A genome-wide screen identifies mts and sgg as candidate modifiers of FUS-ALS in vivo

We performed a genome-wide screen to identify candidate modifiers (suppressors) of FUS toxicity in Drosophila (Fig. 1a). We previously showed that motor neuron-specific expression of wild-type or mutant human FUS leads to a severe eclosion phenotype [5]. We therefore generated a recombinant stock of UAS-FUS (R521G) and the D42-Gal4 motor neuron driver line. This stock was maintained over a TM6B-Gal80 balancer, which suppresses expression and is marked with a visible marker. We crossed this screening stock to flies from Bloomington Drosophila Stock Center (BDSC) deficiency kit, which consists of a selected set of molecularly defined genomic deletions. We focused on deletions on chromosomes X, 2 and 3, crossing 473 deficiency lines to the screening stock and assessing whether eclosion was rescued. When crossed to a control background, the D42-Gal4 > FUS(R521G) flies were not able to eclose, and therefore, any eclosion could be considered as a rescue. Using this approach, we identified 58 candidate deficiencies that attenuated the mutant FUS pupal lethality.

Fig. 1

mts and sgg are modifiers of FUS toxicity in vivo. a Schematic of the experimental set up of the genetic screen in the Drosophila FUS model. Candidate genetic modifiers are identified in an unbiased screen, using the BDSC deficiency kit, the Exelixis kit, as well as specific RNAi lines and null mutants against the genes of interest. A complimentary literature-based approach is also used. Eclosion of a fly from the pupal case is categorized as a rescue. In total, 24 genes are identified as candidate modifiers of FUS toxicity. b RNAi-mediated genetic knockdown of sgg rescues the FUS-induced fly eclosion phenotype. w + v- w1118 crossed to FUS serves as a control. (N = 10 crosses/condition) c RNAi-mediated genetic knockdown of mts rescues the FUS-induced fly eclosion phenotype. w1118 crossed to FUS serves as a control. (N = 10 crosses/condition). d Pharmacological inhibition of sgg by lithium chloride (LiCl) rescues the eclosion phenotype in FUS flies. D42-Gal4/+ serves as a control. (N = 5 crosses/condition). e Pharmacological inhibition of mts by okadaic acid (OA) rescues the eclosion phenotype in FUS flies. D42-Gal4/+ serves as a control. (N = 6 crosses/condition). Values in b, c, d, and e are represented as the mean ± SEM. Statistical comparisons between controls and RNAi conditions (b, c) were determined using one-way ANOVA with Sidak’s multiple comparisons test or Kruskal–Wallis test with Dunn’s multiple comparisons for each group. Statistical comparisons between controls and treated conditions (d, e) were determined using one-way ANOVA with Sidak’s multiple comparisons or Kruskal–Wallis test with Dunn’s multiple comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001, ns not significant

To identify which genes are mediating the rescue effect of the large deficiencies, we first screened with smaller deletions. In total, we screened 167 smaller deletions and found 24 that modified the pupal lethality. Finally, to identify the actual genes responsible for the modifying effects, we used RNAi lines from the ‘VIENNA Drosophila Research Center’ (VDRC) as well as lines carrying mutations in candidate genes which were likely to result in loss-of-function. We found 18 modifying genes out of 88 extensively tested RNAi lines that attenuated mutant FUS-induced pupal lethality (Table S1). Among these genes were mts, the Drosophila ortholog of human PPP2CA, and sgg, the Drosophila ortholog of human GSK3B.

As a complimentary approach to identify modifying genes, we searched the available literature and found 24 candidate genes which have previously been linked to FUS-associated ALS. Screening this list using RNAi lines and putative null alleles allowed us to identify six more genes involved in FUS-induced neurotoxicity, among them sgg (Table S1). As a consequence, we ended up with a total of 24 candidate modifiers, with mts being identified through the unbiased approach, and sgg through both approaches.

Genetic and pharmacological inhibition of mts and sgg rescue eclosion and lifespan in flies

Our approach yielded a total of 24 genes which modify FUS toxicity when their expression is reduced. To prioritize hits for follow-up, we investigated whether any of these genes fall into a common pathway. mts is the catalytic subunit of the PP2A phosphatase complex in Drosophila, while sgg is the ortholog of GSK3B and has recently been linked to ALS/FTD [58, 60]. PP2A has been proposed to directly reduce GSK3 inhibitory phosphorylation in human cells, and thus acts as an activator of GSK3 [35].

