Supercomplex Assembly Determines Electron Flux in the Mitochondrial Electron Transport Chain

See allHide authors and affiliations

Science  28 Jun 2013:
Vol. 340, Issue 6140, pp. 1567-1570
DOI: 10.1126/science.1230381

Respiration Refined

Cells derive energy from redox reactions mediated by mitochondrial enzymes that form the electron transport chain. The enzymes can form large complexes, known as supercomplexes, whose function has been controversial. Lapuente-Brun et al. (p. 1567) discovered that a mouse protein, supercomplex assembly factor I (SCAFI), specifically modulates assembly of respiratory complexes into supercomplexes. Formation of the supercomplexes appears to cause electrons to be processed differently, depending on the substrate from which they are derived.


The textbook description of mitochondrial respiratory complexes (RCs) views them as free-moving entities linked by the mobile carriers coenzyme Q (CoQ) and cytochrome c (cyt c). This model (known as the fluid model) is challenged by the proposal that all RCs except complex II can associate in supercomplexes (SCs). The proposed SCs are the respirasome (complexes I, III, and IV), complexes I and III, and complexes III and IV. The role of SCs is unclear, and their existence is debated. By genetic modulation of interactions between complexes I and III and III and IV, we show that these associations define dedicated CoQ and cyt c pools and that SC assembly is dynamic and organizes electron flux to optimize the use of available substrates.

Electron transport in the mitochondria cannot be fully explained by the classical fluid model (1) or the solid model, which proposes that the respirasome is the only functional mitochondrial electron transport chain (mETC) (2, 3). However, both models can be regarded as extremes of a more dynamic situation in which the respirasome [containing complexes I, III, and IV (CI, CIII, and CIV)], the other supercomplexes (SCs) (CI and CIII or CIII and CIV), and free respiratory complex (RC) populations coexist (4). Although CI is unstable in the absence of CIII or CIV (57), definitive evidence for physical association between RCs in vivo has been lacking, and their putative functions remain unclear (2, 4, 812).

To investigate interaction between CI and CIII, we used mitochondria from mouse fibroblasts with constitutively abnormal low expression of CIII (S cells) (supplementary materials and fig. S1). Blue-native gel electrophoresis (BNGE) detected fewer CI-containing SCs in S cells than in control cells; moreover, no free CIII was detected (Fig. 1A), suggesting that CI sequesters the limited amount of CIII. Oxygen consumption was measured in permeabilized S cells incubated with either the mETC substrates glutamate and malate, which generate intramitochondrial NADH (reduced form of nicotinamide adenine dinucleotide) to feed electrons to CI, thus promoting respiration through CI, CIII, and CIV, or succinate, which feeds electrons to CII, via flavin adenine dinucleotide (FAD), promoting respiration through CII, CIII, and CIV (fig. S1D). Combined with the use of specific inhibitors, this allowed us to measure respiration through alternate routes to CIV and also to spectrophotometrically determine electron flux through mETC components, both individually (CI, CII, CIII, or CIV) and in combination (CI with CIII or CII with CIII) (fig. S1F). The low content of CIII in S cells disrupted respiration from succinate and the activity of CII combined with CIII, indicating that the limited CIII is unavailable to electrons from CII (or other sources that deliver electrons to FAD, such as glycerol-3-phosphate dehydrogenase) (Fig. 1B). We ruled out that changes in coenzyme Q (CoQ) levels would explain this behavior (fig. S1G).

Fig. 1 Separate CoQ pools supply electrons from CI and CII to CIII.

(A) Immunoblot of assembled supercomplexes in digitonin-permeabilized mitochondria separated by BNGE and probed with monoclonal antibodies for CI (anti-NDUFB6), CIII (anti-Core2), CIV (anti-COI), and CII [anti-SDHA (I30)]. (B) Box plot representation of mitochondrial complex activities, either alone or in combination with CIII, in C, S, and heteroplasmic ND6 mutant cells (lacking the CI subunit ND6-M, and thus partially depleted for CI) (1921). Activities were measured spectrophotometrically in the presence of specific electron donors and complex inhibitors (see supplementary materials). When cells were grown in medium containing glucose instead of galactose—to allow them to survive through glycolysis—free CIII was detectable in S cells. SC, supercomplexes containing CI; III2+IV, supercomplex III+IV; III2, dimeric CIII. (C) BNGE immunoblot of CIII-containing supercomplexes in S cells grown with glucose or galactose. (D) BNGE immunoblots showing distribution of CI, CIII, CIV, and CII in control cells (C) and S cells shRNA-depleted for the CI subunit NDUFS3 (S) at the indicated multiplicity of infection (MOI). IV and IV*, nonsuperassembled complex IV. (Inset) SDS–polyacrylamide gel electrophoresis (SDS-PAGE) immunoblot of NDUFS3 (actin was the loading control). (E) Activities of isolated CI, CI+CIII, and CII+CIII in control cells and NDUFS3-interfered S cells (CI-Kd). Respiration capacity of S cells was lower than in (B) because they were in glucose medium. (F) Ubiquinone exists in independent pools, one for NADH (blue) and other for FAD-containing enzymes (brown). Data in (B) and (E) are presented as box plots (horizontal bar, median; box limits, 25th and 75th percentiles; whiskers, 10th and 90th percentiles; dots, observations outside the 10th and 90th percentiles). *Significant differences by analysis of variance (ANOVA). Paired differences were assessed by the post hoc Fisher’s protected least significant difference test (PLSD); detailed statistical analysis for all figures is described in the supplementary materials.

