Summary

Sorting nexins (SNXs) are key regulators of the endosomal network. In designing an RNAi-mediated loss-of-function screen, we establish that of 30 human SNXs only SNX3, SNX5, SNX9, SNX15 and SNX21 appear to regulate EGF receptor degradative sorting. Suppression of SNX15 results in a delay in receptor degradation arising from a defect in movement of newly internalised EGF-receptor-labelled vesicles into early endosomes. Besides a phosphatidylinositol 3-phosphate- and PX-domain-dependent association to early endosomes, SNX15 also associates with clathrin-coated pits and clathrin-coated vesicles by direct binding to clathrin through a non-canonical clathrin-binding box. From live-cell imaging, it was identified that the activated EGF receptor enters distinct sub-populations of SNX15- and APPL1-labelled peripheral endocytic vesicles, which do not undergo heterotypic fusion. The SNX15-decorated receptor-containing sub-population does, however, undergo direct fusion with the Rab5-labelled early endosome. Our data are consistent with a model in which the EGF receptor enters the early endosome following clathrin-mediated endocytosis through at least two parallel pathways: maturation through an APPL1-intermediate compartment and an alternative more direct fusion between SNX15-decorated endocytic vesicles and the Rab5-positive early endosome.

Introduction

The endocytic network is a major cellular system required for nutrient uptake, cell signalling and receptor downregulation (Huotari and Helenius, 2011). It begins at the cell surface where a number of molecularly distinct endocytic events select a range of cargoes for internalisation (McMahon and Boucrot, 2011). Newly formed cargo-enriched endocytic vesicles fuse with the early endosome where sorting decisions are initiated, ultimately leading to cargos being sorted for lysosomal-mediated degradation or recycling to either the plasma membrane or the trans-Golgi network (TGN) (Hsu and Prekeris, 2010; Hurley and Hanson, 2010; Johannes and Wunder, 2011).

Historically, the early endosome was considered to comprise a single-membrane compartment formed through homotypic fusion of newly formed endocytic vesicles (Huotari and Helenius, 2011). However, increasing evidence argues that the early endosome, and the upstream newly formed endocytic vesicles, should be viewed as a heterogeneous collection of discrete compartments that differ in their location, dynamics, morphology and biochemical composition (Huotari and Helenius, 2011). Indeed, newly endocytosed cargo such as tranferrin receptor and epidermal growth factor (EGF) receptor (EGFR), undergo differential sorting into at least two distinct early endosome populations, a small population of dynamic rapidly Rab5 maturing early endosomes and a larger population of static slowly maturing early endosomes (Lakadamyali et al., 2006). Heterogeneity also exists at the level of compartment identity. The Rab5 effector APPL1 (adaptor protein containing pleckstrin homology domain, PTB domain and leucine zipper motif 1) labels a peripheral phosphatidylinositol 3-phosphate (PtdIns3P)-negative endocytic compartment that is a transient station for entry of EGFR into the early endosome (Miaczynska et al., 2004; Zoncu et al., 2009). Linking this compartment to late stages of clathrin-coated pits and newly formed clathrin-coated vesicles is OCRL (oculocerebrorenal syndrome of Lowe), an APPL1- and clathrin-binding inositol 5-phosphatase (Choudhury et al., 2005; Erdmann et al., 2007; Mao et al., 2009). A further OCRL-binding protein, Ses1/2 (also known as IPIP27A/B), associates with the OCRL compartment but because of mutually exclusive binding defines an endocytic sub-population downstream of the APPL1 station (Swan et al., 2010; Noakes et al., 2011). These distinct sub-populations are therefore considered to represent stages along a common maturation pathway that links newly formed endocytic vesicles with the PtdIns3P-enriched early endosome (Huotari and Helenius, 2011).

In the present study, we have designed an RNAi-mediated loss-of-function screen to explore the internalisation and endo-lysosomal degradative sorting of the EGFR. We have specifically targeted members of the human sorting nexin (SNX) family, a large group of evolutionarily conserved phosphoinositide-binding proteins that play central roles in the organisation and dynamics of various compartments within the endocytic network (Cullen, 2008; Cullen and Korswagen, 2012; Teasdale and Collins, 2012). Of the 30 SNXs examined, we establish that targeting of SNX3, SNX5, SNX9, SNX15 and SNX21 generate detectable phenotypes. In describing the detailed analysis of SNX15 (Barr et al., 2000; Phillips et al., 2001), we reveal a role for this sorting nexin in regulating the entry of newly internalised EGFR into the PtdIns3P-enriched early endosome. By describing the association of SNX15 with plasma-membrane-associated clathrin-coated pits and clathrin-coated vesicles, and identifying that this is mediated by a direct interaction with clathrin, we provide mechanistic details of a route through which EGFR can transit from clathrin-mediated endocytosis to the early endosome without proceeding through an APPL1-intermediate compartment. Together these data further define the underlying molecular heterogeneity in endosomal sub-populations that define the early endocytic network.

Results

Sorting nexin loss-of-function screen reveals a role for SNX15 in EGFR trafficking

To define the role of SNXs in endocytic sorting of the EGFR we designed a SNX RNAi loss-of-function screen. By using an anti-EGFR antibody and confocal microscopy we followed the time course of loss of activated EGFR as it underwent lysosomal degradation. Within a cellular population (n>200 cells), the extent of EGFR degradation, 90 minutes after addition of 100 ng/ml EGF, was quantified as the ratio of the immunofluorescent EGFR signal to a DAPI nuclear stain. To perform the screen we coupled this protocol with a previously described SNX targeting SMARTpool RNAi library (Wassmer et al., 2007). Biochemical analysis has quantified that RNAi suppression of SNX1 or SNX2 has no discernable effect on EGFR degradation (Carlton et al., 2004), and so we set a threshold describing a positive phenotype as values greater than observed for SNX1 or SNX2 suppression (Fig. 1A). Based on these criteria, of the 30 SNXs tested only five gave an EGFR phenotype. Proving the validity of the screen, the most pronounced effects were observed following suppression of SNX9, a protein with an established role in endocytosis (Lundmark and Carlsson, 2009), and SNX5, which through an association with the type Iγ phosphatidylinositol 4-phosphate 5-kinase i5 (PIPKIγi5), is required for EGFR sorting into intraluminal vesicles of the late endosome/multivesicular body, a prerequisite for lysosomal-mediated degradation (Sun et al., 2013). The remaining SNXs were SNX3, SNX15 and SNX21. The suppression of SNX3 and SNX21 each generated relatively minor effects on EGFR degradation (Fig. 1A). As a component of the SNX3 retromer, SNX3 regulates early endosome-to-TGN retrograde transport of Wntless (Harterink et al., 2011; Zhang et al., 2011). This SNX has also been implicated in the formation of multivesicular bodies, although, and consistent with the data from the current RNAi screen, suppression of SNX3 only marginally slowed the rate of EGFR degradation (Pons et al., 2008). For SNX21 no functional data are available, although strong expression in liver suggests a role in the development of this organ (Zeng et al., 2002).

Fig. 1.

siRNA loss-of-function screen to identify SNXs with roles in degradative sorting of EGFR. (A) HeLa cells transfected with siRNAs targeting individual SNXs were stimulated with 100 ng/ml EGF for 90 minutes. The total intensity of the EGFR immunofluorescent signal was measured using an ArrayScan II 96 well wide-field fluorescence imaging system. Of the 30 SNX family members screened, five exhibited increased EGFR levels. Data from >200 cells per condition. (B) Domain organisation of SNX15, including the clathrin-binding box identified in the present study. (C–E) HeLa cells transfected with control or SNX15 SMARTpool siRNA were stimulated with 100 ng/ml EGF or PDGF for the indicated times prior to immunoblotting with antibodies against EGFR, PDGFR, SNX15, phospho-ERK (42/44), or as a loading control tubulin or actin. (F) A degradation time course of EGFR following EGF stimulation revealing a decreased rate of EGFR degradation in SNX15-suppressed cells versus control. Data are from three independent experiments (means ± s.e.m. P≤0.015).

Fig. 1.

siRNA loss-of-function screen to identify SNXs with roles in degradative sorting of EGFR. (A) HeLa cells transfected with siRNAs targeting individual SNXs were stimulated with 100 ng/ml EGF for 90 minutes. The total intensity of the EGFR immunofluorescent signal was measured using an ArrayScan II 96 well wide-field fluorescence imaging system. Of the 30 SNX family members screened, five exhibited increased EGFR levels. Data from >200 cells per condition. (B) Domain organisation of SNX15, including the clathrin-binding box identified in the present study. (C–E) HeLa cells transfected with control or SNX15 SMARTpool siRNA were stimulated with 100 ng/ml EGF or PDGF for the indicated times prior to immunoblotting with antibodies against EGFR, PDGFR, SNX15, phospho-ERK (42/44), or as a loading control tubulin or actin. (F) A degradation time course of EGFR following EGF stimulation revealing a decreased rate of EGFR degradation in SNX15-suppressed cells versus control. Data are from three independent experiments (means ± s.e.m. P≤0.015).

