The C-terminal fragment of LRRK2 with the G2019S substitution increases the neurotoxicity of mutant A53T α-synuclein in dopaminergic neurons in vivo

Background: Alpha-synuclein (α-syn) and leucine-rich repeat kinase 2 (LRRK2) likely play crucial roles both in sporadic and familial forms of Parkinson’s disease (PD). The most prevalent mutation in LRRK2 is the G2019S substitution, which induces neurotoxicity through increased kinase activity. There is likely an interplay between LRRK2 and α-syn involved in the neurodegeneration of dopaminergic (DA) neurons in the substantia nigra (SNpc) in PD. However, the mechanisms underlying this interplay are ill-dened. Here, we investigated whether LRRK2 G2019S can increase the neurotoxicity induced by a mutant form of α-syn (A53T mutation) in DA neurons in vivo . Methods: We used a co-transduction approach with adeno-associated virus (AAV), AAV2/6 vectors encoding human α-syn A53T and the C-terminal portion of LRRK2 (ΔLRRK2), which contains the kinase domain, with either the G2019S mutation (ΔLRRK2 G2019S ) alone or the D1994A mutation (ΔLRRK2 G2019S/D1994A ), which inactivates the kinase activity of LRRK2. The AAVs were co-injected into the rat SNpc and histological evaluation was performed at 6- and 15-weeks post-injection (PI). Results: The majority of SNpc neurons co-expressed ΔLRRK2 and human α-syn A53T after transduction. ΔLRRK2 G2019S alone produced no cell loss at 15-weeks PI. Injection of AAV-α-syn A53T alone or mixed with a control AAV coding for GFP produced a signicant loss of DA neurons. Co-injection of AAV-α-syn A53T with AAV-ΔLRRK2 G2019S instead of GFP slightly exacerbated that neuronal loss We also studied the inactive form, ΔLRRK2 G2019S/D1994A at 6 weeks PI. Injection of AAV-ΔLRRK2 rat brain sections revealed for IBA1 immunoreactivity in the SNpc at low (upper images) and high (lower images) magnication in the different groups. antibody and the EM48 antibody. A, Camera Lucida representation of the rostro-caudal extension of striatal lesions produced by the mutant Htt fragment as seen using NeuN IHC. Grey spots in the striatum represent area with loss of NeuN staining (lesions). B, Histograms of the volumes of the striatal lesions for the four groups. C, Photomicrographs of the mouse striatum showing that lesions superimposed with HA tag-positive areas in the 3 ΔLRRK2 groups. D and E shows the quantication of mutant Htt inclusions (ubiquitin positive volume) and aggregates (number of EM48-positive aggregates). Note the absence of major effects of ΔLRRK2 constructs on the size of mutant Htt-induced lesions, and a signicant increase in the number of EM48 aggregates in the ΔLRRK2G2019S group. Results are expressed as the mean ± SEM. N = 10-12/group. One-way ANOVA and post hoc Fisher's PLSD test. *, p<0.05, Scale bars: A, 1 mm; C, 500 µm.


Background
Parkinson's disease (PD) is a neurodegenerative disorder affecting approximately seven million people worldwide. Early in the course of the disease, the most obvious symptoms are movement-related, including shaking (resting tremor), rigidity, and slowness of movement [1,2]. The neuropathological hallmarks of PD are characterized by the progressive loss of dopaminergic (DA) neurons in the substantia nigra pars compacta (SNpc) and the presence of neuronal aggregates (Lewy bodies) and dystrophic Lewy neurites containing the protein α-synuclein (a-syn) [3]. There is currently no treatment to delay this neurodegeneration and the cause of a-syn aggregation and the preferential death of DA neurons is unknown. PD is mainly a sporadic neurodegenerative disorder but approximately 10% of the cases are of genetic origin and several genes have been identi ed as causative factors [4].
Duplication, triplication, and rare mutations (A53T, A30P, E46K, H50Q, G51D, A53E) in the SNCA gene encoding the a-syn protein have been found in families with dominantly-inherited PD and are associated with early-onset forms, with an ampli cation of a-syn aggregation [4][5][6][7]. The A53T [8], A30P [9] and E46K [10] substitutions have been the most studied so far. Compelling evidence shows that a-syn takes center stage in PD and plays a key role via various aggregated forms, including abnormally phosphorylated aggregates that produce multiple cellular alterations, eventually leading to the death of DA neurons [11].
Mutations in leucine-rich repeat kinase 2 (LRRK2) are the most common genetic cause of both familial and sporadic PD [12,13]. There are also variants in the LRRK2 locus that are considered to be risk factors for developing PD [14,15]. The most prevalent mutation in LRRK2 is the G2019S substitution, accounting for 5 to 6% of familial PD and 1 to 2% of de novo genetic PD cases [16,17]. The cases of patients harboring the G2019S and other mutations are clinically indistinguishable from idiopathic PD cases, including the presence of Lewy bodies (LBs) in most cases [18,19]. Although G2019S patients show clinical manifestations similar to those of sporadic patients [20], several studies have shown subtle differences [21,22]. Some have reported the presence of LBs in symptomatic LRRK2 mutation carriers and LRRK2 can be found in LBs [23], although this is still a subject of debate, as other neuropathological studies have instead reported the absence of detectable LBs in a sub-population of PD patients with LRRK2 mutations [24]. In general, although LRRK2 variants or mutations are considered to be risk factors for developing PD, the onset of symptoms in LRRK2 carriers has been found to be similar to that of idiopathic PD cases [25,26]. Furthermore, the underlying mechanisms of LRRK2 neurotoxicity are still unknown. However, it is now generally accepted that the G2019S mutation increases LRRK2 kinase activity (both autophosphorylation and phosphorylation of exogenous kinase substrates) and that neurotoxicity originates from such increased activity [27,28].
The central role of a-syn in PD pathogenesis has led to the hypothesis of a functional, and possibly physical, interaction between LRRK2 and a-syn (for a review [20,29]). Indeed, LRRK2 toxicity may require the presence of a-syn and, conversely, the presence of variant/mutant LRRK2 could increase the risk and/or impact of a-synucleopathy in PD. The level of kinase activity of LRRK2 could thus be a modi er of a-syn toxicity. If this is true, the therapeutic implication would be extremely important: the regulation of LRRK2 kinase activity could be theoretically bene cial in slowing disease progression not only in individuals harboring LRRK2 mutations, but also in idiopathic PD. Recent experiments in transgenic mouse models of LRRK2 and a-syn support these hypotheses. The results of experiments in genetic models of mutant or wildtype LRRK2, in particular the effect of the pharmacological blockade of the kinase activity of LRRK2 G2019S in these models, suggest that LRRK2 may increase a-syn toxicity [30][31][32].
To date, it is not known how LRRK2 (especially the G2019S mutation) exerts a protoxic effect on a-syn toxicity. In particular, the exact role of the kinase domain has not been completely demonstrated but pharmacological intervention suggests that the kinase activity of LRRK2 likely contributes to the synergy (for a review [20]). In addition, it is not known whether the potentiation of a-syn toxicity by the presence of LRRK2 G2019S is related solely to a cell-autonomous mechanism, or if LRRK2-expressing cells that surround DA neurons, especially microglial cells, astrocytes, and cells of the immune system likely play a role [33][34][35] Here, we address these questions in a relevant cellular context by studying the effect of the C-terminal domain of human LRRK2 harboring the G2019S mutation (DLRRK2 G2091S ) or its inactive form, DLRRK2 DK (mutations G2019S plus D1994A, called DK), on the neurotoxicity of human a-syn with the A53T mutation (α-syn A53T ). We performed experiments using adeno-associated viruses (AAVs) that lead to the overexpression of the various forms of DLRRK2 and human a-syn A53T alone or in combination in DA neurons of the SNpc in adult rats. Quantitative histological evaluation showed that although DLRRK2 G2091S alone induced no loss of DA neurons, it could signi cantly increase a-syn A53T -induced neurotoxicity probably through a mechanism involving the catalytic activity of the kinase domain.

