Mitochondrial Protein Interaction Mapping Identifies Regulators of Respiratory Chain Function

Abstract

Mitochondria are essential for numerous cellular processes, yet hundreds of their proteins lack robust functional annotation. To reveal functions for these proteins (termed MXPs), we assessed condition-specific protein-protein interactions for 50 select MXPs using affinity enrichment mass spectrometry. Our data connect MXPs to diverse mitochondrial processes, including multiple aspects of respiratory chain function. Building upon these observations, we validated C17orf89 as a complex I (CI) assembly factor. Disruption of C17orf89 markedly reduced CI activity, and its depletion is found in an unresolved case of CI deficiency. We likewise discovered that LYRM5 interacts with and deflavinates the electron-transferring flavoprotein that shuttles electrons to coenzyme Q (CoQ). Finally, we identified a dynamic human CoQ biosynthetic complex involving multiple MXPs whose topology we map using purified components. Collectively, our data lend mechanistic insight into respiratory chain-related activities and prioritize hundreds of additional interactions for further exploration of mitochondrial protein function.

Keywords: C15orf48; C2orf47; DHRS4.

Figure 1. AE-MS methodology

(A) Schematic workflow of the AE-MS method. (B) Localization of all FLAG-tagged constructs was established based on MLS-GFP and anti-FLAG fluorescence microscopy (see Figure S1). Venn diagrams report the percentage of the MitoCarta+ list (Table S1) associated with each category. (C) Galactose induces mitochondrial respiration. Ratio of rates of oxygen consumption (OCR) and extracellular acidification (ECAR) in HEK293 (left) and HepG2 (right) cells grown in 10 mM glucose (Glu) or galactose (Gal) for 24 hours prior to assay. OCR/ECAR is proportional to mitochondrial vs. glycolytic flux (* indicates t-test p-value < 0.05). Error bars indicate ± SEM.

Figure 2. Overall analyses of our AE-MS approach

(A) CompPASS scoring accuracy. CompPASS scores were calculated for all bait-prey interactions, including those involving only mitochondrial prey (gray), and those determined a priori to be high-confidence PPIs based on the literature (black). At each score threshold, the percent of remaining PPI per bin was calculated. Scores to the right of the vertical gray bar exceeded the threshold set for this study and are counted as high-confidence interactions. (B) Quantitative scoring enriches for high-confidence interactors. Top: Schematic of the results of CompPASS score filtering. Bottom left: Heat map where white indicates a prey was not observed, and shades of gray indicate quantified abundance. Prey proteins are in rows, and FLAG-tagged baits are in columns. Data are averages from 6 replicates per cell line (3 glucose, 3 galactose). Bottom right: Heat map showing scores above CompPASS threshold. Preys and baits are organized in the same order as in left heat map. Black indicates a score above the threshold. (C) Venn diagram of high-confidence PPIs from each cell line. Mitochondrial interactions above threshold in HEK293 (blue) and HepG2 (green) cells are indicated (see Figure S2). (D) Venn diagram of high-confidence PPIs from each carbon source. Mitochondrial interactions above threshold in glucose (orange) and galactose (purple) cells are indicated. (E) Histogram of the fold change abundances of PPIs between carbon sources. Select dynamic PPIs involving proteins from the coenzyme Q biosynthetic pathway (see Figure S7) are indicated.

Figure 3. C17orf89 is required for complex I assembly

(A) Schematic of top-scoring C17orf89 interactions. Arrows originate from bait proteins and point to high-confidence interactors. (B) Immunoblot of immunoprecipitated FLAG-tagged C17orf89, LYRM5, and MLS-GFP with anti-FLAG (red) or anti-NDUFAF5 (green). (C) Activity of complex I in C17orf89 knock down (kd) and control (c) HEK293 cell lines and control lines. Error bars indicate ± SD (see also Figure S3). (D) Measurement of oxygen consumption rate (OCR) for the same kd or c lines as in (C) using a Seahorse Extracellular Flux Analyzer (FCCP, carbonyl cyanide p-trifluoromethoxyphenylhydrazone). Error bars indicate ± SEM. (E) Complex I activity in kd and c cell lines after transfection with GFP-FLAG (negative control) or C17orf89-FLAG (rescue). Error bars indicate ± SD.

