Repurposing lipoic acid changes electron flow in two important metabolic pathways of Escherichia coli

Abstract

In bacteria, cysteines of cytoplasmic proteins, including the essential enzyme ribonucleotide reductase (RNR), are maintained in the reduced state by the thioredoxin and glutathione/glutaredoxin pathways. An Escherichia coli mutant lacking both glutathione reductase and thioredoxin reductase cannot grow because RNR is disulfide bonded and nonfunctional. Here we report that suppressor mutations in the lpdA gene, which encodes the oxidative enzyme lipoamide dehydrogenase required for tricarboxylic acid (TCA) cycle functioning, restore growth to this redox-defective mutant. The suppressor mutations reduce LpdA activity, causing the accumulation of dihydrolipoamide, the reduced protein-bound form of lipoic acid. Dihydrolipoamide can then provide electrons for the reactivation of RNR through reduction of glutaredoxins. Dihydrolipoamide is oxidized in the process, restoring function to the TCA cycle. Thus, two electron transfer pathways are rewired to meet both oxidative and reductive needs of the cell: dihydrolipoamide functionally replaces glutathione, and the glutaredoxins replace LpdA. Both lipoic acid and glutaredoxins act in the reverse manner from their normal cellular functions. Bioinformatic analysis suggests that such activities may also function in other bacteria.

Reduction-oxidation reactions play key roles in many essential metabolic pathways. Some redox reactions involve the oxidation or reduction of thiol residues either in an enzyme's cysteines or in small redox-active molecules. In Escherichia coli, the thiol-disulfide biology of the cell is compartmentalized; the majority of protein thiols in the periplasm are oxidized (disulfide bonded), and the majority of protein thiols in the cytoplasm are reduced. However, for cytoplasmic enzymes that use cysteines in catalysis of reductive reactions, disulfide bonds do form, albeit transiently. These bonds are rapidly reduced, restoring the enzyme's activity.

Two pathways maintain protein thiols in the reduced state in the cytoplasm of E. coli (1). In the thioredoxin pathway, thioredoxin 1 (encoded by trxA) and thioredoxin 2 (trxC) reduce oxidized substrates. The resulting oxidized thioredoxins are then reduced by thioredoxin reductase (encoded by trxB) to regenerate their activity. In the glutathione/glutaredoxin pathway, glutaredoxins 1 (grxA), 2 (grxB), and 3 (grxC) can reduce oxidized substrate proteins. The small molecule thiol glutathione (GSH) and glutathione reductase (gor) provide electrons to maintain glutaredoxins in the reduced state.

The importance of the TrxB and Gor pathways is indicated by the finding that when null mutations in certain genes of both pathways are combined, E. coli cannot grow. The reason for this synthetic lethality is that the essential enzyme ribonucleotide reductase (RNR) must be reduced by these pathways to maintain its activity. In certain synthetic lethal combinations of mutations (e.g., gor trxB), but not others (trxA trxC grxA), cell growth can be restored by the addition of an exogenous reducing agent, such as DTT. The reducing agent restores electron flow into one or the other of these defective thiol-redox pathways and in turn enables the reduction of RNR. The growth of the trxA trxC grxA mutant strain can be restored simply by overexpression of RNR, which then can be reduced by the weak reductant, glutaredoxin 3, indicating the key relevance of RNR to the growth defect (2, 3).

We have previously exploited the essentiality of RNR reduction to select for mutations that suppress the growth defect of strains defective in both the thioredoxin and glutathione/glutaredoxin pathways. For example, when the DTT required for growth of a strain that lacks TrxB and Gor is removed, suppressor mutations allowing growth arise at a high frequency. These suppressors carry mutations in the gene ahpC, which encodes a peroxidase (peroxiredoxin), AhpC. The mutations in ahpC altered the substrate specificity of the enzyme, converting it from a peroxidase to a disulfide reductase, and restoring electron flow to the glutathione/glutaredoxin pathway (4, 5).

