Hyperactive Intracellular Calcium Signaling Associated with Localized Mitochondrial Defects in Skeletal Muscle of an Animal Model of Amyotrophic Lateral Sclerosis*

Amyotrophic lateral sclerosis (ALS) is a fatal neuromuscular disorder characterized by degeneration of motor neurons and atrophy of skeletal muscle. Mutations in the superoxide dismutase (SOD1) gene are linked to 20% cases of inherited ALS. Mitochondrial dysfunction has been implicated in the pathogenic process, but how it contributes to muscle degeneration of ALS is not known. Here we identify a specific deficit in the cellular physiology of skeletal muscle derived from an ALS mouse model (G93A) with transgenic overexpression of the human SOD1G93A mutant. The G93A skeletal muscle fibers display localized loss of mitochondrial inner membrane potential in fiber segments near the neuromuscular junction. These defects occur in young G93A mice prior to disease onset. Fiber segments with depolarized mitochondria show greater osmotic stress-induced Ca2+ release activity, which can include propagating Ca2+ waves. These Ca2+ waves are confined to regions of depolarized mitochondria and stop propagating shortly upon entering the regions of normal, polarized mitochondria. Uncoupling of mitochondrial membrane potential with FCCP or inhibition of mitochondrial Ca2+ uptake by Ru360 lead to cell-wide propagation of such Ca2+ release events. Our data reveal that mitochondria regulate Ca2+ signaling in skeletal muscle, and loss of this capacity may contribute to the progression of muscle atrophy in ALS.

ALS 2 is a neuromuscular disease characterized by degeneration of motor neurons and muscle atrophy. Mutations in the superoxide dismutase (SOD1) gene are associated with ϳ20% of inherited ALS cases and represent the most prevalent cause for familial ALS (1). Transgenic mice with overexpression of ALS-associated human SOD1 mutations develop syndromes similar to those of human ALS patients (2). Numerous studies using ALS transgenic models show that accumulations of mutant SOD1 inside mitochondria are likely the cause of the functional impairments in motor neurons (reviewed in Ref. 1). Morphological and biochemical analyses reveal defective mitochondria in skeletal muscle of ALS patients (reviewed in Ref. 3), which are also observed in the ALS transgenic models (4 -7). Although many studies are focused on the neurodegenerative aspect of ALS, the function of defective mitochondria in muscle degeneration during ALS progression has not been extensively examined.
Muscles use Ca 2ϩ as a messenger to control events ranging from activation of contraction to cell death. Defective intracellular Ca 2ϩ signaling and homeostasis have been linked to skeletal muscle dysfunction during aging (8,9) and in muscular dystrophy (mdx) (10 -14). In skeletal muscle, Ca 2ϩ release and uptake are mainly controlled by the sarcoplasmic reticulum (SR), which forms a network that is intimately associated with the mitochondria organelle. This close spatial proximity between SR and mitochondria, together with the ability of mitochondria to take up Ca 2ϩ from the cytosol, suggests that mitochondria could play an important role for modulating intracellular Ca 2ϩ signaling in muscle cells (reviewed in Ref. 15). Whether mitochondrial Ca 2ϩ uptake modulates physiological Ca 2ϩ transients in skeletal muscle and whether alteration of the mitochondrial Ca 2ϩ -buffering capacity contributes to muscle dysfunction in pathophysiological conditions are unresolved questions.
Here we examine mitochondrial function and Ca 2ϩ signaling in skeletal muscle derived from SOD1 G93A transgenic mice (G93A) (2). In the G93A muscle, we identify localized defects in mitochondrial structure and function that are associated with hyperactive Ca 2ϩ release from the SR at the lesion site. The muscle fiber segments with defective mitochondria appear at the neuromuscular junction (NMJ) region and are found in young mice prior to the onset of overt disease. Our data suggest that myogenic defects in mitochondria-mediated control of SR Ca 2ϩ release can potentially contribute to the progression of muscle wasting in ALS. (presymptomatic stage) and 3-4 months (stage of disease onset with apparent ALS symptoms) (2) were used in this study. Individual muscle fibers were isolated from these mice following the protocol of Wang et al. (12) ] i (fluo-4 AM) and ⌬⌿ (TMRE) in Krebs solution were exposed to a 170 mosM hypotonic solution containing (in mM) 64 NaCl, 5 KCl, 10 Hepes, 10 glucose, 2.5 CaCl 2 , 2 MgCl 2 , pH 7.2, for 2 min and then returned to Krebs solution. Ca 2ϩ signals were monitored before, during, and after the osmotic shock following the protocol developed by Wang et al. (12).

