Runx1 Orchestrates Sphingolipid Metabolism and Glucocorticoid Resistance in Lymphomagenesis

ABSTRACT The three‐membered RUNX gene family includes RUNX1, a major mutational target in human leukemias, and displays hallmarks of both tumor suppressors and oncogenes. In mouse models, the Runx genes appear to act as conditional oncogenes, as ectopic expression is growth suppressive in normal cells but drives lymphoma development potently when combined with over‐expressed Myc or loss of p53. Clues to underlying mechanisms emerged previously from murine fibroblasts where ectopic expression of any of the Runx genes promotes survival through direct and indirect regulation of key enzymes in sphingolipid metabolism associated with a shift in the “sphingolipid rheostat” from ceramide to sphingosine‐1‐phosphate (S1P). Testing of this relationship in lymphoma cells was therefore a high priority. We find that ectopic expression of Runx1 in lymphoma cells consistently perturbs the sphingolipid rheostat, whereas an essential physiological role for Runx1 is revealed by reduced S1P levels in normal spleen after partial Cre‐mediated excision. Furthermore, we show that ectopic Runx1 expression confers increased resistance of lymphoma cells to glucocorticoid‐mediated apoptosis, and elucidate the mechanism of cross‐talk between glucocorticoid and sphingolipid metabolism through Sgpp1. Dexamethasone potently induces expression of Sgpp1 in T‐lymphoma cells and drives cell death which is reduced by partial knockdown of Sgpp1 with shRNA or direct transcriptional repression of Sgpp1 by ectopic Runx1. Together these data show that Runx1 plays a role in regulating the sphingolipid rheostat in normal development and that perturbation of this cell fate regulator contributes to Runx‐driven lymphomagenesis. J. Cell. Biochem. 118: 1432–1441, 2017. © 2016 The Authors. Journal of Cellular Biochemistry Published by Wiley Periodicals, Inc.

