A New Metabolic Link

Lipoic acid is an essential enzyme cofactor that requires covalent attachment to its cognate proteins to confer biological activity. The major lipoylated proteins are highly conserved enzymes of central metabolism, the pyruvate and α-ketoglutarate dehydrogenase complexes. The classical lipoate ligase uses ATP to activate the lipoate carboxyl group followed by attachment of the cofactor to a specific subunit of each dehydrogenase complex, and it was assumed that all lipoate attachment proceeded by this mechanism. However, our previous work indicated that Escherichia colicould form lipoylated proteins in the absence of detectable ATP-dependent ligase activity raising the possibility of a class of enzyme that attaches lipoate to the dehydrogenase complexes by a different mechanism. We now report that E. coli and mitochondria contain lipoate transferases that use lipoyl-acyl carrier protein as the lipoate donor. This finding demonstrates a direct link between fatty acid synthesis and lipoate attachment and also provides the first direct demonstration of a role for the enigmatic acyl carrier proteins of mitochondria.

Lipoic acid is a cofactor required for function of the citric acid cycle (1). Two key enzyme multienzyme complexes use covalently bound lipoic acid to carry reaction intermediates between the active sites of the complexes. These enzymes are the pyruvate dehydrogenase complex (PDC) 1 that is responsible for the synthesis of the acetyl-CoA needed to prime the cycle and the ␣-ketoglutarate dehydrogenase complex (KGDC) that catalyzes the C 5 to C 4 step of the cycle (2). Covalently attached lipoic acid not only provides substrate channeling between the different active sites of these complexes but also maintains the activated carboxyl group in thioester linkage for delivery to coenzyme A. Covalent attachment of lipoic acid to the E 2 subunits of these enzyme complexes is required for activity in vivo. The cofactor is attached via an amide linkage formed between the lipoate carboxyl group and the ⑀-amino group of a specific lysine residue of a well conserved E 2 subunit N-terminal domain (2). The first enzyme shown to attach lipoic acid to the lipoyl domains of PDC catalyzes a two-step reaction in which ATP activates lipoic acid to lipoyl-AMP followed by transfer of the lipoyl moiety to the appropriate E 2 lysine residue (3). Lipoate attachment is also required for activity of the glycine cleavage enzyme (GCV, the plant enzyme is often called glycine decarboxylase) (4,5) and the branched chain ␣-keto acid dehydrogenase complex (2).
All lipoate attachment was thought to proceed by the classical ATP-dependent mechanism, and Escherichia coli contains such a lipoate ligase (LplA protein) that functions primarily in utilization of exogenously supplied lipoic acid (6,7). E. coli strains having null mutations in the lplA gene lack the ability to attach exogenously added lipoic acid to protein and also lack detectable ATP-dependent lipoate ligase activity (6,8). However, these strains retain the ability to attach endogenously synthesized lipoate to PDC and KGDC in vivo indicating the presence of a second enzyme (6,7). We report that this enzyme represents a second class of lipoate attachment enzyme that utilizes the lipoyl thioester of the fatty acid synthetic protein, acyl carrier protein (ACP), as the source of lipoic acid.
ACPs are an extensive family of small (70 -80 residues) acidic proteins modified by covalent attachment of 4Ј-phosphopantetheine to a centrally located serine residue. ACPs carry acyl groups via thioester linkage to the 4Ј-phosphopantetheine sulfhydryl group and were first demonstrated to function in the fatty acid synthetic pathways of bacteria and plants (reviewed in Refs. 9 and 10). Subsequently, the ACPs of fatty acid synthesis have been shown to act as acyl donors in the synthesis of phospholipids (9), lipid A (11), the acylated homoserine lactone signaling molecules of bacteria (12), and in the activation of bacterial toxins (13). Other ACPs function in the synthesis of polyketide (14) and polypeptide antibiotics (15), cell wall polymers (16), and compounds regulating bacteria-plant interactions (17).
