Molecular mechanism for recognition of the cargo adapter Rab6GTP by the dynein adapter BicD2

A model of the Rab6^GTP/BicD2-CTD complex was obtained by structure prediction with AlphaFold2

To establish a structural basis for cargo recognition by BicD2, we used the software ColabFold (Mirdita et al, 2022), which combines the homology search of MMseqs2 with AlphaFold2 (Jumper et al, 2021; Evans et al, 2022 Preprint) to predict a structural model of the Rab6^GTP/BicD2-CTD complex. The complex was previously characterized by size-exclusion chromatography coupled with multi-angle light scattering and shown to form a 2:2 hetero-tetramer (Noell et al, 2018). The structure of the C-terminal cargo-binding domain of BicD2 has been determined; it forms a homodimeric, parallel coiled coil (Liu et al, 2013; Terawaki et al, 2015; Noell et al, 2019). The structure of Rab6 with the ligand GTP bound is also established (Bergbrede et al, 2005; Eathiraj et al, 2005). We used a structure of GTP-bound Rab6a with the Q72L mutation, which locks it in the GTP-bound state, as a template for the AlphaFold2 prediction (PDB ID 2GIL) (Martinez et al, 1994; Matanis et al, 2002; Bergbrede et al, 2005).

The structural model of the Rab6^GTP/BicD2 complex with the highest predicted local distance difference test (pLDDT) scores, and the lowest predicted aligned error (PAE) is shown in Fig 1A. The remaining predicted models are shown in Fig S1.

The pLDDT is a per-residue confidence metric, which estimates how well the prediction would agree with an experimental structure based on a local distance difference test (Jumper et al, 2021; Tunyasuvunakool et al, 2021), which estimates whether the predicted residue has similar distances to neighboring C-alpha atoms as observed in the experimental structure. The pLDDT in the highest ranked model of the Rab6^GTP/BicD2-CTD complex is above 85–98% with the exception of a few residues at the N- or C-terminus, indicating high confidence in the structure (Fig 1C and D). The PAE gives a distance error for every residue pair, and estimates the error of the position of residue x in Å if the predicted and actual structures are aligned at residue y. For complexes, it is an important error estimate that assesses the respective positioning of individual subunits. For the most part, the PAE of the highest ranked model of the Rab6^GTP/BicD2-CTD complex is below 5–10 Å, reflecting a high degree of confidence in the prediction (Fig 1B). Only the N-terminal portion of the BicD2-CTD has a somewhat elevated error, likely reflecting the flexibility of the coiled coil. Notably, the C-terminal part that is bound to Rab6 has a lower PAE.

Furthermore, we also used AlphaFold2 (Jumper et al, 2021) to predict a 2:2 complex of full-length BicD2 and Rab6^GTP (Fig S2). The model of the minimal Rab6/BicD2-CTD overlays well with the same domains in the predicted structure of the full-length complex, and these domains have reliable PAE and pLDDT scores. However, the PAE and pLDDT scores of the remaining domains of the full-length complex suggest that the prediction of the remaining structure is not reliable (Fig S2). Thus, we focused on the reliable model of the minimal Rab6^GTP/BicD2-CTD complex.

The highest ranked model of the minimal Rab6/BicD2-CTD complex is also supported by available biochemical data. In the model, Rab6^GTP binds to the ∼30 C-terminal residues of human BicD2, which is in line with the previously mapped minimal binding site consisting of residues 755–802 of BicD2 (Liu et al, 2013; Terawaki et al, 2015). Overall, the structures of the BicD2-CTD and Rab6^GTP in the complex are very similar to the original structures of the individual proteins (Bergbrede et al, 2005; Noell et al, 2019) (Fig S3A and B), but this is to be expected because AlphaFold2 is trained on the structures in the Protein Data Bank (Jumper et al, 2021).

Before our study, the BicD2 binding site on Rab6 had not been mapped. In our model, BicD2 binds to a site on Rab6 that is formed by a β-strand, a short α-helix, and a coil region (Fig 1A). The GTP is close to the BicD2 binding site but does not engage in the interaction. We analyzed the Rab6^GTP/BicD2-CTD complex for non-covalent interactions, which are summarized in Fig S3D. The presented contact residues are likely reliable, whereas specific salt bridges or hydrogen bonds are not necessarily expected to be accurately predicted.

