LRRK2 causes centrosomal deficits via phosphorylated Rab10 and RILPL1 at centriolar subdistal appendages

The Parkinson’s disease-associated LRRK2 kinase phosphorylates multiple Rab GTPases including Rab8 and Rab10, which enhances their binding to RILPL1 and RILPL2. The nascent interaction between phospho-Rab10 and RILPL1 blocks ciliogenesis in vitro and in the intact brain, and interferes with the cohesion of duplicated centrosomes in dividing cells. We show here that various LRRK2 risk variants and all currently described regulators of the LRRK2 signaling pathway converge upon causing centrosomal cohesion deficits. The cohesion deficits do not require the presence of RILPL2 or of other LRRK2 kinase substrates including Rab12, Rab35 and Rab43. Rather, they depend on the RILPL1-mediated centrosomal accumulation of phosphorylated Rab10. RILPL1 localizes to the subdistal appendages of the mother centriole, followed by recruitment of the LRRK2-phosphorylated Rab protein to cause the centrosomal defects. These data reveal a common molecular pathway by which alterations in the LRRK2 kinase activity impact upon centrosome-related events.


Introduction
Autosomal-dominant mutations in the leucine rich repeat kinase 2 (LRRK2) gene cause familial Parkinson´s disease (PD), and coding variants in the same gene can act as risk factors for sporadic PD. Known pathogenic LRRK2 mutations produce a protein with increased kinase activity (1,2), raising the possibility that kinase inhibitors may be useful to treat LRRK2-related PD. Therefore, understanding the downstream effects of enhanced LRRK2 kinase activity is of interest to understand disease pathomechanism(s) as well as potential therapeutics.
LRRK2 phosphorylates these substrates in a conserved region of the switch 2 domain, which leads to impaired interactions with various effector and regulatory proteins (3). We have previously shown that this interferes with the physiological functions of these Rab proteins as key regulators of distinct membrane trafficking events (6,7). Importantly though, the phosphorylated Rab proteins also show enhanced binding to a set of novel effector proteins (4), raising the possibility that these nascent interactions may contribute to the pathobiology of LRRK2.
LRRK2-phosphorylated Rab8 and Rab10 bind with great preference to RILPL1 and RILPL2 (4), two poorly characterized proteins reported to regulate ciliary content (8). An important and direct consequence of the LRRK2-mediated phosphorylation of these Rab proteins is a decrease in primary cilia in various cell types in vitro as well as in the intact mouse brain (4,9,10). We previously showed that pathogenic LRRK2 not only causes ciliary deficits, but also deficits in the cohesion between duplicated centrosomes in a manner mediated by the phosphorylated Rab proteins and by RILPL1 (10,11). Such centrosomal cohesion deficits can also be observed in peripheral cells derived from LRRK2 PD patients as compared to healthy controls (12). Thus, ciliary deficits in brain, and centrosomal cohesion deficits in dividing cells are distinct cellular reflections of the same LRRK2-mediated phospho-Rab/RILPL1 interaction.
Recent studies have described upstream regulators of the LRRK2 kinase activity including Rab29 and vps35. Rab29 is a protein which modulates PD risk (13), and under certain conditions is able to stimulate the LRRK2 kinase activity (14)(15)(16). A point mutation in vps35, the cargo binding component of the retromer complex, causes autosomaldominant late-onset familial PD (17)(18)(19), and potently activates the LRRK2 kinase activity as assessed by Rab10 phosphorylation in cells and tissues (20). Conversely, the PPM1H phosphatase acts as a downstream regulator to counteract LRRK2 signaling by selectively dephosphorylating Rab8 and Rab10 (21). Finally, LRRK2 harbours several protein coding variants which modulate risk for sporadic PD, potentially mediated by subtle alterations in the LRRK2 kinase activity (2).
Here, we show that distinct modulators of the LRRK2 signaling pathway converge upon causing centrosomal cohesion deficits in cultured cells. The pathogenic LRRK2mediated centrosomal cohesion deficits are independent of the presence of Rab12, Rab35, Rab43 or RILPL2, but depend on the RILPL1-mediated centrosomal accumulation of phosphorylated Rab10. Correlated light and electron microscopy (CLEM) indicates that RILPL1 directs the phosphorylated Rab proteins to subdistal appendages of the mother centriole, where they may interfere with proper centrosomal cohesion and ciliogenesis as observed in the context of pathogenic LRRK2.

