Introduction

The Golgi apparatus is the central organelle in charge of modifying, packaging, and sorting proteins in the secretory pathway. It is estimated that one-third of the human genome encodes for proteins whose function depend on their proper traffic through the Golgi¹. Golgi trafficking is a finely regulated process that depends on the specific budding and fusion of vesicles². Vesicular fusion is a selective process where multiple proteins participate to deliver cargoes to their proper acceptor intracellular compartment². The Complexes Associated with Tethering Containing Helical Rods (CATCHR) are an evolutionarily conserved family of proteins that have been proposed to play key roles in regulating vesicular fusion at different stages of the secretory pathway^3,4. The CATCHR family is composed of five members of Multisubunit Tethering Complexes (MTCs) that are composed of structurally related subunits: NZR, Exocyst, COG, GARP, and EARP. The NZR complex is composed of three subunits and it has been proposed to facilitate the tethering of vesicles on Endoplasmic Reticulum (ER) membranes⁵. The Exocyst complex is composed of eight subunits and it has been proposed to tether vesicles at the Plasma Membrane (PM)⁶. The COG and GARP complexes carry out their tethering functions on Golgi membranes, the COG complex functioning in tethering vesicles at multiple Golgi subcompartments^7,8 and the GARP complex functioning at the Trans-Golgi Network (TGN)^9,10. The EARP complex has been proposed to work as a tethering complex in the endolysosomal (EL) system¹¹ (Figure 1A).

Figure 1.

CATCHR complexes interact with multiple core components of the membrane trafficking machinery, including small GTPases, Coiled-Coil tethers (CCTs), SNAREs, and Sec1/Munc18 (SM) proteins^12–15. Although the molecular mechanisms by which CATCHR complexes coordinate vesicle tethering and fusion remain incompletely understood, it is well established that they are essential for the accurate delivery of transport carriers to their target membranes. Consistent with this central role, mutations in CATCHR subunits cause lethal disorders in humans and depletion of a single subunit is embryonically lethal in mammalians^16–20. Likewise, knockout (KO) and knockdown (KD) cellular models of CATCHRs deficiency exhibit profound defects in intracellular trafficking, Golgi organization, and protein glycosylation ^21–24. Together, these observations underscore the fundamental importance of CATCHR complexes for membrane trafficking throughout eukaryotes. Despite their central role in membrane trafficking, it remains unclear how individual CATCHR complexes are integrated into the broader trafficking machinery to confer pathway specificity. Although numerous pairwise interactions have been described, a systems-level view of the molecular networks surrounding CATCHR complexes has been lacking. To address this knowledge gap, we generated the first comprehensive proximity-interaction map of the Golgi-associated CATCHR complexes COG, GARP, and EARP. Using endogenously expressed TurboID-tagged subunits, we identified high-confidence known interaction partners and established trafficking factors associated with each complex, while also uncovering previously unrecognized interactions and novel candidate regulators of membrane trafficking. These findings provide new insights into the organization of CATCHR-dependent trafficking pathways and reveal distinct molecular modules that cooperate with COG, GARP, and EARP to regulate vesicle tethering and fusion. Our analysis reveals that each Golgi-associated CATCHR complex is embedded within a distinct molecular module that reflects its trafficking specialization. For the EARP complex, this module is centered on the sorting and recycling endosome interface and includes the VPS33B-VIPAS39 SM complex together with regulators of RAB11-dependent recycling, refining the subcellular context in which EARP operates. For GARP, we identify CCDC186 as a bona fide vesicle tether. Mitochondrial relocalization assays combined with electron microscopy demonstrate that CCDC186 is sufficient to ectopically capture transport vesicles, providing direct functional evidence for its tethering activity. Unexpectedly, we identify the tumor suppressor WWOX as a previously unrecognized component of the Golgi trafficking machinery. WWOX localizes to the medial Golgi together with the COG complex, and its depletion disrupts Golgi morphology and both N-and O-linked glycosylation, revealing an essential role in Golgi homeostasis.

Collectively, our study establishes the first comprehensive proximity-interaction map of the Golgi-associated CATCHR complexes COG, GARP, and EARP. Beyond validating known interactions, it defines molecular architecture of CATCHR-dependent trafficking pathways, uncovers previously unrecognized regulators of Golgi and endosomal transport, and provides a conceptual framework for understanding how distinct tethering complexes achieve pathway specificity during vesicle tethering and fusion.

Results

The CATCHR-TurboID constructs are functional and label their proposed compartments and partners

To explore the landscape of the Golgi CATCHRs, we used the proximity labeling system TurboID. TurboID is a fast proximity labeling enzyme engineered from the promiscuous mutant of the bacterial biotin ligase BirA²⁵. This system allows the detection of specific transient and stable interactions as it covalently labels (biotinylates) any protein with surface exposed lysine residues within a ∼10 nm radius from the TurboID-tagged protein. Additionally, this system requires short labeling time, which reduces background noise. To decrease the probability of detecting artifacts, we stably expressed the C-terminally tagged TurboID CATCHR protein into the corresponding KO hTERT-RPE1 cell line. Six cell lines were created: COG4-TurboID and COG6-TurboID to account for the two lobes of the COG complex; VPS52-TurboID and VPS53-TurboID, two of the shared subunits of the GARP and EARP; VPS50-TurboID and VPS54-TurboID to account for the unique subunits of the EARP and GARP, respectively. All constructs were expressed under the regulation of the COG4 promoter region, which allows the expression of tagged CATCHR subunits at near-endogenous levels^22,24. For each construct, a single clone was selected based on its ability to fully rescue the knockout phenotype and to express the TurboID fusion protein at levels comparable to those of the corresponding endogenous protein. One of the cellular defects associated with the depletion of COG complex subunits²⁶ as well as the depletion of VPS54 and VPS53 (GARP subunits)¹⁰ is the increased binding of the lectin HPA, which binds to exposed Tn antigen of underglycosylated proteins. As shown in Figure 1B and 2C, this defect is rescued by the TurboID-tagged protein. For the VPS52 subunit of the GARP and EARP complex, one of the defects associated with its depletion is increased levels of the lysosomal protein LAMP2, a phenotype rescued by VPS52-TurboID. KD of the EARP subunit VPS50 results in decreased abundance of the other subunits that make up the complex¹¹. Similarly, VPS50KO cells have decreased abundance of VPS52 and VPS53, which is rescued by VPS50-TurboID (Figure 1D). To further validate the functionality of the CATCHR-TurboID constructs, we performed immunofluorescence (IF) to assess their localization and labeling region (Figure 1E). As expected, both COG subunits and the GARP subunit VPS54-TurboID localize and biotinylate proteins in the Golgi. Differently, VPS50-TurboID, VPS52-TurboID, and VPS53-TurboID show mostly endosomal localization. This localization is expected for VPS52-TurboID and VPS53-TurboID as these subunits mainly exist as part of the EARP complex¹¹. Additionally, similar to VPS50, VPS52-TurboID and VPS53-TurboID get recruited to RAB4A-positive endosomes (Supplementary Figure 1). Although the localization of VPS52-TurboID and VPS53-TurboID is similar to VPS50-TurboID, closer inspection of the biotinylated region reveals discrete perinuclear puncta that colocalize with the Golgi marker GOLGA4/P230 (Figure 1E), which is consistent with these two subunits also being part of the GARP complex. The specificity of proximity labeling by each CATCHR-TurboID construct was determined by comparison with similarly expressed GFP-TurboID, which is diffusely distributed throughout the cytoplasm and does not preferentially localize to any intracellular compartment. For this reason, GFP-TurboID provides appropriate background control for identifying compartment-specific proximity interactions of peripheral membrane proteins such as the CATCHR complexes. Collectively, these data demonstrate that the CATCHR-TurboID fusion proteins retain biological function, localize correctly within the cell, and enable highly specific mapping of the molecular environments surrounding Golgi and EL CATCHR complexes.

