How an overlooked gene in coenzyme a synthesis solved an enzyme mechanism predicament

Coenzyme A (CoA) is an essential cofactor throughout biology. The first committed step in the CoA synthetic pathway is synthesis of β‐alanine from aspartate. In Escherichia coli and Salmonella enterica panD encodes the responsible enzyme, aspartate‐1‐decarboxylase, as a proenzyme. To become active, the E. coli and S. enterica PanD proenzymes must undergo an autocatalytic cleavage to form the pyruvyl cofactor that catalyzes decarboxylation. A problem was that the autocatalytic cleavage was too slow to support growth. A long‐neglected gene (now called panZ) was belatedly found to encode the protein that increases autocatalytic cleavage of the PanD proenzyme to a physiologically relevant rate. PanZ must bind CoA or acetyl‐CoA to interact with the PanD proenzyme and accelerate cleavage. The CoA/acetyl‐CoA dependence has led to proposals that the PanD‐PanZ CoA/acetyl‐CoA interaction regulates CoA synthesis. Unfortunately, regulation of β‐alanine synthesis is very weak or absent. However, the PanD‐PanZ interaction provides an explanation for the toxicity of the CoA anti‐metabolite, N5‐pentyl pantothenamide.

the tightly linked E. coli livJ and livK genes and found the deletions engendered a β-alanine requirement. In sequencing the livJ-livK region, a small open reading frame (ORF) was found between the two liv genes (Adams et al., 1990). Since the map position agreed with the Ortega report, this resolved the differing map positions and showed that E. coli and S. enterica each have two genes required for β-alanine synthesis. The gene linked to panC encodes aspartate-1-decarboxylase, the enzyme that converts aspartate to β-alanine ( Figure 1) (the E. coli and S. enterica enzymes are 96% identical) (Cronan, 1980;Williamson & Brown, 1979). The role of the Ortega gene is more involved. Aspartate-1-decarboxylase is synthesized as an inactive proenzyme (pro-PanD) (Williamson & Brown, 1979) which must be processed to form the pyruvyl cofactor of the active PanD decarboxylase. The pyruvyl cofactor performs the same Schiff base chemistry as the pyridoxal-phosphate cofactor of many other decarboxylases (Recsei & Snell, 1984). A conundrum was that the autocatalytic conversion of pro-PanD to form the active enzyme was extremely slow in vitro (Ramjee et al., 1997;Smith, 1988). Several days at 37°C was required for autocatalytic processing to convert pro-PanD to the active enzyme; a rate much too slow to provide sufficient β-alanine for rapidly growing cells. In contrast, the other two E. coli pyruvyl cofactor enzymes are processed in minutes (see below). Hence, physiology argued that some factor was required to speed up conversion of pro-PanD to the active enzyme.
Two groups of investigators independently deduced that the gene originally mapped by Ortega and coworkers (Ortega et al., 1975) encodes the protein that accelerates pro-PanD processing Stuecker, Hodge, et al., 2012). This gene (original name yhhK) has been called both panZ and panM. However, the x-ray crystal structures use the PanZ designation, so panZ seems the accepted gene name. PanZ greatly accelerates conversion of the asparatate-1-decarboxylase proenzyme to the active enzyme in vitro Stuecker, Hodge, et al., 2012).
Mutational inactivation of either panD or panZ results in βalanine auxotrophy. This is because the panD gene encodes the inactive aspartate-1-decarboxylase, pro-PanD, and panZ encodes the protein that converts pro-PanD to the active enzyme. It is satisfying that all the above reports were correct. However, it remains a puzzle why some investigators isolated only panD strains and others only panZ strains given that the two ORFs are the same size, the mutant phenotypes are the same and chemical mutagens were used. If a single group of investigators had isolated and mapped both panD and panZ mutants, the issue would likely have been settled many years ago. Note that PanZ is not strictly necessary for β-alanine synthesis because high level PanD overexpression in a panZ deletion strain permits growth (Stuecker, Hodge, et al., 2012). Presumably, slow autocatalysis of the overproduced protein provides sufficient asparatate-1-decarboxylase activity for β-alanine synthesis.