The reaction we report not only provides a new role in central metabolism for ACP but is an unusual reaction in that an enzyme cofactor (lipoic acid) is first bound by thioester linkage to another cofactor, 4Ј-phosphopantetheine, the covalently attached prosthetic group of acyl carrier protein. This lipoyl-ACP is then used to donate lipoic acid to the enzyme subunits of PDC, KGDC, and GCV. An alternate way of viewing the reaction is that lipoic acid is transferred from one protein to another. We have also demonstrated the presence of this new lipoate transferase (the formal name is lipoyl-[acyl-carrier-protein]-protein N-lipoyltransferase) in mitochondria of both plant and fungi thus providing a physiological role for the ACPs reported in mitochondria of mammals (18), plants (19 -21), and fungi (19,22,23).
Purification of ApoPDC/KGDC-A mixture of the apo (unmodified) PDC and KGDC complexes was purified from E. coli strain TM136 (lplA::Tn10 lipB::Tn1000dkan) grown on glucose minimal medium supplemented with acetate and succinate (6). This strain lacks the ability to attach either lipoate or octanoate to protein, and thus the PDC and KGDC from this strain are not lipoylated and lack activity (6). The mixed enzyme complexes were purified by differential centrifugation (25). When fully lipoylated in vitro the complexes had specific activities of 1.5 and 8.1 units/mg in the PDC and KGDC assays, respectively, where 1 unit is defined as the activity required to reduce 1 mol of 3-acetylpyridine adenine dinucleotide/min at 25°C.
Activation and Assay of the Apoenzyme Complexes-The ligation substrate was 0.2 unit of apoPDC in 10 mM sodium phosphate buffer (pH 7.0) containing 0.3 mM dithiothreitol. The reaction (50 l, final volume) also contained 80 g of lipoyl-ACP and 10 -100 g of extract protein (the extracts were from strains that lacked PDC activity due to deletion of the genes encoding the AceE and AceF subunits). The reactions were incubated for 10 -30 min at 37°C (E. coli) or 25°C (mitochondrial lysates), then stopped by bringing the volume to 500 l with 120 mM Tris-HCl (pH 8.5). The reactions were then assayed (25) for PDC or KGDC activity in a total volume of 1 ml using 3-acetylpyridine adenine dinucleotide as the electron acceptor and either 4 mM pyruvate or 4 mM 2-ketoglutarate as the substrate. The acyl-ACPs and lipoyl-ACP were synthesized with acyl-ACP synthetase (26) except that in the latter case 0.7 mM lipoic acid was substituted for the fatty acid substrate. E. coli cell extracts were made as described previously (7). Purified mitochondria from pea (Pisum sativum cotyledons) and Neurospora crassa were gifts from Wada et al. (21).
Gel Shift Assay-The N-terminal lipoyl domain (L1) of the dihydrolipoyltransacetylase subunit of E. coli PDC was produced in the unmodified form and purified as described previously (27). The gel shift assay was done as described previously (27). The E. coli biotin domain protein was the gift of Chapman-Smith et al. (28).
Assay of Transfer of Octanoate from ACP to Lipoyl Domain-The transfer of octanoate was tested by utilizing tritium-labeled octanoate. Octanoyl-ACP was synthesized using [8-3 H]octanoate (20 Ci/mmol) and was then used to modify the lipoyl domain using conditions identical to those of the gel shift assay. The reaction products were then subjected to electrophoresis using 15% SDS-polyacrylamide gels, and the tritiumlabeled bands were detected by fluorography.

RESULTS
Detection of a Novel Lipoate Transferase Activity-Two assays were used to detect attachment of lipoic acid (lipoylation) to the apo forms of PDC and KGDC. In the first assay lipoic acid attachment was measured by conversion of the inactive apo forms of PDC or KGDC (Fig. 1) to the active holo forms whereas the second assay followed the acylation-dependent shift in the electrophoretic mobility of a purified 80-amino acid residue apolipoyl domain from the E. coli PDC ( Fig. 2A). The second assay is much less sensitive but has the advantage that it can be used with any acyl donor because the mobility shift is due to loss of the positive charge of the lysine residue. Both assays detected an enzyme activity in E. coli cell extracts that catalyzed the transfer of lipoic acid from lipoyl-ACP to lipoyl domains. Moreover, this activity was present in extracts of E. coli lplA null mutants which lack detectable ATP-dependent ligase activity (7,8) (Figs. 1 and 2B), and unlike the ATP-dependent ligase (8) the activity was not inhibited by chelating agents. As anticipated from prior in vivo studies (6) the enzyme was also active with octanoyl-ACP (Fig. 2C). [ 3 H]Octanoyl-ACP was used to demonstrate direct transfer of the acyl chain from ACP to the lipoyl domain (Fig. 3), and the stoichiometry of the reaction was appropriate (1 mol of octanoyl-ACP was consumed per 1.4 mol of lipoyl domain modified, the discrepancy is attributed to hydrolysis of a portion of the octanoyl-ACP thioester linkages during electrophoresis). Conversion of the apo form of E. coli PDC to the enzymatically active holo forms requires attachment of lipoate to a specific lysine residue of the lipoyl domains (1,29). Therefore, activation of apoPDC (Fig. 1) demonstrated accurate modification of the lipoyl domains (activation of KGDC was also readily detected).
