α-Synuclein plays a central causative role in Parkinson's disease (PD). Increased expression of the P-type ATPase ion pump PARK9/ATP13A2 suppresses α-Synuclein toxicity in primary neurons. Our data indicate that ATP13A2 encodes a zinc pump; neurospheres from a compound heterozygous ATP13A2−/− patient and ATP13A2 knockdown cells are sensitive to zinc, whereas ATP13A2 over-expression in primary neurons confers zinc resistance. Reduced ATP13A2 expression significantly decreased vesicular zinc levels, indicating ATP13A2 facilitates transport of zinc into membrane-bound compartments or vesicles. Endogenous ATP13A2 localized to multi-vesicular bodies (MVBs), a late endosomal compartment located at the convergence point of the endosomal and autophagic pathways. Dysfunction in MVBs can cause a range of detrimental effects including lysosomal dysfunction and impaired delivery of endocytosed proteins/autophagy cargo to the lysosome, both of which have been observed in cells with reduced ATP13A2 function. MVBs also serve as the source of intra-luminal nanovesicles released extracellularly as exosomes that can contain a range of cargoes including α-Synuclein. Elevated ATP13A2 expression reduced intracellular α-Synuclein levels and increased α-Synuclein externalization in exosomes >3-fold whereas ATP13A2 knockdown decreased α-Synuclein externalization. An increased export of exosome-associated α-Synuclein may explain why surviving neurons of the substantia nigra pars compacta in sporadic PD patients were observed to over-express ATP13A2. We propose ATP13A2’s modulation of zinc levels in MVBs can regulate the biogenesis of exosomes capable of containing α-Synuclein. Our data indicate that ATP13A2 is the first PD-associated gene involved in exosome biogenesis and indicates a potential neuroprotective role of exosomes in PD.
Parkinson's disease (PD) is a progressive age-related movement disorder that results primarily from the selective loss of midbrain dopaminergic (DA) neurons while more widespread pathology can occur, especially at later stages of the disease (1). Although the molecular mechanisms underlying PD are unknown, considerable evidence supports the involvement of the cytosolic protein α-Synuclein in a central causative role. α-Synuclein is a predominant component deposited in Lewy Bodies, the pathological hallmark of PD (2). Allele duplication or triplication of the wild-type α-Synuclein gene (SNCA) results in autosomal dominant PD with a severity proportional to the degree of α-Synuclein over-expression (3) whereas missense mutations in SNCA (e.g. A53T) are linked to dominantly inherited forms of PD (4). Most significantly, genome-wide association studies in multiple populations found that sporadic PD, which comprises 80–90% of patients, has a very significant association with the SNCA locus (5,6).
Modified or elevated expression of α-Synuclein can act in a toxic manner, and thus reducing intra-neural α-Synuclein levels or suppressing its negative effects is an important therapeutic endeavour. Approaches undertaken to suppress the toxic effects of α-Synuclein include: impairing the degeneration-associated phosphorylation of α-Synuclein, si/shRNA-mediated knockdown of α-Synuclein expression or enhanced autophagy as a mechanism to decrease α-Synuclein levels (7,8). An unbiased screen in an α-Synuclein-based PD model system identified over-expression of the yeast gene YPK9 as a suppressor of α-Synuclein-induced toxicity (9). YPK9 is an orthologue of human ATP13A2, and the over-expression of ATP13A2 was subsequently shown to rescue α-Synuclein-induced neurodegeneration in multiple models, including rat midbrain primary DA neurons (9).
ATP13A2 was identified as encoding a predicted polytopic P-type ATPase transporter of unknown ion specificity (10) which, when mutated, resulted in early onset Kufor–Rakeb syndrome, an autosomal recessive juvenile onset form of L-DOPA-responsive parkinsonism that exhibits clinical features of PD (11,12). ATP13A2 is highly expressed in the human brain with highest expression in the substantia nigra pars compacta (SNc), a region that displays progressive DA neurodegeneration in PD (12). Furthermore, surviving DA neurons in the SNc of sporadic PD patients expressed ATP13A2 mRNA at 5- to 10-fold higher levels than controls (12) whereas ATP13A2 protein levels were also found to be increased in surviving SNc neurons of PD patients relative to controls (13). This is an intriguing observation given that ATP13A2 over-expression in primary DA neurons suppresses α-Synuclein toxicity (9) and raises the possibility that elevated ATP13A2 expression may contribute to the survival of the SNc neurons in sporadic PD. Identifying ATP13A2’s function could provide potential therapeutic approaches to enhance neuronal survival and ameliorate the neural degeneration underlying PD.
ATP13A2 and its yeast orthologue YPK9 are members of the P5 subfamily of ATPases, which utilize ATP to pump inorganic cations such as metal ions (10). To identify the metal ions likely to be transported by ATP13A2, yeast cells lacking YPK9 (ypk9Δ) were challenged with different metal ions and showed an increased sensitivity to manganese relative to control cells (9), suggesting that YPK9 and its orthologue ATP13A2 might transport manganese. However, neither ATP13A2−/− patient cells nor human cells with reduced ATP13A2 expression have yet to be tested for sensitivity to manganese or to any other metal.
The localization of epitope-tagged ATP13A2 to lysosomes (12–15) resulted in an examination of lysosomal dysfunction in cells with reduced ATP13A2 expression where decreased lysosomal proteolysis (14,15), impaired mitochondrial function (13,16) and decreased autophagic flux were observed. Autophagy, specifically macroautophagy, is a lysosomal degradation process that removes damaged and long-lived proteins, protein aggregates, and damaged organelles such as mitochondria (17,18). By serving to prevent accumulation of toxic material within the cell, it plays a critical neuroprotective role, as evidenced by its inactivation resulting in extensive neurodegeneration (19). The hallmark of autophagy is the sequestration of cytoplasm into a LC3-labelled double-membrane vesicle termed an autophagosome, which then fuses with degradative compartments in the endosomal/lysosomal pathway to result in the degradation of the inner vesicle and its contents (20). Perturbed lysosomal function and/or impaired autophagy in ATP13A2-deficient cells may contribute to the neurodegeneration of ATP13A2−/− patients as autophagy can deliver α-Synuclein to lysosomes for degradation (21). Conversely, the suppression of α-Synuclein toxicity by increased ATP13A2 expression (9) might be mediated by enhanced lysosomal delivery of α-Synuclein.
