Oxysterol binding protein (OSBP) extracts cholesterol from the ER to deliver it to the TGN via counter exchange and subsequent hydrolysis of the phosphoinositide PI(4)P. Here, we show that this pathway is essential in polarized epithelial cells where it contributes not only to the proper subcellular distribution of cholesterol but also to the trans-Golgi sorting and trafficking of numerous plasma membrane cargo proteins with apical or basolateral localization. Reducing the expression of OSBP, blocking its activity, or inhibiting a PI4Kinase that fuels OSBP with PI(4)P abolishes the epithelial phenotype. Waves of cargo enrichment in the TGN in phase with OSBP and PI(4)P dynamics suggest that OSBP promotes the formation of lipid gradients along the TGN, which helps cargo sorting. During their transient passage through the trans-Golgi, polarized plasma membrane proteins get close to OSBP but fail to be sorted when OSBP is silenced. Thus, OSBP lipid exchange activity is decisive for polarized cargo sorting and distribution in epithelial cells.

Epithelial cells line the surfaces of our body and are differentiated to form tightly connected cellular sheets that act as biological barriers. As a consequence of tight junction formation, the epithelial plasma membrane is divided into distinct apical and basolateral domains, which are equipped with different sets of proteins and lipids. While glycosylphosphatidylinositol (GPI)-anchored proteins are mostly present on the apical surface, cadherins and integrins are located on the basolateral plasma membrane domain to form adherent junctions and focal adhesions, respectively (Paladino et al., 2006; Caceres et al., 2019; Keller et al., 2001; Lebreton et al., 2019). Additionally, compared with the basolateral plasma membrane, the apical membrane is more ordered and enriched in cholesterol, demonstrating the polarized distribution of membrane lipids within the epithelial plasma membrane (Gerl et al., 2012).

To achieve and maintain this cell polarity, cargoes are sorted in the trans-Golgi network (TGN) and directed toward the apical and basolateral membranes using distinct secretory routes. Although TGN sorting signals and their recognizing machinery have been well-characterized, the molecular mechanisms of clustered cargo trafficking are still largely unknown (Ramazanov et al., 2021; Boncompain and Weigel, 2018). It has recently been suggested that cargo-specific sorting and subsequent trafficking routes are primarily regulated by the dynamic adaptation of the TGN to the particular cargo (Boncompain and Weigel, 2018). Furthermore, apical and basolateral cargoes are able to segregate from each other and follow distinct cognate routes even in non-epithelial cells, suggesting that apicobasal cargo sorting is not limited to polarized cells and might be governed by the cargo itself (Yoshimori et al., 1996).

It has long been recognized that the lipid environment affects protein sorting and trafficking (Keller and Simons, 1997, 1998). Lipid composition defines the biophysical properties of the membranes, which might facilitate the formation of different TGN subdomains and the recruitment of specific groups of proteins leading to the assembly of cargo-specific machinery (von Blume and Hausser, 2019; Klemm et al., 2009). For instance, apically sorted GPI-anchored proteins are clustered in cholesterol-rich membrane regions in the TGN, which drives their lateral segregation from other cargo types (Paladino et al., 2004; Simons and Ikonen, 1997; Zurzolo and Simons, 2016). However, our understanding of the link between lipid environment and cargo sorting has remained fragmented due to the difficulty in tuning the lipid composition of organelles both specifically and locally.

Recent advances in the characterization of membrane contact sites (MCS) offer new strategies to study the impact of lipid composition on membrane traffic steps. MCS are regions of close apposition between organelles, which are enriched in lipid transfer proteins (Wu et al., 2018). Oxysterol-binding protein (OSBP)—a member of the OSBP-related protein family (ORPs)—is recruited to the ER-TGN MCS to transfer cholesterol from the endoplasmic reticulum (ER) to the TGN, thereby controlling the lipid composition of trans-Golgi membranes (Antonny et al., 2018; Mesmin et al., 2019). This transfer activity requires the metabolic energy of the lipid phosphatidylinositol-4-phosphate (PI(4)P), which has an unequal distribution between the ER and TGN (Antonny et al., 2018; Mesmin et al., 2013, 2017). Due to the activity of TGN-resident PI4-kinases, PI4KIIα and PI4KIIIβ, the PI(4)P concentration is high in the TGN, while it is subjected to degradation in the ER by the phosphatase Sac-1. When OSBP transfers a cholesterol molecule to the TGN, it extracts one PI(4)P from the TGN and transfers it back to the ER. As such, ER to TGN forward transfer of cholesterol occurs owing to the backward transfer and subsequent hydrolysis of PI(4)P (Mesmin et al., 2013, 2017).

