uracil degradation pathway

Abstract Pantothenate is an indispensable vitamin precursor of the synthesis of coenzyme A (CoA), a key metabolite required in over 100 metabolic reactions. β‐Alanine (β‐ala) is an indispensable component of pantothenate. Due to the metabolic relevance of this pathway, we assumed that orthologous genes for ß‐alanine synthesis would be present in the genomes of bacteria, archaea, and eukaryotes. However, comparative genomic studies revealed that orthologous gene replacement and loss of synteny occur at high frequency in panD genes. We have previously reported the atypical plasmid‐encoded location of the pantothenate pathway genes panC and panB (two copies) in R. etli CFN42. This study also revealed the unexpected absence of a panD gene encoding the aspartate decarboxylase enzyme (ADC), required for the synthesis of β‐ala. The aim of this study was to identify the source of β‐alanine in Rhizobium etli CFN42. In this study, we present a bioinformatic analysis and an experimental validation demonstrating that the source of β‐ala in this R. etli comes from β‐alanine synthase, the last enzyme of the uracil degradation pathway.


Hodge
proposed two classes of ADC based on the type of cleavage of the zymogen (pro-ADC). Class I of the ADC cleavage requires the MRF (Maturation Regulatory Factor) acetyl-CoA sensor and has been found only in gammaproteobacteria. ADC Class II is an autocatalytic cleavage and is found in a wide number of bacterial phyla. Since the majority of archaea lack homologues of the E. coli K12 acetyl-CoA synthesis pathway genes, the mechanism of pantothenate/CoA biosynthesis has not been completely deduced in these organisms.
Although prokaryotes and eukaryotes have an indispensable requirement for β-ala for the synthesis of coenzyme A (CoA), the pathways involved in its synthesis are very diverse. The uracil fermenting bacterium Clostridium uracilicum degrades uracil to β-ala. Uracil or thymine is first converted to dihydrouracil. The dihydropyrimidinase enzyme catalyzes the hydration of dihydrouracil to produce N-carbamoyl-β-ala, which is hydrolyzed to β-ala, CO 2 , and NH 3 , by β-ala synthase (Campbell, 1957).
The reductive degradation of pyrimidine as a source of β-ala was supported by genetic and biochemical analyses in several bacteria, including Clostridium uracilicum (Campbell, 1957) and Clostridium botulinum (Hilton, Mead, & Elsden, 1975). Although the reductive degradation of pyrimidines has also been implicated as the de novo source of β-ala in E. coli auxotrophs, lack of response to dihydrouracil indicated that in these bacteria, the major pathway for β-ala synthesis was the decarboxylation of aspartate catalyzed by ADC.
Genschel (Genschel, 2004) performed a phyletic analysis for the occurrence of E. coli and human genes for in pantothenate and CoA synthesis across 47 completely sequenced genomes, 20 from the Bacteria, 16 from Archaea, and 11 from Eukarya. This study revealed a mosaic of orthologues with 20 to 70% amino acid identities. At least one protein was missing from each of the 47 analyzed genomes.
Comparative genomics using the E. coli pantothenate pathway genes as query against the 20 sequenced bacterial genomes revealed multiple gaps that may represent distantly related homologues due to the absence of, at least, one gene per surveyed bacterial genome (Genschel, 2004).
The Rhizobiales order is a heterogeneous group of Gramnegative bacteria, taxonomically located within the alphaproteobacteria division. Some of its members are facultative diazotrophs that associate with leguminous plants to carry out symbiotic nitrogen fixation. Others are pathogens of plants or animals (Martínez-Romero & Caballero-Mellado, 1996). Our model is Rhizobium etli CFN42, which was originally isolated from bean root nodules (Martínez-Romero, 2003). Its genome consists of a circular chromosome and six large plasmids ranging in size from 194 to 642 Kb (Gonzalez et al., 2006).
In the course of examining Rhizobium etli CFN42 plasmids for the presence of housekeeping genes encoding essential functions, we found that both panC and panB genes were clustered together on the 642-kb replicon p42f. We demonstrated that both are indispensable for the synthesis of pantothenate (Villaseñor et al., 2011; Figure   A1). Surprisingly, we did not find homologues of the E. coli panD gene in the genome of R. etli CFN42. Since strain CFN42 grows in minimal medium without exogenous pantothenate or β-ala, it was assumed that it is a pantothenate prototroph.
Agrobacterium fabrum C58 (formerly A. tumefaciens C58), a plant pathogen that induces tumors in numerous plants, was the only member of the Rhizobiales order included in Genschel's study.
According to this analysis, A. fabrum C58 lacks ketopantoate reductase (KPR, EC 1.1.1.1.169) but has a putative ADC detected by a BlastP search. We performed BlastP searches in order to gain insight on the presence or absence of ADC in the genomes of rhizobial reference strains.
Several questions arise from the presence or absence of ADC in R. etli CFN42. Is the absence of ADC an exclusive characteristic of strain CFN42 or is it a widespread characteristic of the Rhizobiales order or perhaps the alphaproteobacteria?
The aim of this work was to identify the enzyme that synthesizes β-ala and replaces the function of ADC, allowing R. etli CFN42 to be a β-ala prototroph. We also performed an in silico analysis of the alphaproteobacteria group to understand the occurrence, diversity, and evolution of the enzymes involved in β-ala synthesis.

