PPAR gamma activation is neuroprotective in a Drosophila model of ALS based on TDP-43

Abstract

Amyotrophic Lateral Sclerosis (ALS) is a progressive neuromuscular disease for which there is no cure. We have previously developed a Drosophila model of ALS based on TDP-43 that recapitulates several aspects of disease pathophysiology. Using this model, we designed a drug screening strategy based on the pupal lethality phenotype induced by TDP-43 when expressed in motor neurons. In screening 1200 FDA-approved compounds, we identified the PPARγ agonist pioglitazone as neuroprotective in Drosophila. Here, we show that pioglitazone can rescue TDP-43-dependent locomotor dysfunction in motor neurons and glia but not in muscles. Testing additional models of ALS, we find that pioglitazone is also neuroprotective when FUS, but not SOD1, is expressed in motor neurons. Interestingly, survival analyses of TDP or FUS models show no increase in lifespan, which is consistent with recent clinical trials. Using a pharmacogenetic approach, we show that the predicted Drosophila PPARγ homologs, E75 and E78, are in vivo targets of pioglitazone. Finally, using a global metabolomic approach, we identify a set of metabolites that pioglitazone can restore in the context of TDP-43 expression in motor neurons. Taken together, our data provide evidence that modulating PPARγ activity, although not effective at improving lifespan, provides a molecular target for mitigating locomotor dysfunction in TDP-43 and FUS but not SOD1 models of ALS in Drosophila. Furthermore, our data also identify several ‘biomarkers’ of the disease that may be useful in developing therapeutics and in future clinical trials.

Introduction

Amyotrophic Lateral Sclerosis (ALS) is a progressive neurodegenerative disease that leads to paralysis and death within 2–5 years after diagnosis (1). It is a debilitating neuromuscular disease affecting upper and lower motor neurons for which there is no cure, and the available treatments are palliative at best. Several loci, including SOD1, TDP-43, FUS, C9ORF72 and profilin, have been implicated in ALS (2–6). In particular, TDP-43 protein has been linked to a vast majority of ALS cases owing to its presence in pathological aggregates. In addition to being a marker of pathology, TDP-43 can harbor disease causative mutations that lie primarily in the C terminal domain of the protein (7–11). TDP-43 is an RNA-binding protein that normally resides within the nucleus, can shuttle into the cytoplasm via a Nuclear Export Signal and harbors a prion-like domain within its C-terminus (12,13). TDP-43's normal functions include various aspects of RNA processing such as splicing, transport and translation (14–17), with several studies providing evidence for RNA dysregulation in disease (18–20).

Evidence exists to indicate that ALS is a proteinopathy accompanied by abnormalities in diverse cellular processes including oxidative stress response, defects in ER-mitochondrial interaction, apoptosis and dysregulation of cellular metabolism among others (21–28). Findings that ALS patients exhibit hypermetabolism accompanied by dramatic weight loss have led to the hypothesis that cellular and/or systemic metabolism are involved in the onset and the progression of the disease (29,30). Additional observations supporting this idea include studies in SOD1 mice showing that a high fat diet is protective (31) and a recent Phase 2 clinical trial that found high calorie diets to be safe and well-tolerated in patients, with an apparent improvement in lifespan (28) In another study, diabetic patients had a later onset and slower progression of ALS (32). Taken together, these data indicate that some aspects of cellular metabolism may be linked to disease progression.

To identify compounds and molecular targets with neuroprotective potential in ALS, we performed a drug screen using a Drosophila model of ALS based on TDP-43 (33–35). In screening 1200 FDA-approved compounds for their ability to rescue TDP-43-induced lethality, we identified several antidiabetic drugs including thiazolidinediones (TZDs), which bind with high affinity and stimulate the activity of the nuclear receptor PPARγ (36,37). Peroxisome proliferator-activated Receptors (PPARs) are ligand-activated transcription factors belonging to the nuclear hormone receptor superfamily (38). They control several physiological processes including glucose and lipid metabolism as well as growth and differentiation. Upon ligand binding, PPARγ forms heterodimers with Retinoid X receptors (RXRs), then binds and activates the transcription of target genes involved in lipid metabolism. PPARγ is expressed in several cell types, including adipose tissues, cells of the immune system and parts of the brain including microglia and astrocytes, the sites that contribute to anti-inflammatory response in the CNS (39). Interestingly, activation of PPARγ has been shown to have anti-inflammatory and neuroprotective effects in ALS and related neurodegenerative conditions (39). Here, we determine the effects of pioglitazone in multiple Drosophila models of ALS including TDP-43, FUS and SOD1. While pioglitazone could rescue TDP-43 and FUS-dependent phenotypes in motor neurons, it had no effect in the SOD1 model. Interestingly, the protective effects of pioglitazone were observed when TDP-43 was expressed in motor neurons or glia, but not in muscles, indicating that PPARγ activation is required in the nervous system to mitigate TDP-43 toxicity. In contrast to its positive effects on locomotor function, pioglitazone did not improve, and in some cases, shortened lifespan. Using a pharmacogenetic approach, we also show that E75 and E78, the predicted Drosophila homologs of PPARγ, are required for mediating the neuroprotective effect of pioglitazone on TDP-43 toxicity in vivo. Metabolic profiling experiments identify a subset of metabolites that are altered in the context of TDP-43 and can be restored by pioglitazone in a variant dependent manner. Taken together, these data indicate that PPARγ activation in neurons and glia is partially neuroprotective and can restore a subset of metabolic alterations in ALS that may serve as useful biomarkers in the development of therapeutic strategies and future clinical trials.

