A loss-of-function screen reveals SNX5 and SNX6 as potential components of the mammalian retromer

In mammalian cells, there are two distinct mannose-6-phosphate receptors (MPRs) - the cation-dependent and cation-independent MPRs (CD-MPR and CI-MPR, respectively) (Ghosh et al., 2003). These receptors function by associating with the mannose-6-phosphate tag on newly synthesised hydrolytic enzymes, thereby assisting their transport from the trans-Golgi network (TGN) to the endosomal compartment. Once delivered, the hydrolases dissociate and proceed on to lysosomes, while the receptors undergo efficient retrieval back to the TGN. The mechanism by which these MPRs undergo this retrieval was, until recently, unclear. Early work identified Rab9-TIP47 and PACS-1-AP1 as candidate protein complexes in the retrieval of the CI-MPR from endosomes to the Golgi (Riederer et al., 1994; Diaz and Pfeffer, 1998; Wan et al., 1998; Crump et al., 2001). More recently however, attention has focused on a distinct protein complex known as the retromer, which appears to play a pivotal, evolutionarily conserved role in endosome-to-TGN recycling of the CI-MPR (reviewed by Seaman, 2005).

In yeast, the retromer modulates the retrieval of the carboxypeptidase Y receptor Vps10p from prevacuolar endosomes back to the late-Golgi (Seaman et al., 1997; Seaman et al., 1998) (reviewed by Seaman, 2005). The retromer itself consists of two subcomplexes. The sorting nexins Vps5p and Vps17p form a membrane-bound coat that is able to associate with a second, distinct subcomplex composed of Vps26p, Vps29p and Vps35p. This subcomplex is cargo-selective as a result of Vps35p associating with the cytosolic region of Vps10p (Nothwehr et al., 2000; Burda et al., 2002). Orthologues of the yeast retromer have been identified in a number of organisms including S. pombe, plants, C. elegans and mammals, suggesting a fundamental and conserved function for this complex (Kurten et al., 1996; Haft et al., 2000; Iwaki et al., 2006; Oliviusson et al., 2006; Prasad and Clark, 2006).

Mammalian orthologues have been identified for all of the yeast proteins with the exception of Vps17p (Haft et al., 2000). Independent studies have established that mammalian VPS35 directly associates with the CI-MPR and that small interfering RNA (siRNA)-mediated suppression of the VPS26-VPS29-VPS35 subcomplex perturbs endosome-to-TGN retrograde transport of this receptor (Arighi et al., 2004; Seaman, 2004). Since the CI-MPR performs the equivalent role to Vps10p in the transport of lysosomal hydrolases, such data argue for an evolutionarily conserved function of the retromer. Consequently, siRNA-mediated suppression of sorting nexin 1 (SNX1) - the mammalian orthologue of Vps5p - also perturbs endosome-to-TGN retrieval of the CI-MPR (Carlton et al., 2004).

The conserved function and similar subunit composition of the retromer suggests the presence of a mammalian orthologue of yeast Vps17p. One potential candidate is sorting nexin 2 (SNX2) (Haft et al., 2000). Supporting this, SNX1 and SNX2 can form homo- and heterodimers in vitro and in vivo (Haft et al., 1998; Kurten et al., 2001) in a manner similar to Vps5p and Vps17p (Seaman and Williams, 2002). Further evidence comes from genetic studies. Mice lacking either SNX1 or SNX2 are viable and fertile, whereas embryos deficient in both arrest in mid-gestation (Schwarz et al., 2002). Moreover, genetic evidence strongly suggests that SNX2 plays a more crucial role than SNX1 in retromer activity during embryonic development (Griffin et al., 2005). Although such data argue for a role of SNX2 in retromer function, other data are inconsistent with this. SNX2 does not appear to associate with VPS35 (Gullapalli et al., 2004), and siRNA-mediated suppression of SNX2 does not significantly perturb endosome-to-TGN retrieval of the CI-MPR in HeLa cells (Carlton et al., 2005a). However, SNX2 might play a more important role in retromer function in other cell lines or tissues.

