Cryo-EM structure of the Mon1–Ccz1–RMC1 complex reveals molecular basis of metazoan RAB7A activation

Significance

Rab GTPases are critical to specify cellular organelles and to regulate intracellular trafficking. As important molecular switches, the activity of Rab GTPases is spatiotemporally controlled by many proteins, including guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins. The Mon1–Ccz1 complex is the only known RAB7A GEF, which is indispensable for endocytic trafficking and autophagy. In contrast to unicellular organisms, metazoans have a third member in the complex, RMC1. The cryogenic-electron microscopy structure of the Mon1–Ccz1–RMC1 complex and accompanying functional studies presented here, help to define the evolutionally conserved and metazoan-specific mechanisms in regulating RAB7A activation.

Abstract

Understanding of the evolution of metazoans from their unicellular ancestors is a fundamental question in biology. In contrast to fungi which utilize the Mon1–Ccz1 dimeric complex to activate the small GTPase RAB7A, metazoans rely on the Mon1–Ccz1–RMC1 trimeric complex. Here, we report a near-atomic resolution cryogenic-electron microscopy structure of the Drosophila Mon1–Ccz1–RMC1 complex. RMC1 acts as a scaffolding subunit and binds to both Mon1 and Ccz1 on the surface opposite to the RAB7A-binding site, with many of the RMC1-contacting residues from Mon1 and Ccz1 unique to metazoans, explaining the binding specificity. Significantly, the assembly of RMC1 with Mon1–Ccz1 is required for cellular RAB7A activation, autophagic functions and organismal development in zebrafish. Our studies offer a molecular explanation for the different degree of subunit conservation across species, and provide an excellent example of how metazoan-specific proteins take over existing functions in unicellular organisms.

Rab GTPases (Rabs) and their accessory proteins play indispensable roles in many cellular processes, including autophagy, phagocytosis, pinocytosis, lysosome biogenesis and ciliogenesis, and are critical to define the identity of various cellular organelles, where they serve as key organizers to recruit specific accessory proteins to different organelles, and regulate many steps of vesicular trafficking (1–3). Consequently, mutations or dysregulation in the Rabs and their accessory proteins contribute to various human diseases, including neurologic and neurodegenerative diseases, cancer, and lipid storage disorders (1).

The spatial and temporal activation of Rab proteins are tightly regulated by guanine nucleotide exchange factors (GEFs), which facilitate the transformation of Rabs from the inactive GDP-bound to the active GTP-bound form, and GTPase-activating proteins (GAPs), which function oppositely (4, 5). Whereas many RabGAPs contain a classic Tre-2/Bub2/Cdc16 domain, RabGEFs are more diverse in terms of structure and catalytic mechanism. Currently known RabGEFs include Tri Longin Domain (TLD) proteins, DENN proteins, VPS9 proteins, Sec2 proteins, TRAnsport Protein Particle (TRAPP) complex, and others (REI-1 and Retinitis pigmentosa GTPase regulator (RPGR)) (4, 5).

The Mon1–Ccz1 complex is the sole known GEF for RAB7 and belongs to the TLD RabGEF family as each protein contains three Longin Domains (LDs) (6). The complex, by activating RAB7A (or Ypt7 in fungi), is a critical regulator of the RAB5-RAB7A transition in endosomal maturation, and perhaps mitophagy (7, 8). Homozygous deletion of Mon1 in Drosophila leads to embryonic lethality, emphasizing that tight regulation of RAB7A activity is indispensable for organismal development (9). Mon1 and Ccz1 are evolutionarily conserved and can be found in every eukaryotic organism that we have surveyed (Fig. 1A). In contrast, metazoans have an additional non-TLD protein, RMC1 (also known as C18orf8 or Bulli in flies). RMC1 is exclusively found in metazoans, including vertebrates and invertebrates, but is absent in unicellular eukaryotic organisms (Fig. 1A). RMC1 forms an integrated trimer with Mon1 and Ccz1, and is critical for RAB7A activation, autophagy and lysosomal cholesterol export (10–12). Like the loss of Mon1, deletion of RMC1 in flies impairs RAB7A localization and endosomal maturation (10). However, the loss of RMC1 results in a much milder phenotype in comparison with Mon1.

Fig. 1.

Cryo-EM structure of the metazoan Mon1–Ccz1–RMC1 complex. (A) Cartoon representation incorporating the results of multiple phylogenetic analyses, including Mon1, Ccz1, and RMC1. The gene families that contain Mon1 and Ccz1 are present across eukaryotes, while RMC1 only exists in metazoans. (B) Schematic domain organization of DmMon1–Ccz1–RMC1 subunits. WD40, α-solenoid and LDs (LD1 to LD3) are indicated. Dm: D. melanogaster. (C) Gel filtration profile of DmMon1–Ccz1–RMC1 complex. The horizontal axis represents elution volume, and the vertical axis represents ultraviolet (UV) absorption. The peak and estimated molecular mass of targeted proteins are labeled. The Coomassie blue-stained sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS/PAGE) gel shows the peak fraction of the DmMon1–Ccz1–RMC1 complex from gel filtration. (D) Final Cryo-EM map of the DmMon1–Ccz1–RMC1 complex is shown at 2.87 σ threshold with each subunit colored, respectively (pink for DmRMC1, blue for DmMon1 and green for DmCcz1). (E) Shown are the cryo-EM model and density (2.28 σ threshold) of the DmMon1–Ccz1–RMC1 complex.

An earlier crystal structure of the fungal Mon1–LD1/Ccz1–LD1/Ypt7 (the fungal orthologue of RAB7A) complex has revealed that the LD1 domains of Mon1 and Ccz1 form the catalytic core and are sufficient for the GEF activity of the complex—a mechanism conserved in both fungi and metazoans (13). More recently, the cryogenic-electron microscopy (cryo-EM) structure of the fungal Mon1–Ccz1 complex and accompanying biochemical studies have further illustrated that the Mon1–Ccz1 dimer and the Mon1–Ccz1–RMC1 trimer can be recruited to membranes via the cooperative interaction of phospholipid PtdIns(3)P and RAB5-GTP with Mon1 (7, 14). Nonetheless, the structure of the complete Mon1–Ccz1–RMC1 complex, and more importantly, the functions of the metazoan-specific subunit, RMC1, remain elusive.

In this study, we report a high-resolution cryo-EM structure of the Mon1–Ccz1–RMC1 complex from Drosophila. RMC1 forms a scaffold to contact the LD2 and LD3 of Ccz1, and LD3 of Mon1, opposite of the RAB7A-binding site. Disruption of the assembly of RMC1 with Mon1 and Ccz1 decreases Mon1–Ccz1 stability, RAB7A activation, and endosome maturation. Moreover, using zebrafish as a model organism, we demonstrate that the Mon1–Ccz1–RMC1 assembly is critical for autophagy and organismal development. Together, our work provides molecular insight into the assembly of the Mon1–Ccz1–RMC1 complex and reveals the similarities and differences in regulating RAB7A activation in unicellular organisms and metazoans.

