Abstract Genomic, transcriptomic, and proteomic approaches have been employed to gain insight into molecular underpinnings of aging in laboratory animals and in humans. However, protein function in biological systems is under complex regulation and includes factors in addition to abundance levels, such as modifications, localization, conformation, and protein-protein interactions. We have applied new quantitative chemical cross-linking technologies to uncover changes in the muscle mitochondrial interactome contributing to functional decline in aging in female mice. Statistically significant age-related changes in protein cross-links relating to assembly of electron transport system complexes I and IV, activity of glutamate dehydrogenase, and coenzyme-A binding in fatty acid beta-oxidation and TCA enzymes were observed. These changes showed remarkable correlation with measured CI based respiration differences within the same young-old animal pairs, indicating these cross-link levels offer new molecular insight on commonly observed age-related phenotypic differences. Each observed cross-link can serve as a protein conformational or protein-protein interaction probe in future studies making this dataset a unique resource for many additional in-depth molecular studies that are needed to better understand complex molecular changes that occur with aging. Keywords: mitochondria, protein-protein interactions, protein conformations, protein-ligand interactions, quantitative cross-linking mass spectrometry __________________________________________________________________ Aging is a complex process involving several interconnected features that contribute to the progressive decline in function, vulnerability to chronic disease, and ultimately death.^[36]1 Among the hallmarks of aging is mitochondrial dysfunction, first proposed as a major component of aging in 1956.^[37]2 In muscle, aging is accompanied by declines in mass and strength. Decreases in mitochondrial function are thought to be primary mediators of age-related muscle loss.^[38]3 Many phenotypes of aging have been observed in mitochondria including changes in reactive oxygen species (ROS) production, electron transport system (ETS) efficiency and respiration, ATP production, mitochondrial quality control, mitochondrial biogenesis, and mitophagy.^[39]4 Mitochondrial function is among the most significant changes accompanying muscle aging on the cellular level.^[40]3,[41]5,[42]6 Muscle mitochondria have been a primary focus of aging research due to their central role in maintaining metabolic and redox homeostasis, regulating metabolite levels, meeting energy demand during exercise, and the relative ease of muscle biopsy in humans compared to other tissues. Large-scale approaches have been employed to study muscle aging in laboratory animals and humans, including deep quantitative proteomic and transcriptomic profiling of several age groups in humans^[43]7,[44]8 and mice.^[45]9-[46]11 However, protein function is maintained under complex biological regulation and includes factors in addition to abundance levels, such as modifications, localization, conformations, and protein-protein interactions (PPIs). While large-scale studies have been applied to investigate differential mitochondrial protein modifications with age,^[47]12-[48]14 quantitation of large-scale changes in protein conformations and PPIs, collectively referred to here as the interactome, have previously not been possible. Recently developed quantitative cross-linking mass spectrometry technologies (qXL-MS) were applied to elucidate interactome changes in aged female murine skeletal muscle mitochondria contributing to age-related mitochondrial functional decline. Reported here are results from initial investigations of female murine muscle mitochondrial interactomes that enable identification of statistically significant changes associated with aging. Recently advanced isobaric quantitative protein interaction reporter (iqPIR) technologies enabled reproducible detection of age-related mitochondrial interactome changes.^[49]15 Muscle mitochondria from young and old mice were isolated and cross-linked with iqPIR molecules ([50]Fig. 1). Young and old mitochondrial samples were paired, processed, and analyzed to quantify age-related mitochondrial interactome changes. Before cross-linking, mitochondrial protein yield was measured together with functional measurements such as oxygen consumption rates on Complex I (CI) and Complex II (CII) substrates and citrate synthase (CS) activity. This allowed the initial correlation of age-related mitochondrial phenotypic or functional changes with molecular level interactome remodeling. In addition to changes in PPIs, quantifying site-specific interaction of the iqPIR reporter molecules provides new insights into protein activity by identifying changes in 1) protein structure associated with activity, as in glutamate dehydrogenase, described below and 2) substrate binding to protein active sites. Among these data, significant age-related decreases in cross-link levels within the antenna domain of glutamate dehydrogenase (DHE3) were observed that correlated with decreased glutamate and malate driven respiration. Similarly, CI late-stage assembly and binding of cytochrome c oxidase subunit NDUFA4 (NDUA4) subunit to complex IV (CIV) was impaired and correlated with decreased CI-linked respiration. Traditional methods, such as blue native polyacrylamide gel electrophoresis (BN-PAGE) can distinguish large assemblies but lack resolution to provide quantitative differences between late stage assemblies.^[51]16 Moreover, BN-PAGE can only enable visualization of complexes that survive extraction and interactions of ETS complex subunits like NDUA4 appear highly dependent on extraction conditions. Thus, qXL-MS is uniquely suited to study changes in complex assembly and composition and provide biological insight on ETS dysregulation observed with aging. Finally, as previously shown with qXL-MS data,^[52]17 each identified link can be targeted with parallel reaction monitoring (PRM) methods in other labs to visualize conformational and interactome changes with many other perturbations or interventions. Therefore, in addition to biological insight on large-scale age-related protein conformation and interaction changes discussed below, these data can serve as a resource for many additional studies to better visualize molecular changes underpinning age-related mitochondrial functional decline. Figure 1. Experimental workflow. [53]Figure 1. [54]Open in a new tab Gastrocnemius muscle was excised from 4 young (6 months) and 4 old (30 months) female mice and mitochondria were isolated. Each mitochondrial pellet (n = 8) was resuspended, and part of the homogenate was used to measure oxygen consumption for CI: glutamate and malate (G/M), CII: succinate and rotenone for inhibiting CI (S/R), and CI&CII: succinate, glutamate, and malate (S/G/M). Mitochondria from the same homogenate were then cross-linked with binary iqPIR reagents: mitochondria from old mice (n = 4) were crosslinked with reporter heavy (RH) and mitochondria from young mice (n = 4) were crosslinked with stump heavy (SH) iqPIR molecules. Mitochondria were then lysed, proteins were reduced, alkylated, and mixed in a 1:1 ratio based on total protein mass for each young old mouse pair and digested with trypsin overnight. Cross-linked peptides were by strong cation exchange and avidin capture of biotin tag on the cross-linker. Peptides were then separated by LC and MS2 spectra were collected for peptides with charge greater or equal to 4. The data were processed, and abundance of each cross-linked peptide pair was determined using newly developed iqPIR informatics. The dataset was uploaded to XLinkDB to view cross-linked peptides, quantitation, protein and complex structures and networks among other dataset features. Results Generation of mitochondrial interactome of aged muscle Samples from four old (30 months) and four young (6 months) female mice were cross-linked and then paired into four direct comparisons of individual old and young samples (one young and one old mouse in each pair designated P1, P2, P3, P4) for mass spectrometry analysis. The pairings were assigned based on mitochondrial protein content of each sample to maximize protein amounts. In total, 1864 cross-linked peptide pairs, hereafter referred to as cross-links, were identified at 1% cross-link level FDR from these four pairs. 