Animals and Isolation of Single
Image Processing and Data Analysis-iTool (IDL, ITT Visual Information Solutions) was used to display the images and to measure the average fluorescence intensity in regions of interest within muscle fibers. Averages over different fibers (or fiber regions) are represented as mean Ϯ S.E. Statistical significance was determined using Student's t test.
Transmission Electron Microscopy-FDB muscles were dissected and pinned down to a Sylgard dish in Krebs solution. After removing the connective tissue, the muscle was fixed in 0.1 M sodium cacodylate, pH 7.4, plus 4% paraformaldehyde and 2.5% glutaraldehyde for 4 h at room temperature. Fixed samples were embedded using standard techniques. Ultrathin sections were cut by diamond knife and examined by electron microscopy (17).

ALS Muscle Fibers Show Depolarized Mitochondria near
NMJ-Individual FDB muscles isolated from G93A mice and age-matched wild-type (WT) controls were used for structural and functional studies. Using electron microscopy, we found that in WT muscle, most mitochondria are of uniform size and are located within the sarcomeric I band (Fig. 1A, panel 2), a pattern that is similar to previous studies (8,15). The G93A muscle displayed abnormal mitochondria that can appear within the sarcomeric A band (Fig. 1A, panel 1). Some mitochondria in G93A muscle display internal vacuoles (Fig. 1A, panels [3][4][5], as has been previously reported in the motor neurons of ALS transgenic mice (18,19). A, electron microscopy images of FDB muscles from 3-month-old G93A and WT mice. In the WT, mitochondria (arrows) aligned within the sarcomeric I band and had uniform sizes (panel 2). In contrast, G93A muscle (panel 1) had enlarged mitochondria with vacuoles (panels 3-5), invading the sarcomeric A band. B, live muscle fibers probed simultaneously with Mito-Tracker Deep Red (panels 1 and 3) and TMRE (panels 2 and 4). A segment with depolarized mitochondria is identified in the G93A fiber (panel 2, bracket). The corresponding segment in the MitoTracker Deep Red image shows fuzzy staining with loss of contrast, due to morphological alterations of mitochondria (panel 1). Both mitochondrial probes stain the WT fiber equally and homogeneously (panels 3 and 4). C, overlays of TMRE (green) and ␣-BTX images (red) of dually stained G93A FDB fibers. Areas of mitochondrial defects had variable size, but they always faced the NMJ.
For live-cell imaging of mitochondrial morphology and membrane potential, FDB fibers were simultaneously incubated with MitoTracker Deep Red and TMRE. MitoTracker Deep Red stains the mitochondrial network, and TMRE is a voltage-sensitive fluorescent indicator for mitochondrial transmembrane potential, ⌬⌿. The distribution patterns for MitoTracker Deep Red and TMRE are uniform in WT muscle (Fig. 1B, panels 3 and 4). In contrast, the G93A muscle fiber displays localized defects in mitochondria, as shown by the regional loss of TMRE staining (Fig. 1B, panel 2) and fuzzy labeling by MitoTracker Deep Red at the same fiber segment (Fig. 1B, panel 1). This alteration in TMRE staining suggests substantial loss of ⌬⌿ in the affected region. 10 -60% of the muscle fibers isolated from G93A mice at the age of disease onset (3-4 months) showed depolarized mitochondria in fiber segments of various lengths (n ϭ 20 mice). This phenotype was rarely observed in the age-matched WT mice. Such defective mitochondria were also observed in four young G93A mice investigated at the age of 37 days but not in WT mice at the same age (n ϭ 4). This finding constitutes the first evidence of specific defects in mitochondrial structure and function in live muscle fibers of transgenic ALS mice that can occur prior to onset of neurodegeneration and muscle atrophy.
