Neurogenic atrophy induces muscle hydroperoxide production and atrophy
Aging in mice and humans is associated with an age-related loss of motor neurons and sarcopenia4,5. Our previous studies show that loss of muscle mass in response to denervation is accompanied by increased mitochondrial generation of hydroperoxides11,13. To determine the effect of hydroperoxide production on the loss of muscle mass during aging, we compared gastrocnemius muscle mass and basal hydroperoxide production in permeabilized gastrocnemius muscle fibers harvested from young, middle aged, and old C57BL/6 J mice. Figure 1a shows a loss of gastrocnemius muscle mass evident first at 26 months of age that continues into advanced age (32 months), while Fig. 1b shows a clear association between basal hydroperoxide production rates and loss of gastrocnemius mass (Fig. 1a,b). The increase in basal hydroperoxide production rate correlates with the amount of muscle atrophy in aging mice and in several other advanced atrophy models, including mice lacking the Nrf2 antioxidant response transcription factor (Nrf2−/−), mice with accelerated neurogenic sarcopenia (Sod1−/−), and mice with motor neuron disease (SOD1G93A) (Fig. 1c). These data support a direct relationship between basal hydroperoxide production rate and neurogenic muscle atrophy.
To more directly interrogate the relationship between loss of innervation, muscle atrophy and hydroperoxide generation, we measured changes in hydroperoxide production rates and muscle mass in a model of sciatic nerve transection. This model is relevant to aging because the proportion of denervated fibers increases with age6. Denervation by surgical transection of the sciatic nerve was performed in one hindlimb (denervation) and compared to the effect of sham surgery in the contralateral hindlimb of adult, male C57BL/6 J mice. The sciatic nerve transection model allows us to measure time-dependent changes after loss of innervation in samples composed entirely of denervated fibers. We measured the muscle response to nerve transection at 1, 2, 4, and 7 days following surgery. Muscle mass declines significantly between 4 and 7 days after loss of innervation, while hydroperoxide production is increased between 2 and 4 days (Fig. 1d–f). By 7 days post denervation, muscle mass is decreased by 20% and hydroperoxide production rates are increased compared to sham in both permeabilized muscle fibers (1,468%) and isolated mitochondria (987%). These results are supported by a previous finding that basal hydroperoxide production is increased in denervated tibialis anterior (TA) muscle prior to atrophy14.
We also compared sex-specific effects of denervation on gastrocnemius muscle from female and male mice seven days after surgery. Gastrocnemius muscle from both sexes shows a significant decline in mass at 7 days post-denervation (Supplemental Fig. 1a). Male mice have a higher initial gastrocnemius mass and a greater loss of total muscle mass; however, the relative loss of mass is equivalent in both sexes (Supplemental Fig. 1b-c). The greater loss in muscle mass in male mice corresponds to higher hydroperoxide production rates in isolated mitochondria and permeabilized fibers from denervated muscles in male mice compared to female mice (Supplemental Fig. 1d-e).
Loss of innervation induces transcriptional changes
To understand the molecular events underlying the effects of denervation, we measured denervation induced gene expression changes in muscle using RNAseq analysis of denervated muscle samples across a time course following sciatic nerve transection. We found that an acute gene expression response to denervation begins within 12–24 h followed by a chronic response beginning between 2 and 4 days, which coincides with the observed increase in LOOH production by the cPLA2 pathway (Fig. 2a). Further changes in gene expression increase up to 14 days (Fig. 2a,b). A similar number of genes are upregulated compared to the number that are downregulated at each time point during the chronic response from days 2 through 14 (Fig. 2a). Four genes are significantly upregulated throughout the entire time course (Arpp21, Gadd45a, Gdf5, and Myog), and 68 of the genes that are differentially regulated beginning at 2–4 days persist throughout the measured chronic phase (Fig. 2b). By day 14, 35 biological processes (including processes related to muscle cell function, metabolism, intracellular signaling, and ion transport) are significantly affected through overrepresented differentially expressed genes (Supplemental Files). 88 canonical pathways show significant change in activity including activation of Production of Nitric Oxide and Reactive Oxygen Species in Macrophages and ERK/MAPK Signaling and significantly different expression of Calcium, AMPK, NRF2 Antioxidant Transcription Factor, p53, and GADD45A signaling pathways (Supplemental Files). Signaling from ATF4 as an upstream regulator of gene expression changes was elevated within 2–4 days (Fig. 2c). ATF4 was previously associated with age-related muscle atrophy but our findings now suggest that loss of innervation to muscle fibers plays a key role in inducing this pathway15. Of the differentially regulated genes, we observed a significant enrichment of genes containing transcription factor binding sites in their promoter regions for MEF2 beginning at 2–4 days and SIX2 at 7 days (Fig. 2d). In addition to upregulation of LOOH production, loss of innervation is associated with sweeping chronic changes in gene expression.
