Molecular and cellular dissection of the oxysterol-binding protein cycle through a fluorescent inhibitor

ORPphilins are bioactive natural products that strongly and selectively inhibit the growth of some cancer cell lines and are proposed to target intracellular lipid-transfer proteins of the oxysterol-binding protein (OSBP) family. These conserved proteins exchange key lipids, such as cholesterol and phosphatidylinositol 4-phosphate (PI(4)P), between organelle membranes. Among ORPphilins, molecules of the schweinfurthin family interfere with intracellular lipid distribution and metabolism, but their functioning at the molecular level is poorly understood. We report here that cell line sensitivity to schweinfurthin G (SWG) is inversely proportional to cellular OSBP levels. By taking advantage of the intrinsic fluorescence of SWG, we followed its fate in cell cultures and show that its incorporation at the trans-Golgi network depends on cellular abundance of OSBP. Using in vitro membrane reconstitution systems and cellular imaging approaches, we also report that SWG inhibits specifically the lipid transfer activity of OSBP. As a consequence, post-Golgi trafficking, membrane cholesterol levels, and PI(4)P turnover were affected. Finally, using intermolecular FRET analysis, we demonstrate that SWG directly binds to the lipid-binding cavity of OSBP. Collectively these results describe SWG as a specific and intrinsically fluorescent pharmacological tool for dissecting OSBP properties at the cellular and molecular levels. Our findings indicate that SWG binds OSBP with nanomolar affinity, that this binding is sensitive to the membrane environment, and that SWG inhibits the OSBP-catalyzed lipid exchange cycle.

ORPphilins are small natural bioactive molecules that have strong inhibitory activities on the growth of several human cancer cell lines. Despite their structural diversity, these compounds may have a related mechanism of action because they present a similar sensitivity profile on the NCI-60 cancer cell panel (1). It is proposed that ORPphilins directly target oxysterol-binding protein (OSBP) 4 and OSBP-related protein 4 (ORP4), which are two members of a conserved family of lipidtransfer proteins that transport key lipids such as cholesterol or phosphatidylserine between intracellular membranes (2,3). OSBP is a cytosolic protein containing an N-terminal region capable of bridging the membrane of the endoplasmic reticulum (ER) to that of the trans-Golgi network (TGN), and a C-terminal OSBP-related domain (ORD) responsible for the lipid transfer activity (4). We have previously demonstrated that OSBP directs cholesterol transport from the ER to the TGN through coupled counter-exchange and hydrolysis of the phosphoinositide PI(4)P at membrane contact sites between these two organelles (5). By using the ORPphilin OSW-1 on cells whose endogenous level of OSBP exceeds that of ORP4 by 2 orders of magnitude, we have shown that OSBP activity makes a major contribution to the intracellular trafficking of cholesterol (6). As a result, OSBP impacts on the physicochemical properties of cellular membranes. The other potential target of ORPphilins, ORP4, is the closest OSBP ortholog and is also described as a cholesterol/PI(4)P exchanger (7), among other assigned roles (8). However, ORP4 is virtually absent from most tissues except the brain, testis, and some transformed tissues, whereas OSBP is ubiquitously expressed.
Among the ORPphilins described by Burgett et al. (1) are the small molecules schweinfurthin A and B, which were shown to compete with oxysterol binding to OSBP and ORP4. Interestingly, however, unlike other ORPphilins, the affinity of these compounds was 40 times higher for OSBP than for ORP4, sug-gesting that schweinfurthins have a better binding specificity to OSBP (1). It has also been shown that cell treatment with schweinfurthin A could shift OSBP localization to the Golgi area (1). The compounds from the schweinfurthin series (including vedelianin and schweinfurthins A-Q) are natural prenylated stilbenes isolated from the leaves or fruits of different species of the tropical genus Macaranga (9 -14). Schweinfurthins harboring a hexahydroxanthene moiety are promising because they have a powerful and selective antiproliferative activity against some human cancer cell lines (10,15). For example, schweinfurthins show GI 50 values in the nanomolar range on cell lines derived from glioblastomas, whereas they have only limited activity on lung cancer cell lines, such as A549 (13,16). However, this differential cytotoxicity is still poorly understood (15). Previous studies on the mechanism of action of these drugs have shown that they interfere with intracellular trafficking and lipid metabolism. For instance, the compound 3-deoxyschweinfurthin B reduced cellular cholesterol levels by decreasing its synthesis and increasing its export (17). Likewise, treatment with schweinfurthin G (hereafter abbreviated as SWG) resulted in the disruption of the TGN and the depletion from the cell surface of some lipids (GM1, sphingomyelin, and cholesterol) known to promote membrane order and domain formation (18). In these conditions, the PI3K-Akt-mTOR signaling pathway was down-regulated (18).
At the molecular level, do schweinfurthins bind directly to OSBP? A biotinylated analogue of SWG has been shown to bind to OSBP, ORP1L, and ORP1S, albeit with low affinities (in the micromolar range) (18), which contrasted with the nanomolar concentrations needed to observe a cellular effect. More surprisingly, a bioactive fluorescently-tagged SWG showed a punctate intracellular distribution (18,19), partially localized with an endosomal marker, but far from the ER/TGN interface where OSBP concentrates. Nevertheless, an interesting relationship between schweinfurthins and OSBP has been established by the fact that OSBP knockdown sensitizes cells to schweinfurthins, and conversely, that OSBP overexpression increases their resistance to schweinfurthins (1).
Altogether, these previous observations suggest that schweinfurthins may have an inhibitory effect on OSBP, but whether this effect is direct or indirect is not well-characterized. Here we dissect the fine interaction of the highly active compound SWG with OSBP using in vitro reconstitution of membrane systems and cellular imaging approaches. We reveal and take advantage of the fact that SWG has intrinsic fluorescence properties. This allows us first to track SWG within cells with good spatial and temporal precision, and second to perform OSBP binding kinetics. We report the nanomolar affinity of SWG for OSBP, its sensitivity to membrane interfacial environment, and provide evidence that SWG is able to inhibit the entire lipid exchange cycle catalyzed by OSBP by entering its ORD domain.