Here we show that endogenous human ATP13A2 contributes to zinc homeostasis, most likely by acting as a zinc transporter, and is found in the multi-vesicular body (MVB), a late endosomal compartment that represents a dynamic and important convergence point of the endosomal and autophagic pathways prior to lysosomal delivery. Alternatively MVBs can fuse with the plasma membrane to release extracellular vesicles termed exosomes that have the potential of intercellular transfer and have recently been implicated in a number of neurodegenerative diseases (22). We report that elevated ATP13A2 expression reduces intracellular α-Synuclein levels via increased externalization of exosome-associated α-Synuclein whereas reduced ATP13A2 expression resulted in decreased levels of exosomal α-Synuclein.
Yeast ATP13A2 (YPK9) expression levels alter zinc and manganese homeostasis
In an expanded examination of metal ion sensitivity, we found that yeast cells lacking the ATP13A2 orthologue YPK9 (ypk9Δ) displayed sensitivity to manganese (ionic form Mn2+) (Fig. 1), as had been observed previously (9), but displayed greater sensitivity to zinc (ionic form Zn2+). Conversely, YPK9 over-expression in wild-type yeast cells conferred resistance to both Mn2+ and Zn2+ (Fig. 1). Together these data suggest that Ypk9, and potentially human ATP13A2, encodes an ATP-driven pump capable of transporting Zn2+ and Mn2+. ypk9Δ cells were also tested for sensitivity to the transition metals cobalt and iron, but no significant sensitivity was detected for these two metals (data not shown).
Reduced human ATP13A2 expression results in zinc but not manganese sensitivity
To determine whether reduced ATP13A2 expression in human cells increased sensitivity to Zn2+ and/or Mn2+, we knocked down ATP13A2 expression in the human DA neuroblastoma SHSY5Y cell line that can be efficiently differentiated to neuron-like cells. To confirm ATP13A2 knockdown, we generated anti-ATP13A2 polyclonal antibodies that were capable of detecting endogenous levels of ATP13A2 protein. Immunoblotting found that these affinity-purified antibodies recognized endogenous levels of ATP13A2 in SHSY5Y cells as evidenced by a single band at the predicted molecular mass (128 kDa) that co-migrated (Supplementary Material, Fig. S1A) with V5 epitope-tagged ATP13A2 (12). SHSY5Y cells transfected with a pool of four human ATP13A2 siRNAs displayed significantly lowered ATP13A2 protein expression relative to control cells transfected with non-targeting siRNA (Supplementary Material, Fig. S1B), demonstrating both successful ATP13A2 knockdown and antibody specificity. The most effective ATP13A2 siRNA was used to generate a lentiviral shRNA plasmid to create a stable ATP13A2 knockdown SHSY5Y cell line (shPK9). Reduced ATP13A2 expression in these cells was confirmed by immunoblot (Supplementary Material, Fig. S1C) where ATP13A2 protein expression levels were reduced by ∼80% (Supplementary Material, Fig. S1D).
Differentiated shPK9 cells treated with 125 µm Zn2+ suffered a 1.9-fold reduction in viability (P < 0.001) with respect to control cells (Fig. 2A) yet displayed no increased sensitivity to Mn2+ (Fig. 2B). These data demonstrate that, in contrast to the loss of yeast YPK9 (ypk9Δ), reduced human ATP13A2 expression sensitizes neuron-like cells to Zn2+ but not to Mn2+, supporting a role for ATP13A2 in Zn2+ homeostasis.
ATP13A2−/− patient human olfactory neurospheres are sensitive to zinc but not to manganese
To further investigate ATP13A2’s role in Zn2+ homeostasis, human olfactory neurospheres (hONs) from an ATP13A2 compound heterozygous patient (designated ATP13A2−/−; (11)) were generated and the extracted proteins immunoblotted. Endogenous ATP13A2 was readily detectable in control hONs but not in ATP13A2−/− patient extracts (Fig. 2C). ATP13A2−/− patient and control hONs were then tested for sensitivity to Zn2+ and Mn2+ using a viability assay. ATP13A2−/− patient hONs were 10-fold more sensitive to 125 µm Zn2+ than control cells (P < 0.0001) (Fig. 2D) but interestingly displayed no increased sensitivity to 5000 µm Mn2+ relative to control cells (Fig. 2E). Our data from human ATP13A2−/− patients further support the model that ATP13A2 plays a role in Zn2+ homeostasis.
ATP13A2−/− patient hONs accumulate zinc
Further evidence of zinc dyshomeostasis in ATP13A2−/− patient hONs was sought through the use of synchrotron X-ray fluorescence microscopy (µ-XRF), which examines total intracellular element content (23). ATP13A2−/− patient hONs had very similar Zn2+ and Mn2+ levels to those of control hONs (Fig. 2F). However, when challenged with 100 µm Zn2+, ATP13A2−/− patient hONs displayed a 60% increase in total intracellular Zn2+ content (P < 0.05) compared with control hONs (Fig. 2F), suggesting that ATP13A2 may contribute to Zn2+ efflux from the cell. Consistent with the lack of sensitivity to Mn2+ (Fig. 2E), treatment of ATP13A2−/− patient hONs with 1 mm Mn2+ did not result in increased manganese accumulation relative to controls (Fig. 2G).
Elevated ATP13A2 expression in rat primary cortical neurons confers zinc resistance
To determine whether elevated ATP13A2 expression conferred Zn2+ resistance, control and ATP13A2 over-expressing rat primary cortical neurons (Fig. 3A) were challenged with Zn2+. Morphological assessment using reduced neurite length as an indicator of neurotoxicity (24,25) found that treatment of control neurons with 180 µm Zn2+ caused the retraction of neurites to an average length of 18 µm (Fig. 3B and D). In contrast, elevated ATP13A2 expression conferred Zn2+ resistance as indicated by the neurites preserved to an average length of 96 µm (P < 0.05) (Fig. 3C and D) and a 2.4-fold increase in cell viability (Fig. 3E) (P < 0.01).
Association of endogenous ATP13A2 with autophagosomes and MVBs
Identifying the subcellular localization of ATP13A2 would provide critical clues as to how ATP13A2 impacts Zn2+ homeostasis and assist in understanding how elevated ATP13A2 expression suppresses α-Synuclein toxicity. We created a SHSY5Y cell line that stably expressed ATP13A2-V5–6xHis (12) and staining with V5 antibody localized ATP13A2-V5–6xHis to lysosomes as indicated by co-localization with the lysosomal marker proteins LAMP1 and LAMP2 (Fig. 4A).
