Fatty acid synthase causes drug resistance by inhibiting TNF-α and ceramide production.

Fatty acid synthase (FASN) is a key enzyme in the synthesis of palmitate, the precursor of major nutritional, energetic, and signaling lipids. FASN expression is upregulated in many human cancers and appears to be important for cancer cell survival. Overexpression of FASN has also been found to associate with poor prognosis and higher risk of recurrence of human cancers. Indeed, elevated FASN expression has been shown to contribute to drug resistance. However, the mechanism of FASN-mediated drug resistance is currently unknown. In this study, we show that FASN overexpression causes resistance to multiple anticancer drugs via inhibiting drug-induced ceramide production, caspase 8 activation, and apoptosis. We also show that FASN overexpression suppresses tumor necrosis factor-α production and nuclear factor-κB activation as well as drug-induced activation of neutral sphingomyelinase. Thus, TNF-α may play an important role in mediating FASN function in drug resistance.

Overexpression of FASN has been associated with poor prognosis and higher risk of recurrence of human cancers of the breast (9)(10)(11)(12) and prostate ( 13 ), as well as many other types of cancers ( 1 ), suggesting that increased FASN expression may also contribute to disease progression and treatment failure. In particular, FASN overexpression has been found in a drug-selected breast cancer cell line, and this elevation contributes to cellular resistance to Adriamycin (doxorubicin) and mitoxantrone in breast ( 14 ) and to gemcitabine in pancreatic cancer cells ( 15 ). However, the mechanism by which FASN contributes to drug resistance is currently unknown.
In this study, we investigated the molecular mechanism of FASN-mediated drug resistance using FASN overexpression stable clones of MCF7 cells and found that FASN overexpression causes resistance to multiple anticancer drugs as well as to ␥ -irradiation via inhibiting treatmentinduced ceramide production, caspase 8 activation, and apoptosis. Further studies showed that FASN overexpression suppresses tumor necrosis factor (TNF)-␣ production, nuclear factor (NF)-B activation, and drug-induced activation of neutral sphingomyelinase (nSMase). Hence, FASN overexpression may cause drug resistance by inhibiting TNF-␣ expression, which in turn suppresses activation of NF-B and nSMase and, thus, reduced ceramide production, caspase 8 activation, and apoptosis.
Abstract Fatty acid synthase (FASN) is a key enzyme in the synthesis of palmitate, the precursor of major nutritional, energetic, and signaling lipids. FASN expression is upregulated in many human cancers and appears to be important for cancer cell survival. Overexpression of FASN has also been found to associate with poor prognosis and higher risk of recurrence of human cancers. Indeed, elevated FASN expression has been shown to contribute to drug resistance. However, the mechanism of FASN-mediated drug resistance is currently unknown. In this study, we show that FASN overexpression causes resistance to multiple anticancer drugs via inhibiting drug-induced ceramide production, caspase 8 activation, and apoptosis. We also show that FASN overexpression suppresses tumor necrosis factor-␣ production and nuclear factor-B activation as well as drug-induced activation of neutral sphingomyelinase. Thus, TNF-␣ may play an important role in mediating FASN function in drug resistance. - Supplementary key words tumor necrosis factor • NK-kappa B • neutral sphingomyelinase • doxorubicin Fatty acid synthase (FASN) is a key enzyme in the biosynthesis of palmitate, a precursor of some biologically important lipids, such as phosphoinositides (for a review, see Ref. 1 ). Mammalian FASN is a homo-dimer of a multifunctional protein of ‫ف‬ 270 kDa containing seven catalytic domains: ␤ -ketoacyl synthase, malonyl/acetyltransferase, dehydrogenase, enoyl reductase, ␤ -ketoacyl reductase, acyl carrier protein, and thioesterase . Under normal physiological conditions, circulating free fatty acids (FAs) from diet are the preferred source, and de novo FA synthesis is minimally needed. Thus, the FASN expression level is low rate change of OD 340nm was corrected for the background rate of NADPH oxidation. The change in concentration of NADPH during oxidation was calculated using the following equation: where ⌬ C is change in the concentration of NADPH, ⌬ A is change in absorbance, and E is extinction coeffi cient of NADPH (E 340nm = 6.22 mM Ϫ 1 ·cm Ϫ 1 ). FASN activity was expressed as nmol NADPH oxidized/min/mg protein.

