A non-canonical GTPase interaction enables ORP1L-Rab7-RILP complex formation and late endosome positioning

Endosomal transport represents the primary mode for intracellular trafficking and signaling transduction and thus has to be tightly controlled. The molecular processes controlling the endosomal positioning utilize several large protein complexes, one of which contains the small GTPase Rab7, Rab-interacting lysosomal protein (RILP), and oxysterol-binding protein–related protein 1 (ORP1L). Rab7 is known to interact with RILP through a canonical binding site termed the effector-interacting switch region, but it is not clear how Rab7 interacts with ORP1L, limiting our understanding of the overall process. Here, we report structural and biochemical investigation of the Rab7-ORP1L interaction. We found that, contrary to prior studies, the interaction between Rab7 and the N-terminal ankyrin repeat domain (ARDN) of ORP1L is independent of Rab7's GTP- or GDP-binding state. Moreover, we show that Rab7 interacts with ORP1L ARDN via a unique region consisting of helix3 (α3) and 310-helix 2 (η2). This architecture leaves the canonical effector-interacting switch regions available for RILP binding and thus allows formation of the ORP1L-Rab7-RILP tripartite complex. Mutational disruption of the interacting interface between ORP1L and Rab7 compromised the ability of ORP1L-Rab7-RILP to regulate the late endosome positioning. Collectively, our results again manifested the versatility in the interaction between GTPase and its effector.

Endosomal transport represents the primary mode for intracellular trafficking and signaling transduction and thus has to be tightly controlled. The molecular processes controlling the endosomal positioning utilize several large protein complexes, one of which contains the small GTPase Rab7, Rab-interacting lysosomal protein (RILP), and oxysterol-binding proteinrelated protein 1 (ORP1L). Rab7 is known to interact with RILP through a canonical binding site termed the effector-interacting switch region, but it is not clear how Rab7 interacts with ORP1L, limiting our understanding of the overall process. Here, we report structural and biochemical investigation of the Rab7-ORP1L interaction. We found that, contrary to prior studies, the interaction between Rab7 and the N-terminal ankyrin repeat domain (ARD N ) of ORP1L is independent of Rab7's GTP-or GDP-binding state. Moreover, we show that Rab7 interacts with ORP1L ARD N via a unique region consisting of helix3 (␣3) and 3 10 -helix 2 (2). This architecture leaves the canonical effectorinteracting switch regions available for RILP binding and thus allows formation of the ORP1L-Rab7-RILP tripartite complex. Mutational disruption of the interacting interface between ORP1L and Rab7 compromised the ability of ORP1L-Rab7-RILP to regulate the late endosome positioning. Collectively, our results again manifested the versatility in the interaction between GTPase and its effector.
The protein family of Rab small GTPase contains key regulators of intracellular trafficking processes (1). For example, Rab7 plays a central role in early-to-late endosome conversion, late endosome-lysosome fusion, and transport of late endosomes (LEs) 4 along microtubules (2)(3)(4)(5)(6). Like various other Rabs, Rab7 cycles between a GTP-bound active state (GTP-Rab7) and a GDP-bound inactive state (GDP-Rab7). This cycle is controlled by a guanine nucleotide exchange factor complex and a GTPase-activating protein. GTP-Rab7 localizes to late endosomes where it interacts with distinct Rab7 effectors to control the late endosome-related events. For example, GTP-Rab7 interacts with Rab-interacting lysosomal protein (RILP), which forms a complex with dynein-dynactin motor and thus drives movement of late endosomes toward microtubule minus ends (2,3,7,8). Alternatively, GTP-Rab7 interacts with FYVE and coiled-coil domain-containing protein 1 to drive transport of late endosomes toward microtubule plus ends (4).
Recent reports revealed that late endosomal positioning is regulated by the cholesterol levels in the late endosomes. The cholesterol level in the late endosomes is sensed by oxysterolbinding protein-related protein 1 (ORP1L) (9). ORP1L contains an N-terminal ankyrin repeat domain (ARD), a pleckstrin homology domain (PHD), a FFAT motif (two phenylalanines (FF) in an acidic tract), and a C-terminal oxysterol-binding protein (OSBP)-related domain (ORD). When cholesterol molecules in the late endosomes are present at high levels, ORP1L binds to cholesterol through the ORD and adopts a conformation that enables the interaction of ARD in ORP1L with the Rab7-RILP complex. The latter further recruits the dyneindynactin complex, resulting in a clustering of late endosomes at the microtubule-organizing center. When cholesterol molecules are present at low levels, ORP1L fails to bind cholesterol, thus allowing exposure of the FFAT motif, which interacts with the endoplasmic reticulum (ER)resident vesicle-associated membrane protein (VAMP)-associated protein (VAP). As a consequence, the dynein-dynactin complex dissociates from Rab7-RILP, facilitating the cro ARTICLE formation of ER-LE membrane contact sites and distribution of LEs at the cell periphery (9).
The interaction between Rab7 and RILP has previously been studied using X-ray crystallography. Analysis of the Rab7-RILP structure at high resolution showed that GTP-bound Rab7 interacts with RILP using its "effector-interacting switch region" and Rab subfamily motifs 1 and 4 (10). These effectorinteracting switch regions represent canonical binding sites recognized by most Rab effectors (11)(12)(13). However, little structural detail is established as to how ORP1L and RILP simultaneously interact with Rab7 during cholesterol-dependent positioning of late endosomes. Here, we set out to identify the molecular details of this mechanism by performing both structural and biochemical analyses of ORP1L-Rab7-RILP complex. Our structural data clearly show that ORP1L ARD N binds to a noncanonical region of Rab7, which is positioned away from the effector-interacting switch region, thus enabling Rab7 to bind both ORP1L and RILP in a simultaneous manner.

