Dysregulated LRRK2 Signaling in Response to Endoplasmic Reticulum Stress Leads to Dopaminergic Neuron Degeneration in C. elegans

Mutation of leucine-rich repeat kinase 2 (LRRK2) is the leading genetic cause of Parkinson's Disease (PD), manifested as age-dependent dopaminergic neurodegeneration, but the underlying molecular mechanisms remain unclear. Multiple roles of LRRK2 may contribute to dopaminergic neurodegeneration. Endoplasmic reticulum (ER) stress has also been linked to PD pathogenesis, but its interactive mechanism with PD genetic factors is largely unknown. Here, we used C. elegans, human neuroblastoma cells and murine cortical neurons to determine the role of LRRK2 in maintaining dopaminergic neuron viability. We found that LRRK2 acts to protect neuroblastoma cells and C. elegans dopaminergic neurons from the toxicity of 6-hydroxydopamine and/or human α-synuclein, possibly through the p38 pathway, by supporting upregulation of GRP78, a key cell survival molecule during ER stress. A pathogenic LRRK2 mutant (G2019S), however, caused chronic p38 activation that led to death of murine neurons and age-related dopaminergic-specific neurodegeneration in nematodes. These observations establish a critical functional link between LRRK2 and ER stress.


Introduction
Parkinson's disease (PD) is a major neurodegenerative disease that results from the loss of dopaminergic (DAergic) neurons in the substantia nigra of patients. The leading genetic cause of PD is mutation of leucine-rich repeat kinase 2 (LRRK2) [1,2], which is associated with both familial and idiopathic PD [3,4], and represents a potential therapeutic target [5]. The biological functions of LRRK2 remain poorly defined and the molecular mechanisms by which LRRK2 pathogenic mutations contribute to neurodegeneration are largely unknown [6,7]. Transgenic animal models featuring wild type (WT) and mutant forms of human LRRK2 have been generated in nematodes [8,9], flies [10][11][12] and rodents [13][14][15]. In these models, LRRK2 was found to interact with components involved in the autophagylysosomal pathway [16] or protein quality control [15,17], modulate oxidative stress [8,17], regulate protein synthesis [18] and mediate the microRNA pathway [19], indicating that multiple mechanisms may underlie LRRK2 pathology [7,19]. Moreover, confusing and even conflicting experimental results have been reported. For example, observations made in various animal models with gain-or loss-of LRRK2 kinase activity have led to conclusions that LRRK2 kinase activity is protective, deleterious or dispensable for neuronal survival [8,10,[20][21][22]. Therefore, it is important to precisely define the signaling pathways of LRRK2 and their distinct contribution to DAergic neuron viability.
Mutations of human a-synuclein (haSyn) or exposure to neurotoxins such as 6-hydroxydopamine (6-OHDA) also causes DAergic neuron degeneration in humans and animal PD models [23][24][25][26][27]. Recently, a pathophysiological interplay between LRRK2 and a-synuclein was demonstrated by experiments in which overexpression of LRRK2 enhanced pathogenic a-synuclein-induced neurophathological abnormalities in transgenic mice [28]. The molecular mechanism(s) underlying this important observation and other reported interactions between PD genetic/ environmental factors remain unclear, however.
The nematode, C. elegans may constitute a useful model to study genetic mechanisms underlying its pathology. For example, nematodes were used to study the function of LRK-1, the sole nematode homolog of LRRK2, in synaptic protein sorting [29] and organism survival after exposure to mitochondrial toxins [8]. But, the role of LRRK2 kinase activity in maintaining DAergic neuron viability has not been studied in nematodes, although expression of human pathogenic LRRK2 in nematodes leads to DAergic neuron degeneration and motor activity deficit [30].
In the work reported here, we investigated the molecular mechanism by which LRRK2 impacts the viability of DAergic neurons of C. elegans, a human neuroblastoma cell line and murine cortical neurons. We found that deficient LRRK2 signaling increased haSyn-or 6-OHDA-mediated neuroblastoma cell death or nematode DAergic neuron degeneration. LRRK2 executes this cytoprotective role by supporting synthesis of GRP78, a chaperone that plays a key role in promoting cell survival following ER stress [31], possibly signaling through the p38 pathway. On the other hand, human pathogenic G2019S mutant LRRK2, which exhibits enhanced kinase activity and causes chronic p38 activation in human cells, displayed adult-onset, progressive, DAergic-specific neurodegeneration that was dependent on functional p38. Overactive LRRK2 signaling was also found to promote death of G2019S LRRK2-expressing primary murine cortical neurons through p38. These data suggest that ''appropriate'' LRRK2 activity in neurons is cytoprotective, whereas overactive LRRK2 is degenerative. Interestingly, this mechanism of LRRK2 is conserved between nematodes and human cell lines. Together, we identified an unexpected functional link between LRRK2 signaling and ER stress response.

