Immunological and biosynthetic studies of the human pyruvate dehydrogenase complex

The human pyruvate dehydrogenase multi-enzyme complex (PDC) catalyses the oxidative decarboxylation of pyruvate, transferring the resultant acetyl group to coenzyme A. It belongs to the family of 2-oxoacid dehydrogenase complexes that includes the 2-oxoglutarate dehydrogenase (OGDC) and branched-chain 2-oxoacid dehydrogenase complexes (BCOADC). Each assembly consists of multiple copies of three distinct component enzymes termed E1, E2 and E3. Human PDC also contains an accessory subunit (E3BP) that mediates stable E3 integration into the E2 ‘core’ of the complex. Human E2-PDC consists of the following domains: two tandemly-repeated, amino-terminal lipoyl domains in each of which the lipoic acid cofactor is attached covalently in amide linkage to a specific lysine residue; an E1-binding domain and a carboxy-terminal catalytic core domain. E3BP, an E2-related polypeptide, consists of a single lipoyl domain, an E3-binding domain and a carboxy-terminal region with no known enzymatic function. Primary biliary cirrhosis (PBC) is a chronic autoimmune liver disease in which inflammatory infiltration of the intrahepatic bile ducts leads to damage of the biliary epithelial cells followed by fibrosis, cirrhosis and ultimately liver failure. The disease is characterised by the presence of antimitochondrial antibodies (AMA), the production of which occurs at a very early stage of the disease. The major mitochondrial autoantigens have been identified as key constituents of the 2-OADCs, primarily the E2 and E3BP subunits of human PDC. AMA to these polypeptides are present in the serum of more than 95% of patients with PBC. PD1 and PD2 are monoclonal antibodies (mAbs) secreted by individual patient-derived hybridomas (IgG/λ) that interact with a common lipoylation-dependent epitope found on both E2 and E3BP lipoyl domains although the precise antigenic determinant has not been fully defined to date. To address this issue, three amino acid residues adjacent to the lipoylated lysine residue of the inner lipoyl domain of E2-PDC, ILD-PDC were mutated systematically to the equivalent residues found in the non-reactive lipoyl domain of Arabidopsis thaliana plastid E2-PDC. In parallel, the non-reactive lipoyl domain of the E2 enzyme of the human 2-oxoglutarate dehydrogenase complex, LD-OGDC was mutated at several amino acids around the lipoylated lysine residue to the equivalent residues found in the reactive lipoyl domains of E2 and E3BP-PDC in attempts to restore the level of Ab recognition to that of the ILD-PDC. These studies have permitted us to determine the key residues involved in PD1 and PD2 recognition and also their influence on the extent of lipoylation. By using Western blot analysis and ELISA for ILD-PDC and LD-OGDC mutants, the epitope recognized by these mAbs was located to the C-terminal side of the key lipoyl-lysine residue of the domain. Interestingly, there was no specific amino acid, apart from the lipoyl-lysine, that could be considered as essential to Ab recognition but there was a cumulative effect of multiple mutations. Native gel analysis permitted us to study the lipoylation status of these mutants through the separation of lipoylated and non-lipoylated domains. These studies have confirmed that, apart from the lipoyl-lysine residue, there is no specific motif or individual residue that is essential for lipoylation. These data were all consistent with the hypothesis that a precise structural cue involving the presentation of the lipoyl-lysine residue at the tip of a type I -turn was a prerequisite for recognition by the lipoylating enzyme(s). The lipoylation status of the various domains was confirmed by subjecting all mutants to modification using mPEG maleimide, a thiol group reagent (Mr 5000). In contrast to a previous report, it was observed that the minor holodomain form (20%) of wild type ILD-PDC produced in E. coli in the absence of exogenous lipoic acid represented octanoylated rather than lipoylated domain as it was not amenable to mPEG maleimide (mPEG) modification. Q-TOF mass spectrophotometry was employed for confirmation of the identity of the octanoylated ILD-PDC produced in vivo. Moreover, it was realized that PD1 and PD2 do not interact exclusively with lipoylated ILD-PDC as the octanoylated domain, lacking the dithiolane ring, also gave a signal of equivalent intensity on Western blotting. However, modification of lipoamide thiols of ILD-PDC with bulky substituents, mPEG, 4-hydroxy-2-nonenal, N-ethylmaleimide and iodoacetamide yielded modified forms of the ILD that were undetectable when probed with these mAbs. It is suggested that this loss of recognition is not attributable to the importance of the dithiols in Ab recognition per se but rather to the fact that these substituents may block or mask the epitope from Ab