LRRK2 functions in synaptic vesicle endocytosis through a kinase-dependent mechanism

Leucine-rich repeat kinase 2 (LRRK2) is a large multidomain protein harboring a kinase, GTPase and several protein-binding domains (Cookson and Bandmann, 2010). Mutations in this protein are linked to familial as well as sporadic forms of Parkinson's disease (Paisán-Ruíz et al., 2004, Zimprich et al., 2004). For example, the G2019S mutation, which causes elevated kinase activity, is responsible for 4% of familial Parkinson's disease worldwide (Healy et al., 2008). Interestingly, this mutation was also found in some apparently sporadic cases. Several other mutations (e.g. R1441C/H/G and I2020T) in the GTPase and kinase domains have also been shown to segregate with Parkinson's disease (Zimprich et al., 2004; Healy et al., 2008).

Despite the apparent clinical association between LRRK2 mutations and Parkinson's disease, it remains enigmatic how such mutations lead to disease progression and the biological function of LRRK2 itself also still needs to be clarified further. Although several proteins have been identified as kinase substrates of LRRK2, none of these substrates have been independently reported in vivo. Substrates are involved in different processes, such as neuronal survival, cytoskeletal rearrangement, autophagy and apoptosis (Berwick and Harvey, 2011; Gillardon, 2009; Plowey and Chu, 2011; Martin et al., 2014). LRRK2 has also been shown to regulate neurotransmission, by interacting with synaptic proteins that modulate clathrin-mediated endocytosis of synaptic vesicles, such as Rab5b, snapin and N-ethylmaleimide-sensitive factor (NSF) (Piccoli et al., 2011; Shin et al., 2008; Yun et al., 2013; Piccoli et al., 2014). Moreover, we revealed an essential role for LRRK in synaptic vesicle endocytosis at Drosophila melanogaster neuromuscular junctions through phosphorylation of endophilin A (EndoA) (Matta et al., 2012). EndoA is an evolutionary conserved protein involved in clathrin-mediated endocytosis (Ringstad et al., 1997). It drives the formation of synaptic vesicles by sensing or inducing plasma membrane curvature (Gallop et al., 2006; Masuda et al., 2006) and facilitates clathrin uncoating once synaptic vesicles are generated (Milosevic et al., 2011; Verstreken et al., 2002). Consequently, impairment of EndoA function results in severe deficits in synaptic vesicle endocytosis (Gad et al., 2000; Milosevic et al., 2011; Schuske et al., 2003; Verstreken et al., 2002). Our study identified EndoA as a substrate of LRRK and demonstrated that LRRK has an essential role in synaptic vesicle endocytosis in Drosophila by phosphorylating EndoA at S75 (Matta et al., 2012).

Given that flies express only one LRRK protein, which might encompass the function of both mammalian LRRK1 and LRRK2, we specifically assessed the role of LRRK2 in synaptic vesicle endocytosis in mammals. We demonstrate that LRRK2 knockout leads to impairments in clathrin-mediated synaptic vesicle endocytosis and neurotransmission and the presence of endocytic intermediates and abnormal vesicle morphology and number, similar to Lrrk knockout in Drosophila. Furthermore, we show that mammalian EndoA1 (SH3GL2), a neuron-specific variant of EndoA (Giachino et al., 1997), is a kinase substrate of LRRK2, confirming the conservation between Drosophila and the mammalian system. Finally, we show that LRRK2 phosphorylates EndoA1 not only on residue S75, identical to the residue phosphorylated in Drosophila, but on T73 as well. Taken together, our findings advance the understanding of the evolutionary conserved biological function of LRRK2 and its pivotal role in clathrin-mediated endocytosis.

Our group recently reported that knockout of LRRK, the Drosophila melanogaster orthologue of LRRK1 and LRRK2, causes an impairment in clathrin-mediated endocytosis of synaptic vesicles, indicating that there is a role for this protein in synaptic vesicle formation (Matta et al., 2012). To determine whether this role in endocytosis is contained within the LRRK2 protein, we performed dynamic assays of endocytosis in wild-type and LRRK2-knockout primary striatal neurons by using sypHy, a fusion construct of synaptophysin and super ecliptic pHluorin. At the onset of stimulus (20 Hz for 15 s), exocytosis caused a rapid increase in sypHy fluorescence (Fig. 1Aii) which, after cessation of stimulus, slowly returned to baseline (Fig. 1Aiii). As shown in Fig. 1B, the decay of fluorescence was delayed in LRRK2-knockout cells when compared to wild-type controls. The fluorescence decay, which reflects endocytosis, was assessed by fitting the trace with a single exponential time constant (τ_decay). Quantification shows that τ_decay is significantly increased in LRRK2-knockout neurons (wild-type, 68.96±5.07; LRRK2 knockout, 89.29±3.92; mean±s.e.m., P<0.0001, Fig. 1C), indicating that endocytosis of synaptic vesicles is impaired in these cells. However, no significant differences in exocytosis were observed under these conditions (wild-type, 10.00±0.69; LRRK2 knockout, 9.30±0.63; Fig. 1D). These results demonstrate that the role of LRRK2 in synaptic vesicle endocytosis, a mechanism previously identified in Drosophila Lrrk mutants, is conserved in mammalian neurons.

