Desmin and αB-crystallin interplay in the maintenance of mitochondrial homeostasis and cardiomyocyte survival

Desmin, the muscle-specific intermediate filament protein, forms a three-dimensional scaffold that interconnects the contractile apparatus to the nucleus, cellular organelles and the sarcolemma (Capetanaki et al., 2007). The desmin scaffold is a primary target for cardiomyopathy and heart failure both in mice and humans (Goldfarb et al., 1998; Capetanaki et al., 2007, 2015). Mice deficient for desmin develop both skeletal and myocardial defects that are strongly linked to impaired mitochondrial morphology and function (Milner et al., 1996, 1999, 2000; Li et al., 1996; Papathanasiou et al., 2015). These mitochondrial abnormalities lead to cardiomyocyte death and myocardial degeneration, accompanied by inflammation and fibrosis, resulting in dilated cardiomyopathy (DCM) and heart failure (Milner et al., 2000; Mavroidis and Capetanaki 2002; Capetanaki et al., 2007; Psarras et al., 2012; Mavroidis et al., 2015). In humans, over 70 desmin mutations have been linked to desmin-related cardiomyopathies, of which DCM is the most frequent (Capetanaki et al., 2015; van Spaendonck-Zwarts et al., 2011). In addition, desmin aggregate formation due to caspase cleavage is an important mediator of TNF-α-induced heart failure (Panagopoulou et al., 2008). Importantly, regardless of the specific mutations, human hearts undergoing DCM-linked failure display mitochondrial and other structural abnormalities, similar to those in desmin-deficient mice and mice undergoing TNF-α-induced failure. Furthermore, the R120G mutation in the α-crystallin Β-chain (αB-crystallin; encoded by CRYAB), an abundant small heat-shock protein, also causes the same desmin-aggregate-related DCM and heart failure with similar mitochondrial defects (Wang et al., 2001; Vicart et al., 1998). Mitochondrial abnormalities seem to be a common feature of diseases in which desmin and/or αB-crystallin deficiencies are implicated. Previous observations suggest a physical association of desmin with mitochondria, as well as an involvement of desmin in mitochondrial homeostasis (Stromer and Bendayan, 1990; Reipert et al., 1999; Hnia et al., 2011), but the underlying mechanisms remain elusive. αΒ-crystallin associates with all three cytoskeletal networks – microfilaments (Bennardini et al., 1992; Singh et al., 2007), microtubules (Arai and Atomi, 1997; Houck and Clark, 2010) and intermediate filaments (Nicholl and Quinlan, 1994; Iwaki et al., 1989; Wisniewski and Goldman, 1998) (reviewed in Wettstein et al., 2012) – and specifically desmin (Bennardini et al., 1992; Nicholl and Quinlan, 1994; Perng et al., 1999). There is also evidence that a fraction of αΒ-crystallin is located in mitochondria (Fountoulakis et al., 2005; Martindale et al., 2005; Maloyan et al., 2005; Mitra et al., 2013); however, its role there is not completely understood.

Mitochondria are double-membrane-bounded organelles that are involved in important cellular processes. Proper assembly, maintenance and function of the various mitochondrial complexes, including oligomers of optic atrophy 1 (OPA1) (Frezza et al., 2006), F₁F₀-ATP synthase (Strauss et al., 2008) and the recently identified mitochondrial contact site and cristae organizing system (MICOS) (Harner et al., 2011; Hoppins et al., 2011; von der Malsburg et al., 2011) are vital to inner membrane organization, and to proper and efficient mitochondrial function. Mitofilin (Mic60; also known as IMMT) (Gieffers et al., 1997; Odgren et al., 1996), the core component of MICOS, interacts with several inner and outer mitochondrial membrane (IMM and OMM, respectively) proteins (von der Malsburg et al., 2011; Darshi et al., 2011; Xie et al., 2007; Harner et al., 2011; Hoppins et al., 2011; van der Laan et al., 2012), thus maintaining IMM morphology and contact sites between IMM and OMM, and affecting mitochondrial function, mitochondrial (mt)DNA integrity and protein and metabolite trafficking.

To address the role of desmin and αB-crystallin, and their potential interplay in mitochondrial homeostasis and cardioprotection, we investigated the consequences of αΒ-crystallin overexpression in the desmin-deficient cardiomyopathy model. We demonstrated that cardiac-specific αB-crystallin overexpression rescues desmin-deficient heart failure and that this takes place through maintenance of proper mitochondrial composition, structure and function, through inhibition of abnormal mitochondrial permeability transition pore (mPTP) activation and through cardiomyocyte death prevention. Moreover, we found that among other sites, both desmin and αΒ-crystallin localize at sarcoplasmic reticulum (SR)–mitochondria-associated membranes (MAMs), where they interact with VDAC. Furthermore, we have unraveled a potential link of desmin and αB-crystallin with mitochondrial morphology and bioenergetics through their interaction with Mic60 and ATP synthase.

In order to examine the effect of αΒ-crystallin overexpression in the Des^−/− myocardium, we generated transgenic mice overexpressing αΒ-crystallin specifically in the heart under the control of the α-myosin heavy chain (αMHC) promoter. Among different lines generated, two were chosen for further investigation, namely αΒCry4 and αΒCry6, with high and low αB-crystallin expression levels respectively (Fig. 1A; Fig. S1A). These mice were crossed with Des^−/− mice to generate αΒ-crystallin overexpressing Des^−/− mice (Des^−/−αΒCry). Quantification of total αB-crystallin protein levels from hearts of wild-type, Des^−/− and transgenic lines revealed 1.3-fold higher αB-crystallin levels in Des^−/− mice, a 5.4-fold increase in line αΒCry4 and a 2.6-fold increase in line αΒCry6 in comparison to wild type.