If physical association of CI with CIII creates separate populations of CIII molecules (the bound population dedicated to CI, the unbound to other enzymes), depletion of CII would not increase CIII-bound CI nor the flux between CI and CIII, but depletion of CI would release CIII and increase electron transport through CII and CIII. Small interfering RNA (siRNA)–mediated depletion of CII in control fibroblasts reduced activity of CII and CII with CIII without affecting CI or CI with CIII (fig. S1, H and I). In contrast, in a cell line in which CI was stably depleted by 60% (13), CIII activity was normal and CII activity slightly increased, although combined activity of CI and CIII was normal and the linked activity of CII with CIII was increased (Fig. 1B). These results support the idea that electron flux from CI to CIII takes place within SCs, whereas electrons flow from CII to CIII not associated with CI. If this is so, depletion of CI in S cells would release CIII to reactivate electron transfer from CII. To test this, we used siRNA to deplete the CI subunit NDUFS3 from S cells. BNGEs revealed that depletion of CI released CIII from CI-SC (Fig. 1, C and D). This was accompanied by decreased electron flux through CI and CIII and increased flux through CII and CIII (Fig. 1E). CIII thus associates preferentially with CI in CoQ-containing supercomplexes (those containing CI and III or those with CI, III, and IV), establishing preferential electron flux from CI to CIII. To confirm this, we depleted CIII from control cells with siRNA (fig. S2A). Low CIII induced specific loss of supercomplex III+IV (fig. S2B) and a greater decrease in the combined activity of CII with CIII than that of CI with CIII (fig. S2C). Supercomplexes therefore define two functional CoQ populations: CoQ dedicated to transferring electrons originating from NADH (CoQNADH), which is trapped in SCs containing CI, and free CoQ in the inner mitochondrial membrane for use by CII and other enzymes that use FAD (CoQFAD) (Fig. 1F).

To investigate interaction between CIII and CIV, we screened for proteins present in SCs but not in free complexes (fig. S3, A and B). Mitochondrial samples prepared in the presence of digitonin, which preserves the integrity of RC and SC associations, were fractioned by BNGE, and individual bands were analyzed by proteomic methods. The screen identified cytochrome c oxidase subunit VIIa polypeptide 2-like (Cox7a2l) in the respirasome and in SC CIII+CIV, but not in free complexes III or IV (fig. S3). In parallel, we fortuitously discovered a Cox7a2l mutation in a screen of immortalized mouse fibroblasts derived from six littermates with mixed genetic background (C57BL/6 and 129Sv strains) (fig. S4A). Three cell lines had the reported pattern of SC interactions on immunoblots after one-dimensional BNGE (SC+ cells). However, the other three lines (SC cells) showed no association between CIII and CIV (fig. S4, A and B). A search for single-nucleotide polymorphisms (SNPs) differentiating SC+ and SC cells identified most in chromosome 17, which contains the Cox7a21 gene (fig. S4C). The SC lines were homozygous for a Cox7a2l version 6 base pairs shorter than that found in heterozygosis in SC+ cell lines (fig. S5A); these proteins encode proteins of 111 and 113 amino acids, respectively. Our proteomic analysis detected only the long isoform (fig. S3D). Immunoblotting detected Cox7a2l protein only in SC+ cell lines, indicating that the short isoform may be unstable (fig. S5B). Purebred C57BL/6 mice contained the short allele and 129sv mice the long one, both in homozygosis, and accordingly these mice exclusively showed the SC or SC+ phenotype, respectively (Fig. 2, A and B). Based on these findings, we propose that Cox7A2l be renamed supercomplex assembly factor I (SCAFI). The yeast proteins rcf1 and rcf2, previously proposed to be supercomplex III:IV assembly factors, also affect individual CIV assembly (1416).

Fig. 2 Characterization of Cox7a2l in supercomplex-competent (SC+) and SC cell lines and animals.

