Upregulation of cellular palmitoylation mitigates α-synuclein accumulation and neurotoxicity

Abstract

Background:

Synucleinopathies including Parkinson disease (PD) are characterized by α-synuclein (αS) cytoplasmic inclusions. αS-dependent vesicle-trafficking defects are important in PD pathogenesis, but their mechanisms are not well understood. Protein palmitoylation, post-translational addition of the fatty acid palmitate to cysteines, promotes trafficking by anchoring specific proteins to the vesicle membrane. αS itself cannot be palmitoylated as it lacks cysteines, but it binds to membranes, where palmitoylation occurs, via an amphipathic helix. We hypothesized that abnormal αS membrane-binding impairs trafficking by disrupting palmitoylation. Accordingly, we investigated the therapeutic potential of increasing cellular palmitoylation.

Objectives:

We asked whether upregulating palmitoylation by inhibiting the depalmitoylase acyl-protein-thioesterase-1 (APT1) ameliorates pathologic αS-mediated cellular phenotypes and sought to identify the mechanism.

Methods:

Using human neuroblastoma cells, rat neurons and iPSC-derived PD patient neurons, we examined the effects of pharmacologic and genetic downregulation of APT1 on αS-associated phenotypes.

Results:

APT1 inhibition or knockdown decreased αS cytoplasmic inclusions, reduced αS serine-129 phosphorylation (a PD neuropathological marker), and protected against αS-dependent neurotoxicity. We identified the APT1 substrate microtubule-associated-protein-6 (MAP6), which binds to vesicles in a palmitoylation-dependent manner, as a key mediator of these effects. Mechanistically, we found that pathologic αS accelerated palmitate turnover on MAP6, suggesting that APT1 inhibition corrects a pathological αS-dependent palmitoylation deficit. We confirmed the disease relevance of this mechanism by demonstrating decreased MAP6 palmitoylation in neurons from αS gene triplication patients.

Conclusions:

Our findings demonstrate a novel link between the fundamental process of palmitoylation and αS pathophysiology. Upregulating palmitoylation represents an unexplored therapeutic strategy for synucleinopathies.

Introduction

Parkinson disease (PD) is the second most common neurodegenerative disorder,¹ and it lacks a disease-modifying treatment. This critical unmet need stems largely from an incomplete understanding of the basic mechanisms of molecular pathogenesis. Cytoplasmic inclusions known as Lewy bodies are characteristic of PD and are composed principally of α-synuclein (αS), a 14 kDa protein highly expressed in brain.² The pathophysiology of the synucleinopathies, which also include dementia with Lewy bodies (DLB), critically involve αS: point mutations^3,4 and increased gene dosage of wild type αS^5,6 each cause severe familial forms of PD/DLB.

An accumulating body of work suggests that pathological αS disrupts intracellular vesicle trafficking. ER-to-Golgi, endosomal, and lysosomal pathways have all been implicated.^7–10 Precisely how this occurs is unknown. One key possibility relates to αS structure. αS forms an amphipathic helix in the presence of small vesicles. This allows αS to bind dynamically to these highly curved membranes.^11,12 Some have therefore hypothesized that vesicle trafficking is disrupted when excess binding of pathologic αS to vesicle membranes alters their curvature, interfering with their fusogenic properties.^13,14 Consistent with this idea, recent data suggest that Lewy bodies, which have been classically described as composed of αS protein fibrils, also contain large amounts of dysmorphic lipid membranes, vesicles, and organelles.¹⁵

We noted that palmitoylation, the post-translational addition of a long-chain fatty acid (usually palmitate) to cysteines, increases the hydrophobicity of proteins to enable them to accumulate at curved membranes.^16-18 This occurs primarily because of lipid packing defects in such curved membranes, which create binding sites for palmitoylated proteins in the cytoplasmic leaflet, an energetically favorable interaction.^16,19 Although αS itself cannot be palmitoylated as it lacks cysteines, its binding to similarly curved membranes via an amphipathic helix led us to consider a possible link of αS to palmitoylation. In support of this connection, synucleinopathies involve defective vesicle trafficking, while palmitoylation promotes trafficking in multiple ways. Palmitoylation-dependent redistribution of proteins may act as an anterograde trafficking signal through the Golgi apparatus by concentrating protein cargo at the curved cisternal rim and promoting vesicle budding.²⁰ Palmitoylation of numerous specific proteins also regulates membrane vesicle trafficking.²¹ For example, the microtubule binding protein MAP6 is palmitoylated and transported on vesicles to the axon, where it promotes trafficking.^22,23 A link to αS is supported by the finding that saturated fatty acids including palmitic acid, reduce αS oligomers²⁴ and inclusions²⁵ in neurons.

