MON2 Guides Wntless Transport to the Golgi through Recycling Endosomes

Endocytic cargos are transported to recycling endosomes (RE) but how these sorting platforms are generated is not well understood. Here we describe our biochemical and live imaging studies of the conserved MON2-DOPEY complex in RE formation. MON2 mainly co-localized with RE marker RAB4B in peripheral dots and perinuclear region. The peripheral RE approached, interacted with, and separated from sorting nexin 3 (SNX3)-positive early endosomes (EE). Membrane-bound DOPEY2 was recruited to RE dependent upon MON2 expression, and showed binding abilities to kinesin and dynein/dynactin motor proteins. MON2-knockout impaired segregation of RE from EE and led to a decreased tubular recycling endosomal network, whereas RE was accumulated at perinuclear regions in DOPEY2-knockout cells. MON2 depletion also impaired intracellular transferrin receptor recycling, as well as retrograde transport of Wntless during its passage through RE before delivery from EE to the Golgi. Together, these data suggest that the MON2 drives separation of RE from EE and is required for efficient transport of endocytic cargo molecules.


Introduction
A balance between synthesis and degradation of proteins is required to maintain the homeostasis of cellular processes (Cullen and Steinberg, 2018). The endosomal system contributes to this balance through vesicular sorting of proteins for degradation and recycling to various membrane compartments. Mammalian cells contain a complex set of endosomal compartments, including early endosomes (EE), late endosomes (LE), recycling endosomes (RE), and lysosomes (Klumperman and Raposo, 2014). The EE are the first carrier vesicles into which plasma membrane (PM) proteins and lipids enter when internalized (Jovic et al., 2010). EE undergo structural and functional maturation to form either LE (Scott et al., 2014), whose cargos are transported to lysosomes for degradation or RE, whose cargos are sorted and transported back to the PM, or the Golgi (Taguchi, 2013). Although RE play pivotal roles in maintaining cellular homeostasis, the mechanism for how they regulate protein and lipid sorting is not well understood.
RE are mainly localized at the perinuclear regions as transient vesicular/tubular structures as well as peripheral dot-like puncta (Grant and Donaldson, 2009;Maxfield and McGraw, 2004). The formation and maintenance of endosomal compartments involve a host of different Rab GTPases (Pfeffer, 2013;Pfeffer, 2017) whose GTPbound state at membranes is required for vesicle formation from the donor or vesicle fusion with the acceptor. RAB4 is found at tubular structures of EE or RE (D'Souza et al., 2014;de Wit et al., 2001;Perrin et al., 2013;Sonnichsen et al., 2000), while RAB11 is mainly enriched in RE (Green et al., 1997;Ullrich et al., 1996). To date, several multiprotein complexes have been reported that cooperate to generate and move tubules from endosomes (Gautreau et al., 2014). Rab GTPases and kinesin provide the mechanical force for tubularization and fission of tubular/vesicular structures from the EE membrane (Delevoye et al., 2014;Etoh and Fukuda, 2019;Shakya et al., 2018).
Most kinesins mediate transport toward the plus-ends of microtubules (Hirokawa et al., 2009). However, RE are clustered around the microtubule-organizing center (MTOC) (Maxfield and McGraw, 2004) located at the minus-ends of microtubules (Martin and Akhmanova, 2018). Thus, it is likely that coordination of other factors and motor proteins regulates RE tubular network and movement.
Yeast Mon2 is a peripheral membrane protein in the trans-Golgi network (TGN) (Efe et al., 2005) that, with its cytosolic binding partner, Dop1, is implicated in endosome-to-Golgi transport (Gillingham et al., 2006). We recently described a role for yeast Dop1 in retrograde transport from the TGN to the Golgi (Zhao et al., 2019). The yeast TGN is assumed to be functionally equivalent to mammalian EE and RE (Day et al., 2018). Mon2 and Dop1 has been highly conserved through evolution. In metazoans, there are two Dop1 orthologues, named DOPEY1 and DOPEY2 (Barbosa et al., 2010;Mahajan et al., 2019;McGough et al., 2018). Even DOPEY1 is monomer and DOPEY2 is self-interacted, both DOPEY1 and DOPEY2 were reported to interact with mammalian MON2 (Mahajan et al., 2019;McGough et al., 2018). DOPEY1 interacts with the KLC2, the light chain of kinesin-1 motor protein (Mahajan et al., 2019), suggesting that MON2-DOPEY complex mediates vesicular/tubular movement.
Retrograde transport of Wntless (Wls), which is a critical cargo receptor for Wnt secretion (Banziger et al., 2006), from EE to the Golgi requires mammalian MON2- DOPEY2 (McGough et al., 2018). After internalization of Wls from the PM, it is transported to EE, and then transported to the Golgi (Harterink et al., 2011;Port et al., 2008). In Caenorhabditis elegans, Wls is transported from RE to the Golgi (Bai and Grant, 2015). However, it remains unclear whether Wls is directly transported from EE to the Golgi or through RE in mammalian cells. It is because RE are mainly localized at the perinuclear Golgi regions as transient vesicular/tubular structures as well as peripheral dot-like puncta (Grant and Donaldson, 2009;Maxfield and McGraw, 2004).
Recent findings suggest that both the TGN and RE are functionally overlapping and RE attach to the trans-side of Golgi stacks under nocodazole treatment (Fujii et al., 2020).
It is still not clear how MON2-DOPEY proteins are involved in Wls retrograde trafficking.
Here, we show that the MON2 mainly localizes at RE in human embryonic kidney 293 (HEK293) cells. Membrane-bound DOPEY proteins were recruited to RE dependent upon MON2 expression. MON2 knockout (MON2-KO) impaired the segregation of RE from EE. Wls passed through RE dependent upon MON2 for transport to the Golgi. These findings suggest that the MON2-DOPEY complex regulates cargo transport by regulating RE movement.

