Long Chain Base Tolerance in Saccharomyces cerevisiae Is Induced by Retrograde Signals from the Mitochondria*

Saccharomyces cerevisiae cells lacking their mitochondrial DNA (ρ0 cells) respond to this loss of genetic information by induction of a program of nuclear gene expression called the retrograde response. Expression of genes involved in multidrug resistance and sphingolipid biosynthesis is coordinately induced in ρ0 cells by the zinc cluster transcription factor Pdr3p. In this report, we identify a membrane protein involved in control of intracellular levels of a sphingolipid precursor as a transcriptional target of the Pdr3p-mediated retrograde response. These sphingolipid precursors are called long chain bases (LCBs) and increased LCB levels are growth inhibitory. This membrane protein has been designated Rsb1p and has previously been shown to act as a LCB transporter protein and to be a component of the endoplasmic reticulum. These earlier studies used an amino-terminal truncated form of Rsb1p. Here we employ a full-length form of Rsb1p and find that this protein is localized to the plasma membrane and is modified by N-linked glycosylation. Two glycosylation sites are present in the Rsb1p and both are required for normal LCB resistance. Mutational analysis of the RSB1 promoter revealed that two Pdr3p binding sites are present and both of these are required for normal retrograde induction of transcription. LCB tolerance is strongly increased in ρ0 cells but this increase is ablated in ρ0 rsb1Δ cells. Together, these data indicate Pdr3p activation of RSB1 transcription is an important feature of the retrograde response allowing normal detoxification of an endogenous sphingolipid precursor.

Sphingolipids are important lipid constituents of eukaryotic membranes. Key intermediates in the biosynthesis of sphingolipids are the long chain bases (LCBs). 2 In Saccharomyces cerevisiae, these LCBs are dihydrosphingosine and phytosphingosine (PHS) (reviewed in Ref. 1). These LCBs are utilized by ceramide synthase to form ceramide that is a well known bioactive lipid (reviewed in Refs. 2 and 3). However, LCBs and their phosphorylated forms (LCBPs) have also recently been found to serve as potent signaling molecules. LCBs appear to act as inhibitors of proliferation, whereas increased LCBP levels stimulate growth (recently reviewed in Ref. 4). Control of the biological levels of these signaling lipids is an essential feature for ensuring both normal lipid composition of membranes and proper metabolic regulation.
The central importance of these signaling and structural lipids is likely to explain the multitude of regulatory mechanisms modulating their levels. These include enzymes like ceramidases that break down ceramide (5,6) and the LCBP lyase that cleaves these phosphorylated lipids into ethanolamine and a long chain aldehyde (7). More recent experiments in the yeast S. cerevisiae have identified a new membrane protein that is thought to efflux LCBs out of the cell. This protein has been designated Rsb1p (resistant to sphingoid bases) as it was identified as a high-copy suppressor of the PHS hypersensitivity of a dpl1⌬ strain (8). Biochemical experiments provided evidence that Rsb1p stimulates the transport of radiolabeled dihydrosphingosine out of cells.
Further analyses of Rsb1p made an interesting connection between this putative transporter protein and multidrug resistance in S. cerevisiae (9). Disruption of the ATP-binding cassette transporter-encoding gene PDR5 elevated PHS tolerance and induced transcription of the RSB1 gene. PDR5 is a major determinant of multidrug resistance and is transcriptionally regulated itself by the zinc cluster-containing transcription factors Pdr1p and Pdr3p (see Ref. 10 for a recent review). Previous microarray experiments have indicated that PDR5 and RSB1 (systematic locus name YOR049c) are among the most highly induced transcripts by the presence of hyperactive alleles of either PDR1 or PDR3 (11,12). Analysis of pdr5⌬ cells argued that loss of this ABC transporter led to the induction of Pdr1p but not Pdr3p activity with accompanying elevation in RSB1 transcription (9). This transcriptional activation required the presence of at least one Pdr1p/Pdr3p response element (PDRE) located upstream of RSB1. Sucrose gradient experiments (9) coupled with previous indirect immunofluorescence experiments (8) led to the assignment of Rsb1p as an integral membrane protein of the endoplasmic reticulum.
We have previously found that two genes encoding enzymes involved in sphingolipid biosynthesis are transcriptionally responsive to loss of the mitochondrial genome ( 0 cells). The signaling pathway linking nuclear gene expression with mitochondrial status has been designated retrograde regulation (see Ref. 13 for a recent review). LAC1, encoding a component of ceramide synthase (14,15), and IPT1, encoding the inositol phosphotransferase enzyme (16), are both transcriptionally upregulated in 0 cells (17,18). Importantly, 0 cells also strongly induce expression of both PDR5 and RSB1 in response to loss of the mitochondrial genome in a strictly Pdr3p-dependent fashion (19). Previous studies have established that 0 cells are highly resistant to drugs like cycloheximide because of this transcriptional induction of PDR5 (20 -22). To determine whether retrograde signaling in 0 cells also leads to increased PHS tolerance, we have examined the PHS phenotype of 0 cells. Our experiments demonstrate that 0 cells induce RSB1 transcription in a Pdr3p-dependent fashion and exhibit a large increase in resistance to exogenous PHS challenge. Analyses of the RSB1 locus and protein indicate that the sequence deposited in the Saccharomyces Genome Data base contains an error in the amino terminus that leads to the production of a truncated protein lacking 28 amino acids from the N terminus. A full-length form of Rsb1p containing these additional 28 residues localizes to the plasma membrane, and fully complements rsb1⌬ cells, unlike the truncated membrane protein. Two potential N-linked glycosylation sites are present in this 28-amino acid region and site-directed mutagenesis of these sites demonstrated that both must be present for normal Rsb1p function. These experiments provide new insight into the likely role of Rsb1p in control of PHS levels as well as illuminating the extensive interaction between the Pdr regulatory system and genes involved in sphingolipid homeostasis.