We wished to determine whether pathways of electron transfer to RNR in addition to the AhpC enzyme described above could be evolved in E. coli. However, the frequency of suppressor mutations in the ahpC gene was so high in the trxB gor strain that it masked other potential suppressor mutations. We report here the isolation of suppressor mutations of a trxB gor ahpCF triple-deletion strain, in which no ahpC suppressor mutations can occur. This new class of suppressor mutations all map to the gene for the oxidative enzyme lipoamide dehydrogenase, lpdA. Lipoic acid, a small-molecule thiol-redox compound with a disulfide bond, ordinarily acts as an oxidant; it is a cofactor for three multienzyme complexes in E. coli and is covalently attached to lysine residues on proteins in these complexes in the form of lipoamide. Our evidence indicates that these mutations suppress the redox defect by lowering the ability of LpdA to convert the reduced dihydrolipoamide to the oxidized lipoamide, leading to increased amounts of dihydrolipoamide in these protein complexes, which then serves as a source of electrons for the reduction of oxidized glutaredoxins. Further, it appears that in these suppressor strains, oxidized glutaredoxins are now required for a functional TCA cycle because they oxidize dihydrolipoamide, providing a substitute for the missing LpdA activity. In effect, this combination of mutations results in alternative electron transfer steps for two essential intracellular pathways—that for reduction of ribonucleotide reductase, and the oxidation steps in the TCA cycle—leading to a striking change in the fundamental redox biology of the cell.

Based on these studies, we suggest that there may be other bacterial species that use dihydrolipoamide as a source of cytoplasmic reducing power. Our genomic bioinformatic analyses of a large number of bacterial species indicate in which organisms lipoamide/dihydrolipoamide might serve a function outside of its canonical roles in metabolism.

Results

Mutations in the Gene for Lipoamide Dehydrogenase Restore Growth to a trxB, gor, ahpCF Mutant.

To obtain novel suppressor mutations of a strain lacking thioredoxin reductase (TrxB) and glutathione reductase (Gor), we used a triple mutant, SMG89, deleted for trxB, gor, and ahpCF. Construction of the triple-mutant strain, selection for suppressor mutations that restored growth, and mapping of the suppressor mutation are described in SI Materials and Methods and Fig. S1. We mapped the suppressor mutation in one of the strains, SMG123, to the region containing a known oxidoreductase, lpdA. Sequencing of lpdA in SMG123 revealed a point mutation that changed a conserved glycine residue into an aspartate (G186D). The five other suppressor strains from the genetic selection also contained mutations in lpdA. All six mutations (Gly183Cys, Gly183Ser, Ser164Phe, Ser164Tyr, Gly271Asp, and that of SMG123) changed small, uncharged amino acids into larger and/or charged residues (Fig. 1A). Gly271Asp and Ser164Phe were each isolated in two independent selections.

Fig. 1.

Mutations in the gene encoding lipoamide dehydrogenase suppress the trxB gor ahpCF mutant. (A) Crystal structure of the P. putida LpdA homolog showing the region surrounding the NAD⁺ cofactor (depicted in red) and the residues found mutated in the trxB gor ahpCF suppressor strains. (B) Growth of the suppressor strain SMG123 (trxB gor ahpCF lpdA_G186D) and the quadruple-mutant strain MAF180 (trxB gor ahpCF lpdA). Cultures were grown in triplicate, in NZ amine media (solid lines) or NZ amine media + lipoic acid (dashed lines) at 37 °C, and growth was followed by measuring the increase in OD₆₀₀ over 12 h. A representative curve is shown; the growth curve was repeated three times. (C) Expression of wild-type lpdA and two suppressor mutants, lpdA_G186D and lpdA_G271D, cloned under the control of an IPTG-inducible promoter on a low copy-number plasmid. Strains carrying the indicated plasmids were streaked on rich NZ media containing spectinomycin to select for the plasmid and 1 mM IPTG, and were grown at 37 °C for 2 d.

