Unbiased screen for interactors of leucine-rich repeatkinase 2 supports a common pathway for sporadic andfamilial Parkinson diseaseAlexandria Beilinaa, Iakov N. Rudenkoa, Alice Kaganovicha, Laura Civierob, Hien Chauc, Suneil K. Kaliad,Lorraine V. Kaliae, Evy Lobbestaelf, Ruth Chiaa, Kelechi Ndukwea, Jinhui Dingg, Mike A. Nallsh, International Parkinson’sDisease Genomics Consortium1, North American Brain Expression Consortium1, Maciej Olszewskii, David N. Hausera,j,Ravindran Kumarana, Andres M. Lozanod, Veerle Baekelandtf, Lois E. Greenei, Jean-Marc Taymansf, Elisa Greggiob,and Mark R. Cooksona,2

aCell Biology and Gene Expression Section, gComputational Biology Core, Laboratory of Neurogenetics, and hMolecular Genetics Section, National Institute onAging/National Institutes of Health, Bethesda, MD 20892; bDepartment of Biology, University of Padova, 35131 Padua, Italy; cDivision of Brain, Imaging, andBehaviour Systems Neuroscience, dDivision of Neurosurgery, Department of Surgery, and eDivision of Neurology, Department of Medicine, Toronto WesternResearch Institute, University Health Network, University of Toronto, Toronto, ON, Canada M5T 2S8; fLaboratory for Neurobiology and Gene Therapy, KULeuven, 3000 Leuven, Belgium; iLaboratory of Cellular Physiology, National Heart, Lung and Blood Institute/National Institutes of Health, Bethesda, MD20892; and jBrown University/National Institutes of Health Graduate Partnership Program, Department of Neuroscience, Brown University, Providence,RI 02912

Edited by Gregory A. Petsko, Weill Cornell Medical College, New York, NY, and approved December 20, 2013 (received for review September 27, 2013)

Mutations in leucine-rich repeat kinase 2 (LRRK2) cause inheritedParkinson disease (PD), and common variants around LRRK2 area risk factor for sporadic PD. Using protein–protein interaction arrays,we identified BCL2-associated athanogene 5, Rab7L1 (RAB7, memberRAS oncogene family-like 1), and Cyclin-G–associated kinase as bind-ing partners of LRRK2. The latter two genes are candidate genes forrisk for sporadic PD identified by genome-wide association studies.These proteins form a complex that promotes clearance of Golgi-derived vesicles through the autophagy–lysosome system bothin vitro and in vivo. We propose that three different genes forPD have a common biological function. More generally, data in-tegration from multiple unbiased screens can provide insight intohuman disease mechanisms.

BAG5 | GAK | trans-Golgi | autophagy

Genetics contribute to the pathogenesis of Parkinson disease(PD) in two ways. Mutations in several genes can cause

inherited PD (1), and risk factor variants contribute to the risk ofsporadic PD (2). Some genes contribute to both mechanisms.These pleomorphic risk loci (3) include genes that encodeα-synuclein and leucine-rich repeat kinase 2 (LRRK2) (4).However, risk factors for sporadic PD identified by genome-wide association studies (GWASs) (5–9) actually nominate largegenomic loci with multiple candidate genes (10). These loci mayinclude variants that change amino acids or affect disease riskthrough gene expression (11). Also, whether all of the genes forPD converge on a small number of biological pathways is unknown(1). It is, therefore, important to develop unbiased approachesthat would resolve whether genes for PD have similar biologicalfunctions and understand the mechanism(s) of disease risk.Here, we examine one genetic cause of PD (LRRK2) and showthat identifying protein interaction partners can clarify diseasemechanisms.

ResultsLRRK2 contains several protein–protein interaction motifs (showndiagrammatically in Fig. 1A). To identify protein interactions, weprobed protein–protein arrays with biotinylated Glutathione-Stransferase (GST)-LRRK2970–2527, which was kinase-active andtherefore, likely folded (SI Appendix, Fig. S1). Biotinylated GSTwas used as a negative control to exclude false positives that mightnonspecifically bind to many proteins. We also included arrayswhere LRRK2 was coincubated with GDP or GMPPcP. Theseexperiments recovered several candidate interactors, including

three members of the Bcl2-associated athanogene (BAG) do-main cochaperones (12) (SI Appendix, Table S1 and Datasets S1and S2). Of these candidates, we chose to focus on BAG5, be-cause it is expressed in neurons and enhances dopaminergicneuronal death (13).Some known interactors of LRRK2, such as 14-3-3 proteins

(14, 15), could not have been recovered with the truncatedLRRK2 probe that lacks the S910/S935 14-3-3 binding sites.Therefore, we repeated the experiments with full-length Flag-tagged LRRK2 (16) with Flag-tagged green fluorescent protein(GFP) as a negative control. We confirmed interaction with theBAG cochaperones (SI Appendix, Table S2 and Dataset S3) andrecovered 14-3-3 proteins as well as another well-characterizedLRRK2 interactor, STUB1/CHIP (17). This experiment alsoyielded several additional candidates, including two candidatesthat are in risk loci for sporadic PD: Rab7L1 (5, 6) and GAK(7, 8). Because of our interest in the genetic basis of PD, wechose to consider Rab7L1 (RAB7, member RAS oncogenefamily-like 1) and Cyclin-G–associated kinase (GAK) in sub-sequent experiments.We first validated individual interactions between LRRK2

and each candidate. Flag-tagged BAG5 coimmunoprecipitatedendogenous LRRK2 (Fig. 1B), and reciprocally, endogenous

Significance

Understanding loci nominated by genome-wide associationstudies (GWASs) is challenging. Here, we show, using the spe-cific example of Parkinson disease, that identification of protein–protein interactions can help determine the most likely candi-date for several GWAS loci. This result illustrates a significantgeneral principle that will likely apply across multiple diseases.

Author contributions: A.B., S.K.K., I.P.D.G.C., N.A.B.E.C., A.M.L., V.B., L.E.G., J.-M.T., E.G.,and M.R.C. designed research; A.B., I.N.R., A.K., L.C., H.C., S.K.K., L.V.K., E.L., R.C., K.N.,J.D., M.A.N., I.P.D.G.C., N.A.B.E.C., M.O., D.N.H., and R.K. performed research; E.L. con-tributed new reagents/analytic tools; A.B., I.N.R., A.K., J.D., M.A.N., I.P.D.G.C., N.A.B.E.C.,E.G., and M.R.C. analyzed data; and A.B., I.N.R., A.K., S.K.K., J.-M.T., E.G., and M.R.C. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

See Commentary on page 2402.1Consortia members are listed in Supporting Information.2To whom correspondence should be addressed. E-mail: Cookson@mail.nih.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1318306111/-/DCSupplemental.

2626–2631 | PNAS | February 18, 2014 | vol. 111 | no. 7 www.pnas.org/cgi/doi/10.1073/pnas.1318306111

BAG5 coimmunoprecipitated with LRRK2 in stable cell lines(SI Appendix, Fig. S2A). Endogenous BAG5 coimmunoprecipi-tated with endogenous LRRK2 in the mouse brain (SI Appendix,Fig. S2B). Similarly, we found that endogenous LRRK2 coim-munoprecipitated Flag-tagged Rab7L1 (Fig. 1C) or GFP-taggedGAK (Fig. 1D). Collectively, these results show that LRRK2 caninteract with three candidate proteins (BAG5, Rab7L1, andGAK), including endogenous protein in cells and the brain.It is likely that different proteins have different sequence

requirements for binding LRRK2. For example, when 14-3-3proteins were dissociated from LRRK2 by treating with theLRRK2 kinase inhibitor LRRK2-IN1 (18), endogenous BAG5remained bound (SI Appendix, Fig. S2C). To address the se-quence requirements for BAG5-LRRK2 interactions further,we used a DARA mutant BAG5, where key aspartate and argi-nine residues in the BAG domains were replaced with alanine,diminishing heat shock cognate protein 70 (Hsc70) binding (13).DARA-BAG5 had lower binding to LRRK2 than WT BAG5 (SIAppendix, Fig. S2D) [mean intensity of BAG5 relative toLRRK2 (± SEM) of 0.49 ± 0.023 for WT and 0.21 ± 0.037 forDARA; P = 0.0097, two-tailed paired t test, n = 3 independentexperiments], showing that recruitment of Hsc70/heat shock pro-tein 70 (Hsp70) stabilizes the BAG5-LRRK2 complex.Considering LRRK2, the C-terminal Ras of complex protein

(COR) domain was sufficient for interaction with BAG5 (SIAppendix, Fig. S2E). LRRK1, with 46% similarity in the CORdomain to LRRK2, also interacted with BAG5 (SI Appendix, Fig.S2F). These results show that the BAG5-Hsc70/Hsp70 complexis recruited to the COR domain of LRRK proteins. Addingrecombinant BAG5 to LRRK2 did not affect kinase activity(SI Appendix, Fig. S2G).To confirm the interactions using an additional independent

technique, GST-BAG5 was used to pull down myc-LRRK2 fromcells (SI Appendix, Fig. S3A). This technique also confirmed thatthe COR domain of LRRK2 was sufficient for BAG5 interaction(SI Appendix, Fig. S3B). Therefore, the LRRK2-BAG5 inter-actions were supported by protoarrays, coimmunoprecipitation,and GST pull downs.Some LRRK2 interactors are also substrates for LRRK2

(19, 20). When testing whether BAG5 was a LRRK2 substrate, werecovered a copurifying autophosphorylated protein of ∼150 kDa(SI Appendix, Fig. S4 A and B). Given the results in the secondprotoarray experiment with full-length LRRK2, we consideredGAK (143 kDa) as a candidate for this unknown kinase. Sup-porting this hypothesis, we found that GAK coimmunoprecipi-tated with WT BAG5 but not DARA mutant, suggesting thatthis interaction is also Hsp70-mediated (Fig. 1E). Furthermore,endogenous BAG5 was coimmunoprecipitated with GAK in themouse brain (SI Appendix, Fig. S4C). Deletion of the PTEN-likedomain of GAK disrupted and expression of the PTEN-like do-main was sufficient for interaction with both BAG5 and LRRK2(SI Appendix, Fig. S4D).These results do not discriminate whether BAG5 and LRRK2

bind independently to GAK or whether the three proteins arepresent in a cocomplex. To address these possibilities, we per-formed sequential coimmunoprecipitation experiments. GAK wasrecovered from BAG5 purification like in the single coimmuno-prectipitation; then, LRRK2 was coimmunprecipitated from theGAK-BAG5 complexes (Fig. 1F). Furthermore, the cocomplexwas more easily recovered with WT than DARA BAG5, againsuggesting that Hsp70/Hsc70 stabilizes the overall assembly.