EXPERIMENTAL PROCEDURES
Yeast Strains and Media-The genotypes of the yeast strains used in this study are listed in Table 1. Yeast transformations were performed using the lithium acetate procedure (23). Cells were grown in cultures containing YPD (2% yeast extract, 1% peptone, 2% glucose) under nonselective conditions or appropriate SC media under selective conditions (24). PHS resistance was measured by spot test assay on plates with different concentrations of PHS with Nonidet P-40 included as described (8). ␤-Galactosidase activity determinations using o-nitrophenyl-␤-D-galactopyranoside hydrolysis was measured as described previously (25).
To introduce the dpl1⌬ allele, we amplified the corresponding region from the genomic DNA of BY4742 dpl1-⌬::kanMX4 with primers ACAC-AGTTGCCTATCGTTTATCGCC and AATTCAAGGTGATACAAG-TCGTCGTCC, transformed into the strains of interest. Transformants were selected on plates containing geneticin (200 g/ml) and the genotype confirmed by PCR with the 5Ј primer, TGCCGCAAATGGTACACGGT-TTAG, and the 3Ј primer, CAATGGTAGAAAGACACACACCTGCG.
To generate the rsb1-⌬1::natMX4 allele, the plasmid p4339 (from Brenda Andrews) (26) was used as a template to amplify the natMX4 cassette. Primers were used that contained 40 nucleotides of RSB1 sequence either upstream or downstream of the RSB1 open reading frame along with 20 nucleotides that bind upstream or downstream of the natMX4 cassette. This cassette was transformed into S. cerevisiae. Transformants were selected on plates containing nourseothricin (200 g/ml) and screened by PCR analysis.
Plasmids-The 0.915-kbp RSB1 promoter fragment was amplified using primers 5Ј-ctagaattcttagagcgcgtgttgaaatatagtcac-3Ј (Ϫ1000 RSB1for) and 5Ј-cgcgggatccattttgaatttctcaacgt-3Ј (rsb1-prom-del-1) having an EcoRI and a BamHI site, respectively. The EcoRI/BamHIdigested amplicon was then cloned into vector pSEYC102, which contains the lacZ gene, digested with the same enzymes. This clone was called pTA2 and was used for the lacZ assays. The RSB1 promoter was obtained from pTA2 by EcoRI/BamHI cleavage and the released promoter fragment was cloned in pRS316, yielding pKGTA1. This clone was used as a template to generate the different versions of the RSB1 open reading frame (ORF) clones. The sequence originally present in the Saccharomyces Genome Data base utilizes the second ATG present in the RSB1 ORF at residue 29 as the initiator codon. The form of Rsb1p expressed from this ATG is referred to as the ⌬28 form here. The RSB1 ORF beginning from this downstream ATG (⌬28 form) was amplified using primers 5Ј-atggatccgtccaacgcaacaaat-3Ј and 5Ј-ccgtagcggccgctgatataggcttcaatc-3Ј with BamHI and NotI sites and cloned into pCR2.1 TOPO (Invitrogen), yielding pKG ⌬28. The BamHI/NotI 1.6-kbp fragment from pKG⌬28 was cloned in pKGTA1 digested with the same enzymes, yielding pSLP5. This construct (pSLP5) was used as a base template for the construction of the RSB1 and the upstream ATG RSB1 (uATG) clones. The uATG clone was constructed by placing a BamHI site at the ATG for the RSB1 ORF using the primer 5Ј-atggatccggtaccgaaccttcgt-3Ј and introducing a NotI site 3Ј to the ORF as described above for the ⌬28 mutant with the same downstream primer. The resulting PCR fragment was cloned into pCR2.1 TOPO to form pSLP1. A 0.625-kbp fragment released on digestion with BamHI/XbaI was cloned into pSLP5 digested with the same enzymes to construct the full-length ORF and produce pSLP3. This strategy results in both the uATG and ⌬28 forms of Rsb1p being expressed from the natural ATG of RSB1 but the ⌬28 mutant lacks the wild-type Rsb1p residues corresponding to positions 2-28 in the primary translation product. To construct the wild-type RSB1 clone in pRS316 without the BamHI site, primers 5Ј-ctagaattcttagagcgcgtgttgaaatatagtcac-3Ј and 5Ј-ccgtagcg gccgctgatataggcttcaatc-3Ј with EcoRI and NotI sites were used to amplify the entire RSB1 ORF and this PCR fragment was cloned into pCR2.1 TOPO to generate pSLP2. The 1.9-kbp fragment released on digestion of pSLP2 with EcoRI was cloned in EcoRI-digested pSLP5, yielding pSLP4. All clones generated were sequenced to confirm their proper construction.
