Sorting nexins anchor trafficking machines to membranes by binding phospholipids. The paradigm of the superfamily is sorting nexin 3 (SNX3), which localizes to early endosomes by recognizing phosphatidylinositol 3-phosphate (PI3P) to initiate retromer-mediated segregation of cargoes to the trans-Golgi network (TGN). Here we report the solution structure of full length human SNX3, and show that PI3P recognition is accompanied by bilayer insertion of a proximal loop in its extended Phox homology (PX) domain. Phosphoinositide (PIP) binding is completely blocked by cancer-linked phosphorylation of a conserved serine beside the stereospecific PI3P pocket. This “PIP-stop” releases endosomal SNX3 to the cytosol, and reveals how protein kinases control membrane assemblies. It constitutes a widespread regulatory element found across the PX superfamily and throughout evolution including of fungi and plants. This illuminates the mechanism of a biological switch whereby structured PIP sites are phosphorylated to liberate protein machines from organelle surfaces.
All endocytosed cargos are delivered to the early endosome, where they are sorted and transported to the plasma membrane, late endosome or TGN. The retromer relies on sorting nexin proteins in order to segregate the cargo proteins destined for the TGN or the plasma membrane. The sorting nexins associated with the retromer are distinguished by their architectures. The SNX3 subtype1 contain a PX domain but lack multimerization motifs, and may be found in vesicular, rather than tubular, carriers. The SNX1 subtype contain a Bin-Amphiphysin-Rvs (BAR) domain2 and form dimeric complexes capable of sensing and/or remodeling membrane curvature3. How either type and their attached trafficking machines dynamically engage cellular membranes to control cargo trafficking remains unclear.
SNX3 is the best understood short sorting nexin that consists only of a single conserved PX domain and no flanking structural domains. Its cellular localization is principally determined by the specific interaction of its PX domain with PI3P4, as presented at the cytosolic leaflet of the early endosome. Mutations within the conserved PI3P pocket of SNX3 abolish the membrane binding and compromise its endosomal functions4,5. The binding of SNX3 to endosomes is a prerequisite for association with the retromer complex6,7,8 and positions the complex for cargo packaging and transit5. The mechanism regulating this membrane complex assembly remains an unsolved mystery that is critical for the fidelity of subcellular trafficking9. The recruitment of SNX3 to the cytosolic leaflet of the early endosome is a key determinant for productive retromer formation and it serves as a potential point for regulated binding and release. Visualizing how native SNX3 engages PI3P-containing membranes provides an opportunity to explain how retromer targeting and PX protein assemblies are controlled. As PI3P is constitutively present and abundant at the early endosome where a multitude of proteins recognize it, its level of production is unlikely to determine whether proteins membrane localize. Instead, we propose that another master regulatory switch must prevail, the identity of which remains unknown.
There are no structures showing sorting nexins inserted into membranes or in a regulated state. The structure of the PX domain of the distantly related yeast Grd19p protein has been reported10, the human retromer structure has been solved7, and coordinates of SNX3 and SNX12 PX domains coordinates have been deposited as 2YPS and 2CSK, respectively. However, key loops and terminal elements are missing in earlier structures, membrane binding data is lacking and the regulatory elements remain undefined. Moreover, no structure of any full-length sorting nexin interacting with a membrane mimic has been determined.
Here, we have characterized solution structures of native human SNX3 protein states, elucidating their interactions with each individual membrane component in order to better understand how membrane recognition is mediated and controlled. Membrane interaction by full length SNX3 is necessarily orchestrated by the specific recognition of the PI3P head group, followed by deep insertion of proximal hydrophobic elements into the bilayer. A conserved serine phosphorylation site at the rim of the PI3P binding site proved critical for switching off the interaction. Mutant forms mimicking the phosphorylated and non-phosphorylated states show that the interaction between SNX3 and early endosomal membrane is regulated in cells by Ser72 phosphorylation, which abolishes PI3P binding in vitro. This appears to constitute an all-or-none phosphorylation switch for dynamically regulating reversible endosome attachment for retromer assembly and detachment. Moreover, this highly structured regulatory element near the PIP docking site is shared with other sorting nexins that mediate membrane trafficking, suggesting a broad utility for controlling how proteins selectively recognize organelles and move cargo.
Solution structure of human SNX3
In contrast to SNX-BAR proteins, which form oligomers3, SNX3 is an obligate monomer. The full length SNX3 protein remains monomeric and monodispersed in physiological solutions, as does the closely related SNX12 PX domain based on the single peaks seen in analytical ultracentrifugation (AUC) experiments. Their sedimentation coefficients of 1.81 and 1.88 s for SNX3 and SNX12-PX, respectively, are consistent with molecular masses of 15–21 kDa for SNX3 and SNX12-PX, respectively (Supplementary Fig. 1a). This compares favorably with their theoretical molecular masses of 19.9 and 17.8 kDa. Narrow peak widths in 1H, 15N resolved NMR spectra (Supplementary Fig. 1b and ref. 11) are consistent with the SNX12 and SNX3 proteins being monomeric in solution, or when bound to PI3P molecules.
The structure of the 162 residue human SNX3 protein was elucidated under physiological solution conditions. The restraints used included 4984 distances derived from the volumes of crosspeaks in 15N-resolved and 13C-resolved nuclear overhauser enhancement (NOE) spectroscopy experiments, 188 backbone dihedral angles, and 88 hydrogen bond restraints (Table 1). The structure includes a classical PX fold preceded by 25 unstructured N-terminal residues and is followed by 16 residues exhibiting irregular structure based on the NMR data (Fig. 1 and Supplementary Fig. 1c). Conversely, the C-terminus is structured and interfaces with the α1, α3, and α4 helices via the Ile150, Tyr154, and Ile159 side chains, which interdigitate with the hydrophobic core of the PX domain. This juxtaposes the flexible N-terminus and structured C-terminus far from where the membrane is engaged, thus constituting a platform for tertiary interactions with the retromer alongside the 3-stranded β sheet (Fig. 1b), consistent with previous mutational studies6. Hence this structure of functionally intact SNX3 reveals the entire protein's structured elements and dynamic features, including those that are missing in earlier structures10 or the retromer complex. The packed C-terminal hydrophobic and aromatic positions are conserved throughout SNX3 vertebrate, invertebrate, and plant evolution (Fig. 1c), suggesting a pervasive need for the interactions of these core residues.
