Low abundance of the matrix arm of complex I in mitochondria predicts longevity in mice

Mitochondrial function is an important determinant of the ageing process; however, the mitochondrial properties that enable longevity are not well understood. Here we show that optimal

assembly of mitochondrial complex I predicts longevity in mice. Using an unbiased high-coverage high-confidence approach, we demonstrate that electron transport chain proteins, especially

the matrix arm subunits of complex I, are decreased in young long-living mice, which is associated with improved complex I assembly, higher complex I-linked state 3 oxygen consumption rates

and decreased superoxide production, whereas the opposite is seen in old mice. Disruption of complex I assembly reduces oxidative metabolism with concomitant increase in mitochondrial

superoxide production. This is rescued by knockdown of the mitochondrial chaperone, prohibitin. Disrupted complex I assembly causes premature senescence in primary cells. We propose that

lower abundance of free catalytic complex I components supports complex I assembly, efficacy of substrate utilization and minimal ROS production, enabling enhanced longevity.

Mitochondria perform vital functions in life such as they provide energy (ATP), generate iron–sulphur clusters, participate in Ca2+ regulation and have major roles in fatty acid and

amino-acid metabolism. They also produce reactive oxygen species (ROS) as metabolic by-products that can damage cellular components and participate in cellular signalling pathways.

Mitochondria have long been regarded as crucial in the ageing process. This was originally associated with the (mitochondrial) oxidative stress theory of ageing1. However, a recent research

has revealed a more complex picture by uncoupling lifespan extension from a decrease in mitochondrial ROS generation or oxidative damage in multiple species2,3, and other relevant aspects

associated with mitochondrial function such as stress responses, apoptosis and cell senescence have also been attracting attention.

Mitochondrial biogenesis is an extremely complex process integrating the intra-mitochondrial synthesis of 13 electron transport chain (ETC) proteins coded by the mitochondrial DNA with far

more than 1,000 additional proteins that are synthesized in the cytosol as pre-proteins and imported into mitochondria through specific transporters. There are several assembly factors for

the ETC, and mitochondrial protein chaperones such as prohibitin (PhB) are thought to be involved in a dynamic balance in the composition of the mitochondriome that determines mitochondrial

function4. The dynamics of the mitochondriome5 is further determined by targeted turnover6.

We aimed to identify in an unbiased approach the proteome composition of mitochondria that might predict longevity in mammals. We reasoned that such a ‘longevity-enabling phenotype’ of

mitochondria should be (i) evident in mitochondria from long-living animals at young age (that is, before the death of the first member of the cohort); (ii) common in disparate longevity

models and tissues; and (iii) counteracted by ageing. Reduced abundance of ETC proteins and of matrix arm subunits of complex I in particular was found in young mice from long-living mouse

cohorts, which was associated with improved complex I assembly, higher state 3 oxygen consumption rates and decreased complex I-linked superoxide production, and was reversed in old mice.

Moreover, suppression of complex I assembly decreased mitochondrial oxidative metabolism, increased mitochondrial ROS production and precipitated premature senescence. We suggest that

partially assembled matrix arm subcomplexes of complex I, which contain the catalytic subunits, might be detrimental as they only contribute to ROS formation without facilitating electron

transport activity and proton pumping. Minimizing the abundance of free catalytic complex I components might be an essential feature of a longevity-enabling mitochondrial phenotype in

mammals.

To perform an unbiased analysis of mitochondrial protein composition as potential predictor of longevity, we analysed the mouse liver proteome at young age (100% cohort members alive) in two

long-lived mice models: the ICRFa strain, a C57Bl/6 substrain7 with significantly extended longevity under standard ad libitum (AL)-fed condition (P=0.017; Fig. 1a); and C57Bl/6 subjected

to dietary restriction (DR, 60% of AL food intake) from 3 months of age onwards. DR significantly extended lifespan beyond both AL C57Bl/6 (log-rank test, P=0.0073) and AL ICRFa mice

(P=0.0085; Fig. 1a). We addressed ageing-related changes by comparing young (7–8 months) and old (30–32 months) ICRFa mice. DR mice were significantly lighter than the AL C57Bl/6, AL ICRFa

and old ICRFa mice, whereas there were no differences in body weights among the non-DR mice groups (analysis of variance, P=0.05; Supplementary Fig. 1). Food intake was not different between

AL ICFRa and AL C7Bl/6 mice, suggesting that the extended lifespan in ICRFa mice is not caused by a concealed DR effect.