To confirm that mutations in lpdA were responsible for the suppression of the cytoplasmic redox defect, we cloned lpdA_G186D and lpdA_G271D into a low-copy plasmid under the control of an isopropyl β-d-thiogalactopyranoside (IPTG)-inducible promoter. However, neither of the plasmid-encoded mutant lpdA alleles suppressed the growth defect of the trxB gor ahpC strain (Fig. 1C). Furthermore, expression of a plasmid-encoded wild-type copy of lpdA abolished growth of the suppressor strain SMG123, suggesting that the suppressor mutations were recessive and that they may have reduced LpdA activity. These results led us to ask whether a complete knockout of lpdA might also suppress the redox defect of strain SMG89. We constructed a quadruple-deletion strain (trxB gor ahpC lpdA; SI Materials and Methods). This strain (MAF180), although it grew very poorly, did grow to some extent without the addition of an exogenous reductant (Fig. 1B), indicating weak suppression. The addition of (oxidized) lipoic acid to the growth media did not improve the growth of the quadruple mutant in liquid culture (Fig. 1B), but did improve growth on solid media as measured by larger colony size (Fig. S2). Plasmid expression of lpdA_G186D or lpdA_G271D improved the growth of the quadruple mutant containing the lpdA deletion considerably, confirming that these mutant alleles of lpdA suppressed the redox defect as long as the wild-type LpdA was not present (Fig. 1C).

Suppressor Mutations in lpdA Decrease Lipoamide Dehydrogenase Activity.

lpdA encodes lipoamide dehydrogenase, an oxidative enzyme that plays a role in central metabolism as the E3 component of the pyruvate dehydrogenase and α-ketoglutarate dehydrogenase complexes, and as the l-protein of the glycine cleavage complex. LpdA oxidizes dihydrolipoamide moieties covalently bound to lysine residues in proteins in these complexes, thus regenerating active, oxidized lipoamide that has been reduced in the process of oxidizing its substrates. Electrons are transferred by LpdA from dihydrolipoamide to an FAD moiety bound to the enzyme, which, in turn, transfers electrons to NAD⁺.

We mapped the amino acid alterations of LpdA that caused suppression onto the structure of a Pseudomonas putida homolog of LpdA (6). All six amino acid changes altered amino acids that are close to or within the binding site of NAD⁺, the external oxidant of LpdA (Fig. 1). The FAD binding site is also close to this region. The proximity of the amino acid changes isolated as suppressor mutations to the cofactors essential for LpdA activity led us to ask whether the suppressor mutations disrupted LpdA function.

Whereas wild-type E. coli can grow on minimal glucose, strains that lack lpdA cannot; acetate and succinate are required to support the growth of this mutant on minimal glucose media (7). To assess the effect of the suppressor mutations on LpdA activity, we transduced the mutations into an otherwise wild-type background (DHB4) and tested for the ability of these strains to grow on minimal glucose media. Like the lpdA null mutant, our suppressor mutants in the wild-type background do not grow on minimal glucose unless supplemented with acetate and succinate (Fig. 2A).

Fig. 2.

A suppressor mutation in the lpdA gene reduces dihydrolipoamide oxidation. (A) Growth of strains with various lpdA alleles on minimal glucose media. Strains were streaked on minimal glucose media containing leucine and isoleucine, and grown at 37 °C for 1–2 d. (B) Measurement of SucB bound reduced dihydrolipoamide in DHB4 (wild-type) and the suppressor strain SMG123 (trxB gor ahpCF lpdA_G186D). Total cellular protein isolated from stationary phase growing E. coli was treated with the thiol alkylating agent biotin-maleimide, and serial dilutions of the treated protein samples were added to plates coated with anti-FLAG antibody to specifically bind the SucB (encoding a C-terminal FLAG tag) protein. Biotin maleimide-thiol content in dihydrolipoamide bound to SucB was estimated by streptavidin-HRP treatment followed by monitoring A450.

The phenotypic properties of these strains indicate that the suppressor mutations lower lipoamide dehydrogenase activity, which should result in the accumulation of reduced lipoamide (dihydrolipoamide) to higher levels than wild-type cells. To detemine whether this was the case in the trxB gor ahpCF lpdA suppressor strains, we measured the amounts of dihydrolipoamide covalently bound to a FLAG-tagged SucB, the E2 protein of the α-ketoglutarate dehydrogenase complex, using a direct sandwich ELISA. We did this by comparing the content of reduced thiols in SucB in the wild-type and a suppressor strain. Because SucB contains no cysteine residues, free thiols detected could only be due to the thiols in dihydrolipoamide. We detected the free thiols by reacting them with a biotin-maleimide alkylating agent, which would only form covalent bonds with the reduced form of lipoamide. Strikingly, we saw an 8- to 11-fold increase in the amount of the reduced form of lipoic acid, dihydrolipoamide, bound to the SucB complex in SMG123 over that seen in the wild-type strain, DHB4 (Fig. 2B and Fig. S3).