Plasmids
The plasmids for target gene KO were constructed as previously described (Liu et al., 2018). Human genes were amplified from cDNAs isolated from HEK293 or HeLa cells. Alternatively, cDNAs from Ultimate™ ORF LITE clones (Thermo Fisher) were used as templates. In-fusion PCR cloning or restriction enzyme digestion followed by ligation was used for plasmid construction. Full-length DOPEY1 was ligated into the XhoI/MluI site of pME-Hyg-3FLAG to generate pME-Hyg-DOPEY1-3FLAG. pME-Hyg-DOPEY2-3FLAG was obtained by in-fusion cloning of full-length DOPEY2 into the EcoRI/MluI site of pME-Hyg-3FLAG. pME-Hyg-DOPEY1-FLAG-EGFP and pME-Hyg-DOPEY2-FLAG-EGFP were obtained by replacement of the 3FLAG fragment in pME-Hyg-DOPEY1-3FLAG or DOPEY2-3FLAG with the FLAG-EGFP fragment, respectively.
The plasmid pME-Wls-EGFP was obtained by in-fusion cloning of full-length Wls into the EcoRI/MluI site of pME-EGFP. To construct pME-Hyg-HA-Wls as described previously (Varandas et al., 2016), we inserted the HA fragment into the largest extracellular loop and C-terminus of Wls. The plasmid mScarlet-Giantin was obtained from Addgene (#85048). The plasmid mNeonGreen-Giantin was obtained from Addgene (#98880). pME-KLC2-3HA was obtained by in-fusion cloning of full-length KLC2 into the XhoI/MluI site of pME-3HA. pME-3HA-DCTN1 was obtained by infusion cloning of full-length DCTN1 into the MluI/NotI site of pME-3HA. pME-sOGT-3HA was obtained by in-fusion cloning of full-length sOGT into the MluI/NotI site of pME-3HA.