Based on this reasoning, we hypothesized that one mechanism by which pathogenic αS can disrupt vesicle trafficking may be by interfering with the palmitoylation of one or more vesicle-binding proteins. To investigate this concept, we took advantage of available pharmacologic inhibitors of acyl-protein-thioesterase-1, a key depalmitoylase expressed in brain with a role in dendritic spine formation.²⁶ Here we show that inhibiting APT1, which enhances palmitoylation, ameliorates multiple aspects of αS cytopathology, including vesicle- and αS-rich inclusions, abnormal αS phosphorylation, and neurotoxicity. We show that these effects are principally mediated by the APT1 substrate MAP6. Further, PD-causing forms of αS disrupt the palmitoylation of MAP6, including endogenously in iPSC-derived neurons from familial PD patients. These data uncover a hitherto unknown feature of αS pathobiology with direct therapeutic potential.

Materials and Methods

Stable lines and induced neurons.

αS 3K-YFP expression in M17D cells was induced with 1 μg/mL doxycycline as previously described.²⁷ The established 2132 iPSC line from a previously clinically characterized healthy individual²⁸ was transduced with TetO-Ngn2-Puro²⁹ to establish “NR” (neurogenin-2 + rtTA) iPSCs. NR iPSCs were then transduced with pLVX-EF1a/αS-IRES-mCherry lentiviral plasmids for WT and E46K.²⁷ All iNs were grown on poly-L-ornithine/laminin plates (Biocoat). Neurogenin-2 differentiation was done as described.^29,30 The SNCA triplication and corrected iPSC lines (AST23) were a generous gift of the Kunath lab (University of Edinburgh).³¹

Imaging, cell growth, and inclusion quantification.

Inclusion formation was recorded in 4-hour-intervals in live M17D cells in 96-well plates on an IncuCyte Zoom machine (Essen). Constitutive mCherry, total YFP integrated intensity and YFP inclusion integrated intensity were quantified using the IncuCyte software. Parameters are described in detail in Supporting Information.

dsiRNA knockdown.

dsiRNA constructs (3 each) for APT1 (aka LYPLA1), Gα13 (GNA13), MAP6, and αS were obtained as pre-designed dsiRNAs from IDT with ID#s detailed in Supporting Information. Transfections were done using Lipofectamine RNAiMAX (Invitrogen). For M17D cells, transfection of dsiRNAs was done the day following plating. For iNs, transfection was performed on DIV5.

Acyl Resin-Assisted Capture (RAC) method for detection of palmitoylated proteins.

Initial steps were performed in an identical manner to the acyl biotin exchange assay as described.³² Following re-suspension of the blocked protein pellet in hydroxylamine containing buffer, samples were added to thiopropyl sepharose 6B (GE) and incubated for 1.5 hours to capture palmitoylated proteins.³³ After elution with 1% beta-mercaptoethanol, proteins of interest were detected by immunoblotting.

Statistical analysis.

We performed one-way ANOVA with Dunnett’s multiple comparisons test with the exception of Figure 5B, in which an unpaired two-tailed t-test was performed, and Figures 4G and 5G,H, in which a paired two-tailed t-test was performed. All analyses were performed using GraphPad Prism Version 8 following the program’s guidelines. Graphs include means +/− SEM. Criteria for significance, routinely determined relative to the appropriate control condition indicated in the figure legend, are: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Sufficient experiments and replicates were analyzed to achieve statistical significance, and these judgements were based on earlier, similar work.

Figure 5. αS disrupts the palmitoylation cycle of MAP6.

