Mechanism and regulation of cargo entry into the Commander endosomal recycling pathway

Abstract

Commander is a multiprotein complex that orchestrates endosomal recycling of integral cargo proteins and is essential for normal development. While the structure of this complex has recently been described, how cargo proteins are selected for Commander-mediated recycling remains unclear. Here we identify the mechanism through which the unstructured carboxy-terminal tail of the cargo adaptor sorting nexin-17 (SNX17) directly binds to the Retriever sub-complex of Commander. SNX17 adopts an autoinhibited conformation where its carboxy-terminal tail occupies the cargo binding groove. Competitive cargo binding overcomes this autoinhibition, promoting SNX17 endosomal residency and the release of the tail for Retriever association. Furthermore, our study establishes the central importance of SNX17-Retriever association in the handover of integrin and lipoprotein receptor cargoes into pre-existing endosomal retrieval sub-domains. In describing the principal mechanism of cargo entry into the Commander recycling pathway we provide key insight into the function and regulation of this evolutionary conserved sorting pathway.

Fig. 1. SNX17 directly binds to Retriever.

A Purified His-tagged Retriever was mixed with purified SNX17(WT) or SNX17(L470G) and incubated with anti-His-tag TALON® Superflow beads. Input mixtures (left) and protein bound to the beads after washing (middle) were analysed by SDS-PAGE followed by Coomassie staining and western blotting. SNX17 bound to the beads was quantified and normalised to the level of VPS35L (right). n = 3, 1-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d. B AlphaFold2 predictions show a high confidence interaction between the unstructured carboxy-terminal region of SNX17 and Retriever (Fig S5A).

Fig. 2. SNX17 binds to the VPS35L-VPS26C interface of Retriever.

A HEK293T cells were transiently co-transfected with mCherry-SNX17 and either GFP, VPS35L-GFP or VPS35L-GFP mutants that perturb VPS35L-VPS26C (VPS35L(R248A)) and VPS35L-VPS29 (VPS35L(L35D)) associations prior to GFP-nanotrap isolation and quantitative western blot analysis of protein band intensities. n = 3, 1-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d. B HEK293T cells were transiently co-transfected with GFP, or VPS35L-GFP or VPS35L-GFP mutants that target SNX17 binding, and mCherry-SNX17. Protein lysates were then used in GFP-nanotrap experiments. Below, the quantitative analysis of protein band intensities is shown. n = 3, 2-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d., only changes with p < 0.05 are shown. C, D HEK293T cells were transiently co-transfected with GFP or GFP-SNX17 or GFP-SNX17 mutants in all conserved residues of the terminal ⁴⁶⁵IGDEDL⁴⁷⁰ motif (C) or conserved ⁴⁵⁹NFAF⁴⁶² motif (D) that target Retriever binding. Protein lysates were then used in GFP-nanotrap experiments. Below, the quantitative analysis of protein band intensities is shown. n = 3, 2-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d., only changes with p < 0.05 are shown.

Fig. 3. SNX17-Retriever coupling is essential for Retriever-cargo retrieval to plasma membrane.