ELISA
TNF-␣ level in the medium was measured using Human TNF-␣ ELISA kit (Thermo Fisher Scientifi c, Rockford, IL) according to the manufacturer's instructions. Briefl y, vector-transfected or FASN1 and FASN2 clones of MCF7 cells were seeded at 1 × 10 5 cells/well in 48-well plates and on the next day treated with 1 µM doxorubicin or control vehicle for 6 h followed by collection of medium and determination of TNF-␣ protein concentration.

Cytotoxicity assay
The cytotoxicity of the cells to anticancer drugs or ␥ -irradiation was determined using the sulforhodamine B (SRB) colorimetric assay as previously described ( 6 ). Briefl y, cells in 96-well plates were cultured for 24 h and treated with different concentrations of anticancer drugs for three days. The culture medium was then removed, and the cells were fi xed and stained by 0.4% (w/v) sulforhodamine B (Sigma) in 1% acetic acid followed by washing with 1% acetic acid to remove the unbound SRB. The bound SRB was then solubilized with 10 mM Tris base, and the OD 570 nm was determined using a 96-well plate reader (MRX; Dynex Technologies, Chantilly, VA).

Luciferase reporter assay
For the assay of TNF-␣ promoter activity, MCF7 cells with stable FASN overexpression or MCF7/AdVp3000 cells with stable FASN knockdown and their respective control cells were seeded at 5 × 10 4 cells/well in 24-well plates and transfected with the fi refl y luciferase reporter construct driven by the TNF-␣

Materials
All electrophoresis reagents and polyvinylidene difl uoride (PVDF) membranes were purchased from Bio-Rad (Hercules, CA). Doxorubicin, mitoxantrone, cisplatin, etopside, paclitaxel, vinblastine, verapamil, and DTT were from Sigma (St. Louis, MO). Cell culture mediums DMEM and typsin-versene mixture were purchased from Media Tech (Herndon, CA) or Lonza (Walkersville, MD). Monoclonal antibody against human FASN and ceramide were purchased from BD Biosciences (San Jose, CA) and Enzo Life Sciences (Farmingdale, NY), respectively. Monoclonal antibody against cleaved poly(ADP-ribose) polymerase (PARP), caspase-8, cleaved caspase-8, and cleaved caspase-9 were obtained from Cell Signaling (Danver, MA). G418 was from Invitrogen (Carlsbad, CA). Specifi c siRNA targeting human FASN mRNA (sc-43758) was purchased from Santa Cruz Biotechnology. Specifi c siRNAs targeting human nSMase-2 and FASN were purchased from Qiagen and Santa Cruz Biotechnology, respectively. The NF-B reporter expression construct and its negative control plasmid of the PathDetect NF-B Cis-Reporting System were from Agilent Technologies (Santa Clara, CA). All other chemicals were of molecular biology grade from Sigma (St. Louis, MO) or Fisher Scientifi c (Chicago, IL).
Western blot analysis was done as described previously (12)(13)(14). Briefl y, cell lysates were separated by SDS-polyacrylamide gel electrophoresis followed by transfer to a PVDF membrane. The blot was then probed with a monoclonal antibody to FASN, cleaved PARP, AKT, NF-B p65, c-FLIP, or caspases, followed by reaction with horseradish peroxidase-conjugated secondary antibodies. The signal was captured by X-ray fi lm using enhanced chemiluminescence.

FASN activity assay
FASN activity was determined using a protocol previously described ( 14 ). Briefl y, 120 g of particle-free supernatant of cell lysate was mixed with a buffer containing 200 mM potassium phosphate (pH 6.6), 1 mM DTT, 1 mM EDTA, 0.24 mM NADPH, and 30 M acetyl-CoA in a fi nal volume of 0.3 ml, and the reaction was monitored at 340 nm for 3 min to measure background NADPH oxidation. After addition of 50 M malonyl-CoA, the reaction was monitored for an additional 15 min to determine FASN-dependent oxidation of NADPH using a Synergy H1 hybrid reader with Gen5 2.0 software (BioTek Instruments). The and FASN-overexpression MCF7 cells after treatment with or without 1 M doxorubicin for 48 h. PARP cleavage, another indicator of apoptosis, was analyzed using Western blot probed with an antibody that detects only the large fragment of the cleaved PARP to avoid potential confusion from detection of the uncleaved PARP.

