Cell density-dependent reduction of dihydroceramide desaturase activity in neuroblastoma cells.

We applied a metabolic approach to investigate the role of sphingolipids in cell density-induced growth arrest in neuroblastoma cells. Our data revealed that sphingolipid metabolism in neuroblastoma cells significantly differs depending on the cells' population context. At high cell density, cells exhibited G0/G1 cell-cycle arrest and reduced ceramide, monohexosylceramide, and sphingomyelin, whereas dihydroceramide was significantly increased. In addition, our metabolic-labeling experiments showed that neuroblastoma cells at high cell density preferentially synthesized very long chain (VLC) sphingolipids and dramatically decreased synthesis of sphingosine-1-phosphate (S1P). Moreover, densely populated neuroblastoma cells showed increased message levels of both anabolic and catabolic enzymes of the sphingolipid pathway. Notably, our metabolic-labeling experiments indicated reduced dihydroceramide desaturase activity at confluence, which was confirmed by direct measurement of dihydroceramide desaturase activity in situ and in vitro. Importantly, we could reduce dihydroceramide desaturase activity in low-density cells by applying conditional media from high-density cells, as well as by adding reducing agents, such as DTT and L-cysteine to the media. In conclusion, our data suggest a role of the sphingolipid pathway, dihydroceramides desaturase in particular, in confluence-induced growth arrest in neuroblastoma cells.


Cell-culture and cell-cycle analyses
The SMS-KCNR neuroblastoma cell line was obtained from Dr. C. Pat Reynolds (Texas Tech University, Amarillo, TX). The SMS-KCNR cells were grown in RPMI Medium 1640 (with L-glutamine, and supplemented with 10% fetal bovine serum 100 IU/ml penicillin and 100 IU/ml streptomycin) and incubated at 37°C and 5% CO 2. Cell culture media and supplements are from Invitrogen (Carlsbad, CA). To achieve different cell densities, we started the cell cultures with different cell numbers, i.e., in 10 cm cell culture dishes, we used 0.5 × 10 6 and 4 × 10 6 neuroblastoma cells, which, after 48 h of incubation, reached 50% (low) and 90% (high) cell densities, respectively. To ensure consistency for all experiments, we used the same conditions for initial cell numbers and incubation time. DTT 0.1 M solution (Invitrogen) and L -cysteine (Sigma, St. Louis, MO) 200 mM stock solution in water were added to low-density cells cultured for 24 h. Subsequently, cells were cultured for additional 24 h before in situ dihydroceramide desaturase activity was measured. For fl ow cytometry cell-cycle analyses, cell samples were trypsinized and centrifuged, and then cell pellets were washed twice with ice-cold 1× PBS and fi xed in 3 ml of 70% ethanol at 4°C. On the day of the analyses, cells were treated with 20 g/ml DNase-free RNase A and stained with 100 µg/ml propidium iodide (PI) for 30 min. Subsequently, the samples were analyzed with a FACStarplus fl ow cytometer (BD Biosciences).