SNX1 and SNX2 are members of the mammalian sorting nexin family, an emerging group of proteins that have been implicated in modulating cargo sorting within the endocytic pathway (reviewed in Worby and Dixon, 2002; Carlton et al., 2005b). Currently, 30 mammalian sorting nexins have been identified, all of which contain a diagnostic sorting-nexin-phox (PX)-homology domain (Carlton et al., 2005b; Seet and Hong, 2006). In order to examine whether any other members of this family may constitute a functional equivalent of yeast Vps17p, we designed a loss-of-function screen of retromer-mediated endosome-to-TGN retrieval of the CI-MPR. We present here the data from this screen and our subsequent detailed analysis that revealed a potential role for sorting nexin 5 and sorting nexin 6 (SNX5 and SNX6) in retromer-mediated retrograde transport of the CI-MPR.

To define which of the 30 mammalian sorting nexins may be involved in retromer function, we decided to establish an RNAi loss-of-function screen. The screen was based on imaging the steady-state distribution of the CI-MPR because, in HeLa cells, perturbing the function of the retromer induces a clear redistribution of CI-MPR from its normal TGN steady-state distribution into peripheral punctate early endosomes (Arighi et al., 2004; Carlton et al., 2004; Seaman, 2004). Thus, by assessing the steady-state distribution of the CI-MPR one can examine retromer function. To establish the screen, we designed a SMARTpool RNAi library (Dharmacon) targeting each sorting nexin and components of the mammalian retromer. For a given gene product, the corresponding SMARTpool comprised four distinct RNA duplexes, targeting different regions of the gene, and a fifth sample in which the four individual duplexes were pooled together. To visualise the distribution of the CI-MPR we used an ArrayScan II system (Cellomics), which is a 96-well-based, wide-field fluorescence microscopic imaging system that is designed for high-content automated screening.

For the screen we used HeLa cells, because reverse transcriptase (RT)-PCR revealed that transcripts for 29 out of the 30 sorting nexins were detectable (data not shown). As we were unable to detect a clear transcript for SNX28, a limitation of the current screen was that we could not address whether this gene plays a role in retromer function.

In the screen HeLa cells were transfected with SMARTpool RNA duplexes, and after 48 hours they were fixed and stained for endogenous CI-MPR. For each condition, a total of 150-300 cells were imaged using the Arrayscan II system. Cells were visually analysed and arbitrarily divided into two classes as detailed in Materials and Methods.

Analysis of the data from the screen revealed a perturbation in the steady-state distribution of the CI-MPR towards peripheral punctae in HeLa cells transfected with SMARTpool siRNA targeting SNX1 and the other retromer components VPS26A, VPS29 and VPS35 (Fig. 1). Such data are entirely consistent with the published literature, and provide proof of principle that our method of quantification allowed defects in steady-state CI-MPR distribution to be identified. Besides these expected results, the loss-of-function screen also revealed a major effect generated by suppression of SNX5 and SNX6 (Fig. 1). Neither protein has previously been linked to trafficking of the CI-MPR. As their suppression gave clearly the strongest phenotypes, we chose to investigate them in greater detail.

Suppression of SNX27, SNX28 and SNX30 had a relatively mild effect on the steady-state distribution of the CI-MPR. However, given that these proteins show no sequence homology to yeast Vps17p apart from the diagnostic PX domain, we chose in this particular study not to analyse further their potential role in retromer-mediated CI-MPR trafficking.

For the other sorting nexins, no prominent effect on the steady-state distribution of the CI-MPR was observed. In addition, we noted that suppression of SNX15 and SNX18 induced a significant loss in cell number, presumably through a high rate of cell death, as well as pronounced changes in cell morphology and Golgi integrity.

To confirm the level of suppression of SNX5 and SNX6 achieved under these conditions, HeLa cells were treated with each component of the SNX5- and SNX6-specific SMARTpools for 48 hours prior to determining the level of the corresponding mRNA by RT-PCR (Fig. 2). This revealed that for each siRNA of the SMARTpool, a high level of suppression for the corresponding sorting nexin was achieved. Unfortunately, it was not possible to assess the degree of silencing by western blotting, because the anti-SNX5 antibody cross-reacted with other cellular proteins in HeLa cells and the anti-SNX6 antibody did not produce a sufficiently strong signal in western blotting. Nevertheless, it was possible by immunofluorescence to show that the knock-down of SNX6 caused severe depletion of cellular SNX6 (see supplementary material Fig. S3).

