A PX-BAR protein Mvp1/SNX8 and a dynamin-like GTPase Vps1 drive endosomal recycling

Membrane protein recycling systems are essential for maintenance of the endosome-lysosome system. In yeast, retromer and Snx4 coat complexes are recruited to the endosomal surface, where they recognize cargos. They sort cargo and deform the membrane into recycling tubules that bud from the endosome and target to the Golgi. Here, we reveal that the SNX-BAR protein, Mvp1, mediates an endosomal recycling pathway that is mechanistically distinct from the retromer and Snx4 pathways. Mvp1 deforms the endosomal membrane and sorts cargos containing a specific sorting motif into a membrane tubule. Subsequently, Mvp1 recruits the dynamin-like GTPase Vps1 to catalyze membrane scission and release of the recycling tubule. Similarly, SNX8, the human homolog of Mvp1, which has been also implicated in Alzheimer’s disease, mediates formation of an endosomal recycling tubule. Thus, we present evidence for a novel endosomal retrieval pathway that is conserved from yeast to humans.


INTRODUCTION
Recycling of sorting components (e.g., receptors, SNAREs, transporters etc.) in the endo-lysosome system is essential for the normal assembly and function of the lysosome. The best characterized recycling cargo is yeast Vps10. Vps10, the first member of the Sortilin receptor family, is a transmembrane (TM) protein receptor that sorts carboxypeptidase Y (CPY) into vesicles at the Golgi (Marcusson et al., 1994). After CPY-containing vesicles are transported to the endosome, the endosome matures and fuses with the vacuole, delivering soluble CPY to the vacuole lumen. Unlike CPY, which is released from the Vps10 receptor in the endosome, Vps10 is not delivered to the vacuole. It is recycled from the endosome back to the Golgi by retromer, making Vps10 available for additional rounds of CPY sorting. Retromer is an evolutionary conserved protein coat complex composed of five proteins: Vps5, Vps17, Vps26, Vps29, and Vps35 ( Figure 1A, 1B and S1A; Seaman et al., 1997Seaman et al., , 1998. It deforms the endosomal membrane to form cargo-containing recycling tubules/vesicles. In humans, loss of retromer function alters the cellular localization of hundreds of TM proteins. A mutation in VPS35 has been associated with Alzheimer's disease and Parkinson's disease (Vilariño-Güell et al., 2011;Zimprich et al., 2011;Rovelet-Lecrux et al., 2015;Small and Petsko, 2015). Retromer also has been identified as an essential host factor for SARS-nCoV2 infection (Daniloski et al., 2020). Thus, retromer-mediated endosomal recycling has been linked to diverse pathologies (Teasdale and Collins., 2012;Small and Petsko., 2015;McMillan et al., 2017).
Mvp1 is a yeast SNX-BAR protein identified as a multicopy suppressor of dominant-negative VPS1 mutations (Ekena and Stevens., 1995). A previous study proposed that Mvp1 is required for retromer-mediated recycling (Chi et al., 2014). The human homolog of Mvp1 is SNX8 (Dyve et al., 2009). Interestingly, two single nucleotide polymorphisms within the SNX8 gene locus are associated with late-onset Alzheimer's disease (AD) (Rosenthal et al., 2012). Xie et al. report that the SNX8 expression level is significantly lower in AD patients and APP/PS1 AD mouse brain (Xie et al., 2019).
Overexpression of SNX8 suppresses the accumulation of fragments of amyloid precursor protein (Aβ).
Additionally, patients lacking SNX8 were shown to have heart development defects, intellectual disability, learning and language delay, and severe behavioral problems related to the hyperactive-impulsive and inattentive area (Vanzo et al., 2013;Rendu et al., 2014;Mastromoro et al., 2020). Although SNX8 has been linked to several diseases, its molecular function is still poorly characterized. Here, we show that Mvp1 recycle membrane proteins in a retromer-and Snx4-independent manner in yeast. We also demonstrate that SNX8 mediates formation of endosomal recycling tubule in humans. Thus, we propose Mvp1 mediates a conserved endosomal recycling pathway which is mechanistically distinct from the retromer and Snx4 pathways. This study reveals that yeast has three major SNX-BAR endosomal recycling pathways; Retromer, Snx4, and Mvp1.

The endosomal localization of Vps55 requires Mvp1.
To characterize the function of Mvp1 and test the hypothesis that Mvp1 is involved in membrane protein recycling, we tested the requirement of Mvp1 function for endosomal TM cargo proteins. To maintain endosomal localization, these proteins need to be recycled back to the Golgi mainly by retromer before 4 the endosome fuses with the vacuole. Indeed, the endosomal t-SNARE Pep12 accumulates on the vacuole membrane in retromer-defective vps35Δ cells, because retromer-mediated endosome-to-Golgi retrograde trafficking is impaired ( Figure S1B). We hypothesized that TM proteins that remain properly localized to the endosome in a retromer independent manner might be cargo for Mvp1. We examined several endosomal TM proteins' localization in vps35Δ cells and found Vps55 still localized on the endosome even in retromer mutants.
Vps55-GFP localized on the endosome in cells lacking Ykr078w, another poorly characterized SNX-BAR protein ( Figure S1E). These results suggest that the endosomal localization of Vps55 requires Mvp1.