The RSB1 clones described above were all tagged at their COOH termini with a 3ϫ hemagglutinin (HA) cassette using the method described by Longtine et al. (27). The forward primer contained 50 nucleotides of the COOH-terminal RSB1 coding sequence followed by the F2 sequence (27), the reverse primer contained 50 nucleotides of pRS316 sequence followed by the R1 primer sequence. With these primers the 3ϫ HA-kanMX6 sequence was amplified from the corresponding pFA6a plasmid and co-transformed into wild-type cells along with RSB1 clones pSLP5 (⌬28 RSB1), pSLP3 (uATG RSB1), and pSLP4 (wt RSB1) linearized with NotI at the 3Ј-end of the RSB1 ORF. Transformants were selected on medium lacking uracil and then tested for geneticin (200 g/ml) resistance. The 3ϫ HA-tagged versions of the above clones were rescued from the above strain backgrounds and sequenced to confirm the fusion junction. The RSB1 clones (pSLP3, pSLP4, and pSLP5) were also tagged with enhanced GFP (eGFP) using the method described previously by Gerami-Nejad et al. (28).
Construction of the RSB1 N-Glycosylation Mutants-The aspargine residues at Asn-3 and Asn-6 were mutated to glutamine residues by site-directed mutagenesis using various primers. Primers incorporating the desired point mutations were synthesized and used for PCR amplification using Hi Fidelity Taq polymerase (Invitrogen). All the primers used had a BglII site placed immediately downstream of the ATG of the RSB1 ORF. The absence of the BglII site was used to confirm the generation of the final mutant clones. The reverse primer 5Ј-ATAAGAATgcggccgccCAATAAAGAGAACCTGTGGC-3Ј, with a NotI site used for amplification, was common for all the mutants. The following forward primers were used to generate the specific mutants: N3Q, 5Ј-ttcaaaatgagatctgTCCCAAGCAACAAATAATAC-3; N6Q, 5Ј-ttcaaaatgagatctgTCCAACGCAACACAAAATACGTTAGGCAG-T-3Ј; and N3Q,N6Q, 5Ј-ttcaaaatgagatctgTCCCAAGCAACACAAAA-TACGTTAGGCAGT-3Ј. The wild-type RSB1 control for all the mutants also had a BglII site at the true ATG for RSB1 ORF and was amplified using the forward primer 5Ј-ttcaaaATGagatctgTCCAACG-CAACAAATAATAC and the common reverse primer to generate the clone pSLP29. PCR amplicons (384 bp) generated using the different forward primers in combination with the common reverse primer using pSLP35 as the template DNA were cloned into TOPO 2.1 vector. These clones were then digested with BglII and NotI and the 384-bp fragment was cloned into the BamHI/NotI-digested uATG RSB1-3ϫ HA clone (pSLP35). They were all transferred to pSLP35 finally as a 1.0-kb KpnI fragment to generate the final clones, namely, pSLP28-N3,6Q-3ϫ HA pSLP32-N3Q-3ϫ HA, and pSLP33-N6Q-3ϫ HA. All the clones were sequenced to confirm the incorporation of the mutations. To construct eGFP fusion proteins, pSLP28 and pSL29 were digested with KpnI and cloned into KpnI-digested uATG RSB1-eGFP (pSLP10) to yield pSLP31 and pSLP30, respectively. Construction of the Reporter Plasmids-The wild-type RSB1-lacZ fusion plasmid (pTA2) was generated as described above. Overlap extension PCR mutagenesis was used to alter the PDREs in RSB1 (29). Briefly, PCR were performed with a wild-type RSB1 template and two different mutagenic primers (mutant bases are in uppercase) for either PDRE 1 (gcacaatcttttacaTCCtCGagagcattcttgctccg) or PDRE 2 (gctcg-gactctttTCtagaGAaagatatggtctcc) and the primer rsb1-prom-del-1 (see above). These products were purified and mixed with a product generated using the primers Ϫ1000 RSB1for (see above) and PDREcom (CAACTATCGGAGCTGCCACA). Each pair of PCR products was then amplified using the two outside primers (Ϫ1000 RSB1for and rsb1prom-del-1) with the resulting single mutant RSB1 promoter cloned into pSEYC102 as an EcoRI/BamHI fragment. A point mutation introduced in PDRE 2 generating a XbaI site yielded pTA5. Replacement of PDRE 1 with a XhoI restriction site yielded pGK3. A mutant plasmid lacking both PDREs (pGK2) was produced using pTA5 as a template and the PDRE 1 mutant primer (see above) along with the primer PDRE1XhoIanti (cggagcaagaatgctCTcgAggatgtaaaagattgtgc) in a similar overlap extension strategy.
Immunoblotting-Cells were grown to an A 600 of 1 and whole cell extracts were prepared by the TWIRL buffer (8 M urea, 5% SDS, 10% glycerol, and 50 mM Tris, pH 6.8) extraction method. 5 OD units of each sample were electrophoresed on SDS-PAGE, transferred to nitrocellulose, and Western blotted using monoclonal anti-HA (1:1000) (Covance) and the vacuolar membrane protein Vph1p (1:1000) antibodies (Molecular Probes).
Fluorescence Microscopy-Transformants in SEY6210 ϩ and SEY6210 0 containing the wild-type N3Q,N6Q and ⌬28 RSB1-eGFP fusion plasmids were grown to an A 600 of 0.5-0.8. 1 ml of cells was centrifuged and the pellet was resuspended in 500 l of kill buffer (100 mM Tris, pH 8.0, 1% NaN 3 ). Cells were then visualized for eGFP fluorescence and Nomarski optics using an Olympus (Tokyo, Japan) BX-60 microscope with a ϫ100 oil objective. Images were captured using a Hamamatsu (Shizuoka, Japan) ORCA charge-coupled device camera.