The SNX3 structural ensemble converges, with the PX domain complete with its C-terminal extension exhibiting a backbone root mean squared (r.m.s.) deviation of 0.37 Å (Fig. 1b, Table 2). An additional helix connects PPII to α2 and is supported by close Gly94:Phe97, Gly94:Leu98, and Ala96:Arg99 distances. This helix (α*) is missing from the earlier PX structures10, but is functionally significant in that it positions exposed hydrophobic and basic side chains to form the membrane insertion loop (MIL) next to the β1–β2 hairpin loop. Both loops are positioned to contact the lipid bilayer based on analysis by the Membrane Optimal Docking Area (MODA) program12, as validated by lipid titrations (see below). The canonical PI3P binding pocket is bounded by the β1-β2 sheet, α2 and MIL. It presents a positive surface with contributions from Arg70, Lys95, and Arg118 sidechains that line the PI3P pocket. The opposite side of the domain is overwhelmingly negatively charged, with acidic Glu30, Asp32, Asp111, Asp112, Glu116, Glu117, and Glu123 residues contributing to SNX’s electronegative pole (Fig. 2) that would naturally orient away from the negatively charged membrane surface, simultaneously positioning the basic pole towards the phospholipid bilayer surface.
Nonspecific bilayer and specific PI3P interaction
The step-wise mechanism by which SNX3 recognizes phospholipids to initiate retromer assembly on the membrane was examined by NMR titrations of individual components. To study initial non-specific sampling of the membrane,15N-labeled SNX3 was titrated with either dodecylphosphocholine (DPC) or diheptanoyl phosphatidylcholine (DHPC) mixed with CHAPS at a 3:1 molar ratio. Similar chemical shift perturbation (CSP) patterns were observed with either mixed micelle (Fig. 2). In both cases, the perturbations mapped exactly onto the MIL, as delimited by residues Ala96 and Asp111. Both micelles were tested as the DHPC induces larger perturbations that were easier to resolve, and contains a more biologically representative headgroup, while DPC forms smaller micelles at lower concentrations that allow more complete saturation of the complex. Large perturbations identified the key residues involved in the interaction as being Leu98, Gln100, Phe103-Gly105, and Ile109, which present exposed sidechains for concerted bilayer insertion. However, the affinity is weak in the absence of PI3P, with saturation not being reached even at a six-fold excess of micelle, and with an apparent binding affinity of over 2 mM for DHPC micelles (Table 3). Moreover, only a predominantly electrostatic interaction was apparent, with Gln100′s side-chain and Gly105 exhibiting paramagnetic relaxation enhancements (PREs) in presence of micelles spiked with equimolar 5-doxyl PC spin label (Fig. 3a). Together, this shows that SNX3 inserts weakly and reversibly via its MIL to PIP-free bilayer surfaces. The domain’s dipole (Fig. 2) simultaneously reinforces bilayer binding through long range electrostatic forces, placing the termini and acidic cap far from the membrane surface and potentially available for retromer interaction.
The specific recognition of PI3P by 15N-labeled SNX3 was monitored by stepwise addition of soluble forms of the lipid ligand. For maximum resolution, the inositol-1,3-diphosphate headgroup [Ins(1,3)P2] or dibutanoyl (c4) PI3P were added (Supplementary Fig. 3), allowing discrimination of the effect of the glycerol and acyl chain moieties. Both derivatives yielded similar CSPs that mapped to the canonical PIP binding pocket. Significant perturbations were exhibited by residues Val39 in β1, Phe46-Thr47 in β2, Arg70-Ser72 in α1, Asp113, F114, Asn116, Arg118, Lys119 and Leu122 in α2, and by MIL residues Arg99-Leu101, Phe103, Gly105 and Asp107, while Ser72′s resonance underwent line broadening. Similar binding constants of 169.6 μM for Ins(1,3)P2 and 158.9 μM for c4-PI3P were observed, implying that SNX3 primarily recognizes the inositol headgroup of PI3P and has limited contacts with its glycerol moiety or acyl chains. A corollary is that altering the pocket responsible for specific PI3P headgroup recognition would offer the most potential for exerting regulatory control over SNX3-endosome association.
PI3P anchors SNX3 deep into the bilayer
The ternary complex was assembled by progressive addition of PI3P molecules and micelles to SNX3. Assignment of the complex involved titrating Ins(1,3)P2 to levels approaching saturation of the protein:micelle complex. This allowed tracking of the resolved 1H,15N signals (Supplementary Fig. 2a) and comparison with the slow exchange binding data obtained with c4 and c8-PI3P ligands13 (Supplementary Fig. 2b). Perturbations seen in Arg70, Ser72, Leu92, Arg99, Phe103, Gly105, and Gln100ε resonances of the ternary complex mirrored those previously observed with micelles or PI3P derivatives alone. The similar patterns of perturbations induced by the individual and combined components show that the PI3P pocket and MIL interactions are local and could conceivably provide complementary effects on membrane specificity and affinity, respectively.
