Mga2 Transcription Factor Regulates an Oxygen-responsive Lipid Homeostasis Pathway in Fission Yeast*

Eukaryotic lipid synthesis is oxygen-dependent with cholesterol synthesis requiring 11 oxygen molecules and fatty acid desaturation requiring 1 oxygen molecule per double bond. Accordingly, organisms evaluate oxygen availability to control lipid homeostasis. The sterol regulatory element-binding protein (SREBP) transcription factors regulate lipid homeostasis. In mammals, SREBP-2 controls cholesterol biosynthesis, whereas SREBP-1 controls triacylglycerol and glycerophospholipid biosynthesis. In the fission yeast Schizosaccharomyces pombe, the SREBP-2 homolog Sre1 regulates sterol homeostasis in response to changing sterol and oxygen levels. However, notably missing is an SREBP-1 analog that regulates triacylglycerol and glycerophospholipid homeostasis in response to low oxygen. Consistent with this, studies have shown that the Sre1 transcription factor regulates only a fraction of all genes up-regulated under low oxygen. To identify new regulators of low oxygen adaptation, we screened the S. pombe nonessential haploid deletion collection and identified 27 gene deletions sensitive to both low oxygen and cobalt chloride, a hypoxia mimetic. One of these genes, mga2, is a putative transcriptional activator. In the absence of mga2, fission yeast exhibited growth defects under both normoxia and low oxygen conditions. Mga2 transcriptional targets were enriched for lipid metabolism genes, and mga2Δ cells showed disrupted triacylglycerol and glycerophospholipid homeostasis, most notably with an increase in fatty acid saturation. Indeed, addition of exogenous oleic acid to mga2Δ cells rescued the observed growth defects. Together, these results establish Mga2 as a transcriptional regulator of triacylglycerol and glycerophospholipid homeostasis in S. pombe, analogous to mammalian SREBP-1.

up-regulated under low oxygen but are not Sre1 targets (9). Therefore, additional low oxygen-responsive pathways remain to be discovered. Given that fission yeast adaptation to low oxygen is a model for low oxygen responses in pathogenic fungi, identification of these pathways could highlight novel targets for inhibitors of fungal pathogenesis (14 -17).
Here, we report the results of a screen of 2601 fission yeast non-essential haploid deletion mutants for genes required for growth in low oxygen and in the presence of cobalt chloride. In this report, we define Mga2 as a transcriptional activator required for growth under both low oxygen and cobalt chloride conditions. mga2 has homologs in Saccharomyces cerevisiae that are ER membrane-bound transcriptional activators required for expression of the ⌬9 fatty acid desaturase OLE1 (18,19). We demonstrate that fission yeast Mga2 regulates a low oxygen-responsive gene expression program distinct from Sre1. Genes regulated by Mga2 include the fatty-acid synthases fas1 and fas2, the fatty acid desaturase ole1, and the long chain fatty acid CoA ligase lcf1, all of which are homologs of SREBP-1 targets in mammals. We find that mga2 is required to maintain TAG and glycerophospholipid homeostasis. Therefore, Mga2 regulates a second low oxygen response pathway in S. pombe that is analogous to the function of SREBP-1 in mammals.

Experimental Procedures
Materials-We obtained general chemicals and materials from Sigma or Fisher. Other sources include the following: yeast extract, peptone, and agar from BD Biosciences; S. pombe haploid deletion collection version 1 from Bioneer; cobalt(II) chloride and amino acid supplements from Roche Applied Science; Moloney murine leukemia virus reverse transcriptase from New England Biolabs; RNA STAT-60 from Tel-Test; GoTaq qPCR Master Mix from Promega; oligonucleotides from Integrated DNA Technologies; horseradish peroxidaseconjugated, affinity-purified donkey anti-rabbit IgG from Jackson ImmunoResearch; IRDye donkey anti-rabbit and from Li-Cor; prestained protein standards from Bio-Rad; and fatty acid-free bovine serum albumin from SeraCare Life Sciences.
Antibodies-Rabbit polyclonal antibody against amino acids 1-260 of Sre1 (anti-Sre1 IgG) was generated using a standard protocol as described previously; we purified the antigen with an N-terminal polyhistidine tag and a tobacco etch virus protease cleavage sequence from Escherichia coli using nickelnitrilotriacetic acid-agarose (Qiagen). We then cleaved with tobacco etch virus protease (Invitrogen) to remove the histidine tag. We isolated Sre1-specific antibodies from rabbit serum by affinity chromatography using NHS-Sepharose resin (Pierce) conjugated to the polyhistidine-tagged Sre1 antigen (8). Speci-ficity of this antibody was assayed by loss of immunoreactivity in an sre1⌬ strain.
We generated monoclonal antibody 5B4 IgG1 to Sre1 (amino acids 1-260) using recombinant protein that was purified from E. coli by nickel-affinity chromatography (Qiagen). We immunized BALB/c mice with this antigen and screened for immunoreactivity by ELISA and Western blotting. We fused spleen cells from immunopositive mice with SP2/0 myeloma cells to make monoclonal antibodies. We identified positive clones by ELISA screening using the immunizing antigen. After dilution cloning, antibody specificity was tested by immunoblotting against S. pombe extracts from cells overexpressing Sre1. We determined isotype using the mouse isotyping kit (Roche Applied Science). We purified final antibodies from either tissue culture supernatant or ascites fluid using protein-G-Sepharose (GE Healthcare).