A VPS35L KO RPE1 cells were lentivirally transduced with GFP, VPS35L-GFP or VPS35L-GFP(R248A). The stably expressed VPS35L(R248A) localizes to endosomes and can partially rescue COMMD1 localization. Scale bars correspond to 20 µm. Pearson’s coefficients were quantified from 3 independent experiments (GFP-VPS35 coloc. wt: n = 114 cells, R248A: n = 107 cells, VPS35-COMMD1 coloc. wt: n = 117 cells, R248A: n = 106 cells). Pearson’s coefficients for individual cells and means are presented by smaller and larger circles, respectively, colored according to the independent experiment. The means (n = 3) were compared using a two-tailed unpaired t-test. Error bars represent the mean, s.d. B VPS35L KO RPE1 cells were lentivirally transduced with GFP, VPS35L-GFP or VPS35L-GFP(R248A). The stably expressed VPS35L(R248A) failed to rescue Itgα5 missorting as evidenced by increased co-localisation with lysosomal marker LAMP1. Scale bars correspond to 20 µm. Pearson’s coefficients were quantified from 3 independent experiments (wt: n = 125 cells, R248A: n = 113 cells). Pearson’s coefficients for individual cells and means are presented by smaller and larger circles, respectively, colored according to the independent experiment. The means (n = 3) were compared using a two-tailed unpaired t-test. Error bars represent the mean, s.d. C Cell surface proteins were biotinylated in stably rescued RPE1 cells and enriched with streptavidin pull-down to analyze surface protein levels. The VPS35L(R248A) mutant failed to rescue cell surface levels of Itgα5, Itgβ1 and LRP1, but stabilised VPS26C. The quantitative analysis of protein band intensities is shown. The band intensities were normalised to the respective cell surface N-cadherin levels. n = 5, 2-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d., only changes with p < 0.05 are shown.

Fig. 4. Intramolecular association between SNX17 FERM domain and ⁴⁵⁹NFAF⁴⁶² autoinhibits SNX17 binding to cargo and Retriever.

A Overlay of the FERM domain of SNX17 bound to P-selectin (PDB: 4GXB), LRP1 (Supplementary Fig. 5C) and the intramolecular SNX17 peptide (Supplementary Fig. 5B) (as predicted by AlphaFold2) showing a clear overlap in peptide occupancy within the canonical cargo binding pocket. B The C-terminus of SNX17 contains an ⁴⁵⁹NFAF⁴⁶² motif highlighted in bright pink that is predicted by AlphaFold2 (Supplementary Fig. 5B) to bind into the canonical cargo binding pocket, in addition several residues in the extreme C-terminus (⁴⁶⁷DEDL⁴⁷⁰) are predicted to stabilize this interaction. C HEK293T cells were transiently co-transfected with GFP, or GFP-SNX17 or GFP-SNX17 mutants in the ⁴⁵⁹NFAF⁴⁶² motif to target its intra-molecular association with SNX17-FERM domain. Protein lysates were then used in GFP-nanotrap experiments. Below, the quantitative analysis of protein band intensities is shown. n = 3, 2-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d. D HEK293T cells were transiently co-transfected with GFP, or GFP-SNX17 or GFP-SNX17 mutants in the FERM(F3) domain to target its intra-molecular association with the⁴⁵⁹NFAF⁴⁶² motif. Protein lysates were then used in GFP-nanotrap experiments. Below, the quantitative analysis of protein band intensities is shown. n = 3, 1-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d., only changes with p < 0.05 are shown.

Fig. 5. Cargo binding to SNX17 relieves autoinhibition and promotes association with Retriever.

A Modelling of full-length SNX17 with the cytoplasmic tail of LRP1^4460-4490 predicts a perturbation in the intramolecular interaction defined by LRP1 preferentially binding to the cargo binding pocket releasing the disordered carboxy-tail of SNX17. B Isothermal titration calorimetry was performed with the cargo recognition peptide of ⁴⁴⁶⁶LRP1⁴⁴⁷⁹ and ⁷⁵⁵APP⁷⁶⁸ against either the full-length SNX17 (residues 1-470) or a construct which lacked the C-terminal tail (residues 1-390) (SNX17ΔC). In each case the peptide showed an ~2-fold decrease in binding when the C-terminal tail was absent. C Using a synthetic peptide of the C-terminus of ⁴⁵²SNX17⁴⁷⁰ we confirmed that this region can interact with SNX17ΔC, albeit with relatively low affinity, and that this interaction was inhibited when using the full-length construct or the SNX17ΔC construct following pre-incubation with LRP1. K_d ± SEM was calculated from triplicate data. D Purified His-tagged Retriever was mixed with purified SNX17 (WT) and 3 µM−30 µM of LRP1 (NPXY) or LRP1 (NPXA) peptide. The mixtures were incubated with anti-His-tag TALON® Superflow beads, then input mixtures and protein bound to the beads after washing were analyzed by SDS-PAGE followed by Coomassie staining and western blotting. SNX17 bound to beads was quantified and normalised to the level of VPS35L (right). n = 3, 1-way ANOVA with Dunnett’s multiple comparison test, data presented as mean values and error bars represent s.d., only changes with p < 0.05 are shown. E RPE1 cells were transiently transfected with GFP-SNX17 wild-type, or SNX17(V380D) or SNX17(F462A) mutants that decrease or enhance cargo binding, respectively. Fixed cells were examined with confocal microscope, and the localisation of the GFP-tagged constructs was compared to the localisation of early endosome marker EEA1. Scale bars correspond to 20 µm. Pearson’s coefficients were quantified from 3 independent experiments (wt: n = 62 cells, V380D: n = 60 cells, F462A n = 64 cells). Pearson’s coefficients for individual cells and means are presented by smaller and larger circles, respectively, coloured according to the independent experiment. The means (n = 3) were compared using a 1-way ANOVA with Dunnett’s multiple comparison test. Error bars represent the mean, s.d.