R UNX1 (AML1) is one of the most frequently affected genes in human leukemias where it is subject to mutational processes including chromosomal translocation, gene amplification, and point mutation [Osato, 2004;De Braekeleer et al., 2009]. In acute myeloid leukemias, diverse chromosomal breakage and fusion events result in truncation of the RUNX1 gene product and deletion or replacement of the C-terminal transactivation domain with heterologous sequences. Point mutation with apparent loss of function or dominant negative activity is a feature of immature M0 AML  suggesting that RUNX1 may act as a tumor suppressor in this lineage and that the oncogenic fusions operate primarily by a dominant negative mechanism [De Braekeleer et al., 2009]. A tumor suppressor role for Runx1 in myeloid leukemia is also suggested by mouse models where Runx1 deletion is induced, for example, in Flt3-ITD expressing mice [Mead et al., 2013]. However, the notion that RUNX1 is simply a suppressor whose loss of functions confers an automatic growth advantage is challenged by more recent observations that human leukemia cells bearing the common RUNX1-ETO fusion cannot tolerate loss of the remaining wild-type RUNX1 allele [Ben-Ami et al., 2013]. Moreover, although the frequent TEL-RUNX1 fusion in childhood B-ALL is often complemented by loss of the wild-type TEL allele, the unaffected RUNX1 allele is generally intact and is in fact more likely to show copy number gain [Niini et al., 2000].
Early evidence from mouse models showed that all three members of the Runx family can act as targets for transcriptional activation in retrovirus-induced lymphomas [Stewart et al., 1997[Stewart et al., , 2002Wotton et al., 2002]. Moreover, transgenic over-expression leads to predisposition to lymphoma and is strongly synergistic with other oncogenes (Myc,Myb,Pim1) or loss of p53 tumor suppressor function [Vaillant et al., 1999;Blyth et al., 2001Blyth et al., , 2006Shimizu et al., 2013;Huser et al., 2014]. Although oncogenic activity has been most extensively documented for Runx2, ectopic expression of Runx1 has been shown to collaborate strongly with Myc in T and B-cell lymphomagenesis [Blyth et al., 2009]. Although a small proportion of primary human T-cell acute lymphoblastic leukemias (ALL) carry point mutations in RUNX1 [Grossmann et al., 2011], evidence for an oncogenic role for this gene in childhood B-ALL is suggested by the high expression of RUNX1 [Niini et al., 2000] and high level amplification in a poor prognosis subset [Robinson et al., 2003]. These findings reinforce the hypothesis that the Runx gene family can operate as tumor suppressors or as oncogenes depending on the context in which misregulation occurs.
Clues to the contextual factors that influence the outcome of Runx gain or loss have come from studies in mouse and human fibroblasts where integrity of the p53 pathway determines the response to ectopic Runx expression. Normal primary fibroblasts undergo senescence-like growth arrest in response to ectopic Runx expression, while cells in which the p53 pathway is disabled instead display enhanced survival and oncogenicity [Wotton et al., 2004;Kilbey et al., 2007;Wolyniec et al., 2009]. Moreover, ectopic Runx expression in immortalized Ink4a/Arf null fibroblasts revealed a novel link between oncogenic transcription factors and sphingolipid metabolism [Wotton et al., 2008]. Several enzymes involved in sphingolipid metabolism (Sgpp1, Ugcg, Siat9) were identified as direct Runx targets and regulated in a manner consistent with elevated Sphingosine-1-phosphate (S1P) and reduced ceramide levels, changes which were also confirmed at the biochemical level [Wotton et al., 2008;Kilbey et al., 2010].
Bioactive lipids mediate many essential roles in normal development and disease. S1P is among the best characterized and is recognized as a regulator of key processes in cancer development including survival, proliferation, and angiogenesis [Takabe et al., 2008]. The precursors of S1P, sphingosine, and ceramide are mediators of cell growth arrest and apoptosis [Zhang et al., 1991;Olivera and Spiegel, 1993;Cuvillier et al., 1996]. Regulation of the balance between these interconvertible lipids forms the basis of the sphingolipid rheostat and the means by which ceramide, sphingosine, and S1P control cell growth and survival in response to environmental and cellular cues. Multiple enzymes control the rheostat and the biochemical processes that generate or degrade its key components [Van Brocklyn and Williams, 2012]. However, despite the wealth of literature on the biology and biochemistry of these processes, links between sphingolipid signaling and cancer gene function, have emerged only recently [Kilbey et al., 2010;Heffernan-Stroud et al., 2012;Burns et al., 2013;Kajiwara et al., 2014]. The evidence of Runx regulation of sphingolipid metabolism in fibroblast survival stimulated our interest in its relevance to lymphomagenesis where there is strong evidence for an oncogenic role. We examined cell lines from p53 null and Em-Myc lymphomas where ectopic Runx expression has been shown to collaborate strongly in oncogenesis [Blyth et al., 2001[Blyth et al., , 2009. Our results demonstrate a consistent role for Runx1 in control of sphingolipid metabolism in lymphomas as well as in normal development. Moreover, we show that Runx1 regulates glucocorticoid sensitivity in lymphoma cells and identify an established direct target gene, Sgpp1, as an integrator of sphingolipid metabolism and responses to glucocorticoids.

GENERATION OF TRANSGENIC MICE AND CELL LINES
MxLoxP mice (Mx1Cre þ Runx1 fl/fl ) were created as previously described [Borland et al., 2016]. Mx1Cre positive and negative offspring were treated at seven weeks of age with six injections of 600 mg polyinosinic polycytidylic acid (pIpC), or an equivalent volume of PBS, on alternate days and sacrificed for tissue samples two days after the final pIpC treatment.
Animal protocols used in this work were evaluated and approved by the University of Glasgow Ethics and Welfare Committee and were carried out under Home Office License (approval granted September 2012, license number PPL 60/4408) as governed by the Animal Scientific Procedures Act, 1986 and EU Directive 2010.

CELL CULTURE AND TRANSFECTIONS
Cell lines were established from p53 À/À thymic lymphomas or EmMyc splenic lymphomas as previously described [Borland et al., 2016]. Transfections were carried out using Superfect Transfection Reagent (Qiagen, Crawley, UK) according to the manufacturer's instructions. Conditions for transfection have been previously described (21). The 3s cell line was derived from an Mx1Cre þ Runx1 fl/fl EmMyc þ p53 þ/À tumor treated with and without 10U/ml IFNb (R&D systems) for 8 h to excise Runx1. For growth analyses cells were plated at 5 Â 10 5 /ml (T cell lymphoma) or 2.5 Â 10 5 /ml (B cell lymphoma) in a 12-well plate in triplicate. Live/dead cell counts were carried out using a hemocytometer and Trypan blue as a vital indicator.

