key: cord-0316127-yq79dd8o
authors: Simonetti, Boris; Daly, James L.; Simón-Gracia, Lorena; Klein, Katja; Weeratunga, Saroja; Antón-Plágaro, Carlos; Tobi, Allan; Hodgson, Lorna; Lewis, Phil; Heesom, Kate J.; Shoemark, Deborah K.; Davidson, Andrew D.; Collins, Brett M.; Teesalu, Tambet; Yamauchi, Yohei; Cullen, Peter J.
title: ESCPE-1 Mediates Retrograde Endosomal Sorting of the SARS-CoV-2 Host Factor Neuropilin-1
date: 2022-01-22
journal: bioRxiv
DOI: 10.1101/2022.01.20.477115
sha: b40e1135427cdf0413d43204c0ac295391554d30
doc_id: 316127
cord_uid: yq79dd8o

Endosomal sorting maintains cellular homeostasis by recycling transmembrane proteins and associated proteins and lipids (termed ‘cargoes’) from the endosomal network to multiple subcellular destinations, including retrograde traffic to the trans-Golgi network (TGN). Viral and bacterial pathogens subvert retrograde trafficking machinery to facilitate infectivity. Here, we develop a proteomic screen to identify novel retrograde cargo proteins of the Endosomal SNX-BAR Sorting Complex Promoting Exit-1 (ESCPE-1). Using this methodology, we identify Neuropilin-1 (NRP1), a recently characterised host factor for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, as a cargo directly bound and trafficked by ESCPE-1. ESCPE-1 mediates retrograde trafficking of engineered nanoparticles functionalised with the NRP1-interacting peptide of the SARS-CoV-2 Spike protein. ESCPE-1 sorting of NRP1 may therefore play a role in the intracellular membrane trafficking of NRP1-interacting viruses such as SARS-CoV-2.

the electron-dense 3,3'-diaminobenzidine (DAB), we observed predominant labelling of the TGN upon activation of HRP-TGN46 (Fig. 1B,C and Extended Data Fig. 1C,D) .

The streptavidin-based affinity isolation of whole cell lysates following biotinylation showed specific labelling of endogenous proteins spanning the secretory pathway, including the Golgi-localised precursor isoform of the lysosomal hydrolase cathepsin D (CTSD); the glycosylation enzyme polypeptide N-acetylgalactosaminyltransferase 2 (GALNT2); the retrograde TGN-resident cargo cation-independent mannose 6phosphate receptor (CI-MPR); the endoplasmic reticulum (ER) protein calnexin; and cell surface receptors such as integrin-α5 (ITGA5) and N-cadherin (NCAD), that traverse the biosynthetic pathway en route to the plasma membrane (Fig. 1D) . SILACbased mass spectrometry revealed a list of 1237 proteins reproducibly enriched by streptavidin affinity isolation following HRP-TGN46 labelling ( Fig. 1E , Extended Data 

ESCPE-1 is an endosomal coat complex implicated in the retrograde trafficking of transmembrane proteins such as the CI-MPR 13,18-20 . ESCPE-1 consists of a heterodimer of SNX1 or SNX2, which are functionally redundant, associated with either SNX5 or SNX6, which are also functionally redundant and mediate direct binding to the cytosolic tails of transmembrane cargo 18-21 ( Fig. 2A) . Recent advances have revealed the identity of transmembrane cargoes that undergo sequencedependent recognition by ESCPE-1 for sorting to the plasma membrane 20,22 , yet comparatively less is known about cargoes re-routed towards the TGN.

To identify the transmembrane proteins that undergo ESCPE-1-dependent retrograde transport, SILAC-based quantitative proteomics was used to compare HRP-TGN46mediated biotinylation in Scramble (Scr) siRNA-treated and double SNX5+SNX6 siRNA-treated HeLa cells (due to functional redundancy both SNX5 and SNX6 must be suppressed to perturb ESCPE-1-dependent trafficking) 18, 19 . Retrograde trafficking of TGN46 has been demonstrated to be independent of 19 . Double SNX5+SNX6 knockdown did not perturb HRP-TGN46 labelling at the whole cell lysate level (Extended Data Fig. 2A) .

