The Autonomous Glycyl Radical Protein GrcA Restores Activity to Inactive Full-Length Pyruvate Formate-Lyase In Vivo

ABSTRACT

During glucose fermentation, Escherichia coli and many other microorganisms employ the glycyl radical enzyme (GRE) pyruvate formate-lyase (PflB) to catalyze the coenzyme A-dependent cleavage of pyruvate to formate and acetyl-coenzyme A (CoA). Due to its extreme reactivity, the radical in PflB must be controlled carefully and, once generated, is particularly susceptible to dioxygen. Exposure to oxygen of the radical on glycine residue 734 of PflB results in cleavage of the polypeptide chain and consequent inactivation of the enzyme. Two decades ago, a small 14-kDa protein called YfiD (now called autonomous glycyl radical cofactor [GrcA]) was shown to be capable of restoring activity to O₂-inactivated PflB in vitro; however, GrcA has never been shown to have this function in vivo. By constructing a strain with a chromosomally encoded PflB enzyme variant with a G734A residue exchange, we could show that cells retained near-wild type fermentative growth, as well as formate and H₂ production; H₂ is derived by enzymatic disproportionation of formate. Introducing a grcA deletion mutation into this strain completely prevented formate and H₂ generation and reduced anaerobic growth. We could show that the conserved glycine at position 102 on GrcA was necessary for GrcA to restore PflB activity and that this recovered activity depended on the essential cysteine residues 418 and 419 in the active site of PflB. Together, our findings demonstrate that GrcA is capable of restoring in vivo activity to inactive full-length PflB and support a model whereby GrcA displaces the C-terminal glycyl radical domain to rescue the catalytic function of PflB.

IMPORTANCE Many facultative anaerobic microorganisms use glycyl radical enzymes (GREs) to catalyze chemically challenging reactions under anaerobic conditions. Pyruvate formate-lyase (PflB) is a GRE that catalyzes cleavage of the carbon-carbon bond of pyruvate during glucose fermentation. The problem is that glycyl radicals are destroyed readily, especially by oxygen. To protect and restore activity to inactivated PflB, bacteria like Escherichia coli have a small autonomous glycyl radical cofactor protein called GrcA, which functions to rescue inactivated PflB. To date, this proposed function of GrcA has only been demonstrated in vitro. Our data reveal that GrcA rescues and restores enzyme activity to an inactive full-length form of PflB in vivo. These results have important implications for the evolution of radical-based enzyme mechanisms.

INTRODUCTION

Pyruvate formate-lyase (PflB) is the best characterized member of the large glycyl radical enzyme (GRE) superfamily (1, 2), which also includes the enzymes anaerobic ribonucleotide reductase (3), ketobutyrate formate-lyase (4), benzylsuccinate synthase (5), glycerol dehydratase (6), and trans-4-hydroxy-l-proline dehydratase (7). PflB catalyzes the reversible homolytic and coenzyme A (CoA)-dependent cleavage of the C1-C2 bond of pyruvate into acetyl-CoA and formate (2, 8, 9). The radical-based mechanism of the PflB reaction allows acetyl-CoA formation without the concomitant generation of reducing equivalents, which remain associated with formate. The energetic benefit afforded to a fermenting bacterium like Escherichia coli through formate production is that it can be released readily as formic acid from the cell or by disproportion of formate to CO₂ and H₂, catalyzed by the formate-induced formate hydrogenlyase (FHL) complex (10).

The radical on PflB is generated by the pyruvate formate-lyase-activating enzyme (PflA), which belongs to the superfamily of radical S-adenosylmethionine (AdoMet) enzymes. PflA uses a [4Fe-4S] cluster to cleave AdoMet reductively to generate a 5′-deoxyadenosyl radical, which then stereospecifically abstracts a H-atom from glycine at position 734 (G734) on the polypeptide chain of PflB (11, 12). The electron required for reductive cleavage of AdoMet is derived from reduced flavodoxin (13). Consequently, the complete system necessary for glycyl radical formation, and ultimately carbon-carbon cleavage by PflB, is independent of ATP hydrolysis and likely represents a very old mechanism that was evolved to make and break the stable carbon-carbon bond of pyruvate to form formate and acetyl-CoA (14).

The glycyl radical in PflB and other GREs is kinetically stable but extremely oxygen sensitive (2, 15). Structural studies on the inactive, nonradical form of PflB have revealed that G734 is buried within the enzyme, approximately 8 Å from the surface in an αβ-barrel that offers some protection from solvent and, therefore, oxygen (8, 16). Residue G734 is located on the tip of a hairpin loop and is proximal to two adjacently located and catalytically essential cysteine residues (C418 and C419) (9), which are also located on the tip of an opposing hairpin loop (8). The radical on G734 is transferred to C419 to form a thiyl radical, which is responsible for C-C bond cleavage, releasing a formyl radical anion and the acetyl group. The acetyl group is transferred from C419 to C418 when C419 recaptures the radical from the formyl radical anion, thus allowing the catalytic cycle to continue (9).