Confocal microscopy
For fixed samples, glass-bottom dishes (#D35-10-1-N; Cellvis, Mountain View, CA) or sterile glass coverslips in culture dishes were coated with 0.1% gelatin (#G7041; Sigma-Aldrich) for 15 min at room temperature, followed by complete removal of excess gelatin. After transfection, cells were seeded into 24-well plates with sterile glass coverslips or into glass-bottom dishes. When the cell density reached about 60% confluence, the cells were fixed with −20°C cold methanol as described previously (Bhattacharyya et al., 2010). Alternatively, the cells were fixed with 4% paraformaldehyde solution (#P0099; Beyotime, Shanghai, China) at room temperature for 20 min, permeabilized with Saponin (#P0095; Beyotime) for 10 min, and blocked with blocking buffer (#P0260; Beyotime) for 1 h at room temperature. The cells were then incubated with primary antibodies diluted in QuickBlock™ Primary Antibody Dilution Buffer (#P0262; Beyotime) overnight at 4°C. After five gentle washes with phosphate-buffered saline (PBS), the cells were incubated with an Alexa Fluorconjugated secondary antibody (Thermo Fisher) at room temperature for 1 h. Finally, the cells were washed gently eight times with PBS and mounted on slides in Antifade Mounting Medium (#P0126; Beyotime) with or without Hoechst 33258 dye. Images were obtained with a Nikon Eclipse Ti-E inverted confocal microscope equipped with a Plan Apo 100×/1.4 NA oil objective and a C2si scan head camera. NIS-Element AR software (Nikon, Tokyo, Japan) was used to operate the microscope. MCCs were calculated using the JACoP plugin of ImageJ (https://imagej.nih.gov/ij/). We first set the appropriate threshold value of the pixels for both color channels (e.g., Red vs Green).
The M1 coefficient represented the ratio of the "sum of pixels from the former channel (Red) that colocalizes with the latter channel (Green)" to the "total intensity of pixels in the former channel (Red)". M2 measured the opposite channel to M1.

Live-cell microscopy
Cells were cultured and replated onto glass-bottom dishes precoated with 0.1% gelatin. When the cell density reached about 60% confluence, the cells were imaged in pre-warmed FluoroBrite™ DMEM (#A1896701; Thermo Fisher) supplemented with 10% fetal bovine serum. Images were obtained with a Leica SP8 (Wetzlar, Germany) scanning confocal microscope equipped with a Plan Apo 60×/1.4 NA oil objective and solid-state lasers. We set the frame size to 512512 pixels, the digital zoom to 1.5-2.5 for one or two cells, and the pinhole to 1.8 Airy units with bidirectional mode for scanning. The highest sensitivity detector was used for the weaker fluorescence of TagRFP-MON2. All live image videos can be downloaded from: https://drive.google.com/drive/folders/1beD9ONP9k8yKECHoApZhTgk1qomEHxlz? usp=sharing

Wls internalization assay
For internalization of surface-labeled HA-Wls, transfected cells were replated into 24-well plates, rinsed with ice-cold PBS, and incubated with a rabbit anti-HA antibody (#3724; Cell Signaling Technology) diluted 1:250 in OPTI-MEM (#31985062; Thermo Fisher) for 15 min. After a gentle wash with PBS to remove unbound antibody, an internalization assay was performed by placing the cells in normal medium at 37°C.

Cell fractionation
Cells stably expressing DOPEY1-3FLAG or DOPEY2-3FLAG (110 7 ) were placed in 1 ml of hypotonic buffer (10 mM Tris-HCl, pH 7.4, 1 mM Dithiothreitol, 0.2 mM MgCl2, 5 mM KCl, protein inhibitor cocktail, 1 mM PMSF), allowed to swell on ice for 5 min, and lysed by 20 strokes in a Dounce homogenizer. The nuclei were removed from the homogenate by centrifugation at 1000×g for 15 min at 4°C. Sucrose was added to the supernatant to a final concentration of 0.25 M, followed by centrifugation at 10,000×g for 1 h at 4°C. The resulting pellet (P10) containing the PM and endoplasmic reticulum membrane was resuspended in 1 ml of lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, protease inhibitor cocktail, 1 mM PMSF, 1% NP-40), and centrifuged at 100,000×g for 1 h at 4°C. The obtained pellet (P100) containing the Golgi, lysosomes, endosomes, and other membrane vesicles was resuspended in 1 ml of lysis buffer. Finally, the supernatant (S100) containing cytoplasmic proteins was adjusted to 1 ml with lysis buffer. Tubulin was used as the control for the fractionation.