| FORMATI ON OF THE PanD PYRU VOYL COFAC TOR
PanD is a pyruvoyl-dependent enzyme as first shown by Williamson and Brown (1979). The pyruvoyl cofactor is formed via an intramolecular rearrangement of a serine residue of the primary translation product (called the π-chain) leading to chain cleavage and formation of the covalently bound cofactor from the serine residue (Smith, 1988;Williamson & Brown, 1979). Serine residue 25 of the πchain attacks the peptide bond between G24 and S25 and the newly formed N-terminal serine is rearranged to generate an ester intermediate (Albert et al., 1998;Smith, 1988;Williamson & Brown, 1979) ( Figure 2b). The "as isolated" PanD is a tetramer and contains a mixture of fully processed, unprocessed and partially processed (ester intermediate) subunits (Albert et al., 1998) (Figure 2a). The α and β-subunits remain tightly associated after cleavage and constitute the active PanD, although the unstructured β-subunit has no known role in β-alanine synthesis (Figure 2b). The unprocessed Pro-PanD subunit is a six-stranded β barrel capped by small α-helices at each end. The active sites lie between adjacent subunits. These interfaces are extensive and largely hydrophobic consistent with the extreme resistance of the tetramer to denaturation (Albert et al., 1998). For the detailed mechanism of PanD activation, see (Albert et al., 1998; The role of β-alanine in pantothenate synthesis. β-Alanine is synthesized by PanD and ligated to D-pantoate to give pantothenate (vitamin B5). The reaction proceeds by formation of pantoyl adenylate from pantoate and ATP. The first intermediate in pantoate synthesis is derived from α-ketoisovalerate, an intermediate in the valine and leucine synthetic pathways (Leonardi & Jackowski, 2007;Webb et al., 2004). D-Pantothenate is converted to CoA in four reactions catalyzed by the CoaABCDE enzymes (see Figure 4). β-Alanine is the only β-amino acid in nature and its synthesis by PanD is the first committed step of CoA biosynthesis. The IUPAC name for β-alanine is 3-aminopropanoic acid.

| S TIMUL ATION OF PanD PROENZME MATUR ATI ON BY PanZ
PanZ is expressed as a monomeric protein in a complex with CoA or acetyl-CoA, depending on the culture conditions (Monteiro et al., 2012). The PanZ sequence is annotated as a member of the GCN5-like N-acetyltransferase (GNAT) family of enzymes that catalyze protein acetylation (Stuecker, Tucker, et al., 2012). This raised the possibility that PanZ acetylated PanD. However, PanZ lacked the conserved glutamate residue that initiates GNAT acetylation arguing against acetyl transfer from acetyl-CoA to PanZ (Stuecker, Tucker, et al., 2012). This inference was confirmed by direct experiments demonstrating that acetylation does not occur. These were the ability of a nonhydrolyzable analog of acetyl-CoA, ethyl-CoA, to stimulate PanD β-alanine synthesis in vitro and the lack of PanZ catalyzed transfer of acetyl groups from 1-[ 14 C]acetyl-CoA to PanD (Stuecker, Tucker, et al., 2012).
A problem in studying the role of CoA/acetyl-CoA in activation of PanD by PanZ is that these molecules copurify with PanZ. PanZ is not an enzyme but a stoichiometric facilitator of PanD activation (Monteiro et al., 2012;Stuecker, Hodge, et al., 2012).
Incubation of PanD with PanZ followed by assay of β-alanine synthesis showed a 24-fold increase in activity compared to PanD alone (Stuecker, Hodge, et al., 2012). In the presence of CoA or acetyl-CoA PanZ forms a 1:1 complex with each of the PanD monomers of the tetramer (Monteiro et al., 2012) (Figure 3). Binding of CoA or acetyl-CoA puts PanZ in a PanD binding conformation (Stuecker, Hodge, et al., 2012). In the most recent x-ray crystal structures, one molecule of acetyl-CoA is bound to each PanZ-PanD complex thus forming a hetero-octamer (tetrameric PanD plus four PanZ monomers) ( Figure 3). The PanZ-acetyl-CoA complex binds the C-terminal region of PanD. Formation of the PanD-PanZ complex stabilizes a conformation of PanD in which the serine-25 hydroxyl group is in close proximity to the adjacent carbonyl group, thus facilitating the attack of the G24-S25 peptide bond that initiates activation (Monteiro et al., 2012(Monteiro et al., , 2015. This argues that the very slow activation of PanD in the absence of PanZ is due to the high mobility of the PanD Nterminal segment. This mobility allows many conformers, but few are properly aligned for cleavage (Schmitzberger et al., 2003). A parallel case is the DNA damage (SOS) response in which an activated form of the RecA protein binds several repressor proteins and accelerates their slow autocatalytic cleavage (Lima-Noronha et al., 2022).
However, PanD differs in that the new N-terminus is remodeled into a pyruvoyl moiety.

