Novel ethyl methanesulfonate (EMS)-induced null alleles of the Drosophila homolog of LRRK2 reveal a crucial role in endolysosomal functions and autophagy in vivo

Parkinson’s disease (PD) is a common and devastating neurodegenerative movement disorder. The leucine-rich repeat kinase 2 (LRRK2) gene is an important therapeutic target for PD because it is the most common known genetic determinant of the disease (Dawson et al., 2010; Kett and Dauer, 2012). LRRK2 mutations were originally identified as the causative factor in dominantly inherited forms of PD linked to the PARK8 locus (Paisán-Ruíz et al., 2004; Zimprich et al., 2004) and, more recently, sequence variation at the LRRK2 locus has been associated with an increased risk of developing sporadic PD in genome-wide association studies (Satake et al., 2009; Simón-Sánchez et al., 2009). LRRK2 encodes a large multi-domain protein characterized by leucine-rich repeats, a GTPase domain and a kinase domain (Bosgraaf and Van Haastert, 2003). The cellular functions of LRRK2 remain unclear because it has been linked to multiple diverse cellular processes, including mitochondrial function (Smith et al., 2005), regulation of transcription (Kanao et al., 2010) and translation (Gehrke et al., 2010; Imai et al., 2008; Martin et al., 2014), Golgi protein sorting (Sakaguchi-Nakashima et al., 2007), apoptosis (Ho et al., 2009), and regulation of the dynamics of actin (Jaleel et al., 2007; Parisiadou et al., 2009) and microtubules (Gandhi et al., 2008; Gillardon, 2009; Kett et al., 2012; Lin et al., 2009).

Understanding the normal cellular functions of LRRK2 is vital because the mechanisms mediating the pathogenicity of mutant forms of LRRK2 are likely to be related to the physiological functions of the wild-type protein. Moreover, although inhibitor-based therapies targeting LRRK2 have emerged as a prime therapeutic target in PD (Lee et al., 2012), the effects of inhibiting endogenous LRRK2 are not clear. Several recent studies have suggested a role for LRRK2 and its homologs in lysosomal function and in membrane trafficking in the endolysosomal and autophagy pathways; however, these interpretations derive largely from studies of overexpression of LRRK2 and its pathological mutant forms. This raises the question of whether endogenous LRRK2 plays a role in endolysosomal processes under physiological conditions.

Despite the diversity of cellular functions to which LRRK2 has been linked, LRRK2 and its homologs have been consistently found to localize to intracellular membranes in the endolysosomal pathway (Alegre-Abarrategui et al., 2009; Berger et al., 2010; Biskup et al., 2006; Hatano et al., 2007; Higashi et al., 2009; Shin et al., 2008). We have previously reported that the protein encoded by the Drosophila homolog of LRRK2, lrrk, localizes specifically to the membranes of late endosomes and lysosomes, and to a lesser extent to early endosomes, in vivo (Dodson et al., 2012). We have found that Lrrk physically binds to the late endosomal protein Rab7, and overexpression of a PD-causing mutant form of lrrk results in Rab7-mediated lysosomal positioning defects (Dodson et al., 2012). Interestingly, mammalian LRRK2 has recently been found to interact physically with Rab7L1 (Beilina et al., 2014; MacLeod et al., 2013), a homolog of Rab7 (Shimizu et al., 1997). Via this interaction, LRRK2 has been linked in overexpression-based studies to retromer-dependent endosome-to-Golgi membrane transport via interactions with Vps35 (MacLeod et al., 2013), and to Rab7L1-dependent lysosomal clearance of Golgi-derived vesicles (Beilina et al., 2014). Moreover, multiple studies have reported that overexpression of either wild-type or pathogenic mutant forms of LRRK2 result in the accumulation of aberrant lysosomal and autophagosomal structures (Alegre-Abarrategui et al., 2009; Bravo-San Pedro et al., 2012; Gómez-Suaga et al., 2012; MacLeod et al., 2006; MacLeod et al., 2013; Orenstein et al., 2013; Plowey et al., 2008; Ramonet et al., 2011).

Although these studies suggest a role for LRRK2 in endolysosomal processes, what is lacking is strong evidence for a physiological role for LRRK2 in endolysosomal functions in vivo. The strongest evidence to date comes from analysis of renal epithelial cells in LRRK2 knockout mice, showing changes in the numbers of autophagosomes and autolysosomes, as well as accumulation of lipofuscin granules, which can be suggestive of lysosome dysfunction (Tong et al., 2012). Loss-of-function of Drosophila lrrk has been reported to cause defects in synaptic vesicle endocytosis at the larval neuromuscular junction (NMJ) that is associated with decreased uptake of the tracer FM1-43, which is internalized via endocytosis (Matta et al., 2012). However, this experimental system is limited because the NMJ lacks the resolution for detailed subcellular analysis. In our previous work, we used the best-characterized Drosophila lrrk loss-of-function allele, which results from insertion of a PiggyBac transposable element within a lrrk intron (Imai et al., 2008; Lee et al., 2007; Matta et al., 2012; Wang et al., 2008), and found only mild defects in the endolysosomal pathway under endogenous conditions. Our data suggested that endogenous lrrk does play a role in regulating lysosome positioning, but only in a sensitized genetic system in which a constitutively active form of rab7 is expressed (Dodson et al., 2012).