Results
The DAergic neurotoxicity of 6-OHDA is potentiated in lrk-1 mutant nematodes To investigate the molecular mechanism by which LRRK2 impacts neuron viability, we chose to use the DAergic neurotoxin, 6-OHDA, for its experimental convenience, and nematodes for their readily manipulated genetics. 6-OHDA has been used previously in mammals [32] and C. elegans [25] to induce DAergic neuron degeneration and study PD pathology. In nematode hermaphrodites, 6-OHDA-induced DAergic neuron degeneration was dose-dependent and required a functional dopamine transporter (DAT) [25]. Degeneration of these cells in nematode hermaphrodites following either 6-OHDA treatment or haSyn expression was evidenced by Electron Microscopy (EM) or tyrosine hydroxylase (TH) immuonstaining. Degeneration was also visualized in vivo through DAergic neuron-specific expression of the fluorescent markers, GFP or DsRed [25,33]. Using this method, we observed dose-dependent 6-OHDA-induced DAergic neuron degeneration in our DsRed expressing nematode line and found that this degeneration could be prevented by co-treatment with the DAT blocker, imipramine ( Figure S1). Nematode hermaphrodites have a total of eight DAergic neurons: 4 CEPs, 2 ADEs and 2 PDEs. All of these DAergic neurons showed similar 6-OHDA-induced, imipramine-blockable degeneration, although only the DAergic neurons located in the nematode head (CEPs and ADEs) are shown in Figure S1A-F.
To test whether LRK-1, the sole nematode homolog of LRRK2 [29], plays a role in maintaining DAergic neuron viability, we examined the effect of 6-OHDA treatment on lrk-1 mutant nematodes as compared to wild type (Bristol N2) nematodes. We found that a concentration of 6-OHDA (2 mM) that produced little or no DAergic neuron degeneration in wild type nematodes ( Figure 1A and Figure S1), induced substantially more severe DAergic neuron degeneration in several nematode strains with loss-of-function lrk-1 mutations ( Figure 1A and Figure S1G-I). The tested mutations did not themselves affect DAergic neuron viability (vehicle-treated wild type and mutant nematodes had similar numbers of DAergic somas, e.g., in Figure S1H) or alter the dopamine-regulated nematode food response (data not shown). These results indicate that LRK-1 protects against the DAergic neurotoxicity of 6-OHDA in nematodes.
Human LRRK2 can functionally substitute for LRK-1 in protecting nematode DAergic neurons from 6-OHDAinduced degeneration LRRK2 is expressed in mammalian DAergic neurons [34,35]. Although the nematode LRRK2 homolog, LRK-1, was previously reported to have a pan-neuronal expression pattern [29], whether it is truly expressed in nematode DAergic neurons was not addressed in previous studies [8,29]. Therefore, we generated a nematode line expressing a fusion protein composed of the Nterminus of nematode LRK-1 linked to GFP (LRK-1N::GFP, driven by the natural LRK-1 promoter) [29] and DsRed driven by the DAergic-specific promoter of dat-1 (P dat-1 ), the sole dopamine transporter homolog in nematodes [33]. Analysis of these animals showed that lrk-1 was expressed in all eight of the nematode DAergic neurons together with DsRed ( Figure S2). This finding supports the possibility that LRK-1 plays a physiological role in maintenance of DAergic neuron viability.
Having shown that lrk-1 mutant nematodes display increased sensitivity to the neurotoxic effects of 6-OHDA, we next tested whether expression of human LRRK2 could reverse this effect. We expressed WT and K1906M mutant (kinase inactive) [36] forms of human LRRK2 driven by either P dat-1 or a pan-neuronal promoter (P H20 [37]) in lrk-1 mutant (km17) nematodes. Expression of WT, but not K1906M, LRRK2 in DAergic neurons was sufficient to functionally substitute for LRK-1 and rescue the cells from 6-OHDA-induced degeneration ( Figure 1A). These observations indicate that LRRK2 expressed in DAergic neurons protects against neurotoxin-induced DAergic neuron degeneration, and the kinase activity of LRRK2 may be important for this role of LRRK2. Together, these data establish C. elegans combined with an appropriate concentration of 6-OHDA as a model to identify potential functional partners of LRRK2.
Two MAPKKs lie upstream of p38, MKK3 and MKK6, and the nematode homolog of both is sek-1 [40]. We found that nematodes with a loss-of-function mutation in sek-1 consistently displayed enhanced susceptibility to 6-OHDA-induced neurotoxicity ( Figure 1B and Figure S3 panel E). We then generated lrk-1;pmk-1 and lrk-1;sek-1 double mutant nematodes and found that these double mutants were no more sensitive to 6-OHDA-induced DAergic neuron degeneration than their single mutant counterparts (lrk-1, pmk-1 or sek-1). They also did not exhibit a DAergic neuron degenerative phenotype in the absence of 6-OHDA ( Figures 1B, S3 and data not shown). These results suggest that LRRK2, MKK3/6 and p38 function in the same genetic pathway.
Based on our finding that p38 is involved in protection of DAergic neurons from 6-OHDA-induced degeneration, we examined whether p38 becomes activated in the presence of the neurotoxin. As shown in Figure 1C, treatment of 6-OHDA for 24, but not 4 hr resulted in LRK-1-dependent phosphorylation of p38 on Thr191/Tyr193 (equivalent to Thr190/Tyr192 in humans). These data indicate that LRK-1 functions upstream of PMK-1/p38 to maintain DAergic neuron viability in the presence of 6-OHDA.
LRRK2 acts through the p38 pathway to protect human neuroblastoma cells from the toxicity of 6-OHDA To confirm that LRRK2 acts upstream of p38 in human cells, we chose to use the SH-SY5Y human neuroblastoma cell line since it is widely used for PD-related research and is sensitive to 6-OHDA [43]. Previously, 6-OHDA-induced p38 activation was observed in both SH-SY5Y cells and NM9D cells, another human neuroblastoma cell model of DAergic neurons [43,44]. Here, we found that 6-OHDA induced transient activation of p38 in SH-SY5Y cells that peaked about 15 minutes after 6-OHDA exposure before disappearing and was then followed by a second stronger peak of activation at 4 hours ( Figure 2A, upper panel). This second peak was 6-OHDA-dose-dependent ( Figure 2A, lower panel), and was not observed in HEK293 cells (data not shown), a human kidney cell line.
To address whether toxins other than 6-OHDA trigger LRRK2/p38, we treated cells with H 2 O 2 , an oxidant that induces oxidative stress. In both MIX LRRK2 KD cells and SH-SY5Y cells transfected with control vector, two-phase p38 activation similar to the case of 6-OHDA exposure was detected ( Figure S4). The second p38 activation was found also to be reduced in LRRK2 KD cells ( Figure S4).
To examine whether the LRRK2 kinase domain is required for p38 activation following 6-OHDA treatment, we transfected cells from a 39-UTR LRRK2 KD SH-SY5Y cell line ( Figure S5A  (circles) were exposed to the indicated concentrations of 6-OHDA for 24 hours and cell survival was assessed with a XTT-based calorimetric assay. Data represent the mean 6 SEM of 4 independent experiments. **, p,0.01 by two-way ANOVA, and *, P,0.05; **, P,0.01 by t-test. doi:10.1371/journal.pone.0022354.g002 transfected LRRK2 KD cells was similar to that in unreconstituted LRRK2 KD cells ( Figure 2C), suggesting that it resulted from the activity of residual endogenous LRRK2 in the LRRK2 KD cell line. Taken together, these data indicate that, as in nematodes, LRRK2 activates the p38 pathway in response to 6-OHDA exposure in human neuroblastoma cells. The neurotoxicity of 6-OHDA is potentiated by either knock-down of LRRK2 expression or inhibition of p38 activity To test the impact of LRRK2 and p38 signaling on degeneration of human neurons, we examined the viability of control and MIX LRRK2 KD SH-SY5Y cells exposed to 6-OHDA with an XTTbased calorimetric assay, a method previously used to quantify 6-OHDA-induced cell death [43]. We found that 100 mM 6-OHDA induced cell death more rapidly in LRRK2 KD lines than in control cells ( Figure 2D and S5E). Dose response experiments also demonstrated that LRRK2 KD cells were more sensitive to 6-OHDA than control vector-infected cells ( Figure 2E). We tested the role of p38 by using a specific inhibitor of p38 kinase activity, PD169316 [47]. MIX LRRK2 KD SH-SY5Y cells and vectortransfected SH-SY5Y cells were treated with 100 mM 6-OHDA together with either control vehicle or PD169316 for 24 hours. As shown in Figure 2F, inhibition of p38 led to enhanced 6-OHDAinduced death of control SH-SY5Y cells, but not LRRK2 KD cells. It is possible that 6-OHDA already caused ,80% death of LRRK2 KD cells, it is difficult for p38 inhibition to potentate more cell death. This observation is consistent with the concept that LRRK2 signals through p38 to protect cells against 6-OHDA toxicity. We next found that 39-UTR LRRK2 KD SH-SY5Y cells reconstituted with K1906M mutant LRRK2 were more vulnerable to 6-OHDAinduced toxicity than 39-UTR LRRK2 KD cells reconstituted with WT LRRK2 ( Figure 2G). The smaller difference in cell survival following 6-OHDA exposure between WT-transfected and K1906M LRRK2-transfected LRRK2 KD SH-SY5Y cells as compared to the difference between control and LRRK2 KD SH-SY5Y cells ( Figure 2E) is consistent with the respective differences in 6-OHDA-mediated p38 activation in the various cell lines ( Figure 2B and Figure 2C). Together, these data support the concept that LRRK2 signals through p38 to protect against 6-OHDA-induced neurotoxicity and that both the signaling and function of LRRK2 and the p38 pathway are conserved between nematodes and humans.