533 cross-links are interprotein (formed by lysine residues originating from two distinct proteins) and 1331 are intraprotein (from the same protein). Mapping all identified intralinks on to recently predicted structures^[55]18 shows that 931 (89%), the overwhelming majority of intralinks, agree with the models: Euclidean distances between alpha carbons of cross-linked lysine residues are less than or equal to 35 angstroms ([56]Extended Data Fig. 1a). Interlinks are formed between two proximal proteins, so it is expected that these two proteins are localized together within mitochondria. 80% (393) of identified interlinks are between inner membrane associated proteins and only two interlinks are between proteins that are not expected to colocalize (matrix and outer membrane) based on submitochondrial localization information from Mitocarta 3.0 with each pair of interlinked proteins ([57]Fig. 2a).^[58]19 Mapping of intralinks shows that most intralinks are from matrix and inner membrane proteins. Overrepresentation of interlinks from inner membrane proteins is most likely due to many multisubunit protein complexes associated with the inner membrane ([59]Fig. 2a). Figure 2. Quantitative cross-linking enables detection of reproducible changes in the interactome of aging mitochondria. [60]Figure 2. [61]Open in a new tab a. Sub-mitochondrial localization of protein pairs identified in interprotein links (left) and proteins with intralinks (right). b. Distributions of log[2] ratios (old/young) of confidently quantified XLs for each biological replicate. Pairwise correlation plots (c) of confidently quantified cross-links (no missing values and 95% confidence interval < 0.5 for each ratio) with Pearson’s R values between 0.54 and 0.81 and heat map (d) show reliable and reproducible quantitation of cross-links. e. Confidently quantified cross-links (no more than 2 missing values and 95% confidence interval < 0.5 for each ratio) with significant changes (Bonferroni corrected p-value < 0.05 from two-sided t-test). f. KEGG and GO Pathways enriched in the cross-links with statistically significant changes (FDR corrected p-value < 0.01) arranged by strength: log[10](observed / expected). This measure describes how large the enrichment effect is. It is the ratio between the number of proteins in the network that are annotated with a term and the number of proteins that we expect to be annotated with this term in a random network of the same size. g. Protein yield from mitochondrial enrichment of both gastrocnemius muscles is decreased with age and (h) CS activity per mg protein in the mitochondrial enrichments is increased with age. Each rep of young and old mitochondria is denoted by a number next to the data point. i. Maximum oxidative phosphorylation (OXPHOS) capacity of the ETS from isolated mitochondria with CI substrates (glutamate and malate), CII substrates (succinate and rotenone), or combined CI&CII substrates (glutamate, malate, and succinate) measured as oxygen consumption rate (OCR) at saturating ADP concentrations normalized to units of CS activity. Statistical significance in g-i determined by unpaired two-tailed student’s t-test. Mean ± SD. Exact p-values from two-sided Student’s t-test are shown for comparisons that are statistically significant. Median normalized log[2] ratios of confidently quantified crosslinks (no more than 1 missing value across biological replicates and 95% confidence < 0.5) in aged mitochondria compared to young, follow normal distribution, except biological replicate 4 ([62]Fig. 2b). On average, at least 7 ions were used to derive each log2ratio ([63]Extended Data Fig. 1b) increasing confidence in each cross-link quantitation. To produce a log[2] ratio for a given cross-link, it must be present in both channels in a pair: reporter heavy (RH) cross-link in old sample and stump heavy (SH) cross-link in young. On average, 1156 cross-links are quantified in each replicate (1229, 1099, 1001, 1297 respectively), meaning that the majority of the cross-links were formed in both samples, making it unlikely to form by chance. Quantitation with iqPIR technologies and informatics showed excellent reproducibility based on observed pairwise Pearson’s R values between 0.5 and 0.76 ([64]Fig. 2c). Redundancy in cross-link quantitation exists in some cases because multiple peptide sequences with redundant linkage can be formed during sample processing due to trypsin missed cleavage events during digestion or from methionine oxidation. These cases offer internal quality control on quantitation. Log[2] ratios for such multiple cross-links generally show excellent agreement ([65]Extended Data Fig. 1c) and further increase confidence in quantitation for a residue pair. Cross-links quantified in every sample show remarkable agreement across the four biological replicates ([66]Fig. 2d). Reproducibility and robustness of quantitative values produced by the iqPIR method enabled identification of cross-links with statistically significant changes (Bonferroni corrected p≤ 0.05) in aging mitochondria ([67]Fig. 2e). There is no submitochondrial enrichment of significantly changed links compared to all cross-links ([68]Extended Data Fig. 1d). Analysis of KEGG pathways and GO biological processes show enrichment in proteins involved in glutamate metabolism, tricarboxylic acid (TCA) cycle, and oxidative phosphorylation ([69]Fig. 2f and [70]Extended Data Fig. 1e). As expected, total mitochondrial protein recovered from each gastrocnemius following isolation was lower in aged samples due to muscle atrophy and decreased input ([71]Fig. 2g). However, CS activity expressed per mg protein was significantly increased following mitochondrial enrichment in aged samples ([72]Fig. 2h). This finding is consistently reported for isolated muscle mitochondria in the literature, and CS activity is a better metric of mitochondrial content when comparing across ages than mitochondrial protein content in enriched fractions.^[73]20Respiration rates in aged mitochondria are significantly decreased when expressed per CS activity ([74]Fig. 2i, [75]Extended Data 1f). Complex I and Complex IV integrity are impaired with age The present mitochondrial interactome studies resulted in identification of many cross-links originating from ETS complexes and supercomplexes (SC) or respirasome interlinks. Respiratory electron transport and CI biogenesis have been reported as the top pathways affected in aging muscle on a transcriptome level but also the pathways that have the lowest correlation between transcript and protein levels, making the interpretation of its role in aging muscle more complicated.^[76]21 Decreased NAD^+/NADH is a hallmark of cell senescence and aging in muscle tissues and is driven in part by CI activity.^[77]22 CI consists of a membrane embedded part and protruding matrix arm that each assemble independently.