Yeast Culture-Yeast strains are described in Table 1. S. pombe cells were grown to exponential phase at 30°C in rich YES medium (0.5% (w/v) yeast extract plus 3% (w/v) glucose supplemented with 225 g/ml each of uracil, adenine, leucine, histidine, and lysine (20)). YES ϩ CoCl 2 medium was prepared by dissolving cobalt(II) chloride in H 2 O and adding to a final concentration of 1.6 mM in YES medium. For fatty acid supplementation, fatty acids were added to 1 mM in YES medium from a 12.7 mM stock in 12% (w/v) fatty acid-free BSA.
BSA Conjugation of Free Fatty Acids-Fatty acid-free BSA (24% w/v) was made by adding 12 g of fatty acid-free bovine serum albumin to 35 ml of 150 mM NaCl in six 2-g doses over 5 h. pH was adjusted to 7.4 with 5 N NaOH, and the volume was brought up to 50 ml with 150 mM NaCl. This solution was diluted 1:1 with 150 mM NaCl to produce 12% fatty acid-free BSA prior to use as an experimental control.
Fatty acid conjugation to BSA was performed by dissolving 319 mol of the conjugating fatty acid in 2 ml of EtOH. 100 l of 5 N NaOH was added to precipitate sodium salt of the fatty acid. EtOH was evaporated under nitrogen gas. 10 ml of 150 mM NaCl were added to the dried fatty acid, and the solution was heated until the fatty acid dissolved. The solution was then stirred and slowly cooled to just before the fatty acid precipitated, at which point 12.5 ml of ice-cold fatty acid-free BSA (24% w/v) was added. The solution was stirred for 10 min, and the final volume was adjusted to 25 ml with 150 mM NaCl.
S. pombe Deletion Collection Screen-The Bioneer Haploid Deletion Mutant Library version 1.0 was screened as described previously (21). 2601 deletion mutants were streaked for single colonies on YES or YES ϩ CoCl 2 medium and placed at 30°C in the presence of oxygen. An additional YES plate was grown at 30°C in anaerobic conditions using the BBL GasPak system (BD Biosciences). Mutants were compared with wild-type and sre1⌬ strains and scored for growth after 7 days. Full screen results can be found in supplemental Table S1. Sre1 Cleavage Assay-Cells were grown in YES medium to exponential phase inside an InVivo 2 400 hypoxic work station (Biotrace, Inc.) and collected for protein extraction and immunoblotting as described previously using polyclonal or monoclonal 5B4 anti-Sre1 antibody as indicated (8). SDS-polyacrylamide gels were equally loaded for total protein, which was quantified using the BCA protein assay (Pierce). Consistent loading was confirmed following electroblotting by staining the membrane with Ponceau S. Blots were imaged using enhanced chemiluminescence and film or the Odyssey CLx infrared imaging system (Li-Cor), as noted in the figure legends.
Microarray-Data in supplemental Tables S4 and S5 represent the log2 fold changes in expression between sre1⌬ and sre1⌬ mga2⌬ cells grown anaerobically for 1.5 h. Microarray analysis was performed as described previously (11). Total RNA was isolated using RNA STAT-60 and amplified and labeled using the RNA amplification and labeling kit (Agilent Technologies) with oligo(dT) primers (System Bioscience) and cyanine CTPs (PerkinElmer Life Sciences). RNA was fragmented, denatured, and hybridized to a custom Agilent array at 60°C for 17 h. Two dye-reversal hybridizations were performed with cells cultured on different days (biological replicates), yielding two data points per probe. Arrays were scanned by an Agilent G2505B scanner, and features were flagged using Agilent Feature Extraction software. These data were imported into Partek Genomics Suite as the median of the arrays' three replicate probes' g/r processed signals. These log2 signal values were quantile-normalized, and the dye-swapped samples were compared in a one-way ANOVA. Because there were only two biological replicates, a pseudo p value is presented alongside a significance analysis of microarrays (SAM) q-value. SAM parameters were ⌬ ϭ 1.87646, false discovery rate (median) ϭ 0.05447. Mga2-dependent genes showed lower expression in sre1⌬ mga2⌬ versus sre1⌬ cells. The microarray gene expression data described in this study have been deposited in the NCBI Gene Expression Omnibus (22) and are accessible through GEO Series accession number GSE60544.
Quantitative PCR-Yeast cells (1 ϫ 10 8 ) growing exponentially in the presence or absence of oxygen were pelleted, resuspended in RNA STAT 60, and vortexed. Samples were mixed with chloroform to 16% (v/v) and then centrifuged at 10,000 ϫ g for 15 min at 4°C. The aqueous fraction was mixed with isopropyl alcohol to 33% (v/v) and then centrifuged at 10,000 ϫ g for 10 min. Precipitated RNA was washed with 75% (v/v) EtOH and air-dried. cDNA was synthesized following DNase and reverse transcription instructions for Moloney murine leukemia virus reverse transcriptase. RT-qPCR was performed using the indicated primers and GoTaq qPCR Master Mix. Each reaction was performed with two technical replicates per the five biological replicates. Error bars are 1 S.D.
Lipid Extraction-Lipid extraction was performed as described previously (23)(24)(25). Six biological replicate cultures of wild-type and mga2⌬ cells growing exponentially in the presence of oxygen (6.8 ϫ 10 7 -1.3 ϫ 10 8 cells) were pelleted and washed twice in 155 mM ammonium acetate and then frozen under liquid nitrogen. Lipids were extracted at 4°C by breaking cell pellet with beads, and then 20 internal lipid standards were added to the lysate. Samples were subjected to two-step extraction with chloroform/methanol. The collected lipid extracts were then vacuum-evaporated and dissolved in chloroform/methanol. Shotgun Lipidomics-Mass spectrometry was performed as described previously (23)(24)(25). Extracts were analyzed on a LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific) equipped with the robotic nanoflow ion source TriVersa Nano-Mate (Advion Biosciences).