Figure 2.

Golgi CATCHRs integrate within functionally analogous trafficking modules

CATCHR-proximal proteins were labeled by short (15 minute) incubation of cells with biotin followed by affinity purification of biotinylated proteins on streptavidin magnetic beads²⁵. The identity of the biotinylated proteins was detected using label-free Mass Spectrometry (MS). The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE²⁷ partner repository with the dataset identifier PXD071676 and 10.6019/PXD071676. The curated datasets are available in Supplementary Table 1. The labeling experiments were composed of four to five repeats. To assess the quality of the MS results, we first corroborated that the highest hits of each screen were the subunits that make up their corresponding complex, which was indeed the case (Supplementary Figure 2A). We next assessed whether each CATCHR-TurboID constructs preferentially labeled proteins residing in the intracellular compartment where the corresponding complex is known to function. Specifically, we expected the COG and GARP complexes to preferentially label Golgi-associated proteins, whereas EARP was expected to enrich proteins associated with EL membranes. To test this, proteins significantly enriched by each TurboID construct were compared with a curated database of Golgi and EL proteomes^28,29. The resulting datasets were visualized as enrichment plots showing the subcellular distribution of the identified proteins (Figure 2A, Supplementary Table 2). As expected, VPS50-TurboID and VPS52-TurboID showed significant enrichment of endolysosomal proteins, consistent with the established localization of the EARP complex. In contrast, VPS53-TurboID and VPS54-TurboID labeled both Golgi and EL proteins, reflecting the dual role of VPS53 as a shared subunit of GARP and EARP and the dynamic localization of the GARP complex at the TGN-endosome interface. Likewise, COG4-TurboID significantly enriched Golgi-localized proteins, whereas COG6-TurboID exhibited a similar, although not statistically significant, preference for Golgi proteins. The lack of statistical significance for COG6-TurboID resulted from the strong enrichment of a small number of endolysosomal proteins, including EEA1 and IFT20. Overall, these analyses demonstrate that each TurboID constructs selectively labels proteins within the expected intracellular compartment, confirming the high specificity of the proximity-labeling strategy.

Analysis of each of the datasets reveals the enrichment of expected partners as shown in the volcano plots in Figure 2B. The highest hits of all subunits of the investigated CATCHRs were components of the trafficking machinery, including CCTs (Figure 3A), SNAREs and SM proteins (Figure 3B), and other trafficking-related proteins (Figure 3C). Western blot (WB) analysis confirmed the specificity of TurboID labeling (Figure 3D).

Figure 3.

Previous analysis of COG complex localization and function indicated that COG is associated with rims of multiple Golgi cisternae and orchestrates multiple steps in intra-Golgi trafficking^30,31. Consistent with its localization, both COG4-TurboID and COG6-TurboID specifically biotinylated all known Golgi-localized CCTs. In addition to previously described COG-interacting CCTs, such as GOLGA5/Golgin-84³², CUX1/CASP¹³, GOLGA2/M130¹³, TMF1¹³, and USO1/p115³³, our analysis identified several previously unrecognized COG-associated CCTs, including GCC1, GOLGB1/Giantin, TRIP11/GMAP-210, GCC2/GCC185, GOLGA4/Golgin-245, GOLGA3/Golgin-160, and GOLGA1/Golgin-97. Specific proximity labeling of TRIP11/GMAP-210 and CUX1/CASP by both COG4-and COG6-TurboID is shown in Figure 3D. These findings substantially expand the known network of interactions between the COG complex and CCTs. COG4- and COG6-TurboID labeled Golgi CCTs in a similar, yet distinct, hierarchical patterns, indicating that the two COG lobes occupy overlapping yet spatially distinct molecular environments. In descending order of enrichment, COG4-TurboID preferentially labeled GOLGA5/Golgin-84, GCC1, CUX1/CASP, GOLGB1/Giantin, GOLGA2/GM130, TRIP11/GMAP210, and GCC2/GCC185, whereas COG6-TurboID showed particularly strong enrichment of CUX1/CASP, TRIP11/GMAP210, GOLGA5/Golgin-84, TMF1, GOLGB1/Giantin, and GCC2/GCC185. This graded labeling pattern suggests that COG is embedded within an organized Golgi tethering network, and that COG4- and COG6-containing lobes engage partially distinct subsets of Golgi CCTs.

