Mutations in the Dimer Interface of Dihydrolipoamide Dehydrogenase Promote Site-specific Oxidative Damages in Yeast and Human Cells

Background: Dihydrolipoamide dehydrogenase (DLD) is an enzyme involved in energy metabolism also known to possess two cryptic enzymatic activities of as yet undefined significance.

Results: DLD mutations that enhance these cryptic activities promote oxidative damage to specific mitochondrial targets.

Conclusions: The cryptic activities of DLD stimulate production of reactive oxygen species.

Significance: Oxidative damage may contribute to disease severity in patients with DLD mutations.

Abstract

Dihydrolipoamide dehydrogenase (DLD) is a multifunctional protein well characterized as the E3 component of the pyruvate dehydrogenase and α-ketoglutarate dehydrogenase complexes. Previously, conditions predicted to destabilize the DLD dimer revealed that DLD could also function as a diaphorase and serine protease. However, the relevance of these cryptic activities remained undefined. We analyzed human DLD mutations linked to strikingly different clinical phenotypes, including E340K, D444V, R447G, and R460G in the dimer interface domain that are responsible for severe multisystem disorders of infancy and G194C in the NAD⁺-binding domain that is typically associated with milder presentations. In vitro, all of these mutations decreased to various degrees dihydrolipoamide dehydrogenase activity, whereas dimer interface mutations also enhanced proteolytic and/or diaphorase activity. Human DLD proteins carrying each individual mutation complemented fully the respiratory-deficient phenotype of yeast cells lacking endogenous DLD even when residual dihydrolipoamide dehydrogenase activity was as low as 21% of controls. However, under elevated oxidative stress, expression of DLD proteins with dimer interface mutations greatly accelerated the loss of respiratory function, resulting from enhanced oxidative damage to the lipoic acid cofactor of pyruvate dehydrogenase and α-ketoglutarate dehydrogenase and other mitochondrial targets. This effect was not observed with the G194C mutation or a mutation that disrupts the proteolytic active site of DLD. As in yeast, lipoic acid cofactor was damaged in human D444V-homozygous fibroblasts after exposure to oxidative stress. We conclude that the cryptic activities of DLD promote oxidative damage to neighboring molecules and thus contribute to the clinical severity of DLD mutations.

Introduction

An increasing number of enzymes with well characterized metabolic functions are known to possess cryptic activities of largely undefined biological relevance (1–3). An interesting example is provided by dihydrolipoamide dehydrogenase (DLD),² an enzyme that functions in several different multienzyme complexes across species (4) and is well studied as the E3 component of the mitochondrial pyruvate dehydrogenase (PDH) and α-ketoglutarate dehydrogenase (α-KGDH) complexes, which are central to energy metabolism and are implicated in severe mitochondrial disorders (5, 6). The PDH and α-KGDH complexes possess unique E1 (α-keto acid decarboxylase) and E2 (dihydrolipoyl transacetylase) components, whereas they share DLD as their E3 component (Fig. 1A). DLD functions in these complexes as a homodimer of ∼50-kDa subunits (Fig. 1C) and catalyzes the oxidative regeneration of a lipoic acid cofactor covalently bound to E2 with production of NADH (Fig. 1A).

FIGURE 1.

Dihydrolipoamide dehydrogenase and diaphorase reaction schemes and structure of human DLD. A, catalytic cycle of PDH and α-KGDH. In each complex, E1 catalyzes the decarboxylation of its α-keto acid substrate (pyruvate or α-ketoglutarate; denoted R-CO-COO⁻) and acylation of a lipoic acid cofactor covalently bound to E2. Then, E2 transfers the acyl derivative to coenzyme A (CoA-SH) to generate acetyl-CoA or succinyl-CoA and reduced lipoic acid (5). The E3-catalyzed step involves a disulfide exchange site (Cys⁴⁵ and Cys⁵⁰ in C) on DLD close to a tightly bound FAD cofactor (59). DLD transfers reducing equivalents from lipoic acid to its disulfide exchange site, then to FAD, and finally to NAD⁺, which yields NADH and regenerates oxidized lipoic acid. TPP, thiamine pyrophosphate cofactor. The figure was adapted from Ref. 42. B, diaphorase reaction scheme. Reducing equivalents are transferred from NADH to the FAD cofactor and then directly to the electron acceptor (molecular oxygen shown). Diaphorase activity is highest when the catalytic disulfide is reduced (60) as shown. C, crystal structure of human DLD homodimer (Protein Data Bank code 1ZMC). The two identical subunits are shown as gray and purple ribbons, respectively. FAD and NAD⁺ cofactors are shown as yellow and light blue spheres (dimer) or sticks (inset), respectively. Residues in the dihydrolipoamide dehydrogenase active site (Cys⁴⁵, Cys⁵⁰, His⁴⁵², and Glu⁴⁵⁷) and the proteolytic active site (Ser⁴⁵⁶ and Glu⁴³¹) are shown as blue and yellow sticks, respectively. The figure was generated with PyMOL.

DLD is also known to possess diaphorase activity, which is the ability to catalyze the oxidation of NADH to NAD⁺ using different electron acceptors (Fig. 1B) (7–10). In this capacity, DLD is believed to primarily have a pro-oxidant role, achieved by reducing O₂ to superoxide radical and hydrogen peroxide (11, 12) or by reducing Fe³⁺ to Fe²⁺, which can in turn result in the production of highly toxic hydroxyl radical (7). Disease-causing mutations were recently shown to stimulate the ability of human DLD to produce superoxide radical and hydrogen peroxide in vitro (12). Independent studies showed that DLD is an important source of reactive oxygen species (ROS) also in living cells, particularly under conditions that increase the NADH/NAD⁺ ratio (13, 14). Indeed, mitochondria isolated from DLD+/− mice (lacking one copy of the DLD gene) were found to produce significantly less hydrogen peroxide than mitochondria isolated from DLD+/+ controls (13). Similarly, yeast DLD was shown to be responsible for increased oxidative stress and reduced life span in yeast cells with low NAD⁺ availability (15).

Our group reported previously that DLD can also function as a serine protease (16). We identified a highly conserved catalytic dyad (S456-E431) buried in the DLD dimer interface (Fig. 1C). Substituting alanine for Ser⁴⁵⁶ or Glu⁴³¹ abolished the proteolytic activity, whereas replacing other residues within the dimer interface, including a human pathogenic mutation (D444V), enhanced the activity. Although DLD could cleave at least one mitochondrial protein, the iron chaperone frataxin, to a shorter product in vitro, the natural substrates and in vivo role of DLD proteolytic activity remained undefined.

To date, approximately 15 DLD mutations have been identified in human patients with a great deal of clinical heterogeneity ranging from fatal multisystem disorders to milder tissue-specific conditions (Refs. 17–24; mutations analyzed in this study, Refs. 17–22). Interestingly, these clinical phenotypes do not completely correlate with the loss of dihydrolipoamide dehydrogenase activity (23), suggesting that additional mechanisms may contribute to disease pathophysiology. Particularly severe phenotypes are associated with a cluster of mutations in the DLD homodimer interface (Fig. 2, A and B). These mutations have been implicated in destabilization of the DLD dimer (25), decreased binding to the pyruvate or α-ketoglutarate dehydrogenase complexes (18, 26), reduced protein stability (19, 22, 23), or production of superoxide and hydrogen peroxide in vitro (12). Here, we show how different pathogenic mutations affect the three activities of DLD in vitro and how they contribute to mitochondrial oxidative damage in yeast and human cells. Our work emphasizes the potentially complex consequences of mutations in a multifunctional enzyme.

FIGURE 2.

