Rapid Affinity Purification of Tagged Plant Mitochondria (Mito-AP) for Metabolome and Proteome Analyses

A focused analysis of the Arabidopsis mitochondrial proteome and key metabolites is facilitated by a rapid tag-based affinity isolation of the whole organelle from small amounts of plant material.

Abstract

The isolation of organelles facilitates the focused analysis of subcellular protein and metabolite pools. Here we present a technique for the affinity purification of plant mitochondria (Mito-AP). The stable ectopic expression of a mitochondrial outer membrane protein fused to a GFP:Strep tag in Arabidopsis (Arabidopsis thaliana) exclusively decorates mitochondria, enabling their selective affinity purification using magnetic beads coated with Strep-Tactin. With Mito-AP, intact mitochondria from 0.5 g plant material were highly enriched in 30–60 min, considerably faster than with conventional gradient centrifugation. Combining gradient centrifugation and Mito-AP techniques resulted in high purity of >90% mitochondrial proteins in the lysate. Mito-AP supports mitochondrial proteome analysis by shotgun proteomics. The relative abundances of proteins from distinct mitochondrial isolation methods were correlated. A cluster of 619 proteins was consistently enriched by all methods. Among these were several proteins that lack subcellular localization data or that are currently assigned to other compartments. Mito-AP is also compatible with mitochondrial metabolome analysis by triple-quadrupole and orbitrap mass spectrometry. Mito-AP preparations showed a strong enrichment with typical mitochondrial lipids like cardiolipins and demonstrated the presence of several ubiquinones in Arabidopsis mitochondria. Affinity purification of organelles is a powerful tool for reaching higher spatial and temporal resolution for the analysis of metabolomic and proteomic dynamics within subcellular compartments. Mito-AP is small scale, rapid, economic, and potentially applicable to any organelle or to organelle subpopulations.

One challenge in science is to shift the scale in which observations can be made. From tissues via cells to compartments within a cell (e.g. organelles), the scale of scientific observation becomes progressively smaller, and the resolution must increase drastically to allow ever deeper insights into the details of biological processes.

Mitochondria are essential for cellular metabolism, because they are the major ATP-exporting organelles in plants, harbor central biochemical pathways such as the citrate cycle and the respiratory chain, represent a central hub of amino acid metabolism, and support photosynthesis and photorespiration. Objectives in plant mitochondria research are inter alia to create an inventory list of proteins and metabolites residing in these organelles as well as to investigate the dynamics of these molecules, for example upon exposure to different treatments and stresses. To address these objectives, the enrichment of mitochondria relative to other components of the cell is required. The reduction of complexity concomitant with purification also fosters the detection of proteins and metabolites with low abundances, because background and interfering matrix are removed prior to analysis.

To accomplish the isolation of mitochondria from leaf material, centrifugation-based methods have been developed that achieve a high degree of purity (Day et al., 1985; Keech et al., 2005). Nevertheless, a full overview of all mitochondrial proteins is still needed. To generate an inventory of proteins from a particular compartment, proteomic profiling of samples differentially enriched in this compartment was shown to be a powerful technique (Eubel et al., 2007; Kraner et al., 2017). An even greater challenge is the analysis of subcellular metabolite pools, for example the mitochondrial pools, in particular if the same metabolites are present in several compartments, maybe even in grossly different amounts. Highly purified organelles (mitochondria) are needed to ensure that detected metabolites are truly derived from this organelle and not from other copurified compartments.

To capture the mitochondria as close as possible to their in vivo state, the swiftness of the isolation procedure is even more important than the purity. After removing the mitochondria from their cellular context during cell rupture, posttranslational protein modifications are expected to change quickly within the organelle. Steady-state concentrations of metabolites will also be altered by metabolite conversion, degradation, or leakage from the mitochondria (Roberts et al., 1997; Agius et al., 2001).

Currently, the state-of-the-art technology for large scale mitochondrial isolation is a combination of differential and density gradient centrifugation (DGC) of extracts from cell or callus cultures (Klein et al., 1998) or green leaves (Day et al., 1985; Keech et al., 2005). Although this technique provides comparatively high amounts of mitochondria with reasonably high purity, the procedure is laborious and time-consuming, and it requires a large amount of starting material. A centrifugation-based method on a smaller scale is less time consuming, but yield and purity of the obtained mitochondrial preparation were not reported (Boutry et al., 1984). Following isolation by DGC, mitochondria can be further purified by free-flow electrophoresis using the surface charge on the organelles as an alternative handle for differential isolation (Eubel et al., 2007). Additionally, a fractionation technique using rapid filtration has been employed to swiftly separate plant organelles, including mitochondria, for metabolite analysis, thus minimizing leakage or conversion of metabolites (Lilley et al., 1982; Gardeström and Wigge, 1988). This method requires specialized equipment and the preparation of protoplasts prior to isolation, which might have an undesired impact on the cells. Nonaqueous fractionation is another method for the separation of subcellular metabolite pools and proteins (Gerhardt and Heldt, 1984; Arrivault et al., 2014). This technique has the advantage of rapidly quenching biological activities in the material prior to fractionation, allowing the assessment of metabolites and proteins close to their native state. Whether nonaqueous fractionation can reproducibly separate cytosolic from mitochondrial proteins and metabolites is still under debate (Arrivault et al., 2014; Fürtauer et al., 2019).

Alternative approaches for mitochondria isolation have been developed more recently that do not rely on the general physicochemical characteristics of the organelle, but rather use affinity purification techniques targeting specific proteins present on the mitochondrial surface. An antibody coupled to magnetic beads directed against TOM22, a mitochondrial import receptor subunit in the outer membrane, allowed the isolation of native mitochondria from mammalian cells (Hornig-Do et al., 2009; Afanasyeva et al., 2018). Recently, an innovative approach employing a recombinant protein for the isolation of mitochondria from mammalian cells with affinity purification was established and used to quantify metabolites and proteins (Chen et al., 2016). The key feature of this technology is the heterologous expression of a chimeric protein with a C-terminal domain integrated into the outer mitochondrial membrane and an N-terminal cytosolic domain consisting of a GFP for microscopic localization and a hemagglutinin-tag (HA) for immunoprecipitation. Employing magnetic beads coupled to anti-HA antibodies, mitochondria decorated with this fusion protein can be affinity purified.

Here, we established an affinity purification scheme for mitochondria from Arabidopsis (Arabidopsis thaliana), which we call Mito-AP. Plants pose several unique challenges for Mito-AP, because they possess chloroplasts, which need to be selected against during purification, and because plant extracts are a difficult matrix for any purification method. Mito-AP yields highly enriched mitochondria, requires only a fraction of the starting material typically used for conventional mitochondrial isolation, is scalable, and can be performed considerably faster than the classical centrifugation-based methods. The protocol described here introduces several innovations, which substantially lower the cost of this technique and allow the nondestructive removal of the mitochondria from the magnetic beads.

RESULTS

The Locus At1g55450 Encodes a Suitable Adapter Protein for Mito-AP

The expression of a recombinant chimeric protein on the mitochondrial surface enabling affinity purification is a key feature of Mito-AP. The C-terminal domain of such a protein should be integrated into the outer mitochondrial membrane and the N-terminal domain, including a GFP and an affinity tag, should face the cytosol as a handle for the affinity purification employing magnetic beads (Fig. 1). In contrast to Chen et al. (2016), who used a similar approach for HA-tagged mammalian mitochondria employing a magnetic affinity resin decorated with anti-HA antibodies, we aimed to use the Strep-tag/Strep-Tactin interaction to reduce procedure costs and to potentially allow for mild affinity elution. The Twin-Strep tag (IBA Lifesciences) was chosen, because it supposedly binds more strongly to Strep-Tactin than the single Strep tag.

Figure 1.

Workflow of Mito-AP. An ectopically expressed protein, which integrates into the outer mitochondrial membrane, fused with an N-terminal Strep-tag and a GFP-tag facing the cytosol can be used as an anchor for affinity purification of mitochondria with Strep-Tactin-coated magnetic beads.

We did not employ the chimeric protein described by Chen et al. (2016), because plant cells have an elaborate organelle targeting system preventing the misimport of mitochondrial proteins into the chloroplast and vice versa. It therefore cannot be excluded that the mammalian adapter protein may be partially mistargeted to chloroplasts. In fact, plants ectopically producing mammalian mitochondrial proteins sometimes erroneously direct these proteins into chloroplasts (Maggio et al., 2007). We performed a literature survey to identify plant proteins that are specifically integrated with their C-terminal domain in the outer mitochondrial membrane and feature an N-terminal domain facing the cytosol. Three candidate proteins belonging to the class of tail-anchored proteins were identified: (1) Cb5-d described by Maggio et al. (2007); (2) a putative methyltransferase encoded at locus At1g55450 and described by Marty et al. (2014); and (3) TOM22-V encoded at locus At5g43970.

