Limonoid Compounds Inhibit Sphingomyelin Biosynthesis by Preventing CERT Protein-dependent Extraction of Ceramides from the Endoplasmic Reticulum*

Background: Pharmacological inhibitors of sphingolipid metabolism and transport are useful for both biological and therapeutic research. Results: High throughput microscopy-based screening identified limonoids as inhibitors of sphingomyelin biosynthesis by preventing the membrane extraction of ceramide. Conclusion: Some therapeutic properties of limonoids might be related to their effects on ceramide metabolism. Significance: This study provides insights into the role of the ceramide domains in sphingolipid metabolism. To identify novel inhibitors of sphingomyelin (SM) metabolism, a new and selective high throughput microscopy-based screening based on the toxicity of the SM-specific toxin, lysenin, was developed. Out of a library of 2011 natural compounds, the limonoid, 3-chloro-8β-hydroxycarapin-3,8-hemiacetal (CHC), rendered cells resistant to lysenin by decreasing cell surface SM. CHC treatment selectively inhibited the de novo biosynthesis of SM without affecting glycolipid and glycerophospholipid biosynthesis. Pretreatment with brefeldin A abolished the limonoid-induced inhibition of SM synthesis suggesting that the transport of ceramide (Cer) from the endoplasmic reticulum to the Golgi apparatus is affected. Unlike the Cer transporter (CERT) inhibitor HPA-12, CHC did not change the transport of a fluorescent short chain Cer analog to the Golgi apparatus or the formation of fluorescent and short chain SM from the corresponding Cer. Nevertheless, CHC inhibited the conversion of de novo synthesized Cer to SM. We show that CHC specifically inhibited the CERT-mediated extraction of Cer from the endoplasmic reticulum membranes in vitro. Subsequent biochemical screening of 21 limonoids revealed that some of them, such as 8β-hydroxycarapin-3,8-hemiacetal and gedunin, which exhibits anti-cancer activity, inhibited SM biosynthesis and CERT-mediated extraction of Cer from membranes. Model membrane studies suggest that 8β-hydroxycarapin-3,8-hemiacetal reduced the miscibility of Cer with membrane lipids and thus induced the formation of Cer-rich membrane domains. Our study shows that certain limonoids are novel inhibitors of SM biosynthesis and suggests that some biological activities of these limonoids are related to their effect on the ceramide metabolism.


To identify novel inhibitors of sphingomyelin (SM) metabolism, a new and selective high throughput microscopy-based screening based on the toxicity of the SM-specific toxin, lysenin, was developed. Out of a library of 2011 natural compounds, the limonoid, 3-chloro-8␤-hydroxycarapin-3,8-hemiacetal (CHC), rendered cells resistant to lysenin by decreasing cell surface SM.
CHC treatment selectively inhibited the de novo biosynthesis of SM without affecting glycolipid and glycerophospholipid biosynthesis. Pretreatment with brefeldin A abolished the limonoid-induced inhibition of SM synthesis suggesting that the transport of ceramide (Cer) from the endoplasmic reticulum to the Golgi apparatus is affected. Unlike the Cer transporter (CERT) inhibitor HPA-12, CHC did not change the transport of a fluorescent short chain Cer analog to the Golgi apparatus or the formation of fluorescent and short chain SM from the corresponding Cer. Nevertheless, CHC inhibited the conversion of de novo synthesized Cer to SM. We show that CHC specifically inhibited the CERT-mediated extraction of Cer from the endoplasmic reticulum membranes in vitro. Subsequent biochemical screening of 21 limonoids revealed that some of them, such as 8␤-hydroxycarapin-3,8-hemiacetal and gedunin, which exhibits anti-cancer activity, inhibited SM biosynthesis and CERT-mediated extraction of Cer from membranes. Model membrane studies suggest that 8␤-hy-

droxycarapin-3,8-hemiacetal reduced the miscibility of Cer with membrane lipids and thus induced the formation of Cerrich membrane domains. Our study shows that certain limonoids are novel inhibitors of SM biosynthesis and suggests that some biological activities of these limonoids are related to their effect on the ceramide metabolism.
Recent progress in sphingolipid (SL) 3 research highlights the role of these complex lipids in cell growth, differentiation, and apoptosis (1). In addition, SLs have attracted considerable attention because it was shown that they participate in the formation of lipid rafts in biomembranes (2). These lipid domains, which are characterized by a tight packing with a relatively high degree of lateral mobility, are in a permanent associated/dissociated equilibrium state in the membrane (3).