To determine whether the observed mitochondrial defect in G93A fibers was associated with NMJ, we simultaneously imaged the mitochondrial lesion and the NMJ by staining live G93A muscle fibers with TMRE and ␣-BTX, a ligand of the nicotinic acetylcholine receptor in postsynaptic membranes. As shown in Fig. 1C, although the size of affected segments with depolarized mitochondria varied among fibers, in every case (n ϭ 50), the fiber segments with depolarized mitochondria always included the muscle side of the NMJ. Thus, the localized mitochondrial defect at the NMJ is a distinct phenotype of the ALS muscle. This defect may represent an early event in the pathogenesis of ALS.
Hyperactive Ca 2ϩ Release in Fiber Segments with Depolarized Mitochondria-The finding of localized mitochondrial defects in G93A fibers presents a unique opportunity to test whether changes in mitochondrial function can affect intracellular Ca 2ϩ signaling as we can compare Ca 2ϩ release activity in regions with or without depolarized mitochondria in the same muscle fiber. For this purpose, we used osmotic stress-induced Ca 2ϩ sparks as an index for the integrity of the intracellular Ca 2ϩ release machinery (8, 12, 20 -22). Intact FDB fibers isolated from G93A and WT mice were loaded with fluo-4 AM and TMRE for measurement of intracellular Ca 2ϩ release and simultaneous monitoring of mitochondrial ⌬⌿. In the resting condition, spontaneous Ca 2ϩ sparks were rarely observed in G93A or WT cells. Transient exposure of muscle fibers to a hypotonic solution caused cell swelling in both WT and G93A muscle fibers and frequent Ca 2ϩ release events upon return to isotonic solution (Fig. 2). In the WT muscle, the mitochondria were uniformly labeled by TMRE, indicating normal polarization of mitochondrial membranes ( Fig. 2A, panel 1). Consistent with previous observations (12,20), the osmotic stress-induced Ca 2ϩ sparks were confined to the peripheral region of the WT fiber. In contrast, G93A fibers displayed more frequent Ca 2ϩ release events that were not restricted to the periphery of the G93A fiber, specifically within areas of the muscle fiber that had defective TMRE labeling (Fig. 1B, panels 3 and 4). These hyperactive Ca 2ϩ release events are similar to those observed in mdx muscle (12), suggesting that ALS muscle may share features with dystrophic phenotypes.
For quantitative comparison, the fluorescence intensity of fluo-4 in G93A muscle fibers was averaged in regions with normal or depolarized mitochondria. The mean of these averages over six completed experiments is plotted in Fig. 2C. Clearly, the defective regions responded to osmotic shock with greater local increase in cytosolic Ca 2ϩ . Greater increase in cytosolic Ca 2ϩ , likely caused by greater Ca 2ϩ release in response to stimuli in regions with depolarized mitochondria, appears to be a characteristic feature of the affected muscle in the ALS mouse model. This Ca 2ϩ regulatory defect is likely to result from mitochondrial defects. Indeed, we examined the rest- Areas with Polarized Mitochondria Stop the Propagation of Ca 2ϩ Waves-Local defects in mitochondrial ⌬⌿ reduce the driving force for Ca 2ϩ uptake, and loss of the Ca 2ϩ -sequestering capacity of mitochondria may underlie the hyperactive Ca 2ϩ release observed within the defective region in the G93A fibers. A reduced mitochondrial Ca 2ϩ uptake in the presence of random activation of Ca 2ϩ release channels might allow individual channel openings to progress to a local release event or spark via recruitment of neighboring channels by Ca 2ϩ -induced Ca 2ϩ release (24). If this was the case, one would predict that normally polarized mitochondria play a basal containment role to prevent local activation and otherwise control Ca 2ϩ release responses. More support for these views was obtained from the following observations.
Although WT muscle does not show Ca 2ϩ waves at rest or even after osmotic shock (12,22), we found that a portion of the G93A muscle fibers showed Ca 2ϩ waves after osmotic stress (Fig. 3A). These Ca 2ϩ waves often originated in regions with depolarized mitochondria and never appeared elsewhere in the muscle fiber with normal mitochondrial potential. Fig. 3B shows longitudinal scan images of a G93A muscle fiber, which allows comparing the evolution of Ca 2ϩ waves in areas with different mitochondrial polarization. Clearly, Ca 2ϩ waves that originate within the segment with depolarized mitochondria can only penetrate into the nearby normal region for distances of less than 30 m (n ϭ 4). The limited propagation of Ca 2ϩ waves in areas with normal mitochondria of the G93A muscle suggests that mitochondria with polarized membrane may have the ability to take up Ca 2ϩ at a rate that is sufficient to stop the propagation of Ca 2ϩ waves.