Because cholesterol is recognized as a major lipid that controls cargo trafficking (von Blume and Hausser, 2019; Dukhovny et al., 2009; Sugiki et al., 2012), we surmised that OSBP should affect the sorting and the subsequent trafficking of cargoes that depend on the local amount of this lipid. Here, using various proteomics and cell biology approaches, we report that OSBP regulates polarized cargo trafficking in epithelial cells. We found that the post-Golgi trafficking of numerous apical and basolateral proteins depends on OSBP. Moreover, live-cell experiments suggest that OSBP dynamics at the TGN membranes drive the sorting of cargoes with polarized distribution. These observations provide a better understanding of cargo sorting for epithelial polarity, which has multiple impacts on pathological aspects as well, notably in cancers of epithelial origins.

We previously reported that OSBP contact sites are highly dynamic in hTERT-RPE1 cells—a cell line with a well-developed TGN, ideal for real-time imaging. They show traveling waves that are sensitive to changes in PI(4)P synthesis as well as to perturbations of membrane saturation and protein crowding (Mesmin et al., 2017; Jamecna et al., 2019). Notably, gradual inhibition of PI4KIIIβ by PIK93 increases the amplitude of these waves, reflecting a tug-of-war between PI(4)P synthesis by two PI4-kinases (PI4KIIα and PI4KIIIβ) and OSBP-dependent consumption of PI(4)P (Mesmin et al., 2017). We aimed to determine whether cell surface cargoes in the TGN follow the OSBP oscillations by exploiting the Retention Using Selective Hooks (RUSH)-system (Boncompain et al., 2012). For this, we coexpressed a RUSH plasmid expressing EGFP-GPI together with the PI(4)P probe mCherry-PHOSBP, which can be used as a reporter of OSBP contact sites in hTERT-RPE1 cells (Mesmin et al., 2013). We released EGFP-GPI from the ER by biotin addition and then enhanced the PI(4)P waves by adding PIK93 when the cargo started to accumulate at the TGN. Strikingly, we found that Golgi release of EGFP-GPI followed the dynamics of OSBP oscillations as we observed that the curves corresponding to mCherry-PHOSBP and EGFP-GPI were in phase (Fig. 1, A–C). This observation suggests that TGN dynamics of surface cargoes are synchronized with the OSBP cycle.

Next, we aimed to determine whether OSBP dynamics affect apicobasal cargo sorting as well. For this, we used the epithelial cell line MDCK, which showed noticeable PI(4)P waves with high amplitudes already at steady state (Fig. 1, D and E). In these cells, the amplitude of the PI(4)P waves reported by the mCherry-PHOSBP probe diminished upon PIK93 addition, probably due to the dissociation of OSBP from the TGN membranes (Fig. 1, D and E; and Fig. S1 A). We exploited this effect of PI4KIIIβ inhibition in MDCK cells to test the contribution of PI(4)P dynamics to apicobasal cargo sorting. We cotransfected MDCK cells with RUSH constructs expressing apically sorted EGFP-GPI and basolateral mCherry-CDH1 model cargoes. Following the induction of cargo release by biotin addition, we synchronized the two cargoes at the TGN by a 1.5 h temperature block at 19.5°C. Following this step, cells were either fixed directly or further incubated at 37°C for 40 min with or without PIK93 before fixation to let the cargoes leave the TGN. After the temperature block, EGFP-GPI and mCherry-CDH1 colocalized at the TGN (Fig. 1 F). Following a 40 min incubation at 37°C, EGFP-GPI and mCherry-CDH1 in the DMSO-treated control cells were detected mostly in distinct post-Golgi vesicles, indicating that the two cargo types had segregated from each other (Fig. 1, F and G). When the cells were treated with PIK93 after the temperature block, the percentage of post-Golgi vesicles containing both cargoes increased significantly, indicating a defect in cargo sorting (Fig. 1, F and G).

We also assessed the effect of PI4KIIIβ inhibition on the development of MDCK cysts. When MDCK cells were left to develop cysts in the Matrigel matrix for 72 h, most of the cells from the DMSO-treated control sample developed cysts containing a single central lumen, indicating an efficient polarity establishment. In the presence of PIK93, however, a significantly higher number of cysts with luminal defects were detected. Typically, these cysts displayed multiple lumens and contained mislocalized surface cargoes, suggesting a defect in polarized cargo sorting (Fig. 1, H and I; and Fig. S1 B).

