Involvement of the gapA- and epd (gapB)-Encoded Dehydrogenases in Pyridoxal 5′-Phosphate Coenzyme Biosynthesis in Escherichia coli K-12

Abstract

We show that epd (gapB) mutants lacking an erythrose 4-phosphate (E4P) dehydrogenase are impaired for growth on some media and contain less pyridoxal 5′-phosphate (PLP) and pyridoxamine 5′-phosphate (PMP) than their epd⁺ parent. In contrast to a previous report, we found that gapA epd double mutants lacking the glyceraldehyde 3-phosphate and E4P dehydrogenases are auxotrophic for pyridoxine. These results implicate the GapA and Epd dehydrogenases in de novo PLP and PMP coenzyme biosynthesis.

Pyridoxal 5′-phosphate (PLP) is an essential coenzyme used by many enzymes involved in amino acid metabolism and by glycogen phosphorylases (reviewed in references 7, 12, and 16). PLP is thought to be synthesized in Escherichia coli by the convergence of two pathways (Fig. 1) (10, 18, 19, 25). The two branches lead to the synthesis of 4-phosphohydroxy-l-threonine (4PHT) and 1-deoxy-d-xylulose 5-phosphate, which are condensed by the PdxA and PdxJ enzymes to form pyridoxine 5′-phosphate (PNP) (Fig. 1) (6, 23). PNP is oxidized by the PdxH enzyme to form the active PLP coenzyme (24, 29, 39) (Fig. 1). PLP is converted to pyridoxamine 5′-phosphate (PMP) by the half-reaction of transaminases (4, 12) (Fig. 1). PMP is recycled back to PLP by the second half-reaction of transaminases and by PdxH oxidase (29, 39).

FIG. 1.

Pathway for PLP and PMP coenzyme biosynthesis in E. coli K-12. Enzymes that catalyze the steps in the pathway are indicated by their genetic symbols and are boxed. Branch 1 takes E4P to 4PHT, and branch 2 provides DXP, which is condensed with 4PHT to form PNP. PNP is oxidized to the active coenzyme PLP, which can be converted to PMP by transaminases. Oxidation of E4P to 4PE is the first step of branch 1 and is catalyzed by the E4P dehydrogenase activities of the GapA and Epd (GapB) enzymes. See text for details.

Overwhelming genetic and biochemical evidence implicate 4PHT as an obligatory intermediate that provides the phosphate ester group of PNP (13, 35, 37). 4PHT biosynthesis is thought to start with d-erythrose-4-phosphate (E4P), which is oxidized by an E4P dehydrogenase to 4-phosphoerythronate (4PE) (Fig. 1, branch 1). 4PE is further oxidized by the PdxB dehydrogenase and transaminated by the SerC (PdxF) enzyme to form 4PHT (Fig. 1). Three pieces of evidence support this scheme. First, tktA tktB double mutants, which lack transketolase activity and cannot synthesize E4P or the six aromatic amino acids and vitamins (Fig. 1), are pyridoxine (PN) auxotrophs (38). Second, purified PdxB enzyme oxidizes 4PE in a nonsustained reaction (36). Last, the SerC (PdxF) enzyme uses 4PHT as a substrate in the reverse transamination reaction (13). As expected from this scheme, pdxB and serC (pdxF) mutants are PN auxotrophs, but no single mutation that blocked the first E4P dehydrogenase step in branch 1 of the pathway was identified (10, 23).

Previous studies of epd (gapB).

We proposed and confirmed that the gapB gene, which we renamed epd, encoded a nonphosphorylating E4P dehydrogenase (36). The E4P dehydrogenase activity of the Epd enzyme was verified by Boschi-Muller et al. (5), who further identified amino acids in Epd required for E4P dehydrogenase activity and showed that Epd has low-level phosphorylating and nonphosphorylating glyceraldehyde 3-phosphate (G3P) dehydrogenase activities in the presence and absence of inorganic phosphate, respectively. However, this G3P dehydrogenase activity is not sufficient to allow the growth of gapA mutants, which lack the major G3P dehydrogenase of E. coli (9, 20). Conversely, Boschi-Muller et al. showed that the GapA dehydrogenase of Bacillus stearothermophilus has a low level of phosphorylating E4P dehydrogenase activity, but this result was not extended to the E. coli GapA enzyme (5).

