Гибридные эндосомные пальто содержат различные классы сортирующих нексинов.

Текст сохранён. Часть иллюстраций пока доступна только в первоисточнике; загрузка будет проверена повторно.

Introduction

Retromer is a key player in endosomal recycling. This protein complex was discovered in yeast as a pentamer, consisting of two subcomplexes: a heterodimer of the SNX-BAR sorting nexins Vps5 and Vps17, and a heterotrimer of Vps26, Vps29 and Vps35 (Seaman et al, 1998 ). Today, the term Retromer is used to refer to this heterotrimer alone. Retromer is recruited to endosomal membranes through SNX proteins. In yeast, Retromer binds the SNX-BAR dimer Vps5-Vps17 or Snx3 (Seaman et al, 1998 ; Horazdovsky et al, 1997 ; Harrison et al, 2014 ; Strochlic et al, 2007 ). In mammalian cells, Retromer binds SNX3 (Harterink et al, 2011 ; Vardarajan et al, 2012 ; Chen et al, 2013 ), but it interacts with the homologues of the yeast Vps5/Vps17 dimer only indirectly. These homologues, the SNX-BARs SNX1/SNX2 and SNX5/SNX6, form the ESCPE complex (Simonetti et al, 2017 , 2019 ; Lopez-Robles et al, 2023 ). Unlike the yeast SNX-BARs, ESCPE-1 interacts with Retromer through an additional sorting nexin, SNX27 (Lopez-Robles et al, 2023 ; Chandra et al, 2025 , 2022 ; Guo et al, 2024 ; Simonetti et al, 2022 ).

CryoET structures of Retromer bound to either SNX-BARs or Snx3 have been solved (Leneva et al, 2021 ; Kovtun et al, 2018 ; Chen et al, 2025 ; Kendall et al, 2022 , 2020 ). These structures revealed that Retromer can oligomerize to form arch-like structures that cross-link SNX proteins. They show Retromer associating with SNX3 and SNX-BARs in different ways and cross-linking them in different patterns (Fig. 1C ). For example, while Vps26 contacts the membrane in Snx3/Retromer coats, this is not the case in the SNX-BAR/Retromer coats, where Vps26 has no contact with lipids and binds the membrane only indirectly through the SNX-BAR layer. Both the SNX-BAR and the SNX3-based Retromer coats show an SNX layer that covers the membrane extensively but with limited regularity (Leneva et al, 2021 ; Kovtun et al, 2018 ). The binding sites for Snx3 and the SNX-BARs are positioned such that Retromer could only bind one or the other type of sorting nexin, but not both at the same time.

In current models, SNXs usually define distinct recycling pathways and contribute to the selection of specific sets of cargo proteins (Buser and Spang, 2023 ; Weeratunga et al, 2020 ; van Weering et al, 2010 ). For example, yeast Vps10 is mainly ascribed to the Retromer/Vps5-Vps17 complex, while Ear1, Ste13, Kex2 or Ftr1 are addressed by the Retromer/Snx3 complex (Strochlic et al, 2008 , 2007 ; Voos and Stevens, 1998 ; Harrison et al, 2014 ; Suzuki et al, 2019 ). Similarly, in mammals, SNX27- and SNX3-Retromer complexes recycle specific cargo to the plasma membrane and the TGN, respectively (Temkin et al, 2011 ; Lauffer et al, 2010 ; Yang et al, 2018 ; Ghai et al, 2011 ; Lee et al, 2016 ; Steinberg et al, 2013 ; Harterink et al, 2011 ; Wassmer et al, 2009 ; Cullen and Korswagen, 2012 ; McGough et al, 2014 , 2018 ; Tian et al, 2021 ; Cui et al, 2018 ). Sorting occurs at the endosome surface, where coat components can be enriched in domains, which generate tubulovesicular cargo carriers (Varandas et al, 2016 ; Thompson et al, 2007 ; Derivery et al, 2012 , 2009 ; Antón-Plágaro et al, 2024 ; Puthenveedu et al, 2010 ). However, there is a priori no reason why different coat complexes might not be recruited simultaneously into the nascent endosomal carriers, since many of them are quite compatible with the highly curved surface of these tubulovesicular structures. This might lead to the formation of heterogeneous coats, where a single tubule contains a series of homogeneous subdomains with different SNX/Retromer coats, or to “hybrid coats” in which different types of coat components interact and are co-recruited in the same coated domain (Gopaldass et al, 2024 ).

The concept of hybrid carriers is compatible with the fact that the spectra of cargos that depend on SNX3 or the SNX-BARs overlap, both in yeast and in mammalian cells (Bean et al, 2017 ; Steinberg et al, 2013 ). This might reflect the existence of distinct cargo exit routes that are redundant, perhaps due to similarities in the sorting signals for different SNXs (Strochlic et al, 2007 ; Ma and Burd, 2019 ), but it is also consistent with multiple classes of cargos and SNXs jointly using a single endosomal carrier for exiting the endosome. In mammalian cells, evidence for and against a role of common carriers incorporating SNX-BAR and SNX3 cargo has been reported. β2AR, which depends on SNX27 for returning to the plasma membrane, and Wntless, which requires SNX3 for retrograde transport to the Golgi, co-localise on the same endosomes and partition into the same leaving carriers, though with differing enrichment factors (Varandas et al, 2016 ). On the other hand, knockdown of the Drosophila orthologues of SNX1/SNX2 and SNX5/SNX6, with which SNX27 interacts and cooperates in sorting cargo (Simonetti et al, 2022 ; Guo et al, 2024 ; Simonetti et al, 2023 ; Chandra et al, 2025 , 2022 ), was reported not to influence Wntless function (Harterink et al, 2011 ). A caveat is that the efficiency of the knockdown was only 80% at the RNA level, and the effects on the protein levels were not tested. The same study found that mammalian Wntless enters only around 20% of SNX-BAR tubules but instead leaves endosomes in smaller SNX3-containing structures.

We set out to probe the potential for generating hybrid coats utilising an in vitro system (Gopaldass et al, 2023 ) and the sorting nexins Vps5-Vps17 and Snx3, because structures of their respective coats have been solved (Kovtun et al, 2018 ; Leneva et al, 2021 ). The in vitro system offers the advantage to monitor coat growth from purified coat components on synthetic supported membrane tubes (SMTubes) in real time in an optically well-defined setting. This allows quantitative assessments of coat radius, speed of growth, and the density of membrane coverage. Here, we exploit this system to show that yeast SNX-BARs and Snx3 readily form adaptable hybrid coats that can incorporate cargos for both classes of sorting nexins and have properties that are distinct from those of homogeneous SNX-BAR or Snx3 coats. We confirm the relevance of these findings in vivo by showing Retromer-mediated interaction between SNX-BARs and Snx3 and by analysing the interdependency of Snx3 and the SNX-BARs for sorting their respective cargos.

Results

We previously described the formation of SNX-BAR and Retromer-SNX-BAR coats from purified proteins and synthetic supported membrane tubes on the surface of glass slides (SMTubes) (Gopaldass et al, 2023 ). To extend the system and ultimately enable the analysis of potential hybrid coats, we determined the conditions for Snx3 coat formation on SMTubes using recombinant Snx3 purified from E. coli and Retromer (Vps26-Vps29-Vps35) and SNX-BAR (Vps5-Vps17) complexes purified from yeast (Fig. EV1 ) (Purushothaman et al, 2017 ; Gopaldass et al, 2023 ). As cargo is necessary to stabilise the interaction between Snx3 and Retromer (Leneva et al, 2021 ; Lucas et al, 2016 ), we synthesised small peptides that contain the sorting signals and correspond to the cytosolic tails of Ste13 and Ear1, two bona fide Snx3 cargos. The peptides carried an N-terminal HIS 6 tag to recruit them to the membrane via Ni-NTA lipids, and a C-terminal cysteine for maleimide-mediated coupling to fluorescent dyes. The coat proteins were labelled by expressing them as fusion proteins with red or green fluorescing tags, with the tags being fused to Vps17 in the SNX–BAR complex and to Vps29 in Retromer.