In the case of Cb5-d, the C-terminal sequence conferring mitochondrial localization has been well defined (Hwang et al., 2004). Therefore, this domain was tested, whereas for the other candidates, full-length proteins were used for the construction of fusion proteins. The expression of the constructs and the subcellular location of the three resulting N-terminal Twin-Strep-tag-GFP fusion variants were assessed in transient assays performed in Nicotiana benthamiana. Despite the use of a strong 35S promoter for the respective constructs, Cb5-d and TOM22-V expression were barely detectable by confocal microscopy. By contrast, a clear and strong signal consistent with a mitochondrial location was observed for the protein belonging to the S-adenosyl-l-Met-dependent methyltransferase superfamily encoded at At1g55450. We selected this protein, herein referred to as ADAPTER, for our approach (Supplemental Fig. S1).

A homozygous transgenic Arabidopsis line with intermediate expression (line 11) of the 35S:Twin-Strep-tag:GFP:ADAPTER construct was generated. Confocal microscopy analysis suggested that the mitochondrial membrane was exclusively decorated by this chimeric protein, whereas signals from other organelles, in particular from chloroplasts, were not observed (Fig. 2A). To verify this localization, we created Arabidopsis plants expressing the 35S:Twin-Strep-tag:GFP:ADAPTER construct as well as a construct for mCherry directed to the mitochondria by the ScCoxIV peptide fused to the N terminus (Maarse et al., 1984). In these plants (line 1), GFP fluorescence from the ADAPTER:GFP fusion surrounded the mCherry signal located in the mitochondrial matrix (Fig. 2B).

Figure 2.

Subcellular localization of the Twin-Strep-tag:GFP:ADAPTER protein in stable transformed Arabidopsis. A, Confocal fluorescence microscopy images of the vascular tissue in leaves expressing the 35S:Twin-Strep-tag:GFP:ADAPTER construct. Shown are GFP (left) and autofluorescence (middle) of chloroplasts, and an overlay of both channels (right). B, Images from leaf vascular tissue coexpressing 35S:Twin-Strep-tag:GFP:ADAPTER (GFP, left) and 35S:ScCoxIV:mCherry (mCherry, middle), and an overlay of both channels (right). Scale bars = 8 μm.

Continuous GFP fluorescence from the whole rosette was observed in the Twin-Strep-tag:GFP:ADAPTER line 11, which was selected for all further experiments (Fig. 3A). The highly uniform expression in this line indicates that gene silencing did not occur, which would have resulted in individuals with patchy fluorescence. Silencing was probably prevented by the use of a GFP gene with an intron and by selecting a line with only moderate fluorescence.

Figure 3.

Comparison of growth and GFP fluorescence between the wild type (Col-0) and a stable transformed plant carrying the 35S:Twin-Strep-tag:GFP:ADAPTER construct. A, Photographs (top) and GFP fluorescence images (bottom) of wild-type (left) and transgenic plants (right) 21 d after imbibition (dai). Scale bars = 0.5 cm. B, Growth comparison of wild-type (wt) and transgenic (ADAPTER) plants at 35 dai (scale bar = 3.5 cm) and during development at 9 dai (C), 16 dai (D), and 21 dai (E). Pictures are representative of three plants within each experimental group. Scale bars = 1 cm (C–E).

Ectopic production of the ADAPTER fusion protein did not change the normal physiological appearance of the plant at any stage during development (Fig. 3, B–E). A recent study demonstrated that a short photoperiod enhances growth phenotypes caused by mutations in a mitochondrial protein (Pétriacq et al., 2017). Therefore, we asked the question whether a potential impact of the ADAPTER fusion protein on mitochondrial function is revealed under short-day conditions. Leaf surface and weight did not differ significantly between the wild-type and the Twin-Strep-tag:GFP:ADAPTER plants (Supplemental Fig. S2, A and B), suggesting that ectopic production of the ADAPTER fusion protein does not interfere with mitochondrial function. Additionally, we did not observe any effect on the root length of seedlings and the germination rate (Supplemental Fig. S2, C and D). Therefore, mitochondria isolated from the transgenic lines are likely representative of mitochondria from the wild type.

Outer Mitochondrial Membranes Can Be Isolated Together with the Mitoplast by Mito-AP

Stable homozygous 35S:Twin-Strep-tag:GFP:ADAPTER plants also possessing a transgene expressing the ScCoxIV:mCherry protein (line 1) were used for affinity purification of mitochondria with commercially available magnetic beads carrying the Strep-Tactin protein (IBA Lifesciences) on their surface. The beads were washed and then analyzed by confocal microscopy. Circular fluorescent structures were observed (Fig. 4) that were not detected on control beads incubated with extracts from wild-type plants. The fluorescence had the spectral emission characteristics of mCherry, suggesting that (1) mitochondria from total cell extracts had been isolated; and (2) these mitochondria maintained their integrity, still confining the mCherry protein to the mitochondrial matrix (Fig. 4). This analysis does not provide a quantitative assessment of mitochondrial integrity, and the presence of outer membranes without a mitoplast or otherwise damaged mitochondria cannot be ruled out (for a more quantitative assessment of intact mitochondria, see the last paragraph of the “Results” section).

Figure 4.

Colocalization of markers for mitochondrial outer membrane and matrix after affinity purification. Representative confocal fluorescence microscopy images of magnetic bead clusters with bound mitochondria originating from plants coexpressing 35S:Twin-Strep-tag:GFP:ADAPTER (GFP, left) and 35S:ScCoxIV:mCherry (mCherry, middle), and an overlay of both channels (right). Scale bars = 10 μm.

For extracting and washing the beads, an ammonium bicarbonate/sodium chloride buffer previously suggested to be advantageous for downstream liquid chromatography-mass spectrometry (LC-MS) applications with mitochondria isolated from mammalian cells was used (Chen et al., 2016). To ensure that this buffer does not have a negative impact on mitochondrial integrity, we compared it to the mannitol buffer employed for the isolation of plant mitochondria by DGC (Klein et al., 1998). As a control, mitochondria were also exposed to pure water, resulting in hypoosmotic shock and mitochondrial rupture. Mitochondria were isolated using DGC, pelleted by centrifugation, and resuspended in either mannitol (original buffer from the DGC protocol), potassium chloride/bicarbonate, sodium chloride/bicarbonate, or water, and then incubated for 20 min on ice. The preparations were centrifuged, and both the pellet and the supernatant were analyzed by SDS gel electrophoresis. A high amount of protein in the supernatant was only observed from mitochondrial preparations that had been incubated in water (Supplemental Fig. S3), suggesting that mitochondrial integrity was compromised under these conditions. By contrast, all other mitochondrial treatments resulted in only minor amounts of protein in the supernatants. Since there was no difference between mitochondrial preparations exposed to mannitol or sodium chloride/bicarbonate buffer, we concluded that the sodium chloride/bicarbonate buffer is suitable for isolation of plant mitochondria.

The Surface Area of the Isolation Matrix Rather Than Its Loading with Strep-Tactin Limits the Yield of Mitochondria

Initial experiments were conducted with commercially available magnetic beads of 1-μm diameter decorated with Strep-Tactin (IBA Lifesciences), which were developed for the isolation of entire cells. Although the 1-μm beads are suitable for mitochondrial isolation in principle (Supplemental Fig. S4), it might be advantageous to use smaller beads, which have been suggested to increase the yield of mitochondria (Chen et al., 2016). However, Strep-Tactin-coated nanoparticles are not commercially available. Therefore, we generated Strep-Tactin-coated iron oxide particles in house by chemically linking Strep-Tactin (IBA Lifesciences) to iron oxide beads (Chemicell). When testing the separation of magnetic beads of different sizes from the liquid phase with an external magnet, we observed that beads with a 100-nm diameter were more difficult to separate than larger beads with a 200-nm diameter.

Magnetic beads are available either coated with starch, whose hydroxyl groups can be activated with cyanogen bromide, or coated with a sugar polymer possessing carboxyl groups, which can be activated with carbodiimide. Both activated groups react with the amino groups of Strep-Tactin, allowing coupling of the protein to the matrix. We performed both coupling regimes and compared the performance of the resulting matrices. With 200-nm starch-coated beads, a higher purity of mitochondria was obtained than with the other matrices. Purity was judged by the relative enrichment of proteins predicted to be located in mitochondria (SUBA4 database; Hooper et al., 2017). Proteins were identified and quantified using shotgun proteomics, where protein abundance was determined by a label-free quantification algorithm (Schwanhäusser et al., 2011) providing intensity-based absolute quantification (iBAQ) values (Supplemental Fig. S5). Compared to the commercial 1-μm beads, the 200-nm starch-coated beads performed only slightly better in terms of enrichment of mitochondrial proteins (Supplemental Fig. S5). However, with equal bead volumes used in the purification, a 3-fold higher protein yield in the samples derived from the 200-nm material was obtained, thus confirming the initial hypothesis that higher yields can be obtained with smaller beads (Supplemental Fig. S6).

We also quantified the amount of Strep-Tactin coupled to the surface of equal volumes of the commercial 1-μm and the in-house-made 200-nm matrices and observed ∼20-fold more Strep-Tactin on the 1-μm beads (Supplemental Fig. S7). It appears that for isolation of mitochondria, the 5-fold greater surface area of the 200-nm beads is more important for the yield than is the absolute number of Strep-Tactin moieties on the surface.

