Runx Regulation of Sphingolipid Metabolism and Survival Signaling

The Runx genes comprise a three-member family of mammalian transcription factors that regulate developmental cell fate decisions in which cells are selected to proliferate, differentiate, survive, or die. Unusually, these genes have been shown to act either as oncogenes or as tumor suppressors according to lineage and host cell genetic context (1). The transcriptional targets for Runx regulation are therefore of considerable interest and have already been shown to include factors involved in lineage-specific development as well as regulators of cell cycle checkpoints (2–4).

Notably, all three Runx genes have been implicated as oncogenes in lymphoid neoplasia in retroviral mutagenesis screens (5–8), whereas both Runx1 and Runx2 are potently synergistic with Myc (9, 10). The role of Runx in this context seems to be, at least in part, due to inhibition of Myc-induced apoptosis (11). Further evidence of an oncogenic role for Runx2 is provided by its ectopic expression in breast and prostate cancers in which it is associated with metastasis and survival in the bone environment (12, 13), whereas Runx1 promotes tumorigenicity in fibroblasts (14). Runx-mediated survival has also been reported in vitro. In lymphoid cells, Runx1 confers resistance to apoptosis induced by CD3 ligation (15) or dexamethasone (4), whereas in fibroblasts, all three Runx genes were shown to enhance survival under the stress of medium exhaustion (4, 14).

Despite this evidence for potent prosurvival activity, the relevant pathways regulated by Runx are not well known. Our previous gene array analysis revealed a paucity of well-established regulators of apoptotic pathways among the significantly regulated targets. However, a set of 50 genes selected on the basis of significant regulation by all three Runx-regulated genes included three genes encoding enzymes involved in sphingolipid metabolism (4). Each of these genes [Sgpp1, Ugcg, St3gal5(Siat9)] has been reported to play a role in regulating survival and apoptosis (16–18), suggesting that they may be important targets and determinants of Runx oncogenic potential. We now show that all three genes act as direct targets for Runx activation or repression. Moreover, biochemical analyzes show that Runx expression causes significant changes in intracellular sphingolipids and resets the ceramide-sphingosine-1-phosphate “rheostat” (19) in favor of cell survival.

NIH3T3 fibroblast-derived cells expressing Runx1, Runx2, and Runx3 genes from the pBabe retroviral vector system were maintained as previously described (4). Transfections were carried out using the Superfect Transfection Reagent (Qiagen) according to the manufacturer's instructions. Conditions for transfection have been previously described (4). Live/dead cell counts were carried out using a hemocytometer and trypan blue as a vital indicator. Graphs were generated with Sigma-Plot, and error bars relate to SDs. For tumor necrosis factor-α (TNF-α) experiments, cells were treated with 10 ng/mL TNF-α (Peprotech, Inc.) in the presence of 10 μg/mL cycloheximide (Sigma) for the indicated times.

Runx-expressing and control fibroblasts were grown to confluence in duplicate wells (7 d). Cells expressing the RUNX1-ER construct were treated for a further 1 to 24 hours with 4-OHT. cDNA preparation and the microarray assay were performed as described (4). For quantitative real-time PCR (qt-RT-PCR), aliquots (5 μL) of cDNA were amplified in triplicate using primers for murine endogenous control Hprt or primers for murine Sgpp1, St3gal5 (Qiagen QuantiTect Primer Assays), or Ugcg (779F 5′-tttgctcagtacattgctgaagatta-3′ and 861R 5′-acttgagtagacattgaaaacctccaa-3′). Relative quantification was carried out and calibrated to vector control samples (20).

DNA-binding reactions were performed as previously described (21) using 10 μg of nuclear extracts and 4 ng of ³²P- labeled double-stranded oligonucleotide probe. Reactions were resolved by electrophoresis on 4% (w/v) 0.5× Tris-borate EDTA acrylamide gels at 4°C, after which the gels were fixed and dried and complexes visualized by autoradiography. The following double stranded oligonucleotides were used as probes: Sgpp1 site A, 5′-cctttgcagaccacagctgt-3′; Sgpp1 Site Amut, 5′-cctttgcatagcacagctgt-3′. Where indicated, a 50× excess of cold competitor DNA or 2 μL monoclonal Runx2 antiserum (MWG #D130-3) or Runx3 antiserum (a kind gift from Y. Ito, Institute of Molecular and Cell Biology, Singapore) were added to the reaction 15 minutes before the addition of the probe.

