Whole cell formaldehyde cross-linking simplifies purification of mitochondrial nucleoids and associated proteins involved in mitochondrial gene expression

Abstract

Mitochondrial DNA/protein complexes (nucleoids) appear as discrete entities inside the mitochondrial network when observed by live-cell imaging and immunofluorescence. This somewhat trivial observation in recent years has spurred research towards isolation of these complexes and the identification of nucleoid-associated proteins. Here we show that whole cell formaldehyde crosslinking combined with affinity purification and tandem mass-spectrometry provides a simple and reproducible method to identify potential nucleoid associated proteins. The method avoids spurious mitochondrial isolation and subsequent multifarious nucleoid enrichment protocols and can be implemented to allow for label-free quantification (LFQ) by mass-spectrometry. Using expression of a Flag-tagged Twinkle helicase and appropriate controls we show that this method identifies many previously identified nucleoid associated proteins. Using LFQ to compare HEK293 cells with and without mtDNA, but both expressing Twinkle-FLAG, identifies many proteins that are reduced or absent in the absence of mtDNA. This set not only includes established mtDNA maintenance proteins but also many proteins involved in mitochondrial RNA metabolism and translation and therefore represents what can be considered an mtDNA gene expression proteome. Our data provides a very valuable resource for both basic mitochondrial researchers as well as clinical geneticists working to identify novel disease genes on the basis of exome sequence data.

Fig 1. Validation of TwinkleFLAG IAP following whole cell cross-linking.

HEK293 Flp-In T-Rex cells expressing either TwinkleFLAG or a mitochondrially targeted Luciferase FLAG (mtLucFLAG) were induced for 36 hrs with 3 ng/ml doxycycline, harvested, samples equalized by protein content and incubated for 10 min with 1% formaldehyde (FA) for whole cell crosslinking. Following cross-linking, cells were lysed and FLAG-tagged protein purified using FLAG immunoaffinity resin. Precipitated complexes were analysed using Western blot analysis (see M&M and main text for full details). Results (A, B) show that proteins of the mtDNA maintenance machinery are enriched with cross-linking in TwinkleFLAG expressing cells. (C) ρ° HEK293 Flp-In T-Rex cells expressing TwinkleFLAG were established and crosslinked samples of TwinkleFLAG expressing cells were compared with their mtDNA-containing parental cells also expressing TwinkleFLAG. Results show a very substantial decline in levels of co-purifying TFAM and mtSSB, in the absence of mtDNA.

Fig 2. Whole cell cross-linking followed by IAP enriches for mitochondrial and nucleoid associated proteins.

Protein complexes purified using FLAG-tag targeted isolation from 3 independent biological repeats using various batches of TwinkleFLAG (Twinkle) and mtLucFLAG (Luc) cells, treated either with or without FA and further processed as described in Fig. 1, were analysed by shotgun mass spectrometry. Using MaxQuant, LFQ values were derived and ratio’s calculated comparing TwinkleFLAG versus mtLucFLAG witout cross-linking (-XL) with crosslinking (+XL) as well as TwinkleFLAG +XL versus -XL and mtLucFLAG +XL versus -XL. Protein lists were compiled based on a ≥2 fold increase in LFQ values in at least 2 out of 3 experiments (see S1 Table). (A) Gene Ontology (GO)_SLIM_Cellular Compartment (CC) (see also M&M ) annotation was used to calculate percentages of mitochondrial proteins in each set. This analysis illustrates that all crosslinked sets (being either with TwinkleFLAG or mtLucFLAG) showed approximately 70% mitochondrial annotation whereas the TwinkleFLAG versus mtLucFLAG -XL showed only 28% mitochondrial annotation. (B) To identify potentially interesting proteins we compared all 4 generated lists simultaneously using Venny (http://bioinfogp.cnb.csic.es/tools/venny/index.html), that generates a 4-way Venn diagram and separate lists for all intersecting and non-intersecting parts of the diagram. The region for potentially interesting proteins, being enriched with TwinkleFLAG +XL compared to respective controls is further outlined in red. The resulting list of 168, used for later comparison (see Fig. 3) is separately given alphabetically by gene name in S4 Table (first sheet: ‘Biol repeats enriched all’). S1 Table, in addition is sorted in such a way that the same 168 proteins are the first 168 proteins listed in the LFQ comparison sheet (sheet 3).

Fig 3. Q Exactive mass spectrometry analysis following Triton X100 based affinity purification.