As the ArrayScan II system relies on wide-field imaging, we confirmed the effect of SNX5 or SNX6 suppression on the steady-state distribution of the CI-MPR by confocal microscopy (Fig. 3, supplementary material Fig. S2). Compared with control cells, in those cells suppressed for SNX5 and/or SNX6 a clear redistribution of the CI-MPR from its normal steady-state TGN localization to peripheral punctae was observed. Under either condition, colocalization was observed between these CI-MPR-labelled punctae and the early endosome marker EEA1. Interestingly, in those cells suppressed for both SNX5 and SNX6, the CI-MPR redistribution into early endosomes was even more pronounced. It should also be noted that, although the steady-state distribution of TGN marker TGN46 and the cis-Golgi marker GM130 were unperturbed in SNX6 suppressed cells, both markers showed an altered distribution in cells suppressed for SNX5 (Fig. 4). This suggests that suppression of this sorting nexin has effects not only on endosome-to-TGN transport but also the organization of the Golgi complex.

To establish whether the alteration in the steady-state distribution of the CI-MPR resulted from a perturbation in the kinetics of early endosome-to-TGN transport, we performed a series of antibody-uptake experiments. Using HeLaM cells stably expressing a CD8-CI-MPR chimera (Seaman, 2004), we determined the extent of delivery of anti-CD8 antibody to the TGN46 labelled TGN under conditions of either individual or joint suppression of SNX5 and/or SNX6 (Fig. 5A). In control cells, after labelling of CD8-CI-MPR at the cell surface with anti-CD8 antibody, a time course of internalization revealed that the CD8-CI-MPR had been transported through the early endosome and the bulk had reached the TGN46-labelled TGN after an incubation period of approximately 30 minutes (84.5±8.7%). In HeLaM cells suppressed for either SNX5 or SNX6, or jointly for SNX5 and SNX6, the anti-CD8 antibody was able to undergo internalization into the early endosome, but just 42.3±6.8%, 44.0±6.4% and 45.6±11.3%, respectively, of anti-CD8 antibody had reached the TGN after 30 minutes of uptake (Fig. 5B). It should be noted that, although the TGN46-labelled compartment was perturbed in its morphology in SNX5-suppressed or SNX5-SNX6-suppressed cells, we retained the kinetic quantification of CD8-CI-MPR transport as the degree of colocalization between anti-CD8 antibody and the TGN46-labelled structures. Overall, these data are similar to those observed upon SNX1 suppression (Carlton et al., 2005a), and establish a defect in the kinetics of early endosome-to-TGN transport of the CI-MPR in SNX5- and/or SNX6-suppressed cells.

SNX1 and the other known components of the mammalian retromer have previously been shown to be associated with the limiting membrane of the early endosome (Arighi et al., 2004; Carlton et al., 2004; Seaman, 2004). If, as the functional data suggest, SNX5 and SNX6 are associated with retromer-mediated endosome-to-TGN transport, one would predict that they should show a similar localization. Indeed, endogenous SNX6 showed a high degree of colocalization with the early endosomal marker EEA1 and with SNX1 (Fig. 6A,B). Significantly less colocalization was observed with the late endosomal marker LAMP1 (Fig. 6C). Unfortunately, we were unable to study the localization of endogenous SNX5, because our antibody failed to detect this protein by immunofluorescence (data not shown). To overcome this, we transiently transfected HeLa cells with a plasmid encoding GFP-tagged full-length SNX5 or, for comparison, GFP-tagged full-length SNX6, and examined their colocalization with endogenous SNX1. This revealed that GFP-SNX5 and GFP-SNX6 were both associated with cytosolic punctae that showed a partial colocalization with endogenous SNX1 (Fig. 6D,F) - data similar to those obtained when examining endogenous SNX6. But in contrast to GFP-SNX6 (data not shown), GFP-tagged SNX5 showed little colocalization with EEA1.

