A mechanism for retromer endosomal coat complex assembly with cargo

Significance

The endosomal system is a network of organelles that play key roles in nutrient uptake, protein and lipid sorting, and signal transduction. Integral membrane proteins are delivered to endosomes via trafficking from the plasma membrane and the secretory pathway, and many of these proteins are then returned from the endosome for reuse. The selection and packaging of many integral membrane proteins into transport carriers that export cargo from the endosome requires a protein complex called “retromer,” whose function protects organisms from metabolic defects, Charcot–Marie–Tooth neuropathy 2B, Parkinson’s disease, and Alzheimer’s disease. We elucidate the minimal requirements for targeting of retromer to the endosome membrane and show that this mechanism facilitates retromer recognition of a cargo protein.

Abstract

Retromer is an evolutionarily conserved protein complex composed of the VPS26, VPS29, and VPS35 proteins that selects and packages cargo proteins into transport carriers that export cargo from the endosome. The mechanisms by which retromer is recruited to the endosome and captures cargo are unknown. We show that membrane recruitment of retromer is mediated by bivalent recognition of an effector of PI3K, SNX3, and the RAB7A GTPase, by the VPS35 retromer subunit. These bivalent interactions prime retromer to capture integral membrane cargo, which enhances membrane association of retromer and initiates cargo sorting. The role of RAB7A is severely impaired by a mutation, K157N, that causes Charcot–Marie–Tooth neuropathy 2B. The results elucidate minimal requirements for retromer assembly on the endosome membrane and reveal how PI3K and RAB signaling are coupled to initiate retromer-mediated cargo export.

Sorting of cargo within the endosome determines whether it will be retained and ultimately degraded via lysosome-mediated turnover, or exported via a plasma membrane recycling or retrograde pathway that directs cargo to the TGN or recycling endosome. Genetic dissection of endosomal retrograde pathways in budding yeast (Saccharomyces cerevisiae) led to the identification of an endosome-associated protein complex termed retromer, composed of a Vps5–Vps17 heterodimer and a trimeric complex of the Vps26, Vps29, and Vps35 proteins (1). The retromer trimer, also called the “cargo recognition complex,” is the core functional unit of retromer, serving as a platform for recruiting many other factors to the endosome (2), and we shall refer herein to the trimer as retromer. It is now appreciated that retromer constitutes an ancient, evolutionarily conserved protein sorting complex that operates in multiple endosomal cargo export pathways (2, 3). Hence, elucidating the molecular mechanisms that underlie retromer function is key for understanding the endosomal system.

The formation of a vesicular transport carrier is typically initiated by a GTPase module that elicits recruitment of a coat protein from the cytosol to a particular site on the membrane. Retromer is an effector of RAB7A [henceforth referred to as RAB7 (human) or Ypt7 (yeast)], a GTPase regulator of endosome dynamics and depletion of RAB7-GTP in cells results in a substantial loss of endosome-associated retromer (4–9). In addition to GTPase signaling modules, interactions of coat proteins with membrane lipids, such as phosphoinositides, contribute to coat assembly by increasing the avidity of membrane binding. There is no evidence that retromer directly recognizes membrane lipids (10, 11). Instead, retromer membrane recruitment is attributed to its association with any of several different sorting nexins (3), which are peripheral membrane proteins defined by the presence of a Phox homology (PX) domain that recognizes phosphatidylinositol 3-phosphate (PtdIns3P), a signature component of endosomal membranes. However, a formal test of this hypothesis is lacking. Retromer binds sorting nexin 3 [henceforth referred to as SNX3 (human) or Snx3 (yeast)] (12–15), and in SNX3 knockdown cells, less retromer is associated with endosomes (14). In this study, we show that SNX3 and RAB7 are coordinately recognized by retromer and that these interactions are sufficient to recruit retromer to a membrane where they poise retromer to capture integral membrane retrograde cargo.