Of the previously established list of proteins biotinylated by HRP-TGN46, 46 proteins were significantly depleted (p < 0.05 Log2 fold-change < -0.26) in the SNX5+SNX6 siRNA HRP-TGN46-labelled proteome compared to Scramble siRNA-treated cells, of which 25 contained transmembrane-spanning domains (Fig. 2B, Extended Data Fig.   2B and Supplementary Tables 6-8) . STRING analysis identified a network of proteinprotein interactions, including a cluster of proteins containing Neuropilin-1 (NRP1) alongside integrin-α5, integrin-β8, CD44 and the receptor tyrosine kinases MET and EPHA2 (Fig. 1C) . Biological processes and molecular function categories pertaining to cellular adhesion and migration, transmembrane receptor kinase activity, and virus receptor activity were significantly enriched (Fig. 2D, Supplementary Tables 9-11 ).

Moreover, comparison of these proteins with published ESCPE-1 interactors and cell surface cargoes revealed a consistent enrichment of biological pathways (Extended Data Fig. 2C ).

NRP1 is a co-receptor for a range of extracellular ligands, including members of the vascular endothelial growth factor (VEGF) and semaphorin families, and was recently identified as a host factor that facilitates SARS-CoV-2 infection 23,24 . Moreover, NRP1 was also enriched in previously published interaction networks of SNX5, SNX6 (and the neuronal SNX6 paralogue SNX32), though its surface levels were unaffected by ESCPE-1 inactivating mutations 19,20 (Extended Data Fig. 2D ). Expression of an extracellular/luminally GFP-tagged murine NRP1 (GFP-Nrp1) revealed predominant localisation to the plasma membrane and an internal population that colocalised with SNX1-and SNX6-decorated endosomes, and with the TGN markers TGN46, Golgin-97 and CI-MPR (Fig. 2E,F) . Moreover, we confirmed by immunofluorescence staining that the TGN-resident pool of endogenous NRP1 was reduced in HeLa cells lacking ESCPE-1 subunits (Fig. 2G) .

The extracellular/luminal GFP-tag on the Nrp1 construct was amenable for an antibody uptake assay to facilitate the chase of internalised GFP-Nrp1 from the cell surface into intracellular compartments following endocytosis (Extended Data Fig. 3A ). At early timepoints GFP-Nrp1 colocalised with intracellular vesicles, a sub-population of which were SNX6-positive endosomes, and progressively colocalised with TGN46 at later timepoints ( Fig. 3A, Extended Data Fig. 3B ). 3D reconstruction of the TGN46 signal revealed a population of internalised GFP-Nrp1 within the TGN (Extended Data Fig.   3C ).

When we repeated the uptake assay in a previously described SNX5+SNX6 double KO HeLa cell line 20 , GFP-Nrp1 showed a decreased rate of colocalisation with TGN46 ( Fig. 3B) . Blocking of protein synthesis with cycloheximide (CHX) revealed an increased rate of endogenous NRP1 turnover in SNX5+SNX6 KO cells, consistent with enhanced lysosomal degradation in the absence of retrograde endosomal sorting ( Fig. 3C) . Importantly no significant change of GFP-Nrp1 trafficking was observed at earlier time-points, suggesting that internalisation and early endocytic trafficking of GFP-Nrp1 is not altered in the SNX5+SNX6 KO cells (Fig. 3B) .

ESCPE-1 mediates retrograde endosomal trafficking by driving the biogenesis of cargo-enriched tubulovesicular membrane carriers that emanate from endosomes and couple to dynein/dynactin motor complexes for transport towards the TGN 1,25-27 . Live imaging of cells co-expressing GFP-Nrp1 and mCherry-SNX1 revealed enrichment of the receptor in tubular profiles emanating from SNX1 endosomes (Fig. 3D , Extended Data Fig. 3D, Supplementary Videos 1,2) . To assess whether these were indeed tubular carriers undergoing endosome-to-TGN trafficking we co-expressed GFP-Nrp1 alongside a mCherry-CI-MPR construct, which is the prototypical ESCPE-1 retrograde cargo 18-20 . Fluorescently tagged Nrp1 and CI-MPR colocalised in the same tubular profiles (Fig. 3E, Extended Data Fig. 3E, Supplementary Video 3) , and endogenous SNX1 was found to decorate portions of these transport carriers (Fig. 3E) . We conclude that ESCPE-1 mediates tubular-based sorting of NRP1 from endosomes to the TGN, through a similar pathway to that of CI-MPR.