Mvp1 is an endosomal coat complex for Vps55 recycling.
Mvp1 has been characterized as a retromer-associated SNX-BAR protein (Chi et al., 2014). However, its function in the retromer independent pathway has never been studied. Endogenously expressed Mvp1-GFP localized to the endosome as previously reported (Figure 2A; Chi et al., 2014). To test if Mvp1 also localizes to the vacuole membrane, we examined the localization of Mvp1 in cells lacking Pep12, which is essential for endosome assembly (Becherer et al., 1996). In pep12Δ cells, Mvp1-GFP lost its endosomal localization and was distributed to the cytoplasm ( Figure S2A), suggesting that Mvp1 specifically localizes on the endosome.
The PX domain of Mvp1 binds to PI3P, allowing its specific localization on the endosome ( Figure 2B).
The conserved residue R172 located in the PX domain is responsible for PI3P binding (Figure S2B and 5 S2C;Chi et al., 2014). The R172E mutation severely impaired the endosomal localization of Mvp1 ( Figure 2C). When the R172E mutant was expressed in mvp1Δ cells, Vps55-GFP localized to the vacuole membrane ( Figure 2D and 2E), indicating that the endosomal localization of Mvp1 is required for appropriate Vps55 localization.
In addition to the PX domain, Mvp1 has a BAR domain, which induces/stabilizes membrane curvature ( Figure 2B). Since the BAR domain's membrane remodeling activity requires its homoor hetero-dimer formation (Frost et al., 2009), we tested which SNX-BAR protein makes a dimer with Mvp1. We expressed FLAG-tagged SNX-BAR proteins in yeast and examined their binding with Mvp1-GFP.
Mvp1-GFP coimmunoprecipitated with Mvp1-FLAG, but not with other SNX-BAR proteins ( Figure 2F), suggesting that Mvp1 forms a homodimer but not heterodimers. Based on the recent cryo-EM structure of the Mvp1 (Sun et al., 2020), we mutated conserved residues located on the dimer interface predicted to distrupt dimer formation ( Figure S2B). The combination mutant I346E/Q468E/W496E had a defect in dimer formation and failed to rescue the mislocalization of Vps55-GFP in mvp1Δ cells ( Figure 2D, 2E, and 2G). We confirmed that I346E/Q468E/W496E mutants still localize on the endosome ( Figure 2C).
These results suggest that the membrane deformation activity of Mvp1 is required for the endosomal localization of Vps55.

Mvp1 recognizes Vps55 through a specific sorting motif.
To determine the sequence motif required for Vps55 retrieval, we performed mutational analysis of the cytoplasmic region of Vps55. We generated a series of Vps55-GFP mutants in which 3 or 4 consecutive amino acids of the cytoplasmic region were replaced with alanine residues and examined their localization ( Figure 3A and S3A). Since 2-4A and 5-8A mutants failed to express, we were not able to analyze residues 2-8 of Vps55. Of the tested mutants, 60-63A and 64-67A mutants exhibited a severe defect in Vps55 recycling. In contrast, 68-71A, 72-75A, 133-136A, and 137-140A did not show a striking defect. Next, we mutated single resides to alanine in the 60-67 region ( Figure 3B and S3B). Of 6 the mutants tested, Y61A, T63A, F66A, and M67A stabilized Vps55-GFP on the vacuole membrane.