| Bacterial strains, media, and growth conditions
The characteristics of the bacterial strains and plasmids used in this study are listed in Table 1. Bacterial growth was started from glycerol stocks (20%, stored at −70°C) propagated in plates of PYrich medium (per L, 5 g peptone, 3 g yeast extract, 1 ml of CaCl 2 and 15 g agar). Rhizobium strains were grown at 30°C in three different media: (a) PY-rich medium, (b) chemically defined mineral medium (MM), and (c) chemically defined mineral medium plus 1 μM calcium pantothenate (MMP) or 1 μM β-ala, added from filter sterilized stocks. Base MM containing 10 mM succinate as carbon source, 10 mM NH 4 Cl as nitrogen source, 1.26 mM K 2 HPO 4 , and 0.83 mM MgSO 4 was adjusted to pH 6.8 and sterilized. After sterilization, the following components were added to the final con-

| Analysis for the occurrence of 12 proteins involved in pantothenate synthesis and phylogenetic analysis of putative ADC enzymes found in alphaproteobacteria
We selected 204 alphaproteobacteria to analyze for the presence and absence of 12 proteins related to the pantothenate synthesis and transport. The protein FASTA files (faa) for each of the genomes were downloaded from the RefSeq NCBI database. Protein sequences with an expectation value (E) of 10 −3 or less were considered as putative homologues. We used Proteinortho v5.15 to obtain the clusters of orthologous proteins from the 204 protein GabT. Finally, we determined which alphaproteobacteria were represented in each protein cluster (Table S1).
For phylogenetic analysis, we used the Pfam v31.0 database to determine which proteinortho clusters represent the ADC proteins.
A total of 37 homologues belonging to alphaproteobacteria sequences were tested with a group of nine external sequences listed in parameters ( Figure A2).

| Cloning and sequence analysis of amaB gene, mutants, and complemented mutant
The amaB (RHE_CH03290) gene was overexpressed in E. coli DH5-

| Overexpression and purification of wild-type β-ureidopropionase AmaB
The transformant BL21(DE3) strain was grown in LB medium supplemented with 100 mg/ml of carbenicillin. A single colony was transferred into 10 ml of LB medium with carbenicillin at the above-mentioned concentration in a 100-ml flask. This culture was incubated overnight at 37°C with shaking. Five hundred milliliters of LB medium with 100 mg/ml of carbenicillin was inoculated with 5 ml of the overnight culture in a 1-liter flask. After 3 hr of incubation at 37°C with vigorous shaking, the optical density at 600 nm (OD600) of the culture was 0.3-0.5. For induction of β-alanine synthase gene expression, isopropyl-β-D-thiogalactopyranoside (IPTG) was added to a final concentration of 0.1 mM and incubation was continued at 30°C for an additional 6 hr.

| Enzyme assays
The standard enzymatic reaction was carried out with purified AmaB at a final concentration of 1 mg/ml along with 125 mM 3-ureidopropionic acid and 10 mM MgCl 2 dissolved in 100 mM sodium phosphate buffer, pH 8.0, in a 3 ml reaction volume (Martinez-Gomez et al., 2008). The reaction mixture was incubated at 30°C for 60 min, with the apoenzyme preincubated (1 hr) at 4°C with 2 mM of NiCl 2 , and 500 µl samples were stopped for every 15 min, by the addition of 50 µl of 3% TCA. After centrifugation, the presence of β-ala in the resulting supernatants was estimated by high-performance liquid chromatography (HPLC).