Because PIK93 is reported to inhibit some PI3Ks as well, we repeated these experiments using a more specific inhibitor PI4KIIIbeta-10-IN (Knight et al., 2006). Similar to PIK93, this compound triggered the partial dissociation of OSBP from the TGN membrane (Fig. S1 C) and impaired apicobasal RUSH cargo sorting (Fig. S1, D and E). Furthermore, the presence of PI4KIIIbeta-10-IN significantly increased the frequency of cysts with morphological defects (Fig. S1, F and G), confirming the importance of PI(4)P metabolism in cargo sorting.

Next, we investigated whether OSBP silencing also leads to apicobasal sorting defects. After confirming efficient silencing (Fig. S1 H), we tested the dynamics of CDH1 and EGFP-GPI trafficking upon OSBP depletion. Both cargoes showed similar dynamics upon siOSBP treatments (Fig. 1, J and K; and Fig. S1 I). In control cells nucleofected with siNT, the cargoes typically reached the TGN within 30–40 min after ER release; thereafter, they left the TGN and appeared at the cell surface. Although cargoes in OSBP-silenced cells were able to leave the ER, we observed significant delays in the post-Golgi trafficking, indicating that OSBP activity is necessary for efficient cargo release from the TGN (Fig. 1, J and K; and Fig. S1 I).

To test whether this delay in transport is coupled to sorting defects as well, we quantified EGFP-GPI and mCherry-CDH1-containing post-Golgi vesicles following temperature block release in control siRNA-transfected and OSBP-silenced cells. The number of post-Golgi vesicles containing both cargo types significantly increased upon OSBP silencing, indicating that OSBP activity contributes to the segregation of apical and basolateral cargoes (Fig. 1, L and M).

Finally, we compared the morphology of 3D-polarized cysts developed from control and OSBP-silenced MDCK cells. While only ∼35% of the cysts showed irregular morphology in the control siRNA-nucleofected cells, ∼50% of the OSBP-silenced cyst population showed luminal morphology defects. Typically, these cysts developed multiple lumens and displayed apical multipolarity where one cell formed several apical surfaces, presumably due to the mis-sorting of apical cargoes (Fig. 1, N–P; and Fig. S1 J).

We concluded that the OSBP/PI(4)P cycle at TGN membranes drives the segregation of cargoes and, as such, governs their proper trafficking to establish epithelial cell polarity.

Because OSBP regulates apicobasal sorting and cargo exit at the TGN, we assessed whether polarized cargoes approach OSBP-containing MCSs during their Golgi trafficking. For this aim, we coexpressed four model RUSH cargoes with various topologies and final destinations with the catalytically inactive form of OSBP, mCherry-OSBPHHK>AAA, which can be used as a passive reporter for OSBP at the contact sites. These chosen cargoes were CDH1–E-cadherin (basolateral), DSG2–Desmoglein2 (non-polarized), EGFP–GPI (apical), and GP135–Podocalyxin Like (apical). We initiated their release from the ER by biotin addition; then cells were subsequently kept at 19.5°C for 1.5 h to accumulate cargoes in the trans-Golgi. Following this, samples were switched back to 37°C to monitor cargo export dynamics by live imaging. We exploited the blocking effect of the ORPphilin OSW-1 on OSBP. Mechanistically, ORPphilins—such as OSW-1 and SWG—inhibit OSBP-mediated lipid exchange, thereby increasing the level of PI(4)P at the TGN, which results in the forced recruitment and subsequent stabilization of inactive OSBP at contact sites (Mesmin et al., 2017).

At 19.5°C, cargoes colocalized with mCherry-OSBPHHK>AAA, indicating their accumulation at the TGN. In control cells, the shift to 37°C released the cargoes from the TGN, while mCherry-OSBPHHK>AAA still localized to the TGN. Treatments with OSW-1 strongly impaired TGN export of the CDH1 and DSG2 cargoes, as demonstrated by a persistent co-localization with OSBP over time (Fig. 2, A and B; and Fig. S2, A and B). In contrast, the apically-sorted proteins EGFP-GPI and GP135 were able to leave the Golgi, albeit at a slower rate than in controls (Fig. 2, C and D; and Fig. S2, C and D). As expected, OSW-1 treatments triggered mCherry-OSBPHHK>AAA recruitment to the Golgi due to inhibition of PI(4)P turnover (Fig. 2, A–D; and Fig. S2, A–D). These colocalization experiments suggest that both apical and basolateral cargoes require OSBP-dependent MCSs during their transit at the Golgi apparatus; however, basolateral model cargoes appear more sensitive to OSW-1.

