An Unexpected Role for the Periplasmic Phosphatase PhoN in the Salvage of B₆ Vitamers in Salmonella enterica

Nutrient salvage is a strategy used by species across domains of life to conserve energy. Many organisms are unable to synthesize all required metabolites de novo and must rely exclusively on salvage.

ABSTRACT

Pyridoxal 5′-phosphate (PLP) is the biologically active form of vitamin B₆, essential for cellular function in all domains of life. In many organisms, such as Salmonella enterica serovar Typhimurium and Escherichia coli, this cofactor can be synthesized de novo or salvaged from B₆ vitamers in the environment. Unexpectedly, S. enterica strains blocked in PLP biosynthesis were able to use exogenous PLP and pyridoxine 5′-phosphate (PNP) as the source of this required cofactor, while E. coli strains of the same genotype could not. Transposon mutagenesis found that phoN was essential for the salvage of PLP and PNP under the conditions tested. phoN encodes a class A nonspecific acid phosphatase (EC 3.1.3.2) that is transcriptionally regulated by the PhoPQ two-component system. The periplasmic location of PhoN was essential for PLP and PNP salvage, and in vitro assays confirmed PhoN has phosphatase activity with PLP and PNP as substrates. The data suggest that PhoN dephosphorylates B₆ vitamers, after which they enter the cytoplasm and are phosphorylated by kinases of the canonical PLP salvage pathway. The connection of phoN with PhoPQ and the broad specificity of the gene product suggest S. enterica is exploiting a moonlighting activity of PhoN for PLP salvage.

IMPORTANCE Nutrient salvage is a strategy used by species across domains of life to conserve energy. Many organisms are unable to synthesize all required metabolites de novo and must rely exclusively on salvage. Others supplement de novo synthesis with the ability to salvage. This study identified an unexpected mechanism present in S. enterica that allows salvage of phosphorylated B₆ vitamers. In vivo and in vitro data herein determined that the periplasmic phosphatase PhoN can facilitate the salvage of PLP and PNP. We suggest a mechanistic working model of PhoN-dependent utilization of PLP and PNP and discuss the general role of promiscuous phosphatases and kinases in organismal fitness.

INTRODUCTION

Vitamin B₆ is a collective term for six compounds, including pyridoxal (PL), pyridoxine (PN), pyridoxamine (PM), pyridoxal 5′-phosphate (PLP), pyridoxine 5′-phosphate (PNP), and pyridoxamine 5′-phosphate (PMP) (Fig. 1). Among the six vitamers, PLP is the biologically active form that plays a critical role in prokaryotic, eukaryotic, and archaeal metabolism. PLP can be synthesized de novo via the deoxyxylulose 5′-phosphate (DXP)-independent pathway, which exists in Saccharomyces cerevisiae and Bacillus subtilis, among other organisms (1, 2). Alternatively, the DXP-dependent pathway is used to synthesize PLP (Fig. 1) in gammaproteobacteria, such as Salmonella enterica serovar Typhimurium (referred to hereafter as S. enterica) and Escherichia coli. In addition to de novo synthesis, PLP can be salvaged from exogenous pyridoxal (PL), pyridoxine (PN), and pyridoxamine (PM) from the environment (Fig. 1). Further, some Pseudomonas, Arthrobacter, and Mesorhizobium loti species can utilize PN or PL as a sole source of carbon and nitrogen using two different catabolic pathways (3).

FIG 1.

Synthesis and salvage of pyridoxal 5′-phosphate (PLP). In S. enterica and E. coli, PLP is synthesized from erythrose-4-phosphate and 1-deoxy-d-xylulose 5-phosphate or salvaged from other B₆ vitamers in the environment (1, 2). E. coli can convert glycolaldehyde (GA) to 4-phosphohydroxy-l-threonine and thus bypass the early steps in PLP synthesis (4). The relevant pathways are shown schematically, and gene products that catalyze each step are indicated beside the arrows. The dashed arrow under PdxI reflects the absence of this enzyme from S. enterica (13).

For organisms relying on the DXP-dependent pathway, a mutation in pdxJ (encoding a PNP synthase, EC 2.6.99.2) blocks de novo PLP biosynthesis and prevents growth in the absence of an exogenous B₆ vitamer (Fig. 1). In contrast, a lesion in pdxB (encoding an erythronate-4-phosphate dehydrogenase, EC 1.1.1.290) in E. coli can be bypassed by at least three mechanisms, one of which utilizes glycolaldehyde (GA) (4). Many organisms, including humans, lack the ability to synthesize PLP de novo and depend on salvage to obtain this cofactor (1, 2). The canonical vitamin B₆ salvage pathway includes a PNP/PMP oxidase (PdxH; EC 1.4.3.5), which also catalyzes the final step of DXP-dependent PLP biosynthesis (Fig. 1) and is conserved among many bacteria and eukaryotes (1, 2). Although both PNP and PMP can serve as substrates for this enzyme, PdxH from E. coli was reported to prefer PNP to PMP (5), while the PdxH homolog from rabbit liver was shown to not favor one over the other (6). Additional enzymes ascribed a role in vitamin B₆ salvage include PdxK (PN/PL/PM kinase; EC 2.7.1.35) (7) and PdxY (PL kinase; EC 2.7.1.35) (8). In vitro, PdxK from E. coli phosphorylated the three unphosphorylated B₆ vitamer substrates, while PdxY had low activity and was specific for PL (8, 9). In addition, PdxK could act on 4-amino-5-hydroxymethyl-2-methylpyrimidine, a precursor of thiamine (vitamin B₁) (10), highlighting the promiscuity often found in kinase and phosphatase enzymes.