Fig. 1. A LRRK2 protein complex. (A) Ideograms of LRRK2 and candidateprotein interactors. LRRK2 shows the location of the mutations used in thisstudy, including pathogenic mutations (red) and T1348N or K1906M hy-pothesis-testing mutants (black). Domains include leucine-rich repeat (LRR),Ras of complex protein (ROC), and COR as well as kinase and WD40 domains.The protoarray baits used in each experiment are boxed. BAG5 has five BAGdomains. GAK has PTEN-like domain (PTEN), Clathrin binding domain (CBD),and J domain. Rab7L1 has a single Ras-like domain. (B–E) Single-proteininteractions. Coimmunoprecipitations using antibodies to (B and C) Flag or(D) GFP from cell lysates expressing (B) BAG5, (C) Rab7L1, or (D) GAK with (B)mock-transfected cells, (C) Flag-β-glucoronidase (GUS), or (D) GFP as nega-tive controls probed for endogenous LRRK2 (arrows), Hsc70 (closed arrow-head), and tagged proteins (open arrowheads). IB, immunoblotting; IP,immunoprecipitation. (E) IP of BAG5 probed for endogenous GAK (arrow-heads). (F and G) Cocomplex formation between LRRK2, GAK, BAG5, andRab7L1. (F) Lysates were immunoprecipitated first with Flag for BAG5 and

then with GFP for GAK, and they were probed for LRRK2 and interactingpartners. (G) Flag-tagged Rab7L1 complexes probed for endogenous LRRK2,GAK, Hsc70, and Bag5. For all panels, markers on the right of the blots arein kilodaltons, and results are representative of at least three independentexperiments.

Beilina et al. PNAS | February 18, 2014 | vol. 111 | no. 7 | 2627

GEN

ETICS

SEECO

MMEN

TARY

To address whether the interactions require all components orcan be formed in the absence of any one member, we knockeddown each component in turn and performed coimmunopre-cipitation experiments for the remaining partners. These results(SI Appendix, Fig. S4 E–G) showed that interactions remainedeven when BAG5, GAK, or LRRK2 was depleted from cells.We next considered whether Rab7L1 might be part of the

same protein complex. Rab7L1 was coimmunoprecipitated withLRRK2, and this binding was not significantly altered by muta-tions in LRRK2 (SI Appendix, Fig. S5 A and B). However, de-letion of the N-terminal HEAT [Huntingtin, elongation factor3 (EF3), protein phosphatase 2A (PP2A), TOR1] region ofLRRK2 weakened interaction dramatically (SI Appendix, Fig. S5C and D).Rab7L1 is a single-domain protein, and therefore, we did not

generate truncation constructs that would likely be misfolded.Instead, we asked whether LRRK2 interactions depended on thenucleotide state of Rab7L1. To this end, we generated T21N andQ67L mutants in Rab7L1 that were predicted to lack GTPbinding and show impaired GTP hydrolysis based on structuralmodeling (SI Appendix, Fig. S6 A–C). As expected, T21N wasdeficient in GTP binding (SI Appendix, Fig. S6D) [F(4,15), P =4.5 × 10−13, one-way ANOVA, n = 4 independent experiments].Unexpectedly, Q67L had an ∼10-fold higher GTP/GDP disso-ciation rate than WT Rab7L1 (SI Appendix, Fig. S6 E and F), andboth mutants showed altered mobility on size exclusion chro-matography (SI Appendix, Fig. S5 G and H). Therefore, bothRab7L1 mutants act as loss-of-function variants, likely becauseof failure to retain GTP. Both T21N and Q67L Rab7L1 showed

enhanced interaction with LRRK2 (SI Appendix, Fig. S6 I and J)[one-way ANOVA; F(4,10), P = 4.71 × 10−9, n = 3 independentexperiments], suggesting that Rab7L1 binds LRRK2 in both theGTP- and GDP-bound states. However, GTP binding seemed tobe important for some functions of Rab7L1, because both Rab7L1mutants showed diminished (Q67L) or abolished (T21N) Golgilocalization (SI Appendix, Fig. S6K). Endogenous Rab7L1 couldsimultaneously coimmunoprecipitate endogenous LRRK2, GAK,Hsp70, and BAG5 (Fig. 1G). Collectively, these results suggestthat all five proteins are in a single complex.We next asked whether the formation of a single-protein com-

plex implied that these proteins have similar effects in cells. Weused neurite shortening, which is exaggerated by pathogenicLRRK2 mutations (21–24). WT BAG5 caused neurite short-ening, but this effect was diminished with the DARA mutantBAG5 (SI Appendix, Fig. S7 A–C) [F(2,158) = 109, P < 2 × 10−16

by one-way ANOVA; n > 140 cells across three independentexperiments]. Expression of full-length GAK, but not the de-letion constructs incapable of binding LRRK2 (SI Appendix,Fig. S7D) [F(3,215) = 5.97, P = 6.3 × 10−4 by one-way ANOVA;n > 49 cells per construct across two independent experiments),and expression of WT Rab7L1, but not the GTP binding-deficient Q67L variant, also decreased neurite length (SI Appendix,Fig. S7E) [F(2,435) = 55, P < 2 × 10−16 by one-way ANOVA;n > 125 cells across three independent experiments]. The effectof Rab7L1 was diminished in LRRK2 KO neurons (SI Appendix,Fig. S7F) [two-way ANOVA, F(1,281) = 21, P = 0.00076 forgenotype; F(1,281) = 2.7, P = 0.06 for Rab7L1 variant; F(1,281) =21, P = 2.04 × 10−5 for interaction; n > 70 cells counted per

Fig. 2. (A) LRRK2 complex and LRRK2 mutations promote clearance of trans-Golgi–derived vesicles. Clustering and clearance of trans-Golgi by LRRK2 andinteracting partners. Primary neurons were transfected with the indicated constructs and stained for transfection markers (green) and trans-Golgi (GLG1;blue). Right shows higher magnification of the circled neurons, whereas adjacent cells (asterisks) show normal Golgi morphology, including clustered Golgi inRab7L1 (Lower Right) transfections and a LRRK2-transfected neuron with minimal residual Golgi staining (Center Right). (B) Blinded counts of cells withnormal Golgi (blue), clustered Golgi (orange), or minimal or no apparent Golgi staining (white) as a proportion of all counted cells (n > 50 cells in threeindependent cultures per construct, P values are by Fisher exact test for proportions against GUS-transfected cultures corrected for the number of testsapplied). (C and D) Pathogenic mutations in LRRK2 increase clearance of the trans-Golgi network. HEK cells were transfected with indicated LRRK2 variants orGUS as a negative control and stained with antibodies to TGN46. Counting of different Golgi phenotypes after transfections with different LRRK2 variants (C)without or (D) with cotransfection of Rab7L1 and statistical analysis were performed as in B. (E) Minimal functional complex includes LRRK2, BAG5 GAK, andRab7L1. Cells were transfected with GUS (negative control for transfection), BAG5, LRRK2, GAK, or Rab7L1 and simultaneously treated with siRNA against theother components. Counting of different Golgi phenotypes after transfections with different combinations was performed as in C and D.

2628 | www.pnas.org/cgi/doi/10.1073/pnas.1318306111 Beilina et al.

condition), suggesting that the effects of Rab7L1 require expressionof LRRK2.We next asked where each protein might be in the cell. LRRK2

was present in microsomes as reported (25), and endogenousBAG5 and GAK were also present in the same subcellularfractions (SI Appendix, Fig. S8A). To provide additional resolution,we expressed each interactor as tagged proteins in neurons, al-lowing us to use the same antibody for visualization. LRRK2 wascytosolic and vesicular (18, 26), and BAG5 was cytosolic andnuclear, whereas GAK and Rab7L1 were largely vesicular (SIAppendix, Fig. S8B). When coexpressed, LRRK2 and GAK showedvesicular colocalization. WT Rab7L1 promoted vesicular LRRK2localization, but the Q67L Rab7L1 variant that was deficient inGTP binding prevented this effect (SI Appendix, Fig. S8C).These results show that a cocomplex of these proteins mayform in vesicular compartments of the cell and that the effectsrequire GTP binding by Rab7L1.To further understand the possible function of the complex,

we considered what type of vesicular compartments to which itmight localize. Both Rab7L1 (27, 28) and GAK (29, 30) are re-ported to localize to the trans-Golgi network (TGN). WT Rab7L1expressed alone (SI Appendix, Fig. S6K) or with LRRK2 (SIAppendix, Fig. S9 A and B) partially colocalized with markers ofthe TGN but not retromer (31) (SI Appendix, Fig. S9C). Whenwe coexpressed both Rab7L1 and LRRK2, TGN staining wasdiminished and therefore, appeared clustered in the perinucleararea. Quantification showed that this phenotype was strongerfor TGN (TGN46, trans-Golgi network integral membrane protein2) than cis-Golgi (GM130, Golgin subfamily A member 2) markers(SI Appendix, Fig. S6 A and B) [F(3,8) = 56.9, P = 9.7 × 10−6 forTGN46; F(3,8) = 9.5, P = 0.0051 for GM130; one-way ANOVA).These results suggest that Rab7L1 directs LRRK2 to TGN-derived vesicles.In LRRK2-transfected neurons, the TGN was again clus-

tered, or dramatically, only residual staining was visible (Fig. 2A).Counting cells showed that there was a significant change in theproportions of cells with clustered or cleared Golgi after expres-sion of LRRK2 or each interaction partner—BAG5, Rab7L1, orGAK (Fig. 2B, P values) (two-sided Fisher exact test correctedfor the number of tests performed). Validating these resultsbiochemically, HEK cells expressing LRRK2 had lower amountsof the TGN marker TGN46 (SI Appendix, Fig. S10 A and B) (P =0.00038, t test with Welch correction, n = 5 samples) but not thecis-Golgi marker GM130 (SI Appendix, Fig. S10C) (P = 0.07).Considering the timing of these events, the recruitment of LRRK2to TGN46-positive vesicles by Rab7L1 was complete by 24 h aftertransfection, but clearance of Golgi increased up to 72 h (SIAppendix, Fig. S11). Therefore, physical recruitment of LRRK2to the TGN precedes removal of TGN-derived vesicles.The clearance of the TGN potentially provides a cellular mea-

sure of the function of LRRK2 and binding partners. We nextasked whether point mutations in LRRK2 affect Golgi clearance,thus allowing us to infer what functional roles of LRRK2 areimportant. All pathogenic mutations in LRRK2, including thosemutations that increase kinase activity (G2019S) or decrease GTPturnover (R1441C and Y1699C), enhanced Golgi clearance (Fig.2C, P values). However, hypothesis-testing mutations that di-minish GTP binding (T1348N) or are kinase-inactive (K1906M)did not support Golgi clearance. The effects of LRRK2 muta-tions on Golgi clearance were enhanced by Rab7L1 (Fig. 2D,P values). These results suggest that the known enzymaticfunctions of LRRK2 (GTP binding and kinase activity) areimportant in the clearance of Golgi by a LRRK2 complex.To define the minimum functional unit of the complex, we

transfected cells with BAG5, LRRK2, GAK, or Rab7L1 andknocked down each of the other components in turn. We foundthat knockdown of any single component was able to rescue theeffects of increased expression of other components on Golgi

turnover (Fig. 2E, P values). Given that knockdown of each com-ponent did not disrupt protein interactions (SI Appendix, Fig.S4 E–G), these results show that, even if a complex forms, fullfunction requires all components.To confirm these results in vivo, we used lentiviral vectors to

express enhanced green fluorescent protein (eGFP)-tagged G2019SLRRK2 in the striatum of adult mice. We found more cells withminimal staining for GLG1, a TGN marker suitable for immu-nostaining in rodent tissue, after expression of mutant LRRK2compared with controls (Fig. 3) [F(2,12) = 131.4, P = 6.93 ×10−9; one-way ANOVA, n = 3 animals per group]. These resultssuggest that increased expression of LRRK2 has similar func-tional effects in the mouse brain as in cellular systems.We next sought to understand the mechanisms involved in