Viral construction and production
Adeno-associated viral vectors (AAVs). AAV6 viral particles were obtained by encapsidation of AAV2 recombinant genomes into serotype 6 AAV capsids as described previously [36]. Brie y, viral particles were produced by co-transfection of HEK-293T cells with (1) an adenovirus helper plasmid (pXX6-80), (2) an AAV packaging plasmid carrying the rep2 and cap6 genes, and (3) a plasmid encoding a recombinant AAV2 genome containing the transgene expression cassette. Seventy-two hours following transfection, viral particles were puri ed and concentrated from cell lysates and supernatants by ultracentrifugation on an iodixaniol density gradient followed by dialysis against PBSMK (0.5 mM MgCl2 and 1.25 mM KCl in PBS). The concentration of vector stocks was estimated by real-time-PCR following the method described by Aurnhammer et al. [37] and expressed as viral genomes per ml of concentrated stocks (Vg/ml). AAVs coding for human ∆LRRK2 (WT, G2019S, and G2019S plus D1994A mutation, i.e. "kinase dead"), αsyn A53T , and GFP under the PGK1 (mouse phosphoglycerate kinase) promoter were produced.
Lentivirus. DNA sequences encoding GFP and the C-terminal part of human LRRK2 (kinase,K; ROC-CORkinase, RCK; RCK plus the WD40 domain, called hereafter "DLRRK2") were synthesised and inserted into the self-inactivated vector (SIN) backbone containing the WPRE element (W) and the murine PGK promoter. We generated lentivirus vectors LV-GFP, and LV-DLRRK2 coding for the WT form or G2019S forms of the fragments. Viral particles were produced as described elsewhere [38]. All the SIN vectors were pseudotyped with VSV glycoprotein G. Brie y, the viral particles were produced in HEK-293T cells by a four plasmid transient transfection system [39]. The supernatant was collected 48 hours later and ltered. High-titre stocks were obtained by ultracentrifugation. The pellet was re-suspended in 1% BSA in PBS, frozen and stored at −80°C. Particle content of the viral batches was determined by ELISA for the p24 antigen (Gentaur, Paris, France). LV-DLRRK2 vectors were used at a concentration of 100 ng/μl p24. LV-Htt171-82Q was used at a concentration of 150 ng/µl of p24. LVs coding for LV-DLRRK2 forms (WT, G2019S or the dead kinase G2019S/D1994A were used at a concentration of 100 ng/µl of p24. For intracerebral infections, animal were anesthetized (100 mg/kg ketamine and 10 mg/kg xylazine). Local analgesia included subcutaneous lidocaine (5 mg/kg). A total volume of 2 µl of LV or AAV suspension was injected into the mouse striatum, as described [40], at the following stereotaxic coordinates: 1.0 mm anterior and 2.0 mm lateral to bregma, at a depth of 2.7 mm from the dura, with the tooth bar set at 0.0 mm. The mice were then left for one to two hours in a heated (30°C) ventilated box, until they had woken up and recovered fully from anesthesia. Post-surgery analgesia included acetaminophen (Doliprane) in drinking water for 48 h (1.6 mg/ml).

