The membrane proteins, Spt23p and Mga2p, play distinct roles in the activation of Saccharomyces cerevisiae OLE1 gene expression. Fatty acid-mediated regulation of Mga2p activity is independent of its proteolytic processing into a soluble transcription activator.

The Saccharomyces OLE1 gene encodes the Delta-9 fatty acid desaturase, an enzyme that converts saturated fatty acyl-CoAs into cis-Delta-9 unsaturated fatty acids. OLE1 gene expression is regulated by unsaturated fatty acids, which repress transcription and destabilize the OLE1 mRNA. Expression of OLE1 is activated by N-terminal proteolytic fragments of two homologous endoplasmic reticulum membrane proteins, Spt23p and Mga2p. Disruption of either gene does not significantly affect cell growth or fatty acid metabolism; cells that contain null alleles of both genes, however, are unsaturated fatty acid auxotrophs. An analysis of spt23Delta and mga2Delta strains shows that Spt23p and Mga2p differentially activate and regulate OLE1 transcription. In glucose-grown cells, both genes activate transcription to similar levels of activity. Expressed alone, Mga2p induces high levels of OLE1 transcription in cells exposed to cobalt or grown in glycerol-containing medium. Spt23p expressed alone activates OLE1 transcription to levels similar to those in wild type cells. OLE1 expression is strongly repressed by unsaturated fatty acids in spt23Delta or mga2Delta cells, under all growth conditions. To test if OLE1 expression is controlled by fatty acids at the level of membrane proteolysis, soluble N-terminal fragments of Spt23p and Mga2p that lack their membrane-spanning regions (Deltatm) were expressed under the control of their native promoters in spt23Delta;mga2Delta cells. Under those conditions, Mga2pDeltatm acts as a powerful transcription activator that is strongly repressed by unsaturated fatty acids. By comparison, the Spt23pDeltatm polypeptide weakly activates transcription and shows little regulation by unsaturated fatty acids. Co-expression of the two soluble fragments results in activation to levels observed with the Mga2pDeltatm protein alone. The fatty acid repression of transcription under those conditions is attenuated by Spt23Deltatm, however, suggesting that the two proteins may interact to modulate OLE1 gene expression.

diseases. Because of the importance of lipids to numerous cellular functions, eukaryotes have developed complex mechanisms for regulating lipid biosynthetic activity and membrane lipid composition.
The Saccharomyces OLE1 1 gene is a central enzyme in cellular lipid metabolism. It encodes a ⌬-9 fatty acid desaturase, an intrinsic membrane enzyme that converts saturated fatty acyl-CoA substrates to monounsaturated fatty acid species by an oxygen-dependent mechanism (1)(2)(3). The unsaturated fatty acid products of Ole1p compose 75-80% of the fatty acyl groups in membrane lipids, and the regulation of its activity plays a dominant role in governing the composition and physical properties of membranes in growing cells. Expression of the OLE1 gene is regulated by a number of physiological and nutritional controls, including nutrient fatty acids (3)(4)(5)(6) and molecular oxygen (7)(8)(9). We have shown previously that unsaturated fatty acids exhibit strong repressive effects on OLE1 expression at the levels of transcription (6) and mRNA stability (5). Fatty acid-mediated repression of transcription is mediated through at least two elements in the OLE1 promoter. The fatty acid regulated (FAR) sequences (6) act as a primary transcription activation region under aerobic conditions. It can activate and confer fatty acid-mediated repression on a foreign gene when placed upstream of the Saccharomyces CYC1 basal promoter elements. We have also identified a second low oxygen response element (LORE) that is positioned ϳ200 bases downstream from the FAR element (7). That element acts with the FAR element to strongly activate transcription in response to hypoxic conditions and on induction by cobalt, which is thought to interfere with an unidentified oxygen sensor. Tandem copies of the LORE element that are placed in a heterologous promoter exhibit hypoxic induced transcription activation and can be repressed by unsaturated fatty acids (7,8).
Recent studies (10,11) have identified regulatory circuits that employ intrinsic membrane proteins that act as lipid sensors. These control gene activity by a process that involves regulated intramembrane proteolysis (10,11), which involves the proteolytic release of a fragment of an integral membrane protein that can serve as a transcription factor. The most well studied example is that of the sterol regulatory element-binding protein (SREBP), which is an integral endoplasmic reticulum (ER) membrane protein. Under conditions of sterol depletion, SREBP is cleaved by two membrane-bound proteases, releasing a soluble N-terminal polypeptide that contains DNA binding and transcription activation domains that participate in the activation of a number of lipid metabolic genes (10).
