A Non-Canonical Interface of SNX9 PX Domain Selectively Sequesters PI(3,4)P2 Lipids, Protecting Them from Hydrolysis

A Non-Canonical Interface of SNX9 PX Domain Selectively Sequesters PI(3,4)P2 Lipids, Protecting Them from Hydrolysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A Non-Canonical Interface of SNX9 PX Domain Selectively Sequesters PI(3,4)P2 Lipids, Protecting Them from Hydrolysis Feng-Ching Tsai, Jeriann Beiter, Chieh-Ju Sung, Shan-Shan Lin, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6573900/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Plasma membrane remodeling processes are tightly regulated by the spatiotemporal distribution and dynamic conversion of phosphoinositidyl lipids (PIPs). This regulation is controlled by the recruitment of proteins such as sorting nexin 9 (SNX9), a key mediator of late-stage endocytosis and macropinocytosis. Using live cell imaging, in vitro reconstitution assays, and molecular dynamics simulations, we investigated how SNX9 distinguishes between PI(3,4)P 2 and PI(4,5)P 2 , and the physiological relevance of this selectivity. Our results revealed that during macropinocytic membrane ruffling, SNX9 is recruited in a spatiotemporally coordinated manner with PI(3,4)P 2 , but not with PI(4,5)P 2 . While SNX9 induces comparably weak mechanical remodeling on model membranes containing either PIP 2 species, it exhibits a clear selective binding to PI(3,4)P 2 , mediated by a non-canonical interface. Through mutational analysis of key residues involved in this sequestration, we further demonstrated that SNX9 protects PI(3,4)P 2 from hydrolysis. Together, these results reveal a previously unrecognized mechanism of SNX9-PIP 2 lipid interaction that underscores the pivotal role of SNX9 in coordinating membrane remodeling processes. Biological sciences/Biophysics/Membrane structure and assembly Biological sciences/Cell biology/Membrane trafficking/Membrane curvature Biological sciences/Biophysics/Computational biophysics Biological sciences/Cell biology/Membrane trafficking/Endocytosis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Introduction Cells rely on the dynamic and localized synthesis of phosphoinositide (PIP) lipids to generate specialized membrane regions that recruit PIP-associated peripheral proteins at precise locations and times. This intricate PIP synthesis regulation enables numerous membrane processes, especially in endocytic pathways, such as clathrin-mediated and clathrin-independent endocytosis, macropinocytosis, phagocytosis, and pinocytosis. During these endocytic processes, kinases and phosphatases catalyze PIPs to produce different PIP species sequentially. For example, in macropinocytosis, the enrichment of PIP lipids proceeds from PI(4,5)P 2 to PI(3,4,5)P 3 , then PI(3,4)P 2 , and finally PI(3)P, a marker of early endosomal membranes 1 – 3 . Similarly, clathrin-mediated endocytosis begins with PI(4,5)P 2 enrichment, followed by PI(3,4)P 2 , and then PI3P 4 , 5 . PI(4,5)P 2 is known to be abundant at the plasma membrane, where its local concentration often increases during endocytosis. In contrast, other PIPs like PI(3,4)P 2 are synthesized transiently and at much lower concentrations at endocytic sites, with PI(3,4)P 2 level being ~ 40-fold lower than PI(4,5)P 2 6 . To recognize and bind specific PIP lipids, many membrane-associated proteins have PIP-binding domains such as PX ( p hox h omology), PH ( p leckstrin h omology), FYVE ( F ab1, Y OTB, V ac1 and E EA1), and C2 domains 7 . These domains often display affinity for multiple PIP species. For instance, various PX domains bind both PI(4,5)P 2 and PI(3,4)P 2 , though with different affinities 8 . Discriminating between similar PIPs, such as PI(3,4)P 2 and PI(4,5)P 2 , is challenging due to their nearly identical headgroups, which share the same charges, sizes, and protonation constants (Fig. 1 A) 9 , 10 . This overlapping specificity of similar PIPs raises fundamental questions: How do protein domains achieve selective recognition of specific PIP lipids? How might PIP-binding proteins impact PIP conversions, possibly by affecting phosphatase accessibility and thus modulating endocytic processes? Addressing these questions is fundamental for deciphering the regulation of membrane dynamics in cells. The crystal structures of PIP-associated domains typically reveal a single PIP bound in a canonical binding pockets 11 – 14 . However, there is growing evidence that many protein domains interact with lipids in a promiscuous, multivalent manner 15 , 16 . Superstoichiometric lipid binding, where one protein domain can simultaneously bind multiple lipids of the same species, often involves additional secondary binding pockets or polybasic patches and motifs that promote such interactions 16 – 18 . For instance, recent molecular dynamics (MD) simulations have shown that domains like the FERM domain of ezrin can bind simultaneously to multiple PI(4,5)P 2 lipids, and the PH domain of GRP1, with at least two PIP 3 lipids 19 – 22 . However, concurrent with superstoichiometric binding, many peripheral membrane proteins contain PH and PX domains that are relatively weak binders yet still exhibit certain PIP selectivity 8 , 23 . How these PIP-binding domains achieve superstoichiometric yet selective PIP binding remains an open question in understanding the molecular mechanisms of lipid-protein interactions. In this study, to reveal how certain proteins achieve both superstoichiometric and selective PIP binding, we investigated how the PX-BAR domain of sorting nexin 9 (SNX9) interacts with PI(4,5)P 2 and PI(3,4)P 2 , two key PIP lipids in SNX9-mediated endocytosis and macropinocytosis 24 , 25 . SNX9 consists of an N-terminal Src Homology 3 (SH3) domain, a long unstructured motif termed the “low complexity” (LC) domain, and a C-terminal PX-BAR unit, which comprises a PX domain paired with a B in/ A mphiphysin/ R vs (BAR) domain (Fig. 1 , B and C). BAR domains are known to bind to negatively charged lipids, including PIPs and phosphoserine (PS) lipids; moreover, many BAR domains have been shown to be able to sense and generate curved membranes 26 . SNX9 has been shown to play a prominent role in clathrin-mediated endocytosis and macropinocytosis, by facilitating plasma membrane deformation and by recruiting partners such as dynamin and actin nucleation promoting factor N-WASP 25 , 27 , 28 . While SNX9’s PX domain is thought to assist in targeting PIP-rich membranes, its PIP binding selectivity remains unclear 24 . Many studies have investigated SNX9’s PIP preference: some indicate PI(3,4)P 2 and PI(4,5)P 2 as predominant, while others suggest a preference for PI3K products such as PI(3)P, PI(3,4)P 2 , and PIP 3 8,29,30 . Other work indicates that SNX9 binds both PI(3,4)P 2 and PI(4,5)P 2 weakly, and is recruited to membranes through a general electrostatic mechanism 31 . While the high resolution SNX9 PX-BAR crystal structure indicates the binding of the short-chain PI(3)P to the canonical binding pocket of the PX domain, it does not clarify selective binding mechanism for other PIP 2 lipids (PDB ID: 2RAK 11 ). Unlike PI(3)P, which has a smaller charged headgroup, the more complex PIP 2 headgroups suggest that electrostatic and steric factors might allow the PX-BAR domain to achieve specific PIP binding, particularly with the additional consideration of superstoichiometric lipid binding. In this study, we conducted live-cell imaging to quantify the temporal recruitment of SNX9 to macropinocytic membrane ruffles in relation to PIP enrichment. We observed that SNX9 is recruited synchronously with PI(3,4)P 2 but is delayed relative to PI(4,5)P 2 . To reveal how SNX9 distinguishes between PI(3,4)P 2 and PI(4,5)P 2 , we performed in vitro reconstituted assays. We found that SNX9 binds both PIPs with comparable affinities, as well as generates and senses membrane curvature irrespective of the two PIP 2 species. Motivated by these findings, we next examined how SNX9 modulates PI(3,4)P 2 and PI(4,5)P 2 dynamics at the molecular scale. Atomistic MD simulations of SNX9 PX-BAR domain with model membranes revealed a non-canonical binding interface in the PX domain that selectively favors PI(3,4)P 2 over PI(4,5)P 2 , and drives PI(3,4)P 2 sequestration. We hypothesized that this selective sequestration may influence the conversion of PI(3,4)P 2 during macropinocytosis. Indeed, mutational analysis in cells confirmed that this non-canonical interface can interfere with the downstream phosphatase INPP4B’s ability to convert PI(3,4)P 2 into PI3P. Collectively, our findings suggest that the selectively sequestration of PI(3,4)P 2 by SNX9 may not only facilitate its recruitment to the plasma membrane, but also provides a “protective” effect, potentially preventing PI(3,4)P 2 from being prematurely converted into PI(3)P during macropinocytosis and endocytosis. Results SNX9’s Recruitment to Ruffling Membranes is Coincident with PI(3,4)P 2 but Not PI(4,5)P 2 To investigate the interplay between SNX9 and PI(4,5)P 2 or PI(3,4)P 2 during macropinocytosis, we conducted live-cell imaging to monitor their spatiotemporal distribution upon stimulation with platelet-derived growth factors (PDGF) in serum-starved cells. The initial stage of macropinocytosis involves membrane ruffle formation, a process where SNX9 plays a crucial role 32 . We performed time-lapse imaging of cells overexpressing fluorescently tagged SNX9 alongside fluorescent PH-domain probes for PI(4,5)P 2 (PLCd-PH-EGFP) or PI(3,4)P 2 (NES-EGFP-cPHx3) to track the recruitment and dynamics of each molecule at the sites of ruffle formation (Fig. 2 , A and C). Our results show that SNX9 co-localizes spatially with both PIP 2 reporters at PDGF-induced membrane ruffles (Fig. 2 , B and D). However, quantitative analysis of SNX9 and PIP 2 fluorescence intensity profiles during the first 60 seconds of ruffle formation shows distinct enrichment kinetics. Specifically, for PI(4,5)P 2 , we observe a delayed appearance of SNX9 fluorescence relative to PI(4,5)P 2 at the same membrane ruffle site (Fig. 2 E). In contrast, SNX9 and PI(3,4)P 2 fluorescence signals appear simultaneously, with no detectable lag (Fig. 2 F). The observed delayed, yet spatially colocalized, appearance of SNX9 relative to PI(4,5)P 2 is consistent with the established role of PI(4,5)P 2 as an upstream factor during endocytosis and macropinocytosis 1 . However, the simultaneous appearance of SNX9 with PI(3,4)P 2 , as opposed to the lag observed with PI(4,5)P 2 , suggests a potential selective association between SNX9 and PI(3,4)P 2 at macropinocytic ruffles. Our finding suggests that SNX9 displays a specificity for PI(3,4)P 2 over PI(4,5)P 2 . This observation is consistent with previous studies indicating that PI(3,4)P 2 promotes SNX9 membrane association at the late stage of clathrin-mediated endocytosis, thereby facilitating dynamin recruitment by SNX9 to drive membrane fission 31 , 33 . SNX9 is a Weakly Scaffolding BAR Protein that Senses and Generates a Wide Range of Membrane Curvatures on both PI(3,4)P 2 - and PI(4,5)P 2 -Containing Membranes Given that SNX9 is a BAR domain protein, we next asked how different PIP 2 species in membranes impacts the expected ability of SNX9 to sense and generate curved membranes. We purified full length SNX9 and labelled it with Alexa fluorophore dyes for detection by confocal microscopy. We first determined whether SNX9’s binding affinity to flat membranes differs between PI(4,5)P 2 and PI(3,4)P 2 . To do this, we used giant unilamellar vesicles (GUVs) with diameters of about 5 µm or larger. GUV membranes were composed of eggPC, supplemented with 10 mol% DOPS, 10 mol% DOPE, 15 mol% cholesterol, and 0.5 mol% Bodipy ceramide, along with either 8 mol% PI(4,5)P 2 or PI(3,4)P 2 . Upon incubating SNX9 with GUVs, we observed the formation of outward membrane tubules on the GUVs, confirming SNX9’s ability to deform membranes and generate membrane tubules, a characteristic feature of BAR domains (Fig. 3 , A and B). To estimate the binding affinities of SNX9 on PI(4,5)P 2 - and PI(3,4)P 2 -containing GUVs, we measured the surface density of SNX9 on GUV membranes as a function of its bulk concentration (Fig. 3 , C and D) 34 . By fitting to the Hill equation, we estimated dissociation constants of 31 nM for PI(4,5)P 2 and 47 nM for PI(3,4)P 2 . These comparable affinities are consistent with previous findings for the PX-BAR domain of SNX9 on small unilamellar vesicles 31 . To evaluate SNX9’s ability to sense membrane curvature, we generated cylindrical membrane nanotubes from SNX9-coated GUVs using optical tweezers, with tube radii that can be tuned by changing membrane tension via micropipette aspiration 34 (Fig. 3 , E and F). We quantified the enrichment of SNX9 on the nanotubes relative to the flat GUV membranes by calculating the sorting ratio, as previously described for other BAR domains 34 . Our data show that SNX9’s sorting ratio increases as the membrane tubes are thinner – that is, as curvature increases - within the experimentally accessible range of ∼ 0.0125 nm − 1 to ∼ 0.15 nm − 1 (corresponding to tubes radii of ~ 80 nm and 7 nm, respectively) (Fig. 3 , G and H). Furthermore, by comparing the sorting ratio of SNX9 at a relatively low surface density (on average 1% surface coverage, i.e. protein areal density 200 mm − 2 ), where enrichment effects are expected to be the most pronounced, we observed that SNX9 are enriched on both PI(4,5)P 2 and PI(3,4)P 2 membrane tubes, as indicated by the sorting ratios larger than 1 (Fig. 3 , G and H). Because the full-length SNX9 contains a PX-BAR domain, to analyze its sorting behavior, we applied the curvature mismatch model 35 , which has been shown to be more appropriate for full length BAR protein amphiphysin, and truncated BAR-PH domain of b2-centaurin than the spontaneous curvature model 36 . This analysis enables us to estimate the spontaneous curvature of membrane-bound SNX9 ( \(\:{\stackrel{-}{C}}_{p}\) ) and the associated elastic constant \(\:\stackrel{-}{\kappa\:}\) - which reflects the energetic cost of membrane deformation and indicates the strength of SNX9’s mechanical capacity to induce curvature. Our analysis indicates that the intrinsic curvatures of membrane-bound SNX9 for PI(4,5)P 2 and PI(3,4)P 2 containing membranes are comparable (1/ \(\:{\stackrel{-}{C}}_{p}\) ~ 6 nm and ~ 7 nm, respectively), as are their elastic constants ( \(\:\stackrel{-}{\kappa\:}\) ~ 4 k B T and ~ 5 k B T, respectively). These findings suggest that, at a macroscopic level, SNX9 has a similar sorting capacity and mechanical effect on membranes, regardless of the PIP 2 species. Moreover, the relatively low elastic constants imply that SNX9 is a weaker membrane scaffolding protein as compared to other BAR proteins such as amphiphysin, b2-centaurin, and IRSp53 36 . To gain molecular scale insights into how SNX9 organizes and deforms PI(4,5)P 2 - and PI(3,4)P 2 -containing membranes, we performed cryo-electron microscopy (cryo-EM). A heterogeneous solution of Large Unilamellar Vesicles (LUVs) with diameters ranging from 50 nm to 1 µm was prepared by resuspending dry lipid films composed of DOPC and 15% DOPS, supplemented with 10% PI(4,5)P 2 or PI(3,4)P 2 . Besides spherical vesicles, this protocol also yields tubulated vesicles in roughly 15% of the vesicles 37 . These vesicles were incubated with 1 mM SNX9 for at least 1 hour at room temperature before being plunge frozen onto EM grids. Our cryo-EM results revealed dense SNX9 binding along membrane tubes under both PIP 2 conditions, whereas the remainder of the liposome membranes exhibited a lower density of SNX9 (Figs. 4 , A and B, Left panel). Notably, unlike previous studies where BAR domains, such as endophilin, displayed a well-registered alignment on membrane tubes, SNX9 exhibits a more disorganized binding pattern (Fig. 4 , A and B, Left panel, arrows) 38 . This less ordered organization likely reflects the influence of the PX domain and disordered regions present in full-length SNX9. Although we attempted to purify the SNX9 BAR domain to further test this hypothesis, the isolated domain proved too unstable for performing experiments 11 . Quantitative analysis of tube diameters revealed that SNX9-decorated tubes had median diameters of 16.7 nm for PI(4,5)P 2 - containing membranes and 11.4 nm for PI(3,4)P 2 -containing membranes. In contrast, bare membranes exhibited significantly larger medians of 35.5 nm for PI(4,5)P 2 and 32.9 nm for PI(3,4)P 2 (Fig. 4 C). These data confirm that SNX9 can deform membranes and constrict tubes to diameters near 15 nm, consistent with the spontaneous curvature measured in our tube pulling experiments. We noted that to compare the intrinsic curvatures obtained by tube pulling experiments with those from Cryo-EM, a half-bilayer thickness (2 nm considering tube radius and 4 nm, diameter) should be added, resulting in a tube diameter of 16 nm and 18 nm for PI(4,5)P 2 and for PI(3,4)P 2 , respectively. For both PIP 2 conditions, we observed peanut-shaped vesicles characterized by a narrow neck-like region at their center. At these necks, the membranes appeared less defined and more diffuse than in the remainder of the vesicles, suggesting local SNX9 enrichment at these necks and reinforcing its role in membrane remodeling (Fig. 4 , A and B, Right panel). Notably, these neck regions exhibited a broader diameter distribution compared to the SNX9-decorated tubes, indicating that SNX9 functions as a weak scaffolding protein capable of accommodating a wide range of membrane curvatures (Fig. 4 C). This observation aligns with our tube pulling experiments, where we observed a wide distribution of sorting ratios at given membrane curvatures (Fig. 3 , G and H). Of note, the stability of these neck-like structures was confirmed by overnight incubation experiments. Together, our Cryo-EM findings demonstrate that SNX9 preferentially associated with curved, tubular membranes in both PI(4,5)P 2 and PI(3,4)P 2 membranes. Moreover, SNX9 not only binds loosely to membrane tubes but also induces the formation of membrane necks, progressively transforming vesicles into peanut-shaped structures with pinched necks and ultimately into thin tubes. SNX9 PX-BAR Drives Distinct Distributions of PI(3,4)P and PI(4,5)P Lipids To explore the molecular mechanisms by which SNX9 interacts with membranes containing PI(4,5)P 2 and PI(3,4)P 2 , we conducted all-atom molecular dynamics (AA-MD) simulations of the dimeric SNX9 membrane remodeling region, including the PX-BAR domain (residues 200–595) with model membranes containing the following compositions: DOPC:DOPS 80:20; DOPC:DOPS:PI(3,4)P 2 80:15:5; and DOPC:DOPS:PI(4,5)P 2 80:15:5 (Fig. 5 A). The structure of the membrane remodeling region dimer was predicted using AlphaFold2, to facilitate the realistic inclusion of the C-terminal amphipathic helix and missing loops 39 . By probing the unbiased interactions between SNX9 PX-BAR and model membranes, we aimed to reveal whether there is a difference in the way that SNX9 interacts with PI(3,4)P 2 and PI(4,5)P 2 . Specifically, we sought to investigate how SNX9 PX-BAR influences the local dynamics of PIP 2 lipids upon binding. Although the timescales achievable by AA-MD (a total of 5 microseconds for each composition) limits direct measurement of kinetic rate constants, the simulations enable detailed analysis of nanoscopic phenomena such as lipid diffusion. To this end, we calculated the total distance traversed by all lipids and visualized the trajectory traces over time, using threshold distances as a color-coded guide (Fig. 5 , B-D, Supplementary Fig. 1-S4). These trace analyses demonstrate that SNX9 PX-BAR sequesters multiple PIP 2 lipids in its immediate vicinity. The sequestration effect of PIP 2 is most pronounced at the time-averaged positions of the PX domains and the center of the BAR domain (CBD) (Fig. 5 , B and C, Supplementary Fig. 1–4), with SNX9 PX-BAR overlay). These findings indicate that these regions are most strongly implicated in PIP 2 binding. While the PX domain has been previously identified as the primary site of PIP 2 binding, the CBD binding has not been reported in lipid binding. Notably, previous mutational assays have implicated the CBD region in autoinhibition of SNX9, but no lipid selectivity has been reported 40 . Structural predictions of SNX9 linker region implicated in autoinhibition position the linker region near the CBD. This alignment suggests that the positively charged CBD may mediate nonspecific interactions with PIP 2 (Supplementary Fig. 5). Given the plausible non-specific CBD-lipid binding, we focused on the PX domain to investigate the mechanism underlying SNX9’s selectivity for PI(3,4)P 2 and PI(4,5)P 2 . We observed that in the upper leaflet of the membranes where SNX9 PX-BAR is bound, there is super-stoichiometric sequestration of PIP 2 lipids (Fig. 5 , B and C). The averaged binding ratios are 3.6 ± 0.5 : 1 for PI(3,4)P 2 and 3.4 ± 0.8 : 1 for PI(4,5)P 2 (PIP 2 : SNX9 PX domain, Supplementary Table 1). These binding ratios, exceeding a value of unity, challenge the canonical lock-and-key interaction paradigm that assumes a single PIP 2 lipid fits into a tight binding pocket. Instead, they suggest that SNX9 PX domain can engage multiple lipids simultaneously through a multivalent binding mechanism. Supporting this, we found that only two out of ten of the individual PX domains in our simulations showed a binding occupancy of less than two lipids (Supplementary Table 1). The similarity in average binding ratios between PI(3,4)P 2 and PI(4,5)P 2 can be attributed to this multivalency, as having multiple binding subdomains that simultaneously bind to the membrane will be the greatest influence on average interactions. To determine whether the observed superstoichiometric binding is due to specific SNX9 PX-BAR and PIP 2 interactions rather than the intrinsic properties of the PIP 2 lipids, we analyzed the average dynamics of the lipids. As shown in Fig. 5 , B and C, we observe significant differences in the distance traveled by the PIP 2 lipids near the SNX9 PX-BAR domain compared to unbound PIP 2 lipids in the upper leaflet and those in the lower leaflet (Supplementary Fig. 6). We note that there is no significant difference in the total distance traveled between unbound PIP 2 lipids in the upper leaflet and in the lower leaflet (Supplementary Fig. 6). Additionally, no significant spatial variation of PIP 2 lipids in lower leaflet was observed, which indicates the absence of trans-bilayer coupling due to SNX9 PX-BAR binding. Comparing between PI(3,4)P 2 and PI(4,5)P 2 in the lower leaflet, there is no significant difference in the radius of gyration, a measure of lipid motion, indicating that the dynamics of the individual PIP 2 lipids are similar on average (Supplementary Fig. 7). Additionally, we quantified the average network connectivity of PIP 2 clustering as measure of average cluster size and found no significant differences either between the bilayers containing PI(3,4)P 2 versus PI(4,5)P 2 nor between the upper (SNX9 bound) and lower (no protein bound) leaflets of each bilayer (Supplementary Fig. 8). The lack of observable clustering among PIP 2 is consistent with previous findings, given the ionic composition of our simulations and the local sequestration effect observed 41 . Taken together, our results indicate that the observed lipid selective sequestration is not driven by intrinsic differences in PI(3,4)P 2 and PI(4,5)P 2 behaviors. Instead, the specificity arises from direct interactions between SNX9 PX-BAR and PIP 2 lipids, particularly those mediated by the PX domain. To further exclude explanations of the local sequestration based on membrane-mediated