Tissue processing
For all procedures, rats were rst deeply anesthetized by iso urane inhalation, followed by the intraperitoneal injection of a lethal dose of sodium pentobarbital.
Rats were transcardially perfused with 300 ml 4% paraformaldehyde (4% PFA) in phosphate buffer saline (PBS -0.1 M phosphate buffer, 9 g/L NaCl) at a rate of 30 ml/min. After perfusion, the brain of each rat was quickly removed and immersed in ice-cold 4% PFA/PBS for at least 24 h, before transfer to 15% sucrose in PBS for 24 h and then 30% sucrose in PBS the next day, for cryoprotection. The brains were then cut into 40-μm sections on a freezing microtome (SM2400, Leica, Germany). Serial sections of the striatum and midbrain were stored in antifreeze solution (30% glycerol/30% ethylene glycol in PBS) and stored at -20°C until use.
Mice were deeply anesthetized by the intraperitoneal injection of sodium pentobarbital solution (50 µg per gram body weight). They were then transcardially perfused with 100 ml 4% PFA in PBS at a ow rate of 8 ml/min. The brains of the animals were removed, post-xed overnight in the same solution, then cryoprotected by immersion in 30% sucrose in PBS for 36 hours. Free-oating 30-µm serial coronal sections from throughout the striatum were collected with a freezing sliding microtome. Brain slices were placed in a storage solution (30% glycerol, 30% ethylene glycol in PBS) and stored at -20°C before use.

Immunohistological analysis and quanti cation
Immunohistochemistry Sections were removed from the antifreeze solution and washed in PBS. Endogenous peroxidase activity was quenched by transferring them to 1% H 2 O 2 and incubation for 30 min at room temperature (RT) and washing them three times with PBS for 10 min each. The sections were then blocked by incubation with 4.5% normal goat serum for 30 min in PBS-T (0.2% Triton X-100 in PBS) and then incubated overnight with primary antibody in 3% normal goat serum in PBS-T at 4°C with gentle shaking.
For histological evaluation using rat brain sections, the following primary antibodies were used for the present study: anti-tyrosine hydroxylase (TH) antibody: MAB318 clone LNC1, Merk-Millipore, 1:3000; antihemagglutinin tag (HA), Covance clone 11, 1:1000; anti-human α-synuclein, syn 211, 1:1000; antiphospho-α-synS129, ab51253, Abcam, 1:5000]. The next day, the sections were removed from the primary antibody solution, washed three times, and incubated for 1 h at RT with the appropriate biotinylated The rat and mouse sections were then incubated with DAB for 30 s to 1 min and after dehydration mounted on slides in Eukitt mounting medium.

Cell counting
Optical fractionator sampling was carried out on a Zeiss AxioPlan microscope. Midbrain dopaminergic neurons were outlined on the basis of TH immunolabelling with reference to a coronal atlas of the rat brain (Paxinos and Watson, 6 th edition). TH-positive cells were counted by unbiased stereology in the entire SNpc and the number of positive neurons per section was calculated using the Mercator Software (Explora Nova, France). We placed 100 × 100 μm grids in a systematically random manner, 80 × 80 μm apart, with a 3-µm offset from the surface of the section. Quanti cation was performed on 12 serial sections spaced by 200 µm, corresponding to the entire SNpc.
The phosphorylation of α-syn on S129 (p-synS129) was evaluated by counting the number of p-synS129positive neurons in the SNpc using stereology methods. The SNpc was delimited by Nissl staining and the grids (250 x 250 µm) placed, with a space of 100 x 100 µm. Quanti cation was performed on six serial sections spaced by 400 µm, corresponding to the entire SNpc. In the striatum, a threshold was applied to select only the p-synS129-positive neurons by immunostaining and quanti cation performed on three slices, corresponding to the beginning, middle, and end of the striatum.

Immuno uorescence
The procedure used was similar to that for immunohistochemistry, but without the incubation in 1% H 2 O 2 .
The primary antibodies used for the immuno uorescence procedure were the same as previously described (IBA1, Wako, 1:1000). Sections were rst incubated with the primary antibody overnight at 4°C.
The next day, they were incubated with a uorescent secondary antibody (Alexa Fluor 594-labeled goat anti-rabbit IgG or Alexa Fluor 488-labeled goat anti-rabbit IgG (1:1000, Life Technologies)) for 1h at RT. Sections were then washed and incubated overnight at 4°C with another primary antibody. Finally, they were incubated with a second uorescent secondary antibody (Alexa Fluor 488-labeled goat anti-mouse IgG or 594-labeled goat anti-mouse IgG (1:1000, Life Technologies)) for 1h at RT. The sections were stained with DAPI, washed, and mounted in a uorescence mounting medium. Images were acquired with a laser confocal microscope (SP8, Leica, Germany) or an epi uorescence microscope (DM6000, Leica, Germany).

Thio avin-S staining
A double-staining protocol was used to verify that accumulation of positive p-synS19 inside cells could colocalize with aggregated form of a-syn. The immunostaining procedure for p-synS19 and DAPI staining was performed on oating sections before the Thio avin-S (Thio-S) staining. Floating sections were washed in PBS and mounted on Superfrost Plus slides. Slides were place in holders and dive into 70% EtOH and 80% EtOH, for 1 min each. Then, Slides were incubated in Thio-S diluted at 1% in distilled water for 7 min. The Thio-S solution must be protected from light, ltrated before use, and should be stored at 4°C. Then, slides were washed in 80% EtOH, 70% EtOH and distilled water for 1 min each before being coverslipped with the uorescence-mounting medium.

Colocalization
The percentage of co-localization between ΔLRRK2 and α-syn was determined by counting the number of cells co-expressing both ΔLRRK2 and α-syn proteins divided by the number of cells expressing α-syn alone. Images were acquired with a laser confocal microscope (SP8, Leica, Germany). On the same acquisitions, the levels of ΔLRRK2 and α-syn proteins were evaluated on three coronal sections in the SNpc. Twenty cells co-expressing both ΔLRRK2 and α-syn proteins were delineated per animal using image J software and the mean uorescence intensity in the red and in green channels (corresponding to ΔLRRK2 and α-syn proteins, respectively) was measured in each cell.