In order to structurally characterize the ternary complex, the exchange kinetics were tuned by varying the length of the PI3P acyl chains. The presence of zero, short or medium length chains progressively decreases the ligand exchange rates, with c4-PI3P titration into micelle-saturated SNX3 exhibiting a slower off rate and higher affinity than the headgroup alone (Fig. 2, Supplementary Figs. 2b & 3). This reflects the stabilization provided by insertion of its acyl chains into the micelle in parallel with the headgroup binding to the SNX basic pocket. Ternary complex restraints were obtained from ligand titrations with DHPC and DPC-based while collecting15N-resolved and 13C-resolved spectra with cryoprobe-equipped 800 and 900 MHz NMR spectrometers. The orientation of SNX3 on PI3P-containing micelles was defined using the intense 13C signals, particularly from interfacial methyl groups. The resonances of the ligand-bound protein were assigned from c4-PI3P titrations, yielding 13C-resolved intermolecular NOE distances between Lys95 and H4 and H6 of PI3P, between Ile109δ1 and H1 and H3, and between Arg118γ and H4, thus defining the headgroup orientation (Fig. 3b). The slowly exchanging SNX3:micelle complex formed using c8-PI3P allowed resolution of the bound and free state peaks, with introduction of 5-doxyl PC broadening the signals of MIL residues Phe103, Gly105 and Asp107 and Phe110 (Fig. 3a and Supplementary Fig. 2c) compared to non-paramagnetic controls. Thus these SNX3 elements insert deeply into PI3P-loaded micelles, in contrast with PC-only micelles. Hence, SNX3 relies on PI3P to anchor stably into endosomal membranes, while PI3P’s absence leaves SNX3 free to dislodge after briefly sampling the bilayer. Comparison of the NOE patterns between the free and mixed micelle-bound SNX3 states indicated no major conformational changes, as corroborated by their similar circular dichroism (CD) spectra (Supplementary Fig. 1d) and local CSPs observed for SNX3′s states (Fig. 2, Supplementary Figs. 2 & 3). Together this yielded all the CSP, PRE, and NOE restraints needed to dock SNX3′s structure to the single PI3P and 54 DPC molecules within the micelle model (Table 1) in order to define how such sorting nexins structurally recognize the endosome surface.
Structural basis of membrane recognition by SNX3
How SNX3 orients itself onto membrane surfaces to anchor retromers was calculated stepwise using NMR restraints14. First, the c8-PI3P headgroup was docked into its binding pocket using intermolecular distances, chemical shift changes and co-crystal contacts. This showed a canonical ligand orientation, with Arg70, Lys95, and Arg118 residues forming hydrogen bonds with the inositol phosphate and hydroxyl groups, and Ile109 contacting the ring (Fig. 3b, c). The structure of SNX3 bound to C8-PI3P and micelles, as calculated by HADDOCK (Fig. 3d), showed that SNX3 inserts deeply, with only 33.7 Å separating the centers of the micelle and PX domain, as opposed to the 38.7 Å separation seen in the Vam7p:micelle complex1. This large interface buries 3404 ± 262 Å2 and is defined by a higher density of intermolecular restraints than has been measured for other PX domains10,15,16 (Fig. 3d; Table 1). The SNX3 MIL is particularly long, with residues including Arg99, Gln100 or Arg104 exhibiting contacts with PI3P and seven PC molecules to orient the protein on the membrane in such a way that its β-sheet remains accessible to the retromer complex6. The extensive PI3P-micelle interface also includes the β1-β2 hairpin loop, with Arg43 making electrostatic contacts with PC headgroups, the specific ligand interactions of the α1 and α2 residues, and bilayer insertion of the PI3P acyl tails. Together this yields a unique insertion angle of 31.0° ± 5.6° for the protein’s long axis into the micelle interior (Table 2) using established protocols14. The conservation of the interfacial residues across the superfamily (Fig. 4) along with MODA-based identification of the respective membrane interaction surfaces12 in each available PX domain structure indicates that other sorting nexins employ similar membrane binding modes.
PI3P site phosphorylation blocks localization
A potential mechanism of regulation of SNX3′s fast and slow release from weak and strong interactions with bilayers that are either PI3P-free or loaded, respectively, was investigated. We proposed that the conserved Ser72 position (Fig. 4), which is situated near the PI3P-binding Arg70 residue (Fig. 5a & Supplementary Fig. 4), could play a crucial role. Situated at the start of α1, Ser72 is solvent-exposed, helical in structure and environmentally hypersensitive, exhibiting NMR signal variations when the pH or buffer changes. The role of Ser72 was of interest as it is predicted to be phosphorylated17 as evidenced in various tissues18,19,20, and confirmed by mass spectrometry analysis of purified SNX3 purified from HeLa cells (Supplementary Fig. 5). The SNX3 structure indicates that phosphorylation here would prevent ligand binding by blocking inositol 3-phosphate entry and negating the optimal membrane docking propensity here.
To test this PIP-stop hypothesis, SNX3 mutants mimicking constitutively non-phosphorylated (S72A) and phosphorylated (S72E) states were designed and confirmed as folded by CD (Supplementary Fig. 1d). Compared to the wild-type form, selective membrane binding activity was retained by S72A mutant but abolished by the S72E mutant (Fig. 5b, Supplementary Figs. 6c, d). This indicates that introducing a negative charge here prevents PI3P recognition, presumably by compromising the Arg70-mediated contacts. Moreover, phosphorylation of Ser72 specifically blocks the protein from endosome docking, as validated in cellular assays. In particular, the non-interactive SNX3S72E mutant is cytosolic, much like the R70A mutant that cannot bind membranes because of a defective PI3P-binding site5, whereas the constitutively membrane-active SNX3S72A mutant is distributed on endosomes like the wild type protein (Fig. 6). We previously showed that SNX3 overexpression inhibits endosome maturation and delays the degradation of endocytosed EGF receptor5. Consistent with these findings, expression of SNX3S72A, like the WT protein, significantly delayed EGF receptor degradation, when compared to neighboring untransfected cells (Fig. 6a, c, d; Supplementary Figs. 7, 9). By contrast, the S72E mutant had essentially no effect on EGF receptor degradation, like SNX3R70A or free GFP (Fig. 6a, c, d). The submicromolar affinity and high specificity for PI3P-containing bilayers would retain the SNX3 protein by endosomal membranes, with its membrane-embedded Ser72 and slow off-rate leaving limited room for release. Once freed, if only transiently, protein kinases could act on the Ser72 motif in α1 to prevent SNX3 re-attachment to the membrane. Based on MODA analysis this negation of the dominant site of membrane docking in the retromer structure7,12, this could dislodges the entire assembly from the endosome, thus allowing recycling to the TGN until reversal of the cycle by a dephosphorylation event (Fig. 7).