It is known that SNX15 chronic overexpression alters the gross morphology of several endosomal compartments with resultant effects on multiple endosomal transport pathways including endocytosis and recycling to the plasma membrane and TGN (Barr et al., 2000; Phillips et al., 2001). More recently, genome-wide (Collinet et al., 2010) and targeted (Pedersen et al., 2012) RNAi screens have independently identified SNX15 as a regulator of degradative EGFR sorting, although the details of its role have yet to be described [the latter screen specifically targeted the human proteome for proteins that contain phosphoinositide-binding PX and FYVE domains (Pedersen et al., 2012)]. Human SNX15 is a 342 amino acid protein composed of a central region that includes a proline-rich stretch of three PPPxxP repeats (residues P142 to P181) and an acidic cluster (E255EEEDGE261), flanked by an N-terminal SNX-PX domain (F9 to G126) and a C-terminal microtubule-interacting and trafficking (MIT) domain (A267 to P342; Fig. 1B) (Phillips et al., 2001). An alternative splice variant, SNX15 isoform B, is identical apart from the removal of 86 amino acids between E222 and S308 that include the acidic cluster and the N-terminal 42 residues of the MIT domain (Phillips et al., 2001). The ubiquitous expression profile of this sorting nexin (Phillips et al., 2001) and presence of such diverse protein–protein interaction motifs are highly suggestive of SNX15 functioning as an important scaffold during phosphoinositide-mediated EGFR sorting. We therefore chose to examine its role in more detail.

SNX15 suppression perturbs post-internalised trafficking of EGFR

We confirmed the efficiency of the SNX15-targeted SMARTpool and two individual RNA duplexes that targeted both isoforms of SNX15 (supplementary material Fig. S1). To define the kinetics of EGFR degradation we serum starved SNX15 suppressed HeLa cells and performed western analysis of receptor levels after addition of 10 ng/ml EGF (Fig. 1C,D). Compared with control cells SNX15-suppressed cells showed an approximate twofold elevated steady-state level of EGFR, and a statistically significant (P≤0.015) delay in EGFR degradation that was most evident at early time points after EGF addition (Fig. 1C–F; supplementary material Fig. S1). A similar elevation in steady-state level of PDGFR was also observed (Fig. 1C,D). Analysis of the ERK/MAPK cascade, at the level of p42/p44 ERK, revealed no statistically significant effect on the activation and duration of signalling following EGF addition (Fig. 1E). Finally, FACS analysis revealed that the kinetics of EGFR internalisation was unperturbed upon SNX15 suppression, establishing that the function of SNX15 lay downstream of EGFR endocytosis (Fig. 2A).

Fig. 2.

SNX15 regulates trafficking of internalised EGFR. HeLa cells transfected with scrambled or SNX15 siRNA were stimulated with 100 ng/ml EGF (A,B,D,E) or Alexa-Fluor-488–EGF (C) for the indicated time periods. (A) Flow cytometric analysis of cell-surface EGFR. Values are the mean fluorescence ± s.e.m. of three independent experiments as a percentage of levels in non-stimulated control cells. The raw un-normalised fluorescent data at t = 0 were 54,179±5242 and 63,164±942 fluorescent units for control versus SNX15-suppressed, respectively (n = 3 with >20,000 cells from each condition were examined per n number). In control cells the reproducible increase in cell surface EGFR after 2 minutes of EGF addition was statistically significant (P≤0.015). (B) Cells were fixed and immunostained for EGFR 5 minutes after EGF addition (nuclei visualised with DAPI). Representative image showing peripherally localised EGFR in SNX15-suppressed compared with control cells. (C) Stills from live-cell movies of internalised Alexa-Fluor-488–EGF (supplementary material Movies 1–4), depicting peripheral static vesicles in the SNX15-suppressed cells in contrast to motile EGF-enriched vesicles in control cells. Scale bars: 20 µm. Peripheral Alexa-Fluor-488–EGF-enriched vesicles were manually tracked for 20 seconds using Volocity software. The graph shows the mean values ± s.e.m. from three experiments (10 tracks recorded per experiment). (D,E) Colocalisation of immunostained EGFR with EEA1 or LAMP1 was quantified from confocal images using Volocity software. Scale bars: 10 µm. Graphs show means ± s.e.m. obtained from >30 cells, P-values are from paired t-tests between respective means.

Fig. 2.

SNX15 regulates trafficking of internalised EGFR. HeLa cells transfected with scrambled or SNX15 siRNA were stimulated with 100 ng/ml EGF (A,B,D,E) or Alexa-Fluor-488–EGF (C) for the indicated time periods. (A) Flow cytometric analysis of cell-surface EGFR. Values are the mean fluorescence ± s.e.m. of three independent experiments as a percentage of levels in non-stimulated control cells. The raw un-normalised fluorescent data at t = 0 were 54,179±5242 and 63,164±942 fluorescent units for control versus SNX15-suppressed, respectively (n = 3 with >20,000 cells from each condition were examined per n number). In control cells the reproducible increase in cell surface EGFR after 2 minutes of EGF addition was statistically significant (P≤0.015). (B) Cells were fixed and immunostained for EGFR 5 minutes after EGF addition (nuclei visualised with DAPI). Representative image showing peripherally localised EGFR in SNX15-suppressed compared with control cells. (C) Stills from live-cell movies of internalised Alexa-Fluor-488–EGF (supplementary material Movies 1–4), depicting peripheral static vesicles in the SNX15-suppressed cells in contrast to motile EGF-enriched vesicles in control cells. Scale bars: 20 µm. Peripheral Alexa-Fluor-488–EGF-enriched vesicles were manually tracked for 20 seconds using Volocity software. The graph shows the mean values ± s.e.m. from three experiments (10 tracks recorded per experiment). (D,E) Colocalisation of immunostained EGFR with EEA1 or LAMP1 was quantified from confocal images using Volocity software. Scale bars: 10 µm. Graphs show means ± s.e.m. obtained from >30 cells, P-values are from paired t-tests between respective means.

Movement of newly formed EGFR-containing endocytic vesicles into the endosomal system is perturbed by SNX15 suppression

To examine EGFR trafficking immediately proceeding endocytosis we visualised the endosomal distribution of endogenous EGFR following EGF stimulation. In contrast to control cells, where at early time points the majority of EGFR-labelled puncta are dispersed in the cytosol with enrichment in the juxtanuclear region (Fig. 2B), in SNX15-suppressed cells EGFR was retained in puncta that lay more peripherally dispersed in close proximity to the plasma membrane (Fig. 2B). Moreover, live-cell confocal imaging using Alexa-Fluor-488–EGF also established that in contrast to control cells, where peripheral EGF-labelled puncta were observed to move into the cell interior, the corresponding EGF-positive puncta displayed a reduced level of movement following SNX15 suppression (Fig. 2C; supplementary material Movies 1–4).

To further confirm a defect in the movement of post-internalised EGFR into the endosomal system, additional quantification revealed a delay in arrival of EGFR into EEA1-positive early endosomes and LAMP1-positive late endosomes upon SNX15 suppression, which correlated with the perturbed EGFR degradation observed at these early time points (Fig. 2D,E). Together these data are consistent with SNX15 being required for the movement of newly formed EGFR-containing endocytic vesicles into the early endosomes and hence the endo-lysosomal degradative pathway.

SNX15 resides on the PtdIns3P-enriched early endosome

To define the subcellular localization of SNX15 we initially screened a series of available anti-SNX15 antibodies for their ability to detect endogenous SNX15 for immunofluorescent microscopy. Using various fixation and permeabilisation conditions all antibodies, however, proved unsuccessful (data not shown). Mindful of the fact that chronic overexpression of SNX15 leads to gross morphological changes of several endosomal compartments (Phillips et al., 2001), we established a lentiviral-based transduction protocol that in HeLa cells allowed for the low level expression of a GFP-tagged SNX15 transgene (van Weering et al., 2012a). Under these conditions, the level of GFP–SNX15 expression did not adversely alter endosomal or lysosomal morphology as defined by various markers (Fig. 3A–F). Confirming previous work looking at endogenous SNX15 in COS-7 cells (Barr et al., 2000), GFP–SNX15 was associated with cytosolic puncta that overlapped with markers of the early endosome (EEA1 and Rab5) and early-to-late transition endosome (SNX1; Fig. 3A–C) but not the late endosome/lysosome (CD63 and LAMP1) or TGN (TGN36/48; Fig. 3D–F). Endosome association required the activity of a wortmannin-sensitive PI 3-kinase (Fig. 3I,iii) and, consistent with this, full length SNX15 and the isolated SNX-PX domain bound specifically to PtdIns3P, determined through lipid presentation in the form of fat-blots (PIP strips) or low molar per cent phosphoinositide-supplemented sucrose-loaded liposomes (Fig. 3G,H). Moreover, a site-directed mutant targeting a conserved arginine residue essential for PtdIns3P binding to other PX domains, Arg51, resulted in a cytosolic distribution of SNX15 R51A that correlated with a loss in PtdIns3P binding (Fig. 3G,H,Ii–iii). Finally, as SNX15 contains an MIT domain, a recently suggested Ca2+-dependent phosphoinositide-binding site for PtdIns3P and phosphatidylinositol 4-phosphate (PtdIns4P) (Iwaya et al., 2013), and a known protein–protein interaction motif that for UBPY/USP8 is required for endosomal association (Row et al., 2007), we also generated SNX15-ΔMIT, a deletion mutant lacking the entire MIT domain (residues 265–342). Deletion of this domain did not perturb endosome association, nor was the isolated MIT domain able to associate with cytosolic puncta (Fig. 3I,iv–vi). Overall, SNX15 is associated with the PtdIns3P-enriched early endosome through the classic recognition of this phosphoinositide by its SNX-PX domain.

Fig. 3.