LRRK2 and the p38 pathway protects DAergic neurons against haSyn-mediated degeneration
As for neurotoxins such as 6-OHDA, haSyn-induced death of neurons may also involve stress response pathways [48]. To further explore the physiological role of LRRK2 and the p38 signaling in neuron degeneration, we used a DAergic neuron specific haSyn expressing nematode model that exhibits ageingassociated DAergic neuron degeneration and motor deficit, key PD pathogenic features [33]. As shown in Figures 3A-B, loss of DAergic neurons in haSyn expressing nematodes was exacerbated in loss-of-function mutants of lrk-1 (several different alleles), pmk-1/ p38, and sek-1/MKK6 and in double mutants of lrk-1+pmk-1/p38 and lrk-1+sek-1/MKK6, but not in mutants of mpk-2/ERK and jnk-1/JNK. These data indicate that LRRK2 and the p38 signaling protects against DAergic neuron degeneration induced by both 6-OHDA and haSyn.
Importantly, neuronal expression of human WT and kinase inactive (K1906M) LRRK2 driven by P H20 in km17 nematodes either fully or partially rescued the enhanced susceptibility of these animals to haSyn-induced neurotoxicity ( Figure 3C). This observation suggests that LRRK2 kinase activity plays a critical role in protecting DAergic neurons from haSyn-mediated degeneration, but that a kinase-independent function of LRRK2 may also contribute to this effect. Consistent with this notion, it was previously shown that LRRK2 was required for protection of C. elegans [8] and Drosophila [21] from death induced by mitochondrial toxins, such as rotenone and paraquat, but that this effect was independent of LRRK2 kinase activity. These observations also suggest that LRRK2 mediates whole organism death induced by oxidative stress or by haSyn DAergic neuron degeneration via distinct mechanisms, being completely or partially independent of LRRK2 kinase activity, respectively.
LRRK2 and p38 signaling modulate GRP78 synthesis to potentiate cell survival Endoplasmic reticulum (ER) stress is associated with the cytotoxicity of both 6-OHDA and haSyn in mammalian cells [46,48]. Therefore, to further explore the biological role(s) of LRRK2 and the p38 signaling, we examined its relationship to Glucose Regulated Protein 78 (GRP78, also called Binding Immunoglobulin Protein (BiP)), an ER chaperone that is critical for cell survival in the face of ER stress [31]. We found that shRNAmediated suppression of LRRK2 markedly compromised 6-OHDA-induced upregulation of GRP78 at both the translational ( Figure 4A) and transcriptional ( Figure 4B) levels. The specificity of this effect was confirmed by the finding that knock-down of LRRK2 did not affect 6-OHDA-induced transcriptional upregulation of calreticulin ( Figure S6), an ER chaperone protein associated with cell death triggered by extracellular signals [49,50]. Thus, modulation of the ER stress response by the LRRK2 signaling cascade appears to be selective. Increased levels of GRP78 were first observed ,3 hours after addition of 6-OHDA to cell cultures, which coincided in time with LRRK2-dependent activation of p38 ( Figure 2A). 6-OHDA-induced upregulation of GRP78 protein was also found to be reduced by p38 inhibition ( Figure 4C) and dependent on LRRK2 kinase activity ( Figure 4D). These findings suggest that GRP78 lies in the LRRK2-dependent pathway leading from 6-OHDA to cell death. Consistent with this hypothesis, shRNA-mediated knock-down of GRP78 in SH-SY5Y cells ( Figure  S5C) exacerbated 6-OHDA-induced cell death, and shRNAmediated double KD of GRP78 and LRRK2 displayed a similar 6-OHDA-indcuced cell death rate to that observed in single KD of either GRP78 or LRRK2 ( Figure 4E and Figure S5D). Heterologous expression of subtilase A, a functional blocker of GRP78 [51], in LRRK2 KD cells led to cell death or severe unhealthy conditions, indicating synthetic lethality (data not shown).
To further investigate the link between LRRK2 signaling and ER stress, we tested the effect of tunicamycin, a widely used agent that triggers ER stress in many organisms [52], on the viability of control and MIX LRRK2 KD SH-SY5Y cells. Treatment with either 3 or 6 mM tunicamycin resulted in significant cell death in MIX LRRK2 KD, but not control SH-SY5Y cells, as evidenced by the substantial population of cells with sub-G0 DNA content in LRRK2 KD cultures exposed to this agent ( Figure 5A and Figure  S7A-B). Consistent with this observation, tunicamycin-induced upregulation of GRP78 was compromised in LRRK2 KD cells ( Figure 5B). Based on these results, we conclude that LRRK2 signaling leads to upregulation of GRP78 synthesis, which serves to support cell survival in the face of ER stress.
Does LRRK2/LRK-1 have a functional connection with ER stress in vivo? We first examined the effect of 6-OHDA treatment on the abundance of HSP-4::GFP (zcIs4), a functional fusion protein of HSP-4, a nematode ortholog of GRP78, and GFP. It was previously demonstrated that expression of this fusion protein was induced by tunicamycin treatment or heat shock [53]. We found that 6-OHDA exposure enhanced HSP::GFP fluorescence ( Figure 5C). Interestingly, fluorescence of HSP-4::GFP was significantly elevated, suggesting increased ER stress in lrk-1 mutant nematodes, and 6-OHDA exposure did not further increase HSP::GFP fluorescence.
We next examined 6-OHDA-mediated neurodegeneration and used two loss-of-function mutants of HSP-3 and HSP-4, the only orthologs of GRP78 [51]. As shown in Figure 5D, the double mutant of lrk-1;hsp-3 or lrk-1;hsp-4 exhibited slightly more 6-OHDA-mediated DAergic neuron degeneration than single mutants of lrk-1 or hsp-3, but not statistically significant. In addition, triple mutation of lrk-1;hsp-3;hps-4, but not any double mutation of these three mutations, was found to be lethal. Animals carrying this triple mutation grew slower, could not lay eggs and died before day 6, indicating synthetic lethality.
We next examined the contribution of GRP78 in haSynmediated nematode DAergic neuron degeneration by testing the effect of loss-of-function mutation of HSP-4 on nematode DAergic neuron integrity maintenance, which is currently unknown. We found that haSyn induced more severe DAergic neuron degeneration in the hsp-4 mutant background ( Figure 5E).
Taken together, we conclude that LRRK2/LRK-1 supports GRP78-mediated cell survival in vitro and in vivo.
Expression of human pathogenic G2019S mutant LRRK2 in nematodes elicits adult-onset, ageing progressive, DAergic-specific neurodegeneration G2019S is the most common LRRK2 mutation associated with PD [54]. It has been shown that G2019S LRRK2 has enhanced capacity for in vitro phosphorylation of MKKs as compared to WT LRRK2 [36]. However, it remains unclear how this enhanced kinase activity contributes to G2019S-associated cytotoxicity. Having found that LRRK2 signaling counteracts neurodegeneration in the experiments described above, we set out to test the impact of G2019S mutant LRRK2 expression on this pathway in C. elegans. Therefore, we generated wild type nematode lines with pan-neuronal expression of WT LRRK2, K1906M or G2019S mutant LRRK2 (driven by the P H20 promoter) as well as DAergic neuron-specific expression of DsRed and command interneuronspecific expression of yellow fluorescent protein (YFP, driven by the nmr-1 promoter [55]). The expression of WT LRRK2, K1906M or G2019S mutant LRRK2 in wild type and lrk-1 mutant were confirmed by qRT-PCR ( Figure S8). Although G2019S LRRK2 expression in nematodes did not alter the number of identified DAergic neurons in larvae, it did result in mild but statistically significant ageing-associated progressive degeneration of DAergic neurons ( Figure 6A). Command interneurons expressing G2019S did not display this phenotype ( Figure 6B). In addition, pan-neuronal expression of WT or kinase inactive LRRK2 induced little or no DAergic neuron degeneration ( Figure 6A). Importantly, G2019S LRRK2-mediated DAergic neuron degeneration was reduced in nematodes with loss-offunction lrk-1 mutation ( Figure 6A), although this reduction was not statistically significant. Thus, DAergic-specific neurodegeneration in G2019S LRRK2 expressing nematodes was associated with LRK-1/LRRK2 kinase activity. Although it is possible that the different expression levels of G2019S in DAergic neurons and command interneurons contributed to DAergic-specific neurodegeneration in our experiments, the obtained results are consistent with the established fact that LRRK2 is associated with PD [54].