Results

Structure of the Metazoan Mon1–Ccz1–RMC1 Complex.

To gain molecular insight into the assembly and function of the Mon1–Ccz1–RMC1 complex, we expressed and purified recombinant Drosophila melanogaster (Dm) Mon1–Ccz1–RMC1 complex in insect cells. The complex was eluted from size exclusion column with an estimated mass of 180 kDa, consistent with the size of a complex containing one copy of each component (calculated MW of 180 kDa, Fig. 1 B and C). Using single particle cryo-EM, we obtained a three-dimensional (3D) reconstruction of the complex at 3.25 Å resolution (Fig. 1D and SI Appendix, Figs. S1 and S2). The map allowed us to build a model that revealed interactions among all three polypeptides (Fig. 1E and SI Appendix, Figs. S1 and S2). Several large loops, which are predicted to be flexible, were not visible in the map, including the N terminus of DmMon1 (~110 residues), a region likely responsible for RAB5 binding (15) (Fig. 1 B–E).

DmRMC1 consists of two structural domains. The N-terminal segment of RMC1 (residues 1 to 355) folds into a seven-bladed β-propeller that belongs to the WD40 family. The DALI protein structure comparison server (http://ekhidna2.biocenter.helsinki.fi/dali/) revealed that the WD40 domain is most similar to that of yeast Atg18. However, RMC1 lacks the “Phe-Arg-Arg-Gly” motif found in Atg18, which is required for lipid binding (16). The C-terminal segment consists of an array of α-helices and shows an α-solenoid-like fold, which was well resolved in the cryo-EM map (Fig. 1D). Except for the first short helix, the remaining 14 helices of the α-solenoid were arranged into seven bundles, with each bundle containing two antiparallel aligned helices (Fig. 1E). In our structure, the C-terminal α-solenoid of DmRMC1 mediates the interaction with DmMon1 and DmCcz1, with a buried surface area of 2,010.9 Å² (Fig. 1 D and E). In contrast, the WD40 domain is oriented away from DmMon1 and DmCcz1, preventing it from being involved in the interaction. The WD40 domain adopts multiple conformations (Movie S1), thus resulting in a lower resolution density map. Based upon the structure of the DmMon1–Ccz1–RMC1 complex, we truncated human RMC1 from either the C or N terminus and tested the ability to assemble with Mon1 and Ccz1 (SI Appendix, Fig. S3A). Consistent with our structure, the GFP-tagged C terminus of RMC1 (residues 354 to 642), robustly coimmunoprecipitated exogenous Mon1 and Ccz1, like the full-length protein. In contrast, the GFP-tagged N terminus of RMC1 (residues 1 to 370) did not (SI Appendix, Fig. S3A).

Structural Comparison of Metazoan Mon1–Ccz1–RMC1 and Fungal Mon1–Ccz1.

To identify similarities and differences between metazoan Mon1–Ccz1–RMC1 and fungal Mon1–Ccz1, we overlaid our cryo-EM structure with those from Chaetomium thermophilum (Ct) Mon1–Ccz1 (14) and CtMon1–LD1/Ccz1–LD1/Ypt7 (13) (Fig. 2 and SI Appendix, Fig. S3). The sequence identity is 31% between DmMon1 and CtMon1, and 15% between DmCcz1 and CtCcz1, respectively (SI Appendix, Fig. S3B). Despite the low protein sequence identities, the overall configurations of DmMon1 and DmCcz1 are structurally similar to their fungal counterparts, with a rmsd of 2.4 Å and 3.2 Å (393 aligned Cα atoms for Mon1 and 351 aligned Cα atoms for Ccz1, respectively). Similar to CtMon1 and CtCcz1 (Fig. 2 A and B), DmMon1 and DmCcz1 each contains three triangularly organized LDs, and the DmMon1–Ccz1 complex adopts a pseudo twofold symmetry.

Fig. 2.

Structural comparison of metazoan Mon1-Ccz1-RMC1 structure with fungal Mon1-Ccz1 structure. (A) Schematic domain organization of CtMon1-Ccz1 subunits. LDs (LD1 to LD3) are indicated. Ct: Chaetomium thermophile. (B) Cryo-EM structure of CtMon1–Ccz1 complex (PDB:7QLA). LD domains of CtMon1 and CtCcz1 are labeled as indicated. (C) Overlay of the DmMon1–Ccz1–RMC1 complex with the CtMon1–Ccz1 complex. Subunits of metazoan complex are colored as in Fig. 1, and subunits of fungal complex are shown in gray. The 15 α-helices of RMC1 α-solenoid are labeled in order. (D) A zoom-in view of the second α-helix of DmCcz1-LD3 with DmRMC1. (E and F) A zoom-in view of DmMon1 LD2 and LD3 domains with DmRMC1 in comparison with CtMon1. The rotation angle between DmMon1 and CtMon1 is shown as indicated. (G) A zoom-in view of DmCcz1 LD2 domain with DmRMC1 in comparison with CtCcz1. The rotation angle between DmCcz1 and CtCcz1 is shown as indicated.

Notably, limited sequence conservation coincides with multiple biological functions conserved throughout eukaryotic organisms. First, previous structural analysis reveals that the LD1 domains of Mon1 and Ccz1 form a conserved surface to accommodate the small GTPase RAB7A/Ypt7 (13). In our cryo-EM structure, the presence of RMC1 does not alter the conformation of LD1 domains (SI Appendix, Fig. S3C), emphasizing that Mon1 and Ccz1 are the major catalytic subunits to promote nucleotide exchange of RAB7A/Ypt7. Our structural analysis is consistent with the biochemical data that the GEF activity of the Mon1–Ccz1–RMC1 trimer was comparable to that of the Mon1–Ccz1 dimer (7). Second, multiple basic residues in the LD2 and LD3 domains of fungal Mon1 are implicated in phospholipid and membrane binding (14). Many of these residues are highly conserved within metazoan Mon1, such as K284, R304, K306, K508, and K515 in DmMon1 (SI Appendix, Fig. S4), suggesting that the metazoan Mon1–Ccz1–RMC1 complex is likely recruited to membranes in a manner analogous to the fungal Mon1–Ccz1 complex. Third, the LC3-interacting region motif in Ccz1 C terminus is conserved in fungi and metazoans (SI Appendix, Fig. S5), indicating a critical role of the Mon1–Ccz1 dimer and Mon1–Ccz1–RMC1 trimer in the formation of autophagosomes (17).