Recently Della Seta et al. (9) reported that E. coli gapB single and gapA gapB double mutants do not show a growth requirement for B₆ vitamers, such as PN and pyridoxal (PL), which can be converted to PLP and PMP by a salvage pathway (34, 35). If this finding were correct, then it would argue against a requirement for an E4P dehydrogenase in PLP biosynthesis (Fig. 1). We performed experiments similar to those of Della Seta et al. using what should be equivalent strains in three different E. coli genetic backgrounds. In contrast to their results, we found that gapA gapB mutants are indeed auxotrophic for B₆ vitamers and that gapB mutants are impaired for growth without PN under some growth conditions.

Construction of gapA and epd mutants.

We moved the prototypic gapA1 point mutation of strain DF220 (20) by cotransduction with a linked Tn10 transposon into prototrophic strains W3110 and MG1655 grown in Luria-Bertani (LB) medium containing 1% (vol/vol) glycerol plus 1% (wt/vol) succinate (Table 1). The resulting MG1655 gapA1 mutant TX4125 had the expected phenotype of growth on plates containing minimal salts medium (MM) supplemented with glycerol plus succinate but no growth in 3 days on MM containing 0.4% (wt/vol) glucose (Table 2) (9, 20). The W3110 gapA1 mutant TX3484 grew very slowly on MM containing glucose, suggesting slight leakiness of the gapA1 mutation in some genetic backgrounds (Table 2); however, leakiness would not lead to the PN auxotrophy described below. Della Seta et al. mentioned low infectivity by P1 bacteriophage and spontaneous lysis of their ΔgapA::Cm mutants (9); we did not encounter similar difficulties with gapA1 mutants.

TABLE 1.

Strains and plasmids

Strain or plasmid	Genotypea	Source or reference
E. coli K-12 strains
CAG12074	MG1655 zea-3068::Tn10	32
DF220b	gap-1 (called gapA1 here to avoid confusion with gapB)	20
DF221b	gap-2 (called gapA2 here to avoid confusion with gapB)	20
DH5α	F′ endA1 hsdR17 supE44 thi-1 recA1 gyrA relA Δ(lacZYA-argE)U169 deoR	New England Biolabs
JC7623	recB21 recC22sbc15 ara arg his leu pro thr	22
MG1655	Prototroph	3
NU816	W3110 lacU169 tna2 sup0 prototroph	C. Yanofsky
TX3446c	JC7623 epd (gapB)::Ω(Cmr)	Transformation with pTX457 cut with SalI and SphI
TX3457c	NU816 epd (gapB)::Ω(Cmr)	NU816 × P1vir(TX3446)
TX3470	NU816 zea-3068::Tn10	NU816 × P1vir(CAG12074)
TX3481c	NU816 epd (gapB)::Ω(Cmr) zea-3068::Tn10	TX3457 × P1vir(CAG12074)
TX3459	DF220 zea-3068::Tn10	DF220 × P1vir(CAG12074)
TX3484	NU816 gapA1 zea-3068::Tn10	NU816 × P1vir(TX3459)
TX3491c	NU816 epd (gapB)::Ω(Cmr) gapA1 zea-3068::Tn10	TX3457 × P1vir(TX3459)
TX4074c	MG1655 epd (gapB)::Ω(Cmr)	MG1655 × P1vir(TX3481)
TX4090	MG1655 zea-3068::Tn10	MG1655 × P1vir(TX3481)
TX4092c	MG1655 epd (gapB)::Ω(Cmr) zea-3068::Tn10	TX4074 × P1vir(TX3481)
TX4125	MG1655 gapA1 zea-3068::Tn10	MG1655 × P1vir(TX3484)
TX4134c	MG1655 epd (gapB)::Ω(Cmr) gapA1 zea-3068::Tn10	TX4074 × P1vir(TX3484)
TX4187	DF221 epd (gapB)::Ω(Cmr)	DF221 × P1vir(TX3481)
Plasmids
pHP45-	Plasmid carrying a Ω(Cmr) cassette; Apr	30
Ω(Cmr)	Cmr
pTX419	pBR322 containing 1.3-kb HindIII-EcoRI epd+ (gapB+) fragment; Apr	36
pTX457c	pTX419 epd (gapB)::Ω(Cmr); Ω(Cmr) cassette from pHP45-Ω(Cmr) cloned into the ClaI site of pTX419; Apr Cmr	This study