Two important points have to be considered when studying the biological activities of SLs as follows: first, the specific subcellular localization of the enzymes involved in their metabolism, and second, their biophysical properties that require specific transport mechanisms for movement between the membranes as well as translocation across the bilayer (1,4,5). Ceramide (Cer) occupies a central position in SL metabolism being generated either by de novo synthesis or by acidic or neutral sphingomyelinase activity (6,7). The de novo synthesis of Cer occurs on the cytosolic side of the endoplasmic reticulum (ER) (8) by a family of ceramide synthases (CerS), each member synthesizing Cer having different acyl chain lengths (9). Next, Cer is specifically transported by the Cer transfer protein (CERT) (10) to the trans-Golgi region where the synthesis of sphingomyelin (SM) occurs via the action of SM synthase 1 (11) on the luminal side of the Golgi. CERT extracts Cer from the ER membrane and then transports it to the Golgi in a nonvesicular manner (12). Cer is also transported to the cis-Golgi for the synthesis of glucosylceramide (GlcCer), the precursor of complex glycosphingolipids. GlcCer is synthesized on the cytosolic side of the Golgi by GlcCer synthase (13,14).
SM plays an essential role in cell proliferation (15), and the enzymes regulating SL metabolism have been reported as targets in cancer therapy (16,17). However, the effective use of therapeutic molecules has been hampered by their toxicity. Therefore, to find new types of inhibitors that affect Cer metabolism and transport as well as SM metabolism, we used an original microscopy-based automated assay to screen a chemical library of natural compounds. This type of lipid-specific probebased cell screening appears to be a very efficient technique for high throughput analysis of small compounds that affect lipid metabolism. We recently developed this visual technique coupled to biochemical analysis to successfully identify small molecules that interfere with cholesterol metabolism and transport (18) using the nontoxic cholesterol-binding protein toxin domain 4 (19). In the present screening, lysenin, a SM-specific pore-forming toxin (20,21), was used in the presence of dihydrosphingosine (DHS or sphinganine) to exclude the inhibitors of the serine palmitoyltransferase, which disrupt all SL metabolism (22). Thus, we focused on the biosynthetic steps after DHS synthesis. Screening of a library of 2011 natural small compounds and derivatives revealed that 3-chloro-8␤-hydroxycarapin-3,8-hemiacetal (CHC), a limonoid, selectively inhibited de novo biosynthesis of SM. Subsequent screening of 21 limonoids showed that some of them, such as 8␤-hydroxycarapin-3,8-hemiacetal (HC) and gedunin, a palm tree-derived limonoid with reported anti-malaria and anti-cancer activities (23,24), inhibited SM biosynthesis. The results thus indicate that limonoid compounds are novel inhibitors of SL metabolism and suggest that some of their biological activities are partially explained by their inhibition of Cer metabolism and transport.
Cell Culture-CHO-K1 and HeLa cells were routinely cultured as described previously (18) in medium supplemented with 10% fetal bovine serum (FBS) (referred to as complete medium). Medium with 1% Nutridoma-SP (Roche Applied Science) was used as a serum-free medium to study the drug effect on SL metabolism in CHO or HeLa cells.
Fluorescence Image Analysis by IN Cell Analyzer-CHO-K1 cells in 96-well plates were treated for 48 h with 50 M of drugs in complete medium without or with 1 M DHS. Cells were then treated with lysenin (200 ng/ml) (29) for 30 min. At the end of the incubation, cells were washed with phosphate-buffered saline (PBS) and fixed with 3% paraformaldehyde. They were stained with the fluorescent lipid marker C 5 -DMB-PC to stain the membrane and with DAPI to stain nuclei, to follow cell shape and viability. Fluorescence of each well was observed by IN Cell Analyzer 1000 (GE Healthcare) with a ϫ20 objective. Three to five images per well were acquired and analyzed using IN Cell Analyzer analysis software developer. The fluorescence pattern was also controlled with a Zeiss LSM 510 confocal microscope.
Assay for Intracellular Transport of C 5 -DMB-Cer and Recycling of C 6 -NBD-SM-Cells were incubated with fluorescent lipids as described previously (27, 30 -32 (27) with some modifications as indicated in the supplemental material.
Lipid Analysis and Determination of Lipid Content-At the end of the incubation time, cells were collected and lipids extracted by the method of Bligh and Dyer (33). Lipids were separated on HPTLC plates (27,34), and radioactive spots were quantified with a BAS 5000 image analyzer (Fuji Film Inc., Tokyo, Japan). For determination of the effect of limonoids on phospholipid and SM content, total phospholipid and SM content after HPTLC separation were evaluated by phosphorus quantification (35). See supplemental material for detailed information.
