Mutation in VPS35 associated with Parkinson’s disease impairs WASH complex association and inhibits autophagy

Endosomal protein sorting controls the localization of many physiologically important proteins and is linked to several neurodegenerative diseases. VPS35 is a component of the retromer complex, which mediates endosome-to-Golgi retrieval of membrane proteins such as the cation-independent mannose 6-phosphate receptor. Furthermore, retromer is also required for the endosomal recruitment of the actin nucleation promoting WASH complex. The VPS35 D620N mutation causes a rare form of autosomal-dominant Parkinson’s disease (PD). Here we show that this mutant associates poorly with the WASH complex and impairs WASH recruitment to endosomes. Autophagy is impaired in cells expressing PD-mutant VPS35 or lacking WASH. The autophagy defects can be explained, at least in part, by abnormal trafficking of the autophagy protein ATG9A. Thus, the PD-causing D620N mutation in VPS35 restricts WASH complex recruitment to endosomes, and reveals a novel role for the WASH complex in autophagosome formation.

T he retromer complex is a conserved membrane-associated protein complex that functions in the endosome-to-Golgi retrieval pathway. Retromer consists of a cargo-selective complex (CSC) comprising VPS35, VPS26 and VPS29, along with a sorting nexin dimer consisting of SNX1 or SNX2 with SNX5 or SNX6. Many membrane proteins (often referred to as 'cargo') depend on retromer for their respective localization 1 . A wellstudied cargo protein for retromer-mediated endosome-to-Golgi retrieval is the cation-independent mannose 6-phosphate receptor (CIMPR) that operates as a lysosomal hydrolase receptor, sorting acid hydrolases for exit from the trans-Golgi network (TGN) before returning to the Golgi via the endosometo-Golgi retrieval pathway 2,3 .
Recently, it has been shown that retromer function is important for more than endosome-to-Golgi retrieval. The retromer complex is responsible for the endosomal recruitment of the WASH complex 4 , a protein complex that, together with Arp2/3, mediates actin patch formation on endosomes to facilitate protein sorting 5,6 . WASH complex activity is especially important for the endosome-to-cell surface recycling of certain membrane proteins, including the b2-adrenergic receptor, a5b1 integrin, the glucose transporter GLUT-1, and the T-cell receptor in T-cells [7][8][9] , reviewed previously 10 . Cargo proteins that require the activity of the WASH complex for their trafficking are therefore indirectly dependent on retromer, even if retromer does not participate directly in their sorting. The recruitment of the WASH complex to endosomes is mediated by interactions between the extended tail of the FAM21 protein of the WASH complex and VPS35 in the retromer CSC [11][12][13] . In addition, a protein called FKBP15 (also known as FKBP133 and WAFL), which has been implicated in nerve growth-cone collapse, interacts with both the FAM21 tail and VPS35, although its function with the WASH and retromer complexes is unclear 4,11,14 .
Impaired endosomal protein sorting underlies many neurological diseases. One component of the WASH complex, strumpellin, is mutated in hereditary spastic paraplegia 15,16 . Mutation of another component of the WASH complex, KIAA1033 (also known as SWIP), results in autosomal recessive intellectual disability 17 . In addition, recent reports have revealed that Parkinson's disease (PD) can be caused by a rare autosomaldominant VPS35 mutation that results in an aspartate to an asparagine substitution at residue 620 (D620N) 18,19 . The mechanism through which the D620N mutation exerts its effects is presently unknown.
Macroautophagy (hereafter referred to as autophagy) is a degradative pathway by which the breakdown of cellular proteins releases amino acids necessary for survival, particularly under starvation or stress conditions 20 . Autophagy also removes damaged mitochondria and certain intracellular pathogens, such as Salmonella 21,22 . Moreover, autophagy is of profound importance in the clearance of intracytoplasmic aggregate-prone proteins, such as those formed by the polyglutamine-expanded repeats in huntingtin, the Huntington's disease-causing protein, as well as a-synuclein, the major component of PD-associated Lewy bodies [23][24][25] . In fact, autophagy appears to be compromised in the brains of PD patients 26 , and excess a-synuclein inhibits autophagosome formation 27 . The autophagic pathway is especially sensitive to perturbations of endosomal protein sorting, and thus makes a plausible downstream target of defects associated with retromer-mediated sorting. In fact, depletion of yeast retromer components impairs autophagic activity, although results in mammalian cells are at present unclear 28 .
Here we show that the PD-causing mutation in VPS35 destabilizes the retromer-WASH complex interaction, leading to reduced endosomal localization of the WASH complex. We further show that the WASH complex is necessary for autophagosome formation and that cells expressing the PDcausing VPS35 allele exhibit defects in autophagy, as well as in the trafficking of the multipass transmembrane autophagy protein ATG9A. This provides important mechanistic insights into the pathology of the PD-causing VPS35 mutation and demonstrates a novel role for the WASH complex in autophagy.

Results
VPS35 D620N exhibits impaired binding to the WASH complex. Figure 1a shows human VPS35 residues 610-680 aligned with VPS35 homologues from other well-studied eukaryotes. The D620N mutation falls within the horseshoe-shaped region that comprises many helical repeats and folds around VPS29 (ref. 29). We also highlight the recently identified H675R VPS35 mutation that abolishes binding to VPS29, an interaction that is required for the subsequent association of VPS35 and the WASH complex 12 . As the D620N mutation lies within the VPS29-binding region, we tested whether it affected the interaction between VPS35 and VPS29, as well as with other retromer-interacting proteins.