Specificity of the Lipoate Transferase-The gel mobility shift assay was used to determine the chain length specificity of the enzyme. The lipoate transferase was found to be much more active with the C 8 acid than with acids of shorter or longer chain lengths (Fig. 2C). The C 7 acid showed some activity, but since odd chain length acids are not present in vivo (9) the transferase has an appropriate chain length specificity. The enzyme had no detectable activity with octanoyl-CoA (data not shown). Moreover, octanoyl-CoA failed to inhibit domain modification by octanoyl-ACP, indicating that the detergent character of the acyl-CoA could not explain the lack of activity (data not shown). The specificity of the lipoate transferase for the protein acceptor was tested by use of the apo form of the 87-residue biotin-accepting domain of E. coli acetyl-CoA carboxylase (28). Although this protein has a structure remarkably similar to the lipoyl domain structure (30), upon substitution of the biotin-accepting domain for the lipoyl domain no modification of the biotin acceptor protein was detected by a gel shift assay (28) (data not shown; the biotin domain gave the expected mobility shift upon biotinylation in the presence of E. coli biotin protein ligase, biotin, and ATP).
E. coli lipB Mutants Lack Lipoate Transferase Activity-In vivo studies from this laboratory had indicated that E. coli strains having null mutations in both lplA and a second gene (lipB) were completely defective in the modification of lipoylated proteins (6). Moreover, lipB mutants were specifically defective in the attachment of endogenously synthesized octanoate (6). These data indicated that lipB mutants should be deficient in the transferase activity, and this prediction was confirmed. We were unable to detect lipoate transferase activity in extracts of lipB strains (Figs. 1 and 2B). Introduction of multicopy plasmids in which lipB was appropriately oriented downstream of a variety of powerful promoters gave only small increases in transferase activity. Moreover, in agreement with prior results from this laboratory (31) we were unable to detect high level expression of a protein having the molecular weight of LipB. Attempted optimization of translational initiation also failed to markedly increase transferase activity. Thus, we attribute our failure to detect a protein matching the LipB sequence to the very low level of lipB expression such that we failed to detect the protein band in our purified preparations (i.e. the visible bands are contaminating proteins). This very low cellular abundance is consistent with the codon bias of lipB and may be characteristic of vitamin attachment enzymes. We expect that lipB encodes an essential component of the lipoate transferase, although it remains possible that LipB is a positive regulator required for expression of the enzyme. We think this latter possibility is remote since other organisms including diverse bacteria and yeast contain DNA sequences that potentially encode proteins closely similar to LipB (data not shown). The sequences of regulatory proteins are seldom conserved among diverse organisms whereas enzymes often show strong sequence conservation. It should be noted that LipB may be only one subunit of a multisubunit transferase. If so, this would account for only modest increases in transferase activity upon introduction of multicopy lipB plasmids.
Mitochondria Contain Both Lipoate Transferase Activity and a Lipoyl Donor-Several laboratories have reported that mitochondria contain ACPs (18 -23), and nuclear genes encoding ACP-like proteins having mitochondrial targeting sequences have been detected in plants (20) and fungi (32). The role of these proteins has been a puzzle since the major fatty acid synthetic enzymes reside in other cellular compartments.
