The molecular motor Myosin Va interacts with the cilia-centrosomal protein RPGRIP1L

Myosin Va (MyoVa) is an actin-based molecular motor abundantly found at the centrosome. However, the role of MyoVa at this organelle has been elusive due to the lack of evidence on interacting partners or functional data. Herein, we combined yeast two-hybrid screen, biochemical studies and cellular assays to demonstrate that MyoVa interacts with RPGRIP1L, a cilia-centrosomal protein that controls ciliary signaling and positioning. MyoVa binds to the C2 domains of RPGRIP1L via residues located near or in the Rab11a-binding site, a conserved site in the globular tail domain (GTD) from class V myosins. According to proximity ligation assays, MyoVa and RPGRIP1L can interact near the cilium base in ciliated RPE cells. Furthermore, we showed that RPE cells expressing dominant-negative constructs of MyoVa are mostly unciliated, providing the first experimental evidence about a possible link between this molecular motor and cilia-related processes.

To validate this interaction in the YTH system, we co-transformed the yeast strain L40 with the bait (pBTM116 or pBTM116_MyoVa-GTD) and prey plasmids (pACT2 or pACT2_RPGRIP1L) and evaluated the activation of two reporter genes, LacZ and HIS3. As expected, only the colonies expressing both MyoVa-GTD and RPGRIP1L displayed β -galactosidase activity and grew in presence of 10 mM 3-AT, indicating that RPGRIP1L binds to MyoVa-GTD (Fig. 1c).
As aforementioned, RPGRIP1L contains three C2 domains and there is increasing evidence that they can mediate protein-protein interactions, especially in the ciliary transition zone 28,29 . Therefore, to characterize the RPGRIP1L• MyoVa interaction and to evaluate the role of these C2 domains in MyoVa binding, we performed pull-down assays (Fig. 2a). The three C2 domains were able to interact with MyoVa-GTD, despite their low sequence identity (Fig. 2a,b). Microscale thermophoresis (MST) experiments showed that RPGRIP1L-C2 domains bind to MyoVa-GTD with dissociation constants in the 3-9 μ M range, with the C2 NTERM and C2 CTERM displaying the highest affinity for MyoVa-GTD (Fig. 2c).

RPGRIP1L binds to a conserved site of MyoVa and Vb
GTDs. The GTDs of MyoVa and Vb share a protein-binding site at the face C of lobule II, which is also conserved in the class V myosin Myo2p from yeast 30 . To investigate if RPGRIP1L also binds to this region, we mutated some conserved residues at this site to alanine and performed YTH assays (Fig. 3a), following a strategy similar to that used to map the protein-binding sites of Myo2p 31 . Additionally, we evaluated alanine mutants of residues involved in PTEN recognition (K1757 and K1759) 12 . As a control, residues from the other face of the MyoVa-GTD (face M), including one that is crucial for the binding of MyoVa motor domain (K1781) in the auto-inhibited state 32 , were also mutated. Analysis of these mutants indicated that the residues W1713, Y1721, Q1755 and F1792 are required for the interaction between MyoVa and RPGRIP1L (Fig. 3a). Based on these data, we suggest that the RPGRIP1L-binding site overlaps with those of Kar9 and Inp2 to Myo2p and that of Rab11a to MyoVb 31,33 (Fig. 3b). In agreement with this result, YTH assays showed that RPGRIP1L also interacts with MyoVb-GTD (Fig. 3c), indicating a redundant role for MyoVa and Vb in RPGRIP1L binding.
As MyoVa-GTD can be phosphorylated on residue S1652 by calcium/calmodulin-dependent protein kinase II (CaMKII) 34,35 , which results in its release from melanosomes and inhibition of melanosome transport 36 , we also investigated whether S1652 phosphorylation could affect RPGRIP1L binding. For this purpose, we used phospho-mimetic (S1652E and S1651E/S1652E) and non-phospho-mimetic (S1652A and S1651A/S1652A) mutations previously validated by Karcher and co-workers 36 . YTH analyses showed that RPGRIP1L was capable to interact with both mimetic mutants, indicating that S1652 phosphorylation does not prevent the binding of RPGRIP1L to MyoVa-GTD (Fig. 3d).

