Complementation of mitochondrial electron transport chain by manipulation of the NAD+/NADH ratio

See allHide authors and affiliations

Science  08 Apr 2016:
Vol. 352, Issue 6282, pp. 231-235
DOI: 10.1126/science.aad4017

Taking control of cellular NAD+ concentrations

Cellular concentrations of the nicotinamide adenine dinucleotide (NAD+) are critical for proper metabolism and are often altered in aging and disease. To enable better understanding of these processes, Titov et al. altered the concentration of NAD+ in particular cellular compartments. They did this through expression of a bacterial enzyme targeted to specific compartments of human cells in culture. Their experiments emphasize the important role of the electron transport chain in redox transfer of electrons to NADH, rather than proton pumping, in mitochondrial pathogenesis.

Science, this issue p. 231


A decline in electron transport chain (ETC) activity is associated with many human diseases. Although diminished mitochondrial adenosine triphosphate production is recognized as a source of pathology, the contribution of the associated reduction in the ratio of the amount of oxidized nicotinamide adenine dinucleotide (NAD+) to that of its reduced form (NADH) is less clear. We used a water-forming NADH oxidase from Lactobacillus brevis (LbNOX) as a genetic tool for inducing a compartment-specific increase of the NAD+/NADH ratio in human cells. We used LbNOX to demonstrate the dependence of key metabolic fluxes, gluconeogenesis, and signaling on the cytosolic or mitochondrial NAD+/NADH ratios. Expression of LbNOX in the cytosol or mitochondria ameliorated proliferative and metabolic defects caused by an impaired ETC. The results underscore the role of reductive stress in mitochondrial pathogenesis and demonstrate the utility of targeted LbNOX for direct, compartment-specific manipulation of redox state.

A decline in electron transport chain (ETC) activity has been linked to numerous human disorders, ranging from rare genetic syndromes to common diseases, such as neurodegeneration, cancer, and diabetes, as well as the aging process itself (1, 2). How a decline in ETC activity gives rise to the spectrum of observed pathology cannot be readily explained by a simple deficiency in adenosine triphosphate (ATP) production (1). A key challenge in deciphering mitochondrial pathogenesis stems from the fact that the ETC performs at least two coupled functions: redox transfer of electrons from NADH [the reduced form of nicotinamide adenine dinucleotide (NAD+)] to oxygen and a simultaneous conversion of the free energy of the electromotive force into a proton gradient across the mitochondrial inner membrane. In principle, pathology could stem from an excess of reducing equivalents (termed reductive stress or pseudohypoxia, which includes stalling of NAD+-coupled reactions) or a reduced proton gradient (impairing pH and voltage-coupled processes, including ATP synthesis by the F1Fo-ATP synthase). Currently, there are no methods for dissecting the redox function of the ETC from its proton pumping function.

Here, we report the application of a genetically encoded tool for compartment-specific manipulation of the NAD+/NADH ratio. Our tool is based on the flavin adenine dinucleotide (FAD)–dependent H2O-forming NADH oxidases, which catalyze the four-electron reduction of O2 to two molecules of H2O (Fig. 1A). We focused on bacterial oxidases with specificity for NADH over reduced nicotinamide adenine dinucleotide phosphate (NADPH) (37), whose natural function is protection of redox balance and defense against oxygen toxicity (8). Such oxidases have been successfully expressed in bacteria and yeast for biotechnological applications (911). We screened several H2O-forming NADH oxidases by expressing their human codon–optimized, epitope-tagged versions in cultured human cancer–derived epithelial (HeLa) cells. The enzyme from Lactobacillus brevis (LbNOX) was most highly expressed and had the highest oxidase activity when targeted to mitochondria (fig. S1).

Fig. 1 H2O-forming NADH oxidase from L. brevis (LbNOX).