Chronic p38 activation contributes to G2019S-induced DAergic neurodegeration
Since the ability of LRRK2 to protect nematode DAergic neurons and human neuroblastoma cells from 6-OHDA and haSyn was found to involve p38 activation, we examined whether p38 was also involved in promotion of DAergic neuron degeneration by G2019S mutant LRRK2. We transiently transfected 39-UTR LRRK2 KD SH-SY5Y cells with WT or G2019S LRRK2 and found that expression of the G2019S mutant resulted in substantial basal p38 activation. The G2019Sexpressing LRRK2 KD cells, but not the corresponding WT LRRK2-expressing cells, contained activated (phosphorylated) p38 without exposure to a neurotoxin ( Figure 6C). Both the level of LRRK2 protein ( Figure S5B) and the level of p38 activation induced by 6-OHDA ( Figure 6C) was similar in the WT and G2019S reconstituted cell lines. Our finding that G2019S expression leads to constitutive p38 activation suggested that chronic p38 activation might contribute to DAergic neurodegeneration [56]. To further explore this hypothesis, we crossed a wild type nematode line expressing human G2019S LRRK2 in all neurons with a pmk-1 knockout mutant nematode line. Animals of these two parental nematode lines exhibited similar numbers of DAergic neurons as larvae ( Figure 1B and Figure 6A). Analysis of the progeny G2019S-expressing pmk-1 mutant nematodes showed that although they had fewer identified DAergic neurons as larvae than G2019S-expressing wild type nematodes, they did not display the G2019S-mediated adult-onset DAergic neurodegeneration observed in wild type nematodes ( Figure 6D). Taken together, these results indicate that chronic p38 activation contributes to pathogenic G2019S mutant LRRK2-mediated DAergic neuron degeneration in nematodes.