Human ORP1L (hORP1L) ARD N binds to both GTP-and GDP-bound Rab7
GTP is crucial for the interactions between Rab7 and its downstream effectors. ORP1L was reported to preferentially bind to GTP-Rab7 in mammalian two-hybrid assays (8). To test whether GTP is required for the interaction between hORP1L and Rab7, we used dominant active mutant and dominant negative mutant of Rab7 and performed coimmunoprecipitation assays. According to previous reports, the dominant active mutant, Q67L, specifically binds to GTP and exhibits a much slower rate of GTP hydrolysis compared with the WT form. In contrast, the dominant negative mutant, T22N, exhibits higher binding affinity for GDP than GTP (2,14,15). For example, the interaction between T22N and WD repeat domain 91 (WDR91) or Rubicon was barely detectable (5,16). However, we found that full-length hORP1L coprecipitated with both Q67L and T22N in HEK 293 cells (Fig. 1A), although the amount of the hORP1L protein precipitated by Q67L was higher than that precipitated by T22N.
We next defined the core regions in hORP1L necessary for its interaction with Rab7. It was reported that the ARD of hORP1L is required for Rab7 interaction (8). However, according to DomPred, a domain boundary predication software (17), and Phyre2, a secondary structure prediction website (18), ARD is further separated into two subdomains from Gln-137, which is located on a loop between two helices. Consistent with these results, we found that a 15-18-kDa fragment at the N-terminal ARD was stable in a proteinase digestion assay (Fig. S1). Thus, Figure 1. The interaction between ORP1L ARD N and Rab7 is GTP-independent. A, the cellular interaction between full-length hORP1L and WT or mutated Rab7 (Q67L and T22N) was tested using a coimmunoprecipitation assay in HEK 293 cells. The GFP-Rab7 proteins were immobilized using GFP-Trap beads, and hORP1L, the precipitated protein, was detected using anti-FLAG antibody. B, pulldown of Rab7-Q67L by GST-tagged ORP1L proteins. A truncated form of GST-tagged ORP1L was incubated with purified His-tagged Rab7-Q67L protein in the presence of GTP and then immobilized on GSH-Sepharose beads. After extensive washing, Rab7-Q67L bound to ORP1L was detected using anti-His antibody. C, pulldown of ARD N by WT or mutated Rab7 (Q67L or T22N). The purified GST-tagged Rab7 proteins were incubated with His-tagged ORP1L ARD N (residues 1-136) and GTP or GDP as indicated, and then immobilized on GSH-Sepharose beads. After extensive washing, the ORP1L ARD N protein was detected using anti-His antibody. Data shown in A-C are representative of two independent experiments. D, the dissociation constants for ARD N binding to GTP-or GDP-bound Rab7 are identical as determined by ITC. IP, immunoprecipitation.