To gain insight into whether the endocytic delay observed in mammalian LRRK2-knockout neurons could also be observed at a morphological level, we performed transmission electron microscopy (TEM) of wild-type (Fig. 2A) and LRRK2-knockout (Fig. 2B–E) striatal brain sections. Ultrastructural analysis of knockout synapses revealed a phenotype different from control littermates: a decreased number of synaptic vesicles (Fig. 2B), which were in general more heterogeneous in size with the presence of vesicles with a larger diameter (Fig. 2C, asterisks). Abnormal endocytic intermediates (Fig. 2D, asterisks), including clathrin-coated endocytic intermediates (Fig. 2E), were also evident in some knockout nerve terminals. Quantification of TEM micrographs showed that the mean number of synaptic vesicles per µm² synapse was substantially lower in knockout than in control synapses (Fig. 2F, wild-type, 194.8±5.1; knockout, 146.0±4.6; mean±s.e.m.; P<0.0001). In addition, the average synaptic vesicle diameter was increased in LRRK2 knockouts compared to controls (Fig. 2G, wild-type, 47.59±0.47 nm; knockout 50.15±0.48 nm, P<0.001) and vesicles with a larger diameter were more abundant in knockout than in control synapses (Fig. 2H). These observations are reminiscent of the morphological changes we previously described in Drosophila, where synaptic vesicle size was increased in LRRK-knockout boutons (Matta et al., 2012). Interestingly, a reduction in the number of synaptic vesicles and accumulation of clathrin-coated vesicular profiles were also observed in endophilin triple-knockout mouse synapses (Milosevic et al., 2011) and in fly endoA-knockout boutons (Guichet et al., 2002; Verstreken et al., 2002), suggesting that the morphological defects observed in LRRK2 knockouts and mouse endophilin triple-knockouts might be due to overlapping mechanisms.

The previous findings prompted us to hypothesize that LRRK2 might have an important role in synaptic function. To determine whether knockout of LRRK2 affects neurotransmission, we measured the number of spontaneous excitatory postsynaptic currents (sEPSCs) in wild-type (n = 19) and LRRK2-knockout (n = 22) hippocampal neurons. All neurons in culture appeared healthy and, under basal conditions, the number of sEPSCs was similar in wild-type and LRRK2-knockout cells (Fig. 3A). To measure sEPSCs coming from newly formed synaptic vesicles, sucrose (50 mM) was added to the bath solution as it stimulates the release of the readily releasable pool through Ca²⁺-independent mechanical stress (Rosenmund and Stevens, 1996). In sucrose-treated wild-type cells, the number of sEPSCs was increased compared to non-treated wild-type cells (control, 47.74±14.03; sucrose, 100.84±9.76; mean±s.e.m.). In knockout cells, however, sucrose treatment failed to increase the number of sEPSCs (control, 32.05±8.07; sucrose, 43.14±9.76; Fig. 3B). The effect of sucrose on the number of sEPSCs was calculated by determination of the sucrose versus control ratios (wild-type, 4.07±1.38; knockout, 1.69±0.37; Fig. 3C). These data indicate that LRRK2 modulates synaptic function, as knockout of LRRK2 causes impairment in neurotransmission following the release of the readily releasable pool.

To investigate this further, we verified whether the diminished response to sucrose observed in LRRK2-knockout neurons is due to impairments at the pre- or post-synaptic level. When analyzing the frequency of sEPSCs, which is well known to be dependent on the presynaptic machinery (Marinelli et al., 2003), we observed differences between wild-type and LRRK2-knockout neurons. Under basal conditions, the frequency was similar in wild-type and LRRK2-knockout neurons, whereas the presence of sucrose stimulated neurotransmitter release and significantly increased the sEPSC frequency in wild-type (control, 26.32±4.27; sucrose, 46.64±6.48) but not in LRRK2-knockout neurons (control, 35.44±4.01; sucrose, 30.00±4.67, Fig. 3D). Analysis of sucrose versus control ratios (wild-type, 2.08±0.33; knockout, 0.97±0.18; Fig. 3E) confirmed the increase in frequency in wild-type neurons, but not in LRRK2-knockout neurons. This indicates that LRRK2 has a role in the presynaptic mechanisms that replenish the synaptic vesicle pool to govern high frequency neurotransmission, consistent with a role in endocytosis and vesicle recycling. To evaluate whether LRRK2 also affects the postsynaptic mechanisms, we measured the sEPSC peak amplitude, of which changes are thought to reflect changes in the response of the postsynaptic receptors (Sebe et al., 2003; Thompson et al., 1993). As shown in Fig. 3F, there were no differences in peak amplitude between control and sucrose condition in wild-type (control, −25.90±2.84; sucrose, −25.27±2.47) or LRRK2-knockout cells (control, −19.44±1.97; sucrose, −18.29±1.37). Calculation of the sucrose versus control ratio (Fig. 3G) confirmed that the peak amplitude remained unchanged (wild-type, 1.09±0.17; knockout, 0.99±0.06).

