C6-ceramide nanoliposome suppresses tumor metastasis by eliciting PI3K and PKCζ tumor-suppressive activities and regulating integrin affinity modulation

Nanoliposomal formulation of C6-ceramide, a proapoptotic sphingolipid metabolite, presents an effective way to treat malignant tumor. Here, we provide evidence that acute treatment (30 min) of melanoma and breast cancer cells with nanoliposomal C6-ceramide (NaL-C6) may suppress cell migration without inducing cell death. By employing a novel flow migration assay, we demonstrated that NaL-C6 decreased tumor extravasation under shear conditions. Compared with ghost nanoliposome, NaL-C6 triggered phosphorylation of PI3K and PKCζ and dephosphorylation of PKCα. Concomitantly, activated PKCζ translocated into cell membrane. siRNA knockdown or pharmacological inhibition of PKCζ or PI3K rescued NaL-C6-mediated suppression of tumor migration. By inducing dephosphorylation of paxillin, PKCζ was responsible for NaL-C6-mediated stress fiber depolymerization and focal adhesion disassembly in the metastatic tumor cells. PKCζ and PI3K regulated cell shear-resistant adhesion in a way that required integrin αvβ3 affinity modulation. In conclusion, we identified a novel role of acute nanoliposomal ceramide treatment in reducing integrin affinity and inhibiting melanoma metastasis by conferring PI3K and PKCζ tumor-suppressive activities.

ligand sialyl-Lews a/x at sufficient levels to mediate tethering and rolling of tumor cells 16 . Therefore, they hijack polymorphonuclear neutrophils (PMNs) or fibrin to bridge them into close proximity to the endothelial cells, thereby facilitating their subsequent migration through endothelial cells [17][18][19][20][21] . Like melanoma cells, metastatic breast cancer cell, MDA-MB-231 is negative for sialofucosylated selectin ligands and integrin b 1 and b 2 integrins, like lymphocyte function-associated antigen-1 (LFA-1), Mac-1 and very late antigen-4 (VLA-4). Therefore, they were deficient in binding to endothelial intercellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule-1 (VCAM-1). Integrin a v b 3 was found to be expressed on MDA-MB-231 by other groups 22,23 and ourselves. Integrin a v b 3 plays important roles in breast cancer metastasis 24,25 . Integrin a v b 3 can form bonds with fibrinogen which served as connecting ligands facilitating melanoma and breast cancer adhesion to endothelium in flow 18,25,26 .
To evaluate the effect of short-term nanoliposomal ceramide treatment on breast cancer and melanoma migration, we studied MDA-MB-231 and Lu1205 cell static and flow migration potencies in response to 30 min NaL-C 6 treatment. NaL-C 6 attenuated tumor migration in a dose-dependent manner. By using mutant constructs, pharmacological inhibitors and short interference RNA (siRNA) knockdown, we discovered that NaL-C 6 -mediated PKCf and PI3K phosphorylation and PKCa dephosphorylation were responsible for reduced cell migration. As the activation of PKC isoforms and PI3K were conventionally conceived of augmenting malignancy of tumors, we uncovered a novel role of PKCf and PI3K as tumor suppressors. The strategies of activating PKCf might potentiate the therapeutic effect of nanoliposomal ceramide to treat tumor metastasis.

Results
Acute treatment with C 6 nanoliposome suppressed tumor migration. In previous studies, it was demonstrated that C 6 nanoliposomes at high dosage range and long exposure duration mediated cancer apoptosis and growth arrests 2,4-7 . But it remains elusive whether acute treatment of cancer cells with C 6 nanoliposomes at low dosage range had any impacts on cell phenotypes. We measured MDA-MB-231 and Lu1205 apoptosis after being treated with a variety of doses of NaL-C 6 for 30 min and 12 hr 3,27 . Upon 30 min 20 mM NaL-C 6 incubation, only 4% MDA-MB-231 and 1% Lu1205 cells underwent apoptosis (Fig. 1a). In contrast, 20 mM NaL-C 6 with a long 12 hr of exposure resulted in 35% MDA-MB-231 and 20% Lu1205 apoptosis.
Next, we assessed 4-hr transwell migration of MDA-MB-231 and Lu1205 cells which received nonliposomal (NoL-C 6 ) or liposomal C 6 treatment for 30 min. Liposomal C 6 were more effective to suppress MDA-MB-231 and Lu1205 cell migration than freely administrated C 6 (Fig. 1b). This disparity of nonliposomal and liposomal C 6 effects may be caused by the difference in the resultant accumulation of C 6 within 30 min timescale 2 . Collagen IV has been reported to function as chemoattractant for melanoma migration and enhance breast cancer motility 28 . In the absence of collagen IV, few MDA-MB-231 or Lu1205 cells migrated to the opposite side of the membrane (28 6 5 RFU and 21 6 3 RFU), while 100 mg/ml collagen IV in the bottom well strikingly augmented ghost nanoliposome-treated MDA-MB-231 or Lu1205 cell migration (379 6 47 RFU and 278 6 8 RFU) (Fig. 1c). NaL-C 6 suppressed MDA-MB-231 cell migration in a dose-dependent manner. 5 , 20 mM NaL-C 6 significantly attenuated MDA-MB-231 or Lu1205 transmigration as compared with ghost nanoliposome (p , 0.05).