Because proper functioning of the endolysosomal pathway is crucial for cellular homeostasis, it is puzzling that stronger phenotypes have not been detected in the context of loss-of-function of LRRK2 or its homologs, if indeed it plays a vital role in these processes. Thus, we hypothesized that the published Drosophila lrrk allele might not uncover the full range of phenotypes caused by lrrk loss-of-function. To address this issue, we have generated multiple novel alleles of Drosophila lrrk with in-frame stop codons, which should thus function as true null alleles. We report here striking phenotypes in both the endolysosomal and autophagy pathways in these flies.

The best-characterized loss-of-function mutation in Drosophila lrrk, lrrk^e03680, results from a PiggyBac insertion in a lrrk intron (Fig. 1A) (Imai et al., 2008; Lee et al., 2007; Matta et al., 2012; Wang et al., 2008). Previously, we reported a reduction in lrrk transcript levels caused by the lrrk^e03680 allele by quantitative RT-PCR (Dodson et al., 2012). However, in a more detailed analysis using multiple primer pairs distributed across the lrrk genomic region, we found heterogeneity in lrrk transcript abundance in lrrk^e03680 flies (Fig. 1A). For example, primer pairs spanning the exon 6/7 and exon 9/10 splice junctions yielded a transcript abundance of about 25% of control in lrrk^e03680 homozygous flies, whereas primer pairs spanning the exon 7/8 junction yielded an abundance of 90% of control, and primers spanning the exon 4/5 junction gave a 36% increase in transcript abundance (Fig. 1A). These results were highly reproducible across multiple replicates, and suggest the presence of multiple transcripts that are differentially affected by the PiggyBac insertion, and indeed that some of these transcripts might be upregulated in lrrk^e03680 flies.

These results suggested the possibility that the lrrk^e03680 allele does not represent a true null owing to differentially expressed alternative isoforms. To generate new lrrk nonsense alleles, we used TILLING, a high-throughput method by which point mutations in a gene of interest are identified in a population of chemically mutagenized flies (Till et al., 2003). This yielded four ethyl methanesulfonate (EMS)-induced nonsense mutations in lrrk (these alleles are hereafter collectively referred to as lrrk NS). The truncated proteins encoded by all four of the lrrk NS alleles are predicted to lack the kinase domain, and three of the four would additionally lack the GTPase domain (Fig. 1B). In contrast to the lrrk^e03680 allele, the lrrk NS alleles consistently and reproducibly yielded transcript abundances that were either not significantly changed compared with wild type, or that were modestly reduced to 60–80% of control (Fig. 1A), suggesting that the transcript abundance of the different lrrk isoforms is less differentially affected by the NS alleles than by the lrrk^e03680 PiggyBac insertion. Unfortunately, using the only published antibody to Drosophila Lrrk (Imai et al., 2008), we were unable to detect a specific Lrrk signal in wild-type flies, nor in lrrk overexpression flies, under multiple experimental conditions. Thus, we are unable to assess the net effect of these alleles on Lrrk protein abundance.

Because alleles generated by EMS mutagenesis are likely to also carry background mutations, we exclusively focused our studies on trans-heterozygous combinations of these alleles because they would be highly unlikely to carry similar EMS-induced background mutations. All phenotypes described below were highly similar if not identical in all trans-heterozygous combinations of these four nonsense alleles. lrrk NS flies did not show any discernible defects in dopaminergic neuron number or morphology with age (supplementary material Fig. S1), which is consistent with what has previously been reported for lrrk^e03680 flies (Wang et al., 2008). This is not unexpected given that dominant mutations in LRRK2, rather than recessive loss-of-function mutations, cause PD.

Previously, we and others have reported that lrrk^e03680 homozygous females show a dramatic reduction in female fertility (Dodson et al., 2012; Imai et al., 2008; Lee et al., 2007), which can be rescued by restoring wild-type lrrk function specifically in follicle cells (Dodson et al., 2012). Follicle cells are a somatic epithelial cell monolayer that surrounds the developing oocyte during oogenesis (Fig. 2A) (King, 1970; Spradling, 1993). Follicle cells have a number of advantages that make them an attractive cell biological system, including their large size, ease of accessibility and squamous morphology that aids in visualization of subcellular structures. As expected, all trans-heterozygous combinations of lrrk NS alleles showed reduced female fertility that could be rescued by follicle-cell-specific expression of wild-type lrrk (Fig. 1C). Expression of lrrk^GS, which is analogous to the most common PD-causing mutation in human LRRK2 (G2019S) (Bonifati, 2007; Taylor et al., 2006), also suppressed female infertility in lrrk NS flies (Fig. 1C). This suggests that lrrk^GS retains at least some of the functions of wild-type lrrk, consistent with what we have previously reported (Dodson et al., 2012). Moreover, follicle-cell-specific expression of human LRRK2 also restored fertility to lrrk NS flies (Fig. 1C), suggesting that the human and Drosophila proteins are functionally conserved.