S. enterica and E. coli synthesize PLP de novo via the DXP-dependent pathway and share most enzymes attributed to vitamin B₆ salvage (Fig. 1). In addition to PdxK and PdxY, both species have a cytoplasmic PLP phosphatase (EC 3.1.3.74), encoded by ybhA (11). Mutations in YbhA were identified in E. coli strains that evolved to overcome a defect in de novo PLP synthesis (12), suggesting a role of this enzyme in maintaining intracellular PLP pools. Beyond these enzymes, there are marked differences between S. enterica and E. coli in vitamin B₆ salvage. For instance, the E. coli genome encodes a PL reductase (PdxI; EC 1.1.1.65), which has the ability to convert PL to PN and is absent from S. enterica (13). Further, expression of pdxK in S. enterica is controlled by PtsJ, a MocR-like repressor that is not present in E. coli (14). These differences at a genomic level may provide insights into diverse environmental conditions encountered by S. enterica and E. coli as well as their distinct metabolic strategies to thrive. Adding to the differences between these species, this study uncovered an unexpected role for the periplasmic phosphatase PhoN, which is unique to S. enterica, in the salvage of phosphorylated B₆ vitamers.

RESULTS

S. enterica salvages diverse B₆ vitamers.

The capacity of S. enterica to salvage B₆ vitamers and the contribution of the known salvage enzymes was assessed. Strains defective in de novo PLP synthesis (pdxJ) and derivatives lacking either pdxK, pdxY, or pdxH were grown in minimal NCE glucose supplemented with PL, PN, or PM to a final concentration of 1 μM (Fig. 2). As expected, none of the strains grew in the absence of a B₆ vitamer (data not shown). Addition of PL allowed growth of all mutants (Fig. 2A), consistent with phenotypes reported for E. coli strains defective in PLP synthesis (7, 8). Further, strains lacking pdxK or pdxH failed to grow when PN was provided (Fig. 2B). PM allowed growth of the pdxJ pdxH mutant but not the pdxJ pdxK strain (Fig. 2C), a feature attributed to PLP formation via transamination reactions (15).

FIG 2.

S. enterica mutants lacking pdxJ can salvage B₆ vitamers. Strains lacking pdxJ and designated salvage genes were grown in minimal NCE glucose supplemented with a B₆ vitamer (1 μM) as indicated. Data were obtained from two independent experiments with three biological replicates each. Error bars show standard deviations from the means. Abbreviations: PL, pyridoxal; PN, pyridoxine; PM, pyridoxamine.

Unexpectedly, exogenous PLP and PNP (at 1 μM) allowed growth of S. enterica strains lacking de novo PLP synthesis (Fig. 3). At physiological pH, the phosphate group of these vitamers carries a −2 net charge, suggesting a transporter and/or facilitated diffusion are needed for them to cross the outer and/or inner membrane of Gram-negative bacteria. If transported to the cytoplasm, PLP would be available for immediate use, while PNP must be oxidized by PdxH to generate the active cofactor. This simple scenario, in which salvage of PNP would be independent of PdxK and PdxY, was eliminated by assessing growth of the multiple mutant strains described above. Four strains lacking pdxJ grew to full density with PLP (Fig. 3A). In contrast, growth of the pdxJ pdxK and pdxJ pdxH strains was compromised when PNP was provided as a source of vitamin B₆ (Fig. 3B). The pdxJ pdxK mutant strain had an extended lag before growth resumed, while the pdxJ pdxH strain failed to grow. PMP was not able to support growth of any of the PLP-requiring strains (Fig. 3C) and was not considered in subsequent experiments. In medium supplemented with PMP, all strains displayed initial growth similar to that of the pdxJ pdxK and pdxJ pdxH strains in medium with PNP (Fig. 3B), suggesting that some amount of vitamin B₆ from preculture medium was carried over to the minimal NCE medium.

FIG 3.

PLP and PNP support growth of pdxJ and pdxB mutants. S. enterica strains lacking pdxJ and different salvage enzymes (A to C) and those lacking pdxB and salvage enzymes (D to F) were grown in minimal NCE glucose supplemented with a phosphorylated B₆ vitamer (1 μM) or glycolaldehyde (1 mM) as indicated. Data were obtained from three biological replicates, and error bars represent standard deviations from the means. Abbreviations: PLP, pyridoxal 5′-phosphate; PNP, pyridoxine 5′-phosphate; PMP, pyridoxamine 5′-phosphate; GA, glycolaldehyde.

The growth phenotypes presented above suggested that PdxK and/or PdxY, in addition to PdxH, played a role in the salvage of PLP/PNP. The redundancy of PdxK and PdxY could not be tested in a pdxJ background, since the triple mutant would be unable to synthesize or salvage PLP (Fig. 1). Instead, pdxB strains lacking one or both kinases were grown in minimal NCE glucose with supplementation of PLP/PNP (Fig. 3D and E). As expected, the growth phenotypes of the pdxB pdxK and pdxB pdxY strains with exogenous PLP or PNP were similar to those of the pdxJ pdxK and pdxJ pdxY strains, respectively. Importantly, the pdxB strain carrying mutations in both kinases was not viable when PLP (Fig. 3D) or PNP (Fig. 3E) was provided. S. enterica pdxB strains were able to utilize GA for growth (Fig. 3F), as shown in E. coli (4). Taken together, the data suggest that under the conditions tested, (i) the salvage of PLP/PNP requires at least one functional kinase, and (ii) S. enterica encodes a GA-dependent mechanism to bypass part of the DXP-dependent pathway for PLP biosynthesis (Fig. 1).

PhoN facilitates salvage of PLP and PNP.