Results

Drug screening in a Drosophila model of ALS based on TDP-43 identifies pioglitazone, a PPARγ agonist as neuroprotective

Using a Drosophila model of ALS based on TDP-43 that recapitulates several aspects of the disease pathophysiology (33,35), we designed a drug screen aimed at identifying FDA-approved compounds with neuroprotective potential in vivo (Fig. 1A). We expressed human TDP-43, either wild-type or disease-associated mutants D169G, G298S, A315T and N345K (8,10,40) in motor neurons using the D42-Gal4 driver (41) and found that this results in nearly 100% pupal lethality (Fig. 1A). Few pharate adults eclose but cannot extend their wings (Fig. 1A). This phenotype was used to screen for drug candidates with therapeutic potential as determined by their ability to rescue pupal lethality and produce one or more viable adults with extended wings (Fig. 1A). There are several advantages to this screening strategy: (1) it is performed in vivo, (2) it is based on a robust phenotype (>95% pupal/adult pharate lethality), (3) it can lead to rapid identification of known safe drugs that are already used in humans and (4) can inform the future development of novel small molecules with enhanced neuroprotective capabilities. In screening the Prestwick collection of 1200 FDA-approved drugs at 30–50 μm, we identified several antidiabetic drugs comprising different categories including TZDs, sulfonylureas and biguanides. Of these, we focused initially on the two TZDs identified in the primary screen, namely pioglitazone and troglitazone. Upon secondary screening, pioglitazone was determined to be more potent (data not shown) and selected for further studies.

Figure 1.

Pioglitazone rescues several aspects of TDP-43-dependent toxicity in motor neurons. (A) Drug screen strategy. Overexpression of TDP-43 (wild type or mutants) in motor neurons is lethal at the pupal or pharate adult stage. In screening the Prestwick collection of FDA-approved drugs, this phenotype was used to identify adult survivors with fully extended wings (rescue of lethality). (B and C) Pioglitazone rescues TDP-43-dependent pupal lethality. Wild-type TDP-43 (TDP^WT, B) or disease-associated G298S mutant (TDP^G298S, C) expressed in motor neurons using D42-Gal4 driver result in >95% pupal lethality. D42>TDP larvae are grown on fly food containing different concentrations of pioglitazone or DMSO (as indicated). Surviving adults were normalized to pupae numbers and plotted on the Y-axis, as shown. Error bars indicate SEM. (D and E) Pioglitazone rescues larval locomotor defects caused by TDP-43 overexpression in motor neurons. D42>TDP^WT (D) and D42>TDP^G298S (E) larvae grown on fly food containing different concentrations of pioglitazone or DMSO (as indicated) are assayed for locomotor function using larval turning assays (see Materials and Methods). Note that at the concentration of 1 μm, pioglitazone significantly rescues locomotor dysfunctions for both TDP^WT and TDP^G298S. Error bars indicate SEM. Student's T test is used to assess statistical significance. ***P<0.001. (F and G) Pioglitazone does not improve lifespan of TDP-43-expressing flies. 1 μm pioglitazone is administered during development (F) or to adults (G) as indicated. Kaplan–Meier survival analysis shows no significant change in lifespan under either condition when compared with DMSO-fed controls. (H and I) Western blot shows no change in TDP-43 protein levels with or without pioglitazone. Larvae raised on fly food containing DMSO or 1 μm pioglitazone are crushed in Laemmli buffer and subjected to western blotting and probed for TDP-43 or tubulin for loading control (H). Results show no change in protein levels with or without pioglitazone (I). Error bars indicate SEM. Student's T test is used to assess statistical significance.

To assess the neuroprotective effect of pioglitazone, we chose two transgenic lines (TDP^WT and TDP^G298S) expressing comparable levels of TDP-43 (data not shown). Larvae expressing TDP-43 in motor neurons (D42 >TDP-43) were raised on fly food containing different amounts of drug (1, 10 and 25 μm, see Materials and Methods for details) or DMSO as vehicle control (Fig. 1B and C). As seen in Figure 1B and C, we found that although there was no clear dose response in this assay, lower drug concentrations exhibited higher neuroprotective potential and pupal lethality caused by the expression of either TDP^WT or TDP^G298S in motor neurons was best rescued by 1 μm pioglitazone (see also Table 1). Taken together, these data indicate that pioglitazone mitigates TDP-43-dependent pupal lethality.

Table 1.

Summary of rescue of lethality by pioglitazone

PGZ conc. (μm)	TDPWT				TDPG298S
	DMSO		PGZ		DMSO		PGZ
	Average % rescue	SEM	Average % rescue	SEM	Average % rescue	SEM	Average % rescue	SEM
1	0	0	15.2	0.8	0.7	0.7	19.4	5.0
10	6.8	3.8	1.7	0.8	0.7	0.7	15.0	6.5
25	0	0	0.8	0.7	3.8	2.7	9.7	3.6

Genotype, drug concentration and experimental conditions, as indicated. Average percent rescue was calculated as (number of rescued adults/total number of pupae) × 100%. Notably, lower drug concentrations led to better rescue.

PGZ, pioglitazone containing food; DMSO, vehicle control; SEM, standard error of mean.