A second proteolytic activation mechanism of membrane proteins that controls Saccharomyces cerevisiae OLE1 gene expression has been described recently (12)(13)(14). SPT23 and MGA2 are homologous genes that encode ER-resident membrane proteins. Disruption of either gene has little effect on the growth or production of unsaturated fatty acids, whereas disruption of both creates a synthetic auxotrophy for unsaturated fatty acids due to the loss of OLE1 expression (12). Expression of N-terminal fragments of Spt23p and Mga2p further showed that those polypeptides, although they lack a functional DNA binding domain, serve as co-activators of OLE1 transcription (12,13). Membrane-bound Spt23p and Mga2p are apparently converted to these soluble active forms in vivo via a ubiquitinmediated mechanism that involves the cleavage of a soluble 90-kDa N-terminal fragment from its inactive 120-kDa membrane-bound precursor by the 26 S proteosome (14). Evidence from those studies suggests that Spt23p processing may be selectively regulated by unsaturated fatty acids. The unsaturated fatty acids 16:1, 18:2, and 18:3 inhibited processing of Spt23p to its active 90-kDa form, whereas exposure to 18:1 failed to block the proteolytic step. The lack of inhibition of Spt23p proteolysis by the fatty acid that has the highest melting temperature suggests that the regulation mechanism might detect changes in membrane "fluidity" caused by the incorporation of the fatty acids into membrane lipids.
We have determined recently (15) that Mga2p is essential for the hypoxic induction of OLE1 gene expression and that it is a component of the LORE DNA binding complex under those conditions. Given our previous observations (3-6) that OLE1 is regulated by a wide range of unsaturated fatty acids at the levels of transcription and mRNA stability, we wanted to determine whether the proteolytically processed forms of Spt23p and Mga2p also contribute to controlling OLE1 gene expression. To investigate whether the inhibition of Spt23p processing (and presumably a similar effect associated with Mga2p) is the basis for the fatty acid-mediated regulation of transcription, we tested the ability of cells to regulate OLE1 transcription when only soluble forms of the two proteins were expressed. In this paper, we demonstrate that Spt23p and Mga2p have distinct but overlapping functions with respect to the activation of OLE1 transcription. Evidence is presented that OLE1 expression is strongly repressed by unsaturated fatty acids when activated by a soluble form of Mga2p and that Spt23p may act either as an activator of OLE1 transcription or as a negative modulator of Mga2p fatty acid-mediated regulation.

EXPERIMENTAL PROCEDURES
Strains and Growth Media-S. cerevisiae strains used in this study are shown in Table I. Plasmids constructed for this study are shown in Table II. Standard yeast genetics methods were used for construction of strains bearing the appropriate mutations or gene disruptions. Yeast cells were grown at 30°C in SDt (synthetic dextrose tergitol) drop out medium as described previously (6,16). All growth media used in liquid cultures contained 1% tergitol Nonidet P-40 to disperse fatty acid supplements. Tergitol-Nonidet P-40 is not derived from fatty acids or fatty alcohols and is apparently not metabolized by yeast. 3% glycerol was used as an alternative carbon source for glucose in synthetic dropout media designated as SGt. Tergitol was obtained from Sigma. Fatty acids were obtained from Nu Chek Prep (Elysian, MN). Escherichia coli DH5␣ competent cells were obtained from Life Technologies, Inc. Growth tests were performed by monitoring A 600 of the cultures or by hemocytometer counting.
Cloning and Disruption of SPT23 and MGA2-Strains containing disrupted forms of the SPT23 and MGA2 genes derived from the DTY11A background were constructed by standard yeast molecular biological methods using the cloned native gene sequences. DNA fragments encoding the SPT23 genes and MGA2 genes used for the gene disruptions were cloned by polymerase chain reaction (PCR) using S. cerevisiae strain DTY10A genomic DNA as a template. Oligonucleotide primers that were used for the PCR cloning are shown in Table III. Gene-disrupted strains in cells derived from the BY4741 backgrounds were obtained from Research Genetics Corp. (Huntsville, AL).
The SPT23 gene was isolated using the PCR primer pair SF and SR, which amplifies a fragment extending from bases Ϫ1003 to ϩ2152 of the chromosomal gene. Amplified fragments were ligated into vector pCRscript SKϩ. To assemble the gene disruption construct, a fragment extending from residues Ϫ48 to ϩ1351 was removed by restriction digestion with XbaI and SpeI. The remaining vector fragment containing flanking SPT23 sequences was blunt-ended and ligated with a blunt-ended HindIII fragment derived from the Saccharomyces LEU2 gene to create the vector pCRspt23⌬::LEU2. The chromosomal SPT23 gene was disrupted by transformation of the DTY11A strain with a DNA fragment isolated from the vector by digestion with restriction enzymes PstI and NotI.
