Recycling endosomes are stations that sort endocytic cargoes to their appropriate destinations. Tubular endosomes have been characterized as a recycling endosomal compartment for clathrin-independent cargoes. However, the molecular mechanism by which tubular endosome formation is regulated is poorly understood. In this study, we identified Rab10 as a novel protein localized at tubular endosomes by using a comprehensive localization screen of EGFP-tagged Rab small GTPases. Knockout of Rab10 completely abolished tubular endosomal structures in HeLaM cells. We also identified kinesin motors KIF13A and KIF13B as novel Rab10-interacting proteins by means of in silico screening. The results of this study demonstrated that both the Rab10-binding homology domain and the motor domain of KIF13A are required for Rab10-positive tubular endosome formation. Our findings provide insight into the mechanism by which the Rab10–KIF13A (or KIF13B) complex regulates tubular endosome formation.
Higher eukaryotic cells contain several functionally distinct endosomal compartments, including an early endosome (EE) and a recycling endosome (RE). Their endosomal compartments sort internalized proteins according to their destinations, for example, to lysosomes, the plasma membrane (PM) and the trans-Golgi network (TGN). REs were originally characterized as the vesicles mediating the recycling of internalized cargoes, such as the transferrin receptor (TfR), from EEs to the PM (Maxfield and McGraw, 2004). The endosomal sorting machinery is essential for maintaining the homeostasis that underlies fundamental cellular functions such as intracellular signaling, cell migration and polarity formation (Doherty and McMahon, 2009; Scita and Di Fiore, 2010).
Recent studies have demonstrated the presence of another distinct RE compartment that is characterized by striking tubular-shaped structures (Maldonado-Báez et al., 2013; Sabharanjak et al., 2002). The tubular recycling endosomes, called tubular endosomes, have been shown to be involved in the clathrin-independent endocytic (CIE) pathway and not in the clathrin-mediated endocytic (CME) pathway (Grant and Donaldson, 2009; Mayor and Pagano, 2007). Tubular endosomes are thought to regulate the sorting and trafficking of CIE cargoes such as CD147 and not CME cargoes such as TfR (Eyster et al., 2009; Hattula et al., 2006). However, the significance of their tubular structure remains to be determined, in part because of lack of information about the molecular mechanism of tubular endosome formation.
The Rab family of small GTPases are key regulators of membrane traffic and comprise ∼60 members in mammals. Since each Rab is thought to regulate different intracellular membrane trafficking pathways, that is, the secretory pathway, the endocytic pathway and the recycling pathway, for CME cargoes (Barr, 2013; Fukuda, 2008; Pfeffer, 2013; Stenmark, 2009), it seems reasonable to expect that a set of Rabs also functions in the formation of tubular endosomes. Actually, several Rab family members, namely, Rab8A, Rab11A, Rab22A and Rab35, have been reported to regulate tubular endosome formation (Rahajeng et al., 2012; Sharma et al., 2009; Solis et al., 2013; Weigert et al., 2004). However, the full repertoire of tubular-endosome-resident Rabs and their roles in tubular endosomes remain to be determined.
In this study, we performed a comprehensive localization screening of the Rab subfamily and succeeded in identifying Rab10 as a novel protein predominantly localized at tubular endosomes. Intriguingly, knockout of Rab10 completely disrupted the formation of tubular endosomes. We also identified kinesin motors KIF13A and KIF13B (hereafter denoted KIF13A/B) as novel Rab10-interacting proteins by means of in silico screening and found that their knockout also disrupts tubular endosome formation. Our findings suggest that the Rab10–KIF13 complex regulates the formation of tubular endosomes through its motor activity along microtubule tracks.
Comprehensive localization screening for EGFP-tagged Rabs identified Rab8 and Rab10 as specific markers for tubular endosomes
To prepare a list of Rabs that specifically localize at tubular endosomes, we established a library of HeLaM cells that stably express EGFP-tagged Rab subfamily members (Rab1A–Rab43) and performed a localization screening. First, we defined a tubular endosome as a tubular structure >20 µm in length and manually counted the number of cells containing at least one EGFP-positive tubular endosome (Fig. 1A,B, blue bars). Second, we immunostained HeLaM cells with antibody against MICAL-L1, a well-known tubular endosome marker (Sharma et al., 2009) and then calculated Pearson's correlation coefficient (PCC) values for the relationship between EGFP-tagged Rabs and MICAL-L1 (Fig. 1A,B, orange bars). The results showed that ten of the Rab subfamily members tested (Rab5A, Rab8A, Rab10, Rab11A, Rab13, Rab17, Rab21, Rab22A, Rab23 and Rab35), were localized at tubular endosomes, but that neither EGFP alone nor EGFP–Rab1A was associated with tubular endosomes (Fig. 1B, blue bars; Fig. S1A). It should be noted that tubules positive for eight Rabs (Rab8A, Rab10, Rab11A, Rab13, Rab17, Rab22A, Rab23 and Rab35) were colocalized with MICAL-L1, but that Rab5A-positive tubules and Rab21-positive tubules were not (Fig. S1A), suggesting the presence of subdomains in the tubular endosomes. Because these tubular-endosome-resident Rabs are known to localize at REs, EEs and/or the PM (Babbey et al., 2006; Chavrier et al., 1990; Evans et al., 2003; Hunziker and Peters, 1998; Kauppi et al., 2002; Mrozowska and Fukuda, 2016; Peränen, 2011; Simpson, 2004; Ullrich et al., 1996; Yamamura et al., 2008), tubular endosomes are likely to be part of recycling pathways, as described previously (Maldonado-Báez et al., 2013). Consistent with previous reports (Hattula et al., 2006; Solis et al., 2013; Weigert et al., 2004), tubular localization of Rab8A, Rab11A and Rab22A was also observed, thereby validating the results of our screenings. Two of the tubular-endosome-resident Rabs we identified, Rab8A and Rab10, showed higher values with respect to the percentage of cells with Rab-positive tubular endosomes (more than 80%) and to the PCC value for the relationship between the EGFP-tagged form and MICAL-L1 (more than 0.4) (Fig. 1B), and we therefore decided to focus on them for further analysis.