In Vitro Assay of Ceramide Extraction from Isolated ER Membranes and from Artificial Liposomes-ER membranes were prepared from CHO cells as described (37) and analyzed for the absence of SM synthase activity using C 6 -NBD-Cer as substrate (10). The extraction of Cer from ER membranes was then performed as described (10,36). Various concentrations of limonoids were preincubated with ER membranes containing 3 H-labeled Cer (200 g of proteins) or with hCERT (200 pmol) for 20 min on ice in transport buffer HNE (50 mM Hepes-NaOH (pH 7.5), 100 mM NaCl, 0.5 mM EDTA). The extraction was started by mixing hCERT and labeled ER membranes in HNE (150 l final volume), and the incubation lasted for 30 min at 37°C. Background of extraction was evaluated by addition of BSA (200 pmol) instead of hCERT. Incubation was stopped on ice, and tubes were centrifuged immediately at 100,000 ϫ g for 1 h at 4°C. Lipids were extracted from supernatants and pellets (33) and separated by HPTLC with a solvent mixture of chloroform/methanol/acetic acid (94:5:5, v/v). Radioactive spots were quantified with a BAS 5000 image analyzer.
Extraction of 14 C-labeled long chain Cer from artificial liposomes was performed as described (10). Limonoids or DMSO (control, 0.1% final concentration) were preincubated with lipid vesicles composed of egg yolk PC, egg yolk PE, and N-[1- 16 -Cer) (32:8:0.2, mol/mol) (40 g) (referred as "liposome preincubation") or with hCERT (100 pmol) (referred as "CERT preincubation") for 10 min on ice in HNE. Then the extraction was initiated by mixing hCERT and lipid vesicles in HNE (100 l final volume). The mixture was incubated for 30 min at 37°C. Incubation was stopped on ice, and tubes were centrifuged immediately at 100,000 ϫ g for 30 min at 4°C. The radioactivity of the supernatant and pellet was counted with a scintillation counter, and the radioactivity in the supernatant indicated the amount of Cer extracted from the vesicles.
Measurement of Fluorescence Anisotropy of DPH-DPPC vesicles were incubated with increasing concentrations of HC from 100:1 to 5:1 molar ratio for 15 min at 37°C. Egg PC/egg PE/C 16 -Cer (32:8:2 mol/mol) vesicles were incubated with 8 M HC for 15 min at 37°C. After addition of 0.5 mol % DPH, the fluorescence was monitored as described previously (38). See supplemental material for additional information.
Differential Scanning Calorimetry (DSC)-Lipid films of DPPC, C 16 -Cer, and limonoid HC were formed and then hydrated in 50 mM Hepes-NaOH buffer (pH 7.5) containing 0.5 mM EDTA. The final lipid concentration in the vesicles was 1 mM. DSC thermograms were recorded at a scan rate of 60°C/h for all samples, and each scan was performed at least 15 times. The obtained data were analyzed as described previously (39). See supplemental material for additional information.

Fluorescence Image Screening of a Chemical Library Identifies a Limonoid Compound as an Inhibitor of SL Metabolism-In
this study, we screened a commercial chemical library of natural compounds for inhibitors of SL metabolism using an IN Cell Analyzer 1000 automated fluorescence imaging system (18). For this purpose, we employed the SM-specific toxin lysenin (20,40). After binding to SM-enriched domains in the plasma membrane, lysenin induces pore formation and cell lysis (21). Compounds that reduce the cell surface SM content make the cells resistant to lysenin. Cell number, shape, and fluorescence intensity were analyzed using the IN Cell Analyzer 1000 followed by LSM 510 confocal microscopy ( Fig. 1, A-H). DHS was added to the medium to exclude serine palmitoyltransferase inhibitors and to focus the assay on the steps that take place after DHS synthesis. Lysenin treatment caused cell shrinkage as revealed by Bodipy labeling (Fig. 1E) and a smaller condensed nucleus as monitored by DAPI (Fig. 1F). The nucleus is clearly observable in the differential interference contrast image after lysenin treatment (Fig. 1H). From the initial screen of 2011 compounds, only CHC, a limonoid derivative, significantly inhibited lysenin-induced cell death under these experimental conditions. Bodipy staining showed that cells were well spread (Fig. 1A), and the nucleus shape was normal (Fig. 1B) as confirmed by the differential interference contrast image (Fig. 1D), which was distinct from that in Fig. 1H.