The MGA2 gene was cloned using a similar strategy using the PCR primer pair MF and MR, which amplified a fragment extending from nucleotides Ϫ242 to ϩ2992 of the chromosomal DNA. The gene disruption construct was created by replacing an XbaI/HpaI fragment with the LEU2 gene to create vector pCRmga2⌬::LEU2 by the above procedure. The chromosomal MGA2 gene was disrupted with a DNA fragment derived from the vector by digestion with the restriction enzymes PstI and NotI.
Construction of the spt23⌬;mga2⌬ Strains-Strains containing disrupted forms of both spt23⌬ and mga2⌬ genes were constructed by crossing strains containing a disruption of one of the two genes. The diploid parent strain was sporulated, and haploid progeny were screened by comparison of growth on agar medium containing 0.5 mM 16:1 and 0.5 mM 18:1 or no fatty acids. Candidate strains that did not grow on the fatty acid-free medium were verified as authentic spt23⌬; mga2⌬ strains by PCR analysis of both genes. Construction of SPT23⌬tm and MGA2⌬tm Expression Plasmids-To construct plasmids that express the ϳ90-kDa soluble forms of Spt23p and Mga2p under control of their native promoters, DNA fragments were amplified from their respective genes with oligonucleotide pairs MGApro-F/MGA2rev2 and SPTpro-F/SPT23rev1. Each amplified gene fragment was cloned into pCR-Script SKϩ to yield plasmids pCRmga2N⌬tm and pCRspt23N⌬tm. Fragments containing the promoter and protein-coding sequences for each gene were then isolated by NotI and XhoI digestion and inserted into the centromeric plasmid pRS413 (Stratagene) yielding plasmids pRSmga2N⌬tm and pRSspt23N⌬tm.
␤-Galactosidase Assays-␤-Galactosidase activities of cells containing the lacZ reporter plasmids were performed as described previously (6). Cell densities for the assays were determined by measurement at A 600 , and corrections for light scattering were made at A 550 . At least three independent transformants were assayed for each experiment, and ␤-galactosidase activities are the result of at least three independent experiments. Each experimental assay was performed in duplicate.
Fatty Acid Repression/Derepression Studies-Cells containing OLE1 promoter::lacZ reporter genes were tested according to protocols published previously (3,6) that measure induction of reporter gene activity from the fatty acid-repressed state. Cells were pre-grown in the appropriate SDt dropout medium supplemented with 1% tergitol and 1 mM 18:2. Those cultures were used to inoculate (at 1/10 volume) fresh fatty acid-free medium or medium containing 1 mM 18:2, metal ions, or carbon source. Cells were then incubated for 6 -24 h and harvested by centrifugation for analysis of ␤-galactosidase activity.
RNA Isolation, Quantitative Northern Blot Analysis, and Fatty Acid Analysis-RNA was isolated for Northern blot analysis and quantification by PhosphorImaging as described previously (5). Fatty acid methyl esters were extracted from cell pellets derived from strains grown on fatty acid-free medium and were first washed with cold water. Strains grown on fatty acid-supplemented media were first washed with 1% tergitol in deionized water and then cold water. Methyl esters were analyzed by gas chromatography as described previously (17).

Effects of Disruption of the SPT23 and MGA2 Genes on
Growth and Fatty Acid Desaturase Activity-To examine the independent roles of the SPT23 and MGA2 genes on OLE1 expression, we tested the effects on growth and cellular fatty acid composition in strains that contained disrupted forms of the two genes (Table IV). Although there were reduced levels of 16:1 in the gene disrupted strains, these were compensated for by increases in 18:1, so that all strains contained Ͼ68% total (16:1 ϩ 18:1) unsaturated fatty acids. This indicates that disruption of either gene does not produce strong effects on overall OLE1 expression under normal laboratory growth conditions. A spt23⌬;mga2⌬ strain containing null alleles in both genes, however, exhibited a strict requirement for unsaturated fatty acids on liquid minimal medium, which supports the previous observations (12) that SPT23 and MGA2 have overlapping functions that are essential for OLE1 gene expression.
Previous studies (12) using a strain that contained a disrupted MGA2 gene and a temperature-sensitive allele of SPT23 showed that very high levels of unsaturated fatty acids (ϳ50%) persist in cellular lipids 15 h after a shift from permissive to restrictive temperatures. This suggested that high basal levels of OLE1 expression might occur in the absence of both proteins. To determine whether OLE1 continues to be expressed under those conditions, we examined the effects of fatty acid deprivation on the growth and relative desaturase activity of wild type and spt23⌬; mga2⌬ cells that contained disrupted forms of both genes (Fig. 1,  A and B). To monitor the in vivo activity of the desaturase, the spt23⌬;mga2⌬ and wild type strains were initially grown on medium containing linoleic acid (18:2). Under those conditions,

TABLE III PCR primers used in this study
Oligonucleotides are designated by their homology to target gene sequences and numbered with respect to the position of the 5Ј end of the primer relative to the ϩ1 A of the ATG start codon.