Excess Vps55 is ubiquitinated and degraded rather than recycled.
Excess membrane proteins tend to be eliminated from organelles (Weir et al., 2017). To study the fate of excess Mvp1 cargos on the endosomal membrane, we overexpressed Vps55-GFP. It was sorted into the vacuole lumen and degraded ( Figure 3G and S3C). Consistent with this, Vps55-GFP processing was observed in WT cells by immunoblotting analysis, whereas it was not in cells lacking vacuolar hydrolases PEP4 and PRB1 ( Figure 3H). Vacuolar degradative protein sorting is mainly mediated through the ESCRT complex (Henne et al., 2011). Indeed, Vps55-GFP was not degraded in ESCRT defective vps4Δ cells, suggesting that the ESCRT pathway is required for the vacuolar sorting of Vps55-GFP ( Figure 3H). Cargo ubiquitination is a prerequisite for recognition by the ESCRT machinery (Katzmann et al., 2001). To ask whether Vps55 is ubiquitinated, we immunoprecipitated Vps55-GFP, and were able to detect ubiquitinated forms of Vps55-GFP ( Figure 3I). To ask if ubiquitination is sufficient for Vps55 degradation, we fused ubiquitin to Vps55-GFP. Vps55-GFP-Ub was sorted into the vacuole lumen in an ESCRT-dependent manner, even when expressed at endogenous levels ( Figure 3J and S3D). Collectively, we propose that excess Vps55 is ubiquitinated and degraded through the ESCRT pathway rather than recycled ( Figure 3K).
To study the localization of Vps1, we tagged chromosomal VPS1 with GFP using the LAP (70-residue localization and affinity purification) linker (Guizetti et al., 2011), because direct fusion of GFP to the C-terminus of VPS1 interferes with function (Chi et al., 2004). Cells lacking Vps1 showed a growth defect at 37°C, but Vps1-GFP expressing cells grew as well as the WT cells, suggesting that this Vps1-GFP is functional ( Figure S4A). Consistent with a previous report (Varlakhanova et al., 2018), Vps1-GFP colocalized with endosomal Nhx1-mCherry, but not with Golgi localized Sec7-mCherry ( Figure S4B and S4C). We also confirmed that Vps1-GFP colocalized with Mvp1-mRFP ( Figure S4D). Several Vps1-GFP puncta did not colocalize with Mvp1-mRFP, presumably because it also localized on other organelles (i.e. PM, peroxisome etc.). To study its dynamics, we performed live-cell imaging analysis of Vps1-GFP with mCherry-Vps21. The Vps1-GFP punctate structures on the mCherry-Vps21 marked endosomes were elongated and then divided ( Figure 4D and Movie S2).
Upon dynamin assembly, GTP hydrolysis induces conformational changes in the dynamin ring leading to membrane fission ( Figure 4E; Ferguson and De Camilli., 2012). Subsequently, the dynamin ring is disassembled and reused for another round of scission. The K44A mutation in human dynamin-1 impairs the GTP hydrolysis activity (van der Bliek et al., 1993;Damke et al., 1994). In this mutant, membrane scission of clathrin-coated vesicles is blocked. This lysine residue is widely conserved in dynamin-related GTPases, including Vps1 ( Figure 4F and S4E; Varlakhanova et al., 2018;. To test if the GTP hydrolysis activity of Vps1 is required for Vps55 recycling, we examined Vps55 localization in vps1 K42A mutants. In this mutant, Vps55-mNeonGreen localized to the vacuole membrane ( Figure 4G and 4H), suggesting that the membrane scission activity of Vps1 is required for Mvp1-mediated recycling. Next, we examined the localization of the Vps1 K42A mutant. Vps1 K42A -BFP 8 showed punctate structures on the mCherry-Vps21 marked endosome, which colocalized with Mvp1-mNeonGreen ( Figure 4I). These results suggest that the GTP hydrolysis activity of Vps1 is required for Vps55 retrieval.
To test if Mvp1 is required for endosomal localization of Vps1, we examined its localization in mvp1Δ cells, but Vps1-GFP is still localized on endosomes ( Figure S4F). Since Vps1 is required for retromerand Snx4-mediated recycling, we expressed Vps1-GFP in vps35Δ snx4Δ mvp1Δ triple mutants, in which retromer, Snx4, and Mvp1 complexes are defective (Chi et al., 2014;Lukehart et al., 2013). Since the vacuole morphology was extremely fragmented in vps35Δ snx4Δ mvp1Δ cells ( Figure S4G), it was difficult to determine cargo localization in these mutants. However, we found that the vacuole fragmentation phenotype was partially rescued by supplementing choline or ethanolamine, which are used for lipid synthesis. Under these conditions, we found that Vps1-GFP lost its punctate pattern and instead localized the cytoplasm in vps35Δ snx4Δ mvp1Δ cells ( Figure 4J and 4K). We also confirmed that Vps1-GFP maintained its endosomal localization in vps35Δ cells and snx4Δ cells ( Figure 4K and S4F).
Since SNX-BAR proteins require PI3P for their recruitment to the endosomal membrane, we examined Vps1-GFP localization in vps34Δ cells. It lost its punctate localization and shifted to the cytoplasm ( Figure S4H and S4I). In another series of tests, we found that Mvp1-mNeonGreen was still localized on the endosome even in vps1Δ cells ( Figure S4J). We introduced a mutation in the PI3P binding site of Mvp1 (R172E), which altered its punctate localization in vps1Δ cells, confirming the endosomal localization of Mvp1 ( Figure S4K and S4L). These results suggest that the SNX-BAR proteins including Mvp1 are required for the endosomal localization of Vps1.
Next, we analyzed the binding of Vps1 mutants to Mvp1. We found that the Vps1 K42A mutant interacts with Mvp1 more strongly than WT Vps1 ( Figure 4L). The G436D mutant has a defect in the assembly of Vps1 ( Figure 4E, S4E and S4M), but the binding of this mutant was barely detected. We also prepared recombinant Mvp1 and Vps1-GFP and subjected them to in vitro binding assays. Mvp1 coprecipitated with Vps1-GFP, but not with GFP, suggesting that Mvp1 directly binds to Vps1 ( Figure 4M). Based on these results, we propose that Mvp1 recruits the dynamin-like GTPase Vps1 to the site of vesicle tubule formation to catalyze membrane scission.
Next, we examined if Mvp1 and retromer form distinct recycling vesicles from the endosome ( Figure   S5G). To test this, we biochemically immunoisolated Vps55 containing vesicles and asked if Vps10, a retromer cargo, is also present. In this experiment, we tried to accumulate Vps55 containing recycling vesicles using a temperature-sensitive mutant of Sec18, which is essential for the fusion of recycling vesicles with the Golgi ( Figure 5C; Novick et al., 1980). When we shifted sec18 ts mutants to non-permissive temperature (37°C), Vps55-GFP lost its endosomal localization and was distributed in the cytoplasm ( Figure S5H). We incubated Vps55-FLAG expressing sec18 ts cells at the non-permissive temperature for 60 min and then immunoisolated Vps55-FLAG containing structures. Vps55-FLAG was concentrated in the isolated product, but Vps10 was not ( Figure 5D), nor was Vps21 (endosome) or Pho8 (vacuolar membrane). We also analyzed the isolated Vps55 containing structure by EM. In the EM images, we could observe spherical structures with a diameter of 40-70 nm attached with anti-FLAG magnetic beads ( Figure 5E). When we eluted these structures from anti-FLAG beads by FLAG peptide, we could observe similar structures by EM ( Figure S5I). Similarly, we immunoisolated Vps10-FLAG containing vesicles from sec18 ts cells expressing Vps10-FLAG and Vps55-GFP. In the isolated products, Vps10-FLAG was concentrated, but Vps55-GFP was not ( Figure S5J). Next, we performed live-cell imaging of Vps55-mNeonGreen and Vps10-mCherry. The Vps55-mNeonGreen decorated tubules that emerged and detached from a Vps55-mNeonGreen and Vps10-mCherry positive endosome ( Figure 5F).
On the other hand, we also observed that Vps10-mCherry positive but Vps55-mNeonGreen negative tubules also budded from the endosome ( Figure S5K). Based on these results, we conclude that Mvp1 mediates retromer-independent recycling.
The Snx4 complex recycles v-SNAREs from the endosome in a retromer independent manner. To ask if Mvp1 is involved in Snx4-mediated recycling, we examined the localization of GFP-Snc1. GFP-Snc1 was localized on the PM, Golgi, and endosomes in WT cells, whereas it was sorted into the vacuole lumen in snx4Δ cells ( Figure S5L). In mvp1Δ cells, it did not show vacuole lumen localization, suggesting Mvp1 is dispensable for Snx4-mediated recycling.
To ask if the retromer, Snx4, and Mvp1 complexes function in parallel pathways, we generated mutants lacking Vps35 as well as Snx4 and Mvp1 and compared the single or combination mutants for growth.
Although vps35Δ, snx4Δ, and mvp1Δ single or double mutants alone did not exhibit any observable growth defect at 37°C, vps35Δ snx4Δ mvp1Δ triple mutants failed to grow at 37°C ( Figure 5G). These results suggest that retromer, Snx4, and Mvp1 complexes independently function in endosomal recycling ( Figure 5H). Interestingly, the Na + /H + exchanger Nhx1 still localizes on the endosome in vps35Δ, snx4Δ, and mvp1Δ single mutants, but it accumulated on the vacuole membrane in vps35Δ snx4Δ mvp1Δ triple mutants ( Figure 5I and 5J). In contrast, vps35Δ snx4Δ and vps35Δ mvp1Δ double mutants only exhibited a partial defect, and a snx4Δ mvp1Δ mutant showed no defect. These results suggest that Nhx1 is cooperatively recycled by retromer, Snx4, and Mvp1 complexes, which is consistent with our conclusion that retromer, Snx4, and Mvp1 complexes mediates distinct pathways ( Figure 5H).