In Fig. 1b, cells stably expressing green fluorescent protein (GFP)-tagged wild-type (WT) VPS35 or VPS35 D620N were lysed along with untransfected HeLa cells and incubated with anti-GFP antisera (lanes [1][2][3]. Similarly, HeLa cells were transiently transfected with a panel of GFP-VPS35 constructs, and lysates were incubated with anti-GFP to recover the expressed protein (lanes [4][5][6][7][8][9]. Both stably and transiently transfected WT VPS35 coimmunoprecipitated the VPS26 and VPS29 retromer CSC subunits, along with strumpellin and WASH1 of the WASH complex as well as FKBP15 (lanes 2 and 4). The D620N mutant associated normally with both VPS26 and VPS29 but showed reduced associations with FKBP15 and the WASH complex proteins (compare lanes 2 with 3 and 4 with 6). The effect of the D620N mutant was not, however, as pronounced as that of the H675R mutant, which cannot bind VPS29 and completely fails to associate with the WASH complex (compare lanes 6 and 7). The VPS35 L108P mutation within the conserved PRLYL motif (lane 5) blocked binding to VPS26 (refs 30-32) but did not affect interactions with either VPS29 or the WASH complex 12 . Two truncation constructs of VPS35 confirmed that binding of the Figure 1 | Effect of the D620N mutation on the assembly and protein-protein interactions of the retromer CSC. (a) Schematic diagram showing an alignment of the human VPS35 sequence (residues 610-680) with homologues from other well-studied eukaryotes. The D620N mutation is highlighted along with the H675R mutation that blocks binding to VPS29. The position of the D620N mutation of VPS35 is shown on the three-dimensional structure along with the H675R mutation and the V90 and I91 residues of VPS29 that mediate binding to VPS35 (PDB ID. 2R17). (b) Cells stably expressing WT or D620N GFP-VPS35 or otherwise transiently transfected with a panel of GFP-VPS35 constructs were lysed and the respective GFP-tagged protein recovered by immunoprecipitation (IP). (c) Untransfected HeLa cells or cells stably expressing VPS29-GFP, GFP-VPS35 wild-type or GFP-VPS35 D620N were lysed and the lysates incubated with anti-VPS26 to IP the retromer CSC. (d) Cells stably expressing the VPS29-GFP V90D mutant that cannot assemble with VPS35 were transiently transfected with mCherry-tagged WT VPS35, D620N and H675R constructs, along with an additional control of mCherry-FAM21 tail. Cells were lysed 48 h post transfection, and VPS29-GFP V90D and associated proteins were recovered by anti-GFP native IP. Blots shown are representative of experiments replicated at least twice.
WASH complex and FKBP15 to VPS35 occurs independently of VPS26 but requires the carboxy-terminal region of VPS35 and proper interaction with VPS29.
We next used antisera against VPS26 to immunoprecipitate the retromer CSC and determine the effect the D620N mutation on the association of the CSC with the WASH complex.
Components of the WASH complex (FAM21, strumpellin and WASH1) were readily detected in association with the retromer CSC in HeLa cells, as well as cells stably expressing WT VPS29 or WT VPS35, but not in cells expressing VPS35 D620N (Fig. 1c). The reduced level of endogenous VPS29 that co-precipitated using the anti-VPS26 antibody in the VPS29-GFP lysate was due The data presented in Fig. 1 suggest that the VPS35 D620N mutation does not affect VPS29 binding or CSC assembly, but reduces interaction with the WASH complex and FKBP15. To further explore these effects, cells stably expressing WT VPS35 or the D620N mutant were lysed using buffers of differing stringencies. The HEPES-based buffer has lower stringency, while the phosphate-buffered saline (PBS) lysis buffer has a higher stringency, as it contains more salt and detergent. Following lysis, samples of equal volume of the HEPES and PBS lysates were mixed to produce a lysate of intermediate stringency. To ensure that the GFP-tagged constructs could associate with FKBP15 or the WASH complex without competition, small interfering RNA (siRNA) was used to silence the expression of endogenous VPS35. The GFP-tagged murine VPS35 constructs are resistant to the siRNA targeting endogenous VPS35. As before, WT VPS35 coimmunoprecipitated WASH complex components WASH1, FAM21, strumpellin, as well as FKBP15, with some reduction in the levels of WASH complex proteins and FKBP15 in the higher stringency PBS buffer (Fig. 2a). The D620N mutant, however, exhibited a marked reduction in both WASH complex and FKBP15 binding in the intermediate stringency buffer (H/P) and no association in the higher stringency PBS buffer (P). Levels of VPS26 and VPS29 did not change significantly, indicating that there was no destabilization of the retromer complex due to either the D620N mutation or the different lysis buffers.   The sensitivity of the retromer-WASH complex association to different lysis buffer conditions was confirmed in Supplementary  Fig. 1b, where cells stably expressing either GFP-tagged VPS29 or the FAM21 tail were lysed with either HEPES, PBS or the mixed lysis buffer prior to immunoprecipitation (IP) with anti-GFP. In this experiment, interactions between retromer and both the WASH complex and FKBP15 are lost with increasing lysis buffer stringency, but the interactions that underpin assembly of the retromer CSC are retained.