Analysis of three enzymatic activities of wild type and mutant human DLD proteins in vitro. A, schematic representation of human DLD. Point mutations analyzed in this study are shown in red along with another recently reported mutation (I445M) located in the interface domain (24). MTS, mitochondrial targeting sequence; FAD and NAD, FAD- and NAD⁺-binding domains; Central, central domain; Interface, dimer interface domain. The figure was adapted from Ref. 23. B, residues and intersubunit interactions altered by the interface domain mutations analyzed in this study. Glu³⁴⁰, Asp⁴⁴⁴, Arg⁴⁴⁷, and Arg⁴⁶⁰ are shown as yellow sticks; residues interacting with Asp⁴⁴⁴ and Arg⁴⁶⁰ (Tyr⁴³⁸ and Asp³³³, respectively) are shown as purple sticks. The figure was generated with PyMOL. C, purity of recombinant DLD proteins. 5 μg of recombinant DLD was separated by 12% SDS-PAGE and stained with SYPRO Orange. We demonstrated previously that additional weak bands present in purified DLD preparations are fragments of DLD or oligomers that are not reduced during SDS-PAGE (16). D, dihydrolipoamide dehydrogenase kinetics. 1 nm DLD, 0.2 mm NADH, 0.67 mm NAD⁺, 10 mm EDTA, 0.07% BSA (w/v), and varying concentrations of lipoamide in 50 mm sodium phosphate buffer, pH 6.5 were combined in a 1.0-cm quartz cuvette, and the reaction was monitored spectrophotometrically at 340 nm (disappearance of NADH) for 5 min at 25 °C. Initial reaction velocities (V₀) were calculated from the extinction coefficient of NADH at 340 nm (6.22 mm⁻¹ cm⁻¹). E, diaphorase kinetics. 50 nm DLD, 0.15 mm NADH, and the indicated concentrations of DCPIP in 50 mm sodium phosphate buffer, pH 8.0, 100 mm NaCl were combined in a 0.2-cm quartz cuvette, and the reaction was monitored at 600 nm (reduction of DCPIP) for 5 min. Initial velocities (V₀) were calculated from the extinction coefficient of DCPIP at 600 nm (21 mm⁻¹ cm⁻¹) (8). F, proteolytic activity of DLD proteins was measured using a 13-amino acid fluorogenic peptide derived from the FXN^56–210 N-terminal region (16). Each reaction contained 8 μm DLD and 50 μm peptide in 10 mm Tris-HCl, pH 8.0, 50 mm NaCl. The increase in fluorescence (ΔFluorescence) was monitored with excitation and emission wavelengths of 355 and 495 nm, respectively, in 96-well plates. D–F, values shown are the mean ± S.D. of two experiments as described in detail under “Experimental Procedures.”

EXPERIMENTAL PROCEDURES

Purification and Kinetic Analysis of Recombinant DLD Proteins

cDNAs coding for mature wild type (WT) and mutant DLD proteins (residues 36–509) were generated by site-directed mutagenesis (primer sequences are listed in supplemental Table S1A), and purified proteins were prepared as reported previously (16) except that an additional step was included in the purification procedure as follows. Protein eluted from a nickel affinity column was buffer-exchanged into 10 mm sodium phosphate, pH 7.5; loaded onto a hydroxyapatite column (CHT ceramic type II support; Bio-Rad); and eluted with a linear gradient from 10 to 500 mm sodium phosphate buffer, pH 7.5. All DLD proteins were purified, stored, and assayed using identical procedures. Each purification yielded ∼8 mg of purified protein, which was aliquoted and stored at −80 °C in 10 mm HEPES-KOH, pH 6.8 in the presence of 25% glycerol. For each of the three enzymatic activities under investigation, one protein aliquot was thawed, diluted to the appropriate protein concentration, and assayed in triplicate. This procedure was repeated with a second aliquot. The means of the two triplicate data sets were calculated and analyzed as n = 2 experiments using one-way ANOVA and Dunnett's post-test (GraphPad Prism, GraphPad Software, Inc., San Diego, CA). Dihydrolipoamide dehydrogenase activity was measured spectrophotometrically as described previously (16) with a lipoamide concentration between 0 and 5 mm. Diaphorase activity was measured spectrophotometrically from the reduction of varying concentrations of dichlorophenolindophenol (DCPIP) in the presence of a saturating concentration of NADH (8). Background rates of DCPIP reduction by NADH in the absence of DLD were subtracted. Proteolytic activity was assayed as described (16) using 8 μm DLD.

Generation of Yeast Strains

Yeast strains expressing human DLD variants were generated from the haploid strain YPH500α. First, an lpd1Δ strain was generated by integration of the KanMX4 cassette at the LPD1 locus. Then, cDNAs coding for WT and mutant DLD precursor proteins were synthesized by PCR. To achieve homologous recombination at the LPD1 locus, each cDNA was flanked by ∼60-bp sequences identical to the chromosomal regions immediately upstream and downstream of the LPD1 coding sequence. PCR products were transformed into the lpd1Δ strain, and lpd1Δ::DLD transformants were selected based on their ability to grow on rich medium containing glycerol as the sole carbon source. The accuracy of the chromosomal lpd1Δ deletion and subsequent lpd1Δ::DLD integrations was verified by PCR amplification of genomic DNA and DNA sequencing. Strains expressing a C-terminally FLAG-tagged version of the PDH E2 protein were generated by chromosomal integration of the FLAG coding sequence in the LAT1 locus using TRP1 as a selectable marker and verification by PCR analysis of genomic DNA. Strains expressing a C-terminally FLAG-tagged version of the α-KGDH E2 protein were similarly generated using URA3 as a selectable marker. However, the α-KGDH E2-FLAG protein was undetectable by Western blotting in these strains, which were respiratory-deficient and were not analyzed further. To generate lpd1Δ::DLD yfh1Δ[FXN] strains, a low copy centromeric plasmid expressing human frataxin precursor under the control of the yeast glyceraldehyde-3-phosphate dehydrogenase promoter (27) was first transformed into the lpd1Δ::DLD yeast strains. Then, the yeast frataxin gene (YFH1) was deleted by integration of the HIS3 selectable marker at the YFH1 locus. The yfh1Δ knock-out was confirmed by PCR analysis of genomic DNA. Yeast strains used in this study are listed in Table 3. Primers used for yeast strain construction are listed in supplemental Table S1B.

TABLE 3.

Yeast strains used in this study

Strain	Genotype	Source/Ref.
LPD1 (YPH500α)	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1	63
lpd1Δ	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::KAN	This study
lpd1Δ::WT	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::WT (wild type human DLD)	This study
lpd1Δ::E340K	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::E340K (human DLD variant)	This study
lpd1Δ::D444V	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::D444V (human DLD variant)	This study
lpd1Δ::R447G	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::R447G (human DLD variant)	This study
lpd1Δ::R460G	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::R460G (human DLD variant)	This study
lpd1Δ::S456A	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::S456A (human DLD variant)	This study
lpd1Δ::G194C	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::G194C (human DLD variant)	This study
lpd1Δ::DLD PDH E2-FLAG	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::DLD variant LAT1-FLAG::TRP1	This study
lpd1Δ::DLD yfh1Δ[FXN]	MATα ura3–52 lys2–801amber ade2–101ochre trp-Δ63 his3-Δ200 leu2-Δ1 lpd1Δ::DLD yfh1Δ::HIS3 + YCplac22-GPD-FXN	This study