effects, we explored possible changes in the behaviors of DOPC and DOPS, as well as changes in overall membrane properties. We found no significant difference in the average radius of gyration and the spatial distributions of DOPC or DOPS in membranes containing or lacking PIP 2 lipids; put another way, the presence of PIP2 does not locally change the dynamics of other lipids (Supplementary Fig. 7, 9). Global membrane properties such as mean curvature and membrane packing defects also showed no significant differences between the three membrane compositions tested in the presence of SNX9 (Supplementary Fig. 10–11). These results indicate that neither the behavior of lipid species present in the membrane nor of the global membrane properties are altered in ways that could explain the observed PI(3,4)P 2 selectively sequestering by the PX domain of SNX9. The Membrane Binding Interface of SNX9 PX domain is Larger Than Previously Identified We next sought to identify PX-PIP 2 contacts to determine whether previously characterized canonical binding residues are responsible for PIP 2 selectivity. By analyzing the frequency of direct contacts between the alpha carbon atoms of residues in the PX domain and the central phosphorus of PI(3,4)P 2 and PI(4,5)P 2 (Fig. 6 A), we identified three distinct subdomains interacting with PIP 2 . These subdomains are referred to as the insert loop, the canonical pocket, and the fourth helix (Fig. 6 B). We note that the identification of these three subdomains aligns well with the average binding ratios of 3.6 and 3.4 for PI(3,4)P 2 and PI(4,5)P 2 , respectively, indicating that one or more of these subdomains can interact with multiple PIP 2 lipids. These subdomains are broad, ranging from 11 to 19 residues with strong membrane interactions, and extend beyond the canonical residues previously identified through sequence alignment 8 . Between the large membrane interface and observed superstoichiometric binding, the question of selectivity mechanisms should be addressed within each of the subdomains. The insert loop subdomain (residues 258–270) is rich in lysines and hydrophobic residues. Its conserved motif, x K x(S/T) KΦ xG ΦK S YI (where x represents non-conserved residues, and Φ represents a large hydrophobic residue), is shared by SNX9 and closely related members of the SNX family, SNX18 and SNX33 (Fig. 6 C, red dashed box, Supplementary Fig. 12). The flexibility and composition of this subdomain allows it to penetrate the membrane and interact with negatively charged lipids, anchoring the PX domain to the membrane. It is therefore not surprising that the insert loop shows the highest frequency of contacts with the central phosphate group of both PI(3,4)P 2 and PI(4,5)P 2 . Unlike the canonical pocket and fourth helix subdomains whose membrane interactions are mediated by specific residues, the PIP 2 interactions of the insert loop is broadly across its entire length, reflecting its critical role in PIP 2 interactions (Fig. 6 C). The canonical pocket subdomain (residues 286–293 and 309–320) has been the primary focus in studying PIP 2 selectivity and binding. Crystal structures of SNX9 PX-BAR domain, along with analogous structural studies of other PX domains, have suggested that a conserved set of residues within the pocket stabilize the PIP headgroup 8 , 11 . However, while the canonical pocket of SNX9 does interface with the PIP 2 lipids in the membrane, we observed that the conserved residues Tyr294, Arg296, and Lys300 do not interact with the membrane at all (designated with pink * in Fig. 6 C). Furthermore, compared to the insert loop, the canonical pocket exhibits greater variation in contact frequency with PIP 2 among the sequential residues in the loop, due to its more rigid tertiary structure. The fourth helix subdomain (residues 358–370), which is not traditionally considered as part of the PX domain, lies adjacent to the BAR domain and interacts frequently with the membrane (Fig. 6 C, purple dashed box). Like the canonical pocket, this forth helix subdomain exhibits variation in PIP 2 contact frequency among sequential residues due to its rigid secondary and tertiary structure. While the net charge of the fourth helix is + 2, its four positively charged residues (Lys363, Lys367, Arg368, and Arg370) are aligned along its membrane-facing surface. This alignment strongly favor binding to negatively charged lipids such as the PIPs and PS. The alignment of these positively charged residues along the membrane binding surface of the fourth helix suggests that its protruding sidechains could synergistically coordinate the large headgroup of a PIP 2 lipid. The SNX9 PX Fourth Helix Promotes PI(3,4)P Selectivity Of the three subdomains identified in this study, only the canonical pocket and the fourth helix interact with PIP 2 primarily through electrostatic interactions, specifically involving positively charged residues. Because PIP 2 lipids are strongly negatively charged, − 4 compared to -1 for PI and PS at physiological pH, and have a bulky phosphoinositide headgroup, the specific arrangements of positively charged residues in the canonical pocket and fourth helix are likely key for explaining how the SNX9 PX domain selectively interacts with PI(3,4)P 2 over PI(4,5)P 2 . The electrostatic map of the membrane binding face of SNX9 PX domain clearly shows the importance of these positively charged residues (Fig. 7 A). It also highlights the formation of two distinct pockets of charge surrounding the canonical pocket and the fourth helix. To investigate how these residues could cooperatively bind to the same PIP 2 lipid, we calculated a contact frequency for all positively charged residues based on the Bjerrum length, considering instances when two residues coordinate the same lipid, though not necessarily the same phosphate group (Fig. 7 B, Supplementary Fig. 13) 19 . This cooperative frequency based on electrostatic binding groups provides deeper insight into the specificity of lipid interactions. By identifying residues with cooperative self-contact frequencies greater than 1.5, we can determine those involved in major coordination events (Fig. 7 C). For the canonical pocket, the major coordinating residues were identical for both PI(3,4)P 2 and PI(4,5)P 2 , and include Arg286, Lys288, Lys313, and Arg318 (Fig. 7 D). However, for the fourth helix, we find a broader set of coordinating residues for PI(3,4)P 2 over PI(4,5)P 2 . Specifically, residues Lys360, Lys363, Lys366, Arg,367, Lys368 and Arg371 for PI(3,4)P 2 , and Lys363, Lys366, Arg367 and Arg371 for PI(4,5)P 2 , indicating stronger coordinating interactions for PI(3,4)P 2 (Fig. 7 E). Besides, we observe a stronger interaction network between all the residues in the fourth helix for PI(3,4)P 2 over PI(4,5)P 2 (Fig. 7 E). This reinforces the idea of a more robust coordination for PI(3,4)P 2 , and provides an explanation for how selectivity is maintained, despite superstoichiometric binding ratios. The expansive network of coordinating residues is capable of binding multiple PIP 2 lipids simultaneously. Furthermore, inter-domain contacts, particularly between the fourth helix and the canonical pocket subdomains, are stronger for PI(3,4)P 2 than for PI(4,5)P 2 (Supplementary Fig. 13). In addition to electrostatic differences, we note that the canonical pocket has a significantly smaller cavity volume of 745 Å 3 compared to the volume of the fourth helix subdomain of 905 Å 3 (Supplementary Fig. 14). Docking predictions further suggest that both PI(3,4)P 2 and PI(4,5)P 2 are more likely to bind to the fourth helix subdomain than the canonical pocket (Supplementary Fig. 14). Collectively, our findings indicate that SNX9 PX domain has a selectivity for PI(3,4)P 2 over PI(4,5)P 2 , through the fourth helix of the PX domain. Specific SNX9-PI(3,4)P 2 Interactions Protect PI(3,4)P 2 from Hydrolysis by INPP4B at the Plasma Membrane during Membrane Ruffling Based on our simulation results showing that SNX9 PX-BAR dimers sequester PI(3,4)P 2 , we hypothesized that SNX9 may influence the hydrolysis of PI(3,4)P 2 by INPP4B phosphatase. This hydrolysis, converting PI(3,4)P 2 to PI(3)P, is critical in macropinocytosis for macropinosome closure 5 . By monitoring PI(3,4)P 2 enrichment at the plasma membrane of cells overexpressing membrane-bound INPP4B (INPP4B-CAAX), we assessed whether SNX9 protects PI(3,4)P 2 from INPP4B-mediated hydrolysis through sequestration. Consistent with previous reports, overexpression of wild-type INPP4B (INPP4B-WT), but not the catalytic dead C842S mutant (INPP4B-CS), significantly reduces PI(3,4)P 2 enrichment at plasma membrane ruffles (Fig. 8 , A – C) 42 . Overexpression of SNX9-mScarlet restores PI(3,4)P 2 at the plasma membrane of cells expressing INPP4B-WT (Fig. 8 , B and D). This result highlights SNX9’s protective role in shielding PI(3,4)P 2 from hydrolysis by INPP4B. To reveal the functional relevance of SNX9’s PX interaction networks in coordinating PI(3,4)P 2 that our simulations identified, we performed specific mutations in the SNX9 PX domain, and assessed the consequences of PI(3,4)P 2 conversion in cellular assays with either functional or non-functional INPP4B phosphatase. We found that loss-of-function mutations in SNX9, K313A in the canonical pocket and K366/367A in the fourth helix, fail to rescue PI(3,4)P 2 enrichment in cells expressing INPP4B-WT, despite their proper localization at the plasma membrane (Fig. 8 , B and D). Conversely, gain-of-function mutations in SNX9, K366R and K366/368R both in the fourth helix, more effectively restored PI(3,4)P 2 enrichment the plasma membrane. These mutants recapitulated the protective capacity of wild-type SNX9 INPP4B-mediated hydrolysis (Fig. 8 D). Notably, in cells expressing the non-functional phosphatase, INPP4B C842S, PI(3,4)P 2 enrichment index is comparable among all SNX9 variants, indicating that the observed effects of SNX9 on PI(3,4)P 2 coordination are specific to the functional interaction with INPP4B (Fig. 8 D). Discussion We demonstrated that PI(3,4)P 2 and PI(4,5)P 2 exhibit distinct spatial distributions upon interacting with SNX9 PX-BAR domain. Although both lipids bind superstoichiometrically to SNX9 via the membrane-remodeling PX-BAR domain, our results revealed that the non-canonical Helix 4 interface in the PX domain is a key determinant of the differential PIP 2 localization. This finding provides mechanistic insight into how SNX9 can achieve functional lipid selectivity despite its high overall valency for PIP 2 lipids. Even though the selective binding effect is modest compared to the overall binding of the PX domain, it translates into the preferential spatiotemporal colocalization of SNX9 with PI(3,4)P 2 rather than PI(4,5)P 2 at macropinocytic ruffles in vivo . Given the generality of the underlying biophysical principles, SNX9’s selective interaction with PIP lipids likely represents a general mechanism for the spatiotemporal recruitment of peripheral membrane proteins in processes like macropinocytosis and endocytosis. As exemplified here by SNX9, many proteins with PIP selective domains, such as PX, PH, FYVE, and C2, might exploit a similar dual mode of PIP interaction. To further elucidate the mechanism behind SNX9’s Helix 4 selectivity for PI(3,4)P 2 over PI(4,5)P 2 , we characterized a non-canonical binding interface composed of a network of positively charged residues that coordinate the PIP 2 headgroup. We hypothesized and demonstrated with mutational assays that this Helix 4 interface acts as a molecular checkpoint, preventing the premature hydrolysis of PI(3,4)P 2 to PI(3)P by INPP4B during macropinocytosis (Fig. 9 ). Having a checkpoint to prevent the premature conversion of PI(3,4)P 2 to PI(3)P is especially critical at this juncture in the membrane remodeling process, just prior to membrane fission, to prevent stalled or failed internalization 33 . Using quantitative reconstituted assays, we also showed that full-length SNX9 exhibits membrane curvature sensing and remodeling capabilities, accommodating a wide range of membrane curvatures, independent of the specific PIP 2 species present. Although SNX9 functions as a relatively weak scaffolding BAR protein, its loosely organized assembly on membrane tubes may allow it to stabilize divers structures – from elongated tubes to constricted necks – which is essential for dynamic processes such as endocytosis and macropinocytosis. Collectively, our integrated in silico , in vitro and in vivo findings indicate that SNX9’s dual membrane binding domains - its BAR domain for curvature sensing and its PX domain for regulating local PI(3,4)P 2 concentration – work synergistically for membrane remodeling processes. This dual functionality reinforces SNX9’s central role in membrane dynamics and highlights a broadly applicable mechanism by which peripheral membrane proteins achieve selective recruitment and stabilization at the plasma membrane. Moreover, our Cyro-EM imaging reveals that SNX9-enriched membrane tubes exhibit a broad spectrum of membrane curvatures, with some comparable to those observed in clathrin-mediated endocytosis and dynamin-induced tubes 31 , 43 , suggesting that SNX9 remains membrane-associated throughout the course of remodeling events. Ultimately, SNX9’s ability to coordinate membrane remodeling during lipid conversion with the timely recruitment of downstream effectors via its SH3 domain, such as dynamin and N-WASP, provides a mechanistic basis to ensure coordinated protein recruitment during cellular internalization events. Given the critical role of membrane composition in coordinating membrane remodeling, and considering that enzymatic lipid conversion occurs on a faster timescale than micron-scale membrane remodeling, SNX9’s protective role for PI(3,4)P 2 against pre-mature hydrolysis to PI(3)P may serve as a temporal checkpoint. This checkpoint ensures that the remodeling process reaches a sufficient level of maturity before proceeding to the final stages of internalization and conversion into an endosomal body. Together, our findings underscore SNX9’s mechanistic role in membrane association and remodeling during lipid conversion, suggesting that SNX9 is both recruited by and orchestrates the maintenance of local PI(3,4)P 2 concentrations, as well as the timely recruitment of downstream effectors to facilitate membrane remodeling. While not the focus of our study, the previously observed hetero-dimerization of SNX9 with SNX18 could offer a further expanded repertoire of regulatory capabilities 44 . The magnitude of the PIP 2 selectivity of SNX9 PX-BAR domain can be partially attributed to the non-specific superstoichiometric binding of the PX-BAR to the membrane. Binding to multiple lipids simultaneously in non-specific ways can make it difficult to determine selective lipid binding of the PX domain, which may explain why previous in vitro literature has reported weak binding for the SNX9 PX domain 8 . Superstoichiometric binding is likely a general strategy to recruit peripheral proteins to the membrane surface, while specific binding may play a role in the retention of PIP lipids. We note that this does not imply that lipid composition is the exclusive mechanism of recruitment, as other membrane properties such as curvature also change significantly during membrane remodeling events and thus drive protein recruitment. Indeed, the previously reported interplay of the SNX1 PX domain with the BAR domain as partners in membrane recruitment reinforces the idea that curvature plays a likely role in membrane recruitment for other PX-BAR domains, including SNX9 45 . Given our results suggesting that SNX9 responds similarly to membranes of similar curvature containing either PI(3,4)P 2 or PI(4,5)P 2 , we suggest that curvature may be secondary to PIP 2 selectivity for SNX9 recruitment. In addition, the observed general reduction of lipid mobility around the SNX9 PX-BAR domain due to this superstoichiometric binding may be part of the mechanism of protein-mediated PIP 2 aggregation, which has been previously reported for other BAR proteins 46 – 48 . However, aggregation is a collective effect, and therefore further in silico studies should include a much larger membrane system with at least dozens of PX-BAR domains to investigate this phenomenon. The Helix 4 interface, identified as driving selectivity of SNX9 for PI(3,4)P 2 over PI(4,5)P 2 , provides the opportunity to reflect more broadly on our treatment of peripheral membrane protein binding modes. We have endeavored to emphasize here that even for lipid selective domains such as the PX domain, multiple lipids are simultaneously bound and so determining selectivity is not always a straightforward task, and requires a detailed look at the structure and dynamics of the domain of interest on a membrane. Sequence data is helpful for identifying conserved residues that may play an important role, but should be paired whenever possible with structural analysis. The large, negatively charged headgroup of the PIP 2 lipids define their interactions with peripheral membrane proteins, and distinguishes them particularly from other lipid classes; therefore, the electrostatic interaction mechanism proposed here, coupled with spatial information of the non-canonical binding interface (i.e., how the positively charged sidechains face the membrane and the larger cavity size of the non-canonical interface), is grounded by physical arguments and supported by the mutational assays in cells. While we primarily probed the electrostatic impact by mutating positively charged residues, further work could probe the effect of interface size by introducing bulky, neutral sidechains that would protrude into the adjoining cavity. However, the introduction of such residues could disrupt the protein tertiary structure or lead to insertion of these large neutral residues into the membrane, significantly altering the membrane binding mode as opposed to only probing PIP 2 selectivity. It remains unclear what signal eventually drives SNX9 to dissociate from the membrane and free PI(3,4)P 2 for conversion by INPP4B and other phosphatases. The change in Gaussian membrane curvature as a result of scission could be an important signal to induce a mechanical change in SNX9 organization from a loose organization on tubes to more shallow contacts on spherical vesicles, as has been demonstrated for other BAR domain containing proteins 49 , 50 . Altogether, it is striking how lipid selectivity of peripheral membrane proteins between two nearly identical lipids, PI(3,4)P 2 and PI(4,5)P 2 , arises from simple effects such as electrostatics and spatial structural arrangements despite their high valency, and further how selectivity drives the coordination of complex remodeling processes at the plasma membrane. We anticipate that these conclusions are not unique to SNX9 nor the plasma membrane, but rather provide insights into the biophysics behind selective membrane recruitment for a host of peripheral membrane proteins. Materials and Methods CELL EXPERIMENTS Cell culture, transfection and imaging. COS7 cells were maintained in DMEM supplemented with 10% FBS. Transfection was performed using the plasmid DNA of interest and TransIT (Mirus Bio) according to the manufacturer's instructions. For live-cell imaging, cells were visualized using a spinning-disc microscope (Carl Zeiss) in imaging medium consisting of phenol red-free DMEM supplemented with 10 mM HEPES and 10% FBS. Medium containing 50 ng/mL PDGF-BB was applied to cells to stimulate membrane ruffles. Images were captured every 2 seconds for 10–15 minutes. For fixed-cell imaging, transfected cells were seeded onto coverslips and fixed with 4% paraformaldehyde, followed by observation with an LSM700 confocal microscope (Carl Zeiss). Image quantification. To analyze the spatiotemporal distribution of PIPs and SNX9, a 1 µm² area was cropped from each membrane ruffle for quantification. The signals were normalized to the highest value and plotted over time. The maximal intensity of SNX9-RFP was set as 100% and aligned to 50 sec for synchronization in figure plotting. For PI(3,4)P 2 enrichment index quantification, the index is defined as the ratio between PI(3,4)P 2 -enriched membrane and resting plasma membrane. Images were processed with maximal intensity projection, and the contour of PI(3,4)P 2 -enriched membrane regions was manually defined. A similar area of the resting plasma membrane was also defined. The GFP intensity of PI(3,4)P 2 -enriched ruffles and the resting plasma membrane were measured, and the ratio was calculated.. Statistical analysis. Statistical analysis was conducted using GraphPad Prism 9.0. Data were analyzed with one-way ANOVA. A P-value of < 0.05 was considered statistically significant and is indicated as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. IN VITRO RECONSISTUTION EXPERIMENTS Reagents. Brain L-α-phosphatidylinositol-4,5-bisphosphate (PI(4,5)P 2 , 840046P), 1-heptadecanoyl-2-(5Z,8Z,11Z,14Z-eicosatetraenoyl)-sn-glycero-3-phospho-(1'-myo-inositol-3',4'-bisphosphate) (17:0–20:4 PI(3,4)P 2 , LM1903) 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[biotinyl(polyethyleneglycol)-2000] (DSPE-PEG(2000)-biotin, 880129P), L-α-phosphatidylcholine (Egg, Chicken) (EPC, 840051), 1,2-di-(9Z-octadécénoyl)- sn -glycéro-3-phosphoéthanolamine (18:1 (Δ9-Cis) DOPE, 850725P), 1,2-di-(9Z-octadécènoyl)- sn -glycéro-3-phospho-L-sérine (18:1 PS DOPS, 840035), and Cholesterol (700000P), were purchased from Avanti Polar Lipids. BODIPY-TR-C5-ceramide, (BODIPY TR ceramide, D7540), BODIPYFL C5-hexadecanoyl phosphatidylcholine (HPC*, D3803), and Alexa Fluor 488 C5-Maleimide (AX488) were purchased from Invitrogen. Streptavidin-coated polystyrene beads (SVP-30-5) were purchased from Spherotech. β-casein from bovine milk (> 98% pure, C6905) and other reagents were purchased from Sigma-Aldrich. Culture-Inserts 2 Well for self-insertion were purchased from ibidi (Silicon open chambers, 80209). Protein purification and labelling. Full length SNX9 construct was a gift from Christian Wunder (Institut Curie, France). The full length SNX9 sequence was cloned into the pProEX HTB plasmid (Invitrogen) between the BamHI and NotI cloning sites, adding a 6xHis tag and rTEV protease cleavage site to the N-terminus of the protein. The plasmid was transformed into Rosetta™ 2(DE3) pLysS (Novagen) and expressed in 2YT medium for 4 hours at 37°C with 0.2 mM IPTG. Cells were lysed by sonication in 50 mM Hepes pH 7.4, 300 mM NaCl, 20 mM Imidazole supplemented with complete EDTA-free protease inhibitor cocktail (Roche). Sample was centrifuged at 20,000 xg and proteins purified using TALON® Metal Affinity Resins (Takara Bio). Protein was eluted with 50 mM Hepes pH 7.4, 300 mM NaCl, 300 mM Imidazole. Proteins were further purified over the Superdex 200 10/300GL column (GE Healthcare) in PBS, 0,5mM EDTA pH 8,0. Pure protein was the labelled with 1:1 protein to dye ratio with Alexa 488 C5 Malemide. Labelled samples were dialyzed into PBS, 0,5mM EDTA pH 8,0, 10% Glycerol, and stored at -80°C. Buffer compositions . The salt buffer inside GUVs, named I-buffer, was 50 mM NaCl, 20 mM sucrose and 20 mM Tris pH 7.5. The salt buffer outside GUVs, named O-buffer, was 60 mM NaCl and 20 mM Tris pH 7.5. GUV preparation. GUVs was prepared using the polyvinyl alcohol (PVA) gel-assisted method 51 . A PVA solution (5% (w/w) of PVA in a 280 mM sucrose solution) was warmed up to 50°C before spreading on a coverslip, which was cleaned in advance by rinsing with ethanol and MilliQ water. The PVA-coated coverslip was dried in an oven at 60°C for 30 min. 5–10 µl of the lipid mixture (1 mg/mL in chloroform) was spread on the PVA-coated coverslip, followed by drying