Flux through mitochondrial redox circuits linked to nicotinamide nucleotide transhydrogenase generates counterbalance changes in energy expenditure

Compensatory changes in energy expenditure occur in response to positive and negative energy balance, but the underlying mechanism remains unclear. Under low energy demand, the mitochondrial electron transport system is particularly sensitive to added energy supply ( i.e. reductive stress), which exponentially increases the rate of H 2 O 2 ( J H 2 O 2 ) production. H 2 O 2 is reduced to H 2 O by electrons supplied by NADPH. NADP 1 is reduced back to NADPH by activation of mitochondrial membrane potential – dependent nicotinamide nucleotide transhydrogenase (NNT). The coupling of reductive stress-induced J H 2 O 2 production to NNT-linked redox buffering circuits provides a potential means of integrating energy balance with energy expenditure. To test this hypothesis, energy supply was manipulated by varying flux rate through b -oxidation in muscle mitochondria minus/plus pharmacological or genetic inhibition of redox buffering circuits. Here we show during both non-ADP – and low-ADP – stimulated respiration that accelerating flux through b -oxidation generates a corresponding increase in mitochondrial J H 2 O 2 production, that the majority ( ~ 70 – 80%) of H 2 O 2 produced is reduced to H 2 O by electrons drawn from redox buffering circuits supplied by NADPH, and that the rate of electron flux through redox buffering circuits is directly linked to changes in oxygen consumption mediated by NNT. respirometry (24). Progressive additions of carnitine from 25 m J O 2 by 8-fold with no further increase observed above 5 m M carnitine. Titration of carnitine in the presence of de-energized mitochondria (no substrate) or in the presence of NADH-generating substrates ( e.g. 5 m M glutamate, 2 m M malate) did not affect J O 2 (not shown), indicating the effect of carnitine was not due to a technical artifact or uncoupling.

Body weight is remarkably stable in adults over long periods of time despite daily fluctuations in energy intake and expenditure (1). Consistent with homeostatic weight maintenance, energy expenditure rates increase in humans during periods of weight gain and decrease during periods of weight loss, well in excess of what can be accounted for by changes in fat free mass, the thermic effect of food, and fecal calorie loss (2,3). Remarkably, even when body weight is held stable after active weight gain or loss, energy expenditure remains elevated or depressed, respectively (3). Changes in circulating thyroid hormone (4,5), leptin (6), sympathetic tone (5), and skeletal muscle efficiency (7) have all been implicated, but the underlying mechanism(s) that accounts for these apparent compensatory changes in metabolic efficiency remains unresolved.
The ability of living cells to continuously respond to fluctuations in nutrient availability and energy demand depends on the interplay between metabolic and redox circuits. How these circuits connect and respond in real time is not clear. The dynamic equilibrium among opposing free energy driving forces across the inner mitochondrial membrane (i.e. DG redox -NAD 1 /NADH and FAD/FADH 2 , DG DCmmitochondrial membrane potential, and DG ATPmatrix [ATP]/[ADP]) provides a potential ideal mechanism for sensing cellular energy balance (8). This is obvious with respect to energy demand, where an increase in free [ADP] (i.e. decrease in mitochondrial DG ATP ) generates a corresponding increase in ATP synthase activity, an increase in proton (H 1 ) conductance, a decrease in DG DCm , and an increase in electron flow and oxygen consumption (9). Less appreciated is the fact that the electron transport system (ETS) is equally suited to sense energy surplus, as an excess of NADH and/or FADH 2 supply relative to demand, particularly at rest, increases the pressure head (i.e. DG redox ), DG DCm , and thus reductive stress (i.e. oxidation potential) within the ETS, which exponentially increases the rate of mitochondrial H 2 O 2 emission (10)(11)(12)(13).
Mitochondrial JH 2 O 2 emission reflects the balance between JH 2 O 2 production and JH 2 O 2 reduction to H 2 O. The latter is catalyzed by matrix thioredoxin and GSH peroxidases whose substrates, thioredoxin and GSH, are converted back to their reduced states by thioredoxin reductase and GSH reductase, respectively, using NADPH as the electron source (Fig. 1A). The reduction of NADP 1 back to NADPH in turn is catalyzed, at least in part, by nicotinamide nucleotide transhydrogenase (NNT), an inner mitochondrial membrane protein that uses DC m to establish and maintain a high NADPH/NADP 1 ratio (14).

