|Home | About | Journals | Submit | Contact Us | Français|
Author contributions: M.A.L., M.A.S., D.K., B.C.K., J.L.W., and R.G.K. designed research; M.A.L., M.A.S., Z.X., R.L.N., B.C.K., J.L.W., and R.G.K. performed research; R.L.N. contributed unpublished reagents/analytic tools; M.A.L., Z.X., D.K., and R.G.K. analyzed data; M.A.L., M.A.S., and R.G.K. wrote the paper.
A growing body of research indicates that amyotrophic lateral sclerosis (ALS) patients and mouse models of ALS exhibit metabolic dysfunction. A subpopulation of ALS patients possesses higher levels of resting energy expenditure and lower fat-free mass compared to healthy controls. Similarly, two mutant copper zinc superoxide dismutase 1 (mSOD1) mouse models of familial ALS possess a hypermetabolic phenotype. The pathophysiological relevance of the bioenergetic defects observed in ALS remains largely elusive. AMP-activated protein kinase (AMPK) is a key sensor of cellular energy status and thus might be activated in various models of ALS. Here, we report that AMPK activity is increased in spinal cord cultures expressing mSOD1, as well as in spinal cord lysates from mSOD1 mice. Reducing AMPK activity either pharmacologically or genetically prevents mSOD1-induced motor neuron death in vitro. To investigate the role of AMPK in vivo, we used Caenorhabditis elegans models of motor neuron disease. C. elegans engineered to express human mSOD1 (G85R) in neurons develops locomotor dysfunction and severe fecundity defects when compared to transgenic worms expressing human wild-type SOD1. Genetic reduction of aak-2, the ortholog of the AMPK α2 catalytic subunit in nematodes, improved locomotor behavior and fecundity in G85R animals. Similar observations were made with nematodes engineered to express mutant tat-activating regulatory (TAR) DNA-binding protein of 43 kDa molecular weight. Altogether, these data suggest that bioenergetic abnormalities are likely to be pathophysiologically relevant to motor neuron disease.
In amyotrophic lateral sclerosis (ALS), degeneration of both upper and lower motor neurons leads to progressive weakness and ultimately death. Contributing pathogenic processes include glutamate excitotoxicity, protein misfolding, inflammation, and oxidative stress (Reviewed in Rothstein, 2009). Intriguingly, there is also evidence for defective energy homeostasis in patients and mouse models of ALS. A subpopulation of sporadic (sALS) and familial ALS (fALS) patients possesses higher levels of resting energy expenditure (~16–18% increase) as measured by indirect calorimetry and lower fat-free mass compared to healthy controls (Kasarskis et al., 1996; Desport et al., 2005; Dupuis et al., 2008, 2011; Bouteloup et al., 2009; Funalot et al., 2009). Clinically, hypermetabolism is defined as the ratio of measured resting energy expenditure (mREE) and calculated resting energy expenditure (cREE), or mREE/cREE ≥1.1 (≥10% increase). Hypermetabolism in ALS patients cannot be ascribed to changes in forced vital capacity, muscle fasciculations, fever, hyperthyroidism, or infection (Desport et al., 2001). Similarly, transgenic mouse models overexpressing mutant copper zinc superoxide dismutase 1 (mSOD1) (G86R and G93A mice) show increased energy expenditure by indirect calorimetry, increased lipid and glucose clearance in skeletal muscle, and hypolipidemia (Dupuis et al., 2004; Fergani et al., 2007; Kim et al., 2011) (Lim and Kalb, unpublished observations). Metabolic defects have also been reported previously in another potential model of motor neuron disease. Mice with conditional postnatal ablation of tat-activating regulatory (TAR) DNA-binding protein of 43 kDa molecular weight (TDP-43) develop rapid weight loss and increased fatty acid oxidation, and experience early death (Chiang et al., 2010). At the cellular level, mitochondria from spinal cord and muscle of mSOD1 and TDP-43 mice possess both morphological and physiological abnormalities (Bendotti et al., 2001; Mattiazzi et al., 2002; Cassina et al., 2008; Pedrini et al., 2010; Shan et al., 2010; Xu et al., 2010; Zhou et al., 2010; Braun et al., 2011). Compromised mitochondrial function may contribute to systemic metabolic defects in the organismal level.
One-way cells sense a mismatch between energy supply and demand is through AMPK. A master sensor and regulator of energy balance, AMPK is a heterotrimeric protein, comprised of catalytic α subunits (α1, α2), and β and γ regulatory subunits (β1, β2, γ1, γ2, γ3) (Hardie, 2007). When ATP supply does not keep pace with demand, ATP/AMP ratios fall, leading to the activation of AMPK. It is activated by events that either compromise cellular ATP production (i.e., oxidative stress, hypoxia, nutrient deprivation) or increase ATP consumption (exercise) (Kahn et al., 2005; Long and Zierath, 2006; Hardie, 2007). AMPK activation results in increased fatty acid oxidation and glucose uptake, as well as decreased protein translation, glycogen, and cholesterol synthesis.
AMPK's strategic position at the crossroads of multiple metabolic pathways led us to wonder whether AMPK participated in the metabolic dysfunction associated with neurodegeneration. In this report, we confirm the existence of metabolic defects in mSOD1-induced motor neuron disease and provide evidence that these defects are associated with activation of AMPK. Finally, we present in vitro and in vivo evidence that activated AMPK adversely affects not only mSOD1-induced, but also mutant TDP-43-induced motor neuron disease.
Mixed spinal cord neuron cultures were prepared as described previously (Jeong et al., 2006; Mojsilovic-Petrovic et al., 2006). Briefly, an astrocyte feeder layer was prepared from the cortex of newborn Sprague Dawley rat pups [postnatal day 2 (P2)] and grown to ~80% confluency. Subsequently, dissociated embryonic day 15 (E15) spinal cord neurons were added. One to 2 d later, AraC (5 mm) (catalog #C6645; Sigma) was added for 24 h to arrest astrocyte proliferation. Cultures were maintained in glia-conditioned medium supplemented with the following trophic factors (1.0 ng/ml each): human neurotrophin-3, human neurotrophin-4, human brain-derived neurotrophic factor, human cardiotrophin-1, human glial-derived neurotrophic factor, and rat ciliary neurotrophic factor (Alomone Labs). Half of the culture medium was changed on a biweekly basis.
Wild-type (WT) and mutant SOD1, wild-type AMPK α2, and dominant-negative AMPK α2 (dnAMPK; K45R), as well as Lac-Z cDNAs, were cloned into the PrpUC amplicon plasmid to generate recombinant herpes simplex virus (HSV) as described previously (Neve et al., 1997). The titer of virus used in these experiments was ~3–5 × 107 plaque-forming units/ml.
Male heterozygote G93A mSOD1 mice on the congenic C57BL/6 background (Strain 004435; Jackson Laboratory) were backcrossed with C57BL/6 mice. Nontransgenic (non-Tg) littermates served as control mice. Both male and female G93A mSOD1 mice and control littermates were used for experimentation. Mice were fed a standard diet (Purina 5010) and were housed on a 12 h light/dark cycle. Genotypes were determined by PCR using tail snip DNA. Primers are listed in Table 1. All animals were treated in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Children's Hospital of Philadelphia Institutional Animal Care and Use Committee.
Mixed spinal cord cultures and mouse tissues were lysed or homogenized in Trizol (Invitrogen) to extract RNA. Synchronized young adult Caenorhabditis elegans were lysed in Trizol, and RNA was purified using the RNeasy Micro Kit (Qiagen). All samples were checked for RNA concentration and purity and reverse transcribed to cDNA using the Superscript II Reverse Transcriptase kit (Invitrogen). For peroxisome proliferator activated receptor γ coactivator-1 α (PGC-1α), glutathione peroxidase 1 (Gpx1), iron/manganese superoxide dismutase 2 (Sod2), and glyceraldehyde 3-phosphate dehydrogenase (GAPDH), quantitative real-time PCR (QPCR) was performed on an iCycler Real-Time PCR Detection System (Bio-Rad) using SYBR green PCR mix (Invitrogen). For other samples, QPCR was performed using a StepOnePlus Real-time PCR system (Applied Biosystems) using TaqMan Gene Expression Master Mix (Applied Biosystems) or Power SYBR Green PCR Master Mix (Invitrogen). Two to three independent experiments with three to five samples each were performed for mixed spinal cord culture and C. elegans samples. For QPCR analyses, all data were normalized to appropriate housekeeping genes (GAPDH, 18sRNA, or cdc-42) and the comparative Ct (delta delta Ct) method was used. TaqMan gene expression assays (Applied Biosystems) were used to probe for TBC1D1 (assay ID, Rn01413271_m1) and 18sRNA (assay ID, Hs99999901_s1). Other primers used are listed in Table 1.