Similarly, COG-TurboID analysis confirmed the proximity of the COG complex to the Qa-SNARE STX5 and most components of its cognate SNARE machinery, including recently identified partners VAMP7 and SNAP29³⁴. In addition to established STX5 partners, COG-TurboID consistently labeled STX6, whereas its canonical partners STX16, VTI1A, and VAMP4 were not detected, suggesting the existence of a previously unrecognized STX5/STX6 Golgi SNARE complex. Analysis of Golgi SNARE proteins revealed a distinct hierarchy of enrichment within the COG proximity proteomes (Figure 3B). STX5 was the highest-ranking SNARE identified by COG4-TurboID, followed by BET1L, GOSR1, BET1, and GOSR2. In contrast, COG6-TurboID preferentially enriched BET1L, GOSR1, and BET1, with STX5 ranking slightly lower but remaining among the most highly enriched SNAREs. Both proximity proteomes also recovered the cognate SM protein SCFD1 and additional Golgi SNAREs at lower abundance, demonstrating selective enrichment of the STX5-dependent membrane fusion machinery. In addition to trafficking factors, the COG proximity proteomes identified several proteins involved in Golgi ion homeostasis. Among the most strongly enriched were the Golgi zinc transporters SLC30A5 and SLC30A6³⁵, together with the putative Golgi pH regulator GPR89A/B³⁶. Proper zinc concentration and luminal pH are essential for Golgi glycosylation and trafficking, raising the possibility that the COG complex contributes to the organization of Golgi ion homeostasis in addition to vesicle tethering and membrane fusion.

Unlike COG, the GARP complex is primarily localized and functions at the TGN⁹. In agreement with GARP localization, members of the TGN trafficking machinery are among the highest hits. It has been shown that the yeast GRIP domain-containing Golgin Imh1p interacts with the GARP complex³⁷. However, no CCT has been associated with the mammalian GARP to date. We found that the human homologs of Imh1p, GOLGA1 and GOLGA4, were among the highest GARP proximity hits, suggesting that the mammalian GARP also interacts with TGN CCTs. Additionally, GARP-TurboID specifically labeled the TGN Qa SNARE STX16 and its partners STX6, STX10 and VTI1A, confirming their functional and physical interaction observed in mammalian³⁸ and yeast¹⁴ cells.

In addition to canonical tethering and fusion factors, the GARP proximity proteomes identified several trafficking regulators that have not previously been directly linked to the GARP complex (Figure 3C). Among the highest-ranking proteins were TBC1D23, a TGN adaptor that bridges endosome-derived vesicles to GOLGA1/Golgin-97 and GOLGA4/Golgin-245 and promotes endosome-to-TGN transport³⁹, and CLINT1/EpsinR, a clathrin adaptor involved in AP-1– dependent trafficking between endosomes and TGN⁴⁰. The enrichment of these proteins further defines the molecular environment of the GARP complex and supports its role as a central organizer of TGN trafficking by integrating vesicle tethering with cargo sorting and membrane transport pathways. Specific proximity labeling of GOLGA4 and TBC1D23 by VPS54-TurboID is shown in Figure 3D.

To define the trafficking machinery associated with EARP, we next examined the enrichment of CCTs and membrane fusion proteins in the VPS50-, VPS52-, and VPS53-TurboID proximity proteomes (Figure 3C). Among proteins involved in membrane tethering, the EARP proximity proteomes preferentially enriched GRIPAP1, CCDC186, and RELCH (Figure 3A and 3C). GRIPAP1 is an established sorting/recycling endosome tether⁴¹ whereas RELCH has been shown to be involved in recycling endosome tethering to the TGN⁴². Their concurrent enrichment suggests that EARP is embedded within a specialized tethering network operating at the sorting and recycling endosome interface.

Analysis of membrane fusion machinery revealed selective enrichment of the VPS33B-VIPAS39 (CHEVI) complex⁴³, identified in all three EARP proximity proteomes (Figure 3B). IF analysis confirmed the colocalization of VPS53 and VPS33B in EL compartments (Figure 4A). Moreover, the endogenous VPS33B was specifically biotinylated by VPS53-TurboID (Figure 4B) and VPS50-TurboID (Figure 3D). Since VPS33B is an SM protein that regulates endosomal Qa SNAREs⁴⁴, we validated VPS53-VPS33B interaction by stably expressing a mutant version of VPS53 (VPS53mut-TurboID) deficient in binding to Qa SNAREs. The mutation (D473A K479A K480A) was made in the MUN-like domain of VPS53 at the predicted binding region of STX16 (manuscript in preparation). The goal of this mutation was to prevent interaction with the SNAREs, which we reasoned would prevent interaction with the SM protein. Indeed, this mutation abolished the proximity of VPS33B and VPS53 (Figure 3B). To account for the functionality of this mutant version of VPS53, we probed for VPS50, whose labeling didn’t change, indicating that the mutant VPS53 is incorporated into the EARP complex. This result points to the EARP complex functioning in the trafficking steps mediated by VPS33B/VIPAS39. In addition, EARP preferentially labeled the endosomal Qa-SNARE STX10 together with multiple components of the endosomal SNARE machinery, consistent with its proposed role in endosomal membrane fusion. Moreover, the EARP proximity proteomes were enriched for proteins associated with cargo sorting, recycling endosome transport, and organelle homeostasis. Among the most prominent were PACS1, a phosphofurin acidic cluster sorting protein that mediates endosome-to-TGN cargo trafficking⁴⁵, and MYO5A, an actin-based motor that functions in Rab11-dependent recycling endosome transport⁴⁶. Interestingly, the VPS52-TurboID screen specifically enriched all V-ATPase subunits, a complex that has been shown to be a TGN pH regulator⁴⁷. Together, these findings further define the molecular environment of the EARP complex and support its role in coordinating cargo sorting, vesicle tethering, organelle homeostasis, and recycling at the sorting/recycling endosome-TGN interface.

Figure 4.

Our analysis also detected close proximity of CATCHRs to several components of Golgi associated ubiquitination machinery, including previously validated GARP partner E3 ubiquitin ligase RNF41⁴⁸, (Figure 3C). Interestingly, one of the highest hits for the COG complex is also a putative E3 ubiquitin ligase, RNF121 (Figure 3C), indicating potential interplay between CATCHRs and Golgi ubiquitinated machinery.