Determination of ceramide level
Ceramides were extracted as previously described ( 19 ). Briefl y, MCF7 cells with stable FASN overexpression and vector-transfected control cells were resuspended in 0.5 ml PBS, mixed with 10 µl of 2.5 ng/l 17:0 ceramide as internal standards and 3 ml of methanol/chloroform (2:1), followed by incubation on ice for 10 min. Chloroform (1 ml) and water (1.3 ml) were then added to the mixture and vortexed for 1 min prior to centrifugation at 1,750 g for 10 min at 10°C. The organic phase was transferred to a new glass tube for evaporation of the solvent under nitrogen at room temperature. The dried lipids were dissolved in 1 ml of methanol for ESI-MS analyses using API-4000 (Applied Biosystems/MDS SCIEX, Forster City, CA) with an Analyst data acquisition system. Samples (10 µl each) were delivered into the electrospray ionization (ESI) source through a LC system (Agilent 1100) with an auto sampler. The mobile phase was methanol/water/ammonium hydroxide (90:10:0.1, v/v/v). Standard curves for all ceramides were established quantitative analyses.
The effect of GW4869 on total ceramide production was determined using immunofl uorescence staining as previously described promoter. Renilla luciferase reporter with a TK promoter was used as a control for transfection effi ciency. Forty-eight hours after transfection, cells were washed three times with PBS and lysed in passive lysis buffer (Promega, Madison, WI). Luciferase activity was measured using the dual luciferase reporter assay system according to the manufacturer's instructions (Promega).
For the assay of NF-B activity, the same procedures were followed as described above using the PathDetect NF-B Cis-Reporting System along with the renilla luciferase reporter driven by a TK promoter as a control for transfection effi ciency. Basal NF-B activity was detected 48 h after transfection. For doxorubicin-induced NF-B activity, MCF7/AdVp3000 cells were treated with 0.1 µM doxorubicin for 24 h in the presence of 10 µM FTC after reporter transfection, followed by determination of luciferase activity.

Apoptosis assay
The cell death detection ELISA (Roche, Mannheim, Germany) assay was conducted to determine apoptosis in control Fig. 1. Effect of FASN overexpression on cellular resistance to chemotherapeutic agents and ␥irradiation. Stable MCF7 cells with FASN overexpression (FASN1 and FASN2) and vector-transfected control (Vec) cells were tested for their level of FASN expression using Western blot analysis (A), real-time RT-PCR (B), and FASN activity assay (C), and for their resistance to various anticancer drugs and ␥ -irradiation using SRB cytotoxicity assay (D) as described in Materials and Methods. Dose response of SRB assay was analyzed by GraphPad Prism (version 3.02) to generate IC 50 . Relative resistance factor = IC 50 of FASN overexpressing cells/IC 50 of vectortransfected control cells. CDDP, cisplatin; CPT, camptothecin; DOX, doxorubicin; IR, ␥ -irradiation; MXT, mitoxantrone; PT, paclitaxel; VBL, vinblastine; VP16, etopside. * P < 0.05; ** P < 0.01. nSMase activity assay nSMase activity was determined using the sphingomyelinase assay kit from Echelon Biosciences (Salt Lake City, UT) following manufacturer's instructions. Briefl y, vector-transfected and FASN1 or FASN2 stable clones of MCF7 cells were treated with or without 1 M doxorubicin for 48 h. Cells were then harvested, lysed, and centrifuged at 20,000 g for 10 min to remove insoluble and noncytosolic proteins. About 100 µg of cytosolic proteins were used for determination of nSMase activity.