While the glycyl radical is protected within the enzyme during catalysis, in order to receive the radical from PflA in the first place, the glycine residue must be made accessible at the enzyme surface, which means the loop on which G734 is located has to undergo a significant conformational change to expose it to PflA (17, 18). G734 is located in a so-called C-terminal radical domain comprising residues 697 to 759 on the polypeptide chain (19). This domain has been shown to be highly flexible, and biophysical studies have provided strong evidence indicating that a major conformational change in the protein occurs when PflB interacts with PflA (18).

PflB is a dimer and shows half-of-the-sites activity, whereby only one monomer carries the glycyl radical per dimer (15, 20). Although the glycyl radical is stable when kept away from O₂, facultative anaerobic bacteria, such as E. coli, are frequently exposed to O₂. Contact of the glycyl radical-bearing form of PflB with O₂ results in an oxygenolytic cleavage of the polypeptide backbone between residues 733 and 734, which inactivates the enzyme (2). Just over 20 years ago, the group of Knappe made the remarkable discovery that a small, two-domain ∼14-kDa protein, referred to at that time as YfiD (since renamed glycyl radical cofactor A [GrcA]; https://ecocyc.org), was able to restore pyruvate-cleavage activity in vitro to an oxygenolytically inactivated PflB enzyme (21). The roughly 60-residue C-terminal domain of GrcA shows high amino acid sequence conservation (77% amino acid sequence identity) with the glycyl radical domain of PflB, while the N-terminal domain of GrcA shows no similarity to any characterized proteins in the databases (21, 22). A nuclear magnetic resonance (NMR) structure of GrcA has been obtained (23), and while the well-resolved C-terminal domain shows the anticipated structural homology with the radical domain of PflB, the N-terminal domain is highly flexible and largely unstructured. Meanwhile, a recent detailed in vitro biochemical analysis of GrcA with a recombinantly generated truncated PflB (residues 1 to 733), nearly equivalent to the oxygenolytically cleaved species, has demonstrated that the flexible N-terminal domain of GrcA is important for the restoration of catalytic activity to oxygen-damaged PflB but is not required for catalysis (19). That study led to a plausible mechanism of how GrcA might be able to displace the cleaved radical domain of PflB and to act as an autonomous domain that could receive a radical on its conserved glycine residue 102 (G102) (19) from PflA (22). However, an in vivo demonstration that GrcA functions as an autonomous glycyl radical cofactor to rescue PflB, or any other GRE for that matter, has never been demonstrated. Moreover, experimental data on the ability of GrcA to rescue an inactive but uncleaved, full-length PflB species have never been provided. If this ability could be shown, it would provide strong support for the mechanistic model presented by Andorfer et al. (19), indicating that GrcA is able to cause a large conformational displacement of the PflB C-terminal glycyl-radical domain to replace and take over its function.

The grcA gene (previously named yfiD) was originally identified after it was shown that the synthesis of its gene product is induced under acid stress (24). The grcA (yfiD) gene was identified as a member of the fumarate and nitrate reduction regulator (FNR) regulon (25, 26). Its synthesis was shown to be coregulated by aerobic respiratory control regulator (ArcA) and in response to pyruvate through PdhR (22). Together, these findings provide clear evidence of a response of grcA expression to anaerobiosis/microaerobiosis and to acid stress. Moreover, deletion of the grcA gene resulted in enhanced lactate production (22). Together, all these data are consistent with a role for GrcA in supporting formate production as a product of glucose fermentation.

To address directly the ability of GrcA to rescue inactive, radical-free PflB in fermenting E. coli cells, we have introduced a chromosomal codon exchange in the pflB gene (27) that delivers a gene product in which glycine at amino acid position 734 is converted to an alanine residue. Surprisingly, the strain producing this PflB_G734A variant retains a near-wild-type growth phenotype and an ability to produce formate. We show here that this phenotype is due exclusively to GrcA-dependent recovery of PflB enzyme activity. Our study thus demonstrates that GrcA can rescue inactive, full-length PflB in vivo.

RESULTS

E. coli synthesizing a PflB_G734A exchange variant displays unimpaired anaerobic growth and has an intact pyruvate metabolism.

E. coli strain MC801 carrying a GGC to GCC exchange in codon 734 of the pflB gene (generating PflB_G734A; see Materials and Methods) had a fermentative growth rate that was similar (∼10% growth rate reduction) to that of its isogenic parent, MC4100 (Table 1). In contrast, mutants MC803 (ΔpflB) and 234M1 (ΔpflA) lacking the genes encoding PflB (see Table 2) or its activating enzyme PflA, respectively, grew significantly more slowly (40 to 50% growth-rate reduction) and attained lower final optical density at 600 nm (OD₆₀₀) (Table 1) signifying lower cell yields, which is consistent with both mutants having severely impaired pyruvate catabolism (28). These results indicated that substitution of the glycine for an alanine residue in the radical domain of PflB did not significantly impair PflB function in anaerobic pyruvate catabolism. Western blot analysis using anti-PflB antiserum (4) demonstrated that similar levels of PflB_G734A in strain MC801 were synthesized compared with native PflB in MC4100, while no PflB was detectable in strain MC803 (ΔpflB) (see Fig. S1A in the supplemental material).