Mitochondria Isolated by Affinity Purification Are Strongly Enriched

Next, we compared mitochondria preparations derived from Mito-AP to preparations made using DGC, currently accepted as the gold standard for mitochondria isolation from plants. Both methods were performed with 5-week-old 35S:Twin-Strep-tag:GFP:ADAPTER plants grown under short-day conditions, and both preparations were repeated on three different days. We used 220-fold more plant material for DGC than for Mito-AP (110 g versus 0.5 g; Supplemental Fig. S8, A and B). A similar or even slightly better relative protein yield was obtained with Mito-AP (Supplemental Fig. S8C), but far less total protein was purified due to the lower amount of starting material used.

For the Mito-AP protocol, mitochondria bound to the magnetic beads were further processed in two different ways. In method 1, the beads were boiled in SDS loading buffer for elution of the proteins. In method 2, an affinity elution step with biotin was employed to potentially increase purity, because unspecific proteins bound to the bead surface will not be eluted and can be removed together with the affinity matrix by centrifugation. However, we observed that elution with biotin was not fully quantitative (Supplemental Fig. S9). Therefore, we supported the biotin elution by adding proteinase K to the elution buffer, subjecting all proteins outside of the mitochondria to proteolytic degradation, including the Twin-Strep-tag:GFP:ADAPTER fusion protein connecting the mitochondria to the beads as well as any remaining unspecific protein contaminants. We hypothesized that this step might improve elution and increase the enrichment of mitochondrial proteins in the preparation, but that proteins residing in the outer mitochondrial membrane might be lost.

As expected, most proteins in the crude extracts originated from plastids. The abundance of mitochondrial proteins was strongly increased by DGC as well as by Mito-AP (Fig. 5; Supplemental Table S1). Consistent with our expectation, mitochondrial proteins were not as highly enriched by method 1 as by method 2, which led to >70% of iBAQs originating from mitochondrial proteins. The ADAPTER protein, known to reside on the mitochondrial surface, was only detected in the sample derived from method 1 but not method 2, confirming the efficiency of proteinase K digestion. The addition of proteinase K improved the yield 1.7-fold (Supplemental Fig. S9). One can envisage that such a yield increase could also be achieved by upscaling the method without the proteinase K treatment, resulting in improved yield and purity. Most interestingly, mitochondrial proteins were enriched to a higher degree by Mito-AP (method 2) than by DGC (Fig. 5, A and B).

Figure 5.

Enrichment of mitochondria by different purification methods. Plants were grown under short-day conditions for 4 weeks and subjected to either differential centrifugation, DGC, Mito-AP, or a combination of DGC and Mito-AP (DGC + Mito-AP). iBAQ values for every identified protein were assigned to their respective categories (mitochondrion, plastid, peroxisome, cytosol, or other), referred to total iBAQs of the corresponding sample, and plotted as percentages. The categories were defined by the SUBAcon annotation provided by SUBA4. Categorized iBAQ values are shown from crude extract, samples after only differential centrifugation, and samples after DGC (A); from crude extracts and samples after Mito-AP (method 1, boil-off; method 2, elution with biotin and proteinase K; B); and from samples after DGC and after tandem purification with DGC followed by the two different Mito-AP procedures (C). DGC data in A and C are the same. Error bars represent the sd (n = 3 biological replicates).

However, we had obtained a higher level of purity with DGC in the past (Klodmann et al., 2011; Senkler et al., 2017; Rugen et al., 2019). Also, the variation in mitochondrial protein enrichment across the replicates was high at ∼8.5%, whereas the variation for Mito-AP replicates with both methods was considerably lower (3.3% for method 1 and 2.5% for method 2; Supplemental Fig. S10). Reasons for this are currently unknown. Most likely, the expression of the fusion protein has some impact on the sedimentation behavior of the mitochondria during centrifugation. Since the extraction of the mitochondrial band from the gradient is a manual process and subject to a certain degree of variation in terms of volume, position, and disturbance of the gradient, mitochondrial purity may suffer when the mitochondria of the transgenic line are extracted in the same way as previously performed for the wild type. Thus, minor adjustments to the DGC procedure are expected to yield organelle purities in the transgenic line that are comparable to those reported for wild-type mitochondria.

In a tandem purification experiment, we assessed whether purity might be further increased when mitochondria isolated via DGC are used as input for Mito-AP enrichment. With this approach, purities of >90% were achieved, suggesting that each method removes a different spectrum of contaminants (Fig. 5C).

Mito-AP can be performed considerably faster than DGC, requiring only about 30 min for method 1 and 60 min for method 2. The direct comparison of both methods shows that Mito-AP results in better purity than DGC and appears to be more reproducible. Mito-AP requires only comparatively small amounts of plant material and is quite economic when a Strep-Tactin affinity matrix is used. One Mito-AP employing 6.25 mg of Strep-Tactin-coated 200-nm beads and using 500 mg of Arabidopsis leaf material yields ∼10 μg of mitochondrial protein and costs about 10 euros in consumables. The Strep-Tactin matrix is ∼5-fold less expensive than comparable commercial beads coated with anti-HA antibody, which were employed by Chen et al. (2016) for the affinity purification of mammalian mitochondria.

Prediction of Mitochondrial Protein Localization by Correlation Analysis

Although many studies have contributed to a comprehensive inventory of the mitochondrial proteome, it is an ongoing challenge to compile a complete set of mitochondrial proteins. We hypothesize that this challenge can be addressed by monitoring and correlating the relative abundances of proteins in samples from distinct experimental approaches for enriching mitochondria. This study offered an opportunity to test this idea, because several different strategies were employed for generating samples enriched in mitochondria. In such samples, the relative abundances of proteins known to be located in mitochondria will strongly correlate, because their abundance directly reflects the ratio of mitochondria to other cellular components. By contrast, the relative abundance of a contaminant protein will depend on whether a particular enrichment strategy is suitable for its removal, and therefore, the relative abundance of contaminants may correlate to each other but not to proteins truly localized in mitochondria. Thus, correlation can be used as a tool to predict mitochondrial localization. Proteins that cluster with known mitochondrial proteins in such an analysis are likely to reside in mitochondria as well.

Because the combination of DGC and Mito-AP resulted in a higher mitochondrial purity (Fig. 5C) than any of these techniques alone, each method appears to remove different contaminants. Also, the two different versions of Mito-AP have distinct protein enrichment profiles. To increase the coverage of both mitochondrial and contaminating proteins, we employed a longer (4-h) liquid chromatography separation prior to shotgun MS analysis for all three biological replicates of (1) DGC alone, (2) Mito-AP method 1, (3) Mito-AP method 2, (4) DGC combined with Mito-AP method 1, and (5) DGC combined with Mito-AP method 2.

For the individual replicates of each method, the iBAQ value of each protein was normalized to the total number of iBAQs detected in the respective sample. For each of the five methods, a relative mean iBAQ value for each protein was calculated from the three replicates. With this data, a correlation matrix using Pearson’s correlation coefficient was built. All correlations above a cutoff of 0.99 were displayed as a weighted network placing proteins with stronger correlation closer together.

The correlation analysis revealed one cluster containing 619 proteins that is highly enriched in proteins predicted by SUBAcon (Hooper et al., 2017) to be localized in mitochondria (517 of 619 [82%]; Fig. 6; Supplemental Table S2). Of these 517 proteins assigned to mitochondria, 491 (shown as black triangles in the cluster) have available supporting MS or fluorescence microscopy data. For the remaining 26 proteins, experimental evidence of their mitochondrial localization is provided here (Fig. 6, gray squares). For the other 102 proteins in the above-mentioned cluster of 619 proteins (18%), mitochondrial localization is not predicted according to SUBAcon. For 19 of these, no experimental data are available (Fig. 6, red circles). Therefore, this study provides evidence for mitochondrial location of these proteins despite a different SUBAcon prediction. For 83 of the 102 proteins not assigned to mitochondria by SUBAcon, either MS or fluorescent microscopy data are available (Fig. 6, orange diamonds), and 62 of these proteins have at least once been found in mitochondria by either technique. The remaining 21 proteins are not predicted by SUBAcon to be in mitochondria, nor is there current experimental evidence indicating their presence in mitochondria. Further analysis of these 21 proteins shows that only five were demonstrated by fluorescence microscopy to reside in another compartment. Although we cannot completely resolve this discrepancy, it is possible that in some instances protein variants arising from alternative transcripts might be differentially located. This might also be one possible explanation for the remaining 16 proteins, for which localization information is based on MS analysis of isolated cellular compartments. It is worth noting that for half of these proteins, the experimental evidence is limited to a single MS study.

Figure 6.