(A) αS 3K does not affect incorporation of palmitate (“forward” palmitoylation) of MAP6. M17D cells were transfected with vector control or MAP6-FLAG and subjected to 17-ODYA palmitic acid analog metabolic labeling with or without doxycycline to induce αS 3K-YFP, followed by detection of fluorescent signal as in (4F). WB below demonstrates expression of MAP6-FLAG and αS 3K-YFP in the appropriate lanes. (B) Quantification of A as done in (4F) (N = 4). (C) Pulse chase analysis of palmitate turnover on MAP6 by tandem fluorescent imaging shows increased MAP6 depalmitoylation in the presence of αS 3K-YFP. M17D cells were transfected with MAP6-FLAG with or without doxycycline induction of αS 3K-YFP. Cells were pulse labeled with both 17-ODYA and L-AHA (methionine analog) for 1 hr and chased for the indicated times. (D) Quantification of (C). IR800 fluorescent signal (palmitoylation) was normalized to IR680 (protein turnover) and expressed as percentage of signal at time 0. (N = 3). (E) iNs expressing vector only, αS wt or E46K were analyzed by Acyl-RAC. Hydroxylamine dependence demonstrates specificity of the signal. (F) Quantification of (E). Palmitoylation signal was normalized to input and expressed as percentage vector control. (N = 3). Criteria for significance relative to vector control were *P < 0.05, **P < 0.01, ***P < 0.001. (G) Patient-derived αS gene triplication (trip) iNs and isogenic corrected lines (corr) were analyzed as in (E). Palmitoylation signal was normalized to input and expressed as percentage of the triplication signal. (N = 4). (H) Isogenic corrected line from (G) was treated with dsiRNA against αS and palmitoylation of MAP6 measured as in (G). (N=3).

Figure 4. APT1 substrate MAP6 decreases inclusions in a manner dependent on its palmitoylation.

(A, B) M17D cells were transfected with 3 distinct dsiRNA constructs to Gα₁₃ (A) and MAP6 (B) or NC5 negative control. Inclusions were quantified as in (1B). (N = 25). (C, D) Verification of Gα₁₃ (C) and MAP6 (D) knockdown by WB. Protein levels are expressed as percent NC5 control dsiRNA, normalized to vinculin loading control (N = 3). (E) M17D cells were transfected with vector control, FLAG-tagged MAP6 wild type, or MAP6 C5, 10, 11S palmitoylation deficient mutant and inclusions measured as in (1B). (N = 20). (F) M17D cells stably expressing vector, wild type APT1, or catalytically inactive APT1 S119A were transfected with MAP6-FLAG and metabolically labeled with 17-ODYA palmitic acid analog. Incorporation of 17-ODYA into MAP6-FLAG was detected by click chemistry ligation of an azide-800 infrared dye (N3-800) followed by gel fluorescent scanning (top) and Commassie staining (CBB) (bottom). Blank lanes indicate cells grown in DMSO solvent only without 17-ODYA. Fluorescent band intensity was normalized to CBB and expressed as percent vector control. (N = 3). Statistical comparisons were to vector control. (G) M17D cells stably expressing MAP6 were transfected with control or dsiRNA against APT1 were subjected to acyl-resin assisted capture (RAC) to measure MAP6-FLAG palmitoylation. For A-E, G criteria for significance relative to NC5 negative control dsiRNA were *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Results

Inhibition of APT1 decreases αS 3K and E46K cytoplasmic inclusions

Previous work has shown that amplifying the αS E46K familial PD (fPD)-causing mutation by inserting homologous E-to-K mutations into the two adjacent KTKEGV repeats (E35K+E46K+E61K = “3K”) causes neurotoxicity, αS-rich vesicular inclusions, and L-DOPA-responsive, Parkinson-like motor deficits in mice.^25,27,34 Similarly, doxycycline-regulated expression of YFP-tagged 3K (αS 3K-YFP) in M17D human neuroblastoma cells reveals round cytoplasmic inclusions upon live-cell microscopy. Importantly, known modifiers of αS neurotoxicity reduce such inclusions.²⁵ Using this model system, we asked whether modulating palmitoylation in neural cells by targeting the depalmitoylases APT1 and APT2 affects inclusions as an indicator of αS cytopathology. Inclusions were defined by minimum size and intensity parameters in quantitative fluorescence microscopy, as previously reported.²⁷ The total fluorescence intensity of the αS-YFP+ inclusions (Fig. S1A,B) was measured and normalized to total constitutively co-expressed mCherry as a marker for cell number. Treatment with palmostatin B (PSB), a dual APT1 and APT2 inhibitor,³⁵ ML348, an APT1-specific inhibitor,³⁶ or ML349, an APT2-specific inhibitor,³⁶ each decreased total fluorescence intensity of αS 3K-YFP+ inclusions (Fig. 1A,B) without changing the levels of the αS 3K-YFP protein (Fig. S1C).

Figure 1. Inhibition of APT1 reduces αS 3K-YFP and E46K-YFP cytoplasmic inclusions.