OSBP expression level correlates with SWG sensitivity
The specific cytotoxicity of schweinfurthins for a subset of NCI-60 cancer cell lines is not explained so far. Understanding what differentiates cell lines that do not respond identically to SWG would allow a more detailed molecular analysis of its activity (Fig. 1A). Previous data have shown that modulating the cellular level of OSBP (via knockdown or overexpression) could influence the sensitivity of cells to schweinfurthins (1). We therefore thought that the differential cytotoxicity to schweinfurthins observed between various cell lines could be due to their divergent levels of OSBP expression. We then compared OSBP expression in a panel of several cancer cell lines, to which we added the immortalized retinal pigmented epithelial cell line (hTERT-RPE1; hereafter RPE1) that we previously identified as sensitive for OSW-1 (6), and measured their sensitivity to SWG. We found that the OSBP level varies up to 5-fold between these cell lines (Fig. 1B). OSBP expression is relatively low in human primary glioblastoma cell line U87 MG, as well as in RPE1, whereas it is higher in PC3 (human prostate cancer cell line) and maximal in A549 cells (adenocarcinomic human alveolar basal epithelial cells). We then performed viability tests after treatments with increasing doses of SWG for 72 h, and plotted the half-maximal inhibitory concentrations (IC 50 ) obtained for each cell line according to their OSBP level (Fig.  1C). We observed that the sensitivity to SWG is inversely proportional to the cellular level of OSBP, that is, high expression correlated with resistance. As a control, we used doxorubicin, a compound not related to ORPphilins known to target DNA metabolism. In this case, we do not observe any correlation between OSBP expression level and doxorubicin sensitivity (Fig. 1D), and confirmed that the sensitivity profile for doxorubicin is different from that of SWG (Fig. 1E). These results indicate that OSBP has a protective effect against the cytotoxic activity of SWG. Note that a good correlation between sensitivity/resistance and expression level generally defines drugs with a single protein target (20).