Mixed spinal cord cultures were lysed and mouse tissues were homogenized on ice with lysis buffer (210 mm mannitol, 70 mm sucrose, 10 mm HEPES, pH 7.2, and 1 mm EGTA supplemented fresh with 0.5% w/v fatty-acid free bovine serum albumin (BSA) and protease and phosphatase inhibitors). Lysates were sonicated (particulate matter was removed by centrifugation), and boiled in 1× SDS protein loading buffer and 1% β-mercaptoetanol (BME). To prepare worm lysates, worms were washed off agar plates with worm lysis buffer (50 mm KCl, 10 mm Tris-Cl, pH 8.3, 2.5 mm MgCl2, 0.45% Tween-20, 0.45% NP-40, and 0.01% gelatin, supplemented fresh with protease and phosphatase inhibitors). To wash the bacteria off the worms, samples were centrifuged once and resuspended in lysis buffer. Samples were sonicated and centrifuged at 99,000 × g. The supernatant was boiled in 1× SDS protein loading buffer and 1% BME and saved for probing of soluble proteins. The pellet was resuspended in lysis buffer and centrifuged again at 99,000 × g. The remaining pellet was boiled in SDS and 1% BME and saved for probing of insoluble proteins.
All samples were subjected to SDS-PAGE, transferred to nitrocellulose membrane (Bio-Rad Laboratories), and blocked (5% w/v of nonfat milk or BSA in 1 mm Tris-HCl, pH 7.5, and 0.1% Tween). Membranes were incubated in primary antibody overnight, washed, and incubated with secondary fluorescent antibodies (infrared 800 or 680 anti-mouse IgG or anti-rabbit IgG; LI-COR Biosciences). Western blots were visualized and quantified using the LI-COR Odyssey Infrared Imaging System, and protein bands never saturated detectors. The following antibodies from Cell Signaling Technology were used: anti-AMPK (catalog #2532), anti-phospo-AMPK (pAMPK; catalog #2535), anti-acetyl-coA carboxylase (ACC; catalog #3676), anti-phospho-ACC (pACC; catalog #3661), anti-factor 4E binding protein 1 (4EBP1; catalog #9452), anti-phospho-4EBP1 (p4EBP1; catalog #9451), anti-p70 S6 Kinase (p70 S6 Kinase; catalog #9202), anti-phospho-p70 S6 Kinase (pp70 S6 Kinase; catalog # 9205), and anti-human SOD1 (catalog #2770). Anti-actin antibody (catalog #A206) was from Sigma. Aliquots of samples were saved for protein concentration assays.
Oxygen consumption was measured polarographically in intact cells using a magnetically stirred, thermostatically regulated (37°C) Mitocell Clark-type electrode system using a chamber volume of 150 μl (Model MT200; Strathkelvin Instruments). For these studies, mixed spinal cord cultures were resuspended by trypsinization, neutralized with culture media, washed, and then resuspended in fresh culture medium without serum. An aliquot of each sample was saved for protein quantification. A total of 150 μl of cell suspension was loaded into the chamber. Sequential measurements were made of the basal respiratory rate following addition of 0.5 μg/ml of the mitochondrial ATP synthase inhibitor oligomycin and maximal respiration stimulated by addition of the uncoupler carbonyl cyanide m-chlorophenylhydrazone (CCCP; 30 μm). ATP turnover was calculated by subtracting oligomycin-insensitive respiration from the basal respiration rate to yield oligomycin-sensitive oxygen consumption, and proton leak was determined from oligomycin-insensitive respiration. All values were normalized to protein content and are reported as nanoatoms Oxygen per minute per milligram protein.
We also measured mitochondrial respiration in mSOD1 mouse spinal cord homogenates. P40 and P90 G93A mSOD1 and non-Tg control mouse spinal cords were homogenized with 10 strokes of a dounce homogenizer using a tight pestle to prepare 10% w/v homogenates in ice-cold mitochondrial buffer (210 mm mannitol, 70 mm sucrose, 10 mm HEPES, pH 7.2, and 1 mm EGTA) freshly supplemented with 0.5% w/v of fatty-acid free BSA and protease and phosphatase inhibitors. A total of 150 μl of whole tissue homogenate was loaded into the electrode chamber. Complex I-dependent respiration was measured using 10 mm glutamate plus 2 mm malate followed by addition of ADP (0.2 mm) to stimulate State 3 respiration. The respiratory control ratio (RCR; State 3/State 2) was calculated as the ratio of the rate of oxygen consumption after ADP addition (State 3) to that obtained before addition of ADP (State 2). Sensitivity to oligomycin and CCCP were also tested. All reagents were from Sigma and were of the highest grade possible. The results of this method yielded similar respiration rates compared to isolated mitochondrial preparations.
After 14 d in vitro (DIV), mixed spinal cultures were infected with either HSV-WT SOD1 or HSV-mSOD1 (G37R). Viral stock (2 μl) was added to 1 ml of culture media 24 h or more before the next manipulation. When more than one recombinant HSV vector was used, 1.5 μl of each viral stock was added, yielding a total of 3 μl of viral stocks to 1 ml of media. Previous work has shown that both transgenes are expressed in >95% of neurons in a nontoxic manner under these circumstances (Mojsilovic-Petrovic et al., 2009). To quantify motor neurons, immunostained cells were counted in four to five randomly selected fields in each coverslip. In each experiment, three or more coverslips were used per condition, and the results presented are from at least three independent experiments. Experimenter was blind to all groups.
For the pharmacology experiment, cultures were treated with either Compound C (CC; 5 μm; catalog #171260; Calbiochem), aminoimidazole carboxamide ribonucleotide (AICAR) (10 mm; catalog #123040; Calbiochem), or vehicle (DMSO) every 2 d for 7 d. On the seventh day, cultures were fixed in 4% paraformaldehyde. Motor neurons were identified in mixed culture by immunostaining with anti-SMI-32 antibody (catalog #SMI-32R; Covance), which specifically stains for nonphosphorylated neurofilaments. Only labeled cells with cell body diameter 25 μm or greater were counted. This method has been validated previously as a means of recognizing motor neurons (Mojsilovic-Petrovic et al., 2006, their Fig. 1). All graphs are presented as the percentage of SMI-32-positive neurons in mSOD1-infected compared to WT SOD1-infected cultures treated with the same drugs or constructs.
The following C. elegans strains were obtained from the C. elegans Genetic Center: the N2 Bristol strain, aak-2 null (ok524; strain RB754), sid-1 (pk3321; strain NL3321), and tbc-11 null (ok2576; strain RB1959). The mutant lines Psnb-1::WT SOD1-YFP, Psnb-1::G85R SOD1-YFP, and Psnb-1::G85R SOD1 (line 18)-YFP were generous gifts from Arthur Horwich (Yale University, New Haven, CT) and Jiou Wang (Johns Hopkins University, Baltimore, MD), and were created as described previously (Wang et al., 2009). The mutant lines Psnb-1::TDP-43(WT) and Psnb-1::TDP-43(M337V) were generated as described previously (Liachko et al., 2010). Compound mutant strains, which were null for aak-2 and tbc-11, were confirmed by fluorescence and PCR using primers for aak-2 and tbc-11 (see Table 1). The following strains were used to generate G85R worms expressing sid-1 in the nervous system: sid-1(pk3321) and uIS69 [pCFJ90 (Pmyo-2mCherry); Punc-119sid-1];sid-1(pk3321). The latter was a generous gift from Martin Chalfie (Columbia University, New York, NY) and was generated as described previously (Calixto et al., 2010). Triple-mutant worms, Psnb-1::G85R-YFP;[pCFJ90 (Pmyo-2mCherry); Punc-119sid-1];sid-1(pk3321), were confirmed by yellow fluorescent protein (YFP) fluorescence and PCR primers for sid-1 (see Table 1). A G-to-A mutation in the pk3321 allele of sid-1 creates an ApoI restriction digest site, and this was used for genotyping. For the neurodegenerative assays, TDP-43(M337V) and TDP-43(M337V);aak-2(ok424) worms were crossed into the background strain CZ1200 (a generous gift from Yishi Jin, University of California at San Diego, La Jolla, CA) as described previously. This strain contains an integrated Punc-25::GFP transgene expressed in GABAergic neurons that clearly marks the cell bodies and axons of ventral cord ventral D (VD)- and dorsal D (DD)-type inhibitory motor neurons. All animals were grown on Nematode Growth Media (NGM) plates (~10 ml agar) seeded with Escherichia coli OP50. All animals were maintained at 20°C.
C. elegans locomotor behavior was tested in two different assays: (1) a crawling assay performed on agar plated with OP50 bacteria and (2) a swimming assay performed on a drop of M9 buffer on the surface of an agar plate. A total of 15 to 25 worms were allowed to lay eggs for 4 h. Once larvae reached the L4 stage, locomotor behavior was blindly recorded on a videocamera attached to a Zeiss Stemi SV11 dissecting scope and tracked using the Parallel Worm Tracker program (http://wormsense.stanford.edu/tracker; Miriam Goodman, Stanford University, Palo Alto, CA). Behavior was also recorded at an adult stage (24 h after L4). On average, five replicates containing 7–10 animals per group were tested, and at least two independent experiments were performed for each locomotor assay. To analyze crawling and swimming behavior, animal tracks and size were analyzed on MATLAB (MathWorks) to determine average speed (millimeters per second) and average speed normalized to size.