Isolation of Mitochondria, Western Blotting, and Other Analyses in Yeast

The following growth media were used: YP-Dextrose (1% yeast extract, 2% peptone, 2% dextrose), YP-Glycerol (1% yeast extract, 2% peptone, 3% glycerol), SD (6.7% Bacto yeast nitrogen base without amino acids, 2% dextrose supplemented with required amino acids), and YP-Galactose (1% yeast extract, 2% peptone, 2% galactose, 0.5% dextrose). In all experiments, yeast strains were inoculated from freshly streaked frozen stocks, and cultures were grown at 30 °C with shaking at 225 rpm. For mitochondrial isolation, cultures were grown to early stationary phase and then diluted into fresh medium, and growth continued for the indicated time. Cells were disrupted with glass beads, and the mitochondrially enriched fraction was prepared by differential centrifugation as described (28). Isolation of four to eight strains was performed simultaneously with lpd1Δ::WT included in each set to control for variations between mitochondrial isolations. Mitochondria were aliquoted and frozen at −80 °C. Protein concentrations were measured using a BCA protein assay kit (Pierce). For Western blots, 25 μg of mitochondrial protein was separated on a 10% Tris-glycine SDS-PAGE gel unless otherwise indicated and transferred to a PVDF membrane. Polyclonal antibodies against Nfs1p, Cox4p, Hem15p, and Rip1p were generated in rabbits (29, 30). Antibodies against lipoic acid, α-KGDH E2 and α-KGDH E1 were a generous gift from Dr. Luke Szweda (Oklahoma Medical Research Foundation). Monoclonal antibodies against yeast Cox2p and human PDH E2 were from MitoSciences, and polyclonal antibody against porcine heart DLD was from United States Biological. Protein carbonylation was detected with the OxyBlot kit (Millipore) using 15 μg of mitochondrial protein per sample. Quantification of Western blots was performed using ImageQuant 5.0 software (GE Healthcare). Methodologies to measure cell viability, protein carbonylation, and respiratory function during chronological yeast aging were as reported previously (31, 32). Dihydrolipoamide dehydrogenase activity and citrate synthase activity were assayed in triplicate in three independent mitochondrial preparations. The means of the three independent experiments were analyzed using one-way ANOVA and Dunnett's post-test. Organic acid analysis was performed by gas chromatography/mass spectrometry (GC/MS), essentially as described (33). Briefly, cells were grown in YP-Glycerol medium to A₆₀₀ ∼6, and total cell extracts were prepared by lysis with glass beads. Yeast cell extracts were subjected to oximation with hydroxylamine HCl, acidified with HCl (0.5 m final concentration), and then extracted with ethyl acetate. After evaporation, the dry residue was silylated with N,O-bis(trismethylsilyl)trifluoroacetamide containing 1% trimethylchlorosilane and analyzed using an Agilent 6890 capillary GC/MS instrument with an MSD 5973 detector. Five independent cultures were analyzed for each strain.

Analysis of DLD Proteins in Patient Fibroblasts

Dermal fibroblasts from patients homozygous for the D444V (17) and R447G (18) mutations and control fibroblasts (Coriell GM08398) were routinely cultured in minimal essential medium (Invitrogen) containing 10% fetal bovine serum, 1% non-essential amino acids, and 1% antibiotic-antimycotic at 37 °C in 5% CO₂. In some experiments, cells were cultured in glucose-free RPMI 1640 medium (Invitrogen) containing 10% dialyzed fetal bovine serum, 25 mm HEPES-KOH, 25 mm galactose, and 1% antibiotic-antimycotic. Total cell extracts were prepared by treating cells with 25 mm HEPES-KOH, pH 7.5, 100 mm NaCl, 1.5% (w/v) dodecyl maltoside on ice for 15 min. For fibroblast transfections, cDNAs encoding the full-length WT, D444V, and R447G DLD coding sequences were PCR-amplified and cloned into the BamHI and XhoI sites of vector pCDNA4. The resulting pCDNA4-DLD vectors were confirmed by DNA sequencing. R447G patient fibroblasts (4 × 10⁵ cells) were transfected with 5 μg of pCDNA4-DLD or pmax-GFP control using the Human Dermal Fibroblast Nucleofector kit (Lonza) and program U-023. Transfection with pmax-GFP resulted in >95% transfection efficiency.

RESULTS

Mutations in Dimer Interface Domain of Human DLD Have Mild Effects on Dihydrolipoamide Dehydrogenase Activity in Vitro

We expressed in Escherichia coli the mature form of human DLD, including WT, four interface domain mutants (E340K, D444V, R447G, and R460G), and an NAD⁺-binding domain mutant (G194C) (Fig. 2, A and B; numbering starts from the N terminus of the mature protein after removal of the 35-residue mitochondrion-targeting signal). A two-step purification procedure generated proteins of >95% purity (Fig. 2C) (16). Similar yields were obtained for WT and mutant proteins except for a ∼20% lower yield with the R447G protein (Fig. 2C). The four interface domain mutations decreased dihydrolipoamide dehydrogenase activity to variable degrees relative to WT (Fig. 2D). Specifically, the E340K and D444V mutations led to only modest reductions in activity, whereas the R447G and R460G mutations lowered the activity to ∼50% of WT (Fig. 2D). This decrease in activity resulted in only modest effects on enzyme kinetics (Table 1). In contrast, activities measured in patient cell lines had been much lower relative to controls (23). Because others (26) and we routinely measured activity at 25 °C, we tested the possibility that mutant proteins might exhibit lower activity at a higher temperature. At 37 °C with a non-saturating concentration of lipoamide (1 mm), the E340K, D444V, and R447G proteins retained >90% of the activity measured at 25 °C, whereas the R460G protein retained only 41% (not shown), suggesting that the effects of dimer interface mutations could be intensified in vivo by altered protein stability, particularly for the R460G mutation.

TABLE 1.

Dihydrolipoamide dehydrogenase kinetics of recombinant DLD variants

Kinetic parameters were calculated by non-linear regression analysis of kinetic curves shown in Fig 2D using GraphPad Prism software (GraphPad Software, Inc.), and the classical Michaelis-Menten equation, V₀ = V_max[S]₀/K_m+[S]₀ where [S]₀ is the initial concentration of substrate, V_max is the maximal reaction velocity, and K_m is the observed Michaelis constant. k_cat is V_max divided by the concentration of DLD. Each of the two experiments used to construct the kinetic curves in Fig 2D was fitted independently. The mean ± S.D. for each kinetic constant is shown.

	Vmax	Km	kcat	kcat/Km
	mm min−1mg−1	μm	s−1	m−1s−1
WT	447 ± 23	0.84 ± 0.04	342 ± 18	4.06 ± 0.02 × 105
E340K	398 ± 35	0.72 ± 0.06	305 ± 26	4.23 ± 0.01 × 105
D444V	372 ± 19	0.81 ± 0.09	285 ± 14	3.55 ± 0.23 × 105a
R447G	241 ± 1a	0.73 ± 0.01	185 ± 0.5a	2.54 ± 0.03 × 105a
R460G	179 ± 36a	0.55 ± 0.16a	137 ± 27a	2.54 ± 0.22 × 105a
G194C	209 ± 32a	0.24 ± 0.04a	160 ± 25a	6.84 ± 0.12 × 105a

^a p < 0.05 relative to WT as determined by one-way ANOVA.