(A) Reaction catalyzed by LbNOX. (B) UV-visible spectrum of purified LbNOX. Protein (83 μM FAD active sites) in oxidized form (solid line) and after addition of excess of sodium dithionite, reduced form (dashed line). (Inset) SDS–polyacrylamide gel electrophoresis of purified LbNOX. (C) Simultaneous measurement of NADH and oxygen consumption by LbNOX. NADH and LbNOX were added as indicated by arrows. (D) Dependence of the specific activity (S.A.) of recombinant LbNOX on the concentration of NADH and NADPH. Reported values for Vmax, kcat, and Km for NADH are means ± SD from n = 4 independent experiments. (E) Crystal structure of the catalytic dimer of LbNOX. Each of the two-fold symmetry–related monomers (cyan and green ribbons) contains bound FAD, shown here in spherical (CPK) representation. Details of the catalytic center on the Si-face of FAD and of the substrate selectivity loop are shown in fig. S3, A to C.

We evaluated the biochemical properties of recombinant LbNOX modified to contain a C-terminal FLAG tag and a cleavable N-terminal hexahistidine tag. Purified LbNOX-FLAG has a yellow color in solution and a characteristic ultraviolet (UV)–visible absorption spectrum (major absorption peaks at 371 and 444 nm) consistent with the presence of FAD, which can be reduced upon the addition of sodium dithionite (Fig. 1B). Our recombinant enzyme consumes oxygen and is strictly specific for NADH rather than NADPH with the Michaelis constant (Km) for NADH of 69 ± 3 μM, the maximum velocity (Vmax) of 758 ± 33 μmol·min−1·mg−1, and the turnover number (kcat) of 648 ± 28 s−1, which is more active than previously reported (3, 12) (Fig. 1, C and D). The molecular size of LbNOX-FLAG was determined to be 197 ± 4 kD, which indicates that the protein is a tetramer in solution. Although enzymes in this family often produce H2O2, LbNOX-FLAG produces amounts of H2O2 that constitute only 1 to 2% of the amount of H2O produced during its catalytic cycle (fig. S2A) (4, 6, 7). The apparent Km for O2 of LbNOX-FLAG was below 2 μM (~0.17% O2), as estimated from enzyme-monitored turnover experiments (fig. S2B), which is less than 1/10th of the concentration of oxygen in human venous blood (13). Thus, we expect LbNOX to be active in most animal tissues. The enzymatic properties of LbNOX-FLAG in solution were well founded in the 2.4 Å resolution x-ray structure of this protein that we determined (Fig. 1E, fig. S3, and table S1). Our structure is generally similar to the reported structures of H2O-forming the reduced form of nicotinamide adenine dinucleotide phosphate NAD(P)H oxidases from L. sanfranciscensis (PDB ID 2CDU) and Streptococcus pyogenes (PDB ID 2BC0) (14, 15). However, our structure captures LbNOX in a new state with molecular oxygen (O2) bound and the redox active Cys42 in a reduced form (Cys42-SH) (fig. S3). [See supplementary materials (SM) for a detailed discussion of the x-ray structure.] In conclusion, the high selectivity for NADH over NADPH, negligible H2O2 production, and very low Km for O2 made LbNOX attractive for additional studies in human cells.

To determine whether we could express LbNOX-FLAG safely and efficaciously in various compartments of human cells, we used lentiviral infection to generate HeLa cells that expressed untargeted or mitochondria-targeted human codon–optimized LbNOX-FLAG (referred to as LbNOX and mitoLbNOX henceforth) under the control of a doxycycline-inducible promoter (TRE3G) (Fig. 2A and fig. S1A). We used fluorescence microscopy and cell fractionation to confirm diffuse localization of LbNOX and mitochondrial localization of mitoLbNOX (Fig. 2, B and C). Cells appeared grossly normal without any impact on cell proliferation or production of reactive oxygen species (ROS) (fig. S4, A and B). As expected, expression of LbNOX and mitoLbNOX in HeLa cells increased oxygen consumption by 1.6- and 2.4-fold, respectively (Fig. 2D and fig. S4C). The increase was resistant to ETC inhibitors, which indicates that it resulted from LbNOX oxidase activity and not from the increased ETC activity. Despite similar expression levels (Fig. 2A), mitoLbNOX induced a larger increase in oxygen consumption than LbNOX (Fig. 2D), likely because of the higher concentration of NADH within mitochondria (1618). It is important to remember that in converting NADH to NAD+, LbNOX also consumes protons and oxygen and, therefore, could affect cellular pH or oxygen levels, depending on experimental context.

Fig. 2 Expression and activity of LbNOX in human cells.