Each RNA interference (RNAi) colony was grown overnight in Luria broth containing ampicillin (50 μg/ml), and 200 μl was seeded on NGM plates containing isopropylthiogalactoside (1 mm) to induce dsRNA expression. RNAi clones for aak-2 and tbc-11 were generated as described previously (Fraser et al., 2000) and were obtained from Gary Ruvkun (Harvard University, Cambridge, MA) and Todd Lamitina (University of Pennsylvania, Philadelphia, PA), respectively. For locomotor experiments, L4 animals were placed on RNAi plates overnight. On the next day, young adult animals were transferred to a fresh RNAi plate and allowed to lay eggs for another 4–6 h. Progeny was subsequently tested for locomotor activity.
Assays were performed as described previously by Soukas et al. (2009). Briefly, L4 animals were singled and transferred daily to fresh NGM plates. The total number of hatched embryos was counted 3 d later, and total brood size was determined by summing viable offspring across all days. The reproductive span was recorded from the day of first egg lay until the last day when laid eggs hatched into live offspring. Sterile worms, embryos that failed to hatch, and dead or unfertilized embryos were also recorded.
Timed egg lays were arranged to produce synchronized populations at 48 h (L4) and 72 h (young adult). Live worms were placed on a 3% agarose pad containing 0.01% sodium azide to immobilize the worms. Worms were imaged under fluorescence microscopy and scored for number of GABAergic neurons, gaps in the ventral nerve cord, and gaps in the dorsal nerve cord.
Statistics were analyzed using Statview 5.0 software (SAS Institute). One-way ANOVAs were used in comparisons between two-group data and multivariate ANOVAs, including repeated-measures ANOVAs, were used in comparisons among three or more groups of data. Post hoc tests were performed to confirm significance. The threshold for significance was set at p ≤ 0.05.
We began by asking whether bioenergetic defects can be detected in a mutant SOD1 (mSOD1) in vitro model of motor neuron disease. In previous work, we showed that expressing mSOD1, but not the wild-type human protein (WT SOD1), in mixed spinal cord cultures leads to time-dependent death of motor neurons beginning after 48 h (Mojsilovic-Petrovic et al., 2006). We measured mitochondrial respiration polarographically with a Clark-type oxygen electrode in intact cells from mixed spinal cord cultures expressing either HSV-WT SOD1 (WT SOD1) or HSV-G37R mSOD1 (mSOD1) for 48 and 72 h. Since intact cells are not permeable to ADP and many mitochondrial substrates, analysis of their bioenergetic status cannot be performed in the same direct manner as isolated mitochondria or cell/tissue homogenates (Jekabsons and Nicholls, 2004). However, oligomycin, a mitochondrial ATP synthase inhibitor, is cell permeable and thus can be used to good advantage to examine the energetic status of intact cells.
The basal rate of oxygen consumption minus the rate following addition of oligomycin (ATP turnover, or “respiration supporting ATP synthesis”) provides an estimate of oligomcyin-sensitive ATP turnover in the cell coupled to oxidative phosphorylation and thus provides an indication of the functional capability of mitochondria. In addition, oligomycin-insensitive respiration provides an indication of H+ leak across the inner mitochondrial membrane, which decreases the efficiency of oxidative phosphorylation. CCCP-induced respiration provides a measure of maximum electron transport chain activity under uncoupled conditions.
Cultures infected with G37R mSOD1 virus showed a time-dependent decrease in mitochondrial respiration (F(1,11) = 9.05; p ≤ 0.01) (Fig. 1A,B). Furthermore, at 72 h after infection, mSOD1-infected cultures showed significantly decreased values for mitochondrial respiration compared to WT SOD1 counterparts (F(1,17) = 9.32; p < 0.01) (Fig. 1A,B). More specifically, ATP turnover in mSOD1-expressing neurons decreased by >85% compared to WT SOD1-expressing neurons. Although basal rates of respiration showed a decreasing trend in mSOD1-expressing neurons (9.10 ± 2.08 na O/min/mg in mSOD1 vs 12.44 ± 1.17 na O/min/mg protein in WT SOD1), this was not statistically significant. There was also an increasing trend toward oligomycin-insensitive respiration in mSOD1- versus WT SOD1-expressing neurons (8.10 ± 1.40 na O/min/mg in mSOD1 vs 5.24 ± 1.31 na O/min/mg protein in WT SOD1), which indicates increased H+ leakiness of the inner mitochondrial membrane. There was neither a change in basal level nor oligomycin-insensitive respiration in mSOD1-infected cultures at 48 h. In summary, data consistently reveal impaired mitochondrial respiration in mSOD1-expressing spinal cord neurons, possibly due to proton leak across the inner mitochondrial membrane.
We used a similar assay to measure respiration in a crude mitochondrial preparation from spinal cord of presymptomatic (P40) and symptomatic (P90) mSOD1 (G93A) mice. Mitochondrial respiration was measured in the presence of the complex I-linked substrate pair glutamate plus malate followed by addition of ADP, which stimulates oxygen consumption in well-coupled mitochondria. The rate of oxygen consumption upon addition of ADP and after most ADP has been phosphorylated into ATP (State 3/State 4) yields the RCR. Because it was not always possible to measure a true State 4, RCRs were calculated using State 2 (oxygen consumption upon addition of glutamate and malate and before addition of ADP) or oligomycin-insensitive respiration. These values yielded similar results. The RCR reflects the degree of coupling between substrate oxidation and ADP phosphorylation and thus reports the functional integrity of mitochondria.
Addition of glutamate plus malate induced mitochondrial respiration in P40 and P90 mSOD1 mice and non-Tg control littermates to a similar extent (data not shown). There was an age-dependent decrease in mitochondrial respiration between P40 and P90 mouse spinal cord mitochondria in both non-Tg and mSOD1 mice (F(1,19) = 116.30; p < 0.0001) (Fig. 1C,D). However, ADP elicited a significantly smaller stimulation of oxygen consumption in mSOD1 mitochondria compared to controls, yielding a decreased RCR at both P40 and P90 (Fig. 1C,D). More specifically, at P40, the average RCR value in mSOD1 mitochondria was reduced by ~15% compared to control levels (2.482 ± 0.09 compared to control littermates at 2.96 ± 0.04) (F(1,6) = 14.39; p < 0.01). At P90, the average RCR in mSOD1 mitochondria was reduced by ~40% compared to control levels (1.1 ± 0.04 compared to controls at 1.86 ± 0.15; F(1,13) = 25.93; p < 0.001). We also examined oligomycin sensitivity and obtained similar results when State 3/oligomycin-insensitive rates were used to calculate RCRs. No differences were observed in basal respiration rates before addition of glutamate plus malate (data not shown).
In addition to these mitochondrial abnormalities, P40 mSOD1 spinal cord mitochondria showed increased oligomycin-insensitivity at 5.19 ± 0.61 na O/min/mg protein compared to 1.37 ± 0.77 na O/min/mg protein in non-Tg mice (F(1,7) = 15.50; p < 0.006). There was a trend toward increased oligomycin insensitivity in P90 mSOD1 mice compared to non-Tg controls, but this was not significant. There was no difference in response to the uncoupler CCCP between mSOD1 mice and non-Tg counterparts at both P40 and P90 time points. In summary, these results demonstrate that bioenergetic abnormalities are present in both in vitro and in vivo mSOD1 models of motor neuron disease.
Considering the observed bioenergetic defects, we wondered if the energy sensor and regulator AMPK is activated in mSOD1 models. We began by investigating which AMPK subunits were expressed in our cultures and in rodent spinal cord because manipulation of specific subunits leads to distinct phenotypes (Viollet et al., 2003). Transcript expression of the AMPK α, β, and γ subunits, together with the housekeeping gene GAPDH, were therefore assayed by PCR analyses in mixed spinal cord cultures and mouse spinal cord cDNA (Table 2). Both α1 and α2 subunits, β1 and β2 subunits, as well as two of the three γ subunits were expressed. The γ3 subunit was present only in control cDNA from isolated mouse muscle (Table 2). These data are consistent with previous studies (Turnley et al., 1999).
Given that the AMPK catalytic subunits are present in our in vitro and in vivo mSOD1 models, we then asked whether the activity of AMPK and its downstream targets are altered as a result of observed defects in ATP production. We therefore assessed AMPK activity by immunoblotting for active AMPK (pAMPK at threonine 172), AMPK, and several downstream targets, such as ACC (pACC at serine 79 and ACC), factor 4E binding protein 1 (p4EBP1 at threonine 37/46 and 4EBP1), and p70 S6 kinase (pp70 S6 Kinase at threonine 389 and p70 S6 Kinase). ACC is a direct target of AMPK, and when phosphorylated, results in increased fatty acid oxidation. The proteins 4EBP1 and p70 S6 Kinase are downstream targets of the mammalian target of rapamycin (mTOR) pathway, which regulates protein translation. AMPK activation leads to decreased activity of the mTOR pathway, which leads to decreased phosphorylation of 4EBP1 and p70 S6 Kinase and, consequently, reduced protein translation.