Reaction 1
Thus, in theory, anytime the rate of energy supply outpaces energy demand (i.e. nutrient overload), the resulting increase in JH 2 O 2 production should induce a corresponding counterbalance increase in NNT-mediated energy expenditure. The purpose of the present study was to test this hypothesis by determining the extent to which flux rates through b-oxidation and the ETS during low rates of respiration influence overall/ site-specific H 2 O 2 production rates and whether JH 2 O 2 production is coupled to NNT-mediated energy-consuming redox buffering circuits.

Results
Increased flux through b-oxidation increases JO 2 , JH 2 O 2 , and electron flux through redox buffering circuits To examine whether mitochondrial JH 2 O 2 emission may be coupled to compensatory changes in NNT-mediated energy expenditure, JH 2 O 2 emission was measured by Amplex Ultra Red in mitochondria isolated from hind limb skeletal muscle of C57BL/6N (B6N) mice. Experiments were conducted under non-ADP-stimulated conditions (i.e. state 4 respiration) supported by palmitoyl-CoA in the presence of low (25 mM) or high (5 mM) carnitine. Transfer of long-chain fatty-acyl groups from CoA to carnitine by carnitine palmitoyltransferase-1 (CPT-1) is required for entry into the mitochondria and considered the rate-limiting step for b-oxidation (15,16). We focused on fatty acid oxidation because it has multiple potential reductive stress points (i.e. sites of oxidant production), including dehydrogenation reactions within the b-oxidation pathway (17)(18)(19), the electron transfer flavoprotein-ubiquinone oxidoreductase (19)(20)(21)(22), and within complexes I, II, and III of the ETS (19)(20)(21)(22)(23). Mitochondrial JH 2 O 2 emission was 3-fold greater in the presence of high versus low carnitine (Fig. 1B). The subsequent addition of inhibitors to thioredoxin reductase (auranofin (AF)) and GSH reductase (carmustine, BCNU) induced a marked increase in JH 2 O 2 detection under both carnitine concentrations. These findings demonstrate that H 2 O 2 is produced at much higher rates than what is emitted and that the majority (;70-80%) of H 2 O 2 produced during b-oxidation is reduced to H 2 O (i.e. JH 2 O 2 buffered (%)) by matrix redox buffering circuits (Fig. 1B, right y axis).
We next determined whether the carnitine-induced increase in JH 2 O 2 production is associated with an increase in the rate of oxygen consumption (JO 2 ), as measured by high-resolution respirometry (24). Progressive additions of carnitine from 25 mM to 5 mM in the presence of palmitoyl-CoA dose-dependently increased JO 2 by 8-fold (Fig. 1C), with no further increase observed above 5 mM carnitine. Titration of carnitine in the presence of de-energized mitochondria (no substrate) or in the presence of NADH-generating substrates (e.g. 5 mM glutamate, 2 mM malate) did not affect JO 2 (not shown), indicating the effect of carnitine was not due to a technical artifact or uncoupling. Inclusion of malonyl-CoA, an inhibitor of CPT-1, decreased maximal palmitoyl-CoA plus carnitine-supported JO 2 by 75-80% (Fig. 1D), confirming that carnitine-induced flux through b-oxidation depends on CPT-1 activity. In addition to facilitating transport into mitochondria, carnitine also accelerates b-oxidation flux by providing substrate for carnitine acetyltransferase (CrAT) (20,24,25), a mitochondrial enzyme that converts excess acetyl-CoA to membranepermeant acetylcarnitine esters for export out of the mitochondria (26). Acetyl-CoA is a potent negative regulator of the b-ketoacyl-CoA thiolase reaction, the final step in b-oxidation ( Fig. 1A) (27). To determine whether carnitine increases b-oxidation flux via CrAT, saponin-permeabilized skeletal muscle fiber bundles (PmFBs) were prepared from WT and musclespecific CrAT knockout (CrAT m/2 ) mice and studied during state 4 respiration supported by palmitoylcarnitine. In the absence of additional carnitine, JO 2 was low in both WT and CrAT m/2 tissue (Fig. 1E). The addition of excess carnitine (5 mM) increased JO 2 by ;9-fold in PmFBs from WT mice but had no effect in PmFBs from CrAT m/2 mice. The subsequent addition of malate to facilitate entry of acetyl-CoA into the TCA cycle, and thus relieve acetyl-CoA-mediated inhibition of b-ketoacyl-CoA thiolase in CrAT m/2 mitochondria, increased JO 2 by .8-fold, negating the difference between genotypes. These data confirm that carnitine addition accelerates CrAT activity, relieving acetyl-CoA-mediated product inhibition and thereby increasing flux through b-oxidation.