(A) Sample 2 of the 3 biological repeats (measured for Fig. 2) was measured in triplicate on a Q Exactive Orbitrap. To identify potentially interesting proteins we compared all 4 generated lists simultaneously using Venny, similar as in Fig. 2. The region for potentially interesting proteins, being enriched with TwinkleFLAG +XL compared to respective controls again is further outlined in red. The resulting list of 192, used for later comparisons (see Figs. 3B and 4) is separately given alphabetically by gene name in S4 Table (second sheet: ‘TX100 enriched all’). S2 Table, in addition is sorted in such a way that the same 192 proteins are the first proteins listed in the LFQ comparison sheet (sheet 3). (B) In order to compare different sets of experiments we used area-proportional Venn diagrams (BioVenn[58]). Comparing the enriched set of proteins from three biological repeats (Fig. 2) measured using an LTQ-FT mass spectrometer with series 2 of the biological repeat measurement, measured in triplicate with a Q Exactive Orbitrap mass spectrometer (see above, A), shows a considerable overlap between both experiments. The core set of proteins enriched in both measurements includes many established nucleoid associated proteins.

Fig. 4. X-ChIP based affinity purification provides the most inclusive analysis of nucleoid associated proteins.

(A) Protein complexes using FLAG-tag targeted isolation using TwinkleFLAG (Twinkle) and mtLucFLAG (Luc) cells, from cells treated either with or without FA were isolated using an X-ChIP based isolation buffer. Samples were analysed (in duplicate for TwinkleFLAG + XL, otherwise in triplicate) by shotgun mass spectrometry using a Q Exactive Orbitrap. To again identify potentially interesting proteins we compared all 4 generated lists simultaneously using Venny, similar as in Figs. 2/3. The region for potentially interesting proteins, being enriched with TwinkleFLAG +XL compared to respective controls again is further outlined in red. The resulting list of 366 proteins, used for later comparisons (see 4B/C/D) is separately given alphabetically by gene name in S4 Table (third sheet: ‘X-ChIP enriched all’). S3 Table, in addition is sorted in such a way that the same 366 proteins are the first proteins listed in the LFQ comparison sheet (sheet 3). (B) An area-proportional Venn diagram shows the comparison of the enriched set obtained using TX100 lysis compared to the enriched set obtained using the X-ChIP method. An analysis of the proteins identified as enriched in both sets shows that of these 111, 109 proteins (98%) have a Gene Ontology (GO)_SLIM_Cellular Compartment (CC) annotation (C) while the remaining 2 proteins despite the lack of such an annotation are likely also to be mitochondrial. In contrast, of the remaining 81 proteins identified as enriched exclusively with the TX100 method, only 36% is annotated as mitochondrial, while of the 255 proteins that were found specifically enriched with the X-ChIP method but absent in the TX100 dataset, 88% is annotated as mitochondrial. Again this likely is an underestimation by mis-annotation or the lack of a GO_SLIM_CC annotation. These data combined thus identify the X-ChIP method as the superior method in combination with whole cell cross-linking. (D) Using the X-ChIP method we now compared LFQ values of the 366 proteins obtained with regular HEK293 TwinkleFLAG cells with those obtained from HEK293 TwinkleFLAG ρ° cells. The pie-chart shown here illustrates the distribution of the 366 enriched proteins identified with the X-ChIP method in regular HEK293 TwinkleFLAG and measured in HEK293 TwinkleFLAG ρ° in the following classes: not detected (absent), 95 proteins; ≥2 fold decrease, 163 proteins; no change, 100 proteins or ≥2 fold increase, 8 proteins(see also S3 Table). Light gray boxed text shows abridged lists of proteins in each of the four categories selected from S3 Table, concentrated on proteins involved in mtDNA maintenance and gene expression and including a few other categories discussed in the text such as complex I and V, as well as a few newly identified candidate proteins. A few of the proteins that are considered novel candidate nucleoid associated proteins and that are discussed in the main text are highlighted in red. Although quite a few other proteins have not been described primarily as nucleoid associated these have been described as having a role in mitochondrial gene expression and hence have not been highlighted.

Fig. 5. Comparing whole cell cross-linking TwinkleFLAG immune affinity purification with previous nucleoid isolations.

Comparison with most commonly identified potential mtNAPs as published in[29] with their enrichment in the TwinkleFLAG +XL IAP. The data here is reduced to compare previously published mitochondrial formaldehyde cross-linking followed by nucleoid purification as performed by[22], in which for simplicity reasons both published protein list are combined to one list and the data from ICT1-FLAG IP as performed by[38]. For the full table see Hensen et al[29]. Shown are the comparison of the three biological repeats on the LTQ-FT Ultra, Q Exactive TX100 and X-ChIP method datasets. Green checkmark indicates an ≥ 2 fold increase in the TwinkleFLAG IAP compared to the mtLucFLAG IAP control with cross-linking. A light red cross indicates no difference while a dark red cross indicates undetected protein. Green checkmark indicated with an asterix represent proteins which are increased in TwinkleFLAG compared to mtLucFLAG with cross linking but not compared to non cross- linked TwinkleFLAG control (Twinkle itself is a logical representative of this class). For the ρ° samples we indicate the percentage of protein, based on LFQ ratios, co-purified in the absence of mtDNA.