^aAp^r and Cm^r, resistance to ampicillin and chloramphenicol, respectively. 

^bContains point mutation in gapA (20). 

^cSee Fig. 2 for location of Ω(Cm^r) insertion in epd (gapB). 

TABLE 2.

Growth deficiencies of gapA⁺ epd::Ω(Cm^r) and gapA1 epd::Ω(Cm^r) mutants on MM lacking PN^a

Strain (genotype)	Supple- mentb	Colony areas (mm2) on MM plates containing the following carbon sourcesc:
		Glu- cose	Glycerol + succi- nate	Casamino Acids	Sodium acetate
TX3470 (W3110 gapA+ epd+)	None	3.9	1.7	0.07	0.29
	PN	3.8	1.7	0.07	0.28
TX3484 (W3110 gapA epd+)	None	0.04	0.14	NT	0.20
	PN	0.03	0.15	NT	0.30
TX3481 (W3110 gapA+ epd)	None	3.7	0.35	NG	0.07
	PN	3.8	1.7	0.07	0.27
TX3491 (W3110 gapA epd)	None	NG	NG	NT	NG
	PN	0.03	0.07	NT	0.13
TX4090 (MG1655 gapA+ epd+)	None	4.4	1.5	1.1	0.48
	PN	4.5	1.5	1.1	0.49
TX4125 (MG1655 gapA epd+)	None	NG	0.79	NT	0.22
	PN	NG	0.79	NT	0.23
TX4092 (MG1655 gapA+ epd)	None	3.6	0.75	0.18	0.40
	PN	4.5	1.5	1.1	0.43
TX4134 (MG1655 gapA epd)	None	NG	NG	NT	NG
	PN	NG	0.79	NT	0.22

^aBacteria were grown in LB containing 1% (vol/vol) glycerol, 1% (wt/vol) succinate (disodium hexahydrate), and all appropriate antibiotics with shaking at 37°C for about 24 h. Bacteria were collected by centrifugation and washed twice with (1 × E) minimal salts at 25°C (8). Washed cells were diluted by 10⁶ in (1 × E) minimal salts, and 150 μl of the cell suspensions was spread onto the (1 × E) plates containing the indicated supplements and carbon sources (no antibiotics). The plates were incubated at 37°C. 

^b2 μM PN was added as indicated. Addition of 0.1 mM GA gave the same results as PN addition (data not shown). 

^cFor each medium, areas were calculated from the diameters of at least 50 well-separated colonies. Experimental errors of colony areas were less than 15%. Totals of 0.4% (wt/vol) glucose, 0.2% (vol/vol) glycerol plus 0.2% (wt/vol) succinate, 0.5% (wt/vol) Difco vitamin assay Casamino Acids, and 0.4% (wt/vol) sodium acetate were included in the plates as carbon sources. Diameters were measured after 24 h (eight strains on plates containing Casamino Acids), 48 h (all strains on plates containing glucose or glycerol plus succinate; MG1655 strains on plates containing sodium acetate), or 72 h (W3110 strains on plates containing sodium acetate). NT, not tested; NG, no growth detected. 