^[78]23 In the matrix arm there are many interlinks as well as intralinks that are either unchanged or slightly increased in aged mitochondria consistent with previous reports of elevated complex I subunit expression in aged mouse skeletal muscle.^[79]24 Conversely, the only matrix arm cross-links with age-related decrease are between NADH-ubiquinone oxidoreductase 75 kDa subunit (NDUS1) and NADH dehydrogenase [ubiquinone] flavoprotein 1 (NDUV1) subunits ([80]Fig. 3a, [81]b), yet all intralinks in each subunit are either unchanged or increased ([82]Fig. 3c). Comparison of log[2] ratios of NDUS1 – NDUV1 cross-links with log[2] ratios of all other cross-links involving these proteins, including intralinks from both proteins and their interlinks to other CI subunits, revealed a statistical difference (p-value = 9.6*10^−6, Welch two-sample t-test excluding P1 from comparison and 1.35*10^−5 with all 4 replicates, [83]Fig. 3d, [84]Extended Data 2c). NDUS1 and NDUV1 interlinks with other subunits in the matrix arm and intralinks in all CI subunits except for NADH dehydrogenase [ubiquinone] 1 alpha subcomplex subunit 8 (NDUA8) also do not change or show slight increase with age ([85]Extended Data Fig. 2a, [86]b). Two residues on NDUS1 (K84 and K87) were identified cross-linked to the same residue on NDUV1 (K81). One possible explanation for this observation could involve change in post translational modification (PTM) levels at NDUV1 K81 since it is involved in both links with decreased levels. NDUV1 is a target for desuccinylation by NAD-dependent protein deacetylase sirtuin-5 (SIRT5) and deacetylation by NAD-dependent protein deacetylase sirtuin-3 (SIRT3).^[87]25,[88]26 However, if modification levels at a particular lysine were altered, one would expect that all quantified cross-links involving this residue will change accordingly, indicating a more or less accessible lysine due to the PTM. NDUV1 intra-link (K81-K104) shows no age-related changes, so increased modification of NDUV1 K81 cannot explain the decreased NDUS1-NDUV1 interlinks discussed above. NDUS1 and NDUV1 are among the last subunits to be incorporated in the complex and thus, this interaction is an indicator of fully assembled CI ([89]Fig. 3e).^[90]27 The cross-links are also close to flavin mononucleotide (FMN) binding which is a primary site of CI ROS production, making these subunits especially vulnerable to ROS damage.^[91]28 Figure 3. Assembly of Complex I and Complex IV integrity is affected in aging muscle. [92]Figure 3. [93]Open in a new tab a. Decreased interprotein cross-links between Nduv1 and Ndus1 mapped to CI structure (PDB 6G72) are shown in blue and non-changing intralinks are shown in black. b. Heatmap of log[2] ratios of NDUS1-NDUV1 cross-linked peptide pairs for each biological replicate (P1, P2, P3, P4). c. Heatmap of log[2] ratios of intraprotein cross-linked peptides from NDUS1 and NDUV1. d. Boxplots of cross-linked peptide pairs in biological replicates P2, P3, P4 (NDUS1-NDUV1 XLs n = 14, other links n = 60; visualized as median and 25^th and 75^th percentiles, with whiskers indicating minima and maxima) with Welch’s two sided t-test show statistically significant difference between NDUS1-NDUV1 interlinks and intralinks and interlinks to other CI subunits. e. N-module is assembled from subcomplexes NDUV1-NDUV2 and NDUS1-NDUA2 at the end of whole CI assembly. It is frequently replaced due to ROS damage. f. NDUA4-CX6B1 interlinks mapped to a CIV structure (PDB 5Z62). g. Heatmap of log[2] ratios of cross-linked peptide pairs between NDUA4 and CX6B1 (top) and CX6B1 intralinks (bottom). h. Boxplots of cross-linked peptide pairs in biological replicates P2, P3, P4 (CX6B1-NDUA4 XLs n = 16, other links n = 9; visualized as median and 25^th and 75^th percentiles, with whiskers indicating minima and maxima) with Welch’s two-sided t-test show statistically significant difference between NDUA4-CX6B1 interlinks and Cox6b1 intralinks. i. Log[2] fold change of CI linked oxidative phosphorylation in old samples compared to young for each pair. Mean ± SD. j. Correlation plots between average log[2] ratios of CI (left) or CIV (right) cross-links changing with age and log[2] fold change in CI respiration for each pair with multiple R^2 shown. A recently reported CI salvage pathway to achieve efficient maintenance of CI function is accomplished through replacement of damaged subunits instead of assembling the whole complex from scratch.^[94]29 Differential effects of aging on protein abundance and stability of matrix and membrane proteins of CI have been previously reported and are expected to lead to impaired assembly.^[95]24,[96]30 Taken together, these data indicate impaired assembly or elevated turnover of N-module in old mitochondria. Using in-gel activity assays, we find a significant effect of age on CI activity ([97]Extended Data Fig. 3a, [98]b). Defects in CI assembly have been shown to lead to increased production of superoxide and premature senescence, and lower abundance of matrix subunits can be a predictor of longevity. There have also been recent reports of coordinated assembly of ETS complexes. In particular, Complex III (CIII) was shown to mediate CI assembly.^[99]31 CIII deficiency led to stalling of CI assembly, especially incorporation of the N-module. Interlinks between CI and CIII as well as several CIII subunit intralinks are also decreased in aged mitochondria ([100]Extended Data Fig. 2d, [101]e). Interlinks between cytochrome c oxidase subunit 6B1 (CX6B1) and NDUA4 of CIV were among the cross-links that exhibited the largest age-related level decreases ([102]Fig. 3f, [103]g). NDUA4 has been identified to be a small subunit of CIV rather than CI as previously thought.^[104]32 NDUA4 is not required for CIV assembly and CIV is functional without it, but loss of NDUA4 impairs CIV activity.^[105]33 No NDUA4 intralinks were quantified in this study, but comparing NDUA4-CX6B1 interlinks to CX6B1 intralinks revealed statistically significant decreases of interprotein link levels, p-value = 5.9*10^−6 excluding P1 from comparison and 0.03 with all 4 replicates ([106]Fig. 3h and [107]Extended Data 2f) indicating reduced interaction between these subunits rather than reduced CIV levels. Recently, structural characterization of CIV containing NDUA4 subunits has been possible by judicious choice of complex extraction/purification conditions^[108]33 revealing NDUA4 resides at the CIV homodimer interface and precludes CIV homodimer formation.^[109]32 Balsa et al. showed that stable knock down of NDUA4 reduced both the activity and stability of CIV that could be rescued by myc-NDUA4 expression.^[110]32 Therefore, the observed reduction of NDUA4-CIV interaction indicated by multiple reduced NDUA4-CX6B1 cross-link levels would be expected to decrease CIV stability and activity in mitochondria from old mice. Using in-gel activity assays, we find a significant effect of age on CIV activity ([111]Extended Data Fig. 3c, [112]d). We have also recently reported decrease of NDUA4-CX6B1 interaction in a mouse model of heart failure, suggesting that similar mechanisms might be at play.^[113]34 A more precise role of NDUA4 and its effects on CIV is not yet clear and is a subject of ongoing studies.^[114]35,[115]36 Increases in heterogeneity and differential rates of aging are well established including in C57BL6/J mice.