Flow-regulated cancer migration plays important roles in tumor metastasis 14 . To evaluate the effect of nanoliposomal ceramide treatment on cancer migration under hydrodynamic conditions, we utilized a flow migration device which consists of a modified 48-well Boyden chamber and a flow loop 19,21,29 . To facilitate successful tumor extravasation in flow, a stable adhesion mediated by integrin is required. However, a screening of surface expressions of adhesive molecules revealed that neither MDA-MB-231 nor Lu1205 expressed integrins LFA-1 and Mac-1, the counter-receptors for endothelial ICAM-1 (Table 1) 14,30,31 .These two cell lines express integrin a v b 3 which is a receptor for plasma protein, fibrinogen. Previous studies suggested that fibrinogen serving as a linker for cell-cell adhesion supported integrin a v b 3 -dependent adhesion of melanoma cells to endothelium under flow conditions 18,25,26 . Therefore, in the current flow migration settings, the transendothelial migratory properties of cancer cells were analyzed in the presence of fibrinogen at physiological concentration of 1.5 mg/ml. As expected, fibrinogen promoted the transmigration of substantial amounts of MDA-MB-231 cells at shear stresses of 2 and 4 dyn/ cm 2 (512 6 30 and 235 6 39 cells/0.48 mm 2 of filter) (Fig. 1d). At both shear stresses, NaL-C 6 dramatically attenuated MDA-MD-231 and Lu1205 migration compared with ghost nanoliposome in a dosedependent manner (p , 0.05) (Fig. 1d). At high shear stress (4 dyn/ cm 2 ), 20 mM NaL-C 6 for 30 min exposure resulted in 8.7 and 14.1 fold reduction of MDA-MB-231 and Lu1205 migration. The data indicated that acute treatment with NaL-C 6 (30 min) suppressed shear-dependent tumor migration.
Nanoliposomal C 6 triggered phosphorylation of PKCf and PI3K and translocation of PKCf into cell membrane. Ceramide is sphingolipid metabolite and accumulates in cell membrane lipid bilayer upon elevated sphingomyelinase activity or de novo synthesis 1,27,32 . To assess the effect of acute NaL-C 6 treatment on PKC isoform activation in MDA-MB-231 and Lu1205 cells, we determined the phosphorylation states of three PKC isoforms, PKCa, PKCe and PKCf, which were previously shown to control cancer metastasis 33,34 . Phosphorylation of PKCf threonine 410 residue and PKCa threonine 638 residue in activation loop domain and serine 729 in hydrophobic motifs of PKCe contribute to activation as well as stability of enzymes 10,35 . In MDA-MB-231 and Lu1205 cells, PKCa and PKCe were constitutively phosphorylated at residues Thr638 and Ser729, respectively, while PKCf was marginally phosphorylated at Thr410 (Fig. 2a-c). NaL-C 6 treatment reduced PKCa phosphorylation but enhanced PKCf phosphorylation with a maximum effect at 20 mM in both cell lines ( Fig. 2a and c). In contrast, PKCe phosphorylation level was not changed by NaL-C 6 treatment (Fig. 2b).
Previous studies suggested that phosphorylated PKC isoforms translocate to cell membrane to regulate cell behaviors 11 . Therefore, we evaluated the subcellular distribution of three PKC isoforms in response to nanoliposome treatment. Compared with ghost nanoliposome, NaL-C 6 reduced the amounts of phosphorylated and total PKCa in membrane fraction of tumor cells (Fig. 2d). Total PKCa translocated into cytosolic fraction. In addition, phosphorylated membrane PKCa was decreased with increasing doses of NaL-C 6 treatment. The distributions of both phosphorylated and total PKCe were not affected by NaL-C 6 treatment (Fig. 2e). As a positive control, 12-O-tetradecanoylphorbol-13-acetate (TPA) induced a membrane translocation of PKCa and PKCe. On the other hand, acute NaL-C 6 treatment resulted in translocation of total PKCf from cytosol to cell membrane (Fig. 2f). Concomitantly, phosphorylation levels of PKCf was increased in both cytosol and cell membrane. 20 mM NaL-C 6 treatment resulted in a peak PKCf phosphorylation in the cell membrane.