Results

The structure of yeast Snx3 shows that it is composed of short unstructured regions at the amino and carboxy termini and a PX domain that, when bound to the endosome membrane, presents a prominent three-stranded β-sheet to the cytoplasm (Fig. 1A) (16) that we speculate serves as an interaction surface for binding other factors such as retromer and/or cargo proteins. Because both human and yeast SNX3 associate with retromer (12, 13, 17), we implemented a functional assay in yeast to determine whether residues on this surface are required for the function of yeast Snx3 in vivo, and then extended this information to human SNX3. Two or three codons encoding residues (depending on the length of the strand) with solvent-exposed side chains were changed to alanine (Fig. 1A), and then the mutant genes were used to replace the native SNX3 locus. Live cell fluorescence microscopy of GFP-tagged Snx3-β1 and Snx3-β3 mutant proteins shows that they localize to punctate endosomes, just as wild-type Snx3-GFP (Fig. 1B). However, the Snx3-β2 mutant protein was barely detectable by fluorescence microscopy and immunoblotting, suggesting that it is structurally compromised and degraded as a consequence. Because the Snx3-β1 and Snx3-β3 mutants localize to endosomes, and appear to be as stable as wild-type Snx3, we focused further analysis on these mutants.

Fig. 1.

Mutations on the β1 and β3 strands ablate function but not endosome localization of yeast Snx3. (A) Ribbon diagram of yeast Snx3. The location of mutations on β1 (magenta E37A E39A H41A) and β3 (green R74A K77A) are indicated, and the site of the Gly23Cys mutation used to derivatize Snx3 is indicated in red. The structure is based on coordinates published in Zhou et al (16). (B) In vivo localization of Snx3-GFP. Each mutant Snx3 gene replaced the wild-type locus and is expressed with a C-terminal GFP tag. (C) The Snx3- and retromer-dependent cargo, Ste13, was expressed as a GFP fusion in cells expressing the indicated form of Snx3 and imaged by deconvolution microscopy. For B and C, one image from the approximate center of a Z series is shown. (Scale bar: 5 μm.)

Structural Requirements for Snx3 Function.

To determine whether the β1 and β3 mutations affect Snx3 function in retrograde sorting, the steady-state localization of Ste13, an integral membrane protease that is retrieved from the endosome in a Snx3- and retromer-dependent manner, was examined (18). In wild-type cells, GFP-Ste13 localizes to punctate Golgi and endosome compartments, whereas in snx3Δ cells, it localizes to the vacuole membrane and vacuole-associated endosomes (Fig. 1C). In cells expressing the Snx3 β1 or β3 mutant, GFP-Ste13 localizes prominently to the vacuole membrane and to punctate endosomes surrounding the vacuole, clearly demonstrating that mutations in β1 and β3 result in a loss of Snx3 function in vivo. We note that the appearance of GFP-Ste13 in cells expressing the Snx3-β1 mutant is indistinguishable to that of snx3Δ cells, whereas vacuole localization in Snx3-β3–expressing cells is less prominent, suggesting that the β1 mutations result in a severe impairment of Snx3 function, whereas the β3 mutant is markedly, but not completely, functionally compromised. Differential centrifugation experiments confirm these findings (Fig. S1A).

Two biochemical activities have been ascribed to yeast Snx3: binding to the cytoplasmic portions of the Ste13 and Ftr1 proteins that contain retrograde sorting signals (12, 18) and association with retromer (12, 13). In principle, loss of either function should result in a loss of Snx3-dependent retrograde sorting, so we examined each of these activities. In vivo association of Snx3 with retromer was monitored by coimmunoprecipitation after chemical cross-linking of intact cells (12). In this assay, the Vps29 retromer subunit contains a C-terminal HA epitope tag that permits the entire retromer complex, and associated Snx3, to be immunopurified from detergent solubilized cell lysates (12). By this assay, substantially less of the β1 and β3 mutant Snx3 proteins copurify with retromer (Fig. 2A). Notably, the amount of Snx3-β1 mutant protein that copurified was barely detectable, in accordance with the null phenotype of this mutant with regard to Ste13 localization (Fig. 1C). To monitor cargo recognition, we used a pull-down assay that measures binding of Snx3 and the cytoplasmic segment of Ste13, which contains the retrograde sorting signal. Snx3 was produced by coupled transcription translation in the presence of [³⁵S]methionine and incubated with beads coated with varying proportions of GST or GST-Ste13(1–117) (Fig. S1B). The results show that the amount of wild-type and mutant Snx3 proteins captured by the beads is proportional to the amount of the GST-Ste13 fusion protein, and that both mutant forms of Snx3 were captured by the GST-Ste13 beads essentially as efficiently as wild-type Snx3. We therefore conclude that the β1 and β3 Snx3 mutant proteins are compromised in binding retromer but are not significantly compromised in recognition of cargo.