under vacuum for 30 min at room temperature. The PVA-lipid-coated coverslip was placed in a 10 cm cell culture dish and 0.5 mL of the I-buffer was added on the coverslip, and kept it stable for at least 45 min at room temperature to allow GUV to grow. Sample preparation and observation for measuring dissociation constants. GUVs were incubated with SNX9 at bulk concentrations depending on the designed experiments for at least 15 min at room temperature. Chamber coverslips were passivation with a β-casein solution at a concentration of 5 g.L − 1 for at least 5 min at room temperature. Experimental chambers were assembled by placing a silicon open chamber on a coverslip. Samples were observed using a Nikon C1 confocal microscope equipped with a X60 water immersion objective. Image analysis. Image analysis was performed by Fiji 52 . Florescence images were taken at the equatorial planes of GUVs using identical confocal microscopy settings. We measured the fluorescence intensities of SNX9 on spherical GUVs devoid of tubules. The background intensity of the AX488 channel was obtained by drawing a line with a width of 10 pixels perpendicularly across GUV membranes. We obtained the background intensity profile of the line with the x-axis of the profile be the length of the line and the y-axis, the averaged pixel intensity along the width of the line. The background intensity was obtained by calculating the mean value of the sum of the first 10 intensity values and the last 10 intensity values of the background intensity profile. To obtain SNX9 fluorescence intensity on the GUV membrane, we used membrane fluorescence signals to detect the contour of the GUV. A 10 pixel wide band centered on the contour of the GUV was used to obtain the SNX9 intensity profile of the band where the x-axis of the profile is the length of the band and the y-axis, the averaged pixel intensity along the width of the band. SNX9 fluorescence intensity was then obtained by calculating the mean value of the intensity values of the SNX9 intensity profile, following by subtracting the background intensity. We measured SNX9 surface density on GUV membranes (number of proteins per unit area) by using a previously established procedure 53 . We related the fluorescence intensity of AX488 to that of a fluorescent lipid (BODIPY FL-C5-HPC, named HPC*). We measure fluorescence intensity of HPC* on GUV membranes at a given HPC* membrane fraction. The surface density of the protein on membranes is \(\:{n}_{protein}={n}_{{HPC}^{*}}/\left(\frac{{I}_{AX488}}{{I}_{{HPC}^{*}}}\right)\) , where \(\:\frac{{I}_{AX488}}{{I}_{{HPC}^{*}}}\) is the factor accounting for the fluorescence intensity difference between HPC* and AX488 at the same bulk concentration under identical image acquisition condition. The area density of HPC*, \(\:{\varPhi\:}_{{HPC}^{*}}\) , can be related to its fluorescence intensity, \(\:{I}_{{HPC}^{*}}^{vesicle}\) , by measuring fluorescence intensities of GUVs composed of DOPC supplemented with different molar ratios of HPC* (0.04–0.16 mole%) and assuming lipid area per lipid is 0.7 nm 2 (1120–4480 HPC* per µm 2 ). As such, \(\:{n}_{{HPC}^{*}}=A\times\:{I}_{{HPC}^{*}}^{vesicle}\) , where \(\:A\) is a constant depending on the illumination setting in the microscope. We then obtained the surface density of the protein as \(\:{n}_{protein}=(A\times\:{I}_{protein}^{vesicle})/(\frac{{I}_{AX488}}{{I}_{{HPC}^{*}}}\times\:{n}^{*})\) , where \(\:{n}^{*}\) is the degree of labeling for the protein of interest. Finally, we obtained the surface fraction of the protein \(\:{\varPhi\:}_{protein}=\:{n}_{protein}\times\:{a}_{protein}\) , where \(\:{a}_{protein}\) is the area of a single protein on membranes, \(\:{a}_{BAR\:domain}\cong\:50\) nm 2 35 . To obtain the dissociation constant of SNX9 on GUV membranes, we fitted the surface densities of SNX9 \(\:{\varphi\:}_{v},\) determined from the fluorescence signals, as a function of SNX9 bulk concentration \(\:{C}_{bulk}\) to $$\:{\varphi\:}_{v}=\frac{{\varphi\:}_{max}\times\:{{C}_{bulk}}^{n}}{{K}_{d}+{{C}_{bulk}}^{n}}$$ 1 where \(\:{\varphi\:}_{max}\) is the maximum surface density, \(\:{K}_{d}\) is the dissociation constant and \(\:n\) is the Hill coefficient. Tube pulling experiments. The experiments were performed on a setup with a Nikon C1 confocal microscope equipped with a X60 water immersion objective, micromanipulators for positioning micropipettes and optical tweezers as previously described 53 . To pull a tube, a GUV was held by a micropipette, brought into contact with a streptavidin-coated bead trapped by the optical tweezers, followed by moving away from the bead. Tube radius \(\:R\) was measured by the ratio of lipid fluorescence intensity on the tube and on the GUV as \(\:{R={R}_{c}^{TR}\times\:({I}_{tube}/{I}_{vesicle})}_{membrane}\) , where \(\:{R}_{c}^{TR}=200\pm\:50\) nm is the previously obtained calibration factor for using BODIPY TR ceramide as lipid fluorescence reporter in the same setup by performing a linear fit of membrane fluorescence ratio \(\:{({I}_{tube}/{I}_{vesicle})}_{membrane}\) and lipid radii \(\:R\) measured by \(\:R=\:f/(4\pi\:\sigma\:\) ), where \(\:f\) is the force applied by the optical tweezers to hold the tubes and \(\:\sigma\:\:\) the membrane tension tuned by the micropipette holding the GUVs 35 , 53 . To obtain protein/membrane fluorescence intensity in tube pulling experiments, we defined a rectangular region of interest (ROI) around GUV membranes and around the membrane tube such that the membrane/tube were horizontally located at the center of the ROI. We then obtained an intensity profile along the vertical direction of the ROI by calculating the mean fluorescence intensity of each horizontal line of the rectangle. To account for protein fluorescence outside the GUVs, the background protein intensity was obtained by calculating the average value of the mean of the first 15 intensity values from the top and from the bottom of ROI. A similar procedure was used the membrane background. The protein/membrane fluorescence intensities were obtained by subtracting the background intensity from the maximum intensity value in the intensity profile. SNX9 sorting data analysis. We fit the sorting data using the curvature mismatch model 36 . The free energy of the system is $$\:{F}_{mismatch}^{tube}=2\pi\:RL\left[\frac{\kappa\:}{2{R}^{2}}+\frac{\stackrel{-}{\kappa\:}}{2}{{\varphi\:}_{t}\left(\frac{1}{R}-\stackrel{-}{{C}_{p}}\right)}^{2}+{f}_{s}+{f}_{m}\right]$$ 2 Where R is the tube radius, L is the tube length, \(\:\stackrel{-}{\kappa\:}\) is the bending modulus of the protein-bound membrane, \(\:\kappa\:\) is the bending modulus of the protein-free membrane, \(\:\stackrel{-}{\kappa\:}\) is an elastic coefficient penalizing mismatch between protein and membrane curvature, \(\:{\varphi\:}_{t}\) is the protein areal fraction on the tube, \(\:\stackrel{-}{{C}_{p}}\) is a phenomenological coefficient related to the membrane-bound protein’s intrinsic curvature, and \(\:{f}_{s}\) and \(\:{f}_{m}\) are the energy densities of membrane stretching and protein mixing entropy on membranes. At equilibrium the chemical potentials of the lipids and proteins on the GUV and on the tube are equal, thus an implicit dependence of \(\:{\varphi\:}_{v}\) (protein area fraction on the GUV) on the tube curvature can be written as $$\:\frac{{\varphi\:}_{t}}{{\varphi\:}_{v}}{\left(\frac{1-{\varphi\:}_{v}}{1-{\varphi\:}_{t}}\right)}^{{a}_{p}/{a}_{l}}=\text{e}\text{x}\text{p}\left[\frac{\stackrel{-}{\kappa\:}{a}_{p}}{{k}_{B}T}\left(\frac{\stackrel{-}{{C}_{p}}}{R}-\frac{1}{2{R}^{2}}\right)\right]$$ 3 where \(\:{a}_{p}\) and \(\:{a}_{l}\) is the membrane areas occupied by a membrane bound protein and a lipid, respectively. \(\:S={\varphi\:}_{t}/{\varphi\:}_{v}\) is the sorting ratio and \(\:C=1/R\) the tube curvature. By fitting the sorting data with this equation, one can obtain \(\:\stackrel{-}{\kappa\:}\) and \(\:\stackrel{-}{{C}_{p}}\) . Cryo-electron microscopy experiments. A heterogeneous solution of LUVs was generated by resuspension in buffer (60 mM NaCl and 20 mM Tris pH 7.5) of a dried lipid film composed of 80% DOPC, 10% DOPS and either 10% PI(4,5)P2 or 10% PI(3,4)P2 (molar ratios). 1 µM of SNX9 was incubated with the lipid suspensions at 0.1 mg.mL − 1 at room temperature for an hour. The samples were vitrified on copper holey lacey grids (Ted Pella) using an automated device (EMGP, Leica) by blotting the excess sample on the opposite side from the droplet of sample for 4 s in a humid environment (90% humidity). Imaging was performed by a Glacios microscope (Thermofisher) running at 200 kV and equipped with a Falcon IVi direct detector (Thermofisher). ATOMISTIC MOLECULAR DYNAMICS The initial structure of the truncated SNX9 PX-BAR dimer (residues 201–595) was generated with AlphaFold2 39 to model missing loops and the N-terminal amphipathic helix. The predicted structure was compared to previously solved structures (PDB IDs: 2RAI, 2RAK), and found to have only small deviations from each structure (Fig. 1 ) 11 . The PX-BAR dimer was oriented parallel to the membrane and translated to initially be 3nm above the membrane surface. Three sets of molecular dynamics simulations were conducted on membranes that were initially 22 nm by 22 nm in the xy plane, with compositions of 80:20 mole percent DOPC:DOPS, 80:15:5 mole percent DOPC:DOPS:PI(4,5)P 2 , and 80:15:5 mole percent DOPC:DOPS:PI(3,4)P 2 , and solvated with 0.15M KCl; five replicates of each membrane composition were conducted. Atomistic systems were generated with the CHARMM-GUI platform 54 , 55 , and simulated with the CHARMM-36m force field using GROMACS 2021.5 56 for 1000ns per replicate in the constant NPT ensemble after standard equilibration, for a total of 5 µs per system. Analysis was conducted over the last 750ns of each replicate in Python with MDAnalysis 57 . Visualization was conducted with ChimeraX 58 and Visual Molecular Dynamics 59 . Declarations Acknowledgments We thank Olena Pylypenko and Korbinian Liebl for insightful discussion and Christian Wunder for SNX9 plasmid. F.-C.T. and P.B. is a member of the CNRS consortium Approches Quantitatives du Vivant (AQV), Labex Cell(n)Scale (ANR-11-889 LABX0038) and Paris Sciences et Lettres (ANR-10-IDEX-0001-02). The authors greatly acknowledge the Cell and Tissue Imaging (PICT-IBiSA), Institut Curie, member of the national infrastructure France-BioImaging (https://ror.org/01y7vt929) supported by the French National Research Agency (ANR-24-INBS-0005 FBI BIOGEN). PB team is supported by the Fondation pour la Recherche Médicale (FRM) (FRM EQU202003010307) and by the European Union (ERC, PushingCell, #101071793). Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. Simulations were performed using computing resources provided by the University of Chicago Research Computing Center (RCC), the Department of Defense High Performance Computing cluster (HPCMP), and the National Science Foundation ACCESS cluster. Author contributions: JRB, Y-WL, GAV, and F-CT designed the initial project. JRB, CJS, SSL, NdV, AB and F-CT performed experiments and analyzed results with feedback from Y-WL, GAV and PB. JM and SA purified proteins. JRB and F-CT wrote the original draft with inputs and revisions from AB, PB, Y-WL and GAV. Conceptualization: JRB, Y-WL, GAV, and F-CT Methodology: JRB and F-CT Investigation: JRB, CJS, SSL, NdV, AB, F-CT Resources: JM, SA Visualization: JRB, F-CT Supervision: Y-WL, GAV, F-CT Writing—original draft: JRB, F-CT Writing—review & editing: JRB, F-CT, AB, JM, SA, Y-WL, GAV, PB Funding acquisition: FCT, Y-WL, GAV, PB Competing interests: The authors declare that they have no competing interests References Araki N, Egami Y, Watanabe Y, Hatae T (2007) Phosphoinositide metabolism during membrane ruffling and macropinosome formation in EGF-stimulated A431 cells. Exp Cell Res 313:1496–1507. https://doi.org/10.1016/j.yexcr.2007.02.012 Maekawa M et al (2014) Sequential breakdown of 3-phosphorylated phosphoinositides is essential for the completion of macropinocytosis. Proc Natl Acad Sci USA 111:E978–987. https://doi.org/10.1073/pnas.1311029111 Gillooly DJ, Raiborg C, Stenmark H (2003) Phosphatidylinositol 3-phosphate is found in microdomains of early endosomes. Histochem Cell Biol 120:445–453. https://doi.org/10.1007/s00418-003-0591-7 Di Paolo G, De Camilli P (2006) Phosphoinositides in cell regulation and membrane dynamics. Nature 443:651–657. https://doi.org/10.1038/nature05185 Posor Y, Jang W, Haucke V (2022) Phosphoinositides as membrane organizers. Nat Rev Mol Cell Biol 23:797–816. https://doi.org/10.1038/s41580-022-00490-x Ray J, Sapp DG, Fairn GD (2024) Phosphatidylinositol 3,4-bisphosphate: Out of the shadows and into the spotlight. Curr Opin Cell Biol 88:102372. https://doi.org/10.1016/j.ceb.2024.102372 Overduin M, Kervin TA (2021) The phosphoinositide code is read by a plethora of protein domains. Expert Rev Proteom 18:483–502. https://doi.org/10.1080/14789450.2021.1962302 Chandra M et al (2019) Classification of the human phox homology (PX) domains based on their phosphoinositide binding specificities. Nat Commun 10:1528. https://doi.org/10.1038/s41467-019-09355-y Kooijman EE, Gericke A (2014) Physical chemistry and biophysics of polyphosphoinositide mediated lipid signaling. Chem Phys Lipids 182:1–2. https://doi.org/10.1016/j.chemphyslip.2014.06.004 Kooijman EE, King KE, Gangoda M, Gericke A (2009) Ionization properties of phosphatidylinositol polyphosphates in mixed model membranes. Biochemistry 48:9360–9371. https://doi.org/10.1021/bi9008616 Pylypenko O, Lundmark R, Rasmuson E, Carlsson SR, Rak A (2007) The PX-BAR membrane-remodeling unit of sorting nexin 9. EMBO J 26:4788–4800. https://doi.org/10.1038/sj.emboj.7601889 Ceccarelli DFJ et al (2007) Non-canonical interaction of phosphoinositides with pleckstrin homology domains of Tiam1 and ArhGAP9. J Biol Chem 282:13864–13874. https://doi.org/10.1074/jbc.M700505200 Xu T et al (2020) Molecular basis for PI(3,5)P2 recognition by SNX11, a protein involved in lysosomal degradation and endosome homeostasis regulation. J Mol Biol 432:4750–4761. https://doi.org/10.1016/j.jmb.2020.06.010 Zhou C-Z et al (2003) Crystal structure of the yeast Phox homology (PX) domain protein Grd19p complexed to phosphatidylinositol-3-phosphate. J Biol Chem 278:50371–50376. https://doi.org/10.1074/jbc.M304392200 Kerr D, Suwatthee T, Maltseva S, Lee KY (2024) C. Binding equations for the lipid composition dependence of peripheral membrane-binding proteins. Biophys J 123:885–900. https://doi.org/10.1016/j.bpj.2024.02.031 Moravcevic K, Oxley CL, Lemmon MA (2012) Conditional peripheral membrane proteins: facing up to limited specificity. Structure 20:15–27. https://doi.org/10.1016/j.str.2011.11.012 Lemmon MA (2008) Membrane recognition by phospholipid-binding domains. Nat Rev Mol Cell Biol 9:99–111. https://doi.org/10.1038/nrm2328 Falke JJ, Ziemba BP (2014) Interplay between phosphoinositide lipids and calcium signals at the leading edge of chemotaxing ameboid cells. Chem Phys Lipids 182:73–79. https://doi.org/10.1016/j.chemphyslip.2014.01.002 Ehret T et al (2023) FERM domains recruit ample PI(4,5)P2s to form extensive protein-membrane attachments. Biophys J 122:1325–1333. https://doi.org/10.1016/j.bpj.2023.02.027 Yamamoto E et al (2020) Multiple lipid binding sites determine the affinity of PH domains for phosphoinositide-containing membranes. Sci Adv 6:eaay5736. https://doi.org/10.1126/sciadv.aay5736 Yamamoto E, Kalli AC, Yasuoka K, Sansom MSP (2016) Interactions of Pleckstrin Homology Domains with Membranes: Adding Back the Bilayer via High-Throughput Molecular Dynamics. Structure 24, 1421–1431 https://doi.org/10.1016/j.str.2016.06.002 Le Huray KIP, Wang H, Sobott F, Kalli AC (2022) Systematic simulation of the interactions of pleckstrin homology domains with membranes. Sci Adv 8:eabn6992. https://doi.org/10.1126/sciadv.abn6992 Singh N et al (2021) Redefining the specificity of phosphoinositide-binding by human PH domain-containing proteins. Nat Commun 12:4339. https://doi.org/10.1038/s41467-021-24639-y Yarar D, Surka MC, Leonard MC, Schmid SL (2008) SNX9 activities are regulated by multiple phosphoinositides through both PX and BAR domains. Traffic 9:133–146. https://doi.org/10.1111/j.1600-0854.2007.00675.x Bendris N, Schmid SL (2017) Endocytosis, metastasis and beyond: multiple facets of SNX9. Trends Cell Biol 27:189–200. https://doi.org/10.1016/j.tcb.2016.11.001 Simunovic M, Evergren E, Callan-Jones A, Bassereau P (2019) Curving cells inside and out: roles of BAR domain proteins in membrane shaping and its cellular implications. Annu Rev Cell Dev Biol 35:111–129. https://doi.org/10.1146/annurev-cellbio-100617-060558 Soulet F, Yarar D, Leonard M, Schmid SL (2005) SNX9 regulates dynamin assembly and is required for efficient clathrin-mediated endocytosis. Mol Biol Cell 16:2058–2067. https://doi.org/10.1091/mbc.E04-11-1016 Shin N et al (2007) Sorting nexin 9 interacts with dynamin 1 and N-WASP and coordinates synaptic vesicle endocytosis. J Biol Chem 282:28939–28950. https://doi.org/10.1074/jbc.M700283200 Daste F et al (2017) Control of actin polymerization via the coincidence of phosphoinositides and high membrane curvature. J Cell Biol 216:3745–3765. https://doi.org/10.1083/jcb.201704061 Gallop JL, Walrant A, Cantley LC, Kirschner MW (2013) Phosphoinositides and membrane curvature switch the mode of actin polymerization via selective recruitment of toca-1 and Snx9. Proc Natl Acad Sci USA 110:7193–7198. https://doi.org/10.1073/pnas.1305286110 Schöneberg J et al (2017) Lipid-mediated PX-BAR domain recruitment couples local membrane constriction to endocytic vesicle fission. Nat Commun 8:15873. https://doi.org/10.1038/ncomms15873 Yarar D, Waterman-Storer CM, Schmid SL (2007) SNX9 couples actin assembly to phosphoinositide signals and is required for membrane remodeling during endocytosis. Dev Cell 13:43–56. https://doi.org/10.1016/j.devcel.2007.04.014 Posor Y et al (2013) Spatiotemporal control of endocytosis by phosphatidylinositol-3,4-bisphosphate. Nature 499:233–237. https://doi.org/10.1038/nature12360 Sorre B et al (2012) Nature of curvature coupling of amphiphysin with membranes depends on its bound density. Proc Natl Acad Sci USA 109:173–178. https://doi.org/10.1073/pnas.1103594108 Prévost C et al (2015) IRSp53 senses negative membrane curvature and phase separates along membrane tubules. Nat Commun 6:8529. https://doi.org/10.1038/ncomms9529 Tsai F-C et al (2021) Comparing physical mechanisms for membrane curvature-driven sorting of BAR-domain proteins. Soft Matter 17:4254–4265. https://doi.org/10.1039/d0sm01573c Bertin A et al (2020) Human ESCRT-III polymers assemble on positively curved membranes and induce helical membrane tube formation. Nat Commun 11:2663. https://doi.org/10.1038/s41467-020-16368-5 Mim C et al (2012) Structural basis of membrane bending by the N-BAR protein endophilin. Cell 149:137–145. https://doi.org/10.1016/j.cell.2012.01.048 Jumper J et al (2021) Highly accurate protein structure prediction with AlphaFold. Nature 596:583–589. https://doi.org/10.1038/s41586-021-03819-2 Lo W-T et al (2017) A Coincidence Detection Mechanism Controls PX-BAR Domain-Mediated Endocytic Membrane Remodeling via an Allosteric Structural Switch. Dev Cell 43:522–529e524. https://doi.org/10.1016/j.devcel.2017.10.019 Han K, Gericke A, Pastor RW (2020) Characterization of Specific Ion Effects on PI(4,5)P2 Clustering: Molecular Dynamics Simulations and Graph-Theoretic Analysis. J Phys Chem B 124:1183–1196. https://doi.org/10.1021/acs.jpcb.9b10951 Gewinner C et al (2009) Evidence that inositol polyphosphate 4-phosphatase type II is a tumor suppressor that inhibits PI3K signaling. Cancer Cell 16:115–125. https://doi.org/10.1016/j.ccr.2009.06.006 Liu J, Alvarez FJD, Clare DK, Noel JK, Zhang P (2021) CryoEM structure of the super-constricted two-start dynamin 1 filament. Nat Commun 12:5393. https://doi.org/10.1038/s41467-021-25741-x Park J et al (2010) SNX18 shares a redundant role with SNX9 and modulates endocytic trafficking at the plasma membrane. J Cell Sci 123:1742–1750. https://doi.org/10.1242/jcs.064170 Liao Z, Si T, Kai J-J, Fan J (2024) Mechanism of Membrane Curvature Induced by SNX1: Insights from Molecular Dynamics Simulations. J Phys Chem B 128:2144–2153. https://doi.org/10.1021/acs.jpcb.3c07009 Tsai F-C et al (2022) Activated I-BAR IRSp53 clustering controls the formation of VASP-actin-based membrane protrusions. Sci Adv 8:eabp8677. https://doi.org/10.1126/sciadv.abp8677 Jarin Z, Pak AJ, Bassereau P, Voth GA (2021) Lipid-Composition-Mediated Forces Can Stabilize Tubular Assemblies of I-BAR Proteins. Biophys J 120:46–54. https://doi.org/10.1016/j.bpj.2020.11.019 Picas L et al (2014) BIN1/M-Amphiphysin2 induces clustering of phosphoinositides to recruit its downstream partner dynamin. Nat Commun 5:5647. https://doi.org/10.1038/ncomms6647 Boucrot E et al (2012) Membrane fission is promoted by insertion of amphipathic helices and is restricted by crescent BAR domains. Cell 149:124–136. https://doi.org/10.1016/j.cell.2012.01.047 Isas JM, Ambroso MR, Hegde PB, Langen J, Langen R (2015) Tubulation by amphiphysin requires concentration-dependent switching from wedging to scaffolding. Structure 23:873–881. https://doi.org/10.1016/j.str.2015.02.014 Weinberger A et al (2013) Gel-assisted formation of giant unilamellar vesicles. Biophys J 105:154–164. https://doi.org/10.1016/j.bpj.2013.05.024 Schindelin J et al (2012) Fiji: an open-source platform for biological-image analysis. Nat Methods 9:676–682. https://doi.org/10.1038/nmeth.2019 Sorre B et al (2009) Curvature-driven lipid sorting needs proximity to a demixing point and is aided by proteins. Proc Natl Acad Sci USA 106:5622–5626. https://doi.org/10.1073/pnas.0811243106 Jo S, Kim T, Iyer VG, Im W (2008) CHARMM-GUI: a web-based graphical user interface for CHARMM. J Comput Chem 29:1859–1865. https://doi.org/10.1002/jcc.20945 Feng S, Park S, Choi YK, Im WCHARMM-GUI (2023) Membrane Builder: Past, Current, and Future Developments and Applications. J Chem Theory Comput 19:2161–2185. https://doi.org/10.1021/acs.jctc.2c01246 Lee J et al (2016) CHARMM-GUI Input Generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM Simulations Using the CHARMM36 Additive Force Field. J Chem Theory Comput 12:405–413. https://doi.org/10.1021/acs.jctc.5b00935 Gowers R et al in Python in Science Conference. 