Evidence for a Role of SNX16 in Regulating Traffic between the Early and Later Endosomal Compartments*

Sorting nexins (SNXs) are a growing family of proteins characterized by the presence of a PX domain. The PX domain mediates membrane association by interaction with phosphoinositides. The SNXs are generally believed to participate in membrane trafficking, but information regarding the function of individual proteins is limited. In this report, we describe the major characteristics of one member, SNX16. SNX16 is a novel 343-amino acid protein consisting of a central PX domain followed by a potential coiled-coil domain and a C-terminal region. Like other sorting nexins, SNX16 associates with the membrane via the PX domain which interacts with the phospholipid phosphatidylinositol 3-phosphate. We show via biochemical and cellular studies that SNX16 is distributed in both early and late endosome/lysosome structures. The coiled-coil domain is necessary for localization to the later endosomal structures, as mutant SNX16 lacking this domain was found only in early endosomes. Trafficking of internalized epidermal growth factor was also delayed by this SNX16 mutant, as these cells showed a delay in the segregation of epidermal growth factor in the early endosome for its delivery to later compartments. In addition, the coiled-coil domain is shown here to be important for homo-oligomerization of SNX16. Taken together, these results suggest that SNX16 is a sorting nexin that may function in the trafficking of proteins between the early and late endosomal compartments.

The cellular response to a myriad of external signals is known to be elicited by ligand binding to cognate cell surface receptors which in turn become activated and set in motion a variety of signaling cascades. Regulation of signaling pathways is of vital importance and is achieved at multiple levels. One chief mechanism is by the down-regulation of the activated receptors. This may be achieved by rapid internalization of the activated receptor into the endosomal system. For many of the receptors of nutritional macromolecules, entry into the endosomal system is brief and generally results in their recycling back to the cell surface for re-employment in another round of ligand binding (1). On the other hand, receptors of hormones and growth factors are often retained in the endosomal system and shunted through various intracellular membrane compartments on their way to the lysosome for degradation (2). Traf-ficking through the endosomal system involves the use of membrane-bound intermediates such as vesicles, hence sorting of the correct cargo into the correct vesicle is critical. Many of the proteins involved in this sorting process are hydrophilic peripheral membrane proteins, and their interaction with the membrane is mediated by direct binding to lipids. This is a key event that has recently gained much support. Extensive research has shown that phosphoinositides, particularly the phosphorylated species of phosphatidylinositol (PI), 1 play a role in multiple aspects of membrane trafficking (for reviews see Refs. 3 and 4), and an increasing number of PI-binding motifs with specific lipid binding affinities have been identified. These include the pleckstrin homology domain (5), the FYVE domain (6), the epsin N-terminal homologue domain (7), and the PX domain (8).
In terms of endosomal transport and, in particular, receptor trafficking, the most interesting phosphoinositide is phosphatidylinositol 3-phosphate (PI3-P). Existing predominantly in endosomal membranes, PI3-P is produced by the action of a family of phosphatidylinositol 3-kinases (PI 3-kinase). Three classes of PI 3-kinase are known. Class I PI 3-kinases are regulated by activated receptor tyrosine kinases and G-proteincoupled receptors and can produce PI3-P, PI3,4-P 2 , and PI3,4,5-P 3 (9). Class II PI 3-kinases contain PX and C2 domains in their C-terminal regions and can generate PI3-P and PI3,4-P 2 (9). Finally, class III PI 3-kinases consisting of Vps34 and p150 produce mainly PI3-P and are responsible for providing the bulk of the endosomal PI3-P as both of these proteins are located in the endosomes (10,11). Involvement of the FVYE domain, which binds PI3-P, in membrane trafficking is well established (for review see Ref. 12), with proteins containing the domain, such as EEA1 (13), Rabip4 (14), and Rabenosyn-5 (15), being involved in endosome docking and fusion. Recently, studies with the PX domain proteins SNX3 (16), Vam7 (17), p40 Phox , and p47 Phox (18,19) have found the PX domain to have a preference for PI3-P as well as PI3,4-P 2 . The PX domain is present in a large number of proteins, many of which exist in the GenBank TM data base, and contain either the PX domain alone or with C-terminal coiled-coil motifs (for review see Ref. 20). Collectively, they are now known as the sorting nexins (SNX), in accordance with the naming of the prototypic protein, SNX1 (21). SNX1 was first identified in a yeast two-hybrid screen using the kinase domain of the epidermal growth factor (EGF) receptor as bait. The C-terminal region of SNX1 following the PX domain contains three coiled-coil domains, the last of which is responsible for binding to the lysosomal targeting information of the EGF receptor (21). Overexpression of SNX1 has been shown to enhance the degradation of the EGF receptor (21). In addition, SNX1 and the related proteins SNX2 and SNX4 can co-immunoprecipitate not only the EGF receptor but also other receptor tyrosine kinases, such as receptors for insulin-and platelet-derived growth factor and the receptor for transferrin (22). With the characterization of SNX6, a sorting nexin identified through its interaction with the transforming growth factor-␤ family of receptors (23), sorting nexins were also found to be important for trafficking of these types of receptors. SNX1, -2, and -4 were also found to bind to certain members of this receptor family. This is not surprising as SNX1, -2, -4, and -6 have the ability to interact with themselves and with each other (22,24). All of these sorting nexins exhibit predominantly early endosomal localization and show high degrees of overlap with one another in immunofluorescence studies. Recently, SNX1 was also found to interact with the protease-activated receptor 1, a G-protein-coupled receptor (24). Overexpression of the C-terminal region of SNX1 inhibited the degradation of the protease-activated receptor 1 receptor (24). Another sorting nexin, SNX3, consisting of little more than a PX domain is structurally different from the sorting nexins 1, 2, 4, and 6 mentioned previously and is not able to interact with other members of the sorting nexin family or exhibit their receptor binding capabilities (22). However, SNX3 has been implicated in the recycling of the transferrin receptor and is localized to structures associated with the early and the recycling endosome (16). Therefore, it would appear that sorting nexins are intimately involved in the trafficking of many different families of receptors, regardless of whether they are receptors that are recycled to the cell surface or those sent to the lysosome for degradation.
In this paper we report our biochemical and cellular characterization of SNX16. This putative member of the SNX family was identified as part of a data base search using the PX domain sequence of SNX1. We described the ability of SNX16 to bind PI3-P and the importance of this interaction in mediating its endosomal recruitment. We also show that SNX16, unlike other characterized sorting nexins, is distributed between the early endosome and the later endosome/lysosome compartments. This cellular distribution is dependent on the presence of a single coiled-coil domain, which is also important for homo-oligomer formation and the normal segregation of a receptor tyrosine kinase from the early endosome on its way to the later structures of the pathway. Consistent with its unique distribution in the endocytic pathway, SNX16 does not associate with the SNX1 complex nor with SNX3. SNX16 is the first sorting nexin shown to have such a dual distribution, and the results suggest that it may regulate trafficking between the early and late endosome.

