PPARγ ligand production is tightly linked to clonal expansion during initiation of adipocyte differentiation.

Adipocyte differentiation is orchestrated by the ligand-activated nuclear receptor PPARγ. Endogenous ligands comprise oxidized derivatives of arachidonic acid and structurally similar PUFAs. Although expression of PPARγ peaks in mature adipocytes, ligands are produced primarily at the onset of differentiation. Concomitant with agonist production, murine fibroblasts undergo two rounds of mitosis referred to as mitotic clonal expansion. Here we show that mouse embryonic fibroblasts deficient in either of two cell cycle inhibitors, the transcription factor p53 or its target gene encoding the cyclin-dependent kinase inhibitor p21, exhibit increased adipogenic potential. The antiadipogenic effect of p53 relied on its transcriptional activity and p21 expression but was circumvented by administration of an exogenous PPARγ agonist suggesting a linkage between cell cycling and PPARγ ligand production. Indeed, cell cycle inhibitory compounds decreased PPARγ ligand production in differentiating 3T3-L1 preadipocytes. Furthermore, these inhibitors abolished the release of arachidonic acid induced by the hormonal cocktail initiating adipogenesis. Collectively, our results suggest that murine fibroblasts require clonal expansion for PPARγ ligand production at the onset of adipocyte differentiation.


Retro-and lentiviral transductions
Retroviral transduction was done as described previously with puromycin selection for 2 days ( 19 ). Lentiviral particles were produced as described ( 20 ). Lentiviral transduction was confi rmed by green fl uorescent protein expression .

Electroporation
3T3-L1 preadipocytes were electroporated with UASx4-TK luc and GAL4-PPAR ␥ -LBD 1 day past confl uence according to the protocol described elsewhere ( 21 ). The day after electroporation, cells were treated as indicated, and the following day, cells were harvested and luciferase activity was measured according to the standard protocol.

Arachidonic acid release
Confl uent 3T3-L1 preadipocytes were supplemented with radioactively labeled arachidonic acid ([1-14 C]arachidonic acid, Perkin Elmer) at an activity of 0.1 Ci/ml from day Ϫ 2 to day 0. Before induction of differentiation, cells were rinsed three times with differentiation media. Cells were induced to differentiate as indicated. Two days past induction, cells were frozen to lyse cells and release intracellular free arachidonic acid. Cellular debris was removed by centrifugation. Scintillation fl uid was added to the media, and radioactivity was determined.

RNA purifi cation, reverse transcription, and real-time PCR
RNA was purifi ed using TRIzol (Invitrogen) according to the manufacturer's instructions. Reverse transcription was performed essentially as described elsewhere ( 8 ). Quantitative PCR (qPCR) was performed in 25 l reactions containing SYBR ® Green JumpStart™ Taq ReadyMix™ (Sigma-Aldrich), 1.5 l of diluted cDNA, and 300 nM of each primer. Reaction mixtures were preheated at 95°C for 2 min followed by 40 cycles of melting at 95°C for 15 s, annealing at 60°C for 30 s, and elongation at 72°C for 45 s. Primer sequences are available on request. Expression of TATA box binding protein was used for normalization.

DAPI staining
3T3-L1 cells were induced as indicated followed by fi xation and permeabilization by 90% methanol. Unspecifi c binding was blocked by replacing methanol with 5% skim milk in PBS. Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI; Invitrogen) in fresh blocking solution. Stained nuclei were scored by fl uorescence microscopy. compared with its boost in expression was suggested to be important for timely induction of C/EBP ␣ and PPAR ␥ expression. However, mutational ablation of the phosphorylation sites does not affect the adipogenic potential of C/EBP ␤ ( 12 ), and the delay in DNA binding has since been questioned ( 13 ).
Interestingly, data suggest that increased proliferation capacity is associated with increased adipogenic potential. Mice lacking the two cyclin-dependent kinase inhibitors (CDKIs) p21 and p27 display adipose tissue hyperplasia ( 14 ). Also, MEFs lacking p21 undergo accelerated adipogenesis, although the CDKI was required for the survival of mature adipocytes ( 15 ).
The tumor suppressor p53 is critically involved in cell cycle control as evidenced by the massive and accelerated tumor development in p53-defi cient mice ( 16 ). The p53 controls cell cycling through regulating expression of several genes, including p21. Here we show that MEFs lacking either p53 or p21 undergo spontaneous adipogenesis. The inhibitory effect of p53 on adipocyte differentiation relies on its transcriptional activity. Inhibition of adipocyte differentiation by forced expression of p53 or chemical blockage of mitotic clonal expansion was circumvented by the addition of a PPAR ␥ agonist suggesting that the importance of mitotic clonal expansion is linked to a concomitant production of PPAR ␥ agonist(s) necessary for the initiation of adipocyte differentiation.