(A) Western blot of LbNOX and mitoLbNOX in HeLa cells after 24-hour induction with water or doxycycline (300 ng/ml) (Dox). Representative gel from one of three independent experiments. (B) Subcellular localization of LbNOX and mitoLbNOX in HeLa cells determined by cell fractionation. LRPPRC is a mitochondrial marker and actin is a cytosolic marker. Representative gel from one of three independent experiments. (C) Subcellular localization of LbNOX and mitoLbNOX in HeLa cells determined by using fluorescence microscopy. Tomm20 is a marker of mitochondria. (D) Effect of LbNOX and mitoLbNOX expression in HeLa cells on basal, piericidin-resistant, and antimycin-resistant oxygen consumption measured with an XF24 extracellular flux analyzer. Means ± SEM, n = 3 independent experiments.

We determined the impact of expressing LbNOX or mitoLbNOX on cellular concentrations of NAD+ and NADH (Fig. 3 and fig. S5). We used a genetic sensor, SoNar (19), to measure cytosolic NADH. SoNar is a fusion of circularly permuted yellow fluorescent protein and a modified NADH-binding protein, Rex, from Thermus aquaticus. Binding of NADH to SoNar leads to an increase in fluorescence. Expression of LbNOX or mitoLbNOX decreased the fluorescence signal from SoNar, which indicated that both LbNOX and mitoLbNOX decrease cytosolic NADH (Fig. 3A and fig. S5B). Consistent with this result, intracellular and secreted lactate/pyruvate ratios, traditionally used as proxies for the cytosolic NADH/NAD+ ratio (17), decreased in cells expressing LbNOX or mitoLbNOX (Fig. 3B and fig. S5D). The ratio of total cellular NAD+ to total NADH, based on high-performance liquid chromatography (HPLC) measurements, was increased twofold by mitoLbNOX, whereas LbNOX did not have a significant effect (Fig. 3C and fig. S5C). Perturbation of the total NAD+/NADH ratio likely reflects changes in amounts of mitochondrial NADH, because most of the effect on the ratio resulted from changes in NADH concentration (fig. S5A), and most of the NADH inside the cell is present in mitochondria. The latter effect is supported by fractionation experiments (20) and by the observation that the majority of NAD(P)H autofluorescence in cells comes from mitochondrial NADH (21). In summary, LbNOX and mitoLbNOX can be used to perturb the NAD+/NADH ratio, and our compartment-specific measurements of HeLa cells (Fig. 3, A to C) indicate that, although perturbation of the mitochondrial NAD+/NADH ratio leads to changes in the cytosolic NAD+/NADH ratio, the converse is not true.

Fig. 3 Effect of LbNOX and mitoLbNOX on NAD+/NADH ratios, metabolic fluxes, PDH phosphorylation, and gluconeogenesis.

(A to C) Effect of LbNOX and mitoLbNOX expression in HeLa cells on (A) cytoplasmic NADH concentrations determined with fluorescence microscopy using SoNar-expressing cells (n = 7), (B) intracellular and secreted lactate/pyruvate ratio determined by liquid chromatography–mass spectrometry (LC-MS) (n = 4), and (C) intracellular NAD+/NADH ratios determined by HPLC (n = 4). Student’s t test. ns, P > 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Means ± SEM. (D) Effect of LbNOX and mitoLbNOX expression in HeLa cells on release rate of pyruvate, aspartate, and succinate, determined by comparing concentrations in spent versus fresh media. Student’s t test. ns, P > 0.05; **P < 0.01; ***P < 0.001. Means ± SEM, n = 3 replicates from one experiment. (E) Effect of LbNOX and mitoLbNOX expression in HeLa cells on PDH phosphorylation. Representative gel from one of three independent experiments. DCA, dichloroacetate. (F) Effect of adenoviral transduction of green fluorescent protein (GFP), LbNOX, or mitoLbNOX on primary rat hepatocyte gluconeogenesis in Dulbecco’s modified Eagle’s medium containing no glucose, no glutamine, and no pyruvate and by using either no substrate, 5 mM pyruvate, or 5 mM lactate. One-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test. ns, P > 0.05; *P < 0.05; **P < 0.01; ****P < 0.0001. Means ± SEM, n = 3 (no substrate, pyruvate) or n = 7 (lactate) independent experiments. (G) Effect of LbNOX and mitoLbNOX on the secreted β-hydroxybutyrate/acetoacetate ratio in rat hepatocytes performing gluconeogenesis from lactate as a substrate. Metabolite levels determined by using LC-MS. One-way ANOVA followed by Tukey’s multiple comparisons test. ns, P > 0.05; *P < 0.05; **P < 0.01. Means ± SEM, n = 10 independent experiments.