Correlation analysis of protein abundances in samples from DGC, the two Mito-AP methods, and tandem DGC-Mito-AP purifications. The normalized iBAQ values for all identified proteins from DGC and Mito-AP methods 1 and 2 as well as from tandem DGC + Mito-AP (methods 1 and 2) were determined. An average of the three biological replicates from each method was used to calculate Pearson’s correlation coefficients for all possible combinations of two proteins. The correlations are displayed as an interaction graph using a cutoff value of 0.99. Proteins are nodes and the length of an edge is negatively correlated with the strength of the correlation. Black triangles and gray squares, proteins classified by SUBAcon as being located in mitochondria; gray squares, locations predicted only in silico, without experimental support (by MS or fluorescence microscopy [FM]); orange diamonds and red circles, proteins classified by SUBAcon as being located in other cellular compartments and not mitochondria; red circles, locations predicted only in silico, without experimental support; orange diamonds, experimental data are available. The same color scheme was used for the schematic below the interaction graph, representing the proportions of the respective protein groups.

In summary, in the mitochondrial cluster presented here, we have identified previously uncharacterized candidate proteins that exhibit mitochondrial localization as a starting point for further analysis and experimental assessment.

Mitochondrial Metabolites Can Be Detected and Quantified in Affinity-Purified Mitochondria

Our initial motivation to develop a purification protocol for mitochondria was to quantify nucleotide subpools in these organelles. Therefore, we tested whether nucleotides can be detected when mitochondria are enriched using method 1 and then directly extracted on the beads. Unfortunately, metabolites could hardly be detected in these extracts. The addition of labeled nucleotide standards suggested that metabolites interact with the magnetic beads, preventing their detection in the mass spectrometer. We reasoned that the removal of the magnetic beads before metabolite extraction might alleviate the problem. To this end, method 2 was developed, in which the mitochondria are eluted and the affinity matrix is removed prior to extraction. Extracts from method 2 were analyzed for NAD, NADH, AMP, ADP, ATP, NADP, and nicotinamide mononucleotide (NMN). For all these metabolites, robust signals were recorded that were several orders of magnitude stronger than signals from the negative control, where beads had been incubated with extracts from wild-type plants and treated according to method 2 (Fig. 7).

Figure 7.

Polar metabolites in mitochondria enriched by Mito-AP. Seedlings carrying the 35S:Twin-Strep-tag:GFP:ADAPTER construct (enriched mito.) or Col-0 (neg. control) were grown for 10 d in liquid media and processed with Mito-AP (method 2). Polar metabolites were identified and quantified with a triple quadrupole MS in these preparations. A, NMN, NADP, and ADP Glc. The ADP-Glc content was below the detection limit of 0.65 fg (see Supplemental Fig. S10). B, Adenylates and reduced and oxidized NADH, each shown as a pool in one column, to emphasize that pools of these metabolites can be measured but that the concentrations of individual compounds within those pools do not represent the concentrations in vivo. For the adenylates the column is separated into AMP (lower), ADP (middle), and ATP (upper), whereas for the nicotinamides, NAD is shown in the lower part of the column and NADH in the upper part. The values for ATP were too low to be visualized. Error bars represent the sd (n = 4 biological replicates, derived from independent Mito-APs of distinct samples grown in parallel). Significant differences (at P < 0.05) were determined with Student’s t test and marked with an asterisk. Calculation of the false discovery rate (FDR) was done using the two-stage linear step-up procedure described in Benjamini et al. (2006). The statistical analysis was done with a FDR of 1%. nd, Not detectable. Standard curves for different metabolite concentrations in water were used for absolute quantification.

The results demonstrate that with the protocol presented here, it is possible to obtain sufficient amounts of metabolites, aiding in the characterization of enriched mitochondria, as suggested before (Ikuma, 1970). However, from the NADH/NAD and ATP/ADP ratios, it is clear that the mitochondria lost activity during Mito-AP (Roberts et al., 1997, Agius et al., 2001). This strongly suggests that metabolite concentrations in vivo cannot be determined with Mito-AP, except in the case of stable metabolites, such as lipids, or for entire metabolite groups, such as the adenylates (Fig. 7).

Metabolites were also detected in the negative controls of the recently described Mito-AP from mammalian cells, but unfortunately the actual magnitude of this background noise was not shown (Chen et al., 2016). It is possible that this background arose from metabolites directly bound to the matrix, because Chen et al. (2016) extracted the metabolites from the mitochondria in the presence of the magnetic beads. We minimized such effects by removing the beads prior to extraction, but it cannot be fully excluded that a minor amount of matrix remained in our preparations, resulting in some background.

To assess whether there was any metabolite contamination of our mitochondrial preparations from intact plastids, we used ADP-Glc as a marker metabolite. It is generated by ADP-Glc pyrophosphorylase for starch biosynthesis, which normally occurs in the plastids. Under special circumstances, cytosolic starch biosynthesis might occur (Villand and Kleczkowski, 1994); however, the process has never been observed in mitochondria. Dilutions of ADP-Glc down to 0.67 fg on column were detected with a robust signal-to-noise ratio by our analytical platform (Supplemental Fig. S11), but ADP-Glc was not detected in any of the samples obtained with Mito-AP (Fig. 7), indicating that intact plastids were not present.

Besides adenylates, the lipid composition of enriched mitochondria was also analyzed, and the abundances of individual lipids were compared between the Mito-AP and the negative control derived from magnetic beads incubated with the wild-type extract. Several cardiolipins, known to occur in mitochondria (Schlame et al., 1993; Zhou et al., 2016), were exclusively detected in extracts of the Mito-AP and absent in the control. Two cardiolipins were also detected in the control, but were over three orders of magnitude more abundant in the Mito-AP extracts (Fig. 8; Supplemental Table S3). Ubiquinone (Q), a membrane-soluble electron carrier in the electron transport chain of mitochondria, was strongly enriched in samples containing mitochondria. Qs are classified according to the number of isoprenoid side-chain units, which usually rank from 6 (for yeast [Q6]) to 10 (for humans [Q10]). Interestingly, Q7, Q8, Q9, and Q10 were detected (Fig. 8), although it has been reported that Q9 is the sole form of Q in Arabidopsis (Hirooka et al., 2003; Yoshida et al., 2010; Liu and Lu, 2016). The roles of these additional Qs in plant mitochondria are currently unknown, but it is tempting to speculate that they may function in nonrespiratory pathways, similar to the involvement of human Q10 in signaling (Schmelzer et al., 2007). The strong enrichment of mitochondria by Mito-AP in combination with the sensitive detection by MS probably explains why several Qs could be detected here, thus highlighting that resolution is indeed gained by the focused analysis of isolated mitochondria. Previous studies on cardiolipins (Zhou et al., 2016) and Qs (Yoshida et al., 2010) had the advantage of employing a rapid quenching of the sample, abolishing any possibility for metabolite conversion. The cardiolipin species we observed could previously only be detected using a solid-phase extraction protocol (Zhou et al., 2016), suggesting that Mito-AP maybe useful for the initial detection of metabolites in samples of enriched mitochondria without need for further sample preparation. However, such findings are pending confirmation from samples that are more efficiently quenched than is possible with Mito-AP.

Figure 8.

Enrichment of the mitochondrial cardiolipins (CL) and Qs in mitochondria enriched by Mito-AP. Seedlings carrying the 35S:Twin-Strep-tag:GFP:ADAPTER construct (enriched mito.) or Col-0 (neg. control) were grown for 10 d in liquid media for Mito-AP (method 2). Samples were extracted and lipids were identified and quantified with an orbitrap MS. Error bars represent the sd (n = 5 biological replicates, derived from independent Mito-APs of distinct samples grown in parallel). Significant differences (at P < 0.05) were determined with Student’s t test and marked with an asterisk. Calculation of the FDR was done using the two-stage linear step-up procedure described in Benjamini et al. (2006). The statistical analysis was done with a FDR of 1%. nd, Not detectable.

Chlorophyll, associated with thylakoid membranes of the chloroplast, was also slightly enriched, by ∼2-fold, in Mito-AP samples compared to controls (Supplemental Table S3). When mitochondria were bound to the beads, we observed that the beads became adhesive to each other, probably caused by mitochondria interconnecting several beads. This network of beads may occasionally trap thylakoid fragments which cannot be removed by washing. In fact, we occasionally observed some red auto-fluorescence in our mitochondria preparations, but never corresponding to the size of an intact chloroplast (Supplemental Fig. S12).

Interestingly, galactolipids, which are associated with the chloroplast, were also enriched in the Mito-AP samples. For the majority of these lipid species, only a slight enrichment of 10- to 20-fold was observed, but in one case a 200-fold enrichment was obtained (Supplemental Table S3). As for the chlorophylls, the physical entrapment of plastidic fragments during Mito-AP might explain this observation. It is also conceivable that some of these galactolipids actually reside in the mitochondria, because transfer of galactolipids from plastids to mitochondria has been observed, particularly in conditions of phosphate starvation (Jouhet et al., 2004). In summary, this proof-of-concept study demonstrates the feasibility of metabolite quantification in mitochondria upon isolation using the Mito-AP protocol, which paves the way for further studies addressing problems of metabolite conversion during the isolation procedure.

