The atypical small GTPase RABL3 interacts with RAB11 to regulate early ciliogenesis in human cells

The centrosome contains two cylinder-like structures, termed mother and daughter centrioles. The mother centriole is equipped with distal and subdistal appendages (DA and SDA, respectively), which are absent in the daughter centriole. Centrioles nucleate spindles during mitosis, whereas the mother centriole becomes the base of a sensory organelle called the primary cilium during interphase (Kobayashi and Dynlacht, 2011; Sánchez and Dynlacht, 2016). The primary cilium protrudes from the cell surface of most mammalian cells and contains multiple signaling molecules (Ishikawa and Marshall, 2011). Structural and functional anomalies in primary cilia are implicated in a broad spectrum of genetic diseases and cancers (Eguether and Hahne, 2018; Liu et al., 2018; Reiter and Leroux, 2017). In cultured mammalian cells, serum deprivation from the culture medium induces a transition from the mother centriole to the primary cilium.

A small vesicle, termed the ciliary vesicle (CV), covers the top of the mother centriole in the early stage of the intracellular ciliogenesis in human cells (Sorokin, 1962). Pre-ciliary vesicles (PCVs), which are thought to be derived from the Golgi, are initially attached to the DA of the mother centriole, where they are referred to as distal appendage vesicles (DAVs), and DAVs assemble into the CV (Shakya and Westlake, 2021). The CV extends along with microtubule elongation and finally develops into the ciliary membrane (CM), encapsulating the ciliary axoneme. Thus far, mounting studies have indicated that several proteins regulate CM formation processes in a stepwise manner, including members of the RAB family small GTPases. Myo-Va occupies the PCV and aids in the trafficking of PCVs to the DA (Wu et al., 2018). EHD1, EHD3, SNAP29, PACSIN1, PACSIN2, and MICAL-L1 mediate CV assembly from DAVs (Insinna et al., 2019; Lu et al., 2015). The small GTPase RAB34 is also required for CV assembly step (Ganga et al., 2021; Stuck et al., 2021). The small GTPase RAB8 (which has RAB8A and RAB8B forms) appears to be involved in the extension of CV to develop CM (Lu et al., 2015). Two ciliopathy-related proteins, Talpid3 and CEP290, are necessary for RAB8 recruitment and proper CV development (Kobayashi et al., 2014). RAB8 is activated by a guanine nucleotide exchange factor (GEF), Rabin8 (also known as RAB3IP), and the small GTPase RAB11 (herein referring to both RAB11A and RAB11B unless otherwise specified) (Knödler et al., 2010). RAB11 regulates GEF activity and centrosome targeting of Rabin8 (Knödler et al., 2010; Westlake et al., 2011). Recently, it was shown that lysophosphatidic acid (LPA)–LPA receptor-1-dependent phosphatidylinositol-3-kinase (PI3K)/Akt signaling controls RAB11 activation during ciliogenesis (Walia et al., 2019).

RAB-like 3 (RABL3) is a member of the RABL family, a RAB GTPase subfamily (Colicelli, 2004). RABL proteins are devoid of sites for lipid modification and some conserved residues found in other GTP-binding proteins (Blacque et al., 2018; Homma et al., 2021). RABL3 is upregulated in several cancers and is required for their proliferation and migration (An et al., 2017; Ge et al., 2019; Li et al., 2010; Ma et al., 2021; Pan et al., 2017; Usman et al., 2020; Xu et al., 2021; Zhang et al., 2016). Mutations in RABL3 are correlated with heritable pancreatic cancer (Nissim et al., 2019). A recent study has shown that RABL3 is crucial for lymphoid function, and that its knockout mice are embryonic lethal (Zhong et al., 2020). Interestingly, four of the six RABL proteins are known to be involved in molecular trafficking to and within primary cilia. RABL4 and RABL5 (also known as IFT27 and IFT22, respectively) comprise the intraflagellar transport (IFT) machinery, a universally conserved protein complex that bi-directionally conveys molecules in primary cilia (Nakayama and Katoh, 2018). RABL2 controls anterograde IFT and trafficking of ciliary G-protein-coupled receptors (Dateyama et al., 2019; Kanie et al., 2017; Nishijima et al., 2017). A recent study has shown that RABL2 ensures export of ciliary proteins (Duan et al., 2021). A proteomic analysis using mouse photoreceptor sensory cilia suggested RABL3 as a cilia-related protein (Liu et al., 2007); however, it is unknown whether and how RABL3 is associated with primary cilia.