TABLE 1.

Anaerobic growth rates of the mutants with altered pyruvate metabolism

Strain	Growth rate (μ [h−1])	Final OD600a
MC4100	0.375 ± 0.003	0.644 ± 0.012
234M1 (ΔpflA)	0.192 ± 0.008	0.446 ± 0.012
MC803 (ΔpflB)	0.233 ± 0.002	0.468 ± 0.068
MC801 (pflB-G735A)	0.335 ± 0.004	0.642 ± 0.045
MC901 (ΔgrcA)	0.283 ± 0.006	0.559 ± 0.039
MC903 (ΔgrcA pflB-G735A)	0.197 ± 0.003	0.414 ± 0.054

^aGrowth studies were done in M9-glucose (0.8% wt/vol) minimal medium.

TABLE 2.

Strains and plasmids used in this study

Strain, phage, or plasmid	Relevant genotype or characteristics	Reference or source
Strains
°JW0886	ΔpflB727::kan Δ(araD-araB)567 Δ(rhaD-rhaB)568 rph-1hsdR514 ΔlacZ4787(::rrnB-3) λ- Kanr	52
°JW2563	ΔyfiD747::kan Δ(araD-araB)567 Δ(rhaD-rhaB)568 rph-1hsdR514 ΔlacZ4787(::rrnB-3) λ- Kanr	52
°MC4100	F- araD Δ(argF lac) U 169 ptsF25 deoC1 relA1 fblB530 rpsL 150 λ-	53
°234M1	Like MC4100, but Δact Ω(act::cat pACYC184) Cmr	2
°MC801	Like MC4100, but pflB-G734A	This study
°MC803	Like MC4100, but ΔpflB	This study
°MC901	Like MC4100, but ΔgrcA	This study
°MC903	Like MC801, but ΔgrcA	This study
°DH4100	Like MC4100, but λ(fdhF::lacZ) Kanr	29
°DH201	Like MC4100, but ΔfocA ΔpflB λ(fdhF::lacZ) Cmr, KanR	29
°DH234M1	Like 234M1, but λ(fdhF::lacZ) Cmr, Kanr	This study
°DH801	Like MC801, but λ(fdhF::lacZ) Kanr	This study
°DH901	Like MC901, but λ(fdhF::lacZ) Kanr	This study
°DH903	Like MC903, but λ(fdhF::lacZ) Kanr	This study
Phages
°P1kc	P1 starter lysate originated from E. coli MG1655	Lab collection
°P1:: ΔpflB	Insertion of kan cassette and deletion of pflB, derived from donor JW0886 Kanr	This study
°P1:: ΔgrcA	Insertion of kan cassette and deletion of grcA, derived from donor JW2563	This study
°λ(fdhF::lacZ)	λRS45 including the 232 bp regulatory region of fdhF	54
Plasmids
°pJET1.2	Ampr, subcloning vector	Thermo Fischer Scientific, Waltham, USA
°pMAK705	Cmr, thermosensitive plasmid for DNA replication	44
°pMAK705::pflB-G735A	pMAK705 carrying pflB-pflA fragment with pflB codon 734 (GGC) exchanged for GCC	This study
°pCP20	Helper plasmid to excise kanamycin antibiotic-resistance cassettes	46
°pUC19	Ampr, high copy-no. vector	55
°pUC19::grcA	pUC19 carrying the grcA gene	This study
°pUC19::grcA-G102A	pUC19::grcA-G102A with codon 102 (GGT) exchanged for GCG	This study
°p29	Cmr, pACYC184 vector carrying the focA-pflB operon and pflA gene	47
°p29-C2	p29 with pflB codons 418 and 419 (both TGC) exchanged for (GCC)	This study

This surprising lack of a growth phenotype of strain MC801 suggested that perhaps PflB_G734A was being rescued by the autonomous glycyl radical cofactor protein GrcA. Therefore, we tested this by introducing a ΔgrcA allele into strain MC801, which resulted in a fermentative growth phenotype like that of the ΔpflA mutant 234M1 (Table 1). The same ΔgrcA allele, when introduced into the parental strain MC4100, caused an approximate 25% reduction in the anaerobic growth rate. Together, these results suggested strongly that GrcA was capable of conferring pyruvate-cleavage activity to the PflB_G734A variant allowing it to function in anaerobic pyruvate catabolism.

GrcA-dependent formate production by a strain synthesizing PflB_G734A.