Lipidomics Data Analysis-Lipidomics data analysis was performed as described previously using ALEX software (24,25). Species were annotated using sum composition as follows: ͗lipid class͗͘total number of carbons in the fatty acid moieties͘: ͗total number of double bonds in the fatty acid moieties͘. Sphingolipids were annotated as ͗lipid class͗͘total number of carbons in the long-chain base and fatty acid moiety͘:͗total number of double bonds in the long-chain base and fatty acid moiety͘;͗total number of OH groups in the long-chain base and fatty acid moiety͘. Visualization and calculation of mole % values were performed using the commercially available Tableau software. p values were calculated using a Mann-Whitney U test. p values were corrected for multiple hypotheses post hoc using the Benjamini and Hochberg false discovery rate (26

Equation 3
was calculated based on the fact that that S. pombe only synthesizes saturated and mono-unsaturated fatty acids (27). In the case where fatty acid saturation was determined, lipid species were hydrolyzed in silico prior to normalization (25). For example, for TAG 42:2, two of the fatty acid moieties must be monounsaturated and one must be saturated; this is taken into account when tabulating the number of saturated and unsaturated fatty acids across all TAG species. The supplemental Table S6 lists the full results of the lipidomics analysis.

Identification of Genes Required for Growth under Low
Oxygen and on Cobalt Chloride-Adaptation to low oxygen in fission yeast requires the coordinated regulation of many genes (9). The Sre1 transcription factor only regulates expression of 22% of low oxygen-responsive genes, suggesting that additional oxygen-regulated transcription factors exist (9). Successful low oxygen adaptation can be assayed by growth of fission yeast under low oxygen or on the low oxygen mimetic, CoCl 2 . Indeed, sre1⌬ cells fail to grow under both of these conditions (21). To identify genes required for hypoxic adaptation, we screened 2601 mutants from the S. pombe non-essential haploid deletion collection for growth in these two conditions. We found that 105 gene deletions were sensitive to low oxygen and/or CoCl 2 , exhibiting reduced growth in these conditions (Fig. 1A). This included 38 deletion mutants sensitive to only low oxygen (1.5% of all tested gene deletions, supplemental Table S2), 40 mutants sensitive to CoCl 2 (1.5%, supplemental Table S3), and 27 mutants sensitive to both conditions (1.0%, Table 2). Growth data for the complete screen can be found in supplemental Table S1. Gene ontology term enrichment analysis of deletions sensitive to both low oxygen and CoCl 2 ( Table 2) showed enrichment for genes involved in "retrograde transport, endosome to Golgi" and the retromer complex (p Ͻ 4E-3).
Our previous work identified four of the 27 genes required for growth under low oxygen and on CoCl 2 as dsc1-dsc4, members of the Dsc E3 ligase complex required for Sre1 cleavage in the Golgi (21). Genes required for Sre1 cleavage fail to accumulate the N-terminal transcription factor (Sre1N) in the absence of oxygen (21). To identify genes functioning in the Sre1 pathway, we assayed low oxygen induction of Sre1N in a number of deletion strains, including the remaining 23 mutants required for growth under low oxygen and on CoCl 2 ( Fig. 1B and Table  2). Immunoblotting against the Sre1 N terminus permits detection of both the full-length precursor form of Sre1 and the transcriptionally active cleaved N terminus. Wild-type cells showed FIGURE 1. Screen for gene deletions sensitive to low oxygen and cobalt chloride. A, Venn diagram of deletion collection screen results showing overlap between deletions defective for growth under low oxygen and those defective for growth on CoCl 2 . B, Western blots imaged by film probed with anti-Sre1 IgG of lysates from wild-type cells or the indicated deletion strains grown for 0 or 6 h in the absence of oxygen. P and N denote precursor and cleaved N-terminal transcription factor forms, respectively. Different blots are separated by solid black lines. This experiment was performed one time. C-E, wild-type or the indicated deletion strains carrying empty vector or a plasmid expressing sre1N (C and D) or with sre1N integrated at the endogenous locus (E) were grown as indicated for 4 days. Cells were plated in 10-fold serial dilutions (16 -1.6 ϫ 10 6 cells). Each experiment was performed with at least two biological replicates. EV, empty vector. F, indicated strains were grown for 0 or 4 h in the absence of oxygen, and cell lysates were immunoblotted with monoclonal anti-Sre1 5B4 IgG1 and imaged by Li-Cor. N denotes Sre1 N-terminal form. This experiment was performed with three biological replicates. robust cleavage and Sre1N production after 6 h at low oxygen ( Fig. 1B, lanes 7 and 8), and four of the deletions sensitive to both low oxygen and CoCl 2 (mga2, cpp1, ppr6, and trx2) showed reduced Sre1N accumulation ( Fig. 1B and Table 2). These deletions represent new candidates for genes required for Sre1 activation in S. pombe. The remaining 19 deletions showed normal Sre1N accumulation ( Table 2) and thus are genes that could be involved in low oxygen-responsive pathways distinct from Sre1 target gene transcription.