Since mitochondria contain lipoylated proteins as well as ACP we tested lysates of plant (from pea, P. sativum) and fungal (N. crassa) mitochondria for a lipoate transferase active with lipoyl-ACP. Activity was readily detected in these preparations using either the gel shift assay (data not shown) or the E. coli apoPDC activation assay (Fig. 4). In the latter case we inactivated the endogenous mitochondrial PDCs by addition of ATP. In the presence of ATP an endogenous PDC kinase converts the plant and fungal mitochondrial PDCs to their phosphorylated and inactive forms (33,34) (the kinase has no effect on E. coli PDC). The N. crassa transferase was also active with octanoyl-ACP. We also used apoPDC activation to test for the presence of lipoyl-ACP in the mitochondrial lysates and detected appreciable levels of endogenous lipoyl donors in both the plant and fungal preparations (Fig. 4). To demonstrate a stronger dependence of the N. crassa transferase activity on added lipoyl-ACP we serially diluted the lysate and proportionally extended the incubation time of each dilution (to compensate for dilution); the rationale being to dilute the endogenous lipoyl-ACP to concentrations at which the transferase could not efficiently bind the endogenous substrate. Under these conditions addition of E. coli lipoyl-ACP gave a 6.6-fold increase in apoPDC activation (data not shown). The endogenous lipoyl donor present in N. crassa lysates may be a better transferase All lipoate transferase assays were performed as described under "Experimental Procedures" except that assay components were omitted or substituted as given. The octanoyl-ACP reactions contained 80 g of octanoyl-ACP in place of lipoyl-ACP, and the low lipoyl-ACP reactions contained 5 g (rather than 80 g) of lipoyl-ACP. All PDC assays were performed in the presence of 2 mM ATP to inhibit the endogenous PDC activity of the mitochondrial lysates (33,34).
substrate than E. coli lipoyl-ACP, since addition of a low concentration of E. coli lipoyl-ACP (22 M) decreased the rate of apoPDC activation, presumably through occupation of the transferase active site by the less active E. coli lipoyl-ACP substrate. Moreover, E. coli octanoyl-ACP (but not nonacylated ACP) was a potent inhibitor of the endogenous donor-dependent transferase reaction (22 M gave complete inhibition in the presence of excess apoPDC) consistent with identity of the endogenous donors as lipoyl thioesters of the mitochondrial ACPs.

DISCUSSION
Although lipoic acid attachment is required for activity of several enzymes essential to central metabolism (1,2) and this cofactor is synthesized by most organisms (including plants and probably mammals), the mechanism of lipoic acid synthesis is not understood. Octanoic acid has been shown to be the precursor of the carbon chain (35), but neither the origin nor the mechanism whereby the sulfur atoms are inserted into the hydrocarbon chain are known. Our finding that lipoyl-ACP donates lipoate to PDC suggests that lipoic acid synthesis may proceed through ACP-bound intermediates, the first of which (octanoyl-ACP) is produced by the fatty acid synthetic pathway. The amino acid sequences of PDC and KGDC and of their lipoyl domains are highly conserved throughout biology, and the structures of the lipoyl domains are also highly conserved. This conservation extends to function in that mammalian and plant lipoyl domains are appropriately lipoylated upon expression in E. coli (4,5,36). Therefore, it seems reasonable to propose that the lipoate synthetic pathway is also conserved.
We suspect that the low rates of de novo fatty acid synthesis reported for isolated mitochondria and mitochondrial lysates (20 -23) represents (at least in part) lipoic acid synthesis. Although the lipoylated enzymes, PDC, KGDC, GCV, and the branched chain ␣-keto acid dehydrogenase complex can comprise major fractions of the mitochondrial proteins (e.g. more than one-third of the total plant mitochondrial soluble proteins; Ref. 3), these are very large protein assemblies (often larger than ribosomes) with each complex carrying only a few hundred lipoate molecules. Thus, what appears to be a low rate of fatty acid synthesis could readily suffice for synthesis of lipoic acid. Indeed, a major fatty acid chain synthesized by isolated mitochondria is octanoic acid, some of which is found in thioester linkage to the mitochondrial ACPs (21,23). Moreover, Wada et al. (21) have recently reported that a portion of the octanoic acid synthesized by purified pea and N. crassa mitochondria is covalently attached to the H subunit of GCV, the subunit that normally carries the lipoate cofactor. Since the typical lipoylated enzymes of eucaryotes are located within the mitochondria, we propose that lipoic acid synthesis and attachment proceed within this cellular compartment and that mitochondrial ACP functions as an essential cofactor in this pathway.