Taken together, these data indicate that LRRK2 modulates neurotransmission at the presynaptic level. We believe these effects are specific to loss of LRRK2 as we observed similar defects in neurotransmission when wild-type neurons were treated with LRRK2-IN-1, a well-defined LRRK2 kinase inhibitor (Deng et al., 2011). Treating wild-type cells with different concentrations of LRRK2-IN-1 attenuated the effect of sucrose on the number and frequency of sEPSCs, when compared to DMSO-treated cells (supplementary material Fig. S1). Again, no effect on peak amplitude was observed (supplementary material Fig. S1). These data indicate that the presynaptic function of LRRK2 is dependent on its kinase activity.

Our data suggest that LRRK2 has a presynaptic role in neurotransmission by affecting synaptic vesicle endocytosis. To strengthen the previous results, we compared the effect of LRRK2 knockout to the effect of independently blocking endocytosis. To block endocytosis we treated wild-type (control, n = 15; dynasore, n = 14) and LRRK2-knockout (control, n = 12; dynasore, n = 14) hippocampal neurons with dynasore (40 µM for 5 min, Fig. 4A), a cell-permeable small molecule that is a noncompetitive inhibitor of dynamin 1 (Douthitt et al., 2011; Macia et al., 2006).

Similar to previous data, the presence of sucrose caused an increase in the number of sEPSCs in vehicle pretreated wild-type (control, 24.80±6.35; sucrose, 62.73±17.21; mean±s.e.m.), but not in LRRK2-knockout neurons (control, 45.09±14.00; sucrose, 51.09±12.89, Fig. 4B). Pretreatment with dynasore abolished the effect of sucrose in wild-type neurons (control, 119.57±31.76; sucrose, 147.36±34.02), but did not cause differences in LRRK2-knockout neurons (control, 97.56±15.00; sucrose, 139.75±30.78, Fig. 4B). Calculation of the sucrose versus control ratios (wild-type control, 3.77±0.66; wild-type dynasore-treated, 1.72±0.25; knockout control, 1.55±0.23; knockout dynasore-treated, 1.82±0.40; Fig. 4C) confirmed these results. To ensure that we were looking at presynaptic changes, the frequency and peak amplitude of sEPSCs were analyzed. In control conditions, the frequency of sEPSCs was increased in the presence of sucrose in wild-type (control, 18.98±4.05; sucrose, 35.37±6.43) but not in LRRK2-knockout neurons (control, 29.66±4.74; sucrose, 33.10±5.70, Fig. 4D). Such increase was not present in dynasore-treated cells of either genotype (wild-type control, 48.30±6.52; wild-type with sucrose, 54.17±5.61; knockout control, 40.60±6.38; knockout with sucrose, 46.15±7.66, Fig. 4D). Analysis of the sucrose versus control frequency ratios further strengthened these observations (wild-type control, 2.63±0.47; wild-type dynasore-treated, 1.33±0.18; knockout control, 1.11±0.17; knockout dynasore-treated, 1.18±0.10, Fig. 4E). Peak amplitude was not altered in the presence of sucrose (wild-type control, −22.69±2.76; wild-type with sucrose, −26.18±3.68; knockout control, −20.56±3.24; knockout with sucrose −22.27±4.50, Fig. 4F) and treatment with dynasore did not affect this parameter (wild-type control, −19.13±2.63; wild-type with sucrose, −20.48±3.19; knockout control, −23.62±2.55; knockout with sucrose, −22.42±2.50; Fig. 4F). Analysis of the sucrose versus control ratios confirmed this observation (wild-type control, 1.14±0.05; wild-type dynasore-treated, 1.05±0.03; knockout control, 1.15±0.25; knockout dynasore-treated, 0.96±0.05, Fig. 4G).