Membrane scaffolding activity and Retromer saturation of Snx3 and SNX-BAR coats

While mScarletI Snx3 bound to the SMTubes in the absence of the Ear1 and Ste13 cargo peptides, it required these peptides to recruit Retromer mClover and form concentrated domains on the tubules (Fig. 1A ). Addition of a SNX–BAR cargo, the Vps10 tail peptide, did not lead to domain formation (Fig. EV2 ). The intensity of the fluorescent lipid marker Cy5.5.-PE, which had been integrated into the membrane tubules, was reduced underneath the concentrated Snx3 domains. This indicates that the membrane tubes were constricted in these areas relative to the rest of the membrane, where Snx3 and Retromer did not concentrate into a denser arrangement. We hence refer to these constricted regions with a strong concentration of SNX and Retromer as coated domains. We tested whether these Snx3-Retromer domains recruited cargo using a fluorescent Alexa-547-labelled version of the Ear1 peptide (Ear1 547 ) (Fig. 1B ). Coated domains marked by Retromer mClover strongly accumulated Ear1 547 , confirming that these domains were active for cargo recruitment. All subsequent experiments with Snx3, unless stated otherwise, were performed in the presence of 10 µM of the Ear1 cargo peptide.

SNX-BARs can form dimers and higher-order associations, which at elevated SNX-BAR concentrations can suffice to scaffold a membrane tubule with these proteins alone (Gopaldass et al, 2023 ; Lopez-Robles et al, 2023 ; van Weering et al, 2012 ; Sun et al, 2020 ; Zhang et al, 2021 ). By contrast, Snx3 has only a PI3P-binding PX domain and no BAR domain, and it does not show a tendency for spontaneous oligomerisation. Snx3 therefore requires Retromer to form a coat and impose curvature on the membrane (Leneva et al, 2021 ) (Fig. 1C ). Probably for this reason the structure of the Snx3-Retromer coat shows each Snx3 molecule associated with Retromer (Harrison et al, 2014 ; Leneva et al, 2021 ), unlike the SNX-BAR-Retromer coat, which contains also SNX-BARs that are not linked to Retromer (Kovtun et al, 2018 ). We hence compared the amount of Retromer per tubule length in SNX-BAR/Retromer and Snx3/Retromer coats as a function of the initial membrane tube radius (Fig. 1D,E ). While the amount of Retromer mClover increased with the starting tube radius for SNX-BAR/Retromer coats (Gopaldass et al, 2023 ), there was no such tendency for the Snx3/Retromer coats. They entirely failed to constrict wider tubes. The density of Retromer mClover in the Snx3/Retromer coats was around twice the density found in the SNX-BAR/Retromer coats (Fig. 1E ). We also measured the lipid fluorescence of the tubules under the SNX-BAR and the Snx3 coats (Fig. 1F ). This value is directly proportional to the radius of the tubule and can be calibrated by comparing the signals to those under a dynamin oligomer, which compresses the underlying membrane to a known, invariant radius (Roux et al, 2010 ; Dar et al, 2015 ). Applying this calibration revealed that Snx3 coats were on average slightly wider than the SNX-BAR coats (Fig. 1G ) (20 ± 3.8 nm vs. 19 ± 2.6 nm radius to the outer membrane surface), in line with previous structural studies (Leneva et al, 2021 ). However, Snx3-Retromer coats were more sensitive to the initial radius of the non-constricted tube than the SNX-BAR-Retromer coats (Fig. 1F ). Whereas the radius of SNX-BAR-Retromer coats increased only a little with initial tube radius, the radius of the Snx3-Retromer coat increased more steeply when wider tubes had to be constricted. Thus, the Snx3-Retromer coat may be more adaptable in curvature.

Since the density of Retromer in the Snx3 coat is higher than in the SNX-BAR coat, we tested whether both coats are saturated with Retromer. To this end, we incubated SNX-BAR or Snx3 with either equimolar concentrations of Retromer mClover (25 nM) or with a fourfold excess (100 nM) and measured the amount of Retromer mClover incorporated into the coats for tubes of various starting radius (Fig. 2A–D ). Similar amounts of Retromer mClover were integrated into Snx3 coats, irrespective of whether 25 or 100 nM Retromer mClover were offered (Fig. 2C,D ). SNX-BAR coats integrated up to twice the amount of Retromer mClover when it was present at 100 nM (Fig. 2A,B ). Even at this elevated concentration, the density of integrated Retromer remained sensitive to the curvature of the membrane that must be scaffolded. It increased by another factor of two when the coat had to constrict wider membrane tubes. However, the excess of Retromer (100 nM) enabled the SNX-BARs to constrict much wider tubes than the 1:1 mixture (25 nM). This is consistent with Retromer providing driving force for coat formation (Gopaldass et al, 2023 ), because compressing wider tubes requires more work than compressing thinner tubes. Snx3-based coats behaved very differently. They could oligomerize only on much thinner tubes than the SNX-BAR coats, and even a fourfold excess of Retromer mClover did not convey the capacity of constricting significantly wider tubes to them (Fig. 2D ).

SNX-BAR-Retromer coats do not incorporate new SNX-BARs or exchange SNX-BARs with the pool in solution (Gopaldass et al, 2023 ). We also tested this aspect for Snx3-Retromer coats, using a two-stage experiment. In a first stage, GFP Snx3-Retromer coats were formed, and non-bound GFP Snx3 was washed out. In a second stage, an excess of mScarletI Snx3 was flushed into the chamber. This second wave of red fluorescing Snx3 bound readily to the non-coated regions of the tubes but was not incorporated into regions where GFP Snx3 had pre-formed coats (Fig. 2E,F ). Altogether, our data suggest that both SNX-BAR and Snx3-based Retromer coats are stable and do not readily exchange subunits. Nevertheless, Snx3-Retromer coats can perform less work for membrane scaffolding. They carry a fixed Snx3/Retromer ratio and are saturated with Retromer, unlike the SNX-BAR/Retromer coats, which are more adaptable in terms of their coverage by Retromer.

SNX-BARs and Snx3 can co-integrate into a hybrid Retromer coat

Since Retromer can interact with both SNX-BARs and Snx3, we tested whether Retromer could form hybrid coats incorporating both classes of sorting nexins. When SMTubes were incubated with equimolar amounts of mScarletI Snx3, SNX-BAR GFP and Retromer, we observed the formation of hybrid coats harbouring both mScarletI Snx3 and SNX-BAR GFP (Fig. 3A,B ). These hybrid coats could constrict tubules of up to 125 nm initial radius (Fig. 3C ), resembling in this respect the SNX-BAR coats (Fig. 1F ). However, the hybrid coats were larger, with a radius of 22 ± 4.7 nm (Fig. 3D ), compared to 19 ± 2.6 nm for homogeneous SNX-BAR-Retromer and 20 ± 3.8 nm for homogeneous Snx3-Retromer coats (Fig. 1G ). Hybrid coat formation depended on Retromer. Co-incubating SMTubes with SNX-BAR GFP and mScarletI Snx3 without Retromer led to the formation of coats containing only SNX-BARs but not enriching Snx3 in this zone (Fig. 3E,F ). This suggests that Snx3 is not recruited into hybrid coats by a direct interaction with SNX-BARs but requires Retromer for this. Using fluorescently labelled Ear1 and Vps10 peptides revealed that the Retromer hybrid coats concentrate both peptides (Fig. 3G,H ) and are hence competent for cargo integration.

The structures of Snx3 or SNX-BAR-coated membrane tubules show limited regularity of these coats and some voids, which might provide space to incorporate other proteins (Leneva et al, 2021 ; Kovtun et al, 2018 ). We hence performed staged experiments to test whether Snx3 could integrate into pre-formed SNX-BAR-Retromer coats or vice versa (Fig. 4A–D ). SMTs were incubated with 25 nM SNX-BAR GFP and Retromer to allow the formation of coats that were 1–2 µm long. Unbound proteins were washed away, and an excess (100 nM) of mScarletI Snx3 was added (Fig. 4A,B ). After 5 min of incubation, mScarletI Snx3 was mainly localised to the uncoated regions, but it did not integrate into the pre-formed SNX-BAR-Retromer coats. Similarly, pre-formed mScarletI Snx3-Retromer coats could not integrate an excess of SNX-BAR GFP (Fig. 4C,D ). These observations suggest that the pre-formed homogeneous coats do not contain suitable voids and binding sites and/or are not flexible enough to integrate the other sorting nexin class.