We inserted an omega cassette imparting chloramphenicol resistance (Cm^r) into the single ClaI site of epd (gapB) and crossed the resulting epd::Ω(Cm^r) mutation, which also contained a 4-bp deletion created during cloning, into the bacterial chromosome by transformation with linearized plasmid DNA (Fig. 2; Table 1) (2, 33). The epd::Ω(Cm^r) mutation was crossed into the W3110 and MG1655 genetic backgrounds by generalized transduction (Table 1). Western immunoblotting showed that the Epd enzyme was expressed in the W3110 epd⁺ parent but not in the W3110 epd::Ω(Cm^r) mutant (data not shown). The following G3P dehydrogenase-specific activities were obtained in crude extracts (36) of the W3110 gapA⁺ epd⁺, gapA⁺ epd::Ω(Cm^r), gapA1 epd⁺, and gapA1 epd::Ω(Cm^r) strains: 1,746 ± 53, 1,835 ± 117, 91 ± 4, and 0 nmol per min per mg of protein, respectively. Lack of or low residual G3P dehydrogenase activity in the gapA epd or gapA epd⁺ mutant, respectively, agrees with previous results (9).

FIG. 2.

Structure of the epd (gapB) gene (drawn to scale) at 66.17 min in the E. coli K-12 chromosome (1, 15, 21). epd (gapB) is surrounded by and oriented in the same direction as yggC, whose function is unknown, and pgk, which encodes the glycolytic enzyme phosphoglycerate kinase (1). Promoter mapping studies to be published elsewhere locate the promoters for epd (gapB) (P_epd) and pgk (P_pgk) in the regions indicated by bars. The location of the epd::Ω(Cm^r) insertion mutation constructed in this study (Table 1) is indicated. The epd::Ω(Cm^r) insertion mutation is upstream from the P_pgk promoter region and does not interfere with transcription of the pgk gene (data not shown).

Growth properties of gapA⁺ epd::Ω(Cm^r) mutants.

In contrast to the results of Della Seta et al. (9), we observed that colony formation of the gapA⁺ epd::Ω(Cm^r) mutant was impaired on plates containing MM [Vogel-Bonner (1 × E) (8) or M63 (27)] supplemented with glycerol plus succinate or Casamino Acids as carbon sources (Table 2). This impaired growth was relieved by the addition of PN (Table 2) or glycolaldehyde (GA), which can be converted to 4PHT by an alternative pathway (14, 37). Growth of the gapA⁺ epd::Ω(Cm^r) mutant was not significantly impaired on MM plates containing glucose, acetate, ribose, xylulose, fructose, or gluconic acid or in liquid medium containing glycerol plus succinate or Casamino Acids as carbon sources (Table 2 and data not shown).

We performed high-performance liquid chromatography (HPLC) (31) to confirm that the slower growth of the W3110 gapA⁺ epd::Ω(Cm^r) mutant on MM plates containing glycerol plus succinate was correlated with reduced cellular levels of PLP and PMP (Fig. 3; Table 3). Of the six B₆ vitamers, we detected only PLP and PMP in stationary-phase cells washed from plates after 2 or 3 days and suspended and sonicated in cold 5% metaphosphoric acid (31). Consistent with the growth characteristics (Table 2), the W3110 gapA⁺ epd::Ω(Cm^r) mutant contained only 64% of the PLP and PMP compared to the W3110 gapA⁺ epd⁺ parent, where most of the difference was a decrease in the amount of PMP (Table 3). Likewise, the W3110 gapA1 epd⁺ mutant contained only about 60% as much PMP as the parent; however, both strains contained equal amounts of PLP (Table 3). Finally, by assuming about 6 μl of water per mg of protein (28), we calculate that the stationary-phase W3110 gapA⁺ gapB⁺ parent contained about 76 and 36 μM PLP and PMP, respectively. These amounts are somewhat greater than the 40 μM combined intracellular concentration of PLP and PMP reported previously for E. coli K-12 (11). However, unlike this other report (11), we failed to detect appreciable B₆ vitamers excreted into the growth medium. Thus, the impaired growth of the gapA⁺ epd::Ω(Cm^r) mutant was correlated with a 40% reduction in the amounts of cellular PLP and PMP.

FIG. 3.