Although consistent with the lysosomal localization observed by several groups using transfected ATP13A2-V5–6xHis or ATP13A2-GFP (12–15), a caveat remained that either over-expression via the CMV promoter and/or epitope tagging might have caused mislocalization of ATP13A2-V5–6xHis to lysosomes, a possibility raised by Ramirez et al. (12).
In contrast to ATP13A2-V5–6xHis, endogenous ATP13A2 in differentiated SHSY5Y cells displayed a punctate, vesicular pattern that was clearly distinct, both in subcellular distribution and morphology, from the lysosomal proteins LAMP1 and LAMP2 (Fig. 4B). These data suggest that ATP13A2 exists on non-lysosomal vesicular structures into which it may facilitate the transport of Zn2+. High concentrations of Zn2+ in vesicles/organelles (referred to as vesicular zinc) have been observed previously using zinquin, a lipophillic Zn2+-specific fluorescent probe with affinity for vesicular zinc (26,27). If ATP13A2 facilitates transport of Zn2+ into such vesicles, then a reduction in ATP13A2 expression should reduce vesicular Zn2+ concentration and zinquin staining. We observed a characteristic multi-puncta zinquin staining pattern in differentiated SHSY5Y cells (Fig. 4C), whereas zinquin staining in differentiated shPK9 cells showed a >50% decrease (P < 0.0001) (Fig. 4C and D), indicating that the transport of Zn2+ into these vesicular structures is dependent upon ATP13A2. A large decrease in zinquin staining (P < 0.0001) was also observed in HEK293 cells that were knocked down for ATP13A2 expression (Fig. 4E).
The function of ATP13A2 and the consequences of its dysfunction would be aided by identifying its subcellular localization. Endogenous ATP13A2 did not significantly co-localize with markers of the endoplasmic reticulum (Calnexin), cis-Golgi (GM130), trans-Golgi (p230), early endosomes (EEA1, Rab5-GFP) or late endosomes (Rab7-GFP) by indirect immunofluorescence (Supplementary Material, Fig. S2). However, a close association was apparent between endogenous ATP13A2 and LC3, a marker of autophagosomes, in both differentiated SHSY5Y cells (Fig. 5A) and rat primary neurons (Fig. 5B).
Lysosomal delivery of autophagosomes was pharmacologically inhibited to stabilize and enhance the detection of the ATP13A2-LC3 association. Cells were treated with chloroquine (CQ), an effective inhibitor of the later stage(s) of autophagy, that is thought to act by neutralizing acidic endo-lysosomal compartments and impairing fusion of autophagosomes with these compartments (17). Treatment of differentiated SHSY5Y cells with CQ resulted in a dramatic redistribution of endogenous ATP13A2, such that it surrounded LC3-labelled autophagosomes (Fig. 5C). This association did not represent an autophagosome fused with the lysosome (autophagolysosome) as the ATP13A2-LC3-positive structures were devoid of LAMP1 staining (Fig. 5D). A close association of endogenous ATP13A2 with LC3 is significant as impaired autophagy is closely linked with neurodegeneration and autophagosomes have been found to contain Zn2+ (28).
Elevated ATP13A2 expression reduces intracellular α-Synuclein protein levels
A role for ATP13A2 in autophagy may account for the observed suppression of α-Synuclein toxicity by ATP13A2 over-expression (9) as elevated ATP13A2 levels might enhance autophagic rates and result in decreased levels of α-Synuclein. To test this, a well-characterized SHSY5Y stable cell line that conditionally expresses α-Synuclein [SHSY5Y-αSyn (29)] was infected with lentivirus-encoding untagged ATP13A2 (9). The resulting stable cells (SHSY5Y-αSyn-PK9) and control SHSY5Y-αSyn cells were differentiated, α-Synuclein expression induced and α-Synuclein levels determined by immunoblot. SHSY5Y-αSyn-PK9 cells contained 60% less intracellular α-Synuclein than SHSY5Y-αSyn cells (P < 0.001) (Fig. 6A), consistent with increased autophagic delivery of α-Synuclein to the lysosome for degradation.
Evidence of lysosomal degradation can be demonstrated using CQ to impair lysosomal delivery and proteolysis, thus allowing increased detection of a now stabilized lysosomal substrate (30). Surprisingly, CQ did not stabilize intracellular α-Synuclein levels in SHSY5Y-αSyn-PK9 cells compared with those of SHSY5Y-αSyn cells (Fig. 6B). CQ treatment successfully impaired lysosomal degradative function as evidenced by increased levels of autophagosome-associated LC3-II (Supplementary Material, Fig. S3B), which is normally rapidly degraded upon delivery to lysosomes (31). Furthermore, LC3-II levels in CQ-treated cells were not increased upon ATP13A2 over-expression (Supplementary Material, Fig. S3A), indicating that the reduction of α-Synuclein levels by ATP13A2 was not through enhanced activation of autophagy. Together these data indicate that the reduced level of intracellular α-Synuclein observed with ATP13A2 over-expression was not due to increased delivery to the lysosome. qRT-PCR analysis found that the decreased intracellular α-Synuclein protein levels associated with elevated ATP13A2 levels was not due to decreased α-Synuclein mRNA levels (Supplementary Material, Fig. S3C).
Externalization of α-Synuclein in exosomes by ATP13A2
ATP13A2′s association with autophagosomes in the absence of lysosomal delivery (Fig. 5D) prompted us to examine MVBs, which have been shown to fuse with autophagosomes to produce hybrid structures referred to as amphisomes (32). MVBs are dynamic late endosomes that participate in numerous endocytic and trafficking functions including rapid and complex protein and lipid sorting via inward budding of the MVB outer limiting membrane into the lumen to produce multiple intra-luminal vesicles (ILVs) (33).
Using CQ to preserve these dynamic structures, ATP13A2 displayed extensive association with LBPA (lysobisphosphatidic acid, also termed BMP, Fig. 7A), a lipid highly enriched in ILVs (34,35). ATP13A2 staining was also seen to encircle CD63 (Fig. 7B), a tetraspanin protein also highly enriched in ILVs (36). Triple labelling allowed the detection of ATP13A2 in the same compartment with both LC3 and CD63 (Fig. 7C), strongly suggesting that membrane spanning ATP13A2 occupies the outer limiting membrane of MVBs and amphisomes.