(See also fig. S4.) (A) Polymerase chain reaction (PCR) analysis of Cox7a2l alleles in C57Bl/6J (C57) and 129S2/SvPasCrlf (129) mice. (B) BNGE immunoblot analysis of CIII (anti-Core2) showing complex assembly status in samples from mixed-background littermates (178 and 179) and purebred C57 and 129 mice. (C) BNGE immunoblot of the indicated cell line probed for CIII (anti-Core2). (D) Immunoblot of fibroblast line 77 after second-dimension separation of BNGE-separated proteins by SDS-PAGE; blots were probed for CI (anti-NDUFA9), CII (anti-SDHA), CIII (anti-Core2), and CIV (anti-CoxIV). Supercomplexes formed in cells transfected with 113-amino-acid Cox7a2l are circled.

Mice derived from 129sv:C57BL/6 intercrosses showed normal supercomplex formation or no association of CIII with CIV (Fig. 2B), demonstrating tight correlation between absence of 113 amino acid SCAFI and failed CIII-CIV interaction. The long form of SCAFI and the SC+ phenotype were present in CBA, 129, NZB, and CD1 mice, whereas C57BL/6J and Balb/cJ mice were homozygous for the short form and had the SC phenotype (Fig. 3, A and B). SC cells were unaffected by overexpression of the short form, whereas overexpression of the long form restored the assembly of SCs containing CIV (Fig. 2, C and D).

Fig. 3 Cox7a2l/SCAFI depletion alters mitochondrial respiratory function.

(A) (Top) BNGE immunoblot of CIII (anti-Core2) in different mouse strains, showing complex assembly in liver mitochondria. (Bottom) Genotyping (PCR) of the Cox7a2l allele. (B) BNGE immunoblot of CIV (anti-NDUFA4), showing complex assembly in liver mitochondria. (C) Substrate-driven rates of oxygen consumption and ATP synthesis in liver mitochondria from Cox7a2l/SCAFI-deficient (Short) and Cox7a2l/SCAFI-expressing (Long) mice. (D) Plasticity model of mETC organization, showing CI associations with a dedicated CoQ pool [blue, the respirasome (CI+III+IV) and CI+III] coexisting with CIII+CIV associations and free CII, CIII, and CIV. (Left) Normal situation (CD1 mice), in which Cox7a2l/SCAFI modulates CIV-containing supercomplexes, thereby regulating the proportions of free CIII and CIV and generating three states for CIV. (Right) Extreme situation of C57Bl/6J mice, where SCAFI is absent and no CIV-containing supercomplexes form, making all CIV available to any substrate. (E) Time profile (t-t0) of fumarate production by isolated liver mitochondria from SCAFI-proficient and SCAFI-deficient mice in the presence of succinate or succinate and pyr+mal. Data in (C) and (E) are presented as box plots; marks and symbols same as for Fig. 1. *Significant differences by ANOVA. Paired differences were assessed by PLSD.

Respiration rates were higher in liver mitochondria lacking SCAFI, both when incubated with pyruvate and malate (pyr+mal) (for which electrons are carried by NADH) or succinate (for which electrons are carried by FAD) (Fig. 3C). Respiration was higher in C57BL/6J fibroblasts transformed with short-SCAFI than in those transfected with long-SCAFI (fig. S6, A and B). In contrast, CIV respiration driven by the specific electron donor tetramethylphenylenediamine (TMPD) was similar regardless of SCAFI expression. Thus, cells expressing functional SCAFI do not use the full potential of CIV when fed with glucose (fig. S6B). In permeabilized C57BL/6J-derived fibroblasts expressing short-SCAFI, adenosine triphosphate (ATP) production was near maximal with pyr+mal or succinate as substrates, whereas maximal respiration and ATP production in cells expressing long-SCAFI required substrates for both the NADH and the FAD routes (fig. S6C).

In the absence of SCAFI, no CIV will assemble in SCs, and electrons from NADH or FAD will pass through a single pool of cyt c to a single pool of CIV (Fig. 3D). In contrast, in the presence of SCAFI, a fraction of CIV interacts with CIII and two SCs would form: I+III+IV (the respirasome) and III+IV (Fig. 3D). Because CIV-containing SCs also contain cyt c (4, 17), this will define three CIV populations: one receiving electrons exclusively from NADH (CIVNADH), another from FAD-dependent enzymes (CIVFAD), and a third from NADH and FAD (Fig. 3D). Confirming this, simultaneous exposure of SCAFI+ mitochondria to succinate and pyr+mal had an additive effect on respiration and ATP synthesis, an effect not seen in the absence of CIII+CIV superassembly (Fig. 3C).