Transcription assays were performed as described (22) using 2 × 10⁵ cells per 60-mm-diameter tissue culture dish. The assays shown are representative of at least three independent experiments carried out in duplicate. The luciferase activity was normalized to the protein concentration (Bio-Rad protein assay) present in each extract. pGL3Basic luciferase reporter (Promega) was used in all transcription assays. Expression of luciferase activity was driven by a 362-bp fragment amplified from the murine Sgpp1 promoter spanning two consensus Runx binding sites (−133 to −495). Site-directed mutagenesis (QuikChange II Site-Directed Mutagenesis kit, Stratagene) was used to generate bp mutations in the distal site (gaccacag mutated to tagcacag) or in both sites (gaccacag mutated to tagcacag, and caccgctg mutated to cacgcctg) to determine the efficacy of Runx-mediated Sgpp1 repression. pGL3control (Promega) that contains the SV40 promoter and enhancer driving the expression of the luciferase gene was included in all individual experiments to confirm similar levels of transfection efficiency between experiments.

Fibroblast cell pellets (∼2 × 10⁵ cells) containing 15 ng C17-ceramide as an internal standard were extracted using a modified Folch method [extraction with 4 mL chloroform/2 mL methanol/2 mL 0.88% NaCl for each sample, followed by the extraction of upper phase with 3 mL of synthetic lower phase of chloroform/methanol/0.88% NaCl 2:1:1. The combined lower phases were dried under vacuum and redissolved in 50 μL chloroform/methanol 1:1. Two microliters were injected for liquid chromatography/mass spectrometry (LC-MS) analysis]. Cell culture medium (4 mL from ∼2 × 10⁵ cells) containing 30 ng C17-sphingosine-1-phosphate as internal standard, concentrated to 1 mL at room temperature under vacuum, was extracted with 3 × 1 mL n-butanol, combined, concentrated, and redissolved in 40 μL methanol with 2 μL injected for LC-MS analysis using a Shimadzu IT-TOF liquid chromatography tandem mass spectrometry system. In detail, lipid classes of the Folch extract were separated on a normal phase silica gel column (100Å, 4 μm, 150 × 1 mm, MicoSolv Technology) using a hexane/dichloromethane/chloroform/methanol/acetanitrile/water/ethylamine gradient based on the polarity of head group. Accurate mass (with mass accuracy ∼5 ppm) and tandem MS were used for lipid molecular species identification and quantification. Lipid identity was further confirmed by reference to appropriate lipid standards. Mass spectrometer operation conditions were as follows: ESI interface voltage +4.0 kv for positive ESI and −3.5 kv for negative ESI, heat block temperature of 210^oÑ, nebulizing gas flow of 1.0 L/min, and CDL temperature of 200^oC, with drying gas on. All the solvents used for lipid extraction and LC-MS analysis were LC-MS grade from Fisher Scientific.

Preparation of whole-cell protein extracts was performed as previously described (4). Samples equivalent to 50 μg of protein (Bio-Rad protein assay) were resolved on 10% SDS-polyacrylamide gels and transferred to enhanced chemiluminescence (Amersham) nitrocellulose membranes. The antibodies used were α Phospho-p38 mitogen-activated protein kinase (MAPK) 9211 (Cell Signaling Technology) and α p38 MAPK (C20) sc535 (Santa Cruz Biotechnology).

Functional c-Jun-NH₂-kinase (JNK) activity was measured using a non–radioactive stress–activated protein kinase (SAPK)/JNK Assay kit #9810 (Cell Signaling Technology) according to the manufacturer's instructions. Parallel lysates all contained an equal concentration of protein (Bio-Rad protein assay). Immunoblotting was performed with an anti–phospho-specific cJun (Ser63) antibody as supplied by the kit. The blot was reprobed against α GST(Z5) sc459 (Santa Cruz Biotechnology) to ensure equal loading of c-Jun (1-89) glutathione S-transferase (GST) fusion protein beads.