Retromer, Snx4, and Mvp1 complexes are required for proper function of the endosome.
Retromer-mediated recycling is dispensable for normal growth (Krsmanović et al., 2005). However, vps35Δ snx4Δ mvp1Δ triple mutants exhibited a severe growth defect ( Figure 5G). Hence, we reasoned that general endosomal functions might also be affected in vps35Δ snx4Δ mvp1Δ triple mutants. To evaluate this, we examined Mup1 sorting ( Figure 6A). Mup1 is a methionine permease that localizes to the PM in the absence of methionine (Lin et al., 2008), but upon methionine addition Mup1 is endocytosed, trafficked to endosomes, and sorted into intraluminal vesicles (ILVs) via the ESCRT pathway (Henne et al., 2012). Then, the endosome fuses with the vacuole, which delivers Mup1 to the vacuole lumen. To visualize Mup1 sorting, we fused the pH-sensitive GFP variant, pHluorin, to Mup1 (Miesenböck et al., 1998). When Mup1-pHluorin is sorted into ILVs or the vacuole lumen, its fluorescence is quenched (Prosser et al., 2010). Thus, we can monitor Mup1 sorting by the disappearance of the Mup1-pHluorin signal. After a 60 min treatment with methionine, the fluorescence of Mup1-pHluorin was quenched in the WT cells, whereas Mup1-pHluorin puncta remained stable in vps35Δ snx4Δ mvp1Δ cells ( Figure 6B). We also scored Mup1 sorting by immunoblotting. In WT cells, Mup1-pHluorin was fully processed after 30 min stimulation, whereas full length Mup1-pHluorin remained even after 90 min in vps35Δ snx4Δ mvp1Δ cells ( Figure 6C). We examined the sorting of CPS, which is another transmembrane cargo for the ESCRT pathway (Odorizzi et al., 1998). It was partially defective in vps35Δ snx4Δ mvp1Δ cells ( Figure 6D). To ask which SNX-BAR complexes are responsible, we compared Mup1-GFP sorting in single or combination mutants of retromer, Snx4, and Mvp1.
Although vps35Δ, snx4Δ, and mvp1Δ single mutants did not show a defect, vps35Δ snx4Δ mvp1Δ triple mutants exhibited a strong delay in Mup1 sorting ( Figure S6A). EM analysis revealed that endosomal morphology in vps35Δ snx4Δ mvp1Δ cells was altered ( Figure 6E and S6B). These observations suggest that retromer, Snx4, and Mvp1 complexes are required for proper function of the endosome.
Three endosomal recycling pathways cooperatively contribute to maintain appropriate lipid asymmetry.
Since the role of endosomal recycling has been characterized by retromer mutants, we hypothesized that analysis of vps35Δ snx4Δ mvp1Δ triple mutants might provide new insights into endosomal recycling.
For this purpose, we isolated multicopy suppressors of the temperature-sensitive growth phenotype displayed by vps35Δ snx4Δ mvp1Δ triple mutants and identified, in addition to Vps35, Snx4, and Mvp1, the P4 type of ATPase Neo1 ( Figure 6F, 6G, and S6C). Neo1 can flip phospholipids, especially phosphatidylethanolamine (PE), from the extracellular/lumenal leaflet to the cytoplasmic leaflet of the membrane bilayer, thereby establishing an asymmetric distribution of phospholipids. Neo1 mutants are defective in establishing membrane asymmetry, which leads to hypersensitivity to duramycin, a bioactive peptide that disrupts the membrane through the binding of extracellular PE ( Figure 6H; Takar et al., 2016). To assess the PE asymmetry of the PM, we examined cell growth of vps35Δ, snx4Δ, and mvp1Δ single or combination mutants in the presence of duramycin. Although vps35Δ, snx4Δ, and mvp1Δ single mutants and snx4Δ mvp1Δ double mutants only exhibited a mild defect, vps35Δ snx4Δ and vps35Δ mvp1Δ cells were severely impaired for growth ( Figure 6I). Also, the vps35Δ snx4Δ mvp1Δ triple mutant failed to grow in the presence of duramycin. These results suggest that three endosomal recycling pathways cooperatively contribute to maintain appropriate lipid symmetry.
Appropriate lipid composition/distribution of the PM is essential for cell integrity. Hence, we evaluated the PM integrity of vps35Δ snx4Δ mvp1Δ triple mutants under stress conditions. To score cells for loss of integrity, we used propidium iodide, a membrane-impermeable dye. WT cells and vps35Δ snx4Δ mvp1Δ triple mutants grown at 26°C were barely stained ( Figure S6D and S6E). At 40°C, only a small population of WT cells was stained, because they were resistant to mild heat stress conditions (2 h at 40°C). In contrast, most of vps35Δ snx4Δ mvp1Δ triple mutant cells were stained at 40°C. These results suggest that three endosomal recycling pathways are required to maintain PM integrity under stress conditions.