VPS35 binding to WASH complex is independent of FKBP15. The D620N mutation affects binding of VPS35 to the WASH complex and almost abolishes association with FKBP15. To determine if loss of the VPS35-FKBP15 interaction is destabilizing the retromer-WASH complex interaction, cells expressing WT GFP-VPS35 were treated with siRNA to silence FKBP15 expression. Lysates were then generated using the buffers of differing stringency and GFP-tagged VPS35 recovered by IP. In Fig. 2b, there was no pronounced change in the association of the WASH complex with WT GFP-VPS35 after FKBP15 depletion, indicating that the D620N-mediated loss of FKBP15 interaction with VPS35 does not cause the defective WASH complex association with the D620N VPS35 mutant.
To further delineate the defects associated with the D620N mutant, the cells stably expressing GFP-tagged WT or D620N VPS35 were treated with siRNA to silence the expression of either endogenous VPS35, FKBP15 or the FAM21 or KIAA1033 components of the WASH complex. It has been shown previously that knockdown of one WASH complex protein affects the stability and levels of the others 11,34 . Following knockdown, lysates were incubated with anti-GFP to recover VPS35 (Fig. 2c). For this experiment, twice the amount of material used in previous experiments was analysed using the HEPES-based buffer to create conditions that can be considered the most permissive for retaining retromer-WASH complex interactions. WT VPS35 interacted with the WASH complex in the absence of FKBP15, as demonstrated by the presence of strumpellin and WASH1 in lane 3. Similarly, FKBP15 co-immunoprecipitated with WT VPS35 after siRNA knockdown of FAM21 or KIAA1033, although at reduced levels (see lanes 4 and 5). In contrast, the D620N mutant failed to co-immunoprecipitate FKBP15 in cells where the WASH complex was silenced but retained an interaction with WASH complex proteins when FKBP15 expression was abolished. Thus, binding of the WASH complex to retromer appears to occur largely independently of FKBP15. This is consistent with co-IP of WASH components by GFP-tagged VPS29 after knockdown of FKBP15, as seen in Supplementary Fig. 1c and in previous work 4 .
Although FKBP15 is not strictly necessary for the retromer-WASH complex association, increasing levels of FKBP15 by transient transfection can enhance the retromer-WASH complex interaction in the higher stringency PBS lysis buffer for WT VPS35 but not for the D620N mutant ( Supplementary Fig. 1d).
VPS35 D620N impairs WASH complex recruitment to endosomes. Retromer mediates the endosomal recruitment of both the WASH complex and FKBP15 (ref. 4). VPS35 recruits the WASH complex through interactions with the FAM21 tail [11][12][13] . To test the effect of the D620N mutation on the localization of FKBP15 and the WASH complex, cells stably expressing either WT GFP-VPS35 or the D620N mutant were treated with siRNA to silence endogenous VPS35 expression. After fixation, the cells were double-labelled with a monoclonal anti-GFP antibody and antisera against FKBP15 or FAM21 (Fig. 3a,b). Quantification of the fluorescence intensities indicates that the D620Nexpressing cells have reduced endosomally localized FKBP15 and FAM21 compared with the WT cells, compatible with decreased endosomal recruitment (Fig. 3c).
To more rigorously examine the effect of the D620N mutation on the membrane association of FKBP15 and the WASH complex, we separated cytoplasmic proteins in the supernatant (S) from membrane and membrane-associated proteins in the pellet (P) fraction by centrifugation. Cells expressing the D620N mutant had less membrane-associated FKBP15 and strumpellin both in the presence of endogenous VPS35 and most significantly when endogenous VPS35 was silenced (Fig. 3d,e). The cells expressing WT VPS35 showed some reduction in the membrane association of FKBP15 and strumpellin when endogenous VPS35 was silenced, possibly due to minor inhibitory effects of the GFP moiety added to the amino terminus of VPS35. These data, taken together, demonstrate that the D620N mutation results in a reduction of the membrane association of FKBP15 and the WASH complex.
VPS35 D620N expression mimics WASH-associated phenotypes. Thus far, we have reported an impaired association of the D620N mutant for the WASH complex and FKBP15 but the functional significance has not been elucidated. Previous experiments have shown that siRNA knockdown of FKBP15 does not impede endosome-to-Golgi retrieval of a CD8-CIMPR chimera reporter protein 4 . Similarly, D620N-expressing cells exhibited apparently normal VPS26, SNX1, CIMPR, and LAMP1 localization ( Supplementary Fig. 2a-c).
We observed a cell spreading defect in cells expressing VPS35 D620N, compared with their WT counterparts ( Supplementary  Fig. 3a), which we confirmed in independent clones. Cell spreading defects also occur in cells overexpressing the FAM21 tail, which competes with the endogenous WASH complex for binding to retromer 11 or WASH1 knockdown cells 9 . Thus, the cell spreading defect in the VPS35 D620N cells is consistent with the reduced affinity of the mutant VPS35 for the WASH complex, resulting in WASH complex mislocalization. While CIMPR and retromer exhibited normal localization in the D620N-mutant cells, we observed altered localization of GLUT-1, membrane protein that requires the WASH complex for its steady-state localization to the cell surface 7,35 . The cells expressing the VPS35 D620N mutant displayed reduced cell surface staining of GLUT-1 and increased intracellular staining ( Supplementary Fig. 3b). As the input lysates in Figs 1,2 showed that WASH1 levels remain essentially the same in both the VPS35 WT and D620N cell lines, the WASH-related phenotypes observed in D620N cells are not due to changes in the steady-state level of WASH1 but likely due to WASH complex mislocalization.