Activated caspase-3 was detected using an activated caspase-3 assay kit (#550914, BD Biosciences) according to the manufacturer's instructions; 10,000 events per sample were analyzed on an Epics XL flow cytometer (Beckman Coulter) using the Expo32 software.

In a previous study, we compared the global transcriptional changes in immortalized fibroblasts expressing ectopic Runx1, 2, or 3. In contrast to primary fibroblasts, which enter a state of senescence-like growth arrest in response to Runx (14, 20, 23, 24), these cells undergo a characteristic phenotypic change resembling mesenchymal to epithelial transition and display greatly enhanced survival in response to stresses including medium exhaustion (4). Gene expression microarray analysis revealed a largely overlapping set of target genes with no evidence of genes that were oppositely regulated by each of the three family members. In a common Runx signature set defined by the 50 most significantly regulated genes, we found a preponderance of genes encoding cell surface molecules or secreted factors, but few candidates with an annotated role in control of apoptosis. However, gene ontology analysis indicated a bias toward a small set of genes encoding enzymes involved in sphingolipid metabolism that were of potential significance in this regard (4).

In this study, we have examined Runx regulation of the three sphingolipid enzymes and assessed its potential significance for Runx-mediated survival and oncogenesis. Figure 1A shows the biochemical steps catalyzed by the respective enzymes. The right-hand side of the figure shows the fold change in mRNA levels induced by constitutive expression of Runx1, Runx2, or Runx3 in 3T3 fibroblasts as measured by the Affymetrix gene expression microarray (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE11732) and validated by qt-RT-PCR. UDP-glucose ceramide glycosyltransferase (Ugcg) and ST3 β-galactoside α-2,3-sialyltransferase 5 (St3gal5/Siat9/GM3 synthase) are both markedly upregulated by Runx gene expression, whereas sphingosine-1-phosphate phosphatase 1 (Sgpp1) is downregulated. Fig. 1B (bottom) shows a kinetic analysis of the mRNA levels after activation of an inducible RUNX1-ER construct. The fact that the observed regulation happens rapidly supports the hypothesis that all three genes act as direct transcriptional targets for the Runx genes, with Ugcg and St3gal5 as targets for activation, whereas Sgpp1 is repressed. As a further test for direct regulation, we examined the requirement for protein synthesis by inducing RUNX1-ER in the presence of cycloheximide. Although interpretation of these experiments was complicated by the nonspecific effects of cycloheximide on the stability of some transcripts, it was clear that Ugcg induction and Sgpp1 repression were insensitive to cycloheximide inhibition. St3gal5 induction was preserved at very early time points but was lost after 6 hours due to the nonspecific stabilization of transcripts by cycloheximide (Supplementary Fig. S1).

Screening of the promoter regions of the three genes revealed consensus Runx binding sites close to the transcriptional start site in each case. St3gal5 has three discrete sites (−602, −1,151, and −1834), whereas Ugcg has one site (−345) and Sgpp1 has two closely linked sites (−416 and −348). The Sgpp1 promoter region is the most highly conserved sequence across mammalian species and also contains conserved E box motifs next to the proximal site (B). As shown in Fig. 2A, chromatin immunoprecipitation confirmed Runx binding at each of these sites. Runx2-expressing cells were used in this instance to take advantage of antibodies suitable for use under chromatin immunoprecipitation conditions.

In view of the evidence that Sgpp1 plays a central role in the regulation of the ceramide-sphingosine 1 phosphate (S1P) rheostat and cell survival, we focused further efforts on validating the regulatory consequences of Runx binding. We confirmed that Runx proteins from nuclear extracts bound specifically to both site A and site B in electrophoretic mobility shift assay (data not shown; Fig. 2B). A luciferase expression construct with a basal promoter under control of the Sgpp1 regulatory element (−133 to −495) showed negative regulation in the presence of ectopic Runx, which was reduced by mutation of site A and abolished in the double mutant (Fig. 2C). Basal levels of the double mutant were similar to those of the wild-type promoter in the absence of ectopic Runx expression, suggesting that steady-state levels of Runx in fibroblasts are too low to cause significant repression.