F I G U R E 2
Structures of unprocessed (π) PanD and processed (α + β) PanD forms of E coli PanD. The PDB IDs of structures of the unprocessed (panel a) and processed E. coli PanD (panel b) are 1PPY and 1PYU, respectively. In panel a, all four subunits are the π form (the primary translation product). In panel b, the 24 residue β subunits clipped from the π form appear as four unstructured "noodles" two of which are labeled. The small red-yellow molecules are sulfate ions.

| THE ACHILLE S HEEL OF THE PanD PYRU VOYL COFAC TOR
A disadvantage of the PanD pyruvoyl moiety is mechanism-based inactivation of the enzyme (sometimes called suicide inactivation) (Pei et al., 2017;Smith, 1988). The enzyme loses 90% of its activity in 1 h at 42°C and 1 mM aspartate (Smith, 1988). Inactivation is caused by a side reaction of the catalytic process that occurs only under turnover conditions (presence of aspartate) and is irreversible. Mechanism-based inactivation has been elucidated for E. coli S-adenosylmethionine decarboxylase, an extensively studied pyruvoyl-dependent enzyme (Li et al., 2001). In the catalytic cycle, CO 2 is first released from the Schiff base formed by the reaction of the pyruvoyl group with the substrate. Following decarboxylation, the generated iminium intermediate can be protonated either on the substrate or on the pyruvoyl group. Consequently, an amino group is formed either on the product (the correct catalytic step) or on the pyruvoyl group which irreversibly inactivates the enzyme molecule. The E. coli and S. enterica PanDs are significantly more resistant to mechanism-based inactivation than other bacterial PanDs (Mo et al., 2019).

| THE D IS TRIBUTI ON OF PanZ IN BAC TERIA IS NARROW
PanZ is found only in a small clade of γ-proteobacteria, essentially E. coli and close relatives Stuecker et al., 2015).
Recent work has placed bacterial PanDs into three clades (Mo et al., 2018;Zhao et al., 2022) of which the PanZ-requiring enzymes form a clade. The PanD proteins of other bacteria such as Bacillus subtilis and Helicobacter pylori lack PanZ and unassisted autocatalysis efficiently processes the primary translation products to the active pyruvoyl enzymes (Mo et al., 2018;Pei et al., 2017;Zhao et al., 2022). In eukaryotes, β-alanine is made from aspartate using pyridoxal phosphate rather than a pyruvoyl cofactor (Webb et al., 2004). These enzymes are much larger than PanD and often can decarboxylate other amino acids. Some archaea also use a large pyridoxal phosphate-requiring enzyme to decarboxylate aspartate to β-alanine. Mutational studies show the enzyme is essential for βalanine synthesis in Thermococcus kodakarensis and perhaps in most other Euryarchaeota (Tomita et al., 2014). The T. kodakarensis enzyme binds pyridoxal phosphate weakly (Tomita et al., 2014). This weak binding has the advantage that upon mechanism-based inactivation (John, 1995), the inactivated pyridoxal phosphate molecule can be exchanged for a new molecule and the reaction can proceed.