VPS35 D620N impairs autophagy. Proper endosomal function is important for autophagy, prompting us to test if the VPS35 D620N mutation affects this process. Autophagy can be measured by examining levels of LC3-II, the only protein known to associate specifically with autophagosomes, as its levels correlate with autophagosome numbers. Because changes in LC3-II levels and autophagosome numbers may be due to altered autophagosome formation or degradation, one can uncouple these processes using drugs like bafilomycin A1, which blocks autophagosome degradation 36,37 , thus allowing assessment of changes in formation alone. Interestingly, the stable GFP-VPS35 D620N cell lines showed lower LC3-II levels than their WT counterparts both with and without bafilomycin, indicating defective autophagosome formation (Fig. 4a,b). Furthermore, Supplementary Fig. 4a,b replicates this result in independent clones of GFP-VPS35 WT and D620N cells. In addition, knocking down endogenous VPS35 in both sets of clones stably expressing GFP-VPS35 wild type and D620N did not appear to alter the levels of LC3-II, as the difference between WT and mutant cells remained pronounced, regardless of the presence of endogenous VPS35 ( Supplementary Fig. 4c-e). Upon transient transfection of a fluorescent LC3 construct, we observed fewer punctate LC3 structures in cells expressing GFP-VPS35 D620N, compared with wild-type, indicating fewer autophagosomes (Fig. 4c).
To confirm a functional defect in autophagy, we transfected the cell lines with a hemagglutinin (HA)-tagged aggregation-prone model autophagy substrate, huntingtin exon 1 containing 74 glutamine repeats. The percentage of transfected cells with huntingtin aggregates correlates directly with huntingtin abundance, which is, in turn, autophagy dependent 23 . The D620N-mutant cells consistently had a higher percentage of transfected cells containing aggregates, consistent with impaired autophagosome formation and autophagic clearance (Fig. 4d,e).
VPS35 depletion by RNA interference only resulted in modest changes in LC3-II levels, as seen in Fig. 4h-j, consistent with recent data using VPS29 depletion 38 . This is likely because loss of VPS35 destabilizes the entire retromer CSC and has wide-ranging effects on multiple pathways, including defects in CIMPR retrieval that impact cathepsin D trafficking and lysosome biogenesis. Consequently, VPS35 depletion will have pleiotropic consequences on endosomal protein sorting, resulting in the possible masking of the autophagosome formation defect by other compensatory effects.
Perturbation of WASH complex binding disrupts autophagy. As VPS35 D620N exhibits decreased affinity for the WASH 2,000 ARTICLE complex, we mimicked this reduced binding by overexpressing the FAM21 tail, which displaces the endogenous WASH complex from retromer [11][12][13] . FAM21 tail overexpression reproducibly reduced LC3-II levels both with and without bafilomycin (Fig. 5a,b). The effect was not as pronounced as that in the stably expressing D620N cells, as FAM21 tail was transiently transfected and hence not present in every cell.
In an epistasis experiment where the FAM21 tail was expressed in the VPS35 stable cell lines, we observed that overexpression of the FAM21 tail decreased autophagosome formation (as indicated by reduced LC3-II levels under bafilomycin-treated conditions) in WT cells but not in D620N-mutant cells (Fig. 5c,d). In the mutant line, WASH complex binding to retromer is already impaired and thus overexpression of the FAM21 tail has no additive influence, suggesting that the observed effects on autophagosome formation are indeed due to the decreased association between VPS35 and the WASH complex. The importance of WASH complex activity for autophagosome formation was underscored by finding that siRNA knockdown of WASH1 reduced LC3-II levels with and without bafilomycin (Fig. 5e,f). In addition, depletion of WASH1 leads to fewer GFP-LC3 vesicles, again pointing to the impaired autophagosome formation (Fig. 5g).
While the VPS35 D620N mutation hinders retromer binding not only to the WASH complex but also to FKBP15, knockdown of FKBP15 had a much less robust effect on autophagy than WASH1 depletion, as evidenced by only modest, non-statistically significant effects on endogenous LC3-II levels, no difference in GFP-LC3 vesicle numbers, and no difference in the percentage of cells with mutant huntingtin aggregates (Fig. 5h-j). This is consistent with the hypothesis that FKBP15 is not an essential part of the complex and that that the WASH complex can still bind to retromer in the absence of FKBP15 (Fig. 2b,c).
VPS35 D620N and WASH depletion affect trafficking of ATG9A. The autophagy pathway requires multiple autophagyrelated genes, including ATG9A, a multipass transmembrane protein. Under basal conditions, mammalian ATG9A localizes to the TGN and recycling endosomes, and redistributes to a more peripheral pool after autophagy induction by nutrient deprivation (reviewed by Zavodszky et al. 39 ).