Our data show that elevated ATP13A2 lowers intracellular α-Synuclein levels and that ATP13A2 can associate with autophagosomes and MVBs. Although both autophagosomes and MVBs can fuse with lysosomes for degradation, we found that lysosomal inhibition failed to stabilize ATP13A2-dependent lowering of α-Synuclein levels (Fig. 6B), suggesting lysosomal delivery/degradation is not involved. However, MVBs can also fuse with the plasma membrane and release ILVs extracellularly as exosomes (33). Given that SHSY5Y-αSyn cells have previously been shown to produce α-Synuclein-containing exosomes (37), ATP13A2, in associating with MVBs, might influence the generation of α-Synuclein-containing ILVs and/or the release of ILVs from the cell as exosomes.
To determine whether over-expression of ATP13A2 enhanced externalization of α-Synuclein, cell-free culture medium from SHSY5Y-αSyn and SHSY5Y-αSyn-PK9 cells was concentrated on nitrocellulose membranes and probed with anti-α-Synuclein antibodies. Confirming previous studies (37), immunoblots detected α-Synuclein in the culture medium from SHSY5Y-αSyn cells; however, ∼5-fold more α-Synuclein was detected in culture medium from SHSY5Y-αSyn-PK9 cells (Fig. 6C) (P < 0.0001), indicating that elevated ATP13A2 expression may lower α-Synuclein levels within these cells by externalizing α-Synuclein in exosomes. To further investigate this potential ATP13A2–exosome relationship, we employed HEK293 cells as they have been used extensively to characterize exosomes owing to their high yield capacity (38,39). HEK293 cells were co-transfected with a plasmid expressing human wild-type α-Synuclein and either a vector control or plasmid expressing untagged ATP13A2. Exosomes were then purified from cell-free culture media via ultracentrifugation and the resulting immunoblots successfully probed against five different exosomal markers [Alix, CD63, Flotillin, Hsp70 and Tsg101 (40), Supplementary Material, Fig. S4], which demonstrated successful exosome purification. Elevated ATP13A2 expression resulted in a 3-fold increase (P < 0.05) in the amount of α-Synuclein associated with exosomes (Fig. 6D). The exosome fractions were negative for intracellular endoplasmic reticulum (Calnexin) and Golgi (GM130) markers indicating that the exosome-associated α-Synuclein detected was extracellular and not due to cell lysis (Supplementary Material, Fig. S4).
To further investigate this ATP13A2-enhanced externalization of exosome-associated α-Synuclein, we used density gradients to further purify exosomes from the media of α-Synuclein expressing HEK293 cells, with or without ATP13A2 over-expression. The floating of exosomes from the bottom of the density gradient eliminates contaminants from the subsequent analysis as large protein aggregates, which may co-sediment with exosomes sedimented by 100 000g centrifugation, do not subsequently float in a density gradient (41). Washed P100 fractions containing exosomes were loaded at the bottom of an OptiPrep density gradient and resolved by ultracentrifugation as has been described for enhanced exosomal purification (42). Fractions were collected and immunoblotted for a range of exosome markers [Alix, CD63, Flotillin, Hsp70 and Tsg101 (43)] as well as α-Synuclein (Fig. 8).
Elevated ATP13A2 expression resulted in a very large increase in exosomal α-Synuclein in a peak centred on fraction 4 (Fig. 8A). We examined whether elevated ATP13A2 expression changed the relative concentrations of exosomal markers within these α-Synuclein-containing fractions. Intriguingly, the fractionation pattern of the cytosolic molecular chaperone Hsp70 mirrored that of α-Synuclein with a large increase in a peak centred on fraction 4 that was completely dependent on ATP13A2 over-expression (Fig. 8C). Exosomal markers CD63 and Alix (43) also displayed large ATP13A2-dependent changes in fractions containing α-Synuclein (Fig. 8B and D). Flotillin and Tsg101 levels were also modulated by elevated ATP13A2 expression (Fig. 8E and F).
The role of ATP13A2 in the incorporation of α-Synuclein and exosomal markers into exosomes was further supported by assessing the impact of reduced ATP13A2 expression on this process. Exosomes from the media of HEK293 cells expressing α-Synuclein and either a vector control or an ATP13A2 shRNA knockdown plasmid were isolated (Supplementary Material, Fig. S5) and further purified by fractionation on a density gradient. ATP13A2 knockdown resulted in a >3-fold decrease in the level of exosomal α-Synuclein in the fraction 4-containing peak (Fig. 9A). This peak also displayed decreased levels of exosomal markers CD63 (>2-fold; Fig. 9B), Alix (3-fold; Fig. 9D), Flotillin (Fig. 9E) and Hsp70 (Fig. 9C).
Our data integrate previous observations that ATP13A2 expression is elevated in surviving DA neurons in sporadic PD patients (12) and that over-expression of ATP13A2 relieves α-Synuclein-induced cytotoxicity (9). We position ATP13A2 in MVBs, at the critical convergence point of the endosomal and autophagic pathways, acting to regulate Zn2+ homeostasis and influencing the exosome-associated externalization of α-Synuclein.
ATP13A2 plays a key role in zinc homeostasis
The P5 subfamily of ATPases, of which ATP13A2 is a member, utilize ATP to transport inorganic cations such as metal ions (10,44). Loss of ATP13A2, whether through deletion of the yeast orthologue YPK9, or knockdown of human ATP13A2, conferred sensitivity to Zn2+ (Figs 1 and 2A). Furthermore, hONs derived from an ATP13A2−/− patient (11) lacked detectable ATP13A2 protein levels (Fig. 2C) and displayed significantly increased sensitivity to Zn2+ relative to control cells (Fig. 2D). Upon Zn2+ challenge, ATP13A2−/− patient hONs accumulated 60% more Zn2+ when compared with control hONs (Fig. 2F). Correspondingly, over-expression of either YPK9 in yeast (Fig. 1) or human ATP13A2 in rat primary cortical neurons (Fig. 3) conferred resistance to zinc challenges. Finally, ATP13A2 knockdown in two cell types robustly reduced vesicular zinc levels as detected by zinquin (Fig. 4C–E). Together these data strongly support the proposal that ATP13A2 plays a key role in zinc homeostasis, most likely, given it is a P5 ATPase, by directly transporting Zn2+ from the cytosol into the MVB lumen. Further experimentation is required to directly demonstrate this function. By facilitating transport of cytosolic Zn2+ into the MVB lumen, ATP13A2 assists in maintaining cytosolic Zn2+ homeostasis but can additionally contribute to Zn2+ efflux from the cells as fusion of MVBs with the plasma membrane will release the ionic contents of the MVB lumen extracellularly. The loss of such a Zn2+ efflux would account for the significantly elevated levels of intracellular Zn2+ in ATP13A2−/− hONs when challenged with Zn2+ (Fig. 2F). An alternative explanation is that ATP13A2 expression negatively regulates uptake of extracellular Zn2+, but this model would not account for the reduced zinquin staining upon decreased ATP13A2 expression.