Mitochondria Isolated by Mito-AP Have a Membrane Potential, Respire, and Show Respiratory Control

Previous studies on mitochondria isolation techniques reported that the isolated organelles remained physiologically active (Keech et al., 2005). We also asked the question whether mitochondria isolated by Mito-AP maintain their membrane potential and respiratory control. However, this could not be addressed with the sodium chloride extraction buffer, because it was chosen for its compatibility with downstream LC-MS measurements and its reported potential to suppress metabolic activity (Kong et al., 2018). Different buffers containing either mannitol or Suc as osmotic agents were used instead.

First, we tested whether mitochondria in intact roots of the 35S:Twin-Strep-tag:GFP:ADAPTER plants have membrane potential, which would indicate that the mitochondrial tag does not grossly disturb mitochondrial function or integrity. The fluorescent dye tetramethyl rhodamine methyl ester (TMRM) is only sequestered in mitochondria in the presence of a membrane potential (Brand and Nicholls, 2011; Schwarzländer et al., 2012). TMRM was found in mitochondria of the transgenic line as well as the wild type (Supplemental Fig. S13, A and B). The signal was quenched by adding carbonyl cyanide 3-chlorophenylhydrazone (CCCP), an uncoupling agent that abolishes the membrane potential. The same technique was applied to mitochondria isolated with the Mito-AP protocol using the modified extraction buffer containing 10 mm succinate. In these samples, the green fluorescence originating from the Twin-Strep-tag:GFP:ADAPTER protein coincided with the fluorescence of the TMRM dye in 73% of the cases. This rate dropped to 32% upon addition of the uncoupler CCCP (Supplemental Fig. S14). Moreover, the signal from these mitochondria was considerably weaker. These results suggest that >70% of the mitochondria enriched by Mito-AP are sufficiently intact to maintain a membrane potential.

Next, we assessed the respiration of mitochondria isolated by Mito-AP in a Seahorse Analyzer (Agilent; Rogers et al., 2011) and recorded the oxygen consumption rate (OCR) before and after addition of ADP, as well as in response to the inhibitors oligomycin and antimycin (Fig. 9). A basal respiration rate of 29 ± 3 nmol O₂ min⁻¹ μg⁻¹ protein was observed. Addition of ADP led to a 1.3-fold increase in OCR, which reverted to the basal level after addition of the ATP synthase inhibitor oligomycin. Furthermore, addition of antimycin, an inhibitor of complex III, reduced the OCR to background levels. These data demonstrate that mitochondria with respiratory activity can be isolated by Mito-AP, and that they maintain some degree of respiratory control, as indicated by the before-to-after OCR ratio upon addition of ADP or oligomycin (respiratory control ratio; Fig. 9).

Figure 9.

Activity of mitochondria isolated by Mito-AP from Arabidopsis. OCRs were measured for mitochondria bound to magnetic beads before (basal respiration) and after addition of ADP and in response to oligomycin A and antimycin A. In total, five replicates were analyzed for mitochondria bound to beads and three replicates for the negative control. The negative control represents the basal respiration of the empty beads. Error bars represent the sd. Statistical analysis was done using a one-way ANOVA with Tukey’s honestly significant difference (HSD) mean-separation test. Lowercase letters indicate significant differences (P < 0.05).

These results were confirmed by the morphological assessment of Mito-AP-isolated organelles as deduced by electron microscopy. Mitochondria display outer and inner membranes as well as an ultrastructure consistent with intact organelles isolated in previous studies (Supplemental Fig. S15; de Virville et al., 1998; Logan and Leaver, 2000; Eubel et al., 2007). In some instances, a pronounced gap between the inner and outer membrane can be observed, suggesting either some shrinkage of the mitoplasts or an inflation of the outer membrane. Reasons for this are currently unknown. We also observed grana stacks and other unidentified compartments consistent with plastidic fragments observed in confocal microscopy and the presence of nonmitochondrial proteins in the preparations (Fig. 5; Supplemental Fig. S12).

DISCUSSION

Here we demonstrate that plant mitochondria can be enriched from small amounts of plant material in sufficient amounts to support proteome and metabolome analyses using affinity purification. The application and further development of this technique offers the potential to increase the resolution of proteome and metabolome experiments. Because comparatively little plant material is needed for Mito-AP, the biological context from which the mitochondria are obtained can be well defined. It can be envisioned that mitochondrial isolation may be restricted to only a certain tissue or only tissues in defined developmental states, for example leaves of different ages. Mito-AP also offers the prospect to exclusively decorate mitochondria of certain cell types or following environmental stimuli. For this, the Twin-Strep:ADAPTER protein would need to be expressed under the control of a cell-type-specific promoter or a promoter induced by a particular stimulus. Previous studies have shown the power of promoter-reporter fusions, for example for the elucidation of the root cell type-specific transcriptomes (Moussaieff et al., 2013) and for the isolation of cell-specific mitochondria from Caenorhabditis elegans as well as mouse (Mus musculus; Ahier et al., 2018; Bayraktar et al., 2019). Yet another possibility is the transfer of the affinity purification principle to other compartments of plant cells, for example to peroxisomes or chloroplasts. Along this line, an immunoprecipitation protocol has been developed for the enrichment of plant secretory vesicles to analyze their proteomes (Heard et al., 2015).

The affinity purification protocol presented here has several features that will facilitate its adaptation and use by other laboratories. First, the construct was generated using the MoClo system (Patron et al., 2015), which is highly modular, allowing an independent exchange of different elements including the tags, the promoter, or the ADAPTER itself. Second, in contrast to DGC methods, no specialized equipment such as an ultracentrifuge is required to perform Mito-AP. Third, usage of the Strep-tag affinity system instead of a system based on antibodies (Chen et al., 2016) is cost effective. Mito-AP can be performed with a commercially available matrix already coated with Strep-Tactin. Alternatively, it is straightforward to attach Strep-Tactin to prepared iron oxide particles, further lowering the costs and increasing the yield (Supplemental Fig. S6).

Recently, a simple protocol was published for the economic production of iron oxide nanoparticles with carboxyl groups on the surface, called Bio on Magnetic Beads (Oberacker et al., 2019). We are currently investigating the suitability of these particles for Strep-Tactin loading. The availability of large quantities of affordable affinity matrix would permit an increase in the amount of beads used per preparation. This might improve the yield or accelerate the sample processing, because at high yields the magnetic separations do not need to be fully quantitative and can therefore be performed more swiftly. With shorter processing times, mitochondria could be isolated even closer to their native state. This is still a problem, because the ADP/ATP ratio (Fig. 6) showed that the mitochondria isolated by the current Mito-AP protocol are metabolically inactive (Stitt et al., 1982). An additional solution might be to prepare the mitochondria in the presence of buffers containing a broad selection of mitochondrial metabolite precursors suitable to maintain mitochondrial function during isolation. The successful reactivation of mitochondria with substrates for mitochondrial respiration was already demonstrated in this study (Fig. 9). In comparison to traditional isolation techniques, Mito-AP is performed in small volumes, which considerably reduces the costs of such an approach.

Conceptually, affinity elution from the matrix should increase the purity, because unspecifically bound contaminants will not be released and are removed together with the beads. Indeed, this was observed, because Mito-AP involving an elution step with biotin and proteinase K (method 2) resulted in higher purity than Mito-AP with method 1 (Fig. 5). However, it would be desirable to omit the proteinase K treatment, because it is time consuming and damages the mitochondrial surface proteins. Unfortunately, elution only with biotin is not efficient enough, probably because there are too many interactions between the Twin-Strep-decorated mitochondria and the Strep-Tactin on the matrix. The coating density of the matrix and the decoration density of the mitochondria should be further fine-tuned and optimized to develop a protocol that allows complete affinity elution. This might also reduce the number of beads connected to each other via mitochondria and thus limit the potential physical entrapment of contaminants in this network. Additionally, different bead surfaces should be tested with the aim to suppress the unspecific binding of metabolites, which would further reduce the background currently observed in the controls (Figs. 7 and 8). It is clear that metabolites can be interconverted in a very short period of time (Dietz, 2017) and that therefore the duration of the Mito-AP protocol is still too long to avoid such processes. However, the impact of mutants on the mitochondrial metabolome still can be studied by focusing on relative changes in metabolites rather than absolute amounts. Also, groups of metabolites that can be interconverted into each other can be combined and considered as pools. Alternatively, improved quenching techniques prior to the Mito-AP protocol may address this challenge. Furthermore, Mito-AP will allow the identification of low-abundance metabolites that cannot be detected in a crude extract (Fig. 8) because of ion suppression effects.

Plant mitochondria have been analyzed numerous times by proteomic approaches. However, definition of the complete set of proteins residing in this organelle and the discernment of these proteins from background and false positives is an ongoing struggle. Previous studies indicated an inconsistency between computational prediction for mitochondrial localization and experimental evidence from proteomic studies. Several nuclear-encoded proteins with a clear mitochondrial target peptide were not detected in mitochondria by proteomics (Heazlewood et al., 2005, 2007). To increase sensitivity and to eliminate more false positives, tandem purification strategies for mitochondria have been employed combining DGC with free-flow electrophoresis (Eubel et al., 2007). However, both methods are very time consuming and require specialized equipment and expertise. Here, we used a combination of DGC and Mito-AP and performed a correlation analysis for protein abundance integrating the data from all enrichment strategies. Irrespective of the technique used, true mitochondrial proteins should be enriched and form a closely correlated cluster. This was observed, and in addition to known mitochondrial proteins, previously uncharacterized candidate proteins were found in this cluster (Fig. 6). More detailed analyses are required to validate these mitochondrial protein candidates, which might provide clues about yet unknown processes taking place in plant mitochondria.