Because Rab10 had previously been shown to be localized at REs, the TGN or endoplasmic reticulum (ER) in other cell types (Babbey et al., 2006; English and Voeltz, 2013; Liu et al., 2013), we investigated the subcellular localization of endogenous Rab8A and Rab8B (hereafter denoted Rab8A/B) and Rab10 in HeLaM cells. The same as EGFP–Rab8A and –Rab10, we found that endogenous Rab8A/B and Rab10 were well colocalized with MICAL-L1 at tubular endosomes (Figs 1C,D, 2C; Fig. S1A,B). We also compared the localization of EGFP–Rab10 with several organelle markers to determine its subcellular localization. The results showed that EGFP–Rab10 was significantly colocalized with MICAL-L1 (PCC=0.51±0.04; mean±s.e.m.) as compared with TGN46 (also known as TGOLN2; a TGN marker; PCC=0.32±0.03), GM130 (also known as GOLGA2; a cis-Golgi marker; PCC=0.26±0.04), TfR (an RE marker; PCC=0.04±0.03), EEA1 (an EE marker; PCC=0.13±0.04), LC3 proteins (autophagosome markers; PCC=0.05±0.05), LAMP1 (a lysosome marker; PCC=0.04±0.01), lysobisphosphatidic acid [LBPA; a late endosome (LE) marker; PCC=0.04±0.03] or RTN4 (an ER marker; PCC=−0.03±0.04) (Fig. S2A,B), confirming that Rab10 mainly marks tubular endosomes rather than conventional TfR-positive REs in HeLaM cells. Although Rab10 was partially colocalized with TGN46, brefeldin A treatment, which disrupts the Golgi complex, did not affect the EGFP–Rab10-positive tubules (Fig. S2C,E), but it resulted in a decrease in the PCC for the relationship between EGFP–Rab10 and TGN (PCC=0.04±0.03) (Fig. S2D). These results suggest that EGFP–Rab10-positive tubules are distinct from the tubular carriers in the TGN, which is a part of the pathway for newly synthesized proteins (Chen et al., 2017). The discrepancy between the Rab10 localization results obtained in this study (HeLaM cells) and previous studies (other cell types) may be explained by Rab10 localization being cell or tissue specific.
To determine whether Rab10-positive tubules preferentially transport CIE cargoes rather than CME cargoes, we compared EGFP–Rab10 localization with the localization of Tf and CD147 (also known as BSG), which are internalized into the RE via the CME pathway and the CIE pathway, respectively. To do so, we observed HeLaM cells stably expressing EGFP–Rab10 that had internalized both Alexa Fluor 594-conjugated Tf and anti-CD147 antibody. The results showed that EGFP–Rab10 clearly colocalized with the internalized CD147 at tubular endosomes and did not colocalize with the internalized Tf (Fig. 1E,F). These results are consistent with previous reports that tubular endosomes contain CIE cargoes and not CME cargoes (Sabharanjak et al., 2002; Weigert et al., 2004).
Rab10 is required for tubular endosome formation and its dynamics depends on microtubules
To determine whether Rab8 and Rab10 are required for tubular endosome formation and reveal their relationships, we generated Rab8A and Rab8B double-knockout (Rab8-DKO) and Rab10-knockout (Rab10-KO) cell lines by using the CRISPR/Cas9 system (Fig. 2A,B) and examined the Rab8-DKO cells for the presence of EGFP–Rab10-positive tubules and the Rab10-KO cells for the presence of EGFP–Rab8-positive tubules. EGFP–Rab10-positive tubules were still observed in the Rab8-DKO cells, and their overall localization was unaltered even in the Rab8-DKO cells stably re-expressing Myc–Rab8A (Fig. 2C,E). These results are consistent with the previous finding that Rab8 itself is dispensable for tubular endosome formation (Nakajo et al., 2016; Sharma et al., 2009). By contrast, Rab10 KO resulted in complete dispersion of both the Rab8- and MICAL-L1-positive tubules (Fig. 2D, top panels; Fig. S3A), and re-expression of Rab10 in Rab10 KO cells clearly rescued this phenotype (Fig. 2D, bottom panels, Fig. 2F,G). On the other hand, Rab10 KO had no effect on the level of Rab8 expression or MICAL-L1 expression (Fig. 2B), indicating that Rab10 KO causes disruption of both tubules without affecting the levels of expression of tubular components. To quantitatively evaluate the difference between Rab10-KO cells and Rab10-KO cells stably re-expressing Rab10, we measured the total length of MICAL-L1-positive tubules in each cell. The results showed a significant difference between the population profile of total tubule length of Rab10-KO cells and Rab10-KO cells stably re-expressing cells (Fig. 2H). Intriguingly, other tubular-endosome-resident Rabs, namely, EGFP-tagged Rab5A, Rab11A, Rab13 and Rab22A (Fig. 1B), as well as Rab8A were not localized at the tubular endosomes in Rab10-KO cells (Fig. S3A), suggesting that Rab10 is an essential constituent of tubular endosomes. To determine the hierarchical relationship between Rab10 and Rab22A, a regulator of tubular endosomes (Weigert et al., 2004), we generated a Rab22A and Rab22B double-knockout (Rab22-DKO) cell line (Fig. S3B) and examined the effects of Rab22-DKO on EGFP–Rab10 localization. The results showed that Rab22A/B are required for the formation of EGFP–Rab10-positive tubules (Fig. S3C,D), although Rab10 is also necessary for the formation of EGFP–Rab22A-positive tubules (Fig. S3A). These results indicate that Rab10 and Rab22A/B cooperatively regulate tubular endosome formation.