As mts and sgg were candidate modifiers of FUS-toxicity, we sought to confirm their modifying action in the FUS flies. We first backcrossed several RNAi lines against sgg and mts into a suitable control genetic background (see Materials and methods) for six generations to avoid genetic background effects. Wild-type (WT) FUS and mutant R521G FUS were expressed in fly MNs using the D42-Gal4 driver along with mts or sgg RNAi and eclosion was scored. Knockdown of sgg led to a rescue of the FUS fly eclosion phenotype for both WT and R521G FUS-expressing flies for all 4 RNAi lines tested (Fig. 1b). Knockdown of mts led to a rescue for eclosion in WT and R521G FUS-expressing flies for two out of three of the tested RNAi lines (Fig. 1c), while a third line produced a partial eclosion defect on its own, perhaps explaining its lack of a significant effect (Fig. S1). We next tested the consequence of feeding pharmacological inhibitors to the flies. For this, we used lithium chloride (LiCl), a well-known and widely used GSK3 inhibitor [9, 20], and okadaic acid (OA), a specific inhibitor of PP2A in human cells at concentrations of 1–10 nM [65, 66]. Addition of LiCl to the fly food led to a higher percentage of eclosion for flies expressing FUS in their MNs between a range of doses of 5–15 mM LiCl (Fig. 1d). LiCl only affected eclosion of driver-only (D42-Gal4/+) control flies above 20 mM (Fig. 1d), before becoming developmentally lethal at 100 mM (not shown), suggesting that it is well tolerated by the model. OA similarly rescued the eclosion phenotype of FUS flies, with the best rescue occurring between approximately 10 and 25 nM in fly food (Fig. 1e). Toxicity of OA was only observed above 50 nM (Fig. 1e).

To further validate the modifying ability of mts and sgg, we tested the effect of their knockdown on the lifespan of FUS-expressing flies. We expressed FUS WT or R521G in adult MNs using the D42-Gal4 driver, but to avoid developmental lethality, we included a temperature sensitive Gal80 allele (Gal80-ts), and allowed the flies to develop at 18˚C where FUS expression was suppressed, before moving adults to 25˚C to induce expression. We have previously shown that expression of either WT or mutant R521G FUS in this manner significantly reduces the lifespan of the flies compared to healthy controls, leading to death with a median lifespan of approximately 3 weeks [5]. We found that RNAi-mediated knockdown of mts or sgg led to an extension of the FUS fly lifespan with all of the RNAi lines tested (Fig. 2a, b and Fig. S2). Importantly, knockdown of sgg and mts did not alter the levels of FUS protein in the heads of the flies used for lifespan (Fig. 2c, d), suggesting that they exert their effects independently of FUS expression or stability.

Fig. 2

Inhibition of sgg or mts extend the lifespan of FUS flies. a RNAi-mediated genetic knockdown of sgg extends the shortened lifespan of FUS WT and R521G Drosophila at 25 ℃. w + v- w1118 crossed to FUS serves as control (log-rank test, see TableS2 for statistical information). b RNAi-mediated genetic knockdown of mts extends the shortened lifespan of FUS WT and R521G Drosophila at 25 ℃. w1118 crossed to FUS serves as control (log-rank test, see TableS2 for statistical information). c, d Western blotting and quantification demonstrate that FUS expression remains unaltered in Drosophila heads after RNAi-mediated knockdown of sgg (c) or mts (d) (N = 3, mean ± SEM, one-way ANOVA). e A heterozygous null mutation in sgg (sgg1) leads to a pronounced extension of the shortened lifespan of FUS WT and R521G flies at 25 ℃. w1118 crossed to FUS serves as control (log-rank test, see TableS3 for statistical information). f A heterozygous null mutation in mts (mtsXE2258) leads to a pronounced extension of the shortened lifespan of FUS WT and R521G flies at 25 ℃. w1118 crossed to FUS serves as control (log-rank test, see TableS3 for statistical information). g, h Western blotting and quantification show that FUS expression remains unaltered in FUS WT and R521G Drosophila heads after heterozygous knockout of sgg (g) or mts (h) (N = 4, mean ± SEM, one-way ANOVA). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001, ns not significant

As an alternative approach to investigate whether reduction of mts or sgg would lead to a more pronounced extension of the lifespan, we used null mutant lines of the two genes. As homozygous loss-of-function of these genes is developmentally lethal, we introduced a heterozygous loss-of-function. We observed a pronounced lifespan extension after heterozygous loss-of-function of sgg and mts, confirming the findings using RNAi (Fig. 2e, f). We observed an overall unaltered abundance of FUS protein after sgg and mts heterozygous knockout (Fig. 2g, h), and we additionally confirmed an approximate 50% reduction of sgg protein levels in the heterozygous mutant flies (Fig. S3). As antibodies against mts are not available, it was not possible to also confirm the level of mts protein in the mts null mutant line.