Fig. 6. Retriever resides on recycling sub-domain of endosome and colocalises with the markers of the CCC complex.

A–C VPS35L KO RPE1 cells were lentiviral transduced with GFP, VPS35L-GFP or VPS35L-GFP(R248A). The localisation of GFP-tagged proteins was compared to the localisation of endogenous endosomal markers SNX17, EEA1 (A), COMMD1 (B) or VPS35 (C). Representative confocal microscopy images are shown. The relative distributions of endosomal markers were evaluated in ImageJ by generating fluorescence intensity line profiles. Line profiles of 30 endosomes from 3 independent experiments were analysed in Rstudio, where the lengths of line scans and raw fluorescence intensities were normalised and averaged. The average profiles are shown on the right. Loess curve with 95% confidence interval. Scale bars shown for full image or inset correspond to 20 µm and 2 µm, respectively.

Fig. 7. Retriever depletion perturbs Retromer sub-domain organisation.

A Double immunogold labelling of endogenous SNX17 (15-nm gold) and VPS35L tagged with GFP (10-nm gold) in VPS35L KO RPE1 cells expressing wild-type VPS35L-GFP, showing spatial separation of SNX17 and VPS35L over vacuolar (E) and tubular (arrows) endosomal subdomains, respectively. A representative micrograph is shown (2 technical replicates). B Double immunogold labelling of endogenous SNX17 (15-nm gold) and VPS35 (10-nm gold) in parental RPE1 (panels i–ii) and VPS35L KO RPE1 cells (panels iii–iv). In parental cells, VPS35 (Retromer) is localized to endosomal tubules (arrows), whereas in VPS35L KO cells the majority of VPS35 was found on SNX17-decorated endosomal vacuoles. Representative micrographs are shown (2 technical replicates). Scale bars correspond to 200 nm. E = Endosomal vacuole.

Fig. 8. Model of SNX17-Commander association and its regulation through ØxNxx[Y/F] cargo-density sensing and endosomal sub-domain organization.

SNX17 associates with the endosomal membrane enriched for phosphatidylinositol 3-monophosphate (PI3P). We hypothesize that endosomal localization is enhanced through the binding of transmembrane cargoes containing ØxNxx[Y/F] sorting motifs. With increasing cargo density, the auto-inhibitory conformation defined by the SNX17 tail associating with the cargo binding groove in the FERM domain of SNX17 is displaced, enabling the presentation of SNX17 carboxy-tail to the conserved VPS26C:VPS35L interface. The direct binding of cargo-bound SNX17 to Retriever ultimately leads to Commander-mediated recycling of cargo back to the plasma membrane. For simplicity, other endosomal sorting complexes, such as the F-actin polymerizing WASH complex, ESCPE-1 and Retromer are not shown.