Inhibition of Mitochondrial Ca 2ϩ Uptake Exacerbates the Hyperactive Ca 2ϩ Release-Pharmacological assays were conducted to test whether reduced mitochondrial Ca 2ϩ uptake underlies the hyperactive Ca 2ϩ release activity in G93A muscle. To reduce the driving force of mitochondrial Ca 2ϩ uptake, we used FCCP, a proton ionophore that collapses ⌬⌿. Oligomycin (200 nM) was added to block the reverse activity of mitochondrial ATP-synthase and thereby slowed down the depletion of ATP during FCCP application. Fig. 4A shows representative images of cytosolic Ca 2ϩ and TMRE staining in a time series. Following osmotic stress, the muscle fiber segment with defective mitochondrial potential (as revealed by the reduced TMRE labeling, Fig. 4A, panel 1) produced more Ca 2ϩ release events than the normal region where the staining by TMRE was strong (Fig. 4A, panels 2 and 3). After the application of FCCP (1 M), Ca 2ϩ release events with increased frequency and intensity were observed that expanded into fiber segments originally with normal mitochondria (Fig. 4A, panels 4 -7), leading to global elevation of cytosolic Ca 2ϩ (Fig. 4A, panel 8). The reduced TMRE staining after FCCP treatment (Fig. 4A, panel 9) confirms the reduction of ⌬⌿, concomitant with widespread increase in Ca 2ϩ release activity. After 5 min of FCCP exposure, the averaged fluorescence of TMRE in the normal region decreased to 0.61 Ϯ 0.16 of control, whereas the averaged fluorescence of fluo-4 increased 2.17 Ϯ 0.60-fold (n ϭ 8, p Ͻ 0.001).
We then used Ru360, a specific blocker of the mitochondrial Ca 2ϩ uniporter (25), to see whether direct inhibition of mitochondrial Ca 2ϩ uptake can affect the propagation of Ca 2ϩ release in G93A muscle fibers. As shown in Fig. 4B, active osmotic stress-induced Ca 2ϩ release events were restricted to the region with depolarized mitochondria (Fig. 4B, panels 2 and  3). The application of 20 M Ru360 caused an immediate increase in Ca 2ϩ release activity (Fig. 4B, panels 4 -6). Previous studies have shown that Ru360 can have a biphasic effect on mitochondrial Ca 2ϩ uptake in cardiac myocytes, with a slow onset that requires a prolonged incubation of 28 min (26). When recording was extended to longer times after Ru360 application (Fig. 4B, panels 7-9), we observed a further increase in Ca 2ϩ release, resulting in elevation of fluorescence to 1.25 of control level after 30 min (n ϭ 4, p ϭ 0.08). The mitochondrial potential was not changed even after 45 min in Ru360 (Fig. 4B, panel 10), confirming that the mildly increased Ca 2ϩ release activity was due to block of the Ca 2ϩ uptake pathway. In comparison with the effect of FCCP (Fig.  4A), global elevation of cytosolic Ca 2ϩ was never observed even with the prolonged application of Ru360. One possible explanation for the lesser effect of Ru360 is the presence of other pathways of Ca 2ϩ uptake, including a rapid mode of mitochondrial Ca 2ϩ uptake that is less sensitive than the uniporter to ruthenium red (27).  4 -7). At 280 s, Ca 2ϩ release engulfed the entire fiber (panel 8). TMRE staining in the normal areas also decreased (panel 9). B, increased Ca 2ϩ release activity in a G93A fiber after the application of Ru360 (panels 4 -9). The mitochondrial potential was not changed by Ru360 after 45 min (panel 10).