TGN turnover. The above results are reminiscent of the effectsof expression and activity of another Parkinson-related protein,Parkin, that promotes removal of mitochondria by autophagy(32). LRRK2 has previously been reported to control autophagyin several systems (24, 33, 34). To test whether the autophagy–lysosome system contributed to LRRK2-dependent phenotypes,we treated cells with bafilomycin-A1, which prevents lysosomalacidification. Bafilomycin-A1 blocked the effect of LRRK2 (Fig.4 A and B, P values) and also prevented the turnover of TGN46-positive vesicles when LRRK2 was cotransfected with GAK (SIAppendix, Fig. S12A).Asking whether the effects were specifically related to autophagy,

we found that the effects of LRRK2 were diminished by siRNAagainst the autophagy protein ATG7 (Fig. 4C). Morphologically,cells expressing LRRK2 and treated with bafilomycin-A1 con-tained larger LRRK2-positive structures adjacent to lysosomes (SIAppendix, Fig. S12B). When cotransfected, both Rab7L1 andLRRK2 were found in swollen vesicles near lysosomes (SIAppendix, Fig. S12C). Some areas of TGN46 immunoreactivity

Fig. 3. Acute expression of mutant LRRK2 in adult mouse brain results indiminished staining of TGN markers in vivo. (A) Mouse striatum sectionsfrom animals 2 wk after injection with lentiviral vectors expressing either(Left) eGFP or (Right) G2019S mutant eGFP-LRRK2 stained for GFP (green),GLG1 (red), and TOPRO3 (blue in Bottom). (B) Counting of cells with minimalGLG1 staining in animals transduced with the vectors, with untransducedcells in the same sections also counted as controls. Error bars show SEM.***P < 0.001 (Tukey posthoc test from one-way ANOVA compared witheGFP control).

Beilina et al. PNAS | February 18, 2014 | vol. 111 | no. 7 | 2629

GEN

ETICS

SEECO

MMEN

TARY

were enclosed by swollen LRRK2/Rab7L1-positive vesicles (SIAppendix, Fig. S12D). Collectively, these results support thehypothesis that turnover of TGN vesicles depends on theautophagy–lysosome system.The above results make the assumption that increasing the

expression of LRRK2 or its binding partners models the di-rection of effect of genetic risk in PD. To evaluate whetherexpression of these genes might explain the risk of PD in thehuman brain, we assayed a series of postmortem samples (n ∼ 60)using exome sequencing and RNA-Seq. We examined the locicontaining Rab7L1 (SI Appendix, Fig. S13A), GAK (SI Appendix,Fig. S13B), or LRRK2 (SI Appendix, Fig. S13C) for sequence

variants. Although we reliably detected expression of LRRK2,GAK, and Rab7L1 in the human cerebral cortex, many patho-genic variants were too rare to be analyzed or in regions that didnot code for RNA. However, rs708723 in the 3′ UTR of Rab7L1could be identified directly from RNA-Seq data (Fig. 4D). Therewas lower expression of the C allele in RNA-Seq data in het-erozygous carriers, and the proportions of C alleles in the ag-gregate counts were significantly different between exome andRNA-Seq data (Fig. 4E) (χ2 test, χ2 = 5.9, P = 0.015). TheC allele of rs708723 was associated with a lower risk of PD (oddsratio = 0.914, P = 0.000554 under an additive model). Therefore,higher expression of normal sequence variants is a plausiblemechanism for risk of PD, at least for Rab7L1.

DiscussionHere, we used an unbiased survey of protein interactions ofLRRK2 to identify binding partners and show that these proteinsform a cocomplex that promotes TGN removal by autophagy-dependent mechanisms. TGN clearance also occurs in the brainin vivo, and expression levels of some of the protein componentsare a reasonable mechanism for risk of sporadic PD. Our resultsoverall lead us to propose that several PD genes act in a commonpathway (SI Appendix, Fig. S14).Because LRRK2 has several proposed interaction domains, it

may act as a scaffold for multiple interaction partners. Althoughwe recovered known direct binding partners of LRRK2 (15, 35)and inferred indirect interaction with Hsp70 (17), not all pro-posed literature candidates were identified. This result may in-dicate a false-negative rate with arrays, which do not include allsplice variants of all proteins in the human genome or all post-translational modifications. The approach used here should becombined with other methods, which has been suggested moregenerally for understanding protein function (36).With these caveats, our approach led us to show that LRRK2

promotes the relocalization to and clearance of trans-Golgi–derived vesicles. This effect is presumably related to the knownfunction of LRRK2 in autophagy (18, 24, 33, 37, 38) but can betied mechanistically to disruption of TGN. Interestingly, auto-phagic vesicles may derive from the trans-Golgi network (37);hence, altered turnover of Golgi-derived vesicles would impactautophagy function over time, consistent with effects of LRRK2deficiency in animal models (34).Critically, this activity is enhanced by all pathogenic mutations

in LRRK2. However, LRRK2 mutants do not alter binding to itspartner proteins. This observation suggests that the enzymaticactivity of LRRK2, including GTP binding and kinase activ-ity, may be critical for activity of the complex. Supporting thisidea, hypothesis-testing mutations lacking GTP binding orkinase activity do not promote clearance of the TGN-derivedstructures. It is possible that, after being bound, protein in-teraction partners also influence either LRRK2 kinase activity orGTP affinity. Defining the stoichiometry of the complex andeffects on LRRK2 is an important future goal.The other proteins each seem to be required for LRRK2 to

affect Golgi clearance by autophagy. We have shown that removalof BAG5, GAK, or Rab7L1 does not prevent LRRK2 bindingwith other partners, implying that formation of a partial complexis not sufficient for Golgi clearance. Within the complex, wespeculate that each protein may play a specific role. It is of in-terest that Rab7L1 is localized to the trans-Golgi network, likelyin its GTP-bound form, where it might recruit LRRK2 and othercomponents to cooperatively cause TGN to be engulfed by theautophagosomes. The presence of BAG5 and Hsp70/Hsc70 sug-gests that this process requires remodeling of proteins or proteincomplexes. GAK is also known to cooperate to remodel clathrincomplexes (39).Our results have the broadest implications for mechanisms of

understanding disease risk for loci identified by GWASs. Higher

Fig. 4. Mechanism of turnover of Golgi-derived vesicles involves theautophagy–lysosomal system and is consistent with altered gene expressionin the human brain. (A–C) Selective Golgi clearance involves the autophagy–lysosome system. (A) Cells coexpressing LRRK2 (red) and GUS (blue) stainedfor TGN46 (green) after treatment for 24 h with DMSO or 10 nM bafilomycinA1. A cell with only residual TGN46 is shown in Right. (B and C) Blindedcounts of the proportions of Golgi phenotypes in cells treated with (B)bafilomycin A1 or (C) knockdown of Atg7. (D and E) Gene expression in thehuman brain. (D) Direct assay of rs708723 in Rab7L1 from the brain of anindividual who was heterozygous at this marker using RNA-Seq. (E) Densityplots for the proportion of reads with the C allele for RNA-Seq (red) andexome data for DNA (blue) in brains from individuals heterozygous forrs708723. There is departure from the expected proportion of 0.5 (blackvertical line) for RNA-Seq data.

2630 | www.pnas.org/cgi/doi/10.1073/pnas.1318306111 Beilina et al.

expression of Rab7L1 is associated with higher risk of PDin humans, expression quantitative trait locus identified bymicroarrays, thus providing a reasonable mechanism for theaction of the WT allele. We cannot exclude that other genes ateach GWAS locus may contribute to disease risk, because sev-eral of these loci are complex genetically (9). However, identi-fication of GAK and Rab7L1 as binding partners of LRRK2from an unbiased screen raises the probability that these genesexplain some of the GWAS signal. This concept has importantimplications for other diseases that have been studied byGWASs, because it supports the idea that integrating datasetsfrom multiple unbiased approaches leads to unexpected insightsinto human diseases.

Materials and MethodsDetailed materials and methods are included in SI Appendix. Briefly, iden-tification of protein interactors was achieved by screening two batches ofProtoarrays (Invitrogen) with either truncated GST-LRRK2970–2527 (Invitrogen)or full-length protein prepared as described (16). Protein interactions werevalidated using coimmunoprecipitation in HEK293FT cells or mouse brain

and also GST pull downs in cells or brain. Subcellular fractionation wasperformed as described (25) using LRRK2-transfected HEK293FT cells. Proteinlocalization and effects on turnover of TGN were performed using primarypostnatal cortical neurons with calcium–phosphate transfection (40) or inHEK293FT cells. All counting experiments were performed by an operatorblinded to the transfection construct used; details of blinding and statisticalapproaches are included in SI Appendix.

ACKNOWLEDGMENTS. We thank the Genomic Technologies Section at theNational Institute of Allergy and Infectious Diseases for help with proteinarray scans. This research was supported, in part, by the Intramural ResearchProgram of the National Institutes of Health, National Institute on Aging;the Michael J. Fox Foundation for Parkinson’s Disease (V.B., J.-M.T., andE.G.); KU Leuven Grant OT/08/052A (to V.B.); Fonds WetenschappelijkOnderzoek (FWO)-Vlaanderen Project G.0666.09 (to V.B.); Agentschap voorInnovatie door Wetenschap en Technologie-Vlaanderen Grants SBO/80020(to V.B.) and SBO/100042 (to V.B.); an FWO-Vlaanderen postdoctoral fellow-ship (to J.-M.T.); an FWO-Vlaanderen PhD fellowship (to E.L.); the FundDruwé-Eerdekens managed by the King Baudouin Foundation (J.-M.T.);the Rientro dei Cervelli Program (lncentivazione alla mobilità di studiosistranieri e italiani residenti all’estero) from the Italian Ministry of Educa-tion, University and Research (E.G.), Fondazione Cariplo Grant 2011 0540 (toE.G.); and Fondazione Telethon Grant GGP12237 (to E.G.).