Crystal structural of ORP1L-Rab7
we separated ARD into two regions, ARD N (ORP1L(1-136)) and ARD C (ORP1L(141-237)), and tested their interactions with Rab7. Our pulldown assay results clearly showed that ARD N , but not ARD C , is required for the interaction between hORP1L and Rab7 (Fig. 1B). As shown in Fig. 1C, the in vitro interaction between recombinant ARD N and Rab7 is independent of the GTP-or GDP-binding states of Rab7. To determine the binding affinity between ARD N and Rab7, we performed an isothermal titration calorimetry (ITC) assay. As shown in Fig.  1D, the dissociation constants of ARD N binding to Rab7-WT-GTP, Rab7-Q67L-GTP, and Rab7-T22N-GDP were calculated as 3.2 Ϯ 0.3, 3.5 Ϯ 0.6, and 2.5 Ϯ 0.3 M, respectively. These in vitro results showed that hORP1L binds to Rab7 using ARD N and that their interaction is independent of the nucleotidebinding states of Rab7.

Structure of mouse ORP1L (mORP1L)-ARD N and its complex with Rab7
To determine the structure of ARD N and Rab7 at high resolution, we expressed and purified the ARD N of mORP1L and N-terminal residues 1-176 of dominant active mutant Rab7-Q67L, which binds to ORP1L. The identity of mORP1L and hORP1L in the ARD N region is 86.76%, suggesting that the structures of hORP1L and mORP1L are identical. The elution volumes of the single proteins and their complex in a Superdex 75 gel filtration column suggest that the monomeric ARD N and Rab7 form heterodimers in solution ( Fig. 2A).
Both ARD N and the ARD N -Rab7 complex were crystallized. The crystal structure of ARD N was determined at 2.14 Å using the molecular replacement method (Fig. 2B). Comparable with canonical ankyrin repeat domain structures, the structure of ARD N is composed of four tandem ankyrin repeats (ANK 1-4). Each ANK repeat contains a pair of antiparallel ␣-helices. The ANK repeats pack side by side and constitute a concave structure with the first helixes of the ANK repeats (H1, H3, H5, and H7) in the inner side. The loops between ANK repeats fold into ␤-hairpin structures, known as Fingers 1-3 in this structure (19). The fingers stretch out perpendicularly to the inner side of the concave structure.
In addition, we determined the structure of the ARD N -Rab7 complex at 2.14 Å (Fig. 2C). Different from the reported ARD of the VPS9 domain ankyrin repeat protein, which uses its convex side to bind the switch regions of Rab32 (20), the ARD N of ORP1L binds to Rab7 using the inner side of its concave structure, involving Finger 1 and the four inner ␣-helices (H1, H3, H5, and H7). The ARD N -binding site in Rab7 consists of ␣3, 2, , composed of four ANKs and three fingers. C, overall structure of the ARD N -Rab7 complex. ARD N is shown in green, and Rab7 is shown in light blue. The switch I, interswitch, and switch II regions of Rab7 are shown in purple, cyan, and orange, respectively. GTP is shown as red sticks, and magnesium is shown as a yellow sphere. The secondary structure elements are shown. D, ARD N , Rab7, and RILP associate into a ternary complex in solution. GST-tagged ARD N was incubated with His-tagged RILP and His-tagged Rab7 and then immobilized on GSH-Sepharose beads. After extensive washing, RILP and Rab7 were detected using anti-His antibody. E, structure model of a ternary ORP1L-Rab7-RILP complex shows that Rab7 binds ORP1L and RILP simultaneously using two distinct regions. The RILP dimer is shown in pink. ARD N and Rab7 are shown in colors identical to those used in C.