We performed metabolic profiling on medium in which cells expressing LbNOX or mitoLbNOX had been grown (Fig. 3D and fig. S6, A and B). We identified pyruvate, aspartate, and succinate as three metabolites whose consumption or release was changed more than twofold (Student’s t test; P < 0.01) by either enzyme. These changes are attributable to compartment-specific changes of NAD+/NADH by LbNOX or mitoLbNOX (see SM for discussion). It is notable that LbNOX and mitoLbNOX did not have a significant effect on the uptake of glucose and release of lactate (fig. S6C).

In vitro phosphorylation of mitochondrial pyruvate dehydrogenase (PDH) is regulated by the NAD+/NADH ratio (22), but this has not been shown directly in intact cells. Treatment of HeLa cells with dichloroacetate, an inhibitor of pyruvate dehydrogenase kinases (PDKs), inhibits phosphorylation of PDH, and antimycin treatment, which blocks the ETC and decreases the mitochondrial NAD+/NADH ratio, increases PDH phosphorylation. In agreement with in vitro studies, PDH was almost completely dephosphorylated in the presence of mitoLbNOX but not LbNOX (Fig. 3E and fig. S6D). The data on PDH phosphorylation are consistent with our observation that mitoLbNOX, but not LbNOX, increases the mitochondrial NAD+/NADH ratio in HeLa cells (Fig. 3C).

We expressed LbNOX and mitoLbNOX in primary rat hepatocytes to study gluconeogenesis. The cytosolic NAD+/NADH ratio has been reported to affect gluconeogenesis, although classical approaches relied on indirect methods for manipulating the redox state (23, 24). In our hepatocyte system, rates of gluconeogenesis were significantly higher if pyruvate rather than lactate was used as a substrate, which we attribute to the NAD+/NADH ratio-dependent inhibition of lactate-to-pyruvate conversion (23). Consistent with this hypothesis, rates of gluconeogenesis from lactate were increased to those seen with pyruvate when primary hepatocytes expressed LbNOX, whereas gluconeogenesis from pyruvate was not affected (Fig. 3F). Gluconeogenesis from lactate was also increased by mitoLbNOX. Gluconeogenesis from pyruvate, however, was inhibited by mitoLbNOX, perhaps because strong oxidation of mitochondrial NADH prevents formation of malate (25). In assays of gluconeogenesis using lactate as a precursor, expression of either LbNOX or mitoLbNOX decreased the ratio of secreted β-hydroxybutyrate/acetoacetate (Fig. 3G), the classical proxy for the mitochondrial NADH/NAD+ ratio (17). Although the NAD+/NADH ratio appears to be increased, we cannot exclude the possibility that LbNOX or mitoLbNOX induce hypoxia. LbNOX evidently increased the mitochondrial NAD+/NADH ratio in rat hepatocytes (Fig. 3G) but not in HeLa cells (Fig. 3C), which might reflect differences in cell type or media conditions.

Mammalian cells lacking a functional ETC require the addition of exogenous pyruvate and uridine for cell proliferation (2628). Uridine is required because one of the enzymes in de novo uridine biosynthesis (dihydroorotate dehydrogenase) is coupled to the ETC through coenzyme Q. The requirement for pyruvate, however, has been less clear because it participates in many reactions but has been proposed to rescue cell growth by recycling NAD+ from NADH through cytosolic lactate dehydrogenase (26, 29). If the NAD+ recycling hypothesis is correct, then supplementation with oxaloacetate should have the same effect as pyruvate, because it can be reduced by malate dehydrogenase while recycling NAD+. Oxaloacetate, like pyruvate, rescued the proliferation defect induced by piericidin, whereas malate and lactate did not (Fig. 4A). Alpha-ketobutyrate also rescues the proliferative defect induced by ETC inhibition (30). Furthermore, a large number of α-keto acids can rescue the pyruvate dependence of proliferation in cells with intact ETC (31). These findings support the NAD+-recycling hypothesis, although they are still indirect, as α-keto acids have many metabolic roles.