98–105 (SciPy) Pettersen EF et al (2021) UCSF ChimeraX: structure visualization for researchers, educators, and developers. Protein Sci 30:70–82. https://doi.org/10.1002/pro.3943 Humphrey W, Dalke A, Schulten K (1996) VMD: visual molecular dynamics. J Mol Graph 14:33–38. https://doi.org/10.1016/0263-7855(96)00018-5 Additional Declarations There is NO Competing Interest. Supplementary Files JRBeiteretalSI.pdf JRBeiter et al Supplemenary Information Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6573900","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":462934351,"identity":"6c639554-c794-4b53-a640-a0d8f607cbb7","order_by":0,"name":"Feng-Ching Tsai","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABCklEQVRIie3RP0vDQBjH8d9RaJYrWZ9B2rdwEkjbqW9Fl2RpcHUIcqVgF/+sBy6+BF/ClUCmYtcODu1SHBziIhmq2J4RpztXwfsuDw/ch2c4wOf7gwXSDM2bnbpmBbpWwvUPYXtOUUOiXwkagtOvsy4SXBfPyJ+O+uF9sKnzQXqrso2uIM6shD8mA5RbPlRrNuUlZWqVirmCGEoLGdE4jt5lwcVKsynalD1QgoJjJ6xXei+xYA2Z1B+UigPZQdgJ8WhtyFIy2bmkE0PgInwcA+XhCpvcdW7oWC22mF8JBwkWUYW8GIlloV/rt4teOEtaVX1uJ/vaZAZ9/4jJBYBWZUaona98Pp/vH/cJrTFVWss3EwgAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-6869-5254","institution":"Institut Curie","correspondingAuthor":true,"prefix":"","firstName":"Feng-Ching","middleName":"","lastName":"Tsai","suffix":""},{"id":462934352,"identity":"430f0fdc-f38c-4607-9755-8c6757ba76da","order_by":1,"name":"Jeriann Beiter","email":"","orcid":"","institution":"University of Chicago","correspondingAuthor":false,"prefix":"","firstName":"Jeriann","middleName":"","lastName":"Beiter","suffix":""},{"id":462934353,"identity":"f240b4dc-591e-49ea-9c69-ec65f184196a","order_by":2,"name":"Chieh-Ju Sung","email":"","orcid":"","institution":"National Taiwan University","correspondingAuthor":false,"prefix":"","firstName":"Chieh-Ju","middleName":"","lastName":"Sung","suffix":""},{"id":462934354,"identity":"01d9aaad-8134-4d22-bb20-79f9d23ce8c2","order_by":3,"name":"Shan-Shan Lin","email":"","orcid":"","institution":"National Taiwan University","correspondingAuthor":false,"prefix":"","firstName":"Shan-Shan","middleName":"","lastName":"Lin","suffix":""},{"id":462934355,"identity":"408edd1e-9a9b-4011-965a-fab0e2d87527","order_by":4,"name":"Nicolas de Vuono","email":"","orcid":"","institution":"Institut Curie","correspondingAuthor":false,"prefix":"","firstName":"Nicolas","middleName":"","lastName":"de Vuono","suffix":""},{"id":462934356,"identity":"8984309f-7d87-48d8-8065-1483f60f7cad","order_by":5,"name":"Senthil Arumugam","email":"","orcid":"https://orcid.org/0000-0001-6733-4679","institution":"Monash University","correspondingAuthor":false,"prefix":"","firstName":"Senthil","middleName":"","lastName":"Arumugam","suffix":""},{"id":462934357,"identity":"994c85f8-9187-4c93-8cdc-667dea835ded","order_by":6,"name":"John Manzi","email":"","orcid":"https://orcid.org/0000-0003-2574-4260","institution":"Institut Curie/CNRS","correspondingAuthor":false,"prefix":"","firstName":"John","middleName":"","lastName":"Manzi","suffix":""},{"id":462934358,"identity":"343c39f9-bd27-4333-b502-5dd572f28432","order_by":7,"name":"Aurélie Bertin","email":"","orcid":"https://orcid.org/0000-0002-3400-6887","institution":"Institut Curie","correspondingAuthor":false,"prefix":"","firstName":"Aurélie","middleName":"","lastName":"Bertin","suffix":""},{"id":462934359,"identity":"d01d70ff-e868-4bb8-b46a-3b6156072cd1","order_by":8,"name":"Patricia Bassereau","email":"","orcid":"https://orcid.org/0000-0002-8544-6778","institution":"Institut Curie","correspondingAuthor":false,"prefix":"","firstName":"Patricia","middleName":"","lastName":"Bassereau","suffix":""},{"id":462934360,"identity":"6f66e0d5-d7fb-4884-a795-fbc5539dea14","order_by":9,"name":"Ya-Wen Liu","email":"","orcid":"https://orcid.org/0000-0003-0180-4142","institution":"National Taiwan University","correspondingAuthor":false,"prefix":"","firstName":"Ya-Wen","middleName":"","lastName":"Liu","suffix":""},{"id":462934361,"identity":"13ca3e2d-be5e-402f-b418-c93f7e3e7615","order_by":10,"name":"Gregory Voth","email":"","orcid":"https://orcid.org/0000-0002-3267-6748","institution":"The University of Chicago","correspondingAuthor":false,"prefix":"","firstName":"Gregory","middleName":"","lastName":"Voth","suffix":""}],"badges":[],"createdAt":"2025-05-01 19:15:43","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6573900/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6573900/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":83682498,"identity":"7eaeed14-3e6d-47c4-bc9f-d685fb9b98ea","added_by":"auto","created_at":"2025-05-30 16:30:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":386246,"visible":true,"origin":"","legend":"\u003cp\u003eStructural information of sorting nexin 9 (SNX9), PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. (a) Stereoprojected line structures of the headgroups of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e (\u003cem\u003eTop\u003c/em\u003e) and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (\u003cem\u003eBottom\u003c/em\u003e), with the C3 and C4 or C4 and C5 inositol carbons (respectively) labelled to demonstrate the location of the phosphate functional groups. Line structures obtained from LIPID MAPS® structure database. (b) Predicted structure of the membrane remodeling domain of SNX9 dimer viewed from the side (\u003cem\u003eTop\u003c/em\u003e) and looking top-down (\u003cem\u003eBottom\u003c/em\u003e). The predicted structure has an RMSD from that of the apo crystal structure of 1.196 Å (PDB ID: 2RAI) and of the PI(3)P bound crystal structure of 1.271 Å (PDB ID: 2RAK). (c) Schematic of domains of SNX9, in sequence order. The range from residues 200-595 is collectively known as the membrane remodeling domain. SH3 – SRC Homology 3; AH – amphipathic helix; PX – Phox homology; BAR – Bin/Amphiphysin/Rvs domain.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/311f29cb54d2c1681c0b1304.png"},{"id":83682753,"identity":"afcfddf3-2e9f-489b-a1a2-415bedd231b0","added_by":"auto","created_at":"2025-05-30 16:38:21","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":694559,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSNX9 recruitment to macropinocytic ruffles is coincident with PI(3,4)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e rather than PI(4,5)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e.\u003c/strong\u003e \u003cstrong\u003e(a)\u003c/strong\u003e Representative overlay image of SNX9 (magenta) and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e fluorescent probe (cyan) signals. The white inset box highlights the ruffle shown in (B). \u003cstrong\u003e(b)\u003c/strong\u003e Time series evolution of the signals of SNX9 (magenta) and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e probe (cyan) from a macropinocytic ruffle over the first 60 seconds of ruffle formation. \u003cstrong\u003e(c)\u003c/strong\u003e Representative overlay image of SNX9 (magenta) and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e fluorescent probe (cyan) signals. The white inset box highlights the ruffle shown in (D). \u003cstrong\u003e(d)\u003c/strong\u003e Time series evolution of the signal of SNX9 (magenta) and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e probe (cyan) from a macropinocytic ruffle over the first 60 seconds of ruffle formation. \u003cstrong\u003e(e)\u003c/strong\u003e Relative intensity of fluorescent signals for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e reporter and SNX9 from time series, quantified over 12 ruffles. \u003cstrong\u003e(f)\u003c/strong\u003e Relative intensity of fluorescent signals for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and SNX9 from time series, quantified over 11 ruffles. Data represent mean ± standard error of the mean. Scale bars, 5 μm (A and C), 1 μm (B and D).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/b9dd0c902a1db635c5d80fab.png"},{"id":83681802,"identity":"44c5318d-b8fc-4576-bb25-c91b5654a43e","added_by":"auto","created_at":"2025-05-30 16:22:21","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":246567,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSNX9 generates membrane tubes and senses membrane curvature with comparable affinities on PI(3,4)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e and PI(4,5)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e-GUVs.\u003c/strong\u003e \u003cstrong\u003e(a and b)\u003c/strong\u003e GUVs exhibiting outward tubes generated by SNX9. Green, SNX9 and Magenta, BODIPY TR lipid. \u003cstrong\u003e(c and d)\u003c/strong\u003e Analysis of SNX9’s binding on GUVs. Data fitted to the Hill equation (Equation. 1, Materials and Methods) to obtain the dissociation constant \u0026nbsp;and the Hill coefficient n=1.18 for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e GUVs and \u0026nbsp;and n=1.55 for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e GUVs. Black lines represent the fitted regression model, and dashed black lines, the bounds of the 95% confidence interval.\u003cstrong\u003e \u003c/strong\u003eFor each bulk concentration, PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, N = 4, 4, 4, 1, 4, 2, 3, 1, 3 independent experiments, n = 81, 78, 70, 32, 73, 61, 68, 21, 41 GUVs; PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, N = 1, 2, 2, 2, 3, 3, 1, 3, 3, 1, 1 independent experiments, n = 30, 52, 64, 55, 79, 72, 34, 80, 75, 30, 28 GUVs.\u003cstrong\u003e (e and f)\u003c/strong\u003e Representative confocal image of GUVs with pulled tethers. Tube radius ~11 nm (E) and ~ 9 nm (F). \u003cstrong\u003e(g and h)\u003c/strong\u003e Sorting ratio as a function of tube curvature, \u003cem\u003eC=1/R\u003c/em\u003e. Protein coverage on GUVs, , 1% (orange for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and green for P(3,4)P\u003csub\u003e2\u003c/sub\u003e) and 3% (grey for both PIP\u003csub\u003e2\u003c/sub\u003e). By fitting the sorting curves to Equation 3, we obtained averaged 1/\u0026nbsp;6 nm and 7 nm for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, respectively, and averaged \u0026nbsp;4 \u0026nbsp;and 5 \u0026nbsp;for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, respectively. N = 8 and 2 independent experiments, n = 21 and 9 GUVs, for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, respectively. All scale bars, 5 μm.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/5189cc5702682072b373d095.png"},{"id":83682502,"identity":"8b52a283-87a6-48dd-a64d-b42f001d834b","added_by":"auto","created_at":"2025-05-30 16:30:21","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1226591,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSNX9 induces remodeling of membranes containing either PI(4,5)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2 \u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003eor PI(3,4)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e. (a and b)\u003c/strong\u003e Representative cryo-EM images of full length SNX9 incubated with membranes containing PI(4,5)P\u003csub\u003e2\u003c/sub\u003e or PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and with tube and neck-like structures indicated by white arrows. Scale bars, 100 nm. \u003cstrong\u003e(c)\u003c/strong\u003e Quantification of the diameters of observed membrane structures, including tubes with no apparent SNX9 signals, tubes with apparent SNX9 signals, and neck-like structures with apparent SNX9 signals. n = 28, 68, 31, 10, 6, 60 structures analyzed from data presented \u003cem\u003eTop\u003c/em\u003e to \u003cem\u003eBottom\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/72351ba0ebde56a081595cc7.png"},{"id":83682499,"identity":"89b6d271-5478-4599-b337-933e6a47d735","added_by":"auto","created_at":"2025-05-30 16:30:21","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":886758,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSNX9 PX-BAR locally changes PIP\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e lipid behavior in atomistic molecular dynamics simulations.\u003c/strong\u003e \u003cstrong\u003e(a)\u003c/strong\u003e Top (\u003cem\u003eLeft\u003c/em\u003e) and side (\u003cem\u003eRight\u003c/em\u003e) views of the initial simulation set up. The membrane is shown in a point style colored by atom type, while the SNX9 PX-BAR domain is represented in a cartoon style colored by secondary structure. Visualization performed with VMD. \u003cstrong\u003e(b)\u003c/strong\u003e\u0026nbsp;Representative trajectory trace of the last 750 ns (from 1μs trajectory replicate) of the PI(3,4)P\u003csub\u003e2\u003c/sub\u003e lipids in the upper (\u003cem\u003eTop\u003c/em\u003e) or lower (\u003cem\u003eBottom\u003c/em\u003e) membrane leaflet, colored by total distance traveled. Average position of the SNX9 PX-BAR are shown in wire representation on the upper leaflet. \u003cstrong\u003e(c)\u003c/strong\u003e\u0026nbsp;Representative trajectory trace of the last 750ns (from 1μs trajectory replicate) of the PI(4,5)P\u003csub\u003e2\u003c/sub\u003e lipids in the upper (\u003cem\u003eTop\u003c/em\u003e) or lower (\u003cem\u003eBottom\u003c/em\u003e) membrane leaflet, colored by total distance traveled. Average positions of the SNX9 PX-BAR are shown in wire representation on the upper leaflet. \u003cstrong\u003e(d)\u003c/strong\u003e\u0026nbsp;Histogram of total distances traveled over all replicates for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e \u003cem\u003e(Top)\u003c/em\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e \u003cem\u003e(Bottom)\u003c/em\u003e, with bars colored to match trajectory trace colorations in (B) and (C) as well as in Supplementary Fig. 1-4.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/20f28e49f2bddd6bd5b7f189.png"},{"id":83681795,"identity":"0a012d26-fc4f-4424-be03-0cc6a25c6eb6","added_by":"auto","created_at":"2025-05-30 16:22:21","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":396127,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistinct PIP\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e binding interfaces of the PX domain can be identified from atomistic molecular dynamics simulations of the SNX9 PX-BAR interacting with membranes containing either PI(3,4)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e or PI(4,5)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e. (a)\u003c/strong\u003e Metric for calculating interactions, based on a threshold of 10 Å between the alpha carbon of a residue and the central phosphorus of a PIP\u003csub\u003e2\u003c/sub\u003e lipid. \u003cstrong\u003e(b)\u003c/strong\u003e Cartoon depiction of the SNX9 PX domain, with approximate membrane binding areas highlighted with ovals; oval colors correspond to the boxes outlined in (c). \u003cstrong\u003e(c)\u003c/strong\u003e Tabulated log frequencies of interactions for PX domain residues shows clear groupings corresponding to membrane binding areas, designated as the insertion loop (red), the canonical pocket (yellow), and the fourth helix (purple). Blue stars designate canonical binding residues and pink stars designate additional conserved residues identified via homology, according to \u003csup\u003e8\u003c/sup\u003e.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/7313c90839144aae1cf573fd.png"},{"id":83682501,"identity":"0458c4cd-95da-472c-8454-1df875572d42","added_by":"auto","created_at":"2025-05-30 16:30:21","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":795517,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eElectrostatics play a primary role in PX-PIP\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e interactions, particularly in the helix four subdomain. (a)\u003c/strong\u003e Cartoon and electrostatic surface representation of the SNX9 PX domain, with blue indicating positively charged regions and red indicating negatively charged regions. \u003cstrong\u003e(b)\u003c/strong\u003e Illustration of Bjerrum metric, where a side chain nitrogen atom from an arginine or lysine residue within 7 Å of a PIP\u003csub\u003e2\u003c/sub\u003e phosphate group was considered as interacting. \u003cstrong\u003e(c)\u003c/strong\u003e Cartoon representations of the SNX9 PX domain, with super self-coordinating residues for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e (left) and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (right) highlighted in green and orange (respectively) with side chain stick representations. \u003cstrong\u003e(d)\u003c/strong\u003e PX domain interaction networks for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e interactions (\u003cem\u003eLeft\u003c/em\u003e) and PI(4,5)P\u003csub\u003e2 \u003c/sub\u003einteractions (\u003cem\u003eRight\u003c/em\u003e). The subdomains are outlined to highlight the interactions within a subdomain as well as interdomain contacts. \u003cstrong\u003e(e)\u003c/strong\u003e Fourth helix of the PX domain interaction networks for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e interactions (\u003cem\u003eLeft\u003c/em\u003e) and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e interactions (\u003cem\u003eRight\u003c/em\u003e).\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/1421d73b9f4d6b261a4cb38a.png"},{"id":83681804,"identity":"c1ab6300-82c6-4914-984e-08ef8340ecd8","added_by":"auto","created_at":"2025-05-30 16:22:21","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":1062433,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMutations in the SNX9 PX domain impacts its protective effect on PI(3,4)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e hydrolysis at membrane ruffles.\u003c/strong\u003e \u003cstrong\u003e(a)\u003c/strong\u003e Cartoon representations of PX domain, with mutations emphasized by highlighting the affected residue in red and showing the new sidechain. \u003cstrong\u003e(b)\u003c/strong\u003e Wild-type or \u003cstrong\u003e(c)\u003c/strong\u003e catalytically dead INPP4B-CAAX, C842S mutant (CS) and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e reporter were co-transfected into cells together with SNX9-mScarlet or empty vector. After 24 hour incubation in culture medium, cells were fixed and imaged with z-stack confocal microscopy. Projected images were shown. \u003cstrong\u003e(d)\u003c/strong\u003e PI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment index. The enrichment level of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e is derived by quantifying the fluorescent intensity fold change at plasma membrane between ruffle and non-ruffle area. Each data point represents one cell.\u003cstrong\u003e \u003c/strong\u003ePI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment index in cells expressing indicated SNX9-mScarlet variant together with INPP4B-WT or NPP4B-CS were quantified and compared. Data were analyzed with one-way ANOVA. ***, p \u0026lt; 0.01; ****, p\u0026lt;0.001.\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/9713da7dd78dcfed4cfa3f2d.png"},{"id":83681801,"identity":"c7bc27d3-4e93-4fc3-a53a-56521f58442a","added_by":"auto","created_at":"2025-05-30 16:22:21","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":182587,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSNX9 prevents the premature hydrolysis of PI(3,4)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003e to PI(3)P by INPP4B.\u003c/strong\u003e As membrane internalization progresses (from \u003cem\u003eLeft \u003c/em\u003eto \u003cem\u003eRight\u003c/em\u003e), PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (orange) enrichment at the plasma membrane is followed by the enrichment of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e (green), while SNX9 (pink) is recruited to the membrane (purple). Once at the membrane, SNX9 colocalizes with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, shielding it from hydrolysis by INPP4B. After scission of the newly formed vesicle from the plasma membrane, SNX9 dissociates, allowing for the conversion of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e to PI(3)P to occur.\u003c/p\u003e","description":"","filename":"image9.png","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/2b1459c928ebc2068571483f.png"},{"id":86048583,"identity":"b61f6c04-8407-49ae-a468-53e74ebad610","added_by":"auto","created_at":"2025-07-05 00:18:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7483167,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/417c5275-b0b6-444b-81e1-af5faee8f571.pdf"},{"id":83681805,"identity":"e3ae09fd-34b6-41ee-a24d-a95fa02038fe","added_by":"auto","created_at":"2025-05-30 16:22:21","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":3211486,"visible":true,"origin":"","legend":"JRBeiter et al Supplemenary Information","description":"","filename":"JRBeiteretalSI.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6573900/v1/48463debc22a90fdf02feb74.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"A Non-Canonical Interface of SNX9 PX Domain Selectively Sequesters PI(3,4)P2 Lipids, Protecting Them from Hydrolysis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCells rely on the dynamic and localized synthesis of phosphoinositide (PIP) lipids to generate specialized membrane regions that recruit PIP-associated peripheral proteins at precise locations and times. This intricate PIP synthesis regulation enables numerous membrane processes, especially in endocytic pathways, such as clathrin-mediated and clathrin-independent endocytosis, macropinocytosis, phagocytosis, and pinocytosis. During these endocytic processes, kinases and phosphatases catalyze PIPs to produce different PIP species sequentially. For example, in macropinocytosis, the enrichment of PIP lipids proceeds from PI(4,5)P\u003csub\u003e2\u003c/sub\u003e to PI(3,4,5)P\u003csub\u003e3\u003c/sub\u003e, then PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and finally PI(3)P, a marker of early endosomal membranes \u003csup\u003e\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Similarly, clathrin-mediated endocytosis begins with PI(4,5)P\u003csub\u003e2\u003c/sub\u003e enrichment, followed by PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and then PI3P \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. PI(4,5)P\u003csub\u003e2\u003c/sub\u003e is known to be abundant at the plasma membrane, where its local concentration often increases during endocytosis. In contrast, other PIPs like PI(3,4)P\u003csub\u003e2\u003c/sub\u003e are synthesized transiently and at much lower concentrations at endocytic sites, with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e level being ~\u0026thinsp;40-fold lower than PI(4,5)P\u003csub\u003e2\u003c/sub\u003e \u003csup\u003e6\u003c/sup\u003e. To recognize and bind specific PIP lipids, many membrane-associated proteins have PIP-binding domains such as PX (\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ep\u003c/span\u003ehox \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eh\u003c/span\u003eomology), PH (\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ep\u003c/span\u003eleckstrin \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eh\u003c/span\u003eomology), FYVE (\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eF\u003c/span\u003eab1, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eY\u003c/span\u003eOTB, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eV\u003c/span\u003eac1 and \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eE\u003c/span\u003eEA1), and C2 domains \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. These domains often display affinity for multiple PIP species. For instance, various PX domains bind both PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, though with different affinities \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Discriminating between similar PIPs, such as PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, is challenging due to their nearly identical headgroups, which share the same charges, sizes, and protonation constants (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA) \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. This overlapping specificity of similar PIPs raises fundamental questions: How do protein domains achieve selective recognition of specific PIP lipids? How might PIP-binding proteins impact PIP conversions, possibly by affecting phosphatase accessibility and thus modulating endocytic processes? Addressing these questions is fundamental for deciphering the regulation of membrane dynamics in cells.