Early endosomal association of SNX15 requires binding to PtdIns3P. (A–F) Representative confocal image stacks of GFP–SNX15 with endogenous endosomal markers or ectopically expressed Rab5. Scale bars: 10 µm. (G) Sucrose-loaded liposomes composed of phosphatidylserine, phosphatidylcholine and phosphatidylethanolamine (each at 26.3% w/w) and enriched for the specific phosphoinositides (20% w/w) were incubated with GST–SNX15, GST–SNX15-PX or GST–SNX15 R51A. SNX15–lipid complexes were pelleted by centrifugation and the supernatants (S) and pellets (P) resolved prior to western blot analysis using anti-GST. (H) Specificity of SNX15 PX domain for PtdIns3P as demonstrated by protein–lipid overlay assay. 100 pmol of relevant lipids were spotted onto a nitrocellulose membrane and incubated with purified SNX15 or SNX15 R51A, and protein–lipid interactions were detected using an anti-GST antibody. (I) HeLa cells were transfected with DNA encoding various SNX15 derivatives. (i) Wild-type SNX15 localises to peripheral and perinuclear puncta, independently of its MIT domain (iv). (ii) SNX15 R51A is cytosolic, as is wild-type protein upon inactivation of PI 3-kinase using 100 nM wortmannin (iii). (v) The isolated SNX15 MIT domain is cytosolic. (vi) The domain organisation of SNX15 and described truncation mutants.

Fig. 3.

Early endosomal association of SNX15 requires binding to PtdIns3P. (A–F) Representative confocal image stacks of GFP–SNX15 with endogenous endosomal markers or ectopically expressed Rab5. Scale bars: 10 µm. (G) Sucrose-loaded liposomes composed of phosphatidylserine, phosphatidylcholine and phosphatidylethanolamine (each at 26.3% w/w) and enriched for the specific phosphoinositides (20% w/w) were incubated with GST–SNX15, GST–SNX15-PX or GST–SNX15 R51A. SNX15–lipid complexes were pelleted by centrifugation and the supernatants (S) and pellets (P) resolved prior to western blot analysis using anti-GST. (H) Specificity of SNX15 PX domain for PtdIns3P as demonstrated by protein–lipid overlay assay. 100 pmol of relevant lipids were spotted onto a nitrocellulose membrane and incubated with purified SNX15 or SNX15 R51A, and protein–lipid interactions were detected using an anti-GST antibody. (I) HeLa cells were transfected with DNA encoding various SNX15 derivatives. (i) Wild-type SNX15 localises to peripheral and perinuclear puncta, independently of its MIT domain (iv). (ii) SNX15 R51A is cytosolic, as is wild-type protein upon inactivation of PI 3-kinase using 100 nM wortmannin (iii). (v) The isolated SNX15 MIT domain is cytosolic. (vi) The domain organisation of SNX15 and described truncation mutants.

SNX15 is also associated with clathrin-coated pits and clathrin-coated vesicles

In addition to the targeting to the PtdIns3P-enriched early endosome, it was evident from fixed cell confocal imaging that a sub-population of SNX15-decorated puncta lay juxtaposted to the plasma membrane (Fig. 4A). Indeed, in serum-starved HeLa cells virally expressing GFP–SNX15 and fixed 0, 3, 5 and 10 minutes after EGF stimulation extensive colocalisation was observed between SNX15 and dispersed clathrin-decorated puncta including those lying in close proximity to the cell surface (Fig. 4A). Confirming this, an ultrastructural analysis revealed that SNX15 was associated with cell surface structures that morphologically appeared as electron-dense late-stage clathrin-decorated endocytic pits, and in addition, to clathrin-coated vesicles that lay juxtaposed to the plasma membrane (Fig. 4B). That these structures were indeed coated with clathrin was confirmed by immuno-EM (Fig. 4B). Quantification of the overall plasma membrane association established a strong enrichment of SNX15 at sites of clathrin-decorated endocytic pits (0.199 gold grain per micrometer of plasma membrane versus 2.75 per micrometer in clathrin-coated pits). SNX15 was not observed to associate with electron dense bud-like profiles at the TGN, which correspond to the clathrin-coated structures required for exit from this compartment, and we failed to observe enrichment of SNX15 on the bilayered clathrin coat on endosomal vacuoles that is involved in targeting proteins for ESCRT-mediated sorting into the endo-lysosomal pathway (Hurley and Hanson, 2010) (data not shown). Confirming this, and consistent with the SNX15 MIT domain forming a distinct branch of the MIT domain evolutionary tree (Row et al., 2007), direct yeast two-hybrid analysis failed to reveal any association between SNX15 and a wide array of Class E pathway proteins (Hurley and Hanson, 2010) (supplementary material Fig. S2).

Fig. 4.

SNX15 decorated puncta colocalise with clathrin. HeLa cells expressing GFP–SNX15 were stimulated with 100 ng/ml EGF for the indicated times. Fixed cells were co-immunostained using antibodies against clathrin and EGFR, and imaged using confocal microscopy. (A) GFP–SNX15 associates with clathrin-coated structures close to the cell membrane (0 minutes) and on intracellular vesicles (3 minutes – yellow arrows); co-labelled structures in the perinuclear region are enriched with EFGR (10 minutes – white arrows). Scale bar: 10 µm. (B) Typical micrograph of clathrin-coated pits and vesicles from HeLa cells expressing GFP–SNX15 (5 nm-gold) and immunolabelled for endogenous clathrin (10 nm-gold). Scale bar: 100 nm. (C) HeLa cells co-expressing GFP–SNX15 and DsRed–CLC were stimulated with 100 ng/ml Alexa-Fluor-647–EGF and imaged live using TIRF microscopy. Frames from 10 minutes after stimulation are shown, where SNX15 and CLC colocalise on EGFR-enriched vesicles closely juxtaposed to the plasma membrane (see supplementary material Movie 5). Scale bar: 20 µm. (D) HeLa cells co-expressing GFP–SNX15 R51A and DsRed–CLC were imaged live using TIRF microscopy. GFP–SNX15 R51A retained the ability to associate with peripherally localised clathrin-enriched puncta. Scale bar: 10 µm.

Fig. 4.

SNX15 decorated puncta colocalise with clathrin. HeLa cells expressing GFP–SNX15 were stimulated with 100 ng/ml EGF for the indicated times. Fixed cells were co-immunostained using antibodies against clathrin and EGFR, and imaged using confocal microscopy. (A) GFP–SNX15 associates with clathrin-coated structures close to the cell membrane (0 minutes) and on intracellular vesicles (3 minutes – yellow arrows); co-labelled structures in the perinuclear region are enriched with EFGR (10 minutes – white arrows). Scale bar: 10 µm. (B) Typical micrograph of clathrin-coated pits and vesicles from HeLa cells expressing GFP–SNX15 (5 nm-gold) and immunolabelled for endogenous clathrin (10 nm-gold). Scale bar: 100 nm. (C) HeLa cells co-expressing GFP–SNX15 and DsRed–CLC were stimulated with 100 ng/ml Alexa-Fluor-647–EGF and imaged live using TIRF microscopy. Frames from 10 minutes after stimulation are shown, where SNX15 and CLC colocalise on EGFR-enriched vesicles closely juxtaposed to the plasma membrane (see supplementary material Movie 5). Scale bar: 20 µm. (D) HeLa cells co-expressing GFP–SNX15 R51A and DsRed–CLC were imaged live using TIRF microscopy. GFP–SNX15 R51A retained the ability to associate with peripherally localised clathrin-enriched puncta. Scale bar: 10 µm.

To examine the relationship between SNX15 and clathrin at regions juxtaposed to the plasma membrane we turned to total internal reflection fluorescence (TIRF) microscopy to limit the excitation field to a penetration zone of ∼100 nm from the plasma membrane–buffer–glass interface. In cells stimulated with Alexa-Fluor-647–EGF, SNX15 was associated with EGFR-positive and clathrin-decorated puncta that either appeared within, or resided juxtaposed to the plasma membrane (Fig. 4C; supplementary material Movie 5). The association of SNX15 to these clathrin-labelled structures was independent of an ability to bind PtdIns3P, since the PtdIns3P-binding mutant SNX15 R51A, which from confocal microscopy is predominantly cytosolic (Fig. 3I,ii), retained the ability to associate with these structures when observed by TIRF microscopy (Fig. 4D).

SNX15 directly binds clathrin through an LFDPF clathrin box

To elucidate the molecular mechanism for the phosphoinositide-independent association of SNX15 with clathrin-decorated elements of the early endocytic network we performed proteomics on SNX15 immuno-isolates. GFP-nanotrap pull-downs of cell extracts derived from HeLa cells lentivirally transduced to express GFP, GFP–SNX15, GFP–SNX15-ΔMIT or GFP–SNX15-MIT, revealed a predominant band at ∼190 kDa that was present in GFP–SNX15 and GFP–SNX15-ΔMIT-expressing samples but absent from GFP–SNX15-MIT and the GFP control (Fig. 5A). Mass spectroscopy identified this protein as clathrin heavy chain 1 (CHC; protein score 737). To verify the interaction, we isolated full-length recombinant SNX15 and the recombinant CHC terminal domain (residues 1–579). This confirmed and established the direct nature of SNX15 binding to CHC (Fig. 5B). Deletion mutagenesis revealed that immuno-isolates of SNX15 and SNX15-ΔMIT retained binding to endogenous CHC, whereas this protein was absence in SNX15-MIT and the GFP control (Fig. 5C). Further analysis identified that deletions of the region encoding residues 1–203 of SNX15 resulted in an inability to associate with CHC (Fig. 5C). As SNX15-ΔMIT encodes residues 1–264, the clathrin-binding domain would appear to reside within residues 203–264 that form the central region of SNX15 between the SNX-PX and MIT domains. To define this site further, bioinformatics highlighted the presence of four potential clathrin boxes in full length SNX15: box 1, D54FRKL58; box 2, E184ALDLL189; box 3, L189FNCE193 and box 4, L214FDPF218 (Fig. 5C). Of these, only box 4 resided within residues 203–264 (Fig. 5C). Consistent with this, deletion mutagenesis of each individual box, or a combination of boxes, established that only removal of box 4 (SNX15Δ4) led to a loss of CHC binding (Fig. 5C). Confirming these data, recombinant SNX15Δ4 was unable to associate with purified CHC (Fig. 5B), and alanine-scanning mutagenesis resulted in the isolation of individual SNX15 mutants that lacked CHC binding (Fig. 5D). Of these, L214A, F215A, D216A or P217A all showed a pronounced loss in CHC binding, whereas F218A retained some level of association (Fig. 5D).