Larval locomotor deficits caused by TDP-43 expression in motor neurons are rescued by pioglitazone

Next, we set out to determine an ideal drug concentration using a more sensitive and disease-relevant assay that measures neuromuscular coordination, namely larval turning (see Materials and Methods for details). Larvae expressing TDP^WT or TDP^G298S in motor neurons (D42>TDP-43) were raised on food containing either 1, 5, 10, 25 μm pioglitazone or DMSO and assayed for their locomotor ability using larval turning assays. As shown in Figure 1D and E, pioglitazone significantly improved larval locomotion at 1 μm for both TDP^WT (9.2 ± 0.7 s on pioglitazone compared with 14.5 ± 1 s on DMSO, P = 0.7E − 4) and TDP^G298S (17.3 ± 1.1 s on pioglitazone compared with 25.8 ± 1.8 c on DMSO, P = 0.2E − 3). No rescue was observed at higher drug concentrations, and lower doses were not tested. Importantly, 1 μm pioglitazone had no effect on larval turning time in wild-type control larvae (D42>w¹¹¹⁸, data not shown) indicating that the drug effect is specific to TDP-43-dependent phenotypes. Finally, the neuroprotective effect of pioglitazone was not due to a reduction in TDP-43 protein levels as determined by western blot (Fig. 1H and I). These results indicate that the ideal concentration for rescuing TDP-43-dependent larval locomotor defects in the range tested is 1 μm pioglitazone, which was used for all subsequent experiments.

Pioglitazone exerts no protective effect on lifespan in the context of TDP-43 expression in motor neurons

We next tested whether pioglitazone can also improve the decrease in lifespan caused by TDP^WT or TDP^G298S expression in motor neurons. To this end, we raised larvae expressing either TDP^WT or TDP^G298S in motor neurons on 1 μm pioglitazone and found that the rescued adults had a similar lifespan to larvae raised on DMSO-containing food (Fig. 1F). As obtaining adults on DMSO is very difficult owing to the >95% pupal lethality caused by TDP-43, these studies were performed with 10–15 adult flies (‘escapers’) only. Thus, although 1 μm pioglitazone can improve locomotor function when administered during development, it has no effect on the TDP-43-dependent decrease in adult lifespan. To determine whether pioglitazone can be effective ‘after disease onset’, adult D42>TDP-43 ‘escapers’ raised on normal food (see Materials and Methods for details) were fed either 1 μm pioglitazone or DMSO-containing food after eclosion. As with the developmental feeding experiment, we found that pioglitazone has no significant effect on lifespan (Fig. 1G) at least at the 1 μm concentration, which was determined to be optimal for rescuing larval locomotor defects (Fig. 1D and E). These data are consistent with recent clinical trials in which pioglitazone failed to show improvement in ALSFRS-R scores, a well-established clinical measure of disease progression (42). Our results indicate that while pioglitazone is effective in improving locomotor function when administered during development, it has no effect on lifespan regardless of whether it is provided throughout development or ‘after disease onset’.

Glial toxicity caused by TDP-43 is partially mitigated by pioglitazone

Glial and muscle cells have been previously implicated in ALS pathophysiology (43,44). We have previously shown that TDP-43 expression in glial cells results in cell autonomous locomotor defects and abnormalities in synaptic protein distribution at the neuromuscular junction (35). To determine whether pioglitazone may also mitigate glial toxicity, we performed turning assays on larvae expressing TDP-43 using the pan-glial Repo-Gal4 driver. As shown in Figure 2A, these experiments indicate that 1 μm pioglitazone rescues locomotor function abnormalities caused by TDP-43 expression in glia compared with both DMSO and w¹¹¹⁸ controls. These data demonstrate that pioglitazone exerts neuroprotective effects on TDP-43-induced glial toxicity as it does in motor neurons.

Figure 2.

TDP-43-dependent toxicity in glia, but not in muscles, is partially mitigated by pioglitazone. (A and B) Larval locomotor defects are mitigated by pioglitazone when TDP-43 is expressed in glia, but not in muscles. TDP-43 (wild type or mutant, as indicated) is expressed in glia using Repo-Gal4 driver (A), or in muscles using BG487-Gal4 driver (B). Turning assay on larvae raised on fly food containing 1 μm pioglitazone or DMSO shows that pioglitazone can improve glial toxicity, but not that in muscles. Error bars indicate SEM. Student's T test is used to assess statistical significance. *P< 0.05. (C and D) Pioglitazone does not improve survival of Repo>TDP flies. 1 μm pioglitazone is administered during development (C) or to adults (D) as indicated. Note that Kaplan–Meier survival analysis shows that developmentally administered pioglitazone results in significantly reduced lifespan for TDP^WT flies (C), while not affecting the lifespan under any other condition. ***indicates P< 0.001.

Next, we tested whether 1 μm pioglitazone can also increase the lifespan of adult flies expressing TDP-43 in glia and raised on pioglitazone containing food throughout development. These experiments showed no difference in lifespan for TDP^G298S, whereas TDP^WT-expressing flies lived significantly less on pioglitazone compared with DMSO (Fig. 2C). Interestingly, an increase in hazard ratio for mortality was found in ALS patients who were administered pioglitazone (42). This is in contrast to TDP^WT expression in motor neurons, where there was no difference in lifespan between drug and DMSO-fed groups (Fig. 1F). Furthermore, when pioglitazone was administered to adults expressing TDP-43 in glia, we found no difference in lifespan between the drug-treated flies and DMSO controls (Fig. 2D). These data suggest that although the drug can improve locomotor function caused by TDP-43 toxicity in glial cells, it has either no effect or can be detrimental to lifespan in flies as it is in humans.

