Proximity Biotinylation as a Method for Mapping Proteins Associated with mtDNA in Living Cells

Abstract

A recurring challenge in cell biology is to define the molecular components of macromolecular complexes of interest. The predominant method, immunoprecipitation, recovers only strong interaction partners that survive cell lysis and repeated detergent washes. We explored peroxidase-catalyzed proximity biotinylation, APEX, as an alternative, focusing on the mitochondrial nucleoid, the dynamic macromolecular complex that houses the mitochondrial genome. Via 1-min live-cell biotinylation followed by quantitative, ratiometric mass spectrometry, we enriched 37 nucleoid proteins, seven of which had never previously been associated with the nucleoid. The specificity of our dataset was very high, and we validated three hits by follow-up studies. For one novel nucleoid-associated protein, FASTKD1, we discovered a role in downregulation of mitochondrial complex I via specific repression of ND3 mRNA. Our study demonstrates that APEX is a powerful tool for mapping macromolecular complexes in living cells, and can identify proteins and pathways that have been missed by traditional approaches.

Keywords: APEX; APEX2; FASTKD1; mapping; mass spectrometry; mitochondria; nucleoid; oxphos; peroxidase; proteomics.

Figure 1. Design and Characterization of APEX2 Fusion Constructs Used for Live-Cell Proteomic Mapping

(A) Scheme of APEX2-mediated proximity biotinylation within the mitochondrial nucleoid. APEX2 (green) is genetically targeted to the nucleoid via fusion to Twinkle, a known nucleoid protein. The gray shapes are endogenous proteins residing inside and outside the nucleoid in the mitochondrial matrix. H₂O₂ is added for 1 min to cells preloaded with biotin-phenol (BP; red B = biotin) to initiate labeling. APEX2 oxidizes BP into a phenoxyl radical, which covalently tags proximal endogenous proteins at electron-rich side chains such as Tyr. mtDNA, mitochondrial DNA; OMM, outer mitochondrial membrane; IMM, inner mitochondrial membrane. (B) APEX2 fusion constructs employed in this study. Mito-V5-APEX2 employs a 24-amino-acid mitochondrial targeting sequence (MTS) from COX4 to localize APEX2 throughout the entire mitochondrial matrix. V5 is an epitope tag. (C) Fluorescence imaging of Twinkle-APEX2 localization. Stably expressed Twinkle-APEX2 (green) was visualized in HEK293T cells by anti-V5 staining (Alexa Fluor 647). Nucleoids (red) were visualized with anti-DNA antibody (Alexa Fluor 488). Figure S1D shows an expanded view of the entire cell. Scale bar, 10 μm. (D) Electron microscopic (EM) visualization of stably expressed Twinkle-APEX2 in HEK293T cells. Dark-stained regions indicate APEX2 activity. Staining was performed after cell fixation with diaminobenzidine and H₂O₂ followed by reduced osmium. ER, endoplasmic reticulum; M, mitochondria; N, nucleoid; C, cristae. Scale bars, 200 nm. (E) Fluorescence imaging of Twinkle-APEX2 activity. Live-cell biotinylation was performed for 1 min with BP and H₂O₂ as in (A). Cells were fixed and stained with anti-V5 antibody to visualize Twinkle-APEX2 (green, Alexa Fluor 488) and neutravidin-Alexa Fluor 647 (red) to visualize biotinylated proteins. DIC, differential interference contrast. Scale bars, 10 μm. (F) Western blot analysis of biotinylated proteins from HEK 293T cells stably expressing the indicated APEX2 fusion construct and labeled with BP as in (A). Left: lysates were run on SDS-PAGE and analyzed by Ponceau staining or streptavidin blotting. Right: total eluted protein after streptavidin bead enrichment, visualized by silver staining. Lanes 1 and 4 are negative controls with H₂O₂ or APEX2 omitted, respectively. Black arrowheads point to endogenously biotinylated proteins. Red arrowheads point to proteins preferentially biotinylated by Twinkle-APEX2 over mito-APEX2.