1. Cookson MR, Bandmann O (2010) Parkinson’s disease: Insights from pathways. HumMol Genet 19(R1):R21–R27.

2. Lill CM, Bertram L (2011) Towards unveiling the genetics of neurodegenerative dis-eases. Semin Neurol 31(5):531–541.

3. Singleton A, Hardy J (2011) A generalizable hypothesis for the genetic architecture ofdisease: Pleomorphic risk loci. Hum Mol Genet 20(R2):R158–R162.

4. Cookson MR (2010) The role of leucine-rich repeat kinase 2 (LRRK2) in Parkinson’sdisease. Nat Rev Neurosci 11(12):791–797.

5. Satake W, et al. (2009) Genome-wide association study identifies common variants atfour loci as genetic risk factors for Parkinson’s disease. Nat Genet 41(12):1303–1307.

6. Simón-Sánchez J, et al. (2009) Genome-wide association study reveals genetic riskunderlying Parkinson’s disease. Nat Genet 41(12):1308–1312.

7. International Parkinson’s Disease Genomics Consortium (IPDGC); Wellcome Trust CaseControl Consortium 2 (WTCCC2) (2011) A two-stage meta-analysis identifies severalnew loci for Parkinson’s disease. PLoS Genet 7(6):e1002142.

8. Do CB, et al. (2011) Web-based genome-wide association study identifies two novelloci and a substantial genetic component for Parkinson’s disease. PLoS Genet 7(6):e1002141.

9. Nalls MA, et al. (2011) Imputation of sequence variants for identification of geneticrisks for Parkinson’s disease: A meta-analysis of genome-wide association studies.Lancet 377(9766):641–649.

10. Ku CS, Loy EY, Pawitan Y, Chia KS (2010) The pursuit of genome-wide associationstudies: Where are we now? J Hum Genet 55(4):195–206.

11. Latourelle JC, Dumitriu A, Hadzi TC, Beach TG, Myers RH (2012) Evaluation of Par-kinson disease risk variants as expression-QTLs. PLoS One 7(10):e46199.

12. Kabbage M, Dickman MB (2008) The BAG proteins: A ubiquitous family of chaperoneregulators. Cell Mol Life Sci 65(9):1390–1402.

13. Kalia SK, et al. (2004) BAG5 inhibits parkin and enhances dopaminergic neuron de-generation. Neuron 44(6):931–945.

14. Dzamko N, et al. (2010) Inhibition of LRRK2 kinase activity leads to dephosphorylationof Ser(910)/Ser(935), disruption of 14-3-3 binding and altered cytoplasmic localiza-tion. Biochem J 430(3):405–413.

15. Li X, et al. (2011) Phosphorylation-dependent 14-3-3 binding to LRRK2 is impaired bycommon mutations of familial Parkinson’s disease. PLoS One 6(3):e17153.

16. Civiero L, et al. (2012) Biochemical characterization of highly purified leucine-richrepeat kinases 1 and 2 demonstrates formation of homodimers. PLoS One 7(8):e43472.

17. Ko HS, et al. (2009) CHIP regulates leucine-rich repeat kinase-2 ubiquitination, deg-radation, and toxicity. Proc Natl Acad Sci USA 106(8):2897–2902.

18. Alegre-Abarrategui J, et al. (2009) LRRK2 regulates autophagic activity and localizesto specific membrane microdomains in a novel human genomic reporter cellularmodel. Hum Mol Genet 18(21):4022–4034.

19. Stafa K, et al. (2012) GTPase activity and neuronal toxicity of Parkinson’s disease-associated LRRK2 is regulated by ArfGAP1. PLoS Genet 8(2):e1002526.

20. Xiong Y, Yuan C, Chen R, Dawson TM, Dawson VL (2012) ArfGAP1 is a GTPase acti-vating protein for LRRK2: Reciprocal regulation of ArfGAP1 by LRRK2. J Neurosci32(11):3877–3886.

21. MacLeod D, et al. (2006) The familial Parkinsonism gene LRRK2 regulates neuriteprocess morphology. Neuron 52(4):587–593.

22. Winner B, et al. (2011) Adult neurogenesis and neurite outgrowth are impaired inLRRK2 G2019S mice. Neurobiol Dis 41(3):706–716.

23. Dächsel JC, et al. (2010) A comparative study of Lrrk2 function in primary neuronalcultures. Parkinsonism Relat Disord 16(10):650–655.

24. Plowey ED, Cherra SJ, 3rd, Liu Y-J, Chu CT (2008) Role of autophagy in G2019S-LRRK2-associated neurite shortening in differentiated SH-SY5Y cells. J Neurochem 105(3):1048–1056.

25. Rudenko IN, et al. (2012) The G2385R variant of leucine-rich repeat kinase 2 associ-ated with Parkinson’s disease is a partial loss-of-function mutation. Biochem J 446(1):99–111.

26. Biskup S, et al. (2006) Localization of LRRK2 to membranous and vesicular structuresin mammalian brain. Ann Neurol 60(5):557–569.

27. Helip-Wooley A, Thoene JG (2004) Sucrose-induced vacuolation results in increasedexpression of cholesterol biosynthesis and lysosomal genes. Exp Cell Res 292(1):89–100.

28. MacLeod DA, et al. (2013) RAB7L1 interacts with LRRK2 to modify intraneuronalprotein sorting and Parkinson’s disease risk. Neuron 77(3):425–439.

29. Lee D-W, Wu X, Eisenberg E, Greene LE (2006) Recruitment dynamics of GAK andauxilin to clathrin-coated pits during endocytosis. J Cell Sci 119(Pt 17):3502–3512.

30. Kametaka S, et al. (2007) Canonical interaction of cyclin G associated kinase withadaptor protein 1 regulates lysosomal enzyme sorting. Mol Biol Cell 18(8):2991–3001.

31. Cullen PJ, Korswagen HC (2012) Sorting nexins provide diversity for retromer-dependent trafficking events. Nat Cell Biol 14(1):29–37.

32. Youle RJ, Narendra DP (2011) Mechanisms of mitophagy. Nat Rev Mol Cell Biol 12(1):9–14.

33. Gómez-Suaga P, et al. (2012) Leucine-rich repeat kinase 2 regulates autophagythrough a calcium-dependent pathway involving NAADP. Hum Mol Genet 21(3):511–525.

34. Tong Y, et al. (2012) Loss of leucine-rich repeat kinase 2 causes age-dependent bi-phasic alterations of the autophagy pathway. Mol Neurodegener 7:2.

35. Nichols RJ, et al. (2010) 14-3-3 binding to LRRK2 is disrupted by multiple Parkinson’sdisease-associated mutations and regulates cytoplasmic localization. Biochem J430(3):393–404.

36. Ryan CJ, et al. (2013) High-resolution network biology: Connecting sequence withfunction. Nat Rev Genet 14(12):865–879.

37. Manzoni C, et al. (2013) Pathogenic Parkinson’s disease mutations across the func-tional domains of LRRK2 alter the autophagic/lysosomal response to starvation. Bi-ochem Biophys Res Commun 441(4):862–866.

38. Ramonet D, et al. (2011) Dopaminergic neuronal loss, reduced neurite complexity andautophagic abnormalities in transgenic mice expressing G2019S mutant LRRK2. PLoSOne 6(4):e18568.

39. Lee DW, Zhao X, Zhang F, Eisenberg E, Greene LE (2005) Depletion of GAK/auxilin 2inhibits receptor-mediated endocytosis and recruitment of both clathrin and clathrinadaptors. J Cell Sci 118(Pt 18):4311–4321.

40. Jiang M, Chen G (2006) High Ca2+-phosphate transfection efficiency in low-densityneuronal cultures. Nat Protoc 1(2):695–700.

Beilina et al. PNAS | February 18, 2014 | vol. 111 | no. 7 | 2631

GEN

ETICS

SEECO

MMEN

TARY

1

SI Appendix Beilina et al

Supplementary Materials and Methods

Plasmids and cell lines

Myc- or Flag-tagged LRRK2 and LRRK11,2, Flag-tagged BAG53 and GFP-tagged GAK constructs including deletion constructs 4 have been described previously. Full length Rab7L1 was amplified from human cerebral cortex cDNA using PCR primers: 5’-ATGGGCAGCCGCGACCACCT-3’ and 5’-CTAGCAGCAGGACCAGCTGG -3’. BAG5 was cloned using: 5’-ATGGATATGGGAAACCAACA-3’ and 5’-GTACTCCCATTCATCAGATTTC-3’. Full length GAK was amplified with: 5’-ATGTCGCTGCTGCAGTCGGC-3’ and 5’-TCAGAAGAGGGGCCGGGAGCCC-3’.

PCR products were purified and cloned into pCR8/GW/TOPO vector (Invitrogen) then transferred into pCMV-2xmyc-DEST2, a Gateway-modified p3xFlag-CMV™-7.1 N-terminal Met-3xFlag vector (Sigma), pLenti6/V5-DEST (Invitrogen) using Gateway recombination (Invitrogen). T21N and Q67L Rab7L1 mutations were introduced by the QuikChange II site-directed mutagenesis kit (Stratagene).

Protoarrays

GST-tagged, truncated (970–2527) kinase active LRRK2 was purchased from Invitrogen and used as a probe for initial protein arrays. The GST-LRRK2 protein was first dialyzed against 1x PBS to remove Tris storage buffer and then was concentrated using Centrifugal Filter with 30K membrane (Millipore). Concentrated protein was then labeled using the Biotin-XX Microscale protein labeling kit (Invitrogen). We estimated that ~8 biotin molecules were incorporated per molecule of LRRK2. GST-LRRK2 (970–2527) remained active as a kinase after biotin labeling (Supplementary Fig. 1).

Protoarrays, version 4.1 (Invitrogen, Lot number HA20150) were probed with biotinylated LRRK2

according to the manufacturer’s instructions. In this version of the arrays, target proteins on the arrays were hybridized with GST tags oriented towards the glass surface. Five arrays were screened in two batches. In the first batch, we used hybridization buffer alone and either used 6 mg of GST or 6 mg GST-LRRK2 to probe the arrays. In the second batch, we probed with 6 mg GST-LRRK2 in the presence of 200 mM GDP and 6 mg GST-LRRK2 in the presence of 200 mM GMPPcP, with both nucleotides included in the hybridization buffer. Protein binding was visualized using streptavidin-AlexaFluor647 conjugate and arrays were imaged using an Axon GenePix 4000B fluorescence scanner.

