J. Cell. Mol. Med. Vol 20, No 5, 2016 pp. 794-803 New protein–protein interactions of mitochondrial connexin 43 in mouse heart u~ Amanda Denuc a, Estefanıa Nnez b, Enrique Calvo b, Marta Loureiro b, Elisabet Miro-Casas a, Adela Guar as b, Jesus Vazquez b, David Garcia-Dorado a, * a Vall d’Hebron University Hospital and Reseach Institute, Universitat Autonoma de Barcelona, Barcelona, Spain b Centro Nacional de Investigaciones Cardiovasculares Carlos III, Madrid, Spain Received: October 12, 2015; Accepted: December 17, 2015 Abstract Connexin 43 (Cx43), the gap junction protein involved in cell-to-cell coupling in the heart, is also present in the subsarcolemmal fraction of car- diomyocyte mitochondria. It has been described to regulate mitochondrial potassium influx and respiration and to be important for ischaemic preconditioning protection, although the molecular effectors involved are not fully characterized. In this study, we looked for potential partners of mitochondrial Cx43 in an attempt to identify new molecular pathways for cardioprotection. Mass spectrometry analysis of native immunopre- cipitated mitochondrial extracts showed that Cx43 interacts with several proteins related with mitochondrial function and metabolism. Among them, we selected for further analysis only those present in the subsarcolemmal mitochondrial fraction and known to be related with the respi- ratory chain. Apoptosis-inducing factor (AIF) and the beta-subunit of the electron-transfer protein (ETFB), two proteins unrelated to date with Cx43, fulfilled these conditions, and their interaction with Cx43 was proven by direct and reverse co-immunoprecipitation. Furthermore, a previ- ously unknown molecular interaction between AIF and ETFB was established, and protein content and sub-cellular localization appeared to be independent from the presence of Cx43. Our results identify new protein–protein interactions between AIF-Cx43, ETFB-Cx43 and AIF-ETFB as possible players in the regulation of the mitochondrial redox state. Keywords: connexin 43  apoptosis-inducing factor  electron-transfer protein  mitochondria  cardiomyocyte Introduction Connexin 43 (Cx43) is the predominant connexin isoform in ventricu- Besides their role in the formation of permeable transmembrane lar myocardium. It is widely known to participate in cell-to-cell propa- channels ensuring electrical and metabolic coupling between cells, gation of the electrical impulse as well as in chemical communication Cx43 has been shown to participate in several signalling pathways [6, between adjacent cells. As a pore-forming molecule, it is composed 7]. Normal sodium and potassium currents appear to be dependent of four transmembrane domains as well as intracellular amino- and on Cx43 expression [8, 9], and therefore, any alteration of this protein carboxyterminal domains [1]. Connexins are assembled in groups of can modify the action potential of cardiomyocytes with arrhythmo- six units to form hemichannels, or connexons, that may come into genic consequences. Connexins may also control cell growth and contact with a hemichannel from a neighbouring cell to form a gap inhibit tumorigenicity in different cell types; it has been proved that junction (GJ) channel. These intercellular channels conform a special- overexpression of Cx43 in neonatal rat cardiomyocytes decreases ized membrane area of chemical communication that usually inte- DNA synthesis [10]. In Cx43-deficient mice, embryonic development grates other adhesion structures, such as desmosomes or adherens is compromised and animals die at birth due to swelling and blockage junctions [2]. Importantly, channel activity is determined by the con- of the right ventricular outflow tract from the heart [11]. At the con- duction properties of the connexin involved. Thus, Cx43 connexons nexon level, several mechanisms nonrelated to GJ communication have a minimal diameter pore of 15  A, with a relative anion-to-cation have been proposed to modulate cell death and survival, paracrine selectivity ratio of 1.17, and an intermediate unitary conductance of factors release, volume regulation or activation of signal transduction 100 pS [3–5]. pathways [12]. Although the end effectors of these mechanisms remain poorly established, a complete understanding of connexin function has been achieved when defining its molecular interactors, *Correspondence: David GARCIA-DORADO. like kinases, phosphatases or other molecules regulated by receptor E-mail: dgdorado@vhebron.net stimulation or subcellular localization [13]. doi: 10.1111/jcmm.12792 ª 2016 The Authors. Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. J. Cell. Mol. Med. Vol 20, No 5, 2016 Furthermore, Cx43 is involved in myocardial energy metabolism, pentobarbital (150 mg/kg). Whole hearts were quickly excised, minced tolerance to ischaemia and preconditioning protection—i.e. protec- and homogenized with 5–10 strokes of a teflon pistill in a glass potter tion against cell death obtained by brief non-lethal episodes of at 250 r.p.m. in ice-cold isolation buffer A [in mmol/L: 290 sucrose, 5 ischaemia preceding prolonged severe ischaemia [14, 15]. Most of 3-(N-Morpholino)-propanesulfonic acid (MOPS), 2 ethylene glycol-bis(2- aminoethylether)-N,N,N 0 ,N 0 -tetraacetic acid (EGTA), 10 NaF, 2 Na3VO4 the signalling pathways of ischaemic preconditioning converge on and protease inhibitors cocktail (P8340; Sigma-Aldrich, Saint Louis, mitochondria, and recent reports have shown that in cardiomyocytes MO, USA), pH 7.4]. Samples were centrifuged at 750 g and super- Cx43 is also located at the inner membrane of subsarcolemmal natant obtained handled as the subsarcolemmal mitochondrial fraction mitochondria (SSM), whereas it is nearly absent at interfibrillar mito- (SSM). The sediment of this first centrifugation was further processed chondria (IFM) [16, 17]. In vitro cross-linking studies show mito- with 2 mg proteinase K (P2308; Sigma-Aldrich)/g heart tissue in ice- chondrial complexes with a molecular weight comparable to that of cold isolation buffer B [in mmol/L: 100 KCl, 50 MOPS, 2 EGTA, and sarcolemmal Cx43 hemichannels; in addition, those complexes are 10% bovine serum albumin (BSA), pH 7.4] to obtain the interfibrillar sensitive to carbenoxolone and heptanol, two hemichannel blockers mitochondrial fraction (IFM). Two additional centrifugation steps are [18]. It has also been described that mitochondrial Cx43 (mtCx43) necessary for mitochondrial purification. The first one at 5000 9 g to modulates mitochondrial potassium influx and respiration [18, 19]. obtain an enriched SSM and IFM fraction, followed by a second cen- trifugation at 12,500 9 g in 17% percoll in isolation buffer A. Total Moreover, in the presence of mtCx43, ischaemic preconditioning heart homogenates were attained by solubilization of 100–200 mg of prevents reperfusion-induced respiratory failure and oxidative dam- tissue in lysis buffer [2% n-Dodecyl b-D-maltoside (D4641; Sigma- age, whereas lack of mtCx43 abolishes the protective effect of pre- Aldrich), 20 mmol/L TEA pH 8, 20 mmol/L ethylenediaminetetraacetic conditioning on myocardial infarct size [15, 20, 21]. However, the acid, 10 mmol/L DTT and protease inhibitors cocktail]. All samples molecular effectors of the cardioprotective role of Cx43 still remain were kept on ice for immediate use, and protein concentration was elusive. Mitochondrial Cx43 could regulate mitochondrial volume determined by Bradford assay. and, as a consequence, respiration and reactive oxygen species (ROS) signalling by forming hemichannels similar to those present at the sarcolemma. Alternatively, mtCx43 could regulate mitochon- Co-Immunoprecipitation experiments drial function and cardioprotection by interacting with still unknown proteins. A total of 0.8–1 mg of protein was used for each pull down reaction The aim of this study is to find new molecular partners of with 10 lg of rabbit anti-connexin-43 antibody (C6219; Sigma- mtCx43 that could participate in putative channel-independent Aldrich), mouse anti-apoptosis-inducing factor antibody (MA5-15880; actions. Proteomics analysis identifies apoptosis-inducing factor Pierce-Thermo Scientific, Rockford, IL, USA), rabbit anti-ETFB subunit (AIF) and the beta-subunit of the electron-transfer protein (ETFB), beta antibody (17925-1-AP; ProteinTech, Manchester, UK), respec- tively, according the Pierce Crosslink Immunoprecipitation Kit manu- two unrelated proteins involved in oxidative phosphorylation and facturer’s instructions (26147; Thermo Scientific, Rockford, IL, USA). redox control, as molecular interactors of mtCx43. Our results also Total and mitochondrial heart extracts were treated in non-denaturing confirm that AIF and ETFB are able to interact in a Cx43-indepen- conditions with buffers supplied, and protein-antibody-agarose beads dent manner. complexes incubated at 4°C during 16 hrs with an end-over-end mix- ing. After washing, bound proteins were eluted (low pH) and pre- served for posterior proteomics analysis and Western blotting. Materials and methods Proteomics Mice Protein digestion Adult male and female transgenic Cx43KI32 mice (developed by [22]) The resulting soluble protein extracts were run on a SDS-PAGE gel aged between 16 and 20 weeks (30 g) were used in this study. In these (10% resolving gel and 4% stacking gel) at 50 V. The electrophoresis animals, the coding region of the Cx43 gene has been replaced by that was stopped when the front dye has barely passed into the resolving of Cx32 as confirmed by specific PCR analysis. All protocols were gel, ensuring concentration of all proteins into a unique band. Staining approved by the Ethics Committee on Animal Research of the Hospital was performed using GelCode Blue Stain Reagent (Thermo Scientific). Vall d’Hebron, Barcelona, Spain. The study conforms the European Gel pieces were cut into cubes (2 mm). For the protein digestion, modi- Directive on the Welfare of Research Animals (2010/63/UE). fied porcine trypsin (Promega, Fitchburg, WI, USA) was added at a final ratio of 1:20 (trypsin-protein). Digestion proceeded overnight at 37°C in 100 mM ammonium bicarbonate, pH 8.8. Mitochondrial and total heart extracts Mice heart mitochondria were prepared by differential centrifugation Identification of proteins by liquid chromatography coupled [18, 23]. Briefly, for each pull down assay, 6–8 animals from both to tandem mass spectrometry wild-type (Cx43 WT; Cx43 homozygous genotype) and knock-in mutant The resulting tryptic peptide mixtures were subjected to nano-liquid mice (Cx43 KI Cx32; Cx32 homozygous genotype), considered two sep- chromatography coupled to mass spectrometry (LC-MS) for protein arate groups, were deeply anaesthetized by an i.p. injection of sodium identification. Peptides were injected onto a C-18 reversed phase (RP) ª 2016 The Authors. 