Activation of p38 contributes to cytotoxicity induced by G2019S mutant LRRK2 expression in murine neurons
To validate that activation of p38 contributes to G2019Smediated neurodegeneration, we analyzed the effect of expression of different forms of LRRK2 on the viability of murine neurons. Cortical neurons were dissected from murine E18 embryos, transfected with expression vectors for WT, G2019S, or K1906M LRRK2 or with a control vector and then cultured for 48 h. We examined apoptosis ( Figure 7A-B) and viability ( Figure 7C-D) of the transfected neurons and found that expression of G2019S and WT, but not K1906M, LRRK2 markedly increased apoptosis and decreased survival of murine cortical neurons. This result is consistent with the notion that kinase activity mediates the cytotoxicity of G2019S in murine neurons [57]. In our experiments, expression of WT LRRK2 also increased apoptosis and decreased viability of cortical neurons ( Figure 7A-B), presumably due to the high levels of LRRK2 kinase activity that result from overexpression of the protein. We found that p38 inhibition (using the specific chemical PD169316) potentiated death and decreased viability of neurons transfected with control vectors (data not shown), which is consistent with previous reports [58][59][60]. In contrast, p38 inhibition significantly reduced the cytotoxicity caused by G2019S or WT LRRK2 expression. Thus, exposure to PD169316 resulted in reduced apoptosis ( Figure 7C) and increased viability ( Figure 7D) of neurons transfected with G2019S or WT LRRK2. Based on these observations, we concluded that chronic p38 activation contributes to the cytotoxicity of G2019S LRRK2 in murine neurons.