It has been shown that, upon stress, αΒ-crystallin translocates from a cytosolic pool to Z-discs (Golenhofen et al., 1998), where it colocalizes with desmin and α-actinin. Immunofluorescence microscopy was performed to determine whether desmin deficiency affects the Z-disc localization of αΒ-crystallin. We found that, in the unstressed wild-type myocardium, αB-crystallin colocalized only partially with desmin and α-actinin at the Z-disc, and was also located at the intercalated discs (Fig. S1B). This localization pattern was enhanced under stress as a result of the mice swimming (Fig. 1D). In the absence of desmin, the Z-disc localization of αB-crystallin was mostly lost, but its localization at intercalated discs was retained (Fig. 1D), possibly owing to other potential binding partners at the intercalated discs. Overexpression of αB-crystallin in the Des^−/−αBCry4 myocardium rescued its Z-disc localization (Fig. 1D), and more importantly, a large amount of αB-crystallin was located in mitochondria (Fig. 1Bb). Mitochondrial localization of αB-crystallin also occurred in wild-type cardiomyocytes but could not be easily detected. Under the stress of swimming, the total protein levels of both desmin and αB-crystallin were increased (Fig. 1Ba,Ca,Cb), and this allowed for easier detection of αB-crystallin in the mitochondrial fraction in both wild-type and Des^−/− myocardium (Fig. 1Cc). This was more obvious in the corresponding αB-crystallin-overexpressing myocardium [wild-type (wt)αBCry4 and Des^−/−αBCry4]; however, the ratio of mitochondrial to cytosolic levels remained the same.

Inflammation, fibrosis and calcification are hallmarks of Des^−/− mouse myocardial degeneration and necrosis (Milner et al., 1996; Mavroidis and Capetanaki, 2002; Psarras et al., 2012). Dissection of Des^−/−αΒCry4 mice revealed a complete absence of extensive calcification in the myocardium. Further histological analysis revealed no inflammatory infiltration and a minimal amount of fibrosis in the ventricular walls (Fig. 2A–D). Quantification of histological section images confirmed the significant reduction of fibrosis in the Des^−/−αΒCry myocardium (Fig. 2C). Comparison of the two transgenic lines Des^−/−αΒCry4 and Des^−/−αΒCry6 revealed that a 5.4-fold overexpression of αB-crystallin was more beneficial when compared to a 2.6-fold overexpression. Therefore, the subsequent studies were done using line Des^−/−αΒCry4, designated from now on as Des^−/−αΒCry.

Evans-Blue dye (EBD) analysis was performed to detect cellular membrane permeability. Wild-type cardiomyocytes were impermeable to EBD because their membranes were intact (Fig. 2E,F). In contrast, Des^−/− hearts appeared to have considerably large areas of EBD-positive cells, especially in regions with inflammatory infiltration. In Des^−/−αBCry myocardium there were no EBD-positive cardiomyocytes, leading to the conclusion that αB-crystallin overexpression protects Des^−/− myocardium from necrotic cell death.

Heart failure development with aging is characteristic of Des^−/− hearts. To explore the effect of αΒ-crystallin overexpression on cardiac function, we performed standard morphometric analysis and two-dimensional directed M-mode echocardiography (Fig. 3; Table S1). Des^−/−αBCry mice showed a significant improvement in all parameters tested in comparison to Des^−/− mice. The improvement in left ventricular function of Des^−/−αBCry mice was very extensive, reaching almost wild-type levels (96% of the wild-type fractional shortening; Fig. 3Bc), consistent with the improvement in left-ventricular-end systolic diameter, which was also comparable to wild-type levels (Fig. 3Bb). Posterior wall thickness was also increased by 35% in the Des^−/−αBCry compared to the Des^−/− group, consistent with the absence of degeneration in these hearts (Fig. 3Bd). Finally, the ratio of left ventricular radius to posterior wall thickness, an indicator of left ventricular wall stress, was significantly lower in Des^−/−αBCry mice compared to that in Des^−/− mice (Fig. 3Be).

In order to investigate whether this cardioprotection also occurs under stress conditions, wild-type, Des^−/− and Des^−/−αΒCry mice were subjected to an obligatory swimming exercise protocol for 24 days. The Des^−/− mice had a considerably reduced ability to handle stress because 50% of them died during swimming (in comparison to 0% of wild type) (Fig. 3C). The survival rate of Des^−/−αΒCry mice was 100%, strongly demonstrating the powerful cardioprotective action of αΒ-crystallin overexpression against stress.

In an effort to understand the mechanism by which the overexpression of αΒ-crystallin protects Des^−/− myocardium against degeneration and improves cardiac function, we initially examined its effects on cardiomyocyte ultrastructure. Des^−/− cardiomyocytes are characterized by various early-appearing mitochondrial defects (Milner et al., 2000), followed by other structural abnormalities and loss of muscle integrity (Fig. 3D). Mitochondrial abnormalities include disturbed shape and positioning, fragmentation, aggregation, swelling of the mitochondrial matrix, disruption and fragmentation of the cristae and degeneration (Fig. 3E). In contrast, none of these ultrastructural defects are present in Des^−/−αΒCry mice. Myofibrils were more properly aligned and mitochondria, located parallel to myofibrils, appeared normal, without signs of fragmentation, cristae abnormalities or destruction. These data suggest that mitochondria are an important target of αΒ-crystallin cytoprotection.

