Correction: Mutant TDP-43 Deregulates AMPK Activation by PP2A in ALS Models

Bioenergetic abnormalities and metabolic dysfunction occur in amyotrophic lateral sclerosis (ALS) patients and genetic mouse models. However, whether metabolic dysfunction occurs early in ALS pathophysiology linked to different ALS genes remains unclear. Here, we investigated AMP-activated protein kinase (AMPK) activation, which is a key enzyme induced by energy depletion and metabolic stress, in neuronal cells and mouse models expressing mutant superoxide dismutase 1 (SOD1) or TAR DNA binding protein 43 (TDP-43) linked to ALS. AMPK phosphorylation was sharply increased in spinal cords of transgenic SOD1G93A mice at disease onset and accumulated in cytoplasmic granules in motor neurons, but not in presymptomatic mice. AMPK phosphorylation also occurred in peripheral tissues, liver and kidney, in SOD1G93A mice at disease onset, demonstrating that AMPK activation occurs late and is not restricted to motor neurons. Conversely, AMPK activity was drastically diminished in spinal cords and brains of presymptomatic and symptomatic transgenic TDP-43A315T mice and motor neuronal cells expressing different TDP-43 mutants. We show that mutant TDP-43 induction of the AMPK phosphatase, protein phosphatase 2A (PP2A), is associated with AMPK inactivation in these ALS models. Furthermore, PP2A inhibition by okadaic acid reversed AMPK inactivation by mutant TDP-43 in neuronal cells. Our results suggest that mutant SOD1 and TDP-43 exert contrasting effects on AMPK activation which may reflect key differences in energy metabolism and neurodegeneration in spinal cords of SOD1G93A and TDP-43A315T mice. While AMPK activation in motor neurons correlates with progression in mutant SOD1-mediated disease, AMPK inactivation mediated by PP2A is associated with mutant TDP-43-linked ALS.


Introduction
Amyotrophic lateral sclerosis (ALS) is a progressive and fatal paralysing disorder caused by selective degeneration of upper and lower motor neurons [1,2]. The reason for theselective vulnerability of motor neuronsto mutations in ubiquitously expressed proteins, such as superoxide dismutase 1 (SOD1) and TAR DNA-binding protein-43(TDP 43), remains unclear [1,2]. Factors accounting for this selective neuronal susceptibility in ALS may include the unusual high synthetic, energetic and transport demands of large projection motor neurons [3]. This leads to high ATP consumption and mitochondrial metabolism relative to other cells, rendering these neurons susceptible to energetic defects [4]. Thus, defects in mitochondrial function that occur in ALS patients and mouse models [5,6] will affect ATP production, impairing Na + /K + -ATPase function and maintenance of resting membrane potential in motor neurons [7].
There is considerable evidence for hypermetabolism and impaired energy homeostasis in ALS patients and genetic mouse models.ALS patients show accelerated loss of muscle mass and fat during disease progression [8,9] and elevated energy expenditure at rest [10,11].In mutant SOD1 G86R and SOD1 G93A mice,glucose, ATP and fat metabolism are increased in spinal cord and brain at presymptomatic disease, indicative of energy hypermetabolism [12,13].In symptomatic SOD1 G93A mice, there is increased metabolic acidosis, lipolysis and glycogen accumulation in the CNS [14].Furthermore, a high-fat diet significantly delayed disease onset and increased lifespan in mutant SOD1 mice [12], whilecaloric restriction accelerated disease course [15,16].
There is also evidence for metabolic dysfunction in ALS mediated by abnormalTDP-43. Postnatal depletion of TDP-43 in miceleads todramatic loss of body fat followed by rapid death and Tbc1d1 expression, a gene that mediates leanness andlinked to obesity, is downregulated in the absence of TDP-43 [17]. In TDP-43 A315T mice, there is progressive weight gain, fat deposition, impaired glucose uptake and increased Tbc1d1 expression in presymptomatic mice [18].Together,this suggests that TDP-43 expression level is an important regulator of glucose and energy metabolism, while more importantly, mutant TDP-43 leads to metabolic dysfunction in an opposing manner to mutant SOD1.
Here, we sought to correlate these well established findings ofdefective energy metabolism in mutant SOD1 and TDP-43 mice with AMP-activated protein kinase (AMPK) activation status.AMPKis a widely expressed master metabolic and stress sensor which detects mismatches incell energy supply and demand [19]. It is activated by high AMP:ATP ratio and metabolic stresses that inhibit ATP production or stimulate ATP consumption [19]. AMPK is a heterotrimeric protein consisting of acatalyticand regulatory band c subunits. Binding of AMP to the c subunit stimulates a subunit phosphorylation at the active site threonine-172 bytwo known upstream AMPK kinases,liver kinase B1 (LKB1) and Ca 2+ /calmodulin-dependent protein kinase kinase (CaMKK) as reviewed [19].AMPKinactivation by dephosphorylation has been linked to protein phosphatase 2A (PP2A) [20], PP2C [21], PP1 [22] and recently PPM1E [23].Activation of AMPKreplenishes cellular ATP levelsby promoting increased fatty acid oxidation, glucose uptake and autophagy, while reducing protein translation, glycogen and cholesterol synthesis.
Since AMPK is a master regulatorof energy metabolism and stress induced pathways,it is likely to be engaged in energetic defects in motor neurons in ALS. We therefore examined AMPK activation in SOD1 G93A and TDP-43 A315T mice and neuronal cells expressingdifferent ALS-linked SOD1 or TDP-43 mutants.
Here, we show that AMPK activation is increased in spinal cords of SOD1 G93A mice consistent with energy hypermetabolism, while AMPK activation is severely reduced in CNS tissues of TDP-43 A315T mice, in accordance with weight gain and fat accumulation characteristic of this model. We also uncover a novel regulation of AMPK activation by mutant TDP-43 via PP2A.