For full-length LRRK2, we purified Flag-tagged protein or Flag-GFP as a negative control as described previously1. Six mg of each purified 3xFlag-LRRK2 and 3xFlag-eGFP proteins were used to probe Protoarrays, version 5.03 (Invitrogen, Lot number HA20265) according to the manufacturer’s instructions with the modification that after 3xFlag-tagged protein probing, arrays were probed with monoclonal

ANTI-FLAG ® BioM2−Biotin, Clone M2 (Sigma-Aldrich) antibody, followed by probing with Alexa Fluor® 647 streptavidin (Invitrogen). In this version of the arrays, target proteins on the arrays were hybridized randomly with respect to orientation relative to the glass slide.

Raw array images were converted to text files using GenePix Pro that were then used as input into ProtoArray Prospector (Invitrogen). Within ProtoArray Prospector, binding strength was estimated as Z-scores, i.e. numbers of standard deviations above background fluorescence on the array. Each protein on the array is spotted in duplicate, hence reported values are averaged for both spots. To develop a list of candidate interactors for truncated LRRK2, we filtered the

2

raw data from the first set of experiments to generate a list of proteins that had Z-scores >3 for GST-LRRK2 in all conditions and had Z-scores <3 for both buffer and GST. Similarly, for full-length LRRK2, we filtered the raw data from the second set of arrays for those candidates with Z>3 for LRRK2 and Z<3 for eGFP from the arrays probed in parallel. Because there are differences in manufacture between the two array lots, we did not combine the analyses of the GST-LRRK2 and full-length protein. Raw data, before filtering was applied, is included in the supplementary online material.

Cell culture and transfections

HEK293FT cells (Invitrogen) were maintained in DMEM (Lonza) containing 4.5 g/l glucose, 2 mM l-glutamine, and 10% fetal bovine serum (Lonza) at 37°C in 5% CO2. Transient transfection of cells was performed using PEI (polyethyeleneimine) (Polysciences) or Lipofectamine 2000 (Invitrogen), according to the manufacturer's instructions. To knockdown endogenous genes, cells were co-transfected with the SMARTpool ON-TARGETplus BAG5, GAK, LRRK2, Rab7L1 or scrambled siRNAs for 48 hours using DharmaFECT transfection reagent (Thermo scientific). Transgene expression in siRNA transfected cells were performed 24 hours after siRNA transfection for additional 24 hours using Lipofectamine 2000 (Invitrogen) or DharmaFECT Duo transfection reagents (Thermo scientific).. Primary neurons from cortex were prepared from postnatal day 0 pups and plated onto coverslips precoated with poly-D-lysine (Neuvitro) at 0.7x106 cells/well as described5. Transfections were performed on day 2 in vitro using a modified CaPO4 method6.

Lentiviral transduction and GLG1 staining in mouse brain

HIV-1 derived lentiviral vector particles encoding eGFP or eGFP-LRRK2 G2019S were generated by the Leuven Viral Vector Core (LVVC), Laboratory for Neurobiology and Gene Therapy, essentially as described6. Briefly, after seeding HEK293T cells in 15-cm dishes, we performed a triple transient transfection with the respective

transfer plasmids, a packaging plasmid and an envelope plasmid encoding VSV G. The production was performed in Opti-MEM I (Gibco-Invitrogen, Merelbeke, Belgium) and medium was replaced after 24 h. Cell supernatant containing lentiviral vectors was collected on day 2 and 3 post-transfection and filtered through a 0.45 µm pore size filter (Sartorius, Minisart, Göttingen, Germany). Vector containing medium was concentrated by dissolving the vector pellet obtained by ultracentrifugation in 200 µl PBS. Functional vector titers were determined by transducing HEK-293T cells with the respective vector preparation in a 10-fold dilution series. Three days after transduction, cells were harvested, fixed in 4 % paraformaldehyde (PAF, VWR international prolabo, Leuven, Belgium) and analyzed using a FACSCalibur flow cytometer (BD Biosciences, Erembodegem, Belgium) and the CellQuest software package provided with the instrument. Functional vector titers are expressed as transducing units (TU) per ml (eGFP:2.5 E07 TU/ml and eGFP-LRRK2 G2019S: 4.0 E07 TU/ml). The lentiviral vector transfer construct for eGFP was previously described7 and the LV transfer plasmid expressing eGFP-LRRK2 G2019S was derived from the previously described8 LV-3flag-LRRK2 constructs in which the 3xFlag tag was replaced by an eGFP sequence. Lentiviral vectors (4µl) were injected unilaterally in the striatum of adult mice (4 mice per group, with one animal per group reserved for a negative control without staining, hence n=3 animals were used for the final counts) and perfused two weeks after injection, essentially as described9. Sections (50 mm) were stained for rabbit GLG1 (Sigma, 1:500), mouse GFP (Roche, 1:500), and Topro-3. Autofluorescence was quenched using 0.3% Sudan Black, in 70% ethanol, 10 min. Sections (2 sections per animal) were imaged using 63x objective using a Zeiss LSM510 confocal microscope. Each image represents average projection of seven z-stacks of 0.8 micron each. Approximately 20 cells per section were assessed for GLG1 staining in animals injected with eGFP-LRRK2 G2019S and 80 cells per section in animals injected with eGFP control.

3

Mass spectrometry

Purified BAG5 WT and DARA mutant proteins were separated on 4-20 % SDS-PAGE, stained with Gelcode Blue Stain Reagent (Thermo Scientific). Individual bands were excised and digested with trypsin. Tryptic peptides were analyzed by a LC/MS/MS system with Finnigan Surveyor HPLC connected to Thermo LTQ XL mass spectrometer (Thermo Scientific). Protein identification was performed using MASCOT.

Protein purification and assays of Rab7L1 activity

HEK293FT cells transfected with 3xFlag-Rab7L1 WT, T21N or Q67L plasmid DNAs were solubilized in lysis buffer containing 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 10% Glycerol, 1x Halt phosphatase inhibitor cocktail (Thermo Scientific) and protease inhibitor cocktail (Roche) for 30 min on ice. Lysates were centrifuged (10 min, 20,000xg) and supernatants were precleared with EZview™ Red Protein G Affinity Gel (Sigma-Aldrich) for 30 min at 4°C then incubated with EZview™ Red ANTI-FLAG® M2 Affinity Gel (Sigma-Aldrich) for 1h at 4°C on a rotator. Proteins on beads were washed 6 times with 25 mM Tris-HCl pH 7.5, 400 mM NaCl, and 1% Triton X-100.

For binding assays, 3x-Flag-Rab7L1 fusion proteins bound to EZview™ Red ANTI-FLAG® M2 Affinity Gel (Sigma-Aldrich) were washed twice with Buffer A (20 mM Tris–HCl pH 7.5, 100 mM NaCl, 5 mM MgCl2, 1 mM NaH2PO4, 2 mM DTT) and incubated overnight on ice in Buffer A containing 0.1 µM 32P-a-GTP. Beads were then washed twice in Buffer A to remove unbound nucleotides, added to the Bio-safe II (RPI) scintillation cocktail and binding quantified using scintillation counting for 32P on an LS6500 scintillation counter (Beckman Coulter). Equal aliquots of beads with bound Rab7L1 proteins for WT, T21N or Q67L were collected for western blot analysis to estimate total protein loading.

To assay GTP/GDP retention, agarose-bound Rab7L1 proteins were incubated in Buffer A containing 0.1 µm [8-3H]-GTP or [8,5'- 3H]-GDP overnight on ice and washed twice with Buffer A

to remove unbound nucleotide. Subsequently, proteins were incubated in Buffer A containing a 100-fold excess of unlabeled GTP or GDP for 0, 10, 20 or 30 min shaking at 37°C. After each time point, Rab7L1 proteins were washed twice with Buffer A, and retaining 3H-GTP or 3H-GDP bound to proteins was quantified using scintillation counting. The amount of 3H-GTP or 3H-GDP bound at 10, 20 and 30 min for each sample was calculated as a fraction of initial binding.

Co-Immunoprecipitation

Cells were lysed in IP buffer: 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.3% Triton X-100, 10% Glycerol, 1x Halt phosphatase inhibitor cocktail (Thermo Scientific) and protease inhibitor cocktail (Roche) for 30 min on ice. Lysates were centrifuged at 4°C for 10 minutes at 20,000 xg and supernatant further cleared by incubation with EZview™ Red Protein G Affinity Gel (Sigma-Aldrich) for 30 min at 4°C. For Flag tagged constructs, lysates were incubated with EZview™ Red ANTI-FLAG® M2 Affinity Gel (Sigma-Aldrich) for 1h at 4°C and eluted after six washes with wash buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% Triton X-100, 10% Glycerol) then eluted in 1x kinase buffer (Cell Signaling), containing 150 mM NaCl, 0.02% Triton and 150 ng/µl of 3xflag peptide (Sigma-Aldrich) with shaking for 30 mins at 4°C. GFP-tagged proteins were immunoprecipitated using GFP-nAb™ Agarose resin (Allele Biotechnology) and eluted after six washes with buffer as above by boiling in 4x NuPAGE LDS sample buffer (Invitrogen).

For tertiary IP, HEK293FT lysates were immunoprecipitated with anti-Flag monoclonal antibodies as described above. Equal aliquots of the first IP were collected for each sample (first IP) and eluates from the first IP were then used for the second IP with the Chromotek-GFP-Trap resin (Allele Biotechnology) for 30 min at 4°C. Beads were washed six times with IP wash buffer and immunoprecipitated complexes were eluted by boiling in 4x NuPAGE LDS sample buffer (Invitrogen).

4

SDS PAGE and Western Blotting

Subcellular fractionation from HEK cell lysates was performed as previously described10. Other lysates were generated as described above.

Proteins were resolved on 4–20% Criterion TGX gels (Biorad) in SDS/Tris-glycine running buffer, and transferred to poly(vinylidene fluoride) (PVDF) membranes by semi-dry trans-Blot Turbo transfer system (Biorad) according to manufacturer’s instructions. Membranes were blocked with 5% nonfat milk in Tris-buffered saline containing 0.1% Tween 20 (TBST) and then incubated for 1 hour at RT or overnight at 4°C with appropriate primary antibodies including: mouse anti-Flag M2 (Sigma-Aldrich, 1:5000); mouse anti-c-Myc, clone 9E10 (Roche, 1:2000); rabbit anti-LRRK2, MJF2 (Epitomics, 1:2000), rat anti-Hsc70 (Abcam, 1:2000); rabbit anti-BAG5 (Sigma, 1:2000); rabbit anti-14-3-3 (Santa Cruz Biotechnology, 1:2000); mouse anti-V5 (Invitrogen, 1:2000); mouse anti-GFP (Roche, 1:2000); mouse anti-GAK (MBL, 1:2000); rabbit anti-MEK1/2 (Cell Signaling, 1:5000); rabbit anti-Tom20 (Santa Cruz Biotechnology, 1:2000); rabbit anti-GM130 (Novus Biologicals, 1:2000); sheep anti-TGN46 (Serotec, 1:2000);gamma-adaptin (BD Transduction Laboratories, 1:2000); actin (Sigma, 1:5000); GST (GE Healthcare, 1:5000); CHIP (Santa Cruz, 1:200); 14-3-3 (Santa Cruz, 1:2000). Membranes were washed in TBST (3×5 min) at room temperature (RT) followed by incubation for 1 h at RT with horseradish peroxidase-conjugated anti-mouse or rabbit IgG (Jackson Immunoresearch Laboratories). Blots were washed in TBST (3×5 min) at RT, immunoreactive proteins were developed using enhanced chemiluminescence plus reagent (Thermo Scientific) and imaged on a STORM 860 Molecular Imager (GE).