Loss of innervation induces LOOH production through activation of the cytosolic phospholipase A2 (cPLA2) pathway
We previously reported that in addition to detecting hydrogen peroxide (H2O2), the Amplex Red probe interacts with lipid hydroperoxides (LOOHs)13. Invitrogen has since released an updated probe (Amplex UltraRed, Invitrogen A36006) with improved sensitivity and stability. We report that H2O2 and the LOOHs 13(S)-HpODE and 15(S)-HpETE, two metabolites enzymatically produced in vivo by 12/15-lipoxygenase, react with Amplex UltraRed to produce a concentration-dependent linear increase in fluorescent resorufin16. Notably, H2O2 produces a stronger increase in fluorescence at the same concentrations (Fig. 3a).
Our previous study using isolated mitochondria revealed that inhibition of cytosolic phospholipase A2 (cPLA2) with arachidonyl trifluoromethyl ketone (AACOCF3), an inhibitor that binds to the cPLA2 active site to block its activity, prevented the increase in Amplex UltraRed signal in response to denervation13,17. However, the study by Bhattacharya et al. did not define the total hydroperoxide production in muscle fibers, whether LOOHs were being produced upstream or downstream of cPLA2, or whether cPLA2 produces LOOHs in other neurogenic atrophy models including aging. cPLA2 is a membrane associated enzyme that cleaves AA from membrane phospholipids, thereby increasing AA levels that are in turn converted by enzymatic pathways to eicosanoids, lipid mediators of cell signaling18. LOOHs can be released from membranes by cytosolic phospholipase A2 (cPLA2) or formed downstream of cPLA2 by metabolism of AA into LOOH intermediates18,19,20. To define the molecular identity of hydroperoxides produced by denervation using permeabilized muscle fibers, we used Amplex UltraRed and ex vivo treatment with a combination of antioxidants (catalase, glutathione peroxidase, and superoxide dismutase) and the cPLA2 inhibitor AACOCF3. All conditions contained 2.5 U/ml of exogenous CuZn superoxide dismutase (Sod1) to convert any superoxide (O2−) produced by the mitochondria to H2O2. First, to specifically scavenge H2O2 and not LOOHs produced downstream of cPLA2, we treated denervated muscle fibers with supraphysiological concentrations of exogenous catalase. Treatment of denervated fibers with catalase decreased the hydroperoxide production rate but only up to ~ 30%, even at supraphysiological concentrations of catalase (Fig. 3b). We next treated denervated fibers with exogenous glutathione peroxidase 1 (GPX1), which scavenges both H2O2 and LOOHs21. Combined treatment of denervated fibers with GPX1 and catalase significantly decreased the hydroperoxide production rate by approximately 50% (Fig. 3c). In contrast, the CPLA2 inhibitor AACOCF3 decreased the hydroperoxide signal to a much greater extent than catalase or GPX (approximately 90%) (Fig. 3d) suggesting the greater proportion of the Amplex UltraRed signal is generated by LOOH formed downstream of cPLA2 as opposed to H2O2.