Previous studies have highlighted that SNX1 is enriched on CI-MPR-labelled tubular elements of the early endosome (Carlton et al., 2004). In light of this, we decided to study the dynamic relationship of SNX5 and SNX6 with SNX1 and performed live cell imaging on HeLa cells co-expressing GFP-SNX5 or GFP-SNX6 and mRFP-SNX1. Under these conditions, both SNX5 and SNX6 colocalised with SNX1 on dynamic tubular and vesicular elements of the early endosome (Fig. 7A,B; Movies 1-5, supplementary material). In some instances, SNX1- and SNX5-positive or SNX1- and SNX6-positive tubular carriers were observed to be formed from the tubular elements of the early endosome. These data add further support to the conclusion that SNX5 and SNX6 may be functional components of the retromer.

The association of SNX1 with the VPS26-VPS29-VPS35 cargo-selective subcomplex of the retromer is achieved through its direct association with VPS35 (Haft et al., 2000). To examine the link between SNX5 and SNX6 with the retromer, we therefore addressed whether either protein was able to associate with VPS35. By using immunoprecipitation, HeLa cells transiently co-transfected with FLAG-tagged VPS35 and probing for endogenous SNX5 or SNX6, we were unable to observe any association between either of the sorting nexins and VPS35 (data not shown). In light of evidence that SNX1 and SNX5 may form a complex through the interaction of their C-terminal BAR domains (Liu et al., 2006) - and therefore may form a subcomplex of the mammalian retromer analogous to the Vps5p-Vps17p subcomplex in yeast - we next transiently transfected HeLa cells with FLAG-tagged chimeras of full-length SNX5 or SNX6, and immunoprecipitated these proteins, examining whether they associated with endogenous SNX1 (Fig. 8A). Under these conditions, we were able to detect significant association of endogenous SNX1 in SNX6, but not SNX5, immunoprecipitates. To extend these data, we also immunoprecipitated endogenous SNX5 and SNX6, probing the resultant precipitates with anti-SNX1 antibodies. This confirmed that endogenous SNX1 immunoprecipitated with endogenous SNX6, but again we failed to detect significant amounts of SNX1 in SNX5 immunoprecipitates (Fig. 8B,C). These data suggest that, at least for SNX6, the ability to associate with SNX1 may allow it to associate with the other retromer components.

The possibility of a stable complex formed by SNX1 and SNX6 prompted us to investigate what effect SNX6 suppression has on the SNX1 level. In Fig. 9A it is shown that RNAi-mediated suppression of SNX6 induces a significant loss in the level of endogenous SNX1. Interestingly, a more pronounced reduction in SNX1 levels was obtained upon suppression of SNX5 or when SNX6 was jointly suppressed with SNX5 (Fig. 9A). To ensure that the loss of SNX1 was not a consequence of SNX5 siRNA or SNX6 siRNA targeting non-specifically SNX1 mRNA, despite the difference in sequences (supplementary material Fig. S1), we confirmed these results using RT-PCR establishing that the level of SNX1 mRNA was not perturbed under the experimental conditions (Fig. 9B).

A growing body of evidence is emerging that supports an evolutionarily conserved role for the retromer in regulating endosome-to-Golgi retrograde transport (Seaman, 2005). The currently held view is that the retromer consists of two distinct subcomplexes. One is composed of sorting nexins - in yeast Vps5p and Vps17p, and in mammals at least SNX1. This acts as a membrane-bound coat which drives the formation of vesicular and tubular-based endosomal carriers (Seaman and Williams, 2002; Carlton et al., 2004). The second subcomplex is composed of Vps35p-Vps29p-Vps26p. Here, Vps35p plays a pivotal role in bridging the formation of endosomal carriers with cargo sorting, through an ability to bind both the CI-MPR cargo and the sorting nexin coat complex (Seaman, 2005).