Despite the overall similarity, the LD2 and LD3 domains of DmMon1 and DmCcz1, which support the interface with DmRMC1, are differently arranged compared with their fungal counterparts (Fig. 2 C–G and SI Appendix, Figs. S4 and S5). First, DmCcz1 possesses an additional α-helix (LD3-α2) and a following loop in its C terminus, both of which are unique to metazoans (Fig. 2D and SI Appendix, Fig. S5). The metazoan-specific loop of DmCcz1 contributes multiple residues to directly contact DmRMC1, including D479 and F483, and this interface is further stabilized by DmCcz1 LD3-α2 (Fig. 2D and SI Appendix, Fig. S5). Second, many residues from DmMon1 and DmCcz1 involved in interactions with DmRMC1 are highly conserved in metazoans, but not in fungi. For instance, R449 in DmCcz1 which forms hydrogen bonds with Q535 and D539 in DmRMC1 is conserved in metazoan Ccz1, but substituted by an aspartic acid residue in CtCcz1 (D720, Fig. 3 and SI Appendix, Fig. S5). Third, the LD2 and LD3 domains of DmMon1 and DmCcz1 undergo large conformation changes compared to CtMon1 and CtCcz1 (Fig. 2 E–G). Specifically, the LD2-α4 of DmMon1 undergoes a nearly 90° rotation in comparison with CtMon1 (Fig. 2E). The LD2-α1 and LD2-α3 of DmCcz1, and LD3-α2 of DmMon1, move toward DmRMC1 by roughly 25°, 14°, and 20°, respectively. Whereas the LD3-α1 of DmMon1 moves away about 13° from DmRMC1 to avoid steric clash (Fig. 2 F and G). Together, these conformational changes create a binding surface for RMC1 (Fig. 2 C–G). Lastly, multiple loops in the LD2 and LD3 domains of CtMon1 and CtCcz1 are missing in metazoan proteins (SI Appendix, Figs. S4 and S5), suggesting certain fungal functions might not be required for metazoans during evolution. Taken together, the metazoan Mon1–Ccz1 complex utilizes unique amino acids and adopts multiple conformational changes to evolve the binding surface for RMC1 (Fig. 2 C–G).

Fig. 3.

The assembly of RMC1 with Mon1 and Ccz1. (A) Detailed interactions of the interfaces between DmRMC1 (pink) and DmCcz1 (green). Secondary structures of DmRMC1 and DmCcz1 are labeled as indicated. (B–E) Shown are hotspots1–4, with residues critical for stabilizing the interface shown in sticks. The rotation angles are shown as indicated. (F) HEK293T cells were transfected with GFP-tagged RMC1 WT, RMC1^HQY/AAA or RMC1^QDR/ARE mutants and subjected to GFP-nanotrap for immunoprecipitation. RMC1 and endogenous Ccz1, Mon1B were detected via antibodies against GFP, Ccz1 and Mon1B, respectively. (G) HEK293T cells were cotransfected with GFP-tagged RMC1 and Flag-tagged Ccz1 WT or Ccz1^NL/AA mutant, and then subjected to Flag-nanotrap for immunoprecipitation. RMC1 and Ccz1 were detected via antibodies against GFP and Flag, respectively. (H) Detailed interactions of the interfaces between DmRMC1 (pink) and DmMon1 (blue). Secondary structures of DmRMC1 and DmMon1 are labeled as indicated. (I–K) Shown are interfaces between DmRMC1 (pink) and DmMon1 (blue), with residues critical for stabilizing the interface shown in sticks. The rotation angle are shown as indicated. (L) HEK293T cells were transfected with GFP-tagged RMC1 WT, RMC1^LRR/AEE or RMC1^DY/RA mutants and subjected to GFP-nanotrap for immunoprecipitation. RMC1 and endogenous Mon1B were detected via antibodies against GFP and Mon1B, respectively. (M) HEK293T cells were cotransfected with GFP-tagged RMC1 and Flag-tagged Mon1B WT or Mon1B^SL/AA mutant, and then subjected to Flag-nanotrap for immunoprecipitation. RMC1 and Mon1B were detected via antibodies against GFP and Flag, respectively. Black dashed lines indicate hydrogen bonds or salt bridges.

Structural Comparison of Metazoan Mon1–Ccz1–RMC1 and the CPLANE Complex.

DmRMC1 adopts a crescent-like shape, with the convex side of the middle region of the α-solenoid domain binding to DmCcz1, and its C terminus interacting with DmMon1 (Figs. 2 and 3). Similar α-solenoid and LD interactions have been observed in multiple vesicular trafficking complexes, including the COPI complex (18), the AP2 clathrin adaptor (19), and the ciliogenesis and planar polarity effector proteins (the CPLANE complex) (20). The composition of the CPLANE complex is analogous to that of Mon1-Ccz1-RMC1. Both complexes contain two TLD RabGEF family proteins (Mon1 and Ccz1, or Inturned and Fuzzy in the CPLANE), and their third members (RMC1 or Wdpcp in the CPLANE) are composed of similar structural domains (Fig. 1B and SI Appendix, Fig. S6A). Furthermore, the CPLANE complex functions as a GEF for Rab23 and also binds small GTPase Rsg1 (21, 22). Thus, the CPLANE complex could function by integrating different GTPases, similar to the Mon1–Ccz1–RMC1 complex.

Subunits of the Mon1–Ccz1–RMC1 and the CPLANE complexes share extremely low protein sequence similarity. Structural and sequence comparison reveals that Mon1 is more similar to Fuzzy than Inturned, whereas Ccz1 is closer to Inturned (SI Appendix, Fig. S6 A and B). An overlay of the Mon1–Ccz1–RMC1 Cryo-EM structure and that of CPLANE, by superimposing DmMon1–DmCcz1 and HsFuzzy–Inturned, identifies multiple conserved features between the two complexes (SI Appendix, Fig. S6 B–E). First, Fuzzy–Inturned adopts a pseudo twofold symmetry similar to Mon1–Ccz1, and both dimer interfaces are mediated by their LD1 and LD3 domains. Second, structural analysis and Alphafold prediction (20) indicate that the LD1 domains of Fuzzy–Inturned contacts Rab23, in a manner analogous to the interaction between Mon1–Ccz1 and Ypt7. Third, the non-TLD members (RMC1 or Wdpcp in the CPLANE) and GTPases bind to the opposite surfaces of the TLD proteins. The non-TLD proteins interact with the TLD proteins predominately through their respective α-solenoid domains and the LD2/LD3 domains.