795 Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. nano-column (75 lm I.D. and 50 cm, Acclaim PepMap100; Thermo Sci- mouse anti-Pan Cadherin (C1821; Sigma-Aldrich) and rabbit anti- entific) and analysed in a continuous acetonitrile gradient consisting of TOMM20 antibody (HPA011562; Sigma-Aldrich). 0–30% B in 240 min. and 50–90% B in 3 min. (B = 90% acetonitrile, 0.5% acetic acid). A flow rate of ca. 200 nl/min. was used to elute pep- tides from the RP nano-column to an emitter nanospray needle for real- Immunofluorescent assays time ionization and peptide fragmentation on a Orbitrap Elite mass spec- trometer (Thermo Fisher, San Jose, CA, USA). An enhanced FT-resolu- tion spectrum (resolution = 120,000) followed by the MS/MS spectra Isolation of adult cardiac myocytes from most intense fifteen parent ions was analysed along the chromato- Adult cardiac cells from either wild-type or knock-in Cx43 mice were graphic run (272 min.). Dynamic exclusion was set at 30 sec. For pro- obtained by retrograde collagenase perfusion [15]. Briefly, adult mice tein identification, tandem mass spectra were extracted and charge were deeply anaesthetized by an i.p. injection of sodium pentobarbital state deconvoluted by Proteome Discoverer 1.4.0.288 (Thermo Fisher (150 mg/kg), and the heart was perfused in a Langendorff system Scientific). All MS/MS samples were analysed using SEQUESTTM for 15 min. with a modified Krebs buffer with 0.03% type II collage- (Thermo Fisher Scientific). nase (17456; Serva Electrophoresis, Heidelberg, Germany). Calcium- From an initial list of 493 identified proteins, 78 were proposed as tolerant rod-shaped cardiomyocytes were selected by gravity sedimen- potential candidates according to the different proportion of peptides tation in 4% BSA gradient and plated on laminin-coated cover slips. detected between the SSM and IFM sample, favouring the former. Among this list of 78, eight candidates were finally chosen according to the following criteria (see Table S1 for full list): (i) cytoskeleton proteins Subcellular localization assays were excluded, as we were working mitochondrial specificity; (ii) pro- Isolated cardiomyocytes were fixed and permeabilized with ice-cold ace- teins present in the IFM sample were excluded, as Cx43 is not repre- tone during 8 min. at RT. After washing with PBS + 0.025% Triton X- sented in this fraction; and finally, (iii) proteins were chosen according 100, samples were blocked with 1% BSA in PBS for 45 min. at RT and to the bibliographic data available related with the functional capacity of immunodetected using rabbit anti-Cx43 antibody (C6219; Sigma- the organelle (UniProt Database, UniProt Consortium). Aldrich), mouse anti-AIF antibody (MA5-15880; Pierce-Thermo Scien- tific), rabbit anti-ETFB antibody (17925-1-AP; ProteinTech), mouse anti- SDHA (MS204; MitoSciences) and rabbit anti-VDAC antibody (ab15895; Blue-native electrophoresis and MS distribution analyses AbCam, Cambridge, UK) as primary antibodies and AlexaFluor 488-con- Mitochondria-enriched samples from three different tissues (liver, heart jugated antimouse (A-11001; Molecular Probes, Thermo Scientific, and brain) and from a fibroblasts cell line (fC57) were treated with 4% Rockford, IL, USA) and AlexaFluor 594-conjugated anti-rabbit (A-11012; digitonin to solubilize membrane proteins. Blue-native electrophoresis Molecular Probes) as secondary antibodies. Finally, cardiomyocytes (BNE) was used to separate the native complexes in a gradient gel from were washed twice with PBS and nuclei were stained with Hoeschst 4% to 13% acrylamide. Staining was performed using GelCode Blue 33342 (10 lg/ml). Confocal fluorescent images were captured with a Stain Reagent (Thermo Scientific). Each lane was divided and excised in spectral microscope (Olympus Spectral Confocal Microscopy FV1000, 26 bands covering the whole electrophoretic run. Gel pieces were then Olympus, Center Valley, PA, USA). Images were captured taken into digested with trypsin following the aforementioned protocol, and the account that the pixel size was set to 1/2 of the optical resolution, and resulting tryptic peptide mixtures were subjected to LC-MS analysis for quantitative data were obtained with ImageJ software analysis (JACoP identifying the proteins present on each band. plug-in) [24]. Suborganelle protein localization Proximity ligation assay Cardiomyocytes from either wild-type or knock-in Cx43 mice were iso- To assess the mitochondrial localization of the different proteins of inter- lated, plated in cover slips and incubated with the corresponding primary est, 100–300 lg of SSM protein, either from wild-type or knock-in mutant antibodies as previously described. Preparations were then tested using mice, was incubated for 10 min. on ice with different proteinase K con- the DuoLink kit secondary antibodies conjugated with the corresponding centrations (0.1–0.5 lg/ml; Sigma-Aldrich), in order to partially digest the proximity ligation assay (PLA) probes (DUO92101; Sigma-Aldrich) for outer membrane. Samples were further analysed by Western blotting. 