As PflB catalyzes the CoA-dependent production of formate from pyruvate, a straightforward demonstration that the PflB_G734A variant is functional in vivo is to determine whether intracellular formate is generated. Intracellular formate production can be monitored by using a chromosomal formate-responsive fdhF_P::lacZ reporter system (29), which results in β-galactosidase enzyme activity only when formate is made and accumulates intracellularly (30). After transfection of the strains with a lambda phage carrying the fdhF_P::lacZ reporter (see Materials and Methods and Table 2), β-galactosidase enzyme activity of this series of strains (all with acronym DH-) was determined using cells obtained from anaerobically cultivated strains (Fig. 1A). The results show clearly that only the parental strain DH4100 and strains DH801 (PflB_G734A synthesized) and DH901 (ΔgrcA) produced significant levels of β-galactosidase enzyme activity, while strains DH234M1 (ΔpflA), DH201 (ΔfocA, ΔpflB), and DH903 (pflB-G734A, ΔgrcA) produced barely detectable levels of reporter activity (Fig. 1A); note that the absence of focA has no impact on β-galactosidase enzyme activity in strain DH201. The higher levels of β-galactosidase enzyme activity (reflecting increased intracellular formate levels; 30) determined for strains DH801 and DH901 (roughly 40% to 70% increased β-galactosidase activity) than the level measured in the parental strain are possibly due to an impaired formate efflux by the formate channel FocA, whose function is controlled through interaction with PflB (31). However, this hypothesis will require further analysis to elucidate.

FIG 1.

A strain synthesizing the PflB_G734A variant is capable of formate and H₂ production in vivo. The indicated strains (see below) were analyzed for formate-dependent fdhF_P::lacZ expression measured as β-galactosidase enzyme activity (Miller units) (A), H₂ accumulation (B), extracellular formate (C), and d-lactate (D) production (see Materials and Methods for details). Strains included wt, MC400; ΔpflA, 234M1; ΔpflB, MC803; pflB-G734A, MC801; ΔgrcA, MC901; and ΔgrcA pflB-G734A, MC903. When β-galactosidase activity was measured (A), strains carried a λfdhF_P::lacZ construct and were the corresponding DH derivatives (see Table 2). All experiments were performed in duplicate with minimally three biological replicates.

Intracellular formate accumulation is necessary to induce synthesis of the formate hydrogenlyase complex, which disproportionates formate to CO₂ and H₂ (10, 30). Therefore, the ability of each strain to accumulate H₂ was also assessed after fermentative growth (Fig. 1B). The parental strain MC4100 produced approximately 90.2 ± 2.5 μmol H₂ OD₆₀₀⁻¹, which was similar to the amount produced by the grcA mutant MC901 (83.2 ± 1 μmol H₂ OD₆₀₀⁻¹). Although strain MC801 (PflB_G734A synthesized) clearly produced H₂, the amount generated was half that of the parental strain (Fig. 1B). We do not yet have an explanation for this reduced level of H₂ accumulation by strain MC801. None of the other three strains tested, namely, MC803 (ΔpflB), 234M1 (ΔpflA), or MC903 (ΔgrcA, pflB-G734A), produced any detectable H₂ during anaerobic growth (Fig. 1B), which is in accord with the lack of formate production and, consequently, the absence of PflB activity.

As a further confirmation of PflB-dependent formate production, we determined the amount of formate excreted into the culture medium during exponential-phase anaerobic growth (Fig. 1C). The parental strain MC4100 accumulated 5.2 ± 0.5 mM formate OD₆₀₀⁻¹ in the culture medium during growth, while neither strain MC803 (ΔpflB) nor strain 234M1 (ΔpflB) excreted any detectable formate into the culture medium (Fig. 1C). In agreement with the growth experiments, however, strain MC801 (PflB_G734A synthesized) excreted nearly 4 mM formate OD₆₀₀⁻¹ into the culture medium. The grcA mutant MC901 also produced formate at levels similar to the parental strain during glucose fermentation. In contrast, strain MC903 (ΔgrcA pflB-G734A) also failed to produce and excrete any detectable formate (Fig. 1C).

Finally, a characteristic feature of E. coli mixed-acid fermentation is the production of d-lactate when pyruvate no longer can be cleaved to acetyl-CoA and formate (28). The concentration of d-lactate in the culture medium was therefore determined for all six strains (Fig. 1D). An inverse correlation was observed between the ability of strains to produce formate and their reduction of pyruvate to lactate. Thus, MC803 (ΔpflB), 234M1 (ΔpflA), and MC903 (ΔgrcA, pflB-G734A) excreted high concentrations of lactate (17 to 25 mM lactate OD₆₀₀⁻¹), while the parental strain MC4100 and MC901 (ΔgrcA) produced no detectable lactate (Fig. 1D). Strain MC801 (synthesizing PflB_G734A) generated approximately only 3% the concentration of lactate produced by strain 234M1 (Fig. 1D).

Combined, these results demonstrate unequivocally that a strain synthesizing PflB_G734A is able to generate formate from pyruvate and that formate production is totally dependent on GrcA. Formate production by the parental strain synthesizing native PflB does not depend on GrcA.

Glycine residue 102 on GrcA is essential for rescue of PflB_G734A catalytic activity.