We observed that some ATG9A colocalized with VPS35, which is known to reside primarily on sorting endosomes, suggesting that ATG9A may traffic through compartments at which retromer acts (Fig. 6a). A proximity ligation assay, which yields a signal when two proteins of interest are within 40 nm of each other, indicated that ATG9 and VPS35 are found in the same compartment ( Supplementary Fig. 5a). A signal is observed in untransfected cells, as well as cells stably expressing WT and D620N VPS35. The signal, although weak when compared with the signal obtained for anti-VPS35 and anti-VPS26, is specific and does not occur when anti-VPS35 is used alone. Furthermore, ATG9A colocalizes in part with WASH complex components at endosomes, thus suggesting that its trafficking may be perturbed by the WASH complex mislocalization associated with the VPS35 D620N mutation (Fig. 6b,c and Supplementary Fig. 5b).
In cells stably expressing WT GFP-VPS35, ATG9A colocalized with both the TGN marker TGN46 and VPS35 (Fig. 6d,  Supplementary Fig. 6a, Supplementary Movies 1,2). However, the D620N cells showed increased colocalization between ATG9A and the TGN (Fig. 6e). In the D620N cells, the morphology of the TGN-and with it, ATG9A-changed dramatically from clusters to elongated perinuclear structures that were often circular or semicircular in shape. The aberrant TGN and ATG9A morphology in D620N cells was maintained under starvation conditions, and was present regardless of whether or not endogenous VPS35 has been depleted. Throughout a 2 h starvation period, ATG9A gradually dispersed from its basal perinuclear location in WT cells and fragmentation of the TGN was observed, consistent with previous reports 40 ( Supplementary  Fig. 6b). However, in D620N-expressing cells, ATG9A and the TGN were maintained in the perinuclear area at all time points and did not undergo fragmentation. Aberrant Golgi morphology was confirmed with an independent TGN marker, Golgin-97, as well as the cis-Golgi marker GM130 ( Supplementary Fig. 7a,b). The gross morphological changes of the TGN and ATG9A and their increased colocalization both suggest that normal trafficking of ATG9A is perturbed in D620Nexpressing cells, and perhaps ATG9A is trapped in an abnormal perinuclear compartment.
Upon knockdown of WASH1 itself, colocalization of ATG9A with TGN46 once again was increased in the knockdown cells when compared with control ( Fig. 7a,b, Supplementary Fig. 8a and Supplementary Movies 3-6). TGN46 and ATG9A morphology were also altered, although instead of elongated structures, we observed that ATG9A and TGN46 resided in tight clusters and remained so even upon starvation. This contrasts with control cells, where ATG9A noticeably dispersed upon starvation, and some fragmentation of the TGN46 marker was observed as well. Once again, other markers confirm an alteration of Golgi and TGN morphology ( Supplementary Fig. 8b,c). While the phenotype of WASH1-depleted cells does not perfectly mimic the morphological disturbance observed with VPS35 D620N, it is important to note that the PD mutation does not abolish all WASH complex binding to retromer, and it also does not result in a depletion of WASH1 levels. Thus, the scenarios are not identical, but both serve to illustrate that ultimately, the activity of correctly localized WASH1 is necessary for proper ATG9A localization and trafficking. These effects on ATG9A localization may contribute to the defective autophagosome formation caused by the D620N mutation, as previous studies have demonstrated that genetic and pharmacological manipulations that impede ATG9A redistribution from perinuclear compartments inhibit autophagy 39 .
ATG9A acts early in the autophagy pathway and traffics to compartments that become LC3-positive to interact with phagophores and autophagosomes 41 . ATG9A is thought to deliver membrane to growing phagophores, and its depletion has detrimental effects on autophagy initiation 41,42 . In cells expressing WT GFP-VPS35, many of the LC3-positive autophagic structures also contained ATG9A (Fig. 8a,b,  Supplementary Fig. 9a and Supplementary Movies 7,8). However, a significantly smaller proportion of LC3 vesicles contained ATG9A in cells expressing the D620N-mutant form of VPS35. WASH1 depletion had a similar effect, as knockdown of this protein resulted in fewer GFP-LC3 puncta containing ATG9A (Fig. 8c,d, Supplementary Fig. 9b and Supplementary Movies 9,10). Our results indicate that the VPS35 D620N mutation and WASH1 depletion both impair ATG9A trafficking to autophagosomes.
Although ATG9A mislocalization is a likely contributor to the autophagosome formation impairment seen in VPS35 D620N cells and WASH1-depleted cells, it is unlikely to be the only mechanism by which autophagy is compromised. We examined another early key autophagy protein, ATG16L1, and found no gross alterations in the distribution or appearance of mStrawberry-ATG16L1positive vesicles in either VPS35 D620N cells or WASH1 knockdown conditions (Supplementary Fig. 10) However, it is interesting to note that in VPS35 D620N cells, we observed a decrease in the proportion of ATG16L1 vesicles containing ATG9A, echoing the phenotype observed with the diminished ability of LC3 vesicles to acquire ATG9A. A similar, although non- significant trend was observed upon WASH1 knockdown. Recent work has suggested that ATG9A and ATG16L1 vesicles, which enter the cell via distinct endocytic pathways, coalesce in recycling endosomes 43 . In fact, the recycling endosome marker RAB11 also shows altered morphology in both VPS35 D620N cells and WASH1-depleted cells (Supplementary Fig. 11).