Fluorescence intensity measurement
Striatal dopaminergic innervation at 15 weeks was quanti ed by measuring the uorescence intensity of TH-immunoreactive terminals on three coronal striatal sections. The sections were observed by epi uorescence microscopy at a magni cation of 63X and the uorescence intensity determined using MorphoStrider software (Explora Nova, France).

Microglia area measurement
The area occupied by microglia was evaluated by confocal microscopy at a magni cation of 20X in the dorso-medial part of the striatum and in the SN pars reticulata. A threshold was applied and the area of 20 microglia cells measured per acquisition. Three acquisitions per animal were used.
Image analysis of lesion area.
Observation of sections and calculation of the surface of lesion were performed using a Leica DM6000 equipped with a motorized stage and an automated image acquisition and analysis system (Mercator software, Explora Nova, La Rochelle, France). The area covered by striatal lesions resulting from LV-Htt171-82Q infection was delineated manually using a 10x objective by identifying the border of the lesion on each coronal brain section and the corresponding area was calculated. The volume of the striatal lesion was determined by the Cavalieri method and the number of EM48-positive inclusions was determined as previously described with an inter-section distance of 210 µm (i.e. we used one in every seven sections) [40][41][42]. Automatic detection of EM48-positive objects was performed using the Mercator software to count the number of aggregates on the entire cross-sectional area of the striatum and to measure the size of all the detected EM48-positive objects. Objects with an apparent crosssectional area exceeding 5 µm 2 were reliably detected with this method.

Statistical analysis
Normality of data distribution was tested using the Shapiro-Wilk test and homogeneity of variance was tested with Levene's test using a commercially available software (Statistica, 13.0; Statsoft Inc., Tulsa, Oklahoma, USA). When normality and homogeneity of variance were met, unpaired Student's t-test was used for pairwise comparisons between groups. For comparisons of more than two groups, one-way ANOVA for multiple comparisons was carried out, with Fisher's post hoc PLSD test. In the cases where assumption of normality and/or homogeneity of variance were not met, non-parametric tests where applied: Mann-Whitney and Kruskall-Wallis for comparison of 2 and more groups, respectively. The annotations used to indicate the level of signi cance are as follows: *p < 0.05, **p < 0.01, ***p < 0.001.