Specific Inhibitory Effect of 3-Chloro-8␤-hydroxycarapin-3,8hemiacetal on de Novo SM Biosynthesis-These results suggest that CHC affects the biosynthesis, degradation, or recycling of SM. To study the effect of CHC on de novo SL synthesis, CHO cells were preincubated with various concentrations of CHC for 22 h and then labeled with [ 14 C]serine in the presence or absence of the compound. CHC induced a selective dose-dependent inhibition of [ 14 C]SM synthesis displaying a 50% inhibition of SM synthesis at ϳ5 M ( Fig. 2A). A small decrease in the glycolipids GlcCer and G M3 (N-acetylneuranimyl-lactosylceramide) was observed at high concentrations. CHC did not affect cell viability up to 100 M (data not shown). In contrast to the effect on SM, CHC did not significantly influence the incorporation of [ 14 C]serine into phosphatidylserine (PS) and PE (Fig. 2B). One hour of preincubation with CHC was as effective as 22 h of preincubation to specifically inhibit SM synthesis (Fig. 2C). This is similar to the effect observed with the CERT inhibitor HPA-12 ( Fig. 2C) (27) and suggests that CHC itself but not its metabolites plays a role in the inhibition of SM synthesis.
We also examined the biosynthesis of phosphatidylcholine (PC) because SM is synthesized by the transfer of phosphocholine from PC to Cer. Similar to HPA-12 incubation, CHC treatment did not significantly affect the incorporation of [ 14 C]choline into PC (Fig. 2D). Quantification of the phosphorus content after lipid extraction and TLC analysis indicated that SM content displayed a 40 -50% decrease when CHO and HeLa cells were grown for 2 days in the presence of 5 M CHC (Fig.  2E) suggesting that the compound did not display cell specificity. SM composed ϳ5% of total phospholipids both in CHO and HeLa cells. The total phospholipid content was not significantly altered by a 2-day treatment with 5 M CHC (Fig. 2F).
These results indicate that CHC specifically inhibits SM biosynthesis and thus, as a result, cells became resistant to lysenin. However, this does not rule out the possibility that CHC accelerates the degradation of SM. Determination of the turnover of radiolabeled de novo synthesized SM (Fig. 2G) clearly indicated that CHC did not affect SM turnover, thus excluding the possibility that CHC could stimulate SM degradation by the activation of sphingomyelinase.
Fluorescent SM analogs recycle between the plasma membrane and the recycling endosomes (31,32,41,42). Slow recycling back to the plasma membrane may decrease plasma membrane SM. Thus, the influence of CHC on SM recycling was determined (Fig. 3, A-F). CHO cells preincubated in the presence and absence of CHC were first labeled with the fluorescent SM analog, C 6 -NBD-SM, and the fluorescent lipids were then internalized in the recycling endosomes. Cell surface fluorescence was then quenched with sodium dithionite (30,43), and cells were further incubated in the absence of dithionite. The re-emerging cell surface labeling indicates the recycling of C 6 -NBD-SM to the plasma membrane. The results show that CHC did not significantly affect the recycling of a fluorescent SM analog.
We also measured the effect of CHC on bulk vesicular flow to the plasma membrane via the Golgi apparatus with a GFP-labeled tsO45 mutant of vesicular stomatitis virus G protein (44,45). GFP-vesicular stomatitis virus G accumulates in the ER at a nonpermissive temperature and is subsequently transported to the Golgi and the plasma membrane at a permissive temperature (supplemental Fig. S1). No notable differences were detected between the control (supplemental Fig. S1, A-C) and CHC-treated cells (supplemental Fig. S1, D-F), indicating that CHC did not inhibit the intracellular transport of GFP-vesicular stomatitis virus G. Altogether, these results suggest that CHC specifically inhibits the de novo biosynthesis of SM.
CHC Inhibits the Intracellular Transport of de Novo Synthesized Cer to the Site of SM Synthesis-We then examined whether the formation of Cer was the target of CHC by measuring the CerS activity in vitro. Both CerS-2 and CerS-5 activities were unaffected by the presence of CHC (supplemental Fig.  S2, A and B). Furthermore, CHC did not inhibit the activity of the SM synthase in vitro using as substrates the fluorescent short chain Cer analogs C 5 -DMB-Cer and C 12 -NBD-Cer and the radioactive short chain Cer 14 C-C6:0 Cer (supplemental Fig. S2C).