To further investigate the mechanism by which αΒ-crystallin protects against mitochondrial defects, we analyzed its effects on adult cardiomyocyte reactive oxygen species (ROS) levels using the molecular probe CM-H₂DCFDA (Fig. 4A,B). Des^−/− cardiomyocytes showed a significantly higher degree of ROS over those in the wild-type, whereas Des^−/−αΒCry cardiomyocytes were characterized by a lower degree of ROS. Under oxidative stress conditions, simulated by incubation with H₂O₂, there was a significant decrease of ROS levels in both Des^−/−αΒCry and wtαBCry compared to Des^−/− and wild-type cells respectively, indicative of the protection provided by αΒ-crystallin overexpression. To further investigate the mechanism of anti-oxidant protection by αΒ-crystallin, we measured the levels of glutathione (GSH). In Des^−/− hearts, glutathione levels are decreased by 33% and the GSH:glutathione-disulfide (GSSG) ratio by almost 50% in comparison to wild type (Fig. 4C). Overexpression of αΒ-crystallin increased glutathione and GSH:GSSG levels to 87% and 95%, respectively, of those in wild type, suggesting that, at least partially, the mechanism of anti-oxidant protection by αΒ-crystallin occurs through an increase of reduced glutathione levels.

Mitochondrial swelling and increased ROS levels have been linked to abnormal activation of the mPTP. Therefore, we investigated whether the above demonstrated mitoprotection and cardioprotection by αΒ-crystallin overexpression takes place through inhibition of mPTP activation. First, we determined the sensitivity of isolated mitochondria to increased Ca²⁺ levels in the presence or absence of the mPTP inhibitor cyclosporin A (CsA). As expected, Des^−/− mitochondria swelled more than wild-type mitochondria at the same Ca²⁺ concentration, as measured by the decrease in absorbance at 540 nm (Table S2). By contrast, Des^−/−αBCry mitochondria showed a significant decrease in swelling, indicating that overexpression of αB-crystallin improves the diminished capacity of Des^−/− mitochondria to resist exposure to Ca²⁺. Addition of CsA before the Ca²⁺ challenge significantly abrogated mitochondrial swelling. This finding demonstrates that first, the reduced ability of Des^−/− mitochondria to resist increased Ca²⁺ levels (Weisleder et al., 2004) is mPTP dependent and second, the mechanism of mitoprotection and inhibition of necrotic cell death of Des^−/− myocardium by αB-crystallin overexpression involves suppression of mPTP activation.

In the Des^−/− heart, mitochondrial abnormalities, including swelling, can lead to dissipation of mitochondrial membrane potential (Δψ_m), a critical event in the process of cell death. To address this possibility, we incubated adult cardiomyocytes with tetramethylrhodamine methyl ester (TMRM), a potentiometric indicator that is targeted to mitochondria. The TMRM intensity was 21% lower in Des^−/− cardiomyocytes compared to that in wild type, indicating reduced Δψ_m (Fig. 4D,E). Time-lapse confocal scanning showed a more rapid loss of TMRM in Des^−/− mitochondria relative to that in wild-type mitochondria, consistent with mitochondrial membrane depolarization (Fig. 4F; Fig. S2). αB-crystallin overexpression in Des^−/− myocardium significantly increased both the intensity and retention time of TMRM. To examine whether the accelerated TMRM loss in Des^−/− cardiomyocytes was due to mPTP activation, we treated cardiomyocytes with CsA, significantly attenuating TMRM loss.

In an effort to understand the underlying mechanisms responsible for both the observed mitochondrial defects in the absence of desmin and, most importantly, the unprecedented extensive rescue of these defects by αΒ-crystallin overexpression, we conducted proteomic analysis to identify any crucial alterations in the mitochondrial proteome of the different genotypes. The results revealed diminished levels of enzymes involved in the citric acid cycle, lipid metabolism and fatty acid β-oxidation in the Des^−/− compared to wild-type mitochondria (Table 1; Table S3, Fig. S3). Moreover, in Des^−/− mitochondria, the levels of many subunits and components of the electron transport chain were decreased, along with components of complexes linking different metabolic pathways. The most extensive change was found with the cristae junction protein Mic60, the levels of which were reduced by 78% compared to those in wild type (Table 1, Fig. 5A,B). On the contrary, in the mitochondria of Des^−/−αΒCry myocardium, most of the proteins that showed decreased levels in Des^−/− mice, including Mic60 protein, were apparently restored (Table 1, Fig 5; Table S4). In agreement with the western blotting data (Fig. 1B), proteomic analysis revealed that much higher levels of αΒ-crystallin localize to the mitochondria of the Des^−/−αΒCry hearts in comparison to those of Des^−/− hearts. The levels of some of the differentially expressed proteins were also evaluated by western blotting (Fig. 5A,B). In addition, we investigated the mitochondrial levels of adenine nucleotide translocator (ANT), an IMM protein with known mitochondrial and cardiomyocyte function (Narula et al., 2011; Vogelpohl et al., 2011) that cannot be easily detected by proteomic analysis. This analysis revealed that, in contrast to total ANT levels, mitochondrial ANT levels were reduced in Des^−/− myocardium, a defect ameliorated by αB-crystallin overexpression (Fig. 5A,B).