Integration of site-specific H 2 O 2 production with redox buffering systems To examine the integration of H 2 O 2 production with the redox buffering systems, mitochondrial JH 2 O 2 emission and JH 2 O 2 production were measured during a series of substrate/ inhibitor protocols designed to isolate electron leak to specific site(s) during b-oxidation. Electrons entering the Q-pool from b-oxidation have the potential to leak at the quinone-oxidizing sites of complex III (site III Qo ), complex I (site I Q ), and complex II (site II Q ) (28). The flavin site in complex II (site II F ) also produces H 2 O 2 during b-oxidation as a consequence of increased TCA cycle flux and release of oxaloacetate-mediated inhibition of complex II ( Fig. 2A) (20,29). During respiration supported by PCoA plus carnitine, the addition of malonate to block the forward reaction into complex II (i.e. site II F ) decreased H 2 O 2 emission by ;30% (Fig. 2B), leaving site I Q , site II Q , site III Qo , and the upstream b-oxidation dehydrogenases and flavoproteins as the potential remaining sources. The subsequent addition of AF/BCNU increased the rate of H 2 O 2 detection by 5fold (Fig. 2B), confirming the rate of H 2 O 2 production greatly exceeds the rate of H 2 O 2 emission and that flux through the redox buffering circuits accounts for the difference.
Myxothiazol inhibits electron flow and H 2 O 2 production at site III Qo , causing upstream sites (I F , II F , and the upstream b-oxidation enzymes) to become more reduced and dramatically increasing overall JH 2 O 2 emission (Fig. 2C). The addition of malonate in the presence of myxothiazol decreased JH 2 O 2 emission by nearly 90%, confirming reverse electron flow through complex II accounts for the majority of JH 2 O 2 emission when forward electron flow into complex III is blocked during palmitoyl-CoA plus carnitine-supported respiration (20). The subsequent addition of AF/BCNU induced a 3-fold increase in JH 2 O 2 emission, again indicating the majority of H 2 O 2 produced (i.e. ;70%) is reduced by the redox buffering circuits (Fig. 2C). In separate experiments, the addition of AF/ BCNU after myxothiazol increased JH 2 O 2 by only 1.6-fold ( Fig.  2D), suggesting the efficiency of redox buffering circuits is limited at extremely high rates of H 2 O 2 production. Taken together, these findings reveal that the rate of mitochondrial H 2 O 2 emission elicited by b-oxidation is a stark underestimate of the actual rate of H 2 O 2 production due to the capacity of the matrix redox buffering circuits to reduce H 2 O 2 to H 2 O. The findings also suggest that the efficiency of redox buffering circuits is likely dependent on the integration and capacity of NADPH-generating sources.
Flux through NNT redox circuits contributes to energy expenditure NNT uses the free energy of the mitochondrial protonmotive force to generate and maintain a highly reduced NADP 1 / NADPH ratio, which in turn generates and maintains highly reduced GSH (GSSG/GSH) and thioredoxin (Trx2 ox /Trx2 red ) pools. Electrons are drawn from NADPH through the thioredoxin and GSH redox circuits to reduce H 2 O 2 as a function of the demand imposed by JH 2 O 2 production, which should therefore determine the rate of NNT-mediated proton conductance and thus energy expenditure. To test this hypothesis, proton conductance assays were carried out, similar to the approach used to define proton leak from uncoupling protein activity (12). In this assay, JO 2 is measured during state 4 respiration supported by succinate/rotenone as DC m is progressively decreased by titration of malonate, a complex II inhibitor. A difference in JO 2 at the highest common DC m reflects the difference in proton conductance into the matrix between experimental conditions. Experiments were conducted in the presence of two different b-oxidation flux/H 2 O 2 production rates (induced by 25 mM versus 5 mM carnitine) and thus two different rates of demand on NNT. The higher carnitine concentration induced a leftward shift in the curve and a greater JO 2 (1276 versus 1020 pmol/s/mg of protein) at the highest common DC m (2154 mV) (Fig. 3A), consistent with a greater rate of proton conductance and energy expenditure when flux through b-oxidation is accelerated. Inclusion of AF/BCNU prevented the leftward shift in the proton conductance curve induced by high carnitine (Fig. 3B), confirming that increased electron flux through the thioredoxin and GSH redox buffering circuits accounted for the higher JO 2 .