GFP-nanotrap of GFP-tagged ESCPE-1 subunits revealed that SNX5, SNX6 and SNX32 were able to immunoprecipitate endogenous NRP1, alongside CI-MPR (Fig.   4A ). Consistently, NRP1 was also enriched in SILAC-based interactomes of SNX5, SNX6 and SNX32 19 (Extended Data Fig. 4A ). The interaction with SNX5 and SNX6 was also observed for the NRP1 homologue NRP2 (Fig. 4B, Extended Data Fig.   4B ,C). Conversely, GFP-tagged NRP1 tail immunoprecipitated mCherry-tagged SNX5 and SNX6, and all endogenous ESCPE-1 subunits (Extended Data Fig. 4D,E) .

ESCPE-1 cargoes possess a ФxΩxФxnФ sorting motif in their cytosolic tail (whereby Ф corresponds to hydrophobic residues and Ω represents a central aromatic residue) that folds into a β-hairpin structure (comprised of two strands, denoted βA and βB, interspaced by a flexible loop of variable length, denoted xn) that is recognised by the extended PX domain of SNX5 and SNX6 20,28 . We mapped NRP1 binding to the SNX5 PX domain by engineering a SNX1 chimeric construct where the SNX1 PX domain was replaced with that of SNX5; this successfully immunoprecipitated endogenous NRP1 (Extended Data Fig. 4F) . Mutagenesis of the F136 residue within the extended SNX5 PX domain to aspartate prevented the interaction with NRP1, consistent with its inhibitory impact on CI-MPR binding 20 (Fig. 4C) . The cytosolic tail of NRP1 contains a stretch of residues (894-918), conserved in NRP1 and NRP2, conforming to the ФxΩxФxnФ motif for SNX5/SNX6 PX domain binding (Extended Data Fig. 4G ).

Isothermal titration calorimetry (ITC) showed that a synthetic peptide corresponding to these residues in NRP1 directly bound the PX domain of SNX5 and SNX6 with an affinity of 19.8 µM and 14.3 µM respectively, which is similar in strength to other known cargoes 20 ( Fig. 4D and Extended Data Fig. 4G ). Next, using structures of ФxΩxФxnФ cargoes bound to the SNX5 PX domain 20 , we generated a molecular model for the NRP1:SNX5 PX interaction. This was consistent with the NRP1 cytosolic tail folding into a beta hairpin that docked into the SNX5 PX domain (βA: 898-NYNFELV-904, loop: 905-DG-906, βB: 907-VKLKKD-912) (Fig. 4E) .

To validate the model, we generated a panel of mutants of the residues of the predicted βA and βB in the GFP-tagged cytosolic tail of NRP1 (Extended Data Fig.   4H ). Immunoprecipitation experiments revealed that the mutant forms of the NRP1 tail displayed a reduced interaction with mCherry-SNX6 (Extended Data Fig. 4I-J) . ITC experiments confirmed a dramatic loss of affinity of the NRP1 tail for the PX domain of SNX5 when the aromatic residues Y899 and F901 were mutated (Extended Data   Fig. 4K ). Furthermore, a triple deletion of the βA residues 898 NYN 900 produced the largest reduction in mCherry-SNX6 immunoprecipitation and endogenous SNX5 and SNX6 immunoprecipitation (Fig. 4F) . Importantly, deletion of these residues did not affect PSD-95/Dlg/ZO-1 (PDZ)-binding motif-mediated association with GIPC1, the scaffolding protein that regulates NRP1 internalisation through binding to a C-terminal PDZ-binding motif in NRP1 29-31 (Extended Data Fig. 4L ).