PCR GENOTYPING PROTOCOLS AND PRIMERS
Tissue DNA samples were prepared using the Nucleon Genomic DNA Extraction Kit. Runx1 primers for WT, floxed and deleted alleles, and PCR cycling conditions are as described in Chen et al. [2009]. The PCR products were resolved on a 2.0% TAE agarose gel. Control DNAs includes Balb/c kidney (WT Runx1) and Mx1Cre Runx1 fl/fl spleen (non-excised floxed Runx1, excised Runx1).
QUANTITATIVE REAL-TIME PCR Runx1-expressing and control lymphocytes were plated in triplicate on 6 well plates at 5 Â 10 6 /well in the presence and absence of 1.0 mM dexamethasone for 6 h. RNA extraction and cDNA preparation were performed as described (21). For quantitative real-time PCR, 12.5 ng aliquots of cDNA were amplified in triplicate using primers for murine endogenous control Hprt or primers for murine Sgpp1, Siat9, Nr3c1, Runx1 (Qiagen QuantiTect Primer Assays) or Ugcg (779F 5 0 tttgctcagtacattgctgaagatta 3 0 and 861R 5 0 acttgagtagacattgaaaacctccaa 3 0 ). Relative quantification was carried out and calibrated to vector control samples (18).

LENTIVIRUS PRODUCTION AND GENE KNOCKDOWN
SMARTvector 2.0 lentiviral shSgpp1 and shNC non-coding control particles were purchased from Thermo Scientific (2 Â 10 8 particles/ml). shGAPDH lentiviral particles were included as a positive control for transduction efficiency. Virus supernatants were thawed on the day of transfection and centrifuged onto retronectin-coated plates at 3220 g for 1 h at 4°C (4 Â 10 5 particles/well of a 24-well plate). Supernatants were aspirated from the plates and replaced with fresh supernatant (2 Â 10 5 particles/well) and spun as before. The 10 5 cells were added in an equal volume per well and incubated for 7 h with 8 mg/ml polybrene. The medium was replaced with fresh medium overnight and then removed. A second infection with 2 Â 10 5 virus particles in an equal volume of fresh medium containing 8 mg/ml polybrene was performed for a further 7 h before the cells were dissociated from the retronectin and incubated in fresh medium. After 40 h, the medium was replaced with media containing 2 mg/ml puromycin to allow for selection of virally transduced cells. For qt-RT-PCR and viability counts, puromycin-resistant cells were plated as described in the presence and absence of 1.0mM dexamethasone.

MASS SPECTROMETRY: CERAMIDE AND S1P EXTRACTION AND ANALYSIS FROM CELLS
Ceramide and S1P extraction was performed with a modified Folch approach. Briefly, cells from pellets or pulverized tissues were washed twice with cold PBS and re-suspended in 1.5ml methanol. After addition of 50 ng C17-ceramide and 15 ng C17-S1P as internal standards, 1.5 ml of 0.88% NaCl and 3 ml chloroform were added, the mixture vortexed for 20 s and sonicated for 2 min. Then the sample was centrifuged at 1000g at 4C for 15 mins. The lower phase was collected. The upper phase was re-extracted with 3 ml synthetic lower phase (mix chloroform/methanol/0.88% NaCl at volume ratio of 2:1:1, after phase separation, take the lower phase as synthetic lower phase). The lower phases were combined, dried with SpeedVac (Thermo) and re-dissolved in 60 ml chloroform. A total of 7 ml were injected for LCMS/MS analysis. Another 7 ml of sample was used for S1P analysis as described below with 4000QTRAP (AB Sciex). For LC/MS/MS analysis of ceramide, a Thermo Orbitrap Elite system (Thermo Fisher) was hyphenated with a five-channel online degasser, four-pump, column oven, and autosampler with cooler Shimadzu Prominence HPLC system. Lipid classes were separated on a normal phase silica gel column (2.1 Â 150 mm, 4 micro, MicoSolv Technology) with hexane/dichloromethane/chloroform/ methanol/acetanitrile/water/ethylamine solvent gradient based on the polarity of head group. High resolution (240 k at m/z 400)/accurate mass (with mass accuracy <5 ppm) and tandem MS (CID fragmentation) were used for molecular species identification and quantification and further confirmed by reference to appropriate lipids standards. Orbitrap Elite mass spectrometer operation conditions: for ceramide analysis, heated ESI Source in negative ESI mode, heater temperature: 325°C, sheath gas flow rate (arb): 45, aux gas flow rate (arb): 10, sweep gas flow rate (arb): 0, I spray voltage: 3.0 kV, capillary temperature: 375°C, S-Lens RF level: 70%. Orbitrap mass analyzer was operated as SIM scan mode with mass range: m/z 510-740, mass resolution: 240 k at m/z 400. Ion trap mass analyzer was operated as dependent scan in CID mode at normal scan speed.