We revealed a clear phenotype in LRRK2-knockout neurons, with impairments in clathrin-mediated synaptic vesicle endocytosis and neurotransmission, and the presence of endocytic intermediates and abnormal vesicle morphology and number. Recently, we reported a similar phenotype in Drosophila as the result of changes in LRRK-dependent phosphorylation of EndoA and we specifically identified the serine residue located at position 75 as a LRRK-dependent phosphorylation site (Matta et al., 2012). EndoA is an evolutionary conserved membrane-interacting protein that is widely known to play a role in endocytosis (Milosevic et al., 2011; Verstreken et al., 2002). When comparing the sequences of EndoA in Homo sapiens, Mus musculus, Rattus norvegicus and Drosophila melanogaster, we noticed that the S75 residue is highly conserved across species (Fig. 5C). Hence, we questioned whether LRRK2 can also phosphorylate EndoA in mammals. To determine whether human LRRK2 phosphorylates the mammalian EndoA1, the neuron-specific EndoA isoform (Giachino et al., 1997; Milosevic et al., 2011), we performed an in vitro [³³P]ATP phosphorylation assay with the following recombinant proteins: wild-type EndoA1, EndoB1, which lacks the S75 residue and EndoA1 S75A, in which the serine residue is mutated to an alanine residue. As shown in Fig. 5A, there is a time-dependent increase in phosphorylation of EndoA1 and, to a lesser extent, in the EndoA1 S75A mutant. Even less ³³P incorporation was observed in EndoB1. Consistently, quantification of the phosphorylated versus total substrate ratios indicated that phosphorylation was less efficient when the S75 residue was mutated into an alanine residue or was not present (Fig. 5B). An in vitro [³³P]ATP phosphorylation assay with recombinant wild-type, kinase hyperactive G2019S or kinase deficient D1994A LRRK2 as kinases and recombinant EndoA1 as a substrate shows that EndoA1 phosphorylation was enhanced using G2019S LRRK2, but almost completely abolished with the D1994A LRRK2 variant (supplementary material Fig. S2A,B). To address the specificity of LRRK2-dependent EndoA1 phosphorylation we compared LRRK2 and GSK3β as EndoA1 kinases in vitro. In this phosphorylation assay GSK3β, unlike LRRK2, was not able to phosphorylate EndoA1, although its autophosphorylation was more prominent in comparison to LRRK2 (supplementary material Fig. S2C,D). These observations provide further evidence that EndoA1 is not a universal in vitro substrate for the majority of kinases, supporting the specificity of the LRRK2-driven phenomenon.

The previous analysis indicated that LRRK2 phosphorylates EndoA1 at S75, but suggested this is not the only LRRK2-dependent phosphorylation site in EndoA1. Thus, we hypothesized that T73, which is also conserved in the mammalian protein, but not present in Drosophila EndoA nor mammalian EndoB1 (Fig. 5C) might be another LRRK2-dependent phosphorylation site, as it is positioned in the same functional domain as S75 (Fig. 6A,B). In order to test whether EndoA1 has a dual LRRK2-dependent phosphorylation site at T73 and S75, we performed an additional in vitro [³³P]ATP phosphorylation assay with wild-type EndoA1 and the T73A, S75A and T73A/S75A mutant proteins as substrates. Data in Fig. 6C confirm our hypothesis and show that LRRK2-mediated phosphorylation of EndoA1 was diminished to the same extent by mutating the T73 or the S75 residue (Fig. 6D). As expected, phosphorylation of the double mutant EndoA1 was also significantly decreased in comparison to wild-type EndoA1. However, phosphorylation of double mutant EndoA1 was not significantly different from phosphorylation of T73A or S75A EndoA1. We speculate that this is because mutating either site changes the conformation of the protein, meaning LRRK2 is unable to phosphorylate the other residue. Although our data indicate that the T73/S75 site is crucial for LRRK2-mediated phosphorylation of EndoA1, we cannot rule out other phosphorylation events in vitro.

In the present study, we reveal a role for LRRK2 at the mammalian synapse and provide evidence that the function of LRRK in the regulation of endocytosis in flies overlaps with the role of mammalian LRRK2 in this process. We show that knockout of LRRK2 impairs clathrin-mediated synaptic vesicle endocytosis and neurotransmission. In line with this, previous work by our group provided evidence for the involvement of LRRK in synaptic vesicle endocytosis in Drosophila melanogaster. We identified Drosophila EndoA as a substrate of LRRK2 and through rescue experiments using Drosophila EndoA mutants, we demonstrated that LRRK-dependent EndoA phosphorylation regulates EndoA membrane affinity and consequently modulates synaptic endocytosis (Matta et al., 2012). Here, we show that, like the Drosophila variant, mammalian EndoA1 is also phosphorylated by LRRK2, suggesting that the signaling cascade regulating synapse function translates from fly to mammalian neurons.

We provide evidence for a role for LRRK2 in endocytosis in mammalian cells. Our data indicate that knockout of LRRK2 causes a significant impairment in clathrin-mediated endocytosis in striatal neurons, in agreement with our observations in Drosophila LRRK-knockout neurons and previous studies in hippocampal neurons treated with short hairpin RNA (shRNA) (Matta et al., 2012; Shin et al., 2008). Impaired rates of endocytosis have also been observed in experiments with regulators of endocytosis such as amphiphysin, endophilin, SPIN90 and dynamin (Di Paolo et al., 2002; Kim et al., 2005; Milosevic et al., 2011; Newton et al., 2006; Ferguson et al., 2007). Clathrin-mediated endocytosis is a dynamic process in which numerous proteins act as effectors and/or regulators. The function of many of these proteins depends on interactions with other proteins and thus, the removal of a specific protein can impair the cascade of protein–protein interactions and consequently affect endocytosis. Therefore, it is plausible that the lack of LRRK2 alters clathrin-mediated endocytosis by impairing the function of one or more endocytic proteins.