Since the phosphorylation of PKC is usually associated with PI3K activation, we next determined the phosphorylation states of PI3K in MDA-MB-231 and Lu1205 cells 35,36 . 30 min NaL-C 6 exposure elevated the phosphorylation levels of PI3K in a dose-dependent manner (Fig. 2g). A peaked level of PI3K phosphorylation was observed as a consequence of 20 mM NaL-C 6 treatment. Thus, the activity of PI3K paralleled that of PKCf in response to NaL-C 6 treatment (Fig. 2a).  cells did not undergo apoptosis after receiving acute treatment (30 min) of 5 mM, 10 mM or 20 mM ceramide nanoliposome.12-hr treatment with ceramide nanoliposomes induced tumor cell death. % of cells undergoing apoptosis was analyzed with Annexin-v/7-ADD staining and flow cytometry. Values were mean 6 SEM. n 5 three replicates. **p , 0.01 compared with ghost control. (b-c) Liposomal C 6 delivery augmented the anti-migration activity of C 6 -ceramide. MDA-MB-231 or Lu1205 cells received 1 mM, 5 mM, 10 mM or 20 mM liposomal (NaL-C 6 ) or nonliposomal C 6 -ceramide (NoL-C 6 ) (b) or ghost nanoliposome (c) for 30 min before being used for transwell migration assays. The amounts of migrated cells were determined by calcein AM staining after 4-hr onset of experiment and were expressed as RFU. 100 mg/ml collagen IV was used as chemoattractant in the bottom well. In no collagen IV group, DMEM 1 0.1% BSA was added into chemoattractant wells instead. Results were expressed as mean 6 SEM. n 5 three replicates.*p , 0.05, **p , 0.01 compared with control for each cell type.(d) NaL-C 6 attenuated MDA-MB-231 or Lu1205 cell transendothelial migration in a dosedependent manner under flow conditions. Tumor cells were incubated with 1 mM, 5 mM, 10 mM or 20 mM NaL-C 6 or ghost nanoliposome for 30 min before being introduced into flow migration chamber together with 1.5 mg/ml fibrinogen. The flow migration assay was carried out for 4 hr at shear stress of 2 or 4 dyn/cm 2 . The migrated cancer cells were stained and counted at the bottom of filter. Results were expressed as mean 6 SEM. n 5 three replicates.*p , 0.05, **p , 0.01compared with ghost at the same shear stress. PKCf was essential for nanoliposomal C 6 -regulated tumor migration. Since NaL-C 6 induced PKCf phosphorylation and accumulation of phosphorylated PKCf in cell membrane, we hypothesized that PKCf might play a role in NaL-C 6 -induced suppression of cancer migration. To verify this, we transfected MDA-MB-231 cells with full length PKCf (PKCf FL) construct and dominantly negative PKCf(PKCf DN) construct which is kinase-defective and contains a point mutation in its kinase domain to assess PKCf function in cell migration. A 95% transfection efficiency was achieved at the time of functional assays as measured by fluorescently staining HA (constructs were tagged with hemagglutinin (HA)) ( Fig. 3a). In the presence of ghost nanoliposomes, PKCf FL and PKCf DN overexpression had no effect on cell static transwell migration (Fig. 3b). In sharp contrast, PKCf FL potentiated suppressive effect of 20 mM nanoliposomal ceramide on MDA-MB-231 transmigration, while PKCf DN rescued MDA-MB-231 transmigration suppressed by NaL-C 6 (p , 0.05).
Motility is a required process for invasion of tumor cells through the surrounding stroma. To determine whether PKCf activation was required for regulation of cell motility, wound healing assays were conducted with mutant construct-transfected MDA-MB-231 cells. In consistent with transwell migration assays, PKCf FL or PKCf DN overexpression had no effect on wound healing capability of ghost nanoliposome-treated cells (Fig. 3c). 20 mM NaL-C 6 treatment for 30 min increased the size of initial wounded area. This suggests that NaL-C 6 treatment resulted in actomyosin-mediated contraction of cells. PKCf FL overexpression further reduced the wound healing capacities of nanoliposomal ceramide-treated cells (p , 0.01). Conversely, PKCf DN-transfected cells briskly migrated into the wound area, reaching 60% sealing at 12 hr, after wound scratch (Fig. 3c). Upon TPA stimulation, the wound healing rates of cells receiving vector, PKCf FL and PKCf DN constructs were comparable.