To evaluate tubular endosome dynamics, we performed live-cell time-lapse imaging of HeLaM cells stably expressing EGFP–Rab10. Examination of the images revealed that EGFP–Rab10-positive tubules are highly dynamic structures that frequently extend, branch and retract (Fig. 3A; Movie 1). Previous studies have shown that actin filaments, not microtubules, are required for formation of Rab10-binding protein EHBP1-positive tubules in the intestinal epithelium of C. elegans (Wang et al., 2016), whereas microtubules, not actin filaments, are required for the formation of Rab11A- and Rab22A-positive tubules in HeLa cells (Maldonado-Báez et al., 2013; Solis et al., 2013). To determine whether the formation of Rab10-positive tubules in HeLaM cells depends on actin filaments or microtubules, we treated the cells with cytochalasin D and nocodazole, which are inhibitors of actin polymerization and microtubule polymerization, respectively. Cytochalasin D treatment disrupted actin filaments, but dynamic tubular endosomes were still observed (Fig. 3B,C; Movie 2), whereas nocodazole treatment dispersed the EGFP–Rab10-positive tubules into the cytoplasm (Fig. 3B,C; Movie 3). Consistent with these results, EGFP–Rab10-positive tubules were well colocalized with monomeric Strawberry (mStr)–EMTB (Fig. 3D, arrowheads), which is an ensconsin microtubule-binding domain that can be used as a microtubule probe (Faire et al., 1999). Moreover, EGFP–EMTB localization in Rab10-KO cells seemed unaltered in comparison with its localization in parental cells or Rab10-KO cells stably re-expressing Rab10 (data not shown), indicating that the disruption of tubular endosomes in Rab10-KO cells is not caused by microtubule depolymerization. These results, taken together, indicate that tubular endosomes are formed in a microtubule-dependent manner.
KIF13A/B are novel Rab10-interacting proteins
Since Rab proteins are generally thought to regulate membrane trafficking through the functions of their effectors, we focused our attention on the effectors of Rab10 as a means of gaining insight into the molecular mechanism by which Rab10 regulates tubular endosome formation (Stenmark, 2009). Because microtubules are required for tubular endosome formation (Fig. 3B,C), we hypothesized that microtubule-associated Rab10 effectors are involved in the process. We initially investigated previously reported Rab10 effectors, including proteins containing a bivalent MICAL/EHBP Rab-binding domain (MICAL1, MICAL3, MICAL-cl, MICAL-L1, MICAL-L2, EHBP1, EHBP1L1 and C16orf45) (Rai et al., 2016), myosin-Va/b (Chen et al., 2012; Liu et al., 2013), JIP1 (Deng et al., 2014), Sec16A (Bruno et al., 2016), Evi5 (Fukuda et al., 2008) and Lgl1 (Wang et al., 2011), but we were unable to identify any good candidates, because siRNA-mediated knockdown of these Rab10 effectors had no effect on the formation of EGFP–Rab10-positive tubules (data not shown).
To identify a novel Rab10 effector(s) that associates with microtubules, we performed in silico screening. We performed DELTA-BLAST searches by using the known Rab10-binding domain (RBD) of MICAL1 or EHBP1 (Rai et al., 2016) as bait, followed by PSI-BLAST, and then selected microtubule-associated proteins among the candidate Rab10 effectors. By so doing we succeeded in obtaining a list of proteins containing an RBD homology domain (RHD), and identified the molecular motors KIF13A/B as candidates for novel Rab10 effectors that interact with microtubules (Table S1; Fig. 4A).
KIF13A/B belong to the kinesin-3 family, which is evolutionarily conserved in higher eukaryotes and whose members function as organelle transporters in the dimer state (Miki et al., 2005; Soppina et al., 2014). The sequential BLAST search showed that the amino acid (aa) sequence aa1102–1146 of KIF13A and sequence aa1099–1143 of KIF13B are similar to both RBDs of MICAL1 and EHBP1 (Fig. 4A, orange lines, 4B). To structurally compare the RHD of KIF13A and RBD of MICAL1, we performed 3D-homology modeling of the RHD based on the crystal structure of the Rab10–MICAL1 complex (PDB 5LPN) by using MODELLER software (Šali and Blundell, 1993). As expected, the RHD of KIF13A overlapped the Rab10-binding region of the RBD of MICAL1 (Fig. 4C). To confirm the interaction between Rab10 and KIF13A, we transiently co-expressed either the EGFP-tagged KIF13A or KIF13B tail fragment, which contains the RHD (Fig. 4D), and mStr–Rab10 in COS-7 cells and evaluated their interaction by immunoprecipitation using glutathione–Sepharose beads coupled to a GST-fused GFP nanobody. The results showed that EGFP-tagged KIF13A/B, not EGFP alone, interacted with mStr–Rab10 (Fig. 4E, top panel). We also confirmed the direct interaction between Rab10 and the KIF13A fragment containing the RHD by using the purified components (Fig. S4A), indicating that KIF13A/B are novel Rab10-interacting proteins. We further investigated whether KIF13A preferentially interacts with the GTP-bound form of Rab10. To our surprise, however, KIF13A interacted with both the GTP-bound form and GDP-bound form of Rab10 in vitro (Fig. S4B). By contrast, a constitutively active Rab10(Q68L) mutant colocalized with a KIF13A tail at endosomes in COS-7 cells, whereas a constitutively negative Rab10(T23N) mutant was not (Fig. S4C), suggesting that KIF13A preferentially interacts with the GTP-bound form of Rab10 in living cells.