Immunoprecipitation
Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, protease inhibitor cocktail, 1 mM PMSF, 1% NP-40) on ice for 30 min, and centrifuged at 21,500×g for 10 min to remove any intact cells and cell debris. Anti-FLAG M2 affinity gel (#A2220; Sigma-Aldrich) was added to the supernatant and rotated at 4°C for 1 h. After four gentle washes with PBS, the bound proteins were eluted by FLAG peptide or direct boiling in SDS sample buffer. Samples were analyzed by immunoblotting or silver staining. The bands corresponding to DOPEY2 and MON2 were identified by liquid chromatography-tandem mass spectrometry in which performed by Shanghai Applied Protein Technology Co.,Ltd (Shanghai, China).

SDS-PAGE and immunoblotting
Samples were separated in 8% or 4%-20% SDS-PAGE gels, and transferred to PVDF membranes. The membranes were blocked in 5% milk in TBST buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.05% (v/v) Tween-20) and probed with appropriate antibodies diluted in QuickBlock™ Primary Antibody Dilution Buffer (#P0256; Beyotime) for 1 h at room temperature or 4°C overnight. After three washes with TBST for 10 min each, the membranes were incubated with HRP-conjugated secondary antibodies diluted in 5% milk in TBST buffer for 1 h at room temperature, and washed three times in TBST buffer. Bound antibodies were detected with ECL Substrate (#1705060; Bio-Rad). For reuse of the membranes to detect the loading control, the membranes were treated with stripping buffer (#P0025N; Beyotime). Images were captured using an Amersham Imager 680 Camera System (GE Healthcare, Princeton, NJ) or Tanon 5200 Automatic Chemiluminescence Image Analysis System (Shanghai, China). The band intensities were quantified using ImageJ.

Transferrin recycling assay
HEK293 WT or MON2-KO cells were incubated with CF™488A -coupled Tfn (40 μg/ml) at 4°C for 1 h, followed by rapid washing with acid wash buffer (50 mM glycine, 100 mM NaCl, pH 3.0) for 2 min. The cells were then incubated with prewarmed medium in the presence of excess Holo-Tfn (100 μg/ml) for specified periods of time at 37°C. At the end of the incubation, the cells were gently washed with PBS and fixed with cold methanol for confocal analysis. The amounts of intracellular Tfn in six fields were quantified by ImageJ and normalized by the cell number in each field.
To probe their interaction, we analyzed the localization of DOPEY1, DOPEY2 and MON2. Subcellular fractionation revealed that, as expected, DOPEY1 and DOPEY2 were mainly associated with membranes (Supplemental Figure S1b). GFP-tagged DOPEY1 (DOPEY1-FLAG-EGFP) showed weak co-localization with a Golgi marker, Giantin, with the majority of DOPEY1 localized in punctate foci throughout the cytoplasm ( Figure 1c). In contrast to DOPEY1, DOPEY2 was not detectable at all in the Golgi and was only present in these cytoplasmic foci ( Figure 1c). Because both DOPEY1 and DOPEY2 interacted with MON2 in mammalian cells, we monitored DOPEY1-FLAG-EGFP and DOPEY2-FLAG-EGFP localization as a function of MON2 expression (TagRFP-MON2). As shown in Figure 1d, both DOPEY1 and DOPEY2 were strongly recruited to perinuclear structures after co-expression with TagRFP-MON2. These results indicate that MON2 could associate with both DOPEY1 and 2 proteins on the membranes.