Immunostaining

Cells on coverslips were fixed with 4% paraformaldehyde in 1x PBS, blocked with 5% FBS in PBS with addition of 0.1% Triton. Primary antibodies were diluted in blocking buffer and incubated for 1 hour at RT. After three 5 min washes with PBS, secondary fluorescently labeled

antibodies (Invitrogen) were diluted in blocking buffer and were incubated for 1 hour at RT. Coverslips were thn washed three times with PBS, stained for TOPRO-3 nuclear dye (Invitrogen) and mounted with The ProLong® Gold antifade reagent (Invitrogen). Primary antibodies used for ICC experiments included: mouse anti-Flag M2 (Sigma-Aldrich, 1:500); mouse anti-c-Myc, clone 9E10 (Roche, 1:500); sheep anti- c-Myc (Novus Biologicals, 1:500); rabbit anti-GM130 (Novus Biologicals, 1:200); sheep anti-TGN46 (Serotec, 1:200); rabbit anti-Flag (Sigma-Aldrich, 1:500); anti-V5 (Invitrogen, 1:500); mouse anti-LAMP1 (Developmental studies hybridoma bank, 1:250); rabbit anti-Mannose 6 phosphate receptor (M6PR; Abcam, 1:200); mouse anti-VPS35 (Abnova, 1:200); rabbit anti-GLG1 (Sigma, 1:300). All secondary antibodies used for ICC were from Invitrogen and have been used at 1:500: donkey anti-mouse Alexa Fluor® 488; donkey anti-rabbit Alexa Fluor® 568; donkey anti-sheep Alexa Fluor® 488; donkey anti-mouse Alexa Fluor® 647; donkey anti-mouse Alexa Fluor® 568; donkey anti-rabbit Alexa Fluor® 488.

Golgi morphology experiments

For golgi morphology experiments, transfected cells or neurons were stained for trans-golgi markers TGN46 or GLG1. We used antibodies to TGN46 for human cell lines and GLG1 for mouse neurons and in mouse brain; this was because in our hands the TGN46 antibodies did not stain mouse tissue and, conversely, the GLG1 antibody gave no immunoreactivity in human cells. Cells were imaged on a Zeiss LSM510 Meta confocal using a 100x objective. Three Golgi phenotypes were identified: 1) Normal or diffused with multiple puncta throughout the cell; 2) Clustered or condensed with large golgi clusters, less than 3 clusters per cell; and 3) Cleared, with minimal or no golgi staining throughout the cell. Golgi morphology was counted in between 50-100 transfected cells per sample by an observer who was blinded to the transfection group (see below). Two or three independent experiments were performed and counts aggregated across experiments for a total of 100-300 cells per group.

5

Neurite length measurements

Neurons were fixed in 4% PFA in PBS and stained with either mouse anti-Flag (clone M2, Sigma, 1:500), mouse anti-GFP or mouse anti-Myc (Roche, 1:300) along with either rabbit anti-MAP2 (Santa Cruz Biotechnology, 1:500) or chicken anti-bIII tubulin (Novus, 1:500). After washing, cells were then stained with anti-mouse-Alexa Fluor® 488 and anti-rabbit-Alexa Fluor® 568 (both at 1:500 dilution, Invitrogen). Coverslips were mounted using ProLong Gold (Invitrogen) and examined using a Zeiss LSM510 Meta microscope. Cells were identified by being positive for the appropriate neuronal marker and for transfected protein of interest in a blinded fashion (see below) and images collected along the length of the longest neurite. ImageJ software (http://rsbweb.nih.gov/ij/download.html) was used to determine the neurite length in mm, by tracing the length of the longest neurite from the center of the cells to the end of the neurite. Each experiment was repeated at least three times with independent batches of cells and transfections.

Reproducibility, blinding and statistical analyses

For qualitative experiments, such as western blotting or immunoprecipitation, we repeated each experiment on a minimum of three independent occasions with independent batches of cells and transfections. For key results, e.g., confirmation of protein interaction by immunoprecipitation, results were confirmed by at least two individuals. Experiments based on counting immunostained cells, specifically counts of localization organellar morphology and neurite length, were performed in a blinded fashion. To achieve this, the individual performing the immunostaining mounted coverslips on glass slides labeled only with a numeric code for each cDNA construct. When cells were transfected or co-transfected, we used tagged control proteins (e.g., GFP or GUS) such that we measured only transfected cells in all conditions and that the blinded observer was not aware which slides represented negative control conditions. Codes were recorded and stored sealed, while a second

person performed the analysis. After counts or measurements were completed, numeric codes were replaced with identifiers for the cDNA construct prior to statistical analyses. Where possible, we set objective criteria for cell counts. Analysis of the human genetic data, including quality control parameters, were performed as described previously1,8.

All statistical analyses were performed using R 1,9. For experiments with continuous measures comparing only two groups, we used two-tailed unpaired t-tests with Welsh’s correction for unequal variance. For experiments with continuous measures using more than one group and single explanatory variables, we used one-way ANOVA with Tukey’s honest significant difference test post hoc to compare individual conditions if the overall p value was <0.05. For neurite length measurements, raw values were subjected to a log2 transformation prior to analysis. Where two variables were included in the model, e.g., for time, we used two-way ANOVA. To compare proportions with more than two categories, we used Fisher’s exact test on aggregate counts across experiments with a Bonferroni correction for the number of tests performed. To compare proportions of aligned reads between exome data and RNA-seq, we used Chi-Square test on the aggregate numbers of reads across heterozygous individuals.

Description of raw data tables

Three text files are included that detail the output from Protoarray Prospector for two sets of experiments. The first two data files are from two batches of arrays probed with truncated, GST-tagged LRRK2. In the first data file, three conditions are compared – hybridization buffer alone, GST alone and GST-LRRK2970–2527. In the second data file, two conditions ae compared, GST-LRRK2970–2527 was incubated with either GDP or GMPPcP. For reference the GST image from the first hybridization batch is included in this file. The third data file includes two arrays that were hybridized in parallel with either Flag-tagged eGFP alone or Flag-tagged, full length LRRK2.

6

The upper three blocks of the files include details of the array files used; lot numbers of the arrays and detection wavelengths for fluorescence; and values for the control spots on each array.

The fourth block in each file has the following parameters:

Database ID, Ultimate ORF ID, Description (for file 3 only): Descriptive text for each recombinant protein on the array.

Block, Row, Column, Protein Amount: Location of protein spot on the array

File1 Hit: Flags for potential hits based on Z-score >3

File1 Signal.Used: Raw signal averaged from duplicate spots on the array

File1 Z-Factor : mean value Z-factor for all replicates of a given probe on this array

File1 Z-Score: Measure of the binding of a given protein relative to mean of all protein features on the array, in units of standard deviations

File1 CI P-Value: Greatest value Chebyshev’s Inequality P-value for all replicates of a given probe on each array

File1 CV: Coefficient of variance for a probe on the array

File1 GenePix Flags: Inherited from GenePix

Supplementary References

1. Civiero, L. et al. Biochemical characterization of highly purified leucine-rich repeat kinases 1 and 2 demonstrates formation of homodimers. PLoS ONE 7, e43472 (2012).

2. Greggio, E. et al. The Parkinson disease-associated leucine-rich repeat kinase 2 (LRRK2) is a dimer that undergoes intramolecular autophosphorylation. J. Biol. Chem. 283, 16906–16914 (2008).

3. Kalia, S. K. et al. BAG5 inhibits parkin and enhances dopaminergic neuron degeneration. Neuron 44, 931–945 (2004).

4. Lee, D.-W., Wu, X., Eisenberg, E. & Greene, L. E. Recruitment dynamics of GAK and auxilin to clathrin-coated pits during endocytosis. J. Cell. Sci. 119, 3502–3512 (2006).

5. McCoy, M. K., Kaganovich, A., Rudenko, I. N., Ding, J. & Cookson, M. R. Hexokinase activity is required for recruitment of parkin to depolarized mitochondria. Hum. Mol. Genet. (2013).doi:10.1093/hmg/ddt407

6. Jiang, M. & Chen, G. High Ca2+-phosphate transfection efficiency in low-density neuronal cultures. Nat Protoc 1, 695–700 (2006).

7. Ibrahimi, A. et al. Highly efficient multicistronic lentiviral vectors with peptide 2A sequences. Hum. Gene Ther. 20, 845–860 (2009).

8. Baekelandt, V. et al. Characterization of lentiviral vector-mediated gene transfer in adult mouse brain. Hum. Gene Ther. 13, 841–853 (2002).

9. Daniëls, V. et al. Insight into the mode of action of the LRRK2 Y1699C pathogenic mutant. J. Neurochem. 116, 304–315 (2011).

10. Lobbestael, E. et al. Immunohistochemical detection of transgene expression in the brain using small epitope tags. BMC Biotechnol. 10, 16 (2010).

11. Rudenko, I. N. et al. The G2385R variant of leucine-rich repeat kinase 2 associated with Parkinson’s disease is a partial loss-of-function mutation. Biochem. J. 446, 99–111 (2012).

12. International Parkinson Disease Genomics Consortium et al. Imputation of sequence variants for identification of genetic risks for Parkinson’s disease: a meta-analysis of genome-wide association studies. Lancet 377, 641–649 (2011).

7

Fig. S1. Biotinylated LRRK2 is kinase active.

Truncated, GST-tagged LRRK2 (lane 1) was dialyzed and concentrated (lane 2) then biotinylated (lane 3) and subjected either to autophosphorylation and autoradiography (upper panel) or blotting for GST (lower panel). Autophosphorylation was detected in all samples; ratios of 33P to GST are shown below each lane. Data is representative of duplicate biotinylation experiments. Markers on the right of each panel are in kilodaltons.

8

Fig. S2. Additional co-immunoprecipitation data for BAG5

(a) LRRK2 variants interact with BAG5. Cells lines stably expressing Flag-tagged LRRK2 variants were lysed and subjected to immunoprecipitation with FLAG-beads and probed for LRRK2 (top blot), BAG5 (middle) or 14-3-3 proteins (lower blot) as a positive control for LRRK2 interaction. Vector only cells are a negative control, shown in the first and fifth lanes.