1 hr 37°C. For ligation and amplification reactions, DuoLink in situ detec- tion kit recommendations were followed. Cell images were obtained with spectral microscope (Olympus Spectral Confocal Microscopy FV1000) as stated above. Positive cross-reactivity was quantified in background-sub- Western blot tracted images using BlobFinder software (developed by: Carolina W€ahlby and Amin Allalou at CBA, Uppsala University) [25]. Protein samples (25–30 lg of total protein/lane) were mixed 1:1 with 29 protein loading buffer (60 mmol/L Tris-HCl pH 6.8, 10% glycerol, 2% SDS, 1% bromophenol blue, 10% beta-mercaptoethanol), boiled 5 min. at 98°C and loaded onto 12% SDS-PAGE gels. Proteins were Statistics detected by Western blotting using: rabbit anti-Cx43 antibody (C6219; Sigma-Aldrich), mouse anti-AIF antibody (MA5-15880; Pierce-Thermo Data are expressed as mean  S.E.M. Student’s t-test was used to Scientific), rabbit anti-ETFB antibody (17925-1-AP; ProteinTech), mouse compare Manders’ coefficients and PLA quantification values between anti-Na+/K+ ATPase alpha-1 antibody (05-369; Millipore, Billerica, MA, Cx43 WT and Cx43 KI Cx32 transgenic animals. Differences were USA), mouse anti-SDHA (MS204; MitoSciences, Eugene, OR, USA), considered significant when P < 0.05. 796 ª 2016 The Authors. Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. J. Cell. Mol. Med. Vol 20, No 5, 2016 Results Mitochondrial Cx43 co-immunoprecipitates with AIF and ETFB Mitochondrial Cx43 interacts with proteins Because no previous interaction has been described so far for these related with the respiratory electron chain proteins, we next proceeded to confirm the proteomics results by Western blot analysis of the immunoprecipitated samples (Fig. 1A). To examine new molecular partners of Cx43, a first set of For this purpose, we used purified SSM mitochondrial extracts from immunoprecipitation experiments was done in non-denaturing Cx43 WT mice, including the IFM fraction as a negative control. A conditions to preserve and analyse potential protein–protein inter- good recovery of endogenous Cx43 was achieved, and a faint band actions. Cardiac subsarcolemmal (SSM) and interfibrillar (IFM) was detected in the eluted samples, both for ETFB and AIF proteins mitochondrial extracts were purified from Cx43 WT mice hearts (Fig. 1A, lanes 4–6 arrows), although AIF content on SSM Cx43- and 1 mg of total protein was processed with a specific anti- immunoprecipitated sample was barely perceptible. In addition, when Cx43 antibody. IFM elution sample obtained was used as a nega- performing reverse co-immunoprecipitation to analyse the presence tive control, as Cx43 is virtually absent in this mitochondrial pop- of Cx43 in the anti-ETFB and anti-AIF eluted samples, Cx43 was pre- ulation [17]. Protein content of the eluted fractions was analysed sent in both mitochondrial extracts, most notably in the SSM fraction, by mass spectrometry and final results are summarized in validating the proteomics results (Fig. 1A, lanes 7–12 arrows). Table 1. We were unable to prove any co-immunoprecipitation of Cx43 To sum up, from an initial list of 78 protein candidates, eight with NDUFS6 (data not shown), and so it was discarded from our were finally chosen based on its involvement in mitochondria func- study. tion. When candidates were examined focusing on its main biologi- Remarkably, the AIF-immunoprecipitated fraction showed reactiv- cal function, three of them had special interest for being ity against ETFB, pointing to a new protein interaction between these associated to mitochondrial respiration, as Cx43 is described to two proteins (Fig. 1A, lane 9 asterisk). have an impact on respiratory complex I activity and oxygen con- sumption [19]. Ndufs6 is an accessory subunit of the mitochon- drial membrane respiratory chain NADH dehydrogenase (Complex Mitochondrial protein amount or sub-organelle I) that is believed not to be involved in catalysis; Etfb, in associa- distribution of ETFB and AIF is not dependent on tion with Etfa, acts as an electron carrier from fatty acid beta-oxi- dation to the electron transport chain via ETF-ubiquinone the presence of Cx43 oxidoreductase; and finally, Aifm1 is mostly known as a caspase- independent apoptogenic molecule but, in the absence of specific Evidence from the immunoprecipitation assay prompted us to investi- death signals, it has a NADH oxidoreductase activity (biological gate if Cx43 content could either modulate protein amount or sub- process extracted from UniProt Database). localization of ETFB and AIF in mitochondria. ETFB is supposed to Table 1 