We next compared the trafficking of GFP-Nrp1 and GFP-Nrp1 ΔNYN in previously characterised NRP1 KO HeLa cells 23 . Although the total and surface levels of GFP-Nrp1 and GFP-Nrp1 ΔNYN Nrp1 were comparable, the retrograde trafficking of the mutant was reduced when compared to GFP-Nrp1 (Fig. 4G, Extended Data Fig. 4M) and resulted in the increased degradation of the mutant receptor (Fig. 4H) .

Consistently, in SNX5+6 double KO cells, GFP-tagged SNX5 rescued NRP1 turnover rate to wild-type levels, whereas rescue with GFP-SNX5(F136D), which does not immunoprecipitate NRP1, failed to do so (Fig. 4C,I) . These data establish that NRP1 possesses a canonical ESCPE-1 binding motif, and that direct interaction between NRP1 and ESCPE-1 is required for the correct retrograde trafficking of the internalised receptor.

We recently established that the extracellular b1 domain of NRP1 directly associates with a multibasic C-terminal motif (termed the C-end Rule motif 32 ) in the furinprocessed SARS-CoV-2 Spike S1 subunit to facilitate infectivity 23 . Therefore, we investigated whether ESCPE-1 could associate with the SARS-CoV-2 Spike (S) protein via NRP1. In HEK293T cells stably expressing a C-terminally truncated SARS-CoV-2 S gene (SARS-2 SΔ1256-1273), GFP-SNX5 and GFP-SNX6 coimmunoprecipitated bands corresponding to S1 and uncleaved S alongside endogenous NRP1 (Fig. 5A,B) . SARS-CoV-2 S does not contain a putative ФxΩxФxnФ motif within its cytosolic tail, suggesting that co-immunoprecipitation likely occurs through an intermediate protein. Importantly, the association between SNX5 and Spike was abrogated by the SNX5(F136D) mutation incapable of binding NRP1, consistent with Spike binding being mediated through an NRP1-dependent mechanism (Fig. 5B) .

To investigate whether the interaction between NRP1 and ESCPE-1 may regulate retrograde trafficking of internalised NRP1-bound viral particles, we engineered a minimal system comprising approximately coronavirion-sized (80 nm) silver nanoparticles coated with the C-end Rule (CendR) peptide motif of the SARS-CoV-2 S1 protein (Ag-SARS-CoV-2, sequence: TNSPRRAR, original isolate) 23,24 . Silver nanoparticles decorated with CendR sequences specifically bind NRP1 at the cell surface, and are internalised through an NRP1-dependent mechanism 24,32,33 .

Accordingly, NRP1-expressing PPC-1 cells incubated with Ag-SARS-CoV-2 demonstrated colocalisation with NRP1 at the cell surface at early timepoints ( Fig.   5C ). Adherence at the cell surface was followed by internalisation of nanoparticles and transit towards the perinuclear region, visualised by early colocalisation with the endosomal marker VPS35 and accumulating colocalisation with CI-MPR and TGN46 over 3 hours (Fig. 5C,D) . This phenomenon was ESCPE-1 dependent, as SNX5+SNX6 KO PPC-1 cells demonstrated a significant decrease in Ag-SARS-CoV-2 colocalisation with TGN46 and Golgin-97 (Fig. 5D, Extended Data Fig. 5A) . These data suggest that ESCPE-1 can govern the endosomal trafficking of internalised NRP1 ligands, raising the possibility of a role for the complex in SARS-CoV-2 infection (Extended Data Fig. 5B ).

We and others recently demonstrated that SARS-CoV-2 directly interacts with NRP1 at the cell surface to enhance cellular infection 23, 24 . In the present study, through the development of an unbiased proteomic screen to discover novel retrograde endosomal cargoes, we identified a range of potential transmembrane cargoes for the evolutionarily conserved ESCPE-1 complex. From this screen, we validate NRP1 as an interactor of ESCPE-1. ESCPE-1 directly engages the cytosolic tail of NRP1 in a sequence-dependent manner and mediates tubulovesicular endosomal sorting of NRP1 to the TGN. Additionally, we identify a wider functional network of transmembrane proteins perturbed by SNX5+SNX6 depletion, including integrin-α5 and integrin-β8, and the receptor tyrosine kinases MET and EPHA2, raising the possibility that ESCPE-1-mediated retrograde sorting may play a role in directional cell migration and signalling 34-36 . NRP1 has a reported role in mediating integrin internalisation and trafficking 30,37 , and associates with receptor tyrosine kinases including the MET receptor to modulate their signalling outputs 38,39 . Our study thus provides new mechanistic insight into the previously described role of retrograde endosomal sorting in cell migration 10,36 .