Locomotor defects caused by TDP-43 in muscles are not rescued by pioglitazone

Next, we sought to determine whether pioglitazone may also exert beneficial effects on muscle-dependent locomotor phenotypes caused by TDP^WT or TDP^G298S expression in larvae using the muscle-specific Gal4 driver, BG487 (45). In contrast to our results in motor neurons and glia, we found that 1 μm pioglitazone did not rescue TDP-43-dependent larval turning defects when expressed in larval muscles (Fig. 2B). Because BG487-Gal4 drives expression only in larval body wall muscles 6/7 in an anterior to posterior gradient, we also tested the effect of pioglitazone in the context of mhc-Gal4, which drives expression strongly in all larval muscles and obtained similar results (data not shown). To determine whether higher concentrations of drug may be required to elicit protection in muscles, we performed turning assays on larvae expressing TDP-43 (wild type or mutant) using the muscle driver BG487 and fed 5–25 μm pioglitazone containing food. These results show that TDP-43-dependent locomotor defects caused by muscle-specific expression cannot be rescued by pioglitazone, neither at 1 μm, the optimum concentration for motor neuron or glial toxicity, nor at higher doses ranging from 5–25 μm (Supplementary Material, Fig. S1A and B, respectively). Although we cannot exclude the possibility that even higher or perhaps lower doses of pioglitazone may be required to mitigate muscle toxicity, our data suggest that the protective effects of pioglitazone are likely restricted to the nervous system.

FUS-dependent toxicity in motor neurons is partially mitigated by pioglitazone

Having established that pioglitazone rescues several aspects of TDP-43-mediated toxicity in motor neurons and glia, we next asked whether it could also rescue toxicity in a Drosophila model of ALS based on human FUS (46), another RNA-binding protein linked to ALS that shares some functional aspects with TDP-43 (47). As shown in Figure 3A, expression of both human FUS^WT and disease-associated mutant FUS^P525L in motor neurons results in impaired larval locomotion, which is rescued by 1 μm pioglitazone. As is the case with TDP-43, administration of pioglitazone does not alter FUS protein levels (Fig. 3E and F). These data indicate that the beneficial effect of pioglitazone extends to FUS-dependent toxicity in motor neurons at least in regards to locomotor dysfunction.

Figure 3.

Pioglitazone partially improves toxicity caused by FUS, but not SOD1 expression in motor neurons. (A and B) Larval locomotor defects caused by FUS, but not SOD1 expression in motor neurons, are improved by pioglitazone. Human FUS (A) and SOD1 (B), wild type or mutant (as indicated), are expressed in motor neurons using D42-Gal4 driver. Note that 1 μm pioglitazone only improves FUS-dependent larval locomotor defects, but not that caused by SOD1. Error bars indicate SEM. Student's T test is used to assess statistical significance. ** and ***indicate P< 0.01 and P< 0.001, respectively. (C and D) Pioglitazone does not improve survival of D42>FUS flies. 1 μm pioglitazone is administered during development (C) or to adults (D) as indicated. Note that Kaplan–Meier survival analysis shows that developmentally administered pioglitazone results in significantly reduced lifespan for FUS^WT flies (C), while not affecting the lifespan under any other condition. *indicates P< 0.05. (E and F) Western blot shows no change in FUS protein levels with or without pioglitazone. Larvae raised on fly food containing DMSO or 1 μm pioglitazone are crushed in Laemmli buffer and subjected to western blotting and probed for FUS or tubulin for loading control (E). Results show no change in protein levels with or without pioglitazone (F). Error bars indicate SEM. Student's T test is used to assess statistical significance.

We also tested whether pioglitazone can improve lifespan in adults expressing FUS in motor neurons. As with Repo>TDP-43 experiments (Fig. 2C), developmental feeding of pioglitazone resulted in a significantly shorter lifespan for FUS^WT flies, but not FUS^P525L flies (Fig. 3C). No difference was observed between the two variants following adult feeding (Fig. 3D). Overall, these data indicate that although pioglitazone can rescue some aspects of the disease, it is not effective in improving lifespan in adult flies.

Pioglitazone is not protective in a Drosophila model of ALS based on SOD1

As pioglitazone was previously shown to be protective in a SOD1 mouse model of ALS (48,49), we next asked whether it could also mitigate locomotor defects in a Drosophila model (50). These experiments showed that in contrast to the results with TDP-43 and FUS models, pioglitazone does not rescue the larval turning phenotype caused by the expression of SOD1^G85R mutant in motor neurons (Fig. 3B). Interestingly, unlike TDP-43 and FUS, the expression of wild-type human SOD1 (SOD1^WT) in motor neurons does not result in a larval turning phenotype compared with w¹¹¹⁸ controls (Fig. 3B). These findings are in contrast to previous studies in mice where oral administration of pioglitazone to animals expressing SOD1^G93A showed both improved locomotion and increased lifespan (48,49). Although we used a different SOD1 mutation than in the mouse study, SOD1^G85R and SOD1^G93A mice share several phenotypic features including motor neuron degeneration, paralysis of fore- and hind-limbs and muscle atrophy, leading to rapid progression of the disease (51,52). Taken together these results indicate that pioglitazone rescues larval locomotion defects caused by motor neuron expression of TDP-43 and FUS, but not SOD1, and suggest that distinct molecular mechanisms may underlie different sub-types of motor neuron disease.