To test whether a failure to produce Sre1N caused the observed growth defects in the absence of these four genes (mga2, cpp1, ppr6, and trx2), we expressed sre1N from a plasmid and assayed growth on CoCl 2 . Notably, none of these deletion strains was rescued by sre1N expression (Fig. 1, C and D), suggesting that either the Sre1N production defect occurs after cleavage (e.g. Sre1N is highly unstable) or that the deletion mutant is CoCl 2 -sensitive for reasons unrelated to Sre1N function.
Of the four genes, we focused our studies on mga2 because Mga2 is a putative transcriptional activator in fission yeast based on homology to two related transcriptional activators in S. cerevisiae, MGA2, and SPT23 (18). Therefore, Mga2 was a candidate for a new regulator of a low oxygen response. To confirm that the defect in growth of mga2⌬ cells on cobalt chloride was not due to the observed Sre1 cleavage defect, we expressed the sre1N transcription factor from the endogenous sre1 locus and assayed transcription factor activity and growth on CoCl 2 . As with the plasmid-based sre1N previously examined, expression of sre1N from the endogenous locus did not rescue CoCl 2 growth of mga2⌬ cells (Fig. 1E). Importantly, full activation of Sre1 production under low oxygen requires positive feedback regulation in which Sre1 stimulates sre1 transcription (28). In the absence of mga2, Sre1N up-regulated its own expression under low oxygen through positive feedback to wild-type levels (Fig. 1F). These data suggest the following: 1) that mga2⌬ cells are defective for proteolytic activation of Sre1 precursor, rather than at a step downstream of cleavage, and 2) that the Sre1 cleavage defect cannot fully account for the observed growth defects of mga2⌬ cells on CoCl 2 . Based on our interest in the hypoxic transcriptional response and the known function of S. cerevisiae mga2 homologs as transcriptional activators, we further analyzed the requirement for mga2 in fission yeast low oxygen adaptation.
Cell Growth Requires Mga2 in the Presence and Absence of Oxygen-To characterize the requirement for mga2 in response to changes in environmental oxygen, we measured

Mga2 Regulates Low Oxygen Lipid Homeostasis
growth in liquid culture of wild-type and mga2⌬ cells in either the presence or absence of oxygen. mga2⌬ cells showed significantly reduced growth compared with wild-type cells in both conditions and further reduced growth in the absence of oxygen compared with the presence of oxygen ( Fig. 2A). Consistent with the observation that functional Sre1N transcription factor does not rescue CoCl 2 growth in the absence of mga2 (Fig. 1, E and F), Sre1N did not rescue either the plus oxygen or minus oxygen liquid growth defects (Fig. 2B). Given that sre1⌬ cells exhibit wild-type growth in the presence of oxygen (9), these growth assays demonstrate that mga2 functions in adaptation to low oxygen and CoCl 2 through Sre1-independent pathway(s).
Mga2 Functions as a Transcriptional Regulator of Lipid Metabolism-Uncharacterized in S. pombe, mga2 is the sole fission yeast homolog of S. cerevisiae MGA2 and SPT23. S. cerevisiae MGA2, but not SPT23, is required for low oxygen induction of transcription of the ⌬9 fatty acid desaturase OLE1 and modulates the stability of OLE1 mRNA in response to fatty acid availability (19,29). S. cerevisiae Mga2 also up-regulates transcription of other genes involved in fatty acid and sterol homeostasis under oxidative stress, including ERG1, FAS1, ELO1, FAA4, and ATF1 (30). Although a combined deletion of both MGA2 and SPT23 in S. cerevisiae is lethal, deletion of mga2 in S. pombe was slow growing but viable under normoxia (Figs. 1D  and 2A). The low oxygen sensitivity of the mga2⌬ mutant led us to investigate Mga2-dependent transcription in S. pombe under low oxygen.
To determine the transcriptional targets of S. pombe Mga2, we performed genome-wide expression profiling under low oxygen. We compared gene expression in sre1⌬ cells to sre1⌬ mga2⌬ cells after 1.5 h of growth in the absence of oxygen. We deleted sre1 in both strains so that observed changes in gene expression would be due to loss of mga2 and not indirect effects of the Sre1 cleavage block that occurs in mga2⌬ cells (Fig. 1B). Of the 5057 genes examined, 291 (5.8%) were significantly decreased in mga2⌬ cells by SAM. The full list of Mga2-dependent genes can be found in supplemental Tables S4 and S5. Gene ontology term enrichment analysis of the decreased genes identified fatty acid metabolism and lipid metabolism as affected pathways (p value Ͻ0.0001). Table 3 lists genes that decreased expression at least two standard deviations from the mean in mga2⌬ cells. These included long-chain fatty acid CoA ligases (lcf1 and lcf2), TAG lipases (ptl1 and ptl2), and the ⌬9  fatty acid desaturase (ole1). Additionally, Mga2 controls expression of the biotin transporter vht1 and biotin synthase bio2, as well as hem1. Biotin is required for acetyl-CoA carboxylase function, and heme is a prosthetic group for Ole1 in S. cerevisiae (31). Although transcriptome profiling for S. cerevisiae MGA2-dependent genes under low oxygen conditions has not been performed, homologs of known S. cerevisiae Mga2-dependent genes were decreased in our mga2⌬ cells (Table 3 and  supplemental Table S5). Our microarray results suggest that fission yeast Mga2 regulates many lipid metabolism genes.