The PIP-stop motif is highly conserved across SNX3 homologs in plants, fungi, invertebrates and vertebrates, as well as across much of the sorting nexin superfamily (Figs. 1c and 4). Indeed, phosphorylations of the corresponding serines have been identified in proteins including SNX1 and SNX1221,22. To test their biological impacts, mutations of the corresponding serine residues were investigated. As predicted, the SNX1S188E mutant distribution was cytosolic in contrast to the endocytic localization of the wild-type form, while the control SNX12S73A mutant retained the wild-type punctate pattern of wild-type SNX12 (Supplementary Figs. 8, 10). Thus, they mirrored the SNX3 pattern, consistent with parallel regulatory functions. Hence, the PIP-stop mechanism appears to constitute a common way of controlling sorting nexin function, and infers a general way to modulate membrane attachment by selective phosphorylation of structured PIP binding sites.
Binding of sorting nexins to endosomal membranes leads to recruitment of retromers destined to the TGN1,6 and multivesicular body genesis5. To understand this mechanism and its regulation, structures of SNX3 states were solved, illuminating how weak sampling and tight anchoring of bilayer surfaces occur. Membrane recognition includes a long MIL which includes an extra helix between the PPII motif and α2 that mediates both superficial, non-specific interaction with phospholipid surfaces as well as deep hydrophobic insertion into bilayers. An adjacent pocket recognizes PI3P and acts as the principal anchor, which leads to interfacial engagement of the β3-α1 loop and the beginning of α2 to form a strong membrane attachment and unique angular positioning on the surface of endosomes.
The structural characterization of the complete SNX3 protein in free, lipid-bound, micelle-inserted and mutated states exposes recurrent functional and regulatory features. The complete membrane interface of a sorting nexin has been elucidated, showing how a unique MIL element shared with SNX1, SNX2, SNX4, and SNX12 (Fig. 4) engages bilayers broadly to mediate endosomal sorting. Of these proteins SNX12 appears most like SNX3, while in SNX1 and SNX2 the hydrophobic and cationic character of the MIL is enhanced, perhaps endowing them with unique phospholipid binding modes including potential PI(3,5)P2 binding23. A pair of unusual sorting nexins, SNX5 and SNX6, differ in their lipid binding motifs and modes, and bind to PI(4,5)P2 and PI4P, respectively24,25.
An unusual feature seen in SNX3′s solution structure is the extended C-terminal element that juxtaposes the termini. Although C-terminal extensions pack into other SNX structures, they generally involve α-helical segments as in SNX1126 and SNX33 (pdb 4AKV), rather than the extended segment found in SNX3 as well as SNX1, SNX12 and SNX1427. This forms part of the retromer docking surface7 that allows positioning a defined distance off the membrane surface (Fig. 7).
Attachment of a phosphate to a structured PI3P site was found to compete directly with phospholipid ligand binding. This PIP-stop could control membrane binding of a sorting nexin molecule or an entire retromer. Whether this event regulates the formation, cycling or disassembly of such complexes remains to be determined. Moreover, the enzymes that create PIP-stops remain a mystery, as protein kinases are usually thought to add phosphates to induce conformational changes or create protein binding sites. Here conformational changes or other proteins are not required, the PIP-stop simply repulses negatively charged lipid bilayers. This opens new perspectives for understanding how protein association with membrane targets are controlled in cellular systems. Indeed phosphates are added to the PIP binding sites of diverse sorting nexins, revealing a widespread mechanism that is conserved across most of the SNX superfamily. This post-translational modification abolishes membrane binding not only by SNX3 but also by other sorting nexins including SNX1 and SNX12, suggesting a common mechanism for controlling protein assembly and disassembly on membranes.
Our results suggest that the identity of the residue corresponding to SNX3 Ser72 could play a determining role in the specific recognition of different organelle membranes. For instance, the PX domains of SNX5 and SNX6 have glutamates at this position and are reported to recognize PI4P and PI(4,5)P2 lipids24,25, which are concentrated in the trans-Golgi network and plasma membranes, respectively. In contrast SNX14 has a leucine here, displays cytoplasmic localization28 and lacks 3-phosphoinositide binding27. Thus the PIP-stop residues also could contribute to the PIP specificity and subcellular membrane destination of PX superfamily members.
Diverse species consistently show the importance of phosphorylation of SNX3 Ser72or its equivalent phosphorylatable residue. In Drosophila SNX3 this is the primary phosphorylated site within the entire protein29,30. In budding yeast the homologous serine is the sole phosphorylated residue in the full length protein30,31. Indeed selective phosphorylation of SNX3 Ser72 (or its homologous residue) is particularly conserved throughout evolution32. We propose that the context of the PIP-stop, including its presence in the α1 helix by the stereospecific PIP3P pocket, predisposes this element to preferential phosphorylation over other sequences such as those that are disordered or unstructured. In other words, kinases could prefer such structured substrates, or alternatively phosphatases could find them less accessible, thus allowing PIP-stops to become common phosphorylation sites.
Cancer may preferentially involve deregulation of PIP-stop elements. In SNX3 hyperphosphorylation is seen on Ser72 in leukemia20, lymphoma33, breast cancer34, and neuroblastoma cells35, which would presumably alter the trafficking of diverse cargo. The corresponding residue is hyperphosphorylated in SNX1 in leukemia and breast34, lung36, and skin cancer37. Phosphorylation of this position is the most frequent modification of SNX12 including in breast cancer36, gastric cancer38, leukemia20, lymphoma33, and neuroblastoma35. Melanoma cells frequently exhibit phosphorylation of the corresponding serine of SNX12, SNX17, and SNX2137. Lung cancer cells exhibit hyperphosphorylation of the corresponding serine in SNX1, SNX2, SNX3, and SNX2136,39. Thus, PIP-stops appear to be frequently altered in a variety of cancers, suggesting a novel point of causation and possible intervention.