Discussion
Accumulated evidence in the literature indicates that multiple mechanisms may underlie LRRK2 pathology. For example, LRRK2 interacts with components involved in the autophagylysosomal pathway [16] or protein quality control [15,17], modulates oxidative stress [8,17], regulates protein synthesis [18] or mediates the microRNA pathway [19]. In this study, we identified a functional interaction between LRRK2 and GRP78, a key molecule that promotes survival of cells under ER stress. Insufficient LRRK2 kinase activity resulted in more vulnerable nematode neurons and human neuroblastoma cells, possibly through p38 signaling. Overactive LRRK2 kinase activity, however, led to nematode DAergic neuron degeneration and mammalian primary neuron death, partially through chronic p38 activation.
Specifically, we found that LRRK2 supported GRP78 upregulation in human neuroblastoma cells exposed to 6-OHDA, and that GRP78/HSP-4 protected these cells and nematode DAergic neurons against the toxicity of 6-OHDA or haSyn. This cytoprotective mechanism of LRRK2 required its kinase activity and could execute through p38 signaling either directly or indirectly [9,38,61,62], evidenced by data present in this and previous studies. For example, 6-OHDA treatment induced LRK-1-dependent p38 activity in nematodes. Moreover, lrk-1 epistatically interacted with components of the p38 pathway in 6-OHDAor haSyn-mediated cell death or nematode neurodegeneration. Previously, it was also shown that p38 provides a survival signal during ER stress [52,63]. However, an alternative explanation may exist.
While accumulated evidence has illustrated the importance of the kinase domain of LRRK2 to PD pathology, its molecular pathogenic target(s) associated with PD pathology is unknown. One possibility is that LRRK2 functions as a MAP kinase kinase kinase (MAPKKK). In support of this hypothesis, in vitro kinase assays demonstrated that LRRK2 catalyzes the phosphorylation of several MAP kinase kinases (MAPKK) [36]. Furthermore, LRRK2 was reported to regulate the extracellular signal-regulated kinase (ERK) pathway to attenuate H 2 O 2 -induced oxidative stress [64,65]. In LRRK2-mediated protection against mitochondrial stress in C. elegans, LRRK2 interacted with MKKs [9]. Here, we found that LRRK2/LRK-1 functions upstream of p38 in maintaining DAergic neuron integrity. Contrary to the concept that LRRK2 functions as a MAPKKK, an ex vivo kinase substrate tracking and elucidation (KESTREL) assay identified moesin, an anchor protein between the actin cytoskeleton and the plasma membrane, as a substrate of LRRK2 [66]. A recent study showed that moesin phosphorylation by LRRK2 regulates neuronal morphogenesis by promoting actin cytoskeleton rearrangement [67]. Alternatively, LRRK2 was found to phosphorylate 4E-BP to mediate overall protein translation [12] and forkhead box transcription factor FoxO1 to enhance the transcription activity of FoxO1 [18], and directly interact with RNA-induced silencing complex (RISC) [19], all related to neurodegeneration in Drosophila. It is possible that LRRK2 has multiple kinase substrates, multiple functions and multiple mechanisms contributing to cell physiology.
To further define the role of LRRK2 in maintaining neuron integrity, we also overexpressed constitutive kinase active (G2019S) and kinase dead (K1906M) LRRK2 in nematodes or murine primary cortical neurons. In nematodes, we found that expression of G2019S nematodes, but not K1906, LRRK2 led to p38-dependent neurodegeneration that was reduced in a lrk-1 mutant background. Consistently, G2019S expression in human blastoma cells activated p38 and chronic p38 activation was previously reported to be cytotoxic in cell lines [56]. It is interesting that we examined two lrk-1 mutant nematode lines expressing G2019S, one generated from the genetic cross of a lrk-1 mutant and a wild type nematode line expressing G2019S LRRK2; and another one generated by injecting G2019S LRRK2 construct into lrk-1 mutant nematodes. We found that the transgenic nematode line generated by this cross displayed no DAergic neurodegeneration ( Figure 6A) while the transgenic nematode line generated by injection showed some DAergic neurodegeneration ( Figure 1A). The amount of G2019S LRRK2 DNA construct injected into wild type and lrk-1 mutant nematodes to generate these two transgenic nematode lines was the same. The exact reason for this discrepancy is not clear, but it is possible that lrk-1 mutant nematodes tolerated more injected G2019S LRRK2 construct to survive [8]. However, this more G2019S expression induced DAergic neuron degeneration. If so, this observation is consistent with the conserved function in DAergic neuron integrity maintenance between nematode LRK-1 and mammalian LRRK2. In further support of this notion, chronic p38 activation elicited by expression of G2019S LRRK2 contributed to G2019Smediated murine primary cortical neuron death.
It is interesting that heterologous G2019S-mediated neurodegeneration was observed in DAergic neurons but not command interneurons. Although this specificity is consistent with the relevance of LRRK2 with PD, it should be noted that different levels of LRRK2 expression in these two types of neurons can also explain our results. Interestingly, a small portion of the LRK-1 N-terminus was found to promote the expression of GFP in a majority of neurons, especially emphasizing DAergic neurons [29]. It is possible that LRRK2 expression has similar behavior.
All together, our observations indicate that regulated LRRK2 activity is important for the integrity of neurons of both nematodes and mammals. Interestingly, putative kinase loss-or gain-offunction LRRK2 mutations are both associated with PD [3] and LRRK2 kinase activity has been shown to be tightly regulated [68].
It is worthy to note that neuronal expression of human WT and kinase inactive (K1906M) LRRK2 driven by P H20 in km17 nematodes rescued the enhanced sensitivity to a-synuclein neurotoxicity, fully or mildly ( Figure 3D). This observation suggests that LRRK2 kinase activity plays a critical role to defend against a-synuclein-mediated DAergic neuron degeneration, but a kinase-independent role of LRRK2 may also contribute. Consistently, LRRK2-mediated protection against mitochondrial toxins, such as rotenone or paraquat, in the survival of C. elegans [8] or Drosophila [21] was found to be LRRK2 kinase dispensable. These observations also suggest that LRRK2 mediates oxidative stressinduced whole organism death and a-synuclein-mediated DAergic neuron degeneration via distinct mechanisms.
We were initially surprised to identify a functional interaction between LRRK2 and GPR78, an ER-resident chaperone that plays a central role in ER stress survival [69]. But, the cytotoxicty of 6-OHDA and haSyn was found to associate with ER stress in mammalian cells [48] and nematodes (evidence presented here). Moreover, we found that LRRK2 also is important for human neuroblastoma cell death mediated by tunicamycin, a chemical that causes accumulation of unfolded proteins by inhibiting protein glycosylation [70]. Although the interplay between LRRK2 and GRP78 seems to be conserved between nematodes and mammals, this notion should be tested in mammalian animal models and its pathological relevance is not clear. It is worth to note that LRRK2 mainly associated with the ER of human neurons and displayed a Nissl-like pattern. Interestingly, such a Nissl-like pattern of LRRK2 was specifically disorganized in DAergic neurons of idiopathic PD cases and associated with LBs [71].
Genetic model organisms such as C. elegans and Drosophila constitute useful PD models. For example, expression of GFP in DAergic neurons led to a nematode PD model that may serve as a chemical screen platform for drugs that prevent or slow down neurotoxin-mediated DAergic neuron degeneration [25]. Heterologous expression of haSyn-expressing PD C. elegans and Drosophila models helped understand the important pathogenic role of haSyn [72][73][74][75][76]. Recently, heterologous expression of LRRK2 in C. elegans and Drosophila confirmed the important role of LRRK2 in mitochondrial function [8] or identified the regulation of microRNA-mediated translational repression by LRRK2 [19]. Here, we identified a functional connection between LRRK2 and GRP78 using nematode. Given their facile genetics, these C. elegans and Drosophia PD models should continuously make contributions to some aspects of PD-related research.