Fatty Acid Control of OLE1 Expression by Spt23p and Mga2p
wild type desaturase activity is repressed, and over multiple generations, the endogenous 16:1 and 18:1 products of the enzyme are replaced in cellular lipids by the exogenous 18:2. Cells were then transferred to fatty acid-free medium to follow the induction of Ole1p-dependent desaturase activity. When the spt23⌬;mga2⌬ cells were transferred to fatty acidfree medium, growth proceeded at wild type rates for ϳ2 generations before slowing, and stationary growth occurred ϳ3.5 generations after the transfer. In both strains the most abundant fatty acids at the time of transfer were 18:2 (ϳ70 weight %) and the saturated species 16:0 and 18:0 (ϳ20 weight %) (Fig.  1B). Very low cumulative levels of 16:1 and 18:1 were found in the wild type (ϳ7%) and spt23⌬;mga2⌬ (ϳ5%) strains, suggesting that OLE1 might be expressed at low basal levels, even under fatty acid-repressed conditions. The wild type cells exhibited the previously observed pattern of OLE1 induction after transfer to fatty acid-free medium. Eight hours after the transfer, levels of the assimilated 18:2 dropped ϳ5-fold, and the 16:1 and 18:1 products of the induced desaturase increased to greater than 60% of the total fatty acid mass. Under the same conditions, the spt23⌬;mga2⌬ cells showed no significant increases in 16:1 and 18:1 levels, indicating that OLE1 expression was not induced by fatty acid depletion significantly above the previously observed basal levels. At the same time, there was a striking increase in saturated fatty acids and a reduction in 18:2 in the spt23⌬;mga2⌬ strain, which would be expected if saturated fatty acid synthesis continued in the absence of induced desaturase activity.

OLE1 mRNA Levels in spt23⌬ and mga2⌬ Cells Are Regulated by Unsaturated Fatty Acids in a Manner Consistent with
Wild Type Cells-The small differences in unsaturated fatty acid levels in wild type, spt23⌬, and mga2⌬ cells suggested that either gene could independently activate and regulate OLE1 expression under normal laboratory growth conditions. This was confirmed by quantitative Northern blot experiments (Fig.  2) that showed that disruption of either gene did not affect OLE1 steady state mRNA levels in either fatty acid-free or 18:2-supplemented growth medium containing glucose as a carbon source. Under the conditions used in the experiment, OLE1 mRNA levels were strongly suppressed by exposure to the unsaturated fatty acid.
In Spt23p-deficient Cells, Mga2p Elicits a Strong OLE1 Transcription Activation Response to Glycerol and Cobalt-To test the roles of Spt23p and Mga2p on OLE1 transcription, a lacZ reporter gene that was linked to 934 bases of the OLE1 promoter was transformed into the spt23⌬ and mga2⌬ genedisrupted strains. Cells were grown under conditions that we had observed previously (6, 7) to alter OLE1 transcription activity, including repression by unsaturated fatty acids and activation in response to cobalt and the non-fermentable carbon source, glycerol.
The effects of the gene disruptions on reporter gene activity in cells grown on glucose and glycerol carbon sources are shown in Fig. 3. To prevent residual ␤-galactosidase activity from masking the effects of fatty acid repression, cells were initially grown for 12 h in the appropriate medium containing 1 mM 18:2. Aliquots were then washed with fatty acid-free medium and grown for 6 h prior to the assays in either fatty acid-free medium or medium containing 1 mM 18:2.
Wild type, spt23⌬, and mga2⌬ mutant cells grown on glucose-containing medium showed no significant differences in reporter gene activity under derepressed or 18:2 repressed conditions. Under the conditions used in the experiment, the 18:2 repressed reporter gene activity to ϳ20% of the levels seen in derepressed cells. Distinct phenotypic differences were observed, however, in cells that were grown on glycerol. In the wild type and mga2⌬ strains, transfer to glycerol medium without fatty acid resulted in a 1.5-2-fold induction in reporter gene activity that was repressed by 18:2 to levels similar to those in glucose-grown cells. Deletion of the SPT23 gene, however, resulted in a striking 6-fold induction of reporter activity in fatty acid-free glycerol medium. Furthermore, the addition of 18:2 in the medium repressed the reporter gene activity to the same levels found in the wild type and mutant cells, resulting in a 20-fold increase in the range of transcription repression.
Previous studies (7,9) have shown that OLE1 expression is also strongly induced by cobalt. Again, Mga2p and Spt23p appear to play distinct roles in the induction response (Fig. 4, a  and b). Exposure of logarithmic phase cells to 400 M cobalt chloride for 6 h resulted in a 4 -5-fold induction of reporter gene activity in wild type and spt23⌬;MGA2 cultures. By contrast, cells containing the disrupted MGA2 gene showed no significant response under the same conditions. Reporter activity in all of the strains exposed to cobalt was repressed by 18:2. Cells containing only the functional Mga2p showed the most dramatic range of repression (Ͼ20-fold) between fatty acid-free cultures and those exposed to 18:2.