Cell culture, differentiation
Wild-type and p53-defi cient MEFs were generous gifts from Dr. Stephen N. Jones and Dr. Guillermina Lozano. The p21defi cient and corresponding wild-type MEFs were kindly supplied by Dr. Jean-Christophe Marine. For spontaneous differentiation, MEFs were left confl uent for 10 days with media (AmnioMax basal medium, Life Technologies Inc.) supplemented with 7.5% fetal bovine serum, 7.5% AmnioMax-C100 supplement, 2 mM glutamine, 62.5 µg/ml penicillin, and 100 µg/ml streptomycin, changed every second day. 3T3-L1 preadipocytes were grown and differentiated as described previously ( 17 ). Roscovitine (LC Laboratories), purvalanol (MERCK), paclitaxel (LC Laboratories), or rosiglitazone (kindly supplied by Novo Nordisk) were added to the media to the indicated concentrations throughout differentiation. For Oil Red O staining, cells were washed in PBS, fi xed in 3.7% paraformaldehyde for 1 h, and stained as described ( 18 ).
The mdm2 gene is amplifi ed in 3T3-L1 cells ( 22 ), arguing that this cell line is less well suited for analysis of the effects of p53 on adipocyte differentiation and function. We therefore used MEFs to examine the impact of p53 on adipose conversion.
MEFs normally require a hormonal cocktail in order to induce adipose conversion. However, in contrast to wildtype MEFs, we observed that p53-defi cient MEFs underwent spontaneous adipocyte differentiation (supplementary Fig. IA, B). Importantly, ectopic expression of p53 effi ciently inhibited the spontaneous adipogenesis of p53-defi cient MEFs (supplementary Fig. IC, D). Additionally, knockdown of p53 in wild-type MEFs leads to spontaneous formation of adipocytes (supplementary Fig. IE-G). Therefore, our data and those of others ( 23,24 ) point to an inhibitory role for p53 in adipose conversion.
The p53 exerts both transcriptional and nontranscriptional effects. We sought to determine whether the inhibitory effect on adipogenesis relied on transcriptional activity of p53 by ectopic expression of either wild-type or a DNA binding-defi cient mutant. Ectopic expression of wild-type p53 inhibited hormonally induced adipocyte differentiation of wild-type MEFs as determined by triglyceride staining with Oil Red O ( Fig. 1A ) and adipocyte marker gene expression ( Fig. 1B ). The inhibitory effect was dependent on the ability of p53 to bind to DNA, as a p53 mutant with

EdU incorporation
The p53 Ϫ / Ϫ MEFs were cultivated in Amniomax, seeded in chamberslides, and 1 day after reaching confl uence incubated with EdU (Click-iT® EdU Cell Proliferation Assay, Invitrogen) for 4 h. Four days later, cells were fi xed with paraformaldehyde and permeabilized, and incorporated EdU was visualized by immunohistochemistry essentially as recommended by the manufacturer.

SYBR Green growth assay
3T3-L1 cells (grown in DMEM) or MEFs (grown in Amniomax) were transduced with empty vector (pBABE), p53wt, or p53 R175D mutant using standard protocols. After 3 days of selection with puromycin (3 g/ml), cells were seeded in 48-well plates and induced to differentiate according to standard protocols. At the time of induction (day 0) and 4 days later (day 4), cells were harvested in 1% SDS; the amount of accumulated DNA was quantifi ed by staining with SYBR Green, and fl uorescence measured on an EnVision multi-label reader (PerkinElmer).