Transposon mutagenesis identified two loci that were necessary for a pdxJ mutant to use exogenous phosphorylated B₆ vitamers as a source of PLP. Approximately 20,000 colonies of a pdxJ mutant, each carrying a randomly located Tn10(d)Tc insertion, were screened for growth with exogenous PL coupled with the inability to use exogenous PLP. Insertions in colonies with the appropriate growth response were transduced into a fresh genetic background, and the original phenotype was confirmed. Sanger sequencing of nine reconstructed isolates found that eight carried insertions in phoN and one had an insertion in phoP. The phoN gene encodes a nonspecific acid phosphatase (EC 3.1.3.2) in the periplasm and is regulated by the PhoPQ two-component system (16–18). phoQ encodes a histidine sensor kinase (EC 2.7.13.3) in an operon and is located downstream of phoP, which encodes a transcriptional response regulator. Deletions of phoN, phoP, and phoQ were individually placed in a pdxJ background, and growth phenotypes with PLP and PNP were assessed (Table 1). Growth of the pdxJ strain with phosphorylated B₆ vitamers was eliminated by each of the three deletions. Significantly, providing phoN in trans restored growth of all of the double mutants (pdxJ phoN, pdxJ phoP, and pdxJ phoQ) when exogenous PLP or PNP was added to the medium. Growth of the pdxJ phoQ mutant was also restored by expressing phoQ in trans. In contrast, expression of phoP in trans failed to allow growth of the pdxJ phoP mutant with PLP and PNP, while expressing both phoP and phoQ restored growth. In total, these data supported the conclusions that (i) PhoN was necessary and sufficient for salvage of PLP or PNP, (ii) the phoP mutation was polar on phoQ, and (iii) the role of phoPQ was regulatory. Consistent with this, heterologous expression of phoN (pDM1603; Fig. 4A) in an E. coli pdxJ mutant conferred growth on both phosphorylated B₆ vitamers, while the strain with the vector-only control failed to grow (Fig. 4B and C). This result emphasized a critical role for PhoN, which is not encoded by the E. coli genome (19), in the salvage of PLP/PNP under the conditions tested.

TABLE 1.

Growth of pdxJ mutant derivatives^a

Strain	pCV1	Final OD650b
		PLP	PNP
pdxJ	VOC	0.60 ± 0.01	0.62 ± 0.02
pdxJ phoN	VOC	0.10 ± ≤0.01	0.08 ± ≤0.01
	phoN	0.61 ± ≤0.01	0.58 ± 0.01
pdxJ phoP	VOC	0.09 ± ≤0.01	0.08 ± ≤0.01
	phoN	0.58 ± ≤0.01	0.57 ± 0.01
	phoP	0.13 ± 0.01	0.08 ± ≤0.01
	phoQ	0.09 ± ≤0.01	0.08 ± ≤0.01
	phoPQ	0.60 ± ≤0.01	0.25 ± 0.14
pdxJ phoQ	VOC	0.09 ± ≤0.01	0.08 ± ≤0.01
	phoN	0.58 ± 0.01	0.57 ± 0.01
	phoQ	0.50 ± 0.01	0.44 ± 0.12
	phoPQ	0.55 ± 0.02	0.52 ± 0.03

^aS. enterica pdxJ strains carrying an empty pCV1 vector control (VOC) or pCV1 expressing different gene constructs were grown in minimal NCE glucose supplemented with 1 μM PLP or PNP and induced with 0.02% arabinose. The final growth yield is shown.

^bFinal optical density at 650 nm (OD₆₅₀) was determined after 24-h incubation at 37°C. Data were obtained from two independent experiments with three biological replicates each. Representative result from one experiment is shown. Abbreviations: PLP, pyridoxal 5′-phosphate; PNP, pyridoxine 5′-phosphate.

FIG 4.

Periplasmic location is required for the in vivo role of PhoN in PLP and PNP salvage. (A) Relevant PhoN constructs and their general structures are depicted. Each was cloned into pCV1 at the BspQI sites. (B to E) E. coli pdxJ (B and C) and S. enterica pdxJ phoN (D and E) strains carrying the designated vectors were grown in minimal NCE glucose supplemented with 1 μM PLP or PNP and 0.02% arabinose to induce expression. Abbreviations: P_BAD, arabinose-inducible promoter; RBS, ribosomal binding site; ssPhoN, signal sequence of S. enterica PhoN; ssTorA, signal sequence of E. coli TorA; Δ(1-20)PhoN, mature S. enterica PhoN; PLP, pyridoxal 5′-phosphate; PNP, pyridoxine 5′-phosphate.

A periplasmic location is required for PhoN function in vivo.

The first 20 amino acids of PhoN make up a putative Sec-dependent signal sequence directing PhoN to the periplasm, as predicted by SignalP 5.0 (20). The DNA sequence encoding these 20 codons was removed to generate a plasmid that expressed the mature PhoN (pDM1606; Fig. 4A). Plasmid pDM1606 did not support growth of a pdxJ mutant of E. coli or a pdxJ phoN mutant of S. enterica on PLP or PNP (Fig. 4B to E). Sequence encoding the Tat-dependent signal peptide from E. coli TorA (trimethylamine-N-oxide reductase 1; EC 1.7.2.3) (21) was fused to phoN lacking the Sec-dependent signal sequence to generate plasmid pDM1607 (Fig. 4A). Plasmid pDM1607 supported growth of the E. coli pdxJ and S. enterica pdxJ phoN strains in minimal medium with added PLP or PNP to an extent similar to that of the wild-type PhoN (Fig. 4B to E). Taken together, these results demonstrated the periplasmic location of PhoN was required for function in vivo under the conditions tested.

PhoN has phosphatase activity on PLP and PNP in vitro.

PhoN was previously characterized for its phosphatase activity on the nonphysiological substrate p-nitrophenyl phosphate (p-NPP) (22). Results described above suggested PhoN would act on phosphorylated B₆ vitamers. PhoN-6×His was overexpressed in E. coli and purified by nickel-affinity and size-exclusion column chromatography. Purified PhoN-6×His ran as two major bands in the range of ∼30 kDa on SDS-PAGE gel (Fig. 5A), which is consistent with the predicted molecular mass of pre-PhoN (30.1 kDa) and mature PhoN (27.7 kDa). Peptide mass fingerprinting identified both protein bands as S. enterica PhoN. The protein characterized was 89% pure and contained both pre-PhoN and mature PhoN, present at a ratio of ∼3:1.

FIG 5.