PNGase F Treatment-Crude cell extracts were prepared as described for sucrose density gradient fractionation from cells expressing various forms of HA-tagged Rsb1p in 0 SEY6210. After centrifuging the cell extract at 2,000 ϫ g for 3 min to remove unbroken cells, the supernatant was further centrifuged at 12,000 ϫ g for 45 min. The membrane-enriched pellet was resuspended in 30 l of Tris-HCl buffer, pH 7.5, ϩ 5 l of denaturation buffer ϩ 5 l of 10ϫ protease inhibitor mixture (Roche) and incubated at 37°C for 10 min. After denaturation, 5 l of G7 buffer ϩ 5 l of Nonidet P-40 and 250 units of PNGase F were added. All the buffers including Nonidet P-40 were supplied by the manufacturer (New England Biolabs). Control samples were treated identically except that no PNGase F was added. The reaction mixtures were incubated at 37°C for 16 -24 h. After SDS-PAGE and transfer to a nitrocellulose membrane, the blot was probed with anti-HA antibody (1:1000). Primary antibodies were detected with horseradish peroxidase-conjugated anti-mouse secondary antibody (1:2000) followed by measurement of chemiluminescence (Amersham Biosciences).

PHS Resistance Depends on Mitochondrial Status-Previous
DNA microarray experiments have suggested that RSB1 gene expression is strongly induced in S. cerevisiae cells lacking their mitochondrial genome ( 0 status) (19,21). To examine if 0 activation of RSB1 transcription can influence PHS tolerance, isogenic ϩ and 0 strains were generated that varied according to the presence of the RSB1 gene. These strains were then compared for their relative PHS resistance by spot test assay (Fig. 1).
Wild-type, ϩ cells were able to tolerate up to 10 M PHS, whereas rsb1⌬ strains failed to grow above 5 M. Cells lacking the mitochondrial genome exhibited robust growth at 20 M but this phenotype was dependent on the presence of the RSB1 gene. These data are consistent with the idea that transcriptional induction of RSB1 in 0 cells leads to PHS resistance. Other genes are likely to participate in this 0 elevation of PHS tolerance as the 0 rsb1⌬ cells continued to grow at 10 M, a concentration ϩ rsb1⌬ were unable to tolerate.
RSB1 Expression Is Induced in 0 Cells in a Pdr3p-dependent Manner-The data described above are consistent with a model in which Pdr3p induces RSB1 transcription in 0 cells leading to elevated PHS resistance. To test this model, we constructed a lacZ gene fusion to the putative ATG for the RSB1 open reading frame based on sequence information from SGD. This clone was introduced into isogenic ϩ and 0 S. cerevisiae cells but we were unable to detect any ␤-galactosidase activity (data not shown). Comparison of the DNA sequences encoding the amino-terminal protein sequences from other Saccharomyces species suggested that the reported sequence for the S. cerevisiae RSB1 gene might contain an extra adenine residue 16 bp upstream from the predicted ATG. Removal of this adenine residue would extend the aminoterminal domain of Rsb1p by 28 residues that show strong sequence conservation with other fungal Rsb1p homologues. We directly sequenced this region of S. cerevisiae RSB1 and confirmed that the correct sequence lacked this adenine residue (data not shown). Additionally we constructed a lacZ gene fusion to the new ATG present at the start of the RSB1 ORF and found that this fusion expressed ␤-galactosidase activity. We have used this gene fusion as an indicator plasmid to evaluate RSB1 expression in the studies reported below.
This RSB1-lacZ reporter gene was introduced into a series of isogenic strains to evaluate RSB1 expression in response to loss of the mitochondrial genome and different alleles of PDR1 and PDR3. We also tested the PHS tolerance of these same strains (Fig. 2).
The RSB1-lacZ plasmid was induced in 0 cells to 26 units/OD compared with 7 units/OD in ϩ cells. Removal of the PDR3 gene eliminated this 0 -mediated induction. Similar to the behavior of other 0 -induced genes regulated by Pdr3p (22), 0 pdr1⌬ cells supported an even higher level of RSB1 expression than 0 cells with expression rising to 35 units/ OD. Loss of both PDR1 and PDR3 from 0 cells caused no further decrease than that seen in 0 pdr3⌬ strains alone.
These levels of RSB1 expression were generally consistent with the PHS tolerance of each strain. The 0 pdr1⌬ cells were the most resistant to this LCB followed by 0 cells. In ϩ cells, the pdr1⌬ pdr3⌬ mutant was significantly more PHS sensitive than either single mutant strain. The pdr1⌬ pdr3⌬ mutant was also more sensitive to PHS (growth strongly inhibited at 5 M) than the rsb1⌬ strain. This suggests the existence of another gene regulated by the Pdr pathway that is involved in PHS tolerance. In support of this idea, the 0 pdr1⌬ pdr3⌬ strain was more sensitive to PHS than the 0 pdr3⌬ strain even though these two mutants drove essentially identical levels of RSB1-lacZ expression. This discrepancy is consistent with the possibility that a Pdr-regulated gene might be contributing to PHS tolerance along with RSB1.