FASN overexpression causes resistance to multiple chemotherapeutic agents and ␥ -irradiation
We have shown previously that ectopic overexpression of FASN increases cellular resistance to doxorubicin and mitoxantrone ( 14 ). To determine whether FASN overexpression contributes to multidrug resistance, we tested the response of two previously established MCF7 clones with ectopic overexpression of FASN (FASN1 and FASN2) to multiple anticancer drugs in comparison with control cells transfected with vector (Vec). Fig. 1A -C shows that both FASN1 and FASN2 clones have increased FASN mRNA and protein levels as well as FASN activity compared with the Vec control cells. However, the two clones do not appear to differ signifi cantly in FASN expression and activity. Fig. 1D shows that both FASN1 and FASN2 cells are signifi cantly more resistant than the control Vec clone to doxorubicin, mitoxantrone, etopside, camptothecin, and cisplatin with 1.5-to 3-fold increases in relative resistance factor. However, the relative resistance factor between FASN1 and FASN2 cells is not signifi cantly different, consistent with FASN expression level and activity between the two cells. Interestingly, FASN overexpression had no signifi cant effect on cellular response to vinblastine and paclitaxel. These observations suggest that FASN overexpression may cause cellular resistance to DNA-damaging drugs but not to microtubule modulators, such as vinca alkaloids, in MCF7 cells. To further test this notion, these cells were treated with ␥ -irradiation followed by survival analysis. Fig. 1D shows that both FASN1 and FASN2 cells are also signifi cantly more resistant to ␥ -irradiation than the Vec control cells.

FASN overexpression protects cancer cells from druginduced apoptosis via inhibiting caspase 8 activation
To determine whether FASN overexpression protects cells from drug-induced apoptosis, we fi rst tested the druginduced apoptosis rate of FASN1 and FASN2 stable clones compared with Vec control cells using the cell death detection ELISA kit that quantifi es the level of DNA fragmentation. As shown in Fig. 2A , both FASN1 and FASN2 clones had signifi cantly lower levels of doxorubicin-induced DNA fragmentation compared with the Vec control cells. We next analyzed cleaved PARP, another indicator of apoptosis ( 21,22 ), using Western blot and found that the production of the 89 kDa cleaved PARP was signifi cantly less in response to doxorubicin treatment in FASN1 and ( 20 ). Briefl y, MCF7 cells were treated with 10 M GW4869 or 5% methane sulfonic acid vehicle for 30 min, and then treated with 1 M ADR or vehicle DMSO for 24 h. Cells were then harvested, fi xed, and permeabilized using the BD Cytofi x/Cytoperm Plus kit (BD, Franklin Lakes, NJ), followed by staining with monoclonal anti-ceramide antibody and FITC-conjugated secondary antibody. DAPI was used to stain total DNA as a control. Fluorescence was measured using a Synergy H1 plate reader with FITC at excitation/emission of 485/535 nm and DAPI at 350/470 nm. FITC fl uorescence intensity was adjusted by that of the DAPI.