In this study, we investigated normal diploid human cells depleted of RABL3 and discovered that RABL3 is required to assemble primary cilia. A proteomic approach identified RAB11 as a RABL3-binding protein. RABL3 directly interacted with and was stabilized by RAB11 in vitro. RABL3 accumulated around the centrosome during early ciliogenesis. Furthermore, RABL3 depletion impaired CV formation, but ectopic RABL3 expression promoted CV formation, phenocopying what was seen with ablation and overexpression of RAB11, respectively. Altogether, these results suggest that RABL3 cooperates with RAB11 and regulates the CV formation during early ciliogenesis in human cells.

To investigate the consequences of RABL3 depletion in normal diploid human cells, we generated RABL3-mutated retinal pigment epithelial (RPE1) cells by CRISPR/Cas9-mediated gene editing. Sequence analysis indicated heterozygous mutations in two clones, Rabl3-1 and Rabl3-2 (Fig. S1A). A four-nucleotide deletion in one allele led to a premature stop codon before the G4 domain in Rabl3-1, whereas a three-nucleotide in-frame deletion resulted in replacement of two amino acids, glutamine and asparagine with one amino acid (histidine) in another allele. This QN motif is conserved in vertebrate RABL3 (Fig. S1B), suggesting that this mutation seriously impairs RABL3. In contrast, mutations in both alleles lead to premature stop codons in Rabl3-2. We found a substantial decrease in RABL3 protein levels in both clones by immunoblotting analysis, although RABL3 was faintly detected in Rabl3-1 as predicted by sequence analysis (Fig. 1A).

We then investigated primary cilia formation in RABL3-mutated cells. Primary cilia were visualized by immunofluorescence experiments with two specific antibodies against glutamylated tubulin (Glu. Tub.) and ARL13B. The wild-type (WT) RPE1 cells assembled primary cilia when induced to quiescence by removing the serum from the culture medium (Fig. 1B,C). In contrast, Rabl3-1 and Rabl3-2 cells formed significantly fewer primary cilia than WT cells (Fig. 1B,C). We next conducted rescue experiments to verify that the de-ciliation phenotype in RABL3-mutated cells was due to the loss of RABL3. Ectopic expression of RABL3 significantly restored primary cilia in Rabl3-1 and Rabl3-2 cells (Fig. 1D,E). In contrast, the RABL3/S20N mutant, which is expected to fail to bind GTP (i.e. is a negatively (non-active) locked mutant), did not ameliorate the phenotype (Fig. 1D,E). These results suggest that RABL3 with GTP-binding capacity is required for ciliogenesis. We also performed siRNA-mediated knockdown experiments. The protein expression of endogenous RABL3 was considerably decreased by introducing two individual siRNAs targeting RABL3 (Fig. 1F). We observed a significant reduction in primary cilia assembly in these RABL3-knockdown RPE1 cells (Fig. 1G). In contrast, siRNA-mediated silencing of RABL3 failed to influence the number of primary cilia in RABL3-mutated cells (Fig. S1C). Collectively, these results strongly suggest that RABL3 is a requisite for primary cilia formation in RPE1 cells.

To uncover the mechanistic role of RABL3 in primary ciliogenesis, we performed a proteomic screen to identify RABL3-interacting proteins. Immunoaffinity chromatography and subsequent mass spectrometric analysis identified a peptide common to the two close paralogs of the small GTPase RAB11, RAB11A and RAB11B, in RABL3 immunoprecipitation (Fig. S2A,B). Interestingly, RABL3 has been identified as an interactor of RAB11A in a comprehensive RAB-interactome analysis using dendritic cells (Li et al., 2016). We co-expressed Flag-tagged RABL3 and GFP-fused RAB11A in HEK293T cells and performed anti-Flag immunoprecipitation to test whether they interact in cells and found that RABL3 co-precipitated with RAB11A (Fig. 2A). We subsequently performed similar experiments using cells co-expressing wild-type RABL3 (RABL3/WT) or RABL3/S20N, and wild-type RAB11A (RAB11A/WT), its actively locked Q70L mutant, or negatively locked S25N mutant, and found efficient co-precipitation of RABL3/S20N and RAB11A/Q70L (Fig. 2A). Reciprocal immunoprecipitation using an anti-GFP antibody confirmed similar binding properties (Fig. 2B). These results suggest that the negative form of RABL3 preferentially interacts with the active form of RAB11A. In addition, RABL3/S20N co-precipitated with RAB11A but not with RAB23 (Fig. 2C), suggesting that RABL3 specifically associates with RAB11A.