We prepared lysates from mixed spinal cord cultures infected with either HSV-WT SOD1 or HSV-G37R mSOD1. In mSOD1-infected cultures, pAMPK expression was significantly increased compared to WT SOD1 counterparts, with no overall change in AMPK expression (F(1,7) = 23.52; p < 0.003; four independent experiments) (Fig. 2A,B). More specifically, the ratio of pAMPK/AMPK was increased by ~50% in mSOD1-infected cultures. We also observed an approximate twofold increase in the ratio of pACC/ACC, which is consistent with increased AMPK activation (F(1,14) = 7.14; p ≤ 0.02) (Fig. 2A,B). Furthermore, the downstream targets of the mTOR pathway, 4EBP1 and p70 S6 Kinase, showed an ~40% decrease in protein phosphorylation when normalized to overall protein levels, which is consistent with an activated AMPK state (Fig. 2A) (quantitative data not shown; two independent experiments). We next examined spinal cord lysates from G93A mSOD1 mice and control counterparts. At P40, there was no clear change in the expression patterns of pAMPK, AMPK, or its downstream target ACC (Fig. 2C,E). However, in P90 mSOD1 spinal cord, pAMPK expression increased, with no change in overall AMPK levels, resulting in a 90% increase in the pAMPK/AMPK ratio (F(1,14) = 148.80; p ≤ 0.001) (Fig. 2. D,E). AMPK activation was also associated with a 60% increase in the ratio of pACC/ACC compared with controls (F(1,7) = 68.69; p ≤ 0.0002) (Fig. 2D,E). An increased pAMPK/AMPK ratio was specific to the spinal cord, and observed in neither another nervous system tissue, such as the cerebellum, nor a peripheral tissue, liver.
AMPK regulates many transcriptional and translational targets that may play a role in metabolism, aging, and neurodegeneration. One such target of particular interest is PGC-1α, a transcriptional coactivator involved in mitochondrial biogenesis, antioxidant activity, and fiber-type switching (Puigserver and Spiegelman, 2003; Finck and Kelly, 2006). Furthermore, PGC-1α has been shown to be regulated by AMPK (Jager et al., 2007; Irrcher et al., 2008). Since PGC-1α dysregulation has been implicated in other neurodegenerative diseases, such as Huntington's disease and Parkinson's disease (Cui et al., 2006; Weydt et al., 2006, 2009; Chaturvedi et al., 2009; McConoughey et al., 2010; Hathorn et al., 2011; Shin et al., 2011), we wondered about its potential role in in vitro and in vivo mSOD1 models. We assessed the expression of PGC-1α and its downstream targets, GPx1 and Sod2, enzymes involved in detoxifying free radicals. Compared to HSV-WT SOD1-infected cultures, mixed spinal cord cultures infected with HSV-G37R mSOD1 did not show any changes in the levels of PGC-1α or its transcriptional target GPx1 (Sod2 was not detectable in our mixed spinal cord cultures) (Fig. 3A). Furthermore, no differences were observed in PGC-1α, GPx1, and Sod2 in spinal cord and hindlimb muscle from P120 mice (Fig. 3B,C).
Is enhanced AMPK activity in mSOD1-expressing cultures beneficial or detrimental to neuronal health? We began by treating HSV-infected G37R mSOD1 mixed spinal cord cultures for 48 h with the AMPK antagonist CC (5 μm), a competitive inhibitor of AMP binding sites, or vehicle (DMSO). To test drug efficacy, we performed immunoblots on cell lysates infected with WT or mSOD1. In both conditions, CC administration resulted in decreased abundance of pAMPK (F(1,3) = 314.14; p ≤ 0.003) and pACC (F(1,3) = 9411.19; p ≤ 0.0001), with no change in the abundance of the nonphosphorylated species (Fig. 4A,B). The effect of Compound C was dose dependent (across 0, 5, 10 and 100 μm dosages) (data not shown). A dose of 5 μm CC was then selected since it was the lowest dose with which a significant decrease in both pAMPK/AMPK and pACC/ACC levels were observed.
We next conducted mSOD1-induced toxicity assays in mixed spinal cord cultures infected with either HSV-WT SOD1 or G37R mSOD1. Cultures were treated with either CC (5 μm) or DMSO every 2 d for 7 d (three doses total). By ANOVA, significant groups were found (F(3,88) = 7.33; p ≤ 0.0002). Mixed cultures infected with mSOD1 and treated with DMSO demonstrated ~50% survival in SMI-32(+) neurons compared to WT SOD1-infected cultures treated with DMS0 (p ≤ 0.002). This mSOD1-induced motor neuron death was completely blocked by CC administration (F(1,4) = 18.88; p ≤ 0.01; three independent experiments) (Fig. 4C). We also tested the effects of the AMPK activator AICAR (10 mm) administration and found that it had no effect on mSOD1-infected cultures. Furthermore, mSOD1 cultures treated with AICAR showed significantly decreased motor neuron survival compared to CC-treated cultures (F(1,46) = 85.77; p < 0.001). There were no adverse effects of drug administered to cultures infected with WT SOD1 and treated with CC (data not shown).
In our second in vitro experiment, we used the same mixed culture conditions and toxicity assay to genetically manipulate AMPK activity using a dominant negative construct of AMPK as described previously (Mu et al., 2001). HSV vectors were engineered to express WT AMPK or dnAMPK. To test efficacy of these constructs, we performed immunoblots on cell lysates infected with WT or mSOD1. In both conditions, coinfection with a dnAMPK construct resulted in decreased abundance of pAMPK (F(1,3) = 38.21; p < 0.03) and pACC (F(1,6) = 5.69; p ≤ 0.05 for pACC/ACC), with no change in the abundance of the nonphosphorylated species (Fig. 4D,E).
We then performed mSOD1 toxicity assays, wherein cultures were coinfected with either HSV-WT SOD1 or G37R mSOD1 plus either dnAMPK, WT AMPK, or Lac-Z. By ANOVA, significant groups were found (F(5,88) = 4.55; p ≤ 0.001). Mixed cultures expressing mSOD1 + Lac-Z showed ~60% survival in SMI-32(+) neurons (p < 0.002 compared to WT SOD1 plus Lac-Z). Remarkably, cultures coinfected with mSOD1 plus dnAMPK showed close to 100% survival (p < 0.04 compared to mSOD1 plus Lac-Z; three independent experiments) (Fig. 4F). Coinfection of mSOD1 plus WT AMPK had no effects on survival and also showed a trend toward decreased motor neuron survival (p = 0.09 compared to mSOD1 plus dnAMPK). Coinfection of WT SOD1 with Lac-Z, WT AMPK, or dnAMPK had no effects on survival.
In sum, we have shown that bioenergetic defects are found in in vitro and in vivo models of mSOD1, and that these defects are associated with increased AMPK activity. Inhibiting AMPK activity, pharmacologically using the antagonist Compound C or genetically using a dominant negative construct, is neuroprotective in mSOD1-infected mixed spinal cord cultures.
To investigate the role of AMPK in an in vivo model of mSOD1 toxicity, we used a nematode C. elegans engineered to express human WT SOD1 and mSOD1 exclusively in neurons (Psnb-1::WT SOD1-YFP and Psnb-1::G85R SOD1-YFP). We will hereafter refer to these animals as WT SOD1 and G85R worms, respectively. These animals show ALS-like pathology, such as locomotor defects and protein aggregation (Wang et al., 2009). However, these animals possess a normal lifespan, and do not show detectable neuron death (Wang et al., 2009). These animals therefore model the motor neuron dysfunction that precedes death in ALS (Gould et al., 2006; Ilieva et al., 2008). C. elegans expressing human WT SOD1 also possess mild locomotor defects compared to the N2 control strain; however, worms expressing the mutant G85R version possess far more dramatic locomotor defects (Wang et al., 2009).
Given our results that attenuating AMPK activity is beneficial in vitro, we wondered whether placing the G85R mSOD1 animal into the background of an AMPK α2 null animal would improve disease pathology. The gene aak-2 is the ortholog of the AMPK α2 subunit in nematodes, and there are several null strains available. We used the ok524 aak α2 deletion mutant (aak-2) and generated double mutants. We will hereafter refer to these animals as G85R;aak-2. We compared the locomotor behavior of G85R;aak-2 worms to G85R worms on two kinds of locomotor assays: a crawling assay on agar plates containing OP50 bacteria and a swimming assay. For each assay, at least two independent experiments were performed containing at least five replicates each. On the crawling assay, significant differences existed among groups during L4 (F(4,28) = 5.68; p < 0.002) and adult stages (F(4,22) = 7.44; p ≤ 0.0006) (Fig. 5A,B). We observed the modest locomotor defect in WT SOD1 worms compared to the N2 control strain at the L4 stage (p < 0.03), which persisted, but to a lesser degree, during the adult stage (p = 0.08). As anticipated, G85R worms also performed significantly worse compared to WT SOD1 counterparts during both time points. On average, G85R worms moved at only half the average speed of WT SOD1 worms during the L4 stage (0.021 ± 0.003 mm/s in G85R vs 0.045 ± 0.006 mm/s in WT SOD1), which persisted into adulthood (0.021 ± 0.004 in G85R vs 0.063 ± 0.009 mm/s in WT SOD1; p < 0.001 in L4 and adult worms). We also observed enhanced locomotor activity in adult aak-2 worms compared to the N2 strain; however, this effect was erased when average speed was normalized to animal size (data not shown). Double-mutant G85R;aak-2 worms demonstrated a significant improvement in average crawling speed compared to G85R worms despite still being significantly less compared to aak-2 controls (p < 0.01 in L4 animals and p < 0.003 in adults). Enhanced locomotor behavior of double mutants was seen in L4 animals (0.021 ± 0.005 mm/s in G85R vs 0.042 ± 0.003 mm/s in G85R;aak-2; p ≤ 0.01) and improved to a greater degree in adult worms (0.021 ± 0.004 mm/s in G85R vs 0.09 ± 0.007 mm/s in G85R;aak-2; p < 0.001). This beneficial effect also persisted until later stages of adulthood, specifically 168 h after synchronization, and was observed even when average speed was normalized to body size (data not shown).