In this study, we have focused on examining the role of sorting nexins in mammalian retromer function. In particular, we have addressed whether the mammalian retromer possesses a functional equivalent of Vps17p, or whether SNX1 alone fulfils the role of the yeast Vps5p-Vps17p subcomplex (Carlton et al., 2005a). We have described the use of an RNAi-mediated loss-of-function screen to define the role of other mammalian sorting nexins in retromer-mediated endosome-to-TGN retrograde transport of the CI-MPR. This screen revealed that of all the mammalian sorting nexins, SNX5 and SNX6 gave the most pronounced retromer-like phenotype with regard to the steady-state distribution of the CI-MPR. More detailed confocal imaging, together with the analysis of CI-MPR trafficking, established that, like in SNX1-suppressed cells, RNAi-mediated suppression of SNX5 and/or SNX6 reduces the efficiency of endosome-to-TGN retrograde transport of the CI-MPR. Such data suggest that these sorting nexins are either directly or indirectly associated with retromer-mediated endosome-to-TGN transport. SNX6 has previously been suggested to regulate the function of the transforming growth factor β (TGF-β) family of receptors (Parks et al., 2001), the oncogene Pim-1 (Ishibashi et al., 2001) and the translationally controlled tumor protein TCTP (Yoon et al., 2006). SNX5 has been proposed to regulate epidermal growth factor (EGF) receptor degradation (Liu et al., 2006), play a role in macropinosome formation (Kerr et al., 2006) and act as an adaptor for the clathrin isoform CHC22 (Towler et al., 2004). However, this is the first evidence providing a functional link between the two proteins and the retromer complex.

A variety of data argue that SNX6 may be a component of the retromer. First, there is significant colocalization between this sorting nexin and the tubular and vesicular elements of the SNX1-positive endosome (Parks et al., 2001). Second, endogenous SNX1 can probably associate with endogenous SNX6 through their C-terminal BAR domains. Finally, and perhaps most significantly, the RNAi-mediated suppression of SNX6 results in a significant reduction in the level of endogenous SNX1, a loss that appears to arise through a destabilization of the protein rather than transcriptional control. This reduction suggests that, in HeLa cells, endogenous SNX6 coexists with SNX1 in a membrane-bound complex that modulates retromer-mediated endosome-to-TGN retrieval of the CI-MPR.

What evidence exists for SNX5 as a retromer component? From our data, it is clear that RNAi-mediated suppression of SNX5 induces a defect in endosome-to-TGN retrieval of the CI-MPR, and that this also induces a dramatic loss in the levels of SNX1. Such data, together with the colocalization of endogenous SNX1 and ectopically expressed SNX5, argue that, as with SNX6, SNX5 might also act as a component of the retromer. Interestingly, GFP-SNX5 shows little colocalization with EEA1, as reported previously by Merino-Trigo et al. (Merino-Trigo et al., 2004), suggesting that SNX5 and SNX6 reside on slightly different populations of early endosomes, most probably differing in their degree of maturation. However, whether the alteration of CI-MPR distribution observed upon SNX5 suppression arises solely from a disruption of retromer function is, for the following reason, unclear: we have shown that SNX5 suppression has a clear effect on the morphology of the Golgi complex - a phenotype that is not observed upon suppression of SNX6. It is thus possible that the alteration in CI-MPR distribution is a consequence of a more general defect in transport between the endocytic network and the Golgi complex, in addition to the retromer-mediated defect in endosome-to-TGN retrieval caused by the reduction of SNX1 level. It should be noticed that Liu and colleagues have recently shown that, by using co-immunoprecipitation of ectopically expressed proteins and yeast two-hybrid analysis, SNX1 associates with SNX5 (Liu et al., 2006). Under overexpression conditions of SNX5, we were not able to detect an association between endogenous SNX1 and SNX5. However, because a possible physiologically relevant interaction between SNX1 and SNX5 might be short-lasting and dependent on membrane association, it may be difficult to co-immunoprecipitate the two proteins. Additional work will be required to establish whether endogenous SNX5 associates with any established component of the retromer.

In summary, the data presented here are consistent with SNX6 and possibly SNX5 functioning as part of the membrane-bound retromer coat complex. In this context, these two sorting nexins appear to be functionally equivalent to the yeast protein Vps17p. Finally, our data established that, by designing an RNAi loss-of-function screen, one can effectively probe for the involvement of sorting nexins in a given endosomal pathway. By designing other screens using cargoes that define distinct endocytic sorting routes, it will be possible to dissect the role of mammalian sorting nexins in endosomal sorting. In doing so, we will further analyse the role of phosphoinositides in endosomal biology and define some of the molecular machineries that drive cargo sorting within this complex membraneous network.