The C-terminal domain (63 amino acid residues) of the 127 amino acid residue GrcA protein shares 77% amino acid identity with the C-terminal radical domain of PflB (21–23). In particular, the sequences around G102 in GrcA and G734 in PflB are highly conserved (see Fig. S2 in the supplemental material), suggesting that G102 is the residue that receives the radical from PflA (21). To assess any role of G102 of GrcA in the rescue of PflB_G734A catalytic activity, the grcA gene was cloned into vector pUC19 delivering pUC19::grcA (see Table 2). The DNA sequence of codon 102 in grcA was then mutated (GGT→GCG) to decode as alanine, delivering plasmid pUC19::grcA-G102A. These plasmids were introduced separately into strain MC903 (ΔgrcA, pflB-G734A), and H₂ production of the strains after anaerobic growth was assessed (Fig. 2A). The results show clearly that while plasmid pUC19::grcA fully complemented the H₂-negative phenotype of strain MC903, plasmid pUC19::grcA-G102A, encoding GrcA_G102A, when introduced into the strain failed to restore gas production (Fig. 2A). This result demonstrates that the glycine residue at position 102 is essential for GrcA to be able to function in recovery of the catalytic activity of PflB_G734A in strain MC903.

FIG 2.

Pyruvate-cleavage activity of PflB_G734A is dependent on GrcA. Quantitative determination of cumulative H₂ production by the indicated strains is shown. The strains were cultivated anaerobically in glucose-M9 minimal medium, and H₂ was measured by gas chromatography as indicated in the Materials and Methods. (A) H₂ production was measured for the following strains: wt, MC400; ΔpflB, MC803; pflB-G734A, MC801; ΔgrcA pflB-G734A, MC903; and for MC903 transformed with pUC19 (empty vector control), with pUC19::grcA (including grcA and its upstream region) and pUC19::grcA-G102A coding for GrcA_G102A. (B) Cumulative H₂ formation was analyzed in strain MC803 (ΔpflB) transformed with plasmid p29, which carries the parental focA, pflB, and pflA genes, or p29-C2, which carries parental focA and pflA. But, pflB has exchanges in codons 418 and 419 (see Materials and Methods for details). The positive control (wt) was the parental strain MC400. All experiments were carried out with minimally three biological replicates.

GrcA does not obviate the catalytically essential cysteine residues in PflB.

Glycine residue 734 is only the storage location for the radical on native PflB, while homolytic cleavage of pyruvate necessitates the participation of the cysteine residues at positions 418 and 419, whereby C419 is converted to the thiyl radical necessary for catalysis, while the acetyl group is transferred to C418 after pyruvate cleavage (2, 8, 9). To determine whether C418 and C419 were required for formate production by PflB, both codons within pflB on plasmid p29 (see Materials and Methods and Table 1) for these residues were mutated to decode as alanine residues, delivering plasmid p29-C2 (product, PflB_C418A/C419A). Introduction of p29 (focA⁺ pflB⁺pflA⁺) and p29-C2 into strain MC803 (ΔpflB) followed by anaerobic growth and determination of accumulated H₂ revealed that only plasmid p29 was capable of restoring H₂ production, and consequently formate generation, to the mutant, while plasmid p29-C2 failed to restore H₂ production (Fig. 2B), despite a high-level production of PflB_C418A/C419A (Fig. S1B). This result indicated that GrcA was unable to restore activity to the PflB_C418A/C419A variant, demonstrating that the cysteine residues are essential for catalytic activity, even when PflB is reactivated by GrcA.

The results presented so far in the current study with PflB_G734A suggest that GrcA can restore activity to inactivated, full-length, i.e., uncleaved, PflB in vivo. As oxygenolytic cleavage of the radical-bearing species of native PflB results in C-terminal truncation of the polypeptide by 25 amino acid residues (2, 27), and in vitro studies have shown that purified GrcA can rescue oxygen-damaged PflB (19, 21), we wished to determine whether the PflB_G734A variant is also subject to oxygen-dependent cleavage of the polypeptide chain after GrcA-dependent restoration of activity. A Western blot using anti-PflB antiserum was performed with crude extracts derived from strains MC4100 (parental, wild type), 234M1 (ΔpflA), MC803 (ΔpflB), and MC801, which synthesizes PflB_G734A (Fig. 3). The blot revealed that the extract derived from MC803 (ΔpflB) showed, as anticipated, no signal, while that from MC4100 showed a truncated PflB species of ∼81 kDa, characteristic of O₂-derived cleavage of the polypeptide chain (2). As a control, PflB in the pflA mutant migrated ∼3 kDa more slowly at 84 kDa, as expected for the full-length polypeptide because the glycyl radical cannot be formed in the mutant (Fig. 3). The extract derived from strain MC801 revealed a single polypeptide species with a mass equivalent to full-length, uncleaved PflB (Fig. 3), indicating that no oxygenolytic cleavage of PflB_G734A occurred.

FIG 3.

PflB_G734A is not subject to oxygenolytic C-terminal cleavage. Polypeptides in crude extracts (30 μg of protein) derived from the indicated strains were separated in denaturing SDS-PAGE in an 8% (wt/vol) TG Prime SERVA gel. The gel was run until proteins that were smaller than 70 kDa migrated through the gel to separate the intact (upper arrow labeled 1) from the oxygenolytically cleaved (lower arrow labeled 2) PflB polypeptide. The top panel shows a silver-stained gel, while the bottom panel shows a Western blot challenged with anti-PflB antiserum. The migration position of PflB is indicated on the right-hand side of the top panel, and the migration positions of the molecular mass markers (PageRuler prestained protein ladder; Thermo Fisher Scientific) are indicated in kDa on the left of each gel or blot.