To explore the biochemical consequences of altered transcription of genes encoding sphingolipid enzymes, we examined the ceramide/S1P profiles by MS in the presence and absence of ectopic Runx. Levels of extracellular S1P were found to be essentially undetectable in serum-free cultures of control cells. However, although the data were somewhat variable, we were able to detect S1P in the culture medium of cells expressing Runx with values ranging from 0 to 0.88 ng S1P/5 mL culture medium of 2 × 10⁵ cells cultured serum free. In contrast, we consistently observed a marked reduction in the level of several long-chain ceramides in the presence of ectopic Runx (Fig. 3A). Ceramides 16:0, 24:1, and 24:0 are typical of the salvage pathway of ceramide synthesis, suggesting that Runx expression opposes sphingosine recycling through this route (17). The most dramatic effect observed was on the levels of 16.0 ceramide (Fig. 3A), which has previously been reported to respond to Sgpp1 expression and is implicated as a critical component of ceramide-induced apoptosis (17, 25, 26).

To investigate the consequences of long-chain ceramide depletion for Runx-dependent cell survival, we examined the functional activity of p38^MAPK and JNK, which play central roles in stress-induced cell death and apoptotic signaling by ceramide (27–29). For this purpose, NIH3T3 fibroblasts expressing ectopic Runx1 or control vector were grown under conditions of medium exhaustion, and adherent cells were harvested every 1 to 3 days for analysis of phospho-p38^MAPK expression and JNK activity. Consistent with our previous data (4), expression of ectopic Runx conferred protection against stress-induced cell death under these growth conditions (Fig. 3B). Furthermore, Runx-dependent cell survival correlated with reduced activation of the p38^MAPK and JNK signaling pathways as measured by the accumulation of phosphorylated p38^MAPK and c-Jun, respectively (Fig. 3C and D).

TNF-α has been reported to induce cell death through an accumulation of intracellular ceramide and downstream activation of the JNK signaling pathway (28). To determine whether this response is modified by ectopic Runx1, we examined the effects of TNF-α on the survival of vector control and Runx1-expressing NIH3T3 fibroblasts. Cycloheximide was included in short-term assays to suppress the synthesis of short-lived antiapoptotic proteins that can reduce the effects of TNF-α (30). As shown in Fig. 4A, morphologic differences were observed as early as 2 to 4 hours after TNF-α and cycloheximide treatment, with vector control cells rounding up and detaching from the substratum far more readily than their Runx1-expressing counterparts. Quantitation of adherent and nonadherent cells by trypan blue exclusion confirmed greater survival in the presence of ectopic Runx1 (Fig. 4B). The effect was exaggerated after 5 hours of exposure to TNF-α and cycloheximide (Fig. 4B) but lost by 24 hours when widespread cell death was observed throughout (data not shown). To determine whether Runx1 was protecting against apoptosis, fluorescence-activated cell sorting (FACS) analysis was performed on caspase-3–labeled cells grown in the presence and absence of TNF-α for longer time periods. A highly significant difference between Runx and control cultures was observed from the earliest time point up to 48 hours (Fig. 4C), with Runx-expressing cells showing a marked resistance to induction of caspase-3 over the time course.

These data suggest that modulation of sphingolipid enzymes by ectopic Runx resets the sphingolipid rheostat against intracellular ceramide-induced apoptosis and in favor of cell survival.

The reduction of intracellular ceramide in Runx-expressing cells may be due to the cumulative action of the three target enzymes leading to the synthesis of glycosylated forms or gangliosides (Ugcg, St3gal5) or to the accumulation of S1P (Fig. 5). S1P functions as a lipid signal that can promote proliferation and suppress ceramide-mediated apoptosis. S1P is actively transported outside the cell where it signals through a family of G protein–coupled receptors. In addition, S1P seems to be able to act through as yet uncharacterized intracellular receptors (31). Although the detection of extracellular S1P was variable in Runx-expressing fibroblasts, it is probable that this pathway is constitutively activated because other pathways of S1P removal are found in cells and S1P was essentially undetectable in control cultures. To explore this hypothesis, we tested whether Runx-expressing cells retained sensitivity to growth regulation by exogenous S1P. NIH3T3 fibroblasts expressing ectopic Runx1 or control vector were grown to confluence and cultured under conditions of medium exhaustion for 8 days in the presence or absence of increasing concentrations of S1P. Viability counts were then performed on detached cells in the supernatant and the adherent cultures harvested to determine the phosphorylation status of p38^MAPK.