HPLC chromatograms of B₆ vitamers extracted from strains TX3470 (gapA⁺ epd⁺ parent) (top), TX3481 [(gapA⁺ epd::Ω(Cm^r)] (middle), and TX3484 (gapA1 epd⁺) (bottom). Strains were grown on plates as described in footnote a to Table 3, and B₆ vitamers were partially purified by metaphosphoric acid and dichloromethane extractions and were resolved by reverse-phase, ion-pair HPLC on an Ultremex 3 C₁₈ column (150 by 4.6 mm) (Phenomenex, Inc.) fitted with a guard column (30 by 4.6 mm) at 0.5 ml per min as described elsewhere (31). The tracings show the fluorescence intensity (excitation at 328 nm; emission at 393 nm) of postcolumn adducts between the B₆ vitamers and sodium bisulfite (31). The elution positions and fluorescence yields of the six B₆ vitamers (PLP, PMP, PNP, PL, PM, PN) were determined by using pure compounds as standards (data not shown). Only PLP and PMP (top panel) were detected in extracts of E. coli K-12. Two peaks that are not B₆ vitamers were routinely detected (top panel) and were used along with protein amounts of the cells before extraction to normalize PLP and PMP amounts in different preparations. Amounts of PLP and PMP were determined (Table 3) by integration of peaks and comparison of areas with those of the PLP and PMP standard curves, which were linear over this range of concentrations. The fluorescence yield of the PMP-bisulfite adduct was approximately 5.6-fold greater than that of the PLP-bisulfite adduct (data not shown). This difference accounts for the fact that PLP peaks, which are apparently smaller than PMP peaks, actually correspond to more PLP than PMP (Fig. 3; Table 3).

TABLE 3.

Amounts of B₆ vitamers in gapA⁺ epd⁺, gapA-1 epd⁺, and gapA⁺ epd::Ω(Cm^r) strains

Strain	Amt (pmol) of B6 vitamer per mg of protein (%)a
	PLP	PMP	PLP + PMP (%)
TX3470 (W3110 gapA+ epd+)	455 ± 99 (67)	218 ± 8 (33)	673 (≡100)
TX3484 (W3110 gapA1 epd+)	451 ± 18 (67)	126 ± 23 (19)	577 (86)
TX3481 [W3110 gapA+ epd::Ω (Cmr)]	329 ± 74 (49)	103 ± 6 (15)	432 (64)

^aBacteria were grown on plates containing (1 × E) minimal salts, 1% (vol/vol) glycerol, and 1% (wt/vol) succinate at 37°C for 48 h (TX3470 and TX3481) or 72 h (TX3484) (see Table 2). Cells were washed from plates with ice-cold (1 × E) minimal salts, and suspensions were adjusted to a turbidity of 100 Klett (660 nm) Units. Cells from 50 ml of suspension were collected by centrifugation, washed with (1 × E) minimal salts, and centrifuged again, where all steps were performed at 4°C. Cells were suspended in 2.5 ml of ice-cold metaphosphoric acid under yellow lights, and B₆ vitamers were extracted and resolved by reverse-phase, ion-pair HPLC (Fig. 3) (31). Protein amounts were determined by the D_c (Lowry) protein assay (Bio-Rad Laboratories) with bovine serum albumin as a standard in cell suspensions that were brought to 0.5% (wt/vol) in sodium dodecyl sulfate and boiled for 10 min. 

PN auxotrophy of gapA epd double mutants.

Most importantly, we found that the gapA1 epd::Ω(Cm^r) double mutant was auxotrophic for PN on MM containing glycerol plus succinate (Table 2). After 2 days of incubation, we could not detect growth of the gapA1 epd::Ω(Cm^r) double mutant, and after 3 days, we detected very small colonies which may have arisen by leakiness of the gapA1 point mutation. Growth was restored by the addition of PN or GA (Table 2). Unexpectedly, gapA1 mutants of MG1655 and W3110 grew on MM containing 0.4% sodium acetate as the carbon source (Table 2), whereas it was reported that the original DF220 gapA1 mutant in the K10 genetic background failed to grow on MM containing acetate (20). Nonetheless, the MG1655 and W3110 gapA1 gapB::Ω(Cm^r) double mutants were again auxotrophic for PN on MM containing acetate (Table 2).