We next investigated the direct interaction between RABL3 and RAB11. To this end, we purified recombinant RABL3 and RAB11A proteins from bacterial lysates (Fig. S2C). We then performed a pulldown assay using glutathione Sepharose with buffer containing Mg²⁺, and found that GST–RAB11A specifically pulled down RABL3 (Fig. 2D, left two lanes), demonstrating their direct binding. In contrast, adding a non-hydrolyzable GTP analog GTPγS or GDP substantially abrogated their association (Fig. 2D, middle and right lanes). This result shows that RABL3 and RAB11 fail to bind to each other in the same guanine nucleotide forms. Furthermore, we performed a similar in vitro binding assay using nucleotide-preloaded proteins and detected specific binding between GDP-loaded RABL3 and GTPγS-loaded RAB11A (Fig. 2E). Consistent with previous immunoprecipitation assays, these results suggest that the negative form of RABL3 interacts with the active form of RAB11A.

A previous study has shown that RABL4 (IFT27) binds to the small GTPase ARL6 and prevents the aggregation of ARL6 in vitro (Liew et al., 2014). Because of the analogous RABL-small GTPase interaction, we performed a similar turbidity assay using the recombinant RABL3 and RAB11A proteins to test mutual chaperone activity. The optical density at 350 nm was monitored to detect the insoluble (precipitated) proteins (Liew et al., 2014). Although RABL3 was precipitated upon incubation at 37°C for 12 min, the precipitation was significantly reduced in the mixture of RABL3 and GST–RAB11A (Fig. 2E). In contrast, GST–RAB11A was soluble in this assay. These results suggest that RAB11 stabilizes RABL3 in vitro. On the other hand, GST–RAB11A failed to accelerate dissociation of GDP from RABL3 in vitro (Fig. S2D), suggesting that RAB11A is not a GEF for RABL3.

We assessed the subcellular localization of RABL3 in RPE1 cells by performing immunofluorescence assays. Multiple puncta were detected throughout the cells with a RABL3 antibody (Fig. 3A, upper left), and these puncta were almost completely absent in Rabl3-2 cells (Fig. 3A, lower left), indicating that the RABL3 antibody specifically recognizes endogenous RABL3 in RPE1 cells. Endogenous RABL3 incrementally accumulated in the vicinity of the centrosome during induction to quiescence for 6 h in RPE1 cells but not in Rabl3-2 cells (Fig. 3A, right, Fig. 3B). These RABL3-positive puncta around the CEP164-positive mother centriole partially overlapped with RAB11 signal (Fig. 3C). The colocalization between RABL3 and RAB11 was significantly increased by 6 h of serum starvation (Fig. 3D,E). In addition, a fraction of RABL3 colocalized with GM130 (also known as GOLGA2), a canonical marker of the Golgi, after 6 h of serum starvation (Fig. 3F). These results collectively indicate that RABL3 is partly located around the centrosome and the Golgi and colocalizes with RAB11 during early ciliogenesis. Furthermore, to evaluate guanine nucleotide-form-dependent localization of RABL3 during ciliogenesis, RPE1 cells expressing Myc-tagged RABL3/WT or S20N were cultured in the serum-starved medium for 6 h. Although RABL3/WT was distributed uniformly or marginally around the centrosome, RABL3/S20N was clearly concentrated at the centrosome (Fig. 3G). These observations suggest that the negative form of RABL3 is recruited to the centrosome during early ciliogenesis. We further examined guanine nucleotide-form-dependent colocalization between RABL3 and RAB11A. In line with our binding experiments (Fig. 2), the most efficient overlap was observed when RABL3/S20N and RAB11A/Q70L were co-expressed in RPE1 cells after 6 h of serum-starvation (Fig. S3), suggesting that the negative form of RABL3 links to the active form of RAB11 at the centrosome during early ciliogenesis.