PhoN has phosphatase activity on PLP. (A) After purification, PhoN protein was separated on a 14% SDS-PAGE gel and stained with Coomassie blue. Two major bands were visible at ∼30 kDa, corresponding to the molecular mass of pre-PhoN and mature-PhoN. (B) Michaelis-Menten kinetics of PhoN were determined by plotting the initial velocity of PL formation at 37°C against various PLP concentrations. Reaction mixtures contained 50 mM triethanolamine-HCl (pH 7.4) and 50 nM PhoN in addition to PLP. Data were obtained from two independent experiments with at least three technical replicates, and error bars are shown. Abbreviations: PLP, pyridoxal 5′-phosphate; PL, pyridoxal.

PhoN was assayed with PLP as a substrate by monitoring the disappearance of PLP over time as a decrease in absorbance at 390 nm. The initial velocity of each reaction was determined with PLP concentrations from 0 to 1.5 mM (see Fig. S1 in the supplemental material). The K_m and k_cat for PLP, derived from the Michaelis-Menten saturation curves, was 0.35 ± 0.03 mM and 27.19 ± 0.71 s⁻¹, respectively (Fig. 5B). As a reference, another PLP phosphatase, YbhA (Fig. 1), has similar binding affinity for PLP (K_m = 0.37 ± 0.05 mM), while its catalytic turnover (k_cat = 1.0 ± 0.04 s⁻¹) is ∼27 times slower than that of PhoN (23). Unlike PLP, hydrolysis of PNP and PMP cannot be easily monitored spectrophotometrically. The ability of PhoN to hydrolyze these vitamers was determined by endpoint assays using 50 nM PhoN with PLP, PNP, PMP, and p-NPP provided as substrates at 0.1 mM. PhoN had significant activity on each of the provided substrates with the exception of PMP (Table 2). PhoN had the highest specific activity with the generic phosphatase substrate p-NPP, followed by PLP and PNP. Under the conditions tested, PhoN exhibited only low activity with PMP, consistent with the inability of PMP to support growth of a pdxJ mutant (Fig. 3C).

TABLE 2.

Phosphatase activity of PhoN

Substratea	Sp actb (μmol min−1 mg−1)	P valuec
PLP	9.40 ± 1.52	<0.0001
PNP	2.93 ± 0.52	<0.0001
PMP	0.034 ± 0.004	0.04
p-NPP	18.43 ± 1.50	<0.0001

^aAbbreviations: PLP, pyridoxal 5′-phosphate; PNP, pyridoxine 5′-phosphate; PMP, pyridoxamine 5′-phosphate; p-NPP, p-nitrophenyl phosphate.

^bActivity was measured in 50 mM triethanolamine-HCl (pH 7.4) at 37°C. Data were obtained from two independent experiments with at least three technical replicates each.

^cStatistical significance with respect to no enzyme control was determined by two-tailed unpaired Student's t test using GraphPad Prism 7.0c (GraphPad Software, La Jolla, CA).

How does a phosphatase contribute to generating PLP required for growth?

The combination of growth phenotypes and in vitro assays convincingly showed that PhoN can participate in the salvage of PLP and PNP in S. enterica and E. coli. The requirement for characterized kinases in PhoN-dependent salvage suggested that the phosphorylated vitamers were not transported across the inner membrane. A working model depicting incorporation of exogenous PLP is shown in Fig. 6. The model has several tenets based on the data: (i) PhoP and PhoQ are required (likely indirectly [24]) for expression of phoN (Fig. 6A); (ii) the Sec-dependent system translocates PhoN to the periplasm (Fig. 6B), where it is folded to the mature form (Fig. 6C); (iii) PLP crosses the outer membrane, perhaps through a porin (Fig. 6D); (iv) PL produced by PhoN (Fig. 6E) crosses the inner membrane (Fig. 6F); and (v) PdxK and/or PdxY generate PLP to complete salvage (Fig. 6G). Salvage of PNP is viewed as similar, with an additional step of converting PNP to PLP by PdxH (Fig. 1).

FIG 6.

Working model of PhoN-dependent PLP/PNP salvage pathway in S. enterica. (A) Expression of phoN requires that the PhoQ sensor kinase phosphorylates the PhoP response regulator. (B and C) The PhoN peptide is translocated to the periplasm via the Sec-dependent system (B), where it is processed to the mature form (C). (D and E) PLP enters the periplasm, possibly through a generic porin(s) (D), and PhoN dephosphorylates PLP to PL (E). (F and G) Via an unidentified mechanism, PL is transported into the cytoplasm (F), where PdxK/PdxY phosphorylates PL back to PLP (G). PNP salvage mirrors that of PLP but requires an additional step in which PNP is oxidized to PLP by PdxH (Fig. 1). Abbreviations: PLP, pyridoxal 5′-phosphate; PNP, pyridoxine 5′-phosphate; PL, pyridoxal. Image was created with BioRender (Toronto, ON, Canada).

Promiscuous phosphatases and kinases permeate vitamin B₆ salvage.

Although PhoN was required to utilize PLP or PNP in the experiments described above, growth of both E. coli and S. enterica pdxJ mutants was independent of PhoN when the vitamers were provided at 10-fold higher levels (10 μM) (Fig. 7). There are a number of formal possibilities that could explain this result. At high concentrations, (i) PLP and PNP may trickle into the cytoplasm, (ii) PLP and PNP may be dephosphorylated by other promiscuous, but low-efficiency, phosphatases, and/or (iii) sufficient spontaneous hydrolysis to PL and PN may occur to support growth. These possibilities were not distinguished by the work described here, as they were thought to be less relevant to physiological conditions that would be encountered by the enteric bacteria than the lower concentrations.

FIG 7.

At high concentration, PLP and PNP are utilized independently of PhoN. (A and B) E. coli pdxJ (A) and S. enterica pdxJ phoN (B) strains carrying an empty pCV1 vector control were grown in minimal NCE glucose supplemented with 1 μM to 10 μM PLP or PNP and induced with 0.02% arabinose. Data were obtained from two independent experiments with three biological replicates each. Error bars show standard deviations from the means. Abbreviations: PLP, pyridoxal 5′-phosphate; PNP, pyridoxine 5′-phosphate.