Small GTPases and their regulators were also differentially represented across the CATCHR proximity proteomes (Figure 3C). The EARP-specific subunit VPS50-TurboID selectively labeled the TGN-localized ARF guanine nucleotide exchange factor ARFGEF1/BIG1, whereas the GARP-specific subunit VPS54-TurboID additionally enriched ARFGEF2/BIG2, RAB1B, and RAB3B, consistent with its broader association with Golgi and TGN trafficking. In contrast, the shared GARP/EARP subunits VPS52- and VPS53-TurboID recovered a more extensive set of GTPase regulators, including ARFGAP1, ARHGEF1, ARFGEF1/2, GBF1, and the small GTPases RAB1A, RAB12, RAB3B, and RAB23, reflecting the common molecular environment shared by the GARP and EARP complexes. The COG proximity proteomes exhibited a distinct profile, with COG4-TurboID preferentially labeling ARFGEF1, ARHGAP10, RAB1A/B, RAB7A, RAB10, RAB11B, RAB28, and RABL3, whereas COG6-TurboID enriched ARFGAP3, ARFGEF1, ARHGEF2, GBF1, RAB1A, and Rab-like protein RABL3. Interestingly, another Rab-like GTPase RABL6 was detected in both the GARP and EARP proximity proteomes, suggesting that it may shuttle between the trans-Golgi network and sorting/recycling endosomes, thereby coordinating trafficking at this membrane interface. Together, these findings demonstrate that each CATCHR complex associates with a characteristic repertoire of small GTPases and GTPase regulators, reflecting their distinct functional roles within the secretory and endolysosomal trafficking pathways. The lower enrichment of RAB proteins relative to Golgi CCTs and SNAREs suggests that Rab GTPases occupy a more peripheral and dynamic position within the CATCHR trafficking modules.

Vesicle coat and cargo-sorting proteins were differentially, but not exclusively, distributed across the CATCHR proximity proteomes. EARP, particularly VPS50-TurboID, showed strong enrichment of AP-3 and AP-4 adaptor subunits together with PACS1 and MYO5A, consistent with a recycling endosome-associated sorting environment. GARP/VPS54-TurboID enriched TGN-associated sorting factors, including CLINT1/EpsinR, AP-4 subunits, and COPI components. Notably, AP-4 subunits were also recovered in the COG4- and COG6-TurboID datasets, indicating that AP-4 is not EARP-specific but may represent a broader Golgi/endosomal trafficking module. COG proximity proteomes additionally contained GOPC, SEC23IP/SEC24 proteins, CAV1, and GGA3, suggesting links to COPI/COPII-adjacent and clathrin-associated Golgi trafficking pathways.

Taken together, analysis of the proximity proteome of the CATCHRs suggests that these complexes integrate in higher order trafficking modules composed of CCTs, SNAREs and SM proteins, as well as trafficking and organellar homeostasis regulators. By the defining the molecular modules of the CATCHRs, we looked deeper into the datasets to find novel members of the trafficking machinery.

CCDC186 is a vesicular tether of the endolysosomal system working in the GARP domain

One of the highest hits detected in the GARP and EARP screens was the protein CCDC186, a protein with no known trafficking functions in mammalian cells. CCDC186 is evolutionarily conserved in Metazoans and ubiquitously expressed in human tissues. AlphaFold⁵¹ predicts CCDC186 as an elongated structure, which when in dimers, resembles a CCT. We investigated whether this protein could be a novel CCT functioning in the GARP and/or EARP domains. CCDC186 is primarily Golgi localized, colocalizing with VPS54 (Figure 4C). To analyze its sub-Golgi localization, Golgi ministacks in nocodazole-treated cells were stained for CCDC186, cis-Golgi resident GOLGA2/GM130, and TGN resident TGOLN2/TGN46. Just like VPS54, CCDC186 localizes to the TGN (Supplementary Figure 3A). To assess its potential function as a vesicular tether, we ectopically expressed CCDC186 on the mitochondria with either its N- or C-terminus exposed to the cytosol. The construct that exposes the C-terminus of CCDC186 to the cytosol aggregates multiple mitochondria in a large blob (Supplementary Figure 4B), a phenotype characteristic for other CCTs in this setup⁵². Indeed, electron micrographs show that the ectopic expression of CCDC186 results in tethering of ∼60 nm vesicles to mitochondria membranes (Figure 4D). IF investigation of potential cargo of CCDC186-tethered vesicles indicated specific relocalization of CI-M6PR/IGF2R and not B4GalT1 (data not shown), suggesting that CCDC186, like the GARP complex, participates in tethering of endosome-to-TGN trafficking intermediates. On the other hand, anchoring CCDC186 to the mitochondria by its C-terminus results in smaller aggregated mitochondria which is likely due to protein-protein bridges between mitochondria-attached CCDC186 monomers (Supplementary Figure 3C). These results demonstrate that CCDC186 serves as a CCT that localizes to the TGN.

WWOX is a novel member of the trafficking machinery working in the COG domain

WWOX (WW domain-containing oxidoreductase) has been described as a multifunctional protein associated with neuronal development^53,54 and cancer biology⁵⁵. Evidence shows that downregulation of WWOX is associated with poor prognosis of cancer patients, and different mechanisms have been proposed to explain its pathogenic mode of action^56,57. WWOX was enriched in COG-TurboID proximity samples of RPE1 cells (Figure 2B and 3C) and HEK293T cells (data not shown). Commercial anti-WWOX antibodies did not show any specific signal for IF prompting us to use a myc-WWOX for localization studies. IF revealed that WWOX colocalizes with COG8 on Golgi membranes (Supplementary Figure 5A), specifically to the mid Golgi, tending towards the cis side (LQ=0.27, n=20, 3 cells, Figure 5A, and Supplementary Figure 4B). To assess if WWOX serves any function in Golgi trafficking, we knocked it down using siRNA (Supplementary Figure 4C) used in previous studies⁵⁸. KD of this protein caused a slight but significant decrease in the Golgi area (Figure 5B and Supplementary Figure 4D). This observation led us to investigate how Golgi physiology and trafficking machinery are affected in the absence of WWOX. Similar to the defects associated with depletion of the COG complex, WWOX KD caused N- and O-glycosylation defects as revealed by the increased plasma membrane binding of the lectins HPA (Figure 5C and 5D) and GNL (Supplementary Figure 4E), which bind to the exposed Tn antigen and higher terminal mannose of underglycosylated proteins, respectively. Additionally, we found that the coatomer subunit COPB2 was redistributed off the Golgi (Figure 5E and Supplementary Figure 4F), while SLC30A6, a Golgi resident protein, remained unchanged (Supplementary Figure 4G). Mislocalization of COPB2 has also been observed in the acute inactivation of the COG complex²⁴. Taken together, our results show that WWOX plays a key role in trafficking as its depletion results in altered Golgi morphology and glycosylation defects.