(a) Kaplan-Meier survival curves (right censored) of male C57Bl/6 (red) and ICRFa (green) mice under ad libitum (AL) feeding and of C57Bl/6 mice under 40% DR from an age of 3 months onwards

(blue). Group sizes were (censored events in brackets): C57Bl/6 AL 310 (172), C57Bl/6 DR 241 (157) and IRCFa 2391 (1,061). (b) State 3 oxygen consumption rates of purified liver mitochondria

from C57Bl/6 mice under AL and DR, respectively, with pyruvate/malate (top) and succinate (bottom) as substrate. Data are mean±s.e.m., n=4 animals per group. (c) H2O2 release from purified

liver mitochondria from C57Bl/6 mice under AL and DR, respectively. Data are mean±s.e.m., n=4. Top: pyruvate/malate plus rotenone, indicating maximum ROS production from complex I (CI).

Bottom: succinate plus rotenone, indicating ROS production from sites other than complex I. (d) Numbers of confidently changed proteins in all comparisons. (e) Molecular mass distributions

for differentially enriched proteins under DR (blue), in young ICFRa (green) and in old ICRFa (pink).

We studied liver mitochondrial function in C57Bl/6 during ageing in both AL- and DR-fed mice. Oxygen consumption rates under phosphorylating condition (state 3) using either a complex

I-linked substrate (pyruvate plus malate) or a complex II-linked substrate (succinate; Fig. 1b) progressively declined with age in AL mice but were maintained under DR up to an age of 37

months, that is, beyond the maximum lifespan of the AL mice. Hydrogen peroxide release from complex I was higher under AL than under DR up to an age of 30 months. However, DR only postponed

the age-dependent increase and did not prevent it: DR mice at 37 months of age (that is, at equivalent survival rate to 30-month-old AL mice) showed hydrogen peroxide release from complex I

as high as AL mice at 30 months (Fig. 1c). There were only minor differences in hydrogen peroxide release from sites other than complex I (Fig. 1c). In a comparison between ICRFa and C57Bl/6

mice, longer lifespan in ICRFa was also associated with lower hydrogen peroxide release from complex I in isolated liver and brain mitochondria (Supplementary Fig. 2).

Proteomic analysis was performed in two separate 6-plex tandem mass tag (TMT) labelling experiments using one-pooled sample in each for cross-standardization. After imposing a 1% false

discovery rate (FDR) filter, we identified 7,047 and 6,704 peptides in the two experiments, respectively, and quantified a total of 631 proteins (531 and 497 proteins per experiment)

(Supplementary Data 1–3). This represents about half of the mouse mitochondrial proteins in the MitoCarta database and is similar to published data4. The molecular weight distribution of the

identified proteins mirrored that of all mitochondrial proteins according to the MitoCarta and MitoMiner8 databases (Supplementary Fig. 3). Our analysis quantified 57 out of the 96 known

mouse ETC proteins, namely 67% of complex I proteins, 50% of complex II, 63% of complex III, 46% of complex IV and 76% of complex V. One-hundred seventy-seven proteins showed confidently

different abundances in at least one of the three comparisons performed (Supplementary Data 1–3). The number of differentially abundant proteins was the largest (131 proteins) in the AL–DR

comparison in C57Bl/6 mice, whereas there were only 57 and 52 differentially abundant proteins for the young C57Bl6–ICRFa and the young–old ICRFa comparisons, respectively (Fig. 1d). In

general, proteins that were less abundant in the longevity models (and more abundant in old mice livers) tended to be smaller than those that were more frequent in long-living mice (Fig.