DISCUSSION

Our observation that a PflB_G734A amino acid exchange variant retained pyruvate cleavage activity in vivo was initially unexpected because all GREs studied to date have an essential glycine residue that functions both as a radical storage location and as the source of the radical that is ultimately transferred to generate the catalytic thiyl radical (32–35). It was shown here that the autonomous glycyl radical cofactor protein GrcA restored pyruvate formate-lyase activity to the full-length inactive PflB_G734A variant by supplying the missing glycyl radical. Our study demonstrates unequivocally that GrcA indeed functions in vivo as an autonomous glycyl radical domain, which was previously only surmised based on in vitro experiments performed with purified proteins (19, 21). The data also imply, however, that GrcA also can rescue the activity of inactivated full-length PflB, which has not been exposed to O₂. This conclusion is also supported by the observed reduction in growth of a grcA mutant during fermentation, which suggests that even when O₂ is not present, GrcA helps maintain PflB, and possibly also other GREs, in a catalytically active, radical-bearing state. This result also indicates that despite its demonstrated longevity in vitro (2, 18), radical quenching nevertheless needs to be countered even in the anaerobic cell, most likely during acid-induced stress (24).

It is nonetheless likely that the main function of GrcA is to restore activity to O₂-inactivated PflB, especially in facultative anaerobes, such as E. coli. This idea is not only congruent with the similar expression patterns of grcA (22, 25) and pflB (36) but is also reflected in the phylogenetic codistribution of grcA in almost all bacterial and eukaryal species with a pflB gene (Fig. 4). With the exception of the deeply branching Thermotogales order, every other class or phylum whose genome encodes a PflB-like GRE also carries a grcA gene; notably, the GRE in Thermotogae is predicted to belong to the glycerol dehydratase, and not the PflB, family of GRE (37). While PflB and GrcA are distributed widely among the bacteria, they are most significantly prevalent in the proteobacteria, firmicutes, and actinobacteria (Fig. 4). Strikingly, both proteins are also found in some fungi (Opisthokonta) and green algae (Viridiplantae), which are aerobes.

FIG 4.

Co-occurrence of the grcA and pflB genes. An analysis of the co-occurrence of the grcA and pflB genes, encoding the autonomous glycyl radical cofactor protein GrcA and pyruvate formate-lyase PflB, is shown. The analysis was performed using the STRING online tool (version 11.5) (51). The asterisk designates the only example of an order, Thermotogales, in which no co-occurrence of a grcA gene with that encoding a PflB-like protein is found.

Both GrcA and PflB are absent from some very deeply branching bacterial orders/phyla, which include many strict anaerobes, and from the archaea (Fig. 4). However, it should be noted that because the gene sequence of E. coli pflB was used to search for phylogenetically similar sequences, any similarity with PflB in these microorganisms might be visible only when the gene products are compared at the level of protein tertiary structure, as exemplified when PflB and anaerobic ribonucleotide reductase (NrdD) are compared (38). Nevertheless, GrcA is notably absent from these phyla that also lack an E. coli-like PflB, suggesting that GrcA evolved to help protect PflB (GRE) enzymes from O₂-dependent radical quenching.

The grcA gene is not colocalized with the pflB and pflA genes on the E. coli chromosome but is found near genes whose products are involved in nucleotide and DNA metabolism, suggesting a second function for the protein, possibly in rescuing anaerobic ribonucleotide reductase (NrdD) (32, 39). Clearly, while selective pressure ensures that pflB and pflA are maintained and acquired together, e.g., via horizontal gene transfer, it is not the case for grcA and supports a function for GrcA in delivery of the glycyl radical to, and thus rescue of, other GREs within the bacterium. As well as PflB and NrdD, E. coli also synthesizes ketobutyrate formate-lyase (TdeE) (4) and has the coding capacity for two other predicted GREs, namely, PflD and f810 (37).

Despite the separate chromosomal locations of the grcA and pflB genes, nevertheless, their respective expression is coordinated in response to both O₂ and anaerobiosis by ArcA and FNR (22, 36) and to pyruvate levels (22, 40), which in the case of grcA is achieved via PdhR (22). How grcA expression responds to acidic pH (24) remains to be established, but it is conceivable that this response is via sensing of increased organic acid concentration (22). Overall, this transcriptional regulation thus ensures that GrcA and PflB are always present at similar levels under micro-oxic and anoxic conditions. The fact that both proteins are converted to their glycyl radical-bearing forms by PflA (22) guarantees that PflB retains activity under these conditions, regardless of whether it is inactivated inadvertently by oxygenolytic cleavage or by other means of radical quenching.