To confirm the microarray results, we assayed low oxygen expression of candidate Mga2 target genes in wild-type and mga2⌬ cells by quantitative real time PCR. Expression of ole1, lcf1, vht1, ptl1, fas1, and fas2 increased in the absence of oxygen, and this induction required mga2 (Fig. 3A). Expression of ole1, lcf1, and fas2 was also reduced even in the presence of oxygen, indicating that mga2 is required for gene expression both in the presence and absence of oxygen.
To assess the oxygen regulation of all Mga2 targets, we compared our microarray results with known data sets. A previous study in our laboratory identified S. pombe genes up-regulated after 1.5 h of low oxygen treatment (9). 37% of Mga2 target genes were up-regulated under low oxygen in that study ( Fig.  3B and supplemental Table S5). This accounts for 21% of all low oxygen up-regulated genes (9). If we consider those genes most dependent on Mga2 (those showing expression reduced more than 2 S.D. from the mean in the absence of mga2), 68% of these genes were up-regulated under low oxygen in our previous study (Table 3, underlined IDs). Only 21% of Sre1 target genes were also regulated by Mga2 under low oxygen, confirming that Mga2 promotes a low oxygen-responsive pathway that is distinct from the Sre1 pathway (Tables 3, supplemental Table S5, and Fig. 3B) (9). We conclude that S. pombe Mga2 regulates lipid metabolism gene expression in the presence of oxygen and that there is an additional requirement when oxygen is limiting.
Maintenance of Lipid Homeostasis Requires Mga2-We hypothesized that failure to up-regulate genes involved in lipid biosynthesis in mga2⌬ cells may lead to alterations in lipid homeostasis. To measure levels of cellular lipids, we performed quantitative mass spectrometry-based lipidomics on wild-type and mga2⌬ cells grown in the presence of oxygen (23). We examined cells in the presence of oxygen because oxygen is required for fatty acid desaturation and ergosterol synthesis (32)(33)(34), so the absence of oxygen would alter lipid synthesis regardless of strain background. Additionally, Mga2 target genes showed reduced expression in mga2⌬ cells in the presence of oxygen (Fig. 3A). Full data and data analysis of this lipidomics experiment can be found in supplemental Table S6.
Our lipidomics experiment showed that in mga2⌬ cells, diacylglycerol and TAG levels decreased, whereas glycerophospholipid levels varied, with PA and PE increasing, phosphatidylinositol decreasing, and phosphatidylserine and cardolipin levels unchanged (Fig. 4A). Interestingly, these alterations in glycerophospholipids were largely not reflected in the lysophospholipids that are inputs and products of glycerophospholipids. Lysophosphatidylcholine and lysophosphatidylethanolamine levels decreased, despite the fact that phosphatidylcholine was unchanged and PE levels were higher in mga2⌬ cells (Fig. 4A). Although levels of both ceramide and mannosylinositol phosphorylceramide were unchanged, levels of the intermediate inositol phosphorylceramide increased in the absence of mga2 (Fig. 4A). Levels of ergosterol increased in mga2⌬ cells, whereas lanosterol decreased (supplemental Table S6).
In the absence of mga2, glycerophospholipids decreased in chain length, especially in phosphatidylcholine, PA, and PE, although phosphatidylinositol chain lengths increased slightly (Fig. 4B). No significant changes occurred in lysophospholipid chain length (supplemental Table S6). Decreased acetyl-CoA carboxylase cut6 and enoyl reductase tcs13 expression (Table 3) may play a role in the observed decreases in fatty acid chain length. Taken together, these data demonstrate widespread lipidome disruption in the absence of mga2 and thus position this transcriptional activator as a crucial regulator of TAG and glycerophospholipid homeostasis.
Glycerophospholipid Saturation Increases in the Absence of Mga2-MGA2 and SPT23, the S. cerevisiae homologs of mga2, regulate the ⌬9 fatty acid desaturase OLE1 required for fatty acid desaturation. Notably, Mga2 in fission yeast also regulated the ⌬9 fatty acid desaturase ole1 (Fig. 3A). To determine whether decreased expression of ole1 resulted in altered glycerophospholipid saturation in mga2⌬ cells, we examined fatty acid saturation in our lipidomics experiment. All glycerophos- pholipids measured, as well as lysophosphatidic acid, lysophosphatidylinositol, lysophosphatidylethanolamine, LPS , diacylglycerol, and TAG showed increased saturation in mga2⌬ cells (Fig. 5A). Together, these data suggest that even in the presence of oxygen, mga2⌬ cells exhibit strongly decreased levels of lipid molecules with unsaturated fatty acid moieties.
To test whether the observed plus and minus oxygen growth defects resulted from reduced fatty acid desaturation, we grew wild-type and mga2⌬ cells in liquid medium supplemented with monounsaturated oleic acid (18:1) conjugated to BSA (oleic acid) or with BSA alone. Oleic acid is the most abundant unsaturated fatty acid in fission yeast, representing about 75% of all fatty acids in the cell (35). Oleic acid or BSA was added to the media at the 0-h time point, and cell density was measured for the following 12 h. Oleic acid addition had no effect on growth of wild-type cells but rescued growth of mga2⌬ cells in both the presence and absence of oxygen (Fig. 5B).
Although the oleic acid-treated mga2⌬ cells did not reach a density equal to that of wild type after 12 h, calculation of the cellular growth rates showed rescue. Oleic acid-treated mga2⌬ cells exhibited a doubling time similar to that of BSA-treated mga2⌬ cells during the first 6 h of treatment, but they grew like wild-type cells during the second 6 h of treatment (Table 4). These data suggest that normal growth of mga2⌬ cells requires a lag phase, perhaps to allow oleic acid uptake and incorporation into cellular lipids.