Altogether, our results show that genetic and pharmacological inhibition of either mts or sgg can rescue FUS-induced developmental neurotoxicity. Moreover, genetic reduction of mts or sgg activity in adult flies can extend lifespan. These two modifying genes appear to act independently of FUS expression. We next sought to determine whether inhibition of the human orthologs of these genes can rescue toxicity in a human cellular model.

Pharmacological inhibition of PP2A and GSK3 rescue hallmark ALS-associated phenotypes in iPSC-derived sMNs

To further confirm the modifying capacity of PP2A and GSK3, we investigated whether their inhibition could rescue hallmark FUS-ALS phenotypes. Cytoplasmic mislocalization of FUS is a pathological hallmark of ALS that has been linked with protein toxicity and neuronal death [17, 36, 73]. Therefore, we decided to investigate whether pharmacological inhibition of PP2A and GSK3 could rescue this phenotype. To assess the modifying effect of PP2A and GSK3 on mislocalization, we used a well-established iPSC line with a de novo point mutation (P525L) in FUS from a 17-year-old ALS patient [29, 61]. This patient line was compared with its corresponding CRISPR-Cas9 gene-edited isogenic P525P control [29, 61]. In P525P FUS isogenic controls, FUS was mostly localized in the nucleus (Fig. 3a). In P525L FUS mutant iPSC-derived sMNs, we observed cytoplasmic mislocalization of FUS, which was rescued after a 48 h treatment with 1 mM LiCl (Fig. 3b). Given that LiCl can produce off-target effects [20], we also validated our findings using a highly selective GSK3 inhibitor: tideglusib [38, 42, 53, 55]. Tideglusib (TD) is a non-ATP competitive inhibitor of GSK3, which was well tolerated in phase 2 clinical trials for progressive supranuclear palsy (PSP) [67] and is being investigated in clinical trials for treating GSK3 hyperactivity in Alzheimer’s disease [38, 55]. Recently, it has also been suggested as a potential treatment for ALS [42, 58,59,60]. Excitingly, a 48 h treatment of the P525L FUS sMNs with 15 μM tideglusib also led to a significant rescue of FUS mislocalization (Fig. 3b), confirming that the phenotype alleviation is not due to aspecific effects of LiCl, but thanks to GSK3 inhibition. To inhibit PP2A, we treated the FUS MNs with 1 nM okadaic acid (OA) for 72 h, resulting in a significant rescue of FUS mislocalization (Fig. 3b).

Fig. 3

PP2A and GSK3 pharmacological inhibition rescue FUS cytoplasmic mislocalization in patient iPSC-derived motor neurons. a. Representative confocal images showing FUS distribution in patient iPSC-derived motor neurons with the P525L mutation, as well as in P525P isogenic controls after treatment with lithium chloride (LiCl, 1 mM 48 h), tideglusib (TD, 15 μM 48 h) or okadaic acid (OA, 1 nM 72 h). Scale bars 10 μm. b Quantification of nuclear/cytoplasmic ratios (N/C ratio) fluorescent intensity of FUS in motor neurons shows a rescue of FUS mislocalization after treatments. Each dot represents one analyzed cell. Three different colors indicate data combined from three independent differentiations (70–80 cells/condition). Data are shown as Grand mean; Kruskal–Wallis with Dunn’s multiple comparisons test. ****p < 0.0001, ns not significant

We next determined the effect of pharmacologically inhibiting PP2A and GSK3 on NMJ formation. For this, we used a human-derived co-culture system that is well established in our lab, combining iPSC-derived sMNs and myotubes in microfluidic devices, which allow us to study a functional human NMJ in a compartmentalized system [61, 62]. Mutant FUS-associated NMJ impairment is a phenotype well characterized in this model, as we observe a lower number of NMJs per myotube in the P525L mutant when compared to the P525P isogenic control [61]. On day18 of differentiation, we treated our motor neuron—myotube co-culture system with LiCl (1 mM 48 h), TD (15 μM 48 h), or OA (1 nM 72 h) in both the motor neuron and myotube compartments. On day 28, we fixed the cells and assessed the number of NMJs by performing immunocytochemistry against NMJ markers. Using confocal microscopy, we observed that all three treatments had a positive effect on NMJ formation in the P525L mutant co-cultures, demonstrating the strong modifying capacity of GSK3 and PP2A (Fig. 4 and Fig. S4).