Considering the recent identification of NRP1 as a SARS-CoV-2 host factor, the mechanistic basis of its intracellular trafficking by ESCPE-1 opens interesting avenues for future investigation. SARS-CoV-2 exhibits two distinct cellular entry pathways: either direct fusion with the plasma membrane or internalisation into endosomal compartments 40 . It is presently unknown whether the SARS-CoV-2 virions that are internalised through endocytosis can hijack NRP1-and ESCPE-1-dependent endocytic trafficking to subvert innate cellular defenses, or whether the Spike protein alone undergoes this trafficking step. We demonstrate that ESCPE-1 coimmunoprecipitates the SARS-CoV-2 Spike through a sequence-dependent mechanism likely dependent on NRP1. Furthermore, engineered nanoparticles displaying the CendR motif of SARS-CoV-2 Spike hijacked ESCPE-1-dependent retrograde trafficking. ESCPE-1 couples to the dynein-dynactin complex to provide a mechanical pulling force that aids the biogenesis of tubular endosomal carriers that traffic towards the perinuclear region 25-27 . This process may conceivably provide additional energy input and membrane tension to facilitate the virus endosomal uncoating, as seen in influenza A virus 41, 42 . Indeed, modelling of the SARS-CoV-2

Spike protein binding to NRP1 and ACE2 has suggested that NRP1 facilitates S1/S2 separation, a prerequisite for membrane fusion 43 .

Interestingly, additional pathogens hijack ESCPE-1 cargo recognition to promote intracellular survival 2, 3, 44 , and SNX5 has also recently been identified as a key regulator of innate cellular immunity against a range of viruses 45 . Genome wide CRISPR screens and biochemical studies have identified multiple endosomal sorting machineries that facilitate SARS-CoV-2 infection, including components of the retromer, retriever and COMMD/CCDC22/CCDC93 (CCC) complexes 46-51 . Here, we suggest that ESCPE-1 sequence-dependent cargo sorting also plays a role in regulating the endosomal dynamics exploited during SARS-CoV-2 infection, potentially expanding the scope of pathogens that exploit can this complex.

A fascinating and impactful emerging theme is that the NRP1 pathway may influence infection by a wide range of viruses through the recognition of CendR motifs on viral glycoproteins 32,52 . Our findings therefore highlight the possibility that multiple viruses converge upon a NRP1-and ESCPE-1-dependent intracellular trafficking pathway within the endosomal network. We previously demonstrated that pharmacological inhibition of the SARS-CoV-2-NRP1 interaction limits infection in cell culture 23 . Future work will be required to appreciate the importance of NRP1 trafficking in SARS-CoV-2 biology, and the wider range of pathogens that exploit this receptor to mediate infectivity.

Bärlocher 

We 

For western blotting, cells were lysed in PBS with 1% (v/v) Triton X-100 and protease inhibitor cocktail. The protein concentration was determined with a BCA assay kit (Thermo Fisher Scientific), and equal amounts were resolved on NuPAGE 4-12%

precast gels (Invitrogen). Blotting was performed onto polyvinylidene fluoride membranes (Immobilon-FL; EMD Millipore) followed by detection using the Odyssey 