Next, we analyzed the dynamics of actin cytoskeleton and focal adhesion, which are required to maintain cell shapes and promote cell migration. Ghost-treated cells exhibited thick stress fibers which traversed the cell body (Fig. 4a). In addition, in vector, PKCf FL and PKCf DN-transfected cells, the appearance of actin cytoskeleton had no obvious difference. In contrast, in NaL-C 6 -treated cells, filamentous actin assembled around cell periphery, with only a few thin stress fibers located within cell body. The actin morphology in cells transfected with PKCf FL was comparable to that in cells transfected with vector. However, PKCf DN overexpression restored the morphology of actin stress fibers displayed by ghost-treated cells. The stress fibers   became more robust and organized. This implied that NaL-C 6 regulate the dynamics of actin cytoskeleton via PKCf. By staining paxillin, a focal adhesion marker, we showed that in ghost-treated cells, vector, PKCf FL and PKCf DN overexpression did not lead to the difference in focal adhesion distributions (Fig. 4a). They all displayed bright punctate focal adhesions which were colocalized with the end of thick stress fibers. When cells were treated with NaL-C 6 , focal adhesion stainings became dim. Small focal adhesions were visible at cell periphery and they almost disengaged with thin stress fibers. PKCf FL transfection did not change the appearance of focal adhesions. On the contrary, PKCf DN overexpression reverted the loss of focal adhesion complexes and thick stress fibers induced by NaL-C 6 treatment. Focal adhesions moved from the cell periphery to cell body where they were associated with thick stress fibers. Quantitative analysis of the average focal adhesion size and number in a cell revealed that in NaL-C 6 -treated cells, PKCf FL decreased but PKCf DN increased the size and number of focal adhesions (Fig. 4b-c).  Disassembly of focal adhesion complexes is usually accompanied by the dephosphorylation of paxillin molecules. To investigate whether NaL-C 6 treatment and PKCf disturbance may regulate the phosphorylation states of paxillin, MDA-MB-231 cells were stained with anti-paxillin antibody and 4G10, an antibody against phosphotyrosine on proteins. In ghost-treated cells, 4G10 staining appeared as large dots and colocalized with paxillin in cell body, implying that paxillins were tyrosine phosphorylated (Fig. 5a). Upon NaL-C 6 treatment, paxillin and 4G10 dots became smaller and localized at cell periphery (Fig. 5b). The line-scanned fluorescence intensity profiles of paxillin were not correlated with those of 4G10, suggesting that paxillin was dephosphorylated. While PKCf FL transfection did not alter the phosphorylation states of paxillin, PKCf DN overexpression resulted in the presence of larger 4G10stained dots in cell body where they were colocalized with focal adhesions. The line-scanned fluorescence intensity profiles of paxillin were in phase with those of 4G10 ( Fig. 5c-d). This data implied that NaL-C 6 regulated focal adhesion disassembly and paxillin dephosphorylation in a PKCf-dependent manner.
At a shear stress of 4 dyn/cm 2 , NaL-C 6 administration abrogated cell migration ability with only a small number of cells migrating through the human umbilical vascular endothelial cell (HUVEC) monolayer ( Fig. 6a middle vs left). PKCf DN eliminated the inhibitory effect of NaL-C 6 on cell migration ( Fig. 6a right vs middle). At shear stresses of 2 and 4 dyn/cm 2 , PKCf DN overexpression rescued NaL-C 6 -suppressed cell extravasation (p , 0.01). To further determine the role of PKCf in regulating cell migration, we analyzed PKCf phosphorylation levels in cells remaining in circulation and undergoing transmigration after 4-hr flow migration assays. As shown in Fig. 6c, nanoliposomal C 6 -ceramide upregulated PKCf phosphorylation levels in flowing cells in a dose-dependent manner. In contrast, PKCf phosphorylation in migrated cells was unresponsive to up to 20 mM NaL-C 6 treatment. These results implied that PKCf participated in NaL-C 6 -regulated cell migration in flow.
PI3K inhibition rescued tumor migration suppressed by acute C 6 nanoliposome treatment. To determine the role of PI3K in NaL-C 6induced suppression of cancer migration, we assessed transwell migration of MDA-MB-231 cells which received PI3K inhibitors, wortmannin and LY294002. Wortmannin and LY294002 significantly attenuated the transmigration of ghost-treated cells (Fig. 7a). Nevertheless, NaL-C 6 -treated cells receiving wortmannin or LY294002 migrated more vigorously than those receiving DMSO. Wortmannin and LY294002 increased the amounts of migrated cells by 29% and 64%, respectively. In addition, wortmannin and LY294002 promoted the sealing of the wound for NaL-C 6 -treated cells (Fig. 7b). Of note, wortmannin which is a more potent inhibitor for PI3K had a larger impact on cell motility than LY294002. Wortmannin exposure completely reverted the cell contraction phenotype induced by NaL-C 6 treatment and promoted cell longitudinal migration. Wortmannin or LY294002 treatment significantly increased NaL-C 6 -suppressed MDA-MB-231 migration at shear stresses of 2 and 4 dyn/cm 2 (Fig. 7c). At a shear stress of 4 dyn/ cm 2 , PI3K phosphorylation levels were boosted by NaL-C 6 treatment in flowing cells but not migrated cells (Fig. 7d).This suggested that PI3K was critical for ceramide nanoliposomeregulated cell migration.