Recently, it has been described that LRRK2 could also modulate synaptic vesicle exocytosis, either by regulating presynaptic vesicle release or by interacting with snapin (Piccoli et al., 2011; Yun et al., 2013). Yun and colleagues claimed that LRRK2-dependent phosphorylation of snapin decreased the extent of exocytotic release in hippocampal neurons as measured by using vGlut-phluorin assays. However, an earlier study from the same group failed to show an effect of LRRK2 knockdown on synaptic vesicle exocytosis (Shin et al., 2008). Our data in striatal neurons are in line with this study (Shin et al., 2008), as we do not observe any effect on synaptic vesicle exocytosis in LRRK2-knockout neurons. The use of different neuronal populations (hippocampal vs striatal) and different systems to measure exocytosis (vGlut-phluorin versus sypHy) could possibly explain these divergences.

Our findings that LRRK2 affects endocytosis prompted us to hypothesize that LRRK2 consequently might have an important role in neurotransmission. Indeed, we observe that knockout of LRRK2 leads to impaired presynaptic function upon sucrose stimulation. As all sEPSCs measured after sucrose perfusion are coming from newly formed vesicles, it is tempting to speculate that this impairment in neurotransmission is driven by defects in synaptic vesicle endocytosis. Hence, we propose that slowed synaptic vesicle endocytosis in LRRK2-knockout neurons might be the cause of impaired neurotransmission, especially during intense neuronal activity, where the vesicle replenishment is crucial for effective neurotransmitter release. In fact, we found similar effects in synaptic transmission when cells are treated with dynasore, a compound that has been demonstrated to inhibit endocytosis through dynamin-dependent and -independent mechanisms (Macia et al., 2006; Park et al., 2013). Hence, as results in LRRK2-knockout neurons are similar to results seen in dynasore-treated neurons, it is likely that these results are also due to the effect of LRRK2 knockout on endocytosis. Interestingly, treatment with dynasore increased the number of sEPSCs detected in both genotypes in the presence and absence of sucrose. This could be explained by an increase in transmitter release, as dynasore has been described to elevate resting intra-terminal Ca²⁺ concentrations, causing an increase in the probability of transmitter release (Chung et al., 2010; Douthitt et al., 2011).

In addition, pharmacological inhibition of LRRK2 kinase activity with LRRK2-IN-1 causes similar impairments in synaptic transmission, indicating that such impairments are, at least partially, dependent on the LRRK2 kinase domain. However, one needs to keep in mind possible off-target effects of LRRK2-IN-1, as it has been previously shown that in LRRK2-knockout neurons the effect of LRRK2-IN-1 mimics the effect of LRRK2 ablation (Luerman et al., 2014). Further studies in LRRK2-knockout neurons would be needed to shed light on possible off-target effects of LRRK2-IN-1 on synaptic vesicle endocytosis. Interestingly, synaptic transmission is also impaired in mouse triple-knockout neurons of endophilin, an important regulator of synaptic vesicle endocytosis (Milosevic et al., 2011). In line with our data, an in vivo study in LRRK2 mutant animals has suggested an involvement of LRRK2 in neurotransmitter release (Tong et al., 2009). Our data further suggest that this function of LRRK in flies is mediated by LRRK2 and not LRRK1 in the mammalian system, as LRRK2-IN-1 does not inhibit LRRK1 kinase function (Kramer et al., 2012). Future studies in LRRK1-knockout neurons could strengthen this hypothesis further.

Defects in clathrin-mediated endocytosis can also be observed at the ultrastructural level. LRRK2-knockout synapses show a decreased number of synaptic vesicles that are heterogeneous in size and present various aberrant endocytic intermediates, including clathrin-coated intermediates. Similar phenotypes are observed after inhibition of dynamin function using dynasore (Newton et al., 2006) and in Drosophila mutants of endocytic proteins such as EndoA, AP-180/lap, StonedB and Intersectin/Dap160 (Dickman et al., 2005; Koh et al., 2004; Verstreken et al., 2002; Verstreken et al., 2003; Zhang et al., 1998; Ferguson et al., 2007; Guichet et al., 2002). These observations are also similar to the morphological changes described in Lrrk Drosophila mutants, where synaptic vesicle diameter was increased in knockout boutons (Matta et al., 2012). Furthermore, a reduction in the number of synaptic vesicles and an accumulation of clathrin-coated vesicular profiles are also observed in endophilin triple-knockout mice (Milosevic et al., 2011). As the ultrastructural defects observed in these mice are very similar in LRRK2 and endophilin-knockout synapses, one could speculate that LRRK2 regulates endophilin function during endocytosis in mammals. Interestingly, a decreased number of synaptic vesicles was also reported in a knockout model of α-synuclein, another protein involved in Parkinson's disease (Cabin et al., 2002) that is reported to have a role in clathrin-mediated endocytosis (Ben Gedalya et al., 2009). In addition, endophilin has also been reported to bind with high affinity to Parkin, another protein linked to Parkinson's disease (Trempe et al., 2009). Therefore, one could speculate that LRRK2, α-synuclein and Parkin function in a common pathway at the synapse and disturbances of this pathway have a role in Parkinson's disease. Interestingly, defects in synaptic transmission are frequently and consistently observed in both dominant and recessive forms of Parkinson's disease (Kitada et al., 2009; Nakamura and Edwards, 2007; Tong et al., 2009), although the molecular mechanisms remain unknown. Further studies are now needed to determine whether this intriguing common synaptic function of LRRK2 and other Parkinson's disease risk factors contributes to Parkinson's disease pathology associated with these genes.