Autophagy is known to protect against pro-apoptotic insults, and autophagy disruption predisposes to cell death 44 . A previous study indicated that the D620N VPS35 mutation is associated with impaired protection of neuronal cells from the mitochondrial toxin MPP þ (ref. 45). We found that knockdown of WASH1 in SH-SY5Y neuroblastoma cells decreases autophagosome formation and increases cell death (Fig. 9).

Discussion
The retromer complex recruits the WASH complex to endosomes through VPS35 binding to the FAM21 protein 4,[11][12][13] . Our data suggest that a major molecular effect of the VPS35 D620N mutation is to reduce retromer association with the WASH complex as well as FKBP15, without affecting CSC assembly. The WASH complex functions to promote F-actin patch formation on endosomes and its activity is associated with endosomal protein sorting of multiple cargo proteins 10 . Indeed, we find that cells with the D620N VPS35 mutation have decreased membraneassociated WASH complex components and phenotypes such as impaired cell spreading and GLUT-1 mislocalization, seen in cells with compromised WASH-complex function 9,11 .  in (a,c), 10 mm. In b,d, colocalization was measured for the entire area of the cell slice in each image. In order to ensure that results were not confounded by diffuse nuclear staining often observed in the LC3 channel, the analyses were repeated with the nuclei excluded, and all of the results remained statistically significant.
The most obvious disease-relevant phenotype that we associated with the D620N mutation is impaired autophagosome formation. Similar autophagy defects were seen in WASH1 knockdown cells or in cells overexpressing the FAM21 tail, which inhibits association of the WASH complex with VPS35 by competition [11][12][13] . While disturbed retromer-WASH complex association may have multiple consequences for autophagosome formation, we have shown that the D620N mutation, as well as WASH1 depletion, affects ATG9A localization and trafficking. We recently showed that ATG9A is internalized from the plasma membrane by clathrin-mediated endocytosis and travels via early endosomes to the recycling endosome 43 . The data shown here indicate that the VPS35 D620N mutation or WASH1 depletion result in ATG9A becoming trapped in a perinuclear compartment positive for a TGN marker. It is, however, difficult to distinguish between proteins in the TGN and recycling endosomes even by electron microscopy 46 , raising the possibility that the ATG9A is retained in the recycling endosome and is unable to continue its normal trafficking to phagophores and autophagosomes. In fact, the recycling endosome marker RAB11 also shows perturbed morphology in both VPS35 D620N-expressing and WASH1depleted cells. Alterations in ATG9A trafficking result in a smaller proportion of LC3-positive autophagic structures acquiring ATG9A and likely contributes to the decreased autophagy phenotype. Other perturbations that cause similar ATG9A trafficking defects also impair autophagosome formation 40,[47][48][49] . Interestingly, autophagosome formation deficits associated with defective ATG9A trafficking are also seen with a-synuclein overexpression, which models other autosomal-dominant forms of PD caused by a-synuclein gene duplications 27 . Although no obvious changes were observed in the distribution of another important early autophagy protein, ATG16L1, it is possible that other autophagy components might be affected.
Recent studies have reported that the D620N mutation in VPS35 can alter CIMPR localization and impair cathepsin D trafficking, resulting in lysosomal enlargement 50,51 . We found no alterations in CIMPR or LAMP1 staining between cells stably expressing WT or D620N-mutant VPS35. Furthermore, the decreases in LC3-II levels in the VPS35 D620N cells point to defective autophagosome formation, rather than degradation, which would be expected with lysosome perturbations. It is nonetheless possible that the VPS35 D620N mutation leads to additional phenotypes-including those affecting lysosomal function and/or phenotypes specific to neuronal cells-that are not explored here.
Our data suggest that the D620N mutation is a partial loss of function mutation of just one role of VPS35: that of binding the WASH complex and FKBP15, and not a generalized VPS35 hypomorph. Unlike a knockdown of VPS35 that abolishes retromer function and leads to mislocalization of many proteins that traffic between the endosome and either Golgi or cell surface, the D620N mutation appears to affect only those aspects of retromer function that are linked to the WASH complex. The GLUT-1 protein is especially dependent on the activity of the WASH complex for its localization 7,35 and hence localization of GLUT-1 is perturbed in cells expressing the D620N mutation. Similarly, cell spreading is also tightly linked to WASH complex function and the D620N cells exhibit a cell spreading defect 9,11 . The retromer complex can still function in endosome-to-Golgi retrieval without the WASH complex, as demonstrated in yeast where the WASH complex is not conserved and yet retromer works to retrieve Vps10p from endosomes to the Golgi.
Patients carrying this D620N mutation only manifest disease in their 6th decade 18,19 . If the D620N mutant were to drastically affect the localization of a great many membrane proteins, one would expect a more severe disease and earlier onset. Nonetheless, we cannot formally exclude that the D620N mutation has other effects in addition to impaired WASH complex binding and a resultant autophagy deficiency, and that some of these unexplored phenotypes could be due to a gain of function of the mutant protein.