Next, we tested the effect of OSBP inhibition on MDCK cysts. Strikingly, overnight treatment of polarized MDCK cysts with the OSBP-blocking drug SWG led to a drastic loss of cell polarity as the lumens disappeared and the cysts acquired a grape-like morphology consisting of rounded cells showing no asymmetrical marker distribution (Fig. 2 E and Fig. S2 E). Interestingly, we detected less E-cadherin and β-catenin signals at cell–cell junctions in SWG-treated cysts. Additionally, GP135 was detected in plasma membrane domains facing the extracellular matrix as well (Fig. S2 E).

We tested the effect of SWG on MDCK cells under conditions where their polarity is minimal; that is when they form 2D colonies. To observe morphology changes of MDCK colonies, we performed overnight time-lapse imaging after the addition of DMSO or the OSBP inhibitor SWG (Fig. 2 F and Video 1). In the DMSO-treated control, we observed basal expansion of the colonies due to cell proliferation; however, after a few hours of SWG treatment, the colonies started to scatter, indicating that OSBP inhibition perturbs cell–cell and cell–substrate anchoring functions.

Altogether, these results indicate that OSBP-engaged MCSs govern apical and basolateral sorting routes and contribute to the maintenance of epithelial polarity.

One caveat of the RUSH assay is to rely on overexpressed cargoes, which might saturate the endogenous sorting machinery of the cell and could lead to sorting defects. Therefore, we aimed to determine the repertoire of endogenous cargoes that transiently become close to OSBP during their sorting. For this, we chose a proximity ligation strategy in which the promiscuous biotin ligase TurboID was fused to the C-terminus of OSBP (Fig. S3 A). Upon biotin addition, OSBP-TurboID biotinylates proteins close to OSBP. Following pull-down with streptavidin beads, we identified these proteins by mass spectrometry (Branon et al., 2018).

We performed the OSBP-TurboID assays either in the absence or in the presence of ORPphilins to take advantage of the effect of these drugs on OSBP localization and lipid exchange activity. When OSBP is soluble in the cytoplasm, a high number of cytoplasmic proteins should be biotinylated. When OSBP is recruited to the MCSs, both ER and TGN-localized proximity proteins can be screened because the TurboID tag is appended to the OSBP C-terminal region, which transiently interacts with the two organelles. Due to its preference for disordered membranes, OSBP slides toward PI(4)P-rich regions along the TGN upon lipid exchange (Mesmin et al., 2017). When ORPphilins inhibit the lipid-transfer activity of OSBP, cholesterol levels drop and PI(4)P levels rise in the TGN membranes, which then provides new anchoring points for OSBP. This forced recruitment of OSBP leads to the formation of crowded, non-dynamic MCSs along the ER-TGN interface (Fig. 3 A; Burgett et al., 2011; Mesmin et al., 2017; Péresse et al., 2020). In good agreement with this, in MDCK cells stably expressing OSBP-TurboID, the fusion bait was detected both in the cytoplasm and in βGalT1-positive Golgi structures (Fig. 3 B). Upon treatment with OSW-1 or SWG for 60 min prior to biotinylation, OSBP-TurboID showed increased recruitment to TGN membranes at the expense of the soluble form (Fig. 3, B and C). Remarkably, incubating the OSBP-TurboID-expressing cells for 2 h in a biotin-free medium after a 10 min biotin exposure led to the appearance of the biotinylation signal at the cell surface, whereas no such signal was observed when the cells were fixed directly after the 10 min biotin treatment (Fig. S3 B). This observation provided an additional hint that a large proportion of cell surface proteins visit OSBP-containing MCSs during their trafficking.

For proteomic analysis, we exposed MDCK cells stably expressing OSBP-TurboID to either DMSO, SWG, or OSW-1 for 60 min followed by biotin addition for the last 10 min of the ORPphilin treatments. We verified successful protein biotinylation and purification by streptavidin-HRP blotting and silver staining of the elution fractions (Fig. S3, C and D). Following on-bead trypsinization, peptides were subjected to mass spectrometry and proteins were identified by subsequent proteomics analyses.