Fig. 4

PP2A and GSK3 inhibition improves ALS-associated NMJ impairments. a Representative confocal micrographs of NMJs formed by FUS-P525L cells with LiCl, TD, and OA treatments. NMJs are impaired in FUS-P525L ALS, and inhibition of PP2A or GSK3 improves the phenotype (see Fig. S4 for additional images). Scale bars 10 μm. b Quantification of NMJ-like structures for all conditions, based on the neurite/presynaptic marker morphology and/or based on Btx (red)-SYP/NEFH (green) co-localization per myotube. Each dot represents one analyzed myotube. Three different colors indicate data combined from three independent differentiations (70–80 myotubes/condition). Data are shown as Grand mean; N = 3 biological replicates; Kruskal–Wallis test with Dunn’s multiple comparisons test. *p < 0.05; ****p < 0.0001, ns not significant

Due to the polarization and length of sMNs, a proper regulation of axonal transport is essential for their function [44]. Defects in axonal transport are considered an early event in ALS pathogenesis, preceding axon retraction and denervation of the muscle [23, 41]. We investigated whether PP2A or GSK3 inhibition could rescue mitochondrial transport deficits in our P525L FUS iPSC-derived sMNs. We performed live cell imaging of mitochondria at day 30 of differentiation. We quantified the total number of mitochondria per 100 μm of neurite, as well as the total number of stationary and moving mitochondria (Fig. S5), allowing us to calculate the percentage of moving mitochondria. Compared to P525P FUS isogenic controls, the percentage of moving mitochondria was significantly lower in P525L sMNs (Fig. 5). After treating the cells with LiCl for 48 h, to inhibit GSK3, tracking analysis showed a clear rescue of mitochondrial transport defects (Fig. 5a, d). Moreover, a similar rescue of mitochondrial transport was observed after treatment with tideglusib (Fig. 5b, e), confirming the modifying capacity of GSK3 inhibition. In addition, OA treatment for 72 h similarly led to a significant improvement of mitochondrial transport (Fig. 5c, f). Although we occasionally observed small differences in the total number of mitochondria per neurite, the drug treatments consistently reduced the number of stationary mitochondria and increased the number of motile mitochondria (Fig. S5).

Fig. 5

Pharmacological inhibition of PP2A and GSK3 rescue mitochondrial transport deficits. a–c Example kymographs (time-distance plots) of mitochondria (MitoTracker Green) after treatment of 30-day-old P525P isogenic control motor neurons and P525L mutant motor neurons with 1 mM lithium chloride (LiCl) for 48 h (a), 15 μM tideglusib (TD) for 48 h (b), and 1 nM okadaic acid (OA) for 72 h (c). Mock conditions were untreated. Scale bars 30 μm d–f. Percentage of mitochondria that are motile in the motor neurons (day 30) comparing isogenic control and mutant with or without LiCl (d), TD (e), and OA (f). Each dot represents one analyzed neurite. Three different colors indicate data combined from three independent differentiations (70–80 neurites/condition). OA and TD treatments were performed at the same time, so the mock data are the same for both conditions. Data shown as Grand Mean; N = 3 differentiations; Kruskal–Wallis with Dunn’s multiple comparisons test. ****p < 0.0001, ns not significant

The IC50 of OA for PP2A is 0.07–1 nM, which is the range of concentrations that we have used to treat the cells (1 nM). However, OA may also inhibit PP1 in the nanomolar range (IC50 3.4 nM). To confirm our results, we tested a second inhibitor of PP2A, LB-100, a recently developed molecule that is being studied in cancer clinical trials [12, 32], and which is part of a family of compounds with stronger specificity for PP2A compared to PP1 (IC50 PP2A = 0.4 μM, IC50 PP1 = 80 μM) [39]. Addition of LB-100 to the fly food at concentrations greater than 5 μM rescued the eclosion of the FUS flies and had no apparent toxic effects on healthy controls (Fig. S6). In addition, we tested LB-100 in iPSC-derived sMNs and we observed a significant rescue of mitochondrial movement after treating the cells with 1 μM LB-100 (Fig. S6). Altogether, our results demonstrate that PP2A and GSK3 are modifiers of FUS-induced neurotoxicity both in Drosophila and human cellular models.