Mutations in Dimer Interface Domain of Human DLD Enhance Diaphorase and Proteolytic Activity in Vitro

Earlier studies suggested that dimer interface mutations might uniformly increase diaphorase activity via global destabilization of the DLD dimer (34, 35). Although each of the four mutations analyzed here led to higher diaphorase activity, this effect was large for R460G, minimal for R447G, and intermediate for E340K and D444V (Fig. 2E). Each of the homodimer interface mutations increased the enzyme V_max but variably affected the affinity of the enzyme for substrate (K_m). K_m was unaffected by the D444V and R460G mutations, leading to a 2- and 3-fold increase, respectively, in k_cat/K_m, a measure of enzyme efficiency (Table 2). The E340K mutation increased K_m, resulting in only a modest increase in k_cat/K_m, whereas R447G increased both V_max and K_m inconsistently (leading to the high standard deviations in Table 2), resulting in an overall modest decrease in k_cat/K_m.

TABLE 2.

Diaphorase kinetics of recombinant DLD variants

Kinetic parameters were calculated as described in the legend to Table 1 from the kinetic curves shown in Fig. 2E.

	Vmax	Km	kcat	kcat/Km
	mm min−1mg−1	μm	s−1	m−1s−1
WT	4.5 ± 0.6	111 ± 14	1.93 ± 0.27	1.73 ± 0.03 × 104
E340K	11.4 ± 1.6a	173 ± 9	4.85 ± 0.68a	2.81 ± 0.54 × 104
D444V	8.2 ± 0.04a	106 ± 4	3.51 ± 0.02a	3.32 ± 0.13 × 104
R447G	8.7 ± 1.8a	286 ± 92a	3.69 ± 0.78a	1.32 ± 0.15 × 104
R460G	15.1 ± 0.7a	123 ± 1	6.41 ± 0.28a	5.20 ± 0.19 × 104a
G194C	1.7 ± 0.3	14 ± 8	0.71 ± 0.11	5.50 ± 2.22 × 104a

^a p < 0.05 relative to WT as determined by one-way ANOVA.

The G194C mutation decreased both the maximal dehydrogenase and diaphorase activity relative to WT DLD (Fig. 2, D and E), consistent with impaired NAD⁺/NADH cofactor binding to the enzyme. In both cases, V_max and K_m were decreased, resulting in higher k_cat/K_m values (Tables 1 and 2).

We reported previously that the D444V mutation increased the proteolytic activity of DLD (16). Here, we confirmed this observation and found an even greater (∼3-fold) increase in proteolytic activity with the R447G mutation but no significant changes with the other interface domain mutations or G194C (Fig. 2F). Thus, all of the DLD mutations analyzed decreased to various degrees dihydrolipoamide dehydrogenase activity; in addition, dimer interface mutations were more frequently associated with enhanced proteolytic and/or diaphorase activity.

Human DLD Proteins with Pathogenic Mutations Maintain Respiratory Function in S. cerevisiae Lacking Yeast DLD Homologue, Lpd1p

Danner and co-workers (36) showed previously that wild type human DLD could substitute for the yeast DLD homologue, Lpd1p, in Saccharomyces cerevisiae and further characterized the effects of two human DLD mutations (K37A and P453L) in this system. We used a similar system to model the effects of human DLD mutations in vivo. By use of homologous recombination, we generated haploid lpd1Δ::DLD strains (Table 3) in which the coding sequence of the yeast LPD1 gene was precisely replaced with the coding sequence of the human DLD cDNA (WT and mutant versions). Thus, each lpd1Δ::DLD strain carried one copy of the DLD cDNA stably integrated into the yeast nuclear genome under the control of the endogenous LPD1 promoter (Fig. 3A and Table 3). As determined by Western blotting with a polyclonal antibody against porcine heart DLD, expression of human DLD in the lpd1Δ::WT strain appeared comparable with expression of yeast Lpd1p in the parental LPD1 strain (Fig. 3B). This conclusion was supported by the presence of similar levels of dihydrolipoamide dehydrogenase activity in the LPD1 and lpd1Δ::WT strains (Fig. 3D). None of the mutations under investigation altered DLD expression levels relative to WT DLD (Fig. 3B).

FIGURE 3.

Analysis of wild type and mutant human DLD proteins in yeast. A, scheme illustrating the methodology used for creation of lpd1Δ::DLD yeast strains. Homologous recombination was used to sequentially delete the coding sequence of the LPD1 gene (coding for yeast DLD, Lpd1p) followed by integration of individual cDNAs coding for WT and mutant human DLD proteins. See “Experimental Procedures” for details. B, analysis of DLD expression. The indicated yeast strains were grown for 48 h in YP-Glycerol medium, the mitochondrially enriched fraction was prepared, and mitochondrial protein was analyzed by Western blotting using anti-DLD and anti-lipoic acid (LA) antibodies. The two bands detected by anti-LA antibody correspond to the lipoic acid cofactors bound to the PDH E2 and α-KGDH E2 proteins as shown. The same membrane was also probed with anti-Nfs1p antibody as a loading control. C, respiratory growth of lpd1Δ::DLD yeast strains. Yeast cells were grown to early stationary phase in YP-Dextrose, washed with water, and diluted to an A₆₀₀ of 0.1 in YP-Glycerol. Growth was monitored by A₆₀₀ (denoted OD₆₀₀) measurements for three independent isolates of each DLD mutant strain; mean ± S.D. is shown for each of these strains. One isolate each was analyzed for the LPD1 and lpd1Δ strains. D, effects of DLD mutations on dihydrolipoamide dehydrogenase activity and citrate synthase activity. Mitochondria were lysed with 0.1% Triton X-100 for 10 min on ice. Dihydrolipoamide dehydrogenase activity was measured using 25 μg of mitochondrial protein, 0.2 mm NADH, and 0.93 mm lipoamide in 50 mm sodium phosphate buffer, pH 6.5. The reaction was monitored spectrophotometrically at 340 nm for 10 min at 25 °C. Values shown are the mean ± S.D. of three independent experiments as described in detail under “Experimental Procedures.” *, p < 0.05 relative to WT activity (determined by one-way ANOVA). Citrate synthase activity was measured according to standard protocols (61).

As reported by others (36), the lpd1Δ strain was viable on fermentable carbon sources but could not utilize glycerol (Fig. 3C), a carbon source that yeast cells metabolize only aerobically through the tricarboxylic acid cycle and oxidative phosphorylation. In contrast, all of our lpd1Δ::DLD strains grew well on glycerol and actually reached a higher optical density than the LPD1 strain (Fig. 3C). We detected similar levels of dihydrolipoamide dehydrogenase activity in lpd1Δ::WT and LPD1 mitochondria but variably decreased activity in mitochondria isolated from the four interface domain mutants, ranging from 87% residual activity in lpd1Δ::R447G to only 21% residual activity in lpd1Δ::R460G (Fig. 3D). Moreover, only 45% residual activity was present in the NAD⁺-binding domain mutant, lpd1Δ::G194C. In contrast, the activity of another mitochondrial enzyme, citrate synthase, was normal or increased in the mutant strains compared with the lpd1Δ::WT and LPD1 strains (Fig. 3D).

We were unable to measure PDH or α-KGDH activities in the lpd1Δ::DLD strains; we attribute this to dissociation of human DLD from the endogenous complexes during solubilization of mitochondria. As an alternative measure of the function of PDH and α-KGDH, we investigated α-keto acid levels (33) in the lpd1Δ::D444V and lpd1Δ::R460G strains, which had shown significant reductions in dihydrolipoamide dehydrogenase activity (Fig. 3D) despite only minimally reduced respiratory growth rates (Fig. 3C). Organic acid analysis of yeast cell extracts revealed that both lpd1Δ::D444V and lpd1Δ::R460G accumulated only slightly higher levels of pyruvate and α-ketoglutarate compared with lpd1Δ::WT (Table 4). Pyruvate levels were actually higher in the LPD1 strain (Table 4), consistent with the lower growth rate of this strain on glycerol compared with the lpd1Δ::DLD strains (Fig. 3C). From these data, it appeared that the dihydrolipoamide dehydrogenase deficits observed in the interface domain and NAD⁺-binding domain mutants did not impair PDH or α-KGDH complex function at least not to an extent sufficient to compromise flux through the tricarboxylic acid cycle (Table 4) and slow down respiratory growth (Fig. 3C). Defects in DLD can also affect the activity of the branched-chain α-keto acid dehydrogenase complex, which in turn leads to accumulation of the branched-chain amino acids and respective branched-chain α-keto acids (4). We measured the latter metabolites by organic acid analysis and did not observe significant differences between the lpd1Δ::D444V and lpd1Δ::R460G strains and the lpd1Δ::WT control (data not shown).