Mouse monoclonal anti-SNX1, anti-LAMP1 and anti-EEA1 antibodies were purchased from BD Bioscience, goat polyclonal anti-EEA1, anti-SNX5 and anti-SNX6 antibodies were from Santa Cruz, mouse monoclonal anti-Flag and anti-β-actin antibodies were from Sigma, mouse monoclonal anti-CI-MPR and sheep polyclonal anti-TGN46 antibodies were from Serotec, mouse monoclonal anti-CD8 antibody was from Ancell, rabbit polyclonal anti-CI-MPR antibody was a gift from Paul Luzio (University of Cambridge). Mouse monoclonal anti-GM130 antibody was a gift from Jon Lane (University of Bristol). Secondary antibodies for immunofluorescence were purchased from Invitrogen, horseradish peroxidase (HP)-coupled antibodies from Amersham.

Plasmids encoding GFP-tagged SNX5, Flag-tagged SNX5 and SNX6 were kindly provided by Rohan Teasdale (University of Brisbane), and are described elsewhere (Teasdale et al., 2001). The open reading frame of SNX6 was PCR-amplified from the Flag-SNX6 plasmid and cloned into pEGFP-c1 (Clontech) for GFP localization. Mutation E352K in the template plasmid was reverted by site directed mutagenesis to wild type SNX6.

The scrambled- and SNX1-siRNA were described in Carlton et al. (Carlton et al., 2004). SiRNA targeting SNX5 and SNX6 were: SNX5 D1 3′-cuacgaagcccgacuuugauu-5′, SNX5 D2 3′-caaacaaagcucuggauaauu-5′, SNX5 D3 3′-gagcaaagacgucaaguuguu-5′, SNX5 D4 3′-uaacagagcuccuccgauauu-5′, SNX6 D1 3′-gaugaagaccucaaacuuuuu-5′, SNX6 D2 3′-uaaaucagcagauggaguauu-5′, SNX6 D3 3′-caagaagaguugcugcauuuu-5′, SNX6 D4 3′-ccgcggacuuaaagcaauauu-5′. Compare supplementary material Fig. S1 for an alignment of the siRNA sequences SNX5 D1 and SNX6 D2 with their target sequences and SNX1.

SMARTpool siRNA designed against human sorting nexins were designed and purchased from Dharmacon. For transient transfection, HeLa cells were seeded in six-well plates at a density of 1×10⁵ cells per well prior to transfection with 5.9 nM of the relevant siRNA, using HiPerFect (Qiagen) on the following day according to the manufacturer's instructions. In the case of double suppression, 5.9 nM of each of the siRNA was used for transfection. Cells were incubated for a further 48 hours and either processed for immunofluorescence, RT-PCR or western blotting.

Screenmate 96-well plates (Matrix) were coated with collagen I (Sigma) and seeded with 3500 HeLa cells per well. On the next day cells were transfected with SMARTpool siRNA for the 30 sorting nexins, Vps26A, Vps29, Vps35 and a scrambled control at a concentration of 12.7 nM using the HiPerFect reagent (Qiagen). At 48 hours after transfection cells were fixed and immunostained as detailed below. Cells were analysed using the Arrayscan II system (Cellomics), images were acquired with a 40× objective under epifluorescence illumination. Images were analysed by visual inspection and cells arbitrarily divided into two classes: one with compact steady-state localization of the CI-MPR in the TGN, the other one with fragmentation/vesiculation of the Golgi and/or dispersed MPR staining throughout the cytoplasm. More than 150 cells were examined per SMARTpool used, unless otherwise indicated.

Cells were fixed with 4% formaldehyde in PBS and permeabilized using 0.1% Triton X-100 (Sigma) for all staining, except for LAMP1 localization, for which 0.1% Saponin (Sigma) was used for permeabilization. Incubation with the primary antibodies was followed by incubation with fluorescently labelled secondary antibodies. For nuclear staining either DAPI or Hoechst 33342 was used. Cells were imaged using either a TCS-NT, TCS-SP2 or an AOBS-SP2 confocal laser scanning microscope (Leica), equipped with 63× 1.32 or 63× 1.4 oil immersion objectives. Images were analysed using the Leica confocal software.

HeLa cells were co-transfected using 1 μg GFP-SNX5/GFP-SNX6 and 1 μg of mRFP-SNX1 per well in a six-well cell culture dish, using GeneJuice (Novagen) according to the manufacturer's protocol. Live cell imaging was performed using an UltraView MultiUser Confocal Optical Scanner (PerkinElmer Life Sciences) with a 60× 1.42 oil immersion objective.