An important consideration is whether mitochondria isolated by Mito-AP are of the same quality as that described using centrifugation-based methods. Since the quantities of mitochondria isolated by Mito-AP are comparatively low, standard methods for quality control of mitochondria, for example an oxygen electrode assay to assess respiration, cannot be applied easily. Here, we demonstrate oxygen consumption of plant mitochondria using a Seahorse Analyzer, which is uniquely suited for minute sample amounts. Recently, the oxygen consumption of mammalian mitochondria bound to beads was observed using the same technique; however, in contrast to our study, the impact of ADP on respiration was not reported (Ahier et al., 2018), suggesting that the methodology for assessing isolated mitochondria with this technology is still in its infancy. In our hands, the method required optimization of the amount of mitochondria and the concentration of ADP used. Currently, the minimum assay temperature in the Seahorse analyzer cannot fall below 29°C, which differs from previously described procedures for assessing the activity of plant mitochondria. The respiration rates and the respiratory control we report here (Fig. 9) are lower than that described previously for Arabidopsis (Kerbler et al., 2019). However, it is possible that these differences can in part be explained by the current imperfections of the assessment technology, which needs further optimization.

Electron microscopy pictures suggested the presence of predominantly intact mitochondria with an outer and inner membrane in the Mito-AP preparation (Supplemental Fig. S15), but the presence of strongly fragmented mitochondria cannot be excluded. In line with this result, we demonstrated that 73% of the isolated mitochondria showed a membrane potential (Supplemental Fig. S14). Although no abnormal phenotypes were observed in the Twin-Strep-tag:GFP:ADAPTER plants, another explanation for a lower respiration rate/respiratory control is that expression of the Twin-Strep-tag:GFP:ADAPTER protein has an impact on mitochondrial functionality. This potential effect might be attenuated by further reducing the abundance of the protein with a weaker promoter or by removing the catalytic domain of the ADAPTER protein.

In summary, this study shows that Mito-AP allows rapid enrichment of plant mitochondria from small amounts of plant material. This work should, however, be regarded as a proof-of-concept pilot study that requires further experimentation to determine whether tagging the mitochondria interferes with their function, especially under stress conditions. Using this protocol, the quantification of mitochondrial metabolite subpools in the absence of any significant contamination was achieved, and previously uncharacterized candidate proteins for the complete mitochondrial proteome were identified. We envisage that affinity purification of organelles will increase the sensitivity of metabolome and proteome approaches. The potential of this method to target particular organelle subpopulations, in combination with the short processing time required for affinity isolation, will increase the spatial and temporal resolution for the investigation of plant organelles in the future.

MATERIALS AND METHODS

Plant Material and Cultivation

For proteomic analysis, Arabidopsis (Arabidopsis thaliana) plants were cultivated in soil (Steckmedium, Klasmann-Deilmann) in a climate chamber under short-day conditions (8 h light/16 h darkness, 22°C, 85 μmol s⁻¹ m⁻² light, and 65% humidity), as described in Senkler et al. (2017). Plants were harvested after 6 weeks. For metabolome analysis, plants were cultivated at 22°C in sterile 100-mL flasks (10 mg seeds per flask) in liquid culture (0.5 × Murashige and Skoog, 0.125% [w/v] MES, pH 5.8) in a shaker with an artificial light source emitting 45 μmol s⁻¹ m⁻² light. For the first 3 d, the shaker (New Brunswick Innova 42, Eppendorf) was set to 30 rpm and subsequently changed to 50 rpm. Plants were harvested after 10 d. For phenotypical characterization, plants were cultivated in 8-cm pots filled with soil (Steckmedium, Klasmann-Deilmann) in a climate chamber under long-day conditions (16 h light/8 h darkness, 22°C day, 20°C night, 100 μmol s⁻¹ m⁻² light, and 70% humidity). For leaf surface area and rosette fresh weight analysis, plants were cultivated in trays with soil (Steckmedium, Klasmann-Deilmann) under short-day conditions as described above. For root length and germination assays, plants were grown on modified Murashige and Skoog medium (3 mm CaCl₂, 1.5 mm MgSO₄, 1.25 mm KH₂PO₄, 18.7 mm KNO₃, 0.1 mm FeSO₄, 0.1 mm Na₂EDTA, 0.13 mm MnSO₄, 0.1 mm BO₃, 0.03 mm ZnSO₄, 1 μm Na₂MoO₄, 0.1 μm CuSO₄, 0.1 μm NiCl₂, 0.5 g L⁻¹ MES, and 8 g L⁻¹ phytoagar, pH adjusted to 5.7 with KOH) under long-day conditions as described above.

Subcellular Localization and Confocal Microscopy

For subcellular localization, the constructs were coexpressed in Nicotiana benthamiana leaves for 4 d and analyzed by confocal microscopy as described by Dahncke and Witte (2013). For confocal microscopy, the Leica True Confocal Scanner SP8 microscope equipped with an HC PL APO CS2 40× 1.10 water immersion objective (Leica Microsystems) was used. Acquired images were processed using Leica Application Suite Advanced Fluorescence software.

Magnetic Bead Preparation

The activation of the polysaccharide resins was described previously (Kohn and Wilchek, 1984) and coupling of Strep-Tactin to the magnetic beads was performed according to a protocol provided by Chemicell, with minor modifications. In detail, 400 μL uncoated magnetic beads (25 mg mL⁻¹) with a GlcA polymer matrix (fluidMAG ARA, 4115-5, Chemicell) were mixed with 150 μL freshly prepared activation buffer containing 10 mg 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide and mixed in a shaker for 10 min at room temperature (RT). Between the different steps, the beads were always pelleted using a magnetic separator. After activation, the beads were washed twice by suspending them in 1 mL water. For coating with Strep-Tactin, the bead pellet was resuspended in 250 μL water, 20 μL Strep-Tactin solution was added (5 mg mL⁻¹ in phosphate-buffered saline [PBS]; 2-1204-005, IBA Lifesciences), and the slurry was incubated for 2 h under constant shaking. The beads were washed three times with 1 mL water and stored in 400 μL 0.05% (w/v) sodium azide solution.

500 µL beads with a starch polymer matrix (25 mg mL⁻¹; Chemicell, fluidMAG –D, 4101-5) were washed once with 1 mL freshly prepared activation buffer containing 0.2 m sodium hydrogen carbonate (pH range 8.4 to 8.7), pelleted, and resuspended in 500 µL activation buffer. After addition of 100 µL 5 m cyanogen bromide in acetonitrile, the beads were mixed and incubated for 10 min on ice, then washed twice with 1 mL PBS buffer and resuspended in 500 µL PBS. To ensure a homogenous coating, the beads were placed for 2 min in an ultrasonication bath. For coating, 500 µL magnetic beads were incubated with 20 µL Strep-Tactin solution (5 mg mL⁻¹ in PBS; 2-1204-005, IBA Lifesciences) for 2 h under constant shaking. The beads were washed three times with 1 mL water and stored in 500 μL 0.05% (w/v) sodium azide solution.

Cloning

Cloning was performed using the MoClo system described by Engler et al. (2014). For this, several intermediate vectors in addition to already published vectors had to be created. The primers used are listed in Supplemental Table S4.

Turbo-GFP was amplified from pICSL50016 using P626 and P820 and cloned into pAGM1287, resulting in the vector pAGM1287_TurboGFP+Intron_without_Stop (V151). The TOM22-V fragment sequence was synthesized by Integrated DNA Technologies (IDT) and cloned into pAGM1301, resulting in the vector pAGM1301_level_1_TOM22-V (V152). The Cb5-d fragment sequence was synthesized by IDT and cloned into pAGM1301, resulting in the vector pAGM1301_level_0_Cb5-d (V153). The codon-optimized DNA sequence coding for the Twin-Strep-tag (28 amino acids: WSHPQFEK-GGGSGGGSGG-SA-WSHPQFEK) was synthesized by IDT and cloned into pAGM1276, resulting in the vector pAGM1276_Twin-Strep (V154). Amplification of the ADAPTER (encoded at locus At1g55450) using complementary DNA from Arabidopsis was performed by using the primers P1023 + P1024 and P1025 + P1026, integrating a mutation to remove a BbsI recognition site without changing the amino acid sequence. PCR products were digested, ligated, and cloned into pAGM1301, resulting in the vector pAGM1301_level_0_ADAPTER (V159).