PI3K-regulated PKCf phosphorylation and PKCa dephosphorylation were crucial for NaL-C 6 -suppressed cancer migration. To further verify the role of PI3K and PKCf in regulating cell migration, MDA-MB-231 and Lu1205 cells were transfected with PI3K or PKCf siRNA. PI3K and PKCf siRNA effectively knocked down target gene expressions (Fig. 8a). PI3K knockdown reduced the phosphorylation level of PKCf and elevated phosphorylation level of PKCa in both NaL-C 6 -treated MDA-MB-231 and Lu1205 cells. This may suggest that PI3K was activated upstream of PKCf and PKCa in response to acute NaL-C 6 treatment.
PI3K or PKCf knockdown rescued MDA-MB-231 and Lu1205 transwell migration, wound healing and flow migration potencies ( Fig. 8b-d). At 4 dyn/cm 2 , PI3K and PKCf knockdown increased MDA-MB-231 cell migration by 4.7-and 4.3-fold, respectively. PKCa is known to regulate myosine light phosphorylation and actin contraction, thereby promoting cell migration. Since PKCa phosphorylation was downregulated in response to NaL-C 6 treatment, we hypothesized that PKCa deactivation was also required for suppression of cell migration. To test the hypothesis, we transfected cells with constitutively active PKCa (PKCa CAT). PKCa CAT overexpression enhanced MDA-MB-231 and Lu1205 transwell migration and flow migration capacities (Fig. 8e-f).
Integrin a v b 3 affinity modulation was involved in PKCf and PI3K-dependent suppression of migration by NaL-C 6 . Cell attachment to endothelium and development of shear-resistant bonds were critical for tumor extravasation in flow. To assess whether the suppression of the transmigration in flow by NaL-C 6 was caused by disruption of cell firm adhesion, we employed cell detachment assay. To conduct this assay, fibrinogen was coated as substrate in petri dish prior to parallel plate chamber assembly. MDA-MB-231 cells treated with NaL-C 6 for 30 min were allowed to settle onto the coated fibrinogen before step-load shears were exerted. 20 mM NaL-C 6 exposure reduced the number of bound cells with increasing shear rate (0 , 1600 sec 21 ) (Fig. 9a). In contrast to PKCaDN and PKCe DN, PKCf DN transfection rescued the suppression of cell adhesion by NaL-C 6 at each shear rate (Fig. 9a).
Since MDA-MB-231 and Lu1205 cells express integrin a v b 3 which can form bonds with fibrinogen and promote tumor firm adhesion in flow, we next investigated the effect of integrin a v b 3 siRNA knockdown on cell shear-resistant adhesion. Compared with scrambled siRNA control, integrin a v b 3 siRNA knockdown significantly reduced cell adhesion (NaL-C 6 1vec1scr vs NaL-C 6 1vec1siR) (Fig. 9b). In the cells which were depleted of integrin a v b 3 with siRNA, PKCf DN transfection failed to restore cell adhesion ability suppressed by NaL-C 6 (NaL-C 6 1vec1siR vs NaL-C 6 1PKCf DN1siR). PI3K inhibition by wortmannin considerably increased the amounts of bound cells compared with DMSO (Fig. 9c). Integrin a v b 3 knockdown abrogated the effect of wortmannin on cell adhesion.
Integrin affinity is regulated by divalent cations. Addition of Mn 21 or removal of Ca 21 results in increased ligand-binding affinity and adhesiveness of a v b 3 integrin 37,38 . WOW-1 antibody which specifically recognizes a v b 3 activation-epitope was used to probe integrin binding affinity modulated by ions 25 . In the presence of Mn 21 , WOW-1 binding increased by 2.5-fold, while in the presence of Ca 21 , WOW-1 binding decreased by 3-fold (Fig. 9d). Mn 21 restored MDA-MB-231 shear-resistant adhesion suppressed by NaL-C 6 . On the other hand, Ca 21 addition significantly reduced the adhesion of PKCf DN-transfected and NaL-C 6 -treated cells. These results implied that affinity modulation of integrin a v b 3 was required for PI3K and PKCf-dependent cell adhesion weakening mediated by acute NaL-C 6 treatment.