Lipid extraction and mass spectrometry measurements
Cell pellets were fortifi ed with internal standards for endogenous ceramide and dihydroceramide (C 13 /C 16 ceramide, C 17 /C 16 ceramide, and C 17 /C 24:1 ceramide); for endogenous sphingomyelin (C 17 sphingomyelin, C 12 sphingomyelin, and C 12 dihydrosphingomyelin); for endogenous monohexosylceramide (C 8 glucosylceramide); for C 17 ceramide and C 17 dihydroceramide (C 13 /C 16 ceramide and C 13 /C 22 ceramide); for C 17 sphingomyelin (C 8 sphingomyelin, C 12 sphingomyelin, and C 12 dihydrosphingomyelin); and for C 17 monohexosylceramide (C 8 glucosylceramide and C 12 glucosylceramide). Subsequently, lipids were extracted twice with 2 ml ethyl acetate/isopropanol/water (60/30/10 v/v) solvent, dried under a stream of nitrogen, and resuspended into 150 l 1 mM NH 4 COOH in 0.2% HCOOH in methanol. ESI/ MS/MS analyses of endogenous and C 17 -sphingosine backbone sphingolipid species were performed on a Thermo Finnigan TSQ Quantum triple quadrupole mass spectrometer operating in a multiple reaction-monitoring positive ionization mode using a modifi ed version of the published protocol ( 24 ). Samples were injected on the HP1100/TSQ Quantum liquid chromatography/mass spectrometry (LC/MS) system and were gradienteluted from the BDS Hypersil C8, 150 × 3.2 mm, 3 m particle size column, with a 1.0 mM methanolic ammonium formate / 2 mM aqueous ammonium formate mobile phase system. Peaks corresponding to the target analytes and internal standards were collected and processed with Xcalibur software. Quantitative analyses were based on the calibration curves generated by spiking an artifi cial matrix with known amounts of the target analyte synthetic standards and an equal amount of the internal standard. The target analyte/internal standard peak areas ratios were plotted against analyte concentration. The target analyte/internal standard peak area ratios from the samples were similarly normalized to the internal standard and compared with the calibration curves using a linear regression model. The phosphate contents of the lipid extracts were used to normalize the MS measurements of sphingolipids. The phosphate contents of the lipid extracts were measured with a standard curve analysis and a colorimetric assay of ashed phosphate ( 25 ).
De novo sphingolipid synthesis starts with condensation of serine and palmitoyl CoA, catalyzed by serine-palmitoyl transferase. The product, 3-keto-dihydrosphingosine, is fi rst reduced to dihydrosphingosine, which is then combined with a fatty acid CoA by ceramide synthase to produce dihydroceramide ( 1,8 ). The fatty acid CoA can vary in chain length, and different ceramide synthase (CerS) isoforms (six in mammals) have different specifi city toward the length of the fatty acid CoA ( 9 ). Thus, the dihydroceramide product and, subsequently, ceramide and the complex sphingolipids have fatty acid chain-length variability. The next step of de novo sphingolipid synthesis is catalyzed by dihydroceramide desaturase, which converts dihydroceramide to ceramide, a precursor of complex sphingolipids ( 10,11 ). Breaking down of ceramide to sphingosine and a fatty acid is catalyzed by ceramidases ( 12 ). Sphingosine can be reacylated by ceramide synthase to produce ceramide with a different chain length ( 13 ) or activated by sphingosine kinases to sphingosine-1-phosphate (S1P) ( 14 ). S1P can be deactivated by S1P phosphatase or broken down by S1P lyase to hexadecenal and ethanolamine phosphate in the last degradation step of the sphingolipid pathway ( 15 ).
The sphingolipid metabolites most studied for their role as second messengers are ceramide and S1P: ceramide as a promoter of cell death and S1P as a promoter of cell proliferation (reviewed in Refs. 1 and 15 ). On the other hand, until recently, the precursor of ceramide, dihydroceramide, was considered a bio-inert molecule, and a short-chain (C 2 ) dihydroceramide analog was used as a negative control in experiments to study the effects of ceramide ( 16,17 ). Recent studies using depletion or inhibition of dihydroceramide desaturase, which uses dihydroceramide as a substrate, changed the way we think about this concept ( 10,(18)(19)(20)(21)(22). Fenretinide and resveratrol treatments are reported to inhibit dihydroceramide desaturase and increase endogenous dihydroceramides, resulting in autophagy in cancer cells ( 10,20,22,23 ). Treatment of tumor cells with celecoxib, a COX2 inhibitor, has also been shown to increase some dihydroceramide species by inhibiting dihydroceramide desaturase. In the same study, celecoxib also reduced VLC ceramides. The authors suggested that sphingolipid changes resulting from dihydroceramide desaturase inhibition contribute to the COX2-independent, antiproliferating effect of celecoxib ( 19 ). Moreover, in a recent study, oxidative stress was shown to lead to accumulation of dihydroceramide due to dihydroceramide desaturase inhibition in several cell lines ( 21 ). Depletion of dihydroceramide desaturase by siRNA has shown to result in accumulation of dihydroceramides and G0/G1 cell-cycle arrest in neuroblastoma cells, suggesting a role for dihydroceramides in growth inhibition ( 10 ).
In this work, we showed for the fi rst time that saturation cell density results in signifi cant reduction of dihydroceramide desaturase activity, likely due to increased levels of free thiols in the media. The decreased dihydroceramide desaturase activity at high cell density led to an increased dihydroceramide/ceramide ratio, suggesting a role of increased dihydroceramides and/or reduced ceramide in contact inhibition. dodecanoyl]4,5 dihydrosphingosine bromide, which was synthesized and kindly provided by the Lipidomics Core Facility of MUSC ( 26 ), to the media for 30 min. Cells were washed three times with 1× PBS and cell pellets were subjected to immediate lipid extraction followed by MS measurement of the product C 12 -CCPS (D-erythro -2-N -[12'-(1"-pyridinium) dodecanoyl]sphingosine bromide). The results were normalized to the phosphate contents of the lipid extracts.
In vitro dihydroceramide desaturase activity assay SMS-KCNR cells were grown at low (50%) and high (90%) densities for 48 h before the media was removed. Cells were washed twice with ice-cold PBS, collected by scraping, and then total cell homogenates were prepared as described previously ( 21 ). Briefl y, cell pellets were resuspended in 1 ml 5 mM Hepes buffer (pH 7.4, containing 50 mM sucrose) and homogenized by passing through an insulin syringe. Cell homogenates were centrifuged (250 g ) 5 min at 4°C to remove unbroken cells. The in vitro assay was performed as described previously ( 21 ). Briefl y, the assay was performed for 20 min at 37°C using the total cell homogenate (400 g) as an enzyme source, 2 nM N -octanoyl-  ( 27,28 ). For calculations, the amount of radioactivity was reduced by half to account for the amount of radioactivity relevant to enzyme activity; the enzyme extracts only one of the two hydrogen atoms and only one is "randomly" labeled with 3 H in the substrate.