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe crystal structures of PIP-associated domains typically reveal a single PIP bound in a canonical binding pockets \u003csup\u003e\u003cspan additionalcitationids=\"CR12 CR13\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. However, there is growing evidence that many protein domains interact with lipids in a promiscuous, multivalent manner \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Superstoichiometric lipid binding, where one protein domain can simultaneously bind multiple lipids of the same species, often involves additional secondary binding pockets or polybasic patches and motifs that promote such interactions \u003csup\u003e\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. For instance, recent molecular dynamics (MD) simulations have shown that domains like the FERM domain of ezrin can bind simultaneously to multiple PI(4,5)P\u003csub\u003e2\u003c/sub\u003e lipids, and the PH domain of GRP1, with at least two PIP\u003csub\u003e3\u003c/sub\u003e lipids \u003csup\u003e\u003cspan additionalcitationids=\"CR20 CR21\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. However, concurrent with superstoichiometric binding, many peripheral membrane proteins contain PH and PX domains that are relatively weak binders yet still exhibit certain PIP selectivity \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. How these PIP-binding domains achieve superstoichiometric yet selective PIP binding remains an open question in understanding the molecular mechanisms of lipid-protein interactions.\u003c/p\u003e \u003cp\u003eIn this study, to reveal how certain proteins achieve both superstoichiometric and selective PIP binding, we investigated how the PX-BAR domain of sorting nexin 9 (SNX9) interacts with PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, two key PIP lipids in SNX9-mediated endocytosis and macropinocytosis \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. SNX9 consists of an N-terminal Src Homology 3 (SH3) domain, a long unstructured motif termed the \u0026ldquo;low complexity\u0026rdquo; (LC) domain, and a C-terminal PX-BAR unit, which comprises a PX domain paired with a \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eB\u003c/span\u003ein/\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eA\u003c/span\u003emphiphysin/\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eR\u003c/span\u003evs (BAR) domain (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, B and C). BAR domains are known to bind to negatively charged lipids, including PIPs and phosphoserine (PS) lipids; moreover, many BAR domains have been shown to be able to sense and generate curved membranes \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. SNX9 has been shown to play a prominent role in clathrin-mediated endocytosis and macropinocytosis, by facilitating plasma membrane deformation and by recruiting partners such as dynamin and actin nucleation promoting factor N-WASP \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. While SNX9\u0026rsquo;s PX domain is thought to assist in targeting PIP-rich membranes, its PIP binding selectivity remains unclear \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Many studies have investigated SNX9\u0026rsquo;s PIP preference: some indicate PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e as predominant, while others suggest a preference for PI3K products such as PI(3)P, PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and PIP\u003csub\u003e3\u003c/sub\u003e \u003csup\u003e8,29,30\u003c/sup\u003e. Other work indicates that SNX9 binds both PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e weakly, and is recruited to membranes through a general electrostatic mechanism \u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. While the high resolution SNX9 PX-BAR crystal structure indicates the binding of the short-chain PI(3)P to the canonical binding pocket of the PX domain, it does not clarify selective binding mechanism for other PIP\u003csub\u003e2\u003c/sub\u003e lipids (PDB ID: 2RAK \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e). Unlike PI(3)P, which has a smaller charged headgroup, the more complex PIP\u003csub\u003e2\u003c/sub\u003e headgroups suggest that electrostatic and steric factors might allow the PX-BAR domain to achieve specific PIP binding, particularly with the additional consideration of superstoichiometric lipid binding.\u003c/p\u003e \u003cp\u003eIn this study, we conducted live-cell imaging to quantify the temporal recruitment of SNX9 to macropinocytic membrane ruffles in relation to PIP enrichment. We observed that SNX9 is recruited synchronously with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e but is delayed relative to PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. To reveal how SNX9 distinguishes between PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, we performed \u003cem\u003ein vitro\u003c/em\u003e reconstituted assays. We found that SNX9 binds both PIPs with comparable affinities, as well as generates and senses membrane curvature irrespective of the two PIP\u003csub\u003e2\u003c/sub\u003e species. Motivated by these findings, we next examined how SNX9 modulates PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e dynamics at the molecular scale. Atomistic MD simulations of SNX9 PX-BAR domain with model membranes revealed a non-canonical binding interface in the PX domain that selectively favors PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, and drives PI(3,4)P\u003csub\u003e2\u003c/sub\u003e sequestration. We hypothesized that this selective sequestration may influence the conversion of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e during macropinocytosis. Indeed, mutational analysis in cells confirmed that this non-canonical interface can interfere with the downstream phosphatase INPP4B\u0026rsquo;s ability to convert PI(3,4)P\u003csub\u003e2\u003c/sub\u003e into PI3P. Collectively, our findings suggest that the selectively sequestration of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e by SNX9 may not only facilitate its recruitment to the plasma membrane, but also provides a \u0026ldquo;protective\u0026rdquo; effect, potentially preventing PI(3,4)P\u003csub\u003e2\u003c/sub\u003e from being prematurely converted into PI(3)P during macropinocytosis and endocytosis.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n\u003ch2\u003eSNX9\u0026rsquo;s Recruitment to Ruffling Membranes is Coincident with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e but Not PI(4,5)P\u003csub\u003e2\u003c/sub\u003e\u003c/h2\u003e\n\u003cp\u003eTo investigate the interplay between SNX9 and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e or PI(3,4)P\u003csub\u003e2\u003c/sub\u003e during macropinocytosis, we conducted live-cell imaging to monitor their spatiotemporal distribution upon stimulation with platelet-derived growth factors (PDGF) in serum-starved cells. The initial stage of macropinocytosis involves membrane ruffle formation, a process where SNX9 plays a crucial role \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. We performed time-lapse imaging of cells overexpressing fluorescently tagged SNX9 alongside fluorescent PH-domain probes for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (PLCd-PH-EGFP) or PI(3,4)P\u003csub\u003e2\u003c/sub\u003e (NES-EGFP-cPHx3) to track the recruitment and dynamics of each molecule at the sites of ruffle formation (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e, A and C). Our results show that SNX9 co-localizes spatially with both PIP\u003csub\u003e2\u003c/sub\u003e reporters at PDGF-induced membrane ruffles (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e, B and D). However, quantitative analysis of SNX9 and PIP\u003csub\u003e2\u003c/sub\u003e fluorescence intensity profiles during the first 60 seconds of ruffle formation shows distinct enrichment kinetics. Specifically, for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, we observe a delayed appearance of SNX9 fluorescence relative to PI(4,5)P\u003csub\u003e2\u003c/sub\u003e at the same membrane ruffle site (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eE). In contrast, SNX9 and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e fluorescence signals appear simultaneously, with no detectable lag (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eF). The observed delayed, yet spatially colocalized, appearance of SNX9 relative to PI(4,5)P\u003csub\u003e2\u003c/sub\u003e is consistent with the established role of PI(4,5)P\u003csub\u003e2\u003c/sub\u003e as an upstream factor during endocytosis and macropinocytosis \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. However, the simultaneous appearance of SNX9 with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, as opposed to the lag observed with PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, suggests a potential selective association between SNX9 and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e at macropinocytic ruffles. Our finding suggests that SNX9 displays a specificity for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. This observation is consistent with previous studies indicating that PI(3,4)P\u003csub\u003e2\u003c/sub\u003e promotes SNX9 membrane association at the late stage of clathrin-mediated endocytosis, thereby facilitating dynamin recruitment by SNX9 to drive membrane fission \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSNX9 is a Weakly Scaffolding BAR Protein that Senses and Generates a Wide Range of Membrane Curvatures on both PI(3,4)P\u003c/strong\u003e \u003csub\u003e \u003cstrong\u003e2\u003c/strong\u003e \u003c/sub\u003e \u003cstrong\u003e- and PI(4,5)P\u003c/strong\u003e \u003csub\u003e \u003cstrong\u003e2\u003c/strong\u003e \u003c/sub\u003e \u003cstrong\u003e-Containing Membranes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGiven that SNX9 is a BAR domain protein, we next asked how different PIP\u003csub\u003e2\u003c/sub\u003e species in membranes impacts the expected ability of SNX9 to sense and generate curved membranes. We purified full length SNX9 and labelled it with Alexa fluorophore dyes for detection by confocal microscopy. We first determined whether SNX9\u0026rsquo;s binding affinity to flat membranes differs between PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e. To do this, we used giant unilamellar vesicles (GUVs) with diameters of about 5 \u0026micro;m or larger. GUV membranes were composed of eggPC, supplemented with 10 mol% DOPS, 10 mol% DOPE, 15 mol% cholesterol, and 0.5 mol% Bodipy ceramide, along with either 8 mol% PI(4,5)P\u003csub\u003e2\u003c/sub\u003e or PI(3,4)P\u003csub\u003e2\u003c/sub\u003e. Upon incubating SNX9 with GUVs, we observed the formation of outward membrane tubules on the GUVs, confirming SNX9\u0026rsquo;s ability to deform membranes and generate membrane tubules, a characteristic feature of BAR domains (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, A and B). To estimate the binding affinities of SNX9 on PI(4,5)P\u003csub\u003e2\u003c/sub\u003e- and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e-containing GUVs, we measured the surface density of SNX9 on GUV membranes as a function of its bulk concentration (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, C and D) \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. By fitting to the Hill equation, we estimated dissociation constants of 31 nM for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and 47 nM for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e. These comparable affinities are consistent with previous findings for the PX-BAR domain of SNX9 on small unilamellar vesicles \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTo evaluate SNX9\u0026rsquo;s ability to sense membrane curvature, we generated cylindrical membrane nanotubes from SNX9-coated GUVs using optical tweezers, with tube radii that can be tuned by changing membrane tension via micropipette aspiration \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, E and F). We quantified the enrichment of SNX9 on the nanotubes relative to the flat GUV membranes by calculating the sorting ratio, as previously described for other BAR domains \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Our data show that SNX9\u0026rsquo;s sorting ratio increases as the membrane tubes are thinner \u0026ndash; that is, as curvature increases - within the experimentally accessible range of \u0026sim; 0.0125 nm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e to \u0026sim; 0.15 nm\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e (corresponding to tubes radii of ~\u0026thinsp;80 nm and 7 nm, respectively) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, G and H). Furthermore, by comparing the sorting ratio of SNX9 at a relatively low surface density (on average 1% surface coverage, i.e. protein areal density 200 mm\u003csup\u003e\u0026minus;\u0026thinsp;2\u003c/sup\u003e), where enrichment effects are expected to be the most pronounced, we observed that SNX9 are enriched on both PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e membrane tubes, as indicated by the sorting ratios larger than 1 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, G and H). Because the full-length SNX9 contains a PX-BAR domain, to analyze its sorting behavior, we applied the curvature mismatch model \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e, which has been shown to be more appropriate for full length BAR protein amphiphysin, and truncated BAR-PH domain of b2-centaurin than the spontaneous curvature model \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. This analysis enables us to estimate the spontaneous curvature of membrane-bound SNX9 (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\stackrel{-}{C}}_{p}\\)\u003c/span\u003e\u003c/span\u003e) and the associated elastic constant \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{\\kappa\\:}\\)\u003c/span\u003e\u003c/span\u003e- which reflects the energetic cost of membrane deformation and indicates the strength of SNX9\u0026rsquo;s mechanical capacity to induce curvature. Our analysis indicates that the intrinsic curvatures of membrane-bound SNX9 for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e containing membranes are comparable (1/\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\stackrel{-}{C}}_{p}\\)\u003c/span\u003e\u003c/span\u003e ~ 6 nm and ~ 7 nm, respectively), as are their elastic constants (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{\\kappa\\:}\\)\u003c/span\u003e\u003c/span\u003e~ 4 k\u003csub\u003eB\u003c/sub\u003eT and ~\u0026thinsp;5 k\u003csub\u003eB\u003c/sub\u003eT, respectively). These findings suggest that, at a macroscopic level, SNX9 has a similar sorting capacity and mechanical effect on membranes, regardless of the PIP\u003csub\u003e2\u003c/sub\u003e species. Moreover, the relatively low elastic constants imply that SNX9 is a weaker membrane scaffolding protein as compared to other BAR proteins such as amphiphysin, b2-centaurin, and IRSp53 \u003csup\u003e36\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTo gain molecular scale insights into how SNX9 organizes and deforms PI(4,5)P\u003csub\u003e2\u003c/sub\u003e- and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e-containing membranes, we performed cryo-electron microscopy (cryo-EM). A heterogeneous solution of Large Unilamellar Vesicles (LUVs) with diameters ranging from 50 nm to 1 \u0026micro;m was prepared by resuspending dry lipid films composed of DOPC and 15% DOPS, supplemented with 10% PI(4,5)P\u003csub\u003e2\u003c/sub\u003e or PI(3,4)P\u003csub\u003e2\u003c/sub\u003e. Besides spherical vesicles, this protocol also yields tubulated vesicles in roughly 15% of the vesicles \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. These vesicles were incubated with 1 mM SNX9 for at least 1 hour at room temperature before being plunge frozen onto EM grids. Our cryo-EM results revealed dense SNX9 binding along membrane tubes under both PIP\u003csub\u003e2\u003c/sub\u003e conditions, whereas the remainder of the liposome membranes exhibited a lower density of SNX9 (Figs.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e, A and B, \u003cem\u003eLeft\u003c/em\u003e panel). Notably, unlike previous studies where BAR domains, such as endophilin, displayed a well-registered alignment on membrane tubes, SNX9 exhibits a more disorganized binding pattern (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e, A and B, \u003cem\u003eLeft\u003c/em\u003e panel, arrows) \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. This less ordered organization likely reflects the influence of the PX domain and disordered regions present in full-length SNX9. Although we attempted to purify the SNX9 BAR domain to further test this hypothesis, the isolated domain proved too unstable for performing experiments \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eQuantitative analysis of tube diameters revealed that SNX9-decorated tubes had median diameters of 16.7 nm for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e- containing membranes and 11.4 nm for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e-containing membranes. In contrast, bare membranes exhibited significantly larger medians of 35.5 nm for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and 32.9 nm for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eC). These data confirm that SNX9 can deform membranes and constrict tubes to diameters near 15 nm, consistent with the spontaneous curvature measured in our tube pulling experiments. We noted that to compare the intrinsic curvatures obtained by tube pulling experiments with those from Cryo-EM, a half-bilayer thickness (2 nm considering tube radius and 4 nm, diameter) should be added, resulting in a tube diameter of 16 nm and 18 nm for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, respectively. For both PIP\u003csub\u003e2\u003c/sub\u003e conditions, we observed peanut-shaped vesicles characterized by a narrow neck-like region at their center. At these necks, the membranes appeared less defined and more diffuse than in the remainder of the vesicles, suggesting local SNX9 enrichment at these necks and reinforcing its role in membrane remodeling (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e, A and B, \u003cem\u003eRight\u003c/em\u003e panel). Notably, these neck regions exhibited a broader diameter distribution compared to the SNX9-decorated tubes, indicating that SNX9 functions as a weak scaffolding protein capable of accommodating a wide range of membrane curvatures (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eC). This observation aligns with our tube pulling experiments, where we observed a wide distribution of sorting ratios at given membrane curvatures (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, G and H). Of note, the stability of these neck-like structures was confirmed by overnight incubation experiments. Together, our Cryo-EM findings demonstrate that SNX9 preferentially associated with curved, tubular membranes in both PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e membranes. Moreover, SNX9 not only binds loosely to membrane tubes but also induces the formation of membrane necks, progressively transforming vesicles into peanut-shaped structures with pinched necks and ultimately into thin tubes.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eSNX9 PX-BAR Drives Distinct Distributions of PI(3,4)P and PI(4,5)P Lipids\u003c/h3\u003e\n\u003cp\u003eTo explore the molecular mechanisms by which SNX9 interacts with membranes containing PI(4,5)P\u003csub\u003e2\u003c/sub\u003e and PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, we conducted all-atom molecular dynamics (AA-MD) simulations of the dimeric SNX9 membrane remodeling region, including the PX-BAR domain (residues 200\u0026ndash;595) with model membranes containing the following compositions: DOPC:DOPS 80:20; DOPC:DOPS:PI(3,4)P\u003csub\u003e2\u003c/sub\u003e 80:15:5; and DOPC:DOPS:PI(4,5)P\u003csub\u003e2\u003c/sub\u003e 80:15:5 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eA). The structure of the membrane remodeling region dimer was predicted using AlphaFold2, to facilitate the realistic inclusion of the C-terminal amphipathic helix and missing loops \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. By probing the unbiased interactions between SNX9 PX-BAR and model membranes, we aimed to reveal whether there is a difference in the way that SNX9 interacts with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. Specifically, we sought to investigate how SNX9 PX-BAR influences the local dynamics of PIP\u003csub\u003e2\u003c/sub\u003e lipids upon binding. Although the timescales achievable by AA-MD (a total of 5 microseconds for each composition) limits direct measurement of kinetic rate constants, the simulations enable detailed analysis of nanoscopic phenomena such as lipid diffusion. To this end, we calculated the total distance traversed by all lipids and visualized the trajectory traces over time, using threshold distances as a color-coded guide (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e, B-D, Supplementary Fig.\u0026nbsp;1-S4). These trace analyses demonstrate that SNX9 PX-BAR sequesters multiple PIP\u003csub\u003e2\u003c/sub\u003e lipids in its immediate vicinity.\u003c/p\u003e\n\u003cp\u003eThe sequestration effect of PIP\u003csub\u003e2\u003c/sub\u003e is most pronounced at the time-averaged positions of the PX domains and the center of the BAR domain (CBD) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e, B and C, Supplementary Fig.\u0026nbsp;1\u0026ndash;4), with SNX9 PX-BAR overlay). These findings indicate that these regions are most strongly implicated in PIP\u003csub\u003e2\u003c/sub\u003e binding. While the PX domain has been previously identified as the primary site of PIP\u003csub\u003e2\u003c/sub\u003e binding, the CBD binding has not been reported in lipid binding. Notably, previous mutational assays have implicated the CBD region in autoinhibition of SNX9, but no lipid selectivity has been reported \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. Structural predictions of SNX9 linker region implicated in autoinhibition position the linker region near the CBD. This alignment suggests that the positively charged CBD may mediate nonspecific interactions with PIP\u003csub\u003e2\u003c/sub\u003e (Supplementary Fig.\u0026nbsp;5). Given the plausible non-specific CBD-lipid binding, we focused on the PX domain to investigate the mechanism underlying SNX9\u0026rsquo;s selectivity for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e.\u003c/p\u003e\n\u003cp\u003eWe observed that in the upper leaflet of the membranes where SNX9 PX-BAR is bound, there is super-stoichiometric sequestration of PIP\u003csub\u003e2\u003c/sub\u003e lipids (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e, B and C). The averaged binding ratios are 3.6\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5 : 1 for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and 3.4\u0026thinsp;\u0026plusmn;\u0026thinsp;0.8 : 1 for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (PIP\u003csub\u003e2\u003c/sub\u003e : SNX9 PX domain, Supplementary Table\u0026nbsp;1). These binding ratios, exceeding a value of unity, challenge the canonical lock-and-key interaction paradigm that assumes a single PIP\u003csub\u003e2\u003c/sub\u003e lipid fits into a tight binding pocket. Instead, they suggest that SNX9 PX domain can engage multiple lipids simultaneously through a multivalent binding mechanism. Supporting this, we found that only two out of ten of the individual PX domains in our simulations showed a binding occupancy of less than two lipids (Supplementary Table\u0026nbsp;1). The similarity in average binding ratios between PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e can be attributed to this multivalency, as having multiple binding subdomains that simultaneously bind to the membrane will be the greatest influence on average interactions.\u003c/p\u003e\n\u003cp\u003eTo determine whether the observed superstoichiometric binding is due to specific SNX9 PX-BAR and PIP\u003csub\u003e2\u003c/sub\u003e interactions rather than the intrinsic properties of the PIP\u003csub\u003e2\u003c/sub\u003e lipids, we analyzed the average dynamics of the lipids. As shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e, B and C, we observe significant differences in the distance traveled by the PIP\u003csub\u003e2\u003c/sub\u003e lipids near the SNX9 PX-BAR domain compared to unbound PIP\u003csub\u003e2\u003c/sub\u003e lipids in the upper leaflet and those in the lower leaflet (Supplementary Fig.