Fig. 2.

Mutations in β1 of yeast and human SNX3 perturb recognition by retromer cargo selective complex. (A) Coimmunoprecipitation of myc epitope-tagged Snx3 with Vps29-HA immunopurified from yeast cell lysates prepared in the presence of 5 mM DSP cross-linking reagent was visualized by anti-myc blot (Lower). Vps29-HA levels were visualized by anti-HA blot (Upper). (B) Human retromer was immobilized and incubated with purified, recombinant SNX3 or SNX3-β1 mutant. Bound fractions were assayed by anti-SNX3 immunoblot (Upper) and staining with CBB (Lower).

To more directly address association of SNX3 with retromer, and to determine whether the structural requirements for the recognition of human SNX3 and retromer are conserved between the yeast and human proteins, we tested the ability of purified human retromer and human SNX3 to associate in vitro. Human retromer was assembled from proteins produced in Escherichia coli and immobilized on beads via an N-terminal GST moiety on GST-VPS35. As expected (13), purified recombinant human SNX3 was retained by these beads (Fig. 2B). Because the yeast Snx3-β1 mutant was the most severely functionally compromised mutant, we prepared a recombinant form of human SNX3 in which the corresponding positions (F28, D30, E32) were changed to alanine, and expressed and purified this protein from E. coli. Importantly, the solubility and purification of mutant SNX3-β1 were identical to wild-type SNX3, suggesting that the mutations do not grossly perturb its structure. When assayed for binding to human retromer (Fig. 2B), a striking loss of binding is observed, in agreement with results for yeast SNX3-β1 mutant. These data show that docking of both yeast and human SNX3 into retromer requires side-chain residues within the β1 strand of SNX3, and they formally establish that association of SNX3 with retromer is essential for SNX3 function in retromer-mediated cargo sorting.

Recognition of SNX3 by Retromer.

To gain insight into the retromer structural requirements for binding of SNX3, we first set out to identify the retromer subunit(s) to which SNX3 binds. Retromer functions as a trimer and the most highly conserved regions of the individual retromer proteins are those that mediate assembly of the trimer, so we considered it important to address this question by using intact (i.e., full-length) proteins. Accordingly, we devised a biochemical approach in which a chemical cross-linking reagent incorporated into SNX3 was used to determine the “nearest neighbor” contacts in the reconstituted retromer–SNX3 complex (Fig. 3A).

Fig. 3.

SNX3 is recognized by VPS35 retromer subunit. (A) SNX3-Mts-Atf-biotin cross-linking strategy: Assembled SNX3*–retromer complex was exposed to UV light to activate the Atf (X) crosslinkage. Snx3* was subsequently released by reduction. An aliquot of the reaction was removed and probed by streptavidin-HRP blotting. The remainder was subjected to proteolysis, and derivatized peptides were identified by coupled liquid chromatography high resolution mass spectrometry (LC/MS). (B) CBB-stained gels of the indicated reactions (Left). Streptavidin-HRP blots of the same fractions (Right). Anti-SNX3 immunoblots of the blotted fractions are shown below the streptavidin-HRP blot. The asterisk indicates the position of free GST. (C) Summary of LC/MS analysis of Mts-Atf-biotin cross-linked products. Portions of the spectrum containing the major cross-linked peptides, which are each derived from VPS35, are shown. The corresponding mass data are shown in the table. The underlined Lys residues were modified with biotin. (D) Schematic diagram of VPS35 and location of SNX3–cross-linked peptides. VPS35 is proposed to be composed of 17 helical repeats (18). The approximate positions of the minimal regions of VPS35 that bind VPS26, VPS29, and RAB7 (red) and modified VPS35 peptides (yellow) are indicated.