Western blot
Total lysates from neuroblastoma cell grown at low (50%) and high (90%) cell densities were separated on an SDS gel and transferred to a nitrocellulose membrane using standard techniques ( 29 ). Dihydroceramide desaturase, DEGS-1 [homolog of Drosophila melanogaster des-1 (degenerative spermatocyte gene-1)], and ␤ actin were labeled with their specifi c primary antibody for 1 h at 25°C. Anti-DEGS-1 antibody (MLD 3906) was a generous gift from Dr. Gordon N. Gill (University of California, San Diego, CA). The anti-cytochrome B5 reductase and anti-cytochrome B5 antibodies were from Novus Biologicals (Littleton, CO). Subsequently, primary antibodies were detected with appropriate secondary antibodies conjugated with horseradish peroxidase (1 h, 25°C) and detected with ECL (Amersham Biosciences, Sweden) according to the manufacturer's protocol.

Ellman's test
Ellman's test was performed for quantitation of free thiol groups in the media ( 30 ). Fifty microliters of 3 mM of freshly prepared dithiobismitrobenzoic acid (Sigma) in 100 mM potassium phosphate buffer, pH 7.2, supplemented with 0.1 mM EDTA was mixed with 50 l conditioned media for 30 min at room temperature (24°C). After the incubation, the absorbance of the samples was measured at 415 nm.

Statistical analysis
Statistical analyses were performed by using Student t -test (SigmaPlot).

RNA isolation, cDNA synthesis, and RT-PCR
Total RNA isolation was performed with RNeasy® Mini Kit (Qiagen, Valencia, CA) according to manufacturer's instructions.
The concentration and quality of total RNA preparations were evaluated spectrophotometrically. cDNA was synthesized from 1 g of the total RNA using Superscript II Kit for fi rst-strand synthesis (Invitrogen, Carlsbad, CA). RT-PCR was performed on a Bio-Rad iCycler detection system using iQ SYBR Green supermix (Bio-Rad, Hercules, CA). Standard reaction volume was 25 l containing 12.5 l iQ SYBR Green supermix, 9.5 l dH 2 O, 0.4 M specifi c oligonucleotide primers ( Table 1 ), and 50 ng cDNA template. Initial steps of RT-PCR were 2 min at 50°C, followed by 3 min hold at 95°C, and 40 cycles consisting of a 10 s melt at 98°C, followed by a 45 s annealing at a temperature (Ta) specifi c for the primer pair ( Table 1 ), and 45 s extension at 72°C. All reactions were performed in triplicate and threshold for cycle of threshold (Ct) analysis of all samples was set at 0.15 relative fl uorescence units. The data were normalized to an internal control gene GAPDH.

Metabolic labeling with C 17 long-chain bases
SMS-KCNR cells were grown at low (50%) and high (90%) cell densities for 48 h before 1 M C 17 dihydrosphingosine [1 mM stock in ethanol; kindly provided by the Lipidomics Core Facility of the Medical University of South Carolina (MUSC)] or 1 M C 17 sphingosine (1 mM stock in ethanol; Avanti Polar Lipids, Alabaster, AL) were added to the growth media for 30 min. Cells were washed three times with 1× PBS, and cell pellets were subjected to immediate lipid extraction followed by MS analysis. The phosphate contents of the lipid extracts were used to normalize the results.