EXPERIMENTAL PROCEDURES
Cloning of SNX16 cDNA-By using the protein sequence of the PX domain of SNX1, we searched the NCBI data base using the BLAST algorithm (25) and identified two homologous human expressed sequence tags (ESTs). The two ESTs (GenBank TM accession numbers AA280333 and AA329466) were purchased from Research Genetics (Huntsville, AL), and plasmid DNA was isolated using miniprep DNA isolation products from Clontech (Palo Alto, CA). The full sequence of these two ESTs was determined, and the 5Ј start site was not found. PCR was employed to amplify further the 5Ј sequence from a 5Ј-rapid amplification of cDNA ends human liver cDNA library, and sequencing of the amplified products identified the start codon. Standard molecular biology techniques were used to construct full-length cDNA. The identity of the full-length clone was confirmed by sequencing, and it contains a 1035-bp coding region, encoding a 343-amino acid residue sorting nexin, now designated SNX16.
Expression Constructs-PCR using VENT polymerase (New England Biolabs) was used to introduce a 5Ј-XhoI site and a 3Ј-XmaI site at the beginning and ending of the SNX16 coding region. The full-length cDNA as well as truncated cDNAs were cloned via these restriction enzyme sites into pDMycneo, which is a modified version of the pCIneo vector (Stratagene). pDMycneo has an insert encoding two Myc epitope sequences inserted 5Ј to the multiple cloning site (16,26). All of the deletion mutants were generated by PCR, and the Y145A point mutant was made by direct mutation of the codon using PCR. Full-length SNX16 was also cloned into pDHAneo, which has the two Myc epitope tags of pDMycneo replaced with two hemagglutinin (HA) sequence tags, as mentioned above. For expression of full-length SNX16 and selected mutants in bacteria as GST fusion proteins, the cDNA was excised from pDMycneo and cloned into the corresponding sites of pGEX4T-1 (Amersham Biosciences). All constructs were confirmed by sequencing.
Antibodies-SNX16 was produced as a GST fusion protein in bacteria and then cleaved with thrombin. The released SNX16 was used to immunize rabbits with Freund's adjuvant (Invitrogen). To affinitypurify antibodies against SNX16, serum from the immunized rabbits was incubated with the antigen, GST-SNX16, chemically coupled to cyanogen bromide-Sepharose (Amersham Biosciences). Bound antibody was eluted with Immuno-Pure IgG elution buffer (Pierce), neutralized with phosphate-buffered saline (PBS), pH 7.4, and then dialyzed against the same solution. The goat antibody against GST was obtained from Amersham Biosciences. Monoclonal antibodies against EEA1, SNX1, and SNX2 were purchased from Transduction Laboratories. The mouse anti-LAMP1 monoclonal antibody (H4A3), developed by J. T. August and J.E.K. Hildreth, was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. Polyclonal anti-Myc antibodies were from Upstate Biotechnology, Inc., and monoclonal antibody against HA was purchased from Roche Applied Science. The secondary fluorescein isothiocyanate (FITC)-, Alexa Fluor 555-, or Alexa Fluor 647-conjugated anti-mouse or anti-rabbit antibodies were from Jackson Immuno-Research and Molecular Probes, respectively.
Tissue Distribution of SNX16 mRNA-Specific primers representing the start and end of the human SNX16 coding region were used for PCR using first-strand cDNA generated from the indicated human tissues obtained from Clontech. PCR conditions were as instructed by the manufacturer, and primers for human glyceraldehyde-3-phosphate dehydrogenase were used as control. PCR products were analyzed on 1.2% TAE-agarose gels.
Tissue Distribution of Rat SNX16 -Lysates from the indicated rat tissues were prepared by homogenization of 1 g of the appropriate organ in 4 ml of homogenization buffer (25 mM Hepes, pH 7.3, 50 mM sucrose, 10 mM KCl, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, 4 g/ml aprotinin, 1 g/ml pepstatin A) with 25 strokes of a drill fitted Teflon glass homogenizer. The homogenates were centrifuged at 1000 ϫ g for 10 min to yield the post-nuclear supernatant which was resolved on 12% SDS-PAGE prior to transfer to nitrocellulose membrane for immunoblotting.
Protein-Lipid Overlay Assay-Protein-lipid overlay assays were based on the procedure described by Dowler et al. (27). Nitrocellulose membranes spotted with 100 pmol of the indicated lipids were obtained from Echelon Research Laboratories and blocked with 3% bovine serum albumin (BSA) in 10 mM Tris/HCl, pH 8.0, 150 mM NaCl, and 0.1% (v/v) Tween 20 (Tris-buffered saline) for 4 h. Membranes were incubated overnight at 4°C with 1 g/ml of the relevant GST fusion protein in the blocking solution, after which the membranes were washed 12 times for 5 min with Tris-buffered saline. For detection of bound protein, the membranes were incubated with goat anti-GST polyclonal antibody, washed 6 times for 5 min with Tris-buffered saline, then incubated with rabbit anti-goat antibody conjugated with horseradish peroxidase, and finally washed again prior to being visualized using SuperSignal West Pico chemiluminescence substrate (Pierce).
Cell Culture and Transient Transfection-A431 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (v/v, Hyclone Laboratories) and antibiotic/antimycotic (Invitrogen) and grown in a 5% CO 2 incubator at 37°C. Transfection was carried out using EFFECTENE (Qiagen) for ϳ5 h and processed for immunofluorescence, Western blot analysis, or immunoprecipitation 24 h after transfection.
Sucrose Density Gradient-A431 cells from four 10-cm plates were harvested by scraping into 500 l of 20 mM Hepes, pH 7.3, 1 mM EDTA, and 8% (w/v) sucrose and lysed by 20 passes through a 26-gauge needle. Post-nuclear supernatant was obtained by centrifugation of the lysate at 1000 ϫ g for 10 min, and 360 l of the post-nuclear supernatant was loaded on the top of a tube containing 360-l aliquots of 40,37,34,31,28,25,22,19,16,13, and 10% (w/v) sucrose in 20 mM Hepes, pH 7.3, 1 mM EDTA, sequentially overlaid. Following centrifugation at 55,000 rpm in an SW60Ti rotor, 250 l were collected from the bottom of the tube, resolved by SDS-PAGE, transferred to nitrocellulose, and then probed with the indicated antibodies as described below.