Mammalian MON2 mainly localize at RE
Since MON2 has been implicated as scaffolding protein that recruits cytosolic DOPEY proteins to the specific membrane structures, we next determined MON2 localization in HEK293 cells. GFP-tagged MON2 (EGFP-MON2) almost completely co-localized with TagRFP-tagged MON2 (TagRFP-MON2), suggesting that MON2 localization is the same whether it is tagged with GFP or RFP ( Figure 2a). Perinuclear MON2 co-localized with the trans-Golgi/TGN marker Golgin97 and cis-Golgi marker GM130 ( Figure 2a). Since endocytic recycling compartments are concentrated between the Golgi ribbon (Saraste and Prydz, 2019) and recent findings revealed that RE attach to the trans-side of Golgi stacks (Fujii et al., 2020), we sought to determine if MON2 is enriched in endosomes rather than the Golgi. To verify where MON2 positions in the Golgi and endosomes, we treated cells with nocodazole to introduce mini-Golgi stacks.
After 3 h nocodazole treatment, TagRFP-MON2 was localized to trans-side of mini-Golgi stacks by comparison with GM130 (cis-Golgi)/Giantin (medial-Golgi) or Giantin /Golgin97 (trans-Golgi/TGN) (Supplemental Figure S2a, b). In addition, MON2 was separated from Golgin97, suggesting that MON2 is not localized in the trans-Golgi/TGN (Supplemental Figure S2b). To further determine the MON2 localization, we compared MON2 to various endosomal markers, including those for EE, LE, and RE. The small Rab GTPases RAB4A and RAB4B are enriched at the tubular subdomain of EE and RE (D'Souza et al., 2014;Perrin et al., 2013), whereas RAB11 is localized to RE (Green et al., 1997;Ren et al., 1998;Ullrich et al., 1996). The degree with which MON2 and each of these markers co-localized was quantified by Manders correlation coefficients (MCCs), which measures the proportionality of pixel intensity (Bolte and Cordelieres, 2006)  We also examined MON2 co-localization with sorting nexin SNX3 as it was reported to interact with MON2 during endosome vesicle/tubule formation (McGough et al., 2018). SNX3 is an EE marker; it specifically binds to phosphatidylinositol-3phosphate on the membrane of EE and mediates SNX3-retromer complex-dependent vesicular/tubular structure formation (Harterink et al., 2011;McGough et al., 2018;Xu et al., 2001). In comparisons of TagRFP-MON2 with EGFP-SNX3, the M1 of MCC was 0.069±0.009 and the M2 of MCC was 0.082±0.012 (n=19 for both, mean±SE), suggesting MON2 and SNX3 rarely co-localized (Figure 2b and Supplemental Figure   S3). A similar lack of co-localization was also observed for other EE and LE markers such as RAB5 and RAB7 (Figure 2b and Supplemental Figure S3). Taken together, our results indicate that MON2 is primarily localized to RE.

RAB4B-positive RE transiently interact with EE
To examine the dynamics and interactions of MON2 with endosomes in living cells, we co-expressed EGFP-SNX3 or EGFP-RAB4B with TagRFP-MON2 in HEK293 cells and imaged the cells by confocal microscopy (Figure 3a, b). EEs form ring-like structures in some cell types (Wilson et al., 2000). Similar to RAB5 (Hirota et al., 2007), EEs showed larger ring-like shapes in some cells when EGFP-SNX3 was overexpressed. These SNX3-EE rings advantageously enabled a clearer observation of MON2 contacts with endosomal membranes (Figure 3c Figure 3d) that approached, made contact with, and separated from the ring-shaped endosomes (Movie 2). These results indicate that RE transiently interact with EE through bidirectional movement.