Ethics statement
All experiments conformed to the Australian National Health and Medical Research Council published code of practice and were approved by the Howard Florey Institute Animal Ethics Committee.

Transgenic mice
Transgenic SOD1 G93A mice derived from the B6SJL-TgN(SOD1*G93A)1Gur line were obtained from the Jackson Laboratory (Bar Harbor, ME) and backcrossed onto a pure C57BL/6 background. Non-transgenic littermates of SOD1 G93A mice were used astheir wild-type (WT) controls.Transgenic TDP-43 A315T mice derived from the B6.Cg-Tg(Prnp-TARDB-P*A315T)95Balo/J line (Jackson Laboratory) were maintained on aC57BL/6 background. Non-transgenic littermates of TDP-43 A315T mice were used as their WT controls.Only male mice were analysed in this study due to the substantial sex differences in survival of TDP-43 A315T mice.

Tissue extraction
Animals were killed by lethal injection (100 mg/kg, intraperitoneal, sodium pentobarbitone) at postnatal day (P) 60 or 90 and lumbar spinal cord, whole brain, kidney and liver were dissected out and snap-frozen. Tissues were homogenised in RIPA lysis buffer containing 50 mM Tris-Cl, pH 7.4, 150 mM NaCl, 0.1% SDS, 1% sodium deoxycholate, 1% TX-100, 1% protease inhibitor cocktail (Sigma) and phosphatase inhibitors (50 mM NaF and 10 mM sodium pyruvate) and sonicated at 50% output for 15 sec, stored on ice for 20 min and centrifuged at 15,800 g for 20 min at 4uC to collect supernatants. Proteins were quantified using the bicinchoninic acid assay kit (Pierce) using bovine serum albumin standards.

Cell lysis
Cells were lysed in buffer containing 20 mM Tris-Cl, pH 7.5, 150 mM NaCl, 1% Triton-X 100, 1% (v/v) protease inhibitor cocktail (Sigma) and phosphatase inhibitors (50 mM NaF and 10 mM sodium pyruvate) for 20 min on ice. Lysates were centrifuged at 14, 000 rpm for 20 min at 4uC to collect supernatants. Proteins were quantified as above for tissue.