LRRK2 risk variants modulate centrosomal cohesion in HEK293T cells
To explore the relationship between centrosomal cohesion phenotypes and LRRK2 variants described to positively or negatively impact PD risk, we transiently transfected HEK293T cells with wildtype LRRK2, with a point mutant described to be non-pathogenic (T1410A), distinct point mutants described to increase PD risk (R1628P, S1647T, N2081D, G2385R), or a pathogenic point mutant which served as a positive control (Y1699C) (22)(23)(24)(25)(26)(27)(28). Compared to expression of wildtype LRRK2, the pathogenic Y1699C-LRRK2 mutant caused a pronounced deficit in centrosomal cohesion, as evidenced by the percentage of cells displaying duplicated centrosomes with a distance further than 1.5 microns apart, which was significantly reduced by transient application of the LRRK2 kinase inhibitor MLi2 (Figure 1A,B). Expression of the non-pathogenic T1410A LRRK2 mutant was without effect, whilst four distinct PD risk variants caused a statistically significant centrosomal cohesion deficit which was reverted by MLi2 ( Figure 1A,B). As assessed by immunoblotting, transient expression of the pathogenic Y1699C-LRRK2 mutant caused a detectable increase in Rab10 phosphorylation as compared to wildtype LRRK2, whilst no detectable differences were observed when expressing the various LRRK2 risk variants ( Figure 1C). However, increased accumulation of phospho-Rab10 in individually transfected cells could be detected by immunocytochemistry when expressing pathogenic LRRK2 or the various risk variants, and such accumulation was reverted in all cases by LRRK2 kinase inhibitor ( Figure 1D).
To analyse the effect of the R1398H mutation in LRRK2 which is protective against PD (25,29,30), we introduced it into either wildtype or pathogenic LRRK2 constructs.
Transient expression of pathogenic G2019S, R1441C, Y1699C, N1437H or I2020T LRRK2 mutants caused a significant deficit in centrosomal cohesion which was attenuated by introduction of the protective R1398H variant in all cases ( Figure S1). Similar results were obtained when introducing synthetic mutations (R1398L or R1398L/T1343V) described to alter Rab10 phosphorylation by modulating LRRK2 GTP binding/hydrolysis (31)(32)(33)(34) (Figure S2). Collectively, these data indicate that risk or protective LRRK2 variants can either negatively or positively impact upon the centrosomal cohesion deficits mediated by the LRRK2 kinase activity.

Rab12, Rab35 and Rab43 are not required for the centrosomal cohesion deficits mediated by pathogenic LRRK2
LRRK2 phosphorylates various endogenous Rab proteins including Rab8, Rab10, Rab12, Rab35 and Rab43 (4), and previous studies have implicated Rab8 and Rab10 in the ciliary and centrosomal deficits mediated by pathogenic LRRK2 (9,10). To study the potential role of Rab12, Rab35 or Rab43, we employed A549 cells where these proteins were knocked out using CRISPR-Cas9 (4) (Figure 2A,B). Consistent with what we have previously shown (10), expression of pathogenic LRRK2 in wildtype A549 cells caused a centrosomal cohesion deficit which was reverted by MLi2, whilst wildtype or a kinaseinactive LRRK2 mutant were without effect ( Figure 2C). Similar centrosomal cohesion deficits were observed when pathogenic LRRK2 constructs were expressed in A549 cells deficient in either Rab12 ( Figure 2B,D), Rab35 ( Figure 2E) or Rab43 ( Figure 2F). In all cases, the deficits were reverted by MLi2, and were not observed when expressing wildtype or kinase-inactive LRRK2 mutant. The pathogenic LRRK2-mediated increase in the percentage of split centrosomes was paralleled by an increase in the overall mean distance between duplicated centrosomes ( Figure 2G). Furthermore, and as assessed by immunoblot analysis, expression of pathogenic LRRK2 caused similar increases in the levels of phospho-Rab8 and phospho-Rab10 in wildtype cells or in cells deficient in either Rab12, Rab35 or Rab43 ( Figure 2H). Thus, and in contrast to Rab8 and Rab10 (10), the centrosomal cohesion deficits mediated by pathogenic LRRK2 are not dependent on the presence of Rab12, Rab35 or Rab43.