Egg chamber development in Drosophila is divided into 14 stages, and follicle cells normally undergo programmed cell death at the culmination of egg chamber development (King, 1970; Spradling, 1993). In lrrk NS flies, however, TUNEL staining revealed premature follicle cell death at stages 11–13 (Fig. 2C vs 2B; Fig. 2F). As with all other lrrk NS follicle cell phenotypes reported below, there was marked heterogeneity in the number of TUNEL-positive follicle cells per egg chamber, with some lrrk NS egg chambers severely affected and others appearing completely normal. Thus, quantification has been employed for lrrk NS follicle cell phenotypes where possible, such that the magnitude of the phenotype is averaged over all egg chambers. Premature follicle cell death in lrrk NS flies was rescued by targeted expression of lrrk or human LRRK2 in follicle cells, and by a single copy of a lrrk genomic rescue transgene (Dodson et al., 2012) (Fig. 2F). lrrk^e03680 homozygous flies did not show premature follicle cell death (supplementary material Fig. S2C), which is consistent with our hypothesis that this allele might not uncover the full range of lrrk loss-of-function phenotypes. Premature follicle cell death in lrrk NS flies was associated with increased staining for activated caspase-3 (called Drice in Drosophila) (Fig. 2E vs 2D), and double labeling with activated caspase-3 and TUNEL confirmed that caspase activation was specific to dying cells (Fig. 2E″ vs 2D″). Moreover, follicle cell death in lrrk NS flies was suppressed by expression of the caspase inhibitor Drosophila inhibitor of apoptosis 1 (diap-1) (Hay et al., 1995) (Fig. 2F). Collectively, these data suggest that premature follicle cell death in lrrk NS flies is caspase-dependent.

As we have previously reported, Lrrk localizes predominantly to the membranes of late endosomes and lysosomes, and to a lesser extent to early endosomes, in vivo (Dodson et al., 2012). We thus first asked whether premature follicle cell death in lrrk NS flies is associated with defects in the late endosomal and lysosomal compartments. Using multiple late endosomal and lysosomal markers, including Rab7 (Fig. 3C vs 3B), Lamp1:GFP (Fig. 3C′ vs 3B′) and the acidophilic dye Lysotracker (Fig. 3E vs 3D), we found that many lrrk NS follicle cells showed markedly enlarged late endosomal/lysosomal structures. There did not seem to be any appreciable alteration in lysosome position in lrrk NS flies: the enlarged lysosomes occurred in both the perinuclear region as well as the cell periphery. Expression of the caspase inhibitor diap1 had no significant effect on lysosome enlargement in lrrk NS flies (supplementary material Fig. S3), suggesting that lysosome enlargement is not a consequence of caspase activation. In contrast to the massively enlarged lysosomal structures observed in lrrk NS flies, lrrk^e03680 homozygous flies showed a mild increase in the size of the Rab7-positive late endosomal compartment in a subset of cells (supplementary material Fig. S2F) and no discernible change in lysosome morphology by Lysotracker staining (supplementary material Fig. S2I) (Dodson et al., 2012). Interestingly, the massive accumulation of Lamp1:GFP within enlarged lysosomes in lrrk NS flies (Fig. 3C′) suggests that their expansion in size is associated with, and perhaps due to, a defect in the degradative properties of the lysosome, because Lamp1:GFP is normally degraded rapidly upon reaching the lysosome (Pulipparacharuvil et al., 2005).

The enlarged-lysosome phenotype of lrrk NS flies could be rescued by follicle-cell-specific expression of either Drosophila lrrk (Fig. 3H vs 3E; Fig. 3J) or human LRRK2 (Fig. 3J). Previously, we reported that the pathogenic Lrrk^GS (G2019S) mutation alters the function of Lrrk such that Lrrk^GS promotes perinuclear lysosome transport, whereas equivalent expression of Lrrk^WT (wild type) inhibits this process (Dodson et al., 2012). Interestingly, expression of lrrk^GS suppressed the enlarged-lysosome phenotype caused by lrrk NS as efficiently as did expression of lrrk^WT (Fig. 3E–J). The finding that the suppression of lysosome enlargement in lrrk NS flies was identical between lrrk^GS and lrrk^WT indicates that Lrrk^GS retains at least some of the functions of Lrrk^WT, at least with respect to the Lrrk-dependent processes controlling lysosome size. However, perinuclear lysosome clustering was still observed with expression of lrrk^GS (Fig. 3G,I), but not lrrk^WT (Fig. 3F,H), regardless of whether these were expressed in the wild-type or lrrk NS background. This is consistent with a model we previously proposed in which Lrrk^GS retains at least some of the normal functions of the wild-type protein, while also possessing neomorphic effects (Dodson et al., 2012).

On transmission electron microscopy (TEM) analysis, the expanded lysosomes in lrrk NS flies contained prominent large clear inclusions reminiscent of lipid (Fig. 4B), as well as undigested cytosolic contents, including intact mitochondria (Fig. 4C). These findings were confirmed with immunofluorescence analysis, in which the enlarged Lysotracker-positive lysosomes in lrrk NS flies also contained lipid, as labeled by the lipophilic dye BODIPY 493/503 (Fig. 4E), as well as mitochondria, as labeled by GFP fused to a mitochondrial targeting sequence (mitoGFP) (Fig. 4G). BODIPY 493/503 (Fig. 4D)- or mitoGFP (Fig. 4F)-positive signals were never seen within Lysotracker-positive lysosomes in stage-matched wild-type follicle cells. Thus, the aberrantly enlarged lysosomes in lrrk NS cells accumulate both lipid and intact mitochondria, suggesting a primary defect in the degradative properties of lysosomes in lrrk NS flies.