Next, we examined the possibility that CHC inhibits the transport of Cer from the ER, where it is synthesized, to the Golgi apparatus, where it is converted to SM by SM synthase. BFA induces fusion of the Golgi apparatus with the ER (46) and thus abolishes the inhibitory effect of the CERT inhibitor HPA-12 (36). BFA treatment of CHO cells induced a 2-3-fold increase in SM synthesis (Fig. 4) as reported previously (47). SM synthesis in the presence of BFA was comparable in CHCtreated and control cells. The BFA treatment similarly sup- pressed the inhibition of SM biosynthesis induced by HPA-12 ( Fig. 4) (27). In contrast, SM synthesis was inhibited both in the presence and absence of BFA in cells treated with the ceramide synthase inhibitor fumonisin B1 (data not shown). These results indicate that CHC does not inhibit SM synthase and hence CHC, similar to HPA-12, appears to exert an effect on the intracellular transport of endogenously formed Cer from the ER to the Golgi.
De novo synthesized Cer is transported to the Golgi apparatus by CERT (10, 36). The transport of an exogenously added fluorescent short chain Cer analog, C 5 -DMB-Cer, to the Golgi apparatus is also dependent on CERT activity (27,48). We then examined the effect of CHC on the intracellular distribution of C 5 -DMB-Cer (49). First, CHO cells were labeled with C 5 -DMB-Cer at 4°C in the absence and the presence of the drugs (Fig. 5, A-C). After washing, cells were chased at 37°C with and without the drugs. Before the chase, the fluorescence was distributed at the plasma membrane in the control as well as drugtreated cells. After the chase, the fluorescence accumulated in the perinuclear region in the control and CHC-treated cells (Fig. 5, D and E, respectively). In contrast, in HPA-12-treated cells, the fluorescence did not accumulate at the perinuclear area ( Fig. 5F) (27). These results suggest that CHC did not inhibit the transport of the exogenously added C 5 -DMB-Cer to the Golgi apparatus. This is further supported by the observation that CHC did not modify the conversion of C 5 -DMB-Cer to C 5 -DMB-SM (Fig. 5G).
The naturally occurring Cer are highly hydrophobic due to their long acyl chains. Consequently, their physical properties differ from those of their fluorescent or short chain counterparts. Indeed, model membrane studies showed that fluorescent short chain Cer analogs underwent spontaneous membrane transfer much faster (50, 51) than natural C 16 -Cer (52). Thus, it is speculated that CHC did not affect the transport of C 5 -DMB-Cer due to the spontaneously rapid transfer of the lipid.
Next, we studied the effect of CHC on the conversion of the endogenously formed Cer to SM by metabolically labeling the de novo synthesized Cer pool. For this purpose, CHO cells were pulse-labeled with [ 3 (53). Cells were then incubated with and without drugs at 4°C and chased at 37°C in the presence of the CerS inhibitor fumonisin B1. In contrast to the fluorescent short chain Cer, CHC inhibited the formation of SM from endogenous Cer to a similar extent as HPA-12 treatment (Fig. 5H) (27). This result indicates that CHC inhibited the transport and, as a consequence, the conversion of endogenously formed Cer to SM.

H]DHS at 15°C to allow the formation of [ 3 H]Cer without further conversion to [ 3 H]SM
CHC Inhibits the CERT-mediated Cer Extraction from Isolated ER Membranes-The initial step of CERT-mediated transport of Cer from the ER to the Golgi apparatus requires the extraction of Cer from the ER. To determine the influence of CHC on Cer extraction from the ER membrane, we used an in vitro assay utilizing CHO ER membranes containing long chain [ 3 H]Cer formed from SPH and palmitoyl-CoA (10). First, we confirmed that recombinant hCERT extracted [ 3 H]Cer but not [ 3 H]SPH from the ER membrane (Fig. 6). Interestingly, preincubation of the ER membrane with CHC inhibited Cer extraction similar to the CERT inhibitor HPA-12 (Fig. 6). This indicates that CHC interferes with the CERT-mediated extraction of long acyl chain Cer from ER membrane.

Screening of 21 Additional Limonoids Identified as New Inhibitors of SL Metabolism-
The results indicate that CHC is a new type of inhibitor of SL biosynthesis. Because CHC belongs to a large family of natural compounds, called limonoids, we examined the effect of other natural and derivatized limonoids on SL biosynthesis (see Fig. 7 for limonoid structure).   half of the de novo synthesis of SM. CHC and HPA-12 were used as positive controls. As observed previously with CHC, these limonoids did not affect the biosynthesis of other phospholipids, e.g. PS, PE (supplemental Fig. S3A), and PC (supplemental Fig. S3B). Furthermore, the SM content was significantly decreased when CHO cells were grown for 2 days in the presence of 5 and 10 M HC (Fig. 8B), whereas the total phospholipid content was unchanged (Fig. 8C) as observed previously with CHC (see Fig. 2, E and F).