Figure 4. C17orf89 stabilizes NDUFAF5 and is depleted in a case of CI deficiency

(A) Immunoblots of mitochondrial proteins in cells treated with siRNA for CI and CIAF genes. C17orf89 knockdown results in loss of CI subunits and a marked depletion of NDUFAF5 (red box). See also Figure S4. (B) RNAseq analysis of C17orf89 expression in 96 cell lines from patients with respiratory chain dysfunction (RPKM, Reads Per Kilobase of transcript per Million mapped reads). Arrow indicates a line with severe loss of C17orf89 expression. (C) Respiratory chain complex analyses of the patient line indicated in (B), revealing an isolated CI deficiency (P, patient; C, control). Error bars represent mean +/− SD, n=25. (D) Proposed model of C17orf89–NDUFAF5 complex function in CI assembly.

Figure 5. LYRM5 forms a complex with ETF

(A) Schematic of top-scoring LYRM PPIs. LYRM5-ETF interactions are shaded. (B) Validation of the interaction between LYRM5 and both ETFA and ETFB. C-terminally FLAG-tagged GFP, IVD, NDUFA4, and LYRM5 were immunoprecipitated (IP) from HEK cells and immunoblotted (IB) with anti-ETFA and anti-ETFB (upper) or anti-FLAG (lower). LYRM5 enriched for both ETF proteins more efficiently than IVD, a known ETF interactor. (C) IP of LYRM5-FLAG or MLS-GFP-FLAG from HEK293 cells analyzed by Blue Native-PAGE analysis and IB. The same membrane was blotted for ETFB (left) and then FLAG (right) (see also Figure S5). (D) Recombinant N-terminally His-tagged LYRM5 and untagged ETFA/B were co-expressed in E. coli. Purification of LYRM5 by metal affinity chromatography led to the co-purification of ETFA/B. (E) LYRM5 and ETF form a stable complex. Size exclusion chromatography of LYRM5 alone (red), ETF alone (green), or the co-purified LYRM5-ETF complex (blue) noted in (D).

Figure 6. LYRM5 deflavinates ETF

(A) ETF activity in the presence of increasing amounts of LYRM5 (± SD from triplicate measurements). (B) FAD release from ETF upon incubation with varying amounts of LYRM5, measured by fluorescence emission spectroscopy. (C) Visible spectra of ETF in the presence of LYRM5. The flavin visible spectrum shows two shoulder peaks at 420nm and 460nm due to the interaction between FAD and ETF protein residues that are lost upon addition of LYRM5. Spectrum a, b, and c are for LYRM5:ETF ratios of 0, 2, and 4, respectively. For clarity, spectra a and b have been shifted by +0.04 and 0.02 OD units, respectively. (D) Proposed model of the functional interaction between LYRM5 and ETF. Binding of four molar equivalents of LYRM5 (red dot) to ETF (green trace of PDB ID 1EFV) leads to the loss of FAD from ETF (F).

Figure 7. Interaction analysis reveals the dynamic human CoQ-related complex

(A) All CoQ-related proteins used as baits and/or observed as preys in this study are shown as white nodes. Interactions above our score threshold are shown for HEK cells (gray arrows) or HepG2 cells (green arrows). (B) CoQ-related interactions involving COQ8A or COQ8B as bait were assessed for the effect of cellular metabolic status. Interactions more abundant in galactose or glucose media are shown in red or blue, respectively, while those with no clear media effect are in white. (C) Relative abundance of CoQ₁₀ in HEK293 and HepG2 cells after treatment with 10 mM glucose or galactose for 24 hours (*** indicates t-test p-value < 0.001, error bars indicate the 95% confidence interval). (D-E) Representative results of in vitro protein translation and purification of each core CoQ complex protein individually (D) or in pairs (E). Proteins were run on SDS-PAGE and detected by Coomassie stain. See Figure S6 for all interactions observed. (F) Schematic of COQ protein network direct interactions established in vitro. All robust interactions are represented as edges between the COQ protein nodes. (G) In vitro co-purification of all six core COQ complex proteins. (H) Three-dimensional model of predicted COQ complex structure based on in vitro interaction data. Colors are as in Figure 3F.