Discussion
Cancer metastasis is highly coordinated, multistep process, involving tumor undergoing epithelial-mesenchymal transition, traveling in blood stream, lodging onto vascular endothelium and extravasation 14 . By using wound healing, transwell migration and flow migration assays, we revealed that short-term C 6 nanoliposome treatment suppressed melanoma and breast cancer migration without inducing cell apoptosis. In addition, we found that NaL-C 6    www.nature.com/scientificreports    www.nature.com/scientificreports hand, it reduces the affinity of integrin a v b 3 , thereby weakening cell adhesion in flow. The inhibitory activity of PKCf and PI3K in NaL-C 6 -treated cells distinguish them from pro-mitogenic and promigratory activities ascribed to conventional and novel PKC isoforms and PI3K/Akt axis.
In the current study, C 6 -ceramide was delivered in a nanoliposomal form which presents as an effective way of reducing the hydrophobicity of the ceramide and increasing its membrane transport 2 . Upon liposomal administration, ceramide is likely to be inserted into membrane lipid bilayer and localized to structured microdomain where it can be associated with signaling proteins and initiate intracellular signaling cascades 3 . In the current study, we found that in contrast to nonliposomal ceramide, liposomal ceramide more effectively suppressed tumor migration. The roles of sphingolipid metabolites in regulating cell migration remain elusive. Long-chain ceramide C 16 was reported to enhance mouse embryonic stem cell migration in a dose-and time-dependent manner 39 . On the other hand, sphingosine-1-phosphate which can be converted to ceramide via sphingosine-1-phosphate phosphatase and ceramide synthase inhibited chemotactic motility of breast cancer cells 40 . Furthermore, C 2 ceramide was reported to suppress cancer invasiveness through downregulating MMP-2 expression 41 . In the current study, the migration-suppressive effect of nanoliposomal C 6 was a consequence of activating of PKCf and PI3K. It was likely that shortand long-term ceramide treatments orchestrated different signaling pathways to mediate migratory inhibition and apoptosis. Long-term ceramide treatment initiated a pro-apoptotic pathway involving inactivation of Akt 2 , while acute ceramide treatment recruit and activate PI3K. In the current study, we focused on the 30-min acute effect of C 6 nanoliposome. This is because upon adhesion to endothelium, tumor rapidly extravasates with a time period of less than 1 hr 8,9 . The time duration for circulating tumor to interact with nanoliposome is short, especially in face of shear force. Thus, studying tumor responses to acute, rather than prolonged, NaL-C 6 treatment may be a more accurate reflection of physiological conditions.
In agreement with previous studies, the current study showed that NaL-C 6 treatment decreased membrane localization of phosphorylated PKCa 42 . PKCa is known to regulate myosin light chain phosphorylation and actomyosin-mediated cell migration. PKCa activation was critical for focal adhesion formation and integrin-mediated cell migration 43 . Dephosphorylation of PKCa by nanoliposomal ceramide may decrease myosin light chain phosphorylation, preventing cell migration. PKCf was shown to participate in cell adhesion and migration. By using inhibitors, PKCf was suggested to promote epidermal growth factor-mediated breast cancer chemotaxis 44 . PKCf in neutrophils regulated chemoattractantinduced actin assembly, integrin-dependent adhesion and cell migration 45 . PKCf also participated in the migration of CD341 progenitor cells 46 . In the current study, PKCf suppressed tumor migration in response to NaL-C 6 treatment, implying that different stimuli may confer PKC different functionalities.
PI3K is a family of lipid kinases which transduce signals from various growth factors and cytokines by generating phospholipids 47 . In response to stimulation, PI3K is recruited to the membrane by direct interaction of its p85 subunit with tyrosine phosphate motifs on receptors. The activated p110 catalytic subunit generates phosphatidylinositol-3,4,5-triphosphate, which serves as docking sites for several signaling proteins. PI3K was suggested to participate in integrin a v b 3 -mediated melanoma migration, by inducing actin stress fiber formation and enhancing integrin a v b 3 avidity 48 . Activation of PI3K by Cdc42 and Rac1 alters actin organization, leading to increased motility and invasiveness. In the current study, activation of PI3K was negatively correlated with cell migration, suggesting an inhibitory role of PI3K on the migration of nanoliposomal ceramidetreated cells.
Temporal and spatial regulation of cytoskeletal organization and focal adhesion formation plays an essential role in cell migration. It was evident that nanoliposomal ceramide administration had an impact on both cell motility and receptor-mediated adhesion. Cell motility is driven by actin-based protrusion at cell's leading edge 49 . Previous studies showed that mutations in paxillin phosphorylation sites reduced focal adhesion formation 50 . It was also suggested that phosphorylation of paxillin promotes the association of unphosphorylated focal adhesion kinase with paxillin at newly growing focal contact sites, thereby promoting cell motility and migration 51,52 . Paxillin association with stress fiber at adhesion sites may be critical for transmitting propulsive forces and serve as traction points over which cell moves 49 . Therefore, cancer cell migration speed is a function of cell adhesion strength, as regulated by focal adhesion size and number. Highly polymerized cytoskeleton exerts tension on adhe- www.nature.com/scientificreports sion sites and allows retraction of cell, increasing the migration speed of cells. It is likely that ceramide suppresses cell motility by mediating PKCf-dependent stress fiber disassembly and dephosphorylation of paxillin.