Fig. 4 NAD+ recycling rescues proliferation in cells with impaired ETC.

(A) Effect of pyruvate, oxaloacetate, lactate, and malate addition on proliferation of HeLa Tet3G luciferase cells in the presence of 200 μM uridine and in the presence or absence of 1 μM piericidin. Means ± SEM, n = 5 independent experiments. (B) Effect of LbNOX and mitoLbNOX expression in HeLa cells on inhibition of cell proliferation by 1 μM piericidin, 1 μM antimycin, 10 μg/ml chloramphenicol, and 30 ng/ml ethidium bromide (EtBr) in the presence of 200 μM uridine. DMSO, dimethyl sulfoxide. Means ± SEM, n = 3 independent experiments.

We used LbNOX to directly test whether NAD+ recycling is an essential function of the ETC that is required for cell proliferation. We inhibited ETC function, with piericidin (a complex I inhibitor), antimycin (a complex III inhibitor), ethidium bromide (a mitochondrial DNA replication inhibitor), and chloramphenicol (an inhibitor of mitochondrial translation) in HeLa cells supplemented with uridine but lacking pyruvate. HeLa cells cannot proliferate in these conditions (Fig. 4B and fig. S7). Expression of either LbNOX or mitoLbNOX rescued cell proliferation in the presence of these ETC inhibitors, which indicated that regeneration of NAD+ in either cytosol or mitochondria is sufficient to complement ETC activity that is required for cell proliferation (Fig. 4B). Metabolic profiling showed that of the nine metabolites whose uptake or release is affected greater than twofold by antimycin (Student’s t test; P < 0.01), all could be reversed by either LbNOX or mitoLbNOX, which reflects a metabolic rescue (fig. S8) (see SM for discussion). Our metabolic profiling data are complementary to recent studies demonstrating an inhibition of aspartate biosynthesis in cells with dysfunctional ETC (30, 32, 33). As a control, we also showed that the rescue by LbNOX or mitoLbNOX was not attributable to an effect on mitochondrial membrane potential (fig. S9, A to C), nor was it due to a rescue of ETC-derived ATP synthesis (fig. S9, D to G).

Collectively, these studies (Fig. 4 and figs. S7 to S9) underscore the importance of NAD+ recycling by the ETC to support proliferation. In healthy cells, the ETC produces ATP and simultaneously recycles mitochondrial NADH to NAD+, with a secondary oxidation of cytosolic NADH via shuttles. In the absence of a functional ETC, glycolysis is capable of compensating for the lack of ATP production, but it is net redox-neutral. NAD+ recycling is likely key for cell proliferation, because many biosynthetic pathways produce NADH as a byproduct (34). These insights confirm the long-standing hypothesis (26, 29) that pyruvate supplementation rescues proliferation in cells with disrupted ETC by restoring NAD+/NADH balance via the LDH reaction.

In the future, LbNOX and engineered or naturally occurring variants may become valuable tools for studying compartmentalization of redox metabolism. These constructs will allow for a dissection of the relative contributions of redox imbalance and ATP insufficiency to mitochondrial disease pathogenesis. If a substantial amount of the organ pathology of mitochondrial disease stems from reductive stress or pseudohypoxia, then expression of this single polypeptide holds promise as a “protein prosthesis” for the large number of disorders characterized by ETC dysfunction.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 to S3

References (3552)

References and Notes

Acknowledgments: We thank V. Vitvitsky for technical support with HPLC. We thank members of the Mootha lab for valuable discussions and feedback. This work was supported by a T-R01 R01GM099683 grant from NIH. D.V.T. was supported by a Tosteson and Fund for Medical Discovery Postdoctoral Fellowship Award. R.P.G. was supported by a T32DK007191 grant from NIH. V.K.M. is an investigator of the Howard Hughes Medical Institute. The Massachusetts General Hospital has filed a patent application on the technology described in this paper. Atomic coordinates and structure factors have been deposited in the Protein Data Bank with accession number 5ER0.

Stay Connected to Science

Navigate This Article