Fig. 5.

SNX15 contains a conserved modular clathrin-binding box. (A) Tagged proteins form HeLa cells transfected with GFP-SNX15, GFP-SNX15-ΔMIT, GFP-SNX15-MIT or GFP (arrowheads) were isolated from cell lysates using GFP-nanotrap. Immuno-isolates were separated by SDS-PAGE and protein visualised by Coomassie Blue staining. Clear bands specific to GFP–SNX15 and GFP–SNX15-ΔMIT are indicated (arrowheads). Arrows represent position of excised band, subsequently determined to be clathrin. (B) GST fused to the clathrin-terminal domain was isolated from BL21 E. coli onto glutathione resin and incubated with purified recombinant SNX15 prior to isolation of pellet and supernatant fractions. SNX15 associates directly with clathrin. As a negative control, SNX15 was incubated with boiled GST–clathrin resin (B. pellet). (C) The efficiency of various GFP-tagged SNX15 mutants to co-immunoprecipitate clathrin heavy chain (CHC) was assessed in HEK-293T cells using western blot analyses. Deletion of clathrin box 4 perturbs CHC binding to SNX15, whereas cumulative or single deletions of clathrin boxes 1–3 recapitulate the binding of full-length SNX15 or SNX15-ΔMIT. (D) Alanine scanning mutagenesis of clathrin box 4 demonstrates the importance of residues 214–217, but not 218 in conferring the ability of SNX15 to associate with clathrin.

Fig. 5.

SNX15 contains a conserved modular clathrin-binding box. (A) Tagged proteins form HeLa cells transfected with GFP-SNX15, GFP-SNX15-ΔMIT, GFP-SNX15-MIT or GFP (arrowheads) were isolated from cell lysates using GFP-nanotrap. Immuno-isolates were separated by SDS-PAGE and protein visualised by Coomassie Blue staining. Clear bands specific to GFP–SNX15 and GFP–SNX15-ΔMIT are indicated (arrowheads). Arrows represent position of excised band, subsequently determined to be clathrin. (B) GST fused to the clathrin-terminal domain was isolated from BL21 E. coli onto glutathione resin and incubated with purified recombinant SNX15 prior to isolation of pellet and supernatant fractions. SNX15 associates directly with clathrin. As a negative control, SNX15 was incubated with boiled GST–clathrin resin (B. pellet). (C) The efficiency of various GFP-tagged SNX15 mutants to co-immunoprecipitate clathrin heavy chain (CHC) was assessed in HEK-293T cells using western blot analyses. Deletion of clathrin box 4 perturbs CHC binding to SNX15, whereas cumulative or single deletions of clathrin boxes 1–3 recapitulate the binding of full-length SNX15 or SNX15-ΔMIT. (D) Alanine scanning mutagenesis of clathrin box 4 demonstrates the importance of residues 214–217, but not 218 in conferring the ability of SNX15 to associate with clathrin.

Clathrin association is required for SNX15 targeting to clathrin-decorated elements of the early endocytic network

To correlate the direct binding of SNX15 to clathrin with its association to clathrin-decorated elements of the early endocytic network, we performed fixed-cell colocalisation studies. Confocal microscopy revealed that SNX15Δ4 and SNX15 F215A, although retaining association to the early endosome (Fig. 6A), had a greatly reduced colocalisation with clathrin compared with wild-type SNX15 (Fig. 6B). Moreover, TIRF imaging revealed that the SNX15Δ4 mutant had a reduced association with clathrin-decorated puncta in regions juxtaposed to the plasma membrane, which was in stark contrast to the corresponding data for SNX15 wild-type (compare Fig. 6C with Fig. 4C). SNX15 therefore employs a bipartite mechanism for association to early elements of the endocytic network: the direct binding of the L214FDPF218 clathrin box to CHC mediates association to clathrin-decorated endocytic pits and clathrin-coated vesicles, whereas recognition of PtdIns3P is necessary for targeting to the PtdIns3P-enriched early endosome.

Fig. 6.

Binding to clathrin is required for SNX15 to associate with clathrin enriched endocytic vesicles. (A) HeLa cells were transiently transfected to express GFP–SNX15, GFP–SNX15Δ4 or the various GFP–SNX15 point mutants prior to immunostaining for endogenous clathrin. (B) Pearson's correlation co-efficient between GFP–SNX15 and SNX15 mutants with clathrin were quantified from digital confocal images using Volocity. 10 cells were examined per condition (>1000 puncta). Error bars indicate ± s.e.m. Both the clathrin box deletion mutant (Δ4) and each of the SNX15 point mutants exhibited a reduced association with clathrin-enriched puncta. (C) When assessed by TIRF microscopy, unlike the wild-type protein, SNX15Δ4 was not associated with clathrin-positive puncta juxtaposed to the plasma membrane. Scale bars: 10 µm.

Fig. 6.

Binding to clathrin is required for SNX15 to associate with clathrin enriched endocytic vesicles. (A) HeLa cells were transiently transfected to express GFP–SNX15, GFP–SNX15Δ4 or the various GFP–SNX15 point mutants prior to immunostaining for endogenous clathrin. (B) Pearson's correlation co-efficient between GFP–SNX15 and SNX15 mutants with clathrin were quantified from digital confocal images using Volocity. 10 cells were examined per condition (>1000 puncta). Error bars indicate ± s.e.m. Both the clathrin box deletion mutant (Δ4) and each of the SNX15 point mutants exhibited a reduced association with clathrin-enriched puncta. (C) When assessed by TIRF microscopy, unlike the wild-type protein, SNX15Δ4 was not associated with clathrin-positive puncta juxtaposed to the plasma membrane. Scale bars: 10 µm.

SNX15 and APPL1 label distinct sub-populations of peripherally dispersed endosomal vesicles

Association with clathrin-coated pits and peripheral clathrin-coated vesicles as well as the early endosome, together with functional effects on endocytic vesicle movement and EGFR degradation, identify SNX15 as a trafficking component of recently endocytosed vesicles. In light of this, we sought to examine the relationship between SNX15 and APPL1, the later defining a transient station common for entry of internalised cargo into the early endosome (Miaczynska et al., 2004; Zoncu et al., 2009). First we performed a fixed-cell colocalisation analysis between GFP–SNX15 and endogenous APPL1 (Fig. 7A), which revealed a limited overlap between the labelled compartments, and low correlation between the overlapping pixels (Mx = 35±2%; PC = 0.15±0.04; where Mx is the degree of overlap between two channels (%) and PC is Pearson’s correlation between overlapping pixels). Inspection of images revealed that these distinct compartments were highly proximal, but did not appear to overlap (supplementary material Fig. S3). This was further confirmed through live-cell imaging where distinct mCherry–APPL1- and GFP–SNX15-labelled compartments were observed to be juxtaposed to one another, from where they underwent brief periods of circumnavigation prior to dissociation (Fig. 7B; supplementary material Movies 6 and 7). In no instances did we observe heterotypic fusion events between these compartments.

Fig. 7.

SNX15 does not associate with the APPL1 intermediate endosome. (A) Representative confocal image of a HeLa cell expressing GFP–SNX15 and co-immunostained for endogenous APPL1. Quantitative colocalisation analysis (n = 14 cells) revealed a partial colocalisation (Mx = 35±2%), with low correlation between the overlapping pixels (PC 0.15±0.02). (B) HeLa cells transfected with GFP-SNX15 and mCherry-APPL1 (supplementary material Movies 6 and 7) were stimulated with EGF and imaged live using confocal microscopy. Sequential images demonstrate a dynamic relationship in which distinctly labelled compartments undergo brief instances of juxtaposition but not compartment mixing. Scale bars: 2 µm. (C) GFP–SNX15-expressing HeLa cells stimulated with EGF were fixed at specified time points and co-immunostained for endogenous EGFR and APPL1. EGFR independently entered both peripherally localised GFP–SNX15 (t = 3 minutes, white arrows) and APPL1-labelled compartments (t = 3 minutes, red arrows) and continued to be concentrated in GFP–SNX15-positive, APPL1-negative compartments that moved towards the cell centre (t = 7 and 10 minutes, white arrows). Scale bars: 10 µm.

Fig. 7.