\u0026nbsp;6). We note that there is no significant difference in the total distance traveled between unbound PIP\u003csub\u003e2\u003c/sub\u003e lipids in the upper leaflet and in the lower leaflet (Supplementary Fig.\u0026nbsp;6). Additionally, no significant spatial variation of PIP\u003csub\u003e2\u003c/sub\u003e lipids in lower leaflet was observed, which indicates the absence of trans-bilayer coupling due to SNX9 PX-BAR binding. Comparing between PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e in the lower leaflet, there is no significant difference in the radius of gyration, a measure of lipid motion, indicating that the dynamics of the individual PIP\u003csub\u003e2\u003c/sub\u003e lipids are similar on average (Supplementary Fig.\u0026nbsp;7). Additionally, we quantified the average network connectivity of PIP\u003csub\u003e2\u003c/sub\u003e clustering as measure of average cluster size and found no significant differences either between the bilayers containing PI(3,4)P\u003csub\u003e2\u003c/sub\u003e versus PI(4,5)P\u003csub\u003e2\u003c/sub\u003e nor between the upper (SNX9 bound) and lower (no protein bound) leaflets of each bilayer (Supplementary Fig.\u0026nbsp;8). The lack of observable clustering among PIP\u003csub\u003e2\u003c/sub\u003e is consistent with previous findings, given the ionic composition of our simulations and the local sequestration effect observed \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. Taken together, our results indicate that the observed lipid selective sequestration is not driven by intrinsic differences in PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e behaviors. Instead, the specificity arises from direct interactions between SNX9 PX-BAR and PIP\u003csub\u003e2\u003c/sub\u003e lipids, particularly those mediated by the PX domain.\u003c/p\u003e\n\u003cp\u003eTo further exclude explanations of the local sequestration based on membrane-mediated effects, we explored possible changes in the behaviors of DOPC and DOPS, as well as changes in overall membrane properties. We found no significant difference in the average radius of gyration and the spatial distributions of DOPC or DOPS in membranes containing or lacking PIP\u003csub\u003e2\u003c/sub\u003e lipids; put another way, the presence of PIP2 does not locally change the dynamics of other lipids (Supplementary Fig.\u0026nbsp;7, 9). Global membrane properties such as mean curvature and membrane packing defects also showed no significant differences between the three membrane compositions tested in the presence of SNX9 (Supplementary Fig.\u0026nbsp;10\u0026ndash;11). These results indicate that neither the behavior of lipid species present in the membrane nor of the global membrane properties are altered in ways that could explain the observed PI(3,4)P\u003csub\u003e2\u003c/sub\u003e selectively sequestering by the PX domain of SNX9.\u003c/p\u003e\n\u003ch3\u003eThe Membrane Binding Interface of SNX9 PX domain is Larger Than Previously Identified\u003c/h3\u003e\n\u003cp\u003eWe next sought to identify PX-PIP\u003csub\u003e2\u003c/sub\u003e contacts to determine whether previously characterized canonical binding residues are responsible for PIP\u003csub\u003e2\u003c/sub\u003e selectivity. By analyzing the frequency of direct contacts between the alpha carbon atoms of residues in the PX domain and the central phosphorus of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eA), we identified three distinct subdomains interacting with PIP\u003csub\u003e2\u003c/sub\u003e. These subdomains are referred to as the insert loop, the canonical pocket, and the fourth helix (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eB). We note that the identification of these three subdomains aligns well with the average binding ratios of 3.6 and 3.4 for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, respectively, indicating that one or more of these subdomains can interact with multiple PIP\u003csub\u003e2\u003c/sub\u003e lipids. These subdomains are broad, ranging from 11 to 19 residues with strong membrane interactions, and extend beyond the canonical residues previously identified through sequence alignment \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Between the large membrane interface and observed superstoichiometric binding, the question of selectivity mechanisms should be addressed within each of the subdomains.\u003c/p\u003e\n\u003cp\u003eThe insert loop subdomain (residues 258\u0026ndash;270) is rich in lysines and hydrophobic residues. Its conserved motif, x\u003cstrong\u003eK\u003c/strong\u003ex(S/T)\u003cstrong\u003eK\u0026Phi;\u003c/strong\u003exG\u003cstrong\u003e\u0026Phi;K\u003c/strong\u003eS\u003cstrong\u003eYI\u003c/strong\u003e (where x represents non-conserved residues, and \u0026Phi; represents a large hydrophobic residue), is shared by SNX9 and closely related members of the SNX family, SNX18 and SNX33 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eC, red dashed box, Supplementary Fig.\u0026nbsp;12). The flexibility and composition of this subdomain allows it to penetrate the membrane and interact with negatively charged lipids, anchoring the PX domain to the membrane. It is therefore not surprising that the insert loop shows the highest frequency of contacts with the central phosphate group of both PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. Unlike the canonical pocket and fourth helix subdomains whose membrane interactions are mediated by specific residues, the PIP\u003csub\u003e2\u003c/sub\u003e interactions of the insert loop is broadly across its entire length, reflecting its critical role in PIP\u003csub\u003e2\u003c/sub\u003e interactions (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eC).\u003c/p\u003e\n\u003cp\u003eThe canonical pocket subdomain (residues 286\u0026ndash;293 and 309\u0026ndash;320) has been the primary focus in studying PIP\u003csub\u003e2\u003c/sub\u003e selectivity and binding. Crystal structures of SNX9 PX-BAR domain, along with analogous structural studies of other PX domains, have suggested that a conserved set of residues within the pocket stabilize the PIP headgroup \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. However, while the canonical pocket of SNX9 does interface with the PIP\u003csub\u003e2\u003c/sub\u003e lipids in the membrane, we observed that the conserved residues Tyr294, Arg296, and Lys300 do not interact with the membrane at all (designated with pink * in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eC). Furthermore, compared to the insert loop, the canonical pocket exhibits greater variation in contact frequency with PIP\u003csub\u003e2\u003c/sub\u003e among the sequential residues in the loop, due to its more rigid tertiary structure.\u003c/p\u003e\n\u003cp\u003eThe fourth helix subdomain (residues 358\u0026ndash;370), which is not traditionally considered as part of the PX domain, lies adjacent to the BAR domain and interacts frequently with the membrane (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eC, purple dashed box). Like the canonical pocket, this forth helix subdomain exhibits variation in PIP\u003csub\u003e2\u003c/sub\u003e contact frequency among sequential residues due to its rigid secondary and tertiary structure. While the net charge of the fourth helix is +\u0026thinsp;2, its four positively charged residues (Lys363, Lys367, Arg368, and Arg370) are aligned along its membrane-facing surface. This alignment strongly favor binding to negatively charged lipids such as the PIPs and PS. The alignment of these positively charged residues along the membrane binding surface of the fourth helix suggests that its protruding sidechains could synergistically coordinate the large headgroup of a PIP\u003csub\u003e2\u003c/sub\u003e lipid.\u003c/p\u003e\n\u003ch3\u003eThe SNX9 PX Fourth Helix Promotes PI(3,4)P Selectivity\u003c/h3\u003e\n\u003cp\u003eOf the three subdomains identified in this study, only the canonical pocket and the fourth helix interact with PIP\u003csub\u003e2\u003c/sub\u003e primarily through electrostatic interactions, specifically involving positively charged residues. Because PIP\u003csub\u003e2\u003c/sub\u003e lipids are strongly negatively charged, \u0026minus;\u0026thinsp;4 compared to -1 for PI and PS at physiological pH, and have a bulky phosphoinositide headgroup, the specific arrangements of positively charged residues in the canonical pocket and fourth helix are likely key for explaining how the SNX9 PX domain selectively interacts with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. The electrostatic map of the membrane binding face of SNX9 PX domain clearly shows the importance of these positively charged residues (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eA). It also highlights the formation of two distinct pockets of charge surrounding the canonical pocket and the fourth helix.\u003c/p\u003e\n\u003cp\u003eTo investigate how these residues could cooperatively bind to the same PIP\u003csub\u003e2\u003c/sub\u003e lipid, we calculated a contact frequency for all positively charged residues based on the Bjerrum length, considering instances when two residues coordinate the same lipid, though not necessarily the same phosphate group (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eB, Supplementary Fig.\u0026nbsp;13) \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. This cooperative frequency based on electrostatic binding groups provides deeper insight into the specificity of lipid interactions. By identifying residues with cooperative self-contact frequencies greater than 1.5, we can determine those involved in major coordination events (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eC). For the canonical pocket, the major coordinating residues were identical for both PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, and include Arg286, Lys288, Lys313, and Arg318 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eD). However, for the fourth helix, we find a broader set of coordinating residues for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. Specifically, residues Lys360, Lys363, Lys366, Arg,367, Lys368 and Arg371 for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and Lys363, Lys366, Arg367 and Arg371 for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, indicating stronger coordinating interactions for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eE). Besides, we observe a stronger interaction network between all the residues in the fourth helix for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eE). This reinforces the idea of a more robust coordination for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and provides an explanation for how selectivity is maintained, despite superstoichiometric binding ratios. The expansive network of coordinating residues is capable of binding multiple PIP\u003csub\u003e2\u003c/sub\u003e lipids simultaneously. Furthermore, inter-domain contacts, particularly between the fourth helix and the canonical pocket subdomains, are stronger for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e than for PI(4,5)P\u003csub\u003e2\u003c/sub\u003e (Supplementary Fig.\u0026nbsp;13). In addition to electrostatic differences, we note that the canonical pocket has a significantly smaller cavity volume of 745 \u0026Aring;\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e compared to the volume of the fourth helix subdomain of 905 \u0026Aring;\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e (Supplementary Fig.\u0026nbsp;14). Docking predictions further suggest that both PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e are more likely to bind to the fourth helix subdomain than the canonical pocket (Supplementary Fig.\u0026nbsp;14). Collectively, our findings indicate that SNX9 PX domain has a selectivity for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, through the fourth helix of the PX domain.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpecific SNX9-PI(3,4)P\u003c/strong\u003e \u003csub\u003e \u003cstrong\u003e2\u003c/strong\u003e \u003c/sub\u003e \u003cstrong\u003eInteractions Protect PI(3,4)P\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e \u003cstrong\u003efrom Hydrolysis by INPP4B at the Plasma Membrane during Membrane Ruffling\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBased on our simulation results showing that SNX9 PX-BAR dimers sequester PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, we hypothesized that SNX9 may influence the hydrolysis of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e by INPP4B phosphatase. This hydrolysis, converting PI(3,4)P\u003csub\u003e2\u003c/sub\u003e to PI(3)P, is critical in macropinocytosis for macropinosome closure \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. By monitoring PI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment at the plasma membrane of cells overexpressing membrane-bound INPP4B (INPP4B-CAAX), we assessed whether SNX9 protects PI(3,4)P\u003csub\u003e2\u003c/sub\u003e from INPP4B-mediated hydrolysis through sequestration. Consistent with previous reports, overexpression of wild-type INPP4B (INPP4B-WT), but not the catalytic dead C842S mutant (INPP4B-CS), significantly reduces PI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment at plasma membrane ruffles (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e, A \u0026ndash; C) \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. Overexpression of SNX9-mScarlet restores PI(3,4)P\u003csub\u003e2\u003c/sub\u003e at the plasma membrane of cells expressing INPP4B-WT (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e, B and D). This result highlights SNX9\u0026rsquo;s protective role in shielding PI(3,4)P\u003csub\u003e2\u003c/sub\u003e from hydrolysis by INPP4B.\u003c/p\u003e\n\u003cp\u003eTo reveal the functional relevance of SNX9\u0026rsquo;s PX interaction networks in coordinating PI(3,4)P\u003csub\u003e2\u003c/sub\u003e that our simulations identified, we performed specific mutations in the SNX9 PX domain, and assessed the consequences of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e conversion in cellular assays with either functional or non-functional INPP4B phosphatase. We found that loss-of-function mutations in SNX9, K313A in the canonical pocket and K366/367A in the fourth helix, fail to rescue PI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment in cells expressing INPP4B-WT, despite their proper localization at the plasma membrane (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e, B and D). Conversely, gain-of-function mutations in SNX9, K366R and K366/368R both in the fourth helix, more effectively restored PI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment the plasma membrane. These mutants recapitulated the protective capacity of wild-type SNX9 INPP4B-mediated hydrolysis (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003eD). Notably, in cells expressing the non-functional phosphatase, INPP4B C842S, PI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment index is comparable among all SNX9 variants, indicating that the observed effects of SNX9 on PI(3,4)P\u003csub\u003e2\u003c/sub\u003e coordination are specific to the functional interaction with INPP4B (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003eD).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe demonstrated that PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e exhibit distinct spatial distributions upon interacting with SNX9 PX-BAR domain. Although both lipids bind superstoichiometrically to SNX9 via the membrane-remodeling PX-BAR domain, our results revealed that the non-canonical Helix 4 interface in the PX domain is a key determinant of the differential PIP\u003csub\u003e2\u003c/sub\u003e localization. This finding provides mechanistic insight into how SNX9 can achieve functional lipid selectivity despite its high overall valency for PIP\u003csub\u003e2\u003c/sub\u003e lipids. Even though the selective binding effect is modest compared to the overall binding of the PX domain, it translates into the preferential spatiotemporal colocalization of SNX9 with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e rather than PI(4,5)P\u003csub\u003e2\u003c/sub\u003e at macropinocytic ruffles \u003cem\u003ein vivo\u003c/em\u003e. Given the generality of the underlying biophysical principles, SNX9\u0026rsquo;s selective interaction with PIP lipids likely represents a general mechanism for the spatiotemporal recruitment of peripheral membrane proteins in processes like macropinocytosis and endocytosis. As exemplified here by SNX9, many proteins with PIP selective domains, such as PX, PH, FYVE, and C2, might exploit a similar dual mode of PIP interaction. To further elucidate the mechanism behind SNX9\u0026rsquo;s Helix 4 selectivity for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, we characterized a non-canonical binding interface composed of a network of positively charged residues that coordinate the PIP\u003csub\u003e2\u003c/sub\u003e headgroup. We hypothesized and demonstrated with mutational assays that this Helix 4 interface acts as a molecular checkpoint, preventing the premature hydrolysis of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e to PI(3)P by INPP4B during macropinocytosis (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e). Having a checkpoint to prevent the premature conversion of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e to PI(3)P is especially critical at this juncture in the membrane remodeling process, just prior to membrane fission, to prevent stalled or failed internalization \u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eUsing quantitative reconstituted assays, we also showed that full-length SNX9 exhibits membrane curvature sensing and remodeling capabilities, accommodating a wide range of membrane curvatures, independent of the specific PIP\u003csub\u003e2\u003c/sub\u003e species present. Although SNX9 functions as a relatively weak scaffolding BAR protein, its loosely organized assembly on membrane tubes may allow it to stabilize divers structures \u0026ndash; from elongated tubes to constricted necks \u0026ndash; which is essential for dynamic processes such as endocytosis and macropinocytosis.\u003c/p\u003e \u003cp\u003eCollectively, our integrated \u003cem\u003ein silico\u003c/em\u003e, \u003cem\u003ein vitro\u003c/em\u003e and \u003cem\u003ein vivo\u003c/em\u003e findings indicate that SNX9\u0026rsquo;s dual membrane binding domains - its BAR domain for curvature sensing and its PX domain for regulating local PI(3,4)P\u003csub\u003e2\u003c/sub\u003e concentration \u0026ndash; work synergistically for membrane remodeling processes. This dual functionality reinforces SNX9\u0026rsquo;s central role in membrane dynamics and highlights a broadly applicable mechanism by which peripheral membrane proteins achieve selective recruitment and stabilization at the plasma membrane. Moreover, our Cyro-EM imaging reveals that SNX9-enriched membrane tubes exhibit a broad spectrum of membrane curvatures, with some comparable to those observed in clathrin-mediated endocytosis and dynamin-induced tubes \u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e, suggesting that SNX9 remains membrane-associated throughout the course of remodeling events. Ultimately, SNX9\u0026rsquo;s ability to coordinate membrane remodeling during lipid conversion with the timely recruitment of downstream effectors via its SH3 domain, such as dynamin and N-WASP, provides a mechanistic basis to ensure coordinated protein recruitment during cellular internalization events. Given the critical role of membrane composition in coordinating membrane remodeling, and considering that enzymatic lipid conversion occurs on a faster timescale than micron-scale membrane remodeling, SNX9\u0026rsquo;s protective role for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e against pre-mature hydrolysis to PI(3)P may serve as a temporal checkpoint. This checkpoint ensures that the remodeling process reaches a sufficient level of maturity before proceeding to the final stages of internalization and conversion into an endosomal body. Together, our findings underscore SNX9\u0026rsquo;s mechanistic role in membrane association and remodeling during lipid conversion, suggesting that SNX9 is both recruited by and orchestrates the maintenance of local PI(3,4)P\u003csub\u003e2\u003c/sub\u003e concentrations, as well as the timely recruitment of downstream effectors to facilitate membrane remodeling. While not the focus of our study, the previously observed hetero-dimerization of SNX9 with SNX18 could offer a further expanded repertoire of regulatory capabilities \u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe magnitude of the PIP\u003csub\u003e2\u003c/sub\u003e selectivity of SNX9 PX-BAR domain can be partially attributed to the non-specific superstoichiometric binding of the PX-BAR to the membrane. Binding to multiple lipids simultaneously in non-specific ways can make it difficult to determine selective lipid binding of the PX domain, which may explain why previous \u003cem\u003ein vitro\u003c/em\u003e literature has reported weak binding for the SNX9 PX domain \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Superstoichiometric binding is likely a general strategy to recruit peripheral proteins to the membrane surface, while specific binding may play a role in the retention of PIP lipids. We note that this does not imply that lipid composition is the exclusive mechanism of recruitment, as other membrane properties such as curvature also change significantly during membrane remodeling events and thus drive protein recruitment. Indeed, the previously reported interplay of the SNX1 PX domain with the BAR domain as partners in membrane recruitment reinforces the idea that curvature plays a likely role in membrane recruitment for other PX-BAR domains, including SNX9 \u003csup\u003e45\u003c/sup\u003e. Given our results suggesting that SNX9 responds similarly to membranes of similar curvature containing either PI(3,4)P\u003csub\u003e2\u003c/sub\u003e or PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, we suggest that curvature may be secondary to PIP\u003csub\u003e2\u003c/sub\u003e selectivity for SNX9 recruitment. In addition, the observed general reduction of lipid mobility around the SNX9 PX-BAR domain due to this superstoichiometric binding may be part of the mechanism of protein-mediated PIP\u003csub\u003e2\u003c/sub\u003e aggregation, which has been previously reported for other BAR proteins \u003csup\u003e\u003cspan additionalcitationids=\"CR47\" citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. However, aggregation is a collective effect, and therefore further \u003cem\u003ein silico\u003c/em\u003e studies should include a much larger membrane system with at least dozens of PX-BAR domains to investigate this phenomenon.\u003c/p\u003e \u003cp\u003eThe Helix 4 interface, identified as driving selectivity of SNX9 for PI(3,4)P\u003csub\u003e2\u003c/sub\u003e over PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, provides the opportunity to reflect more broadly on our treatment of peripheral membrane protein binding modes. We have endeavored to emphasize here that even for lipid selective domains such as the PX domain, multiple lipids are simultaneously bound and so determining selectivity is not always a straightforward task, and requires a detailed look at the structure and dynamics of the domain of interest on a membrane. Sequence data is helpful for identifying conserved residues that may play an important role, but should be paired whenever possible with structural analysis.