(b). Endogenous LRRK2 and BAG5 were co-immunoprecipitated from mouse brain.

(c). BAG5 interaction is independent of LRRK2 kinase activity. Stable HEK293 cell lines expressing indicated Flag-tagged LRRK2 variants were treated with DMSO or 10 mM LRRK2-IN-1. Immunoprecipitated complexes were blotted for LRRK2 (upper panel), endogenous BAG5 (middle panel) or endogenous 14-3-3 proteins (lower panel).

(d,e). Sequence requirements. (d) V5-tagged WT or DARA mutant BAG5 (closed arrowheads) co-immunoprecipitated with GFP-tagged LRRK2. (e) Myc tagged domain constructs for LRRK2 co-immunonoprecipitated with Flag-tagged BAG5 (closed arrowhead). R/C/K is ROC/COR/kinase.

9

(f). BAG5 interacts with LRRK1 and LRRK2. Immunoprecipitated complexes from cells expressing Flag-tagged GUS, LRRK1 or LRRK2 were blotted sequentially for Flag, BAG5, Hsp70 or 14-3-3 (from top to bottom). Inputs are shown in lanes 1-4 and immunoprecipitated complexes in lanes 5-8.

(g) Addition of BAG5 does not alter LRRK2 autophosphorylation. Full length LRRK2 was subjected to autophosphorylation reaction in vitro (upper panel showing 33P) either alone (lane 1) or in the presence of wild type (lane 2) or DARA mutant (lane 3) BAG5. Lower panel shows Ponceau staining of the membrane for loading.

Fig. S3. GST pulldowns as an additional validation technique.

(a) GST-pulldowns confirm BAG5 interaction with LRRK2. Lysates from cells expressing myc-tagged LRRK2 were subjected to GST pulldowns with GST alone or GST-BAG5. Upper panel shows Myc blot, lower panel shows Ponceau staining for total lysates and GST fusion proteins.

(b) Pull downs were performed with GST or GST-BAG5 from cells transfected with myc tagged LRRK2 domain constructs. Immunoprecipitates were blotted for Myc (upper panel) or stained with Ponceau (lower panel) to show equal loading.

10

Fig. S4. Additional protein interaction data for BAG5, GAK and Rab7L1

(a) Purification of BAG5. Wild type BAG5 (lane 1) or DARA mutant (lane 2) were purified using Flag immunopurification and complexes stained with Coomassie blue. Proteins in the upper band at ~70 kDa were isolated from this gel and identified by mass-spectrometry. Right panel shows a western blot using anti-Hsp70.

11

(b) A copurified kinase in preparations of wild type BAG5. Flag-purified WT or DARA- BAG5 were visualized with Ponceau-S staining (left panel) or subjected to autoradiography after incubation with kinase assay buffer and 32P-ATP (right panel). Arrow indicates BAG5 and the asterisk shows an ~150 kDa autophosphorylated kinase.

(d) BAG5 and LRRK2 both bind to the PTEN-like domain of GAK. Immunoprecipitated complexes with GFP-tagged GAK domains probed for endogenous LRRK2, Hsc70 and BAG5.

(c) Endogenous BAG5 and GAK interact in vivo. Mouse brain samples co-immunoprecipitated for BAG5 and probed for endogenous GAK (upper panel) and BAG5 (lower panel). IgG is a negative control.

(e) Knockdown of BAG5 does not disrupt interactions between GAK and LRRK2. HEK cells were co-transfected with scrambled siRNA or siRNA against BAG5 with no additional plasmids, GFP or GFP-GAK. Inputs (lanes 1-9) and samples after immunoprecipitation (lanes 10-18) were blotted sequentially for endogenous LRRK2, Hsc70, and BAG5 as well as transfected GFP proteins.

(f) Lack of GAK does not prevent LRRK2 from interacting with BAG5. HEK293FT cells stably expressing control shRNA or shRNA against GAK were transfected with Flag-vector, WT or DARA mutant BAG5. Lysates (lanes 1-3 and 7-9) and complexes immunoprecipitated with antibodies to Flag (lanes 4-6 and 10-12) were blotted with antibodies to LRRK2 and GAK (upper panel) and BAG5 and Hsc70 (lower panel). The middle lane (M) shows markers.

(g) Lack of LRRK2 does not prevent GAK from interacting with BAG5. HEK293FT cells stably expressing control shRNA or shRNA against LRRK2 were transfected with Flag-vector, WT or DARA mutant BAG5. Lysates (lanes 1-3 and 7-9) and complexes immunoprecipitated with antibodies to Flag (lanes 4-6 and 10-12) were blotted with antibodies to LRRK2 and GAK (upper panel) and BAG5 and Hsc70 (lower panel).

12

Fig. S5. Mapping of interaction of Rab7L1 to LRRK2.

(a) Binding between WT Rab7L1 and LRRK2 mutants was assessed by co-immunoprecipitation of Myc-tagged Rab7L1 with the indicated Flag-tagged LRRK2 variants. Upper panel shows inputs for Flag and Myc, lower panels show samples after immunoprecipitation probed for Flag (asterisk in the lower Myc blot shows a cross-reactive band of unknown identity).

(b) Quantification was performed by comparing the amount of Rab7L1 to the amount of LRRK2 in the IP. There were no significant differences between variants (F(9,19)=0.639, p=0.751 by one-way ANOVA, n=3 independent experiments). Molecular weight markers on the right of all blots are in kilodaltons.

(c) Rab7L1 binds to the HEAT domain of LRRK2. Tagged (3xFlag) Rab7L1 immunoprecipitates full length 2xMyc-LRRK2 (upper blot, lane 6) and 2xMyc-HEAT domain of LRRK2 (upper blot, lane 8), but not 2xMyc-gus (upper blot, lane5) and diminishes binding to 2xMyc-deltaHEAT LRRK2 (upper blot, lane7). Lysates are shown in lanes 1-4 (upper blot). Samples were blotted for Flag (lower blot) to show immunoprecipitated Rab7L1.

(d) Quantifications for (c) were performed by estimating the ratio between immunoprecipitated LRRK2 to the levels of LRRK2 in inputs. Differences in interaction strength were analzed by one-way ANOVA for all variants with Tukey’s HSD post-hoc test compared to GUS negative control (*, p<0.05; ***, p<0.001; ns= not siginificant; n=3 independent experiments).

13

Fig. S6. Characterization of Rab7L1 mutants.

14

(a-c) Structural model of wild type Rab7L1 (a) showing key residues around the bound non-hydrolysable GTP analog, GppNHp that were mutated to influence function. These include a mutation that should disrupt Mg2+ co-ordination, T21N (b); and a key base for co-ordination and hydrolysis of GTP, Q67L (c).

(d) GTP binding for all variants was measured by loading with radiolabelled GTP and is normalized to total protein loading estimated by western blotting. Each bar represents the mean and SEM between four independent experiments. Statistical significance was estimated by one-way ANOVA with Tukey’s post-hoc test; ***, p<0.001.

(e, f) Guanosine nucleotide retention assays. Rab7L1 WT (solid lines) or Q67L (dashed lines) were loading with radiolabelled GTP (e) or GDP (f) and retention measured as a fraction of initial binding (y axes) as times up to 30 minutes (x axes). Each point is the mean and SD from three independent experiments. Statistical significance was estimated by two-way ANOVA for time versus mutation with Tukey’s post-hoc test for comparison to wild type protein at each time point (**, p<0.01; ***, p<0.001).

(g,h) Cell lysates were separated by size exclusion chromatography from highest to lowest retention time (fractions 1-30) and blotted for Flag-tagged Rab7L1 with (from top to bottom) the WT, T21N or Q67L variants. Quantification across n=3 independent experiments (h) showed accumulation in early fractions for WT Rab7L1 but not T21N or Q67L.

(i,j) Interaction of LRRK2 with Rab7L1 is increased with guanosine nucleotide binding deficient Rab7L1 mutations. HEK293FT cells were transfected with Flag-tagged b-glucoronidase (GUS) or WT or Q67L Rab7L1 were immunoprecipitated for Flag and blotted for endogenous LRRK2 (upper blots) or transfected Flag-tagged proteins (lower panels). Markers on the right of the blots are in kilodaltons. Quantification (j) of the amount of LRRK2 relative to Rab7L1 in the immunoprecipitated complexes was performed for the indicated variants and statistical analysis performed using one-way ANOVA with Tukey’s post hoc test (n=3 independent experiments; ***, p<0.001 vs. WT Rab7L1).

(k) WT but not mutant Rab7L1 localizes to the trans-Golgi. Cells were transfected with WT (upper panels), T21N (middle panels) or Q67L (lower panels) Rab7L1 (red) and stained with two golgi markers (green), GM130 (left panels) or TGN46 (right panels). Cells were counterstained with TOPRO-3 to show the nucleus (blue in all images). Scale bar indicates 10 mm.

15

Fig. S7. Neurite shortening phenotypes and expression of LRRK2 binding partners

(a,b) Representative images of neurite length in BAG5 transfected neurons. (a) Primary neurons transfected with GFP, WT or DARA BAG5, stained with GFP or BAG5 (green) and MAP2 (red). Scale bars represent 50 mm.(b) Primary cortical neurons co-transfected with RFP and either GFP (upper panels), wild type BAG5 (middle panels) or the Hsp70-deficient DARA mutant (lower panels) and stained with GFP or BAG5 (left panels, green in merge) and RFP. Right panels show merged images with nuclear staining (blue). Scale bars represent 50 mm.

(c-e) Neurite length was measured in transfected neurons for BAG5 (c), GAK (d) and Rab7L1 (e) constructs. Numbers in each box are the number of cells measured across 2-3 independent cultures. **, p<0.01; ***, p<0.001 by one-way ANOVA with Tukey’s post-hoc test compared to WT after log(2) transformation.

(f) Quantitation of neurite length per cell in transfected cells shows a significant difference in neurite length for WT Rab7L1 in with type (+/+) neurons but not in LRRK2 knockout (-/-) neurons. ***=p<0.0001; ns=not-significant; compared to WT Rab7L1 for the same genotype.

16

Fig. S8. LRRK2 complex components co-localize to vesicular compartments of neurons

(a) Subcellular fractionation. Lysates (lane 1) from HEK293FT cells transfected with Flag-tagged LRRK2 were separated into cytosolic (lane 2), mitochondrial (lane 3) and microsomal (lane 4) fractions and blotted for endogenous proteins. Markers on the right of the blots are in kilodaltons.

(b) Localization of tagged proteins in neurons. Insets show examples of vesicular staining for LRRK2, GAK and Rab7L1. Scale bar indicates 10 mm.