Substrate promiscuity was also observed for kinases in the vitamin B₆ salvage pathway. An S. enterica pdxJ pdxK mutant was able to grow when 10 μM PN was provided (Fig. 8A), while 10 μM PM had no effect (data not shown). No traces of other B₆ vitamers in minimal medium supplemented with 10 μM PN were detected by high-performance liquid chromatography (HPLC) (data not shown), suggesting that another kinase(s) can phosphorylate PN but not PM. Growth with 10 μM PN was eliminated in a pdxB background with mutations in both PdxK and PdxY (Fig. 8B), indicating the promiscuous activity was due to the latter. It was not established if E. coli PdxY exhibits similar substrate promiscuity, since this enzyme only showed low kinase activity toward PN (9), and interpretation of in vivo function of PdxY was complicated by possible contamination of PL in growth medium supplemented with high PN concentration (100 μM) (8).

FIG 8.

At high concentration of PN, activity of PdxY allows growth of S. enterica strains blocked in de novo PLP synthesis. S. enterica pdxJ (A) and pdxB (B) strains were grown in minimal NCE glucose supplemented with 10 μM PN. Data were obtained from two independent experiments with three biological replicates each. Error bars represent standard deviations from the means. Abbreviation: PN, pyridoxine.

DISCUSSION

PhoN was first characterized as a periplasmic nonspecific acid phosphatase in S. enterica (16). PhoN activity increased in cells grown with limiting carbon, nitrogen, phosphorus, or sulfur (17), and expression of phoN was indirectly regulated by the PhoPQ two-component system (18, 24). Prior to this study, the sole phenotype of a phoN mutant was a slight growth defect when α-naphthyl phosphate or phenyl phosphate was used in phosphorus-limiting minimal glucose medium (18).

Results here showed that PhoN, in combination with the relevant kinases, allowed S. enterica to salvage PLP and PNP. When vitamin B₆ homeostasis is perturbed, some bacteria, such as E. coli and S. enterica, can export excess phosphorylated B₆ vitamers into the environment (25). From an ecological standpoint, the ability to salvage phosphorylated forms of vitamin B₆ excreted from other community members may provide a fitness advantage to S. enterica. Counterintuitively, salvage of PLP and PNP required a phosphatase and a kinase, which resulted in the working model described in Fig. 6. This model is based on data that (i) cells were unable to efficiently take up phosphorylated forms of vitamin B₆ (26), (ii) PhoN dephosphorylated PLP and PNP in vitro (Table 2), and (iii) utilization of PLP/PNP required a functional kinase (PdxK or PdxY) (Fig. 3D and E). Heterologous expression of phoN allowed an E. coli pdxJ mutant to grow with PLP/PNP, making it unlikely that there is a dedicated transport system for PLP/PNP (Fig. 6D). Unphosphorylated B₆ vitamers were suggested to be taken up via facilitated diffusion (Fig. 6F) followed by phosphorylation to trap the molecule (26–28) (Fig. 6G). Additional work is required to rigorously define the mechanism of transport and the transporter(s)/porin(s).

An S. enterica pdxJ pdxK mutant grew with 1 μM PNP (Fig. 3B) but not PN (Fig. 2B) as a source of PLP. In the context of our working model, this result was unexpected, since the salvage of PNP goes through a PN intermediate. In the absence of PdxK, salvage of both PN and PNP is dependent on PdxY (Fig. 8B and 3E); thus, a simple scenario suggests that expression of pdxY is higher in the presence of PNP than PN. Additional experiments are needed to address this hypothesis. It was noted that the growth data here appeared inconsistent with an ³H-labeling study showing that S. enterica did not take up PM (27). A likely explanation for the two results is that the labeling experiment was done at a low concentration of PM (0.1 μM) over a period of 10 min, while the pdxJ mutant strains showed exponential growth with 10-fold more PM (1 μM PM). This interpretation was supported by the finding that PM at 0.1 μM did not support growth of a pdxJ mutant (data not shown).

Substrate promiscuity is a common theme in vitamin B₆ metabolism, where it has been demonstrated for enzymes, including PhoN (here) (16), PdxH (5, 6), PdxK (9, 10), and YbhA (23). PhoN, specifically, had various degrees of activity toward p-NPP, glucose 6-phosphate, β-glycerophosphate, AMP (16), PLP, and PNP (Table 2). It is notable that while both PhoN and YbhA are phosphatases that have the ability to hydrolyze PLP to PL, different cellular localization impacts their respective physiological roles, periplasmic PhoN in salvage and cytoplasmic YbhA in homeostasis of PLP (11, 12). The lack of substrate specificity and unclear regulation suggest that PhoN is not a dedicated salvage enzyme but rather participates when PLP or PNP is in the environment. When PLP or PNP is present at a high level, other promiscuous phosphatases, such as alkaline phosphatase (encoded by phoA in E. coli) (26), can replace PhoN in salvage (Fig. 7). In addition, we provided evidence supporting that PdxY, previously designated a PL kinase, can also use PN as the substrate at high concentration (Fig. 8). In total, the contribution of PhoN to PLP/PNP salvage defined here highlights how enzymes can be recruited to contribute to a robust and adaptive metabolism that allows microbes to thrive in different environments.

MATERIALS AND METHODS

Strains, media, and chemicals.

Strains and plasmids used in this study are shown in Table 3. All strains are derivatives of S. enterica Typhimurium LT2 or E. coli BW25113. Plasmids were propagated in E. coli DH5α (Invitrogen, Carlsbad, CA). Recombinant protein was purified from E. coli BL21AI (Invitrogen, Carlsbad, CA).

TABLE 3.