| Determination of β-ala by HPLC/fluorescence
Determination of β-ala was carried out by HPLC coupled to a Multi

| Orthologues of the canonical L-aspartate-αdecarboxylase enzyme are predominantly absent in α-proteobacteria
A previous study on ADC phylogeny and amino acid conservation analyses revealed that ADCs are present in γ-proteobacterial genomes and most maintain the panCBD synteny (Stuecker et al., 2015). We noticed the absence of the panCBD gene cluster while functionally characterizing panC and panB in rhizobia (Villaseñor et al., 2011). In the present study, BlastP and Psi Blast searches using ADC from E. coli and A. fabrum C58 as query revealed the absence of ADC homologues in R. etli CFN42 and other reference strains (Table S1).
To generalize the absence of ADC homologues in α-proteobacteria, we assessed the occurrence of putative ADCs in the proteome of 204 alphaproteobacteria, 84 rhizobia and 120 members of seven families of alphaproteobacteria (Table S1). The complete proteome of each bacterium was obtained from the NCBI reference sequence collection (RefSeq) and clustered with Proteinortho v5.15, a largescale Blast-based orthology detection tool (Lechner et al., 2011; Figure A2). This analysis only showed 37 putative ADCs from 204 α-proteobacteria genomes.

| Unrooted maximum-likelihood-based tree inferred from the alpha-and gammaproteobacteria ADCs revealed high divergence among them
An important characteristic of the alphaproteobacteria is its genome plasticity, which allows different genome rearrangements, including deletions or duplications (Prell & Poole, 2006;Tiwari & Lata, 2018). We made a phylogenetic analysis to get a wider view of the evolutionary relationship among the ADCs from the α-, β-, γ-, and ε-proteobacteria (Table S2).
The resulting maximum-likelihood-based tree is shown in Figure 1, and the data set is presented in Table S2.  and Hoeflea (2 species). KPHMT was predicted to be present in the other members of Rhizobiales, which have a diversity of habitats.
Two copies of this enzyme were present in 17.8% of rhizobia, mostly in the Rhizobium and Sinorhizobium species. Three copies of KPHMT were found in 4.8% of the Rhizobiales order: three Mesorhizobium species and one in Rhizobium leucaenae (Table S1).
The next step in the pathway is the reduction of α-ketopantoate to produce pantoate. Two enzymes can perform this reduction: KPR (α-ketopantoate reductase, PanE) was found in only 57% of the rhizobial genomes, while KAR (acetohydroxy acid reductoisomerase, ilvC) was present in 95.2% of the genomes (Table S1). Most human, plant, and mammalian pathogens have lost the KPR enzyme.
Interestingly, Candidatus genera lacked both KAR and KPR enzymes in their genomes.
In the last step of the pathway, pantothenate synthetase (PS, PanC) catalyzes the ATP-dependent condensation of D-pantoate with β-ala to form pantothenate. This enzyme was absent in 7.1% of the surveyed genomes, some of which belong to parasites such as Hoeflea and Candidatus (Table S1). Strains with a single copy were found in 88% of the analyzed rhizobia genomes. Two genes were found in 4.7% of Mesorhizobium and Bradyrhizobium species.
The occurrence of putative ADC enzymes is shown in

| Rhizobium etli CFN42 is a pantothenate prototroph
The model of pantothenate synthesis established in E. coli (Cronan, 1980;Leonardi & Jackowski, 2007) indicates that the enzymes missing in rhizobia should cause auxotrophy. Growth assays were done in liquid chemically defined medium with R. etli CFN42 (lacks panD) and Sinorhizobium meliloti 1,021 (lacks panD and panE) wild-type strains, and an R. etli CFN42 plasmid p42f-cured strain (CFNX186) that is defective for growth in chemically defined medium without pantothenate. We found that the wild-type strains were able to grow through three subcultures in minimum medium without β-ala or pantothenate. This shows that even with the absence of panE and/or panD, rhizobia are still able to synthesize β-ala and pantothenate (Figure 2). This prototrophy contrasts with the auxotrophy exhibited by R. etli CFNX186, which lacks panC and panB, as well as plasmid p42f (Brom et al., 1992).