Overall, we identified 1,507 proteins as members of the OSBP proximity landscape. In the DMSO-treated cells, we pinpointed many proteins already known to interact or co-localize with OSBP at MCSs. These include both VAP isoforms (VAPA and VAPB), the PI(4)P generating kinase PI4KIIIβ (PI4KB), and the ER-resident phosphatase SAC1 (SACM1L). Additionally, TMED2, the ceramide transporter CERT1 and the glucosylceramide-transfer protein FAPP2 (PLEKHA8) were also detected, confirming their collaboration with OSBP at the ER-TGN contact sites (Fig. 3 D; Anwar et al., 2022; Kumagai and Hanada, 2019; Mesmin et al., 2019).

To functionally annotate the OSBP proximity proteome, we performed a statistical enrichment analysis in cellular component GO terms. As expected, statistically significant enrichments were found in multiple GO groups encompassing ER and Golgi-resident proteins. However, other GO terms such as “plasma membrane,” “cell junction,” and “anchoring junction” showed significant enrichments as well, suggesting potential cargo clients of OSBP (Fig. S3 E). Besides the numerous adherent junction components (e.g., cadherins and catenins), we identified tight junction proteins (e.g., TJP1 and 2, Occludin, MarvelD2, and Claudin), integrins, CD44, EpCam, and the apical marker proteins PODXL (Gp135) and MUC1, all of which are regulators or determinants of epithelial polarity (Table S1).

We aimed to see whether cargoes with apical or basolateral localization are preferentially enriched in the OSBP proximity. For this, we correlated the abundances of the identified surface proteins among the proximity partners (expressed as -log-transformed P value over the background) with their apicobasal distribution (polarity value). These polarity values were determined by Caceres et al. (2019), who selectively labeled and quantified the apical and basolateral surface proteome of polarized MDCK cells using mass spectrometry. Strikingly, we observed no correlation, indicating that apical and basolateral cargoes are equally detected in the vicinity of active OSBP (Fig. S3 F).

Next, we assessed the effect of the OSBP inhibitors ORPphilins on the proximity proteome of OSBP (Fig. 3, E and F). For this, we divided the hits into three categories according to their fold change upon SWG or OSW-1 treatments over the DMSO-treated control: hits that showed no significant change, hits with increased (up), and hits with decreased (down) abundances. Then, we mapped specific classes of hits to the volcano plots and calculated their distribution along these three categories. Among all detected hits (total), ∼25% of the proteins became more abundant (up) around OSBP upon SWG treatment, ∼30% became less abundant (down), and about 45% of them showed no significant change (Fig. 3 E). Similar changes were detected after OSW-1 treatment (Fig. 3 F). As expected, the majority of ER and Golgi-resident proteins showed increased abundance around OSBP after SWG treatment, while most cytosolic proteins became less abundant (Fig. 3, E and F). These variations were in good agreement with imaging data (Fig. 3 B), showing that the proximity proteome of OSBP changes together with its subcellular localization.

Finally, we assessed the effect of ORPphilin treatments on secretory cargo abundance in the OSBP proximity landscape. For this, we once again used the Caceres dataset to assign apicobasal-distribution values to the identified cargoes in the proximity of OSBP (Caceres et al., 2019). First, we ranked the hits according to their polarized localization across the polarized MDCK plasma membrane. Thereafter, we selected extreme proteins (proteins with either high apical or basolateral localization) and we mapped these polarized surface proteins on the OSBP proximity landscape (Fig. S3 G; and Fig. 3, G and H). Strikingly, the abundance of most of the basolateral cargoes increased in the OSBP proximity upon ORPphilin treatments, whereas the majority of the apical cargoes became less abundant (Fig. 3, G and H). This suggests that when OSBP contact sites extend but are deficient in lipid exchange due to ORPphilin treatments, they prefer to populate the TGN regions that favor the sorting of basolateral cargo proteins.

Overall, the proximity ligation strategy confirmed that endogenous apical and basolateral cargoes become close to OSBP during their Golgi transit. In addition, it suggested that apical and basolateral cargoes are segregated from each other in a manner that correlates with the positioning of OSBP-dependent MCSs on TGN membranes.