SNX15 does not associate with the APPL1 intermediate endosome. (A) Representative confocal image of a HeLa cell expressing GFP–SNX15 and co-immunostained for endogenous APPL1. Quantitative colocalisation analysis (n = 14 cells) revealed a partial colocalisation (Mx = 35±2%), with low correlation between the overlapping pixels (PC 0.15±0.02). (B) HeLa cells transfected with GFP-SNX15 and mCherry-APPL1 (supplementary material Movies 6 and 7) were stimulated with EGF and imaged live using confocal microscopy. Sequential images demonstrate a dynamic relationship in which distinctly labelled compartments undergo brief instances of juxtaposition but not compartment mixing. Scale bars: 2 µm. (C) GFP–SNX15-expressing HeLa cells stimulated with EGF were fixed at specified time points and co-immunostained for endogenous EGFR and APPL1. EGFR independently entered both peripherally localised GFP–SNX15 (t = 3 minutes, white arrows) and APPL1-labelled compartments (t = 3 minutes, red arrows) and continued to be concentrated in GFP–SNX15-positive, APPL1-negative compartments that moved towards the cell centre (t = 7 and 10 minutes, white arrows). Scale bars: 10 µm.

To visualise the relationship between internalised EGFR and the distinct SNX15 and APPL1 sub-populations during the early stages of trafficking when the effect of SNX15 siRNA suppression was most pronounced, we stimulated serum-starved HeLa cells expressing GFP–SNX15 with EGF, and co-immunostained for mCherry–APPL1 and EGFR in cells fixed after 0, 3, 7 and 10 minutes of stimulation (Fig. 7C). The EGFR entered spatially distinct, peripherally localised GFP–SNX15- (green arrows, t = 3 minutes) and APPL1-labelled compartments (red arrows, t = 3 minutes). SNX15 remained associated with the EGFR-labelled compartment as they underwent movement towards the cell centre (green arrows, t = 7 and 10 minutes). To extend this, we also performed live-cell wide-field imaging of HeLa cells expressing GFP–SNX15 and mCherry–APPL1 following addition of Alexa-Fluor-647–EGF. This confirmed that Alexa-Fluor-647–EGF entered peripherally distinct GFP–SNX15- and mCherry–APPL1-labelled compartments (Fig. 8A). Furthermore, within the spatial and temporal resolution of our imaging set-up, these compartments did not appear to exchange their contents during the time course of imaging. These data establish that SNX15 defines an entry route for internalised EGFR into the early endosome that is independent of the APPL1-intermediate compartment.

Fig. 8.

EGF-enriched SNX15-labelled endocytic vesicles mature into Rab5-positive, APPL1-negative endosomes in the cell periphery. (A) HeLa cells virally expressing low levels of GFP–SNX15 and transiently transfected with mCherry–APPL1 were starved for 4 hours prior to stimulation with 100 ng/ml Alexa-Fluor-647–EGF and imaged using live cell microscopy. Selected frames, starting at 3 minutes after Alexa-Fluor-647–EGF addition, depict spatially segregated SNX15- and APPL1-positive endosomes co-labelled for Alexa-Fluor-647–EGF. (B) HeLa cells virally expressing low levels of GFP–SNX15 and mCherry–Rab5 were starved for 4 hours prior to stimulation with 100 ng/ml Alexa-Fluor-647–EGF and imaged live using TIRF-microscopy (100 nm penetration depth). Selected frames, starting at 3 minutes after Alexa-Fluor-647–EGF addition, depict a motile SNX15 and Alexa-Fluor-647–EGF-positive but Rab5-negative vesicle as it moves in from the cell periphery and undergoes fusion with a Rab5-labelled early endosome to become a triple labelled SNX15, Alexa-Fluor-647–EGF and Rab5-positive early endosome. Scale bar: 1 µm.

Fig. 8.

EGF-enriched SNX15-labelled endocytic vesicles mature into Rab5-positive, APPL1-negative endosomes in the cell periphery. (A) HeLa cells virally expressing low levels of GFP–SNX15 and transiently transfected with mCherry–APPL1 were starved for 4 hours prior to stimulation with 100 ng/ml Alexa-Fluor-647–EGF and imaged using live cell microscopy. Selected frames, starting at 3 minutes after Alexa-Fluor-647–EGF addition, depict spatially segregated SNX15- and APPL1-positive endosomes co-labelled for Alexa-Fluor-647–EGF. (B) HeLa cells virally expressing low levels of GFP–SNX15 and mCherry–Rab5 were starved for 4 hours prior to stimulation with 100 ng/ml Alexa-Fluor-647–EGF and imaged live using TIRF-microscopy (100 nm penetration depth). Selected frames, starting at 3 minutes after Alexa-Fluor-647–EGF addition, depict a motile SNX15 and Alexa-Fluor-647–EGF-positive but Rab5-negative vesicle as it moves in from the cell periphery and undergoes fusion with a Rab5-labelled early endosome to become a triple labelled SNX15, Alexa-Fluor-647–EGF and Rab5-positive early endosome. Scale bar: 1 µm.

The SNX15 and EGFR-labelled endocytic sub-population fuse directly with Rab5-labelled early endosomes

To visualise more directly the relationship between the EGFR-containing SNX15 endocytic vesicles and the early endosome, we performed live-cell imaging of HeLa cells expressing GFP–SNX15 and mCherry–Rab5, a marker of the classical early endosome. During the initial stages of stimulation with Alexa-Fluor-647-labelled EGF, 3 minutes after stimulation, Alexa-Fluor-647–EGF extensively colocalised with a population of peripheral GFP–SNX15-labelled endocytic vesicles that were negative for Rab5 (Fig. 8B; supplementary material Movie 8). Over time these SNX15-labelled, Alexa-Fluor-647–EGF-containing vesicles were observed to move and fuse with pre-existing Rab5-labelled early endosomes, thereby forming a triple SNX15-, EGF- (and presumably EGFR) and Rab5-labelled endosomal compartment. We did not observe the loss of SNX15 from these EGF-containing endocytic vesicles during these events, consistent with SNX15 remaining associated with newly formed EGFR-containing endocytic vesicles as they transit into the early endosome independently of the APPL1-intermediate compartment.

Discussion

Here we have provided new insight into the organisation of the early endocytic network that lie between clathrin-mediated endocytosis and the PtdIns3P-enriched early endosome (Shin et al., 2005; Erdmann et al., 2007; Zoncu et al., 2009). We have established that SNX15 is targeted to late-stage clathrin-coated pits and clathrin-coated vesicles through an ability to directly bind clathrin by a non-classical clathrin-binding box. Through an association to the canonical early endosome, a result of binding to PtdIns3P, SNX15 constitutes a directional link on the early endocytic pathway between newly formed endocytic vesicles derived from clathrin-mediated endocytosis and the classical PtdIns3P-enriched early endosome. By using an RNAi loss-of-function screen, we identified SNX15 as a positive regulatory of EGFR transport into the endo-lysosomal degradative pathway, and described how this arises through governing movement of newly formed receptor-containing endocytic vesicles into the endo-lysosomal pathway. Underscoring the heterogeneity of clathrin-mediated endocytosis, the described SNX15 pathway does not proceed through an intermediate APPL1 compartment, and hence defines an additional, more direct entry site to the endosomal network from clathrin-mediated endocytosis.

The SNX-directed loss-of-function screen has shown that, under these experimental conditions, only five SNXs significantly regulate the endo-lysosomal sorting of the activated EGFR. The suppression of SNX3 or SNX21 generated relatively minor effects. Although no data are currently available to describe the function of SNX21, SNX3 is a component of the SNX3-retromer, a complex that regulates early endosome-to-TGN retrograde transport of Wntless (Harterink et al., 2011; Zhang et al., 2011). SNX3 has also been implicated in the formation of multivesicular bodies, although, and consistent with the data from the current RNAi screen, SNX3 suppression only marginally slows the rate of EGFR degradation (Pons et al., 2008). The most pronounced effects were observed upon suppression of SNX9, a protein with an established role in endocytosis including that of the EGFR (Lundmark and Carlsson, 2009), and SNX5, classically considered a component of the SNX–BAR-retromer that regulates early-to-late transition endosome-to-TGN retrograde transport (Wassmer et al., 2007; Wassmer et al., 2009; van Weering et al., 2012a). More recently, however, through an association with PIPKIγi5, SNX5 has been shown to regulate EGFR sorting into intraluminal vesicles of the late endosome/multivesicular body in a manner that is independent of the retromer components VPS26 and VPS35 (Sun et al., 2013). Our data, therefore, confirm a non-retromer role for SNX5 in the degradative sorting of the EGFR.

For the remaining member of the group, SNX15, we have established that it uses distinct mechanisms for recognising the identity of early compartments of the endocytic network: the direct binding to clathrin mediates association to clathrin-decorated endocytic pits and clathrin-coated vesicles, whereas recognition of PtdIns3P is sufficient for targeting to the early endosome. Previous functional analysis of SNX15 has documented that its chronic overexpression, estimated as ∼17-fold higher than endogenous (Barr et al., 2000), leads to gross alteration in the morphology of several endosomal compartments and results in multiple transport defects, many of which normally rely on the availability of clathrin for their function (Barr et al., 2000; Phillips et al., 2001). Interestingly, when we chronically overexpressed GFP–SNX15 by means of transient transfection, and selected those cells where the endosomal morphology was altered, we observed the gross redistribution of endogenous clathrin to GFP–SNX15-labelled rings and globular structures (supplementary material Fig. S4). One interpretation of these data are that the previously described functional effects of SNX15 on transferrin and PDGF receptor endocytosis, insulin receptor processing and the recycling of TGN38 and furin, may primarily arise from an indirect effect on the availability and dynamics of clathrin (Barr et al., 2000; Phillips et al., 2001).