(c) Colocalization of LRRK2 with GAK or Rab7L1 in neurons. Flag-tagged LRRK2 was co-transfected with myc-tagged GAK, WT or Q67L Rab7L1. Insets show vesicular structures and plots below each image are of staining intensity.

17

Fig. S9. Additional data related to clustering and clearance of Golgi.

(a) LRRK2 (red) and Rab7L1 (blue) in transfected alone or together into HEK cells that were stained for the trans-Golgi marker TGN46 (green, upper panels) or the cis-Golgi marker GM130 (green, lower panels). Note partial colocalization with GM130 but stronger colocalization with TGN46, which also shows disrupted morphology when the two proteins are cotransfected. Panels on the right show an enlarged cell, outlined, with clustered TGN46 staining that is also positive for Rab7L1 and LRRK2 (upper panel; lower panel shows TGN46 alone). Two adjacent cells that were not co-transfected are also shown with normal TGN46 staining (asterisks). Scale bar indicates 10 mm.

(b) These effects were quantified by counting cells with relocalization of LRRK2 and statistical analysis performed using one-way ANOVA with Tukey’s post hoc test (n=3 independent experiments with 100 cells counted per experiment; **, p<0.01; ***, p<0.001 vs. WT Rab7L1 for each marker separately).

18

(c) LRRK2 positive structures (green) in cells co-transfected with LRRK2 and Rab7L1 do not colocalize with retromer markers (red) including the cation-independent mannose-6-phosphate receptor (M6PR, left image) or VPS35 (right image). Scale bars indicate 10 mm.

Fig. S10. Biochemical confirmation of turnover of trans-Golgi.

(a) Western blot to confirm turnover of TGN46, but not GM130, in LRRK2 transfected cells. Cells were transfected with Flag-tagged GUS (lanes 1-5) or LRRK2 (lanes 6-10). Twenty-four hours after transfection, cell lysates were blotted sequentially for TGN46, GM130, FLAG for transfected proteins and b-actin as loading control. Markers on the right are in kilodaltons.

(b,c) Graphs show ratio of TGN46 to b-actin (b) or GM130 to b-actin (c) from n=5 individual transfections. ***, p<0.001; ns, not significant (p=0.07), two sample t-test with Welsh’s correction for unequal variance. Representative of duplicate independent experiments with similar sample numbers.

19

Fig. S11. Time course of recruitment of LRRK2 to, and clearance of, the trans-Golgi network.

(a) Cells were transfected with LRRK2 and Rab7L1 and imaged after staining for TGN46 (green) at the indicated times after transfection.

(b) Blinded counts (proportions on the left y-axis) of cells with normal Golgi (blue), clustered Golgi (orange) or cells with minimal or no apparent Golgi stained (white) at the times indicated on the x-axis. Recruitment of LRRK2 was also scored in the same cells (black bars).

20

Fig. S12. Additional data related to mechanism of Golgi turnover by lysosomes

(a) Bafilomycin A1 prevents turnover of TGN-46 vesicles in the presence of LRRK2 with GAK. Cells were co-transfected with LRRK2 and GAK and treated with DMSO (left panels) or 10nM bafilomycin A1 (right panels). Upper panels show staining for LRRK2 (red) and GAK (blue) to identify co-transfected cells (indicated by outlines), lower panels show only TGN46 stain in the same cells.

(b) Effect of bafilomycin A1 on relative localization of LRRK2 and lysosomes.

Cells expressing wild type LRRK2 were treated with DMSO (control; left panels) or 10 nM bafilomycin-A1 (right panels) for 24h and stained for LRRK2 (red, upper panels) and LAMP1 (green, middle panels). After bafilomycin-A1 treatment, vesicles containing LRRK2 (arrow) are seen adjacent to LAMP1-positive lysosomes (asterisk).

21

(c) LRRK2/Rab7L1 vesicles surround lysosomes after bafilomycin-A1 treatment. Cells were transfected with GUS (left) or LRRK2 (right) and treated with 10 nM bafilomycin-A1 for 24h, then stained for Rab7L1 (red), LRRK2 (blue) and LAMP1 (green; merged images are shown in the bottom set of panels).

(d) LRRK2 vesicles engulf trans-Golgi. Cells expressing LRRK2 (red) with Rab7L1 (blue) were treated with bafilomycin A1 (10 nM, 24h) and stained with TGN46 (green). Scale bars represent 10 mm.

22

Fig. S13. Additional human genetic data.

Locus plots for PD risk for markers around Rab7L1/PARK16 (a), GAK (b) or LRRK2 (c) showing markers on the x-axes against –log10 of p values for association with PD risk (y axes). Each point is colored by strength of linkage disequilibrium with the most highly significant marker in the region. The orange line shows recombination rate in centimorgans and positions of known genes in the region are shown below the plot in green.

23

Fig. S14. Proposed model for LRRK2 complex formation and role in turnover of TGN

Schematic of major vesicular sorting pathways within a mammalian cells showing secretory pathway from the TGN and endocytosis from the plasma membrane. Overlaid on this schematic are our proposed regions of localization of LRRK2 (blue), GAK (green) and Rab7L1 (red), suggesting co-complex formation around the TGN and mature endosomes. Once formed, this complex helps traffic golgi-derived vesicles through the autophagy-lysosomal system. Adapted in part from models in Cullen and Korswagen12 and from data in13

24

Table S1 Candidate protein interactors for truncated LRRK21

Z score

Database ID Name Gene symbol

Buffer GST GST-LRRK2

GST-LRRK2 + GDP

GST-LRRK2 + GMPPcP

BC050551 BCL2-associated athanogene 5

BAG5 -0.001 0.6394 31.33 68.41 45.99

BC020221 SH3 and cysteine rich domain

STAC 0.161 2.785 15.66 3.754 3.626

NM_004281 BCL2-associated athanogene 3

BAG3 -0.062 -0.073 14.95 23.95 16.97

NM_006819 Stress-induced-phosphoprotein 1 (Hsp70/Hsp90-organizing protein)

STIP1 -0.574 2.078 10.90 24.27 26.56

NM_153498 Calcium/calmodulin-dependent protein kinase

CAMK1D

-0.113 2.135 8.137 4.873 3.934

BC036089 Myeloid/lymphoid or mixed-lineage leukemia

MLLT3 -0.005 0.5998 5.139 4.485 7.208

NM_004282 BCL2-associated athanogene 2

BAG2 -0.232 -0.222 3.555 10.25 7.602

BC056415 RNA polymerase II associated protein 3

RPAP3 -0.214 2.632 3.075 3.024 4.694

1Candidate interactors were filtered where Z>3 for LRRK2 in all experiments and <3 for GST and buffer

25

Table S2 Candidate protein interactors for full length LRRK2 Z score

Database ID Name Gene symbol GFP LRRK2

NM_003406.2 tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide

YWHAZ -0.02109 14.04168

NM_017949.1 CUE domain containing 1 CUEDC1 0.83579 12.31015

NM_004281.2 BCL2-associated athanogene 3 BAG3 1.30739 10.81284

BC068456.1 14-3-3 protein zeta/delta YWHAZ/D -0.29605 9.93861

NM_003404.2 tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, beta polypeptide

YWHAB 0.62477 8.46554

NM_004282.2 BCL2-associated athanogene 2 BAG2 0.12599 7.64759

NM_006142.1 stratifin SFN 0.08443 7.21162

PV3502 p21(CDKN1A)-activated kinase 6 PAK6 1.06919 6.71068

BC015818.1 lectin, galactoside-binding, soluble, 8 (galectin 8)

LGALS8 1.92125 6.22942

NM_004873.1 BCL2-associated athanogene 5 BAG5 0.36259 5.45035

NM_006007.1 zinc finger, AN1-type domain 5 ZFAND5 2.41843 5.33141

PHC1695 C-X-C motif chemokine 11 CXCL11 2.68541 4.87759

NM_145010.1 chromosome 10 open reading frame 63

C10orf63 0.56561 4.44299

NM_012148.1 double homeobox, 3 DUX3 0.2267 4.32587

NM_012387.1 peptidyl arginine deiminase, type IV PADI4 0.88694 4.29751

XM_379498.2 PREDICTED: Homo sapiens hypothetical LOC401363

LOC401363 1.52639 4.12505

NM_144736.3 Protein midA homolog, mitochondrial

NDUFAF7 2.3465 4.08204

NM_018270.3 MRG-binding protein MRGBP 1.81255 4.01571

BC053836.1 Vasohibin-2 VASH2 1.93405 3.96401

NM_181526.1 myosin, light chain 9, regulatory MYL9 1.57116 3.96127

BC050551.1 BCL2-associated athanogene 5 BAG5 0.44732 3.90867

BC008668.1 cyclin G associated kinase GAK 1.15871 3.90226

BC047776.2 coiled-coil domain containing 43 CCDC43 1.2882 3.72293

BC022888.1 xin actin-binding repeat containing 2 CMYA3 0.53045 3.62777

NM_001819.1 chromogranin B (secretogranin 1) CHGB 0.65034 3.62686

BC002914.1 WAS/WASL-interacting protein family member 1

2.09232 3.55733

26

NM_003929.1 RAB7, member RAS oncogene family-like 1

RAB7L1 1.16831 3.48276

BC007067.1 Kinetochore-associated protein NSL1 homolog

NSL1 1.67187 3.46034

BC006177.1 Metastasis-associated protein MTA1 1.19548 3.44433

NM_004092.2 Enoyl-CoA hydratase, mitochondrial ECHS1 1.6383 3.43609

BC014774.1 BCL2-associated athanogene BAG1 0.45691 3.43015

NM_025104.2 Protein DBF4 homolog B DBF4B 1.83174 3.37113

NM_014904.1 RAB11 family interacting protein 2 (class I) ()

RAB11FIP2 1.24823 3.37022

BC035026.2 family with sequence similarity 47, member B

FAM47B 1.03882 3.33362

BC002369.1 Serine/threonine-protein kinase PLK1 2.71898 3.20003

NM_032017.1 serine/threonine kinase 40 STK40 1.0596 3.13874

BC032347.1 chromosome 8 open reading frame 59

C8orf59 1.56636 3.1241

PV4202 NIMA (never in mitosis gene a)-related kinase 1

NEK1 1.20667 3.08887

NM_005659.1 ubiquitin fusion degradation 1 like (yeast)

UFD1L 1.85731 3.07149

NM_138558.1 protein phosphatase 1, regulatory (inhibitor) subunit 8

PPP1R8 0.815 3.01796

PV3370 megakaryocyte-associated tyrosine kinase

MATK 1.54718 3.00607

NM_005861.2 STIP1 homology and U box-containing protein 1

STUB1 -0.18894 2.94248

1Candidate interactors were filtered where Z>3 for LRRK2 and <3 for GFP