Results
Determination of the experimental conditions to detect a potential synergy between AAV-a-syn A53T and AAV-DLRRK2 G2019S toxicity We investigated whether human LRRK2 can increase the toxicity of human a-syn in DA neurons through cell-autonomous mechanisms. We used serotype 6 AAV capsids, which allow preferential expression in neuronal cells and leads to a high percentage of cells transduced in the injected structure, without excessive diffusion into the surrounding tissue, such as for example serotype 9 [43,44]. In a previous study [45], we showed that the C-terminal portion of human LRRK2 G2019S (DLRRK2 G2019S , aa 1330-2527) retains, at least in part, the biochemical properties of full-length LRRK2 G2019S , including higher kinase activity than the wildtype fragment. In addition, we found that overexpression of the C-terminal portion of human DLRRK2 G2019S in the adult rat SNpc, using AAVs, produced partial (~30%) but signi cant loss of DA neurons at 25 weeks post-transduction, whereas overexpression of the wildtype form of LRRK2 (DLRRK2 WT ) was not toxic [45]. Here, we used a similar approach using a slightly larger fragment (aa 1283-2527) (Fig. 1A).
We injected 4 µl of AAVs solution in all cases. A nal amount of 2.5x10 10 Vg per site and per vector was used. Each AAV was injected unilaterally into the SNpc (2.5 10 10 Vg). In addition to the three experimental groups, a control group received injections of vehicle (PBS/pluronic acid). The integrity of the nigrostriatal pathway was assessed using unbiased stereology to count the number of DA neurons displaying tyrosine hydroxylase (TH) staining in the injected part of the SNpc (Fig. 1). Observation at low-magni cation revealed no major loss of TH-positive cells in any of the groups injected with AAVs encoding the LRRK2 fragments (Fig. 1B). The total number of TH-positive cells in the SNpc did not differ signi cantly between the control group (PBS) and AAV-DLRRK2 WT , AAV-DLRRK2 G2019S , or the AAV coding the dead kinase form ∆LRRK2 G2019S/D1994A (thereafter called ∆LRRK2 DK ) (Fig. 1C). Thus, these results suggest that the ΔLRRK2 fragments alone did not trigger signi cant neurodegeneration of DA neurons at 15 weeks PI.
We wanted to investigate whether AAVs encoding the different ΔLRRK2 constructs could increase the toxicity of AAV-a-syn WT AAV-a-syn A53T . Therefore, we decided on an injection protocol that would lead to mild degeneration, such that a potential "pro-toxic" effect of LRRK2 constructs could be easily detected.
We conducted pilot experiments to determine the appropriate dose (titers) of AAV-a-syn WT and AAV-asyn A53T alone that would lead to progressive and partial loss of DA neurons. Quanti cation of TH-positive cells in the SNpc showed no signi cant loss of DA neurons with AAV-a-syn WT at 15 weeks (not shown). In contrast, we found a moderate loss of DA neurons (~30%) at 12 and 15 weeks after transduction with AAV-a-syn A53T (2.5x10 10 Vg) (Fig. 1D,E). After transduction, DA neurons often displayed accumulation of a-syn phosphorylated at its serine 129 (p-synS129). The cells positive for p-synS19 were also positive for ThioS suggesting that these accumulations were aggregates (Fig. 1F).
Thus, a co-injection protocol with AAV-α-syn A53T and AAV-∆LRRK2 G2019S and the evaluation of DA cell loss at 15 weeks PI appeared to be suitable for the detection of the potential synergy of toxicity between the two pathological transgenes.
Effects of co-transduction with AAV-α-syn A53T and AAV-∆LRRK2 G2019S We next investigated whether the presence of the DLRRK2 fragments G2019S, or G2019S/D1994A -DK) in DA neurons could modify the toxicity of human α-syn A53T using this co-transduction paradigm (2.5x10 10 Vg for each vector).
We rst studied the neurotoxic effects produced by AAV-α-syn A53T in the presence or absence of AAV-∆LRRK2 G2019S at 15 weeks PI. We assessed the co-localization of human a-syn A53T and LRRK2 fragments in the SNpc after co-transduction, as we wanted to investigate the combined effects of asyn A53T and the various LRRK2 fragments in DA neurons. Analysis by confocal microscopy showed that human α-syn expression in the SNpc was high in TH-positive neurons ( Fig. 2A). On average, 70% of neurons co-expressed both human a-syn A53T and the LRRK2 fragments (Fig. 2B).
We also counted the number of SNpc cells showing p-synS129 immunoreactivity, a marker of α-syn aggregation, in the different groups (Fig. 3C). The number of p-synS129-positive cells was signi cantly lower in the group co-infected with AAV-α-syn A53T and AAV-DLRRK2 G2019S than that of the groups infected with AAV-α-syn A53T alone or in combination with AAV-GFP (Fig. 3D). We also normalized the number of p-synS129-positive neurons to the number of TH-positive cells which survived. In this case, there was no difference between group, suggesting that the overexpression of AAV-α-syn A53T and AAV-DLRRK2 G2019S does not markedly modify the presence of p-synS129-in surviving cell soma. We also evaluated p-synS129 immunoreactivity in the striatum, which receives major inputs from the SNpc. Small p-synS129 immuno-positive objects with an elongated form or with a pearl necklace-like shape, reminiscent of neurite-like aggregates were seen in the striatum (Fig. 3E). Consistent with the results obtained in the SNpc, we found signi cantly lower levels of p-synS129 in the striatum of rats co-infected with AAV-α-syn A53T and AAV-DLRRK2 G2019S than in those infected with AAV-α-syn A53T / GFP (Fig. 3E-F).
Then, we evaluated the impact of SNpc cell loss on the level of dopaminergic terminals in the dorsomedial striatum using TH-immuno uorescence in both the α-syn A53T /GFP and α-syn A53T /∆LRRK2 G2019S groups. These measurements were performed in the dorsal striatum (Fig. 4A). TH immunoreactivity in the striatum in both α-syn A53T /GFP and α-syn A53T /∆LRRK2 G2019S groups was 15% lower than in the control group (PBS). This small a-syn A53T -induced loss of TH-positive bers was similar in the GFP and ∆LRRK2 G2019S groups (Fig. 4B, C).
Differential effects of AAV-∆LRRK2 G2019S and AAV-∆LRRK2 G2019S/D1994A on AAV-α-syn A53T toxicity We next investigated whether the effect of AAV-ΔLRRK2 G2019S on AAV-α-syn A53T -induced toxicity was dependent on the kinase activity of the LRRK2 construct. We thus compared the effect of ΔLRRK2 G2019S with that of the dead kinase form AAV-∆LRRK2 G2019S/D1994A (thereafter called ∆LRRK2 DK) . We examined an earlier time point PI (6 weeks) for these experiments. We reasoned that, although the kinase activity of ∆LRRK2 DK is "dead", it may lead to cellular disturbances in long-term experiments because of its potential dominant-negative effect on endogenous rat LRRK2.
We then assessed the loss of DA neurons produced by AAV-a-syn A53T when co-injected with either AAV-ΔLRRK2 G2019S or ∆LRRK2 DK . The loss of DA neurons induced by human α-syn A53T was signi cantly lower in the presence of ∆LRRK2 DK than that in the presence of ΔLRRK2 G2019S (Fig. 7A-B). The number of cells with p-synS129 immunoreactivity was similar in the ∆LRRK2 DK and ΔLRRK2 G2019S groups as shown by quanti cation (Fig. 7C-D). In the striatum, a few p-synS129 immunoreactive bers (arrow heads), reminiscent of dopaminergic bers, were seen in both groups expressing a-syn A53T with no obvious apparent difference in density or size (Fig. 7E).
Finally, we carried out a preliminary characterization of the status of neuroin ammation at this early time point (6 weeks PI), by immunohistochemistry using a validated marker (Iba1), which is highly expressed in activated microglial cells. Indeed, there is a role of neuroin ammation in neurodegeneration in αsyn A53T rodent models. As expected, microglial cells in rats overexpressing human a-syn A53T appeared more reactive as compared to rats injected with vehicle (Fig. 8E, F). The quanti cation of immuno uorescence levels in the SN (Fig. 8A, B) and striatum (Fig. 8C, D) showed that human α-syn A53T signi cantly activated microglia. However, overexpression of ΔLRRK2 G2019S and ΔLRRK2 DK did not have any impact on the microglial activation induced by the mutant human α-syn.
Finally, we investigated whether the "pro-toxic" effect of ΔLRRK2 G2019S on a-syn A53T could also be detected for other aggregating proteins. Using a different approach with lentiviral vectors in mice, we tested whether the different forms of ΔLRRK2 could modify the neurotoxicity produced by the N-terminal domain of human huntingtin (Htt) with a pathological expansion of its poly-glutamine (Q) region (Htt-N171-82Q) [46]. Lentiviral vectors were injected into the striatum of wild-type mice to produce local cell loss within the six weeks following transduction, as previously described [40,47]. The striatal lesions were characterized by the loss of neuronal markers DARPP32 and COX (not shown) as well as NeuN (Fig. 9A-B). Localization of the lesions in the striatum coincided with that of the expression of LRRK2 fragments detected using the HA-tag (Fig. 9C). Quantitative analysis of these histological markers showed that none of the ΔLRRK2 forms signi cantly modi ed the volume of the striatal lesions produced by mutant Htt (Fig. 9B). In addition, overexpression of the mutant Htt-fragment led to the accumulation of inclusions containing ubiquitin (mostly nuclear) (data not shown). Quanti cation of the presence of ubiquitin positive inclusions revealed no difference between groups ( Fig.9 D). We also selectively detected mutant Htt aggregates using the EM48 antibody which speci cally recognizes the aggregated form of the Nterminal domain of mutant Htt [41,48,49]. ΔLRRK2 G2019S signi cantly increased by 44% the number of EM48-positive aggregates when compared to the control group (LV-LacZ), an effect not seen with the WT or -∆LRRK2 DK (Fig. 9E).
Our results show that the synergistic effect of ΔLRRK2 G2019S on the toxicity of human a-syn A53T towards DA neurons depends on its kinase domain. Importantly, ΔLRRK2 G2019S overexpression did not markedly change the neurotoxic effects produced by a different aggregating protein, mutant Htt. Although the comparison between the rat model of SNpc degeneration and the mouse model of striatal lesion must be compared cautiously, our results suggest that the effect of ΔLRRK2 G2019S on a-syn A53T -induced neurotoxicity may be due to speci c molecular mechanisms rather than to a general increase in the vulnerability of the neurons.