In summary, our data reveal a clear phenotype in LRRK2-knockout neurons, with slowed clathrin-mediated endocytosis, impaired neurotransmission and ultrastructural abnormalities. As mentioned before, similar phenotypes are observed in both Drosophila EndoA and Drosophila Lrrk mutants and in endophilin-knockout mice (Matta et al., 2012; Milosevic et al., 2011; Verstreken et al., 2002). As such phenotypes in Lrrk Drosophila mutants are the result of changes in LRRK-dependent phosphorylation of EndoA (Matta et al., 2012), we aimed at studying whether LRRK2 also phosphorylates mammalian EndoA1. In this study, we provide in vitro evidence that EndoA1 is phosphorylated by LRRK2 on at least two sites, T73 and S75. It is possible that additional serine and/or threonine residues in EndoA1 are phosphorylated by LRRK2, because the T73A/S75A mutant protein is still phosphorylated by LRRK2 in vitro. The observation that GSK3β did not phosphorylate EndoA1 suggests that EndoA1 is only phosphorylated by specific kinases, including LRRK2. The fact that LRRK2 does not phosphorylate EndoB1, indicates that LRRK2 has a discrete number of selective substrates.

Both T73 and S75 are located in the EndoA1 BAR domain, which is known for its involvement in membrane curvature (Gallop et al., 2006) and is highly conserved across species. Interestingly, recent evidence indicates that the phosphorylation of S75 constitutes a regulatory mechanism for guiding the curvature state of EndoA1 and could be used to control different types of membrane curvature in vivo (Ambroso et al., 2014). This phenomenon was also observed in our previous study (Matta et al., 2012). Hence, we suggest that LRRK2-dependent phosphorylation of this site regulates the membrane affinity of EndoA1 and consequently its role in clathrin-mediated endocytosis. However, it will be essential to provide evidence that LRRK2 also phosphorylates EndoA1 in vivo. Further research will need to be conducted to resolve this issue. Whether phosphorylation of T73 has a similar function to that of S75 phosphorylation remains to be determined. Compared to autophosphorylation of LRRK2, phosphorylation of EndoA1 appears to be low. However, LC-MS/MS data show that whereas in EndoA1 only T73 and S75 were detected as phosphosites, in LRRK2 more than 20 phosphosites were detected (supplementary material Table S1), explaining the difference in ³³P incorporation. Furthermore, it needs to be kept in mind that the amount of phosphorylation is not directly correlated with functionality in vivo. Besides driving the formation of synaptic vesicles through sensing or inducing plasma membrane curvature (Gallop et al., 2006; Masuda et al., 2006), EndoA1 has another important function in synaptic vesicle endocytosis, that is, facilitating clathrin uncoating once synaptic vesicles have been generated (Milosevic et al., 2011; Verstreken et al., 2002). Morphological phenotypes found in LRRK2-knockout synapses suggest that both EndoA1 functions might be affected. Nevertheless, further work is needed to determine precisely how EndoA1 is affected by LRRK2 in vivo. For example, similar to what was performed before in Drosophila (Matta et al., 2012), it would be interesting to use phosphomimetic or phosphodeficient EndoA1 mutants to investigate whether expression of these mutants would rescue or worsen the endocytosis deficiency phenotype that is observed in LRRK2-knockout neurons. This would further strengthen our hypothesis that LRRK2-dependent EndoA1 phosphorylation is functionally important in synaptic vesicle endocytosis.

In this study, we look at effects of LRRK2-knockout and show that this causes impairments in neurotransmission. However, G2019S, one of the most prevalent LRRK2 mutations in Parkinson's disease, is known to elevate kinase activity (Greggio and Cookson, 2009). Interestingly, transfection studies in human cells have shown that kinase-activating as well as kinase-dead mutations have similar toxic effects (Cookson and Bandmann, 2010). Similarly, Shin and colleagues report that overexpression of LRRK2 slows down synaptic vesicle endocytosis (Shin et al., 2008). This is in accordance with previous work from our group, where we showed that in Drosophila both too much and too little LRRK2 cause impairments in synaptic vesicle endocytosis (Matta et al., 2012). Therefore, it is tempting to speculate that LRRK2 is part of a carefully balanced system, in which both too much and too little protein activity can cause defects in the synaptic vesicle cycle. More research is needed to validate this speculation and it would be interesting to investigate whether our experiments yield similar results in neurons derived from G2019S LRRK2 knock-in models.