PPARγ acts as the molecular target of pioglitazone in vivo, in Drosophila

In humans, pioglitazone activates the nuclear receptor PPARγ, resulting in transcriptional modulation of factors that reduce insulin resistance (53). In the nervous system, PPARγ exerts neuroprotection by reducing inflammation (39,54). Recently, pioglitazone was shown to directly bind mitochondrial proteins suggesting that it may act on additional targets in vivo (55,56). Although Drosophila PPARγ has not been well characterized, there are two predicted homologs, namely E75 and E78 that belong to the nuclear receptor superfamily and are most similar in sequence to the human REV-ERBA receptor, a member of the same subfamily as nuclear PPARs (57). To determine whether pioglitazone exerts its neuroprotective activities in Drosophila by activating PPARγ, we took a pharmacogenetic approach. Indeed, if PPARγ is required for pioglitazone's action, reducing the receptor's expression in the context of TDP-43 should render motor neurons insensitive to the beneficial effects of the drug. To test this, we used a loss of function allele for E75 (Eip75B^Δ51) in the context of TDP^WT or TDP^G298S and found that when E75 expression was reduced by 50%, pioglitazone could no longer rescue TDP-43-dependent larval turning defects (Fig. 4A). Similar results were obtained when E78 was knocked down by RNAi in the context of TDP-43 (Fig. 4B). Knock-down was confirmed by semi-quantitative PCR that showed a reduction of ∼40% in transcript levels compared with controls (Supplementary Material, Fig. S2). Notably, whole larvae were used for PCR whereas the knock-down is targeted using the D42-Gal4 driver, thus it is likely that the actual reduction in E78 expression in motor neurons is greater. These data indicate that PPARγ is the primary in vivo target of pioglitazone in the fly and are consistent with the notion that activating PPARγ mitigates aspects of TDP-43-dependent toxicity. Somewhat surprising is the fact that genetic interaction between TDP-43 and E75/E78 was not significant. Larval turning assays showed a non-significant trend toward worsening of the TDP^WT phenotype by E75 loss of function, whereas no change was observed with E78 RNAi. For TDP^G298S, we observed a significant worsening of the larval turning defect with E78 RNAi whereas no change was found with E75 loss of function. These results suggest that E75 and E78 may need to be knocked down simultaneously to detect a significant effect on TDP-43 toxicity (data not shown).

Figure 4.

E75 and E78, the Drosophila PPARγ homologs, are targets of pioglitazone. Flies expressing TDP^WT or TDP^G298S were crossed with E75 LOF (A) or E78 RNAi (B) virgin females on fly food containing 1 μm pioglitazone or DMSO (as indicated). Larval turning assay was done on larvae expressing TDP-43 or E75 LOF/E78 RNAi alone or together. Pioglitazone rescued TDP-dependent neurotoxicity, which was abolished in the context of E75 LOF and E78 RNAi. Error bars indicate SEM. Student's T test was used to assess statistical significance. *** and *indicate P< 0.001 and P< 0.05, respectively. NS, not significant.

Pioglitazone restores a subset of metabolites dysregulated in the context of TDP-43 proteinopathy

To gain insight into the mechanism by which pioglitazone is neuroprotective, we took a global metabolomic approach using larvae expressing TDP-43 in motor neurons (D42>TDP-43) and D42>w¹¹¹⁸ controls raised on either 1 μm pioglitazone, DMSO or regular food. To determine specific metabolites that were restored by pioglitazone in the context of TDP^WT, we first compared the metabolic profiles of D42>TDP^WT larvae raised on drug food versus DMSO and found 111 alterations among a total of 572 metabolites detected in our samples (see Supplemental Material, Table S1). Next, we asked which of the 111 pioglitazone-specific changes were also caused by TDP^WT expression as determined by comparing the metabolic profiles of TDP-43-expressing larvae and D42>w¹¹¹⁸ larvae raised on regular food. This led to the identification of 28 metabolites altered by both pioglitazone and TDP^WT (Table 2). Of these, 14 metabolites were altered in opposite direction compared with TDP^WT, consistent with a restoration by pioglitazone whereas 8 others appeared to be similarly affected by pioglitazone and TDP^WT suggesting that pioglitazone does not further affect these biochemicals. In addition, we found six metabolites further altered by pioglitazone in the same direction as that of TDP-43, consistent with a potential worsening effect. Among the 28 metabolites, some were also affected by DMSO alone as determined by comparing the metabolic profiles of D42>TDP^WT larvae raised on DMSO and D42>TDP^WT larvae raised on regular food. The fact that DMSO itself can mitigate or further enhance TDP-43-dependent effects made some interpretations difficult as we cannot be certain whether pioglitazone restorations act on TDP-43 or DMSO effects (see Table 2).

Table 2.

The effect of pioglitazone on TDP^WT-specific metabolic alterations

The sole significantly restored metabolite was N-acetylglutamine, which was increased in D42>TDP^WT larvae versus D42>w¹¹¹⁸ on regular food (1.26, P< 0.05) and reduced by pioglitazone (in D42>TDP^WT larvae on pioglitazone versus DMSO, 0.74, P< 0.05, Fig. 5A and Table 2). There was a trend toward higher N-acetylglutamine on DMSO compared with regular food (1.27, P = 0.06); however, because pioglitazone significantly reduced its levels, we considered the effect of DMSO to be negligible. Increased N-acetylglutamine, a stable form of glutamine, suggests alterations in glutamate metabolism, consistent with reports of glutamate excitotoxicity in ALS (58). Pioglitazone also restored the levels of phenylalanylarginine (PAA), a dipepetide found to be elevated in D42>TDP^WT larvae compared with D42>w¹¹¹⁸ on regular food; however, DMSO had a significant effect on this metabolite. Thus, we could not definitively ascertain rescue in this case (see Table 2). It is interesting to note though that increased levels of PAA suggest defects in protein turnover, which are consistent with alterations in protein clearance known to accompany TDP-43 proteinopathy (59). Several metabolites showed trends toward restoration but did not reach statistical significance. Among these, pyruvate, the end product of glycolysis and a key metabolite at the intersection between several metabolic pathways was significantly increased in both D42>TDP^WT and D42>TDP^G298S larvae (1.67 and 1.6, compared with D42>w¹¹¹⁸ on regular food, respectively) and was slightly but not significantly reduced in D42>TDP^WT raised on pioglitazone compared with DMSO (0.68, see Fig. 5B and Table 2).