As demonstrated above, one of the most interesting proteins downregulated in Des^−/− mitochondria is Mic60 (Table 1, Fig. 5A), consistent with the shown cristae morphology defects (Fig. 3D,E). Given that both these defects are rescued by αΒ-crystallin overexpression, we aimed to further unravel the underlying mechanism. We used GST-tagged desmin and co-immunoprecipitation to examine whether desmin interacts with Mic60, and therefore, whether loss of this interaction in Des^−/− cardiomyocytes affects its proper mitochondrial targeting and/or stabilization. Indeed, we found that desmin interacts with Mic60, and as expected, with αΒ-crystallin (Fig. 5C,D). More importantly, we demonstrated that αΒ-crystallin also associates with Mic60 (Fig. 5D), both in wild-type and Des^−/− myocardial extracts, strongly suggesting that Mic60 is one of the αΒ-crystallin targets in the Des^−/− mitochondria. Using similar immunoprecipitation studies, (and immunoprecipitation coupled to proteomic analysis in the case of desmin), we found that the β subunit of the F1 catalytic domain of ATP synthase, another protein important for determining cristae structure (Paumard et al., 2002) and mitochondrial function (Zick et al., 2009), interacts with desmin as well as with αΒ-crystallin (Fig. 5C,D). Loss of this interaction in the absence of desmin, combined with the disrupted cristae and depolarized Δψ_m, is expected to impact the ATP content, which we indeed confirmed to be 40% lower in the Des^−/− cardiomyocytes compared to that in wild type (Fig. 5E). αΒ-crystallin overexpression increased this ATP content to 90% of the wild-type levels.

These findings support the view that the loss of desmin association with these proteins could be a key mediator of pathologies resulting from loss of desmin. Given that αΒ-crystallin also associates with these proteins, its higher levels in the Des^−/−αΒCry cardiomyocytes, in comparison to its basal levels in Des^−/− cells, might be able to compensate for the impact of desmin loss on Mic60 and ATP synthase function.

The question raised by the amelioration of Des^−/− mitochondrial pathology by αB-crystallin overexpression is how can desmin and αB-crystallin, two mainly cytoplasmic proteins, interact with IMM proteins and impact mitochondrial homeostasis. To address this question, we first performed protease accessibility assays to examine whether a fraction of desmin and αB-crystallin is located inside the mitochondria. Crude mitochondria from wild-type hearts were isolated and digested with proteinase K. As shown in Fig. 6A, both desmin and αB-crystallin were degraded by proteinase K, whereas the IMM protein ANT and the OMM protein VDAC remain intact. The digestion pattern of desmin and αB-crystallin resembled that of mitofusin 2, an OMM protein, as detected by an antibody that recognizes a part of mitofusin 2 that is exposed to the cytosol. In contrast, overexpressed αB-crystallin appears to localize inside mitochondria (Fig. 6Ab).

Next, we performed subcellular fractionation of rat and mouse hearts to further determine the localization of desmin. Interestingly, although we found desmin in the SR fraction, it was mainly enriched in the SR–MAMs (Fig. 6B). In contrast, αB-crystallin was found enriched more in the SR fraction and less so in the MAMs and pure mitochondrial fraction. Furthermore, overexpressed αB-crystallin was also enriched in the SR fraction but even more so in the MAMs and in pure mitochondria (Fig. 6C). The localization of desmin was confirmed by using immunogold labeling studies (Fig. 6D; Fig. S4). In addition to the expected places, such as intercalated discs (Fig. S4) and Z-lines, desmin was found to associate with mitochondria and membranous structures, most likely SR, in close proximity to mitochondria.

A previous yeast two-hybrid screening of a heart cDNA library using the entire desmin molecule as bait, conducted in our laboratory, had shown binding of desmin to VDAC (our unpublished results). In addition to its mitochondrial localization, VDAC is also enriched in MAMs, and furthermore it has been found to interact with Mic60 (Hoppins et al., 2011). Therefore, we reconfirmed the binding of VDAC to desmin using both co-immunoprecipitation studies as well as GST pulldown assays (Fig. 7A,B). To determine the corresponding binding regions of VDAC and desmin, we generated several deletion constructs and used them in reciprocal GST pulldown experiments. As demonstrated in Fig. 7D, the N-terminal domain of VDAC is necessary for its binding to desmin, whereas the C-terminal tail domain of desmin and, to a lesser extent, the rod domain, are sufficient for association of desmin with VDAC. The finding of desmin–VDAC binding was further strengthened by double immunogold labeling studies, which confirmed the close proximity of these two proteins in vivo (Fig. 6D; Fig. S4). αB-crystallin was also co-immunoprecipitated with VDAC, consistent with earlier reports (Mitra et al., 2013). The demonstrated associations of desmin and αB-crystallin with VDAC, Mic60 and ATP synthase could somehow facilitate the formation and/or stabilization of a scaffold-like super complex extending from SR–mitochondria contact sites to MICOS and ATP complexes.