Figure 2. Acquisition and Analysis of Mitochondrial Nucleoid Proteomic Data

(A) Design of proteomic experiment. HEK293T cells stably expressing the indicated constructs were labeled with BP and H₂O₂ for 1 min, then lysed. Biotinylated proteins were enriched with streptavidin beads, then digested to peptides on-bead. Resulting peptides were chemically labeled with the indicated TMT labels. Samples were then combined as shown and analyzed by liquid chromatography and tandem MS. (B) Filtering of MS data to produce the final mitochondrial nucleoid proteome. Table shows the number of proteins remaining after each filtering step. The two replicate datasets were intersected to produce the final proteome. Scatterplots in Figure S2A show correlation between TMT ratios across replicates. (C) Left: mitochondrial nucleoid proteome subdivided by protein functional class. Asterisk indicates that two isoforms of this protein are present in the proteome. Nucleoid orphans (proteins not previously linked to the mitochondrial nucleoid) are shown in blue. Right: histograms that illustrate how Filter 1 and Filter 2 cutoffs were determined. True positives (i.e., known mitochondrial nucleoid proteins) are plotted in the green histogram, and potential false positives are plotted in red or gray histograms. Receiver-operating characteristic analysis (Figure S2B) was applied to determine the TMT ratio cutoff value that maximized true positives while minimizing false positives. (D) Graph of all proteins detected in Replicate 1, plotted by extent of biotinylation by Twinkle-APEX2 (x axis) and by relative proximity to Twinkle-APEX2 versus matrix-APEX2 (y axis). Known nucleoid proteins are colored green and non-mitochondrial proteins are colored red. Cutoffs determined in (C) and applied in (B) are shown as dashed blue lines. All five green dots below the horizontal dashed blue line are known to be dual-localized to the nucleoid and elsewhere in the mitochondrial matrix. (E) Mitochondrial (left) and nucleoid (right) specificity of the proteomic list. Further details are provided in Table S1. (F) Schematic showing well-established proteins related to the mitochondrial nucleoid, mitochondrial RNA processing, and protein translation. Proteins are colored according to their detection or lack of detection in our proteome. Red and blue rectangles enclose the minimal proteins sets required for reconstitution of mtDNA replication and transcription in vitro, respectively (Korhonen et al., 2004; Litonin et al., 2010). (G) Western blot detection of nucleoid orphan NDUFS6 in purified nucleoids. HEK 293T cells expressing TFAM-V5 or mito-GFP-V5 were treated with formaldehyde, and crosslinked mitochondrial nucleoids were immunoprecipitated using anti-V5 antibody. Nucleoid protein MGME1 and OXPHOS protein NDUFA8 were also detected as positive and negative controls, respectively. Asterisk indicates the light chain of immunoglobulin G. (H) Co-localization of endogenous C7ORF55 with mtDNA. Left: endogenous C7ORF55 (green, Alexa Fluor 488) was stained in HeLa cells together with mtDNA (red, Alexa Fluor 405). Mitochondria were visualized using MitoTracker (blue) labeling. Quantitation of MitoTracker and C7ORF55 overlap with mtDNA from ≥3 cells is shown as mean ± SD beneath the images. Figure S2C shows expanded view of the entire cell. Scale bar, 5 μm. Right: pixel intensity plot of the dashed line in image on the left.

Figure 3. Expression of the ND3 Subunit of Complex I Is Specifically Regulated by FASTKD1

(A) Fluorescence imaging of FASTKD1-V5. COS-7 cells expressing the indicated constructs and mito-BFP (blue) were fixed and stained with anti-V5 antibody (green, Alexa Fluor 488) and anti-DNA antibody to visualize mtDNA (red, Alexa Fluor 647). Quantitation of V5 overlap with mtDNA from ≥3 cells is shown as mean ± SD on the right. TFAM-V5 and ATPAF2-V5 are positive and negative controls for nucleoid proteins, respectively. Scale bars, 10 μm. (B) Co-localization of FASTKD1-V5 with mtDNA. Top: expanded view of the boxed region in (A). Scale bars, 10 μm. Bottom: pixel intensity plot of the dashed line shown in images above. Pearson’s correlation coefficients for mitoBFP/DNA and V5/DNA are shown on the right. (C) Relative abundance of FASTKD1 mRNA and each of the 15 other mitochondrial mRNAs and rRNAs encoded by the mitochondrial genome, in FASTKD1 shRNA knockdown cells versus controls. qRT-PCR data are shown as mean ± SEM (n = 3). Data were analyzed using a two-tailed Student’s t test, with *p < 0.05, **p < 0.01, and ***p < 0.001; n.s., not significant (p > 0.05). (D) Relative mtDNA copy number in control shRNA knockdown cells compared with FASTKD1 shRNA knockdown cells, measured by qRT-PCR. Data are shown as mean ± SEM (n = 3). Data were analyzed using a two-tailed Student’s t test, with n.s., not significant (p > 0.05). (E) Western blot showing expression of complex I–V markers in FASTKD1 shRNA knockdown cells compared with control shRNA knockdown cells. Three different volumes of each cell lysate were loaded as indicated. Data from a replicate experiment shown in Figure S3C. Quantitation of complex I and II marker proteins (NDUFS6, NDUFB8, and SDHB) is shown on the right. Values are normalized to actin band intensities. Data are shown as mean ± SEM (n = 3) and analyzed using a two-tailed Student’s t test, with *p < 0.05; n.s., not significant (p > 0.05). (F) Relative respiratory complex I and complex II activities in control shRNA knockdown cells compared with FASTKD1 shRNA knockdown cells, measured by colorimetric assays. Equal amounts of lysate were loaded for each sample. Data are shown as mean ± SEM (n = 3) and analyzed using a two-tailed Student’s t test, with **p < 0.01; n.s., not significant (p > 0.05).