The methodology for HRP-TGN46 biotinylation is adapted from the APEX2 biotinylation protocol outlined in 15 . 10x10 6 HRP-TGN46-expressing cells were seeded in a 15 cm plate the day before biotinylation. The next day, cells were incubated in DMEM media supplemented with 500 μM biotin-phenol (BP) and incubated for 30 minutes at 37°C. Hydrogen peroxide (H2O2) was added at a final concentration of 1 mM and distributed by rocking the cell plate. After 45 seconds of H2O2 incubation, the media was removed and replaced with ice-cold, freshly prepared quencher solution consisting of 1 mM sodium ascorbate, 500 μM (±)-6-Hydroxy-2,5,7,8tetramethylchromane-2-carboxylic acid (Trolox, Sigma-Aldrich, 238813) and 1 mM sodium azide in PBS to ensure that the biotinylation reaction does not proceed beyond 1 minute. The quencher solution was left for 1 minute, then discarded, and this washing process was repeated 5 times. Following washes, cells were lysed in RIPA buffer (150 mM NaCl, 0.1% (v/v) SDS, 0.5% (w/v) sodium deoxycholate, 1% (v/v) Triton X-100, protease inhibitor cocktail, 50 mM Tris-HCl, pH 7.5) and lysates spun at 20,000 x g for 10 minutes at 4°C. 50μL aliquots of streptavidin beads were prepared and washed 3 times in RIPA buffer, centrifuging beads between washes at 350 x g.

The cell lysates were mixed with the streptavidin beads and rotated for 1 hour at 4°C.

After incubation, streptavidin beads were centrifuged and the supernatant containing unbound proteins is removed. The beads were washed 7 times (twice in RIPA buffer, once with 1M KCl, once with 0.1M Na2CO3, once with 2M Urea 10mM Tris-HCl pH 8.0, and twice again with RIPA buffer). All buffers were kept ice cold throughout the process. After the final wash step, all supernatant was aspirated off and beads were resuspended in 3X NuPAGE sample buffer supplemented with 2.5% βmercaptoethanol, 2mM free biotin and 20mM DTT.

For surface biotinylation experiments, fresh Sulfo-NHS-SS Biotin (Thermo Scientifics, #21217) was dissolved in ice-cold PBS at pH 7.8 at a final concentration of 0.2 mg / ml. Cells were washed twice in ice-cold PBS and placed on ice to slow down the endocytic pathway. Next, cells were incubated with the biotinylation reagent for 30 minutes at 4˚C followed by incubation in TBS for 10 minutes to quench the unbound biotin. The cells were then lysed in lysis buffer and subjected to Streptavidin beadsbased affinity isolation (GE-Healthcare).

Cells were fixed in 4% ( To obtain differential interference contrast (DIC) images of DAB polymerisation, a Leica DM IRBE inverted epifluorescence microscope (Leica Microsystems) was used.

An initial picture was taken prior to DAB labelling. 

All quantified Western blot are the mean of at least three independent experiments.

Statistical analyses were performed using GraphPad Prism 9 (LaJolla, CA). Graphs represent means and S.E.M. For all statistical tests, p < 0.05 was considered significant and is indicated by asterisks. The raw data files were processed and quantified using Proteome Discoverer software v2.1 (Thermo Fisher Scientific) and searched against the UniProt Human database using the SEQUEST HT algorithm. Peptide precursor mass tolerance was set at 10ppm, and MS/MS tolerance was set at 0.6Da. Search criteria included carbamidomethylation of cysteine as a fixed modification and oxidation of methionine, appropriate SILAC labels and the addition of biotin-phenol (+361.146Da) to tyrosine as variable modifications. Searches were performed with full tryptic digestion and a maximum of 4 missed cleavages were allowed. The reverse database search option was enabled and all data was filtered to satisfy false discovery rate (FDR) of 5%.

For statistical analysis of differential protein abundance between conditions, standard t-tests were used. Volcano plots were plotted using Orange software (University of Ljubljana) or VolcanoseR 53 . For generation of the HRP-TGN46-labelled proteome, proteins that were only identified in the HRP-TGN46 biotinylation condition in ≥ 4 out of 5 repeats and were not identified in negative control conditions and thus could not be statistically analysed, were assumed to be significant hits (Supplementary Table   2 ). Furthermore, 10 proteins significantly enriched in the heavy SILAC condition (HRP-TGN46 + BP -H2O2) relative to the light SILAC condition (WT HeLa + BP + H2O2)

were identified and removed from downstream analyses.