The lysosome is the final degradative compartment for multiple cellular pathways, including both the endocytic and autophagy pathways. The finding of intact organelles within enlarged lysosomes in lrrk NS flies suggests that at least some of these lysosomal contents are derived from the autophagy pathway. The hallmark of autophagy is the formation of double-membrane structures called autophagosomes, which label specifically with the marker Atg8a:GFP (Fig. 3A), the Drosophila homolog of mammalian LC3 (Geng and Klionsky, 2008). Autophagosomes form in the cytosol, engulf cytosolic components and are then transported to the lysosome for degradation (Feng et al., 2014). In lrrk NS follicle cells, autophagosomes were indeed apparent both by TEM analysis (Fig. 4H,I) and by a marked redistribution of Atg8a:GFP from a diffuse pattern in wild-type cells (Fig. 4J′), indicating a relative absence of autophagosomes, to predominantly punctate pattern in lrrk NS cells (Fig. 4K′). Moreover, autophagosomes were not observed in stage-matched wild-type follicle cells by TEM analysis (Fig. 4A). These puncta of Atg8a:GFP in lrrk NS flies often localized to Lysotracker-positive lysosomes, but were also frequently distinct from lysosomes (Fig. 4K″). Thus, lrrk NS cells show an increase in structures derived from autophagy; this increase could be due to either increased autophagy induction, decreased clearance of autophagosomes and autolysosomes, or both.

To evaluate this further, we studied the consequences of suppressing autophagy in lrrk NS flies. Atg7 is a component of the ubiquitin-like conjugation system required for autophagosome formation (Geng and Klionsky, 2008), and removing the function of atg7 in Drosophila has been shown to inhibit starvation-induced autophagy in the fat body (Juhász et al., 2007). Interestingly, we found that removing atg7 function in lrrk NS flies resulted in a significant increase in follicle cell death as assessed by TUNEL staining (Fig. 5A). This was not associated with any significant enhancement in the lysosome-enlargement phenotype observed in lrrk NS flies (Fig. 5B–E). These atg7/lrrk NS double mutants also displayed a mild rough-eye phenotype (Fig. 5I), which was not observed in either single mutant alone (Fig. 5G,H). The increased toxicity observed in atg7/lrrk double mutants versus lrrk NS flies alone is consistent with autophagy induction as a prosurvival response in lrrk NS flies. That this increased toxicity in the double mutants is not associated with enhancement of the lrrk NS lysosome-enlargement defect suggests that the lysosomal defects in lrrk NS flies are independent of the autophagy pathway.

Previously, we reported that Lrrk also physically binds to the early endosomal protein Rab5 (Dodson et al., 2012), which plays a vital role in the maturation of early endosomes to late endosomes and, thus, in endosomal substrate delivery to the lysosome (Huotari and Helenius, 2011). Genetic and physical interactions have also been reported between mammalian LRRK2 and Rab5 in cell-culture models under conditions of LRRK2 overexpression (Heo et al., 2010; Shin et al., 2008). Thus, we investigated whether lrrk also acts to regulate cargo delivery to the lysosome in the endocytic pathway. Interestingly, we found that the early endosomes, marked by either anti-Rab5 (Fig. 6D vs 6C) or anti-hepatocyte-growth-factor-regulated tyrosine kinase substrate (Hrs) antibodies (Fig. 6F vs 6E), were enlarged in lrrk NS follicle cells relative to wild type. In the endocytic pathway, monoubiquitylation serves as a sorting signal for protein cargoes that are bound for the lysosome to be degraded (Piper and Luzio, 2007; Raiborg and Stenmark, 2009). Interestingly, lrrk NS cells showed a marked accumulation of ubiquitylated proteins (as identified by an antibody that recognizes mono- or polyubiquitin conjugated to any target protein, but not free ubiquitin) (Fig. 6H vs 6G; Fig. 6B). Interestingly, this ubiquitin-positive staining often appeared as halos surrounding a non-stained central region (Fig. 6H, inset), suggesting the accumulation of ubiquitylated cargoes on vesicle membranes. Consistent with this hypothesis, the ubiquitylated-protein accumulations in lrrk NS cells colocalized with the expanded anti-Rab5-positive (Fig. 6D″) and anti-Hrs-positive (Fig. 6F″) early endosomes. These structures also accumulated the endocytic tracer dextran (Fig. 6H″), which is taken up from the extracellular space via endocytosis and subsequently trafficked through the endocytic pathway, thus confirming the identity of these structures as endosomes. Moreover, these ubiquitin accumulations consisted primarily of monoubiquitylated proteins, as evidenced by a lack of staining with an antibody that recognizes polyubiquitin chains (Fig. 6J′). This is distinct from what is observed with expression of the dominant-negative proteosome subunit DTS7 (serving as a control), which resulted in the accumulation of predominantly polyubiquitylated proteins recognized by both antibodies (Fig. 6K″). Importantly, the accumulation of monoubiquitylated proteins in lrrk NS flies could be rescued by follicle-cell-specific expression of wild-type lrrk (Fig. 6B). lrrk^e03680 homozygous flies showed a similar, albeit weaker, accumulation of ubiquitylated proteins (supplementary material Fig. S4C versus S4B), which also colocalized with internalized dextran (supplementary material Fig. S4C″) and Hrs (supplementary material Fig. S4F″). Collectively, these results show that lrrk NS flies accumulate aberrantly enlarged early endosomes that are laden with monoubiquitylated endosomal cargo proteins, suggesting a defect in substrate delivery to the lysosome in the endocytic pathway.