We also performed a swimming assay, another test of locomotor ability, wherein we also observed group differences among L4 (F(5,52) = 13.38; p < 0.0001) and adult animals (F(5,42) = 11.10; p < 0.0001) (Fig. 5C,D). Similar to the crawling assay, WT SOD1 worms demonstrated a mild locomotor defect compared to the N2 control worms only at the L4 stage (p < 0.002). There was no significant difference in aak-2 locomotor behavior compared to N2 control worms at both stages of development. However, G85R worms performed significantly worse compared to WT SOD1 counterparts during both time points such that locomotion was reduced by >80% (0.042 ± 0.005 mm/s in G85R vs 0.271 ± 0.032 mm/s in WT SOD1 worms during the L4 stage; 0.047 ± 0.004 mm/s in G85R vs 0.259 ± 0.031 mm/s in WT SOD1 worms during adulthood; p < 0.0007 and p < 0.003, respectively). Although G85R;aak-2 worms still performed significantly worse compared to aak-2 controls (p < 0.0001 and p < 0.0001 in L4 and adult worms, respectively), double-mutant worms revealed a significant twofold improvement in locomotor behavior compared to G85R worms. This pattern was observed at L4 stage (0.042 ± 0.005 mm/s in G85R vs 0.120 ± 0.013 mm/s in G85R;aak-2; p < 0.001) and persisted until adulthood (0.047 ± 0.004 mm/s in G85R vs 0.130 ± 0.016 mm/s in G85R;aak-2; p < 0.002). Furthermore, this improvement was observed as late as 168 h after synchronization and even when data were normalized to body size (data not shown).
We also tested the effects of genetically decreasing aak-2 activity in WT SOD1 worms to test whether its attenuation has a universally beneficial effect on locomotor activity or if its effects are specific in the context of motor neuron disease. C. elegans expressing WT SOD1 in the aak-2(ok524) background, which we will hereafter refer to as WT SOD1;aak-2, do not show any improvement of locomotor activity on a swimming assay compared to WT SOD1 controls, and in fact even show a mild worsening effect (Fig. 5C,D). On the swimming assay, WT SOD1;aak-2 worms showed an ~35% decrease in locomotor activity compared to WT SOD1 control worms. This effect shows a trend at the L4 stage (p = 0.09) and reaches significance in adulthood (0.146 ± 0.032 mm/s in WT SOD1;aak-2 animals vs 0.259 ± 0.031 mm/s in WT SOD1 animals; p < 0.04). Similar results were observed when data were normalized to body size (data not shown). Therefore, the beneficial effect of reducing aak-2 activity appears specific to an ALS-causing mutation. Elimination of aak-2 modestly worsens locomotor behavior in young adult animals expressing a wild-type version of SOD1. This improvement was observed when data were normalized to body size (data not shown). This is consistent with previous studies reporting that metformin, a drug that increases aak-2 activity, promotes health span and locomotor behavior in normal worms (Onken and Driscoll, 2010).
To consolidate the above observations, we studied a second independently generated Psnb1::G85R SOD1-YFP strain, termed line 18, which will hereafter be referred to as G85R(18). The G85R(18) worm expresses the mSOD1 transgene at higher copy number compared to the G85R line used in previous experiments. To test whether or not reducing aak-2 activity will rescue an even more severe disease phenotype, we created G85R(18);aak-2(ok524) double mutants. We then subjected aak-2(ok524), G85R(18), and the G85R(18);aak-2 worms to the swimming assay and observed group differences at both the L4 (F(2,17) = 3.07; p = 0.07) and adult stages (F(2,17) = 3.55; p = 0.05) (Fig. 5E). G85R(18);aak-2 worms demonstrated restoration of locomotor behavior as early as L4 (0.052 ± 0.003 mm/s in G85R(18) worms vs 0.322 ± 0.073 mm/s in G85R(18);aak-2 double mutants; p ≤ 0.03), and this beneficial effect persisted until adulthood (0.055 ± 0.01 mm/s in G85R(18) worms vs 0.218 ± 0.047 mm/s in G85R(18);aak-2 worms; p ≤ 0.04). The percentage of improvement observed upon decreasing aak-2 activity in the G85R(18) worms was even greater compared to that in the G85R worms used originally [for example, compare an approximate threefold improvement in L4 G85R;aak-2 worms with an approximate sixfold improvement in L4 G85R(18);aak-2 worms]. This improvement was observed when data were normalized to body size (data not shown).
In models of neurodegenerative disease, including the G85R mSOD1 worm, aggregation of proteins such as SOD1 have been observed (Gidalevitz et al., 2009; Wang et al., 2009). We therefore asked whether or not the observed protective effect of our genetic manipulation is associated with reduced SOD1 aggregation in the double mutants. There was no difference in the total SOD1 protein content or the ratio of soluble to insoluble SOD1 protein in G85R;aak-2 worm lysates compared to G85R worm lysates (Fig. 5F). In contrast, lysates prepared from WT SOD1 worms showed very little presence of insoluble human SOD1 compared to the soluble protein. This result suggests that the protective effects of reduced AMPK activity in the mSOD1 worm model are not associated with decreased SOD1 protein aggregation.
Another method to decrease gene expression in C. elegans is through RNAi. C. elegans are unique in their ability to exhibit systemic RNAi knockdown such that RNAi bacterial clones can be fed directly to worms to induce silencing of a targeted gene (Timmons and Fire, 1998; Calixto et al., 2010). To our surprise, G85R worms fed an RNAi clone to aak-2 did not show significant improvement in swimming behavior compared to empty vector (EV) RNAi (Fig. 5G). Neurons have been reported to be refractory to the effects of RNAi, and this has been attributed to the lack of neuronal expression of SID-1, a transmembrane protein involved in the passive uptake of dsRNA (Winston et al., 2002; Feinberg and Hunter, 2003; Shih and Hunter, 2011). When sid-1 is expressed in neurons, neurons then respond to RNAi knockdown (Calixto et al., 2010). Furthermore, when sid-1 is placed into the background of a sid-1 null (pk3321) mutant containing a missense allele, the effect of neuronal RNAi is enhanced even further (Calixto et al., 2010). We therefore suspected that RNAi knockdown in the G85R did not affect neuronal expression of aak-2. In support of this, we used an AAK-2::GFP translational reporter worm (Lee et al., 2008) and subjected these animals to RNAi knockdown of aak-2. Consistent with our suspicions, AAK-2 expression was reduced in other tissues, such as the intestine and vulva, but GFP expression remained strong in neuronal tissues, particularly in the nerve ring (data not shown).
We next generated triple-mutant animals: G85R animals expressing sid-1 in neurons in the background of a sid-1 null (pk3321), hereafter termed G85R;sid-1 animals. G85R;sid-1 animals exposed to aak-2 RNAi displayed improved swimming behavior compared to worms exposed to EV RNAi (0.124 ± 0.014 mm/s vs 0.075 ± 0.006 mm/s, respectively; F(1,4) = 9.58; p ≤ 0.04; three independent experiments) (Fig. 5G). This effect occurred only when G85R animals were in the sid-1 background, as shown by a significant interaction effect between genotype (G85R vs G85R;sid-1) and RNAi clone (EV vs aak-2; F(1,6) = 5.99; p ≤ 0.05). Similar results were seen when average speed was normalized to body size. In summary, these data suggest that reduced expression of aak-2 in neurons contributes to its beneficial effects in improving locomotor behavior.
Reproduction and organismal intermediary metabolism are intimately linked (Cardozo et al., 2011; Donato et al., 2011). In nematodes, hermaphrodite reproduction is a complex process that involves the mobilization of lipids to generate, fertilize, and maintain oocytes and developing embryos (Kubagawa et al., 2006; Han et al., 2010). Interestingly, female G93A mSOD1 mice exhibit both bioenergetic abnormalities (Dupuis et al., 2004; Fergani et al., 2007), as well as decreased fertility (Jackson Laboratory) (Lim and Kalb, unpublished observations). In light of this, we wondered whether mSOD1 worms displayed reproductive abnormalities and, if so, whether ablation of aak-2 modified this phenotype. Accordingly, we assessed brood size, reproductive span, and sterility in our mutant worms. A dramatic 50% of G85R worms were sterile, compared to 0% in WT SOD1 worms and in N2 controls (Fig. 6A) (data not shown for N2). From the subset of worms that were fertile, there were group differences that existed across several days of the reproductive span. There was a significant interaction effect between day of reproductive span and genotype (F(36,297) = 22.59; p < 0.0001) (Fig. 6B) and the total number of viable offspring summed across the entire reproductive span (F(4,38) = 23.70; p < 0.0001) (Fig. 6C). Self-fertilized WT SOD1 hermaphrodites showed a modest decrease in the total number of viable progeny compared to N2 control worms (p < 0.001). In contrast, G85R worms exhibited far more severe fecundity defects. G85R worms demonstrated a striking decrease in viable progeny, especially during Days 1 and 2 of adulthood, when normal hermaphrodites are most fertile (p < 0.002 and p ≤ 0.003 compared to WT SOD1 worms for Days 1 and 2, respectively) (Fig. 6B). Moreover, G85R worms possessed less than half the number of viable progeny compared to WT SOD1 controls across the entire reproductive span (42.22 ± 11.80 live offspring from G85R worms vs 133.90 ± 13.60 live offspring from WT SOD1 worms; p < 0.002) (Fig. 6C). Compared to N2 controls, there were no significant differences in the average reproductive span of WT SOD1 and G85R worms (4.2 ± 0.18 d in N2 worms vs 5.0 ± 0.53 d in WT SOD1 worms vs 4.44 ± 0.67 d in G85R worms).