GrcA-dependent restoration of activity to full-length PflB_G734A in vivo does not involve direct transfer of the radical from GrcA to the C-terminal domain of PflB_G734A. This conclusion is clear because no oxygenolytic cleavage of the catalytically active PflB_G734A polypeptide occurs. Rather, the results indicate that the radical on G102 in GrcA is transferred directly to the C418-C419 residues of PflB, as proposed previously (9, 22), because exchange of these residues with alanine completely inactivates PflB in vivo, regardless of whether or not GrcA is present in the cell. Finally, our data also demonstrate unequivocally that, as predicted from in vitro work (19, 22), G102 of GrcA receives the radical from PflA. Thus, based on these data, full-length PflB must also be capable of accommodating the simultaneous binding of both GrcA and PflA, which requires that the protein must undergo a large conformational change, as proposed recently (19).

Modeled superposition of GrcA onto the structure of PflB reveals an excellent fit for residues 697 to 759 and the C-terminal domain of GrcA (residues 64 to 127) (see Fig. S3 in the supplemental material). The intrinsically disordered N-terminal domain of GrcA (23) does not resemble the PflB structure, suggesting that the flexibility of this domain might be important for allowing interaction by allowing “molding” to different GRE substrates. Our in vivo data indicate that the radical domain of PflB must be displaced completely to allow the access of both GrcA and PflA (8, 18, 19), and a recent in vitro study by the Drenan group (19) has provided a plausible and elegant model as to how these three proteins could form such a complex. This proposal is supported by biochemical evidence provided by Broderick and coworkers, which indicates that a major conformational change in the radical domain of PflB is induced by binding of PflA (18, 41).

Once the radical has been introduced onto G734 of PflB, or G102 of GrcA, by PflA, the loop on which it is located must be reinserted into the β-barrel of PflB to afford protection from solvent, as well as to position it close to C419 (8, 16) (see Fig. S3). Pyruvate and its substrate analog oxamate are allosteric activators of PflB activity (16), and they could also be shown to enhance the glycyl radical content of the PflB:PflA complex (41); however, they bind only at the active site of PflB (8, 16). This information has led to the proposal that pyruvate aids reinsertion and thus stabilization of the glycyl radical loop of PflB inside the enzyme (41). As recent evidence has shown that GrcA cannot be converted to the radical species by PflA in the absence of PflB (19), concomitant binding of PflA and GrcA to PflB, together with the substrate pyruvate, could control glycyl radical formation and the ternary complex may be required for activation to proceed in vivo. An association of GrcA and PflA with PflB would also be commensurate with the ready isolation of GrcA (21) (L. Beyer and R.G. Sawers, unpublished data) and PflA (42) in association with PflB during enzyme purification. Ultimately, the nature of such a putative PflB:PflA:GrcA complex and the role of pyruvate in allosteric control of enzyme activation will be best revealed by structural analyses.

The use of glycyl/thiyl radical-based enzymatic mechanisms obviating involvement of either O₂ or ATP suggests an ancient origin for GREs. The evolution of these GRE-based biochemical reactions has been discussed in the context of C-H bond cleavage catalyzed by NrdD as being the first enzyme involved in DNA synthesis (32, 35). The parallels with reversible carbon-carbon bond cleavage catalyzed by PflB in early metabolism are clear (33).

MATERIALS AND METHODS

General cultivation conditions.

The strains used in study are listed in Table 2. Cultivation of bacterial strains for cloning and standard procedures was in Luria-Bertani broth (LB) or on LB agar plates (43). For all other experiments, strains were cultivated anaerobically in M9-minimal medium containing 0.8% (wt/vol) glucose as a carbon source (31). In general, cells were grown anaerobically in 15-mL Hungate tubes at 37°C. The only exception was for the analysis of PflB levels by Western blotting where cells were cultivated in crimp-sealed 500-mL serum bottles. When antibiotics were required, they were added at a final concentration of 50 μg · mL⁻¹ for kanamycin, 100 μg · mL⁻¹ for ampicillin, and 25 μg · mL⁻¹ for chloramphenicol.

Construction of strains and plasmids.

To construct strain MC801, initially an approximately 1,540-bp DNA fragment spanning the pflB-pflA genes (27) was amplified using MC4100 genomic DNA as the template and oligonucleotides pflB_BamHI_fw and pflB_HindIII_re (see Table S1). The DNA fragment was first cloned into pJET1.2 (Table 2). After confirmation of the authenticity of the DNA sequence, the DNA fragment in this plasmid was used as a template for site-directed mutagenesis to exchange codon 734 (GGC for glycine) to GCC (decoding as alanine) using the QuikChange protocol (Agilent Technologies) and the oligonucleotides pflB_G734A_forward and reverse (Table S1). Please note that while the pflB gene has 760 codons that decode as amino acid residues, because the initial fMet residue is absent from the gene product (27), the residue in the PflB polypeptide is conventionally number 734, and so the mutant protein is named PflB_G734A. The mutated DNA fragment was released from the resulting plasmid by digestion with the restriction enzymes BamHI and HindIII, and the pflB-G734A-pflA DNA fragment was ligated into pMAK705, which had been digested previously with BamHI and HindIII and dephosphorylated with alkaline phosphatase. The altered pflB allele was recombined into the chromosome of MC4100 following standard procedures (44), delivering strain MC801. The authenticity of the codon exchange in the pflB gene and the corresponding integration into the genome of MC801 were verified by DNA sequencing of the amplified gene.