^[116]37,[117]38 For both CI and CIV links discussed previously, biological replicate one (P1) deviates from other replicates, showing little or no age-related change in these cross-links ([118]Fig. 3b, [119]g). Overall, CI-linked respiration declined significantly in aged samples ([120]Fig. 2g). Notably however, the aged sample from P1 had no apparent decline in CI-linked respiration compared to young control, which coincides with this pair demonstrating minimal effects of age on CI protein interactions ([121]Fig. 3i). Intriguingly, the magnitude of change in CI assembly cross-links and CX6B1-NDUA4 cross-links in CIV showed strong correlation with decline in respiration on CI substrates across all sample pairs: Pearson’s R^2 0.63 and 0.95 respectively, while showing no correlation with CII respiration ([122]Ext. data table 1, [123]Fig. 3j, [124]Extended Data 2g). DHE3 cross-links associated with activation are decreased DHE3 catalyzes interconversion of glutamate and alpha-ketoglutarate (α-KG) and is encoded by the glud1 gene. A primary DHE3 in vivo function involves catalysis of oxidative deamination of glutamate to produce ammonia and α-KG,^[125]39 a TCA cycle intermediate. α-KG is also involved in regulation of many cellular processes outside of the TCA cycle, such as epigenetic regulation, protein hydroxylation and ATP synthase regulation ([126]Fig. 4a). Connection between DHE3 glycation levels to liver aging has been reported before, but DHE3 abundance levels have not previously been correlated with aging in muscle.^[127]40 In agreement with that notion, two DHE3 intralink levels were quantified that were unchanged in any of the biological replicates, indicating that DHE3 protein levels were unaltered with age. DHE3 exists as a hexamer, comprising a dimer of trimers, and is a subject of intricate and diverse regulatory mechanisms.^[128]41 Each trimer forms a protruding structure where helices from all three subunits form an “antenna” with largely unknown function ([129]Fig. 4b). This antenna is only present in higher organisms and coevolved with the complex regulatory network of DHE3, suggesting the antenna may serve in a regulatory capacity.^[130]42 Decreased homodimeric links in the DHE3 antenna region were also among the largest decreased age-related changes quantified in the present study ([131]Extended Data Fig. 4a). These included multiple cross-linked peptide pairs arising from missed cleavage products and were observed in all biological replicates, with more moderated changes in P1 where functional decreases were also moderated ([132]Fig. 4c,[133]d, [134]Extended Data 4b). Upon substrate binding, DHE3 forms an abortive complex ((DHE3*NAD(P)H*Glu) and its release is facilitated by ADP.^[135]41 Since ADP serves as an activator of DHE3, quantitative cross-linking experiments were also performed with mitochondria isolated from HEK293 cells comparing ADP-treated and control untreated mitochondria. These experiments revealed strong increases in DHE3 antenna homodimer links in both biological replicates ([136]Fig. 4e). The combination of ADP-stimulation with old/young interactome data suggests the possibility that DHE3 activity is repressed in aged muscle mitochondria. If so, this may contribute to the observed reduced malate and glutamate stimulated respiration in mitochondria from aged muscle ([137]Fig. 2i). DHE3 is essential for delivery of NADH to CI during glutamate/malate stimulated respiration. The decreases of antenna cross-links correlate with change in CI respiration and show no correlation with respiration on CII ([138]Fig. 4f, [139]Extended Data 4c). The total ability to fuel respiration using glutamate alone is also decreased with age across a range of glutamate concentrations, and a shift in the kinetic response to glutamate stimulation ([140]Extended Data 4d, [141]e). Figure 4. Cross-link levels associated with glutamate dehydrogenase (DHE3) activation are decreased in aged mitochondria. [142]Figure 4. [143]Open in a new tab a. DHE3 converts glutamate and NAD^+ to NADH and α-KG, a TCA cycle intermediate that is involved in many cellular processes. ADP is a potent DHE3 activator while NADH and GTP inhibit it. b. DHE3 cross-links quantified in aging mouse mitochondria; non-changing intralinks are shown in black and decreasing links in the “antenna” (K477-K477, K477-K480, K480-K480) mapped to one of the trimers in the hexamer (bovine structure 6DHM) are shown in blue. c. Log[2](old/young) ratio for each cross-linked peptide pair in each biological replicate is summarized in heatmap. d. Boxplots of “antenna” ratios and other intralink ratios (antenna XLs n = 19, other links n = 9; visualized as median and 25^th and 75^th percentiles, with whiskers indicating minima and maxima) with Welch two-sided t-test p-value displayed. e. Heatmap of cross-linked peptide pairs quantified in 2 biological replicates of ADP treated HEK293 mitochondria. f. Correlation between average log[2] ratio for DHE3 antenna cross-links and CI respiration with multiple R^2 shown. FAO and TCA enzymes show less accessible substrate sites. Impairment of fatty acid metabolism with aging has been shown in multiple organs and models. Aging mouse heart has a decreased free fatty acid flux, TCA cycle flux, and insulin stimulated anaplerosis.^[144]43 Levels of free fatty acids in blood plasma are decreased with age while triglyceride levels are increased.^[145]44 In addition, muscle contraction leads to a shift in fatty acid oxidation (FAO) and TCA cycle substrate flux and muscle recovery from contraction is impaired with age.^[146]45 FAO and TCA cycle substrates and intermediates show strikingly different patterns in old mice after the unloading and following recovery compared to young mice. Surprisingly though, transcript levels of the proteins involved do not show significant differences, confounding understanding of the mechanisms underlying the changed FAO and TCA cycle fluxes. Many cross-linked peptides from several FAO and TCA enzymes were quantified in this study, including very long-chain specific Acyl-CoA dehydrogenase (ACADV), acetyl-CoA acetyltransferase (THIL), succinate-CoA ligase (ADP/GDP-forming) subunit alpha (SUCA) and succinate-CoA ligase (ADP-forming) subunit beta (SUCB). ACADV, encoded by gene acadvl, catalyzes the first step in beta oxidation. Although the majority of ACADV intralinks showed a slight increase in aged mitochondria, indicating possible slightly elevated protein levels, a subset of four links were significantly decreased in aged muscle mitochondria ([147]Fig. 5a,[148]g, [149]Extended Data 5c). All decreased ACADV links span the binding pocket of fatty acyl-CoA and involve residue K279. The ACADV structure PDB: 2IX5 which contains CoA illustrates that CoA phosphate groups reside within salt bridge formation distance from K279 ([150]Fig. 5b) suggesting that binding of CoA would reduce link formation at this residue. A similar situation was observed with a subset of 8 cross-link levels (out of 18) quantified in THIL, encoded by acat1, which catalyzes the final FAO step, that showed significant age-related decrease ([151]Fig. 5c, [152]h). These link level changes contrast with the remaining 10 THIL cross-links that either slightly increased or showed no change with age. Of the 8 THIL cross-links that decreased with age, 2 span the CoA binding site and K260 which is also within salt bridge distance with CoA phosphate groups as shown ([153]Fig. 5d). Ligand binding can affect cross-link levels within mitochondria as we demonstrated above; both ACADV and THIL age-related decreased links appear statistically enriched in regions involved in ligand binding indicating age-related differences in FAO exist, despite no significant changes in enzyme levels. THIL functions as a tetramer and tetramer formation is linked to higher activity and cancer progression.