| PanZ AND THE CONTROVER S IAL CoA PATHWAY ANTIME TABOLITE , N5 -PENT YL PANTOTHENAMIDE (N5 -Pan)
N-substituted pantothenamides are analogues of pantothenic acid (Clifton et al., 1970). This is a large class of well-studied bacterial growth inhibitors that are possible antimicrobial agents (Spry et al., 2008). The best-studied analogue is N5-pentyl pantothenamide (N5-Pan) (Figure 4) which is an especially potent inhibitor of E. coli growth. In N5-Pan, the terminal carboxyl group of pantothenic acid has been condensed with a five-carbon alkyl amine to give an amide with a five-carbon alkyl chain (Figure 4). Later N5-Pan was shown to function extraordinarily well (10-fold faster than pantothenic acid) with the last five enzymes of CoA biosynthesis F I G U R E 3 Structure of the E. coli PanD-PanZ-acetyl-CoA complex (PDB ID 4CRY). Each of the four subunits of the PanD tetramer is bound to an acetyl-CoA-bound PanZ monomer to form the octameric structure shown. In this structure, Pro-PanD has been fully processed to the active enzyme (α-subunit) hence the β subunit "noodles" are seen. The green spheres are Mg atoms bound to the CoA phosphate groups. When viewed edge on, the structure is nearly planar. The salmon, brown, green, and cyan proteins are PanZ. The other four (interior) proteins are PanD.
(CoaA through CoaE) to give an analogue of CoA having a threecarbon alkyl chain in place of the thiol (Strauss & Begley, 2002) ( Figure 4). The CoA analogue made from N5-Pan is called ethyldethiacoenzyme A (EtdtCoA).

Virtually all the ACP becomes EtdtACP indicating that essentially all
CoA molecules are replaced by EtdtCoA (Thomas & Cronan, 2010;Zhang et al., 2004). Zhang and coworkers (2004) first showed accumulation of EtdtACP in E. coli. They proposed that growth inhibition was due to accumulation of EtdtACP at the expense of functional ACP which blocked fatty acid synthesis. They postulated that EtdtACP was not a substrate for ACP hydrolase (AcpH), an enigmatic enzyme that removes 4′-phosphopantetheine from ACP (Thomas & Cronan, 2005). However, Thomas and Cronan (2010) showed that EtdtACP was a good substrate for AcpH in vitro and in vivo.
Relative to the wild type strain, deletion of the acpH gene shifted the concentration that blocked growth to 2-fold lower and overproduction of AcpH activity shifted the growth-blocking concentration to 2-fold higher. These modest effects on the sensitivity of E. coli to N5-Pan indicate that accumulation of EtdtACP was a factor, but not a major factor, in N5-Pan growth inhibition. Indeed, the main in vivo effect of N5-Pan was an almost complete loss of acetyl-CoA, the major CoA species under the growth conditions used (Thomas & Cronan, 2010). The lack of acetyl-CoA would prevent synthesis of malonyl-CoA, the building block of fatty acid synthesis, and thereby explain the inhibition of fatty acid synthesis reported by Zhang and coworkers (Zhang et al., 2004). Zhang and coworkers reported that "the CoA level in N5-Pan-treated cells remained unchanged within 2 h of treatment, whereas the level of active ACP dropped to below 10% of the control". However, no data on CoA levels were given and an obsolete analytical method was used (Zhang et al., 2004). These workers buttressed their argument by pointing out that protein synthesis was not inhibited. However, CoA depletion blocks protein synthesis by limiting synthesis of glutamate and other amino acids made from tricarboxylic acid cycle intermediates (Keating et al., 1996), and Zhang and coworkers used a medium replete with amino acids (Zhang et al., 2004). Note that overproduction of each of the last five enzymes of CoA biosynthesis (CoaA through CoaE) ( Figure 4) gave no relief of N5-Pan growth inhibition arguing it is the lack of CoA that blocks growth (Thomas & Cronan, 2010).