In contrast, anthothecol (compound 1) and cedrelone (compound 4) inhibited SM biosynthesis (Fig. 8A) in a nonspecific way because they decreased the biosynthesis of PS and PE, respectively (supplemental Fig. S3, A and B). Both compounds were cytotoxic during longer incubation times (22 h, data not shown). Preliminary results indicated that long time treatment (22 h) with 10 M methylangolensate (compound 9), mexicanolide (compound 10), or humilin A (compound 12) did not affect SM biosynthesis in contrast to the CerS inhibitor fumonisin B1 (supplemental Fig. S3C). Odoratone (compound 13) was toxic at concentration below 1 M (data not shown).
Interestingly, as observed previously with CHC and HPA-12 (Fig. 4), BFA treatment abolished the inhibitory activity of the four newly identified limonoids that were active in SM biosynthesis (Fig. 9A) suggesting that they might also affect Cer availability. Next, we selected three structurally distinct limonoids (see Fig. 7 and under "Discussion"), the active HC and Ged as well as the nonactive fissinolide (Fiss, compound 5), to determine their influence on the recombinant hCERT-mediated extraction of Cer from isolated ER membranes. HC as well as Ged reduced Cer extraction in a concentration-dependent manner (Fig. 9B). In contrast, even at concentrations of up to 20 M, Fiss did not affect CERT activity. The presence of HC also interfered with the recombinant hCERT-mediated extraction of Cer from artificial liposomes composed of PC, PE, and [ 14 C]C 16 -Cer (10). HPA-12 was used as positive control, whereas neither DMSO (control, 0.1% final concentration) nor Fiss affected the Cer extraction (Fig. 9B).
Hemiacetals are usually considered unstable. Nevertheless, cyclic hemiacetals, unlike their acyclic counterparts, tend to exhibit a remarkable stability (54), especially if pyranoid or furanoid rings form, such as in the case of glucose and fructose that predominantly exist in their cyclic form in solution. To probe the stability of HC, a cyclic hemiacetal, we subjected it to acidic (pH 5.0) and neutral (pH 7.0) conditions for 24 h at 37°C. Mass spectroscopic (MS) analysis revealed no significant change in either sample (supplemental Fig. S4A). A major peak at m/z 485 [M ϩ H] ϩ was detected corresponding to the intact hemiacetal and a minor peak at m/z 467 corresponding to the fragment without the hydroxyl. This indicates that the cyclic hemiacetal HC and its derivatives can tolerate the pH range of 5-7, which is a range that would be encountered after cellular uptake.
We also showed that the HC content in the cell extracts after 1 and 24 h of incubation was stable (supplemental Fig. S4B), as revealed by MS-MS quantification (precursor ion m/z 485.5, [M ϩ H] ϩ ) with multiple reaction monitoring mode (supplemental Fig. S5). The steady cellular content of HC after 1 and 24 h of incubation is in good agreement with the similar inhibition of SM biosynthesis at short and long time incubation periods (see Fig. 2C). These results suggest that the active limonoids decreased the de novo SM biosynthesis by inhibiting the CERT-dependent extraction of long acyl chain Cer from the ER membrane.
Limonoid HC Induces the Formation of Cer-rich Membrane Domains-The intermembrane Cer transport catalyzed by CERT was recently demonstrated to be markedly reduced when Cer is in a tightly packed environment. Conversely, Cer in fluid membranes was shown to be available for CERT-mediated transfer (55). These results indicate that the membrane matrix surrounding Cer, i.e. Cer miscibility, crucially affects CERT activity. To examine the effect of limonoids on the fluidity of bulk membrane, we measured the fluorescence anisotropy of DPH (38) incorporated into limonoid-containing liposomes (Fig. 10, A and B). DPH localizes to the hydrophobic core of the membrane bilayer, thus providing information on the packing properties of the bulk membrane. DPH anisotropy in DPPC vesicles with or without HC (the range of the molar ratio DPPC/HC was from 100:1 to 5:1) exhibits identical trends between 20 and 60°C with an apparent transition temperature of ϳ41°C (Fig. 10A). Similarly, HC presence in the PC/PE/Cer liposomes (PC/PE/Cer/HC, 32:8:2:0.4) did not modify DPH anisotropy between 20 and 60°C (Fig. 10B).