Fig S3C shows the electrostatic surface potential of Rab6 and BicD2, which highlights that complementary electrostatic interactions are important for stabilizing the complex. BicD2 residues 774–804 form contacts with Rab6 residues 43–82 (Fig S3D), and all Rab6 residues that make contacts with BicD2 are highly conserved (Fig S3E). Interestingly, the binding site of BicD2 spans two regions of Rab6 that have previously been shown to undergo structural rearrangements in the GTP- versus GDP-bound state: Switch 1, which is located in residues 38–50 of Rab6, and Switch 2, which includes residues 67–87 (Garcia-Saez et al, 2006). These results are in line with the observation that BicD2 binds to Rab6^GTP with higher affinity compared with the GDP-bound state (Bergbrede et al, 2009).

Overall, the pLDDT and PAE error plots of the highest ranked model of the Rab6^GTP/BicD2-CTD complex indicate a high degree of confidence in the model, and it also fits well with biological data such as the previously mapped Rab6 binding site on BicD2, as well as previously identified point mutations that diminish binding of Rab6 to BicD2. Notably, the BicD2 binding site is located in two regions of Rab6 that have been previously shown to undergo conformational changes in the active GTP-bound state, potentially explaining why this state has a higher affinity to BicD2 than the inactive GDP-bound state.

The binding sites of BicD2 for Nup358 and Rab6 are structurally distinct but overlapping

Nup358 and Rab6^GTP compete for binding to BicD2, and Nup358 binds to a larger binding site on BicD2 (residues 724–802) compared with Rab6^GTP (residues 755–802), in line with its 20-fold higher affinity (Noell et al, 2018). Yet, the binding mode of these two cargoes to BicD2 is distinct. The core BicD2 binding site of Nup358 is formed by a short cargo-recognition α-helix, which is α-helical in the complex but intrinsically disordered in apo-Nup358 (Gibson et al, 2022). In addition, a short intrinsically disordered region of Nup358 binds to the C-terminal half of the BicD2-CTD in an anti-parallel manner (Gibson et al, 2023). In comparison, the BicD2 binding site of Rab6 is formed by a β-strand, an α-helix, and a coil region (Fig 1E and F). Rab6 binds to the same region of BicD2 as the intrinsically disordered N-terminal residues of Nup358-min, whereas the cargo-recognition α-helix binds on the center of the BicD2-CTD. Thus, the intrinsically disordered domain of Nup358 is the one that competes for binding to BicD2 with Rab6^GTP (Fig 1E and F).

It has previously been shown that the E774A mutation reduces binding to Rab6^GTP and Nup358, whereas the F743I/R747C mutation does not impact Rab6 binding but diminishes Nup358 binding (Noell et al, 2019; Cui et al, 2020; Yi et al, 2023). Our model of the Rab6^GTP/BicD2-CTD complex is in line with these results, because F743 and R747 are N-terminal of the Rab6^GTP binding site, whereas E774 is a contact residue with Rab6^GTP.

To conclude, the Nup358 and Rab6^GTP binding sites on BicD2 are overlapping but structurally distinct, in line with the observation that BicD2 disease mutations affect affinities towards these cargoes differently.

The binding sites of Rab6^GTP for Rab6 and ELKS are overlapping, providing a mechanism for competition

ELKS (also known as Rab6IP2, CAST, or ERC) is a Rab6^GTP interactor that captures Rab6-coated vesicles near plus-ends of microtubules at exocytosis hotspots in neurons and other cells and is a key regulator for protein secretion (Grigoriev et al, 2007; Patwardhan et al, 2017; Fourriere et al, 2019). A structure of a minimal Rab6^GTP/ELKS complex has been previously established, and it was observed that BicD2 and ELKS do not co-localize in cells, indicating that they compete for binding (Jin et al, 2023). Although ELKS has a lower affinity to Rab6 than BicD2, ELKS may form liquid–liquid phase separation condensates that increase its local concentration, are able to capture Rab6 vesicles, and thereby compete with other Rab6 effectors (Jin et al, 2023). However, the molecular basis for competition of BicD2 and ELKS for Rab6 binding was unknown. Here, a least-squares superimposition of the minimal Rab6^GTP/ELKS complex (Jin et al, 2023) with our structural model of the Rab6^GTP/BicD2-CTD complex suggests that both effectors bind to distinct but overlapping binding sites on Rab6^GTP. An analysis of the interface confirms that several Rab6^GTP contact residues engage in binding to both BicD2-CTD and ELKS (Fig S4A–C, Table S1). Competition of the Rab6 effectors BicD2 and ELKS could be an efficient means of organizing a cascade of alternating protein interactions that enable Rab6 to regulate protein trafficking and secretion.