As shown in Fig. 6A, exogenous S1P reduced cell death in a dose-dependent manner in both cell backgrounds, but the effect was more pronounced in the absence of ectopic Runx1, supporting the proposal that S1P signaling was already activated by elevated S1P generation in the Runx-expressing fibroblasts. Cell viability counts confirmed this trend, with the difference between Runx-expressing and control cultures reduced to only 3% in the presence of 1 × 10⁻⁶ mol/L S1P (Fig. 6B). These data support previous studies in NIH3T3 fibroblasts showing a reduction in stress-induced apoptosis in the presence of exogenous S1P (32). Furthermore, the improved survival of control cultures correlated with reduced expression levels of phosphorylated p38^MAPK (Fig. 6C). At higher doses of S1P, expression levels of phospho p38^MAPK were comparable with those in the Runx-expressing cultures, which were relatively refractory to S1P (Fig. 6C). These data suggest that although p38^MAPK inhibition correlates with improved cell survival of control cultures, ectopic expression of Runx has already saturated the relevant survival pathway.

Overexpression of Runx genes has context-specific effects, inducing growth arrest or senescence in most primary cell types, but promoting proliferation and/or survival in cells expressing collaborating oncogenes or mutated tumor suppressor pathways (1, 14, 24, 33). In NIH3T3 fibroblasts, which lack Cdkn2a (Ink4a/Arf) function, overexpression of Runx potently induces survival under stress conditions. Our previous exploration of the likely effector genes in this phenotype revealed a paucity of candidates with established roles in apoptosis or survival, with the notable exception of a small set of genes involved in sphingolipid metabolism that were overrepresented as a gene ontology subset (4). In this study, we have shown that the Runx genes can directly regulate three genes encoding enzymes with central roles in sphingolipid pathways (Fig. 5). Of particular note is the ability of all three Runx genes to repress expression of Sgpp1, a key player in the sphingosine rheostat that governs the interchange between proapoptotic ceramides (34) and S1P, a well-established ligand in survival signaling (19). Moreover, ectopic Runx expression was shown to drive a major reduction in the level of specific intracellular ceramide species, a finding that was associated with reduced induction of stress-associated p38^MAPK and JNK activity. The selective reduction in a subset of ceramide species points to these lipids as being specific substrates for Ugcg and also suggests that these particular ceramides signal through specific target proteins to mediate proapoptotic effects. The selective action of particular species within a lipid class has been previously recognized for diacylglycerol and phosphatidic acid (35). Resistance to death induced by TNF-α, which operates through ceramide, was also confirmed in Runx-expressing cells, whereas the Runx-associated survival phenotype could be at least partially recapitulated by the addition of exogenous S1P. These results suggest that modulation of sphingolipid metabolism underlies the oncogenic and prosurvival effects of ectopic Runx expression in multiple contexts (4, 11, 15) and indicates the possibility of a wider role for Runx in cancer chemoresistance.