Eluted samples from three independent Cx43 immunoprecipitation experiments were analysed and the most regular results were selected from a total of 80 candidates. Gene, UniProt code and complete name are shown, and the biological processes known indicated Gene Protein UniProt Protein name Biological process Gja1 Cx43 P23242 Gap junction alpha-1 protein Cx43 - gap junction channel Eci1 ECI1 P42125 Enoyl-CoA delta isomerase 1 Fatty acid beta-oxidation Pfkm PFKM P47857 ATP-dependent 6-phosphofructokinase Carbohydrate degradation, glycolysis Ndufs6 NDUS6 P52503 NADH dehydrogenase [ubiquinone] Electron transport, respiratory chain iron-sulphur protein 6 Echs1 ECHS1 Q8BH95 Enoyl-CoA hydratase Fatty acid beta-oxidation Acat1 ACAT1 Q8QZT1 Acetyl-CoA acetyltransferase Ketone body metabolism Cs CS Q9CZU6 Citrate synthase Tricarboxylic acid cycle Etfb ETFB Q9DCW4 Electron-transfer flavoprotein subunit beta Electron transport, electron carrier Aifm1 AIF Q9Z0X1 Apoptosis-inducing factor 1 Apoptosis, cell redox homeostasis, respiratory chain complex I assembly ª 2016 The Authors. 797 Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. Fig. 1 Western Blot analysis of protein co-immunoprecipitation and mitochondrial distribution of Cx43, ETFB and AIF in WT and Cx43KICx32 mice. (A) A total amount of 1 mg of heart mitochondrial protein extracts from WT mice was immunoprecipitated using anti-Cx43, anti-AIF and anti-ETFB antibodies. In left panels, lanes from 1 to 3 correspond to the input sample, the initial total (T), subsarcolemmal (SSM) or interfibrillar (IFM) protein extracts. Lanes from 4 to 12 correspond to the output sample—the immunoprecipitated fraction—analysed for the presence of each protein. A rep- resentative IgG control of Cx43-IP protocol is shown in right panels using the total protein extract as a sample. Of note, the Cx43 antibody used recognizes also the phosphorylated protein forms (C0 unmodified, C1 phosphorylated). Also, due to proteinase treatment to obtain the IFM fraction, AIF is partially truncated and noted as a double lower molecular weight band. (B) Protein content of Cx43, AIF and ETFB. Anti-SDHA was used as a mitochondrial loading control. (C) Sub-mitochondrial distribution of Cx43, AIF and ETFB proteins was examined after partial digestion of the outer mitochondrial membrane with 0.1 and 0.5 lg of proteinase K, as indicated. Na+/K+ ATPase reactivity was assessed to rule out any potential contam- ination from gap-junctional membranes, and SDHA and TOMM20 were used as an inner and outer mitochondrial membrane protein control, respec- tively. A representative image of at least three independent replicates is shown. remain in the mitochondrial matrix, and AIF displays an inner mem- differences in protein pattern were found related to Cx43 presence, brane-anchored mature form and an intermembrane-matrix-soluble suggesting that sub-organelle localization of AIF or ETFB does not form just after a proteolytic stimulus [26]. We first studied the total depend on their interaction with Cx43. protein amount in whole heart samples and in SSM and IFM fractions from Cx43 WT and Cx43 KI Cx32 animals, respectively (Fig. 1B). No significant differences in protein content were found between both Sub-cellular localization of Cx43, ETFB and AIF genotypes. in cardiomyocyte mitochondria When exploring the mitochondrial distribution, 1–300 lg of SSM protein extract were digested with 0.1 and 0.5 lg of proteinase K to We next investigated if mitochondrial Cx43 is coincident with sub-cel- eliminate the outer mitochondrial membrane and expose the inter- lular localization of AIF and ETFB (Fig. 2). For this purpose, adult iso- membrane matrix and the inner mitochondrial membrane (Fig. 1C). lated cardiomyocytes from both Cx43 WT or Cx43 KI Cx32 mice were AIF was sensitive to proteinase K action, and proteolysis of the immunolabelled with anti-AIF, anti-ETFB or anti-Cx43. Positive label- anchored mature form was detected as a double band of lower molec- ling of mitochondrial proteins SDHA and VDAC were used as control ular weight (Fig. 1C lanes 3 and 4). Overall, no statistically significant (Fig. 2A). Manders’ coefficients M1 and M2 were quantified with 798 ª 2016 The Authors. Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. J. Cell. Mol. Med. Vol 20, No 5, 2016 Fig. 2 Immunolocalization of Cx43, ETFB and AIF proteins in freshly isolated cardiomyocytes from WT and Cx43KICx32 mice. (A) Confocal micro- scopy images showed a specific localization in cardiomyocyte membrane and mitochondria for Cx43; and presence in mitochondria for AIF and ETFB, in agreement with bibliographic information (upper right snapshots). Co-localization (lower snapshots) of Cx43 with AIF or ETFB, and also AIF with ETFB, is quantitatively evaluated in (B) by analysis of composed images using 8–12 z-stacks for each experimental condition, of at least three independent cardiomyocyte isolation events for each genotype. See text for comment on Manders’ coefficients. ImageJ software (Fig. 2B). On the whole, 51.9  3.3% of Cx43 label- dria. Manders’ coefficient between ETFB and AIF labelling was ling co-localized with ETFB, and 70.4  2.3% with AIF, results that 63.9  1.6% in WT versus 65.5  1.3% in KI mice. Manders’ coeffi- suggest a strongest co-localization with AIF protein in the mitochon- cient between AIF and ETFB labelling was 86.4  1.6% in WT versus ª 2016 The Authors. 