To undergo extravasation in blood vessel, cancer cells must tether and firmly adhere to endothelium. Different molecular constituents are required for multistep cell adhesion 53 . By generating shear-resistant bonds whose dissociation constants are regulated by flow force, integrin mediates firm adhesion of cells 54,55 . Cell detachment assay revealed that silencing integrin a v b 3 compromised the abilities of PI3K inhibitors and PKCf DN to rescue cell adhesion. Therefore, it is likely that the downstream target of PI3K and PKCf is integrinmediated sustained cell adhesion. Integrins are known to exist in distinct activation states, which exhibit different affinities for ligands 37,56 . In general, integrin activation controls cell adhesion. Such control is particularly important in the vasculature, where dynamic flow physically opposes cell attachment. Cell adhesion strengthening through a v b 3 integrin activation enhances cell transmigration in flow. The affinity status of integrin a v b 3 was regulated by ''inside-out'' signaling 31 . Studies indicated that integrin a v b 3 activation status regulated breast cancer migration and metastasis 25,57 . It was suggested that PKCf played a role in integrin activation in response to chemoattractant stimulation 45 . By using integrin affinity regulating ions, we established a causal relationship between NaL-C 6 -mediated suppression of cancer metastasis and the deactivation of integrin a v b 3 . PI3K and PKCf might directly or indirectly decrease integrin a v b 3 affinity to compromise cell shear-resistant adhesion 58 .
In conclusion, acute C 6 -ceramide nanoliposome treatment resulted in a significant reduction in melanoma and breast cancer migration and metastasis. Nanoliposomal ceramide-mediated suppression of cell migration under static and flow conditions was dependent on PI3K and PKCf activation. Cytoskeletal remodeling, focal adhesion disassembly and integrin a v b 3 affinity modulation were involved in this process. Understanding the molecular mechanisms and intracellular pathways downstream of nanoliposomal ceramide may facilitate the development of therapeutic strategies to prevent tumor metastasis. PKCf siRNA (59-GACAGACGCUUGCGCCGAGAC-39) was synthesized by Invitrogen. PI3K siRNA SMARTpool (Thermo Scientific, Rockford, IL) which contained 4 pooled siRNA duplexes with ''UU'' overhang was employed to silence PI3K. MDA-MBA-231 or Lu1205 were plated in a 6-well plate and were 50-70% confluent when siRNA was introduced. Transfection was performed with 50 nM siRNA complexed with 7.5 ml RNAiMax transfection reagent. Medium was changed after 6 hrs. The transfection efficiency to knockdown PKCf or PI3K reached 80% after 48 hrs as assessed by Western blotting.

Methods
Preparation of nanoliposomal ceramide. Preparation of nanoliposomal C 6 -ceramide was previously described with a slight modification. Briefly, DSPC, DOPE, C 6 -ceramide, 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxypolyethylene glycol-2000], and PEG(750)-C 8 were mixed in chloroform at a molar ratio of 3.7551.7553.550.7050.70 and evaporated under vacuum in a water bath at 25uC. After complete removal of the solvent,a thin film formed. Saline was added to the lipid film and hydrated for 30 min to obtain a crude dispersion of liposomes containing 30% ceramide. The resulting solution was then homogenized by sonication for 10 min followed by filtration through a 100-nm polycarbonate membrane with Avanti Mini Extruder (Avanti Polar Lipids). Ghost liposomes were prepared in a similar manner without C 6 -ceramide. The size distributions of nanoliposomes were measured by dynamic light scattering. Zeta potentials of the nanoliposomes were determined to be within the range of 23 mV to 26.5 mV, ensuring a neutral charge on the nanoliposomes. Fluorescent staining. Untransfected or transfected tumor cells were grown on cover slips before experiments. Cells were incubated with ghost nanoliposome or 20 mM nanoliposomal C 6 -ceramide for 30 min. Cells were then washed twice with DPBS and fixed with 5% formaldehyde for 10 min. Cells were permeabilized with 0.3% Triton X-100 in DPBS and blocked for 30 min with 5% BSA. Subsequently, cover slips were incubated with anti-paxillin and 4G10 for 1 hr at room temperature. This was followed by staining with Alexa 488-conjugated anti-rabbit IgG and Alexa 555-conjugated anti-mouse IgG. To image actin filaments, rhodamine-phalloidin (1540; Invitrogen) was incubated with cells for 40 min. Finally, fluorescent staining was visualized with a fluorescence microscope (Olympus IX71). Colocalization of paxillin and actin or 4G10 was processed by ImageJ. Results were representative of three independent experiments.