Figure 5.

Metabolites altered by pioglitazone in the context of TDP^WT. N-Acetylglutamine (A), pyruvate (B), 4-hydroxybutyrate (C) and ALCAR (D) are shown as examples of restoration, restoration trend, worsening and no change by pioglitazone, respectively (see also Table 2). Genotypes and experimental conditions, as indicated. Dashed lines indicate trends with P> 0.05 and P< 0.1, as shown. * and **indicate P< 0.05 and P< 0.01, respectively. DMSO, vehicle control; PGZ, pioglitazone containing food.

Interestingly, saccharopine, an intermediate in lysine metabolism, whereas it trends low in D42>TDP^WT larvae (0.73 compared with D42>w¹¹¹⁸ on regular food, P = 0.08), it is significantly reduced on pioglitazone (0.67 compared with DMSO, Table 2). Similarly, 4-hydroxybutyrate (GHB), a ketone body is significantly reduced by pioglitazone (0.41, P = 0.04) from trending low in TDP^WT (0.54, P = 0.08, Fig. 5C and Table 2). Interestingly, SOD1 mice on ketogenic diet exhibit slower disease progression, which may be attributed to the ability of ketone bodies to generate energy (60). These data suggest that some metabolites worsen in the context of pioglitazone, which may explain the deleterious effect of pioglitazone on lifespan. Additional metabolites of interest include acetylcarnitine (ALCAR), a precursor to carnitine, which is significantly reduced in the context of TDP-43 (A. Joardar and D.C. Zarnescu, unpublished data) and remains low in the context of pioglitazone (Fig. 5D and Table 2). Given carnitine's role in transporting fatty acids into the mitochondria for breakdown, these data suggest a TDP-43-dependent decrease in lipid beta-oxidation. Notably, while there were some trends toward restoration in D42>TDP^G298S larvae, none of the metabolite rescues were statistically significant (see Table 3 and Supplemental Material, Table S2). These findings are consistent with distinct mechanisms underlying TDP^WT versus mutant TDP^G298S toxicity at the cellular level (see model, Fig. 6). Overall, the vast majority of alterations in amino acid or lipid metabolism was not restored by pioglitazone or, was, in some cases worsened, consistent with the drug's ability to rescue some but not all TDP-43-dependent phenotypes in Drosophila.

Table 3.

The effect of pioglitazone on TDP^G298S-specific metabolic alterations

Figure 6.

Proposed model for the neuroprotective effect of pioglitazone. Pioglitazone (or TZDs, in general) activates the nuclear receptor PPARγ homolog E75/E78 in vivo in Drosophila. This neuroprotection is imparted to TDP-43 and FUS, but not SOD1 fly models of ALS. Furthermore, the improvement is tissue specific. In our model, some but not all aspects of cellular metabolism are restored by pioglitazone whereas others are worsened. Additionally, TDP^WT and TDP^G298S exhibit distinct metabolic alterations, implying differential mechanisms of disease pathophysiology.

Discussion

ALS is the third most common form of neurodegeneration following Alzheimer's and Parkinson's diseases (61). Although Riluzole is approved for ALS patients, its benefits are marginal, and at this time, there are no known effective treatments for the disease. There have been several efforts to design therapeutics using the SOD1 mouse, the most commonly used animal model of ALS. However, despite promising preclinical results, these candidate drugs have been disappointing in humans (62,63). To address this significant issue, efforts are being made to develop other animal models of ALS that can be used not only to identify phenotypes and ‘early biomarkers’ of the disease but also will be useful in drug screens for therapeutic purposes. We have previously generated a Drosophila model of ALS based on TDP-43, which recapitulates several aspects of the human disease including locomotor dysfunction and reduced lifespan (33,35). Here, using this model, we show that the antidiabetic drug pioglitazone acts as a neuroprotectant for aspects of TDP-43 proteinopathy by activating the putative Drosophila PPARγ homologs E75 and E78. We also show that pioglitazone mitigates FUS but not SOD1-dependent toxicity in Drosophila, consistent with previous published work showing that distinct mechanisms are likely at work in the context of these different models of ALS (64). Interestingly, pioglitazone did not improve, and in some cases worsened, the lifespan of TDP-43-expressing flies, when administered either during development, or after ‘disease onset’, which is consistent with results from recent clinical trials (42,65). This apparent disconnect is consistent with the effects of pioglitazone on cellular metabolism. As described earlier, while pioglitazone treatment restored some metabolites altered owing to TDP-43 overexpression in motor neurons, others were unchanged or even worsened. This provides a potential explanation for why some phenotypes but not others are rescued by pioglitazone. Aside from the possibility that different drug concentrations may be needed, it remains unclear why pioglitazone is protective in mouse but not fly SOD1 models and, in retrospect, given the similarities between the effect of pioglitazone in Drosophila models of ALS and humans, the fly appears to be a more accurate predictor of clinical trial outcomes.