Strains and plasmids

Strain or plasmid	Description	Sourcea
S. enterica LT2
DM15906	pdxJ662::Km	This study
DM15964	pdxJ664	This study
DM16354	pdxJ664 pdxH669::Km	This study
DM17013	pdxJ664 pdxY667::Cm	This study
DM17017	pdxJ664 pdxK672::Km	This study
DM17067	pdxB680	This study
DM17068	pdxB680 pdxY667::Cm	This study
DM17069	pdxB680 pdxK672::Km	This study
DM17080	pdxB680 pdxK672::Km pdxY667::Cm	This study
DM16983	pdxJ664/pCV1	This study
DM16943	pdxJ664 phoN247::Km/pCV1	This study
DM16944	pdxJ664 phoN247::Km/pDM1603	This study
DM16945	pdxJ664 phoP248::Km/pCV1	This study
DM16946	pdxJ664 phoP248::Km/pDM1603	This study
DM16947	pdxJ664 phoP248::Km/pDM1604	This study
DM17007	pdxJ664 phoP248::Km/pDM1605	This study
DM17008	pdxJ664 phoP248::Km/pDM1620	This study
DM16904	pdxJ664 phoQ245::Km/pCV1	This study
DM16905	pdxJ664 phoQ245::Km/pDM1603	This study
DM16906	pdxJ664 phoQ245::Km/pDM1605	This study
DM17006	pdxJ664 phoQ245::Km/pDM1620	This study
E. coli BW25113
DM16897	pdxJ736::Km/pCV1	This study
DM16898	pdxJ736::Km/pDM1603	This study
DM16941	pdxJ736::Km/pDM1606	This study
DM16942	pdxJ736::Km/pDM1607	This study
Plasmids
pCP20	Temp-sensitive plasmid expressing Flp recombinase (Amr, Cmr)	35
pCV1	Modified pBAD24 (Amr)	37
pDM1603	pCV1 expressing S. enterica PhoN (Amr)	This study
pDM1604	pCV1 expressing S. enterica PhoP (Amr)	This study
pDM1605	pCV1 expressing S. enterica PhoQ (Amr)	This study
pDM1606	pCV1 expressing S. enterica Δ(1-20)PhoN (Amr)	This study
pDM1607	pCV1 expressing E. coli TorA signal sequence fused to S. enterica Δ(1-20)PhoN (Amr)	This study
pDM1620	pCV1 expressing S. enterica PhoP and PhoQ (Amr)	This study
pDM1623	pTEV20 expressing S. enterica 6×His-tagged PhoN (Amr)	This study

^aPhage P22 carrying donor DNA of S. enterica Typhimurium 14028s pdxK::Km, phoN::Km, phoP::Km, and phoQ::Km strains from the SGD-K collection (33) were kindly provided by Anna Karls at the University of Georgia (Athens, GA). E. coli BW25113 pdxJ736::Km mutant from the Keio collection (36) was a gift from Jorge C. Escalante-Semerena at the same institution.

Strains were routinely grown at 37°C on rich medium (nutrient broth for S. enterica, 8 g/liter Difco mix, 5 g/liter NaCl; lysogeny broth for E. coli, 10 g/liter Bacto tryptone, 5 g/liter yeast extract, 5 g/liter NaCl) or minimal no-carbon E (NCE) medium supplemented with MgSO₄ (1 mM), trace elements (29), and glucose (11 mM) as a sole carbon source. Super broth (32 g/liter Bacto tryptone, 20 g/liter yeast extract, 5 g/liter NaCl, 5 mM NaOH) was used for protein purification. Agar was added to 1.5% (wt/vol) for solid media. B₆ vitamers in chloride form were supplemented to a final concentration of 1 μM or 10 μM as indicated. Glycolaldehyde (GA) was added to rich and minimal media at 0.1 mM and 1 mM, respectively (8). When appropriate, antibiotics were added at the following concentrations: kanamycin (Km; 50 μg/ml), chloramphenicol (Cm; 20 μg/ml), ampicillin (Am; 150 μg/ml in rich and 15 μg/ml in minimal media), and tetracycline (Tc; 20 μg/ml in rich and 5 μg/ml in minimal media).

Chemicals were purchased from MilliporeSigma (formerly Sigma-Aldrich, St. Louis, MO) unless otherwise stated. p-NPP was obtained from Thermo Fisher Scientific (Waltham, MA). Restriction enzymes and Taq polymerase were purchased from New England BioLabs (Ipswich, MA). Primers were synthesized by Eton Bioscience (San Diego, CA).

Purification of PNP.

PNP was purified as described previously (30), except Dowex 1X8 was used in place of Amberlite IRC84. Briefly, Amberlite IRA743 resins were washed by following a published protocol, while Dowex 1X8 resins were washed sequentially with 200 ml (10 bed volumes [BV]) water, 100 ml (5 BV) 1 M HCl, and 200 ml (10 BV) water. After a 15-min incubation with NaOH to allow dissociation of the borate-PNP complex, the solution was adsorbed onto Amberlite IRA743 (10 ml in a 1.0- by 20-cm column) and washed with 10 ml (1 BV) water at a flow rate of 1 ml/min. Fractions containing PNP (detected spectrophotometrically at 325 nm) were pooled, and the pH was adjusted to 8.5 with 1 M HCl. The solution was adsorbed onto Dowex 1X8 (20 ml in a 1.5- by 15-cm column), washed with 100 ml (5 BV) water, and eluted with 100 ml (5 BV) 1 M HCl. The eluate containing PNP was pooled and analyzed using HPLC as described previously (25) to determine the purity and concentration of PNP. The purified PNP solution was lyophilized and stored at –20°C for subsequent use.

Strain and plasmid construction.

Primers for strain and plasmid construction are listed in Table 4. In-frame deletions of pdx genes (except pdxK) were generated using the λ Red recombinase system (31) adapted for S. enterica. Phage P22 HT105/1 int-201 (32) carrying donor DNA was propagated in S. enterica Typhimurium 14028s pdxK and pho mutants from the SGD-K collection (33). All mutations were reconstructed in appropriate S. enterica Typhimurium LT2 strains via P22 transduction, and phage-free transductants were isolated as described previously (34). Antibiotic markers were resolved using pCP20 (35) when necessary. E. coli BW25113 strains were generated by electroporating different expression plasmids into the parental pdxJ736::Km mutant from the Keio collection (36).

TABLE 4.