SWG modulates TGN to PM trafficking
Because OSBP is most likely the target of SWG, we next wanted to monitor OSBP intracellular distribution over time after adding SWG. We used RPE-1 cells because they are highly sensitive to SWG and have a wide and extensively tubulated TGN that is well-adapted to imaging. Fig. 1, F and G, show that in control cells, endogenous OSBP is predominantly co-localized with the TGN marker TGN-46, and weakly present in the cytosol. Upon short treatments with low doses of SWG, we noticed that this cytosolic fraction of OSBP shifted partially to the TGN (Fig. S1, A and B). In contrast, more prolonged treatments with SWG lead to the gradual dissociation of OSBP from the TGN, as judged by Pearson's correlation coefficient measurements over time (Fig. 1G), until it became mostly cytosolic after 24 h of treatment. We verified that this effect was not correlated to change in the amounts of OSBP or TGN-46 in response to SWG treatment (Fig. S1C). In agreement with previous observations, long treatments with SWG also lead to TGN disruption (18). In particular, the TGN loses its characteristic tubular morphology and becomes partly dispersed (Fig.  S1, D and E). These short-and long-term effects, first on OSBP distribution and then on TGN morphology, suggest that SWG could quickly and durably affect Golgi trafficking.
To further analyze the effect of SWG on the secretory pathway, we sought to monitor the intracellular transport of OSBP targeting by a fluorescent inhibitor a newly synthesized protein from the ER to the plasma membrane via the Golgi. We therefore used the RUSH system, which allows to trigger the secretion of a cargo (here, the CD59 protein) fused with a streptavidin-binding peptide (SBP) by simply adding biotin to the culture medium (21). Because CD59 is also coupled to GFP, we could track and quantify its trafficking by live cell imaging. Experiments performed with control cells at 37°C in DMEM/F-12 medium showed that CD59 initially retained in the ER (t ϭ 0) crosses the Golgi apparatus within 35 min after biotin addition, and finally reaches the plasma membrane after 90 min (Fig. 1H). Fig. 1I illustrates this wave of secretion by showing the fluorescence intensity in the Golgi region at each time point. Adding SWG does not affect the arrival of CD59 at the Golgi, but strongly delays its exit. This is particularly visible in the temporal projection (Fig. 1H, right panel) showing that CD59 accumulates massively in the perinuclear region over time. Next, we performed RUSH assay with cells treated with siRNA against OSBP or with control siRNA (Fig. S2A). Interestingly, the silencing of OSBP could partially recapitulate the effect observed with SWG: it does not modify the entry of CD59 at the Golgi, but delays its exit (Fig. S2B). Altogether, these experiments suggest that SWG affects the formation of post-Golgi (but not pre-Golgi) transport intermediates, probably by targeting directly OSBP. Indeed, OSBP transfers cholesterol from the ER to the TGN via membrane contact sites, whereas bypassing the cis-Golgi (22), it is therefore in the best position to mediate the effects of SWG.

OSBP targeting by a fluorescent inhibitor SWG inhibits OSBP lipid-exchange activity
To directly assess the effect of SWG on the lipid-transfer activity of OSBP, we conducted in vitro liposome-based reconstitution experiments. We first monitored the transfer of a cho-lesterol fluorescent analogue (dehydroergosterol, DHE) from liposomes mimicking the ER (L E ) to liposomes mimicking the Golgi (L G ) ( Fig. 2A). To facilitate the analysis, we used a short OSBP construct (residues 401-807), corresponding to the

OSBP targeting by a fluorescent inhibitor
ORD domain, which rapidly transfers DHE between liposomes (Fig. 2B, gray trace). We observed that this rate decreased drastically as we increased the concentration of SWG in the reaction mix (Fig. 2B). SWG blocked ORD-catalyzed transfer of DHE with an apparent K i of Ͻ0.1 nM (Fig. 2C), suggesting that SWG binds to the ORD domain of OSBP with a very high affinity. Note that because of this tight binding, the concentration dependence reflects the ratio of [SWG] to [ORD] and is different from a typical situation in which [Inhibitor] Ͼ Ͼ [Enzyme].
Next, we measured the transfer of PI(4)P by the ORD from L G to L E liposomes using a specific probe (PH FAPP ) labeled with a NBD molecule, as previously described (5) (Fig. 2D). The fluorescence of the probe was high at the surface of L G liposomes, but was subsequently quenched as the probe followed PI(4)P relocation to L E liposomes doped with rhodamine lipids (Fig.  2E). As with sterol transfer, SWG decreased the PI(4)P transfer rate in a dose-dependent manner, indicating that this drug inhibits the two transfer steps of the OSBP cycle.
We have previously shown that OSBP contributes significantly to the intracellular sterol distribution and influences the lipid order gradient between membranes along the secretory pathway (6). For this purpose, OSBP consumes massive amounts of PI(4)P at the TGN. We therefore tested whether SWG affects cholesterol distribution and PI(4)P turnover in cells. To do this, we first used RPE-1 cells stably expressing the probe D4-H-GFP derived from the perfringolysin O toxin of Clostridium perfringens, which specifically recognizes cholesterol from 20 mol % in membranes. Previous work has shown that D4-H decorates mainly the plasma membrane at steady state, but that following cholesterol depletion (e.g. caused by the extracellular addition of methyl-␤-cyclodextrin), it translocates to endosomal compartments (23). Accordingly, we found that D4-H-GFP was mostly at the plasma membrane (positive for WGA) in control cells (Fig. 2G). Then, upon treatment with SWG, D4-H-GFP moved partially to endo-lysosomes in a dosedependent manner, indicating that cholesterol levels dropped at the plasma membrane (Fig. 2H). To further examine whether SWG affects other pools of cholesterol, we measured the cholesteryl ester/cholesterol ratio by LC/MS. Indeed, we suspected that, by inhibiting OSBP, SWG could lead to excessive amounts of cholesterol in the ER, which would foster its esterification and incorporation into lipid droplets. As shown in Fig. 2I, a 2-h treatment with 100 nM SWG significantly increased the cholesteryl ester/cholesterol ratio, by about 2-fold, as compared with control DMSO-treated cells. As a comparison, we also tested OSW-1, another OSBP inhibitor, which provided a similar effect. Therefore, these two drugs seem to promote cholesterol esterification, which is consistent with a blockage of OSBP activity. Because OSBP continuously consumes PI(4)P at the TGN, its pharmacological inhibition should protect a substantial fraction of PI(4)P, as previously reported (6). To verify this, we monitored the distribution of the PI(4)P probe P4M-SidM coupled to a fluorophore by live cell imaging. P4M-SidM-GFP stably expressed in RPE-1 cells was mostly cytosolic and decorated slightly the TGN, in control cells (Fig. 2J). However, the addition of SWG caused a sudden 2-fold increase in TGN-associated P4M-SidM-GFP fluorescence, suggesting that the drug is capable of rapidly crossing cell membranes and acting on the PI(4)P pool of the TGN (Fig. 2, J and K). Altogether, these experiments suggest that SWG affects membrane cholesterol levels at multiple steps along the secretory pathway, which coincides with an increase in the level of PI(4)P at the Golgi, suggesting that OSBP is the key cellular target of SWG. Consistently, OSBP silencing also alters cholesterol and PI(4)P distribution significantly as previously reported (6), and as shown in Fig. S2, C and D.
By controlling the turnover of PI(4)P at the Golgi, OSBP regulates its own membrane attachment and ER-TGN contact formation (5). It is therefore possible that SWG facilitates organelle tethering by inhibiting OSBP. To verify this, we used cells overexpressing mCherry-OSBP, GFP-VAP-A, and a TGN marker, TagBFP-␤GalT1. We observed that following the addition of SWG, not only OSBP, but also VAP-A, translocate rapidly and massively to the perinuclear region, suggesting that SWG promotes the formation of ER-TGN contact sites (Fig. S3,  A and B). Note that this observation corroborates our data showing that endogenous OSBP shifted to the TGN after a short treatment of SWG (Fig. S1A). This effect is probably due to the inhibition of OSBP-exchange activity, but it cannot be excluded that SWG also affects structural determinants of contact sites.