We next analyzed whether RABL3 or RAB11 depletion impinges on their mutual localization and expression. Centrosomal localization and protein levels of RAB11 and RAL3 were unchanged in RABL3-mutated and RAB11-knockdown cells, respectively (Fig. S4A,B). These results suggest that the localization and expression of endogenous RABL3 and RAB11 is independent of each other during early ciliogenesis.

We attempted to reveal the role of RABL3 in the primary cilium formation process. We first examined CP110 (also known as CCP110), which localizes to both mother and daughter centrioles in cycling cells and disappears from the mother centriole during early ciliogenesis (Spektor et al., 2007). We found that cells with two CP110 dots were significantly increased in RABL3-depleted cells (Fig. 4A–C), suggesting that RABL3 is required for the CP110 removal from the mother centriole. RAB11 depletion also hampered primary ciliogenesis and the loss of CP110 dots (Figs 1F,G and 4A,B). Next, we explored GFP–Rabin8 vesicles, which accumulate around the centrosome immediately after serum withdrawal in RPE1 cells (Westlake et al., 2011). Ablation of RABL3 suppressed the accumulation of GFP–Rabin8 around the centrosome (Fig. 4D,E). RAB11 knockdown induced a similar phenotype (Fig. 4D,E), consistent with previous reports (Lu et al., 2015; Westlake et al., 2011). These results suggest that RABL3 is a prerequisite for the early steps of primary cilia formation, similar to RAB11.

As GFP–Rabin8 vesicles are thought to provide membrane components to assemble ciliary membrane structures (CV and CM) (Shakya and Westlake, 2021), we established GFP–smoothened (SMO)-expressing RPE1 cells to visualize their formation (Lu et al., 2015). Where an elongated GFP–SMO signal along with the axoneme, indicating CM, was observed in control cells, RABL3 or RAB11 depletion abolished GFP-SMO-positive foci at the centriole (Fig. 5A,B), suggesting that RABL3 and RAB11 are required for GFP–SMO-positive CV (SMO⁺ CV) formation. In contrast to RABL3 and RAB11, silencing of RAB8 allowed SMO⁺ CV formation irrespective of suppressed ciliation (Figs 1F,G and 5A,B). These data suggest that RABL3 and RAB11 play overlapping roles in CV formation, whereas RAB8 is necessary after CV formation, as described previously (Lu et al., 2015).

We further examined CV formation by using an anti-Myo-Va antibody, which is known to mark both the initial CV without GFP–SMO (Myo-Va⁺/GFP-SMO⁻ CV) and the later CV (Myo-Va⁺/GFP-SMO⁺ CV) (Wu et al., 2018). As expected, single Myo-Va-positive dots were detected adjacent to the centriole, and some of them grossly overlapped with GFP–SMO-dots when RPE1 cells were subjected to serum starvation for 6 h (Fig. 5C; Fig. S4C). The number of Myo-Va⁺ CV was significantly lower in RABL3-mutated cells than that in WT cells at 6 h of starvation, whereas this number was equivalent in both cells after 3 h of starvation (Fig. 5D). These results suggest that RABL3 is dispensable for initial Myo-Va⁺/GFP–SMO⁻ CV formation but is required for later Myo-Va⁺/GFP–SMO⁺ CV formation (Fig. S5). A similar phenotype was also induced by the silencing of RAB11 (Fig. 5E). Conversely, ectopic expression of WT or active mutants, but not negative mutants of RABL3 or RAB11A, enhanced Myo-Va⁺ CV formation (Fig. 5F). Moreover, the increased proportion of cells with Myo-Va⁺ CVs seen upon overexpression of RAB11A/QL was partly but significantly inhibited in RABL3-mutated cells (Fig. 5G). Collectively, these results suggest that the RAB11–RABL3 axis positively controls CV formation during ciliogenesis in human cells.