Dihydrolipoamide Is Required for Suppression.

Our findings indicate that the suppressor mutations reduce LpdA activity, thus causing accumulation of the reduced form of lipoamide, dihydrolipoamide. This evidence raised the possibility that dihydrolipoamide was responsible for the suppression and acted as a reductant that compensated for the redox defect of the trxB gor ahpC strain. If this were the case, the pathways for lipoamide biosynthesis should be essential for suppression.

There are two pathways by which lipoylated proteins are generated in E. coli, one using lipoic acid added to growth media and the other using endogenously synthesized lipoate. The lplA pathway uses the lplA gene product to ligate imported lipoic acid onto substrate proteins; the lipA-lipB pathway both synthesizes endogenous lipoate and ligates it onto proteins (8). Neither pathway is essential for the growth of wild-type E. coli on rich medium. However, we wished to determine whether these pathways had become essential for growth of the trxB gor ahpC lpdA_G186D strain.

We constructed derivatives of the suppressor strain SMG123, which lacked either the lipAB or lplA pathway, but had a complementing plasmid (pBAD18-trxB) to allow for growth of the mutant strains (SI Materials and Methods). The trxB gor ahpCF lpdA_G186D lipA/pBAD18-trxB strain (MAF493) is unable to grow unless arabinose is added to induce expression of the complementing trxB (Table 1). The growth defect of strain MAF493 was overcome by addition of lipoic acid to the growth media. The restoration of growth is presumably due to the presence of the lplA pathway. Thus, we constructed versions of the suppressor strain that lacked both pathways for lipoyl protein ligation (lipA lplA and lipB lplA mutant strains). Addition of exogenous lipoate to the growth media did not rescue the growth of either the lipA lplA or lipB lplA strain when expression from the complementing pBAD18-trxB plasmid was shut off.

Table 1.

Requirements for suppression

	Deletion in suppressed strain	trxB expressed	trxB not expressed
	None	++	+
Lipoamide biosynthesis	lipB	+	0
	lipA	+	0
	lplA	++	+
Lipoylated proteins	sucB	+	0
	gcvH	+	0
Glutaredoxins	grxA	+	±
	grxC	++	+
	grxA, grxC	+	0
Thioredoxins	trxA	++	+
	trxC	++	+
	trxA, trxC	++	+

Deletions of genes involved in lipoamide biosynthesis, and of the glutaredoxins and thioredoxin genes, were transduced into SMG181 (the trxB gor ahpC lpdAsupp strain, SMG123, carrying a complementing plasmid, pBAD18-trxB) by P1 transduction. The resulting transductants were tested for their ability to grow on NZ/ampicillin media with either 0.2% glucose (to repress transcription of trxB) or with 0.2% arabinose (to induce trxB). ++, comparable growth to that of SMG123; +, smaller colonies; ±, very small colonies; 0, no colonies observed after 3 d of incubation.

These data suggest that dihydrolipoamide is required for suppression. Our results indicate that each of the lipoylated proteins in E. coli may be individually required for suppression as well. We constructed derivatives of the suppressor strain, SMG181, which lacked the lipoylated proteins SucB and GcvH. These strains, like the version of the suppressor strain that lacked the genes for lipoamide biosynthesis, required arabinose (trxB induction) for growth (Table 1 and Fig. S4). We were unable to construct a version of SMG181 that lacked aceF, suggesting that it too may be essential.

Dihydrolipoamide Restores Electron Flow Through the Glutaredoxin Pathway.

The evidence that lipoamide biosynthesis is required for suppression and that dihydrolipoamide is present in the suppressor strain strengthened our hypothesis that it serves as a source of electrons for reduction of protein disulfides in substrates such as ribonucleotide reductase (RNR). Dihydrolipoamide has a low redox potential (−290 mV)—even more reducing than glutathione (−240 mV)—and is therefore a good potential source of reducing power in the cell (9). Furthermore, in vitro studies have shown that dihydrolipoamide can drive reduction of the glutaredoxins (10).