799 Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. 75.7  3.8% in KI mice. Again, no statistically significant differences protein interaction between AIF and ETFB. However, this same inter- were detected between different genotypes, suggesting a Cx43-inde- play found in mitochondria from transgenic animals lacking Cx43 pendent interaction. suggests that these proteins may interact independently from mtCx43, most probably within a multi-protein complex. These results may serve as a starting point for the investigation of new aspects on Interaction of Cx43, ETFB and AIF in the regulation of mitochondrial function. The identification of Cx43 cardiomyocyte mitochondria interactors has important practical implications as it can help to eluci- date the molecular pathway of ischaemic preconditioning cardiopro- To quantify the possible interaction between our proteins of interest, tection and to propose potential pharmacological targets for the we used an in situ PLA, in which a pair of labelled secondary antibod- development of new therapeutic strategies. ies generates a fluorescent signal only if they are in close proximity Apoptosis-inducing factor is an evolutionarily conserved flavopro- (<40 nm), either in a stable or a transient interaction, being the latter tein, known as a cell death-promoting molecule after proapoptotic a possible explanation for the faint levels detected with co-immuno- signals. Nevertheless, under physiological conditions, AIF is involved precipitation assays. As shown in Figure 3, confocal images obtained in oxidative phosphorylation and redox management [26, 27]. Data after PLA of freshly isolated adult cardiomyocytes demonstrated fluo- gathered from independent studies reveal that AIF-knockout cells and rescent cross-reactivity between Cx43-AIF and Cx43-ETFB (Fig. 3A). tissue-specific knockouts of the Aifm1 gene in the heart and the The number of positive fluorescent spots was 121.25  25.10 and skeletal muscle exhibit a reduction in respiratory complex I protein 65.75  10.89 for Cx43-AIF and Cx43-ETFB pairs, respectively. content or/and activity [28, 29]. Most interestingly, deficient AIF mice Apoptosis-inducing factor-ETFB protein–protein interaction was develop more severe heart injury when subjected to myocardial proved to be persistent and not depending on Cx43, as quantification ischaemia-reperfusion, an effect that has been related to a reduced of the number of fluorescent spots revealed no statistically significant capacity of SSM to scavenge free radicals [30]. Importantly, these differences between genotypes: 533.5  76.82 in Cx43 WT car- evidences are in agreement with functional data on mitochondrial diomyocytes versus 562.5  42.66 in Cx43 KI Cx32 cardiomyocytes complex I activity and ROS handling in previous studies with Cx43- (Fig. 3B). deficient mice [19, 20]. Electron-transfer flavoprotein is a mitochondrial matrix heterodi- mer with alpha- (ETFA, 30 kD) and beta- (ETFB, 28 kD) subunits that AIF and ETFB comigrate in blue-native gels contains one FAD and one AMP molecule per heterodimer [31]. ETF and ETF-ubiquinone oxidoreductase are the two enzymes required for The possible interaction between AIF and ETFB was further character- electron transfer from at least 11 dehydrogenases in the fatty acid ized by blue-native gel electrophoresis followed by mass spectrome- beta-oxidation cascade to the mitochondrial respiratory chain. How- try analysis of each fraction from heart, liver and brain tissues and ever, very few studies have explored the role of ETF or any of its sub- fC57 liver cell line. We found that AIF and ETFB had the same migra- units on mitochondrial respiratory function and ROS generation. In a tion profile in the four samples analysed, suggesting the possible recent study, Rodrigues and Gomes [32] analysed the effect of two interaction and formation of the same macromolecular complex of pathogenic point mutations in human ETFB—which are believed to these proteins (Fig. 4). Moreover, the molecular weight that corre- prevent its interaction with the FAD cofactor and its partner dehydro- sponds to the bands where the two proteins are identified is higher genases—on superoxide and hydrogen peroxide generation. Accord- than the molecular weight of the individual proteins. In addition, we ing to their results, wild-type ETF is able to produce significant found that enoyl-CoA hydratase (ECHS1), a protein implicated in fatty amounts of both superoxide and hydrogen peroxide, but this reactiv- acid beta-oxidation, and citrate synthase (CS), implicated in tricar- ity is decreased when forming part of a tight protein–protein complex. boxylic acid cycle, displayed also the same migration pattern as ETFB The capacity to form such protein complexes depends on ETFB sub- and AIF in the three tissues analysed (Fig. 4). All together, these unit, as disease