Cell Imaging-A431 cells were cultured onto 18 ϫ 18-mm glass coverslips and transiently transfected with the indicated plasmids. All cells were washed twice with PBSCM (PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 ) and fixed with 3% paraformaldehyde in PBSCM for 30 min at 4°C, followed by two rinses with 50 mM NH 4 Cl in PBSCM. Cells were permeabilized with 0.1% (w/v) saponin in PBSCM containing 5% BSA for 30 min at room temperature prior to incubation with the indicated primary antibodies. All primary and secondary antibody incubations were performed in fluorescence dilution buffer (FDB: 5% fetal calf serum, 5% goat serum, and 2% BSA in PBSCM). After mounting the cells were viewed using a laser scanning confocal microscope (Zeiss).
Internalization of EGF-A431 cells were transfected with the indicated plasmids as described above and plated at 1:5 dilution on coverslips the following day. Cells were then serum-starved (Dulbecco's modified Eagle's medium supplemented with 20 mM Hepes, pH 7.3, and 0.2% BSA) 24 h after plating for 3 h. To label cell surface EGF receptors, cells were incubated with 1.6 g/ml EGF conjugated with rhodamine (EGF-rh) for 1 h at 4°C followed by several washes with starvation media to remove unbound EGF-rh. Surface-bound EGF-rh was internalized by incubation with normal growth media at 37°C for the respective periods followed by fixation. Following permeabilization for 5 min with 0.1% (w/v) Triton X-100, cells were processed for triple labeling and analyzed by confocal microscopy as described above.
Immunoprecipitation and Western Blot-A431 cells on 100-mm dishes were transiently transfected with the indicated plasmids and then lysed in 500 l of lysis buffer (25 mM Tris/HCl, pH 7.4, 150 mM NaCl, 1.0% Triton X-100, and Complete EDTA-free protease inhibitor mixture (Roche Diagnostics)). After dilution to 0.5% Triton X-100, immunoprecipitation was performed at 4°C with 25 l of anti-Myc crosslinked to protein A-Sepharose for 4 h. The Sepharose was then washed three times with lysis buffer containing 0.5% Triton X-100, and bound proteins were eluted with SDS-PAGE sample buffer. Immunoprecipitates were resolved by SDS-PAGE and transferred to nitrocellulose for immunodetection. The blots were blocked for at least 1 h with 5% skimmed milk and 20% fetal calf serum in PBST (PBS containing 0.05% Tween 20) and incubated with primary antibody overnight at 4°C. Following several washes with PBST, the blots were incubated with the appropriate secondary antibody conjugated with horseradish peroxidase (Jackson ImmunoResearch) and washed again with PBST prior to detection of antibody complexes with SuperSignal West Pico Chemiluminescence Substrate.

RESULTS
Identification and Tissue Distribution of SNX16 -To identify additional members of the sorting nexin family, we performed a Blast search of the NCBI human EST data base for proteins homologous/identical to the protein sequence of the SNX1 PX domain. A large number of ESTs were identified, two of which were homologous to each other and represent a partial clone of a novel SNX. A rapid amplification of cDNA ends library was used to PCR-amplify the missing 5Ј region. Assembling the additional 5Ј nucleotide sequence with the EST sequence results in the identification of a complete open reading frame of 1035 bp which we have deposited into GenBank TM as SNX16 (GenBank TM accession number AF305779). SNX16 is a 343-residue hydrophilic protein, which contains a central PX domain (residue 101-213) followed by a single putative coiledcoil domain (residue 230 -278) and non-structural C-terminal region (residue 279 -343) (Fig. 1A). Comparison of the human, mouse (GenBank TM accession number NM_029068), and rat (GenBank TM accession number AF305780) SNX16 protein sequences ( Fig. 1B) showed that whereas SNX16 is highly conserved in the middle region of the protein, containing both the PX domain and the coiled-coil domain, both the N-terminal and especially the C-terminal regions show substantial difference, with many of the acidic residues present in the C-terminal region of the human protein absent in the rodent counterparts. Analysis of the human genome has shown 3 transcript variants of SNX16, variant 1 (GenBank TM accession number NM_022133) and 2 (GenBank TM accession number NM_152836) differ in their 5Ј-untranslated region and gives rise to isoform A, simply referred to as SNX16 here, whereas variant 3 (GenBank TM accession number NM_152837) results in a 29-residue deletion of part of the PX domain producing isoform B (Fig. 1B). SNX16 showed little homology to other proteins in regions outside the PX domain. The PX domain showed the highest homology (Fig. 1C) to the PX domains of the cytokine-independent survival kinase (CISK: GenBank TM accession number AF312007 (28)), the uncharacterized PXK protein which contains a kinase-like domain in addition to the PX domain (GenBank TM accession number AF399753), SNX1 (GenBank TM accession number U53225 (21)), and the yeast sorting nexin Mvp1p (GenBank TM accession number U16137 (29)).
To obtain the full-length SNX16 cDNA, we performed high fidelity PCR on a human liver cDNA library. We obtained a PCR product ϳ1050 bp in length, in agreement with the expected size of the SNX16 coding region. The identity of the PCR product was confirmed by sequencing. This 1050-bp fragment is present in diverse tissues, as first strand cDNA generated from eight human tissues subjected to PCR using the same primers employed for the initial cloning of SNX16, and revealed the presence of an ϳ1050-bp band present in most of the tissues tested ( Fig. 2A). The highest level of SNX16 was detected in pancreas, lung, liver, placenta, and heart, with little SNX16 detected in the brain and virtually no SNX16 present in skeletal muscle or kidney under these experimental conditions.
In order to study endogenous SNX16, we prepared specific antibodies by immunizing rabbits with recombinant full-length human SNX16 produced in Escherichia coli. The antibodies were affinity-purified and used to probe an immunoblot of cell lysates from A431 cells or A431 cells expressing Myc-SNX16. The antibodies identified a single protein species migrating at ϳ49 kDa in A431 cells (Fig. 2B, lane 3) as well as the Myctagged SNX16 protein which migrates at ϳ53 kDa (data not shown). SNX16 has a calculated molecular mass of ϳ39 kDa, and the apparent larger size can be explained by the high number of acidic residues present in the C-terminal region of human SNX16. Pre-treatment of the affinity-purified antibody with SNX16 covalently attached to Sepharose completely abolished the detection of this ϳ49-kDa protein and Myc-SNX16 (Data not shown), indicating that the affinity-purified antibodies are specific for SNX16.