We speculated that desmin and αB-crystallin association with Mic60 could affect the formation or stabilization of Mic60-containing complexes. We analyzed crude mitochondria from mouse heart that had been solubilized with digitonin by using blue native PAGE in the first dimension and SDS-PAGE in the second dimension, followed by detection of Mic60 by western blotting (Fig. 7E,F). In the absence of desmin, the number of the protein complexes containing Mic60 was much lower compared to those in wild type, consistent with the diminished Mic60 protein levels found in Des^−/− mitochondria (Table 1, Fig. 5A). Importantly, as determined by using much lower protein levels, the higher-molecular-mass complexes of Mic60 were proportionally less than the corresponding lower-molecular-mass complexes, indicating a possible facilitating role of desmin in the formation or stabilization of these complexes (Fig. 7F; Fig. S4). All these data are consistent with the cristae defects of Des^−/− mitochondria (Fig. 3). Overexpression of αB-crystallin in the absence of desmin restored the abundance of all the different-sized Mic60 complexes to their normal levels. Importantly, desmin and αB-crystallin appeared to be in the same complexes as Mic60 and VDAC, supporting the idea of a functional role in the interface of SR with OMMs and IMMs. Under swimming stress conditions, we obtained similar results with only a slight shift in the molecular mass of VDAC complexes, as observed in the Des^−/− mitochondria (Fig. 7F). This potentially reflects a partial dissociation of Mic60 and VDAC complexes.

We have demonstrated that αB-crystallin overexpression in the Des^−/− myocardium completely halts the development of heart failure. The present results show that, although the mechanisms of cardioprotection might be occurring at various levels, αB-crystallin overexpression definitively provides extensive mitoprotection to Des^−/− cardiomyocytes. The earliest features of the Des^−/− pathology are mitochondrial structural perturbations, evident as early as the second week after birth, before any other cardiac dysfunction arises (Milner et al., 2000). These findings are further supported by additional studies showing that overexpression of Bcl-2, which has a known protective action on mitochondria, provides significant improvement of Des^−/− cardiomyopathy (Weisleder et al., 2004). Our studies demonstrated an increase in oxidative stress, an increased activation of mPTP and a decreased Δψ_m in the absence of desmin. All these mitochondrial defects are corrected in the Des^−/−αBCry myocardium, in which mitochondria have an apparently wild-type phenotype. It has been shown that the αB-crystallin mutation R120G, known to cause desminopathy characterized by desmin and αB-crystallin aggregation, also leads to redox imbalance (Rajasekaran et al., 2007). Defining the extent to which these aggregates contribute to the development and progression of the disease is of great importance. Clearing away the aggregates improves this model of proteotoxicity (Li et al., 2011; Su et al., 2015; Cabet et al., 2015; Pattison et al., 2011; Gupta et al., 2014), but it does not compensate for the loss of function of the aggregated proteins. Previous and present data with the Des^−/− model – which shows no desmin and αB-crystallin aggregates but displays similar cellular and functional defects with those of desmin- and αB-crystallin-related cardiomyopathy (Milner et al., 1996, 1999, 2000; Li et al., 1996; Pattison et al., 2011; Wang et al., 2003; Wang and Robbins, 2006; McLendon and Robbins, 2011) – suggest alternative causes for this pathology, in addition to the gain of toxic function (Bhuiyan et al., 2013).

The association of desmin with αΒ-crystallin and the fact that mutations in either one of them lead to heart failure with common characteristics suggest a potential compensatory interplay between these two proteins in mitoprotection and consequently cardioprotection. Indeed, we found that both proteins are protecting mitochondria by, among other possible mechanisms, maintaining proper composition of mitochondrial proteins. This could take place either though facilitation of efficient protein targeting and transport of the nucleus-encoded mitochondrial proteins into the mitochondria, or through the support of the formation and/or stabilization of mitochondrial protein complexes. Up to now, the former possibility was considered more probable for a desmin-based scaffold, which, although it does associate with mitochondria through plectin (Winter et al., 2008), is mainly cytosolic. The present study identifies a new way by which desmin can regulate mitochondrial homeostasis. The contact sites where SR links to the OMMs and IMMs are established sites for multiple cellular processes, including Ca²⁺ and metabolite transfer, lipid metabolism, mitochondrial shape regulation, and autophagosome and inflammasome formation (Naon and Scorrano, 2014). The desmin scaffold, located at these contact sites, could contribute to all of the above processes, and particularly to the maintenance of SR–mitochondria proximity. This proximity allows the desmin scaffold, together with αΒ-crystallin, to facilitate directed protein and metabolite targeting to mitochondria. As shown here, and as previously suggested by others (Fountoulakis et al., 2005; Martindale et al., 2005; Maloyan et al., 2005), αB-crystallin does localize in and outside of mitochondria, and thus, could support multiple processes. The potential compensatory interplay between desmin and αΒ-crystallin is further supported by the fact that both of them associate with important mitochondrial proteins. Our data indicate that desmin plays a possible role in the formation and/or stabilization of Mic60 macromolecular complexes. In the absence of desmin, this role is restored by overexpressed αB-crystallin. As discussed above, Mic60 is crucial for the formation and function of the extended cristae membrane structures, the main sites of ATP synthesis by oxidative phosphorylation (Gilkerson et al., 2003; Vogel et al., 2006; Wurm and Jakobs, 2006). In addition, it is located at the contact sites (Harner et al., 2011) and has recently been implicated in mitochondrial protein import (van der Laan et al., 2012). It is thus conceivable that the decrease of Mic60 in the absence of desmin could compromise protein import into mitochondria. Our proteomic analysis revealed that the levels of many enzymes involved in aerobic respiration are diminished in the Des^−/− cardiomyocytes (Table 1; Tables S3, S4), whereas αΒ-crystallin overexpression restores them, thus conferring to the cell high-energy electrons in the form of NADH and FADH₂, which can be used for ATP generation by oxidative phosphorylation.