In situ dihydroceramide desaturase activity assay
In situ dihydroceramide desaturase activity was measured as previously described ( 10 ). Briefl y, cells were cultured at low (50%) and high (90%) cell densities for 48 h before addition of 500 nM substrate, C 12   sphingolipid-degrading enzymes, neutral and acid sphingomyelinases, acid and alkaline ceramidase 1, and S1P lyase had increases in message levels, which correlated with the decrease of total ceramide and sphingomyelin at high cell density compared with low cell density. Surprisingly, a number of sphingolipid synthesis enzymes (e.g., serine-palmitoyl transferase, CerS2, CerS4, glucosylceramide synthase, and sphingomyelin synthase 2) also had elevated message levels at high cell density. These data indicate that cell density affected the expression of sphingolipid enzymes at multiple pathway points, suggesting a possible role of the sphingolipid pathway in adjusting the cells to population density.

At high cell density, neuroblastoma cells preferentially synthesized VLC sphingolipids
The changes in the sphingolipid enzyme expression data at different population densities were not fully refl ected in the differences of steady-state sphingolipid levels ( Figs. 2 and 3 and supplementary Fig. VI). To better understand the sphingolipid metabolism differences between low and high cell densities, we performed metabolic labeling with C 17 sphingoid bases, nonnatural precursors of sphingolipids, and analyzed the newly formed nonnatural C 17 sphingolipids by MS ( 32 ). SMS-KCNR neuroblastoma cells were cultured for 48 h at low and high population densities and were metabolically labeled with 1 µM C 17 sphingosine for 30 min, and then lipids were extracted from the cell pellets and subjected to MS. The conversion of C 17 sphingosine to total C 17 ceramide, C 17 sphingomyelin, and C 17 monohexosylceramide was lower in the high-density samples compared with those of low density ( Fig. 4A , D, G ), which agreed with results from steady-state measurements of total sphingolipids ( Fig. 2 ). Interestingly, a clear difference was observed between the labeling of the LC sphingolipids (C 14 -C 18 ) and VLC sphingolipids at different population densities ( Fig. 4B, C, E, F, H, I ). The LC sphingolipids are preferentially synthesized at low cell density, whereas the VLC ones are preferentially synthesized at high cell density. The increased synthesis of VLC sphingolipids at high cell density ( Fig. 4C, F, I ) did not result in increased steady-state VLC sphingolipids (supplementary Figs. II, III, and IV) at high cell density, suggesting increased degradation of VLC sphingolipids as the cell population density increased.
In addition to being a precursor for C 17 ceramide, C 17 sphingosine is a precursor for C 17 S1P as well. Our results showed that conversion of C 17 sphingosine to C 17 S1P was dramatically reduced at high cell density ( Fig. 5 ), indicating a decrease in S1P synthesis. Such reduced S1P synthesis at

High cell density in neuroblastoma cells induced cellcycle arrest and changes in sphingolipid levels
To test whether increased cell density induces growth arrest in neuroblastoma cells, we performed fl ow cytometry cell-cycle analyses in cells cultured at different densities. For our experiments, we used an adherent neuroblastoma cell line, SMS-KCNR. Results showed that densely populated cells had a higher proportion of cells in the G0/G1 phase of the cell cycle and a signifi cantly lower proportion of cells in the G2/M phases of the cell cycle compared with sparsely populated cells ( Fig. 1 ). These results clearly indicate a G0/G1 cell-cycle arrest and a reduced mitotic index in neuroblastoma cells cultured at high cell density.
Changes in the sphingolipid pathway have been implicated in inducing growth arrest ( 31 ); therefore, we investigated whether the G0/G1 cell-cycle arrest at high cell density in SMS-KCNR neuroblastoma cells was accompanied by changes in sphingolipid levels. SMS-KCNR cells cultured at low and high cell densities for 48 h were collected, and lipids were extracted from the cell pellets and subjected to LC/MS analyses. The results showed that the total levels of dihydroceramide were two times higher in densely populated cells compared with sparsely populated cells ( Fig. 2A ), whereas the opposite was true for total ceramide, sphingomyelin, and monohexosylceramide ( Fig. 2B-D ). Accordingly, all species of dihydroceramide were increased at high cell density (supplementary Fig. I), and all species of ceramide, monohexosylceramide, and sphingomyelin were decreased at high cell density (supplementary Figs. II, III, and IV, respectively). Sphingosine and S1P did not significantly change (supplementary Fig. V). These results suggest that the sphingolipid makeup of the neuroblastoma cells depends on cell density, and with increased cell density, the ratio of dihydroceramide/ceramide increased, which can be important in regulating G0/G1 cell-cycle arrest at high cell density.