We also examined the contribution of each of the two Pdr1p/Pdr3p response elements (PDREs) present in the RSB1 promoter to gene expression. Site-directed mutations were prepared in each site individually as well as a double mutant lacking both sites. The most upstream PDRE was designated PDRE 1, whereas the promoter proximal element was designated PDRE 2. These plasmids were introduced into ϩ and 0 cells and analyzed for expression of RSB1-dependent ␤-galactosidase activity (Fig. 3).
The wild-type RSB1-lacZ fusion gene was induced by ϳ4-fold in 0 cells compared with a ϩ background. Loss of either PDRE strongly reduced 0 inducibility to less than 2-fold and the double mutant eliminated this weak residual 0 response. Both PDREs are required for the normal induction of RSB1 seen in 0 cells. Phytosphingosine tolerance responds to PDR gene dosage. A, an isogenic series of strains lacking the indicated PDR genes was grown to mid-log phase and then tested by placing on media containing various concentrations of PHS. B, these same strains were transformed with the RSB1-lacZ reporter plasmid and assayed for ␤-galactosidase activity as described (25). wt, wild-type. Site-directed mutations were constructed in the most upstream of the Pdr1p/Pdr3p response element (PDRE 1) or the more promoter-proximal binding site (PDRE 2) or both (PDRE 1 PDRE 2) in the context of the RSB1-lacZ fusion plasmid. These three mutant PDRE-containing plasmids were introduced into ϩ , 0 , and pdr5⌬ yor1⌬ strains. Transformants were grown to mid-log phase and RSB1-dependent ␤-galactosidase activity was measured as described above. wt, wild-type.
We also transformed this series of RSB1-lacZ reporter genes into cells lacking the ABC transporters Pdr5p and Yor1p. Previous work of others (9) has suggested that loss of these transporters leads to the strictly Pdr1p-dependent activation of RSB1 expression. We did not observe any significant induction of RSB1-lacZ expression in this pdr5⌬ yor1⌬ strain (Fig. 3). Western blot analysis also failed to detect any increase in expression of a Rsb1p-3ϫ HA construct (data not shown). Whereas this pdr5⌬ yor1⌬ strain did exhibit the elevated PHS resistance reported (9), there was no correlation with RSB1 expression in our experiments. The reason for this difference is unknown but may be a result of different strains and/or experimental protocols used for these experiments.
Characterization of Full-length and Mutant Rsb1p Forms-Because the data reported above indicate that the previously predicted form of Rsb1p is actually an amino-terminal truncation mutant, it was important to compare the properties of these two different Rsb1p derivatives. To enable detection of Rsb1p, a 3ϫ HA epitope tag was added to its extreme carboxyl terminus. The wild-type RSB1 gene was prepared by PCR amplification using primers that bound ϳ1000 bp upstream of the ATG and 500 bp downstream of the stop codon. This fragment was cloned into a low-copy number plasmid and the 3ϫ HA tag was inserted in place of the natural termination codon.
Along with this wild-type form of RSB1, we also produced several mutant versions of this gene. A truncation mutant lacking the aminoterminal 28 residues, corresponding to the previously characterized form of Rsb1p (8), was prepared and designated ⌬28 RSB1. Inspection of the sequence of these 28 amino acids also indicated the presence of two potential N-linked glycosylation sites located at residues 3 and 6 in fulllength Rsb1p. Single substitution mutations were generated that removed each site individually (N3Q and N6Q Rsb1p) as well as a double mutant that lacked both sites (N3Q,N6Q Rsb1p). As for the wildtype gene described above, all these mutants were tagged with 3ϫ HA and carried on a low-copy number plasmid under control of the wildtype RSB1 promoter. These Rsb1p derivatives were introduced into yeast cells and analyzed for their expression and function. We also constructed carboxyl-terminal eGFP fusions to the wild-type, N3Q,N6Q, and ⌬28 forms of Rsb1p to compare the localization of these proteins.
Expression of Rsb1p Forms-Western blotting experiments were carried out to determine the expression profile of each Rsb1p form in ϩ and 0 cells. An empty vector plasmid was introduced as a specificity control for anti-HA blotting. Whole cell protein extracts were prepared and equal aliquots were electrophoresed through SDS-PAGE, followed by Western blot analysis (Fig. 4).
Two different forms of wild-type Rsb1p were detected by anti-HA immunoblotting: a single species of ϳ43 kDa and a larger, less discrete form of apparent molecular mass of ranging from 64 to 80 kDa. Interestingly, the ⌬28 Rsb1p form exhibited a single HA immunoreactive species of 40 kDa. The double substitution mutant form of Rsb1p lacking both putative N-glycosylation sites (N3Q,N6Q RSB1) produced a doublet around 43 kDa but no proteins above this molecular mass. The mutant lacking the most amino-terminal of the putative glycosylation sites (N3Q RSB1) generated a disperse group of immunoreactive proteins around 60 kDa and a single polypeptide of 43 kDa. Interestingly, the other single glycosylation site mutant (N6Q RSB1) produced lower levels of the 60-kDa group of proteins but an increased amount of the doublet at 43 kDa. All these proteins were more abundant in samples prepared from 0 cells indicating that retrograde regulation was still operational on all the mutant genes. Vph1p immunoblotting indicated that comparable levels of protein were present in each sample.