In addition to Mic60, F₁F₀-ATP synthase affects cristae membrane structure by imposing curvature through dimerization and oligomerization (Strauss et al., 2008; Dudkina et al., 2006; Minauro-Sanmiguel et al., 2005), thus providing the required conditions to enhance the catalytic capacity of the oxidative phosphorylation system (Strauss et al., 2008). Recently, mitochondrial cristae shape has been found to determine respiratory chain super-complex assembly and respiratory efficiency (Cogliati et al., 2013). Such efficiency minimizes the downstream ROS production discussed above, consistent with the proposed upstream action of αΒ-crystallin. All these data are also consistent with the significant decrease of ATP levels in Des^−/− hearts and their restoration by αΒ-crystallin overexpression. Moreover, recent work indicates that the mPTP can be formed from ATP synthase dimers (Giorgio et al., 2013). The interaction of desmin and αΒ-crystallin with both ATP synthase and VDAC suggests a possible role of these proteins in the regulation of mitochondrial bioenergetics and mPTP opening.

The present data suggest that in normal heart the desmin cytoskeletal network could facilitate the mitoprotective and cytoprotective roles of the endogenous αΒ-crystallin. In the absence of desmin, αB-crystallin loses its Z-disc localization and thus the ability to reach the required high concentrations at a Z-disc-proximal SR– or endoplasmic reticulum (ER)–mitochondrial sites for its proper function. Indeed, much higher levels of αB-crystallin are required to facilitate the same level of protection in the absence of desmin. This is consistent with the presently proposed involvement of both proteins in the ER–mitochondria organizing network, which might constitute only part of the function of these partners in mitochondrial behavior.

Des^−/− mice had been previously generated in the 129SV inbred background (Milner et al., 1996). Transgenic mice overexpressing αΒ-crystallin were generated in the C57BL/6J background, using the αMHC promoter (Genbank ID U71441) (Subramaniam et al., 1991), driving expression of full-length murine αΒ-crystallin cDNA, followed by the bovine growth hormone poly-A sequence and backcrossed onto a pure 129SV mouse strain for several generations. The animals were used independent of sex. The approved procedures for animal care and treatment were followed, according to institutional guidelines following those of the Association for the Assessment and Accreditation of Laboratory Animal Care (AAALAC) and the recommendations of Federation of European Laboratory Animal Science Association (FELASA).

For the identification of transgenes by southern blotting, a fragment of αMHC was isolated and used to generate ³²P-labeled probes by random priming.

For western blotting analysis, proteins were transferred to polyvinylidene fluoride (PVDF) membrane (Whatman) and probed with the following antibodies: anti-αB-Crystallin (Stressgen, SPA223, 1:2000; Santa Cruz, sc-22744, 1:1000), anti-desmin (Abcam, ab8592, 1:1000), anti-GAPDH (Ambion, AM4300, 1:4000), anti-ANT (Santa Cruz, sc-9299, 1:200), anti-ATP5B (Mitosciences, ab5432, 1:500), anti-mitofilin (Novus Biologicals, NB100-1919, 1:1000; Abcam, ab110329, 1:1000), anti-NDUFS2 (Thermo Scientific, PA5-19342, 1:500), anti-SDHA (ab14715, 1:1000), anti-COXIV (ab33985, 1:1000), anti-VDAC1 (ab14734, 1:1000; ab15895, 1:1000; both Abcam), anti-mitofusin-2 (Sigma, M6444, 1:1000) and anti-MnSOD (BD Biosciences, M9920, 1:500) antibodies. Secondary horseradish peroxidase (HRP)-conjugated antibodies were purchased from Sigma-Aldrich. Proteins were visualized using chemiluminescence (ECL kit, Amersham). Films were scanned using a GS800 densitometer (Bio-Rad) and quantified using Quantity one software (Bio-Rad).

For the evaluation of fibrosis, Masson's-trichrome-stained paraffin sections were photographed and subjected to ImageJ software analysis. The results were expressed as the percentage of fibrosis in a given tissue section. For the evaluation of inflammation, hearts were observed under a Stemi 2000-C stereomicroscope (Zeiss) and were graded 0–4, according to the presence of inflammatory infiltrate, as previously described (Psarras et al., 2012) – 0=no inflammatory infiltrate; 1=one inflammatory infiltrate; 2=more than one inflammatory infiltrates appearing in limited areas of the myocardial wall; 3=multiple inflammatory infiltrates extended in multiple areas; 4=multiple extended inflammatory infiltrates eventually covering both the anterior and posterior wall.

The mice were injected intraperitoneally with EBD (250 μg/g of body weight in 1×PBS). After 24 h, the heart was embedded in optimum-cutting temperature (OCT) compound (VWR) and snap frozen in liquid nitrogen. EBD staining was evaluated by using the ImageJ software, and the results were expressed as the percentage of EBD-stained area in the image.

Mice were treated with heparin for 30 min and euthanized, and the heart was perfused with 2.5% glutaraldehyde in 0.1 M phosphate puffer, pH 7.4, fixed overnight and post-fixed with 1% osmium tetroxide for 1 h at 4°C, dehydrated, embedded in epon–araldite resin mixture and allowed to polymerize at 60°C for 24 h. Ultrathin sections (65–70 nm) (Leica EM-UC7 ultramicrotome, Leica) were examined with a Philips 201C transmission electron microscope and photographed on Agfa Copex HDP13 microfilm.