Previously, we found, from predominantly overexpression-based studies, that lrrk inhibits rab7-dependent perinuclear lysosome clustering (Dodson et al., 2012). We therefore asked whether the lysosomal-enlargement phenotype seen with lrrk loss-of-function is due to an interaction between lrrk and rab7. However, we found that lrrk loss-of-function does not phenocopy the lysosome clustering caused by expression of a constitutively active form of rab7 (rab7^CA) (supplementary material Fig. S5B), as would be expected according to this model. Moreover, rab7^CA had no effect on lysosome enlargement in lrrk NS flies (supplementary material Fig. S5E vs S5D,G). In addition, average lysosome size was no different in lrrk NS flies alone compared with lrrk NS flies expressing a dominant-negative version of rab7 (rab7^DN) (supplementary material Fig. S5G). These lrrk NS/rab7^DN flies did not show the massively enlarged lysosomes often seen in lrrk NS flies (supplementary material Fig. S5F vs S5D), but rather showed a more moderate but consistent lysosome-enlargement phenotype, the net result being no overall change in average lysosome size compared with lrrk NS (supplementary material Fig. S5G). This result is not unexpected given that dominant-negative rab7 results in lysosome dispersal and decreased lysosome fusion (Bucci et al., 2000). Taken together, these data suggest that the lysosome enlargement observed in lrrk NS flies is at least partially Rab7-independent.

Recently, it was reported that LRRK2 forms a complex with VPS35 (MacLeod et al., 2013), which is a core component of the retromer complex, involved in retrograde trafficking of cargoes from the endocytic pathway to the trans-Golgi network (Bonifacino and Rojas, 2006). This pathway is required for the retrograde transport of multiple substrates, including the mannose 6-phosphate receptor (MPR) (Arighi et al., 2004; Hierro et al., 2007; Seaman, 2004), a process that also requires the late endosomal GTPase Rab9 (Barbero et al., 2002; Lombardi et al., 1993; Riederer et al., 1994). Overexpression of LRRK2 G2019S (Lrrk^GS) has been reported to disrupt the normal trafficking of the MPR; this disruption can be rescued by overexpression of Vps35 (MacLeod et al., 2013). Inhibition of the retromer pathway increases lysosomal degradation of the MPR and thereby results in swelling of lysosomes due to intra-lysosomal accumulation of undigested material, presumably due to defects in trafficking of hydrolases to the lysosome (Arighi et al., 2004). Because this phenotype is highly reminiscent of what we have observed in lrrk NS flies, we hypothesized that augmenting retrograde transport might suppress lrrk NS phenotypes.

As we have previously reported, Drosophila Lrrk colocalizes with Rab9 at the late-endosome membrane (Dodson et al., 2012). Interestingly, we found that Lrrk-myc co-immunoprecipitated with Rab9-YFP, but did not show significant binding with either YFP alone or with Rab11-YFP (Fig. 7A), which served as controls. Expression of a constitutively active form of rab9 (rab9^CA) in follicle cells resulted in marked suppression of the lysosome-enlargement phenotype observed in lrrk NS cells (Fig. 7D vs 7C; Fig. 7E). Similarly, rab9^CA significantly, although incompletely, suppressed premature follicle cell death in lrrk NS cells (Fig. 7F). One mechanism by which lrrk could be required for rab9-dependent processes is through regulation of the membrane recruitment of Rab9; however, we did not detect any significant difference in the subcellular localization of Rab9 between the wild-type and lrrk-NS-mutant backgrounds (supplementary material Fig. S6). Collectively, these data suggest that augmenting Rab9 function can, at least in part, bypass the need for lrrk.

To further substantiate the finding that overactivation of rab9 can bypass the requirement for lrrk, we examined genetic interactions between lrrk and rab9 in another tissue that lrrk has been shown to act on: the larval NMJ. Previously, it has been reported that lrrk^e03680 mutants show a defect in NMJ morphology characterized by overgrowth of the synapse between the presynaptic motor neuron and postsynaptic muscle cell, with increased branching and number of boutons (Lee et al., 2010). We observed a similar, albeit stronger, phenotype in lrrk NS flies, where the synaptic overgrowth was mostly observed as small supernumerary satellite boutons surrounding the larger normal boutons (Fig. 7H vs 7G; Fig. 7K). This phenotype could be rescued by motor-neuron-specific expression of wild-type lrrk (Fig. 7I vs 7H; Fig. 7K). As observed in follicle cells, expression of rab9^CA significantly suppressed the NMJ morphology defect in lrrk NS flies (Fig. 7J vs 7H; Fig. 7K). Together, these results suggest that lrrk plays a role in retromer-mediated endosomal recycling, and that augmentation of late endosome-to-Golgi retrograde transport can bypass the requirement for lrrk.