Zinc plays an important role in neurobiology and is integral to the homeostasis of the central nervous system; a loss of zinc homeostasis can contribute to the degeneration of neurons that are sensitive to both deficiencies and excesses of zinc. Although total zinc levels are high in the brain, neurons maintain very low levels of cytosolic free zinc (estimated to be in the nanomolar range) through binding to metallothioneins and sequestration in vesicles (45,46). A range of zinc transporters and pumps contribute to zinc homeostasis by transporting Zn2+ across membranes of the plasma membrane and organelles/vesicular compartments (47). The loss in patients of ATP13A2′s capacity to facilitate Zn2+ transportation is likely to have two distinct detrimental consequences: perturbed MVB function owing to diminished MVB luminal zinc levels and correspondingly increased cytosolic zinc concentration. Elevated free zinc levels can potently inhibit mitochondrial membrane potential resulting in ROS generation (48) that triggers the release of metallothionein bound zinc to further elevate cytosolic zinc levels. Environmental zinc exposure has been associated with PD (49) and at least one PD model (MPTP/MPP+) results in increased intracellular free zinc levels that contribute to the associated neurotoxicity (50).
ATP13A2 is less likely to play a significant role in Mn2+ homeostasis; neither ATP13A2 knockdown cells nor ATP13A2−/− patient hONs displayed increased sensitivity to Mn2+ (Fig. 2B and E) and µ-XRF studies also found no difference in intracellular Mn2+ levels between ATP13A2−/− patient hONs and controls when treated with Mn2+ (Fig. 2G). It has been reported (51) that ATP13A2 over-expression can partially protect cultured cells from high Mn2+ levels (2 mm) raising the possibility that ATP13A2 may, at such high Mn2+ concentrations, be additionally capable of transporting Mn2+. However, we found no significant difference in viability between ATP13A2−/− patient hONs and control hONs at Mn2+ levels as high as 5 mm (Fig. 2E).
ATP13A2 regulates zinc levels in MVBs
We found endogenous ATP13A2 in close association with the autophagosome marker LC3 (Fig. 5A and B), and this association was strongly enhanced in the presence of CQ, likely reflecting the dynamic nature of the interaction. ATP13A2-LC3-positive structures lacked lysosomal markers (Fig. 5D) and raised the possibility that ATP13A2 may reside on the autophagosomal membrane. However, ATP13A2 is not an obligate component of autophagosomes as we observed punctate LC3-positive structures that lacked ATP13A2. CQ treatment also revealed an association between ATP13A2 and the MVB markers LBPA and CD63 (Fig. 7). ATP13A2′s close association with both autophagosomes and MVBs is most consistent with ATP13A2 occupying the outer limiting membrane of MVBs that can fuse with autophagosomes to produce amphisomes (Fig. 10A) (32,52). The significant loss of zinquin-stained vesicular zinc in two different human cell lines with reduced ATP13A2 expression (Fig. 4C–E) indicates that ATP13A2 can function to regulate Zn2+ levels in the MVB lumen (Fig. 10B).
ATP13A2 has previously been assigned a lysosomal localization based on staining of over-expressed C-terminally tagged (V5–6xHis or GFP) ATP13A2 (12–15), and we also observed ATP13A2–V5–6xHis in the lysosome (Fig. 4A). However, we did not observe significant co-localization between endogenous ATP13A2 and the lysosome (Fig. 4B), raising the possibility that over-expressed tagged ATP13A2 is mislocalized, a prospect raised by the authors of the original ATP13A2-V5–6xHis/GFP constructs (12). Tagged proteins over-expressed in the secretory/endosomal system are susceptible to mislocalization by either sterically hindering C-terminal localization/retention sequences or saturating sorting machinery. In our studies, the elevated expression of untagged ATP13A2 did not result in lysosomal localization (Supplementary Material, Fig. S6), suggesting that the C-terminal tagging of ATP13A2 may be associated with its lysosomal localization.
MVBs can fuse with both autophagosomes and lysosomes and as such the autophagy–lysosomal pathway defects previously associated with reduced ATP13A2 expression/function could arise as consequences of MVB dysfunction induced by ATP13A2 loss. Impaired fusion of MVBs with lysosomes would result in reduced delivery of lysosomal hydrolases, resulting in the observed decreased lysosomal proteolytic activity and reduced lysosomal clearance rates (14,15). Reduced fusion of MVBs with autophagosomes, or the subsequent fusion of amphisomes with lysosomes, could account for the impaired autophagosomal flux associated with reduced ATP13A2 function (15,53). Similarly, lysosomal delivery and degradation of dysfunctional mitochondria via mitophagy may also be impaired owing to decreased MVB function associated with reduced ATP13A2 activity (16). Additional lysosomal defects could result from Zn2+-deficient MVBs fusing with lysosomes, potentially providing lysosomes with inadequate Zn2+ levels.
The role of ATP13A2 in α-Synuclein externalization
MVBs are dynamic organelles characterized by multiple ILVs enclosed within an outer limiting membrane. Initially regarded as purely prelysosomal structures along the degradative endosomal pathway, MVBs are now known to contribute to numerous endocytic and trafﬁcking functions, including the externalization of ILVs as exosomes (33). MVBs can fuse with: (i) lysosomes to deliver ILVs for degradation, (ii) autophagosomes to produce autophagosome-derived ILVs (AD-ILVs) or (iii) the plasma membrane to release ILVs to the extracellular space as exosomes (Fig. 10A).