Transwell migration. Transmigration was assessed by the ability of cells to migrate through a porous (8 mm) polycarbonate membrane of a transwell device (Corning, NY) towards a chemotactic cue (collagen IV). HUVECs were grown on the filter membrane of transwell inserts to form monolayer. MDA-MB-231 or Lu1205 cells were incubated with NoL-C 6 , ghost nanoliposome, or NaL-C 6 . Thereafter, equal numbers (2.5 3 10 5 ) of cells were plated in the inserts with HUVEC monolayer in serum-free medium containing 1% BSA. Inserts were placed in corresponding wells of a 24-well plate with 100 mg/ml collagen IV in the bottom wells. The cells were allowed to transmigrate across the porous membrane filter for 4 hr. HUVECs and unmigrated cells at the top of the membrane filter were removed by a cotton swab while the cells at the bottom of the filter were fixed in 5% formaldehyde, and stained using calcein AM (Molecular Probes, Eugene, OR). The relative fluorescence units (RFU) for each well were determined by FLx800 Fluorescence Microplate Reader (BioTek, Winooski, VT) with 485 nm excitation and 520 nm emission.
Wound healing assay. MDA-MB-231 or Lu1205 cells were cultured in 6-well plates at 1 3 10 6 cells/well as confluent monolayers. The monolayers were incubated with ghost or ceramide nanoliposomes in the absence of serum for 30 min and wounded in a line across the well with a 200-ml standard pipette tip. The wounded monolayers were then washed twice with serum-free media to remove cell debris. Photographs of a selected area of each scratch were taken at indicated time points using 40X lense Olympus IX71. Wound healing effect was determined by calculating the difference in total pixels of remaining cell-free area and initial wound area.
Flow migration assay. Tumor cell migration was determined using a modified 48-well chemotactic chamber consisting of top and bottom plate separated by a gasket. Prior to each experiment, a monolayer of HUVECs was grown on sterile polyvinylpyrrolidone-free polycarbonate filters (8 mm pore size; NeuroProbe, Gaithersburg, MD, USA) pre-coated with fibronectin (30 mg/ml) for 2 hr (Sigma-Aldrich, St Louis, MO, USA). Prior to the experiments, HUVEC monolayers were activated with TNF-a for 6 hr to allow adhesive molecule levels to reach maximum. The center 12 wells of bottom plate were filled with soluble chemoattractant type IV collagen (100 mg/ml diluted in DMEM with 0.1% BSA; BD Biosciences, San Jose, CA, USA) and surrounding control wells were filled with DMEM containing 0.1% BSA. Previous work has shown that human cancer cells migrate toward collagen IV acting as a chemoattractant 28 . The apparatus was assembled by laying the filter on the bottom plate, followed by a gasket and top plate. The chamber was primed with DMEM and 0.1% BSA to eliminate bubbles from the system. Then, cells were placed in the chamber in the presence of fibrinogen and subjected to a shear flow (2 or 4 dyn/cm 2 ) for 4 hr in an incubator. Migrated cells were stained with Protocol Brand Hema3 solution (Fisher Scientific). Cells in 12 random selected fields were counted using an inverted microscope (Diaphot 330, Nikon) with the NIH ImageJ software.
Cell detachment assay. Cell detachment assay reflects receptor affinity changes and the strength of adhesion. The petri dishes coated with 2.5 mg/ml fibrinogen were assemble in parallel-plate flow chamber and mounted on the stage of an inverted phase contrast microscope Olympus IX71. Tumor cells were transfected with empty vectors or PKC mutant constructs or treated with PI3K inhibitors. After being incubated with ghost or NaL-C 6 for 30 min, cells were perfused into flow chamber. The cell concentrations from the inlet were controlled at similar levels and the fields of view were focused on the same distance from the inlet to avoid variation of initial cell concentrations. Cells were allowed to settle on fibrinogen for 7 min, before being subject to step-load shears (0 , 1600 sec 21 ). Wall shear stress was increased every 10 sec. The percentage of cells remaining bound was determined. In selected experiments, cell detachment assays were conducted in presence or absence of 250 mM Mn 21 or 1 mM Ca 21 .
Statistical analysis. All data were obtained from at least three independent experiments and expressed by means 6 SEM. Statistical significance was determined using Student's t-test or ANOVA. Tukey's test was used in post hoc analysis for ANOVA. A probability value of p , 0.05 or p , 0.01 was considered to be statistically significant.