(A) Pharmacologic inhibition of APT1 and APT2 reduces αS 3K-YFP inclusions. Expression of αS 3K-YFP was induced in M17D cells with doxycycline in the presence of 10 μM palmostatin B (PSB), ML348 or ML349 and inclusions measured after 24 hr in the Incucyte automated video microscopy system, with representative images shown. Arrows indicate representative inclusions. Scale bar, 10 μm. (B) Quantification of inclusions in M17D cells treated with 2.5 or 10 μM of the indicated compounds, normalized to constitutive mCherry levels and expressed as a percentage of DMSO vehicle control. (N = 30 wells). (C) dsiRNA knockdown of APT1 reduces inclusions. M17D cells were transfected with 3 distinct dsiRNA constructs to APT1 or NC5 negative control. Quantification of inclusions analyzed as in (B) (left); verification of APT1 knockdown by WB (right). (N = 25 wells). SCD1 is stearoyl-CoA desaturase 1, a known modulator of inclusions used as a positive control. (D) Inhibition of APT1 reduces neuronal inclusions. Representative images of rat cortical neurons transfected with YFP, αS wt-YFP, or αS 3K-YFP with or without 10 μM ML348 are shown. Scale bar, 20 μm. (E) Reduction in inclusions with ML348. Inclusions were measured by Incucyte and normalized to total YFP signal, expressed as percentage of αS 3K-YFP inclusions treated with DMSO vehicle control. (N = 12 wells). (F) Reduction in αS E46K-YFP inclusions, analogous to (E). (N = 20 wells). All data are means ± s.e.m. Criteria for significance relative to DMSO or NC5 negative control dsiRNA were *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

We noted that PSB and ML348 at 10 μM both decreased αS+ inclusions to a similar degree (~40%), suggesting that APT1 mediates a majority of this effect (Fig. 1B). ML349 treatment produced less of a change in inclusions than ML348 (Fig. S1D). Since there is significant overlap in the protein substrates of APT1 and APT2,³⁷ we focused our attention on APT1. To confirm that ML348 acts through APT1, we knocked down APT1 using three distinct dsiRNA constructs, each of which decreased inclusions (Fig. 1C,S1E), as did knockdown of SCD1, a known modulator of inclusions.^25,38 To determine whether APT1 inhibition decreases inclusions in neurons, we measured αS inclusions in αS 3K-YFP transfected primary rat cortical neurons with or without ML348 treatment (Fig. 1D,E). Consistent with the M17D cell data, we observed decreased αS neuronal inclusions with ML348.

Previous work has shown that the engineered αS 3K mutant results in more robust PD-like phenotypes than the E46K fPD mutant but nevertheless reflects the same fundamental pathophysiology as the naturally occurring mutation.^25,34 To determine if this principle also applies to the APT1-mediated improvement in αS inclusion phenotype, we quantified inclusions in rat cortical neurons transfected with αS E46K-YFP with or without ML348 treatment (Fig. 1F). While the αS E46K-YFP inclusion level is only about 20% that of 3K-YFP, consistent with published data,²⁷ they are also reduced by APT1 inhibition. Thus, reduction of αS 3K-YFP inclusions by APT1 inhibition reflects a property of the fPD-causing E46K mutation.

Inhibition of APT1 ameliorates measures of αS toxicity

Next, we asked how modulation of APT1 affects measures of αS neurotoxicity. Approximately 90% of the αS found in LBs in PD is phosphorylated at serine 129 (pSer129 αS).^39,40 Thus, this modification is often used as a marker of PD-relevant αS neuropathology. To investigate the effect of APT1 on pSer129 αS, we employed the neurogenin-2 induced human neuron (iN) system.^29,30 These iNs were differentiated from the previously characterized clinically normal 2132 iPSC line²⁸ and virally transduced to express ectopic wild-type (wt) or E46K human αS. We confirmed the expression of neuronal markers at DIV 18 (Fig. 2A,B). Treatment of the neurons with ML348 significantly reduced pSer129 αS levels in both wt- and E46K-expressing neurons in a dose-dependent manner (Fig. 2C). This effect was not specific to one iPSC line: ML348 treatment of iNs derived from YZ1 iPSCs⁴¹ similarly reduced pSer129 αS (Fig. 2D). As expected, we observed a step-wise increase in the amount of pSer129 αS relative to total αS with the E46K and then the 3K mutation (Fig. 2E), consistent with extensive published evidence that the 3K mutant amplifies PD-like phenotypes.^25,34,42 To demonstrate specificity of the target, we showed that dsiRNA knockdown of APT1 also reduced pSer129 αS (Fig. 2F).