MASS SPECTROMETRY: S1P EXTRACTION AND ANALYSIS FROM CELL CULTURE MEDIA
A total of 800 ml cell culture media, 200 ml 0.88% NaCl, and 2 ml n-butanol were mixed in pre-cleaned 1.5 ml Eppendorf tubes, with 15 ng C17-S1P as internal standard. After vortexing for 20 s at room temperature and sonication in an ice-cold water bath for 3 min, the tubes were centrifuged at 1500g 4°C for 10 min collecting the upper organic phase. The lower phase was reextracted with 2 ml n-butanol and the organic extractions combined. The lipid extract was dried under vacuum at room temperature with SpeedVac (Thermo), re-dissolved in 60 ml chloroform/methanol/water 2:5:1 (v/v/v) solvent mixture. A total of 7 ml were injected onto a Gemini-NX-3u C18 2.0 Â 150 mm column (Phenomenex) for LC-MS/MS analysis of S1P with Shimadzu Prominence HPLC system hyphenated with AB Sciex 4000 QTRAP mass spectrometer. The S1P separation was carried out with a binary gradient elution at flow rate of 0.25 ml/min with column oven set at 25°C. Mobile phase A: 5% acetonitrile with 1% formic acid and 5 mM ammonium formate; Mobile phase B: 95% acetonitrile with 1% formic acid and 5 mM ammonium formate. The gradient was run from 10% B to 100% B in 10 min. A total of 4000 QTRAP mass spectrometer source/gas parameters: curtain gas: 25; collision gas: high; ionspray voltage: À4500; temperature: 550; ion source gas 1: 30; ion source gas 2: 45; interface heater: on. The following Compound Parameters were used for Multiple Reaction Monitoring (MRM) analysis of each molecular species of S1P. Both Q1 and Q3 mass were set up at unit resolution (Table I).

WESTERN BLOTTING AND ANTIBODIES
Preparation of whole cell protein extracts was performed as described previously (21). Samples equivalent to 30 mg of protein (Bio-Rad protein assay) were resolved on 8% SDS polyacrylamide gels and transferred to enhanced chemiluminescence nitrocellulose membranes (ECL; Amersham, Little Chalfont, UK). The antibodies used were a Runx1 #8229 (Cell Signalling Technologies) and a actin sc1616 (Santa Cruz).

STATISTICAL ANALYSIS
Graphs were generated with Excel as mean values and error bars relate to standard deviations. All statistical comparisons were performed using the Student's t test. Unless stated a significance value ¼ <0.05 is denoted by ( Ã ) and of ¼ <0.01 by ( ÃÃ ).