The pflB and grcA deletion mutants MC803 (ΔpflB) and MC901 (ΔgrcA), respectively, were constructed by P1-mediated phage transduction using MC4100 as the acceptor strain. P1 phage lysates were generated from the respective Keio strains (see Table 2) according to the protocol of Miller (45). Removal of the kanamycin cassette within the mutant allele was done following the method of Cherepanov and Wackernagel using the temperature-sensitive plasmid pCP20 (46). The loss of the cassette was tested by colony PCR using oligonucleotides that annealed approximately 200 bp upstream and downstream of the respective genes (oligonucleotides named del_fw and del_rev) (see Table S1). Strain MC903 (ΔgrcA pflB-G734A) was created by transfection of MC801 (pflB-G734A) with P1::ΔgrcA and subsequent removal of the Kan^r cassette.

Finally, the formate-responsive lacZ reporter was introduced into the chromosomal lambda phage attachment site in different strains using λ phage-mediated transduction of λfdhF_P::lacZ (29) and resulted in strains DH234M1 (ΔpflA λfdhF_P::lacZ), DH801 (pflB-G734A λfdhF_P::lacZ), DH901 (ΔgrcA λfdhF_P::lacZ), and DH903 (pflB-G734A ΔgrcA λfdhF_P::lacZ).

A DNA fragment encompassing the grcA gene, including 30 bp of the upstream sequence, was cloned into the pUC19 vector (Table 2) using BamHI and HindIII restriction sites (oligonucleotides for PCR grcA_HindIII_fw and grcA_BamHI_re) (Table S1). The resulting plasmid pUC19::grcA was then used as a template for mutagenesis to substitute the codon for glycine 102 with one decoding as alanine (Agilent QuikChange method), which delivered plasmid pUC19::grcA-G102A. Replacement of the two codons for the vicinal cysteines (C418 and C419, both TGC codon) with codons for alanine (both GCC codons) was achieved using plasmid p29 (47) (Table 2) as a template for mutagenesis with nonoverlapping oligonucleotides pflB_C419A_C420A_fw and pflB_C419A_C420A_re (Table S1). The PCR fragment was ligated intramolecularly and circularized using the kinase, ligase, and DpnI (KDL) enzyme mix as described by the manufacturer (New England BioLabs). The plasmid was called p29-C2 and resulted in the synthesis of PflB_C418A/C419A.

Assay of β-galactosidase enzyme activity.

Changes in intracellular formate levels were monitored using the β-galactosidase enzyme activity assay (45). Cells were harvested in the exponential growth phase (OD₆₀₀ of ∼0.7 to 0.9) and measurements of the activity were carried out following the established protocol (31). For each experiment, minimally biological replicates were analyzed, and the assay was performed with technical replicates. The data are presented with the standard deviation of the mean.

Analysis of extracellular formate and lactate.

A high-performance liquid chromatography (HPLC) analysis of organic acids in the culture supernatant was carried out according to the method described (48). The data are referenced with respect to the optical density (OD₆₀₀) and depicted with standard deviation of the mean and originate from the analysis of at least three biological replicates.

Analysis of H₂ production.

The analysis of H₂ production was used as an indirect method to draw conclusions regarding intracellular formate levels and was determined by gas chromatography as described previously (48) because formate must be present for H₂ to be produced. The cumulative H₂ production was determined minimally for three biological replicates and is presented in reference to the optical density (OD_{600 nm}). The data are presented with standard deviation of the mean.

SDS-PAGE and Western blot analysis of PflB.

Different pflA, pflB, and grcA mutant strains were cultivated anaerobically in serum bottles and harvested in the exponential growth phase (OD₆₀₀ of ∼0.7 to 0.9). The cells were harvested by centrifugation, suspended in 50 mM Tris-HCl (pH 7), and disrupted by sonification as described (31). The crude cell extract containing PflB, which had been exposed to O₂ during cell disruption, was analyzed by denaturing sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (49) and immunoblotting (50). Aliquots of 30 to 50 μg of protein derived from crude extracts were separated electrophoretically in SDS-PAGE, including 7.5% (wt/vol) polyacrylamide or in an 8% (wt/vol) TG Prime SERVAGel (Serva, Heidelberg Germany). The proteins were visualized either via silver staining using the SERVA silver staining kit for SDS-PAGE or via fluorescence. For fluorescence detection, the protein samples were labeled with 12.5 μg · mL⁻¹ Cy3 dye (Cytiva, Freiburg, Germany) and excited with a wavelength of 535 nm. Proteins were transferred to nitrocellulose membranes as described (31, 50). The membranes were challenged with anti-PflB antiserum (4) in a dilution of 1:1,000. Visualization of PflB relied on the enhanced chemiluminescent reaction (Agilent Technologies).

Computational methods.

Studies on the co-occurrence of grcA and pflB relied on the STRING online tool (version 11.5) using the E. coli sequences as the input (51).

Data availability.

All strains and data presented in the manuscript are available upon request from the corresponding author.