MyoVa interacts with RPGRIP1L at the centrosome. To validate the interaction between endogenous
MyoVa and RPGRIP1L, we performed proximity ligation assays (PLA) in RPE cells, a model system for studying primary cilium formation and function 18 . Since it is well known that a pool of MyoVa [19][20][21][22] and of RPGRIP1L 25,28,37 localize at the centrosome, we investigated whether they interact at this microenvironment in ciliated cells. As expected, the presence of PLA dots evidenced the physical interaction between MyoVa and RPGRIP1L near the primary cilium base (Fig. 4), indicating that the binding of MyoVa to RPGRIP1L can occur at the centrosome and might be involved in cilia-related processes.

Dominant-negative expression of MyoVa inhibits ciliogenesis.
The fact that RPGRIP1L, as well as other MyoVa-binding proteins (PTEN, Rab8, Rab11 and RILPL2), is involved with the regulation of the primary cilium structure and or composition 14,15,18,38 prompted us to investigate the effect of overexpressing two dominant-negative constructs of MyoVa in ciliogenesis, EGFP-GTD and EGFP-mGTD (GTD + 45 upstream amino-acid residues from the medial tail) (Supplementary Figure S1). Interestingly, these two constructs displayed different distribution patterns, being EGFP-mGTD localized in discrete foci near the nucleus, whereas EGFP-GTD was diffusely distributed at the cytoplasm ( Fig. 5a and b), indicating that these additional residues might be critical for GTD targeting. Despite this observation, the overexpression of both constructs strongly suppressed the assembly of primary cilium in RPE cells ( Fig. 5a and b), which did not occur in cells expressing EGFP alone (Fig. 5c). The same phenotype was also observed in melanoma B16 cells (data not shown), indicating that MyoVa might play a role in cilia-related processes.