The fusion of MVBs with the plasma membrane will also release the ionic contents of the MVB lumen extracellularly, including any zinc that could in turn influence interneuronal signalling (54). Any resulting perturbations in intracellular zinc homeostasis could also interfere with normal neuronal signalling and contribute to PD.
The transport of Zn2+ into the MVB lumen is likely to convey functions beyond sequestering excess cytosolic Zn2+. The metal plays an important role in the lumen of the late endosomal system as dyshomeostasis of Zn2+ transport to this compartment has been observed to affect membrane trafficking and/or sorting (55). By modulating luminal concentrations of Zn2+ within MVBs, ATP13A2 might play an important role in a range of functions that could include fusion of autophagosomes with MVBs, or regulating the biogenesis and/or sorting of ILVs within MVBs (Fig. 10B). Although a cytosolic protein, α-Synuclein may enter MVBs to produce α-Synuclein-containing ILVs via autophagy and/or late endosomal micro-autophagy (56). α-Synuclein could be engulfed by a phagophore forming an autophagosome that, upon fusion with an MVB, would result in an α-Synuclein-containing AD-ILV (Fig. 10A). In late endosomal micro-autophagy, cytosolic proteins can associate with molecular chaperones and the MVB limiting membrane and are incorporated into the nascent ILV lumen as it invaginates inward (56) (Fig. 10A). Identifying the proteins associated with α-Synuclein-containing exosomes, and determining whether their levels are modulated by ATP13A2 expression, provided valuable insights as to the mechanism involved.
Elevated expression of ATP13A2 resulted in a large increase in exosome-associated α-Synuclein (Fig. 8A) whose fractionation pattern was mirrored by that of Hsp70 (Fig. 8C), a molecular chaperone previously found in exosomes (57). Hsp70 has been shown to protect against α-Synuclein toxicity in vitro, to reduce the amount of α-Synuclein aggregates in vivo (58,59), and has been found to associate with externalized α-Synuclein (60). Given that Hsc70, the functional equivalent of Hsp70, was found to load ILVs with cytosolic proteins containing the peptide KFERQ, which includes α-Synuclein (56), it is possible that endosomal micro-autophagy is modulated by elevated ATP13A2 expression. ATP13A2 may produce zinc micro-environments in the MVB lumen that influence the clustering of lipid micro-domains and the sorting of α-Synuclein bound to Hsp70 into ILVs (Fig. 10B-1). Clustering of negatively charged lipid head groups within a micro-domain during invagination of a nascent ILV may be unfavourable owing to electrostatic repulsion and ATP13A2, providing luminal Zn2+ might help reduce this charge issue and allow more efficient lipid clustering, appropriate membrane curvature and ILV formation. An increased rate of ILV formation might permit enhanced endosomal micro-autophagy of α-Synuclein-Hsp70 and subsequent release as exosomes. Alternatively, modulation of MVB luminal zinc levels by ATP13A2 may enhance fusion of autophagosomes with MVBs (Fig. 10B-2) or impact the sorting of ILVs within MVBs between the alternative destinations of the plasma membrane and the lysosome (Fig. 10B-3). We observed that modulation of ATP13A2 expression caused corresponding changes of exosomal marker levels in purified exosomes (Figs 8 and 9) providing strong support that ATP13A2 plays a role in exosomal biogenesis.
Chemically induced lysosomal dysfunction, as reflected by large increases in both intracellular α-Synuclein and LC3-II levels, has been observed to result in an increase in α-Synuclein externalization (61). However, the process described by Alvarez-Erviti and co-workers (61) is distinct from that involving elevated ATP13A2 expression as, in contrast, we observed decreased intracellular α-Synuclein levels (Fig. 6B) and no increase in LC3-II levels (Supplementary Material, Fig. S3B), indicating no lysosomal dysfunction.
The role of ATP13A2 in Parkinson's disease
We have shown that elevated ATP13A2 expression results in decreased intracellular α-Synuclein levels and increased α-Synuclein externalization associated with exosomes. This is consistent with reduced ATP13A2 expression being associated with increased levels of intracellular α-Synuclein (15). Our data account for previous observations that elevated ATP13A2 expression suppressed α-Synuclein toxicity in several PD models including primary DA neurons (62) and that DA neurons still viable in the SNc of sporadic PD patients have been found to express ATP13A2 at 10-fold higher levels (12), whereas those cells either unable to compensate with elevated ATP13A2 expression or having naturally lower ATP13A2 expression levels may undergo degeneration (14,63). Our knockdown studies suggest that ATP13A2−/− patients would have an impaired capacity to produce exosomes and externalize α-Synuclein, which would likely contribute to their neurodegeneration.
Why would ATP13A2-dependent increased externalization of α-Synuclein via exosomes be beneficial to neurons when α-Synuclein delivered to MVBs via either autophagy or endosomal micro-autophagy could instead be delivered to the lysosome for degradation? This assumes that lysosomal delivery and degradative capacity are fully functional in aging neurons. However, aging brains exhibit reduced lysosomal function (64) and human PD brain tissue shows an increased accumulation of autophagic vesicles (65,66), suggesting a possible significant impairment in their ability to fuse with lysosomes. α-Synuclein itself might be contributing to this lysosomal dysfunction as α-Synuclein has been demonstrated to impair the lysosomal enzyme glucocerebrosidase (GBA). GBA inhibition in turn caused an accumulation of α-Synuclein aggregates, which caused further lysosomal dysfunction and α-Synuclein accumulation (67). Externalization of α-Synuclein would bypass potentially insufficient neuronal degradative capacity or dysfunctional lysosomes to reduce intra-neuronal α-Synuclein levels and decrease α-Synuclein-associated toxicity. An impaired capacity to externalize α-Synuclein might therefore contribute to the neurodegeneration in ATP13A2−/− patients by increasing the degradative load on neuronal lysosomes, potentially enhancing lysosomal dysfunction and impairing the delivery and/or degradation of autophagosomes or dysfunctional mitochondria (16,53).