FASN overexpression has no effect on c-FLIP level and AKT activation
Because c-FLIP is a well-known inhibitor of caspase 8 activation and its degradation following drug treatment releases caspase 8 for activation ( 23,24 ), we next determined whether FASN overexpression affects drug-induced degradation of c-FLIP using Western blot analysis. As shown in supplementary Fig. I FASN has been shown to have a positive feedback relationship with AKT in multiple studies (for a review, see Ref. 1 ), and AKT is an important regulator of cell survival and apoptosis (for a review, see Ref. 25 ). It has also been shown that doxorubicin increases PI3K-dependent AKT phosphorylation/activation as early as 1 h with a peak at 24 h ( 26 ). Consequently, we tested whether FASN overexpression affects doxorubicin-induced AKT activation and phosphorylation at 24 h following doxorubicin treatment using Western blot analysis. Supplementary Fig. I-B shows that the basal level of phosphorylated AKT is not detectable in all three (Vec, FASN1, and FASN2) cell lines. Following a 24 h doxorubicin treatment, the level of phosphorylated AKT increased dramatically in all three cell lines. However, the level of phosphorylated AKT in FASN1 and FASN2 cells is not different from that in the Vec control cells. Thus, FASN overexpression in MCF7 cells does FASN2 cells than in the Vec control cells ( Fig. 2B ), consistent with the fi ndings using ELISA assay. Together, these results show that FASN overexpression likely inhibits druginduced apoptosis and thus contributes to doxorubicin resistance in MCF7 cells.
We next determined whether FASN affects drug-induced activation of caspases 8 or 9 using Western blot analysis. As shown in Fig. 2C , caspase 8 in the Vec control cells is activated following doxorubicin treatment, whereas this activation in FASN1 and FASN2 clones is inhibited. Caspase 9, on the other hand, failed to be activated by doxorubicin treatment in the cells tested, and there is no difference in caspase 9 activation between the Vec control cells and the FASN1 and FASN2 cells (data not shown). Thus, FASN may protect cells from drug-induced apoptosis by suppressing drug-induced caspase 8 activation.
To determine whether the above observation is cell-line specifi c and whether knocking down FASN in a parental cell line would have the opposite effect on apoptosis and caspase 8 activation, we transiently knocked down FASN expression in human breast cancer cell line MDA-MB-468 as we previously described ( 14 ), followed by doxorubicin treatment and determination of PARP cleavage and caspase 8 activation. As shown in Fig. 2D , FASN knockdown sensitized/enhanced the parental MDA-MB-468 cells to doxorubicin-induced apoptosis and caspase 8 activation. This observation is consistent with our previous fi nding that FASN knockdown in MDA-MB-468 cells causes sensitization to anticancer drugs ( 14 ). Thus, the effect of FASN on drug-induced apoptosis and caspase 8 activation is not cell-line specifi c and confi rms the role of FASN in suppressing drug-induced caspase 8 activation. and C24:0, also appears to be reduced in FASN-overexpression cells. It is noteworthy that FASN1 cells may be more tolerant than FASN2 cells to ceramide because its basal level of ceramide is signifi cantly higher. Although doxorubicin-induced absolute level of total ceramide is not signifi cantly lower in the FASN1 clone, the fold increase in doxorubicin-induced ceramide production is signifi cantly less in FASN1, similar to FASN2, than in the Vec control cells. The individual ceramides also have the similar trend of decrease.
Ceramide can be generated either by de novo synthesis from palmitoyl CoA and serine or by breaking down sphingomyelin by acid or neutral sphingomyelinase (aSMase or nSMase). To determine which pathway is used in doxorubicin-induced ceramide generation and subsequent cell death in MCF7 cells, several selective inhibitors were used in combination with doxorubicin for survival assay. As shown in Fig. 4 , fumonisin B1 and myriosin (inhibitors of de novo ceramide synthesis; Fig. 4A, B ) and desipramine (aSMase inhibitor; Fig. 4C ) had no effect on doxorubicin inhibition of cell survival. However, GW4869, an nSMase inhibitor that can effectively inhibit doxorubicininduced ceramide production in MCF7 cells (supplementary Fig. III), blocked the inhibitory effect of doxorubicin on cell survival ( Fig. 4D ). Because nSMase-2 is the major nSMase in MCF7 cells ( 17,18 ) and it is known to be activated by dunarubicin (an analog of doxorubicin) ( 30 ), we next tested whether knocking down nSMase-2 expression using siRNA could block the inhibitory effect of doxorubicin on cell survival. As shown in supplementary Fig. IV, knocking down nSMase-2 expression signifi cantly attenuated the doxorubicin effect on MCF7 cell survival. Thus, the nSMase → ceramide → caspase 8 → apoptosis pathway may contribute to doxorubicin-induced apoptosis of MCF7 cells, and this pathway may be inhibited by FASN overexpression.
To test this possibility, we compared the nSMase activity among FASN1, FASN2, and the Vec control cells. Fig. 5A shows that the basal nSMase activity had no signifi cant difference among all three cell lines. Upon doxorubicin treatment, the nSMase activity in the Vec control cells is dramatically increased ( Fig. 5B ). However, the doxorubicin-induced increase in nSMase activity is signifi cantly suppressed by FASN overexpression in the FASN1 and FASN2 cells. Thus, nSMase may be a major target and mediator of FASN in doxorubicin resistance.