Upstream and downstream regulators of the LRRK2 signaling pathway impact upon centrosomal cohesion
Previous work has indicated that co-expression of Rab29 and LRRK2 stimulates the LRRK2 activity by recruiting it to the Golgi complex (14,15), and our data indicate that this is associated with centrosomal cohesion deficits (35). To determine whether endogenous Rab29 is required for the pathogenic LRRK2-mediated centrosomal deficits, we employed A549 cells deficient in Rab29. The centrosomal cohesion deficits mediated by pathogenic LRRK2 expression were significantly blunted in Rab29-deficient as compared to wildtype cells ( Figure S3). Since knockout of Rab29 does not seem to influence basal or pathogenic LRRK2 kinase activity (16), these findings suggest that the presence of Rab29 may be important for the LRRK2-mediated centrosomal cohesion phenotype in a manner independent of its ability to regulate the LRRK2 kinase activity.
Vps35 is a key component of the retromer complex which regulates vesicular trafficking to and from the Golgi complex. Strikingly, a point mutation (vps35-D620N) which causes autosomal-dominant late-onset PD (17)(18)(19) hyperactivates LRRK2 through a currently unknown mechanism (20). We next wondered whether vps35 or mutants thereof may impact upon the LRRK2-mediated centrosomal cohesion deficits. Pathogenic LRRK2 expression in vps35-deficient A549 cells (20) caused centrosomal cohesion deficits identical to those observed in wildtype cells, indicating that the presence of endogenous vsp35 is not required for this phenotype ( Figure S3). To determine whether the pathogenic vps35-D620N mutant regulates the centrosomal cohesion deficits by activating LRRK2, we coexpressed wildtype LRRK2 with HA-tagged wildtype or mutant vps35 variants in HEK293T cells. Coexpression of LRRK2 with vsp35-D620N caused a centrosomal cohesion deficit which was reverted by MLi2 treatment (Figure 3A,B) and correlated with an increase in the levels of phospho-Rab10 as assessed by immunoblot analysis ( Figure   3C). In contrast, no effects were observed when coexpressing wildtype LRRK2 with either wildtype vps35, or with two distinct, non-pathogenic vps35 point mutants (vps35-L774M, vps35-M57I) (17,19) (Figure 3A-C). These data support the conclusion that the vps35-D620N mutant activates the LRRK2 kinase as assessed by increased Rab10 phosphorylation to cause centrosomal cohesion deficits.
Lastly, we evaluated the contribution of PPM1H, the protein phosphatase responsible for dephosphorylating the LRRK2-phosphorylated Rab10 and Rab8 proteins (21). Side-by-side comparison revealed that wildtype LRRK2 expression per se was able to cause a centrosomal cohesion deficit in the PPM1H knockout cells, but not in wildtype cells ( Figure 4A,B). The centrosomal deficits were reverted by MLi2, and correlated with an increase in the levels of phospho-Rab10 and phospho-Rab8 as assessed by immunoblotting ( Figure 4C). No additional effects on centrosomal cohesion were observed when expressing distinct pathogenic LRRK2 mutants in the PPM1H knockout as compared to the control cells ( Figure 4C). Therefore, altogether these data support the conclusion that the centrosomal cohesion deficits mediated by the LRRK2 kinase activity are subject to modulation by both upstream and downstream components of the LRRK2 signaling pathway.