RESULTS
The understanding of adipocyte differentiation has been greatly facilitated by in vitro studies using MEFs and the established preadipocyte cell line 3T3-L1. The murine double minute 2 ( mdm2 ) gene encodes the ubiquitin ligase Mdm2, which is a crucial regulator of p53 degradation. impaired DNA binding ability (p53 R175D) ( 25 ) failed to inhibit adipose conversion of wild-type MEFs ( Fig. 1A, B ). Although p53 and p53 R175D were both expressed (supplementary Fig. IIA), only wild-type p53 induced expression of p21 (supplementary Fig. IIB) confi rming the transcriptional inactivity of the p53 R175D mutant. Furthermore, wild-type p53 but not p53 R175D prevented cell division during the early stage of adipocyte differentiation as indicated by measurements of DNA content ( Fig. 1C ). The failure of p53 R175D to inhibit adipocyte differentiation strongly suggested that the transcriptional activity of p53 was required for its antiadipogenic effect.
To relate the possible inhibitory effect of p53 on hormonal induction of adipogenesis in wild-type cells, we analyzed the level of p53 and its phosphorylation status during the early stage of adipocyte differentiation of wild-type MEFs. Phosphorylation of p53 at numerous residues is reported to exert prominent control on the function of p53 ( 26 ). The total level of p53 and several of its N-terminal phosphorylations did not change over the course of adipose conversion ( Fig. 1D ). These phosphorylation events generally exert a stabilizing effect on p53 ( 26 ). However, phosphorylation of serine 389 (serine 392 in human) decreased during the fi rst 2 days of differentiation and then later returned to starting levels ( Fig. 1D ). Interestingly, mutational analyses have shown that phosphorylation of this site is important for maintaining the basal level of expression of numerous p53 target genes ( 27,28 ). It is possible that reduced phosphorylation at serine 389 of p53 promotes adipogenesis by reducing the transcriptional activity of p53 and thereby facilitates clonal expansion.
One of several cell cycle genes whose expression is perturbed by mutation of serine 389 is the CDKI p21 ( 28 ), one of the major players in p53-mediated cell cycle arrest ( 29 ). Expression of p21 was signifi cantly lower in p53-defi cient MEFs compared with wild-type MEFs at confl uence ( Fig. 2A ). Lack of p21 has previously been associated with accelerated hormonally induced adipogenesis and hyperplasic adipose tissue ( 14,15 ). We therefore speculated whether p21-defi cient MEFs would recapitulate the spontaneous adi pogen esis observed in p53-defi cient MEFs. Indeed, maintaining wild-type and p21-defi cient MEFs confl uent for Fig. 2. The p21-defi cient MEFs have increased adipogenic potential. A: Expression of p21 in confl uent wild-type and p53-defi cient MEFs as measured by real-time qPCR. * P < 0.05, Student's t -test. Error bars represent standard deviation. B, C: Wild-type and p21defi cient MEFs were kept confl uent for 10 days with fresh media added every second day. Adipose conversion was assessed by triglyceride staining using Oil Red O staining (B) or adipocyte marker gene expression using real-time qPCR (C). * P < 0.05, Student's t -test. Error bars represent standard deviation. D, E: The p21-defi cient MEFs were retrovirally transduced with either empty vector or vector encoding p53, selected, and kept confl uent for 10 days with media changed every second day. Degree of adipogenesis was scored by adipocyte marker gene expression using Oil Red O staining (D) or adipocyte maker gene expression using real-time qPCR (E). * P < 0.05, Student's t -test. Error bars represent standard deviation. differentiation in the presence of vehicle or the chemical CDKI roscovitine. Increasing concentrations of roscovitine led to a dose-dependent decrease in spontaneous adipose conversion of p53-defi cient MEFs as seen by a decrease in the number of fat droplets ( Fig. 3C ) and adipocyte marker gene expression ( Fig. 3D ). Similar results were observed when treating p53-defi cient MEFs with two other cell cycle inhibitors, paclitaxel and purvalanol (data not shown).