To test this hypothesis, we grew wild-type and mga2⌬ cells under normoxia for 16 h plus oleic acid, then diluted the cells into fresh oleic acid-containing medium, and grew the cells in  Table S6. Error bars represent 1 S.D. p values were calculated using a two-tailed Mann-Whitney U test and corrected for multiple comparisons using Benjamini and Hochberg's method to produce a false discovery rate, q. *, q value Ͻ0.05. LPA, lysophosphatidic acid; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; LPI, lysophosphatidylinositol; LPS, lipopolysaccharide; PA, phosphatidic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol; PIP, phosphoinositol phosphate; PS, phosphatidylserine; DAG, diacylglycerol; TAG, triacylglycerol; CL, cardolipin; Cer, ceramide; IPC, inositol phosphorylceramide; MIPC, mannosylinositol phosphorylceramide. FIGURE 5. Unsaturated fatty acids rescue mga2⌬ growth defect. A, saturation of fatty acyl chains in wild-type and mga2⌬ cells. Saturation of each chain was determined rather than the saturation of a lipid species as a whole (see "Experimental Procedures"). SFA is saturated fatty acids, and MUFA is monounsaturated fatty acids. Error bars are 1 S.D. p values were calculated using a two-tailed Mann-Whitney U test and corrected for multiple comparisons using Benjamini and Hochberg's method to produce a false discovery rate, q. *, q value Ͻ0.05; **, q value Ͻ0.01. B and C, wild-type and mga2⌬ cells were grown for 12 h in the presence (red) or absence (blue) of oxygen, with the addition of BSA alone or oleic acid (18:1) conjugated to BSA. Cell density was measured by absorbance at 600 nm. Data points are average of three biological replicates. Error bars are 1 S.D. B, BSA alone or oleic acid (18:1) was added at time 0. Absorbance at time 0 among the samples ranged from 0.13 to 0.18. To focus on the difference in growth rate among the conditions, we normalized the data for each sample to a value of 0.15 at time point 0 before averaging. The BSA only data are identical to those in Fig. 2A. C, oleic acid (18:1) was added 16 h prior to time 0 and then cells were diluted into fresh oleic acid medium at time 0. Absorbance at time 0 among the samples ranged from 0.09 to 0.14. To focus on the difference in growth rate among the conditions, we normalized the data for each sample to a value of 0.12 at time point 0 before averaging. LPA, lysophosphatidic acid; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; LPI, lysophosphatidylinositol; LPS, lipopolysaccharide; PA, phosphatidic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PS, phosphatidylserine; DAG, diacylglycerol; TAG, triacylglycerol; CL, cardolipin.
the presence or absence of oxygen for another 12 h (Fig. 5C). As expected, preincubation of mga2⌬ cells with oleic acid eliminated the lag phase and mga2⌬ cells showed wild-type growth from the 0-h time point (Fig. 5C and Table 5). Therefore, we conclude that addition of exogenous oleic acid fully rescues growth defects of mga2⌬ cells both in the presence and absence of oxygen.

Discussion
This study represents the first screen of the S. pombe nonessential haploid deletion collection version 1.0 for gene deletions sensitive to low oxygen. We identified 105 gene deletions that resulted in sensitivity to low oxygen and/or the low oxygen mimetic CoCl 2 , with 27 deletions resulting in sensitivity to both. Interestingly, we identified 19 genes that are required for adaptation to low oxygen stress but are not deficient in Sre1N accumulation ( Table 2). These genes may represent other pathways required to adapt to oxygen stress, or they may regulate Sre1 signaling at a post-cleavage step, such as regulating Sre1 binding to promoters. These genes include multiple players of three separate pathways (supplemental Tables S2 and S3).
The first highly represented pathway sensitive to low oxygen and CoCl 2 is transport between the endosome and the Golgi or the vacuole (ryh1, sec28, vsp8, vps17, vps24, vps26, vps29, vps32,  vps35, and vps36). This transport could be required to maintain V-type H ϩ -ATPase at the vacuole. It has been shown in budding yeast that in the absence of V-ATPase function, the cytosol acidifies (36). Normally, the plasma membrane proton pump Pma1 would compensate by hydrolyzing ATP and pumping out protons (37). However, under low oxygen, reduced oxidative respiration may decrease cellular ATP (38), creating conditions that require full V-ATPase activity to maintain cytosolic pH.
The second pathway sensitive to both low oxygen and CoCl 2 is Mediator (med20 and nut2). Highly conserved from yeast to humans, the Mediator complex, composed of 13 subunits in fission yeast, is required for regulated expression of most RNA polymerase II-dependent genes (39,40). A potential role of Mediator in regulating low oxygen adaptation in fission yeast is consistent with data in humans and Caenorhabditis elegans showing that Mediator is required for SREBP transcriptional activation of target genes, as well as lipid homeostasis and fatty acid desaturation (41). Because mediator is required at a postcleavage step in the SREBP pathway, the effects of these mutants may not be apparent in a cleavage assay.