The mutagenesis data described above indicates that β1 of SNX3 is essential for binding retromer, so we targeted a residue (Gly23) that is near, but is not part of, this structural element (Fig. 1A) for conjugation to a chemical cross-linking biotin transfer reagent, tetrafluorobenzoyl-biotinamidocaproyl-methanethiosulfonate (Mts-Atf-biotin). When incorporated into a “bait protein,” the tetrafluorophenyl azide (Atf) moiety can covalently cross-link to a “prey” protein within ∼11 Å upon activation with UV light. Mts-Atf-biotin couples to cysteine residues in a bait protein via a sulfhydryl reactive methanethiosulfonate (Mts) moiety, forming a reducible disulfide bond that upon reduction elicits transfer of the biotin label to the prey protein. To construct a SNX3 Mts-Atf-biotin bait protein, two mutations in SNX3 were engineered: Gly23Cys and Cys140Ser (which eliminates the single cysteine in native SNX3). The expression, solubility, and PtdIns3P binding properties of this mutant form of SNX3 were identical to that of wild-type SNX3, indicating that these mutations did not structurally compromise SNX3. Mts-Atf-biotin was coupled to C23, and varying amounts of modified SNX3 (SNX3*) were incubated with retromer, which had been immobilized on GSH beads via the GST moiety on GST-VPS35. The bound SNX3*–retromer complexes were exposed to UV light, incubated with DTT, and the reaction components were resolved by SDS/PAGE. Streptavidin-HRP blotting of the reaction shows robust UV-induced biotin transfer to VPS35, whereas VPS26 is labeled to a small extent at the highest concentration of SNX3* assayed (Fig. 3B). Neither VPS29, nor GST (derived from GST-VPS35 fusion protein), were labeled above background.

A mass spectrometry-based approach was implemented as an additional means to identify biotin-labeled retromer subunits and to map the site(s) of cross-linking. An aliquot of a cross-linked reaction (1 μM SNX3) was subjected to proteolysis with LysC and Trypsin, and the resulting biotinylated peptides were purified by using streptavidin-agarose and then directly analyzed by liquid chromatography coupled online to high resolution mass spectrometry. In agreement with the streptavidin-HRP blotting results, analysis of purified biotinylated peptides identified two peptides with at least four spectral observations and the highest intensities (Fig. 3C). As expected, both peptides were derived from VPS35. One conjugated peptide, spanning amino acids 205–215, is contained within helical repeat 5 (Fig. 3D), and the second peptide identified spans residues 25–44 (Fig. 3D) and is contained within the minimal segment of VPS35 that assembles with VPS26 (19, 20). These results confirm and extend a recent yeast two-hybrid study that reported that SNX3 can interact with the N-terminal half of VPS35 (21). The N-terminal ∼300 aa of VPS35 containing the binding sites for VPS26, SNX3, and RAB7 comprise its most highly conserved region, underscoring the critical roles that, as we show below, SNX3 and RAB7 play in retromer recruitment to a membrane.

The region of VPS35 containing the SNX3 binding site lies immediately N-terminal to the Rab7 binding site (repeat 6) that we identified (8) and overlaps the minimal VPS26 binding region, raising the possibility that SNX3 may compete with VPS26 and/or RAB7 or, alternatively, that it forms a supercomplex with retromer and RAB7. To distinguish these possibilities, we immobilized GST-RAB7 fusion proteins on beads and incubated them with purified recombinant retromer, SNX3, or retromer and SNX3 together, and then determined which proteins were captured (Fig. 4). All three retromer proteins and SNX3 are recovered on beads coated with the constitutively GTP-bound form of GST-RAB7(Q67L) and, to a lesser extent, with native GST-RAB7, but not by beads coated with the constitutively GDP-bound form of GST-RAB7(T22N). Importantly, control experiments show that SNX3 is not captured by RAB7 beads in the absence of retromer, indicating that retention of SNX3 is mediated by RAB7-bound retromer. This experiment demonstrates that retromer can bind SNX3 and RAB7 simultaneously.

Fig. 4.

SNX3 and RAB7 bind retromer simultaneously. GST-RAB7 fusion proteins were immobilized on GSH beads: GST-RAB7 (native RAB7), RAB7-Q67L (GTP-bound mutant), and RAB7-T22N (GDP-bound mutant). A CBB stained gel (Top) anti-VPS26 blot (Middle), and anti-SNX3 blot (Bottom) are shown.

Requirements for Retromer Recruitment to a Membrane.