Rab6^GTP has multiple other interactors, and we investigated whether they share the same binding site. Least-squares superimpositions of the predicted Rab6^GTP/BicD2-CTD complex with three other structures from the PDB: Rab6 bound to Kif20A (Fig S5A) (Miserey-Lenkei et al, 2017), GCC185 (Fig S5B) (Burguete et al, 2008), or R6IP1 (Fig S5C and D) (Recacha et al, 2009), show that each of these proteins binds to distinct but overlapping binding sites on Rab6; therefore, it is likely that BicD2 competes with these three other interactors for binding to Rab6 (Fig S5).

Rab6^GTP also can bind directly to residues 737–916 of the dynactin-interacting domain p150_glued (Bergbrede et al, 2009). Because a structure of the complex is not available, we used AlphaFold2, and the highest ranked prediction of a Rab6^GTP/p150_glued complex with 2:2 stoichiometry received pLDDT scores mostly above 80 and PAE scores below 10 Å, indicating that the prediction is likely reliable (Fig S6). A least-squares superimposition of the predicted structure of the minimal Rab6^GTP/p150_glued complex with the Rab6^GTP/BicD2-CTD complex is shown in Fig S6, which suggests that p150_glued and BicD2-CTD compete for binding to Rab6^GTP. However, this remains to be experimentally confirmed.

To conclude, our data suggest that all Rab6 interactors analyzed here, including BicD2, ELKS, the dynactin subunit p150_glued, Kif20A, GCC185, and R6IP1, bind to distinct but overlapping binding sites on Rab6^GTP and compete for binding. Thus, these Rab6 interactions are likely organized as a cascade and not formed simultaneously.

The structural model of the Rab6^GTP/BicD2-CTD complex is validated by mutagenesis

Next, we validated the BicD2 binding site on Rab6^GTP by mutagenesis. The contact residues of Rab6^GTP from the AlphaFold2 model (Fig S3D) were mutated to alanine along with the remaining Rab6 residues of the Switch 1 and Switch 2 regions, which undergo structural changes in the GTP-bound state (Fig 2A). For these experiments, the Q72L mutant of Rab6 was used as WT; this mutant is locked in the GTP-bound state that has a higher affinity to BicD2-CTD.

To identify residues that are important for binding of Rab6 to BicD2, binding of BicD2-CTD to GST-tagged Rab6^GTP mutants was assessed by pull-down assays. The elution fractions were analyzed on SDS–PAGE, and the intensities of the gel bands were quantified (Fig 2B and C; a representative full dataset is shown in Fig S7). These experiments revealed several Rab6 residues that are important for binding to BicD2: five residues in the Switch 1 region of Rab6, eight residues in the Switch 2 region of Rab6, and five residues in the connecting interswitch region (red in Fig 2C). It should be noted that three of the Rab6 residues that are contact residues in the AlphaFold2 model and that are also confirmed to be important for binding to BicD2 constitute the invariant hydrophobic triad, which is conserved in the Rab family of proteins and forms a hydrophobic switch region interface: residues F50, W67, and Y82. This invariant hydrophobic triad has been shown to undergo structural changes upon activation in the GTP-bound state that are a determinant for effector recognition, and specifically for BicD2, as we show here (Dumas et al, 1999; Ostermeier & Brunger, 1999; Merithew et al, 2001). Notably, the residues established by mutagenesis that are essential for the Rab6^GTP/BicD2 interaction validate the structural model of the Rab6^GTP/BicD2-CTD complex (Figs 2 and S3, Table S4).