Figure 5.

Discussion

Our work provides the first comprehensive characterization of the landscape of three Golgi/EL CATCHRs: the COG, GARP, and EARP complexes. Using cell lines stably expressing CATCHR-TurboID constructs at near-endogenous levels, we have defined the molecular environment of three evolutionarily conserved CATCHRs. The high specificity of the proximity proteomic datasets relied on the correct localization of the TurboID-tagged CATCHR subunits and preservation of their biological function as demonstrated by rescue of the corresponding KO phenotypes, as well as the use of short labeling times, which minimized secondary and nonspecific biotinylation events. Together, these features enabled precise identification of proteins residing in the immediate molecular environment of each CATCHR complex. As expected, the highest proximity hits of each complex included validated inter-and intra-complex interactions. For instance, each TurboID-tagged subunit specifically biotinylated the other subunits that make up their corresponding complex (Supplementary Figure 2A). Importantly, proximity labeling occurred preferentially within individual COG lobes rather than across the entire COG complex. COG1, COG2, and COG3 were among the most strongly enriched proteins in the COG4-TurboID proximity proteome, whereas COG5 and COG7 were the dominant hits in the COG6-TurboID dataset. Given that all COG subunits are components of a single octameric tethering complex, this differential labeling underscores the high spatial resolution of the TurboID approach. Moreover, these findings provide in vivo support for our model that the COG complex is organized into two subcomplexes (lobes A and B) whose interactions are regulated rather than constitutively fixed⁵⁹.

Importantly, GARP-TurboID labeled STX16, GOLGA1, and RNF41, validated GARP interactors observed in yeast or humans^14,37,48. The strong enrichment of previously characterized GARP partners provided an important internal benchmark that validates the sensitivity and specificity of the TurboID-based proximity-labeling strategy.

Intriguingly, reciprocal proximity labeling between COG and GARP components was observed, with COG-TurboID labeling VPS53 and VPS54 and VPS54-TurboID labeling multiple COG subunits (Supplementary Figure 2A). These findings raise the possibility that Golgi CATCHR complexes do not function as entirely independent entities. One possibility is that COG and GARP cooperate to regulate STX16-dependent membrane fusion events⁶⁰. Alternatively, the two complexes may form hybrid CATCHR assemblies, thereby expanding the repertoire of tethering and fusion machinery available at the Golgi. Such mixed complexes could account for the unexpected redundancy of COG lobe B subunits reported in yeast^12,61 and Drosophila^62,63. One of the most striking features of the CATCHR proximity proteomes was the hierarchical organization of CATCHR-associated trafficking factors. Comparison of the proximity proteomes revealed that each Golgi CATCHR complex is associated with a distinct but evolutionarily conserved trafficking modules composed of CCTs, SNARE and SM proteins, Rab GTPases and their effectors, vesicular coats, as well as trafficking and organellar homeostasis regulators. Among proteins external to the COG complex, Golgi CCTs constituted the most strongly enriched group, with previously established COG interactors, including GOLGA5/Golgin-84, GOLGA2/GM130, GOLGB1/Giantin, and TMF1, ranking among the highest hits, whereas TRIP11/GMAP210, GCC1, GCC2/GCC185, and CASP exhibited progressively lower enrichment. Components of the STX5-dependent membrane fusion machinery formed the next tier of the proximity proteome. Notably, STX5 was the highest-ranking SNARE identified by COG4-TurboID, whereas COG6-TurboID preferentially enriched vesicular components of the cognate SNARE complex, including BET1L, GOSR1, BET1, and GOSR2. Together with previous biochemical studies demonstrating direct interaction between COG4 and STX5^{15,34,64–66}, these findings suggest that the two COG lobes occupy distinct but complementary positions within the membrane fusion machinery.

Alongside several established TGN CCTs, the GARP proximity proteome identified the coiled-coil protein CCDC186, for which we provide direct evidence of vesicle-tethering activity in mammalian cells. The C. elegans homolog of CCDC186, CCCP-1, colocalizes with the EARP complex and is required for dense-core vesicles⁶⁷ biogenesis, whereas the D. melanogaster homolog, Golgin104, functions as a Rab2 effector associated with the Golgi⁶⁸. In addition, the EARP proximity proteomes recovered EIPR1/TSSC1, a protein previously shown to function together with CCDC186 in dense-core vesicle cargo sorting⁶⁹. The concurrent enrichment of CCDC186 and EIPR1/TSSC1 strongly supports the biological relevance of our proximity proteomic datasets and identifies a conserved trafficking module operating at the TGN-endosome interface. Together, these observations indicate that CCDC186 is an evolutionarily conserved CCT that functions predominantly at the TGN while dynamically associating with endosomal membranes, where it cooperates with the EARP complex. Future studies should define the molecular mechanisms governing the dynamic localization of CCDC186 and determine how this CCT, together with EIPR1/TSSC1, coordinates GARP- and EARP-dependent vesicle tethering and cargo transport at the TGN-endosome interface. Additionally, in the EARP neighborhood we also found the SM complex CHEVI, composed of VPS33B and VIPAS39. Our data suggests that EARP may assemble into a larger trafficking module with the CHEVI complex, as its two subunits were among the most enriched proteins in the EARP proximity proteome, and mutations in VPS53 reduced their proximity labeling. Moreover, RAB11FIP5, a Rab11 effector involved in recycling endosome trafficking and part of the FERARI complex⁷⁰, was also strongly enriched. These observations support a model in which EARP cooperates with CHEVI and Rab11-associated factors to coordinate tethering and fusion events at recycling endosomes, an organizational module different than other EL tethering complexes (Supplementary Figure 2B). Future studies should focus on understanding the coordination of the EARP complex with the CHEVI complex.