Published studies provide strong support for the hypothesis that the three novel Runx target genes we have described play a significant role in the observed survival phenotype. For example, RNAi knockdown of Sgpp1 expression was shown to promote cell survival with concomitant accumulation of intracellular and extracellular S1P, whereas ectopic expression of Sgpp1 was reported to drive the conversion of S1P to long-chain ceramides and a subsequent apoptotic response (32). Ectopic expression of Ugcg has been reported to rescue cells from ceramide-induced apoptosis (16). Furthermore, multidrug resistance of breast cancer cell lines has been associated with the shunting of ceramide into a glucosyl ceramide sink and can be reversed by small interfering RNA or oligonucleotide blockade of Ugcg (36). The effects of upregulation of St3gal5 are less readily predicted, as high levels of GM3 are associated with the induction of apoptosis in some cancer cells (18), whereas reintroduction of St3gal5 into lines that fail to synthesize GM3 was shown to enhance their ability to form colonies in soft agar and increase their resistance to stress-induced apoptosis (37). We suggest that the outcome may reflect the steady-state levels of GM3 in different cellular backgrounds. In this regard, it is notable that the upregulation of Ugcg by Runx is much more marked than St3gal5, and that glycosylation of ceramide is also a key step in the synthesis of cerebrosides and gangliosides other than GM3.

Improved survival of Runx-expressing fibroblasts also correlated with reduced levels of intracellular long-chain ceramides. The induction of apoptosis through ceramide production is complex, involving both direct and indirect modes of action and multiple signaling pathways. Mitochondrial damage and caspase release is driven by the formation of ceramide channels in the mitochondrial membrane (38) but also through the activation of stress-activated kinases such as p38^MAPK, which further activate proapoptotic Bax and Bak, and inhibit antiapoptotic Bcl2 and Bcl-XL (39). It has been suggested that the solubility of these target molecules enables ceramide at nonmitochondrial locations such as the endoplasmic reticulum membranes to participate in the apoptotic process (40). Notably, we observed reduced steady-state p38^MAPK phosphorylation and JNK activity in stressed cells expressing ectopic Runx, and both of these kinase pathways have been reported to be induced by intracellular ceramide (29, 39). Further evidence of a role for sphingolipid-mediated changes in the Runx survival phenotype was provided by the greater resistance of Runx-expressing cells to TNF-α–induced apoptosis, which has also been reported to operate through ceramide and JNK signaling pathways (29).

It is notable that we were able to partially recapitulate the survival phenotype of Runx-expressing cells through exogenous administration of S1P to control cells. S1P can mediate survival by binding to a series of cell surface receptors (reviewed in ref. 41) but can also act by a less well-characterized intracellular route (17). The variable detection of free S1P in the medium of Runx-expressing cells could suggest that the latter route is operative here. Nevertheless, it seems that S1P pathways are fully engaged in these Runx-expressing cells, as exogenous S1P caused no additional survival benefit in this context. Sgpp1 is not the only plasma membrane–located enzyme able to degrade S1P, there being a family of lipid phosphate phosphatases able to hydrolyse a range of signaling lipids including S1P and lysophosphatidate (42). The lipid phosphate phosphatases seem to be regulated by substrate availability; thus, although a reduction in Sgpp1 could reduce the degradation of S1P in the proximity of the S1P receptor following its release from the cell, it will not necessarily induce sustained elevation of medium S1P concentration. In support of this, it was reported that reduced expression levels of lipid phosphate phosphatase 1 in ovarian cancer cells was not accompanied by a corresponding increase in extracellular lysophosphatidate (43).

The involvement of sphingolipid enzymes in cancer and chemoresistance (27, 44) suggests a wider role for Runx in these phenomena. Our results establish a link between previous reports of the Runx genes as targets for retroviral activation in an in vivo model of acquired imatinib resistance in CML (45) and studies showing that imatinib resistance can be conferred through changes in the ceramide-S1P rheostat (46). Moreover, the ability of Runx to confer metastatic potential may be mediated by the ability to promote close adhesion (4, 14) and sphingolipid-mediated cell survival under adverse conditions in vivo. Finally, the Runx genes have been widely studied as essential players in lineage-specific differentiation in which they can act as both positive and negative regulators of proliferation according to context (47–49). The possibility that these regulatory steps are mediated through sphingolipid metabolism merits further investigation.

No potential conflicts of interest were disclosed.

We thank Dr. S. Wotton for helpful comments and discussions.

Grant Support: Cancer Research UK and Leukaemia and Lymphoma Research to the Molecular Oncology Laboratory. Lipidomic analyses were performed at the Babraham Institute and supported by grants from Cancer Research UK and the Biotechnology and Biological Sciences Research Council (M.J.O. Wakelam).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.