An analysis of OLE1 steady state mRNA levels (Fig. 4b) also illustrates the effects of Mga2p in the cobalt induction response. In wild type and spt23⌬;MGA2 cells, which express Mga2p, there is an equivalent 5-6-fold increase in mRNA levels of cells grown on fatty acid-free medium. Although reporter gene activity is induced to higher levels in the spt23⌬ cells, the resulting mRNA levels between that strain and wild type are equivalent, suggesting that post-transcriptional mechanisms might operate to maintain the transcripts at similar levels. By comparison, mRNA levels in mga2⌬ cells, which express only Spt23p, show no inductive response to cobalt. Under the conditions of the experiment, mRNA levels in all three strains are repressed by exposure to 18:2, which correlates with reporter gene activities.

FIG. 2. Quantitative Northern blot hybridization of OLE1
mRNA in spt23⌬ and mga2⌬ cells exposed to 1 mM 18:2 and 2 h after transfer to fatty acid-free medium. spt23⌬;MGA2 and SPT23; mga2⌬ cells were pre-grown on 1 mM 18:2 to a cell density of less than 1 ϫ 10 7 cells/ml. Aliquots of the culture were then washed as described under "Experimental Procedures" and transferred to fresh medium containing 1 mM 18:2 or fatty acid-free medium. Total RNA was isolated 2 h after the transfer and subjected to Northern blot analysis. Blots were probed with radiolabeled OLE1 DNA and then stripped and reprobed with DNA complementary to the Saccharomyces PGK1 gene. Images are derived from PhosphorImaging.

FIG. 3. Gene activation and 18:2-mediated repression of OLE1
promoter::lacZ reporter activity in wild type, spt23⌬, and mga2⌬ cells grown on glucose or glycerol. Cells containing the p62 plasmid were grown for at least 6 generations to mid-logarithmic growth phase in synthetic medium containing either 2% glucose ϩ 1 mM 18:2 or 3% glycerol ϩ 1 mM 18:2 to repress reporter gene activity and to reduce residual ␤-galactosidase activity. Cultures were harvested at a density of Ͻ2 ϫ 10 7 /ml, washed with the corresponding fatty acid-free medium, and then transferred to fresh medium with no fatty acids or with 1 mM 18:2. Assays were performed 6 h after the transfer. Data represents the average of three independent transformants performed in triplicate.

The Induction Responses to Cobalt and Glycerol Are Additive
and Synergistic-We showed previously (7) that cobalt can induce OLE1 transcription through the LORE. Given that glycerol requires oxygen-dependent respiration to be metabolized, we asked whether the strong activation of reporter gene activities in the spt23⌬ cells by glycerol and cobalt might be caused by the same function. Fig. 5 shows that the addition of cobalt to cells grown on glycerol stimulates activity to 2-3-fold higher levels than glycerol alone in all three strains, suggesting that the two effects occur by different mechanisms. The spt23⌬; MGA2 cells, which only express Mga2p, exhibit the highest levels of reporter gene activity under those conditions. In those cells, the activity induced by the combination of glycerol and cobalt is the approximate sum of the activities observed in cells that were separately induced. In the wild type and SPT23; mga2⌬ cells, however, induction by the combined stimuli resulted in even higher activities than the sum of the individual responses. The combined glycerol-and cobalt-induced reporter activities in all the tested strains were also strongly repressed by exposure to 18:2. These results suggest that the glycerol and cobalt induction are triggered by separate mechanisms, although their apparent synergism suggests that these processes may be also linked to some common elements that activate transcription.