Fig 2C shows the sequence of Rab6, in which the mutated residues that showed reduced binding in the pull-down assays are colored red. In addition, Rab6 residues that form contacts with both GTP and Mg²⁺ are indicated by an asterisk and Rab6 residues that form contacts with GTP are indicated by a plus sign (see Table S2). It should be noted that mutations of GTP- or Mg²⁺-contacting residues may possibly alter the interaction of Rab6 with GTP, which would in turn impact binding to BicD2.

To exclude that the mutations that decreased binding to BicD2 resulted in misfolding of Rab6^GTP, which could result in lowered binding even if the mutated residues were not engaged in the interaction with BicD2, we characterized the secondary structure of the Rab6^GTP mutants by circular dichroism (CD) spectroscopy.

Representative CD wavelength scans of three mutants overlaid with the WT spectra are shown in Fig 2, and the full dataset of all mutants is shown in Fig S8. Two local minima are observed in the spectra at 208 and 222 nm, which are characteristic of α-helical structures. The CD wavelength scan of the mutants K53A, I79A, and Y42A is very similar to the WT, confirming that the mutation does not result in misfolding or large structural changes compared with the WT (Fig 2D). In total, the secondary structure content of all 19 mutants that reduced binding to BicD2 (red in Fig 2C) was assessed by CD spectroscopy. The CD spectra of all 19 mutants resembled the WT spectra, suggesting that all mutants were correctly folded and did not have significant structural changes compared with the WT (Fig S8). The secondary structure content of all CD spectra was estimated with the program BeStSel and is summarized in Table S3. It should be noted that some minor changes were observed in some of the CD spectra of the mutants, as well as in the derived secondary structure content (Fig S8, Table S3). These differences are not significant, because the experimental error of the molar ellipticity is 3.5–5%. We recently determined this error by calculating the SD of the molar ellipticity at 208 and 222 nm from 10 experiments for which the samples were independently prepared (including the determination of the protein concentration, which is the main source of the experimental error) (Cui et al, 2020).

Fig 3A and B shows a close-up of the structure of the Rab6^GTP/BicD2-CTD complex with the confirmed Rab6 residues that are essential for the interaction with BicD2 in red stick representation. Notably, the interacting residues mainly consist of aromatic residues, as well as some hydrophobic residues: F38, Y42, I46, F50, W67, F75, I79, Y82, I83, and the positively charged R63. These residues project from the surface of Rab6-like fingers. It is likely that the interaction is mainly stabilized by hydrophobic interactions. In addition, the interaction is further strengthened by complementary charged surfaces (Fig S3C).

Furthermore, several residues of BicD2 were previously identified that are important for binding to Rab6^GTP: E774, L782, R783, M784, I786, K789, L790 (Liu et al, 2013; Terawaki et al, 2015). All seven residues form non-covalent interactions with Rab6^GTP in our structural model of the complex (Fig S3D, Table S4). Mutations of BicD2 residues 745, 750, 755, or 756 to alanine, which are N-terminal of the minimal binding site, do not affect binding to Rab6^GTP, as to be expected (Gibson et al, 2023).

It has been established that Rab6 is phosphorylated at position S52 in the G2 phase of the cell cycle by Polo-like kinase 1 (Plk1), which weakens the interaction with BicD2 (Jimenez et al, 2023 Preprint). A phosphomimetic S52D mutation had a similar effect, whereas a phosphonegative mutation S52A had no effect on the interaction (as observed in our pull-down assays as well, Fig 2). Notably, S52 is mediating a contact with BicD2. The close-by residues R63 (positively charged) and Q65 (polar) are important for the interaction between BicD2 and Rab6 and could potentially form a salt bridge and a hydrogen bond with negatively charged residues E797 and E782 of BicD2 (Fig S9A and B). It is conceivable that a phosphorylated S52 residue would form an intramolecular salt bridge with R63 and a hydrogen bond with Q65, which could weaken the interaction overall and there is expected to be like-charge repulsion between phosphorylated S52 of Rab6 and E797 and E783 of BicD2.

To conclude, our results from mutagenesis and CD spectroscopy confirm our structural model of the Rab6^GTP/BicD2 complex.