We considered the possibility that dihydrolipoamide might act analogously to glutathione in suppressor strains, serving as a source of reducing equivalents for intermediary proteins that can reduce RNR, such as the glutaredoxins or the thioredoxins. Therefore, we asked whether thioredoxins or glutaredoxins were required for suppression in the trxB gor ahpCF suppressor strain SMG123. The thioredoxins were not required; however, deletion of grxA (glutaredoxin 1) from SMG181 drastically reduced growth of the strain (MAF502). There was a small amount of growth in the strain carrying the deletion of grxA (tiny colonies after 3 d growth on rich media at 37 °C). Because grxC is up-regulated under conditions of oxidative stress, and because it is known to be reduced by dihydrolipoamide in vitro (10), we asked whether grxC was responsible for the residual growth of MAF502 by introducing deletions of both grxA and grxC into the suppressor strain. This strain failed to grow without expression of trxB from the complementing plasmid, indicating that the glutaredoxins are required for suppression (Table 1 and Fig. S4).

Oxidation of Dihydrolipoamide by the Glutaredoxins.

The indication that dihydrolipoamide provides electrons for the reduction of RNR in the suppressor strains also makes it likely that dihydrolipoamide itself becomes oxidized upon reducing whatever substrate it acts on (likely the glutaredoxins). If this were the case, the very activity of promoting the reduction of RNR (and other substrates) should allow the resultant oxidized lipoamide to participate in the 2-oxoacid dehydrogenase reactions of the TCA cycle. Therefore, we predicted that the trxB gor ahpC lpdAsupp strain would be able to grow on minimal glucose media lacking acetate and succinate, even though it is lacking the activity of an enzyme, LpdA, ordinarily essential for oxidation in the TCA cycle. This prediction was fulfilled; whereas the mutant lpdA alleles in a wild-type background conferred a requirement for acetate and succinate on minimal glucose media, paradoxically, the suppressor strains grew on minimal glucose showing no such requirement (Fig. 3A). Thus, at the same time that the lpdA mutations suppressed the growth defect of the trxB gor ahpC mutant, the alteration of the redox state of glutaredoxins in this same background appears to have provided a source of oxidation of dihydrolipoamide, in effect, replacing LpdA (Fig. 3).

Fig. 3.

Glutaredoxins replace the function of LpdA in the oxidizing cytoplasm. (A) Strains were grown on M63 minimal glucose media (supplemented with leucine and isoleucine) at 37 °C for 2 d. (B) Strains were grown on M63 minimal glucose media (supplemented with leucine and isoleucine) at 37 °C for 2 d, with either 0.2% arabinose (Right) or 0.2% arabinose plus 5 mM acetate and 5 mM succinate (Left) for 3 d at 37 °C. Counterclockwise from the top, the strains are: MAF259 (trxB gor ahpCF lpdA_G186D/pBAD18-trxB), MAF501 (MAF259 grxC), MAF502 (MAF259 grxA), and MAF503 (MAF259 grxA grxC). (C) Proposed scheme for the oxidation of dihydrolipoamide and reduction of glutaredoxins in the suppressor strains.

These results suggested that the glutaredoxins had become essential for the functioning of the TCA cycle in the suppressor strains. We therefore asked whether the trxB gor ahpC lpdA-supp strain could grow on minimal glucose media when grxA and grxC were deleted. On media containing arabinose, trxB is induced from the complementing plasmid, pBAD18-trxB, and the glutaredoxins are not required for the reduction of RNR. The suppressor strain should therefore be able to grow on minimal glucose plus arabinose, independently of the glutaredoxin pathway. However, like an lpdA null strain, the trxB gor ahpC lpdAsupp grxA grxC strain grows only when acetate and succinate are added to supplement the TCA cycle. (Fig. 3B)

Cytoplasmic Disulfide Bond Formation in the Suppressor Strain.