The Cobalt and Glycerol Responses Act through Different OLE1 Promoter Elements-The fatty acid-regulated components of the OLE1 transcription unit include the FAR element, which is located ϳ580 bases upstream of the start codon (6), and a 30-base pair LORE element that is positioned ϳ200 bases downstream of the FAR region. The LORE region is a strong activator of transcription when cells are starved for oxygen (7) or exposed to cobalt. Although these two effects are closely linked, we have shown recently (7) that point mutations within the LORE region can suppress the response to hypoxia without affecting induction by cobalt. Given the differential effects produced by glycerol and cobalt on gene activation, we theorized that the FAR and LORE elements might play specific roles in those responses and thus provide useful tools for investigations into Mga2p and Spt23p functions. To test this possibility, we examined the expression of lacZ reporter genes under control of the isolated FAR and LORE elements and reporter plasmids that contain mutations in those regions FIG. 5. Gene activation and 18:2-mediated repression of OLE1 promoter::lacZ reporter activity in wild type, spt23⌬, and mga2⌬ cells exposed to 400 M cobalt. Cultures were pre-grown in glycerol ϩ 1 mM 18:2-supplemented media containing no cobalt or 400 M cobalt as described in Fig. 3. Cells were harvested and washed with the corresponding fatty acid-free medium and then suspended in either fresh fatty acid-free or 1 mM 18:2-supplemented medium. Assays were performed 6 h after the transfer as described in Fig. 3.   FIG. 4. Gene activation and 18:2-mediated repression of OLE1 promoter ::lacZ reporter activity in wild type, spt23⌬, and mga2⌬ cells exposed to 400 M cobalt. a, p62 lacZ reporter gene activity. Cultures were pre-grown in glucose ϩ 1 mM 18:2-supplemented media containing no cobalt or 400 M cobalt as described in Fig. 3. Cells were harvested and washed with the corresponding fatty acid-free medium and then suspended in either fresh fatty acidfree or 1 mM 18:2-supplemented medium Ϯ 400 M cobalt chloride. Assays were performed 6 h after the transfer as described in Fig. 3. b, steady state OLE1 mRNA levels.
Cultures were pre-grown as described in a, and total cellular RNA was isolated for Northern blot analysis 3 h after the transfer to fresh medium.
within the 934 bases upstream of the OLE1 start codon. The results of that study are shown in Table V.
Plasmids that contained the isolated FAR and LORE elements were not induced by glycerol but were activated by exposure to cobalt. The plasmid containing the LORE element showed the strongest response to cobalt (Ͼ10-fold over glucosegrown cells), whereas the plasmid containing the FAR element was activated slightly more than 2-fold. The two reporter plasmids that contain mutations within the context of the 934-base upstream region were also differentially induced by glycerol and cobalt. Plasmid p62::⌬88, which contains an 88-base pair deletion in the FAR element, is induced ϳ2.3-fold by glycerol and about 4.8-fold by cobalt. Plasmid pAM4, which contains a 3-base substitution in the LORE region that abolishes the hypoxia response but not the cobalt response (7) is not induced by glycerol but is induced ϳ5-fold by cobalt. The complete inactivation of the glycerol response caused by a 3-base substitution in the LORE element within the 934-base sequence, combined with the inability of the isolated LORE element to be activated by glycerol, indicates that elements contained within the LORE region are necessary, but not sufficient, for the glycerol induction response.
Taken together, these data show that the glycerol and cobalt responses are associated with different OLE1 upstream activation sequences. These data further suggest that both responses act through multiple sites within the OLE1 promoter, including some that have not yet been identified. (14) indicated that the proteolytic processing of the membrane-bound forms of Spt23p and Mga2p to their soluble forms is essential for the expression of OLE1. That study showed that Spt23p processing could be inhibited by exposing cells to 16:1, 18:2, and 18:3 but not 18:1. This suggested that increased membrane fluidity, caused by the incorporation of fatty acids with lower melting temperatures into membrane lipids, might be the basis of the fatty acid-mediated regulation of OLE1 expression.

OLE1 Expression Activated by a Soluble Form of Mga2p Is Regulated by 18:2-Previous studies
To clarify the potential regulatory role of the proteolytic processing step on OLE1 expression, we constructed genes that express soluble forms of Spt23p and Mga2p by deleting the regions that encode their C-terminal membrane-anchoring domains. Given the very low levels of expression by the chromosomal SPT23 and MGA2 genes, the truncated genes were placed under control of their native promoters in single copy, centromere-based vectors.
Plasmid Spt23p⌬tm contains 1499 bases of the SPT23 promoter and the N-terminal protein-coding sequence encoding amino acid residues 1-792. Plasmid Mga2p⌬tm contains 904 bases of the MGA2 promoter and encodes N-terminal residues 1-784 of the protein. Expression of either plasmid in a spt23⌬; mga2⌬ strain repaired its unsaturated fatty requirement. This demonstrated that the soluble protein fragment from either gene could activate OLE1 transcription. An examination of p62 OLE1::lacZ reporter gene activity in the same cells, however, showed striking differences in transcription activation levels induced by the truncated forms of Mga2p and Spt23p (Fig. 6). The soluble Spt23p⌬tm protein activated the reporter gene in the spt23⌬;mga2⌬ cells to levels similar (1.0 -1.5-fold) to those seen in the wild type and SPT23;mga2⌬ strains ( Fig. 6, also cf.  Fig. 3). Furthermore, the reporter gene activity induced by the truncated protein was only weakly repressed by 18:2 (ϳ25%). This differed markedly from the strong repression observed in wild type and SPT23;mga2⌬ strains that produce full-length Spt23p (Figs. 3 and 4).