Gene ontology analysis was performed using the PANTHER classification system 54 

The expression plasmid of pGEX-4T-2 containing GST tagged SNX5 and SNX6 PX domain constructs were transformed into Escherichia coli BL 21 (DE3) cells and plated on lysogeny-broth (LB) agar plates supplemented with Ampicillin (0.1mg/mL). Single colony was then used to inoculate 50 mL of LB medium containing Ampicillin and the culture was grown overnight at 37 ºC with shaking at 180 rpm. The following day, 1L

of LB medium containing antibiotics Ampicillin (0.1mg/mL) was inoculated using 10 ml of the overnight culture. Cells were then grown at 37 ºC with shaking at 180 rpm to an optical density of 0.8-0.9 at 600 nm and the protein expression was induced by adding 0.5 mM IPTG (isopropyl-b-D-thiogalactopyranoside). Expression cultures were incubated at 20 ºC overnight and the cells were harvested next day by centrifugation at 4000 rpm for 15 min using Beckman rotor JLA 8.100. Cell pellets were then resuspended in 20 mL (for cell pellet from 1L) of lysis buffer (50 mM HEPES, pH 7.5, 500 mM NaCl, 5% glycerol, Benzamidine (0.1mg/mL), and Dnase (0.1 mg/mL)).

Resuspended cells were lysed by using the cell disrupter (Constant systems, LTD, UK, TS-Series) and the soluble fraction containing the protein was separated from cell debris by centrifugation at 18,000 rpm for 30 min at 4 ºC. The soluble fraction was first purified by affinity chromatography using Glutathione Sepharose 4B resin (GE Healthcare) and the GST tag was cleaved by incubating the protein with Thrombin (Sigma Aldrich) overnight at 4 ºC. Next day the protein was eluted using 50 mM HEPES, pH 7.5, 200 mM NaCl. The eluted protein was then concentrated and further purified by gel filtration chromatography (Superdex 75 (16/600), GE Healthcare) using 50 mM HEPES, pH 7.5, 200 mM NaCl, 0.5 mM TCEP (tri(2-carboxyethyl)phosphine) and the fractions corresponding to SNX5/SNX6 PX were analysed by SDS PAGE.

ITC experiments were carried out by using Microcal ITC200 instrument ( 

The model for the SNX5-NRP1 complex was built based on the 6n5z.pdb structure according to the method outlined in Supplementary Information. 20 ns atomistic dynamic simulations of the modelled complexes were carried out using the amber99sb-ildn forcefield in TIP3P waters and GROMACS 56 (2019.2) according to the method described recently 57 .

Silver nanoparticles labelled with the dye CF555 and functionalised with the biotinylated CendR peptide of the SARS-CoV-2 S1 protein (biotin-Ahx-TNSPRRAR; Ahx = aminohexanoic acid) were prepared as previously described 58,59 . The peptide was purchased from TAG Copenhagen, Copenhagen, Denmark.

PPC-1 cells (10 5 cells) were seeded onto noncoated coverslips ( 

HeLa cells were transfected with GFP-Nrp1. 24 hours after transfection, cells were live imaged using a confocal laser-scanning microscope at 37ºC and incidences of GFP-Nrp1 (greyscale) localisation on tubular structures were observed. Representative frames are displayed in Extended Data Figure 3D .

HeLa cells were co-transfected with GFP-Nrp1 and mCherry-SNX1. 24 hours after transfection, cells were live imaged using a confocal laser-scanning microscope at 37ºC and incidences of GFP-Nrp1 (green) and mCherry-SNX1 (red) colocalisation on tubular structures were observed. Representative frames are displayed in Figure 3D .

HeLa cells were co-transfected with GFP-Nrp1 and mCherry-SNX1. 24 hours after transfection, cells were live imaged using a confocal laser-scanning microscope at 37ºC and incidences of GFP-Nrp1 (green) and mCherry-SNX1 (red) colocalisation on tubular structures were observed. Representative frames are displayed in Extended Data Figure 3E .

Supplementary Table 1 . Complete list of proteins identified by SILAC-based proteomics following proximity labelling by HRP-TGN46. Untransfected HeLa cells were labelled in light (R0K0) SILAC media and treated with BP + H2O2, HRP-TGN46expressing HeLa cells were labelled in medium (R6K4) SILAC media and treated with BP + H2O2, and HRP-TGN46-expressing HeLa cells were treated with BP in the absence of H2O2. N = 5 independent replicates. 10 proteins that were significantly enriched in the heavy (HRP-TGN46 expressing cells incubated with BP in the absence of H2O2) relative to untransfected HeLa cells, some of which were endogenous biotinbinding proteins, are highlighted in red and were removed from subsequent analyses. 