Several of the worms laid eggs that did not mature into viable offspring. Of the eggs laid by G85R animals, 20–50% remained unhatched, in contrast to 0% in WT SOD1 controls during the first 5 d of active reproduction (Fig. 6C). Dead or unfertilized eggs occur at the end of the reproductive period due to the finite amount of sperm produced in C. elegans hermaphrodites. Unfertilized eggs were observed a day earlier in G85R worms, such that >90% of eggs laid by G85R worms on Day 6 were unfertilized, compared to a little over 40% of eggs laid on Day 7 by WT SOD1 animals (Fig. 6D).
Since WT SOD1 and G85R animals only express human SOD1 in neurons and not germ cells, we wondered whether feeding abnormalities, were affecting fecundity in the mSOD1 animals. We recorded pharyngeal pumping rates in adult N2, WT SOD1, and G85R mSOD1 worms. We found no difference in pumping rates between WT SOD1 and G85R worms (154.8 ± 6.8 pumps per minute in WT SOD1 worms vs 153.6 ± 8.3 pumps per minute in G85R worms), although there was a modest decrease in contraction rates in transgenic worms compared to N2 control animals, which pumps an average of 187.2 ± 8.2 pumps per minute (F(2,27) = 5.73; p < 0.009).
We next investigated the consequences of placing the G85R animals into the background of the aak-2(ok524) null animal. Compared to N2 controls, aak-2 null animals did not show significant differences in the number of viable offspring per day (Fig. 6B), the total number of viable offspring across the entire reproductive span (Fig. 6C), and the total reproductive span (4.2 ± 0.22 d in aak-2 animals vs 4.2 d ± 0.18 in N2 controls). Placing the G85R animals into the background of the aak-2(ok524) null worm rescued the fecundity phenotypes observed in the G85R worms. G85R;aak-2 double mutants showed deceased occurrences of sterility compared to G85R animals (9% of G85R;aak-2 animals vs 50% of G85R animals) (Fig. 6A). Of the worms that were fertile, G85R;aak-2 hermaphrodites produced more viable offspring compared to G85R worms throughout most of the reproductive span, especially during Days 1, 2, and 3 of adulthood (p ≤ 0.01 for Days 1, 2, and 3) (Fig. 6B). Furthermore, they produced a larger total number of viable offspring compared to G85R worms (198.20 ± 21.70 live offspring from G85R;aak-2 animals vs 47.22 ± 11.80 live offspring from G85R animals; p < 0.0001) (Fig. 6C). Interestingly, the reproductive span of the G85R;aak-2 animals was significantly different compared to the other groups (F(4,48) = 4.78; p < 0.003); G85R;aak-2 animals exhibited a longer reproductive span compared to G85R animals (6.4 ± 0.54 d vs 4.4 ± 0.67 d, respectively; p < 0.003).
Out of the total eggs laid by G85R;aak-2 worms, only ~10% of these remained unhatched across the first 5 d of active reproduction compared to 20–50% in G85R animals (Fig. 6D). Similar to G85R worms, G85R;aak-2 worms started laying unfertilized eggs at Day 6, but laid a smaller percentage (40% of total eggs laid by G85R;aak-2 were unfertilized vs close to 100% in G85R worms). Altogether, these results indicate that G85R animals possess severe reproductive defects that, like locomotion, are largely rescued by reducing aak-2 activity.
Are the beneficial effects of attenuating aak-2 activity specific to the mSOD1 model or is this effect applicable to other models of motor neuron disease? To answer this question, we used another C. elegans model of motor neuron disease, worms engineered to express human wild-type or mutant versions of TDP-43 exclusively in neurons [Psnb-1::TDP-43(WT) and Psnb-1::TDP-43(M337V)]. We will hereafter refer to these animals as WT TDP-43 and M337V worms, respectively. Similar to G85R mSOD1 worms, M337V mutant TDP-43 worms also display phenotypes that are reminiscent of motor neuron disease, including locomotor defects (Liachko et al., 2010). We placed the WT TDP-43 and M337V worms into the background of the aak-2(ok524) null allele to generate WT TDP-43;aak-2 and M337V;aak-2 double mutants, and then tested these worms on our swimming assay. Significant differences were observed among groups during the L4 (F(5,40) = 27.71; p ≤ 0.0001) and adult stages (F(5,43) = 33.66; p ≤ 0.0001) (Fig. 7A,B). We observed a modest locomotor defect in WT TDP-43 worms compared to N2 worms during the L4 stage (p ≤ 0.002), but not in adult animals. We also observed the even more severe locomotor phenotype of the M337V mutant TDP-43 worms compared to their WT TDP-43 counterparts during the L4 stage (0.071 ± 0.013 mm/s vs 0.217 ± 0.041 mm/s, respectively; p ≤ 0.002) and during young adulthood (0.051 ± 0.002 mm/s vs 0.227 ± 0.023 mm/s in M337V and WT TDP-43 animals, respectively; p ≤ 0.001). In contrast, young adult M337V;aak-2 worms showed a >50% improvement in locomotor behavior compared to M337V worms (0.082 ± 0.011 mm/s vs 0.051 ± 0.002 mm/s, respectively; p ≤ 0.02), which persisted into later stages of adulthood, as late as 144 h after synchronization (0.063 ± 0.008 mm/s in M337V;aak-2 vs 0.042 ± 0.005 mm/s in M337V worms; p ≤ 0.05). This improvement was observed when data were normalized to body size (data not shown). There was no difference in locomotor behavior between WT TDP-43 and WT TDP-43;aak-2 worms both at L4 and adult stages. In summary, similar to mSOD1 models of ALS, reducing aak-2 activity benefits another model of motor neuron disease: C. elegans overexpressing pan-neuronal human M337V mutant TDP-43. Furthermore, this beneficial effect appears specific to an ALS-causing mutation and not to transgenic worms engineered to express WT TDP-43.
Since loss of motor neurons is a key finding in postmortem analysis of the nervous systems of ALS patients, we next wondered whether decreasing aak-2 activity is neuroprotective in vivo. Although motor neuron loss does not occur in the G85R mSOD1 animals (Wang et al., 2009), neurodegeneration in the VD- and DD-type classes of GABAergic motor neurons and a subset of dopaminergic neurons has been recorded in M337V mutant TDP-43 animals (Liachko et al., 2010). To visualize neurons in living animals, we used the Punc-25::GFP reporter transgene driving expression of GFP in all 19 GABAergic motor neurons (Cinar et al., 2005). In a strain carrying the GFP reporter alone (CZ1200), both the dorsal and ventral nerve cords are continuous and contain the normal complement of 19 inhibitory motor neurons (13 VD and 6 DD GABAergic neurons). We generated Punc-25::GFP; TDP-43(M337V) and Punc-25::GFP; TDP-43(M337V);aak-2(ok524), hereafter referred to as M337V and M337V;aak-2 worms, respectively. M337V mutant TDP-43 worms exhibit profound degeneration of the GABAergic motor neuron network (Liachko et al., 2010). At L4 stage, both M337V and M337V;aak-2 animals displayed significant neuronal loss compared to CZ1200 control worms (F(2,57) = 28.92; p < 0.0001) (Fig. 7C). Furthermore, ablation of aak-2 did not significantly alter neuronal loss seen in M337V transgenic animals at this time point (2.1 ± 0.25 lost neurons in M337V vs 1.85 ± 0.25 lost neurons in M337V;aak-2; p < 0.40) (Fig. 7C). At the young adult stage, a time point wherein improved locomotor function was observed in M337V;aak-2 animals, significant neuronal loss was still observed in both worms expressing the M337V mutation compared to CZ1200 controls (F(2,48) = 25.29; p < 0.0001) (Fig. 7C). Moreover, no differences were observed in the number of labeled GABAergic neuronal cell bodies between M337V and M337V;aak-2 worms (3.75 ± 0.46 lost neurons vs 4.45 ± 0.36 lost neurons, respectively; p = 0.19) (Fig. 7C). There were no detectable differences in the number of dorsal and ventral nerve cord gaps in L4 as well as young adult animals (data not shown). These observations suggest that the loss of aak-2 ameliorates motor dysfunction in the M337V mutant TDP-43 worm, but does not prevent their ultimate demise.