Structural basis for the higher affinity of GTP-bound Rab6 to BicD2

To understand the increased affinity of the GTP-bound state of Rab6 towards BicD2 (Bergbrede et al, 2009), we compared the structures of GTP- and GDP-bound apo-Rab6b, which is a close homolog to Rab6a with 94% sequence identity. Structural changes are observed in the Switch 1 and Switch 2 regions of Rab6b^GTP; these regions are located close to the GTP ligand (black in Fig 2A) (Garcia-Saez et al, 2006).

The structure of Rab6^GTP in the complex with the BicD2-CTD is very similar to the structure of apo-Rab6b^GTP (and also apo-Rab6a^GTP, Fig S3B), and no significant structural changes are observed in the protein backbone or the interacting residues (Fig 4B).

Fig 4A shows a least-squares superimposition of the structures of the Rab6^GTP/BicD2-CTD complex with the structure of apo-Rab6b^GDP (Garcia-Saez et al, 2006). Several GTP-dependent structural changes are observed: the loop of Switch 1 and the α-helix of Switch 2 shift their position and undergo structural rearrangements. Notably, the structure of several BicD2 binding residues (colored red) is rearranged. Large-scale structural changes are observed for the hydrophobic and aromatic residues Y42, I46, and F75 and for the positively charged residue R63, which are essential for binding to BicD2. Several other residues that are necessary for the interaction with BicD2 undergo more subtle structural rearrangements: F50, I79, and Y82 (Fig 4A).

We conclude that in the GTP-bound state of Rab6, several aromatic and hydrophobic residues are repositioned compared with the GDP-bound state. These structural rearrangements allow these residues to form a larger number of hydrophobic interactions with BicD2 in the GTP-bound state, thereby increasing the affinity. In the GDP-bound state, these residues are positioned away from the binding site, lowering the binding affinity to BicD2.

Point mutations that disrupt the Rab6/BicD2 interaction strongly diminish the motility of Rab6-positive vesicles and decrease co-migration of these vesicles with BicD2

Because we have identified several point mutations of Rab6, which selectively disrupt the interaction with BicD2, we investigated how these mutations modify the motility of Rab6-positive vesicles and the co-migration of Rab6 with BicD2 through live imaging in HeLa cells.

For the in vitro binding assays above, the Q72L mutant of Rab6 was used, which locks it in the GTP-bound state, as it has a higher affinity towards BicD2. It should be noted that for the live-cell imaging, WT Rab6 was used, which did not carry this mutation. GTP-bound Rab6 is the active form that is anchored in membranes by prenylation, whereas the inactive Rab6^GDP state is released from the membrane into the cytosol by GDI (Goud et al, 1990; Martinez et al, 1994; Grigoriev et al, 2007) (see Fig S10). Thus, Rab6 that localizes to vesicle membranes is expected to be predominantly bound to GTP.

To assess how the mutations modify the motility of Rab6-positive vesicles, we transiently transfected mCherry-tagged Rab6 into HeLa cells and performed live-cell imaging. Fig 5A shows several time points and a kymograph for mCherry (mCh)-Rab6-WT–positive vesicles, which are derived from Video 1. Single vesicles are highlighted by distinctly colored arrows, which suggest normal motility of Rab6-positive vesicles in the WT (Fig 5A, Video 1). Notably, the F38A and Y42A mutations, which selectively disrupt the Rab6/BicD2 interaction in vitro, result in a strong reduction in the motility of the Rab6-positive vesicles (Fig 5B–D and Video 2 and Video 3). To quantify the effect of these mutants on the motility of the Rab6-positive vesicles, we determined the number of trajectories of Rab6-positive vesicles for the WT, the F38A mutant, and the Y42A mutant using automated particle tracking that identifies the trajectories of moving particles (Figs 5D and S11A–D, Video 4 and Video 5). A recent study concluded that the results obtained from this automated analysis were very similar to those determined manually from the analysis of kymographs drawn along Rab6-positive tracks (Grigoriev et al, 2007; Schlager et al, 2014b; Serra-Marques et al, 2020). The results of our analysis showed that the number of motile vesicle trajectories is significantly reduced in both the F38A and the Y42A mutants compared with the WT (Fig 5D).