Despite these similarities, the CPLANE structure varies from that of the Mon1–Ccz1–RMC1 complex in many aspects (SI Appendix, Fig. S6 C–E). First, the α-solenoid domain, but not the WD40 domain of RMC1, is exclusively involved in contacting Mon1 and Ccz1. Unlike RMC1, both domains of Wdpcp are required to contact Inturned and Fuzzy (SI Appendix, Fig. S6D). Second, when Mon1–Ccz1 is superimposed with Fuzzy–Inturned, the main chain direction of RMC1 is almost opposite to that of Wdpcp (SI Appendix, Fig. S6E). Last but not the least, RMC1 engages both Mon1 and Ccz1, forming two interfaces with similar size. In contrast, Wdpcp mainly contacts Inturned, resulting in an interface that is seven times as large as the Wdpcp–Fuzzy interface.

Assembly Mechanism of the Metazoan Mon1–Ccz1–RMC1 Complex.

Next, we examined the detailed interaction between DmRMC1 and DmCcz1 (Fig. 3). Four helices (α6, α8, α10, and α12) from DmRMC1 and a loop between helix α3 and α4, are involved in the interaction between DmRMC1 and DmCcz1 (Fig. 3 A–E). Multiple regions of the LD2 and LD3 domains of DmCcz1, including LD2-α1, LD2-β2 and four loops connecting LD1 and LD2 (Loop1), LD3-β1 and LD3-β2 (Loop2), LD3-β4 and LD3-β5 (Loop3), C terminus of DmCcz1 (Loop4), mediates the interaction with DmRMC1 (Fig. 3 A–E). The interaction between DmRMC1 and DmCcz1 involves multiple hydrogen bonds, salt bridges, and extensive hydrophobic interactions, resulting in a large buried surface area (~1,324 Å²) with four interaction hotspots. 1) The first hotspot is made by the DmRMC1 helix α6 and DmCcz1–LD2-α1 that form an antiparallel two-helix bundle. In addition to multiple hydrophobic contacts formed by their side chains, a hydrogen bond is formed between DmCcz1 E207 and DmRMC1 Y395 (Fig. 3B). 2) Alongside hotspot 1, DmRMC1 helix α8 inserts deeply into a hydrophobic valley formed by Loop1, Loop2, LD2-β2, and LD2-α1 of DmCcz1, thereby forming hotspot 2 (Fig. 3C). Specifically, D502, R505, R506, and Y510 of DmRMC1 are embedded in the hydrophobic groove formed by the alkyl chains of Y188, L189, L191, K193, F196, G226, and S388 of DmCcz1 (Fig. 3C). In addition to hydrophobic contacts, multiple hydrogen bonds are also formed between D502, R505, R506, and Y510 of DmRMC1 and G226, Q387, L189, and L191 in DmCcz1, which further stabilize the interface (Fig. 3C). 3) Hotspot 3 is formed by the DmRMC1 helix α10 and DmCcz1 Loop2, Loop3, and Loop4, and is enriched with hydrophilic interactions, unlike hotspots 1 & 2 (Fig. 3D). For example, multiple salt bridges are formed between D539 and R543 of DmRMC1, and R449 and D479 of DmCcz1 (Fig. 3D). T531 and Q535 of DmRMC1 form multiple hydrogen bonds with the side chains of N385, Q387, and R449 in DmCcz1 (Fig. 3D). 4) The last hotspot consists of helix α12 of DmRMC1 and the C-terminal Loop4 from DmCcz1 (Fig. 3E). Note that Loop4 is specific to metazoan Ccz1, and is missing in fungal counterparts. The interactions are largely mediated by hydrophobic packing of L566 and N569 from DmRMC1 against their neighboring residues F483 and K485 from DmCcz1 (Fig. 3E).

To validate the structure, we incorporated one triple mutation located in human RMC1 helices α8, H519A/Q520A/Y524A (HQY^AAA, corresponding to fly R505/R506/Y510, SI Appendix, Fig. S7), and another triple mutation in α10, Q549A/D553R/R558E (QDR^ARE, corresponding to fly Q535/D539/R543, SI Appendix, Fig. S7). We assessed these mutants in their ability to interact with human Ccz1 and Mon1 (Fig. 3F). As our antibody specifically recognized the human Mon1 orthologue Mon1B, we detected Mon1B in our immunoprecipitation assay. While RMC1 wild-type (WT) robustly retained both Mon1B and Ccz1, the two mutants failed to interact with either of the two partners (Fig. 3F). As these mutants are designed to disrupt the interaction between RMC1 and Ccz1, our findings suggest that Mon1 and Ccz1 cooperatively bind to RMC1 (Fig. 3F). Conversely, a dual mutation in Loop4 of human Ccz1, N477A/L481A (NL^AA, corresponding to fly D479/F483, SI Appendix, Fig. S5), dramatically decreases the interaction with RMC1 (Fig. 3G).

Next to the DmRMC1–DmCcz1 interface, DmRMC1 employs helices α12, α14, and an extended loop following α15 to contact DmMon1, resulting in a significantly reduced buried surface area (~686.6 Å², Fig. 3H). Three loops connecting LD3-β2 and LD3-α1 (Loop11), LD3-β3 and LD3-β4 (Loop12), as well as LD3-β5 and LD3-α2 (Loop13) from LD3 of DmMon1 contact helices α12 and α14 of DmRMC1 in a perpendicular manner, similar to those of DmCcz1 and DmRMC1 (Fig. 3 H–K). Unlike DmRMC1 and DmCcz1, contact between DmRMC1 and DmMon1 is predominantly mediated through hydrophilic interactions. Indeed, three hydrogen bonds are formed between side chains of R565 and K568 in α12 of DmMon1 and the main-chain carbonyls of K474, P495, and V496 (Fig. 3I). Similar to α12, α14 of DmRMC1 also associates with Loop12 and Loop13 of DmMon1 through hydrophilic interactions, including salt bridges formed between DmRMC1 D590 and DmMon1 E476, as well as between DmRMC1 H595 and DmMon1 E494 (Fig. 3J). Notably, the three continuous residues from Loop13 of DmMon1, E494P495V496, are all involved in contacting RMC1 and are conserved in metazoans (SI Appendix, Fig. S4). Last, the loop following DmRMC1 α15 shows a hairpin structure and contacts Loop11 of DmMon1, with DmRMC1 D634 forming a hydrogen bond and a salt bridge with K441 from DmMon1 (Fig. 3K).