To further define the source of the Amplex UltraRed signal in permeabilized fibers, we measured the effect of a series of titrations of mitochondrial electron transport chain (ETC) substrates and inhibitors on hydroperoxide production described in the methods. Adding the mitochondrial ETC inhibitors rotenone and antimycin A greatly increases hydroperoxide production (presumably as a result of O2− and H2O2 from ETC complexes) in permeabilized muscle fibers22,23. We show here that both ETC inhibitors induced hydroperoxide production in both untreated sham and denervated fibers (Fig. 3e). Exogenous catalase and CuZnSOD completely scavenged the source of this signal while AACOCF3 had no effect (Fig. 3e). These results demonstrate that the CuZnSOD and catalase concentrations efficiently scavenge mitochondrial generation of O2− and H2O2, and that AACOCF3 does not inhibit mitochondrial generation of O2− and H2O2 or act as a direct antioxidant scavenger. Together, these data suggest that LOOHs are the primary component of the increased hydroperoxides observed after loss of innervation, their production or release requires cPLA2 activity, and they are not produced from electron transport chain reactive oxygen species (ROS) generation.
To determine the direct scavenging potential in vitro for a number of oxidant scavengers and AACOCF3, we generated a dose–response curve for H2O2 or the 12/15-lipoxygenase generated LOOH 15(S)-HpETE in the presence of several oxidative scavengers and AACOCF3 using the Amplex UltraRed signal as a readout16. Glutathione (GSH) is used to regenerate the antioxidant capacity of GPX1, although it can also serve as a direct antioxidant for free radicals, reactive oxygen species including H2O2, and LOOHs24. SS-31 is a mitochondrial targeted peptide that was originally designed as a mitochondrially-targeted antioxidant that directly scavenges via its tyrosine amino acid residue25. SS-31 was previously reported to completely inhibit H2O2 chemiluminescence at 100 µM and decrease lipid peroxidation by ~ 60% at 1 µM in vitro, although these experiments were carried out with isolated mitochondria or over an incubation period in cell-free conditions25,26. Liproxstatin-1 is a recently reported LOOH scavenger and functions as a positive control27. At the concentration we used in our permeabilized fiber assays, catalase and GPX1 with or without reduced glutathione (GSH) directly scavenge H2O2. In contrast, we observe no direct scavenging of LOOHs by AACOCF3 or SS-31 (Fig. 3f). We also report that Liproxstatin-1 functions as a scavenger of H2O2 in addition to LOOHs (Fig. 3f). For LOOHs, Liproxstatin-1, GPX1 with and without GSH, and GSH work as direct scavengers while SS-31 and AACOCF3 do not (Fig. 3g). AACOCF3 decreases LOOH production by inhibiting cPLA2 without direct scavenging of LOOHs.
The LOOHs produced in skeletal muscle during loss of innervation could be produced by two processes that are upstream or downstream of cPLA2: i.e., oxidative insult to membrane phospholipids creating LOOHs that are released by cPLA2, or metabolism of cleaved AA into eicosanoids that can produce LOOHs as a byproduct through the lipoxygenase and cyclooxygenase pathways19,20. We hypothesized that if denervated fibers exposed to cPLA2 inhibition by AACOCF3 treatment are able to produce hydroperoxides after addition of exogenous AA, this would be evidence that LOOH production is primarily occurring by metabolism of AA into eicosanoids. Exogenous addition of AA to sham and denervated muscle fibers treated with AACOCF3 increased hydroperoxide production rates to levels comparable to untreated denervation fibers (Fig. 3h). This suggests the cPLA2 pathway primarily produces LOOHs in denervated muscle downstream of cPLA2 release of AA and provides evidence that denervation-induced hydroperoxides are produced enzymatically during eicosanoid metabolism of AA and not by the canonical release of LOOHs from membranes formed by oxidative insult. The increase in LOOH production in both sham and denervated muscle treated with AACOCF3 after addition of AA suggests that AA release is the main regulatory step for LOOH production during eicosanoid metabolism.