The mechanism of retromer coat recruitment to the endosome is not known. RNAi-mediated knockdown of SNX3 and RAB7, and expression dominant interfering RAB7 proteins in cells, results in decreased endosome association of retromer (6, 13, 14, 19), implicating them in this process. We began investigating the requirements for retromer membrane recruitment by assaying sedimentation of retromer with liposomes containing PtdIns3P, to which SNX3 binds (22). A small amount of retromer sediments with liposomes in the absence of SNX3, but this value varied (between 2 and 25%) with different preparations of retromer; we therefore considered this binding to be nonspecific and subtracted it from subsequent measurements. Preincubation of SNX3 with liposomes before the addition of retromer resulted in nearly all SNX3 sedimenting with the liposomes; however, only a modest level (ranging from 8 to 22%) of retromer was recruited to liposomes, even when SNX3 was present in substantial (16-fold) molar excess to retromer (Fig. 5 A–C). Importantly, retromer recruitment was due to SNX3 recognition because the SNX3-β1 mutant protein was completely inactive in these assays (Fig. 5B). Given this modest level of recruitment, and the low affinity of the SNX3–retromer interaction suggested by this assay and the pull-down assays (Fig. 2) (13), we conclude that SNX3 recognition alone is insufficient to mediate endosome loading of retromer.

Fig. 5.

Dual recognition of SNX3 and RAB7 recruits retromer to a membrane. (A) Liposomes were preincubated with purified SNX3, SNX3-β1, and/or RAB7-Q67L-His, and then purified retromer was added. Liposomes were collected by centrifugation, and the supernatant (S) and pellet (P) fractions (equal amounts) were loaded onto gels. A CBB-stained gel (Upper) and anti-VPS26 blot (Lower) are shown. The proportions of VPS26 in the pellet fractions, determined from this blot, are indicated below it. (B) Liposomes were incubated with RAB7-Q67L-His (“RAB7^GTP”) or RAB7-T22N-His (“RAB7^GDP”) with and without wild-type SNX3. Cosedimentation of retromer was monitored as described above. (C) Liposomes were preincubated with RAB7^GTP or RAB7^GTP K157N, with or without SNX3, and purified retromer was added. Cosedimentation of retromer was monitored as described above. (D) Retromer sedimentation was assayed with proteoliposomes alone or preincubated with SNX3 and RAB7. The proteoliposomes contained a retromer sorting motif (Tyr-Leu-Met; Center) or a peptide with each residue of this sequence, and surrounding aromatic residues, substituted by alanine.

We next tested whether RAB7 is sufficient to mediate membrane recruitment of retromer. To associate recombinant RAB7 on the surface of liposomes, we constructed versions of RAB7(Q67L) (GTP-bound conformation; “RAB7^GTP” in Fig. 5) and RAB7(T22N) (GDP-bound conformation; “RAB7^GDP”) in which its C-terminal prenylation site was replaced with a 6× histidine tag. Modified RAB7 proteins were incubated with liposomes containing 1,2-dioleoyl-sn-glycero-3-{[N-(5-amino-1-carboxypentyl) iminodiacetic acid] succinyl} [DGS-NTA(Ni)], which contains a head group that is recognized by the 6×His tag. Although RAB7 proteins associated efficiently with these liposomes, only a modest level (ranging from 4 to 20%) of retromer sedimented with the liposomes, even with an eightfold molar excess of RAB7(Q67L) over retromer (Fig. 5 A and B). This recruitment was specific for the GTP-bound conformation of RAB7 [i.e., RAB7(Q67L)], as essentially no recruitment was elicited by RAB7(T22N) (Fig. 5B) or RAB5(Q79L) (Fig. S2A). These results lead us to conclude that RAB7-GTP is insufficient to elicit recruitment of retromer to a membrane. The modest retromer recruitment activity of SNX3 and RAB7 in these experiments, and the close proximity of the SNX3 and RAB7 binding sites on VPS35, led us to consider the possibility that SNX3 and RAB7 might cooperate to recruit retromer. The results of three experiments support this hypothesis. First, when liposomes were preincubated with both SNX3 and RAB7(Q67L), nearly half of the added retromer cosedimented with the liposomes (43, 47, and 41% in the three independent experiments shown in Fig. 5 A–C). This level of binding required both active SNX3 and RAB7-GTP, and was not due to aggregation of the proteins, because the SNX3-β1 mutant and RAB7(T22N) are completely inactive in synergizing with RAB7(Q67L) or SNX3, respectively.