In conclusion, we unravel an essential role of LRRK2, and in particular its kinase domain, in the regulation of clathrin-mediated endocytosis of synaptic vesicles and subsequent neurotransmission in mammals. We show that a mechanism that was previously identified in Drosophila melanogaster is highly conserved across species and relevant in mammalian systems.

LRRK2-knockout C57Bl6/J mice (Charles River, France) and LRRK2-knockout Long Evans rats (SAGE Laboratories, USA) were housed under specific pathogen-free conditions and were used in accordance with the University of Leuven and Janssen Pharmaceutica Animal Ethics Committee. For the endocytosis assay and electrophysiology experiments, wild-type and homozygous LRRK2-knockout animals were used. Pregnant females were killed by decapitation and embryos were isolated. For electron microscopy analysis, LRRK2-knockout animals and wild-type littermates were used. Animals were killed with CO₂ when needed.

Striatal cultures were generated from E17-19 embryos from wild-type and LRRK2-knockout Long Evans rats (SAGE Laboratories, USA). At DIV11 primary neurons were co-transfected with sypHy, a fusion construct of synaptophysin and super ecliptic pHluorin (Granseth et al., 2006) and pRFP-c-RS (Origine) using Lipofectamin-2000 (Invitrogen, Gent, Belgium). SypHy is a pH-sensitive fluorescent reporter that, by analogy with the original synaptopHluorin (synaptobrevin-pHluorin, (Miesenböck et al., 1998; Sankaranarayanan and Ryan 2000) is quenched in the acidic intracellular space of the synaptic vesicles and will only become fluorescent upon exposure to the more basic pH of the extravesicular space. In these experiments, only RFP-positive boutons were assayed for endocytosis in a custom-build stimulation chamber on the stage of a Zeiss Axiovert 200M equipped with a mono-chromator (Poly V) and a cooled CCD camera (PCO, Imago QE), both from TILL Photonics (Gräfelfing, Germany). The assay was carried out as described previously (Kim et al., 2005; Sankaranarayanan and Ryan 2000). Briefly, cells were submerged in 500 µl of HBS buffer (10 mM HEPES, 136 mM NaCl, 2.5 mM KCl, 10 mM D-glucose, 1.3 mM MgCl₂, 25 mM NaHCO₃, 2 mM CaCl₂, pH 7.3). SypHy was excited at 475 nm and its fluorescence emission collected at 525 nm using a 40×, 1.1 NA water immersion objective. Images were acquired every half second for 200 s using TillVision software (TILL Photonics). At frame 15, cells were stimulated for 15 s at 20 Hz. Quantitative measurements of the fluorescence intensity at individual boutons were obtained by averaging a selected area of pixel intensities using custom written macros in Igor Pro (Wavemetrics). Net fluorescence changes (ΔF) were obtained by subtracting the average intensity of the first 15 frames (F0) from the intensity of each frame (Ft) for individual boutons and normalized to F0 (ΔF/F0). Both the fluorescence upstroke and decay were fitted with a single exponential τ (τ_upstroke and τ_decay, respectively). Data are expressed as mean±s.e.m. and statistical significance was assessed with an unpaired two-tailed Student's t-test (GraphPad Prism).

LRRK2-knockout mice (Charles River, France) and wild-type littermates (n = 3 per genotype) were transcardially perfused with 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M cacodylate buffer. Brains were dissected and post-fixed in the same solution overnight at 4°C. Coronal brain sections (300 µm thick) were cut on a vibratome (Leica, Buffalo Grove, IL) and rectangular pieces of tissue comprising the striatum were dissected. The tissue was post-fixed with 1% OsO₄, 1.5% K₄Fe(CN)₆ in 0.1 M cacodylate buffer, rinsed, stained with 3% uranyl acetate and dehydrated in graded ethanols and propyleneoxide, followed by embedding in EMbed812. Ultrathin sections (70–90 nm) were obtained using an ultramicrotome (Leica), mounted on copper grids and treated with uranyl acetate and lead citrate before being imaged using a JEM-1400 transmission electron microscope (Jeol, Zaventem, Belgium), equipped with a 11Mpixel Olympus SIS Quemesa camera. Images were taken at a 30000× magnification, and morphometry and measurements were performed with ImageJ software. For analysis of synaptic vesicle numbers, 14830 vesicles in total from 112 wild-type and 115 knockout synapses from the six brain samples (n = 3 per genotype) were counted. Synapses were selected based on the presence of an active zone and all synaptic vesicles present within the synaptic vesicle cluster were counted. For analysis of synaptic vesicle diameters, a total of 1043 wild-type and 1040 knockout vesicles from the same brain samples used for the synaptic vesicle counts were measured. Statistical significance was assessed by Student's t-test (GraphPad Prism) and data are expressed as mean±s.e.m.