Next, we used the permeabilized fiber protocol and treatment with either catalase or AACOCF3 to identify and compare the relative composition of hydroperoxide production rates in several models of neurogenic atrophy including an aging time-course, accelerated sarcopenia with neuromuscular degeneration (Sod1−/− mice), and end-stage ALS motor neuron disease (SOD1G93A mice). Catalase or AACOCF3 inhibited ~ 50–66% of hydroperoxides in sarcopenic aged mice (Fig. 3i). Age-related hydroperoxides are a mixture of LOOHs produced in the cPLA2 pathway and O2−/H2O2 likely produced by the ETC (Supplemental Fig. 2 Inset). In fibers from Sod1−/−, SOD1G93A, and denervated mice, the increased basal hydroperoxides are almost completely inhibited by the cPLA2 inhibitor AACOCF3 (Supplemental Fig. 2). AACOCF3 inhibits significantly more of the basal hydroperoxides in these models than catalase and in total inhibits ~ 80–90% of hydroperoxides (Fig. 3i). LOOHs produced by the cPLA2 pathway are induced in all tested models of neurogenic atrophy and are the primary component of basal hydroperoxides.
H2O2 scavengers do not protect against denervation-induced muscle hydroperoxide production or atrophy
To more specifically test the contribution of increased H2O2 to neurogenic atrophy pathogenesis, we used two transgenic mouse models (skmMCAT and skmPRDX3) that overexpress mitochondrial H2O2 scavengers in skeletal muscle. The skmMCAT model expresses mitochondrially-targeted human catalase (mCAT) in muscle28. Catalase is an H2O2 scavenger normally localized to the peroxisome. The skmMCAT model uses a flox-stopped version of the construct to express catalase specifically in muscle29. The skmPRDX3 model over expresses peroxiredoxin 3 (PRDX3) in muscle also using a flox-stopped transgene. Peroxiredoxin 3 is an endogenous mitochondrial H2O2 scavenger30. Analysis of skmMCAT expression using Western blot, targeted mass-spectrometry, and qRT-PCR revealed a similar level of mCAT gene expression compared to mouse catalase and approximately a three–ninefold increase in total catalase content in the skeletal muscle of skmMCAT mice compared to controls (Supplemental Fig. 3a-d). mCAT expression did not change protein content of other antioxidant or metabolic enzymes measured by mass spectrometry (Supplemental Table 1). PRDX3 was overexpressed approximately two orders of magnitude in gastrocnemius muscles of the skmPRDX3 model (Supplemental Fig. 3e-f).
To measure the impact of H2O2 generation on muscle atrophy induced by denervation, we performed sciatic nerve transection on control, skmMCAT, and skmPRDX3 mice and analyzed muscle mass and hydroperoxide production rates in isolated mitochondria and permeabilized fibers from gastrocnemius muscles after seven days (Fig. 4a). We find that sham muscle fibers and isolated mitochondria from skmMCAT but not skmPRDX3 mice have decreased hydroperoxide production rates after addition of ETC inhibitors compared to wildtype mice as expected for the increased H2O2 scavenging potential in muscle from these mice (Fig. 4c,e). However, neither skmMCAT or skmPRDX3 mice exhibit decreased peroxide production rates in denervated muscle fibers or isolated mitochondria compared to wildtype mice (Fig. 4b,d). Importantly, increased scavenging of H2O2 in muscle of skmMCAT and skmPRDX3 mice does not protect the mice from denervation induced loss of muscle mass (Fig. 4f). skmPRDX3 mice do tend to show a lower percent of atrophy in the denervated muscle compared to sham control, however this was not statistically significant. Interestingly, the skmPRDX3 mice have significantly smaller sham muscles, and a linear regression analysis shows that the size of the sham muscle partially explains variations in percent atrophy of the denervated gastrocnemius muscle compared to sham (Fig. 4g,h). Muscle fibers from skmMCAT mice and isolated muscle mitochondria efficiently scavenged H2O2 under control conditions, but increased H2O2 scavenging was not sufficient to protect against denervation-induced muscle hydroperoxide production or atrophy. Together our evidence suggests that increased muscle mitochondrial H2O2 production is not causal for muscle atrophy in response to denervation.