The VPS35 mutation also exhibits defective binding to FKBP15 and our data are consistent with the hypothesis that the retromer CSC, WASH complex and FKBP15 form a tripartite complex. The precise function of FKBP15 remains to be determined and is likely to be multifaceted 52,53 . Unlike the WASH complex, FKBP15 is not conserved in either Caenorhabditis elegans or Drosophila melanogaster 4 . Our data suggest that FKBP15 is unlikely to be an important contributor to the autophagy phenotype. Nevertheless, it is conceivable that the loss of retromer-FKBP15 interaction resulting from the D620N mutation may contribute to the pathology of PD in other ways.
Interestingly, another PD gene, DNAJC13 (ref. 54), encodes a protein that has recently been shown to associate with the FAM21 protein of the WASH complex-the same subunit known to interact with VPS35 (ref. 55). Thus, the WASH complex associates with two known PD proteins, suggesting that WASH complex function may be especially important for the pathology of PD.
In conclusion, our data show that the D620N PD-causing VPS35 mutation impairs binding to the WASH complex and ARTICLE FKBP15. We found that loss of WASH complex binding was likely responsible for the impaired autophagosome formation resulting from the D620N VPS35 mutation. This autophagy defect was associated with abnormal ATG9A trafficking. Impaired autophagy is a highly plausible contributor to PD pathogenesis, as it would exacerbate many of the pathologies associated with PD, including protein aggregation, mitochondrial abnormalities, increased levels of reactive oxygen species and enhanced susceptibility to cell death.

Methods
Mammalian cell culture. HeLa and SH-SY5Y cells were maintained in DMEM containing 4,500 mg l À 1 of glucose, and supplemented with 10% fetal bovine serum (FBS), 100 units ml À 1 penicillin, 100 mg ml À 1 streptomycin and 2 mM Lglutamine (all from Sigma-Aldrich) at 37°C and 5% CO 2 . HeLa cells stably expressing GFP-VPS35 constructs, GFP-FAM21 tail or VPS29-GFP were generated by the Seaman Laboratory and cultured in DMEM containing 4,500 mg l À 1 of glucose, and supplemented with 10% FBS, 100 units ml À 1 penicillin, 100 mg ml À 1 streptomycin, 2 mM L-glutamine and 500 mg ml À 1 G418 (Gibco RNA interference-mediated silencing. In IP experiments, the silencing of VPS35, WASH complex components and FKBP15 was achieved using siRNA following previously described knockdown protocols 11  DNA constructs. GFP-tagged VPS35 has been described previously 56 . The D620N mutation was engineered into murine VPS35 in pEGFP C1 using the QuikChange kit (Stratagene) and presence of the mutation was confirmed by DNA sequencing. The construct was subcloned into pmCherry to generate the mCherry-tagged version. The GFP-tagged FAM21 tail construct has also been described previously 11 . The first exon of the huntingtin protein with 74 polyglutamines, tagged with EGFP or HA at the N terminus, in pEGFP-C1 vector (GFP-Q74) or pHM6 vector (HA-Q74) has been described previously 23,57 . GFP-a-synuclein has been described previously 58 , as has mCherry-RAB11 (ref. 43). ATG9A-GFP was a kind gift of Yoshinori Takahashi and was described previously 40 . mStrawberry-ATG16L1, GFP-LC3 and mRFP-LC3 were kindly provided by Tamotsu Yoshimori and were described previously 59,60 . The mRFP-RAB5 construct was described by Vonderheit  Western blotting. Detection for western blotting of native IP samples and accompanying lysates was accomplished using 125 I-protein A and exposure to X-ray film as described previously 62 . In autophagy assays, after lysing cells in sample buffer (62.5 mM Tris pH 6.8, 2% w/v SDS, 10% glycerol, 50 mM DTT, 0.01% w/v bromophenol blue) western blot analysis was performed using standard techniques with an ECL western blotting substrate (Thermo Fisher Scientific) or with direct infrared fluorescence detection on an Odyssey Infrared Imaging System. For ECL western blotting, the following horseradish peroxidase-conjugated secondary antibodies were used: sheep anti-mouse (GE Healthcare, NA931V); donkey anti-rabbit (GE Healthcare, NA934V); rabbit anti-goat (Life Technologies, 61-1620). For infrared fluorescence detection, IRDye-conjugated secondary antibodies were obtained from LI-COR (926-32220, 926-32211, 926-32210). Whole blot images are presented in Supplementary Fig. 12.
Lysates were centrifuged, precleared with 50 ml of a 25% slurry of protein A-sepharose for 30 min and subsequently incubated with antibodies for 90 min.
The mix was incubated with protein A-sepharose for 60 min, after which sepharose beads were recovered by centrifugation, washed four times in lysis buffer, dessicated in a SpeedVac (Eppendorf), resuspended in SDS-PAGE sample buffer and boiled for 5 min.
Membrane association assay. Determination of the amount of membraneassociated strumpellin and FKBP15 was performed using a simple fractionation assay described previously 4,11,63 .