Runx1 PROMOTES S1P RELEASE FROM LYMPHOMA CELLS
We previously reported that ectopic Runx1 depletes intracellular long chain ceramides and opposes ceramide-induced apoptosis in murine fibroblasts [Kilbey et al., 2010]. To determine whether the sphingolipid rheostat is similarly perturbed in a context where Runx1 has established in vivo oncogenic function [Blyth et al., 2001;Shimizu et al., 2013], we introduced Runx1 into two independently derived p53-null T cell lymphoma lines (Fig. 1A). Vector-driven expression led to increased levels of Runx1 protein at 4.3-and 4.5-fold over the endogenous levels in p/m97 and p53/184 cells, respectively (Fig. 1B). Mass spectrometric analysis of ceramide/S1P revealed modest but consistent reductions in the levels of the most abundantly expressed long chain intracellular ceramides ( Fig. 1C-hashed bars) were detected at much lower levels but a significant reduction was preserved when these were included in the total (P ¼ <0.01). A more dramatic effect was observed on extracellular C18-S1P which was induced three to fivefold in the presence of ectopic Runx1 (Fig. 1D). The level of induction was further stimulated by serum (FCS) to activate sphingosine kinase and the conversion of sphingosine to S1P. Indeed by 15 min, seruminduced levels of C18-S1P in the presence of ectopic Runx1 (pink bar) exceeded the sum of ectopic Runx1 (red bar) or FCS (gray bar) alone suggesting that ectopic Runx1 and FCS have additive effects in T lymphoma cells and that absolute levels of Runx1 modulate C18-S1P production in vitro.

ENFORCED DELETION OF Runx1 IMPAIRS S1P RELEASE IN VIVO
To determine whether Runx1 plays an essential physiological role in sphingolipid metabolism in vivo we attempted to delete Runx1 in otherwise genetically normal Runx1 fl/fl Mx1Cre þ conditional knockout mice. A previously validated direct PCR assay was used to measure the proportion of excised to non-excised allele in the tissue [Borland et al., 2016]. As reported, most normal mouse tissues resist enforced deletion of Runx1 [Tober et al., 2013], suggesting that this gene plays an essential maintenance role in many mature cell types. Nevertheless, we were able to achieve partial Runx1 ablation in spleen tissues from 60-day-old Runx1 fl/fl Mx1Cre þ transgenic animals where pIpC induction of Cre recombinase led to Runx1 excision rates of $50-55%. Background excision in Mx1Cre þ mice treated with PBS alone was in the order of 40% presumably due to endogenous interferon activity ( Fig. 2A). Runx1 transcript levels as measured by qt-RT-PCR closely paralleled DNA excision rates in Mx1Cre þ spleens (pale green vs. dark green bars) and were significantly different from control Runx1 fl/fl mice lacking Mx1Cre that showed no response to pIpC treatment ( Fig. 2B; gray vs. black bars). Despite these substantial excision rates, Runx1-excised mice showed surprisingly little perturbation in spleen cell populations [Borland et al., 2016]. Parallel tissue samples were processed for biochemical analysis of C18-S1P by mass spectrometry and revealed a strong correlation between loss of Runx1 and C18-S1P in vivo (Fig. 2C). Splenic tissue levels of C18-S1P were moderately reduced by background excision alone (gray bars vs. pale green bars), but significant depletion was observed in the presence of pIpC (black bars vs. dark green bars). To control for nonspecific effects of the Cre recombinase spleen tissue samples were prepared from Runx1 wt/wt -Mx1Cre þ and Runx1 wt/wt -Mx1Cre À transgenic mice treated with and without pIpC as described previously. In this case, no significant differences in C18-S1P were recorded between the different experimental groups (data not shown), suggesting that loss of Runx1 was responsible for the reduced levels of C18-S1P in pIpC-treated Runx1 fl/fl Mx1Cre þ mice. (D) C18-S1P was analyzed from conditioned media collected from pBabePuro vector control and pBabeRunx1-expressing T lymphoma cells by HPLC mass spectrometry (p/m97 shown). FCS was included to activate Sphk1. The data are means AE SD where n ¼ 4 from one experiment typical of two.