Previously, Chiang et al. (2010) created conditional, postnatal murine knock-outs (KOs) of TDP-43. Upon ablation of TDP-43, the animals become hypermetabolic and die precipitously. Transcriptional profiling revealed that this phenotype is associated with decreased TBC1D1 expression in skeletal muscle. TBC1D1 is a Rab-GTPase activating protein that has been linked to obesity and its inactivation leads to a “lean” phenotype with increased glucose transport and fatty acid oxidation in tissues, especially skeletal muscle (Chadt et al., 2008; Frosig et al., 2010). TBC1D1 is also known to be phosphorylated at serine 237 by AMPK (Chavez et al., 2008), although the functional significance of AMPK-dependent phosphorylation on TBC1D1 activities, such as glucose transport, is unknown. Considering the metabolic similarities between mice with conditional postnatal ablation of TDP-43 and mSOD1 transgenic mice, we wondered whether TBC1D1 activity is decreased in our mSOD1 models and whether the beneficial effects of reduced AMPK activity are associated with alterations in TBC1D1 activity.
We began by monitoring the abundance of TBC1D1 mRNA in our mSOD1 models. Mixed spinal cord cultures infected with G37R mSOD1 for 48 h as well as spinal cord from P120 G93A mSOD1 mice did not reveal any changes in TBC1D1 mRNA levels compared to WT SOD1-infected cultures and non-Tg control mice, respectively (Fig. 8A). Since TBC1D1 is largely expressed in skeletal muscle (Chadt et al., 2008), we also monitored transcript levels in hindlimb muscle from P120 G93A mSOD1 and non-Tg controls. No changes in transcript levels were detected in this tissue at this specific time point (Fig. 8A).
We also investigated expression levels of the worm paralog of TBC1D1, tbc-11, in whole-worm RNA extract from our mutant worm strains. Significant differences were detected among groups (F(4,15) = 9.79; p = 0.004). Compared to N2 worms, WT SOD1 and aak-2 worms did not exhibit any significant changes in tbc-11 mRNA levels (Fig. 8B). In contrast, there was a 70% and 50% increase in tbc-11 expression in young adult G85R worms in comparison to N2 and WT SOD1 controls, respectively (p = 0.003 compared to N2; p < 0.02 compared to WT SOD1) (Fig. 8B). This increase was returned toward N2 control levels by placing the G85R animals into the aak-2 null background (p = 0.002 between G85R and G85R;aak-2 worms) (Fig. 8B). There were no differences in tbc-11 mRNA levels between G85R;aak-2 animals and WT SOD1 animals, as well as compared to aak-2 controls. Therefore, while TBC1D1 message is reduced in mice with conditional postnatal ablation of TDP-43, it is not reduced in the mSOD1 models we have studied.
If reduced TBC1D1 abundance contributes to the hypermetabolism seen in the conditional TDP-43 ablation mouse, one would predict that reducing tbc-11 in our C. elegans models would exacerbate disease. We pursued this issue by attenuating tbc-11 expression using RNAi knockdown. When G85R or G85R;sid-1 worms were fed tbc-11 RNAi, no significant interaction effect was detected between genotype (G85R vs G85R;sid-1) and RNAi clone (EV vs tbc-11) (F(1,10) = 0.45; p = 0.52). Neither G85R and G85R;sid-1 animals displayed a significant worsening or improvement in locomotor behavior when tbc-11 was knocked down. Interestingly, there was a trend toward modest improvement in the G85R;sid-1 animals fed tbc-11 RNAi compared to EV RNAi (0.122 ± 0.017 mm/s vs 0.086 ± 0.009 mm/s, respectively; F(1,8) = 3.41; p = 0.10; five independent experiments) (Fig. 8C).
Since RNAi may lead to insufficient knockdown of gene expression, we used the tbc-11 null strain (ok2576) and created G85R;tbc-11(ok2576) double-mutant worms, hereafter referred to as G85R;tbc-11 worms. The tbc-11 null strain (ok2576) appear phenotypically wild-type and did not demonstrate any locomotor defects compared to N2 worms (data not shown). In a swimming assay, G85R;tbc-11 worms did not demonstrate any significant worsening or improvement in locomotor behavior compared to G85R counterparts (F(1,5) = 2.02; p = 0.23; three independent experiments) (Fig. 8D). Similar results were observed at even later stages of development (data not shown). Furthermore, similar to RNAi knockdown, there was a trend toward modest improvement in swimming behavior (~30% improvement in G85R;tbc-11 worms compared to G85R controls).
We next examined whether tbc-11 is epistatic to aak-2 in our mSOD1 C. elegans by generating G85R;aak-2(ok524);tbc-11(ok2576) triple-mutant worms, hereafter referred to as G85R;aak-2;tbc-11 worms. A significant effect was observed among groups (F(2,12) = 8.55; p < 0.005) (Fig. 8E). Both G85R;aak-2 and G85R;aak-2;tbc-11 animals performed significantly better compared to G85R animals (p < 0.002 and p < 0.03, respectively; two independent experiments). Furthermore, the locomotor behavior of G85R;aak-2;tbc-11 worms was not significantly different from G85R;aak-2 double-mutant worms. Similar results were observed at even later stages of development (data not shown). Altogether, these results indicate that in a mSOD1 model of motor neuron disease, decreased tbc-11 activity neither exacerbates disease progression nor genetically interacts with aak-2 to alter disease progression.
We wondered whether the functional effects of altering TBC1D1/tbc-11 activity may be specific to the TDP-43 model of motor neuron disease. To this end, we examined tbc-11 mRNA levels in WT TDP-43 and M337V mutant TDP-43 C. elegans models of motor neuron disease. No significant differences were detected among groups (F(2,6) = 0.013; p = 0.99) (Fig. 8B). There was no significant change observed in tbc-11 mRNA levels between WT TDP-43 and N2 controls (p = 0.89). Furthermore, WT TDP-43 and M337V worms displayed similar levels of tbc-11 transcript levels, which was, on average, 97% of N2 tbc-11 transcript levels (p < 0.99). Similar to our results with mSOD1 animals, tbc-11 mRNA levels were not reduced in another model of motor neuron disease, mutations in TDP-43.
To explore the functional relevance of tbc-11, we placed the M337V mutant TDP-43-overexpressing worms in the background of the tbc-11(ok2576) null allele to generate double-mutant M337V;tbc-11(ok2576) worms, hereafter referred to as M337V;tbc-11 worms. No significant differences were detected between groups (Fig. 8D). More specifically, no significant improvement in swimming behavior was detected between M337V and M337V;tbc-11 worms (0.068 mm/s ± 0.005 in M337V vs 0.08 mm/s ± 0.003 in M337V;tbc-11 worms; three independent experiments; F(1,5) = 5.48; p < 0.08). This effect persisted until later stages of development (data not shown). Similar to our observations with G85R;tbc-11 worms, M337V;tbc-11 worms demonstrated modest improvements in locomotor behavior, compared to M337V controls, which trended toward significance (p < 0.08). Altogether, these results suggest that in both mSOD1- and mutant TDP-43-overexpression models of motor neuron disease, TBC1D1/tbc-11 does not seem to be pathophysiologically relevant.
Both sporadic and familial ALS patients display bioenergetic abnormalities. In this population of ALS patients, indirect calorimetry reveals increased resting energy expenditure, accompanied by abnormal lipid metabolism (Kasarskis et al., 1996; Desport et al., 2005; Dupuis et al., 2008, 2011; Bouteloup et al., 2009; Funalot et al., 2009). Both G86R and G93A mice, two commonly used mSOD1 mice, as well as conditional TDP-43 KO mice, also show a hypermetabolic phenotype and compromised mitochondrial function (Dupuis et al., 2004; Chiang et al., 2010). Mitochondrial defects have been linked to increased association of the mSOD1 protein with the mitochondrial membrane and interference with the efficiency of oxidative phosphorylation (Deng et al., 2006; Ferri et al., 2006; Israelson et al., 2010; Pedrini et al., 2010). In this report, we confirm that mitochondria from cells expressing mSOD1 have impaired ATP production, and this is associated with activation of the energy sensor AMPK. This observation is consistent with previous literature reporting dysregulation of various metabolic genes in mSOD1 models, including fatty acid synthase, fatty acid transporter (FAT/CD36), glucose transporter 4, p53, FOX03A, mTOR, and nitric oxide synthase, all of which are downstream targets of the AMPK pathway (Gonzalez de Aguilar et al., 2000; Martin, 2000; Lukas et al., 2006; Fergani et al., 2007; Lobsiger et al., 2007; Morimoto et al., 2007; Gonzalez de Aguilar et al., 2008; Martinez et al., 2008; Mojsilovic-Petrovic et al., 2009) (Lim and Kalb, unpublished observations). Furthermore, we show that AMPK activation has adverse effects on genetic models of motor neuron disease. Our findings provide a mechanistic link between bioenergetic abnormalities and motor neuron disease.