To degrade or not to degrade: mechanisms and significance of endocytic recycling

Chlamydia interfere with an interaction between the mannose-6-phosphate receptor and sorting nexins to counteract host restriction

Structural basis for the hijacking of endosomal sorting nexin proteins by Chlamydia trachomatis

Scale bar = 20 µm, Inset = 5 µm. (C) 8 hour surface uptake of GFP-Nrp1 in HeLa cells. The volume of the perinuclear region, including the TGN was reconstructed from serial z-stacks. Scale bar = 20 µm, Inset = 5 µm. (D) HeLa cells were transfected with GFP-Nrp1 and live imaged after 24 hours. Asterisk = examples of Nrp1-positive tubular profiles that emanate from intracellular compartments

Enrichment and coverage of NRP1 in the previously published SILAC-based proteomics of SNX5, SNX6 and SNX32 19 . (B) HEK293T cells were cotransfected to express GFP-tagged NRP1 or NRP2 and mCherry-tagged SNX6 and subjected to GFP-nanotrap. The blot is representative of three independent GFP traps. (C) HEK293T cells were cotransfected to express GFP-tagged NRP1 or NRP2 and mCherry or mCherry-tagged SNX5 and subjected to mCherry-nanotrap. The blot is representative of three independent mCherry traps. (D) HEK293T cells were cotransfected to express GFP, NRP1-GFP or NRP2-GFP, and mCherry, mCherry-SNX5 or mCherry-SNX6 and subjected to GFP-nanotrap

HEK293T cells were transfected with GFP or GFP-NRP1 Tail, and lysates subjected to GFP nanotrap and blotting for ESCPE-1 subunits. Data representative of three independent repeats. (F) HEK293T cells were transfected to express GFPtagged SNX5, SNX1 and a SNX1 chimera generated by the replacement of the SNX1

Px domain with that of SNX5. Data representative of three independent repeats. (G) Sequence alignment of NRP1 and NRP2 from H. sapiens, and Nrp1 from M. musculus, R. norvegicus and D. rerio. (H) Schematics of the NRP1 cytosolic tail sequence that

 Table 8 . List of enriched and depleted proteins in the SNX5+6 siRNA HRP-TGN46-Labelled proteome. Proteins with a fold change of Log2 ± 0.26 and a pvalue of < 0.1 are displayed. Proteins that were depleted from the SNX5+6 siRNA proteome relative to the Scr siRNA proteome are classified by the presence of a transmembrane domain, and putative cytosolic ΩxΦxΩ SNX5/6 interacting motifs.Proteins that were either identified as ESCPE-1 interactors 19 or depleted from a GFP-SNX5(F136D) surface proteome 20 are also indicated. Table 9 . Gene ontology analysis of cellular component categories depleted in the SNX5+6 siRNA HRP-TGN46-labelled proteome. The list of proteins depleted in the SNX5+6 siRNA HRP-TGN46 proteome (Supplementary Table 8) were analysed with PANTHER gene ontology software. Significantly enriched/depleted biological process categories are displayed, ranked by p-value. + = significant overrepresentation, -= significant underrepresentation of categories. Table 10 . Gene ontology analysis of biological process categories depleted in the SNX5+6 siRNA HRP-TGN46-labelled proteome. The list of proteins depleted in the SNX5+6 siRNA HRP-TGN46 proteome (Supplementary Table 8) were analysed with PANTHER gene ontology software. Significantly enriched/depleted biological process categories are displayed, ranked by p-value. + = significant overrepresentation, -= significant underrepresentation of categories. Table 11 . Gene ontology analysis of molecular function categories depleted in the SNX5+6 siRNA HRP-TGN46-labelled proteome. The list of proteins depleted in the SNX5+6 siRNA HRP-TGN46 proteome (Supplementary Table 8) were analysed with PANTHER gene ontology software. Significantly enriched/depleted biological process categories are displayed, ranked by p-value. + = significant overrepresentation, -= significant underrepresentation of categories.