An unexpected finding was the enrichment of multiple vesicle coat proteins and coat-associated adaptors within the CATCHR proximity proteomes. Although the specific repertoire differed among COG, GARP, and EARP, components of the COPI, AP-1, AP-3, AP-4, and GGA coat machineries were detected in close proximity to one or more CATCHR complexes. These observations suggest two, not mutually exclusive, models. First, remnants of vesicle coats may be retained on trafficking intermediates after vesicle formation, allowing CATCHR complexes to recognize coat proteins as an additional determinant of vesicle identity during tethering. Such a mechanism would complement the established roles of Rab GTPases, CCTs, and SNARE proteins in specifying vesicle targeting. Alternatively, vesicle docking and fusion sites may be positioned immediately adjacent to vesicle budding domains within Golgi and endosomal membranes. In this model, CATCHR complexes would not only mediate vesicle tethering but also spatially coordinate vesicle formation, docking, and fusion, thereby coupling successive stages of membrane trafficking. Future studies combining high-resolution imaging and functional perturbation of coat components will be required to distinguish between these models. Importantly, the proximity proteomes identified several regulatory proteins not previously linked to the CATCHRs. For example, the transmembrane RING finger E3 ubiquitin ligase RNF121⁷¹, the oxidoreductase WWOX, the Golgi zinc transporters SLC30A5/ZnT5 and SLC30A6/ZnT6, and the putative Golgi pH regulator GPR89 that were detected in the COG-TurboID screen. Together with WWOX (see below), RNF121 raises the possibility that the COG complex recruits regulatory enzymes in addition to canonical tethering and fusion factors, thereby coupling membrane trafficking to post-translational regulatory and signaling pathways. Likewise, the enrichment of SLC30A5/6 and GPR89 suggests that the COG complex may contribute to maintaining the physicochemical environment of the Golgi, as ion homeostasis and luminal pH are essential for glycosyltransferase activity^72,73. It is also possible that the CATCHRs also recognize ion transporters and pH sensors on acceptor membranes to guarantee fusion to the appropriate compartment. Similarly, VPS52, the shared subunit of the GARP and EARP, could be serving this recognition domain of the complexes, as VPS52-TurboID detected RNF41 and all subunits of the V-ATPase complex. Together, these findings indicate that, in addition to assembling conserved Rab-, CCT-, SNARE-, and SM-dependent trafficking modules, individual CATCHR complexes integrate distinct accessory factors that tailor their functions to specific trafficking pathways. Future studies should determine how these regulatory proteins cooperate with CATCHR complexes and whether disruption of these networks contributes to the glycosylation and trafficking defects characteristic of COG-, GARP-, and EARP-associated disorders.

Among the newly identified COG-associated proteins, WWOX emerged as a particularly compelling candidate regulator of Golgi function. WWOX localized predominantly to the medial Golgi, where it partially colocalized with the COG complex, and its depletion caused Golgi fragmentation, defects in both N-and O-glycosylation, and partial mislocalization of the COPI coat-phenotypes that closely resemble those observed following COG depletion. Consistent with a role in Golgi trafficking, WWOX was previously shown to interact with the ER-Golgi trafficking factor SEC23IP and the Golgi CCT TMF1⁷⁴, both of which were also identified in our COG proximity proteome (Figure 3A and 3C). Although the molecular mechanism of WWOX remains unknown^53,75,76, its tandem WW domains could function as a scaffold to recruit PPxY-containing trafficking proteins to Golgi membranes⁷⁵, whereas its C-terminal short-chain dehydrogenase/reductase (SDR) domain may regulate Golgi physiology through local NAD(H)-or NADP(H)-dependent redox reactions⁷⁷. Future studies should define how WWOX contributes to COG-dependent trafficking and Golgi homeostasis. More broadly, our findings establish an unexpected connection between WWOX and the Golgi trafficking machinery, providing a potential mechanistic link between defects in membrane trafficking and the neurodevelopmental disorders and cancers associated with WWOX dysfunction.

Although TurboID labeling is influenced by protein abundance, lysine accessibility, and interaction dynamics, the reproducible hierarchy observed across all three CATCHR complexes strongly suggests that it reflects the spatial organization of specialized trafficking modules. Remarkably, this hierarchy mirrors the proposed sequence of events during vesicle tethering, beginning with Rab-mediated vesicle targeting, followed by long-range capture by coiled-coil tethers, and culminating in CATCHR-dependent assembly of the cognate SNARE fusion machinery. Together, these findings support a model (Figure 6) in which CATCHR complexes function as central organizing hubs that coordinate successive stages of vesicle tethering and membrane fusion while establishing compartment-specific identity through selective association with CCTs, SNAREs, SM proteins, and pathway-specific regulatory factors. Collectively, our work provides the first comprehensive molecular framework for the organization of Golgi-associated CATCHR complexes and reveals a fundamental principle of membrane trafficking: rather than functioning as isolated vesicle tethers, CATCHR complexes assemble specialized trafficking modules that integrate long-range capture by CCTs, SNARE-dependent membrane fusion, and acceptor membrane molecular cues for selective fusion. By defining these molecular networks and identifying previously unrecognized regulatory components, our study provides a foundation for elucidating the mechanisms that maintain Golgi homeostasis and for understanding how their disruption contributes to congenital disorders of glycosylation, neurodevelopmental diseases, and cancer.

Figure 6.

Materials and methods

Cell line creation and cell culture

Retinal Pigment Epethelial (hTERT-RPE1), HeLa WT, and HEK293T WT cells used for all experiments were purchased from ATCC. WT and KO cell lines were maintained in DMEM/F12 supplemented with 10% FBS. TurboID-expressing cells were maintained in biotin-free DMEM supplemented with 10% FBS and 10 μg/ml of avidin. Cells were incubated in a 37°C incubator with 5% CO2 and 90% humidity.

RPE1 KO cell lines were created as described in previous work from our lab^80–82. These cells were rescued using lentiviral transduction of CATCHR-TurboID constructs. Lentiviruses were made in-house as described previously^80,81. The plasmids used for lentivirus formation along with the packaged plasmids containing the CATCHR-TurboID sequences are listed below (Table 1). All constructs were cloned into a destination vector containing a G418 resistance gene and the COG4 promoter previously described⁸³. Cells were selected with 500 μg/ml of G418. Resulting cells were single cell sorted and individual clones were characterized by Western blot and selected based on the expression level of CATCH-TurboID hybrid closest to the levels of the endogenous protein.

Table 1.

Proximity labeling assays

Cells were grown in a 15 cm dish (MS) or 10 cm dish (WB) maintained in biotin-free media for at least 72 h. To initiate biotinylation, cells were incubated with media supplemented with 100 μM biotin for 15 min for MS analysis, 30 min for WB and IF.