Protein expression and purification
A sequence encoding full length human SNX3 was synthesized (Genescript), cloned into a pET45b vector (Novagen) and overexpressed in E. coli BL21(DE3) cells (Novagen) as a N-terminally His6-tagged protein. Bacterial cultures were grown in Luria-Bertani broth or M9 minimum media supplemented with 15NH4Cl and 2H-labeled, 13C-labeled, or 13C-labeled glucose at 37 °C until an OD600 of 0.6. After addition of 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG), the protein was expressed for 16 h at 18 °C. Cells were harvested by centrifugation (6000×g, 20 min) and resuspended in 20 mM Tris pH 7.5, 100 mM NaCl, 20 mM imidazole, 1 mM NaN3 and 1 mM DTT. The cells were lysed with an Emulsiflex (Avastin) and the soluble protein was purified over a Ni2+-NTA affinity column (GE Healthcare). Fractions containing SNX3 were pooled and applied on a Superdex-75 (GE Healthcare) size exclusion column and eluted with 20 mM sodium phosphate buffer pH 6.5, 100 mM NaCl, 1 mM DTT. The PX domain of SNX12 (SNX12-PX) was expressed in a pET45b vector and purified as with SNX3, except for the cleavage of the N-terminal histidine tag at an Enterokinase site prior to size exclusion chromatography. Full length SNX1 was cloned into a pGEX-5X vector and the GST-tagged protein was purified using a GST-trap column (GE Healthcare), cleaved with Factor Xa and purified on a Superdex-200 column (GE Healthcare). DNA sequences are shown in Supplementary Table 1. Mutations were generated with QuikChange Lightening (Stratagene) and verified by DNA sequencing.
NMR and structure calculation
NMR spectra of 15N, 15N/13C, and 2H/15N/13C labeled protein were acquired at 298 K on 600 and 800 MHz Varian Inova spectrometers equipped with 5 mm cryogenic probes. The spectra were processed with NMRpipe40 and analysed with CCPNMR41. The backbone 1H, 15N, and 13C resonances were sequentially assigned using standard 3D experiments11. Distance restraints were derived from 15N-edited NOESY-HSQC and 13C-edited NOESY-HSQC experiments (τmix = 100 ms) optimized for aromatic or aliphatic groups. Dihedral angle restraints were derived from DANGLE42. Hydrogen bond restraints were obtained by monitoring deuterium exchange of amide signals in HSQC spectra.
Structure calculations were performed by torsion angle dynamics in ARIA2.2 software43. For each of the 8 iterations, 100 structures were generated in vacuum, soaked in water and refined. The representative ensemble consisted of the 20 structures with the lowest experimental energies.
Interactions were monitored by stepwise addition of PI3P and stock solutions of DHPC or DPC micelles prepared with CHAPS in a 3:1 molar ratio. CSPs were calculated as Δ = [(ΔδH)2 + (0.15 ΔδN)2]1/2. Apparent Kd values were calculated by fitting the chemical shift changes (Δ) to Δ = Δmax(LT + PT + Kd − [(LT + PT + Kd)2−4LTPT]1/2) (2PT)−1, where LT and PT are the ligand and protein concentrations. Plots and fittings were carried out with Gnuplot.
Residues inserting into micelles were identified from PREs observed from samples containing a c8-PI3P:protein ratio of 2.5:1 and DHPC:CHAPS (32:11 mM). The PRE values were calculated as (Idia–Ipara) (Idia)−1, where Idia and Ipara are the SNX3 1H-15N-crosspeak intensities in the presence of micelles containing on average one diamagnetic dipalmitoyl phosphatidylcholine (DPPC) or paramagnetic 5-doxyl phosphatidylcholine (PC) molecule per micelle, respectively. Intermolecular and geometric restraints of SNX3 bound to micelles were extracted from a 13C-edited 3D NOESY experiment acquired in D2O (τmix = 100 ms) from a sample containing DPC:CHAPS (8:2.7 mM) and c4-PI3P (1.5 mM).
Experimentally driven docking
Structures of SNX3:PI3P:micelle complexes were generated according to established methods14 using c8-PI3P in the HADDOCK procedure44. The c8-PI3P structure and parameters were designed using XPLO2D45 and implemented in CNSsolve46. The headgroup of c8-PI3P was docked to SNX3 using CSP and intermolecular NOE distances as well as consistency with contacts found in SNX9 (pdb 2RAK), p40phox (pdb 1H6H), and Grd19p (pdb 1OCU) structures. Conserved hydrogen bonds between the sidechains of SNX3 and PI3P were implemented for Arg70…3-PO4, Lys95…1-PO4, and Arg118…4-OH/5-OH. A total of 800 SNX3 structures were calculated from which a subset of 200 were refined in explicit water. An ensemble of 20 of the lowest docking energy SNX3-PI3P structures was selected for micelle docking. Following rotational rigid body minimization of the complex, c8-PI3P was embedded into the micelle. The orientation between the acyl chains of PI3P was restrained and the depth of insertion delimited according to existing DPC micelle47 and 1-palmitoyl-2-oleoyl-PC (POPC) bilayer48 models. Distances between the methyl’s of PI3P were restrained to 11.4 Å based on dioleoyl-PC (DOPC) bilayer models to enforce realistic conformations. Distances from the center of the micelle to the acyl chain methyl were set to 9.71, and to 21.12 Å for the glycerol C3 atom, which corresponds to the average distance with the C8 and C1 atoms of DPC in micelles47. To handle the high temperature dynamics, the distances between guanidinium group of Arg118 and H4 and H5 atoms of PI3P were set to 3 Å. Minimizations were applied to orient the acyl chains, for rigid body insertion of the SNX3:PI3P complex into micelles and then relaxation of the system. The docking protocol yielded 200 structures which were refined in explicit water and ranked according to their energies.
The SNX3 and SNX12-PX samples were prepared in the buffer used for NMR experiments and their monomeric states were evidenced using a XL-I analytical ultracentrifuge (Beckman Coulter). The sedimentation coefficient (s) and distribution [c(s)] were determined from sedimentation velocity experiments with a two-sector cell at 129,024 × g for 17 h at 4 °C. The absorbance of the sample was measured at a wavelength of 280 nm throughout the cell. The partial specific volume of the protein, the viscosity and density of the buffer were calculated in SEDNTERP and used with the coefficient of sedimentation c(s) in the routine SEDFIT49, with approximate molecular weights of the species in solution being deduced from the simplified model of continuous c(M) distribution.