The cytoplasm is normally a reducing environment, where any disulfide bonds formed in proteins are reduced by thioredoxins or glutaredoxins. However, in the absence of TrxB, the thioredoxins accumulate in the oxidized form and can oxidize substrate proteins that are derivatives of proteins missing their signal sequences (11). Even more efficient production of disulfide-bonded proteins occurs in a strain (SMG96) deleted for both the trxB and gor genes and carrying a suppressor mutation in the ahpC gene that restores growth (12). We asked whether the suppressor strain described here would also exhibit high levels of disulfide bond formation in proteins. We introduced a plasmid encoding signal sequenceless tissue plasminogen activator (vtPA), which has multiple disulfide bonds arranged in a complex pattern, as well as a plasmid encoding a signal sequenceless disulfide isomerase (DsbC). We found that there is about a 119.4 ± 10.9 (n = 3) increase in vtPA activity in SMG96 compared with a wild-type strain (DHB4), whereas SMG123 exhibits an ∼234.4 ± 41.4.1 (n = 2) increase compared with DHB4 (approximately a twofold increase over the levels seen in SMG96).

Possible Alternative Roles of Lipoamide Dehydrogenase and Lipoamide in Other Organisms.

Our results suggest that lipoamide may be able to function outside of its canonical role in metabolism, at least in the suppressor strains described here. Past work has shown that E. coli mutants selected for alterations of their thiol redox pathways can exhibit phenotypes that anticipate subsequently discovered novel redox biology in other organisms (12, 13). Therefore, we were inspired to seek evidence that dihydrolipoamide might reduce the glutaredoxins in other bacteria. It was already known that some organisms, such as certain archaea, possess a gene encoding a lipoamide dehydrogenase, but lack its usual substrates such as the dehydrogenase complexes of the TCA cycle (14). Although the role of LpdA homologs in these organisms is unknown, their existence led us to speculate that some organisms might use lipoamide dehydrogenase and lipoamide for novel functions, including possibly the reductive pathways we have found in the trxB gor ahpC suppressor SMG123.

We initiated a bioinformatic analysis of bacterial genomes to identify candidate organisms that might use dihydrolipoamide as a source of reduction for its glutaredoxins. In this search, we identified a list of organisms that lacked the capacity for glutathione biosynthesis, but did possess the genes for glutaredoxins and lipoamide biosynthesis (Table S1). Some of the organisms identified in our search are those that might be expected to use other thiol redox pathways, such as the Bacillus species, which use bacillithiol (15), or the Actinobacteria, which use mycothiol (16), small-molecule thiols analogous to glutathione. However, others may lack the thiol redox pathways and thus require a reductant for their glutaredoxins. Moreover, among these 40-some organisms, we identified a subset that also lacked the genes for the E2 components of the dehydrogenase complexes (Table 2). These organisms presumably do not use lipoamide and dihydrolipoamide dehydrogenase for their canonical metabolic functions, but instead may use (dihydro)lipoamide to carry out other redox reactions.

Table 2.

Bioinformatic analysis of organisms that could potentially use a novel pathway for dihydrolipoamide

The number of homologs of each gene is given in the columns; these organisms lack gshA and the pathway for glutathione biosynthesis, but possess proteins homologous to glutaredoxins (Grx). At the same time, they have the biosynthetic capacity for dihydrolipoamide biosynthesis, but appear to lack homologs of the E2 proteins which in E. coli are lipoylated and function in the 2-oxoacid dehydrogenase complexes. Organisms highlighted in gray possess the E2 homologs that participate in the dehydrogenase complexes, but also have a second homolog of LpdA that contains an N-terminal lipoyl domain, and that could potentially act independently as an oxidoreductase.

In addition to identifying genomes that lacked E2 homologs, we also found that two of the lip⁺ gsh⁻ grx⁺ genomes identified in our bioinformatic search had LpdA homologs with an N-terminal lipoyl domain (Table 2). These LpdA homologs, in effect, could contain their own substrate and thus have an oxidoreductase activity separable from the dehydrogenase complexes to which LpdA normally belongs. Lipoyl domain-containing LpdA homologs have been previously identified in bacteria such as Streptococcus, Neisseria, and Clostridia, among others (17–20). The physiological role of the lipoyl-domain lipoamide dehydrogenases remains to be elucidated, but given that their activity is separable from the dehydrogenase complexes, it may be that these proteins can also function in pathways similar to the one that we have described in this work.