EDITORS' PICK: Redox circuit-mediated energy expenditure
To determine whether the greater proton conductance is mediated by NNT, skeletal muscle mitochondria were isolated from C57BL/6J (B6J) mice, which carry a naturally occurring, in-frame, five-exon deletion mutation within the Nnt gene, rendering the NNT protein nonfunctional (30,31). Maximal b-oxidation flux (5 mM carnitine) induced a 2-fold greater JH 2 O 2 emission in mitochondria from B6J mice compared with normal NNT-expressing B6N mice (Fig. 3C), consistent with the absence of flux through NNT-linked redox circuits in B6J mice. Indeed, when AF/BCNU was added to inhibit the thioredoxin/ GSH redox circuits, JH 2 O 2 production was not different between genotypes (i.e. JH 2 O 2 production increased in B6N but not B6J mitochondria). As such, in the absence of NNT, mitochondria from B6J mice were able to buffer only ;11% of the H 2 O 2 produced (Fig. 3C), consistent with prior findings indicating NNT is the major supplier of NADPH with additional production coming from two NADP 1 -linked TCA enzymes, isocitrate dehydrogenase 2 and malic enzyme 3 (32). Finally, mitochondria from B6J mice displayed a rightward shift in the proton conductance curve and lower JO 2 (1125 versus 1450 pmol/s/mg of protein in B6J versus B6N, respectively) at the highest common DC m (2153 mV) under maximal b-oxidation flux/JH 2 O 2 production conditions, confirming that NNT mediates the increase in proton conductance (Fig. 3D).
Proton conductance assays require specific substrate conditions (i.e. succinate, titration of malonate to progressively inhibit complex II) and are performed under a non-ADPstimulated respiratory state (state 4). To assess the interplay between b-oxidation-induced JH 2 O 2 production and flux through NNT-linked redox circuits under conditions that more closely model in vivo bioenergetics, mitochondria isolated from red skeletal muscle of B6N and B6J mice were subjected to an energy clamp system that incorporates the creatine kinase reaction to stepwise "clamp" ATP/ADP ratios at different values, and thus DG ATP (i.e. the free energy backpressure on ATP synthase), over the entire range of ADP-stimulated respiration (33,34). Respiration was again supported by PCoA (20 mM) and high carnitine (5 mM) to maximize reducing pressure from the b-oxidation pathway. Steady-state JO 2 progressively decreased in mitochondria from B6N and B6J mice as DG ATP became more negative (i.e. as JO 2 approached state 4 values) (Fig. 4A), which generated a progressive increase in JH 2 O 2 emission that was evident only in mitochondria from B6J mice (which lack NNT) (Fig. 4C). At the most negative DG ATP (266.9 kJ/mol), JH 2 O 2 emission was 2-fold higher (Fig. 4F)  JO 2 consumption was 18.6% lower (Fig. 4D) in B6J versus B6N mitochondria, consistent with the absence of an intact NNTlinked redox circuit in B6Js (Fig. 3D). The subsequent addition of BCNU/AF to inhibit redox circuit flux more than doubled the JH 2 O 2 detected in B6N mitochondria (i.e. revealed the actual JH 2 O 2 production), negating the difference between B6J and B6Ns (Fig. 4F). As expected, BCNU/AF decreased JO 2 in mitochondria from both genotypes (Fig. 4D), whereas DC m remained stable (Fig. 4, B and D).