In order to assay lipid interactions POPC and PI3P (98:2 mol:mol) were mixed in chloroform and dried under a nitrogen flux and under high vacuum. Lipid mixtures (2 mM) were obtained by resuspension of the lipids into sodium phosphate 20 mM pH 7, 100 mM NaCl from which 75 μL were mixed with 25 μL of SNX3 (8 μM). After incubating the mixture for 10 min at 25 °C, the liposomes were pelleted, washed three times with the resuspension buffer and loaded on SDS-PAGE gels and bands for quantification by densitometry (Syngene). Liposomes (0.5 mM lipid) were prepared for surface plasmon resonance (SPR) detection in 50 mM Hepes pH 7.2, 50 mM NaCl. Unilamellar vesicles were produced by ten freeze and thaw cycles at 25 °C using liquid nitrogen, followed by extrusion of the vesicle through a 0.1 μm (Avanti) polycarbonate membrane. SPR measurements were carried out on a Biacore 3000 instrument (GE Healthcare). Vesicles were coated on a L1 sensor chip (GE Healthcare) at 5 μL min−1. A reference lane coated with POPC alone was used to measure the specific response for lanes coated with PI3P. Equilibrium measurements were obtained by injecting 85 μL of protein at 3–5 μL min−1. For kinetic measurements, the flow was increased to 30 μL min−1 and 40 μL of protein were injected. Surfaces were regenerated by injections of 10 μL of 10 mM NaOH at 100 μL min−1.
Cell culture and transfection
HeLa-MZ cells, which were provided by Marino Zerial (MPI-CBG, Dresden), and HeLa50 cells were maintained as described. HeLa cells are not on the list of commonly misidentified cell lines maintained by the International Cell Line Authentication Committee. Our HeLa-MZ cells were authenticated by Microsynth (CH), which revealed 100 % identity to the DNA profile of the cell line HeLa (ATCC® CCL-2™) and 100 % identity over all 15 autosomal STRs to the Microsynth’s reference DNA profile of HeLa. Cells are mycoplasma negative as tested by GATC Biotech (Germany). Mutations were introduced in wt SNX3-GFP5 by QuikChange site-directed mutagenesis (Stratagene) using as primers: R70A, 5′-gaatctactgttagaagagcatacagtgactttgaatgg-3′ and 5′-ccattcaaagtcactgtatgctcttctaacagtagattc-3′; S72A, 5′-gaatctactgttagaagaagatacgccgactttgaatggctgcgaagtgaa-3′, and 5′-ttcacttcgcagccattcaaagtcggcgtatcttcttctaacagtagattc-3′; S72E, 5′-gaatctactgttagaagaagatacgaggactttgaatggctgcgaagtgaa-3′ and 5′-ttcacttcgcagccattcaaagtcctcgtatcttcttctaacagtagattc-3′. Cells were transfected transiently with plasmid DNA using FuGENE HD (Promega) or TransIT-X2 Dynamic Delivery System (Mirus) according to the manufacturer’s recommendations.
HeLa-MZ transfected with wt SNX3-GFP or the indicated mutants were starved for 4 h and then incubated in the presence of EGF (100 ng mL−1) for the indicated time periods. Cells were then fixed and stained with antibodies to the EGF receptor, followed by fluorescently labeled secondary antibodies. The EGF receptor was then quantified by automated microscopy using the ImageExpress microscope (Molecular Devices) after segmentation of the granular structures containing the EGFR signal51, using Custom Module Editor™ from MetaXpress™. The obtained mask is then applied to the EGF receptor fluorescent micrographs and the integrated intensity (sum of all pixels intensity) is extracted for each cell. To quantify the distribution of GFP-tagged SNX3 wt and mutants, the endosomes were segmented using the signal of the EGF receptor endocytosed for 15 min., The endosomes containing a significant amount of GFP signal (≥2 × cytosolic background) were filtered over the entire population of structures containing the EGF receptor, in order to establish the mask for endosomes positive for both the EGF receptor and SNX3-GFP. This mask was applied to the GFP micrographs and the integrated intensity (sum of all pixels intensity) was extracted for each cell. Under all conditions, approximately 1000 cells were analysed per condition.
Antibodies and reagents
Mouse anti-GFP (mix of clones 7.1 and 13.1, Roche) (1:1000), rabbit anti-Vps35 (GTX108058, GeneTex) (1:1000), mouse anti-EGFR antibody (BD Transduction Laboratories) (1:200), HRP-labeled secondary antibodies (Sigma-Aldrich) (1:5000), rabbit-anti EEA1 antibody (Enzo) (1:100), fluorescently labeled secondary antibodies (Jackson ImmunoResearch Laboratories) (1:200), Hoechst (1:2500), Human EGF (Sigma). Mouse antibodies against SNX3 (C-16 Santa Cruz Biotechnology) (1:200).
UV circular dichroism
Proteins were dialyzed against 10 mM potassium phosphate buffer, pH 7.2. Data were acquired on a Jacso J-810 spectropolarimeter using a HELLMA cuvette with an optical path length of 0.2 mm. Signals were obtained by averaging 8 scans measured between 190 and 260 nm with a 0.2 nm increment at a sampling rate of 50 nm min−1.
HeLa MZ cells at 60–70% confluency were transiently transfected with indicated GFP constructs; 18 h later cells were scraped in lysis buffer [30 mM Tris–HCl pH 7.4, 1 mM EDTA, 0.5% NP40 and Protease/Phosphatase Inhibitor Cocktail (CST)] and lysed. GFP-Trap MA beads (Chromotek) were used for GFP precipitation. After 45 min incubation at 4 °C with the cleared cell lysates, beads were washed with lysis buffer without detergents. Proteins attached to the beads were eluted by incubating the beads at 95 °C in 2× Laemmli buffer, and subjected to gel electrophoresis and western blotting.
HeLa cells expressing SNX3-GFP were lysed and SNX3-GFP was immunoprecipitated using GFP-trap beads. The immunoprecipitate was resolved by SDS gel electrophoresis, digested with trypsin and analysed by LC-ESI-MS/MSmass spectrometry using a linear trap quadrupole Orbitrap Velos Pro (Thermo Scientific, San Jose, CA, USA) equipped with a nanoAcquity system (Waters, Milford, MA, USA). Hits were generated from raw data using the EasyProt software platform52.