We incorporated one triple mutation in the human RMC1 helix α12, L576A/R580E/R583E (LRR^AEE, corresponding to fly I561/R565/K568), and one dual mutation in helix α14, D605R/Y610A (DY^RA, corresponding to fly D590/H595), and examined their interactions with Mon1B and Ccz1 (Fig. 3L). Unlike RMC1 WT, both mutants failed to interact with Mon1B and Ccz1 (Fig. 3L). Similar to the RMC1 and Ccz1 interaction, disruption of interaction between RMC1 and Mon1B also abolished the interaction between RMC1 and Ccz1 (Fig. 3L). In contrast, alanine substitution of two residues from Loop13 of human Mon1B, S493A/L495A (SL^AA, corresponding to the conserved fly E494V496, SI Appendix, Fig. S4), dramatically reduced the association with RMC1 (Fig. 3M).

We further assessed the subcellular localization of different RMC1 constructs (SI Appendix, Fig. S8 A–D). Comparable to RMC1 WT, the C-terminal α-solenoid domain of RMC1 strongly colocalized with endogenous RAB7A in HeLa cells (SI Appendix, Fig. S8 A and B). In contrast, the WD40 domain of RMC1 displayed much weaker colocalization (SI Appendix, Fig. S8 A and B). Furthermore, RMC1 HQY^AAA and LRR^AEE, which did not interact with Mon1 or Ccz1, exhibited a dramatically decreased colocalization with RAB7A, suggesting that the assembly with Mon1 and Ccz1 is required for endolysosomal (endosomal and lysosomal) localization of RMC1 (SI Appendix, Fig. S8 A and B). On the other hand, we observed that RMC1 displayed a strong colocalization with mitochondrial outer membrane protein TOM20 (SI Appendix, Fig. S8 C and D), consistent with a potential role in mitophagy (8, 23). Interestingly, unlike the endolysosomal localization of RMC1, RMC1 HQY^AAA and LRR^AEE did not decrease RMC1 mitochondrial localization, suggesting RMC1 could be recruited to the mitochondria independent of its assembly with Mon1 and Ccz1 (SI Appendix, Fig. S8 C and D).

Disruption of the Mon1-Ccz1-RMC1 Assembly Leads to Defective RAB7A Activation and Abnormal Endolysosomal Morphology.

We next addressed the impact of disruption of the Mon1–Ccz1–RMC1 assembly on its cellular functions. Using CRISPR/Cas9, we generated HeLa cells that were depleted of RMC1 (knockout, KO) (Fig. 4 and SI Appendix, Fig. S9A). Similar to previous studies (11), RMC1 depletion markedly reduced protein levels of both Mon1 and Ccz1, which could be rescued by reexpression of full-length RMC1 (Fig. 4A and SI Appendix, Fig. S9B). Multiple RMC1 fragments or mutants, including the WD40 domain, HQY^AAA, and LRR^AEE, failed to rescue the reduced levels of Mon1 and Ccz1 in RMC1 KO cells, consistent with their inability to associate with Ccz1 and Mon1 (Fig. 4A). Interestingly, while the RMC1 α-solenoid domain was sufficient for interaction with Mon1 and Ccz1, it did not increase the levels of Mon1 and Ccz1. This suggests that both domains of RMC1 are required to stabilize Mon1 and Ccz1 in cells (Fig. 4A).

Fig. 4.

The Mon1-Ccz1-RMC1 assembly is required for cellular RAB7A activation and endolysosomal morphology. (A) Immunoblotting analysis of lysates from parental, RMC1 KO cells and RMC1 KO cells complemented with Flag-tagged RMC1 WT, WD40 (residues 1 to 370), α-solenoid (residues 354 to 642), Ccz1-binding-deficient mutant HQY^AAA or Mon1-binding-deficient mutant LRR^AEE. Mon1B, Ccz1, GAPDH, and RMC1 were detected by indicated antibodies. (B) GST-RILP was used to pulldown lysates from WT, RMC1 KO cells and RMC1 KO cells complemented with Flag-tagged RMC1 WT, WD40, α-solenoid, HQY^AAA or LRR^AEE mutants. The levels of endogenous RAB7A-GTP in the pull-down samples were analyzed by immunoblotting. RMC1, RAB7A, and GST-RILP were detected using the indicated antibodies respectively. WCL: whole-cell lysates. (C) Immunofluorescence analysis of WT, RMC1 KO cells and RMC1 KO cells transfected with GFP or indicated GFP-tagged RMC1 constructs. Nuclei were stained with DAPI, early endosomes were stained with an antibody against EEA1, and late endosomes/lysosomes were stained with an antibody against LAMP1. (Scale bar: 10 μm.) EEA1′ and LAMP1′ shows the zoom-in view of EEA1 and LAMP1 in white boxes. (D) The average size of early endosomes (EEA1 positive) per cell in (C) was measured using ImageJ. Error bars represent mean ± SE. Statistical difference was determined by Student’s t test between two selected groups. *P < 0.05, **P < 0.01, ****P < 0.0001. (E) The average size of late endosomes/lysosomes (LAMP1 positive) per cell in (C) was measured using ImageJ. Error bars represent mean ± SE. Statistical difference was determined by Student’s t test between two selected groups. *P < 0.05, **P < 0.01, ****P < 0.0001.

We next assessed how different RMC1 mutants influenced RAB7A activation. RAB7A activation promotes binding of its effector, the Rab-interacting lysosomal protein (RILP) and thus RILP binding can be used to assess the activation status of Rab GTPases (24, 25). Deletion of RMC1 almost completely abolished RAB7A activation when compared with parent cells, as determined by the amount of RAB7A retained by the effector protein RILP (SI Appendix, Fig. S9C). Reexpression of RMC1 WT restored the normal RAB7A–GTP level (Fig. 4B). In contrast, all tested RMC1 mutants were unable to increase the RAB7A–GTP level (Fig. 4B). These results are consistent with our and previous observations that RMC1 regulates RAB7A activation via stabilization of Mon1 and Ccz1, while exerting no direct influence on their catalytic activity.

To evaluate the importance of the Mon1–Ccz1–RMC1 assembly for the endosomal and lysosomal functions, we measured their relative sizes in RMC1 KO cells complemented with RMC1 WT or mutants (Fig. 4 C and D and SI Appendix, Fig. S9D). Using early endosome antigen 1 (EEA1) as a marker for early endosome, we observed significantly enlarged endosomal structures in RMC1 KO cells, approximately 2.5 times the size of those in parental cells (Fig. 4D). Reexpression of RMC1 WT effectively reduced the size of early endosomes in RMC1 KO cells, while all mutants continued to display enlarged early endosomes (Fig. 4 C and D). Akin to findings with early endosomes, we found that RMC1 WT, but none of the mutants, restored the size of late endosomes/lysosomes, as determined by immunofluorescence of lysosome-associated membrane glycoprotein 1 (LAMP1, Fig. 4 C and E).