Mutations in RAB7A are associated with Charcot–Marie–Tooth neuropathy 2B (23), and a previous report found that one of these, K157N, results in reduced coimmunopurification of GFP-RAB7A(K157N) with retromer and reduced retromer endosome localization (6). RAB7A(K157N) is predominantly GTP loaded in vivo (24), suggesting that it may not be recognized by retromer and native RAB7A. We addressed this hypothesis by biochemical reconstitution of RAB7-dependent retromer liposome association (Fig. 5C). These experiments confirm that the mutation of K157N, in the context of otherwise wild-type RAB7 or in RAB7(Q67L), is unable to induce significant recruitment of retromer to liposomes individually, similar to RAB7(T22N) (Fig. 5C and Fig. S2B). Moreover, in combination with SNX3, the RAB7(Q67L, K157N) is unable to elicit enhancement of retromer recruitment to liposomes (Fig. 5C). These findings are consistent with the K157N mutation impairing endosomal membrane recruitment of retromer and retromer-mediated sorting.

Having reconstituted membrane targeting of retromer, we sought to test the hypothesis that these interactions prime cargo capture by retromer. We first confirmed that retromer binds directly to a cytoplasmic portion of human diivalent metal transporter 1 isoform II (DMT1-II), an iron transporter that bears a retromer sorting motif (25), and mapped this activity to the C-terminal portion of VPS35 in complex with VPS29 (Fig. S3). Next, we prepared proteoliposomes containing a peptide (250 pmol) derived from DMT1-II containing the retromer sorting motif, or a variant peptide bearing mutations in the retromer sorting motif and nearby aromatic residues (which might also contribute to retromer recognition). Each peptide possesses a membrane spanning segment and a C-terminal cysteine residue to facilitate labeling of the peptide (with a fluorescent dye) so that we could determine the amounts of each peptide that was incorporated into the proteoliposomes in the appropriate orientation. These proteoliposomes were ineffective in recruiting substantial amounts of retromer (Fig. 5D), indicating that retromer does not avidly recognize cargo in this format. We next asked whether inclusion of SNX3 and RAB7 on the proteoliposomes influences retromer recruitment. To clearly distinguish any effect, the amounts of SNX3 (0.7 μM) and RAB7 (1 μM) were reduced to the minimal amounts required to effectively recruit ∼40% of added retromer. In the presence of SNX3 and RAB7, the native peptide had a striking effect, more than doubling the amount (to >80%) of retromer that was recruited to the proteoliposomes (Fig. 5D). This enhancement is due to bona fide cargo recognition because the mutant peptide was entirely without effect.

Discussion

The results presented here define the minimal requirements for robust membrane recruitment of, and cargo capture by, retromer and establish it to function via a general paradigm of vesicle coat assembly, whereby multivalent interactions involving lipid (e.g., phosphoinositide) recognition and a GTPase module initiate cargo sorting and the formation of a transport carrier. At the steady state, SNX3 and RAB7 localize predominantly to early and late endosomes, respectively, indicating that the SNX3- and RAB7-dependent recruitment mechanism is restricted to maturing sorting endosomes as they accrue RAB7 by Rab conversion (2, 13). From these endosomes, the mechanism defined here initiates sorting into retrograde pathways that direct cargo to the TGN and the recycling endosome (13, 17, 25). In recent years it has become apparent that retromer sorts cargo at multiple exit portals within the endosomal system, and our results suggest that multivalent interactions underlie localization of retromer at other exit sites. A candidate component of a mechanism to recruit retromer to the early endosome is SNX27, which is recognized by the VPS26 retromer subunit (26–29), where it functions in a plasma membrane recycling pathway that is regulated by Rab4. Multivalent interactions allow diverse retromer recruitment mechanisms and can explain how retromer functions in organisms, such as Arabidopsis thaliana, which do not possess a SNX3 ortholog.