To directly compare SNX densities in the hybrid and homogeneous coats, we created a GFP Snx3 construct using the same GFP tag as in our SNX-BAR GFP . Retromer coats were generated with either GFP Snx3 or SNX-BAR GFP in the presence or absence of non-tagged versions of the respective other sorting nexin (Fig. 5A,B ). The overall sorting nexin concentration was kept constant (50 nM for pure coats and 25 nM of each sorting nexin class for hybrid coats, both in the presence of 50 nM Retromer) (Fig. 5A–C ). The mean GFP fluorescence per tube length in the GFP Snx3/Retromer coat (404 ± 80 a.u., n = 40) was about twice the mean GFP fluorescence in the SNX-BAR GFP coat (230 ± 40 a.u., n = 40) (Fig. 5C ). Considering that the SNX-BAR complex is a heterodimer of a labelled Vps17 GFP and an unlabelled Vps5 it follows that the SNX densities in Snx3-based and SNX-BAR-based Retromer coats are similar.

The different degrees of labelling of GFP Snx3 (every subunit is labelled) and SNX-BAR GFP (a heterodimer, in which only Vps17 GFP is labelled, but not Vps5) could also be exploited in replacement experiments. To this end, Retromer coats were generated at different ratios of GFP Snx3 over SNX-BAR GFP (Fig. 5D,E ). The total GFP signal in the coats should remain constant if one GFP Snx3 molecule replaced a SNX-BAR dimer, but it should increase if two Snx3 molecules replaced a SNX-BAR dimer. Regarding a coat region of 6 sorting nexins as an example, this unit would contain 3 GFP when made only of SNX-BAR GFP heterodimers. In a coat assembled at a SNX-BAR GFP : GFP Snx3 ratio of 2:1, this unit would contain 4 GFPs (2 from Vps17 GFP , 2 from GFP Snx3); at a 1:2 ratio it would contain 5; and in a pure GFP Snx3 coat it would contain 6. The fluorescence signals of hybrid coats that assembled from GFP Snx3 and SNX-BAR GFP at these ratios followed exactly this expected pattern (Fig. 5D,E ). This result suggests that one Snx3 substitutes one SNX-BAR subunit and that hybrid coats can assemble at variable SNX-BAR/Snx3 stoichiometries. This latter point was further supported when we compared coats formed from mixtures containing GFP Snx3 and unlabelled SNX-BARs at ratios of either 2:0, 2:1 or 1:2. Retromer was kept at the same concentration in all cases (Fig. 5F–H ). The Snx3 GFP signal in the coat decreased in proportion to the SNX-BAR:Snx3 GFP ratio in solution. Hybrid coats thus do not have a fixed composition, suggesting that they are adaptable as a function of available Snx3 and SNX-BARs and/or their cargos.

Hybrid coat formation enhances the capacity for membrane scaffolding

We noticed that hybrid coats could form on and constrict tubes of higher starting radius than homogeneous Snx3 coats (Fig. 5B,D,G ), suggesting that mixed coats are more potent in deforming the membrane and that SNX-BARs might help drive Snx3/Retromer coat formation. This points to a potential synergy between Snx3 and the SNX-BARs in scaffolding the tubes. We used giant unilamellar vesicles (GUVs) to test whether this is the case. On GUVs, coat formation requires more work for membrane deformation because a tubular membrane conformation is not pre-defined as in the case of SMTubes. Tubulation also reduces the volume of the GUV and necessitates extrusion of luminal liquid, requiring additional work. When GUVs were incubated with 100 nM mScarletI Snx3 and Retromer, tubules did not form (Fig. 6A,B ). Substituting a quarter of the Snx3 by SNX-BARs (25 nM SNX-BAR GFP :75 nM Snx3) resulted in vigorous tubule formation on about 60% of the GUVs. This tubulation was stronger than the tubulation observed with a 25 nM SNX-BAR-GFP or with 100 nM mScarletI Snx3 alone. We observed a similar effect when cargos for both Snx3 and SNX-BAR were present simultaneously (Fig. EV3A,B ). Thus, hybrid coat formation allows Snx3 and SNX-BARs to provide higher membrane scaffolding activity than in respective homogeneous coats.

Correlates of hybrid coat formation in vivo

To search for in vivo correlates of hybrid coat formation, we first compared the localisation of Snx3 and SNX-BARs by tagging Snx3, Vps5 or Vps17 with either mNeonGreen (mNG) or yomCherry (Fig. 7A,B ). The major pool of Snx3 mNG colocalized with Vps5 yomCherry (Fig. 7A ). To have a reference point, we performed the same experiment with Vps5 yomCherry and Vps17 mNG as these proteins form a stable heterodimer (Seaman et al, 1998 ; Seaman and Williams, 2002 ) and can therefore provide a benchmark for maximal colocalization (Fig. 7A ). The Pearson correlation coefficients showed similar colocalization between Snx3 mNG and Vps5 yomCherry as between Vps5 yomCherry and Vps17 mNG (Fig. 7B ). Ablation of Vps35, which is necessary for forming hybrid and homogeneous Snx3 coats in vitro (see above), reduced colocalization of Snx3 mNG and Vps5 yomCherry and displaced Snx3 mNG mainly to the vacuolar membrane and the cytosol. Although the resolution of these experiments does not permit to resolve individual tubules and visualise hybrid coat formation in situ directly, the result is at least consistent with this concept because it relates the co-existence of SNX-BAR and Snx3 in the same compartment to coat formation.

To address this point more directly, we assayed the interaction of Snx3 and SNX-BARs. The binding sites of Snx3 and SNX-BARs on Retromer overlap (Leneva et al, 2021 ; Kovtun et al, 2018 ), rendering the simultaneous binding of Snx3 and SNX-BARs to a single Retromer complex unlikely. However, Retromer forms arch-like dimers, which might permit binding of SNX-BARs at one end and of Snx3 at the other. To test this, we generated a strain carrying tagged SNX-BAR (Vps5 HA ), Snx3 (Snx3 V5 ) and Retromer (Vps35 yomCherry ). We also generated a vps35 mutant impaired in Snx3 binding ( vps35 QGRE , with the substitutions Q181A, G182A, R185A, and E186A). These residues were determined by combining information from previous chemical cross-linking studies (Harrison et al, 2014 ) with Alphafold 3 modelling and an assessment of their contribution to the interaction interface using the PDB-PISA software (Fig. EV4 ). We used immunoprecipitation assays to probe whether Snx3 and SNX-BARs interact through Retromer. In the immunoprecipitation experiments, wild-type Vps35 yomCherry pulled down both Snx3 V5 and Vps5 HA . By contrast, V ps35 QGRE -yomCherry lost the interaction with Snx3 V5 but maintained the interaction with Vps5 HA (Fig. 7C,D ). When we pulled down Snx3 V5 from VPS35 yomCherry wild-type cells, both Vps35 yomCherry and Vps5 HA co-adsorbed to the beads (Fig. 7E,F ). The Snx3 V5 -Vps5 HA interaction was reduced in the vps35 QGRE background. This suggests that Retromer can bridge Snx3 and SNX-BARs.

To examine the in vivo relevance of this association, we tagged yeast Retromer cargo proteins and followed their fate in vivo (Fig. 8 ). We used Vps10 as a bona fide SNX-BAR cargo (Bean et al, 2017 ; Purushothaman and Ungermann, 2018 ; Suzuki et al, 2019 ), and Ear1 and Ste13 as Snx3 cargos. In wild-type cells, a Vps10 mNG fusion protein localised to cytosolic puncta which, based on previous work (Suzuki et al, 2019 ; Day et al, 2018 ), should represent yeast endosome/late Golgi (Fig. 8A ). In SNX-BAR knockout cells ( vps5Δ or vps17Δ ), Vps10 mNG colocalised with the vacuolar marker FM4-64. This vacuolar localisation results from a failure to recycle Vps10 to the Golgi (Nothwehr et al, 1999 ; Cereghino et al, 1995 ). In snx3Δ knockout cells, Vps10 mNG localisation was moderately affected, as evident from cells showing weak staining of the vacuolar membrane with Vps10 mNG , in line with previous results (Bean et al, 2017 ). The Increased colocalization between Vps10 mNG and the vacuolar lipid marker FM4-64 confirmed this limited mislocalization to the vacuolar membrane (Fig. 8B ). Furthermore, the vps35 QGRE mutant, which targets the Retromer-Snx3 interaction, showed a partial accumulation of Vps10 mNG on the vacuolar membrane, suggesting that the recruitment of Snx3 to Retromer is relevant for Vps10 sorting (Fig. EV5 ). Vice versa, the Snx3 cargos Ear1 mNG (Fig. 8C,D ) and Ste13 mNG (Fig. 8E,F ) showed significant mislocalization to vacuoles not only in snx3Δ , but also in vps5Δ or vps17Δ cells, suggesting that the Snx3 recycling pathway requires SNX-BARs for its function. We extended these observations to two other Snx3 cargos, Kex2 and Pep12 (Voos and Stevens, 1998 ; Hettema et al, 2003 ), with similar results (Fig. EV6 ). These in vivo observations are in line with our in vitro findings and support the notion that Retromer can link SNX-BARs and Snx3 into hybrid coats that promote recovery of cargos for both classes of sorting nexins.