Mvp1-mediated endosomal recycling is evolutionary conserved.
The human homolog of Mvp1 is SNX8, which also contains a PX domain and a BAR domain ( Figure   7A; Dyve et al., 2009). SNX8 forms a homodimer and exhibits membrane deformation activity in vitro (Weering et al., 2012). Although SNX8 has been linked to several diseases, especially AD (Rosenthal et al., 2012;Xie et al., 2019), its molecular function has not been analyzed in detail. When GFP-SNX8 was expressed in Hela cells, it showed punctate structures that colocalized with the early endosome protein EEA1, as reported previously ( Figure 7B-(i); Dyve et al., 2009). In addition to these puncta, we also observed tubule-like structures that also were labeled by EEA1 ( Figure 7B- (ii)). Live-cell imaging analysis revealed that the tubule-like structures budded from the GFP-SNX8 positive endosome ( Figure   7C, S7A and Movie S3). SNX1 overexpression induces endosomal swelling and tubulation (Carlton et al., 2004). Similarly, when we increased the expression of GFP-SNX8, the endosome marked by GFP-SNX8 were enlarged ( Figure S7B and S7C). Some of these endosomes have long extended tubule structures ( Figure S7B, S7D, and S7E). To analyze the biogenesis of this extended tubule structure, we performed live-cell imaging. Once GFP-SNX8 was concentrated on the endosomal surface, the extended tubules emerged from that site ( Figure 7D and Movie S4). These observations suggest that SNX8 mediates the formation of tubules that bud from the endosome.