Discussion
Linking JH 2 O 2 production to NNT-mediated energy expenditure Mitochondrial (and by extension cellular) redox circuits rely on both the GSSG/GSH (E 0 9 = 2240 mV) and Trx2 ox /Trx2 red (E 0 9 = 2230 mV) redox couples to protect against oxidants. The NADP 1 /NADPH redox couple, which possesses a lower standard midpoint potential (E 0 9 = 2320 mV) serves as the source of electrons for both the GSH and thioredoxin redox couples. In vivo, NNT driven by DC m maintains the NADP 1 /NADPH redox couple at a more negative (reduced) steady state (E h = ;2415 mV), which in turn holds both GSSG/GSH and Trx2 ox / Trx2 red in more reduced steady states (35). This resulting mito-chondrial redox charge is distributed throughout the cell and thought to be responsible for maintaining ;90% of the redoxsensitive thiols in the proteome in a reduced state (36, 37). Once the redox state of the proteome is established, it follows that the rate at which electrons are drawn from the reductive source (i.e. NADPH) through the redox circuits (i.e. Trx red and GSH) is determined by the rate of oxidative input into the circuit (i.e. the rate of H 2 O 2 production and/or oxidation of redox-sensitive protein thiols). In isolated mitochondria under state 4 or low state 3 conditions, the ETS is extremely sensitive to forward reductive stress (i.e. DG redox ), increasing JH 2 O 2 production exponentially with even small increases in DC m (10-12). The system is therefore poised to respond to energy supply outpacing energy demand by coupling the consequent rate H 2 O 2 production directly to electron flux through redox buffering circuits linked to NNT and thus energy expenditure.
The objective of this study was to test this hypothesis by determining whether mitochondrial JH 2 O 2 production stemming specifically from flux through b-oxidation is directly linked, via redox buffering circuits, to corresponding changes in NNT-mediated JO 2 . In vivo, mitochondrial JH 2 O 2 production is thought to increase when flux rate through catabolic pathways, and thus reducing equivalent supply, outpaces metabolic demand (13). In isolated mitochondria, however, this is A.

EDITORS' PICK:
Redox circuit-mediated energy expenditure technically difficult to model due to the inability to recapitulate the same metabolic and redox free energy charges present in vivo. In the present study, we focused specifically on the b-oxidation pathway because flux through the pathway, and thus reducing equivalent supply to the ETS, can be experimentally accelerated by carnitine independent of the TCA cycle and/or ATP demand. The findings reveal that under non-or low-ATP demand states, JH 2 O 2 production is directly related to the rate of flux through b-oxidation, that ;80% of the H 2 O 2 produced is reduced to H 2 O by electrons drawn through the thioredoxin and GSH redox circuits, and that flux through these redox circuits elicits a corresponding increase in NNT-mediated JO 2 (i. e. energy expenditure). Together with a prior study showing a similar link between the pyruvate dehydrogenase complex and NNT (24), the data support the hypothesis that energy balance in mitochondria is continuously sensed and integrated, via JH 2 O 2 production and NNT-linked redox buffering circuits, to compensatory changes in energy expenditure.

Potential contribution of NNT-linked redox buffering circuits to basal proton conductance
Basal proton conductance (i.e. state 4 respiration) in mitochondria is estimated to account for as much as 25% of resting metabolic rate (38), yet the molecular mechanism(s) responsible for what is commonly referred to as proton leak has remained enigmatic (12). Nonspecific proton conductance through the adenine nucleotide translocase (i.e. not associated with ATP/ADP exchange) reportedly accounts for ;50% with the remainder generally attributed to nonspecific proton leak between the lipid bilayer and membrane-bound proteins (12,39,40). It is important to recognize that electron leak from the ETS to O 2 to form superoxide, and its subsequent conversion to H 2 O 2 by superoxide dismutase 2, is thermodynamically favorable under state 4 and low state 3 conditions (13). Thus, the coupling of JH 2 O 2 production to NNT-mediated proton conductance via redox buffering circuits provides a potential additional mechanism contributing to resting metabolic rate (Fig.  5A). Consistent with this hypothesis, energy expenditure during the light cycle is lower in mice that lack NNT (B6J) as compared with genetically similar mice that express NNT (B6N) (24,41,42)