LRRK2-phosphorylated Rab proteins
LRRK2-phosphorylated Rab8 and Rab10 bind with strong preference to RILPL1 and RILPL2 (4), and RILPL1 is required for the centrosomal cohesion deficits mediated by pathogenic LRRK2 (10). To evaluate the involvement of RILPL2, we transfected A549 RILPL2 knockout cells with wildtype or distinct pathogenic LRRK2 mutants. Pathogenic LRRK2 expression caused centrosomal cohesion deficits in RILPL2 knockout cells identical to those observed in wildtype cells, which were reverted by MLi2 in all cases ( Figure 5). In addition, wildtype cells transfected with pathogenic LRRK2 displayed prominent phospho-Rab10 staining in a perinuclear area and in tubular structures, and similar staining was observed in RILPL2 knockout cells ( Figure S4). In contrast, and as previously reported (9), RILPL1 knockout cells transfected with pathogenic LRRK2 showed a diminished perinuclear distribution of phospho-Rab10 accompanied by punctate staining throughout the cytosol ( Figure S4). Together, these data support the conclusion that the centrosomal cohesion phenotype mediated by pathogenic LRRK2 is largely determined by the interaction of phospho-Rab10 with RILPL1, at least in this cell system.
To further study the involvement of RILPL1 in the LRRK2-mediated centrosomal cohesion deficits, we expressed the C-terminal half of the protein (RL1d-GFP), reported to be responsible for its interaction with the phosphorylated Rab proteins (4,9). When expressed in A549 cells, RL1d-GFP displayed a punctate as well as cytosolic localization ( Figure 6A). Strikingly, RL1d-GFP expression completely reverted the centrosomal cohesion phenotype induced by pathogenic LRRK2, whilst not displaying an effect when expressed on its own (Figure 6A-C). Co-expression of RL1d-GFP with pathogenic LRRK2 did not decrease the total levels of phospho-Rab8 or phospho-Rab10 as assessed by immunoblot analysis (Figure 6D). Rather, whilst pathogenic LRRK2 caused a pronounced perinuclear accumulation of phospho-Rab10, co-expression with RL1d-GFP caused the redistribution of phospho-Rab10 to cytosolic RL1d-GFP-positive punctae ( Figure 6E). These data indicate that the centrosomal phospho-Rab10 accumulation depends on the centrosomal localization of RILPL1 which is required to cause the cohesion deficits mediated by pathogenic LRRK2. RILPL1 localizes to subdistal appendages of the mother centriole RILPL1 associates with the mother centriole which becomes the basal body that nucleates the primary cilium (8). Employing two distinct anti-RILPL1 antibodies in HEK293T cells, we corroborated the centrosomal localization of endogenous RILPL1, and the colocalization of RILPL1 with endogenous phosphorylated Rab proteins in cells transfected with pathogenic LRRK2 (Figure S5). Both N-terminally or C-terminally GFPtagged RILPL1 proteins displayed a pericentrosomal localization when expressed in A549 cells (9) (Figure 7A), and an identical pericentrosomal localization was observed when RILPL1 was fused to miniSOG (RILPL1-miniSOG), a tag suitable for correlative light and electron microscopy (CLEM) (36,37). In addition, and as previously reported (10), transient expression of tagged RILPL1 in A549 cells had no effect on centrosomal cohesion as assessed by measuring the mean distance between duplicated centrosomes ( Figure 7B).
In order to image the pericentrosomal localization of RILPL1 with high spatial resolution, we performed miniSOG-induced DAB oxidation which generated a localized polymeric precipitate that could be readily identified by CLEM ( Figure 7C). Using STEM tomography (for which we combined a multiple-tilt tomography approach with the scanning mode of a TEM), we determined that RILPL1 was localized to the subdistal appendages of the mother centriole, hinting that the interaction between phospho-Rab8/phospho-Rab10 and RILPL1 may occur at this localization. The specific DAB labelling was clearly distinguishable in RILPL1-miniSOG-expressing A549 cells as compared to adjacent non-expressing cells that were exposed to the same processing within the photooxidized area. An accumulation of DAB-labeled pericentrosomal vesicles was observed in expressing cells only, suggesting this might be induced by RILPL1 overexpression.