High cell density results in increased message levels of multiple genes from the sphingolipid pathway
To investigate further the changes in sphingolipid metabolism at confl uence, we measured message levels of key enzymes of the sphingolipid pathway at low and high cell densities by RT-PCR in neuroblastoma cells. The sphingolipid genes with at least 2-fold increased expression at high cell density compared with low cell density are shown in Fig. 3 and supplementary Fig. VI. From the The results in Fig. 6C depict decreased synthesis of C 17 dihydrosphingosine-1-phosphate at a high cell density similar to data obtained with C 17 sphingosine labeling ( Fig. 5 ), further confi rming that at high cell density, cells dramatically reduce de novo synthesis of sphingoid base phosphates.
Analyses of the synthesis of the individual C 17 dihydroceramide species ( Fig. 7A -D ) revealed results similar to those obtained with C 17 ceramide synthesis from C 17 sphingosine ( Fig. 4B, C ). At high cell density, cells increased synthesis of VLC dihydroceramide/ceramide and decreased synthesis of LC dihydroceramide/ceramide. Both labeling experiments with C 17 dihydrosphingosine (following de novo synthesis) and with C 17 sphingosine (following recycling) revealed similar trends with respect to dihydroceramide/ceramide synthesis chain length, suggesting that this is most likely due to differences in CerS activities. Different CerS are responsible for synthesis of dihydroceramides/ceramides with different fatty acid chain lengths ( 9 ). This conclusion agrees with the results in Fig. 3 and supplementary Fig. VI: CerS2 and CerS4, responsible for the synthesis of VLC ceramides, had increased message levels at high cell density.
The increased synthesis of VLC C 17 dihydroceramides did not result in increased levels of VLC C 17 ceramides at high cell density when cells were labeled with C 17 dihydrosphingosine ( Fig. 7G, H ). In addition, at high cell density, LC C 17 dihydroceramides did not convert as effi ciently to LC C 17 ceramides as they did at low cell density ( Fig. 7A, B, E, F ). These results further support the hypothesis that high cell density is in agreement with cells undergoing G0/G1 cell-cycle arrest; S1P is known to promote cell proliferation ( 33 ). Interestingly, the message levels of sphingosine kinase and S1P phosphatase (data not shown) and steady-state sphingosine and S1P (supplementary Fig. V) did not change with cell density.

At high cell density, neuroblastoma cells had a metabolic "bottleneck" at the level of dihydroceramide desaturase
Next, we metabolically labeled cells with C 17 dihydrosphingosine to follow de novo ceramide synthesis at different population densities. SMS-KCNR neuroblastoma cells were cultured at low and high cell density for 48 h as described in Materials and Methods, and then labeled with 1 µM C 17 dihydrosphingosine for 30 min. Lipids were extracted from the cell pellets, and C 17 sphingolipids were measured by MS. The results showed that, at high cell density, neuroblastoma cells had increased synthesis of C 17 dihydroceramide ( Fig. 6A ). Interestingly, at high cell density, C 17 dihydroceramide could not be converted to C 17 ceramide as effi ciently as it was at low cell density ( Fig. 6B ). These data support results obtained by measuring steadystate sphingolipids ( Fig. 2A, B ) and suggest that with increased cell density the ratio of dihydroceramide/ceramide also increased, suggesting a metabolic "bottleneck" at the level of dihydroceramide desaturase when cells were densely populated. C 17 dihydrosphingosine, similar to C 17 sphingosine, can also be converted to C 17 dihydrosphingosine-1-phosphate. compared with low cell density, confi rming our hypothesis that dihydroceramide desaturase has lower activity in densely populated cells compared with sparsely populated cells.
Next, we tested the in vitro dihydroceramide desaturase activity in total cell homogenates of cells grown at low and high cell densities. The results from the in vitro activity assay ( Fig. 8B ) were similar to the activity measured in situ ( Fig. 8A ), further confi rming that dihydroceramide desaturase had signifi cantly reduced activity when cells were cultured at high population density compared with low population density. Interestingly, dihydroceramide desaturase message and protein did not change based on cell density ( Fig. 8C, D ).
Dihydroceramide desaturation is a complex reaction, which next to DEGS-1 activity, involves the activities of cytochrome b5 reductase and cytochrome b5 and requires NAD(P)H as an electron donor (supplementary Fig. VII-A) ( 11,28,34 ). To probe the mechanism by which dihydroceramide desaturase activity is regulated under high cell density, we tested whether there was a difference in the protein levels of cytochrome b5 reductase and cytochrome b5 at different cell densities in neuroblastoma cells. Western blot analyses revealed that, as in the case of DEGS-1 protein, the protein levels of cytochrome b5 reductase and cytochrome b5 did not change with cell density (supplementary Fig. VII), thus suggesting indirect inhibition of the enzyme.