This Western blot analysis provided several important new pieces of information relevant to the biogenesis of Rsb1p. First, full-length Rsb1p produces two distinctly different immunoreactive species: a heterodisperse form ranging from 64 to 80 kDa and a more discrete form at 43 kDa. Second, expression of a mutant form of Rsb1p lacking the aminoterminal 28 residues produces a single species of 40 kDa. Third, inactivation of two putative N-linked glycosylation sites present in this amino-terminal 28-residue region prevents the formation of the heterodisperse forms of Rsb1p. Finally, loss of either potential N-linked glycosylation site individually led to production of a new high molecular mass group of immunoreactive species grouped around 60 kDa. Given that these mutant forms of Rsb1p produced a range of polypeptides when present in cells, we assessed the function of each by testing for their ability to complement the PHS sensitivity of a rsb1⌬ dpl1⌬ strain.
Reduced Function of Mutant Forms of Rsb1p-Low-copy number, URA3-containing plasmids expressing wild-type and mutant forms of Rsb1p derivatives were transformed into a rsb1⌬ dpl1⌬ strain. We used this strain as a sensitized genetic background because PHS toxicity depends on interference with nutrient transporter function (31,32). Work from other labs has shown that complementation of auxotrophic markers in S. cerevisiae cells reduces the sensitivity of these strains to PHS (31) and inclusion of the dpl1⌬ lesion facilitates detection of the PHS-sensitive phenotype (8). Transformants were placed on minimal medium containing different concentrations of PHS and their ability to tolerate this long chain base was evaluated (Fig. 5).
Introduction of a plasmid expressing the wild-type form of Rsb1p restored the ability of transformants to grow in the presence of 5 and 7.5 M PHS. The ⌬28 Rsb1p derivative exhibited the greatest defect in PHS resistance although it still maintained some ability to function. Loss of either N-glycosylation site individually reduced the ability of the resulting protein to function compared with wild-type. The double replacement protein (N3Q,N6Q Rsb1p) was slightly more defective than either single mutant but still superior to the ⌬28 form of Rsb1p. Together with the Western blot analysis, these data support the view that Rsb1p must be glycosylated to exhibit full function.
Localization of Wild-type and Mutant Rsb1p-Previous studies using the ⌬28 form of Rsb1p have suggested that this protein is found in the endoplasmic reticulum (8,9). eGFP fusions to wild-type, ⌬28, and N3Q,N6Q Rsb1p were prepared to evaluate the subcellular distribution of these proteins. The addition of eGFP to the COOH termini of these Rsb1p derivatives did not alter their relative ability to complement the PHS hypersensitivity of a rsb1⌬ dpl1⌬ strain (data not shown). Lowcopy number plasmids expressing the Rsb1p-eGFP fusions were introduced into ϩ and 0 cells. Transformants were grown to mid-log phase and visualized by fluorescence microscopy (Fig. 6). Cells were also stained with FM4-64 to ascertain the location of the vacuole as described (33).
All Rsb1p forms were found at the periphery of the cell, consistent with a plasma membrane localization in both ϩ and 0 cells. The fluorescence intensity was higher in the 0 background but no significant difference in localization was seen. Importantly, all three forms of Rsb1p were also found to accumulate intracellularly. This can be most clearly seen in 0 cells where the intracellular eGFP signal overlaps with the FM4-64 fluorescent signature of the vacuolar membrane. Many S. cerevisiae proteins that are localized on the plasma membrane are often routed to the vacuole for their degradation (recently reviewed in Ref. 34). These data suggest this is likely to be true for Rsb1p but further experiments are required to confirm this idea. The central conclusion is that the mutant forms of Rsb1p analyzed here exhibit plasma membrane location that is indistinguishable from the wild-type protein.
To confirm the results using the Rsb1p-eGFP fusions, sucrose gradient fractionation was employed to examine the distribution of epitopetagged Rsb1p proteins between different membrane compartments. Isogenic ϩ and 0 strains expressing either wild-type or ⌬28 Rsb1p-3ϫ HA proteins were grown to mid-log phase and whole cell lysates were prepared. Lysates were separated by sucrose density centrifugation and fractions were collected from these gradients. Aliquots of each fraction were electrophoresed on SDS-PAGE and analyzed by Western blotting with anti-HA and marker antibodies (Fig. 7).
In ϩ cells, wild-type Rsb1p was found to enrich in the densest (bottom) of the gradient in a manner overlapping with Pma1p, the plasma membrane ATPase (35). Both the large and smaller forms of wild-type Rsb1p appeared to fractionate in a similar fashion. Importantly, both of these forms of Rsb1p fractionated with behavior distinct to that of the ER chaperone, Kar2p, which was enriched in the center of these gradients. Strikingly, the ⌬28 form of Rsb1p was also found to enrich in the bottom of the gradient along with Pma1p. This finding suggests that the reduced function seen for this mutant form is unlikely to be explained by a defect in localization. Both wild-type and ⌬28 forms of Rsb1p were expressed at higher levels in the fractions from 0 cells but no changes in their subcellular distributions were detected.
These localization data provide independent support for the assignment of Rsb1p as a component of the S. cerevisiae plasma membrane.
Additionally, the functional defect of the ⌬28 form of Rsb1p is unlikely to be a consequence of mislocalization.