Potential physiological implications of NNT-linked redox buffering circuits
Whether flux through NNT-linked redox circuits contributes to energy expenditure in a physiologically meaningful way depends in part on the stoichiometric coupling of NNT (i.e. the number of H 1 translocated across the mitochondrial membrane per pair of electrons transferred from NADH to NADP 1 and the overall NADPH/O ratio). For ATP synthase, based on structural data from vertebrates (9), the H 1 /ATP stoichiometry of the proton turbine F o component is well defined (;2.7), and P/O ratios ranging from ;2.0 to 2.8 have been empirically measured, depending on the experimental conditions (9,43). NNT is also a complex protein arranged as a homodimer with each component containing a proton-translocating transmembrane domain and two nucleotide-binding domains in the  matrix that mediate the electron transfer from NADH to NADP 1 (44). Early studies using submitochondrial particles or liposomes containing reconstituted NNT suggested a H 1 /2e 2 stoichiometry of ;1, which agrees with the roughly equivalent difference in redox poises of the NADP 1 /NADPH and NAD 1 / NADH couples and DC m in mitochondria under basal conditions. However, the mechanism and stoichiometry underlying the coupling of H 1 translocation to NADPH synthesis when the enzyme is actively generating NADPH remains unknown (44). NNT is thought to undergo significant protein conformational changes in response to changes in nucleotide binding (44,45), which could alter the H 1 /2e 2 stoichiometry. For example, early work using submitochondrial particles suggested that NNT was much less efficient than ATP synthase (46). In the present study, JH 2 O 2 production at the lowest clamped rate of respiration (DG = 266.9 kJ/mol) was ;7.0 pmol/s/mg of protein higher in B6J than B6N mitochondria, which corresponded to ;118.5 pmol/s/mg of protein lower JO 2 in B6J versus B6N mitochondria (Fig. 4, D and F). Assuming this rate of H 2 O 2 production drew the equivalent rate of NADPH oxidation and, in turn rate of NADPH resynthesis by NNT, the effective NADPH/O was ;0.03, implying that when activated, NNT may be quite inefficient with respect to H 1 /2e 2 stoichiometry. There are of course a number of meth-odological limitations and caveats with such empirical data, including the possibility that factors other than the absence of NNT contribute to the difference in JO 2 between B6J and B6N mitochondria. However, in contrast with ATP synthesis, where high efficiency is advantageous, a low efficiency of NADPH synthesis serves to counterbalance the very pressures responsible for increasing JH 2 O 2 production (i.e. reductive stress from energy supply outpacing energy demand) (Fig. 5B).
In conclusion, this study reveals the degree to which mitochondrial bioenergetics couples energy/redox balance to energy expenditure, presumably to minimize disturbances to redox homeostasis throughout the proteome. In vivo, this implies that electron flux through redox buffering circuits enables adjustments in energy expenditure to changes in energy balance in real time, using redox potential within pivotal redox reactions as both the sensor and initiator of the counterbalance response. The cycling of electrons through redox buffering circuits could thus account for a significant portion of resting metabolic rate as well as contribute to the compensatory changes in energy expenditure observed in humans under positive and negative steady-state energy balance conditions (2). The identification of how redox state sensing mechanisms are coupled to compensatory changes in flux through mitochondrial redox buffering circuits also provides potential new avenues for therapeutic development to treat diseases stemming from chronic metabolic imbalance.

Experimental procedures
Animals Male C57Bl/6NJ (B6N) and C57Bl/6J (B6J) mice (Jackson Laboratories (Bar Harbor, ME, USA), 005304 and 00664, respectively) were bred and housed in a temperature-controlled (22°C) facility with a 12-h light/dark cycle. Mice were fed a standard, low-fat pellet diet (Purina ProLab ISOPro RMH 3000, St. Louis, MO, 5P76). At 12-20 weeks of age, mice were anesthetized with intraperitoneal ketamine/xylazine (18/2 mg/ml) injection (5 ml/g body weight) prior to tissue collection and cervical dislocation. All animal procedures were approved by the East Carolina University Institutional Animal Care and Use Committee.