TNF-␣ expression is suppressed by FASN overexpression
To elucidate the signaling pathways further, a nonbiased comparative real-time RT-PCR SuperArray analysis of FASN1, FASN2, and the Vec control cells was conducted. Of the 89 genes analyzed, we found only 2 genes [interleukin (IL)-8 and TNF-␣ ] that had greater than 1.5-fold change consistently in duplicated experiments between the Vec control and the FASN-overexpressing FASN1 and FASN2 cells (see supplementary Tables I and II). This array-based fi nding was verifi ed in another independent FASN-overexpressing clone in duplicated experiments (supplementary Table III). The average reduction in IL-8 not appear to affect doxorubicin-induced AKT activation. Although AKT has been mainly involved in preventing apoptosis, it is also known to activate c-FLIP ( 27,28 ). Together with the data shown in supplementary Fig. I-A, we conclude that FASN does not have an effect on the AKT → cFLP → caspase 8 signaling pathway in MCF7 cells.

FASN overexpression decreases doxorubicin-induced ceramide generation via neutral sphingomyelinase
Another known upstream mediator of drug-induced caspase 8 activation is ceramide ( 29 ). In addition, the metabolism of different classes of lipids is highly intertwined. Thus, we tested the potential effects of FASN on ceramide production. As shown in Fig. 3 , the basal total ceramide level in FASN1 and FASN2 stable clones is slightly higher than in the Vec control cells ( Fig. 3A ), and doxorubicin treatment dramatically induced ceramide level in all cell lines ( Fig. 3B ). However, the doxorubicin-induced increase in total ceramide production is compromised in the FASN stable clone (more signifi cantly in FASN2 clone) compared with that in the Vec control cells ( Fig. 3B, C ). The basal and doxorubicin-induced production of individual ceramide is shown in supplementary Fig. II. It is clear that the C16:0 ceramide is the major constituent of the total ceramide and that doxorubicin-induced C16:0 ceramide production is compromised in both FASN1 and FASN2 cells compared with the Vec control cells. Doxorubicin-induced production of other ceramides, such as C18:1 To determine whether the effect of FASN on TNF-␣ expression is possibly via inhibiting its transcription, FASNoverexpressing MCF7 cells and FASN-knockdown MCF7/ AdVp3000 cells along with their vector or scrambled shR-NA-harboring control cells were transfected with a fi refl y luciferase reporter construct driven by TNF-␣ promoter. A renilla luciferase reporter construct driven by thymidine kinase promoter was cotransfected into these cells as controls for transfection effi ciency. As shown in Fig. 7A , both FASN1 and FASN2 cells show signifi cantly lower relative TNF-␣ promoter activity than do the Vec control cells. MCF7/AdVp3000 cells with FASN knockdown (Si) have signifi cantly higher relative TNF-␣ promoter activity than the control cells (Scr) ( Fig. 7B ).
Because TNF-␣ is a known upstream regulator of and activates NF-B, we hypothesize that the effect of FASN on TNF-␣ expression likely will consequently affect NF-B activity. To test this possibility, we performed a reporter assay of NF-B activity in MCF7/AdVp3000 cells with stable FASN knockdown and its scramble shRNA transfected control cell by transiently transfecting luciferase reporter and TNF-␣ expression in all three stable clones with FASN overexpression was about 16-and 13-fold, respectively.