In the RUSH assays, we found that the Golgi exit of apical cargoes is less susceptible to OSBP inhibitors than that of the basolateral cargoes. The proximity proteome confirmed that both apical and basolateral trafficking routes approach OSBP, and through the impact of ORPphilin treatments, it suggested that apical and basolateral cargoes are sorted across the lipid gradient generated by OSBP. However, the TurboID assay cannot report on the subsequent fate of the identified endogenous cargoes upon ORPphilin treatments. To directly address which cell surface proteins are perturbed by OSBP inhibition, we conducted a second proteomic analysis, now focusing on the dynamics of the MDCK surface proteome upon OSBP targeting.

We monitored the dynamics of the MDCK surface proteome after 0, 6, and 12 h SWG treatments using quantitative mass spectrometry (Fig. S3 H). Compared with time 0, we observed that 17 and 32% of the total detected surface proteome showed significant abundance changes upon 6 and 12 h SWG treatments, respectively (Fig. S3 H and Table S2). To define groups of surface proteins responding similarly to SWG treatment, we gathered the proteins that exhibited similar changes over the time course of the experiment using k-means clustering (Fig. 4 A). Cluster 1 gathers proteins showing a rapid decrease in abundance at the cell surface during the time course of the SWG treatment. Cluster 2 corresponds to proteins showing a gradual increase. Cluster 3 gathers proteins showing a slow decrease in surface expression upon SWG treatment. Last, proteins that showed a transient increase at the cell surface were grouped in Cluster 4 (Fig. 4 B and Table S2). Importantly, the majority of proteins were found in Clusters 1 and 3, indicating that SWG treatment downregulates the surface expression of a large number of proteins (Fig. 4 B).

We examined whether surface expression of apical and basolateral surface proteins show similar sensitivity to SWG treatment. For this, we performed an analysis similar to the one shown in Fig. 3, G and H. Using the Caceres dataset (Caceres et al., 2019), we identified highly apical and basolateral surface proteins and calculated their distribution across the four clusters. Strikingly, most of the basolateral surface proteins were gathered in Cluster 1, whereas the majority of the apical proteins were grouped in Cluster 3 (Fig. 4 C).

To confirm that basolateral cargoes are arrested in the TGN upon OSBP blockage, we localized endogenous E-cadherin in control and ORPphilin-treated MDCK cells using immunofluorescence (Fig. 4 D). As expected, E-cadherin disappeared from the surface after 4 h of ORPphilin treatments, and it concomitantly accumulated in intracellular structures positive for the Golgi resident βGalT1 (Fig. 4, E and F).

It has been established that epithelial cadherins have a very short half-life and that their surface-localized pool undergoes full renewal within hours (Brüser and Bogdan, 2017; Bryant and Stow, 2004; Cavey et al., 2008; McCrea and Gumbiner, 1991). When cultured in Ca2+-free conditions, epithelial cells internalize and subsequently degrade cadherin molecules in a proteasome- and lysosome-dependent manner (Yamada et al., 2005; Shen et al., 2008). However, when Ca2+ concentration is re-established, neosynthesized E-cadherin forms stable cell–cell junctions within a few hours. In our assay, control MDCK cells were able to recover E-cadherin-based cell–cell junctions 2 h following overnight Ca2+ deprivation, whereas E-cadherin accumulated in the Golgi and failed to reach the cell surface in cells treated with ORPphilins during recovery (Fig. 4, G and H).

We concluded that the trafficking of basolateral proteins is more susceptible to OSBP blockage, although surface expression of both apical and basolateral proteins is perturbed by OSBP targeting.

At the molecular level, OSBP is not involved in direct interactions with vesicular trafficking machinery; instead, it directs cholesterol-PI(4)P lipid exchange at ER-TGN contact sites. In hTERT-RPE1 cells, this activity is massive and not only conditions the local distribution of cholesterol and PI(4)P but also determines the gradient of lipid order along the secretory pathway, including the lipid composition of the plasma membrane (Mesmin et al., 2017).

To investigate how this effect applies to the lipid composition of the polarized cell membrane domains, we assessed the effect of OSBP manipulations in polarized MDCK cells. First, we directly analyzed the distribution of cholesterol using cells stably expressing the cholesterol biosensor D4H-EGFP. In control siRNA conditions, the probe D4H-EGFP decorated the cell membranes of the polarized cells and an enrichment of the fluorescent signal was detected at the basolateral domain of the plasma membrane (Fig. 5 A and Fig. S4 A). Upon OSBP silencing, a fraction of cysts exhibited morphological defects (see Fig. 1). Importantly, those cysts also showed altered cholesterol distribution: the D4H-EGFP cholesterol probe decreased at the plasma membrane and appeared in small, cytoplasmic structures (Fig. 5 A and Fig. S4 A). Similarly, the probe disappeared from the plasma membrane of MDCK cysts upon overnight treatment with SWG. Compared with OSBP silencing, the drug treatment exhibited a more potent effect (Fig. 5 B and Fig. S4 B).