Primers

Primer	Sequence (5′ to 3′)	Description or reference
PR1076	AACGCACAGTAAAAACGAAGAAAGATTAACGAGGATTGTCGTGTAGGGCTGGAGCTGCTTC	Inactivation of S. enterica pdxJ
PR1077	GGGCAATCTCTACAATATCCGTTCCCAGGCCGAGAATCGCCATATGAATATCCTCCTTAG
PR1080	GTAACAGGGAGTGAGAAATCACTCCCTTATTTTTGATGTTGTGTAGGCTGGAGCTGCTTC	Inactivation of S. enterica pdxY
PR1081	GGGCAATCTCTACAATATCCGTTCCCAGGCCGAGAATCGCCATATGAATATCCTCCTTAG
PR1374	TTGTCCGTTAACTCTCGTTCTCAAACAGGTACGACAGTCGTGTAGGCTGGAGCTGCTTC	Inactivation of S. enterica pdxB
PR1375	ACACCAAATGCGCCAGTCTTTTCAGCGTCGGTTGATCCAGCATATGAATATCCTCCTTAG
PR1395	GCACAATAGCGCCACCCACTGATTATTTCTGATCAACGCCGTGTAGGCTGGAGCTGCTTC	Inactivation of S. enterica pdxH
PR1396	ATTCATCCGCACCAGTGCTTAAAACAAGATTTTTGCATCTCATATGAATATCCTCCTTAG
PR1320	NNGCTCTTCNTTCATGAAAAGTCGTTATTTAGTATTTTTTCTACC	Construction of pDM1603
PR1321	NNGCTCTTCNTTATCAGTAATTAAGTTTGGGGTGATC
PR1322	NNGCTCTTCNTTCATGATGCGCGTACTGG	Construction of pDM1604
PR1323	NNGCTCTTCNTTATAGCGCAATTCAAAAAGATATCCT
PR1324	NNGCTCTTCNTTCATGAATAAATTTGCTCGCCATTTTC	Construction of pDM1605
PR1325	NNGCTCTTCNTTATTATTCCTCTTTCTGTGTGGGA
PR1328	NNGCTCTTCNTTCATGGAAACAGTGCAACCCTTTC	Construction of pDM1606
PR1321	Listed above
PR1329	NNGCTCTTCNTTCATGAACAATAACGATCTCTTTCAGGCATCACGT	Construction of pDM1607
PR1330	GCACTGTTTCTGCCGCAGTCGCACGTCG
PR1331	CGACGTGCGACTGCGGCAGAAACAGTGCAACCCTTTCATTCTC
PR1332	NNGCTCTTCNTTATCAGTAATTAAGTTTGGGGTGATCTTCTTTACTCAAT
PR1322	Listed above	Construction of pDM1620
PR1325	Listed above
PR1361	NNGCTCTTCNTACATGAAAAGTCGTTATTTAGTATTTTTTCTAC	Construction of pDM1623
PR1362	NNGCTCTTCATTCGTAATTAAGTTTGGGGTGATCTTC
Tn10R	ACCTTTGGTCACCAACGCTTTTCC	Transposon sequencing
Tn10L	TCCATTGCTGTTGACAAAGGGAAT	40, 41
Arb1	GGCCACGCGTCGACTAGTACNNNNNNNNNNGATAT
Arb6	GGCCACGCGTCGACTAGTACNNNNNNNNNNACGCC

Plasmids were constructed by cloning S. enterica pho genes into pCV1 or pTEV20 (37) at BspQI sites by following the described protocol (38). Splicing by the overlap extension (SOE) PCR method (39) was used with slight modifications to create ssTorA-Δ(1-20)PhoN fusion. Briefly, a chromosomal DNA segment encoding the signal sequence of E. coli TorA was amplified for 30 cycles with the PR1329-PR1330 primer pair. A parallel PCR was performed with S. enterica DNA encoding the mature sequence of PhoN using PR1331-PR1332 primers. The generated products had a ∼30-bp overlap and were used as DNA templates for a subsequent round of amplification. The PCR was initiated without primers to allow annealing of the two DNA fragments and enrichment of the fused product. After 5 cycles, PR1329-PR1332 primers were added and the reaction proceeded for another 25 cycles. The fused DNA product was purified using a QIAquick gel extraction kit (Qiagen, Germantown, MD) and cloned into pCV1 as described previously.

Transposon mutagenesis.

A P22 HT105/1 int-201 lysate was generated on a pool of S. enterica cells carrying ∼10,000 independent Tn10d(Tc) insertions and used to transduce a pdxJ662::Km mutant (DM15906) to tetracycline resistance. Tc^r colonies that arose on nutrient agar were replica printed onto minimal NCE glucose plates containing 1 μM PLP or PL as a sole source of vitamin B₆. Colonies that could grow on PL but not PLP were isolated and made phage-free using green indicator agar (34) supplemented with 1 μM PL. Of ∼20,000 colonies screened, 14 putative mutants were identified. The Tn10d(Tc) insertion from each isolate was transduced into the parental pdxJ662::Km background to confirm that the insertion was causative of the growth phenotype. Nine reconstructed mutants showed the expected phenotype and were further analyzed.

Nested colony PCR using degenerate primers (40, 41) (Table 4) was performed to amplify the DNA regions from both ends of the transposon in each reconstructed mutant. PCR products were purified with a QIAquick PCR purification kit (Qiagen, Germantown, MD) and subjected to Sanger sequencing (Eurofins, Louisville, KY). The locations of Tn10d(Tc) insertions were determined by aligning the sequencing results to the S. enterica genome using SnapGene 5.1.4.1 (GSL Biotech LLC, San Diego, CA). Eight insertions were located in phoN (three insertions located at V14, two at Y17, two at G115, and one at A118 of the coding sequence), and one was in phoP between residues Q203 and Y206 of the translated product.

Growth analysis.