The identified clathrin-binding sequence in SNX15, LFDPF, only partially confirms to the canonical clathrin-box motif defined by -[L]-[bulky hydrophobic]-[polar]-[bulky hydrophobic]-[negative]- (Dell'Angelica 2001; Mao et al., 2009). The fifth residue within the SNX15 clathrin box does not play a dominant function in clathrin binding and this is similar to a number of other clathrin boxes, such as the LIDIA box in OCRL (Mao et al., 2009). The proline at position 4, although not classically considered a ‘bulky’ hydrophobic residue, is clearly necessary for clathrin binding to SNX15. Within such clathrin boxes, as far as we are aware, this is a unique amino acid at this position. Interestingly, the LFDPF clathrin box is found in the Rho GEF Vav2 (L93FDPF97), and the periodic tryptophan protein 2 (PWP2H; L735FDPF739). As Vav2 has been implicated as an endocytic regulator for various receptors including that of EGF (e.g. Thalappilly et al., 2010), this protein might also associate with clathrin although the three-dimensional context in which the motif is presented [it is localised within a calponin homology domain (Tybulewicz, 2005)], could be functionally important.

SNX15 joins other sorting nexins that have been reported to associate with clathrin. The disordered region between the SH3 and PX domains of SNX9 directly binds CHC, possibly through non-canonical PWSAW and DWDEDW clathrin boxes, although binding is only observed in truncated and not full length SNX9 (Lundmark and Carlsson, 2003). SNX5 binds a specific clathrin isoform, CHC22, through a binding site located within residues 239–309 of the SNX5 BAR domain (Towler et al., 2004), and an inverted clathrin box, EFELL, present in the SNX-PX domains of SNX1, SNX2, SNX3 and SNX4 has been argued to facilitate clathrin binding (Skånland et al., 2009), although another study has failed to observe these associations (McGough and Cullen, 2013). The equivalent DFRKL (box 1 in SNX15) certainly plays no role in direct binding of SNX15 to CHC.

At the functional level, our study has revealed new insight into the dynamic organisation of endosomal sub-populations at early stages of the endocytic network (Lakadamyali et al., 2006; Erdmann et al., 2007; Zoncu et al., 2009; Swan et al., 2010). Specifically, we have established that suppression of SNX15 expression leads to a pronounced decrease in the dynamics of newly formed endocytic vesicles as they move from the cell periphery and enter the endosomal network. At the level of the EGFR, perturbed endosomal entry leads to a delay in lysosomal-mediated degradation, which is especially evident at early time points after internalisation of the activated receptor. Indeed, the dynamics of early endocytic compartments, in terms of their movement and correct positioning, is known to be crucial for efficient cargo sorting, including the degradative sorting of the EGFR (e.g. Driskell et al., 2007). Precisely how SNX15 regulates early endocytic dynamics remains to be resolved, and ongoing work is seeking to address this point. One can speculate that this may involve coupling to minus-end-directed microtubule motors, especially given the precedent for such associations within the sorting nexin family (Traer et al., 2007; Hong et al., 2009; Skånland et al., 2009; Wassmer et al., 2009). Also, as clathrin regulates a number of recycling and retrograde transport pathways that emerge from the PtdIns3P-enriched early endosome (Johannes and Wunder, 2011; Hsu et al., 2012), it is tempting to speculate that the association between clathrin and SNX15 may also have functional consequences for cargo sorting through these pathways.

Through live-cell imaging we have provided evidence that in the cell periphery SNX15 and APPL1 label distinct sub-populations of early endocytic vesicles. Each population receives internalised EGFR, as defined by accessibility of Alexa-Fluor-647–EGF, but while the two vesicular populations reside in close proximity and display a level of coordinated movement, they do not undergo heterotypic fusion nor, within the spatial and temporal resolution of our imaging system, would it appear that Alexa-Fluor-647–EGF is transferred between them. Importantly, we have observed that the SNX15 and Alexa-Fluor-647–EGF-labelled sub-population can directly fuse with Rab5-labelled early endosomes thereby forming an Alexa-Fluor-647–EGF/EGFR, SNX15 and Rab5-labelled hybrid compartment. Taken together with the mechanistic details of how SNX15 associates with elements of the early endocytic network that extend from clathrin-coated pits and vesicles through to PtdIns3P-enriched early endosomes, we suggest that entry of newly formed EGFR-containing endocytic vesicles into the PtdIns3P-enriched early endosome is achieved through two parallel pathways: maturation through an intermediate compartment defined as clathrin and PtdIns3P negative, Rab5 and APPL1 positive (Zoncu et al., 2009) and an alternative more direct fusion between newly formed SNX15-decorated endocytic vesicles and Rab5-positive early endosome (Rubino et al., 2000). That the APPL1 compartment is considered as an intermediate station for entry of a subset of EGFR-containing endocytic vesicles into early endosomes is entirely consistent with this model (Zoncu et al., 2009). There is clearly a multiplicity of routes through which the internalised EGFR can enter the endosomal network. Further studies employing higher precision imaging will be required to tease apart the complexities and inter-relationships of individual entry routes.

Materials and Methods

Antibodies and other reagents

EGF was from Calbiochem and Alexa Fluor 488 and Alexa-Fluor-647–EGF from Molecular Probes. For immunofluorescence, the following antibodies were used: mouse anti-SNX1 (BD transduction), anti-EEA1 (BD transduction), anti-LAMP1 (Fitzgerald), anti-CD63 (Santa Cruz), anti-EGFR (Abcam), sheep anti-TGN 38/46 (Seroteq), rabbit anti-APPL1 (kind gifts from Pietro De Camilli, Yale University and Philip Woodman, University of Manchester), anti-clathrin (Abcam) and goat anti-EEA1 (Santa Cruz #N-19). Alexa-Fluor-conjugated donkey anti-goat, mouse, rabbit or sheep secondary antibodies were from Molecular Probes. For western blotting mouse anti-GFP (Roche #11814460001), anti-PDGFR β (Calbiochem #Ab-4), anti-α-tubulin (Sigma #T9026), rabbit anti-EGFR (Cell Signaling #2232), anti-phospho-ERK 44/42 (Cell Signaling #9101S) and goat anti-SNX15 (Abcam #ab3990) were used. Mouse anti-biotin (#327906) purchased from Biolegend was used in flow cytometry. HRP anti-goat and anti-rabbit (Jackson Immunoresearch) and HRP anti-mouse (Amersham Biosciences) were used for detection of primary antibodies. LI-COR 680RD/800CW anti-rabbit, -goat and -mouse were used for quantitative detection of primary antibodies.

Tissue culture, transfections and viral expression

HeLa and HEK-293T cells were cultured as previously described (Cozier et al., 2002; Carlton et al., 2004; Wassmer et al., 2007). HeLa cells were transfected with control siRNA (Dharmacon) or siRNA against SNX15 (Dharmacon M-017488) using Hyperfect (Qiagen) according to manufacturers' reverse-transfection protocol to a final concentration of 20 nM. Cells were incubated for 68 hours post-transfection prior to use in assays. Human SNX15 cDNA was cloned into the pEGFP-C1 vector (Clonetech) and mCherry-APPL1 was a kind gift from Professor Martin Lowe (University of Manchester, UK). For viral plasmid transduction, the entire GFP-SNX15 open reading frame from pEGFP-C1-SNX15 was sub-cloned into the XLG3 vector and lentivirus subsequently produced in HEK-293T cells as described previously (van Weering et al., 2012a). HeLa cells were cultured in lentiviral-containing DMEM for 3 days followed by 1 week of standard culturing prior to use. Mutagenesis was performed using the QuikChange site-directed mutagenesis kit (Stratagene).

RNAi loss-of-function screen

To overcome heterogeneity in cell morphology and attachment resulting from treatment with certain SMARTpool siRNAs (Wassmer et al., 2007), screening plates were incubated in 10 µg/ml fibronectin (Sigma) in PBS at 37°C for 1 hour prior to washing twice in PBS. HeLa cells were then seeded onto the fibronectin-coated Screenmate 96-well plates (Matrix) at 3600 cells/well and transfected with 20 nM SMARTpool siRNAs targeting 30 SNX family members. After 68 hours, growth medium was removed and replaced with FCS free D-MEM for 4 hours prior to stimulation with 100 ng/ml EGF for 90 minutes. Cells were then fixed and immunostained using anti-EGFR and an Alexa-Fluor-488-coupled secondary antibody and the nuclear stain, DAPI. Cells were analysed by the Arrayscan II system (Cellomics), capturing epifluorescence images with a 40× objective. To quantify EGFR levels, the total fluorescence emitted from the Alexa Fluor 488 probe was measured using ImageJ and normalised as a function of the number of nuclei present in respective fields, thereby quantifying EGFR levels per cell. Biochemical validation of EGFR degradation kinetics was performed as described previously (Carlton et al., 2004).

Western blotting

Western blots were performed using standard procedures. Detection and quantification was carried out on a Li-Cor Odyssey Infrared scanning system using fluorescently labelled secondary antibodies.