Immunofluorescence. Unless otherwise indicated, cells were cultured for immunofluorescence microscopy on glass coverslips, fixed with 4% paraformaldehyde in PBS for 5 min, permeabilized with 0.1% Triton X-100 in PBS for 5 min, and washed with 0.2% glycine in PBS twice for 5 min each. Coverslips were blocked in 10% FBS in PBS, and subsequently incubated with primary antibodies for 2 h, washed three times with PBS and incubated with Alexa Fluor-conjugated secondary antibodies (Life Technologies) for 30 min. Samples were mounted using ProLong Gold Antifade reagent with 4 0 ,6-diamidino-2-phenylindole (Life Technologies) and imaged at 37°C with a Zeiss LSM 710 laser confocal microscope using a Plan-Apochromat Â 63/1.40 Oil DIC M27 lens and ZEN acquisition software, or in the case of Supplementary Fig 7b, with a Zeiss LSM510 laser confocal microscope using a Â 100 oil immersion lens and LSM510 acquisition software (Carl Zeiss). Images shown represent individual confocal slices, unless stated otherwise. Compressed z-stacks and three-dimensional movies were created with Volocity software (Perkin Elmer). Colocalization was determined by calculating the Pearson's coefficient or Mander's coefficient M1/M2 by tracing individual cells with Volocity software (Perkin Elmer). Thresholds were kept constant within each experiment. For Fig. 3 and Supplementary Fig. 2, the following protocol was employed: cells were fixed for 15 min using 4% paraformaldehyde in PBS, permeabilized with 0.1% Triton X-100 in PBS for 10 min prior to blocking with 3% bovine serum albumin (BSA) in PBS for 20 min. Cells were labelled with antibodies diluted in blocking buffer (3% BSA in PBS) for 1 h, washed and then incubated with Alexa Fluor-conjugated secondary antibodies (Life Technologies) diluted in blocking buffer for a further hour. The cells were mounted using ProLong Gold Antifade reagent (Life Technologies) and viewed at room temperature using a Plan-Apochromat Â 63 lens with oil immersion on a Zeiss AxioPlan epifluorescence microscope equipped with a Hamamatsu CCD camera and SimplePCI acquisition software (Hamamatsu). To quantify the amount of endosomally localized FAM21 and FKBP15, cells were imaged using a Cellomics ArrayScan VTI HCS Reader and the Spot Detector application (Thermo Fisher Scientific) and FKBP15 and FAM21 spot intensity was normalized to GFP-VPS35 and VPS26 signals. The Cellomics Spot Detector application was also used to quantify the number of GFP-LC3 and mRFP-LC3 vesicles per cell.
Proximity ligation assay. The proximity ligation assay kit was obtained from Sigma and used according to manufacturer's instructions. Briefly, cells were seeded onto 13 mm coverslips and allowed to grow in culture for 24 h. After fixation with 4% paraformaldehyde in PBS, permeabilization with 0.1% Triton X-100 in PBS and blocking with 3% BSA in PBS, the cells were incubated with anti-VPS35 and either anti-VPS26 or anti-ATG9A. Following incubation with the primary antibodies, the cells were incubated with secondary antibodies conjugated to oligonucleotide primers. The primers were ligated and then rolling circle amplification was used to create a reaction product that is observable by microscopy due to hybridization of fluorescently labelled nucleotides. Successful production of a DNA product requires that the primary antibodies bind their respective antigens and reside within 40 nm of each other. Coverslips were mounted on slides and imaged by epifluorescence microscopy, as described in the previous section.
Cell spreading. Cells were plated on glass coverslips and fixed in 4% paraformaldehyde in PBS for 5 min at 2, 4 and 6 h after plating. Cells were permeabilized in 0.1% Triton X-100 in PBS for 5 min, and incubated with Alexa Fluor 594conjugated phalloidin for 30 min to mark the cell periphery. The cell perimeter was measured using Cellomics ArrayScan VTI HCS Reader and the Cell Spreading application (Thermo Fisher Scientific). Bright-field images of cells were taken immediately prior to fixation with an EVOS FL Cell Imaging System (Life Technologies).
Aggregation. Cells were plated on glass coverslips and transfected with GFP-Q74 or HA-Q74 for 24 h before fixation with 4% paraformaldehyde in PBS. Coverslips expressing GFP-Q74 were mounted ProLong Gold Antifade reagent with 4 0 ,6diamidino-2-phenylindole (Life Technologies), while coverslips with HA-Q74 were subjected to immunofluorescence against HA before mounting. Using a Nikon Eclipse E600 fluorescence microscope, an experimenter blinded to the identity of the slides scored the percentage of transfected cells containing aggregates. Experiments were performed in triplicate and a minimum of 200 cells were counted in each replicate.
a-synuclein degradation. Cells expressing GFP-VPS35 wild type or D620N were plated in six-well plates and transfected with 1 mg GFP-a-synuclein A53T and 0.5 mg GFP DNA per well. Forty-eight hours after transfection, cells were lysed in sample buffer and the lysates analysed by western blotting. GFP-a-synuclein levels were expressed as a ratio with GFP levels.
Cell survival. SH-SY5Y neuroblastoma cells were depleted of WASH1 using two rounds of siRNA treatment, as described above. Seventy-two hours after the second treatment, trypsinized cells and their growth medium from the last 24 h (containing any detached cells) were spun down and the cells resuspended in 500 ml of PBS on ice. Viability was determined by staining with propidium iodide ReadyProbes reagent (Life Technologies). Cells were analysed by a BD FACSCalibur flow cytometer (BD Biosciences) and Flow-Jo software (Treestar) in the FL2 channel, and cells positive for propidium iodide were classified as dead.
Statistics. Significance levels for comparisons between groups were determined by Student's t-tests.