Imaging SWG reveals its specificity for OSBP
We suspected that SWG could be intrinsically fluorescent because its stilbene motif comprises several conjugated double bonds (Fig. 1A). To test this, we performed fluorimetric analysis of the drug in a saline buffer. As shown in Fig. 3A, SWG exhibited excitation and emission spectra in the near UV range, with excitation and emission peaks at 330 and 410 nm, respectively. We then sought to visualize SWG in a cellular context via its intrinsic fluorescence. Using an imaging system adapted to UV, we directly assessed SWG subcellular distribution in RPE-1 cells after a 30-min treatment at 37°C with 1 M of the drug. Remarkably, SWG was enriched exactly where OSBP is located, that is, in the TGN region marked by TGN-46 (Fig. 3B). Outside the TGN, SWG showed a weaker and diffused signal, as it likely labels membranes due to its partial hydrophobicity. Importantly, SWG distribution contrasted with that of a previously reported fluorescent SWG derivative (called SG-Fluor), consisting of SWG tagged with an additional fluorescent group, which decorated endosomal compartments (18). We next monitored the incorporation of SWG into the TGN over time (Fig. S3C and S4A). Interestingly, after 5-10 min of incubation, SWG fluorescence was already visible in the perinuclear region, suggesting that SWG rapidly crosses the plasma membrane. This was confirmed by experiments showing that SWG could readily diffuse through giant plasma membrane vesicles (Fig.  S4, C and D). The gradual increase (t1 ⁄ 2 ϭ 25 min) of SWG at the TGN level coincided with the recruitment of OSBP in the same region over time (Fig. 3C and Fig. S4B).
We reasoned that if SWG binds specifically to OSBP, the cellular SWG fluorescence intensity should reflect properly the abundance of OSBP. We therefore assessed SWG fluorescence in cells in which OSBP was either silenced or overexpressed (Fig. 3, D and E). We treated RPE-1 cells with siRNA directed against OSBP or with control siRNA for 72 h and then incu-

OSBP targeting by a fluorescent inhibitor
bated the cells with 1 M SWG for 30 min. Although silencing efficiency was variable among cells (Fig. 3D, bottom panel), SWG fluorescence at the TGN was generally much less intense in silenced cells as compared with control cells, indicating that SWG is poorly recruited when low amounts of OSBP are present. Conversely, in OSBP-overexpressing cells, SWG was extremely enriched at the TGN and co-localized with OSBP-mCherry and TGN-46 (Fig. 3E). The level of SWG fluorescence in the cells almost perfectly correlated with that of OSBP-mCherry, suggesting that incorporation of SWG at the TGN is attributable to OSBP. To further verify this hypothesis, we treated the cells with SWG and an increasing dose of OSW-1.