TABLE 4.

Levels of organic acids detected by GC/MS in total cell lysates

Values are expressed in μg/mg of total protein. The mean ± S.D of five cultures is shown. α-KG, α-ketoglutarate.

	Pyruvate	Citrate	α-KG	Succinate
Parental	2.32 ± 0.60a	6.1 ± 3.6	0.95 ± 0.64	0.93 ± 0.26a
WT	1.14 ± 0.15	8.2 ± 2.4	0.76 ± 0.38	0.28 ± 0.10
D444V	1.35 ± 0.19	18.4 ± 2.7a	1.00 ± 0.16	0.27 ± 0.10
R460G	1.68 ± 0.54	10.3 ± 2.9	1.16 ± 0.53	0.23 ± 0.06

^a p < 0.05 relative to WT as determined by one-way ANOVA.

DLD Dimer Interface Mutations Accelerate Loss of Mitochondrial Respiratory Function during Yeast Chronological Aging

We looked for other phenotypic alterations that might be associated with pathogenic DLD mutations. Because DLD had been identified previously as an important source of ROS in a variety of biological contexts (13–15, 37, 38), we hypothesized that increased diaphorase or proteolytic activity might enhance ROS production in yeast. To test this, we analyzed all of our strains during chronological aging, a condition in which yeast cells are forced to survive in a non-dividing but metabolically active state that leads to progressive accumulation of oxidative damage (32, 39), which is believed to mimic aging of postmitotic mammalian cells (40). To achieve this state, we grew our strains on minimal defined medium containing glucose as the sole carbon source (SD medium) (32). Cells initially grew logarithmically by fermentation of glucose, and once they exhausted glucose, they grew more slowly (diauxic shift during which cells switch from fermentation to mitochondrial respiration) and eventually stopped dividing (stationary phase during which cells undergo chronological aging) (Fig. 4A). Growth curves were essentially identical among all strains (Fig. 4A). Mitochondria isolated from the mutant lpd1Δ::DLD strains grown for 24 h in SD medium under the conditions described above exhibited relative (as compared with lpd1Δ::WT) levels of dihydrolipoamide dehydrogenase and citrate synthase activities similar to those exhibited by the mutant strains upon 48 h of growth in rich medium supplemented with glycerol (not shown). The results in SD medium once again demonstrated that DLD mutations variably affected dihydrolipoamide dehydrogenase activity (not shown) but did not affect respiratory function given that under these conditions all strains were able to undergo diauxic shift and enter stationary phase (Fig. 4A), which requires the ability to shift from fermentation to mitochondrial respiration (32, 40). Viability, defined as the number of cells able to resume growth when transferred back to rich medium, decreased progressively during the diauxic shift and the stationary phase but without significant differences between the mutant and control strains (Fig. 4B).

FIGURE 4.

Effects of DLD mutations on mitochondrial function during yeast chronologic aging. A–E, the lpd1Δ::DLD strains were grown at 30 °C in SD medium to early stationary phase, diluted to an A₆₀₀ of 0.1 in fresh SD medium, and cultured for the indicated times. A, growth of lpd1Δ::DLD strains in SD medium. Two independent cultures were analyzed for each strain, and mean ± S.D. of A₆₀₀ (denoted OD₆₀₀) measurements is shown. B, viability of lpd1Δ::DLD strains in SD medium. Aliquots from the cultures in A were removed at 24, 72, and 120 h; diluted to analyze 200 cells per plate; and plated onto YP-Dextrose agar. Plates were incubated at 30 °C for 7 days, and then the total number of colonies was determined (corresponding to 800 total cells analyzed per strain at each time point) and expressed per A₆₀₀ based on the dilution factor used for plating. Data shown are the mean ± S.D. from two independent cultures plated in duplicate and analyzed as n = 2 experiments; one-way ANOVA showed no significant differences in viability (p > 0.05). C, detection of carbonylated proteins. Mitochondrion-enriched fractions were prepared after 24 and 120 h of growth in SD medium. Four strains were processed in parallel with lpd1Δ::WT included in each set as shown. Carbonylated proteins were detected via an OxyBlot kit (Millipore) according to the manufacturer's protocol using 15 μg of mitochondrial protein per sample. D, maintenance of respiratory function was assessed as described (28, 41). The ratio of respiring and respiratory-deficient colonies was determined from the plates used in B, corresponding to 800 total cells analyzed per strain at each time point. Data shown are mean ± S.D. from two independent cultures plated in duplicate and analyzed as n = 2 experiments as described above. **, p < 0.01 as determined by one-way ANOVA. Inset, to assess whether the small difference between WT and S456A strains was statistically significant, five cultures each of lpd1Δ::WT and lpd1Δ::S456A were grown for 120 h in SD medium, and each culture was plated in duplicate, corresponding to 2,000 total cells analyzed per strain. Values shown are the mean ± S.D. of five cultures analyzed as n = 5 experiments. ***, p < 0.001 as determined by Student's t test. E, expression of mitochondrially encoded Cox2p. Mitochondria from strains grown 24 or 120 h in SD medium were analyzed by Western blotting with antibodies specific for mitochondrially encoded Cox2p. The same membrane was reprobed sequentially with antibodies against nucleus-encoded DLD and α-KGDH E1 proteins.

During chronological aging, mitochondria are subjected to increased oxidative damage (31, 41). Accordingly, mitochondrial protein carbonylation increased dramatically from 24 to 120 h in the stationary phase but once again without appreciable differences among the various lpd1Δ::DLD strains (Fig. 4C). Next, we assessed the ability to maintain respiratory function by measuring the proportion of non-respiring cells that accumulated in stationary phase cultures after 24, 72, and 120 h (32, 41). All cultures contained <12% non-respiring cells after 24 h but accumulated much higher proportions after 72 and 120 h (Fig. 4D). The loss of respiratory competency was irreversible; i.e. the vast majority (>99%) of respiratory-incompetent cells remained unable to resume growth on a non-fermentable carbon source after being passaged on rich medium containing glucose in the absence of oxidative stress. In addition, the levels of Cox2p, a subunit of cytochrome c oxidase encoded in the mitochondrial genome, decreased in parallel with the accumulation of non-respiring cells, whereas the levels of DLD and the α-KGDH E1 protein (both encoded in the nuclear genome) remained unchanged or increased over time (Fig. 4E). Accumulation of respiratory-deficient cells and loss of Cox2p immunoreactivity were significantly accelerated in the four interface domain mutant strains as compared with the other strains (Fig. 4, D and E). In contrast, the lpd1Δ::G194C strain did not significantly differ from lpd1Δ::WT at all time points analyzed (Fig. 4, D and E). Given that lpd1Δ::G194C had lower dihydrolipoamide dehydrogenase activity than most interface domain mutants (Fig. 3D), progressive loss of mitochondrial respiratory function and Cox2p protein levels appeared to be independent of this activity.