MON2 is required for RE segregation from EE
To examine MON2 function on RE trafficking, we examined loss of MON2 or DOPEY function on localization of RE labeled by RAB4B. EE and Golgi were marked by TagBFP-SNX3 and Scarlet-Giantin, respectively. All gene KO in this study were confirmed by genome sequence (Supplemental Table. 1) and immunoblotting with available antibodies against MON2, VPS35, and DOPEY1 (Supplemental Figure. S4a).
In  Figure S4c). To verify the RAB4B localization more clearly, cells were treated with nocodazole to introduce mini-stacks. After nocodazole treatment, RAB4B was co-localized with Giantin in D-KO cells, but was concentrated in SNX3-positive structures in MON2-KO cells (Supplemental Figure   S4d). These results demonstrate that deletion of MON2 and DOPEY have distinct phenotypes on RE localization and therefore suggest a more complex relationship than expected (See Discussion part).
In mammalian cells, RE are also present as dense tubular networks (Xie et al., 2016). To determine if MON2 contributes to the formation of this tubular network, we

MON2 is required for recycling of transferrin receptor
The requirement of MON2 for maintenance of RE structures prompted us to investigate if MON2 plays a role in endocytic recycling. One of well-known RE marker is transferrin receptor (TfnR), which continuously recycles between PM and endocytic system (Kobayashi and Fukuda, 2013;Maxfield and McGraw, 2004). When endocytic recycling is impaired, the steady state level of TfnR is altered. We analyzed the TfnR level in MON2-KO and VPS35-KO cells. The steady-state levels of TfnR do not depend on VPS35 (Kvainickas et al., 2017). As shown in Figure 5a, the levels of TfnR was unaffected in VPS35-KO cells. In contrast, the steady-state levels of TfnR were increased in MON2-KO cells, suggesting that the MON2 is affected the TfnR recycling.
To further investigate this phenotype, fluorescently-labeled transferrin (Tfn-488) was internalized for 1 h at 4C in WT and MON2-KO cells, after which the cells were chased and quantified for their intracellular fluorescence. The intracellular fluorescence of Tfn-488 decreased more slowly in MON2-KO cells (t1/2 = ~20 min) than in WT cells (t1/2 = ~10 min) ( Figure. 5b) suggesting the increased levels of TfnR is due to its increased intracellular retention. We next checked the localization of endogenous TfnR by compared with EE marker SNX3 in WT and MON2-KO cells. In WT cells, TfnRpositive dots were separated from SNX3-positive EE (Figure 5c). In contrast, TfnR was accumulated around EE in MON2-KO cells. Quantitative analysis revealed that the colocalization of TfnR with SNX3 was increased (Figure 5d), suggesting that MON2 is required for efficient recycling of TfnR.

Wls passes through RE for retrograde transport to the Golgi
Wls is a critical cargo receptor for the secretion of Wnt proteins (Banziger et al., 2006), and undergoes continuous recycling between the PM, EE, and Golgi (Gasnereau et al., 2011;Port et al., 2008). In EE, the SNX3-retromer complex is needed for recognition and enrichment of Wls (Harterink et al., 2011;Zhang et al., 2011).  Figure S6a, b). These results demonstrate that Wls is accumulated at EE and RE in MON2-KO cells.
Next, transport of HA-Wls to the Golgi was examined. We labeled RE and the Golgi with EGFP-RAB4B and Scarlet-Giantin, respectively. Notably, RAB4B localization that was predominantly around the Golgi in WT cells was decreased in  Figure   S6c, d). Taken together, these results indicate that transport of Wls from RE to the Golgi is impaired by MON2 deletion.

The steady state of Wls-EGFP is mis-localized to lysosome in MON2-KO cells
To analyze where Wls is trafficked in MON2-KO cells, we examined its steady state levels and localization. It is reported that deletion of VPS35 causes Wls mislocalization to lysosomes (Port et al., 2008;Yang et al., 2008). Wls-EGFP localization was analyzed with the endogenous Golgi marker Giantin and lysosome marker LAMP1.
In WT cells, Wls-EGFP was mainly localized at the Golgi (Supplemental Figure S7a).
In contrast, intracellular punctate fluorescent signals of Wls-EGFP appeared in VPS35- whereas it did not colocalize with LAMP1 (Supplemental Figure S7d). In MON2-KO cells, HA-Wls started to co-localize with LAMP1 partially (Supplemental Figure S7d).
Taken together, these results indicate that Wls is missorted to lysosomes when MON2 is deleted.