The ability of SCAFI to segment the mETC into three CIV populations may prevent saturation by one substrate, thereby promoting simultaneous oxidation of multiple substrates at optimum rates. To test this, we evaluated fumarate synthesis in SCAFI+ and SCAFI liver mitochondria supplied with succinate alone or with pyr+mal. With succinate alone, SCAFI mitochondria generated fumarate at higher rates than SCAFI+ mitochondria (Fig. 3E). Simultaneous addition of succinate and pyr+mal did not alter fumarate synthesis in SCAFI+ mitochondria but significantly reduced it in SCAFI mitochondria (Fig. 3E), demonstrating that SCAFI-mediated assembly of CIV into SCs minimizes competitive inhibition of respiration between pyruvate and succinate.

The ratio of NADH:FAD electrons feeding the mETC is higher when glucose is the respiratory substrate and lower for fatty-acid (FA) oxidation. Adjustment of mETC superassembly would provide a mechanism to ensure efficient oxidation of available substrates. To test this, we analyzed respiration capacity through the FAD and NADH routes in liver mitochondria from CD1 and C57Bl/6J mice fed ad libitum or fasted for 18 hours to activate FA degradation. In mitochondria from fed animals of both strains, the respiration rate was higher with succinate than with pyr+mal (Fig. 4, A and B). Simultaneous use of both substrates produced an additive effect on maximum respiration in CD1 (Fig. 4A) but not C57Bl/6J mitochondria (Fig. 4B). In mitochondria from fasted animals, respiration driven by pyr+mal was significantly lower in both strains, whereas succinate respiration was enhanced by fasting in CD1 but not C57Bl/6J mitochondria (Fig. 4, A and B). These results support the existence of separate electron routes and are consistent with the predicted shift to the FAD route upon switching to FA degradation.

Fig. 4 mETC activity and supercomplex distribution in response to fasting.

(A and B) (Top left) Oxygen-dependent (coupled) respiration in isolated liver mitochondria from fed and starved CD1 (A) or C57Bl/6 mice (B) driven by pyr+mal, succinate, or both. (Bottom left) Activities of mitochondrial complexes (CI and CII), individually and in combination with CIII (CI+III and CII+III), expressed as international units (IU) per milligram mitochondrial protein. (Right) BNGE immunoblots showing CIII associations (anti-Core2). CII (anti-SDHA) is shown on the same blot as a loading control. (C) Sites for mitochondrial generation of NADH+H+ or FADH2 from different substrates and processes potentially targeted by starvation. PDHK/D, PDH kinase/dehydrogenase; GDH, glutamate dehydrogenase. (D) Oxygen-dependent (coupled) respiration (left) and ATP production rate (right) driven by pyr+mal or glutamate in isolated liver mitochondria from fed and starved CD1 or C57Bl6 mice. Data in (A), (B) and (D) are presented as box plots; marks and symbols same as for Fig. 1. *Significant differences by ANOVA. Paired differences were assessed by PLSD.

In CD1 mitochondria, fasting reduced maximal CI and CI+CIII activities without influencing CII or CII+CIII (Fig. 4A), but this effect was not seen in C57Bl/6J mitochondria (Fig. 4B). Accordingly, fasting reduced the proportion of CIII assembled with CI (NADH route) only in CD1 mitochondria (Fig. 4, A and B). Pyruvate dehydrogenase (PDH) activity is down-regulated by phosphorylation by PDH kinase during starvation (Fig. 4C) (18). This may explain the lower respiration rate with pyr+mal in both C57Bl/6J and CD1 mitochondria, despite the fact that CI activity is down-regulated only in CD1 mice (Fig. 4, A and C). An alternative way to generate intramitochondrial NADH+H+ is to feed mitochondria with glutamate, which is converted to α-ketoglutarate by glutamate dehydrogenase, an enzyme not regulated by starvation (Fig. 4C). Because starvation reduced CI activity only in CD1 mice, glutamate-dependent respiration should be affected only in this strain. This was confirmed in liver mitochondria from starved CD1 animals, in which pyruvate- and glutamate-driven synthesis of ATP was decreased, whereas only pyruvate-dependent ATP synthesis was decreased in C57Bl/6J-derived organelles (Fig. 4D).

Our results provide genetic evidence for the existence of mitochondrial respiratory superassemblies, supporting the plasticity model of mETC organization and identifying dynamic supercomplex assembly as a mechanism through which cells might adapt to varying carbon sources and tailor the mETC to specific cell-type requirements. The lack of this mechanism in commonly used mouse strains has potential implications for studies of metabolic processes and is of interest for research into human diseases affecting mitochondrial function.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

Table S1

References (2227)


View Abstract

Stay Connected to Science

Navigate This Article