We further examined the effect of limonoids on the physical properties of model membranes using DSC (Fig. 10, C-F). The heating and cooling scans of pure DPPC liposomes (Fig. 10C) displayed a characteristic pretransition peak at 35°C followed by a typical gel (L ␤ ) to liquid crystalline (L ␣ ) phase transition peak at 41°C, as described (56). It is known that Cer exhibits a main endothermic transition at 90°C (57). The addition of HC (1:8 drug/lipid molar ratio) to pure DPPC liposomes abolished the pretransition temperature, but it did not significantly modify the main phase transition peak (Fig 10D). This is in good agreement with the DPH result. In a 1:8 Cer/DPPC mixture, the endothermic peak became broadened (Fig 10E). The cooling thermogram showed the coexistence of a large 41°C peak and a small 52°C peak, indicating the presence of phase-separated DPPC-and Cer-enriched domains (58). The presence of HC in this Cer/DPPC mixture sharpened the transition peak arising from the Cer-enriched domains. In addition, according to the cooling scans, the higher temperature transition peak of the Cer-enriched domains shifted to 55°C in the presence of HC (Fig 10F). These results suggest that limonoids, like HC, reduce the miscibility of Cer in DPPC liposomes by promoting the formation of Cer-rich domains.

DISCUSSION
Limonoids are a large family of natural compounds and have been employed in traditional medicine (59). Recent reports have highlighted the antimalarial and antiproliferative activities of limonoids, such as Ged (23, 60 -62). However, their molecular mechanism of action is not well understood. This study   JULY 13, 2012 • VOLUME 287 • NUMBER 29

Limonoids Inhibit Ceramide Traffic
indicates for the first time that certain limonoids, including Ged, are unique inhibitors of SM biosynthesis.
The screening method was based on the cytotoxicity of lysenin, a SM-specific pore-forming toxin (20,21). Thus, in this system, each compound capable of decreasing the SM on the cell surface would render the cells resistant to lysenin. This lipid-specific protein-based cell screening appears to be a very efficient technique for high throughput analysis of small compounds affecting lipid metabolism (18). In the present screen performed in the presence of DHS, we focused on SM biosynthesis and excluded serine palmitoyltransferase inhibitors because a number of them such as ISP-1/myriocin, sphingofungins, lipoxamycin, and sulfamisterine have been reported (25,(63)(64)(65)(66). Out of a library of 2011 natural products and derivatives, we identified 3-chloro-8␤-hydroxycarapin, CHC, a limonoid derivative. Biochemical analysis indicated that CHC and other limonoids selectively inhibit de novo SM biosynthesis. In the initial steps of SM biosynthesis, DHS is converted into dihydroceramide by CerS. Mammalian cells contain six members of the CerS family (9). Each CerS exhibits a different substrate fatty acid specificity (67,68). CHC did not affect the in vitro activity of CerS2 and CerS5, responsible for the synthesis of the very long chain Cer, i.e. C 22 -C 24 Cer, and C 16 Cer, respectively. This suggests that CHC inhibits a step that takes place after Cer formation.
It is noteworthy that BFA treatment, which induces fusion of the Golgi apparatus and the ER (46), rescued the limonoidinduced inhibition of SM biosynthesis. This demonstrates first that the SM synthase 1 activity is not influenced by these limonoids, and second, it demonstrates that the active limonoids inhibit SM biosynthesis by interfering with Cer trafficking from the ER to the Golgi apparatus, as observed with the specific CERT inhibitor, HPA-12 (27). However, in contrast to HPA-12, these limomoids do not appear to inhibit the transport of the short chain fluorescent Cer analog to the Golgi apparatus or its subsequent conversion to corresponding SM. Nevertheless, the conversion of de novo synthesized Cer to SM was inhibited indicating that these limonoids selectively inhibit the CERT-regulated transport of endogenous long chain Cer. In a separate set of in vitro experiments, we confirmed that these limonoids inhibited the CERT-dependent extraction of endogenously formed long chain Cer from isolated ER membranes, as well as the extraction of long chain Cer from artificial liposomes. Such a differential effect based on the acyl chain length of Cer may be due to differences in the physical properties of long and short chain Cer, the latter of which is known to move relatively more freely between membranes. Model membrane studies showed that a fluorescent short chain Cer analog underwent spontaneous membrane transfer much faster (t1 ⁄ 2 , minute order) (50, 51) than natural C 16 -Cer (t1 ⁄ 2 , days) (52). CERT effi- ciently reduces the transfer time of long chain Cer (53), selectively transferring Cer with C 14 up to a chain length of C 20 (36). In contrast, the rapid spontaneous transfer of C 6 -NBD-Cer (t1 ⁄ 2 ϭ 0.4 min) masks the CERT-mediated transfer (36). Because of its hydrophobicity, natural long chain Cer is embedded in the ER membrane. Its transport to the Golgi requires a two-step process catalyzed by CERT as follows: first, extraction from the membrane, and second, transfer between the ER and Golgi apparatus membrane (12,69). The C-terminal steroidogenic acute regulatory protein-related lipid transfer domain of CERT contains the amphiphilic cavity for Cer binding (in a 1:1 binding ratio) and the pleckstrin homology domain for phosphatidylinositol 4-phosphate binding at the trans-Golgi (36,70). The crystal structure of the CERT-steroidogenic acute regulatory protein-related lipid transfer domain in complex with Cer demonstrates that the size and shape of its cavity controls the chain length limit and stereo-specificity of Cer recognition by CERT. For example, the binding of C 16 -and C 18 -Cer com-pletely fills the hydrophobic part of the cavity, whereas in the case of C 6 -Cer some empty space remains (70). This is in good agreement with the fact that CERT efficiently transfers Cer with a chain length of C 14 up to C 20 .