GSK3 is hyperactive in FUS-ALS due to reduced inhibitory phosphorylation

Recently, reduced GSK3 inhibitory phosphorylation was observed in a FUS mouse model [60], suggesting that GSK3 may become hyperactive in response to a dysfunctional FUS protein. We independently confirmed that reduced GSK3 phosphorylation occurs in the spinal cord of this mouse model at symptomatic stages (Fig. S7). As inhibiting GSK3 was beneficial in our models, we wondered whether this GSK3 hyperactivity was conserved. To determine whether GSK3 is hyperactive in Drosophila, we used nSyb-Gal4 combined with Gal80-ts to drive FUS expression pan-neuronally in adult flies for 7 days, and assessed sgg expression and phosphorylation by western blotting (Fig. 6). Consistent with previous reports [81], we observed two major bands of sgg protein in Drosophila head lysates corresponding to isoforms SGG10 and SGG39 of the protein. Upon WT and R521G FUS expression, we noticed a significant reduction in the levels of phospho-sgg, while the levels of total-sgg protein remained unaltered (Fig. 6a, b). Therefore, the reduced inhibitory phosphorylation suggests that GSK3 is hyperactive in our FUS flies. To translate our Drosophila and mouse findings to human cells, we performed western blotting for phospho-GSK3α/β at residue serine 21/9 in our patient iPSC-derived FUS sMNs on day 30 of differentiation. We found that phospho-GSK3α/β was significantly reduced in mutant P525L FUS sMNs compared to the P525P isogenic controls (Fig. 6c, d), demonstrating that GSK3 inhibitory phosphorylation is also reduced in FUS patient sMNs and suggesting that GSK3 is hyperactive in this model as well.

Fig. 6

GSK3 is hyperactive in FUS-ALS due to reduced inhibitory phosphorylation. a Western blot showing phospho-sgg and total-sgg protein in control vs FUS WT and R521G flies. Control was w1118 crossed to Gal80-ts; nSyb-Gal4. In Drosophila, the two splice isoforms of GSK3-β, SGG39 (upper band) and SGG10 (lower band), are seen. Beta-actin serves as a loading control. b Quantification of panel a shows that phosphorylation of SGG is reduced in FUS WT and R521G Drosophila, for the total-SGG protein and for the individual SGG39 and SGG10 isoforms, while the levels of SGG protein remain unaltered overall (N = 3, mean ± SEM, one-way ANOVA with Sidak’s multiple comparisons test). c Western blot for GSK3α/β pSer21/9 in iPSC-derived motor neurons from FUS-P525L patients and in their CRISPR-corrected P525P isogenic control. Ser21/9 inhibitory phosphorylation of GSK3α/β appears reduced in FUS P525L patient motor neurons compared to isogenic controls, while the total levels of GSK3α/β remain unaltered among the conditions. The numbers 1, 2, and 3 represent three independent differentiations. Alpha-tubulin serves as a loading control. d Quantification of the western blot of panel c shows reduced Ser21 and Ser9 inhibitory phosphorylation for GSK3 α and β, respectively (N = 3, mean ± SEM, two-tailed paired t test) *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

We wondered whether PP2A may also be affected in FUS-ALS models. Given the availability of a suitable mammalian antibody, we performed western blotting to assess the level of the PP2A catalytic subunit (PP2A-C) in the FUS mouse and patient iPSC-sMNs. Excitingly, we found that the level of PP2A-C was increased in both mouse tissue and in the iPSC-sMNs (Fig. S7c, d and Fig. S8).

PP2A acts upstream of GSK3 affecting its phosphorylation, and GSK3 hyperactivity alone is insufficient to drive toxicity