As mentioned previously, the protein sequence of rodent SNX16 is substantially different to that of the human protein in the N-terminal and C-terminal regions, and this is reflected in the apparent molecular size of the protein from different species. When the in vitro translated protein products of both the human and rat clones were illuminated by the SNX16 antibodies (Fig. 2B, lanes 1 and 2), human SNX16 migrated at the size observed for SNX16 in A431 cells (49 kDa), whereas rat SNX16 migrated closer to its calculated molecular mass at 40 kDa, which is expected given the absence of acidic residues in the C-terminal region as compared with human protein (Fig.  1B). To verify the authenticity of the size difference between the human and rodent proteins, an immunoblot of lysates from various tissue culture cells was prepared and probed with the SNX16 antibodies (Fig. 2B, lanes 3-9). As can be seen lysates from cells of human A431, HeLa, and 293T, or monkey COS7 origin yielded a protein of ϳ49 kDa, whereas those of rodent origin AtT20 (mouse), NRK, and L2 (rat) were characterized by the presence of the ϳ40-kDa protein confirming the results seen with the in vitro translated proteins. Immunoblot analysis of various rat tissue samples revealed a restricted distribution of SNX16 (Fig. 2B, lanes 10 -17). Although high levels of protein were detected in skeletal muscle, substantial amounts of SNX16 were also identified in the heart and brain lysates. Interestingly, liver, kidney, and spleen lysates did not show a protein the size of SNX16 but rather revealed the presence of a lower molecular weight band (ϳ36 kDa), also present in skeletal muscle, which may correspond to isoform B mentioned previously.
SNX16 Binds to PI3-P in Protein-Lipid Overlay Assays-Given the recent findings by our laboratory and others (16 -19)  Red shading highlights non-conservative changes between the human and rodent sequences; yellow shading indicates conservative changes; and blue shading highlights the region of SNX16 deleted in isoform B. The number of acidic residues contained in the C terminus of SNX16 is substantially different from the human and rodent sequences, and these are highlighted as follows: #, a change in both rodent sequences; /, a change in only the mouse sequence; /, a change in only the rat sequence. C, alignment of the PX domains from the four closest relatives of SNX16 as determined by a blast search of the GenBank TM data base. Red shading indicates similarity in all five sequences; blue shading highlights similarity in at least four of the sequences, and yellow shading highlights similarity in at least two of the sequences. that the PX domain is capable of binding to phosphatidylinositides (PI), we examined whether SNX16 was capable of binding to any of these phosphorylated lipids. The protein lipid overlay assay (27) involves probing strips of nitrocellulose spotted with various lipids with potential target molecules. This method has become increasingly popular for assessing the lipid binding capability of proteins (30 -33) due to the ease of the method and the availability of commercial strips. As SNX3 has been shown previously to bind PI3-P (16), we first tested for the ability of this protein to bind to PIP strips. As shown in Fig. 3, recombinant SNX3, which is structurally little more than a PX domain, bound exclusively to PI3-P in agreement with our previous findings using the plate binding assay, thereby validating the accuracy of this assay method. We next tested for the ability of SNX16 to bind to the phospholipids. We found that full-length recombinant SNX16 specifically recognized PI3-P. No detectable binding was seen for PI, for PIs phosphorylated at other positions (PI4-P and PI5-P), for the bis-phosphorylated PIs (PI3,4-P 2 , PI3,5-P 2 , and PI4,5-P 2 ) or for the tris-phosphorylated PI (PI3,4,5-P 3 ). Similarly no binding was found for the other lipids tested: lysophosphatidic acid, lysophosphatidylcholine, phosphatidylethanolamine, phosphatidylcholine, sphingosine 1-phosphate, phosphatidic acid, and phosphatidylserine.
Mutation of a conserved tyrosine residue at position 145 to alanine, shown previously (16) to be required for the lipid binding capability of the PX domain of SNX3, abolished the lipid binding of SNX16 confirming that the PX domain was responsible for the lipid binding activity (data not shown). Interestingly, when the PX domain alone of SNX16 was used in this assay, beside the binding to PI3-P, a small amount of binding was also seen for PI3,4-P 2 (Fig. 3B). Deletion of the putative coiled-coil domain does not affect PI3-P binding (Fig. 3B).
SNX16 Localizes to Early Endosomes and the Late Endosome/Lysosome-The SNX16 antibodies discussed earlier were used for immunofluorescence but failed to show any staining, indicating that the epitopes recognized by the antibody are not accessible under the conditions of this assay. Therefore, to study the cellular distribution of endogenous SNX16, we fractionated cells according to compartment density on a 10 -40% sucrose gradient (Fig. 4), a strategy used by Stockinger and co-workers (34) to examine the cellular distribution of SNX17. Early endosomal EEA1, SNX1, and SNX2 as well as late endosomal/lysosomal LAMP1 were used as markers of the respective compartments. SNX16 was found mainly in fractions of lower density (fractions 4 -7), which is similar to the profile of the early endosome marker EEA1 (fractions 5-7), and the early endosome localized sorting nexins SNX1 (fractions 5-8) and SNX2 (fractions 4 -8). The marker for the late endocytic compartment, LAMP1, eluted out in heavier fractions with a profile largely distinct from EEA1. Some SNX16 was also found in these heavy fractions (see below), as was a portion of SNX2. These results suggest that the majority of SNX16 resides in fractions that have similar density endosomal membranes marked by EEA1. In addition, SNX16 may also be present in late endosomes/lysosomes as suggested by its presence in the heavier fractions marked by LAMP1.
As mentioned above, immunofluorescence studies of SNX16 could not be performed; therefore, to examine the morphologi-FIG. 2. SNX16 is present in multiple tissues. A, tissue distribution of human SNX16 transcripts as revealed by PCR (upper panel) using its specific oligonucleotides and first strand cDNA synthesized from the indicated tissues, a product of 1035 bp is expected. PCR of glyceraldehyde-3-phosphate dehydrogenase (G3PDH) (lower panel) was used as a control. B, antibodies produced against human SNX16 in rabbits were affinitypurified and used at 0.1 g/ml to probe immunoblots containing human (lane 1) and rat (lane 2) SNX16 produced by in vitro translation (IVT); 70 g of lysates prepared from the indicated tissue cultured cells of human (lanes 3, 4, and 6), monkey (lane 7), mouse (lane 5), and rat (lanes 8 and 9) origin; and 70 g of lysates prepared from the indicated rat tissues (lanes 10 -17). The antibodies detect a band of ϳ49 kDa in samples of human origin and 40 kDa in samples of rodent origin, which is expected given the difference in acidic residues highlighted in the C-terminal region of SNX16 (Fig. 2). The lower (ϳ35 kDa) band observed in lanes 12-15 may represent SNX16 isoform B, which contains a deletion in the PX domain. The immunoblots were also probed with an antibody against ␤-tubulin as a loading control. cal distribution of SNX16, we decided to perform immunofluorescence investigation on cells expressing Myc-tagged SNX16. A431 cells expressing Myc-SNX16 were fixed, permeabilized, and labeled with polyclonal anti-Myc and monoclonal anti-EEA1 antibodies. As shown in Fig. 5A (a-d), at relatively low expression levels, Myc-tagged SNX16 was present in discrete punctate structures, which reside throughout the cell. SNX16containing structures in the perinuclear region also contained EEA1. However, the peripheral puncta showed very little colocalization with EEA1 (Fig. 5A, d). Cells expressing higher levels of Myc-SNX16 showed some expansion/clustering of the EEA1-containing punctate structures with greater perinuclear localization. At high levels of expression, staining for both Myc-SNX16 and EEA1 was mainly present in the perinuclear region with reduced levels of discrete puncta throughout the rest of the cell (Fig. 5A, e-h). This may represent aggregation of the endosome system due to high level overexpression of SNX16.