Nematode 6-OHDA exposure
Synchronized L3 nematodes were treated with various concentrations of 6-OHDA for 1 hour, collected and then cultured in OP50 seeded NGM plates according to a previously published protocol [25]. Neurodegenerative and biochemical experiments were performed after nematodes reached targeted ages.

Microscopy
All confocal experiments were performed using a Leica TCS SP2 confocal microscope. The spectra used were: DsRed (l ex = 543 nm and l em = 580-630 nm) and GFP (l ex = 488 nm and l em = 510-530 nm). To count fluorescent DAergic neurons, we first immobilized living nematodes on 2% agarose pads with 3 mM sodium azide and then examined them with a Leica DMI3000 microscope according to a published method [33]. Sample images were captured with an Andor iXon EM 885 EMCCD camera and SimImaging (Feng, Z. unpublished software). For HSP-4::GFP fluorescence quantification, HSP-4::GFP fluorescence of each nematode was first obtained with the freehand tool of National Instruments Vision Assistant. Background intensity was next subtracted. Twenty animals from each of the four experimental groups were examined, normalized to wild type control nematodes and graphed. All images were processed and analyzed with National Instruments Vision Assistant 8.5 (Austin, TX).

Quantification of DAergic neurons in nematodes
DAergic neuron degeneration in nematodes was quantified by using a Leica DMI3000 microscope with 406 objective lens. A DAergic neuron was counted as surviving if its fluorescent soma was clearly observed in the expected position of the nematode body and possessed at least one dendrite that was over twice the length of its soma. We used this algorithm because DsRed puncta formed mainly in the DAergic neurites before severe DAergic soma degeneration in our haSyn expressing nematode lines [33]. These puncta can be easily distinguished from DAergic soma by their position in the animals. A similar algorithm was previously used to quantify LRRK2-mediated primary murine neuron degeneration [57]. Because transient transgenic nematode lines were used for some of our experiments, DsRed expression in DAergic neurons could be mosaic. As a result, not all DAergic neurons displayed strong fluorescence to allow unbiased identification of DAergic somas, especially in larval stages when nematode neurons are small and their fluorescent dendrites are slim. In a small proportion of the observed nematode L4 (the 4th larvae stage), i.e., day 0 nematodes, one out of the total eight DAergic somas (mostly the PDEs located in a nematode tail) in a nematode from a strain including wild type could be missed by an observer. To remove this noise, we normalized the mean DAergic soma number for 20-30 animals/strain in each independent experiment to the mean number of DAergic somas observed in similar number of wild type L4, which was obtained in the same experiment as a control. These normalizations resulted in a relative DAergic soma number over 1 for some adult nematodes of strains including Bristol N2. The relative DAergic soma number of wild type L4 nematodes in the same experiment was 1.

Culture of murine embryonic cortical neurons and human neuroblastoma cells
Cortical neurons from E18 murine embryos (Swiss Webster, Charles River, Wilmington, MA) were dissociated and collected as previously described [57]. After transfection (see next section), these embryonic murine neurons were plated on laminin-(Invitrogen, Carlsbad, CA) and poly-L-lysine-(MP Biomedicals, Solon, OH) coated plates and cultured in Neurobasal Medium with addition of GlutaMAX, B-27 supplement and penicillin/ streptomycin (Invitrogen). Human SH-SY5Y neuroblastoma cells were cultured in a 1:1 mixture of modified Eagle's medium and F-12 Ham's medium with 10% fetal bovine serum and incubated in 5% CO 2 at 37uC.