To verify the above fi ndings, we performed a real-time RT-PCR analysis of TNF-␣ expression in the control Vec and FASN-overexpressing FASN1 and FASN2 cells. As shown in Fig. 6A , B , both the basal and doxorubicin-induced TNF-␣ mRNA levels are signifi cantly reduced in FASN1 and FASN2 cells compared with the Vec control cells. To determine whether the production of TNF-␣ protein is also inhibited, we performed an ELISA assay. As shown in Fig. 6C , doxorubicin-induced production of TNF-␣ protein was signifi cantly reduced in the FASN1 and FASN2 cells compared with the Vec control cells. We next performed analysis of TNF-␣ expression using the stable FASN knockdown MCF7/ AdVp3000 cells ( 14 ), and we found that FASN knockdown drastically increased both basal and doxorubicin-induced TNF-␣ expression ( Fig. 6D, E ), consistent with the fi ndings of FASN-overexpression cells. We also tested a different drug, mitoxantrone, in inducing TNF-␣ production and found that FASN knockdown similarly inhibited mitoxantrone-induced TNF-␣ expression ( Fig. 6F ). This fi nding is contradictory to the observations by Menendez et al. ( 32 ) who found that inhibiting FASN activity enhanced the cytotoxicity of docetaxel in the Her2-overexpressing breast cancer cell lines. The same group has also reported that inhibiting FASN expression or activity could sensitize breast cancer cells with a high Her2 level to vinorelbine ( 33 ) and paclitaxel ( 34 ). The reason for this discrepancy is currently unknown; however, it is possible that the high level of Her2 may affect the role of FASN in cellular response to docetaxel.
Although the detailed mechanism is not yet known, FASN has been shown to play a role in regulating gene expression. A previous genomic profi ling analysis of a breast cancer cell line MDA-MB-435 showed that FASN knockdown upregulated the expression of IL-8 ( 35 ), consistent with our fi nding in this study. Because IL-8 is one of the downstream target genes of TNF-␣ signaling via NF-B ( 36 ), the observed reduction of IL-8 in FASN-overexpressing clones may be the result of TNF-␣ reduction. Indeed, we also found that NF-B activity in FASN knockdown cells is signifi cantly upregulated, and its expression in FASNoverexpression cells is downregulated. Thus, the TNF-␣ signaling pathway may be inversely regulated by FASN, and it may mediate FASN-induced drug resistance.
It is noteworthy that the effect of FASN on TNF-␣ promoter activity ( Fig. 7 ) is lesser compared with the effect on the expression of endogenous TNF-␣ ( Fig. 6 ). The reason for this difference is currently unknown. However, it is likely that transfection effi ciency of the reporter construct contributed to the lower effect of FASN on the promoter activity assay. It is also possible that the TNF-␣ promoter in the reporter construct does not constitute the proper structure as does the endogenous promoter for a more effi cient effect. Further, it cannot be ruled out that the endogenous TNF-␣ mRNA is less stable in FASN-overexpression cells, whereas the stability of reporter mRNA is not affected by FASN expression. All these possibilities may contribute to the lower effect of FASN on the reporter assay than on the endogenous TNF-␣ , and further investigations are needed to delineate the detailed mechanism of FASN effect on TNF-␣ expression.
construct driven by a promoter containing NF-B response element followed by treatment with or without doxorubicin and determination of luciferase expression. As shown in Fig. 7C, D , both the basal and doxorubicin-induced NF-B activity are dramatically increased in the cells with FASN knockdown compared with the control cells. Furthermore, we found that the expression level of NF-B in the FASN overexpression cell lines is dramatically reduced compared with that of the Vec control cells ( Fig.  7E ). These observations are consistent with the negative relationship between FASN expression and TNF-␣ expression. Together, these fi ndings suggest that FASN may downregulate the transcription and production of TNF-␣ , which in turn downregulates NF-B level and activity and suppresses doxorubicin-induced ceramide production and apoptosis.