F I G U R E 1
The putative ADCs of alphaproteobacteria found in our occurrence analysis. A maximumlikelihood phylogenetic tree inferred from a subset of 204 genomes, where we extracted only 37 ADCs. The tree shows a monophyletic clade of proteins distantly related to those from γ-proteobacteria and other α-proteobacteria

| Occurrence analysis revealed different pathways that would replace ADC in rhizobia
To identify which enzyme(s) might be responsible for the synthesis of β-ala, we performed bioinformatic analyses of 204 alphaproteobacterial genomes to find possible pathways or genes that could potentially produce this metabolite. Based on a literature search, we selected six genes of interest that encode enzymes of the pyrimidine degradation pathway (AmaB, Dht, PyrD), glutamate decarboxylase (GAD), and the Aam and GabT transaminases ( Figure 3).
The pyrimidine degradation pathway involves three enzymatic steps from uracil to produce β-ala, CO 2 , and NH 3 (Campbell, 1957). In the final step of the pathway, β-ala synthase (AmaB) uses N-carbamoylβ-alanine as substrate. In rhizobia, the in vitro activity of AmaB has been detected in A. fabrum C58 and S. meliloti 1021. The authors showed the production of β-ala from 3-ureidopropionic acid in vitro, in the last step of the pathway (Martínez-Rodríguez, Martínez-Gómez, Rodríguez-Vico, Clemente-Jiménez, & Las Heras-Vázquez, 2010). In archaea, β-ala can be synthesized by a GAD that uses Asp as a substrate.
In summary, our bioinformatic analysis showed that two transaminases and the pyrimidine degradation pathway were encoded in the R. etli CFN42 genome. We did not find any candidate genes for ADC or GAD, nor a complete polyamine degradation pathway.

| AmaB functionally complements strains lacking ADC
In our study, we tested the function of different genes in R. etli, by inactivating those that encode two transaminases (Aam and GabT) and the amaB gene for pyrimidine degradation (Table A1).
Following with the canonical decarboxylation pathway, we found a putative ω-amino acid decarboxylase that was different from the ADC and GAD enzymes. The genes were interrupted using a suicide plasmid, and the resulting mutants were tested for growth in defined medium without β-ala or pantothenate. From this screening, we found that the amaB mutant was auxotrophic for β-ala, while inactivation of the other genes caused no growth deficiency (data not shown).
amaB (RHE_CH03290) is a chromosomal gene annotated as β-alanine synthase. It belongs to the pyrimidine degradation pathway and transforms 3-ureidopropionic acid to β-ala, CO 2 , and ammonia.
We disrupted this gene in R. etli CFN42 and grew the ReAM-1 (amaB -) mutant in mineral medium (MM) without β-ala or pantothenic acid.
The mutant was deficient in growth, indicating a β-ala auxotrophy, and its growth was restored by exogenous β-ala or by introducing the amaB gene in a plasmid (Figure 4).
Similarly, the mutant was complemented with a plasmid-borne copy of the amaB gene from A. fabrum C58. The product of this gene has been shown to have β-alanine synthase activity in vitro (Martinez-Gomez et al., 2008).

| Purified AmaB produces β-ala from 3-ureidopropionic acid in vitro
The his 6 -tag enzyme was purified in an immobilized nickel affinity column under native conditions and had a molecular mass of 60 kDa,
Enzymes of this type are characterized as metalloenzymes that use Ni 2+ and Co 2+ as cofactors in enzyme assays. The reaction mixture contained purified AmaB preincubated with Ni 2+ or Co 2+ , 10 mM MgCl 2 , 100 mM sodium phosphate buffer, and 3-ureidopropionic acid as a substrate. We initially used a TLC system with ninhydrin detection to identify the presence of β-ala (Niederwieser et al., 1971; Figure A4). We observed enzymatic activity with both metal ions, and no product was formed in their absence.
As described below, we also performed our enzymatic assays using an HPLC system to obtain a better resolution.

| Synthesis of β-ala by recombinant AmaB
The R. etli CFN42 amaB gene was heterologously expressed in E. coli strain BL21 (DE3) and recovered by Ni 2+ affinity chromatography, as The time course of recombinant AmaB activity using 3-ureidopropionic acid as substrate and Ni 2+ as cofactor showed that β-ala is synthesized at a linear rate for up to 30 min ( Figure 6a). β-ala was not detected in a reaction assay without recombinant AmaB protein (Figure 6b).
The standard of β-ala overlapped with the peak of the compound synthesized by AmaB, while the α-ala standard did not (Figure 6c).
These results indicated that recombinant AmaB is able to synthesize β-ala.