Figure 2. Inhibition of APT1 reduces pSer129 αS.

(A) IF for neuronal markers demonstrates the neuronal phenotype of human iN cells derived from the 2132 wt iPSC line. Scale bar, 50 μm. (B) IF for neuronal markers in YZ1 human iN cells. Scale bar, 50 μm. (C) APT1 inhibition in human neurons reduces αS phosphorylated at S129 (pSer129 αS). iNs expressing vector only, αS wt or E46K at DIV18 were treated with the indicated doses of ML348 for 48 hr. Shown are a representative result (left) and quantification (right). pSer129 αS band intensity was normalized to total αS and expressed as percentage of the untreated DMSO vehicle control. (N = 5). (D) Analogous to c using iNs derived from the YZ1 iPSC line. (N = 4). (E) pSer129 αS was measured by WB in iNs expressing αS wt, E46K, or 3K, expressed as a ratio of pSer129 αS to total αS. (N = 4). (F) APT1 knockdown in human neurons reduces pSer129 αS. iNs were transfected with dsiRNA against APT1 and pSer129 αS measured as in (B). (wt, N = 5; E46K, N = 7). All data are means ± s.e.m. Criteria for significance relative to DMSO or NC5 negative control dsiRNA were *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

While PD progresses gradually over decades, cellular models may utilize mutant αS (such as E46K or 3K) or over-expressed wt αS to induce short-term toxicity as a proxy indicator.^38,42,43 We re-derived a M17D line with higher levels of doxycycline-inducible αS 3K-YFP, which resulted in inhibition of cell growth, unlike the line used in Figure 1. Treatment of these cells with ML348 rescued both this 3K-dependent cell growth defect (Fig. 3A) and an associated ATP deficit (Fig. 3B) in a dose-dependent manner. In neurons, viral transduction of wt αS results in rapid toxicity.^38,43 Using human iNs transduced with wt αS, we found that APT1 dsiRNA knockdown reduced cytotoxicity as measured by LDH release (Fig. 3C) without affecting αS levels (Fig. 3D,E). Taken together, these results show that downregulating APT1 function either pharmacologically or by genetic knockdown ameliorates multiple measures of αS cytotoxicity across several distinct cellular models.

Figure 3. Inhibition of APT1 reduces αS-dependent cytotoxicity.

(A) APT1 inhibition restores cell growth in M17D cells. αS 3K-YFP was induced the day after plating in the presence of the indicated doses of ML348 or DMSO vehicle control and cell growth measured by confluence in the Incucyte. (N = 10). (B) ML348 restores cellular ATP content. Cellular ATP content was measured 96 hr after induction of αS 3K-YFP. (N = 10). (C) APT1 knockdown protects human neurons from αS wt toxicity. APT1 was knocked down in iNs as in Figure 2f followed by transduction with either vector control or αS wt encoding lentivirus to induce cytotoxicity, measured by LDH release 4d post-transduction. (N = 5). (D) WB confirms knockdown of APT1 and (E) similar levels of αS. All data are means ± s.e.m. Criteria for significance were *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

The APT1 substrate MAP6 modulates pathologic αS inclusions

We sought to determine the mechanism of APT1 inhibition on αS cytopathology. The G-protein α subunit Gα13 and microtubule-associated protein 6 (MAP6) are known palmitoylated substrates of APT1.^26,44 These proteins regulate distinct aspects of neural cell physiology. APT1-mediated depalmitoylation of Gα13 regulates dendritic spine morphogenesis, presumably through the cytoskeletal re-organizing functions of its effector RhoA.²⁶ Palmitoylated MAP6 binds to secretory vesicles which transport it to specific subcellular regions, where it promotes intracellular vesicle trafficking via its microtubule stabilizing function.²³ Importantly, an unbiased palmitoyl-proteome study performed in APT1/2 double knockout mouse brain found that MAP6 was the only protein with a greater than 2-fold increase in palmitoylation levels in the knockout compared to wild type.⁴⁴ This suggests that in the brain, MAP6 is the principal substrate of APT1, APT2, or both.