DISCUSSION
Altered and increased metabolism is one of the hallmarks of cancer ( 31 ). Despite the fact that the use of anticancer agents in appropriate combinations has led to major improvements in the treatment of malignant tumors, resistance to chemotherapy frequently occurs and is a major obstacle in cancer treatment. Hence, understanding metabolic-related drug resistance becomes critical. In this study, we show that FASN overexpression in breast cancer cell line MCF7 causes resistance to DNA damageinduced cell death possibly via inhibition of TNF-␣ production, which in turn increases NF-B activity and reduces drug-induced production of ceramide possibly by nSMase and caspase 8-mediated apoptosis.
Previously, FASN overexpression was found in Adriamycin (doxorubicin)-selected and multidrug-resistant MCF7 cells, and it contributes to the observed drug resistance of these cells ( 14 ). It was also found that FASN overexpression contributes to gemcitabine resistance in pancreatic cancer cells ( 15 ). Interestingly, we did not fi nd that FASN overexpression in MCF7 cells causes resistance to non-DNA-damaging drugs, such as paclitaxel and vinblastine. were transiently transfected with a fi refl y luciferase reporter construct driven by TNF-␣ promoter (A, B) or by a luciferase reporter construct containing a NF-B response element (C, D) followed by treatment without or with 1 µM doxorubicin (and in the presence of 10 µM FTC for MCF7/AdVp3000 cells) for 24 h and dual luciferase reporter assay as described in Materials and Methods. * P < 0.05; ** P < 0.01. E: Western blot analysis of NF-B p65 in FASN overexpression (FASN1 and FASN2) and vectortransfected control (Vec) cells. Actin is used as a loading control. which in turn reduces doxorubicin-induced production of ceramide possibly by reducing nSMase activity, resulting in lowered caspase 8 activation and apoptosis ( Fig. 8 ). Because mitoxantrone, camptothecin, and etopside, similar to doxorubicin, are all topoisomerase II inhibitors, FASN may also use the same mechanism of resistance to these drugs as to doxorubicin (see Fig. 8 ). Indeed, we found that mitoxantrone-induced TNF-␣ expression is signifi cantly increased by knocking down FASN expression. However, how FASN overexpression inhibits TNF-␣ production is currently unknown. Previously, it has been shown that supplementation of palimitate causes cellular resistance to doxorubicin and mitoxantrone ( 14 ). Thus, it is possible that palmitate, the fi nal product of FASN, plays some role in inhibiting TNF-␣ production; this possibility is awaiting further investigation.
It is also noteworthy that the anticancer drugs (doxorubicin, mitoxantrone, camptothecin, and etopside) that FASN causes resistance to are topoisomerase II inhibitors and that they as well as ␥ -irradiation induce double-strand DNA breaks. Cisplatin also causes DNA damage. Thus, it is tempting to speculate that FASN overexpression may protect MCF7 cells against DNA damage or facilitate the repair of damage. In support of this argument, it has been observed that FASN suppresses the expression of DNA damage-inducible transcript 4, a stress-response gene, which may also mediate the function of FASN in resistance to DNA-damaging treatments. We are currently investigating this possibility.
GW4869, among several other inhibitors, was found to be the only inhibitor that improved survival against doxorubicin treatment. GW4869 is a cell-permeable, noncompetitive, specifi c inhibitor of nSMase-2 ( 37 ), and it does not affect aSMase activity or other key TNF-mediated signaling effects, such as NF-B activation ( 38 ), GW4869 has been shown repeatedly to suppress ceramide production. For example, GW4869 signifi cantly inhibited TNF-␣ -induced ceramide production in MCF7 cells ( 38 ), signifi cantly reduced the l-OHP-induced accumulation of ceramides in human colon cancer cell line HCT116 ( 39 ), and inhibited hypoxia-induced ceramide production in rat pulmonary artery smooth muscle cells ( 20 ). Furthermore, it has been shown that anthracycline drugs, such as doxorubicin, induce ceramide production in cancer cells by activating nSMase and that inhibiting nSMase attenuates doxorubicin-induced ceramide production ( 40 ). Thus, we conclude that the FASN function in drug resistance and in suppression of drug-induced ceramide production is likely mediated by nSMase. This conclusion is corroborated by our observation that specifi cally knocking down nSMase-2 signifi cantly attenuates the doxorubicin effect on MCF7 cell survival and that inhibiting nSMase reduces doxorubicin-induced ceramide production.
It is noteworthy that FASN appears to inhibit doxorubicin-induced fold increase in ceramide level in two independent clones, although the absolute level of increase in one of the two clones is also signifi cantly reduced by FASN overexpression ( Fig. 3 ). Currently, it is not clear whether the inhibition in fold increase or absolute level of ceramide is the key to FASN function in doxorubicin resistance. Future studies will be needed to clarify this issue.
It has also been shown that TNF-␣ activates nSMase and initiates a rapid increase in ceramide production ( 37,(41)(42)(43) and that doxorubicin treatment increases TNF-␣ expression ( 44 ). Ceramide-induced apoptosis has also been shown to be mediated by caspase 8 ( 45 ). On the basis of our fi ndings and previous observations, we propose that FASN overexpression suppresses production of TNF-␣ , Fig. 8. Mechanism of FASN action in drug resistance. FASN overexpression causes drug resistance possibly by inhibiting TNF-␣ production, which in turn inhibits drug-induced nSMase activation, production of ceramide from sphingomyelin, activation of caspase 8, and apoptosis.