Discussion

Structural studies revealed many features of Retromer associated with either the SNX-BAR Vps5 or Snx3 (Leneva et al, 2021 ; Kovtun et al, 2018 ; Lucas et al, 2016 ; Kendall et al, 2020 , 2022 ). These coats share the organisation of Retromer dimers into arches that cross-link a SNX layer on a membrane tubule. The available structures support the current view that each SNX that associates with Retromer defines a separate sorting route. However, larger scale analyses of the impact of various endosomal coat components on the steady state distribution of a broad spectrum of cargo have revealed a significant overlap in the cargos affected by different classes of sorting nexins, including Snx3 and SNX-BARs, both in yeast and in mammalian cells (Bean et al, 2017 ; McNally et al, 2017 ; Steinberg et al, 2013 ). Further observations exist that are consistent with the formation of hybrid coats between Snx3 and SNX-BARs, but were not explicitly interpreted from this perspective. In mammalian cells, Retromer cargo for SNX3 (Wntless) and the SNX-BAR-associated protein SNX27 (β2-adrenergic receptor) (Harterink et al, 2011 ; Simonetti et al, 2022 ) were partially found in and left from the same endosomal sorting domains, although they use different sorting nexins and have different destinations (Varandas et al, 2016 ). The sorting signals of Wntless and β2-adrenergic receptor could be exchanged without impairing their entry into endosomal sorting domains (Varandas et al, 2016 ). Also in yeast, Snx3 and Vps17 were reported to partially co-localise, and both were necessary for efficient sorting of the iron transporter Ftr1 (Strochlic et al, 2007 ). Another link is provided by overexpression of the Snx3 cargo Ear1, which led to missorting of the SNX-BAR cargo Vps10 to the vacuole, and by the observation that Ear1 retrieval was compromised in a vps5Δ mutant (Suzuki et al, 2019 ). These studies were not designed to exclude indirect effects and were hence not interpreted in the context of hybrid coat formation. Therefore, models linking one type of homogenous sorting nexin coat with a defined population of cargo and a transport direction have prevailed in the general discussion of endosomal protein sorting. Our direct observation of the hybrid coat formation from purified proteins now provides a physical correlate and a potential explanation for the overlaps in the coat dependencies of cargo transport.

The hybrid coats assembled from Snx3 and SNX-BARs that we have analysed can integrate cargo peptides for both sorting nexin classes, suggesting that they are functional for cargo collection. However, they have properties that are distinct from those of the respective homogeneous coats. While both Snx3 and SNX-BARs oligomerize with Retromer into stable coats on SMTubes, Snx3/Retromer cannot constrict wider tubules, which SNX-BAR-Retromer readily scaffolds. Such differences could of course be due to factors such as the lipid composition of endosomal membranes, membrane tension, as well as the dynamics of coat formation, which may not be the same for the two coats. A more parsimonious, and in our view more likely explanation, is that the scaffolding potential of the two coats reflects their different capacity to self-interact. SNX-BARs self-interact and form a lattice on their own that can deform the membrane, and Retromer adds additional bonds and driving force to this process (Gopaldass et al, 2023 ; Simunovic et al, 2015 ; Lopez-Robles et al, 2023 ). In a Snx3 coat, however, the sorting nexins do not self-interact, and Retromer is the only element that crosslinks them and imposes curvature on the membrane (Leneva et al, 2021 ; Kovtun et al, 2018 ). Integration into hybrid coats apparently compensates for this lack of scaffolding activity of Snx3, allowing it to populate and constrict tubes of similar width as a homogeneous SNX-BAR coat, and to tubulate GUVs that it would otherwise not be able to deform.

Available structures of Snx3- and SNX-BAR Retromer coats suggest that simultaneous binding of the two sorting nexin classes to Retromer might be impossible due to steric clashes (Leneva et al, 2021 ; Kovtun et al, 2018 ; Lucas et al, 2016 ). Hybrid coats could thus form in two ways, which are not mutually exclusive. They might contain interspersed homogeneous zones that are covered only by Snx3-Retromer or SNX-BAR-Retromer and are too small to be resolved by our light-microscopic analysis. Alternatively, Retromer oligomers, which can form through dimerisation of Vps35 or Vps26 (Lucas et al, 2016 ; Leneva et al, 2021 ; Kovtun et al, 2018 ; Kendall et al, 2022 , 2020 ), could cross-link Snx3 and SNX-BARs into a joint, oligomeric structure. This latter possibility is favoured by the Retromer-dependent interaction of Snx3 with the SNX-BARs that we have detected.

The sorting nexin densities of homogeneous Snx3- and SNX-BAR-based coats on SMTs are comparable. After their formation, these homogeneous coats do not readily exchange subunits. They cannot integrate sorting nexins from the respective other class, suggesting that they are stable structures and that they do not contain suitable voids to convert into hybrid coats. The composition of the coats is apparently defined when they are being formed. During formation, however, there is considerable adaptability. Hybrid coats can form at variable Snx3/SNX-BAR ratios. Our replacement experiments suggest that this leaves the overall sorting nexin density unchanged, implying that SNX-BAR and Snx3 subunits occupy similar average membrane areas in the coats. The adaptability of hybrid coat composition is an interesting feature, particularly in combination with a role for cargo in coat formation. Snx3, for example, can bind to membranes by itself, but its integration into coats requires cargo (Fig. 1 )(Lucas et al, 2016 ; Leneva et al, 2021 ). An adaptable coat stoichiometry combined with cargo dependence could offer a simple mechanism for adjusting coat composition to the amounts of cargos for Snx3 and SNX-BAR that must be transported.

Hybrid coats can have advantages. For example, they enhance the scaffolding activity of Snx3, enabling coat formation on less curved membranes that do not support a homogeneous Snx3–Retromer coat, presumably because such a coat cannot provide the required driving force. However, the formation of hybrid coats also has important conceptual implications. We generally assume that cargo recruitment into a homogeneous carrier—mediated by interactions between targeting motifs and coat components—is the major factor determining the cargo’s final destination. When cargos with different destinations enter the same hybrid carrier, targeting must instead occur after carrier formation through additional mechanisms.

Several possibilities have been proposed (Varandas et al, 2016 ). These may include correction through subsequent trafficking steps that remove misdelivered cargo or selective retention at the correct location, for example, through interactions with resident proteins. Biophysical properties of the cargo may also contribute to this retention, such as the length of its transmembrane domain. Organelles and the plasma membrane exhibit distinct, characteristic membrane thicknesses; therefore, the transmembrane domain length must match the compartment in which the cargo resides (Sharpe et al, 2010 ; Watson and Pessin, 2001 ; Cosson et al, 2013 ).

Taken together, these considerations lead to a concept in which a cargo protein’s sorting signal and its targeting information become partially decoupled. The sorting signal may function primarily as an entry ticket into a carrier without necessarily specifying the cargo’s ultimate destination. It may increase the efficiency of recruitment into a coat—serving as an accelerator of transport along a trafficking route—whereas the final location would be determined by a more complex set of interactions and compatibility cues. This final destination might be reached only after several rounds of trafficking and “error correction,” which ultimately leads the cargo to the correct compartment, where it encounters the most favourable set of matching interactions for its retention.