Runx1 PROTECTS LYMPHOMA CELLS AGAINST DEXAMETHASONE-INDUCED APOPTOSIS
We previously reported that each of the three Runx genes can oppose dexamethasone-induced growth inhibition in murine fibroblasts, at least in part by down-regulation of the glucocorticoid receptor gene, Nr3c1 [Wotton et al., 2008]. We tested this observation further in lymphoid cells where dexamethasone induces apoptosis rather than growth stasis and glucocorticoids feature as important therapeutic agents in lymphoma treatment [Bhadri et al., 2012]. Using trypan blue exclusion as a surrogate for viability, ectopic Runx1 was observed to provide a sustained protective effect against dexamethasone that increased markedly from 48 h after dexamethasone exposure (Fig. 3A). The protective effect was even more striking in light of the modest growth inhibition observed with ectopic Runx1 in this cell background (Fig. 3B). Together these data suggest that protection against dexamethasone-induced death is due to increased survival mediated by Runx1. Modulation of pro-and anti-apoptotic enzyme expression to promote apoptosis was previously reported in response to dexamethasone in double positive murine thymocytes [Bianchini et al., 2006]. Of these, Sgpp1 and Ugcg were also identified as direct transcriptional targets of Runx1 in murine fibroblasts, whereas the glucocorticoid receptor Nr3c1 was identified as an indirect target [Wotton et al., 2008;Kilbey et al., 2010]. To determine whether Runx1 opposed dexamethasone-mediated gene expression to promote cell survival we examined transcription of Runx1, Sgpp1, Ugcg, and Nr3c1 by qt-RT-PCR in the presence and absence of dexamethasone and ectopic Runx1. As shown in Figure 3C, endogenous Runx1 was repressed by dexamethasone as previously reported, but the pattern was reversed in the presence of ectopic Runx1. In contrast, Sgpp1 expression was dramatically opposed by ectopic Runx1 in dexamethasone-treated p/m97 cells whereas Nr3c1 expression was not. Indeed, as previously reported in murine fibroblasts, Nr3c1 was slightly downregulated by ectopic Runx1 but was upregulated by dexamethasone in both the presence and absence of ectopic Runx1 [Wotton et al., 2008]. Ugcg expression was refractory to dexamethasone treatment in p53-null thymic lymphoma cells but was induced by Runx1. This finding contrasts with double positive murine thymocytes where Ugcg was reportedly repressed by dexamethasone [Bianchini et al., 2006] but was consistent with the effects of ectopic Runx1 on Ugcg transcription in murine fibroblasts [Wotton et al., 2008;Kilbey et al., 2010]. Together the data support a mechanism whereby the opposition of Sgpp1 transcription and the overexpression of Ugcg might contribute to Runx1-induced survival in the presence of dexamethasone in thymic lymphoma cell lines.  (3). A significant reduction in Runx1 expression was observed between Mx1Cre þ and Mx1Cre À mice indicating background excision (gray vs. light green bar; P ¼ <0.05). Inclusion of pIpC to excise Runx1 generated a further reduction (black vs. green bar; P ¼ <0.01). (C) Remaining splenic tissues were pulverized and analyzed by HPLC mass spectrometry for C18-S1P. Data are expressed relative to 1 mg of pulverized spleen tissue and represent means AE SD of spleens of three mice for each set of conditions.

Runx1 EXCISION PROMOTES DEXAMETHASONE-INDUCED APOPTOSIS AND Sgpp1 TRANSCRIPTION
To further explore the relationship between Runx1 and Sgpp1 in dexamethasone-induced apoptosis, we examined Runx1 fl/fl / Mx1Cre þ B cell lymphoma lines established from a mouse model heterozygous for p53 loss and expressing c-Myc in the B cell compartment (Em-Myc). We exploited interferon b induction of Mx1Cre to mediate in vitro excision of Runx1 and derived a pair of clonally identical daughter cell lines that express low but readily detectable levels of full length Runx1 (Runx1Fl) or the deleted nonfunctional variant (DRunx1Fl- Fig. 4A) and asked whether loss of Runx1 promotes dexamethasone-induced apoptosis and Sgpp1 transcription. Enforced deletion of endogenous Runx1 significantly increased cell death in response to 1.0 mM dexamethasone (Fig. 4B). Moreover, enforced deletion of Runx1 increased expression of Sgpp1 in DRunx1Fl cells (black vs. green bars) and this was further increased in the presence of dexamethasone (hashed black vs. green bars) (Fig. 4C). Parallel cell lines derived from Runx1 wt/wt /Mx1Cre þ / p53 +/À /Em-Myc lymphomas were refractory to interferon b with respect to Sgpp1 and Runx1 expression [Borland et al., 2016] thereby excluding a non-specific effect of the Mx1Cre recombinase as a causative factor for the increase in Sgpp1 expression following Runx1 excision. To determine which functional domains of Runx1 are required to suppress Sgpp1 induction in the presence of dexamethasone, we introduced a panel of Runx1 mutants (Fig. 4D) into DRunx1Fl cells which essentially lack Runx1 expression. qt-RT-PCR analysis revealed that Runx1 repression of Sgpp1 was dependent on intact DNA binding and heterodimerization domains ( Fig. 4D), indicating that Runx1 directly opposes dexamethasoneinduced transcription of Sgpp1.