Reducing agents in the media led to decrease of dihydroceramide desaturase activity
To further probe the mechanism by which dihydroceramide desaturase is reduced under high cell density conditions, we tested whether a factor or factors secreted in the media are responsible for the reduced activity of the enzyme by applying conditioned media from cells cultured at high cell density to cells cultured at low cell density. Conditioned media was applied to low-density cells for 7 h before in situ dihydroceramide desaturase activity assay was performed (see Materials and Methods). The results in Fig. 9A show that adding conditioned media to low-density cells resulted in partial reduction of dihydroceramide desaturase activity.
Dihydroceramide desaturase was shown to be sensitive to both oxidation [addition of H 2 O 2 to the medium ( 21 )] and addition of a reducing agent, DTT, in the in vitro reaction ( 35 ). Next, we tested whether there were differences in the redox state of the conditioned media from cells cultured at low and high cell densities. Ellman's test was used to measure the levels of free thiols in the media. The results in Fig. 9B show that the levels of free thiols were higher in the conditioned media from cells cultured at high cell density and suggested that a reducing agent from the media could be responsible for the decreased desaturase activity at high cell density. To test that, we added two different reducing agents, DTT, which was previously described to inhibit dihydroceramide desaturase in vitro ( 35 ), and L -cysteine, at two different concentrations to cells cultured at low cell density, and then compared the in situ dihydroceramide desaturase activity to untreated dihydroceramide desaturase has lower activity when cells are densely populated.

Dihydroceramide desaturase activity is reduced at high cell density
To test the hypothesis that dihydroceramide desaturase has lower activity in densely populated cells compared with sparsely populated cells, in situ and in vitro dihydroceramide desaturase activity assays were performed with cells cultured at different population densities. First, we assessed the in situ activity by using an assay specifi cally developed for dihydroceramide synthase ( 10 ). This in situ assay utilizes a water-soluble nonnatural analog of the substrate dihydroceramide, C 12 -dhCCPS. SMS-KCNR neuroblastoma cells were grown at low and high cell density for 48 h as described above. C 12 -dhCCPS (500 nM) was added to the cells for 30 min. Subsequently, lipids were extracted from the cell pellets, and accumulation of the nonnatural product of the reaction C 12 -CCPS was measured by MS ( 10 ). As shown in Fig. 8A , the in situ activity of dihydroceramide desaturase was four times lower at high cell density controls. The results in Fig. 9C confi rmed our hypothesis that reducing agents in the media can signifi cantly inhibit dihydroceramide desaturase activity. Collectively, the results in Fig. 9 suggest that the more reduced state of the medium   5. At high cell density, the neuroblastoma cells signifi cantly decrease sphingosine-1 phosphate synthesis. Cells were cultured, labeled, and lipids were extracted and measured as in Fig. 4 . C 17 sphingosine-1-phosphate levels were normalized to cellular lipid phosphate. Error bars represent the results from three independent experiments. from cells cultured at high cell density is the likely cause for the decreased dihydroceramide desaturase activity.