Discussion
The mechanisms leading to the degeneration of DA neurons in LRRK2 mutation gene carriers with PD are unknown. It is generally accepted that the LRRK2 G2019S mutation leads to increased kinase activity which could then lead to cell death [27,[50][51][52].
In addition, a role for LRRK2 in a-syn toxicity has been suggested [30][31][32]. Indeed, neuropathological evaluation of the brains of PD patients with LRRK2 mutations shows in many cases the presence of bona de LBs and Lewy neurites [53]. However, the role of the kinase in the crosstalk between LRRK2 and asyn, especially how the kinase activity of LRRK2 G2019S modulates a-syn neurotoxicity in the SNpc is unknown. In addition, the respective roles of cell-autonomous and non-cell-autonomous mechanisms in this interaction are largely unknown.
Here, we used an AAV-based approach to target SNpc DA neurons and investigated how the C-terminal domain of LRRK2, harboring the G2019S mutation, with increased kinase activity, could modify the loss of DA neurons induced by the overexpression of α-syn A53T in the rat SNpc. Under our experimental conditions, AAV-DLRRK2 G2019S alone do not induce the loss of DA neurons whereas AAV-α-syn A53T alone can produce partial loss. The loss of DA neurons produced by co-expression of DLRRK2 G2019S and αsyn A53T was signi cantly higher than that measured in rats injected with AAV-α-syn A53T alone or coinjected with a control vector (AAV-GFP). Conversely, overexpression of the inactive "dead kinase" form DLRRK2 DK , at levels similar to those of DLRRK2 G2019S , did not alter the toxicity of a-syn A53T . Quantitative characterization of microglial reactivity induced by human a-syn A53T in SNpc and striatum did not show obvious changes attributable to DLRRK2. These novel ndings further support the hypothesis that the Cterminal domain of LRRK2 G2019S is su cient to augment the toxic effects of a-syn A53T through a cellautonomous mechanism involving the catalytic activity of its kinase domain.
The histological evaluation we performed after transduction of the SNpc with AAV-a-syn A53T and AAV-DLRRK2 G2019S shows that both transgenes are overexpressed in DA neurons. In both cases, approximately 70% of the SNpc was infected. After co-injection, co-localization of both transgenes in neurons of the SNpc was found in a large proportion (~77%) of SNpc neurons. Neuropathological evaluation after transduction with AAV-a-syn A53T (15 weeks PI) showed the partial loss of DA neurons, based on the detection of TH-positive cells. This likely re ects neuronal loss, as suggested in our previous work [45].
The relevance of overexpressing the C-terminal domain of LRRK2 versus the full-length protein is debatable and the mechanisms underlying the neurotoxic effect of DLRRK2 G2019S in our models are unknown. Indeed, our DLRRK2 G2019S construct lacks N-terminal domains that are known to play crucial roles in LRRK2 function. We previously showed that overexpression of the DLRRK2 G2019S fragment using AAVs triggers neurodegeneration of DA neurons six months PI, whereas the DLRRK2 WT fragment, expressed at similar high levels, was devoid of obvious neurotoxicity [45]. In this work, we suggested that death of DA neurons produced by DLRRK2 G2019S is likely independent of the interaction with RAB10, since we found that the DLRRK2 fragment was found unable to interact with RAB10, in contrary to full-length LRRK2 fragment [45]. Thus, other signaling pathways have to be considered. It is conceivable that the overexpression of DLRRK2 G2019S leads to abnormally high phosphorylation of substrates when compared to DLRRK2 WT . Indeed DLRRK2 G2019S kinase activity is higher than that of DLRRK2 WT [45] , a phenomenon that is also observed for full-length LRRK2 G2019S [54][55][56][57]. Alternatively, the "protoxic" effect of DLRRK2 G2019S upon a-syn A53T could also result from molecular mechanisms unrelated to the enzymatic activity of the catalytic domains. Changes in protein-protein interactions and/or a modi cation of the conformation of LRRK2 fragments induced by the G2019S substitution may also play a role. In regard to this hypothesis, we now know that LRRK2 interacts with microtubules [58,59] and recent highresolution cryo-EM studies have shown the enzymatic domain of LRRK2 (ROC-COR-Kinase) is su cient for the interaction of LRRK2 with microtubules and their regulation [60,61]. The orientation of the kinase domain from microtubules is different between wild-type LRRK2 and LRRK2 with pathological mutations [61] and therefore in the present study, it is conceivable that the pro-toxic effects of DLRRK2 G2019S are linked to microtubule-related perturbation. Further in vivo studies are required to fully address this hypothesis.
We investigated whether DLRRK2 G2019S effect on a-syn A53T was speci c of these two pathological proteins, or only resulted from a non-speci c "protoxic" effect of DLRRK2 G2019S  neurotoxicity might be speci c, although testing the effect of DLRRK2 G2019S on the toxicity of other pathological proteins is necessary to fully support this hypothesis.
The present experimental paradigm using AAVs allowed us to address the question of the potential cellautonomous exacerbation of a-syn A53T toxicity by the kinase activity of LRRK2 directly in the SNpc and only in neurons. In contrast, other viral vector platforms that could potentially host the full-length LRRK2 ORF [i.e. vectors derived from Herpes Simplex Virus (HSV) or adenovirus] also transduce other cell types in the striatum [62,63]. Here, we directly investigated whether there is a functional interaction between AAV-α-syn A53T and-DLRRK2 G2019S in DA neurons. Our results show the existence of such a "functional" interaction, as overexpression of DLRRK2 G2019S signi cantly enhanced the neurotoxic effects of a-syn A53T in rat SNpc. Lin et al. showed that the overexpression of LRRK2 (wildtype or with the G2019S mutation) in forebrain neurons (striatum and cerebral cortex) increased the toxicity of a-syn A53T in transgenic animals [64,65]. In these double-transgenic mice, the authors found signi cant degeneration of the striatum and cortex and enhanced accumulation of a-syn aggregates. This proved the existence of a functional crosstalk between a-syn and LRRK2 in neurons in vivo when the proteins are expressed at relatively high levels. Pathological transgenes were not expressed in the SNpc and DA degeneration was not evidenced in these models [66]. The CamKIIa promoter used to drive the expression of the tetracycline transactivator (tTA), which activates the TetO promoter of the LRRK2 and a-syn A53T transgenes in these mice, is likely not active in SNpc DA neurons, as endogenous expression of CamKIIa in neurons of the SNpc is lower than that observed in forebrain neurons ( [66] and see also the Allen Brain Atlas, http://mouse.brain-map.org/experiment/show/79490122). In LRRK2 knockout rats, the toxicity induced by AAV coding for a-syn is lower than that in wildtype rats [31]. Daher et al. found no synergy between the transgenes following the crossbreeding of other transgenic models in which the promoters driving LRRK2 G2019S and α-syn A53T expression were different (Prion and CMV respectively) [67]. Indeed, data from the latter work indicate that the expression level of human LRRK2 transgene is low in the SNpc (see Figure 2 in [67]). Neurons that express a-syn A53T are apparently sparse in the SNpc compared to the known density of DA neurons in this structure (see Figure 5 in [67]). These observations and our results suggest that the crosstalk between LRRK2 and a-syn can occur if the two proteins are localized in the same neurons. Thus, our results show that synergy between LRRK2 and a-syn depends, at least in part, on cell-autonomous mechanisms.