The detection of Lewy bodies in embryonic nigral neurons transplanted in PD patients suggested the possibility of host-to-graft disease propagation (68,69). α-Synuclein externalized in cell culture has been observed to negatively impact surrounding neurons (37), supporting the proposal that extracellular α-Synuclein plays an important role in disease progression by amplifying and spreading degenerative changes from neurons to surrounding tissues (70). ATP13A2-mediated externalization of α-Synuclein might benefit the exporting neuron by decreasing intracellular levels of toxic α-Synuclein but could this consequently transfer toxicity to neighbouring neurons and trigger their degeneration? This apparent enigma might be resolved by the possibility that both exosomal and non-exosomal mechanisms are contributing to the externalization of α-Synuclein and that extracellular α-Synuclein not within exosomes is at least partially responsible for the extracellular toxicity. Some of the toxic forms of extracellular α-Synuclein are clearly not contained within exosomes as their toxic effect was blunted with anti-α-Synuclein antibody (37), whereas bacterially produced recombinant α-Synuclein has imparted toxicity to both primary neurons and neurodegeneration in whole animal studies (71,72). Therefore, α-Synuclein externalized within exosomes may provide a benefit to the exporting neuron without conferring a detriment to neighbouring neurons.
Exosomes have been found to contain and transfer selected proteins, mRNA and microRNA into target cells and provide a powerful potential path of cell–cell communication as well as a promising therapeutic tool for multiple diseases. The rapidly expanding role of exosomes is being actively pursued within neurodegeneration, immunology, cancer biology and viral infection including HIV (73). Discovering how ATP13A2 and zinc influences MVBs and exosomes could well provide insights and potential therapeutic approaches not only for PD but also for a broad range of other diseases.
MATERIALS AND METHODS
Yeast culture conditions and spot assays for viability
Saccharomyces cerevisiae were cultured at 30°C in S media (0.5% w/v (NH4)2SO4, 0.17% w/v yeast nitrogen base with amino acid selection) with glucose or galactose (2% w/v). ypk9Δ cells were transformed with YPK9 galactose inducible over-expression plasmid (pBY011-YPK9) (9). Yeast media components were purchased from BD Sciences Difco (Australia).
Cell culture conditions
All cells were cultured at 37°C with 5% CO2. HEK293 and SHSY5Y cells were cultured in DMEM/F12; SHSY5Y-αSyn cells were cultured in RPMI1640, with 10% FBS and 5 mm Glutamax. α-Synuclein over-expression was repressed with 5 µm doxycycline. SHSY5Y and SHSY5Y-αSyn cells were differentiated in 2% FBS in culture media with 10 µm retinoic acid. Cells were treated with 100 µm CQ for 1.5 h. Cortical neurons were cultured on poly-L-lysine-coated coverslips with Neurobasal media and B27 supplement. DMEM/F12, RPMI1640, Neurobasal media and other tissue culture components were purchased from Gibco Life Technologies (Australia). FBS GOLD was purchased from Quantum Scientific VWR (Australia). SHSY5Y cells were from American Type Culture Collection (USA), HEK293FT cells were a kind gift from Professor David James (Garvan Institute of Medical Research, Australia) and SHSY5Y-αSyn cells were a kind gift from Dr Kostas Vekrellis (University of Athens, Greece).
Establishment and culture of hONs has been described (11). Briefly, primary olfactory cells were cultured from a nose biopsy of the proband and an age- and sex-matched control subject with enzymatic dissociation and BM cyclin treatment. Neurospheres were generated by growing primary olfactory cells on a poly-L-lysine-coated plate with supplement of 25 nm basic fibroblast growth factor, 50 nm epidermal growth factor and ITS and then cultured in DMEM containing 10% FBS and antibiotics.
ATP13A2 antibody production
Rabbit polyclonal antibodies were raised against human ATP13A2 (amino acids 72–256 that lacked identity with other human protein/P-type ATPases) and affinity-purified against recombinant ATP13A2 protein. Polyclonal antibody production was conducted by IMVS Veterinary Services (SA, Australia).
Production of ATP13A2 knockdown/over-expression cell lines
siRNA targeting PARK9 was purchased from Dhamacon (Australia). For the knockdown of ATP13A2, shRNA against ATP13A2 was cloned into the lentiviral vector pLL5.0 (74) to produce the plasmid pAC1085. For the over-expression of ATP13A2, full-length ATP13A2 [pENTR-ATP13A2 (9)] was sub-cloned into pLenti6/V5-Dest with gateway technology (Life Technologies, Australia). The empty pLL5.0 vector backbone was used create vector control cells. For lentivirus production, constructs and lentiviral packaging plasmids (pMD2G, pMDLg/pRE and pRSV.Rev) (75) were transfected into HEK293 cells using Lipofectamine 2000 (Life Technologies).
For the knockdown of PARK9, SHSY5Y cells were transfected with PARK9 siRNA by Amaxa electroporation nucleofection (Lonza, Australia). siRNA-mediated PARK9 knockdown was assessed by immunoblotting. For the creation of stable cell lines, lentivirus was collected and concentrated using Amicon column centrifugation. SHSY5Y and SHSY5Y-aSyn cells were infected with lentivirus for 24 h. Stably infected cells were selected for using 2 µg/ml blasticidin. Fluorescence-activated cell sorting (FACS) confirmation and western blots were conducted to assess PARK9 knockdown/over-expression.
Synchrotron elemental analysis of hONs
Sterilized silicon nitride windows (1.5 × 1.5 mm2 × 500 nm) (Silson, UK) were seeded with 3 × 104 hONS cells in 6-well plates and incubated for 24 h, as previously described (76,77). Cells were then treated with 0.1 mm ZnCl2 or 1 mm MnCl2 for 24 h before fixation. X-ray fluorescence elemental distribution maps of single cells were recorded on the XFM beam-line at the Australian Synchrotron (Victoria, Australia).
A monochromatic 9.9-keV X-ray beam was focused using a ‘high resolution’ zone plate, and the fluorescence signal was collected using a single-element silicon drift energy dispersive detector (Vortex EX, SII Nanotechnology, Northridge, CA, USA) for 1 s per spatial point. Individual cells were located using images obtained from an optical microscope positioned in the beam-line downstream from the sample.
The fluorescence spectrum at each spatial point was fit to Gaussians, modified by the addition of a step function and a tailing function to describe mostly incomplete charge collection and other detector artefacts. The analysis was performed using MAPS software (78) standardized by fitting to known spectra from NIST (National Bureau of Standards, Gaithersburg, MD, USA). Elemental area content (counts/cm2) are presented as the mean and the standard deviation in the mean.