It is tempting to speculate that the predictive power of the Drosophila model may lie in the tools that enable motor neuronal versus glial versus muscle-specific expression of the toxic TDP-43 protein. Our results show that pioglitazone mitigates neuronal and glial TDP-43-dependent toxicity but has no effect on the locomotor dysfunction caused by muscle-specific expression of TDP-43. This type of knowledge is easily obtainable in the fly model and can provide helpful information about cell autonomous versus non-autonomous effects as well as the efficacy of candidate drugs in different tissues of interest. While it was shown that pioglitazone reduces inflammation in the glia, its effects in neurons or muscles have not been studied in the mouse prior to human trials (48). Our results indicate that the protective effects of pioglitazone are specific to the nervous system and were not observed in muscles, at least within the limits of our experimental conditions (i.e. tissue-specific levels of expression and drug concentration). These findings suggest that future preclinical studies may benefit from testing candidate therapies in multiple disease models in which tissue specificity and several phenotypic outcomes are easily ascertained.

Pioglitazone has been originally developed for the treatment of type 2 diabetes as PPARγ activation in the liver improves glucose metabolism systemically (66). In the nervous system, activation of the nuclear hormone receptor PPARγ has been shown to have anti-inflammatory and neuroprotective effects (39,54). In our model, we found that pioglitazone can restore a rather limited set of metabolites altered in a TDP-43-dependent manner (see Fig. 6 for model). We found evidence of altered glutamine/glutamate metabolism in TDP^WT flies, as displayed by elevated levels of N-acetylglutamine, which is restored by pioglitazone. Excessive levels of extracellular glutamate in the central nervous system cause hyperexcitability of neurons, ultimately leading to their death. The glutamate transporter GLT1/EAAT2 plays a major role in maintaining extracellular glutamate levels below the excitotoxic concentrations by efficiently transporting this metabolite. Interestingly, astrocytic GLT1/EAAT2 gene is a target of PPARγ, leading to neuroprotection by increasing glutamate uptake (67). Furthermore, pyruvate, which is significantly high in both TDP^WT and TDP^G298S, shows a trend toward reduction upon pioglitazone treatment for TDP^WT. Pyruvate is a central metabolite that lies at the junction of several intersecting cellular pathways including glucose and fatty acid metabolism. It is converted to oxaloacetate by the enzyme pyruvate carboxylase, which is a key step in lipogenesis. Interestingly, PPARγ, the target of pioglitazone, is a direct transcriptional modulator of the pyruvate carboxylase gene (68). Given the fact that ALS patients suffer from massive weight loss, this provides a possible explanation for the potential protective effects of pioglitazone through increased lipogenesis. Taken together, our metabolomics approach provides useful insights for understanding the molecular mechanisms underlying ALS pathophysiology. Interestingly, altered cellular metabolism has previously been implicated in ALS pathophysiology with patients exhibiting signs of hypermetabolism (28–32,69). Notably, our fly model also showed signs of hypermetabolism including an increase in pyruvate, a key metabolite linking glucose metabolism to the TCA cycle. Additionally, the ketone body GHB is reduced in the context of TDP^WT, consistent with a clinical study showing that a ketogenic diet slowed ALS disease progression (60). Given the similarities between the metabolic profile of the Drosophila model and human samples, it will be interesting in the future, to design therapeutic approaches aimed at restoring these common metabolic changes using nutritional supplementation.

In summary, our data show the potential of using the fly model of ALS as a rapid and efficacious system for drug screening in vivo. Our results using FUS and SOD1 fly models of ALS indicate that pioglitazone is effective in mitigating some, but not all forms of the disease, which suggests that stratification of patient populations should be considered in future clinical trials. The primary endpoint tested in prior clinical trials, namely lifespan, was not improved by pioglitazone, which is consistent with our data in Drosophila. Although the results from the clinical trials have not shown much promise in ALS patients, the use of pioglitazone as a tool to dissect molecular mechanisms of the disease remains attractive. Our metabolomic profiling with and without pioglitazone pinpoints pathways that could be targeted either by drugs or by diet modifications. Also, it is possible that chemical modifications of pioglitazone, which was optimized for adipose tissue, skeletal muscle and liver, are needed for increased efficacy in the nervous system. The protective effect of pioglitazone opens up avenues for designing small molecules with modifications around the basic structure of the drug, and testing their potential in vivo, in the fly model. Furthermore, clinical trials have not been stratified for TDP-43 pathology or mutations; thus, significant results may have been missed. Our developmental and adult feeding experiments clearly demonstrate that locomotor function is improved by pioglitazone suggesting that despite its lack of effect on lifespan, PPARγ remains a molecular target with therapeutic potential, perhaps in combination with other strategies based on restoring the metabolic alterations caused by TDP-43 in the nervous system.

Materials and Methods

Drosophila genetics

All Drosophila stocks and crosses were maintained on standard yeast/cornmeal/molasses food at 25°C. Transgenic flies expressing human TDP-43 variants with C-terminal YFP tags were generated as previously described (33,35). GMR Gal4, D42-Gal4, Repo-Gal4, BG487-Gal4 or mhc-Gal4 were used to drive expression in different tissues using the GAL4-UAS system (70) and were obtained from the Bloomington Stock Center together with Eip75B^Δ51/TM6B, P{Ubi-GFP.S65T}PAD2, Tb, y¹ v¹; P{TRiP.JF02258}attP2, w¹¹¹⁸; P{UAS-hSOD1.G85R}2a P{UAS-hSOD1.G85R}2b; P{UAS-hSOD1.G85R}3a P{UAS-hSOD1.G85R}3b, and w¹¹¹⁸; P{UAS-hSOD1}16.2. w¹¹¹⁸; UAS-FUS^WT and w¹¹¹⁸; UAS-FUS^P525L transgenics were gifts from Brian McCabe. For genetic controls, w¹¹¹⁸ was crossed with the appropriate Gal4 driver.