Discussion
Pathogenic LRRK2 activity causes loss of primary cilia in various cell types in vitro, and the intact brain of G2019S or R1441C LRRK2 knockin mice in vivo (9,38).
Conversely, in dividing cells pathogenic LRRK2 activity causes a deficit in the proper cohesion of duplicated centrosomes (10)(11)(12). Both ciliogenesis and centrosomal cohesion deficits are due to an accumulation of LRRK2-phosphorylated Rab proteins at a centrosomal location which is dependent on RILPL1 (9,10). In this study, we show that centrosomal cohesion is also regulated by risk and protective LRRK2 variants, and by modulators of the LRRK2 kinase signaling pathway. Expression of LRRK2 risk variants causes centrosomal cohesion deficits, whilst introduction of a protective variant into distinct pathogenic LRRK2 constructs decreases such deficits. Moreover, a point mutation in vps35 (vps35-D620N) which causes autosomal-dominant PD (17)(18)(19) and activates the LRRK2 kinase (20) causes a pronounced centrosomal cohesion deficit. Knockout of centrosomes. It further predicts that the underlying mechanism may involve the dynamic recruitment and/or displacement of protein(s) necessary to keep duplicated centrosomes in close proximity to each other.
We have previously reported centrosomal cohesion deficits in lymphoblastoid cell lines from G2019S LRRK2-PD patients as compared to healthy controls, and such deficits were also observed in a subset of sporadic PD patients (11,12). In future experiments, it will be interesting to determine whether centrosomal deficits can be detected in peripheral cells from PD patients harboring LRRK2 risk variants or mutations in vps35. In addition, our study shows that expression of N2081D LRRK2 mutant causes kinase activitymediated centrosomal cohesion deficits. Since the N2081D LRRK2 mutation confers risk for PD as well as for Crohn´s disease (28), further studies are warranted to probe for centrosomal deficits in peripheral cells from Crohn´s disease patients, as this may aid in stratifying patients benefitting from LRRK2 kinase inhibitor therapeutics in clinical studies.
Recent work has shown that inducing lysosomal damage causes recruitment of LRRK2 to lysosomes, followed by the lysosomal accumulation of phospho-Rab10 (39)(40)(41)(42). Conversely, mitochondrial depolarization causes the mitochondrial accumulation of Rab10 to facilitate mitophagy, and such accumulation is impaired in the context of pathogenic LRRK2 (43). Our studies were performed in the absence of treatments to induce lysosomal damage or mitochondrial depolarization. Under such normal physiological conditions, we find that the majority of phospho-Rab10 localizes to the centrosome to cause centrosomal cohesion and ciliogenesis deficits. In future experiments, it will be interesting to determine how triggers which lead to lysosomal or mitochondrial damage might impact upon the centrosomal deficits as described here. In either case, our data indicate that it is the RILPL1-mediated localization of phospho-Rab10 to the centrosome which is responsible for the cohesion deficits mediated by pathogenic LRRK2.
Interestingly, the expression of a C-terminal fragment of RILPL1 which localizes to cytosolic punctate structures reverts the cohesion deficits mediated by pathogenic LRRK2 by redistributing phospho-Rab10 from its centrosomal location to those structures, without altering the total levels of phospho-Rab10 as assessed by Western blot analysis. Therefore, the subcellular location of the phospho-Rab10 accumulation, rather than total phospho- Altogether, our data demonstrate that pathogenic LRRK2 mutations, LRRK2 risk variants and modulators of the LRRK2 signaling pathway all converge upon causing centrosomal cohesion and ciliogenesis defects. These deficits are dependent on RILPL1, and directly mediated by the centrosomal accumulation of phospho-Rab10. The localization of RILPL1 implicates the subdistal appendages of the mother centriole as the prime site of action for the LRRK2-mediated phospho-Rab10 accumulation, with downstream effects on centrosomal cohesion and ciliogenesis as described here.

Image acquisition and quantification
Images were acquired on a Leica TCS-SP5 confocal microscope using a 63x 1.4 NA oil UV objective (HCX PLAPO CS) (10), or on an Olympus FV1000 Fluoview confocal microscope using a 60x 1.2 NA water objective lens. Images were collected using single excitation for each wavelength separately and dependent on secondary antibodies, and the same laser intensity settings and exposure times were used for image acquisition of individual experiments to be quantified. Around 10-15 optical sections of selected areas were acquired with a step size of 0.5 m, and maximum intensity projections of z-stack images analyzed and processed using Leica Applied Systems (LAS AF6000) image acquisition software or ImageJ. Duplicated centrosomes were scored as being split when the distance between their centres was > 1.5 m for HEK293T cells (10,11) or > 2.5 m for A549 cells (10), respectively. Mitotic cells were excluded from the analysis in all cases.
Quantification of centrosomal distances was performed by an additional observer blind to condition, with identical results obtained in both cases.

A549 cells were cultured in glass-bottom MatTek dishes (MatTek Life Sciences,
P35G-0-14-C) and transfected with RILPL1-miniSOG-HA using LipoD293™ Transfection Reagent (SignaGen Laboratories, SL100668) as described above. Proteins were allowed to express for 48 hours and cells were processed as previously described (37,47). Briefly, cells were rinsed with pre-warmed HBSS and fixed using pre-warmed 2. Photo-oxidized areas of interest were identified by transmitted light, sawed out using a jeweller's saw, and mounted on dummy acrylic blocks with cyanoacrylic adhesive.
The coverslip was carefully removed, the resin was trimmed, and ultrathin sections (80 nm thick) were cut using a diamond knife (Diatome). Electron micrographs were recorded using a FEI Tecnai™ 12 Spirit TEM (transmission electron microscope) operated at 80 kV. Detection Reagent (GE Healthcare) as previously described (10).