F I G U R E 4
Incorporation of N-pentylpantothenamide (N5-Pan) into ethyldethia-CoA (EtdtCoA) and ethyldethia-ACP (Etdt-ACP) (Thomas & Cronan, 2010). Comparison of the 4′-phosphopantetheine and N-pentyl phosphopantothenamide structures illustrates the replacement of the 4′-phosphopantetheine thiol by the alkyl group. The abbreviations are given to the right of the structures. The dashed box shows the β-alanine residue of pantothenate. Arnott and coworkers (2017) found that PanZ bound CoA and acetyl-CoA with equal affinities and hypothesized that PanZ would accommodate the alkyl extension of EtdtCoA because it would mimic the acetyl group of acetyl-CoA. This proved to be the case.
They then asked if β-alanine addition would reverse N5-Pan inhibition of growth of E. coli. The answer was a qualified yes. β-Alanine supplementation reversed N5-Pan growth inhibition up to about 50 μg/mL N5-Pan whereas 2 μg/mL blocked growth in the absence of supplementation. This agreed with prior workers (Mercer et al., 2008) who reported that addition of β-alanine or pantothenate increased the N5-Pan minimum inhibitory concentration from 6 μM (1.4 μg/mL) to 250 μM and 500 μM, respectively. These data argue strongly that inhibition of β-alanine synthesis is the major target of N5-Pan following conversion of the analog to EtdtCoA. However, excess β-alanine does not completely cancel N5-Pan inhibition indicating that EtdtCoA has other targets. One of these seems likely to be its conversion to EtdtACP, although inhibition of CoA/acetyl-CoA enzymes remains possible. In this regard, it would be interesting to overproduce AcpH in the presence of β-alanine supplementation. Arnott and coworkers (2017) reported two methods to assay N5-PanD inhibition of growth. In one set of experiments, serial dilutions of washed cultures of a wild type E. coli strain were spotted on solid minimal medium containing various concentrations of N5-Pan followed by incubation for ~48 h. A parallel set of plates also received β-alanine. The data obtained were straightforward.
Complete inhibition was seen at 2.3 μg/mL that was fully overcome by β-alanine supplementation. This pattern persisted until 18 μg/mL of N5-Pan. Above this concentration (50, 60, 70, 100 μg/mL were tested), β-alanine supplementation failed to reverse inhibition. A strain in which B. subtilis panD replaced E. coli panD gave results that largely paralleled the results obtained for the wild type E. coli strain supplemented with β-alanine. Inhibition above 18 μg/mL N5-Pan was only partially reversed by β-alanine supplementation. When these investigators switched to liquid medium, they surprisingly found that N5-PanD addition did not affect growth rate but rather the point where growth of the culture stops (plateaus). In careful experiments, they showed that the number of cells inoculated (the seeding density) gave a linear relationship with the plateau point. The cause of these atypical results is unclear but seems likely to reflect the large pool sizes of CoA and its esters. CoA is involved in many critical pathways that may have different thresholds and sensitivities to EtdtCoA. What is clear is that N5-Pan inhibition in liquid media must be measured with caution. It seems that use of solid minimal medium provides the most valid assay probably because colony formation requires many generations (about 10 5 for a small colony). In this regard, Zhang and coworkers (2004) reported that pantothenate addition gave only a 2-fold increase in resistance to N5-PanD. However, the N5-Pan concentrations that blocked growth were about 10-fold higher than those of Arnott and coworkers (2017). Zhang and coworkers (2004) used an amino acid-containing liquid medium which may have resulted in the greater N5-Pan concentrations needed to block growth (see above) and assayed inhibition with a microtiter plate reader which is generally used to measure growth rates rather than yield.