Next, we investigated the subcellular localization of Myc–KIF13A in HeLaM cells. Consistent with the results of the co-immunoprecipitation assays (Fig. 4E), Myc–KIF13A was strongly colocalized with EGFP–Rab10 at tubular endosomes (Fig. 5A,B). By contrast, a KIF13A(ΔRHD) mutant that lacks an RHD (Fig. 5C), had lost the ability to localize at Rab10-positive tubules (Fig. 5A,B,D). Similarly, the wild-type KIF13B tail [EGFP–KIF13B-tail(WT)], but not the ΔRHD mutant, was colocalized with mStr–Rab10 at tubular endosomes (Fig. S4D,E). To determine whether the tubular endosome localization of KIF13A depends on Rab10, we investigated its localization in Rab10-KO cells. As expected, the same as the KIF13A(ΔRHD) mutant in the parental cells, Myc–KIF13A in the Rab10-KO cells did not localize at any tubular structures at all (Fig. 5E,F), although the Rab10-KO itself did not affect the level of KIF13A/B expression (Fig. 5G). These results suggest that KIF13A/B are localized at tubular endosomes through their Rab10 interaction.
Both the RHD domain and the motor domain of KIF13A are required for tubular endosome formation
Previous studies have shown that certain kinesin motors together with microtubules contribute to organelle tubulation through their motor activity along microtubule tracks (Du et al., 2016; Zhou et al., 2013). We therefore hypothesized that KIF13A/B regulate the tubulation of Rab10-resident endosomes through their motor activity. To test this hypothesis, we established KIF13A-KO, KIF13B-KO and KIF13A/B-DKO cell lines (Fig. 6A) and examined the tubular endosomes in these cells. The results showed that neither KIF13A-KO cells nor KIF13A/B-DKO cells contained MICAL-L1-positive tubules (Fig. 6B,C), the same as Rab10-KO cells, which also lacked MICAL-L1-positive tubules (Fig. 2D). There were also fewer MICAL-L1-positive tubules in the KIF13B-KO cells, but the phenotype was much less severe than that of the KIF13A-KO cells (Fig. 6C), indicating that KIF13A plays the predominant role in tubular endosome formation. To further investigate the functional redundancy of KIF13A/B, we evaluated the effect of KIF13B overexpression on tubular endosomes in KIF13A-KO cells. The results showed that KIF13B overexpression in KIF13A-KO cells rescued the dispersion of tubular endosomes, the same as KIF13A overexpression does (Fig. S5), suggesting that KIF13A/B have redundant functions in tubular endosome formation. The level of KIF13B expression in HeLaM cells may be much lower than that of KIF13A, or KIF13A may have greater ability to facilitate tubule formation than KIF13B does, even if their expression level in HeLaM cells is similar.
To further define the importance of the RHD and motor activity of KIF13A in tubular endosome formation, we re-expressed KIF13A mutants (ΔRHD and tail, which lacks an N-terminal motor domain; see also Figs 4D and 5C, respectively) in KIF13A-KO cells. EGFP–Rab10 was not localized at tubular endosomes in KIF13A-KO cells, and re-expression of KIF13A(WT) clearly rescued this phenotype (Fig. 6D,F). By contrast, re-expression of a KIF13A mutant (tail or ΔRHD) did not restore EGFP–Rab10-positive tubular endosomes (Fig. 6D,F), even though equivalent amounts of proteins were expressed under our experimental conditions (Fig. 6E). We, therefore, concluded that both the RHD and the motor domain of KIF13A are required for tubular endosome formation in HeLaM cells.
In the present study, we performed a comprehensive localization screening for Rabs that are localized on tubular endosomes and succeeded in identifying ten Rabs, including six Rab proteins that had not been previously seen at this location. One of them, Rab10, and its interactors KIF13A/B were found to be essential factors for the formation of tubular endosomes in HeLaM cells. Because Rab10-positive tubular endosomes are sensitive to nocodazole (Fig. 3B) and likely depend on KIF13A/B motor activity (Fig. 6), we propose the following model for tubular endosome formation (Fig. 7A): Rab10 recruits KIF13A/B to recycling (or early) endosomes, and the Rab10–KIF13 complex together with microtubules facilitates tubulation through KIF13 motor activity. Although several kinesin-3 family members (e.g. KIF1) are known to interact with organelles through their lipid-binding domains, for example, a pleckstrin homology (PH) domain (Soppina et al., 2014), KIF13A/B lack such a domain. Thus, Rab10 is likely to function as a molecular scaffold for the recruitment of KIF13A/B to tubular endosomes. Moreover, the motor activity of KIF13A has been shown to mediate vesicle tubulation in vitro (Zhou et al., 2013), a finding that supports the validity of our model.