Of the six RABL proteins, five (RABL2A, RABL2B, RABL3, RABL4 and RABL5) are similar to typical small GTPases in size (185–266 aa in humans); in contrast, RABL6 (730 aa in humans) is a relatively large protein. Based on our findings, we suggest that all small GTPase-type RABL proteins play cilia-related functions in human cells. Whereas other small RABL proteins localize to the basal body and/or the ciliary axoneme, RABL3 shows a wider distribution throughout cells. Given that RABL3 has been shown to be involved in multiple cellular events, including KRAS signaling, lymphopoiesis, and cancer cells proliferation and migration (An et al., 2017; Ge et al., 2019; Li et al., 2010; Ma et al., 2021; Nissim et al., 2019; Pan et al., 2017; Usman et al., 2020; Xu et al., 2021; Zhang et al., 2016; Zhong et al., 2020), RABL3 might be located broadly to play distinct roles depending on target molecules, interacting proteins and/or cell types.

The RABL3–RAB11 interaction that we have demonstrated in this study is analogous to other RABL-small GTPase associations in that RABL4 (IFT27) and RABL5 (IFT22) bind to the small GTPase ARL6 (also known as BBS3), and these interactions occur during ciliary events. GTP-bound RABL4 associates with nucleotide-free or GDP-bound ARL6 in human cells (Liew et al., 2014). Both RABL4 and RABL5 stabilize ARL6 (Liew et al., 2014; Xue et al., 2020). Interestingly, RABL4 has been reported to be structurally akin to RAB11 (Bhogaraju et al., 2011). Therefore, our findings and previous reports indicate that similar GTP-bound small GTPases (RAB11 and RABL4) stabilize GDP-bound or nucleotide-free small GTPases (RABL3 and ARL6, respectively), both of which are related to ciliary functions in human cells. However, immunoblotting and immunostaining experiments for RABL3 in RAB11-depleted cells did not detect a decrease in RABL3 expression. This is likely because RAB11 stabilizes RABL3 in a more specific location and/or time-point during ciliogenesis. Our immunoprecipitation and subcellular localization analyses raise a possibility that RAB11 could contribute to the recruitment of GDP-bound or nucleotide-free RABL3 to the centrosome. As RABL3/WT but not RABL3/S20N rescued de-ciliation in RABL3-mutated cells, GTP-binding is required for RABL3 function during ciliogenesis. Therefore, we hypothesize that GDP on RABL3 is probably replaced with GTP by GEF(s) around the centrosome, and then the GTP-bound RABL3 promotes CV formation through effector protein(s). Future work should identify the GEF(s) and effector(s) of RABL3 during ciliogenesis.

Here, we might have uncovered a previously unidentified step in the CV formation pathway in human cells. RABL3 is dispensable for initial CV formation but is required for the following step, denoted ‘CV maturation’ (Fig. S5). Once the initial CV is assembled from DAVs, RABL3 is required for the subsequent CV maturation step, in which several ciliary membrane proteins, such as GFP–SMO, are probably loaded to the CV. RAB11 is also involved in CV maturation through its association with RABL3. This hypothesis agrees with a previous study in which initial Myo-Va⁺ CV formation was unaffected by depletion or overexpression of RAB11 (Wu et al., 2018). After CV maturation, RAB8 promotes CV extension, which is also activated by RAB11 via Rabin8 (Lu et al., 2015). In the present model, RAB11 in the GTP-bound state controls continuous CM formation processes through its association with RABL3 or Rabin8–RAB8. A recent study has shown that serum deprivation triggers primary cilia formation through PI3K/Akt and downstream RAB11 (Walia et al., 2019). Therefore, it is feasible that RAB11 regulates multiple steps in early ciliogenesis, including RABL3-dependent CV maturation. It will be of interest to investigate whether serum deprivation and/or the PI3K/Akt axis impact the RABL3-dependent CV maturation step. Furthermore, as RABL3-mutated zebrafish exhibit developmental phenotypes reminiscent of cilia-loss mutants (Nissim et al., 2019), future genetic studies might lead to the identification of unknown mutations in Rabl3 that cause cilia-related diseases.

HEK293T cells (from Brian D. Dynlacht, New York University School of Medicine, New York, USA) were grown in DMEM (Nacalai Tesque) supplemented with 10% calf serum (Thermo Fisher Scientific) and 100 units/ml penicillin and 100 μg/ml streptomycin (P/S; Nacalai Tesque). hTert-RPE1 (RPE1) (from Brian D. Dynlacht), GFP–SMO-expressing RPE1 cells (generated in this study) and GFP–Rabin8-expressing RPE1 (from Peter K. Jackson, Stanford University, Stanford, USA) cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS; Biosera) and P/S. Serum-starved medium contained DMEM and P/S (without FBS).