We showed previously that substitution of serine 456 with alanine abolished human DLD proteolytic activity without significant effect on dihydrolipoamide dehydrogenase activity in vitro (16). We therefore generated an lpd1Δ::S456A strain to assess whether this strain might accumulate less damage than the lpd1Δ::D444V and lpd1Δ::R447G strains. Interestingly, this strain was essentially indistinguishable from the lpd1Δ::WT strain in terms of respiratory growth rate (Fig. 3C) and viability during chronological aging (Fig. 4B). However, it consistently accumulated a smaller proportion of non-respiring colonies relative to lpd1Δ::WT (Fig. 4D, inset).

DLD Dimer Interface Mutations Enhance Oxidative Damage to Lipoic Acid Cofactor during Yeast Chronological Aging

The lipoic acid cofactor covalently bound to the E2 components of the PDH and α-KGDH complexes had been extensively shown to be susceptible to oxidative damage (for reviews, see Refs. 42 and 43) both following oxidative insult (44) and ischemia-reperfusion (45). We therefore hypothesized that lipoic acid cofactor might be a specific target of ROS produced by DLD. To test this, we used a well characterized antibody that is specific to lipoic acid and does not recognize the oxidatively modified cofactor (45–48). During exponential growth in rich medium supplemented with glycerol, the DLD mutants showed levels of PDH E2 and α-KGDH E2 lipoic acid that were equal to or even higher than those present in the control strains (Fig. 3B), indicating that no appreciable damage to lipoic acid cofactor occurred under these conditions. This result also excluded the possibility that the mutations under investigation might affect the role of DLD in lipoic acid biogenesis (49). In contrast, during chronological aging, the four DLD interface domain mutants showed dramatic loss of immunoreactivity toward lipoic acid (Fig. 5A). Specifically, the immunoreactivity of α-KGDH E2 lipoic acid was almost completely lost in all four interface domain mutants after 24 h, whereas the immunoreactivity of PDH E2 lipoic acid was reduced after 24 h and almost completely lost after 120 h. As compared with lpd1Δ::WT, lipoic acid levels also were reduced in lpd1Δ::G194C although not as drastically as in the interface domain mutants, especially after 120 h (Fig. 5A). Lipoic acid cofactor was similar or even higher in lpd1Δ::S456A relative to lpd1Δ::WT (Fig. 5A).

FIGURE 5.

Effects of DLD mutations on integrity of lipoic acid cofactor during yeast chronologic aging. A, the same mitochondrial preparations analyzed in Fig. 4, C and E, were analyzed by Western blotting with the indicated antibodies. The α-KGDH E1 panel is the same shown in Fig. 4E. B, PDH E2 protein levels. Mitochondrion-enriched fractions were prepared from the indicated lpd1Δ::DLD PDH E2-FLAG strains after 120 h of growth in SD medium as described in the legend of Fig. 4. Lipoic acid (LA), PDH E2-FLAG, and α-KGDH E1 were detected by sequentially blotting the same membrane with specific antibodies.

To rule out the possibility that the loss of E2-bound lipoic acid immunoreactivity might be due to a reduction in E2 protein levels, PDH E2 and α-KGDH E2 were C-terminally tagged with FLAG epitope in the lpd1Δ::D444V, lpd1Δ::R447G, and lpd1Δ::G194C strains as well as lpd1Δ::WT (Table 3). Strains expressing α-KGDH E2-FLAG became respiratory-deficient, possibly due to a negative effect of the tag on α-KGDH E2 function, and could not be analyzed during chronological aging. Strains expressing PDH E2-FLAG were phenotypically normal and were analyzed as described above. After 120 h in the stationary phase, the lpd1Δ::D444V, lpd1Δ::R447G, and lpd1Δ::G194C strains once again showed a reduction in PDH E2 lipoic acid immunoreactivity but similar levels of PDH E2-FLAG protein as compared with lpd1Δ::WT (Fig. 5B).

D444V Mutation Leads to Loss of Lipoic Acid Cofactor in Human Fibroblasts during Elevated Oxidative Stress

We next investigated whether lipoic acid is also a target of oxidative damage in cultured fibroblasts derived from skin biopsies of patients homozygous for the D444V (17) and R447G DLD mutations (18). To increase their reliance on respiratory metabolism, we cultured mutant and control cells in glucose-free medium containing galactose as the carbon source (50). DLD protein levels were ∼70 and ∼10% of control levels in D444V and R447G cells (Fig. 6A), respectively, in contrast to the yeast expression system where mutant and WT DLD proteins were present at similar levels (Fig. 3B). This indicated that the effects of DLD mutations in different cells could be influenced by additional effects on protein biogenesis and/or stability. Despite their drastically reduced levels of DLD protein, both the D444V and R447G cell lines grew at a rate similar to that of the control (Fig. 6A and not shown). Levels of lipoic acid were largely unaffected by the presence of DLD mutations (data not shown). We reasoned that the reduction in the levels of D444V and R447G proteins might have limited damage to lipoic acid, and therefore, we further characterized the mutations in cells with similar DLD protein levels. We transiently transfected the R447G cell line, which expressed almost no endogenous DLD (Fig. 6A), with expression vectors carrying cDNAs coding for WT, D444V, and R447G DLD under the control of a strong promoter. Following transfection, WT and mutant DLD protein levels were ∼2- and 1.7-fold higher, respectively, as compared with the control cell line expressing endogenous WT DLD (Fig. 6, B versus A). When transfected cells were grown on glucose, they exhibited similar levels of lipoic acid (not shown). Next, we grew WT- and D444V-transfected cells for 48 h in medium containing galactose to increase their dependence on mitochondrial metabolism. The immunoreactivity of α-KGDH E2 lipoic acid was inconsistently reduced in the D444V cell line relative to WT (Fig. 6, C and D). However, loss of both PDH E2 and α-KGDH E2 lipoic acid was accentuated by the D444V mutation upon treatment of the cells with the pro-oxidants menadione and iron (Fig. 6, C and D); both compounds are substrates for DLD diaphorase activity in vitro (7, 51, 52). PDH E2 and α-KGDH E2 protein levels were essentially unchanged in untreated cells or after iron treatment (Fig. 6C), suggesting that the loss of lipoic acid immunoreactivity had not resulted from loss of E2 protein. Treatment with menadione caused extensive cell death and yielded insufficient material to verify PDH E2 and α-KGDH E2 protein levels.

FIGURE 6.

DLD expression and lipoic acid integrity in fibroblasts harboring D444V and R447G mutations. A, endogenous expression of WT and mutant DLD proteins. Fibroblasts from normal controls or patients homozygous for the D444V or R447G mutation were cultured in glucose-free RPMI 1640 medium containing 25 mm galactose, and total cell extracts were prepared as described under “Experimental Procedures.” Aliquots (50 μg of total protein) were analyzed by 10% SDS-PAGE and Western blotting using anti-DLD antibody. In B–D, R447G patient fibroblasts were transfected to overexpress WT and mutant DLD proteins as described under “Experimental Procedures.” B, expression of WT, D444V, and R447G DLD in transfected cells. Following transfection with pCDNA4-DLD or pmax-GFP (control), cells were grown in minimal essential medium for 24 h, and cell extracts (10 μg of total protein) were analyzed by Western blotting as described above. C, lipoic acid (LA) integrity under oxidative stress conditions. Transfected cells were grown in RPMI 1640 medium containing 25 mm galactose for a total of 48 h. Following a 4-h recovery from nucleofection, cells were either left untreated or treated with 10 μm menadione or 100 μm ferric ammonium citrate (Fe). Cell extracts were prepared, and aliquots (10 μg of total protein) were analyzed by Western blotting with the indicated antibodies as described above. Results from one of three independent experiments are shown. D, residual lipoic acid cofactor was quantified by band densitometry using ImageQuant software (GE Healthcare). For each of the three conditions analyzed, the lipoic acid band density measured in D444V fibroblasts is expressed as a percentage of that measured in WT fibroblasts subjected to the same treatment. Data shown are the mean ± S.D. of three independent experiments. *, p < 0.05; **, p < 0.01 as determined by Student's t test.