We consider it likely that additional combinations of hybrid coats exist. This may apply to Retromer, because in yeast Snx3 and Snx4 substantially co-localise with the SNX-BAR protein Vps17 (Strochlic et al, 2007 ). More broadly, it is conceivable that other endosomal coat components may also co-enrich on tubules generated at the surface of endosomes. Many such components are sorting nexins with affinities for the highly curved environment of tubular membranes (Teasdale and Collins, 2012 ; van Weering et al, 2010 ; Solinger and Spang, 2022 ; Shortill et al, 2022 ). Thus, once tubules are generated—e.g. through the activity of motors or actin polymerisation—they may attract a mixture of coat components.

Testing this possibility will require careful live-imaging approaches both in vivo and in vitro, as well as continued structural analyses of coat complexes. This will be particularly interesting in metazoan cells, which use a wider and more complex array of coat systems than yeast, that can exhibit different or additional properties.

Methods

Materials

Lipids were purchased from Avanti Polar Lipids (USA): Egg L-alpha-phosphatidylcholine (EPC); 1,2-dioleoyl-sn-glycero-3-phospho-L-serine sodium salt (DOPS); 1,2-dioleoyl-sn-glycero-3-phospho-(1’-myo-inositol-3’-phosphate) (PI3P). 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(Cyanine 5.5) (Cy5.5 PE); 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl] (nickel salt) (DGS-NTA(Ni)). All lipids were dissolved in chloroform and stored under argon at -20 °C and used within 2 months. Phosphatidylinositol phosphates were dissolved in chloroform/methanol/water (20:9:1). Texas Red DHPE (Thermofisher cat. T1395MP) was purchased as a mixed isomer. The para isomer was separated by thin-layer chromatography as previously described (Dar et al, 2015 ).

Cell culture, strains and plasmids

SEY6210 yeast cells were grown at 30 °C in YPD (1% w/v yeast extract, 2% peptone, 2% dextrose) medium. Genes were deleted by replacing a complete open reading frame with a marker (Janke et al, 2004 ; Güldener et al, 1996 ) or CRISPR technique (Laughery et al, 2015 ) (see Tables 1 – 3 for a list of strains, plasmids and PCR primers used in this study). Gene tagging was performed as described (Sheff and Thorn, 2004 ) or with CRISPR (Laughery et al, 2015 ). Strains used for expression and purification of the Retromer complex have been described previously (Gopaldass et al, 2024 ).

Live microscopy

Vacuoles were stained with FM4-64 essentially as described (Desfougères et al, 2016 ). An overnight preculture in SC medium was used to inoculate a 10 ml culture. Cells were then grown in SC at 30 °C and 150 rpm to an OD 600 between 0.6 and 1.0. The culture was diluted to an OD 600 of 0.4, and FM4-64 was added to a final concentration of 10 µM from a 10 mM stock in DMSO. Cells were labelled for 60 min with FM4-64, washed three times in fresh media, and then incubated for 60 min in media without FM4-64. Just before imaging, cells were concentrated by a brief low-speed centrifugation and placed on a glass microscopy slide overlaid with a 0.17-mm glass coverslip. Z-stacks were taken with a spacing of 0.3 µm and assembled into maximum projections. Imaging was performed with a NIKON Ti2 spinning disc confocal microscope with a 100×1.49 NA lens and two Photometrics Prime BSI cameras. Image analysis was performed with ImageJ as described in (Gopaldass et al, 2023 ) for SMTubes. Pearson’s correlation coefficient was used to quantify the colocalization between Vps5, 17 and Snx3 using the “Just Another Colocalization Plug-in” (JACoP) in ImageJ on background-subtracted maximum projected images. A customised Python-based app was used to create masks defining a zone close to the FM4-64 signal around the vacuolar membrane and calculate Pearson’s correlation coefficient of Vps10 and other cargoes on these FM4-64 masks. All performed experiments were repeated at least three times. SD and SEM calculation and plotting were done with PRISM GraphPad software.

Protein immunoprecipitation and Western blotting

The interaction between SNX3, Retromer, and VPS5 was examined using a co-immunoprecipitation (co-IP) protocol adapted from (Suzuki et al, 2019 ). Yeast strains endogenously expressing Snx3-V5, Vps35-yomCherry, and Vps5-HA were inoculated in fresh YPD from saturated pre-cultures, grown overnight at 30 °C and harvested at mid-logarithmic phase (OD₆₀₀ = 0.8–1.0). A total of 50 mL of culture was collected and washed twice with immunoprecipitation (IP) buffer (20 mM HEPES-KOH, pH 7.2, 0.2 M sorbitol, 50 mM potassium acetate, 1× protease inhibitor cocktail (200 µM pefablock, 11 µM leupeptin, 500 µM o-phenanthroline, 7 µM pepstatin A), 1 mM PMSF, followed by centrifugation to pellet the cells.

Cell lysis was performed using 0.5-mm zirconia beads in 300 µL IP buffer at 4 °C for 5 min at 2500 rpm on an IKA ® VIBRAX ® VXR orbital shaker. Subsequently, 300 µL of IP buffer containing either 1.0% Triton X-100 (for Vps35 immunoprecipitation) or 1.5% Triton X-100 (for Snx3 immunoprecipitation) was added to reach final detergent concentrations of 0.5% and 0.75%. Lysates were incubated on a rotating wheel at 4 °C for 2 h, followed by clarification through centrifugation at 500× g for 5 min and then at 17,500× g for 10 min. Protein concentrations in the supernatants were quantified using a NanoDrop spectrophotometer and the absorption at 280 and 205 nm. 10 mg of total protein was used for each immunoprecipitation reaction. Samples were incubated with pre-equilibrated anti-RFP magnetic particles (RFP-Trap Magnetic Particles, Chromotek) or anti-V5 magnetic agarose beads (V5-Trap Magnetic Agarose, Chromotek) for 2 h at 4 °C. After incubation, beads were washed three times with IP buffer containing the appropriate Triton X-100 concentration (0.5% or 0.75%). Prior to the final wash (performed with detergent-free IP buffer), the beads were transferred to fresh microcentrifuge tubes. Bound proteins were eluted by boiling the beads at 95 °C for 10 min in 4× NuPAGE sample buffer supplemented with 10 mM DTT, 2 mM PMSF, and 1× PIC. Protein samples were subsequently resolved by SDS-PAGE on freshly cast 4–18% polyacrylamide gels.

Protein purification

TAP-tagged Retromer complex was extracted from yeast as previously described (Purushothaman et al, 2017 ; Purushothaman and Ungermann, 2018 ). Briefly, a 50 mL preculture of cells was grown for 24 h in YPGal medium. The next day, two 1 L cultures in YPGal were inoculated with 15 mL of preculture and grown for 20 h at 30 °C and 150 rpm to late log phase (OD 600 = 2–3). All the following steps were performed at 4 °C. Cells were pelleted and washed with 1 pellet volume of cold RP buffer (Retromer Purification buffer: 50 mM Tris pH 8.0, 300 mM NaCl, 1 mM MgCl 2 , 1 mM PMSF, and home-made protein inhibitor cocktail (200 µM pefablock, 11 µM leupeptin, 500 µM o-phenanthroline, 7 µM pepstatin A). Pellets were either processed immediately or flash-frozen in liquid nitrogen and stored at -80 °C. For cell lysis, the pellet was resuspended in one volume of RP buffer and passed once through a French press (One shot cell disruptor, Constant Systems LTD, Daventry, UK) at 2.2 Kpsi. DNase I was added to the lysate (final concentration 0.1 mg/mL) followed by a 20 min incubation on a rotating wheel. The lysate was precleared by centrifugation for 30 min at 45,000× g in a Beckman JLA 25.50 rotor and cleared by a 60 min centrifugation at 150,000× g in a Beckman Ti 60 rotor. The cleared supernatant was passed through a 0.2 µm filter and transferred to a 50 mL Falcon tube. In all, 1 mL IgG bead suspension (GE Healthcare, cat 17-0969-01) was added to the supernatant. After 60 min incubation on a rotating wheel, beads were spun down and washed three times with RP buffer and resuspended in 2 mL RP buffer. 250 µg of home-purified HIS-TEV protease from E. coli was added to the beads. After 30 min incubation at 4 °C, beads were centrifuged, the supernatant containing purified Retromer subcomplex was collected and concentrated on a 100 kDa cutoff column (Pierce™ Protein Concentrator PES, 100 K MWCO). The concentrated protein fraction was re-diluted in RP buffer and reconcentrated 3 times. This final step allowed for the removal of TEV protease and a high enrichment of intact complexes.