For patch-clamp experiments, hippocampal neurons were derived from E17–E19 embryos from wild-type and LRRK2-knockout C57Bl6/J mice (Charles River, France). Cells were plated in a density of 50,000 cells per ml on cover glasses. Cultures were maintained in neurobasal medium with B27 supplement (Life Technologies, Gent, Belgium) for 7–12 days, after which recordings were performed in bath solution (140 mM NaCl, 2.4 mM KCl, 10 mM HEPES, 10 mM glucose, 2.5 mM CaCl₂, 1.3 mM MgCl₂, pH 7.3). Pyramidal neurons were voltage clamped under an Axioskop 2FS upright microscope (Carl Zeiss, Nossegem, Belgium) at a holding potential of −70 mV using HEKA software. Electrodes (thick-walled borosilicate glass capillaries with an outer diameter of 1.5 mm and an inner diameter of 0.86 mm, Harvard Apparatus) were pulled with a P97 Flaming/Brown Micropipette puller (Sutter Instruments). Micropipette tips were heated with a micro forge to produce a smooth surface, with a diameter and resistance of 1 µm and 3–5 MΩ. Total sEPSCs were recorded in the absence of TTX to collect all possible synaptic activities including action-potential-mediated vesicle release. Events were recorded with intracellular solution (146 mM K-gluconate, 17.80 mM HEPES, 4 mM Mg-ATP, 0.30 mM Na₂-ATP, 1 mM EGTA and 12 mM phosphocreatine, pH 7.3) (Rost et al., 2010) for 10 s with an EPC10 USB Patch Clamp Amplifier (HEKA). Liquid junctional potential and series resistance were automatically compensated before each recording. To further ensure the recorded cells were indeed neurons, firing patterns were routinely determined at the beginning of each experiment by introducing depolarization steps. Directly after obtaining a whole-cell patch configuration, a 1-min recording of baseline activities was made. After recording baseline activities, bath solution was replaced by hypertonic solution (bath solution with 50 mM sucrose) by perfusion. Sucrose is known to increase exocytosis, and consequently endocytosis, leading to an increase in instantaneous frequency of sEPSCs (Geppert et al., 1994; Rosenmund and Stevens, 1996). After 2 min in hypertonic solution, new recordings were made and all measured sEPSCs are thought to come from newly endocytosed vesicles. For dynasore experiments, cells were incubated in bath solution containing 40 µM dynasore for 5 min before recordings were made. After the first recordings, neurons were incubated in hypertonic solution with added 40 µM dynasore. After 2 min, new recordings were made. All experiments were performed at room temperature. All measurements were analyzed using Igor Pro software (Wavemetrics). Data are expressed as mean±s.e.m. from n cells. Statistical significance was assessed by two-tailed Student's t-test (GraphPad Prism).

Recombinant GST-tagged LRRK2 protein (Life Technologies) and GST-tagged substrates (Abnova, Taipei, Taiwan, or Dario Alessi, University of Dundee, Dundee, UK) were incubated in kinase buffer (50 mM Tris-HCl pH 7.5, 1 mM EGTA, 10 mM MgCl₂, 2 mM DTT, 0.01% Tween-20) in the presence of 1 µM ATP (with 0.5 µCi [³³P]ATP, Perkin Elmer, Zaventem, Belgium) during the indicated time points at 37°C. Reactions were stopped by adding SDS-PAGE sample buffer. Proteins were separated by electrophoresis on 4–12% NuPAGE gradient gels and electroblotted onto nitrocellulose membranes. Phosphorylation of substrates was quantified using Typhoon (GE Healthcare, Belgium) and total protein levels were determined by western blotting using an anti-GST antibody (Sigma-Aldrich, Diegem, Belgium). Data are expressed as mean±s.e.m. Statistical significance was assessed by two-way ANOVA with Bonferroni post-test (GraphPad Prism).

LRRK2 (recombinant protein; Life Technologies) was used in an in vitro phosphorylation assay in the presence or absence of human SH3GL2. Protein samples were separated by SDS-PAGE and stained with Coomassie Brilliant Blue G-250. Excised gel bands were washed with water, followed by 50% acetonitrile and acetonitrile before vacuum drying. Bands were cut in two and one half was fully submerged in digestion buffer (50 mM NH₄HCO₃ in 10% acetonitrile) with 0.1 µg trypsin (Promega) for in-gel digestion. Digestion proceeded overnight at 37°C, following which the generated peptide mixtures were acidified with formic acid. The other half of the protein band was first N-propionylated in-gel before digestion with trypsin, as in this way, side-chains of lysines are blocked (indicated in the supplementary material Table S1 as <prop*>) and no longer recognized by trypsin, which then only cleaves C-terminal to arginines. All peptide mixtures were then analyzed by LC-MS/MS using a LTQ Orbitrap XL mass spectrometer (Thermo Scientific). The generated MS/MS spectra were searched in the human subsection of the Swiss-Prot database by the Mascot algorithm. For phosphopeptide enrichment, TiO₂ Mag Sepharose beads (GE Healthcare) were used prior to LC-MS/MS analysis.

We thank Lucia Chavez-Gutierrez for her advice on the 3D modeling and members of the Patrik Verstreken, Bart De Strooper and Janssen CNS Research laboratories for comments.