^[154]46 We observed decrease in the homodimeric links at the tetramer interface indicating decreased levels of active tetramer. Figure 5. Tricarboxylic acid cycle and fatty acid beta oxidation. [155]Figure 5. [156]Open in a new tab a. Heatmap of log2 ratios of all ACADV cross-linked peptide pairs. b. ACADV cross-link at the CoA binding site mapped to A.Thaliana structure, a short chain specific acyl-CoA oxidase in complex with acetoacetyl-CoA (2IX5, left) and all cross-links mapped on a human structure with cross-links at CoA binding site in the box (2UXW, right). CoA from acetoacetyl-CoA is within the distance to form hydrogen bonds with side-chain nitrogen of K279. c. Heatmap of log2 ratios of THIL cross-linked peptide pairs. d. THIL cross-links mapped to a human structure (2IB8) with zoomed in tetrameric links (left). Cross-links at the CoA binding site were also mapped on a human structure crystalized with CoA (2F2S) showing that K260 is within the salt bridge bond formation distance (right zoomed in panel). e. Heatmap of succinyl-CoA ligase cross-links. f. Succinyl-CoA ligase cross-links mapped on a pig structure of GTP specific succinyl-CoA (4XX0) crystalized with inset view of CoA proximity to SUCA K90 (left). Succinyl-CoA ligase cross-links spanning ATP/GTP binding site mapped to a pig GTP specific structure (2FP4). g, h, i. Boxplots comparing decreased cross-link levels at specific sites to log2 ratios of all other intralinks for ACADV (CoA proximal XLs n = 10, other links n = 36), THIL (CoA proximal XLs n = 10, tetramer XLs = 24, other links n = 36), and SUCA/SUCB1 (CoA proximal XLs n = 8, ADP proximal XLs n = 19, other links n = 18) visualized as median and 25^th and 75^th percentiles, with whiskers indicating minima and maxima. Welch’s two-sided t-test was used for comparisons and exact p-values are shown on the graphs. The final FAO product, Acetyl-CoA, enters the TCA cycle to produce NADH that can then be used by the ETS in oxidative phosphorylation. Significant changes in TCA cycle enzyme cross-link levels in aged mitochondria were also observed, including fumarase (FUMH) ([157]Extended Data Fig. 5a, [158]b), SUCA and SUCB1 ([159]Fig. 5e). Enzymes SUCA and SUCB1 are subunits of succinate-CoA ligase, which converts succinyl-CoA to succinate that is both a TCA cycle intermediate and an ETS substrate. Both SUCA and SUCB1 exhibited age-related decreased cross-link levels within ligand binding regions, including a cross-link at SUCA K90 which is the SUCA CoA binding site ([160]Fig. 5f). Succinyl-CoA ligase also produces ATP (or GTP in some tissues) during generation of succinate, and a nucleotide binding pocket exists in SUCB1. A total of 4 intra-links in SUCB1 were observed with age-related decreased levels, all including K143 cross-links that span the nucleotide binding pocket. Moreover, K108 and K98 in these decreased links exist within a distance compatible with salt bridge formation with GTP phosphate groups as shown in the pig crystal structure (PDB:2FP4). Since other SUCB links appear unchanged with age, these results indicate that age-related conformational differences within the ligand binding sites and not changes in protein levels mediate the observed changes in succinate-CoA ligase. Indeed, with consideration of the entire group of links, observed ratios of cross-links within both CoA and nucleotide binding regions appear significantly different from those in other succinate-CoA ligase regions ([161]Fig. 5i). Although the present age-related changes were measured in young and old murine muscle mitochondria, Wu et al. demonstrated that T cells from rheumatoid arthritis (RA) patients lack sufficient succinyl-CoA ligase activity to maintain balanced TCA cycle metabolic intermediates, implicating acetyl-CoA in controlling pro-inflammatory T cells in autoimmune disease.^[162]47 Cross-links identified in the present interactome data offer new opportunities to investigate succinyl-CoA conformational regulation in RA and possibly many other autoimmune diseases in ways not previously possible. Taken together, these results indicate that this snapshot of the FAO and TCA pathways in aged skeletal muscle demonstrate considerable protein conformational changes linked to reduced ligand binding that are not regulated at the protein level. Discussion Large-scale mitochondrial interactome quantitation together with functional measurements provide new molecular insights on age-associated functional decline in bioenergetics and metabolism. While previous transcriptome and proteome studies have provided unparalleled ability to visualize molecular abundance level regulation important in aging, it is clear other regulatory mechanisms beyond abundance are also involved. The approach presented here combines quantitation of protein, conformation, modification, ligand binding, and protein interaction levels to provide new biological insight on age-related molecular changes. Recently, the importance of studying protein interactions and their role in aging has been brought to the attention of the community.^[163]48 While this initial study is not comprehensive, these efforts have yielded the largest quantitative interactome dataset to define age-related mitochondrial differences thus far and includes 1521 quantified cross-links. To date, changes in glutamate dehydrogenase mRNA or protein levels with aging have not been reported and the studies presented here are consistent with that finding. However, the present quantitative cross-linking data generate new insights on DHE3 interactions in mitochondria and age-related conformational differences that may functionally contribute to age-associated changes connecting TCA cycle, ETS and α-KG effects on lifespan. Multiple reports demonstrate α-KG involvement in lifespan extension in mice^[164]49, flies^[165]50, yeast^[166]51and worms^[167]52. Increase in DHE3 activity has also been shown to accompany caloric restriction and subsequent increased lifespean.^[168]53 Moreover, diet-based lifespan extension in flies appears to be dependent on DHE3 expression.^[169]54 α-KG also promotes myofibroblast differentiation through epigenetic regulation by driving histone demethylation and the role of anaplerotic supply from glutamine and glutamate has been highlighted.^[170]55 Interestingly, α-KG effect on lifespan was stronger in female mice compared to males.^[171]49 The decreases in DHE3 antenna cross-link levels presented here are not the result of protein level changes and indicate PTM and or conformational differences exist in aged muscle mitochondria. Age-related increase in PTM levels is possible since both cross-linked residues, K477 and K480 are targets of the sirtuins SIRT3 and SIRT5, with acetylation levels of these residues increasing more than 8 fold upon SIRT3 knock out.^[172]25,[173]26,[174]56 However, cross-linked sites in the DHE antenna show both decreased levels in aged vs young mitochondria where glutamate respiration is repressed, and increased levels in ADP stimulated mitochondria where glutamate dehydrogenase activity is increased. In addition to the decrease in maximum glutamate/malate stimulated CI respiration, we also observed decreased maximum glutamate only stimulated respiration and lower sensitivity of respiration to glutamate in aged mitochondria compared to young. While not conclusive, these data are also consistent with impaired glutamate metabolism because with only glutamate as a substrate NADH is supplied to the ETS by glutamate dehydrogenase and alpha-ketoglutarate dehydrogenase activity alone. Therefore, these antenna cross-link levels can serve as probes of glutamate dehydrogenase activity in many future studies, including those to help unravel Sirtuin-, diet-, or exercise-mediated lifespan or healthspan extension. For instance, quantitation of DHE3 antenna links prior to age-induced changes in mitochondrial function, with caloric restriction or other interventions can help elucidate pathways that mediate age-related reduction in glutamate respiration. Quantitative cross-linking revealed changes in protein interactions and conformations affecting many facets of metabolism in aging muscle. Increased ROS production and alterations among mitochondrial ETS complexes are among primary aspects under study to better understand age-related mitochondrial functional decline. ETS complexes, especially CI, require coordinated and controlled assembly to achieve functional maturity.