OSBP targeting by a fluorescent inhibitor
This compound also binds to OSBP but is not fluorescent, so it should compete with SWG for OSBP. OSW-1 prevented the TGN labeling by SWG in a dose-dependent manner. The addition of 20, 50, and 100 nM OSW-1 caused a reduction in the amount of SWG at the TGN by 1.9, 3.6, and 7-fold, respectively (Fig. 3E, bottom panel). Alternatively, we first labeled cells overexpressing OSBP-mCherry with SWG and then added the competitor OSW-1 (Fig. 3F). By time lapse imaging experiments, we observed that OSW-1 gradually chased out SWG from the TGN in a dose-dependent manner (t1 ⁄ 2 ϭ 9.2 Ϯ 0.7 min with 50 nM OSW-1), further indicating that the retention of SWG at the TGN is most likely related only to OSBP.
ORPphilins were initially described as being able to bind to OSBP and ORP4, but schweinfurthins A and B exhibited a clear preference for OSBP over ORP4 (1). To assess SWG specificity for OSBP over ORP4, we overexpressed ORP4-mCherry instead of OSBP-mCherry and performed treatment with SWG for imaging (Fig. 3G). ORP4 did not facilitate SWG recruitment, thus demonstrating that SWG selectively targets OSBP.

Biochemical characterization of SWG
Because SWG rapidly penetrates into cells to target OSBP (Fig. S4), we hypothesized that SWG is able to passively interact with membranes due to its partial hydrophobicity. To characterize the interactions of SWG with both membranes and OSBP, we performed in vitro fluorescence measurements of SWG. Spectral analysis revealed that when the drug was switched from an aqueous medium to polar solvents (such as methanol or acetonitrile), its fluorescence increased and this effect was accompanied by a blue shift (Fig. 4A). Interestingly, adding detergent micelles (Thesit; green curves) or liposomes (Fig. 4B, blue curve) further increased the fluorescent signal and structured SWG emission spectrum, indicating that SWG fluorescence is even more sensitive to interfacial environments.
We reasoned that SWG could also experience a change in environment if it enters in the hydrophobic cavity of the ORD of OSBP. We added OSBP constructs to a mixture containing 200 nM SWG and liposomes (0.1 mg/ml), and recorded SWG emission spectra upon excitation at 330 nm (Fig. 4B). Remarkably, the addition of ORD (200 nM) more than doubled the fluorescence intensity of SWG as compared with liposomes alone. In contrast, adding 200 nM of the N-PH-FFAT construct, i.e. the N-terminal region of OSBP, did not promote any fluorescence change. These results strongly suggest that the drug inserts into the binding cavity of the ORD. To further characterize the contact between SWG and OSBP, we performed the 3D structure prediction of the ORD of OSBP by homology modeling, based on the crystal structure of the yeast OSBPrelated protein Osh1 (24), which also exchanges sterol for PI(4)P (Fig. 4C). We noticed that several tryptophan residues are close to the cavity of the ORD (indicated in yellow) and thus in good position to form fluorescence resonance energy transfer (FRET) pairs with SWG. To test this hypothesis, we mixed liposomes with the ORD, with SWG, or with a mixture of ORD and SWG, and performed spectral analysis upon tryptophan excitation at 280 nm (Fig. 4D). As expected, a tryptophan fluorescence signal peaking at ϳ340 nm was observed when the ORD was employed alone (black line), whereas the signal obtained with SWG alone (red line) was weak because it is poorly excited at 280 nm (Fig. 3A). Strikingly, when both the ORD and SWG were present, we observed a massive FRET signal (green line) with peaks characteristic of SWG (as seen in Fig. 4B). The FRET signal was much higher than the theoretical sum of the signals obtained with the ORD or SWG separately (dashed gray line). The FRET signal was also 2-fold greater than the SWG fluorescent signal obtained upon direct excitation at 330 nm under the same conditions. The apparition of FRET was concomitant with partial quenching of tryptophan emission at 340 nm. Altogether, these results provide evidence that SWG interacts with the ORD of OSBP by entering in its lipid-binding pocket, and also suggest that FRET between the ORD and SWG is a useful tool to probe this interaction (Fig. 4E).