Crystal structural of ORP1L-Rab7
and the loop between them, which is located away from the conventional effector-binding switch regions (Fig. 2C). As shown in Fig. S2, we found a conformation change of Finger 1 in ARD N through comparison of the structures of ARD N alone and ARD N -Rab7. The buried interface area between ARD N and Rab7 is approximately 680 Å 2 as calculated by PISA (Proteins, Interfaces, Structures and Assemblies) (21). As the ARD Nbinding site in Rab7 is located away from its nucleotide-binding pocket (Fig. 2C), this structure is consistent with our hypothesis that ARD N -Rab7 interaction is nucleotide-independent.
To verify the interaction between ARD N -Rab7 and RILP, we performed pulldown assays using purified GSH S-transferase (GST)-tagged ARD N (GST-ORP1L ARD N ), His-tagged RILP(241-320), and His-tagged Rab7. As shown in Fig. 2D, in the presence of GTP-bound Rab7, RILP binds to ARD N . To build a model for the ORP1L-Rab7-RILP ternary complex, we superposed the structures of ARD N -Rab7 with Rab7-RILP (Protein Data Bank (PDB) code 1YHN) (10) based on the structure of Rab7 (Fig. 2E). In this model, two ARD N -Rab7 heterodimers were bridged by an RILP homodimer, forming an X-shaped structure with C2 symmetry. Because ARD N and RILP interact with Rab7 at different positions, this model provides structural explanations for the fact that ORP1L and RILP simultaneously interact with Rab7 (3).

The interaction between ARD N and Rab7
To study the amino acids required for ARD N -Rab7 heterodimer formation, we analyzed the structure of ARD N -Rab7 complex, mutated the crucial amino acids in the protein-protein interface, and performed GST pulldown assays. As shown in Fig. 3, A and B, Trp-49 (Finger 1), Tyr-57 (H3), Phe-58 (H3), and Phe-90 (H5) in ARD N constitute a hydrophobic region in the inner face of the concave structure. The aromatic groups in these residues interact with the side chains of Ile-108 and Pro-115 in Rab7. In addition, the positive-charged side chains of Arg-14, Lys-44, and Arg-93 in ARD N interact with the negativecharged side chains of Asp-104, Glu-105, and Glu-116 in Rab7, respectively (Fig. 3A). As shown in Fig. 3A, hydrogen bonds formed between Phe-90, Arg-87, and Tyr-57 in ARD N and Arg-113, Ser-111, and Pro-112 in Rab7. In addition, we performed sequence alignments for Rab7 and other Rab GTPases family members. As shown in Fig. 3C, the ARD-binding sites in Rab7 are not conserved in the Rab family, suggesting that ORP1L specifically binds Rab7 but not any other Rab proteins. This result is consistent with a previous study (22). Compared with the WT ARD N , the binding abilities of ARD N mutants W49A, Y57A/F58A, and F90A with Rab7 were greatly reduced (Fig.  3D). In agreement with this result, mutations at Ile-108 and Pro-115 in Rab7 greatly reduced the interaction between Rab7 and ARD N (Fig. 3F). In addition, the interaction between Rab7 and ARD N was also abolished by mutants affecting the electrostatic interactions and hydrogen bond formation, including R14A, R87A, and R93A in ARD N and D104A, E105A, E116A, S111A, and R113A in Rab7 (Fig. 3, E-G). In contrast, the mutants L98D, E65K/D66K, and Y103A, whose mutation sites are far from the ARD N -Rab7 interface, preserved their binding Crystal structural of ORP1L-Rab7 abilities (Fig. S3A). Although failing to bind ARD N , Rab7 mutants D104A and S111A still interacted with RILP (Fig. S3B), indicating that the structural integrity of Rab7 is not destabilized by these mutations.
To examine the interaction between ORP1L and Rab7 in vivo, we performed coimmunoprecipitation assays. As shown in Fig. 4A, full-length ORP1L-Y57A/F58A mutant failed to coprecipitate with Rab7, and Rab7-S111A mutant did not associate with ORP1L (see also Fig. 1A). In agreement with this result, ORP1L-Y57A/F58A mutant failed to colocalize with Rab7 and was distributed evenly in the cytosol (Fig. 4, B and C). In contrast, the WT ORP1L proteins are recruited to late endosomes, indicated by CD63, an independent endosome marker (23). Moreover, siRNA knockdown of Rab7 abolished the endosomal localization of ORP1L, which was rescued by expression of an siRNA-resistant Rab7. However, expression of siRNAresistant Rab7-S111A failed to recruit ORP1L to late endosomes (Fig. 4, D and E). Taken together, these results confirmed that ORP1L binds to the ␣3-2 region but not the switch regions of Rab7.