DISCUSSION
In this study, we show that muscle fibers derived from the ALS mouse model exhibit localized mitochondrial defects characterized by altered structure and loss of membrane potential, which is accompanied by an increased tendency to release stored Ca 2ϩ from the SR. The increased activity of Ca 2ϩ release is restricted to the area of compromised mitochondrial membrane potential, and pharmacological inhibition of mitochondrial Ca 2ϩ uptake exacerbates this abnormal activity, suggesting that a deficit of mitochondrial Ca 2ϩ uptake is likely a cause of the augmented Ca 2ϩ release activity in the G93A muscle. We also find that Ca 2ϩ waves that start from the defective region in G93A muscle could not propagate much inside the regions with normally polarized mitochondria, revealing the strong mitochondrial buffering capacity of Ca 2ϩ released from the SR. Loss of this mitochondrial control of intracellular Ca 2ϩ may constitute a significant step in the progression of muscle atrophy in ALS.
These results provide the first evidence that the mitochondrial lesion appears first at the NMJ region in the G93A muscle and is an early event of ALS progression. ALS has been described as a "distal axonopathy," which affects the axon and NMJ in G93A mice at the age of 47 days, prior to significant loss of neuronal bodies and the onset of muscle atrophy (28). One of the matters in dispute is the importance of denervation as a pathogenic event contributing to muscle atrophy. The initial defect on the muscle side of the NMJ may reflect the early consequences of denervation. However, this mitochondrial lesion appears in G93A muscle at the age of 37 days, before axonal withdrawal is expected to occur. Therefore, one aspect of the muscle pathology in ALS could result from intrinsic defects in the mitochondria of skeletal muscle.
The NMJ conducts retrograde signals from muscle to nerve, which play critical roles in axonal growth and maintenance of synaptic connections in adult motor neurons (29,30). It was previously found that expression of a muscle-specific insulin-like growth factor 1 (Igf-1) isoform in the G93A model enhanced motor neuronal survival and could delay the progression of the disease (31). A primary muscle effect is supported by a recent study of Dobrowolny et al. (32), who found that muscle-restricted expression of SOD1 G93A directly caused muscle atrophy and mitochondrial dysfunction. Their electron microscopy study revealed swelling mitochondria with vacuoles similar to those that we observed in the muscle of G93A mice, suggesting the possibility that SOD1 mutation-mediated defects in muscle mitochondria contribute to ALS. In this study, we further demonstrate that acutely expressed human SOD1 proteins reach inside muscle mitochondria, where the mutant SOD1 may exert a role in disrupting function (supplemental data 2). The toxicity of mutant SOD1 may be dose-dependent as partial reduction of the expression of mutant SOD1 in muscle did not affect the disease onset or survival in ALS transgenic mice (33).
It has been shown that elevated [Ca 2ϩ ] i could promote the aggregation of mutant SOD1 inside mitochondria and exacerbate the loss of ⌬⌿ in cultured motor neurons (34). Although we did not detect an elevated [Ca 2ϩ ] i at rest, osmotic stress caused the area with defective mitochondria to display increased Ca 2ϩ release activity, which could lead to a sustained elevation of [Ca 2ϩ ] i near the NMJ and accelerate mitochondrial defects near the defective region. Furthermore, nicotinic acetylcholine receptors are abundant in the postsynaptic membrane of the NMJ. There is evidence of an elevated Ca 2ϩ permeability of the nicotinic acetylcholine receptors in adult mammalian muscle (35), and it has been found that nicotinic acetylcholine receptors are more numerous in G93A muscle (31). Thus, it can be expected that following repetitive stimulations, mitochondria near the NMJ will face elevated local [Ca 2ϩ ] and thereby become more susceptible to the deleterious effects of the mutant SOD1. A similar Ca 2ϩ -mediated mechanism may also apply to the neuronal side of the NMJ. Earlier studies by Siklós et al. (36) showed that motor nerve terminals from ALS specimens contained significantly increased levels of Ca 2ϩ . It has also been demonstrated that repetitive action potentials increase cytosolic Ca 2ϩ level rapidly and heavily load local mitochondria in motor terminals of G93A mice (37,38). This may accelerate the mitochondrial damage induced by mutant SOD1 and axonal degeneration, which in turn would foster denervation and its consequences. However, this is only an inference by analogy, not based on specific studies on nerve cells.