1e).

Out of the 13 proteins that were different in abundance in all three comparisons (Fig. 1d), 12 were possibly implicated with longevity because they were less abundant in both longevity

models. Moreover, 11 of them were increased in old mice (Supplementary Table 1), together suggesting that low abundance of these proteins at a relatively young age by either genetic or

environmental manipulation might predispose to longevity. These proteins included both PhB and PhB 2, and the majority (8 out of 13) were components of the ETC (complexes I, III and V;

Supplementary Table 1).

To identify the functional traits of liver mitochondria associated with longevity, we performed mitochondria-specific KEGG pathway enrichment analysis based on the proteins that were more

abundant in liver mitochondria from young C57Bl/6 mice under DR in comparison with AL. A wide range of diverse metabolic pathways, notably those associated with the Krebs cycle and fatty

acid and amino-acid metabolism, was enriched under DR (Supplementary Table 2). However, few of these pathways were similarly enriched in young long-living ICRFa mice or were depleted in old

mice. Not a single pathway changed consistently in all three comparisons (Supplementary Table 2), indicating that most of the metabolic pathways that are more abundant under DR are not

general longevity-assurance mechanisms. In contrast, there were only nine KEGG pathways significantly enriched in an analysis based on proteins less abundant in DR (Supplementary Table 3).

Importantly, the four highest ranking of these were also significantly enriched in an analysis based on less abundant proteins in young ICRFa mice (Supplementary Table 3) and equally so in

an analysis based on proteins more abundant in old animals (Supplementary Table 3). These four pathways (oxidative phosphorylation, Parkinson’s disease, Huntington’s disease and Alzheimer’s

disease) include high proportions of ETC proteins, suggesting that multiple ETC proteins within liver mitochondria are reduced at young age in a longevity phenotype but increased with age.

As we were able to quantify the majority of ETC proteins, we examined the changes of the individual ETC subunits in all three comparisons in detail (Fig. 2a–c). The abundance of numerous

subunits of complexes I, III, IV and V (but not of complex II) was reduced in both long-lived mouse models, which was generally reversed in old mice (Fig. 2a–c). There was no evidence for an

induction of a mitochondrial unfolded protein response in the long-lived mice with low ETC abundance: several mitochondrial heat-shock proteins (HSPD1, HSPA9 and DSPE1) and the Clpp1

protease were not changed in any of comparisons, whereas heat-shock proteins HSPA5 and HSP90D1 were decreased under DR (Supplementary Data 1–3).

Heatmaps show changes in abundance of all ETC proteins in the comparisons young DR versus young AL C57Bl/6 (a), young ICRFa versus young C57Bl/6 (b) and old versus young ICRFa (c). CI and CV

reach out from the inner membrane into the matrix space (upward). Grey symbols represent proteins not identified; and black/white symbols represent abundance that is not significantly

changed. Colours indicate fold change.

Strikingly there was a highly skewed pattern of protein abundance changes within complex I. Reduced abundance of complex I subunits in both long-living mice models occurred almost

exclusively in subunits localized in the matrix arm while membrane arm subunits were hardly affected (Fig. 2a,b). This pattern was partially reversed in the aged mice (Fig. 2c). The

abundances of PhB and PhB 2, which function as chaperones for ETC proteins and/or as structural scaffolds, paralleled the changes in matrix arm subunits with strong statistical confidence

and large fold differences (Supplementary Table 1).

We confirmed these observations in a series of SDS–PAGE western blot experiments, by measuring the ratios of selected matrix subunits (NDUFS3 and NDUFV2) and PhB to the membrane subunits

NDUFB9 and NDUFA9, which is located at the junction between membrane and matrix arm and is involved in stabilizing this junction9 (Fig. 3a). In agreement with the proteomics results, we

found a decreased ratio of matrix subunits and PhB to membrane subunits under DR (Fig. 3b) and in liver mitochondria from young long-lived ICRFa mice (Fig. 3c), which was reversed in aged

liver mitochondria (Fig. 3d).