Expression of the soluble Mga2⌬tm protein, however, resulted in a 25-fold increase over the levels of reporter gene activity produced by its native form in wild type or spt23⌬; MGA2 cells. Furthermore, that activity was very strongly repressed by 18:2. To test the possibility that this suppression might occur after the proteolytic processing step, p62 reporter gene activity was measured in spt23⌬;mga2⌬ cells that co-expressed the soluble forms of the two proteins. In fatty acid-free medium, reporter gene activity was found to be the same as when only the Mga2p⌬tm protein was expressed, suggesting that Spt23p⌬tm had no significant effect on the Mga2p⌬tmmediated activation (Fig. 6). The combined expression of both soluble proteins, however, suppressed the strong 18:2mediated repression of the reporter activity, which remained at 75% of the levels found in fatty acid-free cultures. The resulting fatty acid "repressed" levels in the cells that coexpressed Spt23p⌬tm and Mga2p⌬tm were 6-fold greater than those found in wild type cells grown in fatty acid-free medium and Ͼ60-fold higher than the corresponding wild type fatty acid repressed activity.
Effect of Spt23⌬tm and Mga2⌬tm on Steady State mRNA Levels-Quantitative Northern blots of OLE1 mRNA isolated from strains expressing the soluble Spt23⌬tm and the Mga2⌬tm proteins are shown in Fig. 7. As expected, mRNA levels in the spt23⌬;mga2⌬ strain that contained the empty vector are severely reduced compared with the levels found in wild type cells. Expression of the truncated form of the Spt23⌬tm protein in the gene-disrupted strain results in mRNA levels that are consistent with the above reporter gene activities. Transcript levels in those cells grown on fatty acidfree medium are nearly identical to those found in wild type cells and they remain at the same levels after release from fatty acid repression.
By contrast, mRNA levels in cells that express the soluble Mga2p⌬tm protein are about 5-fold higher than those in the wild type cells, and they are strongly repressed by exposure to 18:2. OLE1 transcript levels in cells that co-express both soluble proteins are ϳ7.5-fold greater than wild type cell levels when grown on fatty acid-free medium, and as predicted from the reporter gene assays, they are unaffected by exposure to 18:2.

DISCUSSION
The regulation of membrane fatty acid composition requires cells to respond to a complex set of stimuli. In addition to physiological controls that monitor the demand for membrane assembly, cells must detect and respond to changes of bilayer fluidity, the availability of nutrient lipids, and changes in the composition of intracellular fatty acid pools to regulate lipid biosynthetic activity. Fatty acid desaturases represent important control points in these metabolic systems because their unsaturated products compose most (70 -80%) of the acyl groups found in membrane lipids. Furthermore, manipulations of their activities can produce radical changes in the physical state of membrane lipid bilayers that, in turn, can affect numerous membrane-associated cellular functions. The finding that the activity of the two homologous membrane proteins, Spt23p and Mga2p, is essential for OLE1 gene activation raises several important questions about their functional roles in the regulation of the desaturase activity. Previous reports (13) have indicated that these proteins act as transcription co-factors. Although they do not have recognizable DNA binding domains, N-terminal elements of the proteins can activate transcription when they are fused to the DNA binding domain of the GAL4 transcription factor (13). Furthermore, overexpression of Spt23p and Mga2p can suppress mutations that affect SNF/SWI-mediated transcription activation, leading to the proposal that they may do so by influencing chromatin accessibility (13).
The studies presented here indicate that under normal growth conditions in aerated glucose medium, Spt23p and Mga2p essentially perform overlapping functions. The independent expression of either gene produces almost equal levels of OLE1 transcription activity and can maintain normal OLE1 steady state mRNA levels under derepressed and unsaturated fatty acid-repressed conditions. Furthermore, disruption of ei-ther gene has little effect on the total unsaturated fatty acid levels in cellular lipids. Significant differences in Spt23p and Mga2p functions become evident, however, under conditions when OLE1 transcription is induced above the glucose-grown basal levels by exposure to glycerol or to cobalt. The highest levels of induction to both stimuli were observed in cells that only express Mga2p, which appears to be the dominant transcription activator. Furthermore, the higher levels of OLE1 induction by cobalt or glycerol in spt23⌬ cells (in which Mga2p is independently expressed) compared with wild type cells (in which both proteins are expressed) suggest that Spt23p partially suppresses Mga2p activity under those conditions. The observation that the soluble form of Spt23p does not suppress Mga2p-dependent activation suggests that this effect could take place at the level of the membrane. This might occur through competition for Rsp5p (14), the ubiquitin ligase that activates the proteolytic conversion of Spt23p and Mga2p to their active, soluble forms.