Although the number of the motile Rab6-positive vesicles is greatly reduced for the mutants, the Rab6-related fluorescence signal in the Golgi area remains strong, suggesting that these mutations do not affect localization of Rab6 to the Golgi membranes. Because membrane-associated Rab6 is mainly in the GTP-bound and active state, the strong localization of the mutants and the WT to Golgi-derived Rab6-positive vesicles suggests that the F38A and Y42A mutations do not affect the equilibrium between GDP- and GTP-bound states of Rab6 and also do not compromise activation and membrane integration of Rab6. In comparison, Fig S10A and C shows micrographs of cells expressing fluorescent fusion proteins of WT Rab6 and the Rab6/Q72L mutant that is locked in the GTP-bound state, and for both conditions, Rab6 localizes robustly to Golgi-derived vesicles, somewhat similar as observed for the F38A and Y42A mutants. In comparison, the Rab6/T27N mutant, which is locked in the GDP-bound state, disperses into the cell, resulting in an easily detectable phenotype (Fig S10B). It should be noted that one of the mutants we tested, Rab6/W67A, which is a contact residue in the AlphaFold2 model, also disperses into the cell, indicating that the mutation could affect either binding to GTP or activation of Rab6, which would subsequently impact membrane integration (Fig 5E). This is to be expected, as W67 is a member of the invariant hydrophobic triad that is conserved in Rab proteins. This triad is involved in the activation mechanism of Rab6, and important for recognition of effectors such as BicD2 (Dumas et al, 1999; Ostermeier & Brunger, 1999; Merithew et al, 2001). We conclude that the Y42A and F38A mutations do not significantly diminish binding of GTP or activation of Rab6, as it would result in their dispersal from the Golgi membranes.

We also quantified co-migration of Rab6- and BicD2-positive vesicles to assess whether the observed disruption of the Rab6/BicD2 interaction by the mutations results in reduced co-migration in the context of live cells. BicD2-GFP and Rab6-mCherry were transiently co-expressed, and the cells were monitored by live-cell imaging. For WT Rab6 and BicD2, Rab6-positive vesicles display normal motility and Rab6 and BicD2 co-localize throughout the experiment, suggesting that they co-migrate as expected (see Video 6, Video 7, and Video 8 and the derived time points and kymograph shown in Fig 6A–C). We can see in the movies, kymographs, and images at distinct time points that Rab6/Y42A-positive vesicles are negative for BicD2 and virtually immotile (yellow arrows, Fig 6D–F, Video 9, Video 10, and Video 11). BicD2-positive/Rab6-negative vesicles, however, display normal motility (blue, purple, and white arrows, Fig 6D–F, Video 9, Video 10, and Video 11). These results suggest decreased Rab6/BicD2 co-migration of the Rab6/Y42A mutant compared with the WT. We also quantified the ratio of co-localization of Rab6 and BicD2, which is significantly decreased in the Y42A mutant compared with the WT (Fig 6G). F38A has a similar effect to Y42A (Video 12).

These experiments confirm that Y42 is a key residue for stabilizing the Rab6/BicD2 interaction in cells and therefore support our model of the Rab6/BicD2 complex, as well as the Rab6 residues that are important for the interaction with BicD2.

To conclude, our data suggest that point mutations that selectively disrupt the interaction between Rab6 and BicD2 in our in vitro binding assays result in severely impacted motility of Rab6-positive vesicles in cells. In addition, these mutations also disrupt co-localization and co-migration of Rab6 and BicD2 in cells, suggesting that they are key residues for stabilizing the Rab6/BicD2 interaction, thereby supporting our model of the Rab6/BicD2 complex.

GST pull-down assays

All expression constructs were cloned and codon-optimized for expression in E. coli as described by the company GenScript, which also performed site-directed mutagenesis (Gibson et al, 2022). The BicD2-CTD expression construct with the N-terminal His₆-tag and the thrombin cleavage site was previously described and contains residues 715–804 of human BicD2 cloned into the pet28a vector (Gibson et al, 2022). The expression construct for full-length human Rab6a/Q72L with the N-terminal GST tag that can be cleaved off by PreScission protease in the pGEX6P1 vector was previously described (Gibson et al, 2023). The Q72L mutant was used as it locks Rab6 in the GTP-bound state and renders it GTPase-deficient (Martinez et al, 1994; Matanis et al, 2002). The full-length GTP-bound Rab6a/Q72L is referred to as Rab6^GTP WT in the study.