\u003c/p\u003e \u003cp\u003eThe large, negatively charged headgroup of the PIP\u003csub\u003e2\u003c/sub\u003e lipids define their interactions with peripheral membrane proteins, and distinguishes them particularly from other lipid classes; therefore, the electrostatic interaction mechanism proposed here, coupled with spatial information of the non-canonical binding interface (i.e., how the positively charged sidechains face the membrane and the larger cavity size of the non-canonical interface), is grounded by physical arguments and supported by the mutational assays in cells. While we primarily probed the electrostatic impact by mutating positively charged residues, further work could probe the effect of interface size by introducing bulky, neutral sidechains that would protrude into the adjoining cavity. However, the introduction of such residues could disrupt the protein tertiary structure or lead to insertion of these large neutral residues into the membrane, significantly altering the membrane binding mode as opposed to only probing PIP\u003csub\u003e2\u003c/sub\u003e selectivity.\u003c/p\u003e \u003cp\u003eIt remains unclear what signal eventually drives SNX9 to dissociate from the membrane and free PI(3,4)P\u003csub\u003e2\u003c/sub\u003e for conversion by INPP4B and other phosphatases. The change in Gaussian membrane curvature as a result of scission could be an important signal to induce a mechanical change in SNX9 organization from a loose organization on tubes to more shallow contacts on spherical vesicles, as has been demonstrated for other BAR domain containing proteins \u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAltogether, it is striking how lipid selectivity of peripheral membrane proteins between two nearly identical lipids, PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, arises from simple effects such as electrostatics and spatial structural arrangements despite their high valency, and further how selectivity drives the coordination of complex remodeling processes at the plasma membrane. We anticipate that these conclusions are not unique to SNX9 nor the plasma membrane, but rather provide insights into the biophysics behind selective membrane recruitment for a host of peripheral membrane proteins.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eCELL EXPERIMENTS\u003c/h2\u003e \u003cp\u003e \u003cb\u003eCell culture, transfection and imaging.\u003c/b\u003e COS7 cells were maintained in DMEM supplemented with 10% FBS. Transfection was performed using the plasmid DNA of interest and TransIT (Mirus Bio) according to the manufacturer's instructions. For live-cell imaging, cells were visualized using a spinning-disc microscope (Carl Zeiss) in imaging medium consisting of phenol red-free DMEM supplemented with 10 mM HEPES and 10% FBS. Medium containing 50 ng/mL PDGF-BB was applied to cells to stimulate membrane ruffles. Images were captured every 2 seconds for 10\u0026ndash;15 minutes. For fixed-cell imaging, transfected cells were seeded onto coverslips and fixed with 4% paraformaldehyde, followed by observation with an LSM700 confocal microscope (Carl Zeiss).\u003c/p\u003e \u003cp\u003e \u003cb\u003eImage quantification.\u003c/b\u003e To analyze the spatiotemporal distribution of PIPs and SNX9, a 1 \u0026micro;m\u0026sup2; area was cropped from each membrane ruffle for quantification. The signals were normalized to the highest value and plotted over time. The maximal intensity of SNX9-RFP was set as 100% and aligned to 50 sec for synchronization in figure plotting.\u003c/p\u003e \u003cp\u003eFor PI(3,4)P\u003csub\u003e2\u003c/sub\u003e enrichment index quantification, the index is defined as the ratio between PI(3,4)P\u003csub\u003e2\u003c/sub\u003e-enriched membrane and resting plasma membrane. Images were processed with maximal intensity projection, and the contour of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e-enriched membrane regions was manually defined. A similar area of the resting plasma membrane was also defined. The GFP intensity of PI(3,4)P\u003csub\u003e2\u003c/sub\u003e-enriched ruffles and the resting plasma membrane were measured, and the ratio was calculated..\u003c/p\u003e \u003cp\u003e \u003cb\u003eStatistical analysis.\u003c/b\u003e Statistical analysis was conducted using GraphPad Prism 9.0. Data were analyzed with one-way ANOVA. A P-value of \u0026lt;\u0026thinsp;0.05 was considered statistically significant and is indicated as follows: *, P\u0026thinsp;\u0026lt;\u0026thinsp;0.05; **, P\u0026thinsp;\u0026lt;\u0026thinsp;0.01; ***, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ****, P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eIN VITRO RECONSISTUTION EXPERIMENTS\u003c/h3\u003e\n\u003cp\u003e \u003cb\u003eReagents.\u003c/b\u003e Brain L-α-phosphatidylinositol-4,5-bisphosphate (PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, 840046P), 1-heptadecanoyl-2-(5Z,8Z,11Z,14Z-eicosatetraenoyl)-sn-glycero-3-phospho-(1'-myo-inositol-3',4'-bisphosphate) (17:0\u0026ndash;20:4 PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, LM1903) 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[biotinyl(polyethyleneglycol)-2000] (DSPE-PEG(2000)-biotin, 880129P), L-α-phosphatidylcholine (Egg, Chicken) (EPC, 840051), 1,2-di-(9Z-octad\u0026eacute;c\u0026eacute;noyl)-\u003cem\u003esn\u003c/em\u003e-glyc\u0026eacute;ro-3-phospho\u0026eacute;thanolamine (18:1 (Δ9-Cis) DOPE, 850725P), 1,2-di-(9Z-octad\u0026eacute;c\u0026egrave;noyl)-\u003cem\u003esn\u003c/em\u003e-glyc\u0026eacute;ro-3-phospho-L-s\u0026eacute;rine (18:1 PS DOPS, 840035), and Cholesterol (700000P), were purchased from Avanti Polar Lipids. BODIPY-TR-C5-ceramide, (BODIPY TR ceramide, D7540), BODIPYFL C5-hexadecanoyl phosphatidylcholine (HPC*, D3803), and Alexa Fluor 488 C5-Maleimide (AX488) were purchased from Invitrogen. Streptavidin-coated polystyrene beads (SVP-30-5) were purchased from Spherotech. β-casein from bovine milk (\u0026gt;\u0026thinsp;98% pure, C6905) and other reagents were purchased from Sigma-Aldrich. Culture-Inserts 2 Well for self-insertion were purchased from ibidi (Silicon open chambers, 80209).\u003c/p\u003e \u003cp\u003e \u003cb\u003eProtein purification and labelling.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eFull length SNX9 construct was a gift from Christian Wunder (Institut Curie, France). The full length SNX9 sequence was cloned into the pProEX HTB plasmid (Invitrogen) between the BamHI and NotI cloning sites, adding a 6xHis tag and rTEV protease cleavage site to the N-terminus of the protein. The plasmid was transformed into Rosetta\u0026trade; 2(DE3) pLysS (Novagen) and expressed in 2YT medium for 4 hours at 37\u0026deg;C with 0.2 mM IPTG. Cells were lysed by sonication in 50 mM Hepes pH 7.4, 300 mM NaCl, 20 mM Imidazole supplemented with complete EDTA-free protease inhibitor cocktail (Roche). Sample was centrifuged at 20,000 xg and proteins purified using TALON\u0026reg; Metal Affinity Resins (Takara Bio). Protein was eluted with 50 mM Hepes pH 7.4, 300 mM NaCl, 300 mM Imidazole. Proteins were further purified over the Superdex 200 10/300GL column (GE Healthcare) in PBS, 0,5mM EDTA pH 8,0. Pure protein was the labelled with 1:1 protein to dye ratio with Alexa 488 C5 Malemide. Labelled samples were dialyzed into PBS, 0,5mM EDTA pH 8,0, 10% Glycerol, and stored at -80\u0026deg;C.\u003c/p\u003e \u003cp\u003e \u003cb\u003eBuffer compositions\u003c/b\u003e. The salt buffer inside GUVs, named I-buffer, was 50 mM NaCl, 20 mM sucrose and 20 mM Tris pH 7.5. The salt buffer outside GUVs, named O-buffer, was 60 mM NaCl and 20 mM Tris pH 7.5.\u003c/p\u003e \u003cp\u003e \u003cb\u003eGUV preparation.\u003c/b\u003e GUVs was prepared using the polyvinyl alcohol (PVA) gel-assisted method \u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e. A PVA solution (5% (w/w) of PVA in a 280 mM sucrose solution) was warmed up to 50\u0026deg;C before spreading on a coverslip, which was cleaned in advance by rinsing with ethanol and MilliQ water. The PVA-coated coverslip was dried in an oven at 60\u0026deg;C for 30 min. 5\u0026ndash;10 \u0026micro;l of the lipid mixture (1 mg/mL in chloroform) was spread on the PVA-coated coverslip, followed by drying under vacuum for 30 min at room temperature. The PVA-lipid-coated coverslip was placed in a 10 cm cell culture dish and 0.5 mL of the I-buffer was added on the coverslip, and kept it stable for at least 45 min at room temperature to allow GUV to grow.\u003c/p\u003e \u003cp\u003e \u003cb\u003eSample preparation and observation for measuring dissociation constants.\u003c/b\u003e GUVs were incubated with SNX9 at bulk concentrations depending on the designed experiments for at least 15 min at room temperature. Chamber coverslips were passivation with a β-casein solution at a concentration of 5 g.L\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e for at least 5 min at room temperature. Experimental chambers were assembled by placing a silicon open chamber on a coverslip. Samples were observed using a Nikon C1 confocal microscope equipped with a X60 water immersion objective.\u003c/p\u003e \u003cp\u003e \u003cb\u003eImage analysis.\u003c/b\u003e Image analysis was performed by Fiji \u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. Florescence images were taken at the equatorial planes of GUVs using identical confocal microscopy settings. We measured the fluorescence intensities of SNX9 on spherical GUVs devoid of tubules. The background intensity of the AX488 channel was obtained by drawing a line with a width of 10 pixels perpendicularly across GUV membranes. We obtained the background intensity profile of the line with the x-axis of the profile be the length of the line and the y-axis, the averaged pixel intensity along the width of the line. The background intensity was obtained by calculating the mean value of the sum of the first 10 intensity values and the last 10 intensity values of the background intensity profile. To obtain SNX9 fluorescence intensity on the GUV membrane, we used membrane fluorescence signals to detect the contour of the GUV. A 10 pixel wide band centered on the contour of the GUV was used to obtain the SNX9 intensity profile of the band where the x-axis of the profile is the length of the band and the y-axis, the averaged pixel intensity along the width of the band. SNX9 fluorescence intensity was then obtained by calculating the mean value of the intensity values of the SNX9 intensity profile, following by subtracting the background intensity.\u003c/p\u003e \u003cp\u003eWe measured SNX9 surface density on GUV membranes (number of proteins per unit area) by using a previously established procedure \u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. We related the fluorescence intensity of AX488 to that of a fluorescent lipid (BODIPY FL-C5-HPC, named HPC*). We measure fluorescence intensity of HPC* on GUV membranes at a given HPC* membrane fraction. The surface density of the protein on membranes is \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{n}_{protein}={n}_{{HPC}^{*}}/\\left(\\frac{{I}_{AX488}}{{I}_{{HPC}^{*}}}\\right)\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{{I}_{AX488}}{{I}_{{HPC}^{*}}}\\)\u003c/span\u003e\u003c/span\u003e is the factor accounting for the fluorescence intensity difference between HPC* and AX488 at the same bulk concentration under identical image acquisition condition. The area density of HPC*, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varPhi\\:}_{{HPC}^{*}}\\)\u003c/span\u003e\u003c/span\u003e, can be related to its fluorescence intensity, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{I}_{{HPC}^{*}}^{vesicle}\\)\u003c/span\u003e\u003c/span\u003e, by measuring fluorescence intensities of GUVs composed of DOPC supplemented with different molar ratios of HPC* (0.04\u0026ndash;0.16 mole%) and assuming lipid area per lipid is 0.7 nm\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (1120\u0026ndash;4480 HPC* per \u0026micro;m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e). As such, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{n}_{{HPC}^{*}}=A\\times\\:{I}_{{HPC}^{*}}^{vesicle}\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:A\\)\u003c/span\u003e\u003c/span\u003e is a constant depending on the illumination setting in the microscope. We then obtained the surface density of the protein as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{n}_{protein}=(A\\times\\:{I}_{protein}^{vesicle})/(\\frac{{I}_{AX488}}{{I}_{{HPC}^{*}}}\\times\\:{n}^{*})\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{n}^{*}\\)\u003c/span\u003e\u003c/span\u003e is the degree of labeling for the protein of interest. Finally, we obtained the surface fraction of the protein \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varPhi\\:}_{protein}=\\:{n}_{protein}\\times\\:{a}_{protein}\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{a}_{protein}\\)\u003c/span\u003e\u003c/span\u003eis the area of a single protein on membranes, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{a}_{BAR\\:domain}\\cong\\:50\\)\u003c/span\u003e\u003c/span\u003e nm\u003csup\u003e2 35\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo obtain the dissociation constant of SNX9 on GUV membranes, we fitted the surface densities of SNX9 \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varphi\\:}_{v},\\)\u003c/span\u003e\u003c/span\u003e determined from the fluorescence signals, as a function of SNX9 bulk concentration \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{bulk}\\)\u003c/span\u003e\u003c/span\u003e to\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:{\\varphi\\:}_{v}=\\frac{{\\varphi\\:}_{max}\\times\\:{{C}_{bulk}}^{n}}{{K}_{d}+{{C}_{bulk}}^{n}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varphi\\:}_{max}\\)\u003c/span\u003e\u003c/span\u003e is the maximum surface density, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{K}_{d}\\)\u003c/span\u003e\u003c/span\u003e is the dissociation constant and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:n\\)\u003c/span\u003e\u003c/span\u003e is the Hill coefficient.\u003c/p\u003e \u003cp\u003e \u003cb\u003eTube pulling experiments.\u003c/b\u003e The experiments were performed on a setup with a Nikon C1 confocal microscope equipped with a X60 water immersion objective, micromanipulators for positioning micropipettes and optical tweezers as previously described \u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. To pull a tube, a GUV was held by a micropipette, brought into contact with a streptavidin-coated bead trapped by the optical tweezers, followed by moving away from the bead. Tube radius \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:R\\)\u003c/span\u003e\u003c/span\u003e was measured by the ratio of lipid fluorescence intensity on the tube and on the GUV as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R={R}_{c}^{TR}\\times\\:({I}_{tube}/{I}_{vesicle})}_{membrane}\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{c}^{TR}=200\\pm\\:50\\)\u003c/span\u003e\u003c/span\u003e nm is the previously obtained calibration factor for using BODIPY TR ceramide as lipid fluorescence reporter in the same setup by performing a linear fit of membrane fluorescence ratio \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{({I}_{tube}/{I}_{vesicle})}_{membrane}\\)\u003c/span\u003e\u003c/span\u003e and lipid radii \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:R\\)\u003c/span\u003e\u003c/span\u003e measured by \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:R=\\:f/(4\\pi\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e), where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:f\\)\u003c/span\u003e\u003c/span\u003e is the force applied by the optical tweezers to hold the tubes and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\:\\)\u003c/span\u003e\u003c/span\u003ethe membrane tension tuned by the micropipette holding the GUVs \u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e,\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo obtain protein/membrane fluorescence intensity in tube pulling experiments, we defined a rectangular region of interest (ROI) around GUV membranes and around the membrane tube such that the membrane/tube were horizontally located at the center of the ROI. We then obtained an intensity profile along the vertical direction of the ROI by calculating the mean fluorescence intensity of each horizontal line of the rectangle. To account for protein fluorescence outside the GUVs, the background protein intensity was obtained by calculating the average value of the mean of the first 15 intensity values from the top and from the bottom of ROI. A similar procedure was used the membrane background. The protein/membrane fluorescence intensities were obtained by subtracting the background intensity from the maximum intensity value in the intensity profile.\u003c/p\u003e \u003cp\u003e \u003cb\u003eSNX9 sorting data analysis.\u003c/b\u003e We fit the sorting data using the curvature mismatch model \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. The free energy of the system is\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:{F}_{mismatch}^{tube}=2\\pi\\:RL\\left[\\frac{\\kappa\\:}{2{R}^{2}}+\\frac{\\stackrel{-}{\\kappa\\:}}{2}{{\\varphi\\:}_{t}\\left(\\frac{1}{R}-\\stackrel{-}{{C}_{p}}\\right)}^{2}+{f}_{s}+{f}_{m}\\right]$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cem\u003eR\u003c/em\u003e is the tube radius, \u003cem\u003eL\u003c/em\u003e is the tube length, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{\\kappa\\:}\\)\u003c/span\u003e\u003c/span\u003e is the bending modulus of the protein-bound membrane, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\kappa\\:\\)\u003c/span\u003e\u003c/span\u003e is the bending modulus of the protein-free membrane, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{\\kappa\\:}\\)\u003c/span\u003e\u003c/span\u003e is an elastic coefficient penalizing mismatch between protein and membrane curvature, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varphi\\:}_{t}\\)\u003c/span\u003e\u003c/span\u003e is the protein areal fraction on the tube, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{{C}_{p}}\\)\u003c/span\u003e\u003c/span\u003e is a phenomenological coefficient related to the membrane-bound protein\u0026rsquo;s intrinsic curvature, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{s}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{m}\\)\u003c/span\u003e\u003c/span\u003e are the energy densities of membrane stretching and protein mixing entropy on membranes. At equilibrium the chemical potentials of the lipids and proteins on the GUV and on the tube are equal, thus an implicit dependence of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varphi\\:}_{v}\\)\u003c/span\u003e\u003c/span\u003e (protein area fraction on the GUV) on the tube curvature can be written as\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:\\frac{{\\varphi\\:}_{t}}{{\\varphi\\:}_{v}}{\\left(\\frac{1-{\\varphi\\:}_{v}}{1-{\\varphi\\:}_{t}}\\right)}^{{a}_{p}/{a}_{l}}=\\text{e}\\text{x}\\text{p}\\left[\\frac{\\stackrel{-}{\\kappa\\:}{a}_{p}}{{k}_{B}T}\\left(\\frac{\\stackrel{-}{{C}_{p}}}{R}-\\frac{1}{2{R}^{2}}\\right)\\right]$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{a}_{p}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{a}_{l}\\)\u003c/span\u003e\u003c/span\u003e is the membrane areas occupied by a membrane bound protein and a lipid, respectively. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:S={\\varphi\\:}_{t}/{\\varphi\\:}_{v}\\)\u003c/span\u003e\u003c/span\u003e is the sorting ratio and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:C=1/R\\)\u003c/span\u003e\u003c/span\u003e the tube curvature. By fitting the sorting data with this equation, one can obtain \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{\\kappa\\:}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{{C}_{p}}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cb\u003eCryo-electron microscopy experiments.\u003c/b\u003e A heterogeneous solution of LUVs was generated by resuspension in buffer (60 mM NaCl and 20 mM Tris pH 7.5) of a dried lipid film composed of 80% DOPC, 10% DOPS and either 10% PI(4,5)P2 or 10% PI(3,4)P2 (molar ratios). 1 \u0026micro;M of SNX9 was incubated with the lipid suspensions at 0.1 mg.mL\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e at room temperature for an hour. The samples were vitrified on copper holey lacey grids (Ted Pella) using an automated device (EMGP, Leica) by blotting the excess sample on the opposite side from the droplet of sample for 4 s in a humid environment (90% humidity). Imaging was performed by a Glacios microscope (Thermofisher) running at 200 kV and equipped with a Falcon IVi direct detector (Thermofisher).\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eATOMISTIC MOLECULAR DYNAMICS\u003c/h2\u003e \u003cp\u003eThe initial structure of the truncated SNX9 PX-BAR dimer (residues 201\u0026ndash;595) was generated with AlphaFold2 \u003csup\u003e39\u003c/sup\u003e to model missing loops and the N-terminal amphipathic helix. The predicted structure was compared to previously solved structures (PDB IDs: 2RAI, 2RAK), and found to have only small deviations from each structure (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. The PX-BAR dimer was oriented parallel to the membrane and translated to initially be 3nm above the membrane surface. Three sets of molecular dynamics simulations were conducted on membranes that were initially 22 nm by 22 nm in the xy plane, with compositions of 80:20 mole percent DOPC:DOPS, 80:15:5 mole percent DOPC:DOPS:PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, and 80:15:5 mole percent DOPC:DOPS:PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, and solvated with 0.15M KCl; five replicates of each membrane composition were conducted. Atomistic systems were generated with the CHARMM-GUI platform \u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e,\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e, and simulated with the CHARMM-36m force field using GROMACS 2021.5 \u003csup\u003e56\u003c/sup\u003e for 1000ns per replicate in the constant NPT ensemble after standard equilibration, for a total of 5 \u0026micro;s per system. Analysis was conducted over the last 750ns of each replicate in Python with MDAnalysis \u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. Visualization was conducted with ChimeraX \u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e and Visual Molecular Dynamics \u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank Olena Pylypenko and Korbinian Liebl for insightful discussion and Christian Wunder for SNX9 plasmid. F.-C.T. and P.B. is a member of the CNRS consortium Approches Quantitatives du Vivant (AQV), Labex Cell(n)Scale (ANR-11-889 LABX0038) and Paris Sciences et Lettres (ANR-10-IDEX-0001-02). The authors greatly acknowledge the Cell and Tissue Imaging (PICT-IBiSA), Institut Curie, member of the national infrastructure France-BioImaging (https://ror.org/01y7vt929) supported by the French National Research Agency (ANR-24-INBS-0005 FBI BIOGEN). PB team is supported by the Fondation pour la Recherche M\u0026eacute;dicale (FRM) (FRM EQU202003010307) and by the European Union (ERC, PushingCell, #101071793). Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. Simulations were performed using computing resources provided by the University of Chicago Research Computing Center (RCC), the Department of Defense High Performance Computing cluster (HPCMP), and the National Science Foundation ACCESS cluster.