The tetrapeptide SS-31 was initially reported to have direct H2O2 antioxidant scavenging and to a lesser extent LOOH scavenging25. More recently, it has been suggested to exert its protective effects, including those in several age-related conditions, by virtue of its targeting to cardiolipin, with enhancement of ETC function, reduced ROS, and improved ATP production31. We treated wildtype mice for 10 days with daily injections of saline or SS-31 (3 mg/kg body weight). After 3 days of treatment, we performed sciatic nerve transection and analyzed muscle mass and hydroperoxide production 7 days later (Fig. 4i). SS-31 does not protect against muscle atrophy or increased basal hydroperoxide production in denervated muscle (Fig. 4j,k). In summary, the results of these three interventions suggest that H2O2 signaling is not required to induce muscle wasting, but instead denervation is primarily associated with increased LOOHs.
Denervation induces cPLA2 activity and downstream eicosanoids in skeletal muscle resulting in LOOH production
We previously reported an increase in cPLA2 protein content in denervated muscle compared to sham muscle13. In the current study, we measured cPLA2 activity in gastrocnemius muscles from sham and denervated mice seven days after sciatic nerve transection. cPLA2 activity is significantly elevated in denervated muscle fibers (Fig. 5a). We also measured cPLA2 activity in gastrocnemius muscles from young 6–9 month and old 28 month wild-type, onset Sod1−/−, and end-stage SOD1G93A mice. cPLA2 activity was significantly elevated in wasting muscle from SOD1G93A mice but not sarcopenic or Sod1−/− mice (Fig. 5b).
To better understand the nature of the LOOH response following loss of innervation, we used a targeted lipidomics approach to quantify the content of eicosanoids in gastrocnemius muscle from young control and seven-day post-denervation mice. We find that multiple eicosanoids downstream of cPLA2 are increased in muscle from 7-day denervated mice (Fig. 5c). In the lipoxygenase pathway, phospholipid hydroperoxide glutathione peroxidase (GPX4) converts LOOH intermediates to hydroxylipids18,20. Of the increased eicosanoids in denervation, 9 out of 16 are hydroxy lipids found downstream of LOOH intermediates in the lipoxygenase pathway.
We chose a targeted panel of eight eicosanoids based on our results in denervated muscle and measured their content in muscle from aged and SOD1G93A mice. Seven out of the eight measured eicosanoids were upregulated in aged and SOD1G93A gastrocnemius compared to control (Fig. 5d,e). We found significant upregulation of eicosanoids in all neurogenic atrophy models compared to control.
In vivo cPLA22 inhibition protects against denervation-induced muscle hydroperoxide production and atrophy
To determine in vivo if increased production of LOOHs generated by cPLA2 plays a causal role in denervation atrophy, we performed sciatic nerve transection surgeries and treated mice with a daily ip injection of 9.5 mg/kg body weight AACOCF3 or corn oil vehicle control for 7 days then measured the effect of cPLA2 inhibition on muscle hydroperoxide production, atrophy, and individual fiber size (Fig. 6a). AACOCF3 is a potent cell-permeable inhibitor of cPLA2 and previous publications demonstrated this dose inhibited cPLA2 in vivo in spinal cords32. Vehicle injected mice show a similar amount of atrophy compared to untreated controls (~ 20% atrophy in denervated muscle compared to sham at 7 days) (Fig. 6b). AACOCF3-injected mice undergo significantly less atrophy (~ 17%) at 7 days, resulting in a rescue of 16% of atrophy (Fig. 6b). We have repeated this experiment in several cohorts of mice, and each cohort saw similar protection from atrophy (data not shown).