Discussion

In this paper, we describe a remarkable shift in two electron transfer pathways important for cellular function. These two redesigned pathways are the result of a combination of mutations that inactivate the reductive thioredoxin and glutathione/glutaredoxin pathways and, separately, alter the function of the oxidative enzyme lipoamide dehydrogenase (LpdA). The lpdA mutations were obtained by in vivo genetic selections in E. coli trxB gor ahpC cells. As a consequence of these changes, (i) the source of electrons to reduce and maintain activity of such enzymes as ribonucleotide reductase is shifted from NADPH to dihydrolipoamide and (ii) the oxidative power to generate the oxidized lipoamide necessary for TCA cycle function is shifted from NAD to oxidized glutaredoxins. In this way, the mutant cells have reversed the normal role of two key redox components, lipoic acid and glutaredoxin. There is currently widespread interest in engineering bacteria to carry out new tasks by generating significant changes in metabolic pathways or by incorporating new metabolic pathways (for examples, see refs. 21–23). Our findings further emphasize the metabolic and genetic plasticity of electron flow pathways in bacteria.

Although reduced lipoic acid is a powerful reductant, the known physiological functions of lipoic acid are oxidative. However, in vitro work has shown that dihydrolipoamide can serve as a source of electrons for the reduction of glutaredoxins (10) and for the transfer of electrons to the M. tuberculosis peroxidase AhpC (24) and the organic hydroperoxide resistance protein Ohr from Xylella fastidiosa (25).

These in vitro studies leave open the question of whether reduced dihydrolipoamide can serve in vivo as an electron donor for enzymatic catalysis. There are many factors (competition, specificity, accessibility, oxidation by LpdA or other oxidants, etc.) that might have limited the use of the strong reductive power of dihydrolipoamide in physiological pathways. Our findings show that reduced dihydrolipoamide can serve as a source of electrons for key reactions in the cell. Because our results establish that these reactions can take place in vivo, we asked whether there are organisms that might naturally take advantage of dihydrolipoamide as a source of electron equivalents. Our bioinformatic analysis raises the possibility that the glutaredoxin/lipoamide coupling in vivo in E. coli may function in several other bacteria.

Several lines of evidence show that the lpdA suppressor mutations are defective in lipoamide dehydrogenase (LpdA) activity: (i) The mutants are all recessive to the wild-type allele of lpdA; (ii) The mutations, when transferred into a wild-type background, all behave phenotypically like lpdA null mutations; (iii) In an lpdA suppressor strain (SMG123), direct assay of the redox state of the lipoamide bound to one of the TCA cycle complexes shows a substantial increase in the amount of reduced lipoamide (dihydrolipoamide) compared with a wild-type strain. The defects in LpdA activity prevent the oxidation of dihydrolipoamide to lipoamide, allowing dihydrolipoamide to accumulate in its reduced state; and (iv) All of the mutations cause changes in or very close to the NAD binding site of LpdA, indicating that they could be interfering with electron flow between NAD and the FAD bound to LpdA.

We have shown that when we introduce into the lpdA suppressor strains mutations that lack the dihydrolipoamide biosynthetic pathway, the strains become dependent on exogenous lipoic acid for growth. Further, the ability of the exogenous lipoic acid to restore growth is dependent on lipoyl ligase (lplA). These results indicate that lipoic acid covalently bound to TCA cycle complexes is essential for the suppressor effect of the lpdA mutations. We are led to conclude that the enzyme-bound dihydrolipoamide has replaced glutathione as a source of electrons for reduction of disulfide bonds in enzymes such as RNR. The dependency of the suppression on glutaredoxins indicates that this reduction is accomplished by the transfer of electrons first from dihydrolipoamide to glutaredoxins and thence to RNR. Our results indicate that the reduction of glutaredoxins by dihydrolipoamide, which had been previously shown to occur in vitro (10), can occur in vivo to a physiologically significant extent.