DISCUSSION
This study reveals that the SNX-BAR protein Mvp1/SNX8 assembles with the dynamin-like GTPase Vps1 to mediate endosomal recycling ( Figure 7E). Mvp1 forms a homodimer and is recruited to the endosome through PI3P binding. The Mvp1 dimer recognizes its cargo, Vps55, through a specific sorting motif. The BAR domain of Mvp1 deforms the endosomal membrane to form a cargo containing tubule like structures. Subsequently, Mvp1 recruits Vps1 to catalyze membrane scission. Importantly, Vps55 retrieval does not require retromer or the Snx4 complex. In human cells, SNX8, the human homolog of Mvp1, facilitates the formation of endosomal recycling tubules. Thus, we propose that Mvp1 mediates a conserved endosomal recycling pathway mechanistically distinct from retromer-and Snx4-mediated recycling.
Chi et al. proposed that Mvp1 functions in the retromer pathway, but no specific cargo for Mvp1 had 13 been identified (Chi et al., 2014). We found that Vps55 (human OB-RGRP) is recycled by Mvp1 in a retromer-independent manner. Through biochemical analysis, we revealed that retromer and Mvp1 mediate the formation of distinct recycling vesicles. Also, although vps35Δ, snx4Δ, and mvp1Δ single mutants did not exhibit any growth defect at 37°C, vps35Δ snx4Δ mvp1Δ triple mutants failed to grow 37°C. Based on these results, we propose that Mvp1 mainly functions in retromer-independent recycling, consistent with Mvp1 not being essential for retromer-mediated recycling. However, it does not exclude the possibility that Mvp1 assembles with the retromer. Since Mvp1 and retromer cooperatively recycle Nhx1 and also contribute to appropriate phospholipid distribution, a subpopulation of Mvp1 molecules may coassemble with the retromer.
The abnormal accumulation of Aβ is a hallmark feature of AD. The Amyloid precursor protein (APP) is cleaved by β-secretase (BACE) at the endosome, and then its fragments (Aβ) are unconventionally secreted (Small and Petsko., 2015). Interestingly, overexpression of SNX8 reduces Aβ levels and rescues cognitive impairment in an APP/PS1 AD mouse model (Xie et al., 2019). However, how SNX8 prevents the accumulation of Aβ has not been determined. In this study, we found that Mvp1 and retromer not only function independently, but also recycle the same cargo (Nhx1). The human retromer complex has been shown to directly or indirectly (possibly through the Sortilin receptor) recycle APP and BACE (Small and Petsko., 2015). Since defects in this retrieval leads to AD's pathology, SNX8 may also recycle these AD-related membrane proteins.
We propose that Mvp1 recruits Vps1 to mediate membrane scission. Since the endosomal localization of Vps1 was altered in vps35Δ snx4Δ mvp1Δ triple mutants, the retromer and Snx4 complexes may also recruit Vps1 to catalyze the fission of recycling vesicles. Future analysis will be necessary to characterize how Vps1 acts on retromer-and Snx4-mediated recycling tubules. In mammalian cells, dynamin mainly localizes on the PM, but also to the endosome (Nicoziani et al., 2000). Inhibiting dynamin causes the tubulation of endosomes (Mesaki et al., 2011). These observations raise the possibility that dynamin-mediated endosome scission is conserved. However, unlike in yeast cells, the WASH complex, an activator of Arp2/3-dependent actin polymerization, forms a complex with retromer, which might be sufficient for membrane scission (Simonetti and Cullen., 2019). How membrane scission is catalyzed in endosomal recycling pathways is a fundamental question in this field.
Recycling of key membrane proteins (SNAREs, receptors, flippases, transporters, etc.) allows cells to maintain organelle identity and function. In this study, we reveal that Mvp1 mediates a novel endosomal 14 recycling pathway. Our findings pave the way toward understanding how cells maintain the composition and function of each organelle. However, several question remains to be answered: (1) To form a uniform size of recycling vesicles, the activity of Mvp1 and Vps1 should be temporally and spatially regulated, but its regulation mechanism(s) remains unknown. (2)

DECLARATION OF INTERESTS
The authors declare no competing interests.                Table   Table S1. Yeast strains used in this study Table S2. Plasmids used in this study

Figure Legend for Supplemental Movies
Movie S1, related Figure 2I.

LEAD CONTACT AND MATERIALS AVAILABILITY
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Scott D. Emr (sde26@cornell.edu).

Yeast Strain and Media
S. cerevisiae strains used in this study are listed in Table S1. Standard protocols were used for yeast manipulation (Kaiser et al., 1994). Cells were cultured at 26°C to mid-log phase in YPD medium [1% (w/v) yeast extract, 2% (w/v) bacto peptone, and 2% (w/v) glucose] or YNB medium [0.17% (w/v) yeast nitrogen base w/o amino acids and ammonium sulfate, 0.5% (w/v) ammonium sulfate, and 2% (w/v) glucose] supplemented with the appropriate nutrients.

Mammalian Cell line
Hela cells were kindly provided by Dr. Anthony Bretscher (Cornell University). Cell lines were verified to be free of mycoplasma contamination and the identities were authenticated by STR profiling.

Cell Culture conditions for mammalian cells
Hela cells were maintained at 37°C and 5% CO2 in DMEM supplemented with 10% fetal bovine serum (FBS), penicillin, and streptomycin.

Plasmids
Plasmids used in this study are listed in Table S2.