Sgpp1 KNOCKDOWN REDUCES DEXAMETHASONE-INDUCED CELL DEATH
To investigate the role of Sgpp1 as a direct mediator of glucocorticoidmediated apoptosis we used a short hairpin (sh) RNA to knock down expression of Sgpp1 in p/m97 thymic lymphoma cells which displayed the most dramatic induction of Sgpp1 in response to dexamethasone ( Fig. 3C-black bar vs. hashed black bar). Retroviral transduction of Sgpp1 shRNA viral supernatants reduced dexamethasone-induced expression of Sgpp1 by $40% compared to the non-coding (NC) shRNA control sequences (Fig. 5A). Stably transduced cells were then cultured in the presence and absence of dexamethasone and cell death measured by trypan blue exclusion. After 36 hr, dexamethasoneinduced death was reduced by $20% in cells receiving Sgpp1 shRNA compared to the NC shRNA control sequence (Fig. 5B). Together these data indicate that Sgpp1 induction is at least partially responsible for death of lymphoma cells mediated by glucocorticoid signaling. Insights gained from this data regarding the interplay between dexamethasone and Runx1 for the regulation of sphingolipid metabolism and cell survival are illustrated in the model shown in Figure 5C.

DISCUSSION
Previous studies in murine fibroblasts identified three sphingolipid metabolism enzymes as direct transcriptional targets of the three members of the Runx gene family [Wotton et al., 2008;Kilbey et al., 2010]. Moreover, the pronounced pro-survival effects of ectopic Runx expression in this context were associated with depletion of intracellular ceramide levels suggesting a shift in the sphingolipid rheostat in favor of pro-survival S1P [Kilbey et al., 2010]. We have now extended these studies to the lymphoid system where ectopic Runx expression has been shown to be potently oncogenic in combination with other cancer gene lesions including p53 loss and Myc over-expression [Vaillant et al., 1999;Blyth et al., 2001Blyth et al., , 2006Shimizu et al., 2013]. In thymic lymphoma cells Runx1 expression consistently regulates sphingolipid metabolism to favor S1P production with concomitant reduction in long chain ceramides. As we found, no positive effect of Runx1 over-expression on cell proliferation in this cell background, these results strongly reinforce previous evidence for enhanced survival as a key mechanism in Runx oncogenesis [Blyth et al., 2006]. Moreover, we have shown that partial deletion of Runx1 in vivo results in a significant and specific reduction in tissue S1P levels, revealing a fundamental physiological role for Runx regulation of sphingolipid metabolism in normal hematopoiesis. The production and distribution of S1P in spleen is known to be tightly regulated [Ramos-Perez et al., 2015] and the importance of Runx1 expression for maintenance of S1P levels described here is likely to be critical for regulation of the immune system.
The regulation of sphingolipid metabolism by the Runx genes is a multigenic phenomenon involving both direct and indirect gene regulation and entails a series of incremental changes in individual genes that combine to achieve significant metabolic shifts [Wotton et al., 2008;Kilbey et al., 2010]. However, this study highlights an additional need to measure transcriptomic changes under stress if the consequences of Runx modulation for sphingolipid metabolism are to be fully understood. In particular, ectopic Runx1 significantly impaired the massive induction of Sgpp1 in T-lymphoma cells in response to dexamethasone, but had no effect on basal expression of the gene in these cells. Similarly, although Sgpp1 was up-regulated in response to deletion of Runx1 in B-lymphoma cells, the differential was more marked in the presence of dexamethasone. Replacement with ectopic Runx1 reversed this effect, but not when the DNA or Cbfb cofactor binding functions were disabled, indicating that suppression most likely operates through the previously mapped binding site in the Sgpp1 promoter [Kilbey et al., 2010].