The glycerol and the cobalt induction processes appear to act through different elements in the OLE1 promoter. Although evidence presented here suggests that glycerol induction requires elements in the LORE region, it also appears to require additional promoter sequences that are distinct from the FAR and LORE elements (6,7). Cobalt stimulates transcription activity through the LORE element and additionally induces activity through the FAR region. These observations suggest that both types of induction act through multiple but distinct sites in the OLE1 promoter region. The observation that Mga2p associates with the LORE element under hypoxic conditions (15) also suggests that Mga2p and Spt23p may associate with additional DNA-binding complexes on the OLE1 promoter and that it will be important to determine those interactions to define the functions of the two proteins.
The observation that proteolytic processing of Spt23p can be inhibited by certain unsaturated fatty acids offers a compelling mechanism for the fatty acid-mediated repression of OLE1 transcription. If the proteolytic release of both proteins from the membrane were blocked by the exogenous fatty acids, then transcription activity could be effectively repressed by a mechanism that responds only to changes in the fatty acid composition of the surrounding membrane lipids. That type of control correlates with recently described (18 -20) regulatory pathways that release other membrane-bound transcription factors through intramembrane proteolysis. These include the SREBP that is regulated by cholesterol and the ATF6 and IRE1 tran- scription factors that are regulated by an unfolded protein response that is triggered by ER stress (18 -20). Observations by Hoppe et al. (14) indicate that, although unsaturated fatty acids exert tight control over the processing of Spt23p, they do not have a strong effect on the proteolytic processing of Mga2p. Given the observations here that Mga2p acts as the primary activator of OLE1 transcription in cobalt-and glycerol-induced cells and is strongly repressed by fatty acids suggests that a membrane fluidity-based mechanism may not be the sole determinant of that regulatory response.
To test further the hypothesis that OLE1 repression occurs through a fluidity-regulated mechanism, we expressed soluble (90 kDa) forms of Spt23p and Mga2p in a strain containing disrupted alleles of both genes. Those experiments also revealed the striking differences in the functions of the two proteins and uncovered a second level of fatty acid-mediated control. Spt23p⌬tm was found to activate transcription at a level similar to the basal activity found in wild type cells, but this was only weakly repressed (ϳ25%) by exposure to 18:2. Mga2p⌬tm, however, activated OLE1 transcription at levels that were 20-fold greater than wild type cells and was strongly repressed by 18:2. The latter observation shows that fatty acid-mediated regulation of OLE1 can occur independent of the proteolytic release of that protein from the membrane. Because Mga2p processing appears to be a stronger transcription activator, and its processing is not suppressed by unsaturated fatty acids in glucose grown cells, it is possible that the primary mechanism of fatty acid-mediated repression occurs downstream of Spt23p/Mga2p proteolytic processing.
The co-expression experiments involving the Spt23p⌬tm and Mga2p⌬tm suggest a mechanism by which the two proteolytically processed proteins might interact, either with each other or with a common set of proteins, to modulate transcription activity. Unlike the effects seen in cells that express the fulllength proteins, Spt23p⌬tm does not attenuate the high levels of transcription activation produced by Mga2p⌬tm when the two proteins are co-expressed. It is particularly striking, however, that it exerts a strong suppressive effect on the fatty acid-mediated repression process. One possible mechanism for that effect might involve an unsaturated fatty acid (or fatty acyl-CoA)-binding protein that associates with the soluble forms of Mga2p⌬tm and blocks its transcription activity. When the soluble form of Spt23p is co-expressed, it might compete for limited supplies of the fatty acid-binding regulator, thus preventing it from binding and suppressing Mga2p-dependent gene activation.
The observation that OLE1 mRNA levels are higher when Spt23p⌬tm and Mga2p⌬tm are co-expressed than when Mga2p⌬tm is expressed alone might be the result of several possible conditions. One is that the OLE1 mRNA levels in cells that only express Mga2p⌬tm may recover slowly following re-lease from fatty acid repression and may not have reached the high steady state levels when analyzed. Alternatively, since OLE1 mRNA levels are affected by both transcription activity and the rate of mRNA degradation (5), post-transcriptional control mechanisms, such as modulation of OLE1 mRNA stability, might be influenced by the expression of the two soluble proteins, resulting in increased levels of transcript. We have recently observed that Mga2p is essential for the fatty acidregulated stability of the OLE1 transcript, suggesting that this protein might play a role in the functional linkage between transcription and mRNA stability. 2 This problem is currently under investigation.
Taken together, the experiments described here suggest that regulation of SPT23/MGA2-activated OLE1 expression acts in multiple cellular locations that may include the nucleus as well as the ER membrane surface. It appears unlikely that membrane fluidity-dependent regulated proteolysis is the sole determinant of this process but rather that the control of OLE1 expression occurs through several different mechanisms that allow desaturase activity to respond to the multiple metabolic and physiological demands for unsaturated fatty acids.