Immunocytochemistry
Cells were cultured for 2 days and fixed with 4% paraformaldehyde (PFA)for 10 min, permeabilised in 0.4% TX-100 in phosphate-buffered saline (PBS) for 10 min, blocked in 5% normal goat serum in PBS for 30 min and incubated with rabbit phospho-AMPK (1:100, Cell Signaling Technology)antibody in blocking buffer overnight at 4uC, washed three-times with PBS, for 10 min and incubated with Alexa Fluor 488 conjugated secondary antibody (1:1,000, Molecular Probes) for 2 hr,stained with Hoechst 33342 (1:10,000, Invitrogen) for 15 min, washed three times with PBS for 10 min and mounted using fluorescent mounting medium (Dako) on glass slides (Thermo Scientific) for microscopy using an Olympus FV 1000 confocal microscope. Images were captured using identical exposure and gain settings. Negative controls without primary antibodies were performed which produced no staining.

Immunohistochemistry
Mice were transcardially perfused with PBS followed by 4% (w/v) PFA in 0.1 M phosphate buffer. Lumbar spinal cords were dissected out, post-fixed in 4% (w/v) PFA for 2 hr, processed, dehydrated and embedded in paraffin and cut into 20 mm transverse sections. Sections were deparaffinised, treated with20 mg/mlProteinase K (Qiagen)in PBS at 37uC for 5 min for antigen retrieval, permeabilised in 0.4% TX-100 in PBS for 10 min, blocked in 5% normal goat serum in PBS for 30 min and incubated with rabbit phospho-AMPK (1:100, Cell Signaling Technology) and mouse NeuN (1:1,000, Millipore, MAB377) antibodies in blocking buffer overnight at 4uC. Sections were washed three times with PBS for 10 min, incubated with Alexa Fluor 488 and Alexa Fluor 594 conjugated secondary antibodies (1:1,000, Molecular Probes) for 2 h, stained with Hoechst 33342 (1:10,000, Invitrogen) for 15 min, and washed three times with PBS for 10 min before mounting using fluorescent mounting medium onto glass slides for confocal microscopy.The number of cells with pAMPK granules was counted from 30 motor neurons per mouse, 3 mice per group, and expressed as a percentage of WT (100%). The number of pAMPK granules per cell was counted from 30 motor neurons per mouse, 3 mice per group.

Statistical analysis
Western densitometry datawere analysed by unpaired t-test or one-way ANOVA with Tukey's post-hoc test.Motor neuron and pAMPK granule quantification was analysed by unpaired t-test. All statistical tests were performed with GraphPad Prism software (version 5.0, GraphPad Software, San Diego, CA).