mutants show higher production of ROS. In agree- results reinforce our data and are in agreement with previous studies ment with a multi-protein scenario, other studies have shown that that describe the association between fatty acid oxidation enzymes fatty acid oxidation enzymes and respiratory chain complexes are and respiratory complexes (see Discussion). physically associated, comprising multifunctional electron-transfer chain complexes [33]. Our data provide a new lead about the management of the redox Discussion status in cardiomyocyte mitochondria showing that mtCx43 interacts preferentially with AIF and also with ETF—through its beta-subunit. In In this study, we investigated the interaction of Cx43 with new molec- the context of available data on the roles of these proteins in the mod- ular partners in mitochondria from adult mouse cardiomyocytes ulation of mitochondrial respiration and ROS production, these under non-stress conditions. By using co-immunoprecipitation results may indicate the participation of mtCx43, ETF and AIF in mul- assays and second-generation proteomics, we provide evidence that ti-protein complexes with important actions on mitochondrial physiol- mtCx43 interacts with AIF and ETFB, two unrelated proteins to date. ogy. Reactive oxygen species and mitochondrial function are strongly This interaction was further supported by colocalization and comigra- interconnected both in normal and dysfunctional states. The concen- tion studies that also demonstrate a previously unsuspected protein– tration of these reactive species shifts the overall redox balance 800 ª 2016 The Authors. Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. J. Cell. Mol. Med. Vol 20, No 5, 2016 Fig. 3 In situ proximity ligation assay of Cx43, ETFB and AIF proteins in freshly isolated cardiomyocytes from WT and Cx43KICx32 mice. (A) Cells treated with no primary antibody (negative control) or with only anti-Cx43 were used to determine the signal specificity, whereas Cx43-PanCadherin and Cx43-SDHA pairs of proteins were used as a positive control. In the lower group of confocal images, Cx43 interacts with AIF and ETFB, as well as AIF with ETFB in a Cx43-independent manner. (B) Assessment of the number of signals detected with composed images using 8–12 z-stacks for each experimental condition, of at least three independent cardiomyocyte isolation events for each genotype. ª 2016 The Authors. 801 Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. Fig. 4 AIF, ETFB, ECHS1 and CS protein patterns in blue-native gels. High correla- tion in the distribution patterns for pro- teins AIF, ETFB, ECHS1 and CS is showed according with their BNE mobility in heart, liver, brain and fC57 the cell line. between their role as signalling molecules to their role in promoting Acknowledgements deleterious effects. The control of this balance may depend on various interacting proteins working together, most probably in localized We thank Angeles Rojas for technical assistance and Marisol Ruiz-Meana for dynamic platforms. critical discussion of the article. AD, EN, EC and ML performed the research; Further studies will be necessary to determine the composition AD, EN, EC, ML, JV and DGD analysed the data; AD, JV and DGD wrote the and functions of these putative multi-protein complexes, and in par- article. All authors approved the final version. This work was funded by CD12/ ticular whether its formation may explain the important modulatory 00493 Sara Borrell grant from the Institute of Health Carlos III to A.D and RIC effect of mtCx43 on mitochondrial respiration and ROS production. RD12/0042/0021 and SAF 2008-03067 from the Spanish Ministry of Science. Our present data indicate that mtCx43 may not be an indispensable element in these complexes, as its absence does not appear to modify the cellular localization of ETFB or AIF, nor the interaction between Conflicts of interest them. The authors confirm that there are no conflicts of interest. Conclusion Supporting information The present study proposes the interaction of mtCx43 with ETFB and AIF to form multi-protein complexes with effects on mitochondrial Additional Supporting Information may be found in the online respiration and ROS signalling. Elucidation of the composition and version of this article: functions of such complexes could have therapeutic implications given the important role of Cx43 on preconditioning cardioprotection. Table S1 Initial list of candidates from the proteomics analysis. 802 ª 2016 The Authors. Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine. J. Cell. Mol. Med. Vol 20, No 5, 2016 References 1. Jansen JA, van Veen TAB, de Bakker JMT, pling. Prog Biophys Mol Biol. 2007; 94: 23. Gnaiger E, Kuznetsov AV. Mitochondrial respi- et al. Cardiac connexins and impulse propa- 219–32. ration at low levels of oxygen and cytochrome gation. J Mol Cell Cardiol. 2010; 48: 76–82. 13. Giepmans BNG. Gap junctions and con- c. Biochem Soc Trans. 2002; 30: 252–8. 2. Nielsen MS, Axelsen LN, Sorgen PL, et al. nexin-interacting proteins. Cardiovasc Res. 24. Bolte S, Cordelieres FP. A guided tour Gap junctions. Compr Physiol. 