Mitochondrial protein interactome elucidated by chemical cross-linking mass spectrometry

Abstract

Mitochondrial protein interactions and complexes facilitate mitochondrial function. These complexes range from simple dimers to the respirasome supercomplex consisting of oxidative phosphorylation complexes I, III, and IV. To improve understanding of mitochondrial function, we used chemical cross-linking mass spectrometry to identify 2,427 cross-linked peptide pairs from 327 mitochondrial proteins in whole, respiring murine mitochondria. In situ interactions were observed in proteins throughout the electron transport chain membrane complexes, ATP synthase, and the mitochondrial contact site and cristae organizing system (MICOS) complex. Cross-linked sites showed excellent agreement with empirical protein structures and delivered complementary constraints for in silico protein docking. These data established direct physical evidence of the assembly of the complex I-III respirasome and enabled prediction of in situ interfacial regions of the complexes. Finally, we established a database and tools to harness the cross-linked interactions we observed as molecular probes, allowing quantification of conformation-dependent protein interfaces and dynamic protein complex assembly.

Keywords: cross-linking; interactome; mass spectrometry; mitochondria; protein interaction reporter.

Fig. 1.

Determination of cross-linked protein interactions in functional mouse mitochondria. (A) Functional mitochondria were cross-linked and lysed, and protein lysates were digested with trypsin, followed by SCX fractionation on the cross-linked peptides. Pooled SCX fractions were enriched with monomeric avidin to capture the biotin-containing cross-linker covalently bound to two peptides. Cross-linked peptides were identified by MS/MS and ReACT (24). (B) The OCR of isolated mitochondria with pyruvate and malate as substrates was measured. After the baseline measurements, 50 µL of 40-mM ADP (4 mM final concentration) and 55 µL of 25 µg/mL oligomycin A (2.5 µg/mL final concentration), a complex V inhibitor, were injected sequentially. The changes in OCR after the addition of ADP (state 3) and then oligomycin A (state 4) were measured by Seahorse XF24. The RCR is expressed as state 3 OCR/state 4 OCR. Data were expressed as the SD from five technical replicates. (C) Protein interactions as determined by large-scale chemical cross-linking analysis of functional mouse heart mitochondria. Nodes represent individual proteins; edges (lines) represent all cross-links identified between two proteins. Nodes are colored according to the number of samples in which each protein interaction was observed. An interactive network depicting site-to-site interactions is available in Dataset S2.

Fig. S1.

Posttranslational modifications and complex interactions determined by XL-MS analysis of functional mitochondria. (A) Conserved mouse ADT1 cross-linked and acetylated sites are shown on the bovine crystal structure (PDB ID code: 2C3E) (45) in the space-filling model. For the intraprotein link between ADT1 cross-linked sites (K49–K147, light purple), acetylation was observed at the nearby residue K52 (green). (B) Network of cross-linked sites from the MICOS/MIB complex. Nodes represent individual cross-linked sites. Edges represent linkages between identified cross-linked sites. (C) XL-MS interactions depicted in relation to protein sequences for the MICOS proteins. Colored bars represent protein sequences, purple edges are intraprotein links; red edges are homo-multimeric links; green edges are interprotein links. UniProt sequence annotations were used to color the protein sequences. (D) Homology model of MIC60 highlighting sites of cross-linking. Identified cross-linked sites are shown as green space-filled residues. MICOS proteins identified in interactions at each site are shown. The low-confidence predictions for the N-terminal structure (amino acids 1–203) are denoted by a dotted line.

Fig. 2.

Cross-linked sites mapped to empirical protein structures. Sites of cross-linking identified in empirical structures are shown along with site–site interaction networks. Cross-links are depicted as space-filled residues and green lines. Cα–Cα distances for the links are shown. Network nodes represent individual cross-linked sites; edges represent identified cross-links. Cross-links highlighted in each structure are shown as green edges in the network (78). (A) Complex III dimer (monomers shown in light gray and dark gray). Homology model of mouse QCR2 overlaid on the yeast complex III structure (PDB ID code: 1KYO). The mouse QCR2 proteins for monomers one and two are shown in blue and teal, respectively. (B) Homology models of mouse SDHA and SDHB were overlaid on the porcine complex II structure (PDB ID code: 4YXD) (79). Identified cross-linked lysines are highlighted in blue. (C) The ETFA–ETFB complex structure (PDB ID code: 1EFV) (78) from human, with conserved lysine sites highlighted in teal. Mouse ETFA (purple) K139 corresponds to human K139, and mouse ETFB (orange) K26 corresponds to human R26. (D) Complex IV structure from bovine with COX7B (magenta) and COX6C (pink) highlighted (PDB ID code: 3ASO) (80). Mouse COX7B K75 (corresponding to bovine K51) and mouse COX6C K68 (corresponding to bovine K65) link the proteins on the IMS side of the complex. (E) The Phyre2 model of ATP8 (red) was superimposed on a helix assigned to ATP8 (yellow) in cryo-EM–derived structures (55) of ATP synthase in rotational state 3 (PDB ID code: 5LQX) and rotational state 1 (PDB ID code: 5LQZ). Cross-linked sites (green space-filled residues) between APT8 (K46 and K48) and ATPD (K136) are compatible with state 3, whereas the cross-link between ATP8 (K48) and ATP5E (K50) is compatible only with state 1. Cross-linked sites involving ATP8 and six other ATP synthase subunits are indicated in the subnetwork and are displayed on a structure in Fig. S4.