Transfection and generation of stable shRNA LRRK2 KD cell lines
For transient LRRK2 transfections, primary murine cortical neurons or human SH-SY5Y cells were transfected using the Amaxa nucleofector system (Lonza, Switzerland) and HEK293T cells were transfected with Lipofectamine 2000 (Invitrogen) following the manufacturers' recommended protocols. Primary murine cortical neurons were cotransfected with LRRK2 expression constructs and pEGFP vector (Clontech) at a 15:1 ratio [57]. To generate SH-SY5Y cell lines stably expressing LRRK2 or GRP78 KD, lentiviral shRNA constructs were expanded through infection of HEK293T cells cultured in 10 cm plates. Viruses were harvested and used to infect SH-SY5Y cells three times in succession. Infected SH-SY5Y cells were then cultured in the presence of puromycin (0.5 mg/ml) to select stable cell lines. The shRNAs used included: for the MIX LRRK2 KD line, a mixture of all five clones (1 mg each) from the LRRK2 shRNA set (Open Biosystems, Lafayette, CO, RHS4533-NM_198578); for the 39 UTR LRRK2 KD line, 5 mg of clone TRCN0000021459 of RHS455-NM_198578 (mature sense sequence: CGTGTGTATGAAGGAATGTTA); for the GRP78 KD line, a mixture of all five clones (1 mg each) from the GRP78 shRNA set (Open Biosystems, RHS4533-NM_005347). . Primary murine neurons were cultured with or without 20 mM PD169316 for 48 h after transfection. In survival assays, GFP-positive (i.e., transfected) viable neurons, possessing neurites twice the length of the soma, were counted in thirty randomly selected 406 microscopic fields [57]. Apoptosis of neurons was analyzed using a TUNEL-based in situ cell death detection kit, TMR Red (Roche). In both apoptosis and viability assays, at least 15 neurons were counted for each microscopic field, which typically had 20-60 neurons. Numbers of TdTmediated X-dUTP nick end labeling (TUNEL)-positive neurons relative to GFP-positive neurons were counted from ten randomly selected 406 microscopic fields. These counting analyses were performed by an investigator blind to the experimental conditions. The percentages of apoptotic or viable neurons in each experimental group relative to those in groups cotransfected with a control expression vector and pcDNA3.1-GFP (15:1) were calculated [57].

Flow cytometry assays
SH-SY5Y cells were cultured to 70-80% confluence and then treated for 16 hours with 3 or 6 mM tunicamycin or 400 nM thapsigargin (positive control). Cells were then washed twice with PBS, treated with 20 mg/ml RNase at 37uC for 30 min and stained with propidium iodide (50 mg/ml propidium iodide, 0.05% Triton X-100, 0.05 NaN 3 in PBS) for 30 min at 4uC. Data were collected using a FACScan Flow Cytometer (BD Bioscience, San Jose, CA) and analyzed with WinMDI (Joe Trotter, Scripps Research Institute) and CellQuest (Becton Dickinson, Franklin Lakes, NJ) software. A small shift in the fluorescence spectrum was observed for tunicamycin treated LRRK2 KD SH-SY5Y cells, presumably due to dead cells.

Western blots
For Western blotting of nematode samples, animals were washed three times with M9 buffer [77] and lysed in HB buffer containing a protease inhibitor and phosphatase inhibitor mixture (Roche) followed by brief sonication [40]. For human cells, cells were treated in 10 cm plates and then collected as described for cell survival assays. Cells then were washed with PBS, harvested and lysed in ratio immunoprecipitation assay (RIPA) buffer containing a protease inhibitor and phosphatase inhibitor mixture. Insoluble cell debris was removed by centrifugation (13,000 rpm, 10 min) with a benchtop centrifuge at 4uC and the resulting supernatants were used for Western blotting. The antibodies used were: anti-nematode p38 (self-developed by N.H. and K.M.) [40]; anti-p38, phospho-specific anti-P-p38, anti-GRP78, anti-eIF2a, phospho-specific anti-P-eIF2a, and anti-MKK6 (Cell Signaling, Danvers, MA); and anti-LRRK2 (Sigma). Antibodies against P-p38 from Cell Signaling were demonstrated previously to specifically recognize nematode P-p38, but not un-phosphorylated p38 [40]. Quantification of Western blot results were obtained using National Instruments Vision Assistant 8.5.

Statistical analysis
Statistical significance was analyzed using Statistica software (StatSoft, Tulsa, OK). T-tests, ANOVA with Bonferroni corrections or Dunnet's post-hoc analyses were used for their appropriate applications as indicated in the Figure legends.  Figure S8 LRRK2 qRT-PCR confirmed the expression of WT LRRK2 and mutant forms K1906M and G2019S in wild type and lrk-1 mutant backgrounds. Wild type nematode lines and lrk-1 mutant nematode lines with panneuronal expression of WT LRRK2, K1906M or G2019S mutant LRRK2 (driven by the P H20 promoter) as well as DAergic neuronspecific expression of DsRed and command interneuron-specific expression of yellow fluorescent protein (YFP, driven by the nmr-1 promoter) were subjected to total RNA extraction, and qRT-PCR analysis. The relative quantity (RQ) value of WT LRRK2, K1906M or G2019S mutant LRRK2 vs. actin-1 were calculated. The error bar was based on the RQ Min/Max confidence level that represents the standard error of the mean expression level (RQ value). (TIF)