Cholesterol is enriched and complexes with sphingomyelin at the apical surface of polarized epithelial cells (Gerl et al., 2012). However, due to its restricted affinity for free, accessible cholesterol pools, D4H does not reliably report apically localized cholesterol (Maekawa and Fairn, 2015; Maekawa, 2017). To test whether OSBP supplies cholesterol toward both apical and basolateral membranes or whether it is selective to one of these domains, we used filipin staining to visualize plasma membrane cholesterol in MDCK cysts treated with SWG at various times. In control cysts, we measured higher fluorescent intensities in the apical membrane domains than in the basolateral segments, confirming the polarized distribution of cholesterol (Fig. 5 C). Remarkably, after 5 h of OSBP inhibition, conditions under which the cyst structure was still visible, significant drops in cholesterol levels in both apical and basolateral plasma membrane regions were observed (Fig. 5 C). After 16 h of SWG exposition, the cysts lost cell polarity and we observed a further drop in cell-surface cholesterol levels (Fig. 5 C).

We also used the plasma membrane-selective and solvatochromic polarity probe NR12A (Danylchuk et al., 2021) to detect changes in membrane lipid order upon OSBP inhibition/silencing. In non-polarized cells, we observed that ORPphilin treatments (Fig. 5 D and Fig. S4, C–E) or OSBP silencing (Fig. 5 E and Fig. S4, F–H) decreased the ratio values of NR12A, indicating that OSBP targeting leads to a less ordered plasma membrane. On filter-polarized MDCK cells, we observed that ratio values corresponding to both apical and basolateral domains shifted toward smaller values following 5 h SWG treatment, again indicating decreased lipid order in both domains of the polarized plasma membrane (Fig. 5 F and Fig. S4, I–K).

Altogether, these various imaging approaches indicate that OSBP supplies both apical and basolateral domains with cholesterol. This wide effect on cholesterol distribution is in good agreement with the effect of OSBP inhibition on both the apical and basolateral trafficking routes as determined before.

Because OSBP regulates epithelial sorting and subsequent trafficking of polarized cargoes that determine apicobasal polarity, we aimed to test whether this feature of OSBP is general across many cell types or restricted to epithelial cells.

First, we determined whether the biochemical activity of OSBP changes upon epithelial-to-mesenchymal transition, i.e., when epithelial cells are losing their apicobasal polarity. For this, we induced EMT in MDCK cells using HGF and in A549 adenocarcinoma cells using TGF-β (Kubiczkova et al., 2012; Hao et al., 2019; Hua et al., 2019). Thereafter, we measured PI(4)P levels at the Golgi area upon OSBP inhibition. By inhibiting OSBP, OSW-1 protects PI(4)P from the OSBP cycle, thereby allowing the estimation of the PI(4)P pool that is consumed specifically by OSBP in the cell (Mesmin et al., 2017). Following OSW-1 addition, Golgi-localized PI(4)P levels increased rapidly and reached a plateau within 10–15 min in both MDCK and A549 cells (Fig. 6, A and B; and Fig. S5, A and B). However, the difference between the steady state and maximal values was larger in control cells than in HGF or TGF-β-induced mesenchymal cells (Fig. 6, A and B; and Fig. S5, A and B). Quantification of the PI(4)P signal before and after OSW-1 addition suggested that OSBP consumed ∼95% of the cellular PI(4)P pool in control MDCK cells compared with only ∼60% in HGF-treated cells. In A549 cells, OSBP consumed ∼75% of the total PI(4)P pool at steady state and only 40% in TGF-β-induced cells. These observations indicated that EMT correlates with a significant reduction of OSBP activity (Fig. 6, A and B). Thus, OSBP is biochemically more active in epithelial cells than in mesenchymal cells.