SNX3′s coordinates have been deposited in the Protein Data Bank, with accession code 2MXC, and its NMR chemical shifts are available in BMRB Entry 25402. The authors declare that all other data supporting the findings of this study are available within the paper and its supplementary information files or from the corresponding author upon reasonable request.
Harterink, M. et al. A SNX3-dependent retromer pathway mediates retrograde transport of the Wnt sorting receptor Wntless and is required for Wnt secretion. Nat. Cell Biol. 13, 914–923 (2011).
van Weering, J. R. T., Verkade, P. & Cullen, P. J. SNX-BAR-mediated endosome tubulation is co-ordinated with endosome maturation. Traffic 13, 94–107 (2012).
van Weering, J. R. T. et al. Molecular basis for SNX-BAR-mediated assembly of distinct endosomal sorting tubules. EMBO J. 31, 4466–4480 (2012).
Xu, Y., Hortsman, H., Seet, L., Wong, S. H. & Hong, W. SNX3 regulates endosomal function through its PX-domain-mediated interaction with PtdIns(3)P. Nat. Cell Biol. 3, 658–666 (2001).
Pons, V. et al. Hrs and SNX3 functions in sorting and membrane invagination within multivesicular bodies. PLoS Biol. 6, 1942–1956 (2008).
Harrison, M. S. et al. A mechanism for retromer endosomal coat complex assembly with cargo. Proc. Natl Acad. Sci. USA 111, 267–272 (2014).
Lucas, M. et al. Structural mechanism for Cargo recognition by the retromer complex. Cell 167, 1623–1635.e14 (2016).
Hierro, A. et al. Function architecture of the retromer cargo-recognition complex. Nature 449, 1063–1067 (2007).
Bean, B. D. M., Davey, M. & Conibear, E. Cargo selectivity of yeast sorting nexins. Traffic 18, 110–122 (2017).
Zhou, C. Z. et al. Crystal structure of the yeast phox homology (PX) domain protein Grd19p complexed to phosphatidylinositol-3-phosphate. J. Biol. Chem. 278, 50371–50376 (2003).
Overduin, M., Rajesh, S., Gruenberg, J. & Lenoir, M. Secondary structure and 1H, 13C, 15N resonance assignments of the endosomal sorting protein sorting nexin 3. Biomol. NMR Assign. 9, 355–358 (2015).
Kufareva, I. et al. Discovery of novel membrane binding structures and functions. Biochem. Cell Biol. 92, 555–563 (2014).
Kutateladze, T. G. et al. Multivalent mechanism of membrane insertion by the FYVE domain. J. Biol. Chem. 279, 3050–3057 (2004).
Dancea, F., Kami, K. & Overduin, M. Lipid interaction networks of peripheral membrane proteins revealed by data-driven micelle docking. Biophys. J. 94, 515–524 (2008).
Bravo, J. et al. The crystal structure of the PX domain from p40phox bound to phosphatidylinositol 3-phosphate. Mol. Cell 8, 829–839 (2001).
Pylypenko, O., Lundmark, R., Rasmuson, E., Carlsson, S. R. & Rak, A. The PX-BAR membrane-remodeling unit of sorting nexin 9. Embo. J. 26, 4788–4800 (2007).
Blom, N., Gammeltoft, S. & Brunak, S. Sequence and structure-based prediction of eukaryotic protein phosphorylation sites. J. Mol. Biol. 294, 1351–1362 (1999).
Brombacher, E. et al. Rab1 guanine nucleotide exchange factor SidM is a major phosphatidylinositol 4-phosphate-binding effector protein of Legionella pneumophila. J. Biol. Chem. 284, 4846–4856 (2009).
Shiromizu, T. et al. Identification of missing proteins in the neXtProt database and unregistered phosphopeptides in the phosphositeplus database as part of the chromosome-centric human proteome project. J. Proteome Res. 12, 2414–2421 (2013).
Weber, C., Schreiber, T. B. & Daub, H. Dual phosphoproteomics and chemical proteomics analysis of erlotinib and gefitinib interference in acute myeloid leukemia cells. J. Proteom. 75, 1343–1356 (2012).
Hoffert, J. D., Pisitkun, T., Wang, G., Shen, R. F. & Knepper, M. A. Quantitative phosphoproteomics of vasopressin-sensitive renal cells: regulation of aquaporin-2 phosphorylation at two sites. Proc. Natl Acad. Sci. USA 103, 7159–7164 (2006).
Villen, J., Beausoleil, S. A., Gerber, S. A. & Gygi, S. P. Large-scale phosphorylation analysis of mouse liver. Proc. Natl Acad. Sci. USA 104, 1488–1493 (2007).
Cozier, G. E. et al. The phox homology (PX) domain-dependent, 3-phosphoinositide-mediated association of sorting nexin-1 with an early sorting endosomal compartment is required for its ability to regulate epidermal growth factor receptor degradation. J. Biol. Chem. 277, 48730–48736 (2002).
Koharudin, L. M., Furey, W., Liu, H., Liu, Y. J. & Gronenborn, A. M. The phox domain of sorting nexin 5 lacks phosphatidylinositol 3-phosphate (PtdIns(3)P) specificity and preferentially binds to phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2). J. Biol. Chem. 284, 23697–23707 (2009).
Niu, Y. et al. PtdIns(4)P regulates retromer-motor interaction to facilitate dynein-cargo dissociation at the trans-Golgi network. Nat. Cell Biol. 15, 417–429 (2013).
Xu, J. et al. Structure of sorting nexin 11 (SNX11) reveals a novel extended phox homology (PX) domain critical for inhibition. J. Biol. Chem. 288, 16598–16605 (2013).
Mas, C. et al. Structural basis for different phosphoinositide specificities of the PX domains of sorting nexins regulating G-protein signaling. J. Biol. Chem. 289, 28554–28568 (2014).
Huang, H. S. et al. Snx14 regulates neuronal excitability, promotes synaptic transmission, and is imprinted in the brain of mice. PLoS One 9, e98383 (2014).