Since APT1 inhibition or knockdown reduced αS inclusions, we reasoned that depletion of its relevant substrate should prevent this beneficial effect. Knocking down Gα13 did not change αS 3K-YFP+ inclusions in M17D cells (Fig. 4A), despite at least 60-70% knockdown of the protein (Fig. 4C). In contrast, 2 of the 3 dsiRNA knockdowns of MAP6 increased inclusions (Fig. 4B). Interestingly, we noted that dsiRNA #2 to MAP6, which did not have an effect on inclusions, also conferred the least degree of knockdown (Fig. 4B,D). These results provide evidence for the specificity of MAP6, the major known substrate of APT1 in brain, in modulating αS 3K inclusions. They also suggest that cells can tolerate loss of MAP6 to a certain degree before measurable changes in αS+ inclusions may be detected by our methods. Consistent with its worsening of inclusions, knockdown of MAP6 increased pSer129 αS, while Gα13 again had no effect (Fig. S2A,B). dsiRNA#3 against Gα13 does slightly decrease total αS levels; however the other two dsiRNAs do not, suggesting this is an off-target effect (Fig. S2A). MAP6 is a particularly interesting candidate for our hypothesis since it associates with vesicles specifically through its palmitoylation and thereby promotes intracellular vesicle trafficking.²³ A recent study also implicated altered glycosylation of MAP6 in an MPTP-based mouse model of PD.⁴⁵

Inhibition of APT1 decreases αS+ inclusions, and loss of this depalmitoylase activity would be expected to increase palmitoylation of its substrate MAP6. Thus, if MAP6 is the relevant APT1 substrate in our studies, palmitoylated MAP6, but not its non-palmitoylated form, would be predicted to decrease αS+ inclusions in the same manner as inhibiting the depalmitoylase APT1. To test this hypothesis, we transfected M17D cells with either wt MAP6 or a palmitoylation-incompetent mutant containing cysteine-to-serine mutations at each of its 3 known palmitoylation sites: C5, C10 and C11. ²³ Consistent with our hypothesis, expression of wild type MAP6 reduced αS 3K-YFP inclusions, while its palmitoylation-preventing mutant did not (Fig. 4E).

We directly confirmed that MAP6 is a substrate of APT1. Expression of wt APT1 nearly abolished incorporation of the palmitic acid analog 17-ODYA into MAP6 in cells, while the catalytically-inactive S119A mutant of APT1 had no effect (Fig. 4F). Similarly, knockdown of APT1 increased MAP6 steady state palmitoylation measured by by the acyl resin-assisted capture (RAC) method (Fig. 4G). Taken together, the above data strongly suggest that palmitoylation of the APT1 substrate MAP6 plays a key role in downregulating αS inclusions.

Pathologic αS disrupts the physiologic palmitoylation cycle of MAP6

We reasoned that the effect of APT1 and its substrate MAP6 on inclusions may fit into two broad scenarios. Pathologic αS may have no effect on the palmitoylation of MAP6, and increased MAP6 palmitoylation as seen with APT1 inhibition simply suppresses the deleterious effects of αS via a separate mechanism. Alternatively, pathologic αS may directly interfere with physiologic MAP6 palmitoylation, and this is corrected by APT1 inhibition.

To distinguish between these two possibilities, we asked if αS 3K disrupts palmitoylation of MAP6. We directly measured incorporation of the palmitic acid analog 17-ODYA into MAP6 by metabolic labeling in the presence or absence of αS 3K expression. Induction of αS 3K with doxycycline did not affect the incorporation of 17-ODYA into MAP6 (Fig. 5A,B). This process may be thought of as the “forward” (or biosynthetic) palmitoylation reaction. However, since APT1 is a depalmitoylase, we asked whether the “reverse” reaction, that is, depalmitoylation, is affected by αS 3K. To answer this question, we performed pulse-chase analyses of MAP6 depalmitoylation by tandem fluorescent imaging and click chemistry of the palmitic acid analog 17-ODYA and the methionine analog L-AHA. In contrast to the “forward” reaction, depalmitoylation of MAP6 was accelerated by the presence of αS 3K (Fig. 5C,D).