Previous studies with SNX1, -2, -4, and -6 have shown that these sorting nexins may be members of a larger SNX complex as they can hetero-oligomerize and have been shown to colocalize with each other in the early endosome. Our result above using the sucrose gradient confirmed these findings as SNX1 and SNX2 show similar elution profiles. As SNX16 showed a similar elution profile to SNX2 on the sucrose gradient, we next sought to determine whether they are co-localizing in endocytic vesicles by immunofluorescence. As can be seen in Fig. 7 (a-d), SNX16 and SNX2 did show some overlap in the perinuclear region. However, similar to that observed with EEA1, the majority of the SNX16 peripheral puncta were distinct from those labeled by SNX2, suggesting that significant amounts of these two sorting nexins reside in distinct compartments.
As indicated above by the sucrose density gradient study, SNX16 was found to elute in fractions that also contained LAMP1; therefore, we proceeded to examine the possibility that SNX16 may be present in the late endosome/lysosome compartment. A431 cells expressing Myc-SNX16 were processed for immunofluorescence using a monoclonal antibody against LAMP1 as a marker for late endosome/lysosome. As shown in Fig. 5B, there is good co-localization between SNX16 and LAMP1. Not only is this evident in the perinuclear region, but the majority of peripheral puncta observed for SNX16 (Fig.  5B, a) overlaps very well with the peripheral puncta (late endosome/lysosomes) marked by LAMP1 (d). Even where the level of Myc-SNX16 was high enough to aggregate the endosome and SNX16 in the perinuclear region, many of the remaining peripheral structures overlapped with LAMP1 (Fig.  5B, e-h). These results collectively suggest that SNX16, at least in steady state, resides in both the early and late endosome/lysosome system.
Deletion of the Coiled-coil Domain of SNX16 Abrogates Its Late Endosome/Lysosome Localization-Because the majority of sorting nexins described show an early endosomal localization and SNX16 appears to exhibit a substantial localization to the late endosome/lysosome, we decided to express a number of deletion mutants of SNX16 to define the structural/molecular basis for its distribution in the late endosome/lysosome. As expected for some but not all sorting nexins, deletion of the entire PX domain of SNX16 totally abolished its association with membrane structures, resulting in distribution to the cytosol (data not shown). Deletion of the N-terminal region immediately preceding the PX domain did not have an effect on the residence of SNX16 in either the early or late endosome/ lysosome, as the pattern of staining observed for SNX16⌬NT by immunofluorescence was indistinguishable from that of the full-length protein (data not shown). However, when the putative coiled-coil domain of SNX16 was removed, the late endosome localization seen with wild type SNX16 was no longer present. As can be seen in Fig. 6, expression of SNX16 lacking the coiled-coil domain (SNX16⌬CT1) (Fig. 3A) resulted in somewhat enlarged punctate structures (Fig. 6, a and e) in comparison to those observed for wild type SNX16. SNX16⌬CT1 also differed from the wild type SNX16 as it co-localized nearly completely with EEA1 in both the perinuclear and peripheral regions (Fig. 6, c and d). In fact, expression of SNX16⌬CT1 altered the labeling pattern of EAA1 compared with that of non-transfected cells with EEA1 showing fewer but enlarged punctate structures (Fig. 6b). Similarly SNX2, which was distinct from the peripheral puncta observed for wild type SNX16, now showed partial overlap with the majority of the peripheral structures observed for SNX16⌬CT1 (Fig. 7, e-h). When double labeling was performed with LAMP1, in direct contrast to the results obtained with wild type SNX16, the peripheral lysosomes illuminated by the LAMP1 antibody showed no significant overlap with those containing SNX16⌬CT1 (Fig. 6, e-h).
To ensure that the difference in localization observed for SNX16⌬CT1 was not due to alteration of the lipid binding characteristics of the protein, the mutant was expressed as a fusion with GST in bacteria and was used in the protein lipid overlay assay described previously. As expected GST-SNX16⌬CT1 bound only PI3-P (Fig. 3B). Taken together, these results suggest that deletion of the coiled-coil domain of SNX16 somehow inhibits its localization to the late endosome/lysosome, resulting in retention of the mutant in the early endosome thereby causing alteration of its structure.