AMPK activation is increased in spinal cords of symptomatic mutant SOD1 mice
We first determined AMPK activationstatus in SOD1 G93A mice at presymptomatic (P60) and symptomatic (P90) ages. SOD1 G93A mice appear normal at P60 and develop hindlimb muscle weakness and wasting at P90. AMPK activation by phosphorylation of a subunit threonine-172 was assessed by immunoblotting for phospho (pAMPK) and total AMPK levels in tissues of SOD1 G93A mice and WT littermates. Western blot analysis showed pAMPK and AMPK bands at 62 kDa as expected (Fig. 1). AMPK activation, determined by pAMPK/AMPK ratio, was similar in spinal cords of P60 mice of both genotypes (Fig. 1A,B). However, AMPK activation in spinal cords of SOD1 G93A mice was 300% greater than in spinal cords of WT mice at P90 (Fig. 1B, p,0.05). There was no change in AMPK activation level in brains of P60 and P90 SOD1 G93A mice, however (Fig. 1C,D).
To determine whether increased AMPK activation is selective to spinal cord, we next examined pAMPK/AMPK ratio in nonaffected tissues such as liver and kidney. AMPK activation was elevated by 60% and 80% in liver of SOD1 G93A mice at P60 and P90, respectively, which was statistically significant at P90 (Fig. 1E,F,p,0.05). AMPK activation was also significantly increased by 80% in kidney of SOD1 G93A mice at P60 (Fig. 1G,H, p,0.01), while unchanged at P90. Thus, AMPK activation occurs at symptom onset in spinal cords of SOD1 G93A mice and in peripheral tissues, suggesting that AMPK signalling and therefore energy depletion is a late event in pathogenesis and not restricted to degenerating neuronsin this mouse model of ALS.
AMPK activation is reduced in spinal cords and brains of mutant TDP-43 mice AMPK activation was then examined in TDP-43 A315T mice at presymptomatic (P60) and symptomatic (P90) stages. TDP-43 A315T mice appear healthy at P60 and develop an abnormal swimming gait at P90 without gross muscle wasting.In contrast to SOD1 G93A mice, AMPK activation wasdrastically decreased by 60% at P60 (p,0.05)and 30% at P90 in spinal cords of TDP-43 A315T mice ( Fig. 2A,B). In brain, pAMPK/AMPKratios were significantly diminished by 80% at P60(p,0.001) and 60% at P90 (p,0.05) in TDP-43 A315T mice (Fig. 2E,F). AMPK activation was not studied in liver and kidney,which do not express the TDP-43 A315T transgene in our hands and others [32].Hence, mutant TDP-43 leads to reduced activation of AMPK in spinal cord and brain which was more severe in the lattertissue, consistent with pronounced cortical pathology in these mice [32].
We next screened expression level of phosphatases that could be responsible for dephosphorylation of AMPK inTDP-43 A315T mice. We examined four candidate AMPK phosphatases:protein phosphatase 2A (PP2A), PP2C, PP1 and PPM1E. The levels of PP2C, PP1 and PPM1E were similar in spinal cords of presymptomatic and symptomatic TDP-43 A315T mice compared to WT (Fig. 2C). However, levels of PP2Aweresignificantly increased by 120% in spinal cords of TDP-43 A315T mice at P60 (p,0.05), positively correlating with AMPK inactivation (Fig. 2C,D). This suggests that PP2A is the likely phosphatase responsible for AMPK inactivation in TDP-43 A315T mice and thisphosphatase remained the focus of this study.We also assessed PP2A expression in SOD1 G93A mice, showing that PP2A levels were significantly diminished by 30% in spinal cords of mice at P90 (p,0.05), showing an inverse correlation with AMPK activation (Fig. 1A,I). This strengthens the evidence that PP2A is modulating AMPK activation in the CNS of mutant SOD1 and TDP-43 mice.

Activated AMPK distribution in spinal motor neurons of normal mice and ALS mouse models
The subcellular distribution of AMPK was examined in spinal cord sections of mice at P90 using immunohistochemistry. In WT mice,there were small pAMPK-positive punctae in the cytoplasm of motor neurons, identified bytheir size, ventral horn locationand NeuN immunoreactivity (Fig. 3, arrowheads). In SOD1 G93A mice, there was increased pAMPK immunoreactivity in the cytoplasm of motor neurons (Fig. 3A), consistent with Western blotting findings (Fig. 1A,B). Increased pAMPK was restricted to motor neurons and was not observed in NeuN-negative Hoechst-positive cells. We quantified the percentage of motor neurons with cytoplasmic Figure 1. AMPK activation is increased in spinal cords of symptomatic mutant SOD1 mice, but not pre-symptomatic mice. Immunoblot analysis of phosphorylated (pAMPK) and total AMPK levels in A spinal cord,C brain,E liver and G kidney of pre-symptomatic (P60) and symptomatic (P90) transgenic SOD1 G93A and age-matched wild-type (WT) mice. Quantification of pAMPK/AMPK ratio level in B spinal cord, D brain, F liver and H kidney from immunoblots normalised to WT mice. Immunoblot analysis of A, PP2A and I, quantification of PP2A level from immunoblots for spinal cord. Data represent mean 6 SEM, n = 4-5 mice per group, *p,0.05 and **p,0.01 compared to WT mice using unpaired t-test. doi:10.1371/journal.pone.0090449.g001 pAMPK granules, showing that almost all motor neurons in WT and SOD1 G93A mice contained pAMPK punctae, consistent with the ubiquitous expression of AMPK (Fig. 3B). However, the number of pAMPK granules was doubled in motor neurons of SOD1 G93A mice compared to WT mice (p,0.05), indicative of increased AMPK activation (Fig. 3C).
In TDP-43 A315T mice, pAMPK immunoreactivestaining in motor neurons was less than in WT mice (Fig. 3D), again concordant with Western results (Fig.2A,B). The number of motor neurons containing pAMPK granules was reduced by 50% in TDP-43 A315T mice (Fig. 3E, p,0.05). Furthermore, there was severe depletion of pAMPK granules by 80% in motor neurons of mutant TDP-43 mice (Fig. 3F, p,0.05), consistent with AMPK inactivation in spinal cords of these mice. Thus, activated AMPK is predominantly found inmotor neurons and localised to cytoplasmic granules in the spinal cord and abnormally expressed in mutant SOD1 and TDP-43mouse models of ALS.