Concomitant with the mitotic clonal expansion, ligand(s) for PPAR ␥ is generated. Interestingly, in vitro adipocyte differentiation of human cells occurs in the absence of mitotic clonal expansion but requires the addition of an exogenous PPAR ␥ agonist ( 30 ). Adipocyte differentiation of C3H10T1/2 cells does not rely on completion of cell division ( 10 ). Effi cient differentiation of these cells is, however, dependent on inclusion of a PPAR ␥ agonist in the differentiation media unless pretreated with bone morphogenetic protein 4 or 7 (31)(32)(33)(34). Of further interest is the observation that MEFs defi cient for the retinoblastoma protein have irregular clonal expansion and require an exogenous PPAR ␥ ligand in order to develop into adipocytes ( 18 ).
Accordingly, we speculated whether clonal expansion was necessary for PPAR ␥ agonist production. To investigate this, we tested whether the inhibitory effect on adipogenesis of either synthetic cell cycle inhibitors or ectopic p53 10 days with medium renewed every second day resulted in spontaneous adipose conversion of the p21-defi cient MEFs, but not the corresponding wild-type MEFs as visualized by Oil Red O staining ( Fig. 2B ) and adipocyte marker gene expression ( Fig. 2C ). In keeping with the fi nding that p21 is required for survival of the mature adipocyte ( 15 ), we observed massive shedding of dead p21-defi cient adipocytes to the medium (results not shown). Finally, supporting the notion that induction of p21 is necessary for the p53-dependent inhibition of adipocyte differentiation, forced expression of p53 did not inhibit spontaneous adipose conversion of p21-defi cient MEFs ( Fig. 2D, E ).
The requirement for p21 and the DNA binding ability of p53 to inhibit adipogenesis suggested that continuous cell cycling might at least in part explain the spontaneous adipogenesis of p53-defi cient MEFs. Confl uent p53defi cient MEFs exhibited signs of increased cell cycling as they increased their DNA content relatively more than their confluent wild-type counterparts ( Fig. 3A ). To address whether the p53-defi cient cells developing into adipocytes had undergone cell division/DNA replication, we pulselabeled confl uent p53-defi cient with EdU and allowed the cells to differentiate spontaneously. More than 93% of labeled cells were adipocytes ( Fig. 3B and data not shown).
To further link cell cycling to adipogenesis in p53 Ϫ / Ϫ cells, we allowed these MEFs to undergo spontaneous adipocyte Fig. 3. Inhibition of cell cycling in p53-defi cient MEFs abolishes their spontaneous adipogenesis. A: Relative increase in DNA amount in confl uent wild-type and p53-defi cient MEFs. DNA amount was measured by SYBR Green fl uorescence and depicted as relative increase in DNA amount after 24 h compared with levels in confl uent cells. * P < 0.05, Student's t -test. Error bars represent SEM. B: Confl uent p53defi cient MEFs were labeled with EdU for 4 h, rinsed, and kept confl uent for 4 days with fresh media added at day 2. Incorporation of EdU and adipocyte differentiation was scored using microscopy. C, D: The p53-defi cient MEFs were kept confl uent for 10 days with fresh media added every second day in the presence of vehicle or increasing concentrations of roscovitine. Adipocyte differentiation was scored by triglyceride staining using Oil Red O staining (C) or adipocyte marker gene expression using real-time qPCR (D). Lowercase letters (a, b, c) denote different signifi cance groups; P < 0.05, one-way ANOVA. Error bars represent standard deviation. tended to increase the intracellular free pool of arachidonic acid ( Fig. 6A ).