Dexamethasone is a synthetic glucocorticoid that induces ceramide and sphingosine to promote apoptosis in mouse thymocytes in a manner that mimics the normal process of T cell selection in the thymus [Cifone et al., 1999;Lepine et al., 2004]. Previous studies in primary mouse thymocytes identified changes in Runx1, Nr3c1, Ugcg, and Sgpp1 as early consequences of dexamethasone treatment compatible with ceramide-induced apoptosis [Bianchini et al., 2006]. Our studies in lymphoma cells partially recapitulate these findings, and indicate a causal chain in which down-regulation of Runx1 is proximal to the other changes and reversed by ectopic expression. Moreover, we have identified positive regulation of Ugcg and repression of Sgpp1 as consistent Runx1 functions. An effector role for Sgpp1 in the depletion of S1P and concomitant dexamethasone-induced death is further supported by shRNA knock down of Sgpp1 which, although far from complete, showed a significant sparing effect.
The importance of S1P for survival and drug resistance has been well documented in hematological malignancies but is more commonly attributed to deregulation of Sphk1 or Ugcg rather than to reduced levels of Sgpp1 [Marfe et al., 2011;Casson et al., 2013]. However, the fundamental importance of Sgpp1 is underlined by the neonatal lethality of gene deletion in mice [Allende et al., 2013] and the effects of modulating expression on cell growth and survival in response to cellular stress. For example, depletion of Sgpp1 promotes ER stress-induced autophagy in eukaryotic cell lines [Lepine et al., 2011] and thermo tolerance in yeast [Mao et al., 1999], whereas ectopic expression of Sgpp1 promotes apoptosis and accumulation of intracellular ceramide in NIH3T3 and HEK293 fibroblasts [Le Stunff et al., 2002]. The relevance of SGPP1 in other cancer types is suggested by its down-regulation in therapy-resistant ovarian and prostate cancers [Helleman et al., 2006;Huang et al., 2013].
The observation that ectopic Runx1 suppresses glucocorticoidinduced death and that deletion of endogenous Runx1 has the opposite effect also helps to elucidate the oncogenic consequences of Runx over-expression in lymphoid cells [Stewart et al., 1997[Stewart et al., , 2002Wotton et al., 2002]. Moreover, glucocorticoids comprise key components of multi-drug regimens for the treatment of acute lymphoid malignancies, where resistance is common but the underlying in vivo mechanisms poorly understood [Bhadri et al., 2012]. Our results predict that perturbations in expression of RUNX family members in human leukemia will be relevant to cell survival and therapy resistance. In support of this hypothesis, human t(8;21) leukemias that carry only a single functional RUNX1 allele are unique among AMLs in their sensitivity to glucocorticoids [Miyoshi et al., 1997]. Moreover, relapsing t(12;21) leukemias harboring the TEL-RUNX1 fusion display mutually exclusive gains of extra wild-type RUNX1 on chromosome 21 and losses of the glucocorticoid receptor Nr3c1. In light of these and our observations reported here it will therefore be interesting to test the synergistic potential of glucocorticoids and Fig. 5. shRNA knockdown of Sgpp1 reduces dexamethasone-mediated apoptosis. (A) qt-RT-PCR analysis of Sgpp1 expression in p/m97 cells stably infected with viral supernatants expressing an Sgpp1 or a non-coding (NC) control shRNA sequence. Cells were grown for 6 h in the presence of 1.0 mM dexamethasone prior to RNA extraction and qt-RT-PCR analysis. The data were calculated as described in Fig. 4C. (B) The same cells were plated in triplicate and grown for 36 h in the presence (Dex) and absence (MtOH) of 1.0 mM dexamethasone and monitored for live/dead counts by trypan blue exclusion. Knockdown of Sgpp1 had no effect on cell viability under control conditions but gave a significant reduction in cell death in the presence of 1.0 mM dexamethasone. (C) Interplay between Runx1 and dexamethasone on the expression of sphingolipid metabolism enzymes involved in the synthesis and breakdown of sphingosine and ceramide and their potential contributions to cell death and survival [Bianchini et al., 2006;Kilbey et al., 2010].
inhibi tors of RUNX-CBFb functions [Illendula et al., 2016] or their key downstream targets in sphingolipid metabolism [Edmonds et al., 2011] in the growing number of cancers where the RUNX genes are implicated as oncogenic drivers [Ito et al., 2015].