Expression of the SNX16⌬CT1 Mutant Delays Early to Late Endosome Trafficking-The prototypic sorting nexin, SNX1 was first identified for its ability to interact with the lysosomal targeting signal of epidermal growth factor receptor (EGFR) (21). SNX1 interacts with EGFR, and its overexpression was shown to enhance the degradation of the receptor, as measured by following the internalization and subsequent degradation of fluorescently labeled ligand EGF. Since this first discovery, the ability of a sorting nexin to either associate with EGFR (22,23) or to affect the degradation of EGFR or its ligand (16,35) has provided a useful system to study sorting nexin function. We have therefore examined the effect SNX16 and its mutants on trafficking of rhodamine-labeled EGF (EGF-rh). In this triple labeling immunofluorescence assay, EGF-rh bound to surfaceexpressed EGFR at 4°C was internalized with incubation at 37°C, followed by fixation of cells at specified times. Fig. 8 (a-d) shows that after 10 min at 37°C, EGF has been internalized to the early endosome. While at the 60-min time point segregation of EGF from the early endosome marked by EEA1 was observed (Fig. 8, i-l). At the 180-min time point decrease of the majority of internalized of EGF-rh was observed, probably due to degradation in the lysosome (Fig. 8, q-t), in cells expressing wild type SNX16. Importantly, SNX16 together with EGF was seen to segregate away from the EEA1-marked early endosomes (Fig. 8, j), an effect that is most obvious after a 45-60-min chase. The observed rate of EGF-rh trafficking after internalization and subsequent degradation in cells expressing wild type SNX16 was indistinguishable from the results obtained from non-transfected cells. However, expression of the coiled-coil domain mutant had a marked effect on the trafficking of EGF-rh. After a 10-min chase, cells expressing SNX16⌬CT1 showed normal internalization to the early endosome as EGF-rh co-localized with both EEA1 (Fig. 8, g and h) and Snx16⌬CT1 (f and h). At 60 min, however, the presence of SNX16⌬CT1 delayed the segregation of EGF-rh from the early endosome, as the majority of EGF-rh was still present in a compartment marked by both EEA1 (Fig. 8, o and p) and the Myc-tagged protein (n and p). The delay of EGF-rh segregation was still observable after 180 min (Fig. 8, w), although some EGF-rh was beginning to segregate away from the early endosome (u-x). The efficient degradation of internalized EGF-rh seen in SNX16-expressing (Fig. 8, q-t) and non-transfected cells was somehow compromised in cells expressing SNX16⌬CT1. In order to quantify the inhibition observed with overexpression of Myc-SNX16⌬CT1, a similar experiment to that described above was performed. In this case, cells expressing either SNX16 or SNX16⌬CT1 were fixed after 30, 90, and 180 min of internalization. By using confocal microscopy the difference in EGF-rh intensity between overexpressing and neighboring control cells was determined for 100 cells at any given time point. Fig. 9 shows the plot of these intensity differences. As can be seen, cells expressing SNX16 show little difference in EGF-rh intensity to that observed for control cells across all time points. However, although the EGF-rh intensity difference from cells overexpressing SNX16⌬CT1 shows little difference to those overexpressing SNX16 at 30 min, the latter time points, where EGF-rh is transported through later compartments of the endosome to the lysosome for degradation, show a substantially higher level of EGF-rh as quantified by the increase in intensity difference for SNX16⌬CT1-expressing cells, indicating that EGF-rh is not being efficiently transported to the lysosome.
The Coiled-coil Domain Is Required for Homo-oligomerization of SNX16 -The formation of homo-and hetero-oligomers is a common feature of many of the sorting nexins studied to date. SNX1 can associate with itself (36) as well as with SNX2 and SNX4 (22). Similarly, SNX6 can associate with itself and with SNX1, SNX2, and SNX4 (23); and SNX15 can also associate with itself, SNX1, and SNX2 and weakly with SNX4 (37). Interestingly, SNX3 has not been found to associate with any other sorting nexins (22,23,36,37). To determine whether SNX16 was able to form homo-and hetero-oligomers, we cotransfected cells with HA-tagged SNX16 and Myc-SNX1, Myc-SNX3, Myc-tagged SNX16, or each of its various deletion mutants (Fig. 10A). Cell lysates were immunoprecipitated with anti-Myc antibodies. The immunoprecipitates were analyzed by immunoblot analysis with anti-Myc antibodies to determine the efficiency of the immunoprecipitation as well as anti-HA antibodies to detect the amounts of co-immunoprecipitated HAtagged SNX16. SNX16 was found to associate with itself (Fig.  10, lane 1). This interaction is not dependent on its ability to bind to the membrane, as mutation in a key residue of the PX domain responsible for PI3-P-binding (RRA) had no effect on its interaction with HA-SNX16 (lane 2). The SNX16 N-terminal region plays no role in the self-association, as its deletion had no effect and the N-terminal region alone could not interact with HA-SNX16 (Fig. 10, lanes 3 and 7, respectively). Deletion of the entire C-terminal region following the PX domain abolished self-association (Fig. 10, lane 4). The C-terminal fragment alone was able to associate with SNX16 (Fig. 10, lane 8). Because this region seemed important for self-association, more truncation mutants were made to delineate the C-terminal region critical for this function. Deletion of the region corresponding to the very extreme C-terminal amino acid residues 295-343 did not effect homo-oligomer formation at all, as FIG. 4. Endogenous SNX16 localizes to the early endosome and the late endosome/lysosome fractions resolved by a sucrose density gradient. A431 cells were harvested, and a post-nuclear supernatant was prepared and loaded on top of a 10 -40% discontinuous gradient. Following centrifugation, fractions were collected from the bottom of the tube. 25 l of each fraction was resolved by electrophoresis and immunoblotted with the indicated antibodies. Fractions are numbered with respect to increasing density. the amount of HA-SNX16 immunoprecipitated by this mutant was similar to that observed with wild type Myc-SNX16 (Fig.  10, lane 6). However, when the region corresponding to the coiled-coil domain (amino acid residues 213-295) was deleted, homo-oligomer formation was abolished in a similar manner to that observed for the entire C-terminal deletion (Fig. 10, lane  5). This result indicates that the coiled-coil domain is a critical determinant for the self-association of SNX16. Furthermore, SNX16 was unable to hetero-oligomerize with either SNX1 or SNX3 (Fig. 10, lanes 9 and 10), suggesting that unlike the other sorting nexins mentioned above, SNX16 is not part of the SNX1 sorting nexin complex. DISCUSSION The sorting nexins are a family of proteins identified by the presence of a PX domain and are implicated in regulating   FIG. 5. SNX16 partially co-localizes with EEA1 and LAMP1. A, A431 cells were transfected to express Myc-SNX16 and processed for indirect immunofluorescence using a rabbit polyclonal antibody against Myc (revealed by a goat polyclonal antibody against rabbit IgG conjugated to FITC) (a and e), and a mouse monoclonal antibody against EEA1 (revealed by a goat polyclonal antibody against mouse IgG conjugated to AlexaFluor 555) (b and f). The presence of yellow staining indicates overlap in the merged pictures (c, d, g, and h), and the overlap is observed preferentially in the perinuclear region, but little overlap in the peripheral structures was observed. B, cells prepared as in A were processed for immunofluorescence with rabbit polyclonal antibody against Myc (a and e) and mouse monoclonal antibody against LAMP1 (b and f). Again yellow staining in the merged images (c, d, g, and h) indicates overlap. Partial overlap of the peripheral structures was observed. Scale bars, 10 m.
protein trafficking and associated signaling pathways. In the present study, we describe the biochemical, cell biological, and functional characteristics of SNX16, a 343-residue hydrophilic protein containing a central PX domain and a C-terminal flank-ing coiled-coil domain. Examination of SNX16 from different mammalian species has found that although high conservation is observed throughout the region containing both the PX domain and coiled-coil domain, substantial difference in the N-  -SNX16 (a-d, i-l, and q-t) or Myc-SNX16⌬CT1 (e-h, m-p, and u-x) were incubated with EGF-rhodamine on ice. Surface-bound EGF was then internalized at 37°C for the indicated times. Cells were fixed and processed for triple labeling terminal and especially the C-terminal regions is observed. In the case of the C-terminal region, human SNX16 contains 11 more acidic residues than does its rodent counterparts, which may result in an ϳ9-kDa increase in the apparent molecular mass of the human species. The physiological relevance of this difference remains to be determined, and our current study focuses on the human protein. Although two isoforms of SNX16 are identified, our studies concentrated on isoform A, as isoform B contains a 29-amino acid residue deletion of the PX domain, which would destroy the character of this domain. Our data suggest that SNX16 is associated with several endosomal structures and may regulate the trafficking of proteins through this pathway.