| DOE S PanZ PL AY A REG UL ATORY ROLE IN CoA B I OSYNTHE S IS?
Overproduction of PanZ inhibits β-alanine synthesis and traps the pyruvoyl moiety as the inactive ketone hydrate (Arnott et al., 2017).
Whether this striking effect happens in vivo is debatable since the cellular content of PanZ seems considerably less than that of PanD.
Mass spectral analyses collected in the PaxDB database (Wang et al., 2012) argue that PanZ levels are 3 to 6-fold lower than those of PanD. Ribosome profiling gives much lower values; 14 to 40-fold less PanZ than PanD (Li et al., 2014). Although two hybrid interaction assays Stuecker, Hodge, et al., 2012) give qualitative indications that PanD and PanZ interact in vivo, this does not seem a stable interaction. PanD purified by gentle procedures is often pro-PanD when analyzed in vitro and PanZ addition is required to obtain active PanD Stuecker, Hodge, et al., 2012;Stuecker, Tucker, et al., 2012). These considerations aside it is striking that in cases where PanD was overexpressed in strains containing only the chromosomal panZ gene, an appreciable amount of PanD was the processed species (Arnott et al., 2017;Mo et al., 2018;Ramjee et al., 1997). These results argue that following facilitation of maturation of one PanD monomer, PanZ dissociates from that monomer and matures another monomer. In vitro studies of the dissociation of PanD-PanZ complexes should be performed.
The activation of PanD by PanZ-CoA has been proposed to regulate synthesis of CoA and acetyl-CoA (Arnott et al., 2017;Stuecker, Tucker, et al., 2012). This may be the case but if so, the regulation is very weak or absent because E. coli has long been known to excrete pantothenate (Davis, 1950). Indeed, E. coli produces and excretes 15fold more pantothenate (hence 15-fold more β-alanine, Figure 1) than the amount required for CoA synthesis (Jackowski & Rock, 1981).
Cross-feeding experiments indicate that S. enterica also excretes pantothenate (LaRossa & Van Dyk, 1989). E. coli and S. enterica express a pantothenate transporter (PanF) so the vitamin can potentially be recovered from culture media. However, in the gut environment or in open water, recovery seems unlikely. Note that B. subtilis which produces active PanD by unassisted autocatalysis excretes pantothenate and lacks an efficient pantothenate transporter (Baigori et al., 1991).
The apparent lack of control of the rate of pantothenate synthesis in E. coli can be rationalized. The pathway requires only a few moderately expressed enzymes (Li et al., 2014), and the amount of cellular material consumed in synthesis of pantothenate is much lower than those of the amino acids from which pantothenate is derived ( Figure 1). Synthesis of pantoate consumes only ~5% of the amount of the α-ketoisovalerate consumed in valine and leucine synthesis whereas the synthesis of β-alanine consumes <2% of the cellular aspartate. Hence, the metabolic cost of regulating pantothenate synthesis could exceed that of the seemingly wasteful overproduction. There are two E. coli pyruvyl enzymes in addition to PanD, the speD encoded S-adenosylmethionine decarboxylase, an enzyme of polyamine synthesis (Tabor & Tabor, 1987) and the psd encoded phosphatidylserine decarboxylase of phospholipid synthesis (Li & Dowhan, 1990). Unlike PanD, the primary translation products of both speD and psd are rapidly processed to the pyruvyl enzymes in vitro. Full processing of SpeD requires 20 min (Tabor & Tabor, 1987), whereas Psd requires only 10 min (Li & Dowhan, 1990).
Assistance by a PanZ-like protein or other factors is not required. Therefore, in E. coli PanD is alone in requiring another protein for activation.

| CON CLUS IONS
What benefit does E. coli derive from the requirement for the PanZ-CoA/acetyl-CoA complex to mature PanD? The proposed tight coupling of β-alanine synthesis and CoA synthesis early in development of CoA synthesis seems the only plausible explanation for the PanZ-CoA/acetyl-CoA requirement. It may be possible to obtain E.
coli PanD proteins that no longer require PanZ for efficient maturation. H. pylori PanD efficiently undergoes autocatalysis in vitro (Mo et al., 2018). The sequences of the E. coli and Helicobacter pylori PanDs are 39% identical and the α-carbons of the crystal structures of the E. coli and H. pylori PanDs can be superimposed to 1.3 Å (Lee & Suh, 2004). Both the cleavage site and the tyrosine residue that provides the proton in the cleavage reaction are conserved in the two proteins. Hence, it may not require large numbers of amino acid residue substitutions in E. coli PanD to relieve the requirement for PanZ.

AUTH O R CO NTR I B UTI O N S
John E. Cronan: Conceptualization; Investigation; Writing-original draft; Writing-review & editing; Formal analysis; Data curation.

ACK N O WLE D G E M ENTS
Contributions from this laboratory were supported by National Institutes of Health grant AI15650 from the National Institute of Allergy and Infectious Diseases.

E TH I C S S TATEM ENT
This review does not include studies involving people, medical records, or human subjects.

FU N D I N G S TATEM ENT
U.S. Department of Health and Human Services >National Institutes of Health >National Institute of Allergy (AI15650).

CO N FLI C T O F I NTE R E S T S TATE M E NT
The author declares that he has no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in PubMed at https://pubmed.ncbi.nlm.nih.gov/. These data were derived from the following resources available in the public domain: PubMed, https://pubmed.ncbi.nlm.nih.gov/.