For MS, after biotinylation cells were placed on ice and washed four times with ice cold PBS. Cells were then lysed with ice cold RIPA buffer supplemented with 100 μM PMSF and Halt protease inhibitor as per the manufacturer’s instructions. Lysate was centrifuged at 4°C for 10 min at 13,000 g to pellet cell debris and clarified lysate (1mg of total protein quantified with BCA assay) was added to streptavidin magnetic beads and allowed the beads to enrich with the biotinylated proteins overnight. After enrichment, beads were washed with RIPA buffer and ice-cold PBS. To remove detergents, the beads were washed with 1 M KCl, 0.1 M Na2CO3, 2 M urea in 10 mM Tris-HCl (pH 8.0), and lastly with 100 mM ammonium hydrogen carbonate (AMBIC). AMBIC was removed, and beads were immediately stored at-80°C until MS processing time.

For WB the biotin labeling was extended to 30 min and the biotinylated proteins were isolated with the same protocol described above. After the protein binding step, the beads were washed with RIPA buffer and PBS and then 2X sample buffer was added and heated to 95°C for 5 min to elute biotinylated proteins.

Proteomics analyses

Protein samples were reduced, alkylated, and digested using filter-aided sample preparation (IP samples) or following chloroform/methanol extraction (input lysates) with sequencing grade modified porcine trypsin (Promega). Tryptic peptides were then separated by reverse phase XSelect CSH C18 2.5 um resin (Waters) on an in-line 150 x 0.075 mm column using an UltiMate 3000 RSLCnano system (ThermoFischer). Peptides were eluted using a 60 min gradient from 98:2 to 65:35 buffer A:B ratio. Eluted peptides were ionized by electrospray (2.2 kV) followed by mass spectrometric analysis on an Orbitrap Exploris 480 mass spectrometer (Thermo). To assemble a chromatogram library, six gas-phase fractions were acquired on the Orbitrap Exploris with 4 m/z DIA spectra (4 m/z precursor isolation windows at 30,000 resolution, normalized AGC target 100%, maximum inject time 66 ms) using a staggered window pattern from narrow mass ranges using optimized window placements. Precursor spectra were acquired after each DIA duty cycle, spanning the m/z range of the gas-phase fraction (i.e. 496-602 m/z, 60,000 resolution, normalized AGC target 100%, maximum injection time 50 ms). For wide-window acquisitions, the Orbitrap Exploris was configured to acquire a precursor scan (385-1015 m/z, 60,000 resolution, normalized AGC target 100%, maximum injection time 50 ms) followed by 50x 12 m/z DIA spectra (12 m/z precursor isolation windows at 15,000 resolution, normalized AGC target 100%, maximum injection time 33 ms) using a staggered window pattern with optimized window placements. Precursor spectra were acquired after each DIA duty cycle.

Buffer A = 0.1% formic acid, 0.5% acetonitrile

Buffer B = 0.1% formic acid, 99.9% acetonitrile

Following data acquisition, data were searched using Spectronaut (Biognosys version 18.3) against the UniProt Homo sapiens database (Proteome ID: UP000005640, Taxon ID: 9606, 3rd version of 2023) using the directDIA method with an identification precursor and protein q-value cutoff of 1%, generate decoys set to true, the protein inference workflow set to maxLFQ, inference algorithm set to IDPicker, quantity level set to MS2, cross-run normalization set to false, and the protein grouping quantification set to median peptide and precursor quantity. Fixed Modifications were set to Carbamidomethyl (C) and variable modifications were set to Acetyl (Protein N-term), Oxidation (M). Protein MS2 intensity values were assessed for quality using ProteiNorm. The data was normalized using Cyclic Loess and analyzed using proteoDA to perform statistical analysis using Linear Models for Microarray Data (limma) with empirical Bayes (eBayes) smoothing to the standard errors. Proteins with an FDR adjusted p-value < 0.05 and a fold change > 2 were considered significant.

Proteomic visualization charts were made using R studio and all codes have been deposited in Github, Lupashin-Lab/wsaragon.

Western blotting

Protein sample loaded into the 8-16% Bio-Red SDS gel and transferred to 0.22 μ nitrocellulose membrane using semi-dry transfer unit (ThermoFischer). Membranes were stained with Ponceau S stain to assess loading and then blocked in Bio-Red blocking buffer for subsequent incubation with primary and secondary antibodies (Table 2). Washed membranes were scanned on LICORBio Odyssey scanner.

Table 2.

Immunofluorescence

Cells were plated on glass coverslips to 80%–90% confluency and fixed with 4% PFA diluted in PBS for 15 min at room temperature. Cells were then permeabilized with 0.1% Triton X-100 in PBS for 1 min, followed by 50 mM ammonium chloride wash for 5 min to remove remaining fixative. Then, cells were blocked with 1% BSA and 0.1% saponin in PBS for 20 min for subsequent incubation with primary antibody (diluted in 1% cold fish gelatin, 0.1% saponin in PBS) for 45 min, washed with PBS and then incubated with fluorescently conjugated secondary antibodies for 30 min and thoroughly washed with PBS to remove unbound antibody. Coverslips were then mounted on glass microscope slides using Prolong Gold antifade reagent (Life Technologies). Cells were imaged with a 63× oil 1.4 numerical aperture objective of an LSM880 Zeiss Laser inverted microscope and Airyscan super-resolution Cocchiaro microscope using ZEN software. Quantitative analysis was performed using single-slice confocal images. All the microscopic images shown are Z-stacked Maximum Intensity Projection images.

Electron microscopy

Samples were processed using a modified version of the protocol developed by Valdivia and colleagues⁸⁴. Briefly, cells were fixed on ice for 20 min in 2.5% glutaraldehyde and 0.05% malachite green (EMS) prepared in 0.1 M sodium cacodylate buffer (pH 6.8). Samples were then post-fixed for 30 min at room temperature in 0.5% osmium tetroxide and 0.8% potassium ferricyanide in 0.1 M sodium cacodylate buffer, followed by incubation in 1% tannic acid for 20 min on ice and 1% aqueous uranyl acetate for 1 h at room temperature. Specimens were dehydrated through a graded ethanol series and embedded in Araldite 502/Embed 812 resin (EMS). Ultrathin sections (50 nm) were prepared using a Leica UltraCut UCT ultramicrotome and post-stained with aqueous uranyl acetate and Reynolds’ lead citrate (EMS). Images were acquired on an FEI Tecnai TF20 transmission electron microscope operated at 80 kV and equipped with an FEI Eagle 4k digital camera controlled by FEI acquisition software.