A third pathway sensitive to only low oxygen in our screen is miRNA biogenesis (arb1, arb2, and dcr1). Interestingly, a number of recent papers have suggested that chronic hypoxia induces down-regulation of miRNA biogenesis to maintain hypoxia-inducible factor-␣ (HIF␣) induction in vascular endothelial and cancer cell lines (42)(43)(44). This adaptation is correlated with tumor progression and poor prognosis (43,44). However, other recent work has identified the micro-RNA miR-210 as a "master" hypoxia-regulated miRNA, exhibiting up-regulation by HIF-1␣ under conditions of hypoxia (45)(46)(47). miR-210 promotes cell cycle progression and evasion of apoptosis while compromising mitochondrial integrity and DNA repair in numerous cancer cell lines (47). Additional work is required to fully understand the complex interplay between miRNA biogenesis and hypoxia adaptation in human cells, as well as in fission yeast, which lack an HIF homolog.
In addition to mga2, this screen identified three genes that were required for Sre1N accumulation but were not rescued by exogenous Sre1N (Fig. 1C). Cpp1 is a subunit of protein farnesyltransferase required for farnesylation of small GTPases such as Ras and Rheb. cpp1⌬ exhibits severe normoxia growth defects and temperature sensitivity as well as morphological defects, indicating a role for farnesylation in a number of processes, one of which may be Sre1 activation (48). Trx2 is a mitochondrial oxidoreductase that helps to maintain the mitochondrial redox state, especially in the absence of glutathione (49,50). Ppr6 is a mitochondrial membrane protein that regulates the mRNA stability of the mitochondrial ATP synthase subunit atp9. In the absence of ppr6, atp9 mRNA is reduced, and cells become sensitive to galactose and antimycin, indicating defects in aerobic respiration (51). Taken together, these data suggest a role for mitochondrial function in regulating Sre1 activation and cellular adaptation to low oxygen and CoCl 2 .
Ryuko et al. (52) previously performed a screen of the 3004 S. pombe deletion collection version 2.0 strains for growth on 3.5 mM CoCl 2 . We used 1.6 mM CoCl 2 in our screen. They found 54 gene deletions sensitive to CoCl 2 and 56 gene deletions resistant to CoCl 2 . Of the 67 deletions that we found sensitive to CoCl 2 , only 11 were also identified by Ryuko et al. (52) as sensitive to CoCl 2 (supplemental Table S3). The small overlap between these studies may be due to the different concentrations of CoCl 2 used, leading to selection for different responses to the chemical. We chose a concentration sufficient to inhibit growth of sre1⌬ cells to mimic hypoxia. By screening under two conditions (low oxygen and CoCl 2 ), we focused our search to those genes most likely to be involved in low oxygen response. Additionally, another group recently screened the 669 strains of deletion collection version 2.0 that are not present in deletion   Fig. 5B. Cells were grown in BSA/oleate or BSA alone for 12 h Ϯ oxygen. Doubling times were calculated using 0-and 6-h or 6-and 12-h optical density readings as indicated, using the formula: doubling time ϭ (culture time ϫ log(2))/log(final OD) Ϫ log(initial OD). Doubling times are displayed as mean Ϯ S.D. for three biological replicates.    (2))/log(final OD) Ϫ log(initial OD). Doubling times are displayed as mean Ϯ S.D. for three biological replicates.

Mga2 Regulates Low Oxygen Lipid Homeostasis
collection version 1.0 (screened in this study) for growth on both CoCl 2 and under low oxygen conditions. They identified 33 genes required for growth under low oxygen and/or CoCl 2 conditions, although they only reported the identity of one, the rhomboid protease rbd2 (53). We focused our studies on the transcriptional activator Mga2, which our screen showed is required for low oxygen adaptation (Fig. 1D). Multiple lines of evidence support the conclusion that Mga2 regulates an oxygen-responsive transcriptional program that is independent from the Sre1 pathway. First, transcriptionally active Sre1N did not rescue low oxygen and CoCl 2 growth of mga2⌬ cells (Figs. 1, E and F, 2B). Second, mga2⌬ cells exhibited a growth defect in the presence of oxygen, whereas sre1⌬ cells have no normoxia growth defects ( Fig.  2A) (8). Third, Mga2 regulates numerous genes not known to be oxygen-dependent, whereas Sre1 activity is only required under low oxygen (Fig. 3, A and B, and supplemental Table S5) (9). Finally, comparison of our list of Mga2 low oxygen gene expression targets with the known list of Sre1 expression targets showed only 22% overlap (most notably csr101, lcf2, fsh2, and hmg1) ( Table 3 and Fig. 3B). Instead, Mga2 regulated genes primarily involved in TAG and glycerophospholipid homeostasis, including the fatty-acid synthases fas1 and fas2, the fatty acid desaturase ole1, and the long chain fatty acid CoA ligase lcf1 (Table 3 and Fig. 3A). This study positions Mga2 at the head of a new oxygen-responsive pathway in fission yeast.
Although Sre1 is regarded as the SREBP-2 analog that regulates sterol synthesis in S. pombe, no SREBP-1 analog regulating TAG and glycerophospholipid synthesis has been described. Given the results described in this study, we propose that mga2 is the SREBP-1 analog in fission yeast. Indeed, the list of Mga2 gene targets is strikingly similar to the known gene targets of mammalian SREBP-1 (8/9 targets conserved, supplemental Table S5) (4,54). Therefore, we conclude that Mga2 serves as the analogous transcription factor to SREBP-1 in S. pombe by coordinately regulating TAG and glycerophospholipid metabolism. Importantly, only 38% of S. pombe anaerobically up-regulated genes are transcriptional targets of either Mga2 or Sre1 (Fig. 3B). Thus, additional transcriptional regulators of low oxygen-responsive pathways, including carbohydrate synthesis and the mitochondrial response, remain to be discovered (9).