Snx3 (with or without a fluorescent tag (GFP or mScarletI) at the N-terminus) was cloned into the pGE-6P vector modified to contain a HRV3C protease cleavage site after the GST. Plasmids were transformed into E. coli BL21. A 50 mL preculture was used to inoculate 2 litres of LB media (37 °C). Cells were grown to an OD 600 of 0.8–0.9. The cultures were then cooled to 16 °C on ice, and IPTG (Roche) was added to a final concentration of 0.5 mM. Cells were incubated overnight at 16 °C, pelleted, washed with ice-cold RP buffer and resuspended in one pellet volume of RP buffer + 1 mM DTT. Cells were lysed in a French Press (One Shot cell disruptor, Constant Systems LTD, Daventry, UK). Insoluble material was pelleted at 30,000× g for 30 min in a Beckman JA25.50 rotor at 4 °C. Supernatants were filtered through a 0.2-μm filter (Sarstedt, Germany), and lysates were passed three times over 1 ml of Glutathion-Sepharose (Qiagen). The resin was washed with 30 ml of RP buffer, and resuspended in cleavage buffer (50 mM Tris pH = 8.0, 300 mM NaCl, 1 mM EDTA, 1 mM DTT), and 150 μg of HRV3C protease was added. Beads were incubated overnight at 4 °C. The supernatant containing purified Snx3 was analysed by SDS gel electrophoresis. For GFP Snx3, the protein was further purified by FPLC to remove unlabelled Snx3.

Proteins were concentrated to ~2 mg/mL, aliquoted in 10 µL fractions and flash-frozen in liquid nitrogen. Proteins were stored at -80 °C and used within 3 months. Thawed aliquots were used only once.

Peptide synthesis and labelling

Peptides were synthesised at the Protein and Peptide Chemistry Facility, University of Lausanne (Switzerland) and analysed by mass spectrometry. The homogeneity of all peptides was >90%, as indicated by analytical HPLC. For fluorescent labelling, 3–4 mg of peptides were dissolved in 100 µL DMSO and incubated at 20 °C for 12 h with an equimolar amount of maleimide-coupled fluorophore (Alexa Fluor™ 488 or 546 C5 Maleimide, Invitrogen). Labelled peptides were then purified by HPLC.

Cargo peptide sequences : numbering indicates the amino acid position in the full-length protein. The sequence in brackets at the end was added to permit labelling with a maleimide-coupled fluorophore.

Ste13: HHHHHHGGGG- 78 MRPRRESFQFNDIENQH

Ear1: HHHHHHGGGG- 475 GKKIINEEINLDSL-(GGC)

Vps10: HHHHHHGGGGGG- 1423 GGFARFGEIRLGDDGLIE-(GGC).

Supported membrane tubes

SMTubes were generated as previously described (Dar et al, 2015 ; Gopaldass et al, 2024 ). Briefly, glass coverslips were first washed with 3 M NaOH for 5 min and rinsed with water before a 60 min treatment with piranha solution (95% H 2 SO 4 /30% H 2 O 2 3:2 v/v). Coverslips were rinsed with water and dried on a heat block at 90 °C. Coverslips were then silanized with 3-glycidyloxypropyltrimethoxysilane (Catalogue no. 440167, Sigma) for 5 h under vacuum, rinsed with acetone and dried. Polyethylene glycol coating was performed by placing the coverslips in a beaker containing PEG400 (Sigma) at 90 °C for 60 h. Coverslips were washed with distilled water and stored for up to 2 months at room temperature in a closed container.

To generate supported membrane tubes, lipids were mixed from 10 mg/mL stocks in a glass vial and diluted to a final concentration of 1 mg/mL in chloroform. The same lipid mix was used throughout this study (5% PI3P, 15% DOPS, 0.1% Texas Red DHPE, 79.5% egg-PC). Lipids were then spotted (typically 1 µL, corresponding to about 1 nmol) on the coverslips and dried for 30 min under vacuum. The coverslip was mounted on an IBIDI 6-channel µ-slide (µ-Slide VI 0.4. IBIDI, Cat. No: 80606). Lipids were hydrated for 15 min with buffer (PBS pH 7.2: 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 )) and SMTs were generated by injecting PBS into the chamber using an Aladdin Single-Syringe Pump (World Precision Instruments, model AL-1000) at a flow rate of 1.5 mL/min for 5 min. SMTs were left to stabilise without flow for 5 min before the start of the experiment. Protein stocks (typically 1–2 µM) were first diluted in PBS and then injected into the chamber at a flow rate of 80 µL per minute. Tubes were imaged with a NIKON Ti2 spinning disc confocal microscope equipped with a 100×1.49 NA objective.

Quantification of SMT fluorescence

SMT fluorescence was quantified with ImageJ as described previously (Gopaldass et al, 2023 ). Briefly, line scan analysis was performed along tubules using ImageJ. For each line scan, a Gaussian curve was fitted, and the maximum height was extracted. Maximum height was then plotted against the tube length for all channels. For quantification of the radius of the tubes, lipid fluorescence values of a tubule underneath a constricted protein domain, extracted from the series of line scans described above, were sorted in ascending order. The curve typically showed two plateaus, the lower corresponding to the constricted state and the higher to the non-constricted one. Plotting the corresponding FP values confirmed that the FP-labelled protein localised to the constricted zone. For each tube, the zones corresponding to the constricted and non-constricted areas were determined manually, and the mean fluorescence value was used to calculate the tube radius or the FP intensity in the corresponding region.

Supplementary information

Peer Review File

Source data Fig. 1

Source data Fig. 2

Source data Fig. 3

Source data Fig. 4

Source data Fig. 5

Source data Fig. 6

Source data Fig. 7

Source data Fig. 8

Figures EV Source Data

Expanded View Figures

Acknowledgements

We thank Arianna Ravera from the DCSR at CI-UNIL for implementing the Python-based app used for image analysis. This work has been supported by grants from the Swiss National Science Foundation to AM (179306, 204713 and 10.006.083), and by EMBO to SRC (ALTF 240-2023).

Expanded view

Author contributions

Navin Gopaldass : Conceptualisation; Data curation; Formal analysis; Validation; Investigation; Methodology; Writing—original draft; Writing—review and editing. Sudeshna Roy Chowdhury : Data curation; Investigation; Methodology; Writing—original draft; Writing—review and editing. Ana Catarina Alves : Data curation; Investigation; Methodology; Writing—original draft; Writing—review and editing. Lydie Michaillat Mayer : Resources; Investigation; Methodology. Véronique Comte-Miserez : Resources; Investigation; Methodology. Andreas Mayer : Conceptualisation; Resources; Supervision; Methodology; Writing—original draft; Project administration; Writing—review and editing.

Source data underlying figure panels in this paper may have individual authorship assigned. Where available, figure panel/source data authorship is listed in the following database record: biostudies:S-SCDT-10_1038-S44318-026-00716-0 .

Data availability

This study includes no data deposited in external repositories.

The source data of this paper are collected in the following database record: biostudies:S-SCDT-10_1038-S44318-026-00716-0 .

Disclosure and competing interests statement

The authors declare no competing interests.

Contributor Information

Navin Gopaldass, Email: navin.gopaldass@unil.ch.

Andreas Mayer, Email: andreas.mayer@unil.ch.

Supplementary information

Expanded view data, supplementary information, appendices are available for this paper at 10.1038/s44318-026-00716-0.

References

Antón-Plágaro C, Chen KE, Guo Q, Liu M, Evans AJ, Lewis PA, Heesom KJ, Wilkinson KA, Collins BM, Cullen PJ (2025) Mapping of endosomal proximity proteomes reveals Retromer as a hub for RAB GTPase regulation. Nat Commun. 16:6990.