AICAR stimulates activation of AMPK in NSC-34 cells
We next studied whether similar changes inAMPK activation occurin motor neuronal NSC-34 cells expressing different ALSlinked SOD1 and TDP-43 mutations. The physiological activation response to the AMPK agonist AICAR was first established in NSC-34 cells. Treatment of NSC-34 cells with AICAR for 2 hr resulted in dose-dependent phosphorylation of AMPK withmaximal stimulation at 2 mM concentration (Fig. 4A). Immunocytochemical analysis of untreated NSC-34 cells revealed that pAMPK was primarily and diffusely localisedto the cytoplasm (Fig. 4B). There was a dose-dependent increase in cytoplasmic pAMPK immunoreactivity with AICAR treatment. AICAR treatment did not adverselyaffect NSC-34 cellviability (Fig. 4C). Thus, activated AMPK is mainly cytoplasmic in NSC-34 cells and motor neurons in spinal cord.

AMPK activation in NSC-34 cells expressing SOD1 mutants
AMPK activation in NSC-34 cells stably transfected with dismutase active (G37R, G93A) or inactive (A4V, G85R) human SOD1 mutants was compared to NSC-34 cells expressing WT human SOD1. Although transient transfection with these mutants results in protein aggregates, ER stress and cell death [24,25,26], these features were not found in these stable cells, whichmay represent a pre-symptomatic ALS statei.e. SOD1 G93A mice at P60. In support of this, Western blot analysisshowed similar levels of Figure 2. AMPK activation is diminished in spinal cords and brains of mutant TDP-43 mice from pre-symptomatic age. Immunoblot analysis of phosphorylated (pAMPK) and total AMPK levels in A, spinal cord and E, brain of pre-symptomatic (P60) and symptomatic (P90) transgenic TDP-43 A315T and age-matched WT mice. Quantification of pAMPK/AMPK ratio level in B, spinal cord and F, brain from immunoblots normalised to WT mice.Immunoblot analysis of C, PP2A, PP2C, PP1 and PPM1E withE, quantification of PP2A level from immunoblots for spinal cord. Data represent mean 6 SEM, n = 4-5 mice per group, *p,0.05 and ***p,0.001 compared to WT mice using unpaired t-test. doi:10.1371/journal.pone.0090449.g002 AMPK activation in cells expressing normal or mutant SOD1, although there was a non-significant trend for diminished AMPK activation (Fig. 5A,B). Likewise, PP2A levels were similar in cells expressing WT or mutant SOD1 (Fig. 5C).These Western blot findings were correlated with immunocytochemical analysis, showing that pAMPK is mainly cytoplasmicin NSC-34 cells expressing normal or mutant SOD1 (Fig. 5D). Therefore, AMPK activation does not occur in presymptomatic neuronal cells or mouse models expressing mutant SOD1.
To determine whether mutant TDP-43 is driving AMPK inactivation by PP2A, we treated NSC-34 cells expressing WT or mutant TDP-43 with the selective PP2A inhibitor okadaic acid (OA). OA treatment at 1 mM led to increased AMPK activation in cells expressing WT, Q331K or M337V TDP-43, which was more pronounced for mutant TDP-43 (Fig. 6E). Hence, mutant TDP-43 reduction of AMPK activation occurs by a PP2A-dependent mechanism.