In our experimental condition, the interplay between DLRRK2 G2019S and α-syn A53T is detected while the two proteins are expressed in high levels. We could not precisely compare the overexpression levels of the transgenes that are of human origin with the endogenous rat proteins. It can be only grossly estimated that expression of human DLRRK2 G2019S and α-syn A53T is likely 5-50 fold higher than those of the rat endogenous proteins. This estimation is based on the previous experiments where other mouse transgenes (DCLK3, Crym, abhd11os, Capucin) were overexpressed with lentiviral vectors or AAVs [40,47,48,68,69]. Thus, it cannot be ruled out that in the presence of physiological levels of expression of LRRK2 and a-syn, the cell-autonomous crosstalk between the two proteins might be of moderate importance in DA neurons.
In our AAV-based model, LRRK2 fragments are expressed only in neurons, which allowed us to investigate the cell-autonomous mechanisms of LRRK2 / a-syn interplay. The other comparable experimental approaches that investigated this interplay were carried out in in models where LRRK2 is expressed in all cells (See Daher and collaborators [31]). In the transgenic animal models and in patients LRRK2 is expressed in cells of different types (i.e. neurons, microglia, oligodendrocytes, astrocytes). In these conditions, non-cell autonomous mechanisms involving interaction of DA neurons with neighboring glial cells and immune cells may have also important roles. For example, it has been recently shown that the seeding of a-syn aggregates by the exposure of neurons to a-syn brils is higher in neurons expressing mutant LRRK2 [70]. More generally, LRRK2 mutation may change the potential propagation of aggregated a-syn species in the brain [71]. The level of LRRK2 activity in microglial cells may also regulate protoxic phenomena associated with α-syn-induced neuroin ammation [30,34,72]. LRRK2 plays a key role in the immune system [73]. A single nucleotide polymorphism (N2081D) in the region coding for the kinase domain of LRRK2 is a major risk factors for Crohn's disease, a form of in ammatory bowel disease [74].
The neuronal mechanisms underlying the synergy between LRRK2 and a-syn are ill-de ned. It is possible that LRRK2 -a-syn A53T interplay involves a direct physical interaction between the two proteins [75]though, indirect effects leading to a functional interplay is more often discussed. The present results indicate that DLRRK2 G2019S does not markedly change p-synS129 immunoreactivity at 6 and 15 weeks PI. This suggests that the "pro-toxicity" produced by the overexpressing the mutant LRRK2 fragment may not be related to a change in a-synA53T bioavailability/expression levels, although a more complete biochemical study is required to further support this hypothesis. In addition, the role of the kinase domain of LRRK2 is unclear. It is generally accepted that the higher kinase activity of LRRK2 G2019S , relative to that of wild-type LRRK2, leads to neurodegeneration through increased phosphorylation of substrates, possibly through multifactorial cellular changes, including the disruption of microtubule assembly, mitochondrial defects, and alterations in protein translation [76]. However, whether the increased kinase activity of LRRK2 mutations plays a key role is still a subject of debate. As already mentioned, various transgenic rodent models expressing LRRK2 G2019S have been developed and extensively characterized.
These models display no (or very limited) degeneration of DA cells in the SNpc. Various experimental approaches clearly demonstrate that the severity of the resulting toxicity is dependent on the level of expression of LRRK2 G2019S [77]. However, there is limited evidence obtained in vivo that shows a relationship between the higher kinase activity of LRRK2 G2019S and neurotoxicity to dopaminergic cells of the SNpc. The overexpression of LRRK2 G2019S (or DLRRK2 G2019S ) in the SNpc was found to induce the loss of DA neurons using HSV and adenovirus models injected into the striatum [27,78], as well as in our previous study with AAV-DLRRK2 G2019S injected into the SNpc [79].
Only a few studies have directly addressed the role of the kinase domain in the interaction between a-syn and LRRK2 toxicity. It was shown that neuroin ammation and neurodegeneration produced by the transduction of the SNpc with AAV-a-syn is signi cantly attenuated in LRRK2 KO rats relative to that in wildtype littermates. In these experiments, the role of the kinase activity was not assessed [30]. More recently, Daher et al. showed that the toxicity of AAV-α-synuclein in the SNpc was higher in transgenic LRRK2 G2019S than wildtype rats. Interestingly, treatment of both genotypes with the LRRK2 inhibitor PF-06447475 reduced the toxicity of a-syn [31]. This suggests that the exacerbation of a-syn toxicity by LRRK2 G2019S could result from elevated catalytic activity of the kinase. Since it has been observed that some LRRK2 inhibitors, including PF, can reduced cellular levels of the protein [80], it is possible that protection by PF-06447475 against the toxicity triggered by injection of AAV-a-syn in LRRK2 G2019S mice may result from the reduction of LRRK2 levels rather than actual inhibition of the catalytic activity of the kinase. Indeed, it has been suggested that the level of expression of the LRRK2 protein could play a determinant role in mutant LRRK2 rather than the kinase activity [77]. However, new generation of inhibitors with protective effects do not reduce LRRK2 levels (see for review ([81]). In our work, the inactive protein DLRRK2 DK did not alter the toxicity of AAV-α-syn A53T , whereas expression of DLRRK2 G2019S increased the toxicity of AAV-α-syn A53T towards DA neurons and we were able to verify that the difference between the toxicity of DLRRK2 G2019S and DLRRK2 DK was not related to a difference in protein levels as evaluated by confocal microscopy.