2012; 2: 2004; 62: 233–45. into subcellular colocalization analysis in 1981–2035. 14. Garcıa-Dorado D, Rodrıguez-Sinovas A, light microscopy. J Microsc. 2006; 224: 3. Veenstra RD, Wang HZ, Westphale EM, Ruiz-Meana M. Gap junction-mediated 213–32. et al. Multiple connexins confer distinct reg- spread of cell injury and death during 25. Allalou A, W€ahlby C. BlobFinder, a tool for ulatory and conductance properties of gap myocardial ischemia-reperfusion. Cardio- fluorescence microscopy image cytometry. junctions in developing heart. Circ Res. vasc Res. 2004; 61: 386–401. Comput Methods Programs Biomed. 2009; 1992; 71: 1277–83. 15. Rodrıguez-Sinovas A, Sanchez JA, 94: 58–65. 4. Wang HZ, Veenstra RD. Monovalent ion Gonzalez-Loyola A, et al. Effects of substi- 26. Modjtahedi N, Giordanetto F, Madeo F, selectivity sequences of the rat connexin43 tution of Cx43 by Cx32 on myocardial et al. Apoptosis-inducing factor: vital and gap junction channel. J Gen Physiol. 1997; energy metabolism, tolerance to ischaemia lethal. Trends Cell Biol. 2006; 16: 264–72. 109: 491–507. and preconditioning protection. J Physiol. 27. Joza N, Pospisilik JA, Hangen E, et al. AIF: 5. Harris AL. Emerging issues of connexin 2010; 588: 1139–51. not just an apoptosis-inducing factor. Ann N channels: biophysics fills the gap. Q Rev Bio- 16. Boengler K, Dodoni G, Rodriguez-Sinovas Y Acad Sci. 2009; 1171: 2–11. phys. 2001; 34: 325–472. A, et al. Connexin 43 in cardiomyocyte 28. Joza N, Oudit GY, Brown D, et al. Muscle- 6. Jiang JX, Gu S. Gap junction- and mitochondria and its increase by ischemic specific loss of apoptosis-inducing factor hemichannel-independent actions of connex- preconditioning. Cardiovasc Res. 2005; 67: leads to mitochondrial dysfunction, skeletal ins. Biochim Biophys Acta. 2005; 1711: 234–44. muscle atrophy, and dilated cardiomyopa- 208–14. 17. Boengler K, Stahlhofen S, van de Sand A, thy. Mol Cell Biol. 2005; 25: 10261–72. 7. Agullo-Pascual E, Delmar M. The non- et al. Presence of connexin 43 in subsar- 29. Brown D, Yu BD, Joza N, et al. Loss of Aif canonical functions of Cx43 in the heart. J colemmal, but not in interfibrillar cardiomy- function causes cell death in the mouse Membr Biol. 2012; 245: 477–82. ocyte mitochondria. Basic Res Cardiol. embryo, but the temporal progression of 8. Jansen JA, Noorman M, Musa H, et al. 2009; 104: 141–7. patterning is normal. Proc Natl Acad Sci Reduced heterogeneous expression of Cx43 18. Miro-Casas E, Ruiz-Meana M, Agullo E, USA. 2006; 103: 9918–23. results in decreased Nav1.5 expression and et al. Connexin43 in cardiomyocyte mito- 30. van Empel VP, Bertrand AT, van der Nagel reduced sodium current that accounts for chondria contributes to mitochondrial potas- R, et al. Downregulation of apoptosis-indu- arrhythmia vulnerability in conditional Cx43 sium uptake. Cardiovasc Res. 2009; 83: cing factor in harlequin mutant mice sensi- knockout mice. Heart Rhythm. 2012; 9: 747–56. tizes the myocardium to oxidative stress- 600–7. 19. Boengler K, Ruiz-Meana M, Gent S, et al. related cell death and pressure overload- 9. Danik SB, Rosner G, Lader J, et al. Elec- Mitochondrial connexin 43 impacts on respi- induced decompensation. Circ Res. 2005; trical remodeling contributes to complex ratory complex I activity and mitochondrial 96: e92–101. tachyarrhythmias in connexin43-deficient oxygen consumption. J Cell Mol Med. 2012; 31. Roberts DL, Frerman FE, Kim J-JP. mouse hearts. FASEB J. 2008; 22: 1204– 16: 1649–55. Three-dimensional structure of human 12. 20. Heinzel FR, Luo Y, Li X, et al. Impairment electron transfer flavoprotein to 2.1- A res- 10. Doble BW, Dang X, Ping P, et al. Phospho- of diazoxide-induced formation of reactive olution. Proc Natl Acad Sci USA. 1996; rylation of serine 262 in the gap junction oxygen species and loss of cardioprotection 93: 14355. protein connexin-43 regulates DNA synthe- in connexin 43 deficient mice. Circ Res. 32. Rodrigues JV, Gomes CM. Mechanism of sis in cell-cell contact forming cardiomy- 2005; 97: 583–6. superoxide and hydrogen peroxide genera- ocytes. J Cell Sci. 2004; 117: 507–14. 21. Ruiz-Meana M, Nu~nez E, Miro-Casas E, tion by human electron-transfer flavoprotein 11. Reaume AG, de Sousa PA, Kulkarni S, et al. Ischemic preconditioning protects car- and pathological variants. Free Radic Biol et al. Cardiac malformation in neonatal mice diomyocyte mitochondria through mecha- Med. 2012; 53: 12–9. lacking connexin43. Science. 1995; 267: nisms independent of cytosol. J Mol Cell 33. Wang Y, Mohsen A-W, Mihalik SJ, et al. 1831–4. Cardiol. 2014; 68: 79–88. Evidence for physical association of mito- 12. Rodrıguez-Sinovas A, Cabestrero A, Lopez 22. Plum A, Hallas G, Magin T, et al. Unique chondrial fatty acid oxidation and oxidative D, et al. The modulatory effects of connexin and shared functions of different connexins phosphorylation complexes. J Biol Chem. 43 on cell death/survival beyond cell cou- in mice. Curr Biol CB. 2000; 10: 1083–91. 2010; 285: 29834–41. ª 2016 The Authors. 803 Journal of Cellular and Molecular Medicine published by John Wiley & Sons Ltd and Foundation for Cellular and Molecular Medicine.