Discussion
In the present work, we revealed that MyoVa interacts with the cilia-centrosomal protein RPGRIP1L at the centrosome. Moreover, we showed structural features of MyoVa required for RPGRIP1L recognition and provided the earliest evidence for a role of MyoVa in the primary cilium development, a centrosome-regulated process.
We demonstrated in vitro that the three C2 domains of RPGRIP1L can recruit the GTD of MyoVa with affinity typical of transient protein-protein interactions (low micromolar K d ), being the C2 NTERM and C2 CTERM domains those that best recognize MyoVa-GTD. In vivo, the affinity between RPGRIP1L and MyoVa might be further enhanced by MyoVa dimerization and the tandem disposition of the three C2 domains in RPGRIP1L, which likely increase the probability of their binding. Attempts to identify residues of RPGRIP1L involved in MyoVa  recognition using in silico predictions and site-directed mutagenesis were inconclusive, indicating the need of a deeper investigation to elucidate the molecular basis of MyoVa recruitment by the C2 domains of RPGRIP1L (Supplementary Figure S2).
C2 domains are recurrent in ciliary proteins from the transition zone, such as RPGRIP1L 29 and CC2D2A 39 , as well as in MyoVa-interacting proteins, such as PTEN and the Rab effectors granuphilin-a/b (Gran-a/b) and rabphilin-3A (Rph-3A) 12,40 . Although the role of PTEN-C2 domain in MyoVa binding is still elusive 12 , the C2 domains of Gran-a/b and Rph-3A interact with an alternatively spliced region of MyoVa tail, but in a region different from the binding site of RPGRIP1L-C2 domains 40 . Together, these examples illustrate an emerging role for C2 domains in linking proteins to the actin cytoskeleton via the recruitment of MyoVa.
Using YTH assays and site-directed mutagenesis, we showed that RPGRIP1L binds to both MyoVa and Vb, being recognized by a conserved region that overlaps with the Rab11a-binding site 33 . Interestingly, active Rab11 (GTP-bound form) is also recruited to the centrosome -specifically to the mother centriole appendages -where it is "turned off " (GDP-bound form) by Evi5 41 . Inactivation of Rab11a induces the release of effector proteins like MyoVa/Vb, suggesting that, in this microenvironment, the association of MyoVa with RPGRIP1L might be favoured over that with Rab11a. In agreement with this hypothesis, the Rab11a• GDP binding to MyoVb-GTD monomer 33 displays a K d 6 times higher than that of RPGRIP1L-C2 NTERM domain to MyoVa-GTD.
Our results also evidenced a direct interaction between MyoVa and RPGRIP1L at the vicinity of the basal body, indicating that the MyoVa• RPGRIP1L complex might play a role in primary cilium development. Genetic diseases related to defects in MYO5A and RPGRIP1L genes are characterized by common neurologic impairments, suggesting they function in correlated pathways in the brain 7,25,28,42,43 . One of such pathways might involve the primary cilium, since the protein RPGRIP1L has been linked to signaling pathways that depend on this organelle and play a key role in brain development (sonic hedgehog and Wnt) or brain function (leptin receptor signaling) 26,27,[44][45][46] . Furthermore, in neuron photoreceptors, RPGRIP1L localizes not only in the connecting cilium but also near the plasma membrane of the calycal processes 28 -microvillus-like projections rich in actin filamentssuggesting a potential role for RPGRIP1L in anchoring membranes to the actin cytoskeleton via MyoVa.
By overexpressing two dominant-negative constructs of MyoVa in RPE cells, we showed that loss of myosin V transport function suppresses ciliogenesis. Together with the fact that most proteins known to bind to MyoVa GTD regulate cilia assembly (Rab11), transition zone establishment (RPGRIP1L), cilia dynamics (PTEN) or cilia composition (RILPL2), our data support a role for MyoVa in cilia-related processes.
In summary, our studies revealed RPGRIP1L as a novel MyoVa-binding protein -the first to be demonstrated to interact with MyoVa at the centrosome -and uncover an unprecedented link between MyoVa and ciliogenesis, providing new perspectives for studies aiming to better understand why defects in MyoVa cause neurological disorders in Griscelli syndrome patients.
Yeast two-hybrid screen (YTH). YTH screen was performed in Saccharomyces cerevisiae strain L40 (trp1-901, his3D200, leu2-3, ade2 LYS2::(lexAop)4-HIS3 URA3::(lexAop)8-lac GAL4) using MyoVa-GTD cloned into pBTM116 (LexA DNA-binding domain, DBD) as bait and a human fetal brain cDNA library (Clontech) cloned into pACT2 (Gal4 activation domain, AD) as prey. Yeast cells were transformed with pBTM116_MyoVa-GTD vector and the library as described by Alborghetti and co-workers 47 . The screen was performed in solid Synthetic Defined Medium without tryptophan, leucine and histidine (SD-WLH) containing 5 mM 3-amino-1,2,4-triazole (3-AT) (Sigma-Aldrich, St. Louis, MO). To identify the preys, the pACT2 plasmids of positive clones were isolated and sequenced. The DNA sequences were then compared with those available in the NCBI data bank using the BLASTX program 48 . The clone identified as encoding for the full-length RPGRIP1L isoform c (NP_001295263.1) was further selected for in vitro and in cell validation and characterization.
Yeast reporter gene assays. To confirm the interaction between pBTM116_MyoVa-GTD and pACT2_ RPGRIP1L, S. cerevisiae L40 cells were transformed with both constructs. As negative controls, we used L40 cells transformed with pBTM116_MyoVa-GTD and empty pACT2 or pACT2_RPGRIP1L and empty pBTM116. Cells were plated in Synthetic Defined Medium without tryptophan and leucine (SD-WL) and then incubated at 30 °C for 3 days. For β -galactosidase activity assay, cells were transferred to Whatman ® 3 MM paper (Sigma-Aldrich), permeabilized with liquid nitrogen and wrapped on a second paper soaked in Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM MgCl 2 , 50 mM β -mercaptoethanol and pH 7.0) containing 2 μ g/mL 5-bromo-4-chloro-3-indolyl-β -D-galactoside (X-Gal; Sigma-Aldrich). Cells were incubated at 37 °C for a couple of hours until the blue color appears, indicating the β -galactosidase activity. For HIS3 activation assay, cells were plated in SD-WLH containing 10 mM 3-AT, incubated for 3 days at 30 °C and imaged.
Cells were disrupted by sonication and centrifuged at 40,000 x g. The supernatant was loaded onto a 5 mL GSTrap FF column (GE Healthcare), pre-equilibrated with lysis buffer, using an Äkta FPLC (GE Healthcare). GST-tagged constructs were eluted using lysis buffer added by 10 mM reduced glutathione (Sigma-Aldrich). The recombinant protein was dialyzed against ligation buffer (50 mM HEPES, 20 mM NaCl, 5% (v/v) glycerol, pH 7.5) and subsequently incubated with 1% (m/m) trypsin (Sigma-Aldrich), at 4 °C, during 30 min under gentle agitation, for GST-tag cleavage. The reaction was stopped with 1 mM phenylmethylsulfonyl fluoride (PMSF) (USB Corporation, Cleveland, OH) and loaded onto a 5 mL HiTrap Q FF column (GE Healthcare) pre-equilibrated with ligation buffer using an Äkta FPLC (GE Healthcare). After washing the resin, the target protein was eluted using a step-gradient from 20 mM to 1000 mM NaCl. Residual contamination with GST was removed by affinity chromatography using a 5 mL GSTrap FF column (GE Healthcare). All purification steps were carried out at 4 °C. MyoVa-GTD was expressed and purified as described by Nascimento and co-workers 30  MST assays 51 were performed using 300 nM of FITC-labeled MyoVa-GTD and a serial dilution of RPGRIP1L-C2 domains in interaction buffer (50 mM HEPES, 150 mM NaCl, 5% (v/v) glycerol, 0.05% (v/v) tween-20 and pH 7.2). Samples were loaded into Monolith TM NT.115 MST Premium Coated capillaries (NanoTemper Technologies, Munich, Germany) and thermophoresis data were measured in the Monolith TM NT.115 device (NanoTemper Technologies) using a LED power of 20% and a MST power of 60%. Initial fluorescence and back diffusion were measured for 5 s whereas the thermophoretic movement was recorded for 30 s. All assays were performed in triplicate and data were processed using the NTAffinity Analysis software (NanoTemper Technologies) and Origin 8.0. The dissociation constant (K d ) was calculated from changes in the normalized fluorescence (F norm ) as a function of the RPGRIP1L-C2 domains concentration. BSA (Sigma-Aldrich), 0.2% (v/v) Triton X-100 (Sigma-Aldrich) in PBS pH 7.4 (wash buffer) and permeabilized with 0.5% (v/v) Triton X-100 in PBS pH 7.4 for 10 min. Free aldehydes were blocked by incubating cells with 10 mM glycine (Promega, Madson, WI) in PBS pH 7.4 for 5 min. Cells were blocked with 3% (w/v) BSA, 0.2% (v/v) Triton X-100 in PBS pH 7.4 for 30 min. PLA was performed using rabbit anti-MyoVa at 0.2 μ g/mL (M4812, Sigma-Aldrich), goat anti-RPGRIP1L at 1 μ g/mL (sc-165400, Santa Cruz) and Duolink ® in situ red starter kit goat/rabbit (DUO92105, Sigma-Aldrich) whereas counterstaining used DAPI, mouse anti-acetylated-α -tubulin at 0.7 μ g/mL (32-2700, Invitrogen, Thermo Fisher Scientific) and chicken anti-mouse 488 at 10 μ g/mL (A21200, Invitrogen, Thermo Fisher Scientific) antibodies. All procedures were performed according to the manufacturer's protocol. ProLong ® Gold Antifade (Life Technologies) was used as mounting medium. Cells were imaged in True Confocal Scanning (TCS) SP8 microscope (Leica, Wetzlar, Germany) at the Biological Imaging facility from the Brazilian Biosciences National Laboratory (LNBio). Images were obtained using oil immersion HC PL APO CS2 63x/1.4 objective lens and 1.4 numerical aperture. Microscopy images were previously analyzed in the LAS AF lite program (Leica). The maximum intensity projection and channel levels correction were performed using the FIJI platform 52 .