Having shown that αS 3K alters the dynamics of MAP6 palmitoylation, we asked if steady-state levels of palmitoylation are changed also, and whether the wt and fPD mutant forms of αS have an effect. We note that increased αS gene dosage (in the form of duplications and triplications of the wt gene) and specific point mutations such as E46K each cause fPD.^4,5 Using human neurons (iNs) stably expressing wt or E46K αS, we measured levels of endogenous MAP6 palmitoylation by acyl-RAC. Expressing excess wt αS or E46K αS each decreased palmitoylation of MAP6 (Fig. 5E,F). In contrast, the palmitoylation of Gα13, which did not affect αS inclusions (Fig. 4A), was unchanged by αS (Fig. 5E,F). In order to assess the relevance of this finding to naturally occurring PD-causing mutations, we performed similar experiments in iNs from a familial PD patient with SNCA (αS) gene triplication.⁵ We found that the doubled endogenous αS gene dosage also resulted in decreased MAP6 palmitoylation compared to the isogenically corrected control line (Fig. 5G). The palmitoylation of Gα13 was again unchanged (Fig. 5G). Conversely, knockdown of SNCA in the corrected (wt) line increased MAP6 palmitoylation, indicating that physiologic levels of αS modulate normal MAP6 palmitoylation (Fig. 5H). We noticed that expression of wt and E46K αS decreased steady state MAP6 palmitoylation to similar extents, despite the more severe phenotype of E46K in terms of αS phosphorylation and cytotoxicity.^34,42 It is possible that there are other mechanisms in addition to alteration of MAP6 palmitoylation that cause αS inclusions and these mechanisms are more adversely affected by E46K than wt αS wt. Taken together, our data demonstrate a novel link between PD-relevant forms of αS and the basic cell biological process of palmitoylation, representing a potential new approach to synucleinopathy therapeutic development, as discussed below.

Discussion

The central role of pathologic αS in altering vesicle trafficking is becoming well-recognized.^13,14 This is in accord with the dynamic membrane-binding properties of αS and the dysmorphic membranous elements found in αS-rich Lewy bodies.¹⁵ Our study shows for the first time that protein palmitoylation, which promotes vesicle trafficking by multiple mechanisms, ameliorates several aspects of αS-induced cytopathology. In particular, inhibition of the depalmitoylase APT1 decreases inclusions of αS and protects cells against various measures of αS-dependent cytotoxicity across distinct cellular models through actions on its known substrate, MAP6.

Our study has direct therapeutic implications. Modulating palmitoylation has not previously been considered in therapeutic development. Inhibiting APT1 may be one such possibility as our data suggest, but the literature indicates many more are likely. A large number of palmitoylated proteins other than MAP6 have critical functions in trafficking, including the potent suppressor of αS toxicity ykt6 (discussed in more detail below). The depalmitoylase for ykt6 and other palmitoylated proteins involved in trafficking is yet unclear. However, the existence of approximately 21 depalmitoylases^46,47 and 23 palmitoyltransferases⁴⁸ in the mammalian genome, all with overlapping but distinct substrates, suggests a complex interplay that remains to be explored in the context of synucleinopathy therapeutic development. Of relevance to our overall hypothesis, upregulation of palmitoylation by ectopic expression of zDHHC palmitoyltransferases enhanced anterograde trafficking through the Golgi, likely by promoting vesicle budding.²⁰

Our study also provides basic cell biological insights. αS contains no cysteines and therefore cannot be palmitoylated. However, we present evidence for a novel functional link between αS and palmitoylation. This may be because both amphipathic helices (such as that contained in αS) and palmitoyl moieties on proteins bind to curvature-induced lipid packing defects of small-vesicle membranes. Competition for such sites may result in excess or mutant αS displacing palmitoylated substrates such as MAP6 to other membrane domains where they may be more susceptible to depalmitoylation.

MAP6 associates with Golgi and vesicle membranes via its three palmitoylated cysteines. Transport of MAP6 on vesicles distributes it to subcellular regions where it further promotes trafficking by stabilizing microtubules. Our data are consistent with a model in which pathologic αS abnormally accelerate turnover of palmitate on MAP6, decreasing normal association of MAP6 with vesicles. MAP6 may then be unable to migrate to its usual subcellular regions to promote trafficking. We postulate that as αS+ vesicles bud from the Golgi in the setting of such impaired trafficking, they accumulate and eventually form the vesicle-rich αS inclusions²⁷ we observe here. These inclusions are reminiscent of Lewy bodies in human disease due to a) the presence of abundant αS, b) dysmorphic lipid structures, and c) morphological similarity to Lewy-like bodies in aged (>16 month old) αS 3K transgenic mice that display multiple PD-like phenotypes.^27,34,49