As our results suggested that mitotic clonal expansion is intimately associated with ligand production, we sought to expression could be circumvented by addition of a PPAR ␥ agonist. Indeed, rosiglitazone was able to restore adipocyte differentiation in 3T3-L1 cells treated with each of the cell cycle inhibitors as seen by Oil Red O staining ( Fig. 4A ) and, as exemplifi ed by paclitaxel treatment, adipocyte marker gene expression ( Fig. 4B ). Along the same lines, rosiglitazone circumvented the inhibition of adipose conversion in wildtype MEFs mediated by forced expression of p53 ( Fig. 4C, D ), suggesting that the p53-mediated constraint of adipocyte differentiation precedes PPAR ␥ agonist(s) production. In support of this, the augmented adipogenesis of p53-defi cient MEFs was inhibited by addition of a PPAR ␥ antagonist ( 24 ). Collectively, these results argue that mitotic clonal expansion is linked to the production of PPAR ␥ agonist(s) during the initial stage of adipose conversion.
The generation of PPAR ␥ ligand(s) during adipocyte differentiation has been verifi ed using reporter assays measuring the activity of an isolated PPAR ␥ -LBD ( 6,8 ). We used a similar strategy to measure PPAR ␥ ligand production in 3T3-L1 cells induced to differentiate in the presence or absence of cell cycle inhibitors. All inhibitors effi ciently reduced PPAR ␥ -LBD-mediated transactivation of the reporter gene ( Fig. 5A ) supporting the idea that inhibition of mitotic clonal expansion reduced the production of PPAR ␥ agonist(s). In support of this, we found that the two inhibitors tested prevented the induction of the classic PPAR ␥ target gene, aP2/ALBP (adipocyte lipid binding protein)/FABP4 (fatty acid binding protein 4), on day 1 following administration of the classic hormonal cocktail, whereas they did not prevent aP2/ALBP/FABP4 induction by rosiglitazone ( Fig. 5B ).
The nature of the true endogenous PPAR ␥ agonist(s) inducing adipocyte differentiation remains enigmatic. It is, however, evident that the cAMP-elevating compound of the hormonal cocktail used to induce differentiation is responsible for initiating agonist production ( 6 ). Interestingly, elevated levels of cAMP were important for increasing the number of cells in the differentiating 3T3-L1 cells ( Fig. 5C ).
We have recently suggested that hepoxilins act as agonists during the initial stages of adipogenesis. The generation of hepoxilins during adipogenesis is likely to rely on a reactive oxygen species (ROS)-mediated conversion of arachidonic acid or structurally similar PUFAs into substrates for the unconventional lipoxygenase, epidermal type-3 lipoxygenase (eLOX3) ( 35 ). The source of these PUFAs was, however, not determined. As cAMP is suggested to be the priming factor for PPAR ␥ agonist production, we speculated whether cAMP could be linked to the release of PUFAs from phospholipids in cellular membranes. Therefore, we attempted to estimate free, endogenous arachidonic acid in 3T3-L1 cells treated with dexamethasone, 3-isobutyl-1-methylxanthine (MIX) , or insulin. In short, 3T3-L1 cells were grown in media containing 14 C-labeled arachidonic acid. Before stimulation, the labeled cells were rinsed thoroughly. Intracellular free arachidonic acid was released to media by freeze-thaw lysing of the cells. Before assaying radioactivity in cell lysate, debris was spun down. Although this assay only led to a rough estimate of released arachidonic acid, only the cAMP-elevating compound MIX