Fig. S2.

In situ mitochondrial protein OXPHOS interaction network. The network of protein interactions determined by large-scale chemical cross-linking analysis of functional mouse heart mitochondria for only the proteins that belong to one of the major OXPHOS complexes. Intercomplex protein interactions are highlighted as red edges.

Fig. S3.

Murine structural models used to measure Cα–Cα distances in eukaryotic complex structures. Structural models of murine mitochondrial proteins derived from homology modeling were overlaid with empirical structures from other eukaryotic species. The rmsd for the model compared with the empirical structure is shown. (A) QCR2 from Saccharomyces cerevisiae (PDB ID code: 1KYO). The homo-multimeric link for QCR2 at K250 in mouse is shown as a magenta spaced-filled residue. The conserved lysine in yeast (K218) is shown as a green space-filled residue. (B) Bovine SDHA (PDB ID code: 4YXD). (C) Bovine SDHB (PDB ID code: 4YXD).

Fig. S4.

Structural model of full-length ATP8 in complex V. The full-length ATP8 model from Phyre2 was overlaid into the cryo-EM–derived complex V structure from P. angusta in rotational state 3 (PDB ID code: 5LQX). The experimentally observed helix assigned to ATP8 is shown in yellow. Cross-linked sites are highlighted as green space-filled residues. The cross-link identified between K48 of ATP8 and K50 of ATP5E exceeds the possible distance for cross-linking (∼35 Å) in state 3 but is possible in rotational state 1 (shown in Fig. 2E).

Fig. S5.

Modeling of full-length ATIF1 interacting with complex V and ADT1 acetylation sites. (A) Overlay of the Phyre2 model of full-length ATIF1 ATP synthase inhibitor (white) on the empirically derived ATIF1 fragment (red) within the context of ATP synthase (PDB ID code: 4Z1M). (B) Sites of cross-linking within the full-length model of ATIF1 to ATP synthase subunits based on the overlaid model in A. Cross-links identified in this study are highlighted for each of the protein components.

Fig. 3.

Determination of supercomplex structures from functional mitochondria. (A) Supercomplex model from rigid body docking (59) of complex I (NDUA2, NDUA4) and complex III (QCR2, QCR6) using cross-linked peptide distance constraints (NDUA2–QCR2; NDUA4–QCR6). Complex I is shown in pink, and complex III is shown in blue. Ribbon models of intercomplex cross-linked proteins are shown within the surface model in the left panel, and the distance constraints used are displayed as gray lines. (B) Workflow for the comparison of a recently published cryo-EM structure (PDB ID code: 5J8K) (21) and the XL-MS–based supercomplex model. The XL-MS–based model was generated without prior knowledge of the cryo-EM model. (C) Comparison of the in situ XL-MS docked supercomplex with the cryo-EM supercomplex model (PDB ID code: 5J8K). Structures were aligned based on complex I. Complex I rmsd = 1.3 Å; complex III rmsd = 2.4 Å.

Fig. S6.

Cross-linked site mapping used to generate the docked model of the complex I–complex III supercomplex. Proteins identified in intercomplex interactions were modeled with Phyre2. Full-length murine homology models were overlaid on conserved proteins in Bos taurus or S. cerevisiae empirical complex structures for complex I and III (PDB ID codes: 5DLW and 1KYO), respectively, to determine structurally conserved residues. The cross-linked sites and conserved residues from homologous proteins in other species are shown (space-fill). The rmsd values from the comparison of the murine models (orange) with their homologs (blue) are shown. (A) Murine NDUA2 model compared with fragment of NDUA2 from B. taurus (PDB ID code: 5LDW). The empirical structure was truncated at R16 of the N terminus (i.e., did not contain K13); therefore R16 was used with a distance constraint of 0–40 Å. (B) Full-length murine NDUB4 was recalcitrant to modeling because of the nature of the extended α-helix with multiple loop regions. Thus, the conserved site of cross-linking is shown surrounding K120 in both bovine and murine NDUB4 proteins. The NDUB4 structure was derived from PDB ID code 5DLW. (C) Murine QCR2 model compared with yeast QCR2 (PDB ID code: 1KYO). K218 from yeast was used in place of K250 from murine QCR2. (D) Murine QCR6 model compared with yeast QCR6 (PDB ID code: 1KYO). QCR6 K83 corresponded to R141 in the yeast structure.

Fig. S7.

Demonstration of qualitative and quantitative capabilities using the in situ mitochondrial protein interaction atlas. (A) Spectral library analysis of cross-linked peptide relationships. One mzXML file from the combined dataset was searched against the full spectral library to render the reciprocal plots. Dot product for each of these plots was 1.000. (B) Workflow demonstrating how identified cross-linked residues from the atlas can be used to generate PRM transitions and allow quantification in the absence of specialized chemical cross-linking platforms. Neither technique requires specialized instrumentation or software or access to ReACT to identify and quantify cross-linked peptides; both techniques allow analyses across a wide range of instrumentation (29, 35). Both the spectral library and the PRM transition library (highlighted in red boxes) can be utilized to probe conformational and protein interaction changes in mitochondria under different conditions and are available in Dataset S3.