How do the Rabs that localize at tubular endosomes contribute to their formation or functions? Based on the results of previous studies on the functions of tubular endosomal Rabs taken together with our own findings in the present study, we propose a model that would explain how these Rabs coordinate tubular endosome formation (Fig. 7B). Since tubular endosome formation has previously been shown to be involved in the Arf6-dependent endocytic pathway for CIE cargoes (Caplan et al., 2002; Radhakrishna and Donaldson, 1997), it seems reasonable to expect the origin and the fate of tubular endosomes or CIE cargoes to be as follows: formation of CIE-cargo-carrying vesicles from the PM (step 1); vesicle-to-tubular endosome transition (step 2); CIE cargo sorting/transport to the PM, other RE compartments (e.g. Tf-positive RE) or the TGN (Step 3) (Fig. 7B).
Our Rab localization screen identified novel MICAL-L1-positive tubular-endosome-resident Rabs (Rab10, Rab13, Rab17 and Rab23) in addition to the known Rabs (Rab8A, Rab11A, Rab22A and Rab35). Rab13 and Rab35 are likely to participate in step 1, because they are mainly localized at the PM (Chaineau et al., 2013; Sakane et al., 2012) and regulate certain types of Arf6-dependent endocytosis (Condon et al., 2018; Dutta and Donaldson, 2015). In step 2, EE-resident Rab22A, in addition to Rab10, is indispensable for tubular endosome formation (Fig. S3D). Rab22A may have a similar function to Rab10, because an interaction between Rab22A and KIF13A has been reported during the course of preparing this manuscript for publication (Shakya et al., 2018). However, since disruption of tubular endosomes in Rab10-KO cells and Rab22-DKO cells was not rescued by overexpression of Rab22A and Rab10, respectively (Fig. S3A,C), Rab10 and Rab22A would be expected to have different functions in tubular endosome formation. In step 3, membrane scission must occur before CIE cargo sorting/transport to other organelles (Fig. 3A). We suggest that the Rab8 proteins are the most likely candidates for the regulator in this process, because they are known to form a complex with EHBP1L1–Bin1–dynamin and regulate membrane scission (Nakajo et al., 2016). Rab11 and Rab17 may also be involved in step 3 (or step 2), because they are mainly localized at Tf-positive REs and regulate its trafficking (Ullrich et al., 1996; Zacchi et al., 1998). It should be noted that Rab11A also interacts with KIF13A through the C-terminal region, which does not contain the RHD region (Delevoye et al., 2014), although Rab11A itself is not necessary for tubular endosome formation (Solis et al., 2013; Weigert et al., 2004). We therefore speculate that KIF13A initially interacts with Rab10 in the process of tubular endosome formation and then interacts with Rab11A in the process of CIE cargo sorting/transport from tubular endosomes. Further studies will be necessary to fully understand the functional relationships or hierarchy of tubular-endosome-resident Rabs in tubular endosome formation and function.
What is the physiological role of Rab10-positive tubular endosomes? There are two tasks that must be accomplished to answer this question: one task is compiling the entire list of tubular endosome cargoes and the other task consists of identifying specific cells that contain tubular endosomes in vivo. A certain acidic amino acid cluster of CIE cargoes has already been shown to be a sufficient sorting signal for tubular endosome localization (Gong et al., 2007; Maldonado-Báez et al., 2013), and, intriguingly, forward genetic screening by the gene-trap method has identified Rab10 as a candidate involved in the trafficking of the acidic-cluster-containing proteins (Navarro Negredo et al., 2017). Thus, it is plausible to consider the acidic-cluster-containing proteins good candidates for the cargoes of Rab10-positive tubular endosomes. On the other hand, the in vivo role of Rab10-positive tubular endosomes is poorly understood, because Rab10-KO mice exhibit embryonic lethality (Lv et al., 2015) and specific cells or tissues containing tubular endosomes have not been identified. Thus, a future detailed analysis of in vivo Rab10 distribution/localization and Rab10 conditional KO mice in various cells or tissues (Vazirani et al., 2016; Zhang et al., 2017) is needed. Identification of acidic-cluster-containing CIE cargoes and specific cells containing Rab10-positive tubular endosomes will be necessary to elucidate the physiological significance of the formation and function of tubular endosomes.
MATERIALS AND METHODS
The anti-Rab22A rabbit polyclonal antibody was raised against GST–Rab22A (Itoh et al., 2008) as an antigen and purified as described previously (Fukuda and Mikoshiba, 1999). The following commercially available antibodies were used in this study (Table S2): anti-human Lamp1 (clone H4A3, sc-20011) and anti-c-Myc (9E10, sc-40) mouse monoclonal antibodies from Santa Cruz Biotechnology (Dallas, TX); HRP-conjugated FLAG-tag mouse monoclonal antibody (M2, A8592) from Sigma-Aldrich (St. Louis, MO); anti-β-actin mouse monoclonal antibody (G043) from Applied Biological Materials (Richmond, BC, Canada); anti-Rab8 (610845; reacting with both Rab8A and Rab8B) and anti-GM130 mouse monoclonal antibodies (35/GM130, 610822) from BD Biosciences (San Jose, CA); anti-Rab10 (clone D36C4, 8127), anti-Rab8 (clone D22D8, 6975; for immunofluorescence) and anti-EEA1 rabbit monoclonal antibodies (clone C45B10, 3288) from Cell Signaling Technology (Danvers, MA); anti-Rtn4/NogoA rabbit monoclonal antibody (AHP1799) from Bio-Rad (Hercules, CA); anti-LBPA mouse monoclonal antibody (clone 6C4, Z-PLBPA) from Echelon Biosciences (Salt Lake City, UT); anti-LC3 (PM036) and anti-GFP (598) rabbit polyclonal antibodies from MBL (Woburn, MA); anti-MICAL-L1 mouse polyclonal antibody (H00085377-B01P) from Abnova (Taipei, Taiwan); anti-KIF13A rabbit polyclonal antibody (A301-077A) from Bethyl Laboratories (Montgomery, TX); anti-CD147 mouse monoclonal antibody (clone HIM6, 306202) from Biolegend (San Diego, CA); anti-KIF13B rabbit polyclonal antibody (NBP1-83398) and anti-Rab22B/31 mouse monoclonal antibody (1C6, H00011031-M01) from Novus Biologicals (Littleton, CO); anti-RFP rabbit polyclonal antibody (600-401-379) from Rockland (Gilbertsville, PA); anti-TGN46 rabbit polyclonal antibody (ab50595) from Abcam (Cambridge, MA); anti-HA rat polyclonal antibody (3F10, 11867423001) from Roche Diagnostics (Indianapolis, IN); anti-TfR mouse monoclonal antibody (clone H68.4, 13-6800) and Alexa Fluor-labeled secondary antibodies from Thermo Fisher Scientific (Waltham, MA).