^[175]57 Therefore, altered CI assembly uncovered in the present study should be investigated further to elucidate its role in ETS pathologies in aging. Altered activity and ligand binding in FAO and TCA cycle enzymes can bring to the forefront the contributions of TCA intermediates and fatty acid metabolism to aging phenotypes, connecting phenotypes and molecular remodeling.^[176]58-[177]60 Taken together, these data enable a system-wide view of the changing mitochondrial interactome landscape linking changes in glutamate dehydrogenase activity together with amino acid metabolism, TCA cycle, and energy production by oxidative phosphorylation ([178]Fig. 6a). All age-related changes highlighted in this figure involve changes in protein conformations and interactions that are not readily attainable through conventional protein abundance level quantitation. Strikingly, many age-related interactome changes are well-correlated to the severity of aging mitochondrial phenotype, as shown with pairwise CI oxygen consumption ratio compared with the magnitude of changes in protein conformations, interactions and ligand binding ([179]Fig. 6b). Figure 6. Interactome remodeling associated with changes in muscle metabolism with aging. [180]Figure 6. [181]Open in a new tab a. Integrated pathways with age-associated changes in protein-protein interactions, protein-ligand interactions, or conformational changes highlighted in this study. b. Correlation between average of log[2] ratios in each biological replicate of cross-links changed with age in DHE3 (purple triangles, R^2=0.58), CIV (red squares, R^2=0.95), ACADV (green rhombi, R^2=0.87), CI (blue circles, R^2=0.63) and CI driven respiration. c. K-means clustering of cross-links with 5 clusters. CI oxphos log[2] change for P1,P2,P3, and P4 are in increasing order. d. Cross-links with age-related changing levels that are discussed in this study (light blue) cluster together (cluster E); non-changing cross-links from the same proteins are in dark grey. In the present manuscript, detailed discussion of only a small subset of cross-links was possible. K-means cluster analysis of all cross-links quantified in at least 2 biological replicates (95% conf. =<1) revealed that cross-links correlating with functional measurements cluster together and non-changing cross-links from these proteins are in a separate cluster together ([182]Fig. 6c, [183]d). Many other cross-links in these proteins display similar patterns to those discussed above with functional measurements, including proteins from the same pathways, such as ACSL1, CMC1, CPT1B, ATPB, and ATP5F1. The entire interactome dataset with quantitation, structures with mapped cross-links, and k-means clustering assignments is available to view online in XLinkDB.^[184]61 Although this report only covers aging mitochondria in a single sex and single tissue, these data provide a unique, detailed, and quantitative view of mitochondrial aging in muscle that can be used to guide future studies unraveling molecular underpinnings of metabolism changes with age. A more complete discussion of limitations can be found in the [185]supplement. Conclusions and Future Directions Mitochondrial protein interactomics provides a new tool to identify age-related changes in mitochondria that contribute to functional decline. This tool can now be applied to address many remaining questions including the interaction of age and sex, tissue- or cell-type specific changes such as muscle fiber type shifts, the impact of geroprotective treatments such as elamipretide or exercise training, and the human interactome. Methods Animal Husbandry. This study was reviewed and approved by the University of Washington Institutional Animal Care and Use Committee. Female 6-month (n = 4) and 30-month-old (n = 4) C57BL6/J mice were received from the Jackson Laboratory and used to measure CI-, CII-, and CI&CII-linked respiration and cross-linked for mass spectrometry. Further characterization of differences in aging mitochondria to assess top hits from the interactomics results were performed in female young (4-6 month, n=4) and old (27-29 month, n=5) C57BL6/J mice from the National Institute of Aging (NIA) colony. All mice were maintained at 21 °C on a 14/10 light/dark cycle at at 30-70% humidity and given standard mouse chow (LabDiet PicoLab^® Rodent Diet 20) and water ad libitum with no deviation prior to or after experimental procedures. Mitochondrial Isolation. The gastrocnemius muscle was dissected, and mitochondrial isolation was performed by differential centrifugation. The whole muscle was homogenized in Mitochondria Isolation Buffer (210 mM Sucrose, 2 mM EGTA, 40 mM NaCl, 30 mM HEPES, pH 7.4) using a high-speed drill in a glass Dounce homogenizer on ice.. The homogenate was centrifuged at 900 x g at 4 °C for 10 minutes. The supernatant was collected and centrifuged at 10,000 x g at 4 °C for 10 minutes. The supernatant was removed, and the mitochondrial pellet was resuspended in ice-cold Respiration Buffer (RB) without taurine or bovine serum albumin (BSA) (1.5 mM EGTA, 3 mM MgCl[2]-6H[2]O, 10 mM KH[2]PO[4], 20 mM HEPES, 110 mM Sucrose, 100 mM Mannitol, 60 mM K-MES, pH 7.1). The respiration buffer for mitochondrial resuspension did not include taurine, because it is an aminoethane sulfonic acid which contains a primary amine that could react with the cross-linker. Isolated mitochondria protein concentration was determined using standard Bradford Assay procedures. Mitochondrial Respiration. CI, CII, and CI&CII-linked mitochondrial respiration were assayed in mitochondria isolated from young (6-mo-old) and old (30-mo-old) female C57Bl6/J mouse gastrocnemius using an Oxygraph 2K dual respirometer/fluorometer (Oroboros Instruments, Innsbruck, Austria). RB with taurine and BSA was used for respiration measurements (1.5 mM EGTA, 3 mM MgCl[2]-6H[2]O, 10 mM KH[2]PO[4], 20 mM HEPES, 110 mM Sucrose, 100 mM Mannitol, 60 mM K-MES, 20 mM taurine, 1 g/L BSA, pH 7.1). Hexokinase clamp (1 U/ml hexokinase, 2.5 mM D-glucose) was used to maintain equilibrium of ATP/ADP at submaximal ADP concentrations.