Overlapping Functions of the Yeast Oxysterol-Binding Protein Homologues

Corresponding author: Jasper Rine, Department of Molecular and Cell Biology, Rm. 401 Barker Hall, University of California, Berkeley, CA 94720-3202., jrine{at}uclink4.berkeley.edu (E-mail)

Communicating editor: M. JOHNSTON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The Saccharomyces cerevisiae genome encodes seven homologues of the mammalian oxysterol-binding protein (OSBP), a protein implicated in lipid trafficking and sterol homeostasis. To determine the functions of the yeast OSBP gene family (OSH1–OSH7), we used a combination of genetics, genomics, and sterol lipid analysis to characterize OSH deletion mutants. All 127 combinations and permutations of OSH deletion alleles were constructed. Individual OSH genes were not essential for yeast viability, but the elimination of the entire gene family was lethal. Thus, the family members shared an essential function. In addition, the in vivo depletion of all Osh proteins disrupted sterol homeostasis. Like mutants that affect ergosterol production, the viable combinations of OSH deletion alleles exhibited specific sterol-related defects. Although none of the single OSH deletion mutants was defective for growth, gene expression profiles revealed that each mutant had a characteristic molecular phenotype. Therefore, each gene performed distinct nonessential functions and contributed to a common essential function. Our findings indicated that OSH genes performed a multitude of nonessential roles defined by specific subsets of the genes and that most shared at least one essential role potentially linked to changes in sterol lipid levels.

ERGOSTEROL, a cholesterol-like lipid, is a major constituent of the yeast cell membrane, where it is present in 3.3-fold molar excess over all phospholipids (ZINSER et al. 1991 ). In eukaryotes, sterols like cholesterol and ergosterol are the bulk isoprenoid products of the mevalonate biosynthetic pathway. The products of the mevalonate pathway exert feedback regulation on their own synthesis at both transcriptional and post-transcriptional levels (GOLDSTEIN and BROWN 1990 ; BROWN and GOLDSTEIN 1997 , BROWN and GOLDSTEIN 1999 ). Oxygenated derivatives of cholesterol, referred to as oxysterols, are particularly potent feedback regulators (KANDUTSCH et al. 1978 ). Oxysterols are biosynthetic metabolites of sterols, steroids, and bile acids and are also produced when sterols are exposed to oxidants. In addition to cholesterol feedback regulation, oxysterols play roles in apoptosis, cellular aging, platelet aggregation, and sphingolipid metabolism (reviewed in SCHROEPFER 2000 ). Two protein families appear to mediate many of the activities ascribed to oxysterols. These proteins include some of the steroid hormone nuclear receptors (reviewed by RUSSELL 1999 ) and another family known as the oxysterol-binding proteins (OSBPs).

The canonical OSBP was purified to homogeneity on the basis of its high affinity for oxysterols (DAWSON et al. 1989A ) and the corresponding gene was subsequently cloned (DAWSON et al. 1989B ). Homologues are present in many eukaryotes, including humans (LEVANON et al. 1990 ), flies (ALPHEY et al. 1998 ), worms (C. ELEGANS SEQUENCING CONSORTIUM 1998), and fungi (JIANG et al. 1994 ; SCHMALIX and BANDLOW 1994 ; FANG et al. 1996 ; DAUM et al. 1999 ; HULL and JOHNSON 1999 ). Because of its binding activity and the potency of oxysterols as feedback regulators, OSBP was proposed to mediate feedback control of the mevalonate pathway (TAYLOR et al. 1984 ). However, the SRE binding protein (SREBP) and SREBP cleavage-activating protein (SCAP) are now known to mediate this role (reviewed by BROWN and GOLDSTEIN 1997 , BROWN and GOLDSTEIN 1999 ) and SREBP is unrelated in sequence to any OSBP. Nevertheless, overexpression of the mammalian OSBP gene causes pleiotropic effects on both cholesterol synthesis and expression of genes encoding some mevalonate pathway enzymes (LAGACE et al. 1997 ). However, the in vivo role of the OSBP family is still unclear.