DISCUSSION
In the current study, we applied a metabolic approach to investigate the role of sphingolipids in confl uenceinduced growth arrest in neuroblastoma cells. Our data showed that, at saturation density, cells underwent complex changes in their sphingolipid metabolism, increasing synthesis of VLC sphingolipids, decreasing synthesis of S1P, and signifi cantly reducing the activity of dihydroceramide desaturase, which resulted in increased ratio of dihydroceramide/ceramide. Moreover, our results showed that reducing agents in the media can decrease dihydroceramide desaturase activity. In addition to suggesting a role for sphingolipids in contact inhibition in neuroblastoma cells, our results revealed an important technical aspect for interpreting sphingolipid measurements, namely, that cell density needs to be considered when comparing sphingolipids in two cell culture samples. cell densities in neuroblastoma cells and pinpointed decreased desaturase activity at high cell density that was not due to decreased message levels of DEGS-1 ( Fig. 8C ).
Current results, which implicate dihydroceramide desaturase-reduced activity and the increased ratio of dihydroceramide/ceramide in G0/G1 cell-cycle arrest at confl uence in neuroblastoma cells, corroborate our previous fi ndings that inhibition by siRNA of DEGS-1, the major desaturase in the SMS-KCNR neuroblastoma cells, leads to G0/G1 cell-cycle arrest ( 10 ) and further confi rm a role for dihydroceramide desaturase-reduced Sphingolipid pathway is a complex interdependent pathway, and our results demonstrated that analyzing only one aspect of it could be misleading. For example, steady-state levels of sphingolipids at low and high cell densities ( Fig. 2 ) could not be explained solely by looking at the message levels of sphingolipid metabolic enzymes ( Fig. 3 ) because changes in message levels of an enzyme do not always result in changes in activity. Our approach, which also included metabolic-labeling experiments ( Figs. 4, 5, 6, and 7 ), further characterized the changes in the sphingolipid pathway at low and high Metabolic labeling with C 17 dihydrosphingosine of neuroblastoma cells cultured at different cell densities (total). SMS-KCNR cells were cultured at low (50%) and high (90%) cell densities. After 48 h, cells were labeled with 1 µM C 17 dihydrosphingosine for 30 min. Lipids were extracted, and C 17 sphingolipid species were measured by LC/MS analyses. C 17 sphingolipid levels were normalized to cellular lipid phosphate. Error bars represent the results from three independent experiments. (A) Total C 17 dihydroceramide (the data used for calculating the total C 17 dihydroceramide is shown in supplementary Table IV). (B) total C 17 ceramide (the data used for calculating the total C 17 ceramide is shown in supplementary Table V). (C) C 17 dihydrosphingosine-1phosphate. Alternatively, there could be additional contributing factors for changing dihydroceramide desaturase activity. In a study showing the inhibitory effect of celecoxib on dihydroceramide desaturase, it has been speculated that membrane rigidity is the indirect factor contributing to the inhibition. It will be interesting to explore in future studies whether membrane rigidity can inhibit dihydroceramide desaturase activity and whether cells increase membrane rigidity at high cell density. Notably, in a study addressing HIV infection, it has been shown that dihydrosphingomyelin increases membrane rigidity ( 38 ). Of note is that in vitro studies have shown that dihydroceramides do not support transbilayer fl ip/fl op of lipids, whereas ceramides do ( 39 ). Perhaps dihydroceramide desaturase-reduced activity and an increased dihydroceramide/ceramide ratio at high density serve as a positivefeedback mechanism to increase membrane rigidity and reduce transbilayer lipid movement. Whether this contributes to the growth arrest or to the reinforcement of cell-to-cell contacts is a question to be addressed in future studies. activity in cell-cycle progression. Interestingly, the confl uence-induced reduction of dihydroceramide desaturase activity was not due to reduced DEGS-1 message (as mentioned above) and protein levels or to the protein levels of cytochrome b5 reductase and cytochrome b5, the other two enzymes participating in the reaction ( Fig. 8 and supplementary Fig. VII). In a study in COS-7 cells, it has been shown that myristoylation targets DEGS-1 to the mitochondria, increasing the conversion of dihydroceramide to ceramide ( 36 ). Subcellular fractionation of neuroblastoma cells grown at high and low densities did not show differences in DEGS-1 subcellular distribution (supplementary Fig. VIII), excluding changes in DEGS-1 localization as a possible mechanism for reduced activity. Rather, our results in Fig. 9 suggest that reducing agents, like free thiols, in the media of high-density cells were responsible for the decreased