Wild-type but Not ⌬28 Rsb1p Is Glycosylated-The appearance of the higher molecular mass form of Rsb1p in cells expressing the wild-type but not ⌬28 or N3Q,N6Q forms of this protein suggested the possibility that the longer Rsb1p derivative was glycosylated. To test this idea, membranes were prepared from cells expressing the wild-type, N3Q,N6Q, or ⌬28 Rsb1p derivatives and subjected to PNGase F digestion. PNGase F will cleave N-linked sugars from their protein backbones (36). After PNGase F digestion, aliquots were resolved on SDS-PAGE and analyzed by Western blotting with anti-HA antibodies (Fig. 8).
As seen above, wild-type Rsb1p produced two different forms of immunoreactive protein: a heterogenous, higher molecular mass species and a more discrete smaller form. PNGase F digestion resulted in an elimination of the heterogenous species with an increase in the intensity of a band with slightly high molecular weight as compared with the discrete, lower molecular weight form that is thought to represent deglycosylated Rsb1p. The difference in molecular mass between these two smaller Rsb1p digestion products may represent partial deglycosylation but is not currently understood. Conversely, no such change in electrophoretic behavior was seen for either the N3Q,N6Q or the ⌬28 Rsb1p derivatives. These data support the conclusion that most, but not all, wild-type Rsb1p is a glycoprotein, whereas no glycosylation could be found in the case of the ⌬28 or the N3Q,N6Q mutants. Two consensus sites (Asn-X-Thr/Ser) for N-linked glycosylation (37) can be found that begin at residues 3 (NAT) and 6 (NNT). As described above, expression FIGURE 5. Complementation via wild-type and ⌬28 forms of Rsb1p. Low-copy number plasmids expressing the indicated forms of Rsb1p were introduced into rsb1⌬ dpl1⌬ cells along with the empty vector plasmid as a control (pRS316). Transformants were grown to mid-log phase and then 2-fold serial dilutions were plated on selective media containing the indicated concentrations of PHS. Cells were incubated at 30°C and photographed. FIGURE 6. Localization of Rsb1p-eGFP fusions. Low-copy number plasmids expressing wild-type or the ⌬28 forms of Rsb1p containing a eGFP fusion to the COOH terminus of each protein were produced. These plasmids were transformed into the indicated genetic backgrounds and evaluated by fluorescence microscopy. wt, wild-type; DIC, differential interference contrast. of either a N3Q or N6Q mutant form of Rsb1p produced two immunoreactive forms of Rsb1p: a series of proteins centered around 60 kDa and a polypeptide of 43 kDa. We subjected samples from cells expressing these single mutant derivatives to PNGase F digestion and found that the 60-kDa series of proteins were sensitive to digestions by this glycosidase. 3 The simplest interpretation of these data is that Rsb1p is glycosylated at both the Asn-3 and Asn-6 positions with this dually glycosylated form appearing around 80 kDa.

DISCUSSION
Loss of the mitochondrial genome triggers global reprogramming of gene expression called the retrograde response (13). Several laboratories have provided evidence that 0 cells dramatically induce expression of ABC transporter-encoding genes like PDR5 along with elevation of transcription of loci encoding enzymes involved in sphingolipid biosynthesis (19 -21, 38). This work extends the link between retrograde regulation and sphingolipid biosynthesis with the identification of RSB1 as a target of Pdr3p regulation in 0 cells. Establishing that 0 cells induce RSB1 via a Pdr3p-dependent mechanism provides further support to the idea that the physiological role for the Pdr regulatory pathway in S. cerevisiae is to modulate the lipid composition of membranes.
Along with this new insight into the transcriptional control of RSB1, several features of the localization of Rsb1p have been clarified. First, we have found and corrected a sequencing error that led to the prediction of an amino-terminal truncated Rsb1p (8). Correction of this error extends the region of sequence similarity with other Saccharomyces species and more importantly is required for the normal glycosylation of Rsb1p to occur. Second, complementation assays indicate that these additional 28 residues are required for normal function of Rsb1p. Third, two distinct N-linked glycosylation sites are found in this amino-terminal region of Rsb1p. Finally, analyses of subcellular localization indicate that full-length, N3Q,N6Q, or the ⌬28 forms of Rsb1p accumulate in the plasma membrane, a location that simplifies the view of Rsb1p action given that this protein has previously been demonstrated to lead to efflux of LCBs out of the cell (8).
Several of the conclusions reported here differ from the previous work of Kihara and Igarashi (8,9). These authors assigned Rsb1p an endoplasmic reticulum localization and did not report any glycosylation of this protein. There are two likely reasons to explain the different conclusions reached by these authors. First, their Rsb1p expression constructs were all based on the incorrect DNA sequence and led to the production of the truncated form of Rsb1p lacking the N-linked glycosylation sites. Second, amino-terminal epitope-tagged forms of Rsb1p were employed that could interfere with the function of Rsb1p. Our expression constructs employed full-length Rsb1p with carboxyl-terminal tags that were demonstrated to normally complement rsb1⌬ strains  . Sucrose gradient analysis of membrane distribution of Rsb1p. Isogenic ϩ and 0 cells were transformed with low-copy number plasmids carrying RSB1-3ϫ HA fusion genes corresponding to the wild-type and ⌬28 forms of this protein. Transformants were grown to mid-log phase and whole cell lysates were prepared. These lysates were centrifuged through a 10 -60% sucrose cushion at 300,000 ϫ g for 12 h. Equal aliquots of each gradient fraction were then collected and concentrated with the addition of trichloroacetic acid. The resulting precipitates were resuspended in loading buffer and separated by SDS-PAGE. After transfer to nitrocellulose membranes, the distribution of proteins was analyzed by Western blotting with the indicated antibodies. Top refers to the least dense fraction of the gradient, whereas bottom indicates the fraction from the densest part of the gradient. The position of glycosylated Rsb1p is indicated by the arrow. Anti-Kar2p (endoplasmic reticulum) and anti-Pma1p (plasma membrane) antibodies were used as controls for the distribution of proteins of known subcellular location. FIGURE 8. The full-length but not ⌬28 form of Rsb1p is glycosylated. Microsomal membranes were prepared from 0 cells expressing wild-type, N3Q,N6Q, or ⌬28 Rsb1p under control of the RSB1 promoter. All Rsb1p forms were expressed from a low-copy number plasmid and contained a 3ϫ HA epitope tag at their COOH terminus. These membranes were then digested at 37°C in the presence (ϩ) or absence (Ϫ) of PNGase F for 16 -24 h. Aliquots were electrophoresed through SDS-PAGE and analyzed by Western blotting with a monoclonal antibody directed against the HA epitope. Molecular mass standards are indicated on the left side in kDa. (Fig. 5). For these reasons, we believe the data reported here more accurately reflect the localization and modification status of Rsb1p in vivo.