For immunogold labeling, mice were perfused with cold 4% paraformaldehyde and 1% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. Hearts were post fixed in the same fixative for 2 h and rinsed in 0.1 M phosphate buffer. The dehydration and embedding of tissue samples was performed using the progressive lowering of temperature (PLT) method using a Leica EM AFS (Leica), as described previously (Robertson et al., 1992).

Mounted ultrathin sections were subjected to the post-embedding immunogold procedure, as previously described (Havaki et al., 2003; Diokmetzidou et al., 2016). All incubations were performed using the automated immunogold labeling Leica EM IGL instrument. Primary antibodies (VDAC, Abcam, ab147234, 1:10; desmin, Abcam, ab8592, 1:20) were incubated overnight at 4°C, and samples were incubated for 1 h at room temperature with 10-nm-gold-conjugated secondary antibodies against polyclonal IgG particles (1:40) or 25-nm-gold-conjugated secondary antibodies against monoclonal IgG (1:40). Sections were examined with a Philips 420 transmission electron microscope at 60 kV and photographed with a Megaview G2 CCD camera (Olympus SIS).

Frozen mouse cardiac sections were used for immunolabeling, as previously described (Diokmetzidou et al., 2016). Primary antibodies against αB-crystallin (Stressgen, SPA223, 1:100; Santa Cruz, sc-22744, 1:50) and α-actinin (Sigma-Aldrich, Α7811, 1:1000) were incubated overnight at 4°C, and with secondary antibodies (conjugated to Alexa-Fluor-594 or Alexa-Fluor-488, Invitrogen) for 1 h at room temperature. The sections were examined using a Leica TCS SP5 confocal microscope equipped with a HC-PL-APO-CS 63× oil objective and LAS-AF acquisition software (Leica).

Mice were exercised twice a day using a swimming protocol, as previously described (Milner et al., 1999). The swimming program began at 10 min/day and was increased by 10 min daily to a maximum of 1 h. The twice-a-day 1 h exercise was continued for 9 days, and subsequently was increased to 90 min per session. The twice-daily 90 min exercise was continued for 9 days. In the case of αB-crystallin localization studies, a three-day protocol, during which the mice were exercised twice for 30 min the first day and twice daily for 1 h the next two days.

Mice were anesthetized through intraperitoneal injection of ketamine (100 mg/kg of bodyweight). Echocardiographic studies were performed using a Vivid 7 GE ultrasound system with a 13-MHz linear transducer. Two-dimensional targeted M-mode imaging was obtained from the short-axis view at the level of greatest left ventricular dimension. Images were analyzed using the Echopac PC SW 3.1.3 software (GE Healthcare). Data were expressed as mean±s.d. for continuous variables and analyzed by using Statview 5.0 (Abacus Concepts). Statistical comparisons were performed using ANOVA with Bonferroni or Dunn post-hoc test, or the unpaired Student's t-test where appropriate.

Isolation was performed using a modified protocol (Diokmetzidou et al., 2016) based on standard procedures (O'Connell et al., 2007).

Adult cardiomyocytes were incubated with 20 nM TMRM (Invitrogen) for 15 min at 37°C, washed and examined by using time-lapse confocal scanning for 20 min with a 1-min interval (TCS SP5, Leica) by excitation at 568 nm and emission at 580–640 nm. 5 μM cyclosporin A (Invitrogen) was added with TMRM to confirm mPTP opening. For the ROS measurement, the cardiomyocytes were incubated with 20 μΜ CM-H₂DCFDA (Invitrogen) for 30 min at 37°C, washed and examined using a 488-nm argon-ion laser excitation and emission at 503–555 nm. A HCX-PL-APO-CS-20 ×0.70 dry UV objective and LAS-AF acquisition software (Leica) were used in both cases. The acquired images were analyzed using the Volocity 5 software (Perkin Elmer).

Ca²⁺-induced swelling of mitochondria was performed as previously described (Weisleder et al., 2004). 5 μΜ CsA was added to mitochondria 5 min before CaCl₂.

Fractionation of mouse or rat hearts and the isolation of MAMs and pure mitochondria were performed by differential centrifugation at 4°C, as previously described (Wieckowski et al., 2009; Diokmetzidou et al., 2016; Frezza et al., 2007).

The GSH concentration was measured in hearts of 6-month-old mice, as previously described (Rahman et al., 2006). The hearts were weighed, homogenized in 5% sulfosalicilic acid and centrifuged at 10,000 g for 10 min at 4°C. The supernatant was removed and used for the GSH assay based on that of Griffith (1980).

ATP content was measured using the Aposensor ATP assay kit (Biovision) in a luminometer (Berthold) according to the manufacturer's instructions.

Mouse cardiac mitochondria were solubilized in native solubilization buffer pH 7 (20 mM Bis-Tris, 500 mM aminocaproic acid, 20 mM NaCl, 2 mM EDTA pH 8, 10% glycerol, 2 mM phenylmethylsulfonyl fluoride, protease inhibitor cocktail) with 4 g digitonin per g of protein for 30 min on ice and analyzed in a 3–12% native gel (Serva). Individual lanes were excised, incubated in equilibration buffer (125 mM Tris-HCl, pH 6.8, and 1% SDS) for 20 min and analyzed in the second dimension by 10% gel SDS-PAGE. Proteins were transferred to PVDF membrane (Whatman) and probed with the desired antibodies.