activity of dihydroceramide desaturase. In support of our fi ndings in neuroblastoma cells, a study in cultured melanoma cells have shown that, at high density, cultures rapidly accumulate free thiols in the media ( 37 ). The results from the metabolic-labeling experiments with C 17 sphingosine ( Fig. 4 ) and C 17 dihydrosphingosine ( Fig. 7 ) revealed that the LC sphingolipids are preferentially synthesized at low cell density, whereas the VLC ones are preferentially synthesized at high cell density. The difference in synthesis did not result in differences in steady-state levels of LC versus VLC sphingolipids at low and high cell densities. The metabolic "bottleneck" at the level of dihydroceramide desaturase could be an explanation as well as an increased selective degradation of VLC sphingolipids at high cell density. As shown in Fig.  3A and supplementary Fig. VI, the message levels of alkaline ceramidase 1 and neutral sphingomyelinase 2 were increased at confl uence, which could result in increased activity. Alkaline ceramidase 1 has been shown to have selectivity toward VLC sphingolipids ( 40 ). In addition, in another study using breast cancer cells, investigators reported that, at confl uence, neutral sphingomyelinase 2 preferentially hydrolyses VLC sphingomyelin ( 7 ). The second study with breast cancer cells revealed a preferential increase of VLC ceramide at confl uence, which our data did not confi rm in neuroblastoma cells. Such differences between the two cell lines could arise from different alkaline ceramidase 1 activity at high cell density.
In addition, we have shown that in neuroblastoma cells, cell density affected the expression levels of multiple enzymes of the sphingolipid metabolic pathway ( Fig. 3A and  supplementary Fig. VI), not just the activity of dihydroceramide desaturase. Taken together, our results suggest that cell density affects sphingolipid metabolism in a complex way, which is cell-type specifi c, and only a comprehensive metabolic approach can characterize it.
An example of such complexity is the discrepancy between S1P steady-state levels and S1P synthesis at different cell densities in neuroblastoma cells. Although the steady-state levels of S1P did not change between sparsely and densely populated cells (supplementary Fig. V-B), our metabolic-labeling experiments showed that the synthesis of S1P ( Fig. 5 ) or dihydrosphingosine-1-phosphate ( Fig. 6C ) was dramatically decreased in densely populated cells. The elevated S1P lyase message suggested possible activation of the enzyme, which could be an explanation of the faster degradation of de novo synthesized S1P or dhihydrosphingosine-1-phosphate. S1P is known to promote cell proliferation , whereas ceramide has been shown to promote cell death and growth inhibition, which leads to emergence of the S1P/ceramide rheostat concept as a regulator of cell growth ( 41 ). Our sphingolipid analysis in neuroblastoma cells revealed similar trends. At high cell density, cells underwent growth arrest, reduced their S1P synthesis, and increased dihydroceramide due to dihydroceramide desaturase-reduced activity. Therefore, it is possible that a "contact inhibition-specifi c rheostat" that depends on S1P synthesis and the dihydroceramide/ ceramide ratio is involved in regulating cell growth in neuroblastoma cells.
The authors thank the Lipidomics Core Facility at MUSC for lipid measurements, the Flow Cytometry Facility at MUSC for fl ow cytometry measurements, and Drs. Cungui Mao and Russell Jenkins for RT-PCR primers. Fig. 9. Reducing factors in the medium caused decrease in dihydroceramide desaturase activity. (A) SMS-KCNR neuroblastoma cells were cultured at low (50%) and high (90%) cell densities for 48 h. During the last 7 h of the experiment, in a subset of the samples cultured at low cell density, the growth media was replaced with conditioned media derived from cells cultured at high cell density. The in situ dihydroceramide desaturase activity was measured by adding 500 nM C 12 -dhCCPS (a dihydroceramide desaturase substrate) to the cells for 30 min. The product of the dihydroceramide desaturase reaction, C 12 -CCPS, was extracted from the collected cell pellets and quantifi ed by MS. Results were normalized to the cellular lipid phosphate. Error bars represent the results from three independent experiments. (B) A standard "Ellman's test" was used for quantitation of the free thiols in the medium of SMS-KCNR cells cultures at low and high cell densities for 48 h. Error bars represent the standard deviation from six independent samples (two experiments). (C) SMS-KCNR cells were cultured at low cell density for 24 h. Reducing agents, DTT or L -cysteine, were added to the media and the cells were cultured for additional 24 h. The in situ dihydroceramide desaturase activity was measured as describe above (see part A). Error bars represent range of two independent experiments.