Here, we provide evidence of a vital physiological role for a LRRK2 homolog in endolysosomal membrane transport and autophagy in vivo, by characterizing novel loss-of-function nonsense alleles in Drosophila lrrk. These lrrk NS alleles cause accumulation of markedly enlarged lysosomes full of undigested lipid and other cytosolic contents, as well as accumulation of autophagosomes and enlarged early endosomes laden with monoubiquitylated cargo proteins. The interactions that we describe between lrrk and rab9 suggest that the role of Lrrk in maintaining proper lysosome function involves the Rab9-dependent late-endosome-to-Golgi pathway.

Whereas the increase in perinuclear lysosome clustering caused by expression of lrrk^GS is rab7-dependent (Dodson et al., 2012), the effects of lrrk loss-of-function on lysosome size are rab7-independent, and rather seem to be mediated through the Rab9-retromer pathway. rab9 promotes retromer-dependent recycling of cargoes from late endosomes to the Golgi (Barbero et al., 2002; Lombardi et al., 1993; Riederer et al., 1994). Defects in retromer-dependent cargo trafficking lead to impaired lysosomal degradation and the accumulation of undigested materials within lysosomes (Arighi et al., 2004), a phenotype highly reminiscent of what is seen in lrrk NS flies. Rab9 and Lrrk colocalize at the late endosomal membrane (Dodson et al., 2012), and, as we show in this work, physically bind to one another. Augmentation of rab9 activity rescues lysosomal enlargement and premature follicle cell death in lrrk NS flies. Enhancing rab9 activity also rescues the NMJ morphology defect in lrrk NS flies, which have an NMJ phenotype identical to that reported with loss-of-function of the retromer component Vps35 (Korolchuk et al., 2007). Thus, in multiple tissues, augmentation of rab9 activity bypasses the need for lrrk, suggesting a role for lrrk in the Rab9-retromer pathway.

Defects in the autophagy pathway have previously been reported in the context of lrrk loss-of-function, but the results have been conflicting, including the finding of decreased autophagosome formation with LRRK2 knockdown in immune cells (Schapansky et al., 2014) and biphasic alterations in autophagy in the kidney of LRRK2 knockout mice, with markers of autophagy increased at 7 months of age and reduced at 20 months (Tong et al., 2012). Here, we report increased autophagic structures in lrrk NS flies, including both autophagosomes and autolysosomes. Although there is probably a component of decreased turnover of autophagic structures due to the lysosome dysfunction that occurs in lrrk NS flies, our data also suggest that autophagy induction is increased in lrrk NS flies, possibly as a prosurvival stress response to lysosome dysfunction or ubiquitylated-protein accumulation. This hypothesis is supported by the increased toxicity in lrrk NS flies when atg7 function is removed. Caspase-dependent apoptosis seems to be the end result when the autophagic response is insufficient, and occurs downstream of both lysosome dysfunction and autophagy induction. We cannot, however, exclude the possibility that lrrk loss-of-function increases autophagy activation via a more direct mechanism. Interestingly, rab9 has been implicated in autophagosome formation in mammalian cells in an alternative atg7-independent autophagy pathway that can be induced by certain cellular stressors (Nishida et al., 2009). That some lrrk NS phenotypes are enhanced by atg7 loss-of-function suggests that autophagy induction in the context of lrrk loss-of-function is atg7-dependent rather than atg7-independent. Still, it remains an intriguing possibility that lrrk could directly regulate autophagy induction, perhaps via its interaction with rab9.

Another important question is whether the defects seen in lrrk NS flies at both the early endosomal compartment and the late endosomal/lysosomal compartments reflect distinct roles for Lrrk at multiple endolysosomal transport steps. For example, via its physical binding to the early endosomal protein Rab5, which we and others have reported (Dodson et al., 2012; Shin et al., 2008), LRRK2/Lrrk might exert a distinct Rab9-independent role in the trafficking of early endosomes. Interestingly, loss-of-function of Drosophila lrrk has also been reported to impair synaptic vesicle endocytosis via an interaction with endophilin A, which is involved in the early membrane tubulation steps involved in vesicle budding from the plasma membrane (Matta et al., 2012). Thus, Lrrk might function as a general regulator of multiple distinct membrane transport steps.