Bodenmiller, B. et al. PhosphoPep–a phosphoproteome resource for systems biology research in Drosophila Kc167 cells. Mol. Syst. Biol. 3, 139 (2007).
Zhai, B., Villen, J., Beausoleil, S. A., Mintseris, J. & Gygi, S. P. Phosphoproteome analysis of Drosophila melanogaster embryos. J. Proteome Res. 7, 1675–1682 (2008).
Chi, A. et al. Analysis of phosphorylation sites on proteins from Saccharomyces cerevisiae by electron transfer dissociation (ETD) mass spectrometry. Proc. Natl Acad. Sci. USA 104, 2193–2198 (2007).
Sridhara, V., Marchler-Bauer, A., Bryant, S. H. & Geer, L. Y. Automatic annotation of experimentally derived, evolutionarily conserved post-translational modifications onto multiple genomes. Database 2011, bar019 (2011).
Rolland, D. et al. Global phosphoproteomic profiling reveals distinct signatures in B-cell non-Hodgkin lymphomas. Am. J. Pathol. 184, 1331–1342 (2014).
Mertins, P. et al. Proteogenomics connects somatic mutations to signalling in breast cancer. Nature 534, 55–62 (2016).
Palacios-Moreno, J. et al. Neuroblastoma tyrosine kinase signaling networks involve FYN and LYN in endosomes and lipid rafts. PLoS Comput. Biol. 11, e1004130 (2015).
Klammer, M. et al. Phosphosignature predicts dasatinib response in non-small cell lung cancer. Mol. Cell Proteom. 11, 651–668 (2012).
Stuart, S. A. et al. A phosphoproteomic comparison of B-RAFV600E and MKK1/2 inhibitors in Melanoma cells. Mol. Cell Proteom. 14, 1599–1615 (2015).
Hornbeck, P. V., Chabra, I., Kornhauser, J. M., Skrzypek, E. & Zhang, B. PhosphoSite: a bioinformatics resource dedicated to physiological protein phosphorylation. Proteomics 4, 1551–1561 (2004).
Wang, Y. T. et al. An informatics-assisted label-free quantitation strategy that depicts phosphoproteomic profiles in lung cancer cell invasion. J. Proteome Res. 9, 5582–5597 (2010).
Delaglio, F. et al. NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 6, 277–293 (1995).
Vranken, W. F. et al. The CCPN data model for NMR spectroscopy: development of a software pipeline. Proteins 59, 687–696 (2005).
Cheung, M. S., Maguire, M. L., Stevens, T. J. & Broadhurst, R. W. DANGLE: a Bayesian inferential method for predicting protein backbone dihedral angles and secondary structure. J. Magn. Reson. 202, 223–233 (2010).
Rieping, W. et al. ARIA2: automated NOE assignment and data integration in NMR structure calculation. Bioinformatics 23, 381–382 (2007).
Dominguez, C., Boelens, R. & Bonvin, A. M. HADDOCK: a protein-protein docking approach based on biochemical or biophysical information. J. Am. Chem. Soc. 125, 1731–1737 (2003).
Kleywegt, G. J. & Jones, T. A. Databases in protein crystallography. Acta Crystallogr. D 54, 1119–1131 (1998).
Brunger, A. T. et al. Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D 54, 905–921 (1998).
Tieleman, D. P., van der Spoel, D. & Berendsen, H. J. C. Molecular dynamics simulations of dodecylphosphocholine micelles at three different aggregate sizes: Micellar structure and chain relaxation. J. Phys. Chem. B 104, 6380–6388 (2000).
Tieleman, D. P., Breed, J., Berendsen, H. J. & Sansom, M. S. Alamethicin channels in a membrane: molecular dynamics simulations. Faraday Discuss. 111, 209–246 (1998).
Schuck, P. Size-distribution analysis of macromolecules by sedimentation velocity ultracentrifugation and lamm equation modeling. Biophys. J. 78, 1606–1619 (2000).
Le Blanc, I. et al. Endosome-to-cytosol transport of viral nucleocapsids. Nat. Cell Biol. 7, 653–664 (2005).
Wallabregue, A. et al. Selective imaging of late endosomes with a pH-sensitive diazaoxatriangulene fluorescent probe. J. Am. Chem. Soc. 138, 1752–1755 (2016).
Gluck, F. et al. EasyProt—an easy-to-use graphical platform for proteomics data analysis. J. Proteom. 79, 146–160 (2013).
Higgins, D. G. & Sharp, P. M. CLUSTAL: a package for performing multiple sequence alignment on a microcomputer. Gene 73, 237–244 (1988).
Jurrus, E. et al. Improvements to the APBS biomolecular solvation software suite. Protein Sci. 27, 112–128 (2018).
The authors thank Cancer Research UK, PRISM EU Sixth Framework Program, MRC, Wellcome Trust, TMIC, Genome Canada, Canada Foundation for Innovation and Campus Alberta Innovates Program for funding (M.O.), the Swiss National Science Foundation, the Swiss Sinergia program, the Polish-Swiss Research Program (PSPB-094/2010), the NCCR in Chemical Biology and LipidX from the Swiss SystemsX.ch initiative (J.G.), the HWB-NMR staff for spectrometer access and discussions, and the Wellcome Trust for support of the UK’s national NMR facility.
The authors declare no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Lenoir, M., Ustunel, C., Rajesh, S. et al. Phosphorylation of conserved phosphoinositide binding pocket regulates sorting nexin membrane targeting. Nat Commun 9, 993 (2018). https://doi.org/10.1038/s41467-018-03370-1
This article is cited by
The cryo-EM structure of the SNX–BAR Mvp1 tetramer
Nature Communications (2020)
PX Domain-Containing Kinesin KIF16B and Microtubule-Dependent Intracellular Movements
The Journal of Membrane Biology (2020)
To degrade or not to degrade: mechanisms and significance of endocytic recycling
Nature Reviews Molecular Cell Biology (2018)
The lipid products of phosphoinositide 3-kinase isoforms in cancer and thrombosis
Cancer and Metastasis Reviews (2018)
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.