Two-dimensional gel electrophoresis was performed as previously described (Makridakis et al., 2010) with the following specifications. Mitochondrial fractions were solubilized in the isoelectric focusing (IEF) sample buffer [7 M urea, 2 M thiourea, 4% CHAPS, 1% dithioerythreitol, 2% ampholytes (pH range: 3–10)] and resolved (100 μg per sample) on 7-cm strips, pH range 3–10 non-linear (Bio-Rad), using the in gel rehydration method. IEF was conducted for about 12,500 volt-hours. Strips were then incubated in equilibration buffer (6 M urea, 50 mM Tris-HCl pH 8.8, 30% glycerol, 2% SDS) containing 0.5% dithioerythreitol for 45 min with gentle agitation, followed by a second incubation with equilibration buffer containing 4.32% iodoacetamide. Second dimensional analysis was performed on a 12% SDS-PAGE gel, and the gels were stained with Coomassie Colloidal Blue stain (Candiano et al., 2004).

Gels were scanned on a GS-800 imaging densitometer (Bio-Rad) in transmission mode, and image analysis was performed with the PD Quest 8 software package (Bio-Rad). Individual protein spot quantification was normalized based on the total intensity of the spots in the gel and expressed in ppm. Statistical analysis was conducted using the Mann–Whitney and Student's t-test.

Protein spots were excised manually or automatically with the use of the ProteineerSp Protein picker (BrukerDaltonics), destained (30% acetonitrile, 50 mM NH₄HCO₃), washed, dried and trypsinized overnight (trypsin, proteomics-grade, Roche). Peptides were extracted (50% acetonitrile, 0.1% trifluoroacetic acid) and applied onto the matrix-assisted laser desorption ionization (MALDI) target using the dry droplet method. Peptide masses were determined by using MALDI time-of-flight mass spectrometry (MALDI-TOF)–TOF-MS (Ultraflex TOF/TOF instrument, BrukerDaltonics). The peak list was created with Flexanalysis v3.3 software (Bruker); smoothing was applied with the Savitzky–Golay algorithm (width 0.2 m/z, cycle number 1), and a signal:noise threshold ratio of 2.5 was allowed. For peptide matching (Mascot Server 2; Matrix Science), the following settings were used: monoisotopic mass, one miscleavage site, carbamidomethylation of cysteine as fixed and oxidation of methionine as variable modifications. Stringent criteria were used for protein identification with a maximum allowed mass error of 25 ppm and a minimum of four matching peptides. False identity probability was usually lower than 10⁻⁵. Data analyzed with the same settings as described above, using a sequence scrambled version of Swiss-Prot data base generated by the decoy-generating script that is available at Matrix Science, provided no identifications.

Desmin mutants – head, amino acids 1–108 (1–324 bp); head-rod, amino acids 1–412 (1–1236 bp); rod, amino acids 109–412 (324–1236 bp); rod-tail, amino acids 109–470 (324–1410 bp); tail, amino acids 413–470 (1237–1410 bp). VDAC open reading frame and mutants – ΔNterm, amino acids 33–296 (99–888 bp) lacking the N-terminal α-helix; Δβ1–β4, lacks amino acids 34–81 (102–243 bp), thus lacking β-barrels 1–4. Sequences were isolated by using PCR with wild-type mouse heart cDNA as template and DNA Q5 high fidelity polymerase and sub-cloned into pGEX-4T-3. The recombinant polypeptides were expressed in BL21 bacteria, and the proteins were isolated as previously described (Kouloumenta et al., 2007; Diokmetzidou et al., 2016)

Mouse cardiac mitochondria (5 mg) were solubilized in native buffer pH 7 (20 mM Bis-Tris, 500 mM aminocaproic acid, 20 mM NaCl, 2 mM EDTA, pH 8, 10% glycerol, 2 mM PMSF, protease inhibitor cocktail), containing 4 g digitonin/g of protein, for 30 min on ice and were pre-cleared for 3 h at 4°C and used as previously described (Diokmetzidou et al., 2016). The bound proteins were eluted by heating at 97°C for 10 min in 2× electrophoresis loading buffer or, in the case of GST pull down, with 20 mM reduced GSH in 50 mM Tris-HCl, pH 8, for 30 min at room temperature. The eluates were analyzed by western blotting.

One hundred μg (1 mg/ml) of crude mouse cardiac mitochondria were suspended in MSE buffer (225 mM mannitol, 75 mM sucrose, 1 mm EGTA, 1 mM Tris-HCl pH 7.4) or osmotic buffer (20 mM KCl) with or without 0.5% Triton X-100 and incubated with 2 μg/ml proteinase K for 1 h on ice. The reaction was stopped with 4 mM (final) PMSF. Samples were loaded with 2× SDS-PAGE loading buffer and analyzed by western blotting.

All data are expressed as mean±s.e.m. unless otherwise stated. Statistical comparisons were performed using ANOVA with Bonferroni or Dunn post-hoc test, or unpaired Student's t-test where appropriate. A P-value<0.05 was considered significant. Graphics and statistical analysis were performed using Sigma Plot 10.0 (Systat Software).

We thank very much S. Pagkakis and E. Rigana for their help with imaging. We are grateful to E. Mavroidis and I. Kostavasili for constant assistance throughout this work. We also thank Prof. William James Craigen and N. Flytzanis for valuable comments on the manuscript.