Fluorescence Microscopy
Fluorescence microscopy was performed using a CSU-X spinning-disk confocal microscopy system (Intelligent Imaging Innovations) or a DeltaVision Elite system (GE Healthcare Life system).
A CSU-X spinning-disk confocal microscopy system is equipped with a DMI 6000B microscope (Leica), 100×/1.45 numerical aperture objective, and a QuantEM electron-multiplying charge-coupled device (CCD) camera (Photometrics). Imaging for yeast cells was done at room temperature in YNB medium using GFP and mCherry channels with different exposure times according to each protein's fluorescence intensity. Imaging for mammalian cells was done at 37°C in FluoroBride DMEM media STAR METHODS (Thermo Fisher) using GFP, mCherry, and far-red-fluorescent dye channels. Images were analyzed and processed with SlideBook 6.0 software (Intelligent Imaging Innovations). A DeltaVision Elite system is equipped with an Olympus IX-71 inverted microscope, DV Elite complementary metal-oxide semiconductor camera, a 100×/1.4 NA oil objective, and a DV Light SSI 7 Color illumination system with Live Cell Speed Option with DV Elite filter sets. Imaging was done at room temperature in YNB medium using GFP and mCherry channels with different exposure times according to each protein's fluorescence intensity. Image acquisition and deconvolution (conservative setting; five cycles) were performed using DeltaVision software softWoRx 6.5.2 (Applied Precision)

Immunoprecipitation for yeast cells
Anti-FLAG-conjugated magnetic beads were prepared according to the manufacturer's protocol. In brief, NHS FG beads (Tamagawa Seiki) were treated with methanol and then incubated with anti-DYKDDDK antibody (Wako) at 4°C for 30 min. The magnetic beads were mixed with 1.0 M 2-aminoethanol, pH 8.0, at 4°C for 16-20 h to quench the conjugation reaction, washed three times with the bead wash buffer [10 mM HEPES-NaOH (pH 7.2), 50 mM KCl, 1 mM EDTA, and 10% glycerol], and stored in wash buffer containing 1 mg/ml BSA (A7030; Sigma-Aldrich).
To examine the Mvp1-Mvp1 interaction, cells expressing Mvp1-FLAG and Mvp1-GFP were grown to mid-log phase were washed twice with wash buffer [50mM Tris-HCl (pH 8.0), 150 mM NaCl, and 10% glycerol]. The cells were lysed in IP buffer [50mM Tris-HCl (pH 8.0), 150 mM NaCl, 10% glycerol, 1 mM PMSF, 1x protease inhibitor cocktail (Roche)] and lysed by beating with 0.5 mm YZB zirconia beads (Yasui Kikai) for 1 min. IP buffer containing 0.2% Triton X-100 was added to the lysate (final concentration of 0.1%), and the samples were rotated at 4°C for 10 min. The solubilized lysates were cleared at 500 x g for 5 min at 4°C, and the resultant supernatants were subjected to a high-speed centrifugation at 17,400 x g for 10 min. The cleared supernatants were incubated with pre-equilibrated anti-FLAG-conjugated magnetic beads and rotated at 4°C for 1 hour. After the beads were washed with wash buffer containing 0.1% Triton X-100, the bound proteins were eluted by incubating the beads in SDS-PAGE sample buffer at 98°C for 5 min.
The Mvp1-Vps1 interaction was examined similarly as the Mvp1-Mvp1 interaction with some modifications. Cells expressing Mvp1-FLAG and Vps1-GFP were washed twice with High Salt wash buffer [50mM Tris-HCl (pH 8.0), 500 mM NaCl, and 10% glycerol]. Cells were lysed in High Salt IP buffer [50mM Tris-HCl (pH 8.0), 500 mM NaCl, and 10% glycerol, 1x protease inhibitor cocktail (Roche)] as above. The lysate was solubilized with 1.0% Triton-X 100 at 4°C for 60 min. After centrifugation, the cleared supernatant was incubated with preequilibrated anti-FLAG-conjugated magnetic beads and rotated at 4°C for 1 h. After the beads were washed with wash buffer containing 1.0% Triton X-100, the bound proteins were eluted by incubating the beads in SDS-PAGE sample buffer at 98°C for 5 min.
Cells were lysed in PBS IP buffer as above. The lysate was solubilized with 0.5% Triton-X 100 for 4°C for 10 min. After centrifugation, the cleared supernatant was incubated with preequilibrated anti-FLAG-conjugated magnetic beads and rotated at 4°C for 2 h. After the beads were washed with wash buffer containing 0.5% Triton X-100, the bound proteins were eluted by incubating the beads in SDS-PAGE sample buffer at 98°C for 5 min.

Immunoprecipitation under denature condition for yeast cell lysate
To analyze the ubiquitination status of Vps55, cells expressing Vps55-GFP were washed twice with 400 mM NEM. Cells were lysed in Urea cracking buffer [50 mM Tris-HCl (pH 8.0), 1% SDS, 8M Urea, 20 mM NEM, 1x protease inhibitor cocktail (Roche)] and lysed by beating with 0.5 mm YZB zirconia beads (Yasui Kikai) for 1 min. High salt IP buffer with 20 mM NEM and 0.2% Triton X-100 was added to the lysate, and the samples were rotated at 4°C for 10 min. The solubilized lysates were cleared at 500 x g for 5 min at 4°C, and the resultant supernatants were subjected to a high-speed centrifugation at 17,400 x g for 10 min. The cleared supernatants were incubated with pre-equilibrated GFP-TRAP_A beads (Chromo Tek) and rotated at 4°C for 1 hours. After the beads were washed with SDS wash buffer [50 mM Tris-HCl (pH 8.0), 250 mM NaCl, 1% SDS, 4M Urea, 5% Glycerol], the bound proteins were eluted by incubating the beads in SDS-PAGE sample buffer at 98°C for 5 min.