SHSY5Y cells were seeded at 2 × 104 cells/well in 96-well plates, differentiated for 48 h then treated with various stress conditions for a further 48 h. Cell viability was assessed by alamarBlue (Life Technologies, Australia). hONs viability assays were performed in 24-well plates seeded at 5 × 104 cells/well in media for 24 h. hONs were grown in serum-free media for an additional 24 h, treated with ZnCl2 for 48 h, and cell viability was assessed by MTT assay.
SDS–PAGE gel and immunoblotting
Proteins were extracted with HES buffer and concentration determined by a BCA assay (Pierce, IL, USA). Protein samples were separated by SDS–PAGE and transferred onto nitrocellulose or PVDF membrane. Immunoblots were incubated with primary and secondary antibody before development with HRP chemiluminescence (Millipore, Australia). Proteins were extracted with HES buffer (2% w/v SDS, 20 mm HEPES, 1 mm EDTA, 250 mm sucrose, Roche cOmplete™ protease inhibitors, pH 7.4). Blots were blocked with 10% milk in TTBS (Tris-buffered Saline, 0.1% Tween) and incubated with primary antibodies for 24 h at 4°C followed by secondary antibodies for 1 h at room temperature. Blots were washed at least three times in TTBS after incubations.
Antibodies against α-Synuclein, β-Tubulin, CD63, LC3 and LAMP1 were purchased from BD Laboratories, Iowa Monoclonal Bank, BioLegend, MBL and Iowa Monoclonal Bank, respectively. Antibodies against LBPA were a kind gift from Dr Robert Parton (University of Queensland, Australia).
Indirect immunofluorescence microscopy
Cells cultured on coverslips were fixed in 4% PFA, blocked and incubated with primary and secondary antibody. Coverslips were washed after incubations and mounted onto slides for microscopy. For zinquin analysis, fixed coverslips were incubated with 20 µm zinquin for 30 min, washed and mounted for microscopy. Fluorescent images were taken on a Zeiss Upright microscope using Axiovision Version 4.8, or a N-SIM Super Resolution Microscope (Nikon) for Figs. 4B and 7B.
Zinquin fluorescence-activated cell sorting
HEK293 cells were seeded in triplicate at 20 × 104 cells per well in a six-well plate. Cells were incubated for 16 h at 37°C prior to transfection with appropriate plasmids. Cells were then incubated for a further 48 h at 37°C. Culture media was then replaced with fresh media containing zinquin at a final concentration of 20 µm and cells incubated for 1 h at 37°C. Cells were then harvested and resuspended in 500 µl FACS buffer (2 mm EDTA, 25 mm HEPES, 1% FBS in PBS), and 10,000 cells per sample were analysed in a 7 Laser SORP LSRII FACS machine (BD Biosciences). Zinquin was detected with a 355-nm laser, and GFP (the ATP13A2 shRNA knockdown plasmid shPK9 was marked with GFP) was detected with a 514-nm laser. Data were then analysed using BD coherent and BDFACSDiva software version 6.1.3 (BD Biosciences) and plotted as medium fluorescent intensity. For cells transfected with shPK9, zinquin fluorescence was only measured in cells that also displayed a GFP signal (an indication of shPK9 expression).
Exosome purification and density analysis
HEK293 cells were seeded into 15 cm dishes and then transfected with appropriate plasmids using Lipofectamine 2000 (Life Technologies). Media was changed 16 h after transfection to exosome-free medium (complete medium containing exosome-depleted FCS), and cells were grown for a further 48 h. Cell-free culture media was harvested as previously described (79). Culture supernatants were collected, and cellular debris was removed by centrifugation at 300 g for 10 min. The supernatant was centrifuged at 5000 g for 30 min at 4°C. The supernatant was collected and centrifuged at 100,000 g for 2 h to pellet exosomes. Exosome pellets were resuspended and washed in PBS before being re-centrifuged at 100,000 g to pellet exosomes. The final exosome-containing pellet was resuspended in PBS and used for either immunoanalysis or density gradient fractionation.
For gradient fractionation of 100,000 g exosomal pellets, a discontinuous step-gradient of 0–50% OptiPrep (AxisShield) was used. The 100,000 g exosomal pellets were resuspended in 50% OptiPrep and step-gradients of OptiPrep in 10% decreases were layered on top. The step gradients were centrifuged at 100,000 g for 18 h at 4°C in a SW41 rotor, and fractions of 400 µl were harvested. Each density gradient fraction was diluted with PBS before being re-pelleted at 100,000 g for 1 h, and the resulting exosomes pellets were resuspended in PBS for further analysis by immunoblot. An Auto-Densi Flow machine (Labconco, USA) was used to create and harvest all density gradients.
Imaging and statistical analysis
Densitometry and microscopy fluorescence intensity were conducted using Macbiophotonics ImageJ (http://www.macbiophotonics.ca/imagej). Neurite length was measured using Simple Neurite Tracer in Fiji (http://pacific.mpi-cbg.de/wiki/index.php/Fiji). Microscopy images were pseudo-coloured and overlayed using Photoshop (Adobe). Graphing and statistical analysis were conducted in Prism (GraphPad).
This work was supported by a project grant from the NHMRC. S.M.Y.K. was supported by the National Health and Medical Research Council and the Estate of the Late Olga Mabel Woolger. B.K.K.C. was supported by an Australian Post-graduate Award. J.B.A. was supported by the Australian Research Council and P.A.L. was supported by the Australian Research Council, including a Professorial Fellowship. C.M.S. is supported by a National Health and Medical Research Council Clinical Practitioner Fellowship and the Ramsay Health Teaching Research Fund. Work at the Australian Synchrotron was supported by the Australian Synchrotron Access Program.
We thank Peter Zalewski for zinquin, Eva Kovacs for pLL5.0 and shRNA advice, Tristan Tan for help with the Nikon N-SIM Super Resolution Microscope and Kostas Vekrellis for the SHSY5Y-αSyn cells and Ornella Tolhurst for help creating SHSY5Y-V5-tagged ATP13A2 cells. We thank Adam Cole and Hovik Farghaian for providing us rat primary neurons. We also thank the beam-line scientists Martin de Jonge, David Paterson and Daryl Howard at the Australian Synchrotron for their help.
Conflict of Interest statement. None declared.