Cleavage of Two Frataxin Isoforms by Isolated Mitochondria Is Enhanced by D444V Mutation

In a previous study, we used one of the isoforms of human frataxin (FXN^56–210) (29, 53) as a substrate to identify proteolytic activities potentially involved in the regulation of mitochondrial iron homeostasis (16). We found that native DLD from mouse liver or pig heart and recombinant human DLD were able to catalyze cleavage of FXN^56–210 to a shorter ∼14-kDa product (FXN^78–210) in vitro. Here, we examined the cleavage of two recombinant frataxin isoforms, FXN^56–210 and FXN^42–210 (29), by mitochondrial extracts isolated from lpd1Δ::WT and lpd1Δ::D444V yeast strains (Fig. 7). Similar to the stepwise cleavage of FXN^56–210 by recombinant DLD in vitro (16), both FXN isoforms were cleaved to FXN^78–210 in an apparent stepwise process. The FXN substrate was first cleaved to two products slightly larger than FXN^78–210; longer incubations resulted in the formation of FXN^78–210. Cleavage of FXN^56–210 (Fig. 7, A and B) occurred more rapidly than that of FXN^42–210 (Fig. 7, C and D); in addition, ∼30% of the FXN^42–210 was not cleaved even after 24 h (Fig. 7D). Processing of both FXN^56–210 and FXN^42–210 occurred more rapidly in the lpd1Δ::D444V strain than in lpd1Δ::WT (Fig. 7, B and D), consistent with the higher proteolytic activity exhibited by the mutant D444V DLD protein in vitro (Fig. 2F).

FIGURE 7.

Proteolytic processing of FXN isoforms by mitochondrial lysate. The lpd1Δ::WT and lpd1Δ::D444V yeast strains were grown for 48 h in YP-Galactose medium, and an enriched mitochondrial fraction was prepared as described (62). Mitochondria were solubilized with 0.1% Triton X-100 on ice for 30 min, and debris was removed by centrifugation. Mitochondrial lysate (10 μg) was incubated alone (Lys) or in the presence of 3 μg of recombinant FXN^56–210 or FXN^42–210 (29) at 37 °C in 10 mm Tris, pH 8.0, 50 mm NaCl. At the indicated time points, the reaction was stopped by boiling the sample in SDS-PAGE sample buffer. Recombinant FXN isoforms were also incubated at 37 °C for 24 h to account for any self-cleavage of FXN (last two lanes in each panel). Proteins were separated by 14% Tris-glycine SDS-PAGE and stained with SYPRO Orange. Experiments were performed three times, and one representative gel is shown. A and C, time course of FXN^56–210 or FXN^42–210 processing. B and D, the residual FXN substrate remaining (relative to the 0-min (′ time point) was determined by band densitometry using ImageQuant software. The mean ± S.D. of three independent experiments is shown. In both B and D, after 1 h and at all subsequent time points, p was <0.05 (determined by Student's t test) for differences between WT and D444V.

These results prompted us to generate lpd1Δ::DLD yfh1Δ[FXN] strains in which human DLD (WT and the D444V, R447G, and S456A variants) and FXN were co-expressed in the absence of the endogenous yeast homologues, Lpd1p and Yfh1p (Table 3 and “Experimental Procedures”). We showed previously that human FXN^1–210 (i.e. the nucleus-encoded precursor) is efficiently imported by yeast mitochondria and processed to the same isoforms normally present in human cells (29). Upon growth in rich medium containing galactose, a very similar pattern of FXN^1–210 processing was observed in mitochondria isolated from each of the lpd1Δ::DLD yfh1Δ[FXN] strains (supplemental Fig. S1A). Relative to strains expressing WT or S456A DLD, the levels of all FXN isoforms were drastically decreased in the two interface domain mutants, consistent with the results obtained with a synthetic peptide or recombinant protein substrates (Figs. 2F and 7). However, these mutants did not accumulate FXN^78–210 or other detectable degradation products (supplemental Fig. S1A). The mature forms of three different unrelated nucleus-encoded mitochondrial proteins were present at similar levels in all strains, indicating that the reduction in FXN isoforms associated with the D444V and R47G mutations had not resulted from impaired mitochondrial protein import and processing (supplemental Fig. S1A). In a subsequent experiment, we allowed the lpd1Δ::WT yfh1Δ[FXN] and lpd1Δ::D444V yfh1Δ[FXN] strains to reach early logarithmic phase, added cycloheximide to the cultures to stop de novo protein synthesis, and then followed processing and turnover of the FXN precursor over time. Instead of isolating mitochondria, we prepared total protein extracts by rapid trichloroacetic acid treatment of yeast cells. At the initial time point, both strains exhibited similar levels of FXN^1–210. At later time points, precursor was progressively converted to FXN^42–210 and FXN^81–210 with no apparent differences between the two strains. The levels of FXN isoforms were equally stable in the two strains over a period of 9 h (supplemental Fig. S1B). It therefore appeared that solution conditions associated with mitochondrial isolation or solubilization might accelerate FXN turnover by mutant DLD.

DISCUSSION

The Three Enzymatic Activities of Human DLD Are Variably and Independently Influenced by Mutations in Different Domains of the Protein

In addition to its essential role as the E3 component of the α-keto acid dehydrogenase complexes, DLD has been known for some time to possess two cryptic enzymatic activities that were revealed under conditions shown to destabilize the DLD dimer (16, 34, 35). To investigate the biological relevance of these activities, we exploited a cluster of human mutations localized to the interface domain of the DLD homodimer that eliminate symmetrical hydrogen bonds predicted to stabilize the enzyme (Fig. 2, A and B) (25). We also used a human mutation localized to the NAD⁺-binding domain and unlikely to affect DLD dimer stability. Each of the four dimer interface mutations independently affected the three activities of DLD (Fig. 2, D–F). There was no correlation between diaphorase activity, which was enhanced to varying degrees by all four interface domain mutations (R460G ≫ E340K > D444V > R447G), and proteolytic activity, which was enhanced by only two of the interface domain mutations (R447G ≫ D444V), notably the mutations that enhanced diaphorase activity the least. These results indicate that each interface domain mutation influences the three activities of DLD independently, most likely through local structural changes. This view is consistent with recent reports that mutations in the dimer interface of human DLD do not alter dimer stability (26, 54).

Adam-Vizi and co-workers (12) recently reported that several pathogenic mutations increased the production of hydrogen peroxide and superoxide by human DLD in vitro. In agreement with our diaphorase activity measurements, this effect was primarily associated with interface domain mutations, whereas most mutations outside the domain either decreased or did not alter ROS production relative to WT DLD (12). The G194C mutation was found to promote ROS production ∼2-fold in vitro (12), and similarly, we found that G194C enhanced to some degree oxidative damage to lipoic acid in yeast. These results may reflect the relatively high diaphorase catalytic efficiency associated with the G194C mutation (Table 2), which may favor ROS production by DLD under certain conditions, e.g. at low substrate concentrations.