Given the well-established role of RAB7A in autophagy (26, 27), we determined the impact of RMC1 loss in autophagy. Of note, RMC1 depletion led to an accumulation of two critical autophagosome proteins SQSTM1/p62 and LC3-II compared with parental cells, suggesting a critical role of RMC1 in basal autophagic flux (SI Appendix, Fig. S10). Taken together, our data suggest the Mon1–Ccz1–RMC1 assembly is critical for stabilizing Mon1 and Ccz1, activating cellular RAB7A activity, maintaining normal endolysosomal morphology, and regulating autophagy.

The RMC1 Assembly with Mon1 and Ccz1 Is Important for Autophagy and Development in Zebrafish.

Previous studies have indicated that Mon1 and RMC1 are critical for embryonic development (9, 10, 28). Loss of Mon1 in flies results in embryonic lethality (9). In contrast, deletion of the C-terminal 181 residues of Mon1, which likely does not impair the interaction with Ccz1 but fails to contact RMC1, leads to multiple developmental defects in flies, including shrinking body size, sterility, and poor motor abilities (28, 29). To assess the importance of the Mon1–Ccz1–RMC1 assembly in metazoan development, we chose zebrafish as a model organism (Fig. 5 and SI Appendix, Fig. S11). Using an antisense morpholino oligonucleotide (MO), we effectively depleted most RMC1 mRNA (SI Appendix, Fig. S11). We found that RMC1 was critical for the swimming ability of zebrafish as RMC1 MO embryos displayed a significantly lower angular velocity relative to WT (Fig. 5A). In addition, depletion of RMC1 dramatically altered body shape of zebrafish and led to curved tails of varying degrees (Fig. 5 B and C). These defects could be largely rescued to the WT level when mRNA encoding human RMC1 was coinjected, indicating that human RMC1 could compensate for the loss of zebrafish RMC1 (Fig. 5 B and C). Conversely, coinjection of mRNAs encoding the WD40 or α-solenoid domains of RMC1 were unable to rescue the defects (Fig. 5 B and C). Similarly, RMC1–HQY^AAA and RMC1–LRR^AEE could not rescue the curly tail phenotype, suggesting that RMC1 associates with Mon1 and Ccz1 to promote normal development in zebrafish (Fig. 5 B and C).

Fig. 5.

The Mon1-Ccz1-RMC1 assembly is critical for development and autophagic functions in zebrafish. (A) Motor abilities of zebrafish larvae injected with control MO (control) and RMC1 MO (MO) were monitored. Swimming abilities of larvae (control or MO) at 72 hpf (hours post fertilization) were recorded in a 48-well plate using the DanioVision system. The larvae were placed in a resting state for 5 min, then stimulated for 11 min by selected point magnifying sound waves. Each circle represents one larva. Error bars represent mean ± SE. About 19 larvae were used in each group. P values were calculated by Student's t test. *P < 0.05. (B) Classification of zebrafish embryos based on their developmental stage, body and tail morphology. Normal, similar to control fish; Delayed, moderately delayed; Malformed, significantly delayed, microcephalic and short-tailed. (Scale bar: 150 μm.) (C) Ratios of embryos in categories of normal, delayed and malformed, upon injection of MO or coinjection of different mRNAs. Control: control; MO: injection of RMC1 MO; MO+WT: coinjection of RMC1 MO and human RMC1 WT mRNA; MO+WD40: coinjection of MO and WD40 mRNA; MO+α-solenoid mRNA: coinjection of MO and α-solenoid mRNA; MO+HYQ^AAA: coinjection of MO and HYQ^AAA mRNA; MO+LRR^AEE: coinjection of MO and LRR^AEE mRNA. The number of embryos used for statistics are shown above each column. (D) CaP axon morphology from 24 hpf embryos injected with control MO, or RMC1 MO alone or together with different mRNAs indicated in (C) at one-cell stage (Top). Bottom: zoom-in views of white box from top pictures. All injections were performed on the Tg [Hb9: GFP]ml2 transgenic zebrafish embryos. (Scale bar: 25 μm.) (E) Statistical results of the length of CaP axons in embryos treated as in (D). For each group, approximately 30 to 40 axons were used for analysis. Error bars represent mean ± SE, ****P < 0.0001; ns, no significant. P values were calculated using one-way ANOVA, Tukey’s multiple comparisons test. (F) Immunoblotting analysis of LC3 and p62 protein levels in injected embryos and uninjected controls. The protein level of LC3 was detected via both long and short exposure. Embryos were injected with RMC1 MO alone or coinjected with mRNAs encoding RMC1 WT or indicated mutants. Protein samples were extracted at 24 hpf (>50 embryos/sample). The LC3-II/LC3-I ratio is shown below.

As depletion of RMC1 impaired swimming ability in zebrafish larvae, we next evaluated the importance of RMC1 for the development of motor neurons using Tg[hb9: GFP]ml2 transgenic zebrafish, which expresses fluorescently tagged early neuronal marker (Fig. 5 D and E). Depletion of RMC1 decreased the length of CaP motor neurons by approximately 20%, which was effectively rescued by coinjection of mRNA encoding human RMC1 WT. In contrast, none of the RMC1 mutants restored the length of the CaP motor neurons (Fig. 5 D and E).

Finally, to assess the importance of the Mon1–Ccz1–RMC1 assembly in regulating autophagy in zebrafish, we examined the expression level of p62 and LC3 in embryos coinjected with RMC1 MO and mRNAs encoding RMC1 WT and mutants (WD40, α-solenoid, HQY^AAA, LRR^AEE). Similar to what we have observed in cultured cells (SI Appendix, Fig. S10), depletion of RMC1 increased the level of p62 and LC3-II, indicating autophagy defects (Fig. 5F). Importantly, RMC1 WT, but none of the mutants we tested, restored the defects (Fig. 5F). Collectively, our study demonstrates that the Mon1–Ccz1–RMC1 assembly is crucial for both autophagy and organismal development in zebrafish.

Discussion

The Rab cascade is critical for many cellular and physiological activities, such as establishing organelle identities, vesicular trafficking, and tumor growth (3, 30, 31). The evolutionally- conserved Mon1–Ccz1 complex mediates RAB5-to-RAB7A transition by acting as an effector of RAB5 and as the GEF for RAB7A (6, 7). RMC1 is a metazoan-specific subunit of the complex, yet its structure and molecular functions remain unclear. We present a high-resolution cryo-EM structure of DmMon1–Ccz1–RMC1 complex, which shows that RMC1 functions as a scaffold to stabilize Mon1 and Ccz1 through extensive interactions. Disruption of the interaction between RMC1 and Mon1 or Ccz1 impairs RAB7 activation in cells, endosomal and lysosomal morphology, and autophagic functions. Our studies illustrate similarities and differences between a critical metazoan complex and its fungal counterpart.