. Effects of mutations on ORP1L-Rab7 interaction and colocalization in cells.
A, left panel, coimmunoprecipitation between full-length GFPtagged WT or mutant ORP1L and FLAG-tagged Rab7. Immunoprecipitations (IP) were performed with anti-FLAG M2-agarose beads, and precipitated proteins were detected with anti-GFP antibody. Right panel, coimmunoprecipitation between GFP-tagged WT or mutant Rab7 and FLAG-tagged ORP1L. Immunoprecipitations were performed with anti-FLAG M2-agarose beads, and precipitated proteins were detected with anti-GFP antibody. B, confocal images of the colocalization of full-length GFP-ORP1L-WT or -Y57A/F58A mutant with mCherry-Rab7 in HeLa cells transfected with these plasmids. Scale bars, 5 m. The graph shows quantification of protein colocalization. Pearson's correlation coefficient is plotted on the y axis. Ն50 cells were scored. Error bars represent S.E. Significance was tested by unpaired Student's t test: ***, p Ͻ 0.001. C, confocal images of the colocalization of full-length GFP-ORP1L-WT or -Y57A/F58A mutant with mCherry-CD63 in HeLa cells transfected with these plasmids. Scale bars, 5 m. D, colocalization between full-length ORP1L and Rab7-WT or -S111A mutant. HeLa cells were treated with control siRNA or Rab7 siRNA to knock down (KD) endogenous Rab7 expression. Then GFP-ORP1L and mCherry-Rab7-WT or -S111A mutant plasmids were cotransfected. Scale bars, 5 m. E, immunoblotting (IB) of Rab7 in HeLa cells treated with control (Ctrl) siRNA or Rab7 siRNA.