Antibodies used in this study were: rabbit anti-RABL3 [1:500 immunofluorescence (IF), 1:1000 western blotting (WB); Abcam, ab196024], mouse anti-glutamylated tubulin (GT335) (1:1000; Adipogen, AG-20B-0020), rabbit anti-ARL13B (1:1000 IF; Proteintech, 17711-1-AP), rabbit anti-GM130 (1:1000 IF; BD Biosciences, #610822), mouse anti-RAB11 (1:100 IF, 1:1000 WB; Millipore, 05-853), rabbit anti-RAB8A (1:1000 WB; Proteintech, 55296-1-AP), mouse anti-centrin (1:1000 IF; Millipore, 04-1624), rabbit anti-CP110 (1:1000 IF, from Brian D. Dynlacht; Chen et al., 2002b), goat anti-CEP164 (1:200 IF; Santa Cruz Biotechnology, sc-240226), rabbit anti-myosin-Va (1:500 IF, Novus, NBP1-92156), mouse anti-β-Actin (1:1000 WB; Santa Cruz Biotechnology, sc-47778), rabbit anti-FLAG (1:1000 WB; Sigma-Aldrich, F7425), rabbit anti-Myc (1:1000 WB; MBL, #562), rabbit anti-GST (1:1000 WB; MBL, PM013), and rabbit anti-GFP (1:1000 WB; Santa Cruz Biotechnology, sc-9996).

To generate Flag–RABL3 or Myc–RABL3, a human RABL3 fragment encoding residues 1–236 was amplified by PCR using forward primer 5′-AAGAATTCATGGCGTCCCTGGATCGGGT-3′ and reverse primer 5′-AAGTCGACTCAGTCATAATGAAGGCTCTT-3′, and sub-cloned into pCMV5-Flag or pCMV5-Myc (from Kenji Kontani and Toshiaki Katada; University of Tokyo, Tokyo, Japan). RABL3/S20N plasmid was generated by PCR-based mutagenesis using forward primer 5′-AATTCGTTAGTCCATCTCCT-3′ and reverse primer 5′-TTTCCCAACACCTGAGTCTC-3′. pEGFP-C2-hRAB11A was obtained from Kenji Kontani and Toshiaki Katada. RAB11A/S25N and Q70L constructs were generated by PCR-based mutagenesis using forward primer 5′-AATAATCTCCTGTCTCGATTTAC-3′ and reverse primer 5′-CTTTCCAACACCAGAATCTC-3′, and forward primer 5′-CTAGAGCGATATCGAGCTAT-3′ and reverse primer 5′-CCCTGCTGTGTCCCATATCT-3′, respectively. To generate Myc–RAB11A, RAB11A/S25N and RAB11A/Q70L, human (h)RAB11A fragments were subcloned into pCMV5-Myc. To generate recombinant GST–RABL3 and GST–RAB11A, hRABL3 and hRAB11A fragments were subcloned into pGEX6P1 (Promega). To generate PX459-hRABL3, annealed oligo (5′-CACCGATGACCAACGACGCAAGTTT-3′ and 5′-AAACAAACTTGCGTCGTTGGTCATC-3′) was inserted into PX459 (pSpCas9(BB)-2A-PuroV2.0) (Addgene #62988; Ran et al., 2013). The plasmid expressing Flag-RABL2B was previously described (Dateyama et al., 2019). pGEX6P1-GFP nanobody was obtained from Yohei Katoh and Kazuhisa Nakayama (Katoh et al., 2015). pEGFP-mSmo was obtained from Addgene (#25395; Chen et al., 2002a). pEGFP-C1-mRAB23 was obtained from Mitsunori Fukuda (Tohoku University, Sendai, Japan).