A previous study characterizing the role of MAP6 palmitoylation argued that ABHD17B, not APT1, was primarily responsible for depalmitoylation of MAP6 in developing hippocampal neurons.²³ However, upon careful examination, these and other data do not contradict the notion of MAP6 being the principal substrate of APT1 in our experimental systems. First, in that paper, ²³ early-stage developing hippocampal neurons were used, in contrast to mature cortical neurons in our experiments. Second, ectopic expression of ABHD17B reduced MAP6 vesicle binding (a palmitoylation-dependent property) but APT1 also had a significant though less robust effect (Figure 8b in that reference).²³ The separate but related question of the identity of the major APT1 substrate(s) was addressed in another study.⁴⁴ Martin and colleagues found in their unbiased palmitoyl-proteome examination of APT1/2 double knockout adult mouse brain that MAP6 was the only protein with a greater than 2-fold increase in palmitoylation in the knockout. This result suggests that MAP6 is the main physiologic substrate of APT1, APT2, or likely both, since they have significantly overlapping specificities.³⁷ The above results are not mutually exclusive. Multiple depalmitoylases may act on MAP6 to varying degrees in specific contexts. For example, ABHD17B may be more important for MAP6 depalmitoylation in developing hippocampal neurons, where APT1 nevertheless has some effect, while APT1 may be more functionally prominent in the adult mouse brain. The methodologies of each study were also different, with one employing an over-expression system²³ and the other, a mouse knockout.⁴⁴

Extensive work by our group and others has shown that physiologic αS exists in part as a tetramer.^50-52 Disease causing αS mutations, including E46K examined in this study, decrease tetramers and create an excess of monomers.⁴² These monomers accumulate in cytoplasmic inclusions, are associated with neurotoxicity,^42,49 and when caused by E46K or 3K mutation, are more membrane associated compared to tetramers.⁴² In light of these properties, the deleterious effects of abnormal αS on MAP6 palmitoylation are likely due to excess αS monomers.

While this manuscript was in preparation, two studies were published which added potential new dimensions to our central hypothesis. One group, using an unbiased approach, identified MAP6 as the sole protein for significantly altered Gal-(β-1,3)-GalNAc glycosylation in an MPTP PD model.⁴⁵ Unlike palmitoylation, the influence of glycosylation on MAP6 function is unknown. However, in the context of our study, this work raises the possibility that several mechanisms of PD pathogenesis may converge on MAP6.

A second group found that αS impairs trafficking of hydrolases to lysosomes by inhibiting the function of the SNARE protein ykt6.⁵³ Ykt6 is regulated by both palmitoylation and farnesylation in a unique manner. Farnesylation of ykt6 traps it in an inactive “closed” conformation by stabilizing a hydrophobic groove.⁵⁴ Subsequent palmitoylation of an adjacent cysteine then permits adoption of an “open” conformation, membrane association, and SNARE activity.⁵⁵ Farnesyltransferase inhibitors restored ykt6-dependent hydrolase trafficking.⁵³ The influence of palmitoylation on ykt6 was not explored. When considered together with our work, it raises the possibility that palmitoylation impacts αS biology through multiple pathways. Further work will be required to determine the role of palmitoylation in this and potentially other aspects of the molecular pathophysiology of the ubiquitous and abundant αS protein in neurons.

Supplementary Material

Figure S1

Figure S1. Inhibition of APT1 does not alter total αS 3K-YFP levels.

(A) Co-stain of YFP inclusions with an αS specific antibody. Scale bar, 1 μm. (B) Demonstration of αS 3K-YFP inclusions with DAPI co-staining. Scale bar, 5 μm. (C) Quantification of αS 3K-YFP levels in cells treated with the indicated compounds, normalized to intensity of the untreated condition. (N = 3). (D) Quantification of inclusions in M17D cells treated with either ML348 or ML349. (N = 30). *P < 0.05 (E) Quantification of αS 3K-YFP levels in cells treated with the indicated dsiRNAs, normalized to NC5 control (N=4). All data are means ± s.e.m.

Figure S2

Figure S2. Knockdown of MAP6 increases pSer129 αS.

(A) M17D cells were treated with dsiRNA against Gα13 as in 4a, followed by Western blotting of pSer129 αS 3K-YFP and total αS 3K-YFP. Levels of pSer129 αS 3K-YFP were normalized to total αS 3K-YFP and expressed as a percentage of the NC5 control dsiRNA. Total levels of αS 3K-YFP were expressed as a percentage of NC5. (N = 4) (B) Western blot of αS 3K-YFP levels with MAP6 knockdown using the two effective dsiRNAs, measured as in S2A. (N = 4) *P < 0.05, **P < 0.01, ***P < 0.001. All data are means ± s.e.m.