PP2A has been suggested to modify the phosphorylation of GSK3, affecting its activity [35]. To assess a possible PP2A-GSK3 interaction in our model, we first tested whether PP2A can regulate GSK3 phosphorylation in Drosophila. To determine whether mts controls sgg phosphorylation, we overexpressed mts using the pan-neuronal nSyb-Gal4 driver. After aging the progeny for 7 days, we assessed GSK3 phosphorylation using western blotting. We found that mts overexpression led to significantly decreased levels of inhibitory phosphorylation of both major sgg isoforms, suggesting that PP2A acts upstream of GSK3 in flies and causes its dephosphorylation (Fig. 7a, b). To determine whether GSK3 hyperactivity is sufficient to lead to neurodegeneration, we overexpressed either wild-type sgg (UAS-sgg) or a constitutively active mutant in which the serine 9 phosphorylation site had been replaced with alanine (UAS-sgg S9 → A). Overexpression of these sgg variants in fly MNs, using the D42-Gal4 driver, caused an eclosion phenotype, milder than the one observed after FUS expression, suggesting that GSK3 hyperactivity alone can drive toxicity but may not be fully responsible for the strong eclosion defect that we observed in FUS-expressing flies (Fig. 7c). Recapitulating the rescue that we previously saw, addition of LiCl in the food rescued the mild eclosion defect induced by sgg overexpression, increasing eclosion rates from ~60 to ~85% (Fig. 7c). On the contrary, LiCl had no effect on the eclosion of the flies expressing the constitutively active form of sgg, suggesting that lithium inhibits sgg via increasing its phosphorylation at Serine 9 (Fig. 7c). To determine whether PP2A inhibition could rescue sgg hyperactivity, we added OA to the food of sgg-overexpressing flies. OA significantly increased eclosion of UAS-sgg flies, but had no effect on the eclosion of the UAS-sggS9 → A flies (Fig. 7d). These data indicate that PP2A and GSK3 act in a common pathway, with PP2A lying upstream of GSK3. Even more importantly, we conclude that PP2A can alter GSK3 inhibitory phosphorylation on this specific Serine9 residue.

Fig. 7

PP2A acts upstream of GSK3, affecting its inhibitory phosphorylation. a Western blot for phospho-sgg and total-sgg protein in control vs. flies overexpressing mts pan-neuronally. Control was w1118 crossed to Gal80-ts; nSyb-Gal4. In Drosophila, the two splice isoforms of GSK3-β, SGG39 (upper band) and SGG10 (lower band), are seen. Beta-actin serves as loading control. b Quantification of panel a shows that phosphorylation of SGG is reduced after mts overexpression in the Drosophila brain, for the total-SGG protein and for the individual SGG39 and SGG10 isoforms, while the levels of SGG protein remain unaltered overall (N = 3, mean ± SEM, two-tailed paired t test). c Overexpression of sgg in the fly motor neurons causes a mild eclosion defect, with ~60% flies eclosing. LiCl in the fly food rescues the eclosion deficit to ~85 to 90%. Overexpression of a constitutively active form of sgg (S9 → A) leads to a mild eclosion defect (~75 to 80%) and LiCl has no effect on this deficit. Statistical comparisons between control (0) and treated conditions (2.5–20) were determined using one-way ANOVA with Sidak’s multiple comparisons (N = 6 vials/condition). d Overexpression of sgg in the fly motor neurons causes a mild eclosion defect. OA in the fly food rescues the eclosion deficit to ~90 to 95%. Overexpression of a constitutively active form of sgg (S9 → A) also leads to a mild eclosion defect (~80%) and OA has no effect on this deficit. Statistical comparisons between control (0) and treated conditions (1–50) were determined using one-way ANOVA with Sidak’s multiple comparisons or Kruskal–Wallis with Dunn’s multiple comparisons test (N = 6 vials/condition). e Western blot for GSK3α/β pSer21/9 in SH-SY5Y cells treated with OA 0–10 nM. Ser21/9 inhibitory phosphorylation of GSK3α/β starts to increase with an increasing dose of OA, while the total levels of GSK3α/β remain generally unaltered. Beta-actin serves as the loading control. f Quantification of the western blot of panel e shows a significant increase of Ser21/9 inhibitory phosphorylation for GSK3α/β after treatment with 5 nM OA and 10 nM OA (N = 3, mean ± SEM, one-way ANOVA with Sidak’s multiple comparisons) *p < 0.05; **p < 0.01; ***p < 0.001; ****p <  0.0001, ns not significant

These results suggest that in Drosophila, one role of PP2A is to modify the activation state of GSK3. To test whether this interaction is conserved in human cells, we treated SH-SY5Y cells with OA, in a concentration range of 0–10 nM. With increasing inhibition of PP2A, we observed increasing levels of phospho-GSK3α/β, indicating that PP2A also controls GSK3 phosphorylation in human cells (Fig. 7e, f).

Increased expression of kinesin-1 is sufficient