The pMRX-IRES-puro retroviral vector was a kind gift from Dr Shoji Yamaoka (Tokyo Medical and Dental University, Japan) (Saitoh et al., 2003). cDNAs encoding the mouse Rab proteins (1A, 2A, 3A, 4A, 5A, 6A, 7A, 7B/42, 8A, 9A, 10, 11A, 12, 13, 14, 15, 17, 18, 19, 20, 21, 22A, 23, 24, 25, 26, 27A, 28, 29, 30, 32, 33A, 34, 35, 36, 37, 38, 39A, 40C, 42/43 and 43/41) (Itoh et al., 2006) were subcloned into the pMRX-puro-EGFP vector. cDNAs encoding Rab10, Rab8A and Rab22A were subcloned into the pMRX-bsr vector, the pMRX-bsr-Myc vector, pmStr-C1 vector (Ohbayashi et al., 2012) and/or pEF-FLAG vector (Fukuda et al., 1999). cDNA encoding EMTB (aa153–419 of human MAP7, transcript variant 4) was amplified from Marathon-Ready human testis cDNA (Clontech-Takara Bio, Shiga, Japan) by PCR with the following pair of oligonucleotides: 5′-GTCCGGACTCGGATCCGCAGTGCGAAGCGAAACAG-3′ (forward primer; homologous sequence with the 3′ end of the linearized vector underlined) and 5′-ATTTACGTAGCGGCCCTAAGAGCCCTCAGGTGGTGTT-3′ (reverse primer; homologous sequence with the 5′ end of the linearized vector underlined), and subcloned into the pMRX-puro-EGFP vector and pMRX-bsr-mStr vector with an In-Fusion HD cloning kit (Clontech-Takara Bio). cDNA encoding LifeAct-mCherry (kindly provided by Dr Kazumasa Ohashi, Tohoku University, Japan) was subcloned into the pMRX-bsr vector. cDNAs encoding the tail domains of mouse KIF13A and KIF13B were prepared as described previously (Ishida et al., 2015). cDNAs encoding the motor domains of mouse KIF13A and KIF13B were amplified from Marathon-Ready mouse brain and testis cDNAs (Clontech-Takara Bio) by PCR with the following pairs of oligonucleotides: 5′-GGATCCATGTCGGATACGAAGGTAAA-3′ (Kif13A motor forward primer; BamHI site underlined) and 5′-CCGCGGAACTGGTTTCTCTCTCAA-3′ (Kif13A motor reverse primer); 5′-AGATCTATGGGAGACTCCAAAGTGAA-3′ (Kif13B motor forward primer; BglII site underlined) and 5′-CAGATCTCGGATAATCCGAG-3′ (Kif13B motor reverse primer). cDNAs encoding the full-length of mouse KIF13A and KIF13B were prepared by combining their motor domain and tail domain constructs, and subcloned into the pMRX-puro HA vector or pMRX-bsr Myc vector. cDNAs encoding mouse KIF13A/B (ΔRHD) mutants were prepared by inverse PCR using a KIF13A/B-harboring vector as a template and the following pairs of oligonucleotides: 5′-CCGATGCAGTGCTGGTGCCCGCCCCTGGCAGCG-3′ (KIF13A ΔRHD forward primer) and 5′-CCAGCACTGCATCGGACCATCTCTCTCTTACA-3′ (KIF13A ΔRHD reverse primer); 5′-TCAATGCAGTGATGGTGCCTTCTGCTGGGAGTG-3′ (KIF13B ΔRHD forward primer) and 5′-CCATCACTGCATTGAGCCATTTTCTACGAAGC-3′ (KIF13B ΔRHD reverse primer). cDNA encoding mouse KIF13A (aa1003–1266) was similarly prepared by PCR using the following pairs of oligonucleotides: 5′-AAGGATCCGATTATGCCGCTGTGGAGCT-3′ (KIF13A AA1003 forward primer; BamHI site underlined) and 5′-TTGTCGACTTACATGGCGGCCGGGTGGCTGA-3′ (KIF13A aa1266 reverse primer; SalI site underlined). cDNA encoding mouse KIF13A (aa1003–1266) was subcloned into the pEF-T7-GST vector (Fukuda et al., 2002). pEGFP-C1-Rab10(T23N) and -Rab10(Q68L) were prepared as described previously (Homma and Fukuda, 2016). pGEX-6P-1-GFP-nanobody (Katoh et al., 2015) was kindly provided by Dr Kazuhisa Nakayama (Kyoto University, Japan).