Inclusion of the additional 28 residues at the Rsb1p amino terminus is critical for the function of this protein. Removal of this region of Rsb1p prevents the normal glycosylation of the protein without obvious effects on steady-state localization as the ⌬28 Rsb1p still accumulated in the plasma membrane. Expression of the ⌬28 form of Rsb1p is reduced compared with wild-type (because of the absence of the glycosylated species) but still readily detectable. Analysis of the site-directed replacements of the two N-linked glycosylation sites in the 28-amino acid region also argues that loss of glycosylation alone is not likely to explain the functional defect seen in ⌬28 as the N3Q,N6Q Rsb1p mutant exhibited an increased ability to confer PHS tolerance than the ⌬28 Rsb1p (Fig. 5). There also appears to be a hierarchy of importance in the two glycosylation sites in the Rsb1p amino terminus. Loss of Asn-6 causes a more complete block to subsequent glycosylation than does loss of Asn-3 (Fig. 4), although we were not able to identify a further loss of function in the N6Q mutant compared with the N3Q Rsb1p. Further work is required to determine the role of glycosylation in Rsb1p function but these experiments strongly suggest that the normal qualitative pattern of subcellular localization remains even in the absence of this post-translational modification.
The sucrose density gradient analysis argues that both the glycosylated and deglycosylated forms of wild-type Rsb1p appear in the plasma membrane. Keeping in view that N-glycosylation is a common property of plasma membrane transporters from mammalian cells (see Refs. 39 -41 for examples), the finding that the putative LCB transporter protein would be glycosylated is unusual in S. cerevisiae as previous analyses of other plasma membrane transporters such as Pdr5p (42), Yor1p (30), Ste6p (43,44), and Gap1p (45) indicated that none of these proteins were detectably glycosylated. Studies on the Mep2p ammonium permease (46) and Zrt1p zinc transporter (47) have indicated that these plasma membrane transporters are N-glycosylated but in both of these latter cases, the only form seen in the plasma membrane is the glycosylated species. Moreover, loss of glycosylation does not affect normal functioning of the Mep2p permease (46), unlike the effect reported for Rsb1p in this study. In view of the role of Rsb1p in LCB homeostasis in S. cerevisiae cells, the relatively unique requirement of Rsb1p for N-glycosylation suggests that this post-translational modification may be directly involved in the regulation of this membrane protein.
The finding that changes in the level of activity of the Pdr pathway leads to changes in PHS resistance implicates Pdr1p and Pdr3p in the detoxification of this natural sphingolipid intermediate. Whereas many of the substrates that are handled by the Pdr pathway represent compounds that S. cerevisiae are unlikely to naturally encounter (cycloheximide, 4-nitroquinoline-N-oxide, azole drugs, etc.), PHS is produced as a result of normal metabolism in this organism. Strikingly, a number of different sphingolipid pathway inhibitors are also substrates for the Pdr pathway. These antifungal agents include rustmicin (48), aureobasidin A (49), and fumonisin B1 (5) and are believed to act as inhibitory analogues of naturally produced pathway intermediates. Perhaps the Pdrregulated transporter proteins normally act on the sphingolipid precursors that these analogues mimic.
Evidence is steadily accumulating in support of the idea that the Pdr pathway has a major role in control of lipid levels in the plasma membrane. Hyperactive alleles of PDR1 and PDR3 were found to trigger the removal of fluorescent phosphatidylethanolamine analogues in a Pdr5p-and Yor1p-dependent manner (50,51). We have previously documented Pdr-dependent regulation of genes involved in sphingolipid biosynthesis (17,18). This study and the work of Kihara and Igarashi (9) place the RSB1 gene clearly in the Pdr regulon. In mammalian cells, ABC transporters have been linked to transbilayer distribution of phospholipids, cholesterol, and sphingolipids (52). Because signals that activate the Pdr pathway (such as in 0 cells) simultaneously induce ABC transporters like Pdr5p and a variety of sphingolipid homeostatic proteins, it is tempting to speculate that Pdr1p-and Pdr3p-mediated transcriptional regulation acts to coordinate biosynthesis and distribution of a variety of membrane components. Future work will be directed toward testing this hypothesis.