Synthetic HPA-12 is structurally similar to D-erythro-ceramide and thus acts as a competitive inhibitor of CERT (27,36,71). In contrast, the structure of limonoids does not resemble Cer or other sphingolipids (Fig. 7). Recently, it was demonstrated that the miscibility of Cer in the membrane affects its CERT-mediated extraction (55). Cer was less efficiently extracted by CERT from tightly packed membranes compared with more fluid ones. This is in line with the DSC analysis, which indicated that the presence of HC reduced the miscibility of Cer and DPPC inducing the formation of Cer-rich domains in the membrane.
Limonoids include a structurally heterogenous group of natural products with a prototypical structure of a furan attached to a tetracyclic core (A-D rings) (59,72). Synthetic access to limonoids represents a tremendous challenge due to the complexity of their molecular structure, as recently highlighted by the completion of the total synthesis of azadirachtin, a natural insecticide extracted from the neem tree (73). The limonoids and derivatives of this study can be roughly divided into three groups based on their structural features (Fig. 7). The first group is composed of the gedunin-like limonoids possessing a predominantly flat and fused core ring system. This group is composed of compounds 1, 4, 7, 13, 14, 15, 21, and Ged (compound 6) as well as the loosely related 9, 17, and 18. The 3-fold decrease of the inhibitory activity of compound 14 compared with 7 suggests a slightly adverse effect of the lactone in ring D. However, a variation in the A ring substitution pattern exerts only a minor influence on the inhibitory activity, as demonstrated by the comparable activity of Ged, compounds 14 and 15. In contrast, the presence of an ␣-hydroxy-␣,␤-unsaturated ketone in ring B was accompanied by reduced specificity of the inhibitory activity. The second group, the carapin-like limonoids, composed of compounds 3, 8, 10, 12, and Fiss (compound 5) feature a bridged ketone between the A and B rings and the lack of an epoxide at ring D. This completely abolished any inhibition of SM biosynthesis. In contrast, in the third group, limonoids featuring multiple bridged ring systems, such as compounds 2, 11, 16, 19, HC, and CHC, tend to inhibit SM biosynthesis. Nevertheless, a high level of bulky substituents in combination with an ortho ester, as in compounds 2, 11, and 19, resulted in only moderate inhibitory activity. Interestingly, potent inhibitors, such as compounds 16, HC, and CHC feature a hemiacetal bridge in their pentacyclic core ring system. Unlike acyclic hemiacetals, their cyclic counterparts tend to exhibit a remarkable stability (54,74,75). This is further supported by the MS analysis indicating a sufficient stability of HC in the pH range of 5-7. In addition, the quite constant cellular level of HC after 1 and 24 h of incubation suggests that the limonoid itself, and not a metabolite, plays an active role in the inhibition of SM biosynthesis. This is further supported by the direct inhibitory effect of HC on the CERT-dependent extraction of Cer in vitro.
It is worth pointing out that SL metabolism has been associated with cancer cell proliferation and Plasmodium development (76 -79). The limonoid-induced inhibition of SM biosynthesis represents a plausible explanation of the anti-cancer and anti-malaria properties of these compounds. We hope that these results will help spur the search for novel natural products with a high degree of selectivity to avoid the problem of multidrug resistance often encountered in cancer and malaria therapy.