ORP1L-Rab7 interaction is indispensable for ORP1L-mediated LE-ER contact site formation and LE positioning
Endosomal ORP1L facilitates the formation of ER-endosome contact sites by sensing cholesterol embedded in late endosomal membranes, which is essential for properly directing and docking of late endosomes to the ER (9). To determine how important ORP1L-Rab7 interaction is for endosomal transport to the ER, we designed ORP1L(1-514) and Rab7 interactiondefective mutants of ORP1L. Because the C-terminal ORD of oxysterol-binding protein-related protein down-regulates its membrane-tethering ability (24), we proposed that ORP1L(1-514), a truncated form of ORP1L without the ORD (residues 548 -940), should displays a heightened ability to bind to Rab7 and VAP-A when compared with its full-length version. As shown in Fig. 5A, overexpression of ORP1L(1-514) rather than the full-length protein significantly stimulated the formation of LE-ER contact sites in HeLa cell as indicated by a strong increase in colocalization of Rab7 with VAP-A (Fig. 5A). However, expression of ORP1L(1-514) with Y57A/F58A mutation failed to induce LE-ER contacts, suggesting that the interaction with Rab7 is essential for ORP1L to promote LE-ER contact site formation (Fig. 5A). As shown in Fig. 5B, overexpression of RILP resulted in clustering of LEs around the perinuclear region in all HeLa cells because RILP drives microtubule minus-end transport of LEs. Coexpression of full-length ORP1L enhanced RILP-induced perinuclear clustering of LEs. However, coexpression of ORP1L(1-514) abolished RILP-induced perinuclear clustering of LEs. In comparison, coexpression of another ORP1L(1-514) mutant containing Y57A/F58A failed to abolish such clustering (Fig. 5B). This finding suggests that an ORP1L(1-514)-induced increase in LE-ER contacts antagonizes RILP-mediated microtubule minus-end transport of LEs and that the interaction of ORP1L with Rab7 is critical for the positioning of LEs.

Discussion
In this study, we found that ARD N (ORP1L(1-136)) is sufficient for ORP1L to interact with Rab7. Further crystal structure analysis of ARD N -Rab7 complex revealed that the ␣3-2 region on Rab7 is required for ORP1L binding. This unique Rab effector-binding mode explained our finding that ORP1L ARD N -Rab7 binding is GTP-independent and how ORP1L-Rab7-RILP ternary complex is formed. The structure was further verified by mutagenesis, in vitro pulldowns, and colocalization analysis in cells.
ORP1L was first discovered as a Rab7 effector that preferentially binds to GTP-bound Rab7 in mammalian two-hybrid assays (8). In agreement, our coimmunoprecipitation assays revealed that the full-length ORP1L interacts with GTP-Rab7 more strongly than GDP-Rab7. In contrast, ORP1L ARD N binds equally well to both GTP-and GDP-bound Rab7in vitro, and the Rab7-S111A mutant failed to interact with either ARD N or the full-length ORP1L, suggesting that Rab7 binds to ARD N independently of Rab7 nucleotide-loaded states. We reason that the discrepancy in binding preference of Rab7 with ARD N and full-length ORP1L might result from the regulatory effects of the C-terminal domains within ORP1L.The regula-tory effects of Rab7 effectors has been previously reported. For example, C-terminal residues 392-747 of WDR91, another Rab7 effector, interact with both GTP-and GDP-Rab7 equally; nevertheless, the full-length WDR91 preferentially binds to GTP-Rab7 (16), suggesting a regulatory function of its N terminus. In addition, it is possible that a third protein or some unknown indirect interactions are involved in the in vivo situation. For instance, it was shown that addition of RILP stabilized and increased ORP1L binding to immobilized GST-Rab7 (3).
Previous studies suggested that Rab GTPases use the switch regions to interact with more than one effector. For example, in the tripartite complex Rab11-phosphatidylinositol 4-kinase (PI4KIII␤)-FIP3, PI4KIII␤ makes contacts with Rab11 on its switch I region, leaving most of the switch regions available for FIP3 to join the complex (25). In the Rab11-Rabin8-FIP3 complex, Rabin8 and FIP3 specifically interact with GTP-Rab11 and share the same switch regions on Rab11 (26). Before this study, it was known that Rab7 forms a tripartite complex with ORP1L and RILP to mediate cholesterol-dependent LE transport and positioning (3,9). In this tripartite complex, Rab7 interacts with RILP using the switch and interswitch regions, the canonical Rab GTPase effector-binding sites (10 -13). However, in contrast to the Rab11-Rabin8-FIP3 model, our structure reveals that the ARD N of ORP1L makes contacts with Rab7 on its ␣3-2 region, which is far away from the switch regions. Thus, the interaction of ORP1L ARD N with Rab7 is not affected by the GTP-or GDP-binding state of Rab7.
Our findings that ORP1L and RILP simultaneously interact with Rab7 by binding to distinct regions explain the specificity of ORP1L-Rab7 interaction and provide a molecular basis for ORP1L-mediated LE-ER contact site formation and LE positioning. Our structure provides an additional interaction mode as to how one Rab GTPase binds two effectors at the same time, which will increase the understanding of their regulatory mechanisms.