We confirmed the conclusion that gapA epd double mutants are auxotrophic for PN in one other way. We simply transduced the epd::Ω(Cm^r) mutation into strain DF221 that contained the gapA2 nonsense allele (20), which is different from the gapA1 point mutation from strain DF220 used in the experiments described above (Table 1). Colonies of the resulting gapA2 epd::Ω(Cm^r) double mutant TX4187 appeared on MM plates containing glycerol plus succinate and PN in 2 to 3 days; however, no colonies appeared after 5 days when PN was omitted from the growth medium (data not shown). DF221 (gapA-2) failed to grow on MM containing glucose after 3 days at 37°C, but it did grow on MM containing acetate (data not shown). These growth properties might be explained by slight leakiness of the gapA1 and gapA2 mutations such that there was sufficient gluconeogensis to allow growth on acetate but insufficient glycolysis to support growth on glucose. Finally, it could be argued that a mutation in another gene that cotransduces with the gapA1 point mutation caused the PN auxotrophy (Table 2). To rule out this hypothesis, we analyzed 20 independent spontaneous mutants of TX3491 and TX4134 [gapA1 epd::Ω(Cm^r)] that grew rapidly as did the gapA⁺ epd::Ω(Cm^r) strain on MM containing 0.4% (wt/vol) glucose. For all 20 mutants, reversion or suppression of the gapA1 mutation not only allowed growth on glucose medium but also alleviated the PN requirement.

Summary.

Together, our data show that the GapA and Epd dehydrogenases are required for de novo PLP biosynthesis, and this involvement supports the pathway depicted in Fig. 1. The contribution of Epd dehydrogenase to 4PE synthesis seems to vary and becomes more significant in colonies growing on certain nonglycolytic carbon sources (Tables 2 and 3). We do not understand why the growth of gapA⁺ epd mutants is impaired on solid but not in liquid media (see above; Tables 2 and 3). In cells grown on glycolytic carbon sources, such as glucose, we observed that the specific activity of G3P dehydrogenase increased at least twofold in crude extracts compared to cells grown in MM containing glycerol plus succinate or acetate (data not shown). Therefore, it seems likely that the GapA enzyme alone is sufficient to carry out PLP biosynthesis in cells growing on glycolytic carbon sources. The involvement of the GapA enzyme in PLP biosynthesis is consistent with low levels of E4P dehydrogenase detected for some GapA dehydrogenases (5), and the enzymatic redundancy of GapA and Epd would explain why mutants deficient in this first step of branch 1 of PLP biosynthesis were never isolated (10, 23). Using purified enzymes, we did not detect feedback inhibition of the Epd E4P or GapA G3P dehydrogenase activities by 4PHT (data not shown). This finding is consistent with the idea that PLP biosynthesis responds to the carbon source and overall metabolic state instead of to the amounts of pathway end products (26).

We can only speculate as to why we obtained results completely different from those of Della Seta et al. (9). In their experiments, the parent and mutant strains seem to have been spread onto plates containing different combinations of antibiotics corresponding to the insertions in their gapA and epd mutants (9). However, it is not immediately clear why antibiotic addition would bypass the requirement for GapA and Epd in PLP biosynthesis. Likewise, it is difficult to see how polarity of the gapA and epd insertion mutations, if any, could bypass the need for an E4P dehydrogenase in PLP biosynthesis. In their experiments, no control was mentioned to test media for traces of B₆ vitamers that would allow growth of the gapA epd double mutant. Finally, it seems possible that their gapA epd (gapB) double mutant may have acquired a suppressor mutation that allowed growth without supplementation with B₆ vitamers. Another partial homolog of gapA, called gapC, is present in E. coli K-12, but gapC is not thought to encode a functional dehydrogenase (5, 17).