Conclusions
Our results show that the C-terminal domain of LRRK2 G2019S containing the ROC-COR, Kinase and WD40 domains is su cient to potentiate the toxicity of human a-syn A53T in DA neurons in vivo and suggest that this effect depends on the kinase domain. This cell-autonomous mechanism may act additively or synergistically with other non-cell-autonomous mechanisms, especially those involving neuroin ammation to trigger the death of DA neurons in PD.

Declarations
Authors consent All authors gave their consent to Emmanuel Brouillet for publication of the present results.
Ethics approval and consent to participate Manuscripts reporting studies involving human participants, human data or human tissue must: Not applicable

Consent to publication
Not applicable

Competing interest
The authors have no competing nancial interests to declare.

Availability of data and material
The different LRRK2 vectors maps and raw data of analyses can be obtained from the corresponding author on request. The authors would be glad to share on a collaboration basis the plasmids for producing the different vectors coding for the LRRK2 fragments described herein. Gwenaëlle Auregan contributed to the stereotaxic surgery in the rat substantia nigra and mouse striatum.
Martine Guillermier contributed to the entire process to obtain Ethical authorizations, performed the stereotaxic surgery in the rat substantia nigra and mouse striatum, and the optimization of anaesthesia for the different experiments.
Suéva Bernier, supervised anesthesia of rats and the stereotaxic surgery in rats.
Caroline Jan supervised the histological works (histochemistry and confocal).
Philippe Hantraye contributed to scienti c discussions on the design of the different experiments, and helped to write the manuscript.
Marie-Christine Chartier-Harlin discussed many conceptual and methodological aspects of the study and contributed to the writing of the manuscript. cell count in B shows signi cant loss compared to control rats of the same age (injected with PBS). Neurons in the SNc after infection were found to be immune-positive for the phosphorylated form of αsyn at serine 129 (P-synS129) and thio avin S (ThioS) uorescence suggesting aggregation of α-synS129. Results are expressed as the mean ± the standard error of the mean (SEM). One-way ANOVA and post hoc Fisher's PLSD test. *, p<0.01. Scale bar in C, 10 µm.  Evaluation of α-syn (in green) transduction in the SNpc after co-injection of AAV-α-synA53T with ΔLRRK2G2019S (ΔLRRK2GS) was determined using delineation of SNpc with TH staining (red). Scale bar: 500 µm. (B) Measurement of the number of neurons expressing both α-synA53T and ΔLRRK2GS from confocal images. The higher magni cation shows cytoplasm localization of ΔLRRK2G2019S.