Dominant-negative overexpression.
For dominant-negative overexpression assays, 3 × 10 4 hTERT RPE-1 cells (ATCC) were incubated overnight on 13 mm diameter cover slips (Knittel, Braunschweig, Germany) into 24-well plates (Corning incorporated Costar ® ). The cells were transfected using Lipofectamine ® 3000 (Thermo Fisher Scientific, Carlsbad, CA). For overexpression, pEGFP-C1 plasmids encoding chicken brain MyoVa-mGTD (residues 1377-1830; CAA77782.1) or MyoVa-GTD (residues 1423-1830; CAA77782.1) fused to EGFP as well as the empty plasmid encoding only EGFP as control were used. MyoVa-mGTD was identical to the one described in 19,20 , except that the insert, previously in pS65T-C1, was transferred to pEGFP-C1. MyoVa-GTD (residues 1423-1830; CAA77782.1) was PCR amplified using as template a chicken brain MyoVa full tail cDNA clone 53 , and the PCR product was inserted into pEGFP-C1 plasmid in fusion with EGFP. The transfections were carried out in a final volume of 500 μ L medium following the manufactures' instructions. After 72 h of incubation for overexpression of EGFP or EGFP-MyoVa-GTD, cells were washed with PBS, fixed with 4% (v/v) paraformaldehyde pH 7.4 for 15 min, washed with PBS and processed for immunofluorescence and data acquisition as described above. For these assays, cells were imaged using a Leica CTR 6000 microscope (Leica).