We measured basal hydroperoxide production in permeabilized fibers from sham and denervated gastrocnemius muscle with and without AACOCF3 treatment. The standard protocol for fiber preparation found no difference in basal hydroperoxide production with AACOCF3 treatment. However, the protection in atrophy in the treatment group suggested some drug effect. Because AACOCF3 is a reversible inhibitor, we hypothesized it may be washed out during the dilution steps in the permeabilization protocol. We performed mechanical separation of fibers from vehicle and treated mice and immediately measured hydroperoxide production rates omitting the permeabilization and washing steps. No difference was found in fibers from vehicle-injected mice with or without permeabilization, therefore this protocol is suitable for measuring basal hydroperoxide production (Fig. 6c). Using this fiber preparation protocol, we observe a significant reduction in basal hydroperoxides (~ 34%) in denervated fibers from AACOCF3-injected mice compared to control (Fig. 6c).
Denervation results in a significant decrease in average muscle fiber cross-sectional area (CSA) in gastrocnemius muscles from both vehicle- (29%) and AACOCF3-treated (20%) mice (Fig. 6d–f). However, we observe a significant treatment effect in response to AACOCF3, which is associated with an increase in individual denervated fiber CSA (27%) compared to vehicle treated mice (Fig. 6d–f). Denervated muscle contains an increased proportion of small muscle fibers compared to sham, but we observe that AACOCF3 treatment increases the proportion of large muscle fibers in both sham and denervated muscles and increases average CSA of denervated fibers compared to vehicle control (Fig. 6e–f). Average CSA area is linearly correlated with gastrocnemius mass, which suggests that denervation primarily induces muscle atrophy through individual fiber atrophy that is partially rescued with AACOCF3 treatment (Fig. 6g). We previously reported that mice lacking 12/15-lipoxygenase (Alox15−/−) were protected from denervation atrophy33. Here we report that the basal hydroperoxide production rate in Alox15−/− denervated fibers is significantly decreased (~ 64%) (Fig. 6h). These experiments provide evidence that LOOHs produced by the cPLA2 pathway play a causal role in neurogenic muscle atrophy and are a viable target to protect against loss of muscle mass in neurogenic atrophy.
AACOCF3 treatment does not decrease cPLA2 pathway protein content
We next measured protein content of cPLA2, calcium-independent mitochondrial phospholipase A2 (iPLA2), and 12- and 12/15-lipoxygenase following denervation and cPLA2 inhibition in vivo. We measured protein content of key components of the cPLA2 pathway in both muscle whole homogenate and isolated mitochondria. Our analysis reveals that loss of innervation increases protein content of cPLA2, iPLA2, and 12- and 12/15-lipoxygenase in whole homogenate of the gastrocnemius muscle (Fig. 7a,b). In addition, AACOCF3 treatment significantly increases 12- and 12/15-lipoxygenase in denervated muscle compared to vehicle-treated denervated muscle (Fig. 7a,b). cPLA2 can be activated by increased intracellular calcium or phosphorylation by MAP kinase34. We observe no difference in phosphorylated cPLA2 content in whole homogenate. We also observe no difference in isolated mitochondrial content of cPLA2, phosphorylated cPLA2, iPLA2, or 12- and 12/15-lipoxygenase. Denervation does not increase activation of cPLA2 by phosphorylation (data not shown). Furthermore, AACOCF3 treatment inhibits cPLA2 activity but does not decrease expression or content of PLA2 pathway enzymes. These results suggest that denervation induces an increase in PLA2 pathway enzyme protein content, which results in the increased cPLA2 activity we observe in denervation.
cPLA2 inhibition mitigates oxidative stress in denervation
We measured several parameters to determine if LOOH production in the cPLA2 pathway contributes to atrophy primarily through oxidative damage or eicosanoid signaling. F2-isoprostanes are a biomarker of oxidative stress formed by non-enzymatic reaction of AA with a free radical35. F2-isoprostane content is increased in denervated muscle, but treatment with AACOCF3 mitigates this increase (Fig. 8a). The cPLA2 knockout mouse were previously identified to have cardiac fiber hypertrophy via modulating IGF-1 pathway signaling36. We measured the content and phosphorylation of several proteins in the IGF-1 signaling pathway. Loss of innervation increases AKT protein content and MAP kinase phosphorylation (Fig. 8b,c). However, treatment with AACOCF3 did not affect these changes after loss of innervation.