^[186]62 Respirometry and fluorometry reagent stocks were prepared according to Oroboros instructions (bioblast.at). Respiration was measured at 37°C with stirring during substrate and inhibitor titrations. To measure CI, CII, and CI&CII-linked respiration, first, 10 μM cytochrome c was added to each chamber to allow measurement of respiration in isolated mitochondria without limiting by membrane damage occurring during isolation. Approximately 35 μg mitochondrial homogenate (~8-11 μL) was added to each 2 mL chamber. CI, CII, and CI&CII-linked respiration were measured in parallel for each sample by adding complex-specific substrates and inhibitors and titrating in ADP. CI-linked respiration was measured by adding 10 mM glutamate and 0.5 mM malate. CII-linked respiration was measured by adding 10 mM succinate and 0.5 μM rotenone. CI&CII-linked respiration was measured by adding 10 mM succinate, 10 mM glutamate, and 0.5 mM malate. The OXPHOS capacities for each substrate condition were determined as the maximum oxygen consumption rate (OCR) measured during a titration of ADP from 5-6000 μM ADP. The background oxygen consumption with de-energized mitochondria was subtracted from all measured functional parameters before reporting final values. Response to glutamate was measured in mitochondria isolated from young (4-6 month, n=4) and old (27-29 month, n=4) female C57BL6/J gastrocnemius muscles using the Oxygraph 2K dual respirometer/fluorometer. RB with taurine and BSA was used for respiration measurements without hexokinase clamp because saturating ADP concentrations were added to the chambers in a single bolus during the experiment. To measure glutamate sensitivity, a sequential titration of 50 μg mitochondrial protein, 2.5 mM ADP, 10 μM cytochrome c, and sequential additions of 1 mM glutamate up to 10 mM glutamate final concentration were performed. The mitochondrial respiration results were analyzed using Microsoft Office Excel and GraphPad Prism 9.9 for Mac OS X (GraphPad Software, La Jolla, CA). For all comparisons, P < 0.05 was considered statistically significant. Comparisons between two groups were analyzed using unpaired two-tailed student’s t-test. Comparisons during ADP or glutamate titrations were analyzed using repeated measures Two-way ANOVA with Sidak’s multiple comparisons. titration kinetics were analyzed using nonlinear regression - [Agonist] vs. normalized response variable slope to calculate EC[50]. Plots depict mean ± standard deviation Citrate Synthase Activity Assay. CS activity is reportedly a more accurate marker of mitochondrial mass than total protein content when performing comparisons across age.^[187]63 CS activity assay was performed on mitochondrial isolations and used to normalize mitochondrial respiration. CS Activity was measured by spectrometric quantitation (412 nm) of 5,5’dithiobis-2-nitrobenzoic acid conversion to 2-nitro-5-thiobenzoic acid in the presence of Coenzyme A thiol generated during citrate production (CS0720, Sigma) as previously described.^[188]64 In-gel Complex I and IV activity assays. In-gel CI and CIV activity were measured in mitochondria isolated from young (4-6 month, n=4) and old (27-29 month, n=5) female C57BL6/J gastrocnemius muscles as previously described.^[189]65 100 μg of isolated mitochondria were loaded in each gel lane. All visible bands for each complex and supercomplex on each gel were quantified In Bio-Rad ImageLab (Version 6.1 for OS X). The background was corrected using the pixels immediately outside each lane volume and the same background cut off filter value for all lanes. The volume values for each band were normalized to the CS activity per mg of protein for the sample in each lane. All values were compared by student's t-test or Two-Way ANOVA with Sidak's post hoc test in GraphPad Prism. Cross-linking of isolated muscle mitochondria. Isolated mitochondria from murine gastrocnemius muscles of 8 mice (4 young and 4 old) were resuspended in cross-linking buffer (170 mM Na[2]HPO[4], pH 8.0) and either reporter heavy (RH) or stump heavy (SH) iqPIR reagent was added;^[190]15 final reaction volume was 100 uL and cross-linker concentration was 10 mM. Cross-linking reaction was allowed to proceed for 30 min at room temperature with shaking to form covalent cross-links between proximal lysine residues within and between proteins and protein complexes in intact mitochondria. Cross-linking buffer was then removed by centrifugation and mitochondrial pellets were lysed in 8M urea. Proteins were reduced with TCEP (30 min RT with shaking) and alkylated with IAA (30 min RT with shaking). Protein concentration of each mitochondrial sample was measured with a Bradford assay using Cytation plate reader. Samples were mixed pairwise (one old and one young using equal amount of protein from each sample making 4 biological replicates total. Protein mixtures were digested with trypsin overnight (1:100 trypsin concentration at 37 C with shaking). Peptides were then acidified with TFA and cleaned using Sep-Pak c18 columns (Waters). Peptides were separated using SCX chromatography (Luna column, Agilent HPLC) into 14 fractions and fractions were pooled together as following: fractions 1 to 5, fractions 6 and 7, fraction 8, fraction 9, fraction10, fractions 11 to 14. Pooled fractions were dried in a SpeedVac and resuspended in ammonium bicarbonate buffer; pH was adjusted to 8.0 with NaOH. Biotinylated cross-linked peptides were captured with monomeric avidin (ThermoFisher Scientific 20228) for 30 min at RT with shaking. The beads were washed with ammonium bicarbonate and peptides were eluted with 0.1% formic acid in 70% ACN, dried down by vacuum centrifugation and resuspended in 20 uL of 0.1% formic acid. Mitochondrial isolation from HEK293 cells and treatment with ADP. HEK293 cells were grown in DMEM media supplemented with 3.5 mg/L glucose, 10% fetal bovine serum, 1% penicillin and streptomycin to confluency. The plates were washed with PBS, cells detached using EDTA 20 mM, centrifuged and washed twice in MgCl[2]. Cells were then resuspended in ice-cold mitochondrial isolation buffer (70 mM sucrose, 220 mM D-mannitol, 5 mM MOPS, 1.6 mM carnitine, 1 mM EDTA at pH 7.4) and homogenized in a glass homogenizer. The homogenate was centrifuged at 600 g for 5 min at 4 C. The supernatant was transferred to a 15 mL tube and centrifuged at 8000 g for 10 min at 4 C. The supernatant was then removed, and mitochondrial pellet was resuspended in 5 mL of mitochondrial isolation buffer and centrifuged at 8000 g for 10 min. The mitochondrial pellet was then resuspended in 200 uL of mitochondrial isolation buffer and split into two. ADP was added to one vial to a final concentration of 1.5 mM. Both samples were incubated at RT for 10 min with shaking. Supernatant was then removed by centrifugation and pellets were resuspended in cross-linking buffer. RH iqPIR cross-linker and ADP was added to ADP treated sample to final concentrations of 10 and 1.5 mM respectively. SH iqPIR cross-linker was added to control sample to a final concentration of 10 mM. The cross-linking reaction was allowed to proceed for 30 min at RT with shaking. The supernatant was then removed by centrifugation and mitochondrial pellets were lysed, reduced, alkylated, combined, digested, and processed for mass spectrometric analysis the same way as murine muscle mitochondria. Mass Spectrometry and data analysis. Four uL of each pooled fraction was loaded on a 60 cm C8 heated column and separated on 2 hour gradient on nanoAcquity HPLC system (Waters) and analyzed with QExactive Plus mass spectrometer (ThermoFisher Scientific). Thermo Scientific Xcalibur software (4.2.28.14) was used for MS data acquisition. MS1 scans were analyzed at 70K resolution with AGC target 1e6, and maximum ion time 100 ms. Top 5 peaks with charge 4 or greater were selected for HCD fragmentation with NCE 30 and MS2 spectra were collected at 70K resolution, 5e4 AGC target, and 300 ms maximum ion time. Raw files were converted to mzXML, and spectra containing cross-linked peptides were determined with Mango software.^[191]66 These spectra were then searched against mouse Mitocarta 2.0 database using Comet search engine^[192]67 and cross-linked peptides were validated with XLinkProphet.^[193]68 Identified cross-links were quantified using iqPIR algorithm and results were uploaded to XLinkDB database. Normalized log[2] ratios and associated p-values based on the Student’s t-test on each quantified ion for every cross-link (t = sqrt(df*mean/std) and p-values calculated with the pt function of R: pt(-abs(t), df) where t is the t-statistic and df the degrees of freedom) were downloaded from XLinKDB and correlation plots between biological replicates, density plots for each replicate, volcano plot indicating significantly changed cross-links, box-plots and t-test comparisons were generated in R using tidyverse package and R markdown is provided.