| D ISCUSS I ON
The relevance of β-ala as a key component of pantothenate synthesis has been well established. However, the diversity of mechanisms described in bacteria and eukaryotes suggests that the synthesis of β-ala has not been totally elucidated.
Pioneer studies performed in E. coli and γ-proteobacteria defined that β-ala was synthesized by the decarboxylation of L-aspartate in a one-step reaction catalyzed by ADC. The concept of a canonical one-step decarboxylation reaction was for many years assumed to be the sole source of β-ala in bacteria.
The genomic era facilitates the comparison of pathways among numerous species (Genschel, 2004); this bioinformatic approach helped us determine the diversity of mechanisms involved in β-ala synthesis. In this study, we found several differences between R. etli (alphaproteobacteria) and E. coli (γ-proteobacteria); the most intriguing was the absence of an ADC homologue in rhizobia. In previous studies, analyses of E. coli and other γ-proteobacteria revealed that β-ala was produced by the decarboxylation of aspartate by aspartate decarboxylase enzyme (ADC; Cronan et al., 1982); in several archaea, β-ala was synthesized by a glutamate decarboxylase (GAD) able to decarboxylate both aspartate and glutamate (Tomita et al., 2014). These data confirm the relevance of one-step decarboxylases, not only in bacteria but also in archaea.
An unusual alternative source of β-ala synthesis is the reductive degradation of pyrimidine. This three-step reaction was found in Clostridium uracilicum (Campbell, 1957) and C. botulinum (Hilton et al., 1975), as well as in E. coli strains: E. coli W, E. coli D2, E. coli 99-1, and E. coli 99-2 (Table S1); in contrast to previous studies, none of them was able to grow in the presence of dihydrouracil and β-ureidopropionic acid (Slotnick & Weinfeld, 1956).
In bacteria belonging to the Rhizobiales order, little is known about the metabolism of β-ala and pantothenate (Villaseñor et al., 2011). The occurrence analysis performed in this work indicates that our model, R. etli CFN42, lacks ADC and GAD, the most common one-step reaction used in bacteria to synthesize β-ala. We suggest that there can be functional redundancy in certain rhizobia strains.
As part of our work, we constructed different single and double mutants in A. fabrum C58 to try to get an auxotrophic strain, but in all cases, the mutants continue to be β-ala prototrophic (data not shown).
Particularly for the Rhizobiales order, we constructed a heat map with their most representative genomes; here, we can associate the loss and prevalence of different pathways, assuming that the decarboxylation pathway is missing in most of rhizobia genomes ( Figure A5). As part of our occurrence analysis, we extended our work to alphaproteobacteria with 120 more genomes from seven different orders (Table S1). We found a correlation between the rhizobia order and alphaproteobacteria. In general, we observed that the pyrimidine degradation pathway (37%) and Aam transaminase (56%) are widely distributed in alphaproteobacteria, as well as in rhizobia (Table S1). We also observed that ADC and GAD enzymes are poorly represented in alphaproteobacteria, with 17% and 6.8%, respectively. This analysis suggests a strong correlation between the loss of the decarboxylation pathway and predominance of the pyrimidine degradation pathway in the Rhizobiales order and in alphaproteobacteria.

Sinorhizobium meliloti and
We also tested the activity of recombinant AmaB in vitro, by HPLC, to confirm the catalytic activity of this protein by showing that it produces β-ala from 3-ureidopropionic acid; this corroborates the presence of alternative pathway in which bacteria produce this essential metabolite.

| CON CLUS IONS
Prokaryotes and eukaryotes require β-ala to synthesize CoA; however, the source of β-ala is quite variable even in bacteria.
For years, it has been assumed that the main source of β-ala in prokaryotes comes from the decarboxylation of aspartate in a single enzymatic step catalyzed by ADC. This reaction was discovered in E. coli and has been assumed to be the main source of β-ala in γ-proteobacteria.
This study in R. etli CFN42 and other alphaproteobacteria revealed a remarkable reduction of ADC orthologs in these bacteria.
The bioinformatics and experimental analyses performed with rhizobia indicate that in these alphaproteobacteria β-ala is synthesized through the reductive pyrimidine degradation pathway.
All these data highlight the metabolic plasticity for β-ala and pantothenate in bacteria.

ACK N OWLED G M ENTS
Mariana López-Sámano is a doctoral student from Programa

CO N FLI C T O F I NTE R E S T
None declared.

E TH I C S S TATEM ENT
None required.