DISCUSSION
During recent years, there has been an explosion in knowledge of adipocyte differentiation. Still, the role of mitotic clonal expansion has been a matter of dispute. Here we provide data arguing that mitotic clonal expansion is important for the concurrent PPAR ␥ ligand production. We demonstrate that an exogenous PPAR ␥ agonist was capable of restoring adipogenesis in cells where cell cycling was ablated either chemically or genetically. Furthermore, the chemical inhibitors of cell cycling prevented the activation of a PPAR ␥ -responsive reporter plasmid.
Since the realization that PPAR ␥ plays a pivotal role in adipocyte development, several attempts have been made to identify the agonists activating PPAR ␥ during adipogenesis ( 37 ). Shortly after the cloning of PPAR ␥ , a derivative of arachidonic acid, 15-deoxy-⌬ 12,14 -prostaglandin J2, was suggested to be the endogenous ligand orchestrating adipose conversion ( 38,39 ). However, other reports have questioned the assumption that 15-deoxy-⌬ 12,14 -prostaglandin J2 is involved in the initiation of adipocyte differentiation ( 6,40 ).
Progress in the quest for the ligands has been made through deciphering pathways involved in ligand production. determine whether cell cycle inhibitors also prevented mobilization of arachidonic acid. Indeed, as determined by estimation of the free cellular arachidonic acid pool, inhibition of cell cycling prevented the increase in the level of free arachidonic acid in cells induced to undergo adipogenesis ( Fig. 6B ). Collectively, these results indicate that mitotic clonal expansion partakes in adipose conversion by facilitating the release of PUFAs used for PPAR ␥ agonist production.
As discussed previously, mitotic clonal expansion has been suggested to be required for activation of C/EBP ␤ ( 11 ). Recently, induction of xanthine oxidoreductase (XOR) expression during adipose conversion was shown to be highly dependent on C/EBP ␤ ( 36 ). As XOR expression can be used as a measure for C/EBP ␤ activity, we quantifi ed induction of XOR in differentiating 3T3-L1 cells treated with vehicle or cell cycle inhibitors. Although the inhibitors blocked MDI-induced adipose conversion ( Fig. 4A, B ), they did not affect the induction of either C/EBP ␤ or XOR during the early stages of adipocyte differentiation ( Fig. 6C ) suggesting that the requirement for mitotic clonal expansion in adipogenesis is not related to expression or activation of C/EBP ␤ . remodeling of the nuclear membrane during the clonal expansion is a prerequisite for effi cient PUFA release by cPLA 2 .
Besides the suggested pathway, it is possible that the elevated release in arachidonic acid refl ected an increase in the release of nitrolinoleic acids. These fatty acids are reported to be potent PPAR ␥ ligands ( 47 ). However, this does not fi t our observation that inhibition of lipoxygenase activity decreases PPAR ␥ ligand generation during adipocyte differentiation ( 8 ). Still, it is possible that nitrolinoleic acids contribute to the PPAR ␥ ligand pool under adipose conversion.
Cancer cells are defi ned by their uncontrolled proliferation. Furthermore, these cells are known to have high levels of ROS rendering them capable of producing PPAR ␥ agonists by the mechanism described previously. The plausible increased generation of PPAR ␥ agonists in cancers would seem counterintuitive considering the antiproliferative effect of PPAR ␥ and its ligands ( 48 ). However, because PPAR ␥ is mutated in several tumors and tumorderived cells lines (49)(50)(51), it is possible that inactivation of PPAR ␥ can circumvent the antiproliferative effects of the PPAR ␥ ligands generated by increased proliferation and ROS levels.
In conclusion, our results provide evidence for a cell cycle-dependent production of endogenous PPAR ␥ ligands necessary for initiation of adipocyte differentiation and with possible implications in relation to other cell types with deregulated cell cycle control.
Although the mechanism was not described, Farmer and colleagues pinpointed C/EBP ␤ as an essential regulator of ligand production ( 41 ). A later study by the laboratory of Friedman strongly suggested that C/EBP ␤ acted through induction of xanthine XOR expression ( 36 ). Still, the exact mechanism by which XOR augmented PPAR ␥ activity remained unresolved.
We have then proposed that lipid autooxidation by ROS generated by XOR results in the production of substrates for eLOX3. eLOX3 converts the peroxidized lipids into hepoxilin-like products that function as endogenous PPAR ␥ agonists during adipose conversion ( 35 ). The dependency on eLOX3 in PPAR ␥ ligand production is consistent with the observation that a lipoxygenase inhibitor, baicalein, decreases PPAR ␥ ligand generation ( 8 ). However, the source of lipids for the autooxidation reaction was unknown. eLOX3 can convert peroxidized fatty acids originating from arachidonic and linoleic acid ( 42,43 ). These fatty acids are present in phospholipids embedded in cellular membranes. Their release is mainly regulated by the actions of phospholipases. Of the known families of phospholipases, PLA 2 , PLC, and PLD, only PLA 2 releases the free fatty acid directly. Interestingly, the activity of cytoplasmic phospholipase A 2 (cPLA 2 ) is upregulated by proliferation ( 44 ). The activity and localization of cPLA 2 are regulated by an activating phosphorylation by p42/p44 MAPKs. This phosphorylation leads to localization of cPLA 2 to the nuclear membrane ( 45 ). cAMP is responsible for the activation of MAPK early during adipogenesis ( 46 ). Based on our recent fi ndings, we hypothesize that cAMP elevates MAPK activity through activation of the exchange proteins directly activated by cAMP ( 17 ). Through its activation of MAPK, the second messenger may thereby lead to the redistribution of cPLA 2 to the nucleus. It is therefore possible that the extensive Fig. 6. Cell cycle inhibitors abolish arachidonic acid release but not C/EBP ␤ transcriptional activity. A: 3T3-L1 cells were labeled with radioactive arachidonic acid, rinsed, and induced as indicated. Radioactivity in the media at day 2 was scored by scintillation after cell lysis. Error bars represent SEM. B: 3T3-L1 cells were labeled with radioactive arachidonic acid, rinsed, and induced as indicated. Radioactivity in the media at day 2 was scored by scintillation after cell lysis. * P < 0.05, one-way ANOVA. Error bars represent SEM. C: 3T3-L1 cells were induced to differentiate with MDI in the presence of absence of 10 M roscovitine, 5 M purvalanol, or 5 M paclitaxel. Expression of C/EBP ␤ and XOR was measured by real-time qPCR. * P < 0.05, one-way ANOVA. Error bars represent standard deviation.