The PX domain is a region of ϳ100 -130 amino acids first identified in SNX1 and the p40 Phox and p47 Phox subunits of NADPH oxidase (8). The PX domain is now recognized as a motif that mediates interaction with phosphorylated phosphatidylinositides. This interaction with phosphatidylinositides may be important for spatial recruitment of proteins to membranes and/or allosteric regulation of protein activity. The PX domains of SNX3 (16), Vam7 (17), and p40 Phox (18,19) were first described to be able to bind PI3-P and that of p47 Phox to PI3,4-P 2 . This view is further supported with studies of other PX domain proteins that show the PX domain of SNX2 (33) binds PI3-P, the PX domain of RGS-PX1 binds strongly to PI5-P and PI3-P, and weakly to PI3,5-P 2 (35), and the PX domain of CISK binds PI3,5-P 2 , PI3,4,5-P 3 , and to some extent PI4,5-P 2 (30). The lipid binding specificity of SNX1 is slightly complicated as the lipid binding specificity varied depending on the type of assay used. However, consensus shows a preference for PI3,5-P 2 and PI3-P (33,38). We have shown here that SNX16 preferentially binds PI3-P. Interestingly, when only the PX domain of SNX16 was tested, some binding to PI3,4-P 2 was also observed. This may be due to increased flexibility in the structure of the isolated PX domain, allowing entry of the bis-phosphorylated lipid into the binding site and may explain some of the weak binding seen with some of the other PX proteins, as isolated PX domains were used in these assays.
Localization studies of SNX16 by both biochemical and cellular means have suggested a novel distribution for it in several endosomal compartments. When we separated membrane structures according to density on a sucrose gradient, we found that SNX16 was present in fractions marked by the early endosomal marker EEA1 and also to those characterized by the late endosome/lysosome marker LAMP1. By using indirect immunofluorescence, we found that Myc-tagged SNX16 resides in discrete punctate structures present in both perinuclear and peripheral regions of the cell. In accordance with the gradient result, perinuclear structures of Myc-SNX16 overlapped quite well with EEA1, whereas the majority of the more peripheral puncta exhibited co-localization or partial overlap with LAMP1.
The majority of PX domain proteins studied so far are believed to be located only in the early endosome. Although the majority of SNX1 and SNX2 structures exhibited similar density with EEA1-containing compartment in the sucrose gradient experiment, SNX1 resides in discrete punctate structures identified as early endosome, which overlap only partially with EEA1 and are for the most part distinct from LAMP1 (33,36). These structures have also been found to contain SNX2 (33), SNX4, SNX5, and SNX6 (23,39), which is not surprising because with the exception of SNX5, these proteins have been found to interact with each other (22,23). SNX3 also resides in punctate structures, although they co-localize more with the transferrin receptor and EEA1 and are identified as structures involved in trafficking from early endosomes to recycling endosomes (16). Recently, SNX17 was identified and similar to the other sorting nexins has an early endosome localization (34). Until now, the only PX domain proteins exhibiting late endosome/lysosome localization are the yeast target SNARE Vam7, which localizes to vacuole membranes (17,40), and SNX15, which shows partial overlap with LAMP1 (41). However, the overlap observed between SNX15 and LAMP1 is only evident when SNX15 is expressed to very high levels and may represent an aggregation of the endosomal system rather than residence in the late endosome/lysosome per se, as these structures also contained markers for the early endosome and trans-Golgi network (41). Vam7 consists of a PX domain and a coiledcoil SNARE domain homologous to that of SNAP-25. Mutation of either of these regions destroyed the vacuolar localization of FIG. 9. SNX16⌬CT1 inhibits EGF degradation. Degradation of rhodamine-conjugated EGF in the lysosome was measured by a loss of fluorescence using Zeiss confocal microscope (LSM510 software provided by the supplier). To quantify the effect of SNX16⌬CT1 on this process, the difference of the average intensity of EGF-rh between cells expressing either SNX16 or SNX16⌬CT1 and neighboring control (nontransfected) cells was determined at the given time points following release of surface-bound EGF-rh as described in Fig. 8 (100  using rabbit polyclonal anti-Myc (visualized with goat polyclonal antibody against rabbit IgG conjugated with FITC) and a mouse monoclonal antibody against EEA1 (visualized with a goat polyclonal antibody against mouse IgG conjugated to AlexaFluor 647), and EGF-rh was visualized directly (a, e, i, m, and u). Distinct transport events of EGF-rh could be followed, as overlap with EEA1 indicates presence in the early endosome, whereas loss of overlap at later time points highlights segregation of EGF away from the early endosome on its way to the later structures. Loss of staining for EGF-rh indicates degradation in the lysosome. The boxed portions were enlarged, and yellow staining indicates co-localization between SNX16/SNX16⌬CT1 and EGF-rh (b, f, j, n, r, and v); pink staining indicates overlap between EGF and EEA1 (c, g, k, o, s, and w); and white staining indicates overlap of SNX16/SNX16⌬CT1, EGF, and EEA1 (d, h, l, p, t, and x). Scale bar, 10 m.
Vam7, suggesting that it is the combined action of both the PX domain and the SNARE domain that is important for vacuolar localization (40).
Examination of the distribution of SNX1 and SNX2 on the sucrose density gradient revealed a surprising result. Whereas SNX1 showed a profile where most of the protein was present at densities corresponding to the early endosome, SNX2 showed a distribution similar to SNX1 in the early endosome region but had a substantial amount of the protein present at densities corresponding to the late endosome/lysosome similar to SNX16 and LAMP1. Localization of SNX2 to these structures has not been reported previously. However, when we used SNX2 in double labeling immunofluorescence studies with SNX16, co-localization between the two proteins showed few of the peripheral puncta of SNX16 overlapping with SNX2. Although this result supports the findings that SNX2 resides in the early endosome (33), it does not rule out the possibility that some SNX2 is also present in other late endosomal structures, which may not be the same as the SNX16 containing late endosomes.