Immunoisolation of Vps55-FLAG or Vps10-FLAG containing structures from yeast cells
To immunoisolate Vps55-FLAG or Vps10 containing structures, cells expressing Vps55-FLAG or cells expressing Vps10-FLAG and Vps55-GFP grown at 26°C were incubated at 37°C for 60 min before harvest. Cells were washed twice with H25S75E5 buffer [25 mM HEPES-NaOH (pH 7.4), 750 mM Sorbitol, 5 mM EDTA]. The cells were lysed in H25S75E5 buffer supplemented with 1x protease inhibitor cocktail (Roche) and lysed by beating with 0.5 mm YZB zirconia beads (Yasui Kikai) for 1 min. The lysates were cleared at 500 x g for 5 min at 4°C twice, and the cleared supernatants were incubated with anti-FLAG-conjugated magnetic beads and rotated at 4°C for 1 hour. After the beads were washed with H25S75E5 buffer, the immunoisolated structures were eluted by incubating the beads in SDS-PAGE sample buffer at 98°C for 5 min. For the EM analysis, immunoisolated structures were not eluted, but subjected to negative staining EM analysis.

Electron Microscopy of immunoisolated Vps55-FLAG containing structure
Immunoisolated Vps55-FLAG containing structures on anti-FLAG conjugated magnetic beads were applied to the carbon-coated electron microscope grid and then was stained with 2% ammonium molybdate and imaged on FEI Morgagni 268 TEM.

Quantitative analysis of cargo localization
Vps55-GFP/mNeonGreen, Vps10-GFP, Kex2-GFP, and GFP-Neo1 localization was classified into two categories: punctate structures and vacuole membrane localization. Cells having both punctate structures and vacuole membrane localization were classified in the vacuole membrane localization category. For each experiment, at least 30 cells were classified, and the data from three independent experiments were used for the statistical analysis.

Quantitative analysis of Mvp1-GFP and Vps1-GFP localization
Mvp1-GFP and Vps1-GFP localization was classified into two categories: punctate structures and cytoplasmic localization. For each experiment, at least 30 cells were classified, and the data from three independent experiments were used for the statistical analysis.
The cells were then rinsed with 5mL spheroplast buffer (0.1 M phosphocitrate, 1.0 M sorbitol). The cells were then re-suspended in 1 mL spheroplast buffer containing 0.25mg/mL zymolyase and incubated at RT for 30min. After spheroplasting, the cells were gently washed twice with 1mL staining buffer (0.1 M sodium cacodylate pH 6.8, 5 mM CaCl2) to remove sorbitol. The cells were embedded in 50 µL of 2% ultra-low-melt agarose, then cut into ∼2mm 3 blocks. The blocks were postfixed/stained in 1mL osmium staining solution (1% OsO4, 1% potassium ferrocyanide, 0.1 M sodium cacodylate pH 6.8, 5 mM CaCl2, 10% formamide) for 1 hr at room temperature. The blocks were washed four times with water, then stained with 1 mL 1% uranyl acetate overnight. The blocks were washed four times with water, then dehydrated through a graded series of ethanol: 50%, 75%, 95%, 2× 100% for 10 min each (1 mL). The blocks transitioned to propylene oxide: 1:1 propylene oxide:ethanol for 10 min, then 2x 5 min in 100% propylene oxide (1mL). The blocks were embedded in 1:1 propylene oxide:epon resin (hard formulation, EMS #14120) and left on a rotator overnight to allow the propylene oxide to evaporate. The blocks were transferred to fresh epon resin and polymerized for 24 hrs at 60°C.
The samples were sectioned at ~70nM. The sections were poststained with 4% uranyl acetate for 10 min, then washed in water. Next, the sections were stained in Reynold's lead citrate for 2 min, then washed in water. The sections were imaged using a FEI Morgagni 268 TEM.

Protein expression and purification
His-SUMO-Vps1-GFP and His-SUMO-Mvp1 were expressed in Escherichia coli Rosetta (DE3) by addition 250 mM IPTG for 16 hours at 16°C and purified using TALON® Metal Affinity Resin (Clontech) according to the manufacture's protocol. The recombinant proteins were eluted with SUMO protease (Milipore) for Vps1-GFP and Mvp1.

In vitro binding assay
Purified proteins were incubated with anti-GFP magnetic beads at 4°C for 30 min. After the beads were washed three times with PBS with 0.1% TX-100, the bound proteins were eluted by incubating the beads in SDS-PAGE sample buffer at 98°C for 5 min.

Transfection for mammalian cells
Transient transfections were carried out using FuGENE (Promega) according to the manufacturer's instructions, and the experiments were performed 48 hours after transfection.