Kinetics of SWG binding to the ORD
We took advantage of the large FRET signal to monitor the binding and dissociation kinetics of SWG to the ORD. We first mixed SWG with ORD (200 nM each) and measured the changes in fluorescence over time at 410 nm, which is the most specific FRET wavelength, upon tryptophan excitation at 280 nm. As shown in Fig. 4F, SWG binding to the ORD was faster when liposomes (1 mg/ml) were present in the cuvette than simply in buffer alone (compare the black and blue curves). As a control, we replaced the ORD by the N-terminal half of OSBP (residues 1-408, N-PH-FFAT) and observed no change in fluorescence (red curve). Interestingly, addition of detergent micelles (Thesit) caused the reaction to be completed in about 2 s (green curve, see also Fig. S5A), suggesting that the association rate of SWG with the ORD is very fast (k on Ͼ 2 s Ϫ1 ) in the presence of an interfacial environment. The effect of membranes or micelles can be explained by the fact that the ORD is purified with a cholesterol molecule in its binding pocket, as shown by mass spectrometric analysis (Fig. S5B). The presence of a membrane interface likely facilitates the dissociation of cholesterol from the ORD, thereby allowing SWG entry. Our results in Fig. 3F suggest that SWG could be removed by OSW-1 from their binding site, so we next monitored the effect of OSW-1 addition (in slight excess; 500 nM) on SWG-loaded ORD in the same conditions as in Fig. 4F, i.e. in buffer supplemented or not with liposomes or micelles. With micelles, the rate of dissociation of SWG from the ORD was very fast (k off Ͼ 2 s Ϫ1 ), in contrast, SWG exit from the ORD in buffer alone or in the presence of liposomes was rather slow (Fig. 4G). Finally, to further analyze the interaction of SWG with the ORD of OSBP, we performed equilibrium measurements using FRET. Fig. 4H shows a titration experiment in which we added increasing amounts of SWG to a mixture containing the ORD (200 nM) and 0.1% Thesit. Below 200 nM, the stepwise addition of SWG induced a large FRET signal. Thereafter, the apparent increase in fluorescence was no longer due to FRET but to SWG direct fluorescence. To reveal the saturation of the FRET signal, we repeated the same experiment, but in the presence of OSW-1 (200 nM) to block the ORD. In this case, the increase in fluorescence came only from SWG increments in the mixture and not from FRET. Fitting the dose-response curve after correction of the nonspecific signal indicated an apparent affinity of SWG for

OSBP targeting by a fluorescent inhibitor
the ORD in the presence of detergent micelles of K d ϭ 35.5 nM. This affinity is much lower than the K i observed in cholesterolexchange assays performed with liposomes (Fig. 2C) suggesting that the micellar environment unlocks the ORD, hence leading to a decrease in affinity and to a dramatic acceleration in the dissociation kinetics of ORD ligands.

Discussion
The discovery of ORPphilins is an important step for the study of ORPs, highlighting the importance of these proteins in cancer, and providing pharmacological tools to characterize their molecular activities and physiological functions. ORPphilins include compounds with a variety of structures, comprising OSW-1, cephalostatin, stelletin, and schweinfurthin subfamilies (1). Although they are described to have the same intracellular targets (OSBP and ORP4), their differential affinity and the way they interact with the cellular environment could have different functional and physiological consequences.
The present work addresses several properties of SWG, a natural product of the schweinfurthin family, which displays antiproliferative activities for specific tumor-derived cell lines. By using a combination of reconstitutions on artificial membranes and cellular approaches, we show that SWG binds directly to the ORD domain of OSBP. This binding prevents the lipid-exchange cycle of OSBP, i.e. PI(4)P consumption and cholesterol transfer, and promotes ER-TGN contact site stabiliza-