Protein expression and purification
The gene of mouse ARD N (ORP1L(1-136)) was inserted into the pRSFDuet-1 vector and expressed as an N-terminal Histagged protein in the Escherichia coli strain BL21 (DE3). Cells were cultured in Luria-Bertani (LB) medium with 100 mg/liter ampicillin at 37°C until the A 600 of the culture reached 0.8 -1.0.

Crystal structural of ORP1L-Rab7
mARD N protein expression was induced by 0.25 mM isopropyl ␤-D-thiogalactopyranoside (Sigma) for 20 h at 16°C. The cells were harvested by centrifugation at 4000 rpm. The pellet was resuspended with lysis buffer (20 mM Tris-HCl, pH 7.5, 400 mM NaCl, 30 mM imidazole) and lysed by sonication. The lysate was centrifuged at 16,000 rpm for 30 min, and the supernatant was loaded to a nickel-nitrilotriacetic acid column (Novagen). After extensive washing with lysis buffer, the mARD N protein was eluted with elution buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 500 mM imidazole) and further purified by anion
The human Rab7 (Rab7(1-176)) with Q67L mutation, Rab7-Q67L, used in crystallization was inserted into the pGEX-6P-1 vector and expressed using the same protocol as for mARD N mentioned above. The harvested cells were resuspended using lysis buffer (20 mM Tris-HCl, pH 7.5, and 200 mM NaCl) and lysed by sonication. The lysate was centrifuged at 16,000 rpm for 30 min, and the supernatant was loaded to a GSH-Sepharose column (GE Healthcare). Contamination was removed by extensive washing using lysis buffer. The GST tag of the target protein can be kept or removed by on-column cleavage using PreScission protease (homemade). The cleavage was conducted at 4°C overnight. The eluted target protein was further purified using a HiTrap Q HP column (GE Healthcare). Other GST-tagged proteins, including ORP1L truncated protein, the WT and mutated ARD N (ORP1L(1-136)), and Rab7 (Rab7(1-176)) used in pulldown and ITC assays, were purified following a similar procedure.

Protein crystallization and structure determination
Purified mARD N was concentrated to 10 mg/ml and subjected to crystallization screens by the hanging-drop vapor diffusion method at 16°C. To set up trials for crystallization, the protein were mixed with precipitant (Hampton Research) at a ratio of 1:1. The best crystal diffracted at 2.14 Å and was obtained in conditions with 200 mM NaCl, 100 mM Bis-Tris, pH 6.5, and 25% PEG 3350. The diffraction data were collected at the Swiss Light Source PX-II beamline at a wavelength of 1.00 Å and processed with XDS (27). The structure of mORP1L ARD N was solved by molecular replacement with the program Phaser (28) using the structure of a designed ankyrin repeat protein (PDB code 4K5B, chain A) as the search model. The initial model was automatically built by Buccaneer (29 -31). After that, the structure was manually modified using Coot (32) and refined with Phenix.Refine (33). The final model has an R work of 0.238 and an R free of 0.240. Data scaling, refinement, and validation statistics are listed in Table 1.
Before setting up the trials of the ARD N -Rab7 complex, mARD N and hRab7-Q67L proteins were purified separately and mixed in molar ratio of 1:1. The complex was supplied with GTP (Amresco) and purified with a HiLoad 16/60 Superdex 75 gel filtration column (GE Healthcare). The purified complex was concentrated to 10 mg/ml and used for the crystallization screen. The best crystal grew in a reservoir solution containing 100 mM HEPES, pH 7.5, and 4% PEG 8000. The diffraction data were collected at the Swiss Light Source PX-II beamline at a wavelength of 1.00 Å and processed with XDS (27). The complex structure was solved by the molecular replacement method with the program Phaser (28) using the Rab7 structure (PDB code 1T91, chain A) and the above solved mARD N structure as initial search models. The structure was further rebuilt and refined using the protocol as described mentioned. The final model has an R work of 0.182 and an R free of 0.209. Data scaling, refinement, and validation statistics are shown in Table  1. All figures were prepared using PyMOL. The coordinates and structure factors have been deposited in the RCSB Protein Data Bank (PDB codes 5Z2N and 5Z2M).