In this study, we demonstrate that downregulating AMPK activity is beneficial in in vitro and in vivo mSOD1 models, as well as in an in vivo mutant TDP-43 model of motor neuron disease. We first establish that defects in mitochondrial ATP synthesis are associated with increased AMPK activity. When AMPK activity is decreased pharmacologically or genetically, in vitro, we observe increased motor neuron protection. In vivo, when C. elegans that possess pan-neuronal expression of G85R mSOD1 or M337V mutant TDP-43 are placed on the background of an AMPK (aak-2) null strain, we observe improved locomotor behavior. Furthermore, ablating aak-2 in G85R mSOD1 worms rescues fecundity defects. This observation is noteworthy for several reasons. Fecundity defects have been reported in ALS patients (Johnson et al., 1995), as well as mouse models of the disease. Reproductive health is also tightly regulated by the metabolic state of organisms (Cardozo et al., 2011; Donato et al., 2011). For example, female mice that are null for the adipocyte hormone leptin (ob/ob mice) are not only obese, hypometabolic, and hyperphagic, but also sterile (Chehab et al., 1996). Furthermore, reproductive span has been linked to several pathways involved in aging and metabolism (such as insulin signaling), as well as mitochondrial function (Hsin and Kenyon, 1999; Branicky et al., 2000; Luo et al., 2010). Reduced fecundity in mSOD1 models may therefore be indicative of whole organism bioenergetic abnormalities. It will be interesting to determine whether the fecundity effect of ablating aak-2 in the mSOD1 worm is a non-cell-autonomous phenomenon.
Although decreasing aak-2 activity improved the locomotor defects of M337V mutant TDP-43 worms, it did not rescue neurodegeneration observed in this model. These results suggest that the beneficial effects of decreasing aak-2 activity may be through improving the functioning of neurons under conditions of stress and proteotoxicity. Several reports highlight the presence of neuronal dysfunction before neuronal death in several models of neurodegenerative disease, such as ALS (Gould et al., 2006; Ilieva et al., 2008), spinal muscular atrophy (Mentis et al., 2011), spinocerebellar ataxia (Barnes et al., 2011; Shakkottai et al., 2011), and Huntington's disease (Spampanato et al., 2008). Therefore, manipulations that may benefit neuronal function may at least be partially independent of processes that lead eventually to neuronal death in these disease models.
What downstream targets of AMPK involved in metabolism may be dysregulated in mSOD1 models? Here, we report that expression of PGC-1α message and its downstream targets are not altered in mSOD1-infected cultures, spinal cord, and hindlimb muscle from symptomatic mSOD1 mice. These results were unexpected since, typically, impaired ATP synthesis should trigger an increase in PGC-1α message, protein, and transcription activity. However, it is also plausible that changes in PGC-1α expression may differ between specific neuronal populations (i.e., sensory vs motor neurons) or cell-types (i.e., neurons vs astrocytes) (Cui et al., 2006) and may therefore be masked in mixed culture or whole tissue preparations. Our results do not rule out changes in PGC-1α protein abundance or alterations in posttranslational modifications. More experiments are needed to investigate the relationship between bioenergetic abnormalities and its effects on AMPK and PGC-1α.
Another gene of interest is TBC1D1. Chiang et al. (2010) generated mice with a conditional postnatal ablation of TDP-43 (TDP-43 KO) and found that these animals develop rapid weight loss and increased energy expenditure. These abnormalities bear striking resemblance to the hypermetabolic defects observed in mSOD1 mice, although the time frame of sign and symptom progression in the TDP-43 KO mice is greatly accelerated. There is a reduction in TBC1D1 expression in the muscle of the TDP- 43 KO mice, and this is of interest because TBC1D1 inactivation leads to increased glucose clearance and fatty acid oxidation in muscles (Chadt et al., 2008). If TBC1D1 expression was suppressed in our mSOD1 and mutant TDP-43 models, it might account for the hypermetabolic phenotype and be pathophysiologically relevant. In mSOD1-infected mixed spinal cord cultures, as well as tissues from symptomatic mSOD1 mice, we found that TBC1D1 message was similar in comparison to controls. In fact, instead of showing decreased expression, TBC1D1 mRNA levels were even slightly elevated. A similar pattern was observed for the expression of tbc-11, the worm paralog of TBC1D1, in whole-worm lysates, such that G85R worms showed significantly increased tbc-11 expression compared to WT SOD1 worms. No significant alteration in tbc-11 message levels was detected between WT TDP-43 and M337V worms. Finally, in several functional assays, we demonstrate that decreasing tbc-11 activity in G85R and M337V mutant worms does not exacerbate disease, and in some cases even leads to a modest improvement. In addition, epistatic experiments suggest that the beneficial effects of aak-2 are independent of tbc-11 activity. Altogether, these results indicate that TBC1D1/tbc-11 is unlikely to participate in the bioenergetic abnormalities observed in mSOD1 mice and does not contribute to motor neuron dysfunction/death in mSOD1 or mutant TDP-43 worm models.
Why might increased AMPK activity be detrimental in models of motor neuron disease? The metabolic phenotype observed in mSOD1 and TDP-43 models matches the outcomes one may anticipate when AMPK is activated, including increased fatty acid oxidation and glucose clearance. The physiological characterization of these models appears to be that of a “starvation” phenotype, wherein ATP levels are low and fat stores are rapidly depleted despite normal to increased food intake. In fact, caloric restriction, which activates AMPK (Baur et al., 2006; Curtis et al., 2006), further shortens lifespan and accelerates disease progression (Hamadeh et al., 2005; Mattson et al., 2007; Patel et al., 2010). On the contrary, feeding mSOD1 mice a high-energy diet remarkably increases lifespan, improves locomotor behavior, increases fat stores, and protects motor neurons (Dupuis et al., 2004; Browne et al., 2006; Zhao et al., 2006). AMPK activation may be initially beneficial to turn on catabolic pathways that can increase ATP levels and therefore restore energy homeostasis. However, chronic AMPK activation might become detrimental when energy stores are exhausted and important anabolic pathways remain shut down. These pathways may include glucose, cholesterol, and fat synthesis pathways, as well as pathways promoting protein translation, which are all crucial for survival. Chronic AMPK activation could be particularly detrimental to neurons with high metabolic demands, such as motor neurons with their relatively large size and long axonal projections. In response to AMPK activation, motor neurons may reallocate resources away from vital anabolic processes, such as axonal transport, as well as protein and lipid synthesis. Over the long run, this may lead to dysfunction and eventually neuronal death.
Previous reports support that excessive AMPK activity may have detrimental effects in stressed conditions. In mouse models of ischemia, AMPK activation has been observed (McCullough et al., 2005; Li et al., 2007; Mukherjee et al., 2008). More importantly, pharmacological and genetic reduction of AMPK has been shown to be neuroprotective in the ischemic mouse brain. In Alzheimer's disease, AMPK activation leads to increased production of Aβ peptides, which contribute to amyloid plaque formation (Chen et al., 2009). In response to glucose deprivation, AMPK activation may also induce apoptotic cell death through its interactions with tumor-suppressor p53 (Jones et al., 2005; Okoshi et al., 2008). It has also been shown previously that inhibition of SNF1, the AMPK homolog in yeast, promotes stress resistance and increases replicative lifespan, independent of caloric restriction pathways (Lu et al., 2011). Although AMPK activation may be beneficial in some settings, such as in the context of cardiovascular diseases, diabetes, and obesity, our data and the evidence cited above suggest that it is reasonable to think that increased and prolonged AMPK activity may paradoxically play a detrimental role in the setting of some neurodegenerative diseases.
In sum, we report the potential role that the energy sensor AMPK may play in models of motor neuron disease. AMPK may therefore provide a link connecting metabolic perturbations, motor neuron degeneration, and muscle pathology observed in ALS patients and animal models. Further experiments to support this link are underway. Decreasing AMPK activity and perhaps other pathways that lead to increased energy expenditure in genetic models of motor neuron disease may potentially be an intriguing therapeutic target to treat ALS patients.
This work was supported by National Institutes of Health Grants F31NS06816502 (M.A.L.), R21NS060754, and RO1NS052325 (R.G.K.), the Department of Veterans Affairs (B.C.K.), and the ALS Association. We thank the generosity of Dr. Morris Birnbaum (University of Pennsylvania, Philadelphia, PA) for the AMPK wild-type and dominant negative plasmids; Dr. David Borchelt (University of Florida, Gainesville, FL) for the wild-type and mutant SOD1 plasmids; Drs. Arthur Horwich (Yale University, New Haven, CT) and Jiou Wang (Johns Hopkins University, Baltimore, MD) for the WT SOD1, G85R mSOD1, and G85R(line 18) worm strains; Dr. Hyeon-Sook Koo (Yonsei University, Seoul, South Korea) for the AAK2::GFP translational reporter strain; Drs. Martin Chalfie, Irini Topalidou, and Xiaoyin Chen (Columbia University, New York, NY) for the sid-1 mutant strains; Dr. Yishi Jin (University of California at San Diego, La Jolla, CA) for the CZ1200 strain; Dr. Miriam Goodman (Stanford University, Palo Alto, CA) for sharing the Parallel Worm Tracker program; and Drs. Gary Ruvkun (Harvard University, Cambridge, MA) and Todd Lamitina (University of Pennsylvania, Philadelphia, PA) for sharing RNAi clones. We acknowledge the Gene Knockout Consortium at Oklahoma Medical Research Center, which provided several of the C. elegans knock-out strains used in our experiments. We also acknowledge Jelena Mojsilovic-Petrovic, Heather Pratt, Peter Krumbhaar, and Daniel Helbig for their excellent technical assistance. Finally, we thank the C. elegans laboratories at the University of Pennsylvania (Christopher Fang-Yen, Todd Lamitina, John Murray, David Raizen, and Meera Sundaram) for their advice and suggestions.
The authors declare no competing financial interests.