The level_1 vector containing a phosphinothricin resistance for plant selection was generated by combining pICH87633, pICH43844, and pICH41421 with the recipient vector pICH47742 in a digestion/restriction reaction with BsaI resulting in pICH47742_Basta_pos2_fwd (V166). For cloning the level_1 ScCoxIV:mCherry vector, pICH45089, pAGM1482, pICSL80007, and pICH41432 were digested/ligated together with the recipient vector pICH47742, resulting in pICH47742 _pos2_fwd_ ScCoxIV:mCherry (V168). For cloning the level_1 vector containing 35S:Twin-Strep-tag:GFP:TOM22-V, pICH41373, pAGT707, V154, V151, V152, and pICH41421 were combined with the recipient vector pICH47732 in a digestion/restriction reaction with BsaI resulting in pICH47732_pos1_fwd_35S:Twin-Strep-tag:GFP:TOM22-V (H405). For cloning the level_1 vector containing 35S:Twin-Strep-tag:GFP:Cb5-d, pICH41373, pAGT707, V154, V151, V153, and pICH41421 and the recipient vector pICH47732 were combined in a digestion/restriction reaction with BsaI resulting in pICH47732_pos1_fwd_35S:Twin-Strep-tag:GFP:Cb5-d (H407). For cloning the level_1 vector containing the 35S:Twin-Strep-tag:GFP:ADAPTER, pICH41373, pAGT707, V154, V151, V159, and pICH41421 were combined with the recipient vector pICH47732 in a digestion/restriction reaction with BsaI resulting in pICH47732_pos1_fwd_35S:Twin-Strep-tag:GFP:ADAPTER (H433).

For the level_2 vector containing 35S:Twin-Strep-tag:GFP:ADAPTER and a Basta resistance gene, H433, V166, and pICH41744 were combined with the recipient vector pAGM4723 in a digestion/restriction reaction with BbsI resulting in the vector pAGM4723_35S:Twin-Strep-tag:GFP:ADAPTER_Basta (H500). The level_2 vector pAGM4723_35S:Twin-Strep-tag:GFP:ADAPTER_ScCoxIV:mCherry (H501) was cloned by combining H433, pICH41744, V168, and pAGM4723 in a digestion/restriction reaction with BpiI. All materials (plants and vectors) are made available upon request.

Leaf Area Quantification

To quantify the leaf area, RGB (red/green/blue) pictures were taken from the same distance every day and loaded into ImageJ. To create a binary image, the pictures colors were split into separate channels (Image → color → split channels). The green channel image was then further processed to reduce background signal by minimum filtering with a radius of 50 pixels followed by maximum filtering with a radius of 50 pixels and a contrast enhancement with a saturation of 0.01%. Subsequently a binary picture was created by setting an auto-threshold. The binary picture was modified (euclidian distance map binary options → close [20 iterations]) and the region of interest was projected onto the scaled original picture to retrieve the leaf area of the respective sample.

Density Gradient Isolation of Mitochondria

A modified version of the protocol of Keech et al. (2005) was used. All steps were carried out at 4°C. Arabidopsis rosette leaves were ground for 10 min in disruption buffer using a mortar and pestle (0.3 m Suc, 60 mm TES, pH 8 [KOH], 25 mm sodium pyrophosphate, 10 mm KH₂PO₄, 2 mm EDTA, 1 mm Gly, 1% [w/v] PVP 40, 1% [w/v] bovine serum albumin [BSA], 50 mm sodium ascorbate, and 20 mm Cys) in the presence of sand. Two milliliters of disruption buffer was used for each gram of plant material. The homogenate was filtered through 4 layers of Miracloth and the dry cake was ground in disruption buffer for another 5 min before being filtered as outlined above. The pooled filtrate was centrifuged at 300g for 5 min and the resulting supernatant was centrifuged at 2,500g for 5 min once or twice, depending on the size of the pellet. A mitochondria-enriched pellet was produced by centrifuging the supernatant for 15 min at 15,100g. The pellet was subsequently resuspended in 1 mL washing buffer (0.3 m Suc, 10 mm TES, pH 7.5 [KOH], and 10 mm KH₂PO₄) and subjected to two strokes in a Dounce homogenizer. Mitochondria were further enriched on continuous Percoll gradients (50% [v/v] Percoll, 0.3 m Suc, 10 mm TES [pH 7.5 with KOH], 1 mm EDTA, 10 mm KH₂PO₄, and 1 mm Gly). Percoll gradients were established by centrifugation of the gradients at 69,400g for 40 min prior to loading. One milliliter of homogenized material was layered upon each of the 12 gradients, followed by centrifugation at 17,400g for 20 min. Mitochondria, concentrated near the bottom of the tube as a cloudy layer, were quantitatively removed using a glass pipette, resuspended in washing buffer, and centrifuged at 17,200g for 15 min. This was repeated two to four times until the pellets were solid. Finally, pellets of all gradients were pooled in a last washing step.

Isolation of Plant Mitochondria with Mito-AP

For Mito-AP, 500 mg plant material was ground using a mortar and pestle with some sand and 1 mL of ice cold, freshly prepared extraction buffer (32 μL mL⁻¹ BioLock [IBA Lifesciences], 100 mm ammonium bicarbonate, and 200 mm NaCl, pH 8) on ice for 3 min. The extract was transferred to a 2-mL reaction tube and centrifuged at 4°C for 5 min at 1,000g. Aliquots of magnetic beads (25 mg mL⁻¹; 500 μL for metabolome analysis and 250 μL for proteome analysis) were prepared by washing beads twice with 1 mL washing buffer (100 mm ammonium bicarbonate and 200 mm NaCl, pH 8) followed by resuspension in 500 μL washing buffer. The supernatant of the extract (800 μL) was added to the beads and incubated for 5 min at 4°C with continuous inversion. The beads were separated from the remaining liquid with a magnetic separator for 2 min. The supernatant was discarded and 1 mL washing buffer was added. Beads were resuspended by inverting and gentle shaking of the reaction tube. The washing was repeated three times. For method 1, the beads were directly incubated with 20 μL 2× SDS loading buffer and heated for 5 min at 95°C. For method 2, the beads were further treated by resuspension in 1 mL elution buffer (50 mm biotin, 100 mm ammonium bicarbonate, and 200 mm NaCl, pH 8). The mixture was incubated for 5 min on ice and inverted every minute. Subsequently, 950 µL digest buffer (200 µg mL⁻¹ proteinase K, 4 mm CaCl₂, 100 mm ammonium bicarbonate and 200 mm NaCl, pH 8) was added and the slurry incubated at RT for 10 min. Proteinase K was inhibited by the addition of 50 μL 100 mm phenylmethylsulfonyl fluoride (in pure isopropanol). The beads were separated for 2 min with a magnetic separator and the liquid was transferred with a cut tip, to prevent damaging the mitochondria, to a new 2-mL reaction tube followed by centrifugation at 4°C for 10 min at 14,500g. The supernatant was carefully removed and the pellet was stored at −80°C for further analysis.

Proteomic Analysis

Proteins from all crude extracts were first quantified with Bradford reagent to ensure a similar loading of SDS gels. Samples were resuspended in 20 μL 2× SDS loading buffer and heated for 5 min at 95°C. The proteins from all samples were additionally quantified by comparison with a BSA standard on the preparative SDS gel (Luo et al., 2006) employing a Li-cor Odyssey FC with Image Studio software. Sample preparation and proteomic analysis were performed according to Rugen et al. (2019). The volume used for resuspension of the peptides was adjusted for each sample to account for differences in the initial protein amounts. In fact, the total number of iBAQs in all samples with enriched mitochondria varied only by a factor of 3 (Supplemental Table S1). We found ∼10-fold fewer iBAQs in crude extracts for differential and density gradient centrifugation (DGC), potentially indicating interference of the extraction buffer with in-gel quantification.

Correlation Analysis

Peptides from DGC, Mito-AP, and DGC combined with Mito-AP (method 1 and 2, respectively) were reanalyzed using a 4-h gradient for LC-MS/MS and otherwise treated as described above. All iBAQs were normalized to the total iBAQs in the respective sample and a correlation matrix, based on the Pearson correlation coefficient, was constructed in R (version 3.6.0; https://www.R-project.org). Correlations with a coefficient >0.99 were visualized by Cytoscape version 3.7.1 (Shannon, 2003), including an expression correlation plugin) with an edge-weighted spring-embedded layout, based on the correlation coefficient. Edges with a higher correlation coefficient are shorter.