Golgi localization studies

To form ministacks, cells were treated with 33 μM nocodazole and stained using the IF protocol described. For the cells transiently expressing myc-WWOX, transfection was done 24 h prior to nocodazole treatment. Location Quotient (LQ) was calculated as described by the Liu lab⁸⁵. Briefly, distance between the centers of mass of TGOLN2 (trans) and GOLGA2 (cis) were used as the denominator and the distance between the center of mass of Myc-WWOX and GOLGA2 was used as numerator, average ratio is the LQ.

Mitochondria relocalization studies

For mitochondrial relocalization experiments, full-length human CCDC186 and a DNA fragment encoding 3×HA-OMP25⁸⁶ were synthesized by GenScript. To generate an N-terminal mitochondrial targeting construct, CCDC186 was subcloned into a previously described TOM20-Dronpa vector⁶⁴. To generate a C-terminal mitochondrial targeting construct, CCDC186 lacking the stop codon was fused in-frame to 3×HA-OMP25 and cloned into pcDNA3.1. RPE1 cells or HeLa cells were transfected 24 h before IF or EM processing.

Data availability

The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD071676 and 10.6019/PXD071676. The curated datasets are available in Supplementary Table 1.

Acknowledgements

We thank our colleagues who generously provided reagents used in this study. We are particularly grateful to Tetyana Kudlyk for technical assistance with the generation of CATCHR-TurboID constructs and to the members of the Lupashin laboratory for valuable discussions and critical feedback. We also thank the UAMS Digital Microscopy, Proteomics, and DNA Sequencing Cores for access to their facilities and for their expert technical support.

Additional information

Funding

This work was supported by the NIH/NIGMS grant R01GM083144 for Vladimir Lupashin. IDeA National Resource for Quantitative Proteomics is supported by NIH/NIGMS grant R24GM137786

Abbreviations

• CATCHR: Complexes Associated with Tethering Containing Helical Rods

• CCT: Coiled-Coil Tether

• COG: Conserved Oligomeric Golgi

• EL: EndoLysosomal system

• ER: Endoplasmic Reticulum

• GARP: Golgi-Associated Retrograde Protein

• EARP: Endosome-Associated Recycling Protein

• IF: Immunofluprescence

• KD: Knock Down

• KO: Knock Out

• MTC: Multisubunit Tethering Complex

• MS: Mass Spectrometry

• NRZ: NAG, RINT1, and ZW10

• PM: Plasma Membrane

• SM: Sec1/Munc18

• SNARE: Soluble N-ethylmaleimide-sensitive factor Attachment protein REceptor

• WB: Western blot

Funding

HHS | NIH | NIH Office of the Director (OD) (R01GM083144)

• Vladimir Lupashin

Additional files

Supplementary Figures. Supplementary Figure 1. Stably expressed VPS50-, VPS52-, and VPS53-TurboID are recruited to RAB4A-positive endosomes. RPE1 cells stably expressing GARP-and EARP-TurboID subunits were transiently transfected with mGold-RAB4A. Immunofluorescence analysis demonstrates VPS50-, VPS52-, and VPS53-TurboID (red) get recruited to RAB4A-positive compartments (green) while VPS54-TurboID (red) does not, supporting the endosomal localization of VPS52-and VPS53-TurboID is due to their association with the EARP complex. Supplementary Figure 2. Specific labeling of individual subunits of Golgi/endolysosomal multisubunit tethering complexes. Dot size represents average protein abundance and color indicates log2 fold enrichment relative to GFP-TurboID controls. (A) VPS50-, VPS52-, and VPS53-TurboID specifically labeled proteins of the EARP complex. Similarly, VPS52-, VPS53-, and VPS54-TurboID labeled proteins of the GARP complex. VPS54-TurboID also labeled the subunits of the COG complex. The COG-TurboID constructs specifically labeled all the subunits that make up the COG complex, in addition to labeling VPS53 and VPS54, subunits of the GARP complex. (B) Subunits of the EARP complex specifically labeled both subunits of the CHEVI complex, while no specific labeling of other endolysosomal tethering complexes was detected (CORVET, HOPS, and FERARI). Supplementary Figure 3. CCDC186 localizes to the trans-Golgi region and it participates in vesicular tethering vis its C-terminus. (A) High-resolution immunofluorescence (IF) analysis of endogenous CCDC186 (red) localization relative to the trans-Golgi marker TGOLN2/TGN46 (blue) and the cis-Golgi marker GOLGA2/GM130 (green). Line plot highlighting the colocalization of CCDC186 with TGOLN2/TGN46. Four representative images are shown. (B) Transient expression of CCDC186 anchored to the mitochondria by either its N-(TOM20-CCDC186) or C-terminus (CCDC186-OMP) in RPE1 cells. Relocalization of both constructs resulted in perturbation of the mitochondrial network. As shown in Figure 4D, N-terminally anchored CCDC186 causes mitochondria connections via vesicles, while C-terminally anchored CCDC186 results in formation of electron-dense bridges (C) between neighboring mitochondria. Supplementary Figure 4. The medial-Golgi protein WWOX is necessary for Golgi morphology and glycosylation. (A) myc-WWOX (red) colocalizes with COG8 (green) forming ring-like structures similar to those seen for (B) GOLGB1/giantin (green). (C) Two biological repeats of the knockdown (KD) efficiency of WWOX siRNA measured with RT-PCR. (D) Immunofluorescence field showing the Golgi area (GOLGA2/GM130) of control (CTRL) and WWOX KD RPE1 cells. Scale bar 50 μm (E) Increase in the plasma membrane GNL binding of WWOX KD cells in comparison to CTRL, consistent with impaired N-glycosylation (n = 30 cells) (F) Representative IF images showing mislocalization of COPB2 in WWOX KD cells relative to control. (G) WWOX KD does not affect the localization of the Golgi-resident protein SLC30A6 (Pearson colocalization coefficient, n = 30 cells).

Supplementary Table 1.

Supplementary Table 2.