Cell culture, transfection, infection and drug treatment
HeLaM cells, COS-7 cells and Plat-E cells (a kind gift from Dr Toshio Kitamura, The University of Tokyo, Japan) were grown at 37°C in Dulbecco's modified Eagle medium (D-MEM) (044-29765; FUJIFILM Wako Pure Chemical, Osaka, Japan) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin G and 100 µg/ml streptomycin in a 5% CO2 incubator. For live-cell imaging, D-MEM with HEPES (044-32955; FUJIFILM Wako Pure Chemical) was used instead of D-MEM with Phenol Red. Transfection of plasmids into HeLaM cells and Plat-E cells was performed by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. To establish stable cell lines, retrovirus production in Plat-E cells and retrovirus infection were performed as described previously (Morita et al., 2000). Plasmid mixtures containing pMRX vector and pLP/VSVG (Thermo Fisher Scientific, MA) were transfected into Plat-E cells and the culture medium containing the desired retroviruses was collected. HeLaM cells were infected by using the retrovirus-containing medium and 5 µg/ml polybrene. Stable cell lines were selected by using puromycin (2 µg/ml for 48 h, Merck Millipore, MA) and blasticidin S (5 µg/ml for 24 h, FUJIFILM Wako Pure Chemical). Cells were exposed for 1 h to 10 µM cytochalasin D or 10 µg/ml nocodazole under the above culture conditions.
CRISPR/Cas9 gene KO
Rab8A/B-, Rab10-, Rab22A/B-, KIF13A-, KIF13B- and KIF13A/B-KO HeLaM cell lines were established by means of the following procedures. pSpCas9(BB)-2A-Puro vector (Addgene #48139) was used to generate a single guide RNA (sgRNA). The following sgRNA sequences were designed by using the CRISPRdirect website (http://crispr.dbcls.jp/): sgRNA sequences targeting human Rab8A (sense, 5′-CACCGATTAGGACCATAGAGCTCGA-3′ and antisense, 5′-AAACTCGAGCTCTATGGTCCTAATC-3′); human Rab8B (sense, 5′-CACCGCTCCTGCTGATCGGCGACTC-3′ and antisense, 5′-AAACGAGTCGCCGATCAGCAGGAGC-3′); human Rab10 (sense, 5′-CACCGGATCGGGGATTCCGGAGTGG-3′ and antisense, 5′-AAACCCACTCCGGAATCCCCGATCC-3′); human Rab22A (sense, 5′-CACCGTTAGCACCAATGTACTATCG-3′ and antisense, 5′-AAACCGATAGTACATTGGTGCTAAC-3′); human Rab22B (sense, 5′-CACCGCATCGTGTGTCGATTTGTCC-3′ and antisense, 5′-AAACGGACAAATCGACACACGATGC-3′); human KIF13A (sense, 5′-CACCGGATATGCAGACCGAGCCAAA-3′ and antisense, 5′-AAACTTTGGCTCGGTCTGCATATCC-3′); and human KIF13B (sense, 5′-CACCGAGTGAACGAGCAACGAAGAC-3′ and antisense, 5′-AAACGTCTTCGTTGCTCGTTCACTC-3′). The sgRNA expression constructs were transfected into HeLaM cells, and 24 h later 2 µg/ml puromycin (Merck Millipore) was added to the culture medium to select transfected cells. Puromycin was removed after an additional 48 h, and the cells were cloned by limited dilution. Clonal lines were isolated and analyzed by immunoblotting with a specific antibody to check for absence of expression of target proteins.
Purification of GST fusion proteins
GST-tagged GFP nanobody was expressed in Escherichia coli JM109 and purified as described previously (Kuroda and Fukuda, 2005).
HeLaM cells were lysed in an SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2% 2-mercaptoethanol, 10% glycerol and 0.02% Bromophenol Blue) and boiled. The samples were subjected to SDS-PAGE and then transferred to polyvinylidene difluoride membranes (Merck Millipore). The membranes were blocked with PBS-T (0.1% Tween 20 in PBS) containing 1% skimmed milk and incubated for 1 h with specific primary antibodies. The membranes were subsequently incubated for 1 h with appropriate HRP-conjugated secondary antibodies, and immunoreactive bands were detected by using enhanced chemiluminescence and X-ray films or the ChemiDoc Touch imaging system (Bio-Rad).
COS-7 cells co-expressing mStr–Rab10 and EGFP or EGFP-tagged KIF13 (A or B) tail were lysed in lysis buffer (50 mM HEPES-KOH, pH 7.2, 150 mM NaCl, 10 mM MgCl2, 1% Triton X-100 and protease inhibitor cocktail; Roche). The lysates were centrifuged at 17,400 g for 10 min at 4°C, and the supernatants were incubated for 1 h at 4°C with 5 µl of glutathione–Sepharose 4B beads (wet volume) (GE Healthcare, Little Chalfont, UK) and 1 µg of GST-tagged GFP nanobody. The beads were washed with a washing buffer (50 mM HEPES-KOH, pH 7.2, 150 mM NaCl, 10 mM MgCl2 and 0.1% Triton X-100) three times and boiled in the SDS sample buffer. To determine the nucleotide preference for the interaction between Rab10 and KIF13A, COS-7 cells co-expressing FLAG–Rab10 and EGFP or EGFP-tagged KIF13A-tail were lysed in the lysis buffer without MgCl2. The lysates were centrifuged at 17,400 g for 10 min at 4°C, and the supernatants were incubated for 1 h at 4°C with 5 mM EDTA and 5 mM GTPγS (or 10 mM GDP). After addition of 10 mM MgCl2, co-immunoprecipitation assays were performed as described above by using GST–GFP nanobody. The samples then were subjected to SDS-PAGE and analyzed by immunoblotting with appropriate antibodies.