We show that RMC1 and RAB7A bind to opposite surfaces of the Mon1–Ccz1 heterodimer, and the addition of RMC1 does not dramatically alter the conformation of Mon1 and Ccz1 LD1 domains, the catalytic core of the complex. Our data are consistent with previous biochemical data that the Mon1–Ccz1–RMC1 trimer and Mon1–Ccz1 dimer display comparable RAB7A GEF activity. Our study also highlights several major differences between metazoan Mon1 and Ccz1 and their fungal counterparts, which might explain why RMC1 is specifically required in metazoan. First, metazoan Mon1 and Ccz1 possesses multiple unique amino acids that are critical to interact with RMC1. For instance, metazoan Mon1 utilizes an “SPL” motif (S493P494L495 in human Mon1B), and Ccz1 harbors a unique Loop4, to contact RMC1. Both regions are only found in metazoans, but are missing in fungi (SI Appendix, Figs. S4 and S5). Second, DmMon1 and DmCcz1 undergo pronounced conformational rearrangements relative to their fungal counterparts, in order to contact RMC1.

The WD40 domain of RMC1, which is not involved in the interaction with Mon1 and Ccz1, is still required for some RMC1 functions. Its high flexibility (Movie S1) suggests a possibility to be an effector domain to interact with other proteins and/or lipids, which might direct RMC1 to different organelles. Importantly, we identified that the WD40 domain is required for the RMC1 localization to the mitochondria, but not endolysosomes (SI Appendix, Fig. S8). Future studies will be necessary to further dissect the biological functions of RMC1, and to determine its targeting mechanisms to different cellular organelles. In conclusion, our study illustrates similarities and differences between a critical metazoan complex and its fungal counterpart, and paves the way for further understanding the regulation of Rab GTPase activities across species.

Materials and Methods

Materials.

Details of all plasmids, sgRNA constructs, gene fragments, reagents, and antibodies used in this study are summarized in SI Appendix.

Molecular Cloning.

All plasmids used in this study are summarized in SI Appendix, Table S4. The gene encoding the human RMC1 and Mon1B, Ccz1, RILP (residues 241 to 320), and fly Mon1, Ccz1, and RMC1 were synthesized and cloned into multiple vectors, including pEGFPC1, pCDNA3.1⁺, pBIG, or pGEX4T1, for expression in mammalian cells, insect cells, or bacteria. Mutagenesis of plasmids in this study were generated through site-directed PCR.

Protein Expression and Purification.

The fly RAB7A GEF complex was expressed and purified as described before (10). Briefly, His-tagged DmMon1, DmCcz1-3×Flag, and DmRMC1 proteins were cloned into pBIG vector, and expressed in sf9 insect cells. The expression of the GEF complex was conducted via the biGBac system with FuGENE6. After 72 h of infection with recombinant baculoviruses encoding DmMon1–Ccz1–RMC1, the cells were collected and lysed in buffer A (20 mM Tris, pH 8.0, 400 mM NaCl, 10 mM imidazole, 10% (v/v) glycerol) supplemented with 1 mM phenylmethylsulfonyl fluoride, and subjected to high-pressure homogenization. The mixture was centrifuged at 15,000 rpm/min for 45 min, and the supernatant was mixed with Ni-NTA beads at 4 °C for 15 min. Elution of the complex was performed by addition of 300 mM imidazole in buffer A. The eluted proteins were further purified by ion-exchange and size-exclusion chromatography. The purified proteins were stored in buffer B (20 mM Tris, pH 8.0 and 200 mM NaCl), concentrated to 1 mg/mL, and used for cryo-EM.

GST-tagged RILP (residues 241 to 320) was similarly expressed and purified as previously described (32). In brief, the expression of GST-tagged RILP was induced by 0.5 mM IPTG at 30 °C for 5 h. The protein was captured by GSH resin in buffer C (20 mM Tris, pH 8.0, 200 mM NaCl, 1 mM DTT and 2 mM MgCl₂).

Cryogrid Preparation and EM Data Collection.

The cryogrid preparation was performed as previously described (33, 34). We applied 3 µL of protein solution onto M023-Au300-R1/1 grids (NanoDim Tech) that had been glow-discharged for 90 s. The grids were blotted for 2.5 to 3 s and quickly cryocooled in liquid ethane via Vitrobot Mark IV (FEI). Girds were loaded onto the Titan Krios cryoelectron microscope (Thermo Fisher) set at 300 kV, with a 50-μm objective aperture, spot size of 6, magnification of 165,000× (calibrated sampling of 0.85 Å per physical pixel), and equipped with a K2 direct electron detector with BioQuantum energy filter set to 20 eV (Gatan). EPU2.9.1 software (Thermo Fisher) was used to automatically collect data.

Image Processing and 3D Reconstructions.

The movie stacks were motion corrected using Motioncor2 (35). After CTF estimation by CTFFIND4 (36), selected micrographs were subjected to EMAN2.31 (37) for neural network particle picking, with a threshold setting of −0.3 used to maximize inclusion of good particles. The resulting picked particles were extracted in Relion 3.1 (38). After two rounds of two-dimensional classifications, the best classes selected by visual examination were subjected to iterative rounds of cryoSPARC 4.1.1 (39) ab initio reconstruction with multiple classes and heterogeneous refinement to further classify particles. The resulting best 3D classes by visual examination were subjected to cryoSPARC nonuniform refinement and local refinement. To improve the density of the RMC1 N terminus, 3D variability analysis was performed as implemented in CryoSPARC 4.1.1, selecting three modes, with a filter resolution of 8 Å. This mode was then used (using the 3D Variability Analysis Display job type) to split the particles into 20 clusters, and a reconstruction calculated for each. Particles belonging to clusters with the RMC1 N terminus high-resolution features were combined, realigned globally by nonuniform refinement, and subjected to another round of local refinement to generate the final 3.25 Å map. Resolution of these maps was estimated internally in CryoSPARC by gold-standard Fourier shell correlation using the 0.143 criterion. Details for each cryo-EM reconstruction could be found in SI Appendix, Fig. S1 and Table S1.

Model Building and Structure Refinement.

The DmMon1, DmCcz1, and DmRMC1 models were initially created using AlphaFold 2 (40). These models were then fitted into the EM density map using UCSF Chimera1.14 (41), and underwent rounds of manual adjustment and real-space refinement using Coot 0.8.9.2 (42) and Phenix 1.19 (43), respectively. The final structures demonstrated good model geometry, with detailed refinement statistics provided in SI Appendix, Table S1. Structural movie and figures were produced with UCSF ChimeraX1.3 (44) and PyMOL (https://pymol.org/2/).

Cell Culture and Transfection.