Western blotting
Western blotting was performed using a standard protocol as described previously (34). The Rab7 rabbit mAb was purchased from Cell Signaling Technology (catalogue number 9367) and used at 1:1000 dilution. Anti-His antibody was from Tiangen Biotech (catalogue number AB102) and used at 1:1000 dilution. Anti-FLAG antibody was from Sigma-Aldrich (catalogue number F1804) and used at 1:1000 dilution. GFP polyclonal antibodies were generated in rabbits by injecting purified His-GFP and used at 1:1000 dilution.

GST pulldown
To test the interactions between ORP1L and Rab7, 60 g of GST-tagged ORP1L were incubated with 20 g of His-tagged Rab7 and 20 l of GSH-Sepharose beads (GE Healthcare) in binding buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 2 mM DTT, and 0.01% Nonidet P-40) for 2 h at 4°C. Beads were extensively washed, and the bound Rab7 was detected by Western blotting using anti-His. The results are shown in Figs. 1B; 2D; and 3, D and E. Results shown in Figs. 1C and 3, F and G, were detected using GST-tagged Rab7 and Histagged ORP1L. The results from an independent replicate are shown in Fig. S4.

Crystal structural of ORP1L-Rab7
Healthcare). ARD N was expressed and purified using a His tag. All the Rab7 proteins were expressed as GST fusion proteins, and the GST tag was removed during the purification process. The proteins were maintained in ITC assay buffer (20 mM Tris-HCl, pH 7.5, and 150 mM NaCl). The exothermic heat of the reaction was measured by 20 sequential 1.5-l injections of ARD N (450 M) into 200 l of Rab7 (25 M), spaced at intervals of 120 s. The heat of dilution was obtained by injecting ITC assay buffer into Rab7 and was subtracted from the heat of reaction before the fitting process. Binding curves were analyzed by nonlinear least-squares fitting of the data using Micro-Cal Origin software.

Immunoprecipitation
Cells were lysed in lysis buffer (20 mM Tris-HCl, pH7.5, 150 mM NaCl, 1% Triton X-100, and 1 mM phenylmethanesulfonyl fluoride). Immunoprecipitations were then performed as described before (35) using individual antibodies. Precipitated proteins were resolved by SDS-PAGE and detected using the different antibodies listed above.

Immunostaining and imaging of mammalian cells
Cells grown on coverslips were fixed in 4% paraformaldehyde followed by permeabilization with 0.2% saponin for 8 min. After extensive washing with phosphate-buffered saline (PBS), coverslips were incubated in blocking buffer 1 (5% BSA and PBS) for 1 h at room temperature and then incubated with primary antibodies in the same buffer at 4°C overnight. Cells were washed extensively again and incubated with Cy3-or FITCconjugated secondary antibodies for 1 h at room temperature. Following another round of thorough washing, cells were sealed with VECTASHIELD mounting medium (Vector Laboratories) for confocal microscopy. Fluorescence images were obtained with an inverted Olympus FV1000 confocal microscope system (IX81) using a 60ϫ 1.42 numerical aperture oil objective. Excitation was achieved using solid-state 488 nm and gas-state 595 nm lasers. All images were taken at 25°C. Colocalization between two channels was calculated using Pearson's coefficient with FV10-ASW 4.0a Viewer software (Olympus).