Plasmid transfection into HEK293T cells was performed using PEI Max (Polysciences) according to the manufacturer's instruction. Plasmid transfection into RPE1 cells was performed using ViaFect (Promega) according to the manufacturer's instruction.

siRNA oligonucleotides used in this study were siRABL3#1 (Ambion, s49874), siRABL3#2 (Ambion, s57630), siRAB8A (5′-GACAAGUUUCCAAGGAACGtt-3′, Sigma-Aldrich), siRAB8B (5′-GACAAGUGUCAAAAGAAAGtt-3′, Sigma-Aldrich), siRAB11A (5′-UGUCAGACAGACGCGAAAAtt-3′, Sigma-Aldrich), siRAB11B (5′-GCACCUGACCUAUGAGAACtt-3′, Sigma-Aldrich). The siRNA for luciferase (siLuc) was as described previously (Kobayashi et al., 2017). For RNAi, 2×10⁴ RPE1 cells were seeded in 24-well plates and cultured for 24 h. After transfection of 20 pmol siRNA using Lipofectamine RNAiMAX (Invitrogen), cells were cultured in normal medium for 24 h and subsequently incubated in serum-starved medium.

Cells were lysed using lysis buffer (50 mM Hepes-NaOH pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 0.5% NP-40, 1 mM DTT, 0.5 mM PMSF, 2 µg/ml leupeptin and 10% glycerol) at 4°C for 30 min. For immunoprecipitation, 1 mg of the resulting supernatant after centrifugation (17,000 g for 10 min) was incubated with anti-Flag agarose beads (Sigma-Aldrich) or GFP-nanobody at 4°C for 2 h. The resin was washed with lysis buffer, and the bound polypeptides were analyzed by SDS-PAGE and immunoblotting. 10 µg of the lysate was loaded in the input lane. Original immunoblots are shown in Fig. S6.

Immunofluorescence microscopy was performed as described previously (Kobayashi et al., 2017). Quantification analysis of fluorescence was performed using ImageJ (Kobayashi et al., 2020). The Pearson's correlation coefficient was determined using the Coloc 2 plug-in in ImageJ.

PX459-hRABL3 plasmid was transfected into RPE1 cells using Lipofectamine 2000 (Invitrogen). After 24 h, cells were cultured in a medium containing 10 µg/ml puromycin (Nacalai Tesque) for 3 days and then singly plated into 96-well plates. Genomic DNA was extracted from surviving cells using QuickExtract DNA Solution 1.0 (epicentre), and mutations were determined via sequencing against amplified PCR products including a guide RNA target sequence. Primers for PCR amplification and sequencing are listed in Table S1.

To generate GFP-SMO-expressing cells, RPE1 cells were transfected with pEGFP-mSmo plasmid and subsequently cultured in a medium containing 1 mg/ml G418 (Nacalai Tesque) for 12 days. GFP–SMO-expressing colonies were identified by detecting GFP fluorescence.

Escherichia coli BL21(DE3) pLysS cells (TOYOBO) harboring pGEX6P1, pGEX6P1-RABL3, or pGEX6P1-RAB11A were cultured in lysogeny broth (LB) with 2% glucose medium at 37°C until the optical density (OD) at 600 nm was 0.3. After 0.2 mM isopropyl β-D-1-thiogalactopyranoside (IPTG, Nacalai Tesque) treatment, cells were subsequently incubated at 37°C for 2 h. The cells were collected by centrifugation (6000 g for 15 min), resuspended in extraction buffer (50 mM Tris-HCl pH 8.0, 100 mM NaCl, 5 mM MgCl₂, 1 mM DTT, 10% glycerol, 0.5 mM PMSF and 2 μg/ml leupeptin), sonicated and then solubilized with Triton-X 100 to a final concentration of 0.5%. The supernatant after centrifugation at 17,000 g at 4°C for 1 h was incubated with glutathione–Sepharose 4B (GE Healthcare) at 4°C for 3 h. After washing with wash buffer (50 mM Tris-HCl pH 8.0, 100 mM NaCl, 5 mM MgCl₂, 1 mM DTT, 10% glycerol and 0.5% Triton X-100), the resin was incubated in wash buffer containing 15 mM Glutathione (Nacalai tesque) at 4°C for 20 min to obtain GST or GST–RAB11A. To obtain untagged RABL3, GST–RABL3 bound to the resin was washed with wash buffer and then incubated with wash buffer containing PreScission Protease (GE Healthcare) at 4°C for 16 h to cleave GST. The eluates containing recombinant proteins were passed through PD-10 columns (GE Healthcare) to exchange buffer with store buffer (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM MgCl₂, 1 mM DTT, 10% Glycerol, 0.1% Triton X-100). For turbidity assays, the proteins were solubilized in store buffer-2 (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM MgCl₂ and 1 mM DTT).