^[194]69 In all boxplots horizontal line represents median, the lower and upper hinges correspond to the first and third quartiles (the 25th and 75th percentiles) and the whiskers extend to the value no further than 1.5 IQR. Pathway enrichment analysis and network of differentially expressed cross-links were generated using STRING web-application. KEGG and GO terms are displayed in the order of strength to describe how large the enrichment effect is. The strength value is the ratio between the number of proteins in the network that are annotated with a term and the number of proteins that is expected to be annotated with this term in a random network of the same size. All displayed terms have FDR corrected p-value less than 0.01.^[195]70 Heatmap of all common cross-links was generated for cross-links quantified with 95% confidence (interval within which one can be sure with 95% confidence that the actual mean value resides, calculated as 1.96 * std / sqrt(num_reps) assuming normal distribution) less than 0.5 in all 4 biological replicates. Cross-links were mapped on available structures with either Euclidean distances or SASD distances calculated by Jwalk.^[196]71 Statistics and Reproducibility No statistical method was used to predetermine sample size. No data were excluded from the analyses. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment. An F-test was used to test for unequal variances in the respirometry data, but no conditions had significant differences in variance. Normality for XL-MS data is presented in [197]fig. 2b. Variance was not measured in XL-MS as the data analysis was performed by pairing samples. Therefore, the variance cannot be calculated for each population. Welch’s t-test was used for comparisons between cross-link populations. Extended Data Extended Data Fig. 1, [198]Extended Data Fig. 1, [199]Open in a new tab a. Histogram of calculated Euclidean distances for all intraprotein cross-links mapped to AlphaFold predicted structures. b. Boxplots of number of ions used for each log[2] ratio in P1, P2, P3, P4. X-axis is capped at 40 for readability but there are ratios in each sample with more than 40 ions. c. Histogram of differences between a mean log2 ratios for cross-linked residue pairs based on multiple cross-linked peptides (cross-links that connect the same lysines, but can be identified in differently cleaved or modified peptides) and each cross-linked peptide pair d. Submitochondrial localization of significantly changed interlinks (left) and intralinks (right). e. STRING network of proteins with significantly changed cross-links. f. Isolated gastrocnemius mitochondria oxphos capacity normalized by amount of protein from the 4 young (6 months) and 4 old (30 months) female mice. Mean ± SD. Extended Data Fig. 2, [200]Extended Data Fig. 2, [201]Open in a new tab a. Heatmap of log2 ratios of all NDUS1 and NDUV1 cross-links and interprotein cross-links to other CI subunits. b. Boxplots of all intralinks in CI subunits by a biological replicate. c. Boxplots of crosslinks downregulated in aging and non-changing Intralinks based on all 4 biological replicates for CI (NDUV1XLs n = 34, other links n = 112; visualized as median and 25^th and 75^th percentiles, with whiskers indicating minima and maxima and CIV (f). Interlinks n = 30, other links n = 14; visualized as median and 25^th and 75^th percentiles, with whiskers indicating minima and maxima; P-values are from Welch two-sided t-test. Structure of supercomplex (PDB 5GUP) with cross-linked CI and CIII subunits highlighted (top) and specific CI-CIII crosslinks mapped to the subunits (bottom); decreased cross-linked are in green. e. CIII cross-links mapped to a bovine structure. Subunits with decreased intralinks highlighted and zoomed in (right). g. Correlation plots of CI and CIV cross-links changing in aging mitochondria and CII OXPHOS with multiple R-squared displayed. Extended Data Fig. 3, [202]Extended Data Fig. 3, [203]Open in a new tab In-gel CI activity assay blot with identified bands. b. Quantification of total CI activity (left; statistical significance determined by unpaired two-tailed student’s t-test; p=0.1726) and CI activity for each identified band (right; statistical significance determined by Ordinary Two-Way ANOVA with Sidak’s post-hoc test; p=0.0011 age main effect). c. In-gel CIV activity assay blot with identified bands. d. Quantification of total CIV activity (left; statistical significance determined by unpaired two-tailed student’s t-test; p=0.1219) and CIV activity for each identified band (right; statistical significance determined by Ordinary Two-Way ANOVA with Sidak’s post-hoc test; p=0.0045 age main effect; p=0.0059 for Band 1 Sidak’s post hoc test). Both activity assays were performed using isolated gastrocnemius mitochondria from young (4-6 month, n=4) and old (27-29 month, n=5) NIA C57BL/6J female mice. Mean ± SD. ns - not significant, **p<0.01 by Sidak’s post-hoc test. Extended Data Fig. 4, [204]Extended Data Fig. 4, [205]Open in a new tab a. Decreased and non-changing cross-link levels in glutamate dehydrogenase highlighted on the volcano plot with Bonferroni corrected p-value=0.05. b. Heatmap of all DHE3 cross-linked peptide pairs with each individual peptide sequence shown. Cross-linked lysine residues are in red. c. Correlation plots of DHE3 antenna cross-links changing in aging mitochondria and CII OXPHOS. e. Normalized glutamate stimulated respiration across a range of glutamate concentrations (left; statistical significance determined by Two-Way RM ANOVA with Sidak’s post-hoc test; p=0.0058 age main effect) and maximum respiration capacity with glutamate stimulation (right; statistical significance determined by unpaired two-tailed student’s t-test; p=0.0009) in young (4-6 mo, n=4) and old (27-29 mo, n=4) female NIA C57BL/6J isolated gastrocnemius mitochondria. f. The kinetics of glutamate stimulated respiration are altered with age (left; nonlinear regression determined by [Agonist] vs. normalized response -- Variable slope) in young (4-6 mo, n=4) and old (27-29 mo, n=4) female NIA C57BL/6J isolated gastrocnemius mitochondria. The amount of glutamate required to stimulate 50% respiration calculated from the nonlinear regression is increased in old (right; statistical significance determined by unpaired two-tailed student’s t-test; p=0.0253). Mean ± SD. *p<0.05, **p<0.01. Extended Data Fig. 5, [206]Extended Data Fig. 5, [207]Open in a new tab Fumarate hydratase cross-links mapped to a E.Coli structure(PDB 4HGV). Decreased cross-link levels are shown in the zoomed in square. b. Heatmap of log[2] ratios of fumarate hydratase cross-links. c. Boxplots for Acadvl cross-links based on all 4 biological replicates. CoA proximal XLs n = 21, other links n = 46; visualized as median and 25^th and 75^th percentiles, with whiskers indicating minima and maxima; P-values are from Welch two-sided t-test. Extended Data Table. 1, Pairs (P1-P4) are numbered the same way for qXL-MS and functional data. Pairs were assigned based on the amount of available protein to maximize input material. Log[2] fold change calculated for each paired biological replicate in mitochondrial yield, CS activity, and oxygen consumption on either CI substrates or CII substrates. Pair Mito yield CS activity CI oxphos CII oxphos P1 −0.712 0.712 0.441 −0.667 P2 −0.878 0.918 −2.1 −0.788 P3 −1.282 0.932 −0.933 0.107 P4 −1.417 1.186 −1.369 −0.697 [208]Open in a new tab Supplementary Material Supplemental Info [209]NIHMS1880499-supplement-Supplemental_Info.pdf^ (340.4KB, pdf) Acknowledgments