Bean BD, Davey M, Conibear E (2017) Cargo selectivity of yeast sorting nexins. Traffic 18:110–122

Buser DP, Spang A (2023) Protein sorting from endosomes to the TGN. Front Cell Dev Biol 11:1140605

Cereghino JL, Marcusson EG, Emr SD (1995) The cytoplasmic tail domain of the vacuolar protein sorting receptor Vps10p and a subset of VPS gene products regulate receptor stability, function, and localization. Mol Biol Cell 6:1089–1102

Chandra M, Collins BM, Jackson LP (2022) Biochemical basis for an interaction between SNX27 and the flexible SNX1 N-terminus. Adv Biol Regul 83:100842

Chandra M, Kendall AK, Ford MGJ, Jackson LP (2025) VARP binds SNX27 to promote endosomal supercomplex formation on membranes. Sci Adv 11:eadr9340

Chen C, Garcia-Santos D, Ishikawa Y, Seguin A, Li L, Fegan KH, Hildick-Smith GJ, Shah DI, Cooney JD, Chen W et al (2013) Snx3 regulates recycling of the transferrin receptor and iron assimilation. Cell Metab 17:343–352

Chen K-E, Tillu VA, Gopaldass N, Chowdhury SR, Leneva N, Kovtun O, Ruan J, Guo Q, Ariotti N, Mayer A et al (2025) Molecular basis for the assembly of the Vps5-Vps17 SNX-BAR proteins with Retromer. Nat Commun 16:3568

Cosson P, Perrin J, Bonifacino JS (2013) Anchors aweigh: protein localization and transport mediated by transmembrane domains. Trends Cell Biol 23:511–517

Cui Y, Carosi JM, Yang Z, Ariotti N, Kerr MC, Parton RG, Sargeant TJ, Teasdale RD (2018) Retromer has a selective function in cargo sorting via endosome transport carriers. J Cell Biol 218:615–631

Cullen PJ, Korswagen HC (2012) Sorting nexins provide diversity for retromer-dependent trafficking events. Nat Cell Biol 14:29–37

Dar S, Kamerkar SC, Pucadyil TJ (2015) A high-throughput platform for real-time analysis of membrane fission reactions reveals dynamin function. Nat Cell Biol 17:1588–1596

Day KJ, Casler JC, Glick BS (2018) Budding yeast has a minimal endomembrane system. Dev Cell 44:56–72.e4

Derivery E, Helfer E, Henriot V, Gautreau A (2012) Actin polymerization controls the organization of WASH domains at the surface of endosomes. PLoS ONE 7:e39774

Derivery E, Sousa C, Gautier JJ, Lombard B, Loew D, Gautreau A (2009) The Arp2/3 activator WASH controls the fission of endosomes through a large multiprotein complex. Dev Cell 17:712–723

Desfougères Y, Neumann H, Mayer A (2016) Organelle size control—increasing vacuole content activates SNAREs to augment organelle volume through homotypic fusion. J Cell Sci 129:2817–2828

Ghai R, Mobli M, Norwood SJ, Bugarcic A, Teasdale RD, King GF, Collins BM (2011) Phox homology band 4.1/ezrin/radixin/moesin-like proteins function as molecular scaffolds that interact with cargo receptors and Ras GTPases. Proc Natl Acad Sci USA 108:7763–7768

Gopaldass N, Chen KE, Collins B, Mayer A (2024) Assembly and fission of tubular carriers mediating protein sorting in endosomes. Nat Rev Mol Cell Biol 25:765–783

Gopaldass N, Leo MGD, Courtellemont T, Mercier V, Bissig C, Roux A, Mayer A (2023) Retromer oligomerization drives SNX-BAR coat assembly and membrane constriction. EMBO J 42:e112287

Güldener U, Heck S, Fiedler T, Beinhauer J, Hegemann JH (1996) A new efficient gene disruption cassette for repeated use in budding yeast. Nucleic Acids Res 24:2519–2524

Guo Q, Chen K, Gimenez-Andres M, Jellett AP, Gao Y, Simonetti B, Liu M, Danson CM, Heesom KJ, Cullen PJ et al (2024) Structural basis for coupling of the WASH subunit FAM21 with the endosomal SNX27–Retromer complex. Proc Natl Acad Sci USA 121:e2405041121

Harrison MS, Hung CS, Liu TT, Christiano R, Walther TC, Burd CG (2014) A mechanism for retromer endosomal coat complex assembly with cargo. Proc Natl Acad Sci USA 111:267–272

Harterink M, Port F, Lorenowicz MJ, McGough IJ, Silhankova M, Betist MC, Weering JRT, van, Heesbeen RGHP van, Middelkoop TC, Basler K et al (2011) A SNX3-dependent retromer pathway mediates retrograde transport of the Wnt sorting receptor Wntless and is required for Wnt secretion. Nat Cell Biol 13:914–923

Hettema EH, Lewis MJ, Black MW, Pelham HRB (2003) Retromer and the sorting nexins Snx4/41/42 mediate distinct retrieval pathways from yeast endosomes. EMBO J 22:548–557

Horazdovsky BF, Davies BA, Seaman MN, McLaughlin SA, Yoon S, Emr SD (1997) A sorting nexin-1 homologue, Vps5p, forms a complex with Vps17p and is required for recycling the vacuolar protein-sorting receptor. Mol Biol Cell 8:1529–1541

Janke C, Magiera MM, Rathfelder N, Taxis C, Reber S, Maekawa H, Moreno-Borchart A, Doenges G, Schwob E, Schiebel E et al (2004) A versatile toolbox for PCR-based tagging of yeast genes: new fluorescent proteins, more markers and promoter substitution cassettes. Yeast 21:947–962

Kendall AK, Chandra M, Xie B, Wan W, Jackson LP (2022) Improved mammalian retromer cryo-EM structures reveal a new assembly interface. J Biol Chem 298:102523

Kendall AK, Xie B, Xu P, Wang J, Burcham R, Frazier MN, Binshtein E, Wei H, Graham TR, Nakagawa T et al (2020) Mammalian retromer is an adaptable scaffold for cargo sorting from endosomes. Structure 28:393–405 e4

Kovtun O, Leneva N, Bykov YS, Ariotti N, Teasdale RD, Schaffer M, Engel BD, Owen DJ, Briggs JAG, Collins BM (2018) Structure of the membrane-assembled retromer coat determined by cryo-electron tomography. Nature 561:561–564

Laughery MF, Hunter T, Brown A, Hoopes J, Ostbye T, Shumaker T, Wyrick JJ (2015) New vectors for simple and streamlined CRISPR-Cas9 genome editing in Saccharomyces cerevisiae. Yeast. 32:711–720.

Lauffer BE, Melero C, Temkin P, Lei C, Hong W, Kortemme T, von Zastrow M (2010) SNX27 mediates PDZ-directed sorting from endosomes to the plasma membrane. J Cell Biol 190:565–574

Lee S, Chang J, Blackstone C (2016) FAM21 directs SNX27–retromer cargoes to the plasma membrane by preventing transport to the Golgi apparatus. Nat Commun 7:10939

Leneva N, Kovtun O, Morado DR, Briggs JAG, Owen DJ (2021) Architecture and mechanism of metazoan retromer:SNX3 tubular coat assembly. Sci Adv 7:eabf8598

Lopez-Robles C, Scaramuzza S, Astorga-Simon EN, Ishida M, Williamson CD, Banos-Mateos S, Gil-Carton D, Romero-Durana M, Vidaurrazaga A, Fernandez-Recio J et al (2023) Architecture of the ESCPE-1 membrane coat. Nat Struct Mol Biol 30:958–969

Lucas M, Gershlick DC, Vidaurrazaga A, Rojas AL, Bonifacino JS, Hierro A (2016) Structural mechanism for cargo recognition by the retromer complex. Cell 167:1623

Ma M, Burd CG (2019) Retrograde trafficking and plasma membrane recycling pathways of the budding yeast Saccharomyces cerevisiae. Traffic 21:45–59

McGough IJ, de Groot REA, Jellett AP, Betist MC, Varandas KC, Danson CM, Heesom KJ, Korswagen HC, Cullen PJ (2018) SNX3-retromer requires an evolutionary conserved MON2:DOPEY2:ATP9A complex to mediate Wntless sorting and Wnt secretion. Nat Commun 9:3737

McGough IJ, Steinberg F, Gallon M, Yatsu A, Ohbayashi N, Heesom KJ, Fukuda M, Cullen PJ (2014) Identification of molecular heterogeneity in SNX27-retromer-mediated endosome-to-plasma-membrane recycling. J Cell Sci 127:4940–4953

McNally KE, Faulkner R, Steinberg F, Gallon M, Ghai R, Pim D, Langton P, Pearson N, Danson CM, Nagele H et al (2017) Retriever is a multiprotein complex for retromer-independent endosomal cargo recycling. Nat Cell Biol 19:1214–1225