Metabolome Analysis

The pellet of mitochondria derived from Mito-AP method 2 was solubilized with 0.5 mL 80:20 MeOH:5 mm ammonium acetate (pH 9.5, adjusted with NH₃), vortexed, and sonicated for 2 min at RT. Samples were centrifuged for 10 min at 40,000g, and supernatants were dried down in a vacuum centrifuge concentrator (RVC 2-25 CD plus, Thermo Fisher Scientific) until dry. The resulting metabolites were resuspended in 11 μL 5 mm ammonium acetate (pH 9.5, adjusted with NH₃). The chromatography of 10 μL solution was performed on a 1290 Infinity HPLC (Agilent Technologies) using a Hypercarb column (50 mm, 4.6-mm diameter; Thermo Fisher Scientific) with mobile phase A consisting of 5 mm ammonium acetate (pH 9.5, adjusted with NH₃) and mobile phase B consisting of pure acetonitrile. With a flow rate of 0.6 mL min⁻¹, the gradient was 0–7 min, 4% to 20% B; 7–11 min 10 s, 20% to 30% B; 11 min 10 s–15 min 10 s, 50% B; 15 min 10 s–15 min 20 s, 50% to 100% B; 15 min 20 s–17 min 42 s, 100% B; 17 min 42 s–18 min, 100% to 4% B; and 4% B until the end of the gradient (23 min). With this gradient, metabolites elute in the following order: NMN (4.6 min), ATP (6.1 min), ADP (6.33 min), AMP (6.35 min), NADP (9.1 min), and NAD/NADH (11.6 min). The masses for the metabolites were quantified on a 6470 triple quadrupole mass spectrometer (Agilent Technologies). The LC-Agilent Jet Stream-electrospray ionization-tandem MS measurements were conducted under multiple reaction monitoring in positive ion mode. Agilent Jet Stream-electrospray ionization source conditions were set as follows: gas temperature, 250°C; gas flow, 12 L min⁻¹; nebulizer gas, 20 psi; sheath gas temperature, 395°C; sheath gas flow, 12 L min⁻¹; capillary voltage, 3000 V; and nozzle voltage, 500 V. Absolute amounts were calculated by comparison with the respective external standard curves of known concentrations. Identification of compounds was confirmed by retention times identical with the external standards. For quantification, the multiple reaction monitoring with the highest signal was used as quantifier. Other transitions, if available, were used as additional qualifiers. Transitions, fragmentors, and collision energies for all compounds can be found in Supplemental Table S5. For quantification, the Agilent Mass Hunter Quantitative Analysis Software was used.

Lipidomic Analysis

The pellet of mitochondria from Mito-AP method 2 was extracted according to a described method for the extraction and analysis of lipids, metabolites, proteins, and starch (Salem et al., 2016). In brief, frozen beads were resuspended in 1 mL extraction buffer (methyl tert-butyl ether:MeOH, 3:1 [v/v]) and the samples were incubated on a shaker for 45 min before a 15-min sonication in an ice-cooled sonication bath was applied. Phase separation was achieved by adding 0.65 mL of water:MeOH (3:1). Lipids were analyzed from the organic (methyl tert-butyl ether) phase by ultra-performance LC-MS. For this purpose, an Acquity iClass (Waters) ultra-performance LC was connected to a Q-Exactive HF (Thermo Scientific) high-resolution MS. Samples were measured in positive ionization mode as described previously (Giavalisco et al., 2011; Hummel et al., 2011; Salem and Giavalisco, 2018;). Data analysis was performed using targeted peak extraction and integration using the Trace Finder software (Version 4.1, Thermo Scientific).

Visualization of the Mitochondrial Membrane Potential

Twelve-day-old seedlings originating from liquid culture were washed twice with water and incubated in 500 μL one-half strength Murashige and Skoog medium containing 40 nm TMRM (dissolved in dimethyl sulfoxide) for at least 15 min. For analyzing the decoupled state of the mitochondria, 40 μm CCCP (dissolved in dimethyl sulfoxide) was added additionally to the solution to a final concentration of 40 μm and incubated for 30 min.

Isolated mitochondria bound to magnetic beads were resuspended in 100 μL mitochondrial assay solution buffer (70 mm Suc, 219 mm mannitol, 5 mm HEPES, 1 mm EGTA, and 0.5% [w/v] BSA, pH 7.2) containing 10 mm succinate and treated as described.

The analysis of the samples was done by using confocal microscopy as described above. The TMRM dye was excited at 552 nm and the emission between 600 and 615 nm was visualized.

Respiratory Activity Measurement

For the respiratory activity, 500 mg plant material was ground up using a mortar and pestle with some sand and 1 mL ice cold, freshly prepared extraction buffer (32 μL mL⁻¹ BioLock [IBA Lifesciences]; 5 mm dithiothreitol, 300 mm Suc, 5 mm KH₂PO₄, 10 mm TES, 10 mm NaCl, 2 mm MgSO₄, and 0.1% [w/v] BSA, pH 7.2) on ice for 2 min. The extract was transferred to a 2 mL reaction tube and centrifuged at 4°C for 5 min at 1,000g. Five hundred microliters of magnetic beads (25 mg mL⁻¹) was prepared by washing twice with 1 mL washing buffer (300 mm Suc, 5 mm KH₂PO₄, 10 mm TES, 10 mm NaCl, 2 mm MgSO₄, and 0.1% [w/v] BSA, pH 7.2), followed by resuspension in 500 μL washing buffer. The subsequent steps were performed as described in “Isolation of Plant Mitochondria with Mito-AP.”

For the measurement of the respiration through succinate dehydrogenase (complex II), the beads were resuspended in respiration buffer 2 (300 mm Suc, 5 mm KH₂PO₄, 10 mm TES, 10 mm NaCl, 2 mm MgSO, and 0.1% [w/v] BSA, 10 mm succinate, 500 μm ATP, and 500 μm n-propyl gallate, pH 7.2) at RT.

The OCR measurements were performed using the Seahorse XFe 96 Analyzer (Agilent Technologies). The XFe96 Sensor Cartridge was prepared according to the manufacturer’s instructions. The ports were filled as follows: port A, 20 μL 60 mm ADP, pH adjusted to 7.2 with KOH; port B, 22 μL 50 µg mL⁻¹ oligomycin A; port C, 24 μL 80 mm antimycin A; port D, 26 μL buffer. All chemicals were prepared in the corresponding respiration buffer.

Resuspended beads were loaded into the XF96 cell culture microplate and the plate was subsequently centrifuged for 5 min at 2,200g. The wells were filled with the corresponding respiration buffer to a final volume of 200 μL. The four corners of the plate were not used for sample measurements.

The respiration assay was performed at 29°C (internal heater was turned off). After equilibration for 12 min, the baseline was measured by mixing for 30 s, waiting 10 s, and measuring for 2 min. The same procedure was done after the injection of the single ports. After the measurements, the protein amount was quantified for every sample.

Transmission Electron Microscopy

Samples, prepared as described above for the respiratory measurements, were fixed in 0.15 m HEPES, pH 7.35, containing 1.5% (w/v) formaldehyde and 1.5% (w/v) glutaraldehyde for 30 min at RT and overnight at 4°C. Samples were immobilized in 2% agarose (w/v) and then incubated in aqueous solutions of 1% (w/v) OsO4 (2 h at RT) and 1% (w/v) uranyl acetate (overnight at 4°C). After dehydration in acetone, samples were embedded in Epon, and 60-nm sections were stained with uranyl acetate and lead citrate (Reynolds, 1963). Images were recorded using a Morgagni transmission electron microscope (FEI) with a side-mounted Veleta charge-coupled device camera.

Accession Numbers

Sequences for the ADAPTER protein and TOM22-V are provided by Araport with the accession numbers At1g55450 and At5g43970, respectively. The Cb5-d sequence is available under the GenBank accession AY578730.

Supplemental Material

The following supplemental materials are available.

• Supplemental Figure S1. Transient expression of different tail-anchored proteins fused to GFP in N. benthamiana.

• Supplemental Figure S2. Quantitative comparison of phenotypical traits between the wild type (Col-0) and a stable 35S:Twin-Strep-tag:GFP:ADAPTER construct.

• Supplemental Figure S3. Integrity of isolated mitochondria in different buffers.

• Supplemental Figure S4. Plant mitochondria bound to a commercial matrix of 1-µm diameter.

• Supplemental Figure S5. Different bead sizes and coating materials influence the enrichment of mitochondria.

• Supplemental Figure S6. Quantification of the relative protein amounts from mitochondria enriched with different matrices.

• Supplemental Figure S7. Quantification of Strep-Tactin bound to different magnetic matrices.

• Supplemental Figure S8. Plant material needed for DGC and Mito-AP and relative mitochondrial protein yield obtained by both methods.

• Supplemental Figure S9. Quantification of the total protein yield obtained from elution with and without proteinase K.

• Supplemental Figure S10. Variation of mitochondrial purity from biological repeats processed by DGC or Mito-AP.

• Supplemental Figure S11. ADP-Glc standard curve.

• Supplemental Figure S12. Contamination with plastidic fragments.

• Supplemental Figure S13. Visualization of the mitochondrial membrane potential in vivo.

• Supplemental Figure S14. Visualization of the mitochondrial membrane potential in vitro.

• Supplemental Figure S15. Electron micrographs of isolated mitochondria.

• Supplemental Table S1. Proteins identified by proteomics for the determination of mitochondrial purity.

• Supplemental Table S2. Proteins identified in a 4-h LC gradient for correlation analysis.

• Supplemental Table S3. Lipids identified by LC-MS in Mito-AP (method 2) from 35S:Twin-Strep-tag:GFP:ADAPTER plants or Col-0 plants.

• Supplemental Table S4. Primer sequences.

• Supplemental Table S5. MRM parameters used for detection of polar metabolites with a triple quadrupole MS.