S. enterica and E. coli strains were precultured in 2 ml rich medium (supplemented with 1 μM PL for S. enterica pdxJ strains lacking plasmids and 0.1 mM GA for pdxB strains) in an Innova 43 shaker (New Brunswick Scientific, Edison, NJ) at 250 rpm for 6 to 8 h. Cells were pelleted, resuspended in equal volumes of 0.85% NaCl, and inoculated at a 40-fold dilution into minimal NCE glucose medium with or without vitamin B₆ supplementation. In trans expression of pho genes from pCV1 was induced by addition of 0.02% (wt/vol) arabinose as indicated. Growth was monitored in a 96-well plate as a function of the optical density at 650 nm (OD₆₅₀) using a BioTek Elx808 model (BioTek Instruments, Winooski, VT). Data were graphed using GraphPad Prism 7.0c (GraphPad Software, La Jolla, CA). Vitamin B₆ concentrations in growth media were analyzed with HPLC as described previously (25).

Purification of recombinant PhoN.

Freshly transformed E. coli BL21AI/pDM1623 cells were precultured in 10 ml lysogeny broth at 37°C in replicates. The overnight cultures were used to inoculate four flasks, each of which contained 1.5 liters of super broth. Growth was monitored until cells reached an OD₆₅₀ of ∼0.8 as determined by a Spectronic 20D+ instrument (Thermo Fisher Scientific, Waltham, MA). Expression of phoN was induced with 0.02% (wt/vol) arabinose and 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside), and the temperature was shifted to 30°C in an Innova 44 shaker (New Brunswick Scientific, Edison, NJ) at 200 rpm. After incubation for 19 h, cells were pelleted (7,000 × g, 15 min, 4°C) and stored at –80°C until purification. All purification and buffer exchange steps were performed at 4°C. Buffer A (50 mM Tris-HCl, 100 mM NaCl, 10 mM imidazole, pH 7.4) containing lysozyme (1 mg/ml), DNase (0.1 mg/ml), and phenylmethylsulfonyl fluoride (1.5 mM) was added to the cell pellet (2.5 ml/g wet weight). Cells were lysed at 20,000 lb/in² using a Constant Systems Limited One Shot cell disruptor (Northants, United Kingdom). Cell lysate was centrifuged (48,000 × g, 1 h, 4°C) and filtered (0.45-μm polyvinylidene difluoride) before being loaded onto a preequilibrated 5-ml HisTrap HP Ni-Sepharose column (GE Healthcare, Chicago, IL) attached to an NGC Quest 10 chromatography system (Bio-Rad, Hercules, CA). The loaded column was washed with 10 column volumes (CV) of buffer A, followed by 5 CV of 4% buffer B (50 mM Tris-HCl, 100 mM NaCl, 250 mM imidazole, pH 7.4), and eluted with a 20-CV gradient of 4 to 100% buffer B at a flow rate of 2 ml/min. Fractions containing recombinant PhoN were confirmed on 14% SDS-PAGE gel and pooled. The protein solution was concentrated using an Amicon Ultra-15 10K centrifugal filter unit (MilliporeSigma, St. Louis, MO) and exchanged into buffer C (50 mM Tris-HCl, 100 mM NaCl, pH 7.4) using a PD-10 desalting column (GE Healthcare, Chicago, IL). The concentrated protein solution was applied to an ENrich SEC 650 (10 by 300 mm) column (Bio-Rad, Hercules, CA), which was preequilibrated with buffer C. Size-exclusion purification was carried out with buffer C according to the manufacturer’s protocol. Fractions containing PhoN (determined by SDS-PAGE) were pooled. The protein solution was exchanged into buffer D (50 mM Tris-HCl, 100 mM NaCl, 10% glycerol, pH 7.4) using a PD-10 desalting column, flash-frozen with liquid nitrogen, and stored at –80°C until use. Protein concentration was determined using a Pierce bicinchoninic acid protein assay kit (Thermo Fisher Scientific, Waltham, MA) with bovine serum albumin as a standard.

In vitro phosphatase assays.

Purified PhoN was thawed from frozen pellets and diluted in assay buffer before each use. Substrates were freshly prepared in assay buffer prior to each experiment. Enzyme assays were performed in two independent experiments with at least three replicates according to a published protocol (42), with modifications. Briefly, reaction mixtures contained 50 mM triethanolamine-HCl (pH 7.4), 50 nM PhoN, and 0 to 1.5 mM PLP in a total volume of 100 μl. The rate of PL formation was inferred by spectrophotometrically monitoring the decrease in absorbance at 390 nm normalized to a 1-cm path length using a SpectraMax 398 Plus plate reader (Molecular Devices, Sunnyvale, CA). The extinction coefficient of PLP was 5.23 × 10³ M⁻¹cm⁻¹, as determined from a 10-point standard curve with 0 to 2 mM PLP in triplicates. Initial velocity was determined at 37°C for the first 100 to 150 s after adding PhoN protein to initiate the reactions. Kinetic parameters were estimated by fitting the data to the Michaelis-Menten equation using GraphPad Prism 7.0c (GraphPad Software, La Jolla, CA).

Specific activity was determined after incubating the enzyme reaction mixtures containing 50 mM triethanolamine-HCl (pH 7.4), 50 nM PhoN, and 0.1 mM PLP, PNP, PMP, or p-NPP at 37°C for 2.5 min. Hydrolysis of PLP was measured as described above. PN and PM formation was quantified using HPLC following extraction with 7 M HClO₄ containing 15 μM 4-deoxypyridoxine as an internal standard (25). Production of p-nitrophenylate (p-NP) was measured spectrophotometrically at 410 nm with a SpectraMax 398 Plus plate reader (Molecular Devices, Sunnyvale, CA) after addition of 100 μl 0.4 M NaOH to the enzyme reaction mixtures (22). The extinction coefficient of p-NP in assay buffer was 12.75 × 10³ M⁻¹cm⁻¹, as determined from a 9-point standard curve with 0 to 0.4 mM p-NP in triplicates. A no enzyme control was used to account for spontaneous hydrolysis of substrates, which was subtracted from the reported activity.

Data availability.

All relevant data are included in the content of the manuscript.