Discussion
We demonstrate several novel findings in this study. Firstly, AMPK activation occurs only at disease onset in spinal cords of mutant SOD1 mice, but is not restricted to this tissue, occurring also in non-affected liver and kidney. Secondly, AMPK activation is severely diminished in spinal cords and brains of mice and motor neuronal cells expressing mutant TDP-43. Thirdly, mutant TDP-43 inhibition of AMPK phosphorylation ismediated by PP2A. Lastly, activated AMPK is predominantly localised tosmall cytoplasmic granules of motor neurons in adult spinal cord.
Elevated AMPK activation inspinal cords of symptomatic SOD1 G93A mice correlateswell with key indices of energy hypermetabolism reported in this model, including increased glucose and ATP consumption [13], lactate metabolism [2], lipolysis [12] and metabolic acidosis [14].The presence of activated AMPK in both CNS and peripheral tissues in mutant SOD1 mice at symptom onset argues against a role for AMPK signalling in disease initiation, although it may be a determinant of disease progression. In support of this, treatment of SOD1 G93A mice with metformin, which activates AMPK, conferred no significant effect on disease onset, but hastened death in female mice only [27]. Consistent with our findings,increased AMPK activation in spinal cords of P90 mutant SOD1 mice, but not juvenileP40 mice, was recently reported [28]. Here, we resolved the time course of AMPK phosphorylation, showing activation at P90, but not at P60, in these mice. However, this study also found that AMPKmediated transcription of its downstream target genes (e.g.PCG-1a, GPx1 and SOD2) was unaltered in spinal cords of SOD1 G93A mice [28] and thus, the role of elevated AMPK signalling in mutant SOD1 mice remains unclear. We suggest that increased AMPK activation in spinal cords of SOD1 G93A mice reflects energy depletion in motor neurons well after they start degenerating and is a late event in ALS pathogenesis.This might be brought about by reactivity to muscle wasting and weight loss at symptom onset [29]. Pharmacological activation of AMPK by AICAR or its inhibition by compound C [19] is therefore unlikely to significantly modify disease in mutant SOD1 mice or provide an effective therapeutic benefit in ALS.
In contrast to increased AMPK activation in SOD1 G93A mice, its activity was drastically diminished in mice and neuronal cells expressing TDP-43 mutants. While this was an unexpected finding, it can be rationalised in the case of TDP-43 A315T mice. Unlike SOD1 G93A mice, TDP-43 A315T mice develop mild motor neuron loss in the spinal cord and do not progressively lose muscle mass or succumb to ALS-like symptoms, but rather die from bowel obstruction caused by gut innervation defects [30,31]. In contrast, TDP-43 A315T mice have pronounced neuronal loss and ubiquitin pathology in cortex, which would be consistent with greater depletion of AMPK activation in brain, compared to spinal cord in this model [32]. Most recently, it was shown that TDP-43 A315T mice develop progressive weight gain, increased body fat and adipocyte hypertrophy [18], which is consistent with a state of energy surplus, and therefore AMPK inactivation as shown here.Interestingly,analysis of neuronal cells expressing different TDP-43 mutants also revealed reduced AMPK activation similar to mice. Expression of PP2A,but not other reported AMPK phosphatases,was induced by mutant TDP-43 and treatment with the PP2A inhibitor okadaic acid reversed this effect in cell culture.In another study, PP2A inhibition by okadaic acid or PP2AC knockdown both restored AMPK phosphorylation in cells exposed to heat shock, supporting that AMPK inactivation is driven by PP2A under cell stress [33]. We therefore concludethat PP2A mediates AMPK inactivation observed in cell and mouse models expressing mutant TDP-43.There is increasing evidence that TDP-43 plays an important role in regulating energy metabolism [17,18] and stress-inducible kinases such as c-Jun Nterminal kinase andextracellular-signal-regulated kinase [34] and modulation of PP2A may therefore provide an important link coupling pathogenic TDP-43 to AMPK activity.
In conclusion, our results demonstrate contrasting effects of mutant SOD1 and TDP-43 on AMPK activation in mouse models of ALS. This is likely to reflect key differences inneurodegeneration, resulting muscle wasting and energy metabolism characteristic of these models. In SOD1 G93A mice characterised primarily by spinal motor neuron loss, muscle atrophy and weight loss, AMPK activation is increased which is indicative of energy hypermetabolism. In TDP-43 A315T mice with mainly cortical motor neuron pathology and minimal spinal motor neuron loss without muscle wasting, AMPK activation is reduced, in keeping with weight gain and fat deposition in this model. Furthermore, we uncover a novel regulation of AMPK activation mediated by PP2A by mutant TDP-43 in vitro and in vivo. Modulation of PP2A levels in mutant TDP-43 models may therefore present an interesting therapeutic approach in ALS.