Mitochondrial isolation
Differential centrifugation was used to isolate skeletal muscle mitochondria as performed previously (47) but with a few modifications. In brief, whole gastrocnemius, tibialis anterior, and quadriceps muscles from both hind limbs were dissected and placed immediately in a Petri dish containing 8 ml of ice-cold mitochondrial isolation medium (MIM): 300 mM sucrose (Thermo Fisher Scientific, BP2201), 10 mM HEPES (Sigma, H3375), and 1 mM EGTA (Sigma, E4378). Muscles were transferred to a separate Petri dish and finely minced with scissors while kept on ice. The minced muscle was homogenized in 8 ml of ice-cold MIM 1 0.5 mg/ml BSA (Sigma, A3803) until consistent (;8 passes) using a tight-fitting Teflon glass homogenizer. The homogenate was centrifuged for 10 min at 800 3 g and 4°C. The resulting supernatant was transferred to an Oakridge tube and centrifuged at 12,000 3 g and 4°C for an additional 10 min. The supernatant was discarded, and the pellet Blue lines indicate the drawing or "pull" of electrons by oxygen through the ETS and b-oxidation pathway and by H 2 O 2 (derived from superoxide, not shown) through the redox buffering circuits from NADPH. Red lines indicate the "push" of electrons as a consequence of reductive stress (oxidation potential indicated by the gray gauges) leading to H 2 O 2 production. JH 1 represents the rate of proton conductance through NNT as a consequence of the rate of NAPDH oxidation. B, high fuel supply; as in A showing increase in reductive stress due to high fuel supply relative to demand increasing JH 2 O 2 production, the rate of electron flux through redox buffering circuits, NNT-mediated proton conductance, and thus energy expenditure.
was rinsed once in 1 ml of MIM. The pellet was gently resuspended in 150 ml of MIM. The resulting mitochondrial preparation was diluted 503 prior to measuring the mitochondrial protein concentration (Pierce BCA Protein Assay, Thermo Fisher Scientific, 23225).

Creatine kinase energetic clamp
Experiments utilizing the creatine kinase energetic clamp technique were performed as described previously (33,34,49). Briefly, JO 2 and DC m were measured simultaneously in O2K respirometers under increasing values of clamped ATP-free energy using known amounts of creatine (Cr), phosphocreatine (PCr), and ATP in the presence of excess creatine kinase (CK). DG ATP was calculated using the equilibrium constant of the CK reaction (i.e. K CK ) from the equation where DG9 o ATP is the standard apparent transformed Gibbs energy (at specific pH, ionic strength, free magnesium, and pressure), R is the gas constant (8.3145 J/kmol), and T is the temperature in Kelvin (310.15). Experiments were performed in Buffer Z with the exception that the KCl was replaced by sucrose (10 mM) to prevent the ionic strength from increasing too high from counterion equivalents in the PCr additions. Buffer Z was also supplemented with ATP (5 mM), Cr (5 mM), PCr (1 mM), CK (20 units/ml), PCoA (20 mM), and carnitine (5 mM). PCr was sequentially added to progressively increase DG ATP . Both DG9 o ATP and K9 CK were calculated at each titration step to account for changes in buffer ionic strength and free magnesium (33). JH 2 O 2 was measured fluorometrically in parallel experiments under identical buffer and CK clamped conditions.

Statistics
Data are presented as means 6 S.E. The data exhibited normal distribution and were analyzed by unpaired Student's t tests with significance set at p , 0.05. Microsoft Excel and GraphPad Prism were used for statistical tests and data presentation.

Data availability
All raw data used to generate the data figures are available upon request from Dr. P. Darrell Neufer (neuferp@ecu.edu). Funding and additional information-This work was supported by National Institutes of Health Grants R01 DK096907 and R01 DK110656 (to P. D. N.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.