OSBP targeting by a fluorescent inhibitor
tion. As a result, post-Golgi trafficking and plasma membrane cholesterol levels are severely affected.
We showed that when there are more or less copies of OSBP present in a cell type, a higher or lower SWG concentration is required, respectively, to mediate a cytotoxic effect (Fig. 1C). This probably reflects a buffering effect: SWG concentration must be adjusted to reduce total OSBP activity below a critical threshold for cell viability. Consistently, it is reported that OSBP overexpression confers cell resistance to schweinfurthins, whereas OSBP silencing sensitizes cells to these drugs (1). First, this excludes a potential toxic effect due to excessive ER-TGN membrane tethering triggered by SWG, because in this case, cytotoxicity would increase with overexpression of OSBP. Second, this may explain the specific cytotoxicity of these drugs for a subset a cancer cell lines. In the future, it would be interesting to verify whether cell sensitivity to various ORPphilins could be predicted according to tissue concentration of OSBP.
High levels of cholesterol in the ER usually leads to two cellular responses: the rapid esterification of excess cholesterol followed by its incorporation into lipid droplets (25), and a slower transcriptional response mediated by the cholesterol homeostatic regulatory machinery (26), including the transcription factors sterol regulatory element-binding protein (SREBP) and liver X-receptor, which stop cholesterol biosynthesis and uptake, and promote its export from the cell. The addition of SWG blocks cholesterol fluxes through OSBP coming from the ER. As a result, we showed an increased cholesterol esterification. Interestingly, previous work has revealed that treatment with 3-deoxyschweinfurthin B induces changes in the expression of genes related to cholesterol homeostasis controlled by SREBP and liver X-receptor (17). Because these regulators are coupled to cholesterol-sensing mechanisms localized at the ER, it is possible that inhibiting OSBP triggers such a homeostatic response. Further studies would be needed, however, to address the direct involvement of the OSBP cycle in cholesterol sensing at the ER (e.g. through the SCAP/ SREBP-2 pathway).
Fluorescent probes are instrumental for tracking biomolecules with good spatio-temporal resolution in living cells by imaging. In particular, the use of fluorescently tagged drugs can shed light on their specific interactions with cellular receptors, and provide molecular insights into pharmacological mechanisms. The major drawback of adding a large fluorescent moiety to compounds is that it can alter their pharmacological properties by modifying their chemical properties and/or their cellular distribution. Several studies have been conducted to make some ORPphilins fluorescent to visualize them in cells. A fluorescent OSW-1 analogue, labeled with the 4-(N,N-dimethylaminosulfonyl)-7-N-methylamino)-2,1,3-benzoxadiazolyl (DBD) group, was shown to distribute at the Golgi and to a lesser extent at the ER, and retains anti-proliferative properties (27). For schweinfurthins, fluorescent probes with differential activity between cell lines have been synthesized, such as schweinfurthin B and F modified by a p-nitro stilbene group, showing fluorescence in the visible spectrum (19). However, these analogues localized to punctate structures that do not resemble the Golgi compartment. Similarly, a fluorescent probe consisting of SWG with an additional p-nitro stilbene group (called SG-Fluor) was localized to punctate structures that overlapped with the endosomal marker LAMP1 (18). This seems not consistent with the subcellular distribution of OSBP, which is mainly found at the TGN and weakly localized to endosomes. In the present study, we exploited the intrinsic fluorescence properties of SWG in imaging and in membrane reconstitution assays. By using a UV-sensitive imaging system, we were able to observe directly SWG in its intracellular site of action, that is, at the TGN. We showed that when OSBP cellular abundance was adjusted (via silencing or overexpression), the cellular incorporation of SWG was modified accordingly, suggesting a high specificity of the drug for OSBP. We also showed that variations in ORP4 levels had little or no impact on SWG cell labeling. These results indicate that SWG is selective toward OSBP. This is consistent with earlier findings showing that schweinfurthin A had a better binding specificity to OSBP as compared with ORP4 (1).
The ORD domain of several ORP/Osh proteins, including OSBP, comprises tryptophans close to the lipid-binding pocket (Fig. 4C). In a previous study, we showed that FRET could be measured between tryptophans of Osh4 and the intrinsically fluorescent sterol DHE once bound in the pocket (28). Here, we took advantage of the fluorescence properties of SWG, which absorption-emission spectra partly overlap with that of DHE, to perform equilibrium and kinetics measurements by using FRET between tryptophans of OSBP ORD and SWG. We have shown that the association and dissociation rates of SWG from the ORD domain were very fast (Ͼ2 s Ϫ1 ). More specifically, SWG entry in the ORD was facilitated when membranes (or micelles) are present, which is probably because the ORD is co-purified with a cholesterol molecule bound in the pocket. Our hypothesis is that the ORD must first interact with the membrane to allow cholesterol exit so that SWG can then enter the pocket. Understanding this mechanism will, however, require further investigation.
Finally, this work invites a comparison between the OSBPtargeting compounds SWG and OSW-1. Both drugs strongly inhibit OSBP-catalyzed lipid transfer with K i Ͻ1 nM and promote ER-TGN contact site formation (6). We have shown that OSW-1 prevents TGN labeling by SWG when both drugs are added simultaneously to the cells, and that SWG can be removed from the TGN when OSW-1 is added afterward, even at low doses. Last, OSW-1 could displace SWG from purified OSBP ORD in vitro. These results suggest that OSW-1 has a stronger affinity for the ORD than SWG, and/or OSW-1 dissociates more slowly from the ORD. In this regard, we reported an apparent affinity of SWG for the ORD in the presence of detergent micelles of K d ϳ 35 nM, which is rather low compared with the measured K i . Akin to what has been observed recently on Osh6 (29), the ORD lid is probably unlocked by the interfacial micellar environment, hence leading to faster exchange kinetics and lower affinities. Further studies of the kinetics of SWG binding to or dissociation from the ORD with liposomes of defined composition should give information on the interfacial interactions involved in lipid exchange. The lack of fluorescence of OSW-1, and its too high affinity for the ORD, makes this inhibitor an almost irreversible ligand and therefore less suited for mechanistic investigations than SWG.