Although our work has focused predominantly on follicle cells in the Drosophila ovary owing to the aforementioned benefits as a cell biological model system, we believe that our results have significance to the function of lrrk in neurons for several reasons. Lysosomal defects in lrrk NS flies can be rescued by human LRRK2, suggesting functional conservation of the endolysosomal functions of the two proteins. Moreover, both follicle cell lysosomal defects and abnormal NMJ morphology in lrrk NS flies are rescued by augmenting rab9 function, suggesting similar defects in rab9-dependent membrane transport processes due to lrrk loss-of-function in both neuronal and non-neuronal tissues. Why might follicle cells be particularly sensitive to the effects of lrrk loss-of-function? Ultrastructural analyses have reported an increase in the number of lysosomes in follicle cells from stage-12 egg chambers (Cummings and King, 1970), the stage at which lysosome defects in lrrk NS flies become apparent. Moreover, this stage-specific increase in lysosome abundance plays a role in the degradation of nurse cells, which are germline cells clonally related to the oocyte that undergo programmed cell death at the culmination of oogenesis (King, 1970; Spradling, 1993). Thus, we speculate that an increased requirement for lysosomal degradation in late-stage follicle cells could underlie the sensitivity of these cells to lrrk loss-of-function.

Our work demonstrates a vital physiological role for endogenous lrrk in endolysosomal membrane transport and lysosome function in vivo, and suggests that the toxicity of the pathogenic mutant lrrk^GS allele is related to dysregulation of its endolysosomal functions. These findings have important implications for the development of inhibitor-based therapies for PD that target LRRK2, because excessive inhibition might result in endolysosomal dysfunction and thus cellular toxicity. Regulation of lysosomal function and autophagy have emerged as key themes in the pathogenesis of neurodegenerative diseases, from inherited diseases of lysosomal storage to diseases such as PD, involving the accumulation of protein inclusions. Together with a striking amount of recent genetic data linking PD risk to genes involved in endolysosomal functions, including β-glucocerebrosidase (Aharon-Peretz et al., 2004; Gan-Or et al., 2008), ATP13A2 (Ramirez et al., 2006), VPS35 (Vilariño-Güell et al., 2011; Zimprich et al., 2011) and RAB7L1 (Gan-Or et al., 2012), our data establishes dysfunction of the endolysosomal pathway as a central pathogenic mechanism in PD.

The UAS-lrrk, UAS-lrrk^GS and CaSpeR-lrrk genomic rescue constructs were previously described (Dodson et al., 2012). For UAS-LRRK2, the human LRRK2 cDNA was introduced into the pTHW gateway vector (which contains an N-terminal HA tag as well as the UASt promoter) by recombination. All cloned PCR products were confirmed by DNA sequencing. The Rab-YFP constructs were obtained from Matthew Scott (Stanford University, Palo Alto, CA).

RNA was isolated from 1-day-old flies using Trizol, and purified using Phase Lock Gel Tubes to remove DNA. Two independent cDNA synthesis reactions were performed for each genotype using Clontech EcoDry Premix cDNA Kit (Double Primed Oligo-dT and Random Hexamers), after normalizing the amount of input RNA across genotypes. Quantitative PCR was performed in triplicate for each cDNA reaction and the six values for each genotype across the two cDNA reactions averaged. All primer pairs spanned an intron as depicted in Fig. 1A. Values were normalized to the geometric mean of three control genes, rpl32, eIF1a and aTub84B, for each genotype, and then normalized to wild type to calculate a fold change in gene expression.

To generate loss-of-function mutations in lrrk, we used TILLING (Till et al., 2003), which is a method for detecting point mutations in a gene of interest following chemical mutagenesis. EMS mutations were generated in a prior screen (Koundakjian et al., 2004), and were recovered using the Drosophila Tilling Service (Fred Hutchinson Cancer Research Center). lrrk¹, lrrk³ and lrrk⁴ are nonsense alleles resulting from substitution of a stop codon for L389, Q622 and W1221, respectively. lrrk², which was recovered from a mixed stock, has a single-nucleotide insertion changing L584 to F, and resulting in a frameshift that generates a stop codon 11 codons downstream. lrrk^e03680 results from the insertion of a PiggyBac transposable element in the intron between exons 5 and 6, and was obtained from the Harvard Drosophila Stock Center. The follicle-cell-specific driver CY2-GAL4 was a gift from Celeste Berg (University of Washington, Seattle, WA). atg7 mutants were obtained from Thomas Neufeld (Juhász et al., 2007). UAS-rab7^DN, -rab7^CA and -rab9^CA are as described (Zhang et al., 2007). UASp-diap1 flies were obtained from Bruce Hay (California Institute of Technology, Pasadena, CA) and Andreas Bergann (University of Massachusetts, Worchester, MA). All other stocks are from the Bloomington Drosophila Stock Center. For experiments involving transgenic flies, multiple lines were generated (Rainbow Transgenic Flies) and tested for each transgene. Drosophila strains were maintained in a 25°C humidified incubator unless otherwise noted.