Discussion
In the present study, we demonstrate that the MON2 regulates RE distribution and movement, which are required for retrograde transport of Wls from EE to the Golgi ( Figure 8). In MON2-KO cells, RAB4B-positive RE were retained around EE. As a consequence, recycling of endocytic cargos like Wls and TfnR, which pass through RE, was impaired.
MON2 was first reported to localize at the TGN in HeLa cells (Mahajan et al., 2013), although a subsequent study revealed that endogenous MON2 may reside in both the Golgi and RE (Mahajan et al., 2019). In the present study, we carefully checked MON2 localization and found that both EGFP and TagRFP and DOPEY2, the localization of RAB4B differed in each of the deletion cells ( Figure   4a and Supplemental Figure S4a). One possible explanation for these markedly different phenotypes is that the bidirectional transport of RE. The N-terminal of DOPEY1 interacts with a motor protein, kinesin-1 (Mahajan et al., 2019). Our preliminary experiment also showed that MON2-DOPEY has binding abilities to kinesin and dynein/dynactin motor proteins (Supplemental Figure S8) (Figure 8).
Recycling of Wls was impaired in MON2-KO cells. Wls continually undergoes recycling between the plasma membrane, EE and the Golgi (Harterink et al., 2011;Port et al., 2008), whereas it has remained unclear whether Wls is directly transported from EE to the Golgi or through RE. We showed that endocytic Wls is transported to MON2and RAB4B-positive RE, indicating that Wls is recycled back to the Golgi through RE.
The SNX3-retromer complex is essential for the recognition of Wls and ATP9A is responsible for the membrane deformation that is required for carrier formation (McGough et al., 2018). It has been reported that ATP9A is also localized at RE (Tanaka et al., 2016). The MON2-DOPEY complex associates with ATP9A and SNX3 (McGough et al., 2018), and could play roles in carrier transport.
On the other hand, knockdown of MON2 accelerates the transport of Furin and CI-M6PR from endosome to TGN (Mahajan et al., 2013). Furin is trafficked to the TGN directly through late endosomes without passing RE compartment (Mallet and Maxfield, 1999), and CI-M6PR is a cargo of SNX-BAR carriers (Kvainickas et al., 2017).
Compared with the long tubular carriers formed by SNX-BAR, the membrane carriers containing Wls, which are formed by the ATP9A-SNX3-retromer, are vesicular (Harterink et al., 2011). The evidence support that retrograde transport of Furin and CI-M6PR is independent of MON2 function. Depletion of MON2 causes SNX3-retromer blockade and may accelerate the transport of Furin and CI-M6PR to the Golgi due to upregulation of other retrograde pathways such as SNX-BAR.
It is well established that several mechanisms including actin-driven, dynein-driven, and ER-endosome contact site-driven processes are involved in the fission of SNX-BAR tubules from LE (Daly and Cullen, 2018 (Kurokawa et al., 2014). One model that our data support is that pre-existing RE interacts with newly synthesized RE at the tubular EE through the bidirectional movement on microtubules, and Wls may be incorporated from EE to RE. Once Wls has been transported to RE, it is delivered to the TGN. These possibilities will be addressed in future investigations.

Acknowledgment
We thank Drs Hideki Nakanishi (Jiangnan University) and Yasuyuki Suda (University of Tsukuba) for their critical reading of the manuscript and comments. This work was supported by grants-in-aid from the National Natural Science Foundation of China

Conflict of interest
The authors declare that they have no conflicts of interest with the contents of this article.         Supplemental Figure S8. The MON2-DOPEY complex interacts with kinesin and dynactin HEK293 cells stably expressing DOPEY1-3FLAG or DOPEY2-3FLAG were transiently co-transfected with TagRFP-MON2 and HA-tagged sOGT, KLC2, or DCTN1. Cell lysates were incubated with anti-FLAG beads and precipitates immunoblotted with the indicated antibodies. Numbers below each lane are the relative levels of HA-tagged proteins, normalized to sOGT-HA, and were quantified by ImageJ.

Figure legends
Supplemental Table S1. Genomic DNA sequences of WT and CRISPR-Cas9 KO target genes.