Rab6^GTP and BicD2-CTD constructs were expressed in the E. coli BL20(DE3)-RIL strain at 37°C as described.

GST pull-down assays of full-length GST-tagged Rab6^GTP and BicD2-CTD were performed as described (Gibson et al, 2023). BicD2-CTD WT (residues 715–804) was purified by a single Ni-NTA affinity chromatography step from 1 liter of cell culture. Rab6^GTP was purified by glutathione Sepharose from 0.5 liter of cell culture, and washed, but not eluted. 1 mM GTP and 2 mM MgCl₂ were added, and the columns were incubated for 30 min. Purified BicD2-CTD was added to the columns with bound Rab6^GTP and incubated for 30 min. The columns were washed and eluted with glutathione elution buffer as described (Cui et al, 2018, 2020; Gibson et al, 2022, 2023). The elution fractions were analyzed by SDS–PAGE, using gels with 16% acrylamide, and stained with Coomassie blue. The gel band intensities of Rab6^GTP and BicD2-CTD were quantified, and the background intensities were subtracted using ImageJ (Schneider et al, 2012) as described (Yi et al, 2023). The ratio of bound BicD2-CTD/Rab6^GTP was calculated and normalized to the WT (WT = 1).

Structure predictions by ColabFold/AlphaFold2

Structure predictions were carried out with the software ColabFold v1.3.0 (Mirdita et al, 2022), which combines the homology search of MMseqs2 with AlphaFold2-multimer (Jumper et al, 2021; Evans et al, 2022 Preprint), in the Google Colab AlphaFold2_mmseqs2 Notebook (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb, accessed on 26 Sep 2022). Amber relaxation was activated; thus, each round of the AlphaFold2 prediction process involved an energy minimization step using the AMBER99SB force field (Hornak et al, 2006), with additional harmonic restraints to maintain the system near the input structure. The restraints are applied independently to heavy atoms, with force constants of 10 kcal/mol Ų (Jumper et al, 2021).

A 2:2 hetero-tetramer composed of two molecules of human Rab6a/Q72L (residues 13–174 or full-length) and two molecules of human BicD2-CTD (residues 715–804 or full-length) was predicted, using the structure coordinates with the PDB ID 2GIL as a template for Rab6^GTP (Bergbrede et al, 2005). It was previously established that the Rab6^GTP/BicD2-CTD complex forms a 2:2 hetero-tetramer (Noell et al, 2018) and that residues 13–174 of Rab6^GTP have the same affinity towards the BicD2-CTD as the full-length Rab6^GTP protein, suggesting that they contain the entire BicD2-CTD binding site (Noell et al, 2018).

Other structures were predicted as 2:2 complexes from the sequences, using the same protocol. For the prediction of the full-length Rab6^GTP/BicD2 complex and the prediction of the Rab6^GTP/p150_glued complex, ColabFold v1.5.2 was accessed on 6/3/2023 and 6/20/2023, respectively.

Structure figures were created by UCSF Chimera and UCSF ChimeraX (Pettersen et al, 2021). Interface residues were identified in structure coordinates with the PDBePISA server, which identifies atoms that are exposed to the other protein molecule rather than the solvent (Krissinel & Henrick, 2007). Results were visualized in UCSF Chimera (Pettersen et al, 2021). Multiple-sequence alignments were performed with T-Coffee version 11 accessed on 8 Oct 2023 (Di Tommaso et al, 2011). Adobe Photoshop was used for figure preparation.