Total RNA was extracted by using the Tri-reagent (Sigma) according to the manufacturer's instructions. First strand cDNA synthesis from 2 μg of total RNA was performed using oligo-dT (15 bases) priming in combination with MLV reverse transcriptase (Promega) according to the manufacturer's protocol. PCR was carried out using Taq polymerase (Roche), oligonucleotides used (at 200 nM) as primers were: SNX1, 5′-ctgctcctcacactctacctcc-3′ and 5′-acactcatttctcttcctcctc-3′; SNX5, 5′-gcggttcccgagttgctgc-3′ and 5′-gtcatctacctccttaactcc-3′; SNX6, 5′-atgatggaaggcctggacg-3′ and 5′-atgatattccaaaagaaatg-3′; GAPDH, 5′-gaacgggaagctcactggcatg-3′ and 5′-gtccaccaccctgttgctgta-3′. After 30, 35 and 40 cycles samples were taken from the PCR reactions and run on a gel for visualization to ensure that the reaction had not reached saturation.

The uptake experiments were performed as previously described (Carlton et al., 2005a). Samples were analysed using a TCS-NT confocal microscope (Leica), and at least five individual situations per specimen in eight stacks with more than 50 cells were recorded and analysed. Degree of colocalization was measured using Metamorph software (Molecular Devices) by acquiring the area of CD8-and TGN46-staining and quantifying the percentage of the CD8-positive area included in the TGN46-positive area.

HeLa cells were washed with PBS and either lysed using lysis buffer 1 for Flag IP [50 mM TrisCl pH 7.4, 150 mM NaCl, 0.5% Triton X-100, 0.05% sodium azide, containing the Complete Protease Inhibitor Cocktail (Roche) and PMSF (Sigma)] or lysis buffer 2 [20 mM TrisCl pH 7.4, 50 mM NaCl, 0.1% Nonidet P40 (Sigma), containing the Complete Protease Inhibitor Cocktail] for IP of endogenous SNX5 or SNX6 protein. Lysates were cleared from cell debris by centrifugation at 16,000 g for 15 minutes at 4°C. 1-2 μg of antibody was added to the supernatants and the solution was incubated on a shaker for 1-2 hours at 4°C. As control in FLAG IP, a mouse monoclonal anti-CD8 antibody was used (because CD8 is not expressed in HeLa cells). In the IP of endogenous SNX5 or SNX6, goat polyclonal EEA1 antibody was used as a negative control or the primary antibody was omitted. Protein-G-coupled agarose beads (Pierce) equilibrated in lysis buffer were added and the slurry was incubated on a shaker for another 1-2 hours at 4°C. The agarose beads were pelleted by centrifugation at 16,000 g for 1 minute, and supernatant and pellet were collected. The pellet was washed three times with ice-cold PBS (Flag IP) or with washing buffer (20 mM TrisCl pH 7.4, 50 mM NaCl) (SNX5 and SNX6 IP). Antibodies and protein complexes were released from the pellet by incubation in the presence of 2% SDS at 37°C for 15 minutes. Proteins in solution were separated from the agarose beads by centrifugation. The supernatant of the immunoprecipitation and the supernatant of the eluted pellet fraction were both precipitated using the Wessel-Flügge method (Wessel and Flugge, 1984). Samples were resuspended in 1×NuPAGE LDS loading buffer (Invitrogen), heated to 100°C for 10 minutes and subjected to SDS-PAGE and western blotting.

This work was funded by the Wellcome Trust and a Medical Research Council Infrastructure Award (G4500006), which established the School of Medical Sciences Cell Imaging Facility. N.A. is supported by the Biotechnology and Biological Sciences Research Council; M.V.B. is supported by the Department of Biochemistry, University of Bristol and PerkinElmer Life and Analytical Sciences. We thank Mark Jepson and Alan Leard (both University of Bristol) for their assistance and ProXara Biotechnology for the use of the Arrayscan II system. We also thank Jon Lane (University of Bristol) for the anti-GM130 antibody, Paul Luzio (University of Cambridge) for the anti-CI-MPR antibody and Rohan Teasdale (University of Brisbane) for plasmids.