Nothwehr SF, Bruinsma P, Strawn LA (1999) Distinct domains within Vps35p mediate the retrieval of two different cargo proteins from the yeast prevacuolar/endosomal compartment. Mol Biol Cell 10:875–890

Purushothaman LK, Arlt H, Kuhlee A, Raunser S, Ungermann C (2017) Retromer-driven membrane tubulation separates endosomal recycling from Rab7/Ypt7-dependent fusion. Mol Biol Cell 28:783–791

Purushothaman LK, Ungermann C (2018) Cargo induces retromer-mediated membrane remodeling on membranes. Mol Biol Cell 29:2709–2719

Puthenveedu MA, Lauffer B, Temkin P, Vistein R, Carlton P, Thorn K, Taunton J, Weiner OD, Parton RG, von Zastrow M (2010) Sequence-dependent sorting of recycling proteins by actin-stabilized endosomal microdomains. Cell 143:761–773

Roux A, Koster G, Lenz M, Sorre B, Manneville J-B, Nassoy P, Bassereau P (2010) Membrane curvature controls dynamin polymerization. Proc Natl Acad Sci USA 107:4141–4146

Seaman MN, McCaffery JM, Emr SD (1998) A membrane coat complex essential for endosome-to-Golgi retrograde transport in yeast. J Cell Biol 142:665–681

Seaman MNJ, Williams HP (2002) Identification of the functional domains of yeast sorting nexins Vps5p and Vps17p. Mol Biol Cell 13:2826–2840

Sharpe HJ, Stevens TJ, Munro S (2010) A comprehensive comparison of transmembrane domains reveals organelle-specific properties. Cell 142:158–169

Sheff MA, Thorn KS (2004) Optimized cassettes for fluorescent protein tagging in Saccharomyces cerevisiae. Yeast 21:661–670

Shortill SP, Frier MS, Conibear E (2022) You can go your own way: SNX-BAR coat complexes direct traffic at late endosomes. Curr Opin Cell Biol 76:102087

Simonetti B, Daly JL, Cullen PJ (2023) Out of the ESCPE room: emerging roles of endosomal SNX-BARs in receptor transport and host–pathogen interaction. Traffic 24:234–250

Simonetti B, Danson CM, Heesom KJ, Cullen PJ (2017) Sequence-dependent cargo recognition by SNX-BARs mediates retromer-independent transport of CI-MPR. J Cell Biol 216:3695–3712

Simonetti B, Guo Q, Gimenez-Andres M, Chen KE, Moody ERR, Evans AJ, Chandra M, Danson CM, Williams TA, Collins BM et al (2022) SNX27-Retromer directly binds ESCPE-1 to transfer cargo proteins during endosomal recycling. PLoS Biol 20:e3001601

Simonetti B, Paul B, Chaudhari K, Weeratunga S, Steinberg F, Gorla M, Heesom KJ, Bashaw GJ, Collins BM, Cullen PJ (2019) Molecular identification of a BAR domain-containing coat complex for endosomal recycling of transmembrane proteins. Nat Cell Biol 21:1219–1233

Simunovic M, Voth GA, Callan-Jones A, Bassereau P (2015) When physics takes over: BAR proteins and membrane curvature. Trends Cell Biol 25:780–792

Solinger JA, Spang A (2022) Sorting of cargo in the tubular endosomal network. Bioessays 44:2200158

Steinberg F, Gallon M, Winfield M, Thomas EC, Bell AJ, Heesom KJ, Tavaré JM, Cullen PJ (2013) A global analysis of SNX27–retromer assembly and cargo specificity reveals a function in glucose and metal ion transport. Nat Cell Biol 15:461–471

Strochlic TI, Schmiedekamp BC, Lee J, Katzmann DJ, Burd CG (2008) Opposing activities of the Snx3-retromer complex and ESCRT proteins mediate regulated cargo sorting at a common endosome. Mol Biol Cell 19:4694–4706

Strochlic TI, Setty TG, Sitaram A, Burd CG (2007) Grd19/Snx3p functions as a cargo-specific adapter for retromer-dependent endocytic recycling. J Cell Biol 177:115–125

Sun D, Varlakhanova NV, Tornabene BA, Ramachandran R, Zhang P, Ford MGJ (2020) The cryo-EM structure of the SNX-BAR Mvp1 tetramer. Nat Commun 11:1506

Suzuki SW, Chuang YS, Li M, Seaman MNJ, Emr SD (2019) A bipartite sorting signal ensures specificity of retromer complex in membrane protein recycling. J Cell Biol 218:2876–2886

Teasdale RD, Collins BM (2012) Insights into the PX (phox-homology) domain and SNX (sorting nexin) protein families: structures, functions and roles in disease. Biochem J 441:39–59

Temkin P, Lauffer B, Jäger S, Cimermancic P, Krogan NJ, von Zastrow M (2011) SNX27 mediates retromer tubule entry and endosome-to-plasma membrane trafficking of signalling receptors. Nat Cell Biol 13:715–721

Thompson A, Nessler R, Wisco D, Anderson E, Winckler B, Sheff D (2007) Recycling endosomes of polarized epithelial cells actively sort apical and basolateral cargos into separate subdomains. Mol Biol Cell 18:2687–2697

Tian Y, Kang Q, Shi X, Wang Y, Zhang N, Ye H, Xu Q, Xu T, Zhang R (2021) SNX-3 mediates retromer-independent tubular endosomal recycling by opposing EEA-1-facilitated trafficking. PLoS Genet 17:e1009607

van Weering JRT, Sessions RB, Traer CJ, Kloer DP, Bhatia VK, Stamou D, Carlsson SR, Hurley JH, Cullen PJ (2012) Molecular basis for SNX-BAR-mediated assembly of distinct endosomal sorting tubules. EMBO J 31:4466–4480

van Weering JRT, Verkade P, Cullen PJ (2010) SNX–BAR proteins in phosphoinositide-mediated, tubular-based endosomal sorting. Semin Cell Dev Biol 21:371–380

Varandas KC, Irannejad R, von Zastrow M (2016) Retromer endosome exit domains serve multiple trafficking destinations and regulate local g protein activation by GPCRs. Curr Biol 26:3129–3142

Vardarajan BN, Bruesegem SY, Harbour ME, Inzelberg R, Friedland R, George-Hyslop PS, Seaman MNJ, Farrer LA (2012) Identification of Alzheimer disease-associated variants in genes that regulate retromer function. Neurobiol Aging 33:2231.e15–2231.e30

Voos W, Stevens TH (1998) Retrieval of resident late-Golgi membrane proteins from the prevacuolar compartment of Saccharomyces cerevisiae is dependent on the function of Grd19p. J Cell Biol 140:577–590

Wassmer T, Attar N, Harterink M, van Weering JRT, Traer CJ, Oakley J, Goud B, Stephens DJ, Verkade P, Korswagen HC et al (2009) The retromer coat complex coordinates endosomal sorting and dynein-mediated transport, with carrier recognition by the trans-Golgi network. Dev Cell 17:110–122

Watson RT, Pessin JE (2001) Transmembrane domain length determines intracellular membrane compartment localization of syntaxins 3, 4, and 5. Am J Physiol-Cell Physiol 281:C215–C223

Weeratunga S, Paul B, Collins BM (2020) Recognising the signals for endosomal trafficking. Curr Opin Cell Biol 65:17–27

Yang Z, Follett J, Kerr MC, Clairfeuille T, Chandra M, Collins BM, Teasdale RD (2018) Sorting nexin 27 (SNX27) regulates the trafficking and activity of the glutamine transporter ASCT2. J Biol Chem 293:6802–6811

Zhang Y, Pang X, Li J, Xu J, Hsu VW, Sun F (2021) Structural insights into membrane remodeling by SNX1. Proc Natl Acad Sci USA 118:e2022614118

Associated Data

Supplementary Materials

Peer Review File

Source data Fig. 1

Source data Fig. 2

Source data Fig. 3

Source data Fig. 4

Source data Fig. 5

Source data Fig. 6

Source data Fig. 7

Source data Fig. 8

Figures EV Source Data

Expanded View Figures

Data Availability Statement

This study includes no data deposited in external repositories.

The source data of this paper are collected in the following database record: biostudies:S-SCDT-10_1038-S44318-026-00716-0 .