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eJRB, Y-WL, GAV, and F-CT designed the initial project. JRB, CJS, SSL, NdV, AB and F-CT performed experiments and analyzed results with feedback from Y-WL, GAV and PB. JM and SA purified proteins. JRB and F-CT wrote the original draft with inputs and revisions from AB, PB, Y-WL and GAV.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConceptualization: JRB, Y-WL, GAV, and F-CT\u003c/p\u003e\n\u003cp\u003eMethodology: JRB and F-CT\u003c/p\u003e\n\u003cp\u003eInvestigation: JRB, CJS, SSL, NdV, AB, F-CT\u003c/p\u003e\n\u003cp\u003eResources: JM, SA\u003c/p\u003e\n\u003cp\u003eVisualization: JRB, F-CT\u003c/p\u003e\n\u003cp\u003eSupervision: Y-WL, GAV, F-CT\u003c/p\u003e\n\u003cp\u003eWriting\u0026mdash;original draft: JRB, F-CT\u003c/p\u003e\n\u003cp\u003eWriting\u0026mdash;review \u0026amp; editing: JRB, F-CT, AB, JM, SA, Y-WL, GAV, PB\u003c/p\u003e\n\u003cp\u003eFunding acquisition: FCT, Y-WL, GAV, PB\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e The authors declare that they have no competing interests\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAraki N, Egami Y, Watanabe Y, Hatae T (2007) Phosphoinositide metabolism during membrane ruffling and macropinosome formation in EGF-stimulated A431 cells. Exp Cell Res 313:1496\u0026ndash;1507. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.yexcr.2007.02.012\u003c/span\u003e\u003cspan address=\"10.1016/j.yexcr.2007.02.012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaekawa M et al (2014) Sequential breakdown of 3-phosphorylated phosphoinositides is essential for the completion of macropinocytosis. Proc Natl Acad Sci USA 111:E978\u0026ndash;987. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.1311029111\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1311029111\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGillooly DJ, Raiborg C, Stenmark H (2003) Phosphatidylinositol 3-phosphate is found in microdomains of early endosomes. Histochem Cell Biol 120:445\u0026ndash;453. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00418-003-0591-7\u003c/span\u003e\u003cspan address=\"10.1007/s00418-003-0591-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDi Paolo G, De Camilli P (2006) Phosphoinositides in cell regulation and membrane dynamics. Nature 443:651\u0026ndash;657. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nature05185\u003c/span\u003e\u003cspan address=\"10.1038/nature05185\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePosor Y, Jang W, Haucke V (2022) Phosphoinositides as membrane organizers. Nat Rev Mol Cell Biol 23:797\u0026ndash;816. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41580-022-00490-x\u003c/span\u003e\u003cspan address=\"10.1038/s41580-022-00490-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRay J, Sapp DG, Fairn GD (2024) Phosphatidylinositol 3,4-bisphosphate: Out of the shadows and into the spotlight. Curr Opin Cell Biol 88:102372. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ceb.2024.102372\u003c/span\u003e\u003cspan address=\"10.1016/j.ceb.2024.102372\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOverduin M, Kervin TA (2021) The phosphoinositide code is read by a plethora of protein domains. Expert Rev Proteom 18:483\u0026ndash;502. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/14789450.2021.1962302\u003c/span\u003e\u003cspan address=\"10.1080/14789450.2021.1962302\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChandra M et al (2019) Classification of the human phox homology (PX) domains based on their phosphoinositide binding specificities. Nat Commun 10:1528. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-019-09355-y\u003c/span\u003e\u003cspan address=\"10.1038/s41467-019-09355-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKooijman EE, Gericke A (2014) Physical chemistry and biophysics of polyphosphoinositide mediated lipid signaling. Chem Phys Lipids 182:1\u0026ndash;2. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chemphyslip.2014.06.004\u003c/span\u003e\u003cspan address=\"10.1016/j.chemphyslip.2014.06.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKooijman EE, King KE, Gangoda M, Gericke A (2009) Ionization properties of phosphatidylinositol polyphosphates in mixed model membranes. Biochemistry 48:9360\u0026ndash;9371. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/bi9008616\u003c/span\u003e\u003cspan address=\"10.1021/bi9008616\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePylypenko O, Lundmark R, Rasmuson E, Carlsson SR, Rak A (2007) The PX-BAR membrane-remodeling unit of sorting nexin 9. EMBO J 26:4788\u0026ndash;4800. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/sj.emboj.7601889\u003c/span\u003e\u003cspan address=\"10.1038/sj.emboj.7601889\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCeccarelli DFJ et al (2007) Non-canonical interaction of phosphoinositides with pleckstrin homology domains of Tiam1 and ArhGAP9. J Biol Chem 282:13864\u0026ndash;13874. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1074/jbc.M700505200\u003c/span\u003e\u003cspan address=\"10.1074/jbc.M700505200\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu T et al (2020) Molecular basis for PI(3,5)P2 recognition by SNX11, a protein involved in lysosomal degradation and endosome homeostasis regulation. J Mol Biol 432:4750\u0026ndash;4761. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jmb.2020.06.010\u003c/span\u003e\u003cspan address=\"10.1016/j.jmb.2020.06.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou C-Z et al (2003) Crystal structure of the yeast Phox homology (PX) domain protein Grd19p complexed to phosphatidylinositol-3-phosphate. J Biol Chem 278:50371\u0026ndash;50376. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1074/jbc.M304392200\u003c/span\u003e\u003cspan address=\"10.1074/jbc.M304392200\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKerr D, Suwatthee T, Maltseva S, Lee KY (2024) C. Binding equations for the lipid composition dependence of peripheral membrane-binding proteins. Biophys J 123:885\u0026ndash;900. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bpj.2024.02.031\u003c/span\u003e\u003cspan address=\"10.1016/j.bpj.2024.02.031\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoravcevic K, Oxley CL, Lemmon MA (2012) Conditional peripheral membrane proteins: facing up to limited specificity. Structure 20:15\u0026ndash;27. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.str.2011.11.012\u003c/span\u003e\u003cspan address=\"10.1016/j.str.2011.11.012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLemmon MA (2008) Membrane recognition by phospholipid-binding domains. Nat Rev Mol Cell Biol 9:99\u0026ndash;111. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nrm2328\u003c/span\u003e\u003cspan address=\"10.1038/nrm2328\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFalke JJ, Ziemba BP (2014) Interplay between phosphoinositide lipids and calcium signals at the leading edge of chemotaxing ameboid cells. Chem Phys Lipids 182:73\u0026ndash;79. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chemphyslip.2014.01.002\u003c/span\u003e\u003cspan address=\"10.1016/j.chemphyslip.2014.01.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEhret T et al (2023) FERM domains recruit ample PI(4,5)P2s to form extensive protein-membrane attachments. Biophys J 122:1325\u0026ndash;1333. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bpj.2023.02.027\u003c/span\u003e\u003cspan address=\"10.1016/j.bpj.2023.02.027\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYamamoto E et al (2020) Multiple lipid binding sites determine the affinity of PH domains for phosphoinositide-containing membranes. Sci Adv 6:eaay5736. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/sciadv.aay5736\u003c/span\u003e\u003cspan address=\"10.1126/sciadv.aay5736\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYamamoto E, Kalli AC, Yasuoka K, Sansom MSP (2016) Interactions of Pleckstrin Homology Domains with Membranes: Adding Back the Bilayer via High-Throughput Molecular Dynamics. \u003cem\u003eStructure\u003c/em\u003e 24, 1421\u0026ndash;1431 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.str.2016.06.002\u003c/span\u003e\u003cspan address=\"10.1016/j.str.2016.06.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLe Huray KIP, Wang H, Sobott F, Kalli AC (2022) Systematic simulation of the interactions of pleckstrin homology domains with membranes. Sci Adv 8:eabn6992. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/sciadv.abn6992\u003c/span\u003e\u003cspan address=\"10.1126/sciadv.abn6992\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh N et al (2021) Redefining the specificity of phosphoinositide-binding by human PH domain-containing proteins. Nat Commun 12:4339. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-021-24639-y\u003c/span\u003e\u003cspan address=\"10.1038/s41467-021-24639-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYarar D, Surka MC, Leonard MC, Schmid SL (2008) SNX9 activities are regulated by multiple phosphoinositides through both PX and BAR domains. Traffic 9:133\u0026ndash;146. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1600-0854.2007.00675.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1600-0854.2007.00675.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBendris N, Schmid SL (2017) Endocytosis, metastasis and beyond: multiple facets of SNX9. Trends Cell Biol 27:189\u0026ndash;200. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tcb.2016.11.001\u003c/span\u003e\u003cspan address=\"10.1016/j.tcb.2016.11.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimunovic M, Evergren E, Callan-Jones A, Bassereau P (2019) Curving cells inside and out: roles of BAR domain proteins in membrane shaping and its cellular implications. Annu Rev Cell Dev Biol 35:111\u0026ndash;129. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1146/annurev-cellbio-100617-060558\u003c/span\u003e\u003cspan address=\"10.1146/annurev-cellbio-100617-060558\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoulet F, Yarar D, Leonard M, Schmid SL (2005) SNX9 regulates dynamin assembly and is required for efficient clathrin-mediated endocytosis. Mol Biol Cell 16:2058\u0026ndash;2067. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1091/mbc.E04-11-1016\u003c/span\u003e\u003cspan address=\"10.1091/mbc.E04-11-1016\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShin N et al (2007) Sorting nexin 9 interacts with dynamin 1 and N-WASP and coordinates synaptic vesicle endocytosis. J Biol Chem 282:28939\u0026ndash;28950. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1074/jbc.M700283200\u003c/span\u003e\u003cspan address=\"10.1074/jbc.M700283200\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaste F et al (2017) Control of actin polymerization via the coincidence of phosphoinositides and high membrane curvature. J Cell Biol 216:3745\u0026ndash;3765. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1083/jcb.201704061\u003c/span\u003e\u003cspan address=\"10.1083/jcb.201704061\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGallop JL, Walrant A, Cantley LC, Kirschner MW (2013) Phosphoinositides and membrane curvature switch the mode of actin polymerization via selective recruitment of toca-1 and Snx9. Proc Natl Acad Sci USA 110:7193\u0026ndash;7198. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.1305286110\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1305286110\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSch\u0026ouml;neberg J et al (2017) Lipid-mediated PX-BAR domain recruitment couples local membrane constriction to endocytic vesicle fission. Nat Commun 8:15873. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/ncomms15873\u003c/span\u003e\u003cspan address=\"10.1038/ncomms15873\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYarar D, Waterman-Storer CM, Schmid SL (2007) SNX9 couples actin assembly to phosphoinositide signals and is required for membrane remodeling during endocytosis. Dev Cell 13:43\u0026ndash;56. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.devcel.2007.04.014\u003c/span\u003e\u003cspan address=\"10.1016/j.devcel.2007.04.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePosor Y et al (2013) Spatiotemporal control of endocytosis by phosphatidylinositol-3,4-bisphosphate. Nature 499:233\u0026ndash;237. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nature12360\u003c/span\u003e\u003cspan address=\"10.1038/nature12360\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSorre B et al (2012) Nature of curvature coupling of amphiphysin with membranes depends on its bound density. Proc Natl Acad Sci USA 109:173\u0026ndash;178. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.1103594108\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1103594108\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePr\u0026eacute;vost C et al (2015) IRSp53 senses negative membrane curvature and phase separates along membrane tubules. Nat Commun 6:8529. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/ncomms9529\u003c/span\u003e\u003cspan address=\"10.1038/ncomms9529\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsai F-C et al (2021) Comparing physical mechanisms for membrane curvature-driven sorting of BAR-domain proteins. Soft Matter 17:4254\u0026ndash;4265. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1039/d0sm01573c\u003c/span\u003e\u003cspan address=\"10.1039/d0sm01573c\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBertin A et al (2020) Human ESCRT-III polymers assemble on positively curved membranes and induce helical membrane tube formation. Nat Commun 11:2663. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-020-16368-5\u003c/span\u003e\u003cspan address=\"10.1038/s41467-020-16368-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMim C et al (2012) Structural basis of membrane bending by the N-BAR protein endophilin. Cell 149:137\u0026ndash;145. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cell.2012.01.048\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2012.01.048\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJumper J et al (2021) Highly accurate protein structure prediction with AlphaFold. Nature 596:583\u0026ndash;589. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-021-03819-2\u003c/span\u003e\u003cspan address=\"10.1038/s41586-021-03819-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLo W-T et al (2017) A Coincidence Detection Mechanism Controls PX-BAR Domain-Mediated Endocytic Membrane Remodeling via an Allosteric Structural Switch. Dev Cell 43:522\u0026ndash;529e524. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.devcel.2017.10.019\u003c/span\u003e\u003cspan address=\"10.1016/j.devcel.2017.10.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan K, Gericke A, Pastor RW (2020) Characterization of Specific Ion Effects on PI(4,5)P2 Clustering: Molecular Dynamics Simulations and Graph-Theoretic Analysis. J Phys Chem B 124:1183\u0026ndash;1196. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jpcb.9b10951\u003c/span\u003e\u003cspan address=\"10.1021/acs.jpcb.9b10951\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGewinner C et al (2009) Evidence that inositol polyphosphate 4-phosphatase type II is a tumor suppressor that inhibits PI3K signaling. Cancer Cell 16:115\u0026ndash;125. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ccr.2009.06.006\u003c/span\u003e\u003cspan address=\"10.1016/j.ccr.2009.06.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu J, Alvarez FJD, Clare DK, Noel JK, Zhang P (2021) CryoEM structure of the super-constricted two-start dynamin 1 filament. Nat Commun 12:5393. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-021-25741-x\u003c/span\u003e\u003cspan address=\"10.1038/s41467-021-25741-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark J et al (2010) SNX18 shares a redundant role with SNX9 and modulates endocytic trafficking at the plasma membrane. J Cell Sci 123:1742\u0026ndash;1750. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1242/jcs.064170\u003c/span\u003e\u003cspan address=\"10.1242/jcs.064170\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiao Z, Si T, Kai J-J, Fan J (2024) Mechanism of Membrane Curvature Induced by SNX1: Insights from Molecular Dynamics Simulations. J Phys Chem B 128:2144\u0026ndash;2153. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jpcb.3c07009\u003c/span\u003e\u003cspan address=\"10.1021/acs.jpcb.3c07009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsai F-C et al (2022) Activated I-BAR IRSp53 clustering controls the formation of VASP-actin-based membrane protrusions. Sci Adv 8:eabp8677. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/sciadv.abp8677\u003c/span\u003e\u003cspan address=\"10.1126/sciadv.abp8677\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJarin Z, Pak AJ, Bassereau P, Voth GA (2021) Lipid-Composition-Mediated Forces Can Stabilize Tubular Assemblies of I-BAR Proteins. Biophys J 120:46\u0026ndash;54. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bpj.2020.11.019\u003c/span\u003e\u003cspan address=\"10.1016/j.bpj.2020.11.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePicas L et al (2014) BIN1/M-Amphiphysin2 induces clustering of phosphoinositides to recruit its downstream partner dynamin. Nat Commun 5:5647. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/ncomms6647\u003c/span\u003e\u003cspan address=\"10.1038/ncomms6647\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoucrot E et al (2012) Membrane fission is promoted by insertion of amphipathic helices and is restricted by crescent BAR domains. Cell 149:124\u0026ndash;136. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cell.2012.01.047\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2012.01.047\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIsas JM, Ambroso MR, Hegde PB, Langen J, Langen R (2015) Tubulation by amphiphysin requires concentration-dependent switching from wedging to scaffolding. Structure 23:873\u0026ndash;881. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.str.2015.02.014\u003c/span\u003e\u003cspan address=\"10.1016/j.str.2015.02.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeinberger A et al (2013) Gel-assisted formation of giant unilamellar vesicles. Biophys J 105:154\u0026ndash;164. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bpj.2013.05.024\u003c/span\u003e\u003cspan address=\"10.1016/j.bpj.2013.05.024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchindelin J et al (2012) Fiji: an open-source platform for biological-image analysis. Nat Methods 9:676\u0026ndash;682. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nmeth.2019\u003c/span\u003e\u003cspan address=\"10.1038/nmeth.2019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSorre B et al (2009) Curvature-driven lipid sorting needs proximity to a demixing point and is aided by proteins. Proc Natl Acad Sci USA 106:5622\u0026ndash;5626. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.0811243106\u003c/span\u003e\u003cspan address=\"10.1073/pnas.0811243106\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJo S, Kim T, Iyer VG, Im W (2008) CHARMM-GUI: a web-based graphical user interface for CHARMM. J Comput Chem 29:1859\u0026ndash;1865. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jcc.20945\u003c/span\u003e\u003cspan address=\"10.1002/jcc.20945\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng S, Park S, Choi YK, Im WCHARMM-GUI (2023) Membrane Builder: Past, Current, and Future Developments and Applications. J Chem Theory Comput 19:2161\u0026ndash;2185. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jctc.2c01246\u003c/span\u003e\u003cspan address=\"10.1021/acs.jctc.2c01246\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee J et al (2016) CHARMM-GUI Input Generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM Simulations Using the CHARMM36 Additive Force Field. J Chem Theory Comput 12:405\u0026ndash;413. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jctc.5b00935\u003c/span\u003e\u003cspan address=\"10.1021/acs.jctc.5b00935\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGowers R et al in \u003cem\u003ePython in Science Conference.\u003c/em\u003e 98\u0026ndash;105 (SciPy)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePettersen EF et al (2021) UCSF ChimeraX: structure visualization for researchers, educators, and developers. Protein Sci 30:70\u0026ndash;82. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/pro.3943\u003c/span\u003e\u003cspan address=\"10.1002/pro.3943\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHumphrey W, Dalke A, Schulten K (1996) VMD: visual molecular dynamics. J Mol Graph 14:33\u0026ndash;38. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/0263-7855(96)00018-5\u003c/span\u003e\u003cspan address=\"10.1016/0263-7855(96)00018-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6573900/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6573900/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePlasma membrane remodeling processes are tightly regulated by the spatiotemporal distribution and dynamic conversion of phosphoinositidyl lipids (PIPs). This regulation is controlled by the recruitment of proteins such as sorting nexin 9 (SNX9), a key mediator of late-stage endocytosis and macropinocytosis. Using live cell imaging, \u003cem\u003ein vitro\u003c/em\u003e reconstitution assays, and molecular dynamics simulations, we investigated how SNX9 distinguishes between PI(3,4)P\u003csub\u003e2\u003c/sub\u003e and PI(4,5)P\u003csub\u003e2\u003c/sub\u003e, and the physiological relevance of this selectivity. Our results revealed that during macropinocytic membrane ruffling, SNX9 is recruited in a spatiotemporally coordinated manner with PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, but not with PI(4,5)P\u003csub\u003e2\u003c/sub\u003e. While SNX9 induces comparably weak mechanical remodeling on model membranes containing either PIP\u003csub\u003e2\u003c/sub\u003e species, it exhibits a clear selective binding to PI(3,4)P\u003csub\u003e2\u003c/sub\u003e, mediated by a non-canonical interface. Through mutational analysis of key residues involved in this sequestration, we further demonstrated that SNX9 protects PI(3,4)P\u003csub\u003e2\u003c/sub\u003e from hydrolysis. Together, these results reveal a previously unrecognized mechanism of SNX9-PIP\u003csub\u003e2\u003c/sub\u003e lipid interaction that underscores the pivotal role of SNX9 in coordinating membrane remodeling processes.\u003c/p\u003e","manuscriptTitle":"A Non-Canonical Interface of SNX9 PX Domain Selectively Sequesters PI(3,4)P2 Lipids, Protecting Them from Hydrolysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-30 16:22:16","doi":"10.21203/rs.3.rs-6573900/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"76b21eaf-befb-48b4-8794-eac3af5f1c45","owner":[],"postedDate":"May 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":49154780,"name":"Biological sciences/Biophysics/Membrane structure and assembly"},{"id":49154781,"name":"Biological sciences/Cell biology/Membrane trafficking/Membrane curvature"},{"id":49154782,"name":"Biological sciences/Biophysics/Computational biophysics"},{"id":49154783,"name":"Biological sciences/Cell biology/Membrane trafficking/Endocytosis"}],"tags":[],"updatedAt":"2025-07-05T00:10:37+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-30 16:22:16","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6573900","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6573900","identity":"rs-6573900","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}