The molecular and biochemical basis underlying the distri-bution of SNX16 between the early endosome and the late endosome/lysosome were then addressed. By using deletion mutants we have shown that the coiled-coil domain of SNX16 is necessary. Similar to the full-length protein, the SNX16 coiledcoil deletion mutant bound exclusively to PI3-P. However, immunofluorescence localization showed that the usual punctate structure was somewhat enlarged as compared with wild type SNX16. The most interesting aspect of the coiled-coil deletion mutant punctate structures is that they are almost completely localized to the early endosome, evident by the near complete overlap with EEA1 and SNX2 and the lack of overlap with LAMP1. Therefore, the consequence of removal of the coiledcoil domain of SNX16 was the inability of the protein to move out of the early endosome to the late endosome/lysosome. Overexpression of this mutant affected the trafficking of proteins destined for degradation in the lysosome. SNX1 was identified by its interaction with the lysosomal targeting signal of the EGF receptor, and overexpression of the protein enhanced the degradation of the receptor (21). Based on these studies we examined the effect of overexpression of SNX16 on the trafficking of the EGF receptor as traced by its FIG. 10. The coiled-coil domain of SNX16 is important for homo-oligomerization. A, diagrammatic representation of the SNX16 deletion mutants. B, A431 cells were co-transfected with HA-SNX16 and Myc tagged proteins: SNX16, each of the various truncation mutants, SNX1, or SNX3. 24 h after transfection, total cell lysates were prepared and incubated with monoclonal antibody against Myc cross-linked to protein A-Sepharose. The immunoprecipitates were resolved by electrophoresis, transferred to nitrocellulose, and probed with rabbit polyclonal antibody against the HA tag (upper panel) to detect the efficiency of co-immunoprecipitation as a measure of oligomerization. Following stripping, the immunoblots were re-probed with rabbit polyclonal antibody against the Myc tag (middle panel) to ensure that immunoprecipitation of the Myc-tagged proteins was effective. 5% of the lysates used in the immunoprecipitations was resolved by electrophoresis and the Western blots probed with the HA antibody (lower panel) to show the expression of HA-SNX16.
ligand. Overexpression of SNX16 generally did not have an effect on the internalization and subsequent degradation of the EGF, with SNX16-positive cells showing normal rates of EGF degradation. When the coiled-coil deletion mutant was expressed, trafficking of internalized EGF was delayed at the level of the early endosome. The internalization of EGF to the early endosome was not effected by expression of the mutant as these cells showed normal levels of EGF in the early endosome after 10 min of internalization. However, the segregation of EGF away from the EEA1 marked early endosome was delayed. Indeed EGF was only beginning to move out of early endosomes in the mutant-expressing cells when normal cells had almost completely degraded the internalized EGF.
Our results suggested that the coiled-coil domain of SNX16 is pivotal to its function, as its absence not only altered the distribution of SNX16, but also delayed the trafficking of internalized EGF. Examination of the coiled-coils of a variety of proteins suggests that this domain may represent one of the principal determinants in oligomerization of proteins (42) and/or interaction with other proteins. Like some other sorting nexins, SNX16 was revealed to form homo-oligomers. By using various deletion mutants we found that the coiled-coil domain of SNX16 is both necessary and sufficient for this homo-oligomerization based on two observations. First, its removal abolished its ability to self-associate, and all of the SNX16 mutants without the coiled-coil region did not show any association with full-length SNX16. Second, the domain on its own could associated with the full-length protein. It is clear from our studies that the PX domain does not contribute to the oligomer formation of SNX16.
With the exception of SNX3 and SNX8 (Mvp1p in yeast), most of the sort nexins characterized so far contain two or three coiled-coil domains following their PX domains (39), and the role these domains play in the function of the sorting nexins is receiving more attention. In the case of SNX5 and SNX1, complete removal of the C-terminal region immediately after the PX domain resulted in loss of membrane localization (33,39). Recent studies have begun to examine the role of coiled-coil domains in sorting nexin function. Wang and co-workers (24) have shown that this region of SNX1 can by itself assemble with the full-length protein, localize to the early endosome, and cause accumulation of the activated G-protein-coupled receptor, protease-activated receptor 1, in the early endosome, preventing its degradation in the lysosome. However, Zhong and co-workers (33) have found that overexpression of a SNX1 mutant lacking the C-terminal coiled-coil region disrupted the degradation of EGFR in the lysosome causing its accumulation in the early endosome. Therefore, deletion of either the PX domain or the C-terminal coiled-coil region of SNX1 has detrimental effects on the trafficking of proteins out of the early endosome. This study also examined the formation of homoand hetero-oligomers of SNX1 and SNX2. The PX domain, but not the coiled-coil domain, of SNX1 was shown to mediate oligomerization with itself and with SNX2, whereas both the PX domain and the coiled-coil domain of SNX2 contain independent oligomerization activity. The N-terminal PX domain of SNX2 can mediate self-association as well as association with SNX1, whereas the coiled-coil domain of SNX2 can only mediate self-oligomerization by interaction with its own coiled-coil domain. In addition, SNX1 and SNX2 have been found to associate with other sorting nexins and other proteins and are present in larger complexes containing other retromer components, such as Vps35, Vps29, and Vps26 (22,23,37,43). SNX6 (23) and SNX15 (37) also exhibit self-association mediated by the PX domain. Both these proteins can also form PX domainmediated hetero-oligomers with SNX1, SNX2, and SNX4. It is therefore important to note that SNX16 is unique as oligomerization of SNX16 was shown to be solely dependent on the presence of the coiled-coil domain. Importantly, we did not observe any hetero-oligomerization of SNX16 with SNX1 or SNX3, suggesting that SNX16 may represent a unique SNX that regulates endosomal traffic through a mechanism different to that involving SNX1 and it oligomeric partners.
In summary, SNX16 seems to be distributed to both early endosome and late endosome/lysosome. The PX domain alone seems to be sufficient for early endosomal recruitment likely due to its specific affinity for PI3-P, whereas the combined action of the PX domain and the coiled-coil domain is essential for its association with late endosomal structures. Deletion of the coiled-coil domain disrupted the formation of SNX16 homooligomers, and cells expressing this mutant exhibited a significant delay in the segregation of internalized EGF away from the early endosome, inhibiting its trafficking to the lysosome and ultimately its degradation. It is tempting to speculate that the coiled-coil domain mediated homo-oligomerization of SNX16 may be the driving force for segregation of SNX16 in the early endosome for trafficking to the later structures of the endocytic pathway. This could be achieved either by causing the formation of the vesicle itself, where the assembly of SNX16 oligomers may cause deformation of the membrane or by recruiting other trafficking proteins to the early endosomes which result in vesicle formation.