In solution, retromer is a ∼210-Å-long flexible rod (19), and our data demonstrate that it becomes initially anchored to the endosome membrane by interactions involving SNX3 and RAB7 at the N-terminal proximal region of VPS35. Mutations in yeast Vps35 that result in a cargo-specific sorting defect of Vps10 map to the C-terminal region of Vps35 (30), and our data show that this region in human retromer is responsible for recognition of DMT1-II (Fig. S3). The location of these interactions poise retromer on the endosome membrane to survey and capture cargo via a binding site that localizes to the opposite end of VPS35. Upon cargo recognition, retromer is stabilized on the membrane, effectively capturing cargo and initiating sorting. Thus, recognition of cargo by this region of retromer would anchor the two ends of the retromer complex to the membrane, and this arrangement likely underlies the substantially increased association of retromer with the membrane. Cargo binding activity has also been reported for the VPS26 subunit of retromer, which binds to the opposite end of VPS35 (31), raising the possibility that recognition of different cargos may drive distinct retromer coat assembly pathways.

An outstanding question regards the relationship between SNX3 and SNX-BAR proteins in retromer recruitment. In yeast, the Vps5–Vps17 SNX-BAR heterodimer, which was originally defined as stoichiometric components of retromer, is required for retromer endosome localization, but Snx3 is not (8, 12). The stable pentameric form of yeast SNX-BAR retromer, however, is distinguished from metazoan retromer, which does not associate avidly with SNX-BAR proteins. Although depletion of retromer SNX-BAR proteins in cultured human cells is reported to increase the cytosolic pool of retromer (32), multivalent recognition of SNX3, RAB7, and cargo by metazoan retromer could serve to recruit retromer on the relatively flat vacuolar membrane domain, juxtaposing retromer near a high local concentration of SNX-BARs that coat tubules. The incorporation of retromer into the SNX-BAR–coated tubular endosomal network may be coordinated with termination of RAB7 signaling by the RAB7 GTPase activating protein, TBC1D5, that binds retromer (6). In further support of this model, we recently found that in yeast cells lacking Ypt7-GTP (the ortholog of RAB7), retromer cargo accumulates in the endosome (8) and, in cultured human cells, acquisition of RAB7 on the endosome by Rab conversion is temporally correlated with the appearance of retromer-coated tubules that bud from the endosome (33). The biochemical reconstitution system established here provides an avenue to definitively address these questions.

Materials and Methods

Cell Culture and Molecular Biology.

All yeast strains were constructed in the BY4742 background (MATα his3Δ 1, leu2Δ 0, lys2Δ 0, ura3Δ 0) and propagated by standard methods (34). Additional information is provided in SI Materials and Methods.

Recombinant Protein Expression, Purification, and Binding Assays.

The purification, processing procedures, and conditions used for protein binding assays are provided in SI Materials and Methods.

Liposome Binding Experiments.

Liposomes were produced from pure synthetic lipids (Avanti Polar Lipids) by mixing dioleoyl-phosphatidylcholine, dioleoyl-phosphatidylserine, nickel salt (DOGS-NTA-Ni), and extrusion through 1-μm pore-size filters by using a miniextruder (Avanti Polar Lipids). Proteoliposomes were produced in a similar manner with peptides added to the lipid mixture before extrusion. For binding assays, liposomes (50 nmol lipid) were incubated with RAB7 and/or SNX3 at room temperature for 60 min, then with retromer at 4 °C for 60 min. Liposomes were collected by centrifugation at 200,000 × g for 20 min at 4 °C, and the supernatants pellet fractions were harvested. More detailed information, including the sequences of the peptides used to produce the proteoliposomes, is provided in SI Materials and Methods.

Cross-Linking Assay.

Purified SNX3 (G23C C140S) was conjugated with Mts-Atf-biotin reagent (Pierce Chemical) according to manufacturer’s instructions. Conjugated SNX3 (SNX3*) was incubated with GSH bead-immobilized retromer in binding buffer (without DTT) in the dark, beads were washed, 20% were removed for the “before UV” control, and the remainder was exposed to UV light. Samples were resolved by SDS/PAGE for CBB staining or immunoblotting with antibodies against SNX3 or Streptavidin-HRP, or biotin-labeled proteins were enriched on streptavidin agarose and subjected to proteolysis and mass spectrometry. Additional information is provided in SI Materials and Methods.

Mass Spectrometry.

The methods are described in SI Materials and Methods.