EGF internalisation assay

To analyse the kinetics of EGF receptor internalisation, cells were transfected with SMARTpool SNX15 and control siRNA. After 67 hours medium was replaced with FCS-free DMEM prior to stimulation with a 2 minute pulse of Alexa-Fluor-488–EGF (200 ng/ml) and processed at 2, 5, 10 and 20 minutes after stimulation. After stimulation, culture dishes were placed on ice, washed twice with ice-cold PBS and then detached using 1 ml of ice-cold PBS supplemented with 5 mM EDTA. Cells were pelleted by centrifugation at 14,000 rpm and re-suspended in 40 µl of ice-cold PBS supplemented with 2% FCS and anti-EGFR (50 ng/µl) antibody (Abcam). Following 30 minutes rotation at 4°C, cells were pelleted and resuspended in 40 µl of ice-cold PBS supplemented with 2% FCS and anti-mouse Ig 488 (100 ng/µl) for a further 30 minutes at 4°C. Cells were washed twice in ice-cold PBS + 2% FCS prior to re-suspension in 0.5 ml PBS supplemented with 1% paraformaldehyde (PFA) and 2% FCS and transferred to flow cytometry tubes (Falcon). All measurements were performed on a FACS CANTO II (BD Biosciences) equipped with an argon ion laser. Twenty thousand cells from each sample were analysed. As a control for background signal, the fluorescence intensity from a sub-set of cells treated in the absence of primary antibody was measured and this value was subtracted from subsequent analyses. The fluorescence intensities of cell surface Alexa-Fluor-488–EGF in control and test samples at t0 were normalised to 100%, and the rate of EGF internalisation within each group measured as the percentage loss in fluorescence.

Protein purification and recombinant binding assays

Cultures of E. coli BL21 cells containing either pGEX6P-C1-GST-SNX15, pGEX6P-C1-GST-SNX15Δ4 or GST-clathrin terminal domain (from Dr Steve Royle, University of Warwick, UK) were grown to an OD600 of 0.8, prior to inoculation with 300 µM isopropyl-beta-D-thiogalactoside (Apollo Scientific). Expression proceeded for a further 18 hours at 25°C. Subsequent isolation of recombinant proteins was as described previously (van Weering et al., 2012b). Direct interaction assays between SNX15 and clathrin, and GFP-nanotrap immunoisolation were performed as detailed previously (McGough and Cullen, 2013; Steinberg et al., 2012; Steinberg et al., 2013).

Immuno-electron microscopy

HeLa cells virally transduced to express low-levels of GFP–SNX15 were grown to 80% confluency in a 10 cm dish. Cells were fixed by the addition of 5 ml 4% PFA (w/v), 0.1% (v/v) glutaraldehyde in 0.2 M phosphate buffer (PB) into the culture medium (5 ml) for 20 minutes and then processed for immunogold labelling of thawed cryosections, as described previously (van Weering et al., 2012a).

Immunofluorescence and live cell imaging

HeLa cells were plated on glass coverslips and fixed in 0.1 M PBS supplemented with 4% (w/v) PFA for 20 minutes, prior to quenching in 20 mM glycine for 5 minutes. Cells were permeabilised using 0.1% Triton X-100 (Sigma) for 5 minutes, except for staining of late endosomal markers, when 0.1% (w/v) saponin (Sigma) for 4 minutes was employed. Cells were blocked by incubation at room temperature for 10 minutes in PBS supplemented with 1% BSA (P-BSA), incubated on droplets of 0.1% P-BSA containing primary antibody for 1 hour, prior to incubation on droplets of Alexa Fluor secondary antibody and, where applicable, DAPI reagent diluted in 0.1% P-BSA. All cells were imaged using a Leica SP5 confocal laser scanning microscope with a 63× 1.40–0.60 PL Apo λBL oil immersion objective. Live-cell imaging was routinely performed at 37°C with cells incubated in CO2-independent medium (Gibco). Fluorescent cells were imaged live on a Hamamatsu EMCCD camera and a Perkin Elmer Ultra VIEW ERS 6FE scanning confocal system attached to a Leica DM I6000 inverted epifluorescence microscope, or for TIRF microscopy on a Hamamatsu C9100-13 black thinned EM-CCD camera attached to a Leica DMI 6000 inverted epifluorescence microscope using 100 nm penetration depth of evanescent field. Rapid switching between excitation/emission wavelengths generally allowed a capture rate of 2–5 frames per second.

Image analysis

Colocalisation measurements were performed using Volocity software as described previously (van Weering et al., 2012a).

Acknowledgements

We thank the Medical Research Council and Wolfson Foundation for supporting the Wolfson Bioimaging Facility at the University of Bristol. We thank Dan Fitzgerald and Ross Fuller for their assistance during early stages of this project.

Author contributions

C.D., J.G.C. and P.J.C. conceived the project and designed the experimental strategy. C.D., I.J.M., S.Y. and J.G.C. carried out the cell biological experiments. E.B. and P.V. carried out the immuno-EM study. K.J.M. performed the proteomics. O.J.H. and M.T.M. performed the particle tracking and statistical analysis. C.D., O.J.M., J.M.-S., M.T.M., P.V. and P.J.C. wrote the manuscript.

Funding

This work was supported by the Biotechnology and Biological Sciences Research Council [grant number BB/I011412/1 to P.J.C.]; the Wellcome Trust [grant number 089928/Z/09/Z]; and a Wellcome Trust four-year PhD studentship [grant number 086777/Z/08/Z to I.J.M.]. Deposited in PMC for release after 6 months.