Our conclusion that oxidized glutaredoxins can functionally replace lipoamide dehydrogenase is based on the finding that though wild-type strains carrying the lpdA suppressor mutations are unable to grow on minimal glucose media unless acetate and succinate are added, counterintuitively, the trxB gor ahpC strain carrying the same lpdA mutants do grow on minimal media. Thus, the functioning of the TCA cycle, which is abrogated in wild-type cells carrying the lpdA mutations, is restored in the trxB gor ahpC mutant carrying the very same lpdA mutations. The rewiring of these electron transfer pathways is shown in Fig. 4.

Fig. 4.

Model for the rewired electron transfer pathways in the trxB gor ahpC lpdAsupp strain. Arrows represent the direction of electron flow. In wild-type cells (A), NADPH is used as a source of reducing power for the thioredoxin and glutathione/glutaredoxin pathways, which reduce disulfide bonds in substrate proteins such as RNR. The oxidative reaction of lipoamide dehydrogenase, LpdA, is used to regenerate active, oxidized lipoamide for the function of the α-ketoglutarate and pyruvate dehydrogenases and the glycine cleavage complex. (B) In the trxB gor ahpC lpdAsupp strains, LpdA is inactive and dihydrolipoamide accumulates. The thioredoxin and glutaredoxin pathways are inactive, and oxidized glutaredoxins accumulate. These glutaredoxins oxidize dihydrolipoamide to lipoamide, allowing for the function of the multienzyme complexes, whereas the glutaredoxins are themselves reduced and able to go on to reduce substrate proteins.

Interestingly, the mutations we isolated as suppressors of the trxB gor ahpC strain were all missense mutations. This finding, along with the result that a deletion of the lpdA gene acts only very weakly as a suppressor, suggests that the restoration of growth to the gor trxB ahpC strain by the lpdA suppressor mutations depends on the presence of an intact LpdA protein even though it is mutant. This property of the mutants could be because the presence of the LpdA protein is necessary to stabilize the structure of the dihydrolipoamide-containing multienzyme complexes or because a small amount of residual activity in the mutant proteins is necessary for suppression. Our results suggest that the latter explanation may be correct, as the expression of a catalytically inactive lipoamide dehydrogenase does not suppress. However, the oxidation of dihydrolipoamide in these suppressor strains seems to be largely due to the glutaredoxins, as the deletion of both glutaredoxins abolishes TCA cycle function.

Finally, we have shown that this class of suppressor strains is able to produce correctly folded, active disulfide bonded proteins in its cytoplasm and is potentially useful for recombinant protein production. Because of slow growth, the use of the strain for this purpose is currently limited. However, addition of lipoic acid to the media gives the suppressor strain SMG123 a moderate increase in growth (Fig. 1C). Therefore, it may be possible to improve the growth of this strain further, while maintaining efficient cytoplasmic disulfide bond formation, by additional mutations or by optimizing media conditions.

Materials and Methods

Strains and Growth Conditions.

Strains and plasmids used in this work are listed in Table S2 and primers are listed in Table S3. All strains were grown on rich media (NZ-amine) (26) at 37 °C unless specified otherwise. Additional details of growth conditions are in SI Materials and Methods.

Determination of Reduced Dihydrolipoamide Levels.

Strains carrying the C-terminally FLAG-tagged sucB were grown to stationary phase in LB media. After cells were lysed with a French press, cell debris was removed by centrifugation and the proteins concentrated using Millipore Ultra-4 Centrifugal Filters (molecular weight cutoff 3,000). Biotin-maleimide (5 mM) was added to the protein concentrate and allowed to react with free thiols for 20–30 min at 37 °C. Excess biotin-maleimide was removed using PD-10 desalting columns (Millipore). Serial dilutions of these samples were then added to ELISA plates that had been coated with anti-FLAG antibody, incubated at room temperature for 1 h, and washed three times with PBS Tween-20. Streptavidin-HRP (Sigma) was added at 1:10,000 in PBS + 1% milk and incubated at room temperature for 1 h. After three washes with PBST, 75 μL ultra TMB was added and the reaction allowed to proceed for 30 min before it was stopped with 75 μL of 2 M H₂SO₄. Absorbance was then measured at 450 nm.