OSBP targeting by a fluorescent inhibitor
Because most ORPphilins (including OSW-1) interact with both OSBP and ORP4, it has been suggested that their anti-proliferative effect is due to ORP4 targeting and not OSBP (7). Indeed, high expression of ORP4 in immortalized cell lines, and in leukemia, leads to cell proliferation (30,31). Moreover, a ϳ90% reduction in the OSBP level does not alter cell proliferation (1). Intriguingly, our data suggest that SWG, which is a potent inhibitor of cell proliferation, selectively targets OSBP and not ORP4. In addition, our study provides a correlation between the level of OSBP expression and the differential sensitivity to SWG, suggesting that OSBP is a potential mediator of cell proliferation. Interestingly, a recent study has shown the direct activation of mTORC1 by OSBP-mediated cholesterol transfer, thereby clearly defining a link between OSBP activity and cell growth signaling (32). SWG might provide a useful tool to dissect this pathway.

Plant material and SWG isolation
The green fruits of Macaranga tanarius were collected in June 2014 at A Luoi, Thua Thiên-Huê Province, Vietnam, and authenticated by N. T. Cuong and D. D. Cuong. A voucher specimen (VN-2371) has been deposited at the Herbarium of the Institute of Ecology and Biological Resources of the Vietnam Academy of Science and Technology, Hanoi, Vietnam. Isolation of SWG was performed as previously described (33). SWG stock in DMSO was kept at Ϫ20°C. Solutions containing SWG were prepared fresh for every experiment.

Cell viability assay
Cells were seeded at 3000 cells/well-density into white wall 96-well-plates then treated on the following day with either SWG or doxorubicin. After 72 h treatment, cell viability was measured using CellTiter-Glo Luminescent Cell Viability Assay (Promega) and the luminescence was detected in a FLUOstar Omega microplate reader (BMG Labtech). Viability values of DMSO-treated cells were considered as 100%, then surviving curves and IC 50 values were determined using GraphPad Prism 7.0 software.

RUSH secretion assay
RPE-1 cells stably expressing streptavidin-KDEL (hook) and SBP-EGFP-CD59 (reporter) (gift from G. Boncompain, S. Miserey-Lenkei, and F. Perez) were used for RUSH assays. Time-lapse imaging was performed at 37°C and secretion was triggered by the addition of 80 M biotin.

Protein expression and purification
The human OSBP ORD fragment (401-807) with a C-terminal His 6 tag was purified from baculovirus-infected Sf9 cells. For cloning and protein expression, see supporting "Experimental procedures". Cell pellets were resuspended in lysis buffer (20 mM Tris, pH 7.5, 300 mM NaCl, 20 mM imidazole, EDTA-free protease inhibitors and phosphatases inhibitors) and lysed with a Dounce homogenizer. After ultracentrifugation (125,000 ϫ g), OSBP-ORD from the supernatant was adsorbed on an HisPur TM Cobalt Resin (Thermo Scientific), submitted to 3 washes with lysis buffer supplemented with 800, 550, and 300 mM NaCl, respectively, and then eluted with 250 mM imidazole-containing buffer. OSBP-ORD fractions were pooled, concentrated on Amicon Ultra (cut-off 30 kDa), then purified on a Sephacryl S300 HK16/70 column (GE Healthcare) using an AKTÄ chromatography system (GE Healthcare). All steps were performed at 4°C. The purified protein fractions were pooled, concentrated, supplemented with 10% glycerol, flash-frozen in liquid nitrogen, and stored at Ϫ80°C. The preparations of NBD-PH FAPP and OSBP N-PH-FFAT (1-408 fragment) have been described previously (5,34).

Data analysis
The signal was first corrected from SWG intrinsic fluorescence, then curves were fitted either with a linear equation (for slow kinetics when SWG is Ͼ200 nM) or an exponential equation to determine the apparent kinetic constant, k. Because SWG has a strong affinity for ORD, the concentration of free SWG cannot be approximated to the total concentration of SWG. Therefore, the inhibition curve was fitted with a quadratic equation to obtain K i .

Microscopy
Widefield microscopy was performed using an Olympus IX83 inverted microscope equipped with a Z-drift compensator, a scanning stage SCAN IM (Märzhäuser), and an iXon3 camera (Andor). For time-lapse imaging, cells plated in 35-mm -Dish (Ibidi) were mounted in a stage chamber set at 37°C (Okolab). TagBFP, EGFP, and mCherry were observed using Chroma fluorescence filter sets (reference 49,000, 39,002, and 39,010). Multidimensional acquisition and analysis was performed with MetaMorph software (Molecular Devices). Cells were placed in an appropriate initial volume of medium (phenol red-free, HEPES; Gibco) so that homogenous mixing of compounds (freshly diluted before use at ϫ2-5 concentration) was achieved upon subsequent additions during the time course measurements. Confocal microscopy was performed with a Zeiss LSM 780 microscope operated with ZEN software using a Plan-Apochromat ϫ63/1.4 Oil objective (Carl Zeiss).