HEK293T and HeLa cells were cultured in high-glucose Dulbecco’s modified Eagles medium, supplemented with penicillin–streptomycin 1% (v/v) (Hyclone) and 10% (v/v) fetal bovine serum. Generally, the transfection in HEK293T cells was conducted through polyethylenimine (PEI), and in HeLa cells was through the Lipofectamine Transfection Kit (Thermo Fisher).

Cell Lines.

Generation of RMC1 KO cell line.

The RMC1 KO HeLa cell line was produced by Ubigene Biosciences. CRISPR/Cas9 technology was used to specifically edit the exon 2 region of RMC1. To target the chosen sites, a pair of oligos were annealed and ligated to the YKO-RP006 vector (Ubigene Biosciences Co., Ltd.). The resulting YKO-RP006-hRMC1[gRNA] plasmids containing each target sgRNA sequence were transfected into HeLa cells using the Neon™ Transfection System (Thermo Fisher Scientific). Positive cells were selected using puromycin 24 to 48 h after transfection. After antibiotic selection, a limited dilution method was used to dilute and inoculate a certain number of cells into a 96-well plate. Single RMC1 KO clones were selected 2 to 4 wk later and validated by Sanger sequencing. The sgRNAs for CRISPR design are shown below:

gRNA1(R):5′-ATGCCACCGTGACACCGGGTCGG-3′

gRNA2(F):5′-GTCATTTGGTCTGCCCTAAGAGG-3′

Stable RMC1 WT and mutant overexpression were achieved using lentiviral transduction system. In brief, HEK293T cells were cotransfected in a 3:1 ratio with a lentiviral expression vector (pLVX) and the packaging vectors pMDlg/pRRE, pRSV-Rev and pMD2.G using PEI. The supernatant was harvested at 48 h posttransfection and further concentrated by 5% (m/v) PEG8000. Concentrated lentiviruses were then used to transduce target cells with addition of polybrene (1:1,000). After 48 h, puromycin was used to select the positive cells with stable transgene expression.

Immunoblotting and Immunoprecipitation.

For immunoblotting in cells, HeLa or HEK293T cells were washed twice using PBS, suspended and lysed with SDS loading buffer. The samples were heated for 10 min at 90 °C and subjected to normal immunoblotting procedures. For immunoprecipitations, cells were washed twice using PBS, resuspended, and lysed with lysis buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA and 0.5% NP40, Beyotime) supplemented with cocktail protease inhibitor in a rotator for 30 min at 4 °C. Samples were then centrifuged at 12,000 rpm/min for 10 min, the supernatants were subjected to indicated beads for at least 4 h at 4 °C. After four washes in lysis buffer, samples were mixed with SDS loading buffer and detected against indicated antibodies using immunoblotting.

For immunoblotting in zebrafish, larvae were anesthetized with tricaine, and then homogenized with an oscillator in RIPA buffer (#9806, Cell Signaling) supplemented with a protease inhibitor cocktail (P8340, Sigma) and phosphatase inhibitor (P5726, Sigma). The tissue was broken by ultrasonic shock. The extracts were then spun down at 4 °C for 30 min at 12,000 rpm/min. The expression level of LC3 and p62 from the supernatants were detected by respective antibodies using immunoblotting.

Immunofluorescence.

Immunofluorescence staining for HeLa cells was performed as previously described (45–47). Cells were fixed on a glass slide with 4% PFA, permeabilized with 0.1% Triton X-100, and blocked with 5% BSA in PBS. Cells were then incubated with indicated primary and secondary antibodies, sequentially. Images were acquired by Olympus FV-3000 confocal microscope and analyzed by NIH ImageJ software. All experiments were repeated at least three times.

GST-RILP Pull-Down Assay.

GST pull-down experiments were performed as previously (48). In Brief, 20 μg GST tagged RILP was mixed with Glutathione Sepharose 4B beads in a pull-down buffer (PB: 20 mM Tris, pH 8.0, 200 mM NaCl, 2 mM MgCl₂ and 0.5% NP40) for 2 h. The indicated cell lysates were then added and incubated overnight at 4 °C. The beads were then washed for four times with 1 mL of PB. RILP and RAB7A were detected by immunoblotting via respective antibodies.

Zebrafish Experiments.

Zebrafish strains and husbandry.

AB and Tg[Hb9:GFP]ml2 transgenic zebrafish were used as previously described (49). Zebrafish (Danio rerio) maintenance followed the approval by the West China Hospital and Sichuan University Animal Care and Use Committee. Both embryos and adult fish were raised at 28.5 °C.

Morpholino injection.

The sequence of the rmc1 morpholino antisense oligomer (MO) was 5′-ATCTAGTTGACTTACCTGAAAGCCA-3′, which was synthesized and purchased from Gene Tools, LLC. MO blocked the splicing of rmc1 mRNA. Each embryo was injected with 5 ng rmc1 morpholino oligo.

Rescue experiment.

To reverse the effects of rmc1 knockdown, 1 μL 100 ng/μL rmc1 WT or mutant mRNAs were coinjected with rmc1 MO into embryos. The coding sequence of WT or mutant RMC1 was cloned into pcDNA 3.1+ vector. The recombinant DNA plasmids were linearized and transcribed into RNAs by the mMessage mMachine T7 kit in vitro.

Gene expression analysis by real-time qPCR.

Total RNA was extracted from about 50 zebrafish embryos and reversely transcribed into cDNA. The primers used for rmc1 gene were 5′-ACAGGTGTTTGCGGTACGAT-3′ (upstream primer) and 5′-TGCGCTGAACCGCTAAGATT-3′ (downstream primer); for β-actin gene were 5′-ATGCCCCTCGTGCTGTTTTC-3′ (upstream primer) and 5′-GCCTCATCT CCCACATAGGA-3′ (downstream primer).

Swimming activity assessments.

Zebrafish larvae’ swimming experiments were performed in a 48-well-plate at 72 hpf. Total distance moved and mean velocity in response to stimulation of vibration (2 min rest/2 min vibration) over five cycles were assessed using the DanioVision system (Noldus).

Statistical Analysis.

All statistical analyses of cellular or animal experiments were done in GraphPad Prism 8 using the statistical test indicated in the figure legends.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

3D variability analysis of the DmMon1-Ccz1-RMC1 complex, related to Figure 1. Conformational flexibility of the DmMon1-Ccz1-RMC1 complex. Projection of principal components of motion, showing rotations about three orthogonal axes.

Data, Materials, and Software Availability

Structure factor and atomic coordinates were deposited to Protein Data Bank with ID code 8JBE. The cryo-EM map was deposited in the Electron Microscopy Data Bank with accession code EMD-36143. All other data are included in the manuscript and/or supporting information.