Drug screen

UAS TDP-43 males were crossed with D42-Gal4 female virgins on fly food containing either DMSO or pioglitazone. For DMSO controls, the same volume of DMSO as the corresponding drug concentration was added. Bromophenol blue was added to a final concentration of ∼0.02% to ensure homogeneity. Crosses were made on drug food with three female virgins and two males in each vial and were maintained at 25°C unless noted. The parents were discarded after 5–7 days and then the vials were screened for adult progeny with straight wings from Day 14 to Day 25. All adults were screened for TDP-43 expression by visualizing the YFP tag. Total number of pupae was counted on Day 25. Percent survival was calculated using the formula (total number of straight-winged adults/total number of pupae) × 100. All experiments were performed in triplicate.

Larval turning assay

Assays were performed as previously described (33,35). Briefly, wandering third instar larvae were placed on room temperature grape juice plates, allowed to acclimate then gently turned ventral side up. The time required for larvae to flip back to dorsal side up and start moving forward was noted as the larval turning time. At least 30 larvae per genotype and condition were tested.

Lifespan analysis with pioglitazone

For developmental feeding experiments, crosses were made on drug food at 25°C. Newly eclosed flies were separated by sex and housed in separate vials with no >4 flies per vial at 25°C. For adult feeding experiments, crosses were made on regular food at 22°C. Immediately after eclosion, flies were separated by sex and placed in fresh vials with drug food at 25°C. Survival plots were generated using the survival and Hmisc packages in R (R Development Core Team, 2013) and Rstudio (R Studio, Inc., Boston, MA, USA) software. Statistical analysis was done using the log-rank test in R.

Metabolomics

D42-Gal4 virgin females were crossed with UAS TDP-43 YFP or w¹¹¹⁸ males and raised on regular food (RF), DMSO or 1 μm pioglitazone (PGZ) containing food. 50–60 third instar larvae (∼50–60 mg) per sample were collected and flash-frozen in liquid nitrogen. Metabolomic and statistical analyses were conducted at Metabolon as described previously (71). Briefly, fly samples (N = 5 for each genotype/treatment) were subjected to methanol extraction and then split into aliquots for analysis by ultrahigh performance liquid chromatography/mass spectrometry (UHPLC/MS) in the positive, negative or polar ion mode and by gas chromatography/mass spectrometry (GC/MS). Metabolites were identified by automated comparison of ion features to a reference library of chemical standard followed by visual inspection for quality control. Peaks were quantified using area-under-the-curve. Following normalization to Bradford protein concentration data were log transformed. Missing values for a given metabolite were imputed with the minimum observed value for each compound based on the assumption that they were below the limit of instrument detection sensitivity. For studies spanning multiple days, a data normalization step was performed to correct variation resulting from instrument inter-day tuning differences.

Western blotting

Western blotting was done following standard protocol. Briefly, protein samples (whole larvae crushed in 2× Laemmli Buffer) were resolved on 4–20% SDS–PAGE gradient precast gels (BioRad) and then transferred to a nitrocellulose membrane (Millipore). Following blocking in 5% fat-free milk in 1× TBST, the following primary antibodies were used for different experiments at 4°C overnight: rabbit anti-GFP (Invitrogen) at 1:3000, rabbit anti-FUS (Abcam) at 1:1000, and mouse anti-tubulin β clone KMX-1 (Millipore) at 1:1000. Secondary antibodies used were as follows: goat anti-rabbit at 1:10 000 or goat anti-mouse at 1:10 000. Proteins were detected using Odyssey and quantified using LICOR Image Studio Lite software. TDP43 and FUS levels were normalized to β-tubulin and represented as fold change. The data are an average of three independent experiments. Error bars indicate SEM.

Semi-quantitative PCR

Total RNA was prepared from fly heads (GMR Gal4>E78^RNAi) using RNeasy Kit (Qiagen) with on-column DNAse treatment. First strand cDNA synthesis was performed with a Superscript III cDNA synthesis Kit (Invitrogen). Semi-quantitative PCR reactions were performed by normalizing E78 to GAPDH levels. Primers used were as follows: CGAGGGTTGCAAGGGATTC (forward) and CACATTGAGGGAGGCAGGAG (reverse).

Statistical analyses

Student's T test was used for statistical analysis of larval turning experiments, western blots and semi-quantitative PCR, whereas log-rank test was used for survival analyses. For metabolomics statistical analyses and data display, any missing values were assumed to be below the limits of detection; these values were imputed with the compound minimum (minimum value imputation). To determine statistical significance, two-way ANOVA with post hoc contrasts (T-tests) was performed in ArrayStudio (Omicsoft) or ‘R’ to compare data between experimental groups; P < 0.05 was considered significant. An estimate of the false discovery rate (Q-value) was calculated to take into account the multiple comparisons that normally occur in metabolomic-based studies, with Q< 0.05 used as an indication of high confidence in a result (72).

Supplementary Material

Supplementary Material is available at HMG online.

Funding

This work was supported by the MDA B2I fellowship 4202380 to A.J., MDA 255293, and funds from the Himelic family foundation and Sandra Harsha's estate to D.C.Z. J.M. was in part supported by HHMI grant # 52006942 through the Undergraduate Biology Research Program at University of Arizona.