For follicle cell experiments, freshly eclosed females were maintained on wet yeast paste for 24 hours prior to ovary dissection, and individual stage 11–13 egg chambers were hand dissected following fixation. The NMJ at abdominal segment 4, muscle 4 was used for all NMJ experiments. All tissues were fixed in either 3.7% formaldehyde or 4% paraformaldehyde in PBS. PBS + 0.1% Triton X-100 (or 0.2% Tween) + 2% BSA was used for blocking and antibody incubations. The following primary antibodies were used for immunocytochemistry: mouse anti-ubiquitin protein conjugates (Biomol, FK2, 1:300), mouse anti-polyubiquitin chains (Biomol, FK1, 1:25), rabbit anti-Rab5 (Abcam, 1:20), rabbit anti-Rab7 [a gift from Yashodhan Chinchore (Harvard University, Boston, MA) and Patrick Dolph (Dartmouth University, Hanover, NH), 1:100], mouse anti-tyrosine hydroxylase (Immunostar, 1:300), guinea pig anti-Hrs [a gift from Hugo Bellen (Baylor University, Houston, TX), 1:100], rabbit anti-cleaved-caspase-3 (Cell Signaling, 1:300), mouse anti-CSP (DHSB, 1:10) and fluorescein-conjugated anti-HRP (1:250). Secondary antibodies were Alexa Fluor 488 or 568 (Molecular Probes, 1:250). For Lysotracker staining, ovaries were dissected and incubated for 15 minutes in 100 nM Lysotracker red DND-99 (Molecular Probes), washed briefly, then mounted and imaged immediately, all in Schneider’s Drosophila Medium (Gibco). For lipid visualization, the Lysotracker staining solution was supplemented with 2 mg/ml BODIPY 493/503 (Molecular Probes). Images were obtained with a Zeiss LSM5 confocal microscope. All follicle cell images shown are from stage-12 or -13 egg chambers.

Freshly dissected stage-12 egg chambers were incubated in 0.5 mM Texas-Red–dextran (3000 MW, Invitrogen) for 15 minutes, washed briefly five times, then incubated for 30 minutes prior to fixation in 4% paraformaldehyde. All steps were performed at 25°C in Schneider’s Drosophila Medium (Gibco).

Ovaries were fixed in 3.7% formaldehyde for 20 minutes, washed in PBS, and stage 11–13 egg chambers hand dissected. Egg chambers were permeabilized for 30 minutes in 50 mM Tris-Cl + 188 mM NaCl + 0.1% Triton X-100 + 3% BSA, washed in PBS, then incubated for 1 hour at 37°C in 100 μl of TUNEL reaction mix according to the manufacturer’s instructions (Roche In Situ Cell Death Detection kit, TMR-red).

All samples were prepared for electron microscopy as described (Deng et al., 2008). 70- to 80-nm sections were stained with uranyl acetate and lead citrate and visualized using a JEOL 100CX transmission electron microscope (UCLA Brain Research Institute EM Facility). At least three individual flies of each genotype were examined for TEM studies.

Freshly sacrificed flies were mounted on their side, with one eye upward, on white tape using clear nail polish. All flies were placed on a rotating platform to permit for orientation under vacuum and were imaged as described (Gross et al., 2008).

Single 0- to 3-day-old females were placed in a vial supplemented with dry yeast along with three sibling males and maintained at 25°C. After 4 days, the flies were removed and progeny were allowed to develop at 25°C for an additional 14 days, at which time the number of adult progeny per vial was quantified.

S2 cells were cultured in Schneider’s Drosophila Medium (Gibco) + 10% fetal bovine serum (Invitrogen) + 1% penicillin/streptomycin (Invitrogen). For immunoprecipitations, cells were plated on 10-cm dishes at a density of 3.6×10⁶. Transfections were performed 24 hours later using the Qiagen Effectene kit according to the manufacturer’s recommendations. pMT-lrrk-9myc was transfected along with pAC-Gal4 and pUASp-Rab:YFP constructs (obtained from Matthew Scott) as indicated. pMT-lrrk-9myc expression was induced by adding 0.5 mM copper sulfate 24 hours after transfection, and cells were harvested 24 hours later.

Cells were lysed in 800 ml of RIPA buffer (Upstate) containing protease inhibitor cocktail (Roche), and immunoprecipitations performed using Dynabeads from Invitrogen according to the manufacturer’s instructions. Rabbit anti-GFP (Invitrogen) or mouse anti-Myc antibodies were used for immunoprecipitation. Bound proteins were eluted at 72°C for 10 min in 1× SDS sample buffer containing 5% 2-mercaptoethanol, and were resolved on a 6% or 10% polyacrylamide gel. Proteins were transferred to an Immobilon membrane (Millipore), and labeled with mouse antibodies against Myc (Millipore, 1:10,000), GFP (Roche, 1:5000), or α-tubulin (Sigma, 1:15,000) and goat anti-mouse HRP. Blots were developed using SuperSignal West Pico Chemiluminescent Substrate (Pierce).

Ubiquitin-positive endosome density was calculated by manually counting the number of ubiquitin-positive puncta (labeled with FK2) per 63× field per egg chamber. Lysosome size was calculated from Lysotracker-stained samples using ImageJ software. All quantifications were performed on randomly selected egg chambers. Error bars represent standard error of the mean (s.e.m.). The two-tailed Student’s t-test was used to determine statistical significance.

Supplementary material available online at http://dmm.biologists.org/lookup/suppl/doi:10.1242/dmm.017020/-/DC1

We are very grateful to P. Dolph, C. Berg, M. Scott, H. Bellen, H. Kramer, B. Hay and A. Bergmann for fly strains, antibodies and DNA constructs; to C. C. Ho and W. Dauer for generating the human LRRK2 construct; to F. Laski for use of his microtome; and to E. Dell’Angelica and B. A. Hay for discussions.