Although this study is the first characterizing the transcriptional regulation of glycerophospholipid and TAG metabolism in fission yeast, glycerolipid regulation is much better characterized in S. cerevisiae. There, PA signals the glycerolipid state of the cell, as all membrane phospholipids and TAG are synthesized from PA. Low levels of membrane PA cause the Opi1 repressor to leave the perinuclear ER membrane and enter the nucleus. There it represses phospholipid synthesis genes to maintain lipid homeostasis (55). Interestingly, there is no known Opi1 homolog in S. pombe, raising the question as to whether a different method is used to sense PA availability. Given the high degree of functional conservation between fission yeast and mammals, studies of phospholipid homeostasis in fission yeast may provide insight into the regulation of this pathway in mammals.
Interestingly, a recent report suggested that the CSL transcription factor Cbf11 binds to the promoters of seven Mga2 target genes in S. pombe (56). Although cbf11 is not itself an Mga2 target gene (supplemental Table S4), deletion of cbf11 results in a low oxygen growth defect in our screen (supplemental Tables S1 and S2), and Cbf11 was found to bind Mga2 in an affinity capture screen for the fission yeast protein interactome network (57). The Cbf11 human homologs RBPJ and RBPJL share ϳ19% protein sequence identity with Cbf11 (clustered in the functional domains), raising the possibility that these proteins perform similar functions. These RBP family transcription factors mediate Notch signaling and are required for pancreas development (58,59). It is intriguing to speculate that fission yeast Mga2 and Cbf11 may act together to regulate phospholipid and TAG metabolism in response to low oxygen. More work is required to determine whether and how these two transcription factors work together to regulate lipid homeostasis.
These results support a model for S. pombe low oxygen lipid homeostasis in which decreased oxygen availability results in reduced ergosterol synthesis. This activates Sre1, increasing expression of enzymes required for ergosterol synthesis and restoring homeostasis (8). At the same time, low oxygen availability also results in reduced fatty acid desaturation. This activates Mga2, increasing expression of ole1 and up-regulating unsaturated fatty acid synthesis. Interestingly, although both SREBP-1 and SREBP-2 are activated by the same machinery in mammalian cells, Sre1 and Mga2 are not processed coordinately (60). Instead, Sre1 is activated through the function of a Golgi-resident Dsc E3 ligase and a rhomboid protease, whereas S. cerevisiae Mga2 and Spt23 are activated by the Rsp5 ubiquitin ligase and the proteasome (21,53,61). Independent activation of Sre1 and Mga2 is supported by the fact that cells lacking a Dsc E3 ligase component show wild-type normoxic growth unlike mga2⌬ cells, indicating that the Dsc E3 ligase is not required for Mga2 activation (data not shown).
Despite differing mechanisms of regulation, both of these pathways are likely product inhibited. Work in mammalian cells has shown that SREBP trafficking by Scap is inhibited by exogenous cholesterol, although the parallel experiment cannot be performed in S. pombe as fission yeast do not uptake ergosterol (8,62,63). Similarly, work in S. cerevisiae showed a block in Mga2 activation when exogenous unsaturated fatty acid was added to the medium (61). Consistent with this result, activation of SREBP-1, but not SREBP-2, is inhibited by addition of exogenous arachidonate (20:4) in HEK293 cells (64). Because both ergosterol and unsaturated fatty acids require oxygen for synthesis, product inhibition of these two pathways would allow indirect sensing of oxygen availability. Additionally, as membrane production requires coordinated supply of both ergosterol and glycerophospholipids, cross-talk between these two pathways is likely. Future studies will examine coordination of these two oxygen-responsive pathways.
An important next step will be to examine whether unsaturated fatty acids regulate Mga2 through protein levels or activation. Alternatively, there could be other modes of regulation of these genes by oxygen. Our observation that Mga2 activates genes under normoxia and that expression is further increased under low oxygen conditions (Fig. 3A) suggests that there may be constitutive activation of Mga2 under all oxygen conditions and that an additional oxygen-responsive transcription factor may cooperate with Mga2 under low oxygen.
Finally, the Mga2 pathway represents a new avenue for antifungal drug discovery. Recent reports have shown that pathogenic fungi must adapt to the low oxygen environment of the host tissue to cause disease (15)(16)(17)65). Importantly, mga2 has apparent homologs in multiple pathogenic fungi, including Aspergillus fumigatus, Aspergillus flavus, Histoplasma capsulatum, Cryptococcus neoformans, Cryptococcus gattii, Candida albicans, and Candida glabrata. These homologs are characterized by an IPT (Ig-like, plexins, transcription factors) transcription factor immunoglobulin-like DNA binding domain, an ankyrin repeat, and a transmembrane domain near the C terminus (66,67). Many of these pathogenic fungi also have SREBP homologs, suggesting overall conservation of these parallel oxygen-adaptation pathways (15,17). Additionally, unlike the SREBP transcription factors, Mga2 has no apparent homologs in humans, limiting the potential toxic effects of drugs targeting this pathway. Future studies will characterize mga2 homologs in these organisms to determine whether they also regulate pathways required for low oxygen adaptation and fungal virulence.
Author Contributions-R. B. conducted the experiments except as noted below, analyzed the results, and wrote the paper. E. V. S. conducted preliminary experiments and provided results for the deletion collection screen and microarray experiment. W. S. assisted with qPCR experiments. S. Z. assisted with the deletion collection screen. H. K. H. B. and C. S. E. performed the lipidomics experiments and analysis. P. J. E. conceived the project and wrote the paper. All authors reviewed the results and approved the final version of this manuscript.