To test whether other members of the ORP family contribute to epithelial and mesenchymal phenotypes, we performed a bioinformatics screen to correlate the expression level of OSBP and the other members of the ORP family across cell lines with different epithelial characteristics (Rajapakse et al., 2018). We extracted the gene expression data of the cell lines included in the Cancer Cell Line Encyclopedia (Barretina et al., 2012), focusing on all ORP genes as well as a panel of epithelial and mesenchymal marker genes (Fig. 6 C). To define a cell-line specific epithelial–mesenchymal signature, we calculated a single epithelial–mesenchymal index (EMI) for each cell line using the marker gene expression data (Fig. 6 D) and we correlated these values with the expression levels of the various ORPs (Fig. S5 C and Fig. 6 E). This analysis indicates that OSBPL2 and OSBP expression levels correlate positively with EMI, while OSBPL8 and OSBPL6 expressions show a negative correlation. Thus, OSBPL2 and OSBP are highly expressed in cell lines with epithelial characteristics, while OSBPL8 and OSBPL6 are rather mesenchymal cell-specific genes.

To experimentally investigate the differential ORP gene expression upon EMT, we triggered EMT by TGF-β treatment in A549 lung adenocarcinoma cells; then, we assessed the relative transcript levels of EMT transcription factors and the top-changed ORPs by RT-qPCR. While snai1- and snai2-inductions dominated the TGF-β-triggered EMT in A549 cells, we observed a concomitant decrease in OSBP and OSBPL2 transcript levels. A significant increase in OSBPL6 messengers was measured, validating our previous findings, whereas no change in OSBPL8 expression was observed (Fig. 6 F).

To test whether this observation can be extended to disease conditions as well, we repeated the analysis using RNAseq data obtained from TCGA lung cancer specimens (Fig. S5, D–G). In agreement with the cell-line collection data, OSBP and OSBPL2 expressions positively correlate with the epithelial features of the given cancer tissue specimens.

To further investigate the epithelial expression of OSBP and OSBPL2, we assessed histology samples of the Human Protein Atlas (Uhlen et al., 2015). As revealed by specific antibodies, OSBP and OSBPL2 proteins exclusively decorate the epithelial cell layers of various tissue samples, such as nasopharynx and urinary bladder, while the skeletal muscle, which has a mesenchymal origin, lacks OSBP and expresses high levels of OSBPL6 (Fig. S5 H).

Loss of epithelial polarization through epithelial-to-mesenchymal transition (EMT) is frequently observed in various stages of malignancies. EMT can lead to metastatic dissemination; therefore, tumors with mesenchymal characteristics are typically coupled to shorter patient survival (Dongre and Weinberg, 2019; Aruga et al., 2018). By analyzing lung adenocarcinoma (LUAD) survival data, we found that high OSBP and OSBPL2 expressions are associated with significantly longer patient survivals, suggesting that downregulation of these genes is a signature of tumor malignancy (Fig. 6, G and H).

Altogether, these analyses suggest that (i) OSBP is highly expressed in epithelial cells, (ii) it is targeted by EMT, and (iii) it might have an impact on the overall outcome of some malignancies.

It has long been shown that the TGN is a major cargo sorting station for the formation of transport intermediates delivered to the apical or basolateral surface of epithelial cells. Furthermore, evidence suggested that such sorting relies not only on protein recognition mechanisms, typically coat/adaptor-mediated vesicle budding, but also on lipid-mediated mechanisms, possibly on cholesterol and sphingolipid enriched regions (Cao et al., 2012; Keller and Simons, 1997; Deng et al., 2016; Boncompain and Weigel, 2018; Lippincott-Schwartz and Phair, 2010). However, despite progress in imaging cargo and lipid sorting at the TGN, tools for modifying or isolating lipid membrane regions are limited, making it difficult to unravel the contribution of lipid-mediated mechanisms on cargo-selective secretion. Taking advantage of the identification of OSBP as a prominent transporter of cholesterol toward the TGN and of OSBP-specific pharmacological tools, we show here that OSBP is essential for proper cargo sorting at the TGN and their subsequent transport in epithelial cells. The observation of cargo waves with the same periodicity as lipid waves arising from the OSBP/PI(4)P cycle suggests that OSBP acts by creating lipid gradients along the extensive tubular network of the TGN, thereby facilitating lateral sorting of cargo proteins.

OSBP burns more PI(4)P in epithelial than in mesenchymal cells. Cysts developed from OSBP-silenced cells display morphological defects. Furthermore, OSBP inhibition by the specific drug SWG leads to a complete loss of epithelial polarity, which is accompanied by reduced cholesterol levels in both the apical and basolateral regions. Altogether, these observations indicate a key function of the lipid exchange activity of OSBP in maintaining the epithelial phenotype.

Targeting OSBP leads to an altered distribution of cholesterol at the polarized plasma membrane; however, this effe