Based on past and present results, one potential model of the pathogenic sequence in ALS skeletal muscle is depicted in Fig.  5. In early presymptomatic stages (Fig. 5A), the situation is sim- FIGURE 5. Proposed pathogenic sequence in muscle of the ALS mouse model. A, in the early presymptomatic stages, the situation is similar to the WT. mito, mitochondria. B, accumulation of mutant SOD1 inside mitochondria results in functional deficits, including a reduced ⌬⌿ and Ca 2ϩ removal by mitochondria. The alterations lead to progressive withdrawal of the nerve terminal. These changes can be self-reinforcing consequences, including loss of mutual trophic signals between muscle and nerve cells, uncontrolled release of Ca 2ϩ from the SR, local cytosolic and mitochondrial increase in [Ca 2ϩ ] (in both cells), and more Ca 2ϩ stress on the normal mitochondria. The process may be characterized as a dual interplay between abnormal mitochondria and uncontrolled SR and between muscle fibers and altered nerve terminals, which eventually leads to destruction of both cells. nAChR, nicotinic acetylcholine receptor.
ilar to the WT. Probably the abnormality at this stage is the accumulation of mutant SOD1 inside mitochondria, an event that may occur simultaneously in muscle and nerve terminals. Accumulation of these defects would result in compromised mitochondrial potential near the NMJ, which would reduce mitochondrial capacity for Ca 2ϩ uptake and lead to abnormal Ca 2ϩ release in this region (Fig. 5B). This abnormal release stresses neighboring normal mitochondria, causing a decrease in their ⌬⌿ and furthering the elevation of [Ca 2ϩ ] i after the stimulation. In this view, the process of muscle degeneration may involve a dual interplay between abnormal mitochondria and uncontrolled SR Ca 2ϩ release in both muscle fibers and nerve terminals, which eventually leads to destruction of both cells.
Mitochondrial Ca 2ϩ uptake is believed to help regulate mitochondrial metabolism and synthesis of ATP so that the demands of muscle contraction are met. Whether mitochondrial Ca 2ϩ uptake modifies Ca 2ϩ signaling during excitationcontraction coupling remains an open question. A recent review summarized 70 publications supporting or arguing against this idea in cardiac muscle (39). In skeletal muscle, the situation is equally unsettled (reviewed in Ref. 15). Although studies show that mitochondria in skeletal muscle can take up Ca 2ϩ during contraction (40,41), it is not known whether altered mitochondrial Ca 2ϩ uptake can play a role in pathophysiological conditions. By demonstrating increased local Ca 2ϩ release in areas of mitochondrial deficiency, the stop of propagation of Ca 2ϩ waves in the neighboring normal areas, and the exacerbation of some of these effects by interfering with mitochondrial Ca 2ϩ uptake, our study constitutes a direct demonstration of the importance of mitochondria in shaping cytosolic Ca 2ϩ signaling in skeletal muscle. Malfunction of mitochondrial Ca 2ϩ uptake likely plays an important role in muscle degeneration of ALS. It must be noted that Ca 2ϩ release was induced by osmotic shock, a nonphysiological stimulus in our experimental condition. Future studies are required to quantify the amplitude and kinetics of mitochondrial Ca 2ϩ uptake during physiologic excitation-contraction coupling.
The present studies do not exclude other potential contributions to the abnormal Ca 2ϩ transient in G93A muscle. A distinct possibility is that the increased Ca 2ϩ level is a consequence of increased membrane permeability to Ca 2ϩ in the area of lesion. In a study documented with supplemental data 1, we found that eliminating external Ca 2ϩ did not stop the osmotic stress-induced hyperactive Ca 2ϩ release in the area of lesion, which suggests that Ca 2ϩ entry through the plasma membrane is not a major contributor to the elevated Ca 2ϩ in the area with depolarized mitochondria. An enhanced Ca 2ϩ release activity could also reflect functional changes in the SR or an altered production of reactive oxygen species. Future studies are required to elucidate the relative contribution of denervation versus a primary muscle effect due to SOD1 mutation in producing defective Ca 2ϩ signaling and triggering the initial mitochondrial defect in ALS muscle.
In summary, the new data demonstrate that a localized mitochondrial lesion in skeletal muscle is an early event during the progression of ALS. Understanding this and other cellular events underlying muscle degeneration will facilitate the development of rational therapeutic approaches to ALS.