CD spectroscopy

CD spectroscopy was performed as described (Gibson et al, 2022). For these experiments, Rab6^GTP was purified by glutathione Sepharose, as described above, but without the addition of GTP. Rab6^GTP was eluted from the column by proteolytic cleavage with PreScission protease (Cytiva) to remove the GST tag (i.e., not by glutathione). Purified Rab6^GTP was incubated for 30 min with 1 mM GTP, and transferred into a buffer of 150 mM NaCl, 2 mM MgCl₂, 10 mM Tris, pH 8.0, and 0.2 mM TCEP by three cycles of dilution and concentration. Rab6 was concentrated to 0.3 mg/ml and flash-frozen in liquid nitrogen as described (Gibson et al, 2022). For Rab6^GDP, 1 mM GDP was added instead of GTP, and for nucleotide-free Rab6, no nucleotide was added.

CD data (CD, HT, and absorbance) were recorded with a Jasco J-1100 CD spectrometer from 250 to 190 nm at 10°C. The quartz cuvette had a pathlength of 0.1 cm. The following parameters were used: data pitch: 0.1 nm; D.I.T.: 2 s; bandwidth: 1.00 nm; scanning speed: 50 nm/min; and accumulations: 8. The buffer baseline was subtracted from the CD wavelength scans, and the raw ellipticity Θ (mdeg) was converted to mean residue molar ellipticity (Θ). The protein concentration used for the conversion was determined from the buffer-subtracted absorbance recording of the CD data at 214 nm, using the extinction coefficient 371,769 M⁻¹ cm⁻¹ (Kuipers & Gruppen, 2007). The data were not smoothened. Three CD wavelength scans from distinct purification batches were recorded for each protein, and representative spectra are shown. The secondary structure content of CD spectra was estimated with the program BeStSel (v1.3.230210, accessed at https://bestsel.elte.hu/index.php on 03/11/2024) (Micsonai et al, 2022).

Live-cell imaging

HeLa cells were cultured in DMEM supplemented with 10% FBS at 37°C with 5% CO₂. Rab6 mutants and the WT were codon-optimized for expression in human cells and cloned into the pmCherry-C1 vector using the EcoRI/BamHI sites (GenScript). The BicD2-GFP plasmid was described in Hu et al (2013). Transient transfections were performed using Effectene (301425; QIAGEN) according to the manufacturer’s protocol. For live-cell imaging, 18 h after the transfection, cells were imaged using IX83 Andor Revolution XD Spinning Disk Confocal System with an environmental chamber at 37°C, a 60x oil objective (NA 1.30), and a 2x magnifier coupled with iXon Ultra 888 EMCCD Camera. Images were taken at the rate of one frame per second for 3 min.

ImageJ version 2.14.0 was used (Schindelin et al, 2012; Schneider et al, 2012). Kymographs were generated using the Multi Kymograph 3.0.1 ImageJ (NIH) plug-in. The quantification of the ratio of co-localization was done in the portion of the image considered for the kymographs using the Coloc2 3.0.6 ImageJ (NIH) plug-in.

The number of motile Rab6-positive vesicles was quantified in 1-min videos (1 frame/s), using the Mosaic Particle Tracker 1.6 ImageJ (NIH) plug-in (Sbalzarini & Koumoutsakos, 2005), considering the following parameters: radius: 6; cutoff: 0; percentile: 0.1; displacement: 10; link range: 2; and trajectories: longer than 10 frames. All quantifications are the result of at least three separate experiments. Comparisons between experimental groups were done using a non-parametric test of Kruskal–Wallis for multiple groups and Mann–Whitney for two groups considering significant differences when the P-value was less than 0.05.

For immunofluorescence of fixed cells, cells were fixed with 4% PFA for 20 min, then permeabilized with Triton X-100 for 10 min. A blocking step was done by incubating with donkey normal serum (017-000-001; Jackson ImmunoResearch Laboratories) for 1 h. Then, incubation with the primary antibody (ab167453; Anti-mCherry, Abcam) was done for 2 h, followed by incubation with the secondary antibody (Cy3-AffiniPure Donkey Anti-Rabbit IgG (H+L); 711-165-152; Jackson ImmunoResearch Laboratories) for 1 h. Antibodies were diluted 1:200 in a blocking solution. Three washes of 10 min each were done after each antibody incubation. DAPI (9564; Sigma-Aldrich) was included in the second wash of the secondary antibody in a dilution of 1:10,000. Cells were mounted using Aqua-Poly/Mount (18606; Polysciences) on glass for imaging.