E. coli BL21(DE3) pLysS cells harboring pGEX6P1-GFP nanobody were cultured in LB with 2% glucose at 37°C until the OD at 600 nm was 0.5. After 0.1 mM IPTG treatment, cells were subsequently incubated at 20°C for 20 h. The cells were collected by centrifugation (6000 g for 15 min), lysed and purified using Glutathione-Sepharose 4B as described previously (Katoh et al., 2015).

0.5 µM RABL3 and 0.05 µM GST or GST–RAB11A were mixed with glutathione–Sepharose 4B in the presence of 10 µM GTPγS (Sigma-Aldrich) or 10 µM GDP (Sigma-Aldrich) in store buffer. The mixture was incubated at 25°C for 1 h, and then rotated at 4°C for 2 h. The resin was washed with store buffer, and the bound polypeptides were analyzed using SDS-PAGE and immunoblotting.

To prepare nucleotide-loaded (Nuc-) proteins, 1.25 µM RABL3 or GST–RAB11A was incubated with 10 µM GTPγS or GDP in EDTA buffer (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 0.5 mM MgCl₂, 5 mM EDTA and 1 mM DTT) at 25°C for 1.5 h. After addition of MgCl₂ to a final concentration of 10 mM, the mixture was further incubated at 25°C for 1 h and loaded onto NAP-5 columns (GE Healthcare) to remove free nucleotide. Nuc-proteins were eluted with Mg buffer (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 10 mM MgCl₂ and 1 mM DTT). Next, 0.1 µM nuc-RABL3 and nuc-GST-RAB11A were incubated with glutathione–Sepharose 4B in Mg buffer-2 (Mg buffer containing 0.1% Triton X-100 and 10% glycerol) at 4°C for 2 h. The resin was washed with Mg buffer-2, and the bound polypeptides were analyzed using SDS-PAGE and immunoblotting.

20 µM RABL3 and 20 µM GST or GST–RAB11 were mixed on ice with store buffer-2. The mixture was then incubated at 37°C, and OD was measured at 350 nm using a Nanodrop 2000c (Thermo Fisher Scientific) every 3 min.

The nucleotide exchange assay was basically performed as reported previously (Kanie and Jackson, 2018). First, 20 µM RABL3 was incubated with 0.4 mM MANT-GDP (Sigma-Aldrich) in EDTA buffer at 25°C for 1.5 h. After addition of MgCl₂ to a final concentration of 10 mM, the mixture was further incubated at 25°C for 1 h and then loaded onto a NAP-5 column. RABL3–MANT-GDP was eluted using Mg buffer. Then, 0.64 µM RABL3–MANT-GDP was mixed with 100 µM GTPγS and 10 µM GST or GST-RAB11A in Mg buffer. The fluorescence was measured every 20 s using a ARVO MX (PerkinElmer).

Immunoprecipitation followed by protein identification using mass spectrometry was performed as described previously (Kobayashi et al., 2011). Briefly, Flag–RABL3 or Flag–RABL2B was expressed in HEK293T cells and immunoprecipitated with anti-Flag agarose beads at 4°C for 3 h. Bound proteins were eluted by incubating with Flag peptide (Sigma-Aldrich) for 30 min, and the resultant eluates were TCA precipitated and separated by SDS-PAGE followed by Coomassie staining. The Coomassie-stained gel was subjected to mass spectrometric analysis performed at the NAIST mass spectrometric laboratory.

The statistical significance of the difference was determined using an unpaired two-tailed Student's t-test or χ-squared test (Fig. 5B). Differences were considered significant when P<0.05. **P<0.01; *P<0.05, ^##P<0.01; #P<0.05.

We thank B. D. Dynlacht (New York University) for anti-CP110 antibody, hTert RPE1, and HEK293 T cells; P. K. Jackson (Stanford University) for GFP-Rabin8-RPE1 cells; Y. Katoh and K. Nakayama (Kyoto University) for pGEX6P1-GFP nanobody; T. Katada and K. Kontani (University of Tokyo) for pCMV5-Flag, pCMV5-Myc, and pEGFP-C2-hRAB11A; M. Fukuda (Tohoku University) for pEGFP-C1-mRAB23. We thank Y. Fukao and R. Kurata (NAIST) for mass spectrometric analysis.