References

Barr
V. A.
,
Phillips
S. A.
,
Taylor
S. I.
,
Haft
C. R.
(
2000
).
Overexpression of a novel sorting nexin, SNX15, affects endosome morphology and protein trafficking.
Traffic
1
,
904
916
.
Carlton
J.
,
Bujny
M.
,
Peter
B. J.
,
Oorschot
V. M.
,
Rutherford
A.
,
Mellor
H.
,
Klumperman
J.
,
McMahon
H. T.
,
Cullen
P. J.
(
2004
).
Sorting nexin-1 mediates tubular endosome-to-TGN transport through coincidence sensing of high- curvature membranes and 3-phosphoinositides.
Curr. Biol.
14
,
1791
1800
.
Choudhury
R.
,
Diao
A.
,
Zhang
F.
,
Eisenberg
E.
,
Saint-Pol
A.
,
Williams
C.
,
Konstantakopoulos
A.
,
Lucocq
J.
,
Johannes
L.
,
Rabouille
C.
 et al. (
2005
).
Lowe syndrome protein OCRL1 interacts with clathrin and regulates protein trafficking between endosomes and the trans-Golgi network.
Mol. Biol. Cell
16
,
3467
3479
.
Collinet
C.
,
Stöter
M.
,
Bradshaw
C. R.
,
Samusik
N.
,
Rink
J. C.
,
Kenski
D.
,
Habermann
B.
,
Buchholz
F.
,
Henschel
R.
,
Mueller
M. S.
 et al. (
2010
).
Systems survey of endocytosis by multiparametric image analysis.
Nature
464
,
243
249
.
Cozier
G. E.
,
Carlton
J.
,
McGregor
A. H.
,
Gleeson
P. A.
,
Teasdale
R. D.
,
Mellor
H.
,
Cullen
P. J.
(
2002
).
The phox homology (PX) domain-dependent, 3-phosphoinositide-mediated association of sorting nexin-1 with an early sorting endosomal compartment is required for its ability to regulate epidermal growth factor receptor degradation.
J. Biol. Chem.
277
,
48730
48736
.
Cullen
P. J.
(
2008
).
Endosomal sorting and signalling: an emerging role for sorting nexins.
Nat. Rev. Mol. Cell Biol.
9
,
574
582
.
Cullen
P. J.
,
Korswagen
H. C.
(
2012
).
Sorting nexins provide diversity for retromer-dependent trafficking events.
Nat. Cell Biol.
14
,
29
37
.
Dell'Angelica
E. C.
(
2001
).
Clathrin-binding proteins: got a motif? Join the network! Trends Cell Biol.
11
,
315
318
.
Driskell
O. J.
,
Mironov
A.
,
Allan
V. J.
,
Woodman
P. G.
(
2007
).
Dynein is required for receptor sorting and the morphogenesis of early endosomes.
Nat. Cell Biol.
9
,
113
120
.
Erdmann
K. S.
,
Mao
Y.
,
McCrea
H. J.
,
Zoncu
R.
,
Lee
S.
,
Paradise
S.
,
Modregger
J.
,
Biemesderfer
D.
,
Toomre
D.
,
De Camilli
P.
(
2007
).
A role of the Lowe syndrome protein OCRL in early steps of the endocytic pathway.
Dev. Cell
13
,
377
390
.
Harterink
M.
,
Port
F.
,
Lorenowicz
M. J.
,
McGough
I. J.
,
Silhankova
M.
,
Betist
M. C.
,
van Weering
J. R.
,
van Heesbeen
R. G.
,
Middelkoop
T. C.
,
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
.
Hong
Z.
,
Yang
Y.
,
Zhang
C.
,
Niu
Y.
,
Li
K.
,
Zhao
X.
,
Liu
J. J.
(
2009
).
The retromer component SNX6 interacts with dynactin p150(Glued) and mediates endosome-to-TGN transport.
Cell Res.
19
,
1334
1349
.
Hsu
V. W.
,
Prekeris
R.
(
2010
).
Transport at the recycling endosome.
Curr. Opin. Cell Biol.
22
,
528
534
.
Hsu
V. W.
,
Bai
M.
,
Li
J.
(
2012
).
Getting active: protein sorting in endocytic recycling.
Nat. Rev. Mol. Cell Biol.
13
,
323
328
.
Huotari
J.
,
Helenius
A.
(
2011
).
Endosome maturation.
EMBO J.
30
,
3481
3500
.
Hurley
J. H.
,
Hanson
P. I.
(
2010
).
Membrane budding and scission by the ESCRT machinery: it's all in the neck.
Nat. Rev. Mol. Cell Biol.
11
,
556
566
.
Iwaya
N.
,
Takasu
H.
,
Goda
N.
,
Shirakawa
M.
,
Tanaka
T.
,
Hamada
D.
,
Hiroaki
H.
(
2013
).
MIT domain of Vps4 is a Ca2+-dependent phosphoinositide-binding domain.
J. Biochem.
153
,
473
481
.
Johannes
L.
,
Wunder
C.
(
2011
).
Retrograde transport: two (or more) roads diverged in an endosomal tree?
Traffic
12
,
956
962
.
Lakadamyali
M.
,
Rust
M. J.
,
Zhuang
X.
(
2006
).
Ligands for clathrin-mediated endocytosis are differentially sorted into distinct populations of early endosomes.
Cell
124
,
997
1009
.
Lundmark
R.
,
Carlsson
S. R.
(
2003
).
Sorting nexin 9 participates in clathrin-mediated endocytosis through interactions with the core components.
J. Biol. Chem.
278
,
46772
46781
.
Lundmark
R.
,
Carlsson
S. R.
(
2009
).
SNX9 – a prelude to vesicle release.
J. Cell Sci.
122
,
5
11
.
Mao
Y.
,
Balkin
D. M.
,
Zoncu
R.
,
Erdmann
K. S.
,
Tomasini
L.
,
Hu
F.
,
Jin
M. M.
,
Hodsdon
M. E.
,
De Camilli
P.
(
2009
).
A PH domain within OCRL bridges clathrin-mediated membrane trafficking to phosphoinositide metabolism.
EMBO J.
28
,
1831
1842
.
McGough
I. J.
,
Cullen
P. J.
(
2013
).
Clathrin is not required for SNX-BAR-retromer-mediated carrier formation.
J. Cell Sci.
126
,
45
52
.
McMahon
H. T.
,
Boucrot
E.
(
2011
).
Molecular mechanism and physiological functions of clathrin-mediated endocytosis.
Nat. Rev. Mol. Cell Biol.
12
,
517
533
.
Miaczynska
M.
,
Christoforidis
S.
,
Giner
A.
,
Shevchenko
A.
,
Uttenweiler-Joseph
S.
,
Habermann
B.
,
Wilm
M.
,
Parton
R. G.
,
Zerial
M.
(
2004
).
APPL proteins link Rab5 to nuclear signal transduction via an endosomal compartment.
Cell
116
,
445
456
.
Noakes
C. J.
,
Lee
G.
,
Lowe
M.
(
2011
).
The PH domain proteins IPIP27A and B link OCRL1 to receptor recycling in the endocytic pathway.
Mol. Biol. Cell
22
,
606
623
.
Pedersen
N. M.
,
Raiborg
C.
,
Brech
A.
,
Skarpen
E.
,
Roxrud
I.
,
Platta
H. W.
,
Liestøl
K.
,
Stenmark
H.
(
2012
).
The PtdIns3P-binding protein Phafin 2 mediates epidermal growth factor receptor degradation by promoting endosome fusion.
Traffic
13
,
1547
1563
.
Phillips
S. A.
,
Barr
V. A.
,
Haft
D. H.
,
Taylor
S. I.
,
Haft
C. R.
(
2001
).
Identification and characterization of SNX15, a novel sorting nexin involved in protein trafficking.
J. Biol. Chem.
276
,
5074
5084
.
Pons
V.
,
Luyet
P. P.
,
Morel
E.
,
Abrami
L.
,
van der Goot
F. G.
,
Parton
R. G.
,
Gruenberg
J.
(
2008
).
Hrs and SNX3 functions in sorting and membrane invagination within multivesicular bodies.
PLoS Biol.
6
,
e214
.
Row
P. E.
,
Liu
H.
,
Hayes
S.
,
Welchman
R.
,
Charalabous
P.
,
Hofmann
K.
,
Clague
M. J.
,
Sanderson
C. M.
,
Urbe
S.
(
2007
).
The MIT domain of UBPY constitutes a CHMP binding and endosomla localisation signal required for efficient EGF receptor degradation.
J. Biol. Chem.
282
,
30929
30937
.
Rubino
M.
,
Miaczynska
M.
,
Lippé
R.
,
Zerial
M.
(
2000
).
Selective membrane recruitment of EEA1 suggests a role in directional transport of clathrin-coated vesicles to early endosomes.
J. Biol. Chem.
275
,
3745
3748
.
Shin
H. W.
,
Hayashi
M.
,
Christoforidis
S.
,
Lacas-Gervais
S.
,
Hoepfner
S.
,
Wenk
M. R.
,
Modregger
J.
,
Uttenweiler-Joseph
S.
,
Wilm
M.
,
Nystuen
A.
 et al. (
2005
).
An enzymatic cascade of Rab5 effectors regulates phosphoinositide turnover in the endocytic pathway.
J. Cell Biol.
170
,
607
618
.
Skånland
S. S.
,
Wälchli
S.
,
Brech
A.
,
Sandvig
K.
(
2009
).
SNX4 in complex with clathrin and dynein: implications for endosome movement.
PLoS ONE
4
,
e5935
.
Steinberg
F.
,
Heesom
K. J.
,
Bass
M. D.
,
Cullen
P. J.
(
2012
).
SNX17 protects integrins from degradation by sorting between lysosomal and recycling pathways.
J. Cell Biol.
197
,
219
230
.
Steinberg
F.
,
Gallon
M.
,
Winfield
M.
,
Thomas
E. C.
,
Bell
A. J.
,
Heesom
K. J.
,
Tavaré
J. M.
,
Cullen
P. J.
(
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
.
Sun
Y.
,
Hedman
A. C.
,
Tan
X.
,
Schill
N. J.
,
Anderson
R. A.
(
2013
).
Endosomal type Iγ PIP 5-kinase controls EGF receptor lysosomal sorting.
Dev. Cell
25
,
144
155
.
Swan
L. E.
,
Tomasini
L.
,
Pirruccello
M.
,
Lunardi
J.
,
De Camilli
P.
(
2010
).
Two closely related endocytic proteins that share a common OCRL-binding motif with APPL1.
Proc. Natl. Acad. Sci. USA
107
,
3511
3516
.
Teasdale
R. D.
,
Collins
B. M.
(
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
.
Thalappilly
S.
,
Soubeyran
P.
,
Iovanna
J. L.
,
Dusetti
N. J.
(
2010
).
VAV2 regulates EGF receptor endocytosis and degradation.
Oncogene
29
,
2528
2539
.
Towler
M. C.
,
Gleeson
P. A.
,
Hoshino
S.
,
Rahkila
P.
,
Manalo
V.
,
Ohkoshi
N.
,
Ordahl
C.
,
Parton
R. G.
,
Brodsky
F. M.
(
2004
).
Clathrin isoform CHC22, a component of neuromuscular and myotendinous junctions, binds sorting nexin 5 and has increased expression during myogenesis and muscle regeneration.
Mol. Biol. Cell
15
,
3181
3195
.
Traer
C. J.
,
Rutherford
A. C.
,
Palmer
K. J.
,
Wassmer
T.
,
Oakley
J.
,
Attar
N.
,
Carlton
J. G.
,
Kremerskothen
J.
,
Stephens
D. J.
,
Cullen
P. J.
(
2007
).
SNX4 coordinates endosomal sorting of TfnR with dynein-mediated transport into the endocytic recycling compartment.
Nat. Cell Biol.
9
,
1370
1380
.
Tybulewicz
V. L.
(
2005
).
Vav-family proteins in T-cell signalling.
Curr. Opin. Immunol.
17
,
267
274
.
van Weering
J. R.
,
Verkade
P.
,
Cullen
P. J.
(
2012a
).
SNX-BAR-mediated endosome tubulation is co-ordinated with endosome maturation.
Traffic
13
,
94
107
.
van Weering
J. R.
,
Sessions
R. B.
,
Traer
C. J.
,
Kloer
D. P.
,
Bhatia
V. K.
,
Stamou
D.
,
Carlsson
S. R.
,
Hurley
J. H.
,
Cullen
P. J.
(
2012b
).
Molecular basis for SNX-BAR-mediated assembly of distinct endosomal sorting tubules.
EMBO J.
31
,
4466
4480
.
Wassmer
T.
,
Attar
N.
,
Bujny
M. V.
,
Oakley
J.
,
Traer
C. J.
,
Cullen
P. J.
(
2007
).
A loss-of-function screen reveals SNX5 and SNX6 as potential components of the mammalian retromer.
J. Cell Sci.
120
,
45
54
.
Wassmer
T.
,
Attar
N.
,
Harterink
M.
,
van Weering
J. R.
,
Traer
C. J.
,
Oakley
J.
,
Goud
B.
,
Stephens
D. J.
,
Verkade
P.
,
Korswagen
H. C.
 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
.
Zeng
W.
,
Yuan
W.
,
Wang
Y.
,
Jiao
W.
,
Zhu
Y.
,
Huang
C.
,
Li
D.
,
Li
Y.
,
Zhu
C.
,
Wu
X.
 et al. (
2002
).
Expression of a novel member of sorting nexin gene family, SNX-L, in human liver development.
Biochem. Biophys. Res. Commun.
299
,
542
548
.
Zhang
P.
,
Wu
Y.
,
Belenkaya
T. Y.
,
Lin
X.
(
2011
).
SNX3 controls Wingless/Wnt secretion through regulating retromer-dependent recycling of Wntless.
Cell Res.
21
,
1677
1690
.
Zoncu
R.
,
Perera
R. M.
,
Balkin
D. M.
,
Pirruccello
M.
,
Toomre
D.
,
De Camilli
P.
(
2009
).
A phosphoinositide switch controls the maturation and signaling properties of APPL endosomes.
Cell
136
,
1110
1121
.

Supplementary information