For direct binding assays, FLAG-tagged Rab10 and T7-GST-tagged GFP or KIF13A (aa1003–1266) were purified from COS-7 cell lysates by using anti-FLAG M2 agarose beads (Sigma-Aldrich) and glutathione–Sepharose 4B beads, respectively. The purified FLAG–Rab10 was eluted with an elution buffer [500 ng/ml 3×FLAG peptide (Sigma-Aldrich) and 5 mM HEPES-KOH, pH 7.2] from the agarose beads. The beads coupled with T7-GST-tagged GFP or KIF13A (aa1003–1266) were incubated overnight at 4°C with the purified FLAG–Rab10 in a binding buffer (50 mM HEPES-KOH, pH 7.2, 150 mM NaCl, 10 mM MgCl2, 5 mM GTPγS and 0.1% Triton X-100). The beads were washed with the washing buffer three times and boiled in the SDS sample buffer.
For the immunofluorescence analysis, cells were cultured on coverslips, fixed with 4% paraformaldehyde in PBS for 10 min, and permeabilized for 10 min with 0.05% saponin in PBS containing 0.1% gelatin. The coverslips were incubated for 1 h with primary antibodies and subsequently incubated for 1 h with appropriate Alexa Fluor-conjugated secondary antibodies. The samples were mounted using ProLong Gold Antifade Mountant (Thermo Fisher Scientific). The stained cells were captured with an FV1000D confocal fluorescence microscope through a 60× oil/1.4 NA Plan Apochromatic objective lens and Fluoview software (Olympus, Tokyo, Japan). Super-resolution images were acquired by using the FV-OSR (Olympus), which is a commercial super-resolution system, of an FV1000D confocal fluorescence microscope.
Live-cell imaging was performed with an Andor Dragonfly spinning disk scanning unit (Dragonfly200) or a Zeiss Yokogawa spinning disk scanning unit (CSU-W1) coupled with an inverted Olympus IX83 microscope equipped with a 60× oil/1.35 NA Plan Apochromatic objective lens. During live-cell imaging, the culture dish was mounted in a chamber (STRG-WELSX-SET; Tokai Hit, Shizuoka, Japan, or STXG-IX3WX-SET; Tokai Hit) to maintain the cells at 37°C and under a 5% CO2 atmosphere. The imaging was immediately started after drug treatment (time=0).
Quantification of tubular endosomes
The number of cells containing at least one tubule >20 µm in length was manually measured. The number of cells containing at least one tubule >20 µm in length was counted in each experiment (more than 20 cells were analyzed in each experiment). Total tubule length (i.e. sum of the length of all tubules) per cell was measured by using Fiji (https://fiji.sc) and the population profile was created (Fig. 2H). Maximum intensity z-projected images were binarized and then skeletonized by LpxLineExtract, which is invoked by Lpx_Filter2d plug-ins (filter=lineFilters, linemode=lineExtract) in the LPixel ImageJ plugins package (https://lpixel.net/services/research/lpixel-imagej-plugins/) (Kuki et al., 2017). The >5 µm skeletonized lines in the images were extracted by using the analyze particles function in Fiji. Total tubule length in each cell was measured by using the measure function in Fiji, and a histogram analysis of the data was performed by using the Excel software program (Microsoft, Redmond, WA).
The PCC values were calculated for more than four maximum intensity z-projected images from each experiment by using Coloc 2, a Fiji plugin for colocalization analysis (http://imagej.net/Coloc_2). The line plot profile of the yellow arrows in Figs 1E and 5A was obtained by using the plot profile function in Fiji.
The RHD of KIF13A/B was identified by using the RBD of MICAL1 (human, isoform 3 of aa939–1080) or EHBP1 (human, isoform 1 of aa1070-1212) as a search query in PSI-BLAST searches, followed by a DELTA-BLAST search. The 3D-homology modeling of the RHD of KIF13A was performed by using the Modeller9.19 software (Šali and Blundell, 1993), incorporated in the UCSF Chimera interactive graphic interface (Pettersen et al., 2004). Rab10–MICAL1 complex (PDB: 5LPN) was used as a template.
All quantitative data are expressed as the mean±s.e.m. Tukey's test, Dunnett's test, unpaired Student's t-test and Pearson's χ2 test were performed to evaluate the statistical significance of differences between samples. P<0.05 was regarded as statistically significant.
We thank Drs Kazuhisa Nakayama, Kazumasa Ohashi, Shoji Yamaoka and Toshio Kitamura for kindly donating materials, and Dr Naonobu Fujita for technical advice and critical reading of the manuscript. We also thank Megumi Aizawa for technical assistance, Yuki Hatoyama, Koki Okuyama and Futaba Osaki for plasmid construction, and members of the Fukuda laboratory for valuable discussions.
Conceptualization: K.E.; Methodology: K.E.; Software: K.E.; Validation: K.E.; Formal analysis: K.E.; Investigation: K.E.; Data curation: K.E.; Writing - original draft: K.E.; Writing - review & editing: K.E., M.F.; Visualization: K.E.; Supervision: M.F.; Project administration: M.F.; Funding acquisition: K.E., M.F.
This work was supported in part by Grant-in-Aid for Scientific Research(B) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan (grant number 15H04367 to M.F.), Grant-in-Aid for Scientific Research on Innovative Areas from MEXT (grant number 17H05682 to M.F.), by Japan Science and Technology Agency (JST) CREST (grant number JPMJCR17H4 to M.F.), by the Japan Society for the Promotion of Science (to K.E.) and by Tohoku University Division for interdisciplinary Advanced Research and Education (to K.E.).
The authors declare no competing or financial interests.