recognition as a result of steric hindrance. In an attempt to investigate the hypothesis that PBC could be induced by aberrant modification of the lipoyl-lysine residue via natural metabolites or xenobiotic agents serving as substrates for E. coli LplA ligase, the non-lipoylated domain of human ILD-PDC (produced in the absence of the exogenous lipoic acid) was modified in vitro with various lengths of saturated fatty acids (C2-C14) and related compounds. These included a branched-chain fatty acid (C8), valproic acid, a common anti-epileptic drug causing steatosis and an unsaturated aliphatic compound, trans-2-nonenoic acid, closely related in structure to a major lipid peroxidation product, 4-hydroxynonenal. It was shown that E. coli lipoyl LplA ligase has a broad substrate specificity that was not exclusive to 8-carbon substrates; it can incorporate a range of saturated fatty acids of varying chain length (C6 to C12) and also branched or unsaturated compounds with variable degrees of efficiency. Western blot analysis was employed to study the cross reactivity between these modified ILD-PDC domains and mAb PD1 or PD2. Incorporation of a fatty acid with a minimal chain length of 8-carbon atoms (octanoate) was necessary to elicit restoration of mAb cross-reactivity. Similar responses were induced by modification with decanoate and dodecanoate as well as valproate and trans-2-nonenoic acid. Interestingly, trans-2-nonenoic acid is a close relative of the major lipid peroxidation product, 4-hydroxynonenal that can be readily converted to 4-hydroxy trans-2-nonenoic acid in vivo. This compound is potentially a suitable substrate for the bacterial lipoylation system. Thus, it is proposed that oxidative stress could lead to aberrant modification of nascent non-lipoylated E2 components by this route resulting in neoantigen formation. Chronic or severe exposure to oxidative stress may also promote energy depletion by causing PDC malfunction and metabolic damage. These events may render modified PDC neoantigens accessible to the immune system, there breaking tolerance and initiating an autoimmune response against native lipoylated E2 and/or E3BP. However, the question as to why AMA are targeted mostly to E2 and E3BP-PDC is still an enigma. The second part of the thesis involved biosynthetic studies of the E2, E3BP and E3 enzymes of the PDC and the role(s) of their N-terminal presequences (matrix targeting signals) in the regulation of protein expression/folding. A variety of precursor constructs of human PDC were engineered to carry out this work. Thus, pre-E2, its N-terminal truncated form and pre-E3BP containing elongated presequences (53-86 amino acids) as well as pre-E3 housing a standard length presequence (35 amino acids) were cloned into pET-14b. In addition, the work was extended to study the behaviour of the hybrid precursors, pE2-E3 and pE3-E2 cloned into the same plasmid. The general findings of this study were as follows: firstly, comparing the levels of expression of each precursor to its mature form through small-scale protein inductions in E. coli at different temperatures, it was observed that both types of presequence as well as the nature of mature protein affect the level of protein expression. Therefore, it was noticed that both types of presequence had no significant effect on the level of protein expression when they were linked to the mature E3 whereas the negative effect of these presequences was apparent when they were linked to mature E2 or E3BP. Secondly, comparing the solubility of precursors with their mature forms, both extended and standard presequences markedly reduced the solubility of precursor constructs by inducing the production of inclusion bodies although the effect appeared to be more marked with the former. Thirdly, on decreasing the rate of the protein synthesis by growing E. coli cultures at lower temperatures, it was possible to minimise the formation of insoluble protein aggregates and achieve partial or indeed complete solubility of precursor forms in some cases. Fourthly, these precursors appeared to retain the ability to fold correctly or at least to initiate the correct folding pathway. Thus both soluble and insoluble fractions of E2 and E3BP precursors contained lipoylated domains as judged by their ability to cross-react with PD2, an indication that these N-terminally located domains had adopted their native conformations. These observations were consistent with the view that N-terminal mitochondrial targeting sequences markedly reduced the rate of protein folding rather than suppressing the folding process completely. In this scenario, precursors would exist as nascent folding intermediates for longer periods compared to their mature equivalents and so would be more prone to aggregation and degradation as observed in this study. Further experiments are planned to test this hypothesis