Dihydrolipoamide Dehydrogenase Deficits Associated with Human DLD Mutations Do Not Impair Mitochondrial Function in Yeast and Human Cells

The structural and functional conservation between yeast and human DLD (36) enabled us to use yeast to model in vivo effects of DLD mutations. Notably, by integrating the wild type human DLD cDNA into the yeast genome under the control of the endogenous LPD1 promoter, it was possible to achieve similar levels of dihydrolipoamide dehydrogenase activity in the lpd1Δ::WT compared with the parental LPD1 strain (Fig. 3D). Despite the fact that residual dihydrolipoamide dehydrogenase activity varied broadly among all lpd1Δ::DLD strains (Fig. 3D), they all grew even better than the LPD1 strain under non-fermentative conditions (Fig. 3C) and were able to undergo diauxic shift and enter stationary phase (Fig. 4A), which requires the ability to shift from fermentation to mitochondrial respiration (32, 40). DLD protein levels were similar among strains (Figs. 3B and 4E), eliminating the possibility that protein overexpression might have compensated for the loss of dihydrolipoamide dehydrogenase activity.

These observations are consistent with a threshold above which the loss of dihydrolipoamide dehydrogenase activity has no significant detrimental effects on yeast respiratory function. In our experimental conditions, the threshold was well below 50%. For example, in YP-Glycerol the lpd1Δ::R460G strain exhibited only 21% residual dihydrolipoamide dehydrogenase activity but an unremarkable growth phenotype and only mildly elevated pyruvate and α-ketoglutarate levels relative to lpd1Δ::WT (Figs. 3, C and D, and 4A and Table 4). Interestingly, a similar situation was observed in patients homozygous for DLD interface domain mutations (D444V, R447G, and a recently reported I445M mutation) (17, 18, 24) in which residual dihydrolipoamide dehydrogenase activity ranged from 15 to 20% (17, 18) or was undetectable (24), whereas biochemical markers of PDH and α-KGDH complex deficiency were only modestly and inconsistently altered. In agreement with these findings, we observed that fibroblasts from patients homozygous for the D444V or R447G mutation grew as well as control fibroblasts in medium supplemented with galactose (data not shown), a condition in which cells with severe pyruvate dehydrogenase deficiency are typically non-viable (50).

Mutations in Dimer Interface of Human DLD Intensify Loss of Mitochondrial Respiratory Function during Yeast Chronological Aging

The lack of obvious effects of DLD mutations on cellular respiratory function together with the knowledge that DLD could be an important source of ROS in vitro (12) and in mitochondria (13–15, 37, 38) prompted us to investigate oxidative damage as an alternative biochemical marker of DLD deficiency during yeast chronological aging. This condition is an established tool to evaluate stress resistance and its effects on mitochondrial function and cellular longevity (32, 40). In the metabolically stressful conditions associated with yeast chronological aging, we observed generalized progressive carbonylation of mitochondrial proteins but without obvious differences between the pathogenic DLD mutants and control strains (Fig. 4C). However, compared with the other lpd1Δ::DLD strains, the four interface domain mutants consistently showed a reduced ability to maintain respiratory function over time (Fig. 4, D and E). We suggest that the largely irreversible loss of respiratory function, associated with loss of mitochondrially encoded Cox2p, at least in part reflected loss of mitochondrial DNA integrity. It is possible that mitochondrial DNA could be specifically damaged in the dimer interface mutants because of the presence of the PDH and α-KGDH complexes in yeast mitochondrial nucleoids (55, 56). Whereas this association may normally be involved in mitochondrial DNA maintenance (55, 56), in the context of the dimer interface mutant strains, it could have exposed mitochondrial DNA to higher levels of ROS generated locally via diaphorase activity. One limitation of this hypothesis is that the four dimer interface mutations increased the diaphorase activity of DLD to different degrees in vitro (Fig. 2E), whereas the extent of mitochondrial damage was higher for the lpd1Δ::D444V strain at the beginning of chronological aging and similar among all four dimer interface mutants at a later stage (Fig. 4D).

Pathogenic Human DLD Mutations Intensify Damage to Lipoic Acid Cofactor of PDH and α-KGDH in Yeast and Human Fibroblasts

The lipoic acid cofactor bound to the PDH and α-KGDH E2 proteins is a molecule susceptible to oxidative damage (Ref. 44; for reviews, see Refs. 42 and 43) in close proximity to DLD and provided an additional way to test whether dimer interface mutations might be associated with locally increased oxidative damage. The availability of an antibody specific to unmodified lipoic acid allowed us to uncover extensive damage to lipoic acid cofactor in the dimer interface mutants and to a lesser extent in the G194C mutant during chronological aging (Fig. 5A). This phenomenon was not associated with loss of E2 protein (Fig. 5B), consistent with localized damage to the cofactor followed by enzymatic removal of the damaged cofactor (57). Human fibroblasts expressing D444V DLD revealed a pattern of damage similar to that of the corresponding yeast strain whereby the integrity of PDH and α-KGDH lipoic acid was largely unaffected during glycolytic growth (data not shown) but was progressively lost during respiratory growth upon exposure to pro-oxidant compounds that have been shown to serve as substrates for DLD diaphorase activity in vitro (7, 51, 52) (Fig. 6, C and D). Together, these findings indicate that damage to the lipoic acid cofactor covalently bound to the E2 protein of PDH and α-KGDH probably represents an important mechanism associated with DLD dimer interface mutations. This mechanism could chronically affect the function of PDH and α-KGDH in tissues with high metabolic rates such as brain, heart, and skeletal muscle and precipitate episodes of acute decompensation under stressful conditions. Transient intramitochondrial accumulation of α-keto acids could further influence mitochondrial function during oxidative insults. These effects together could perhaps explain the variable but generally severe and often eventful phenotypes exhibited by patients homozygous for dimer interface mutations and the fact that biochemical markers of DLD deficiency were not consistently detected in these patients (17, 18).

DLD Dimer Interface Mutations May Accelerate Frataxin Turnover via Unknown Mechanism

Our attempts to clarify the role of DLD proteolytic activity in the turnover of human frataxin gave contrasting results. We observed higher proteolytic activity against recombinant FXN^42–210 and FXN^56–210 in lpd1Δ::D444V relative to lpd1Δ::WT mitochondria (Fig. 7). However, although there was a dramatic decrease in all endogenously expressed FXN isoforms in mitochondria isolated from lpd1Δ::D444V yfh1Δ[FXN] and lpd1Δ::R447G yfh1Δ[FXN] strains, FXN levels were unchanged by the presence of mutant DLD proteins when analyzed in total cell extracts (supplemental Fig. S1, A and B). Further attempts to define the kinetics of FXN turnover during mitochondrial protein import assays in vitro revealed only subtle differences in the pattern of cleavage products accumulated over time (data not shown). We therefore hypothesize that FXN turnover by DLD can be stimulated by an unknown mechanism under stressful conditions possibly involving pH- or redox-dependent modifications of FXN and/or DLD (58).

Conclusions

Interface domain mutations appear to act largely independently of the loss of dihydrolipoamide dehydrogenase activity with damage to lipoic acid cofactor representing a consistent finding associated with these mutations in yeast and human cells. Enhanced diaphorase activity leading to localized ROS production and oxidative modifications of E2-bound lipoic acid cofactor represents an additional pathogenic mechanism common to the four interface domain mutations analyzed here and likely other DLD mutations that enhance diaphorase activity. However, it is important to keep in mind that the pathophysiology of DLD deficiency is much more complex as compared with our yeast and cellular models, and multiple pathologic mechanisms likely contribute to disease severity (18, 22, 26). The role of DLD proteolytic activity remains unclear and will require a better understanding of conditions that accelerate FXN turnover and the identification of additional natural substrates cleaved by DLD.