The localization of OSBP within cells is governed by lipids. When OSBP binds oxysterols, a conformational change in the protein occurs, allowing OSBP to translocate from cytoplasmic vesicles to the Golgi apparatus (RIDGWAY et al. 1992 ). The amino-terminal region of OSBP contains a pleckstrin homology (PH) domain, which binds phosphatidylinositol lipids and thereby targets the protein to Golgi membranes where these lipids are enriched (LEVINE and MUNRO 1998 ). OSBP localization is also sensitive to concentrations of the lipid sphingomyelin (STOREY et al. 1998 ; RIDGWAY et al. 1998 ). Since OSBP localization is inherently linked to cellular lipid distribution, the function of OSBP likely involves some aspect of lipid maintenance in membranes.

A Saccharomyces cerevisiae OSBP homologue, KES1 (referred to here as OSH4/KES1), has also been implicated in the PI-dependent formation of Golgi-derived transport vesicles. Deletion of this homologue in yeast bypasses the requirement for SEC14, an essential gene encoding a phosphatidylinositol/phosphatidylcholine transfer protein (FANG et al. 1996 ). Since Sec14p is otherwise essential for vesicle biogenesis from the Golgi, this genetic suppression suggests that yeast OSBPs are also involved in secretion. Taken together the evidence suggests that this OSBP and perhaps its homologues affect membrane trafficking, possibly by influencing lipid distribution.

The potential roles of the OSBP family are not limited to membrane trafficking. Like oxysterols themselves, OSBPs have been implicated in a diverse variety of cellular processes. OSBP homologues may be involved in tumor metastasis (FOURNIER et al. 1999 ) and in cell-cycle progression (ALPHEY et al. 1998 ). How the OSBPs are involved in these processes is unclear.

None of the Saccharomyces OSBP homologues studied to date encode an essential gene (JIANG et al. 1994 ; SCHMALIX and BANDLOW 1994 ; FANG et al. 1996 ; DAUM et al. 1999 ). However, mutant strains in which some of these homologues were disrupted exhibit phenotypes similar to viable mutants defective in sterol biosynthesis (JIANG et al. 1994 ; DAUM et al. 1999 ). These OSBP deletion mutants are cold sensitive and are resistant to the fungal inhibitor nystatin, whose toxicity is manifested through its binding of the yeast-specific sterol, ergosterol, in the cell membrane. In some deletion strains, sterol lipid content has also been reported to be affected (JIANG et al. 1994 ; DAUM et al. 1999 ). As with its mammalian counterpart, the mechanism by which the yeast OSBP homologues affect sterols is unknown.

To understand OSBP function, we analyzed the entire family of seven OSBP homologues encoded by the S. cerevisiae genome. We used a combination of genetics, genomics, and lipid analysis to analyze the essential and nonessential roles of the yeast OSH (oxysterol-binding protein homologue) genes. With respect to cell growth, the disruption of any single OSH gene caused no overt phenotype. However, single deletion mutations had unique effects on gene expression profiles, indicating that the OSHs performed distinct nonessential roles. Since the elimination of all OSH genes resulted in cell lethality, together the yeast OSHs also performed at least one essential function in common. Moreover, sterol lipid analysis revealed that depletion of all Osh proteins drastically perturbed sterol levels. These results indicated that, collectively, the members of the OSH gene family were essential for the maintenance of sterol lipid composition and for cell viability.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Strains and microbial and genetic techniques:
Culture media and genetic manipulations were as described (ADAMS et al. 1997 ). Yeast rich medium (YPD) was supplemented with excess tryptophan (50 mg/liter) because of tryptophan import defects of some mutants. Cycloheximide-resistant strains (cyh2^R) were selected on solid YPD containing 10 mg/liter cycloheximide. Lovastatin (a gift of James Bergstrom, Merck) was prepared as previously described (DIMSTER-DENK et al. 1994 ). Nystatin (Sigma Chemicals, Inc., St. Louis) was added to cooled agar medium from an ethanol stock (10 mg/ml) and the plates were used within 24 hr. To select for the kanMX4 gene, yeast were grown on YPD containing 200 µg/ml geneticin sulfate (G418) (Gibco BRL Life Technologies, Inc., Rockville, MD; WACH et al. 1994 ). For lipid extractions, gas chromatography (GC) grade methanol and hexane were used (Fisher Scientific, Pittsburgh).

The genotypes of all yeast strains used in this article are shown in Table 1. All strains were congenic with SEY6210, unless otherwise noted. Strains bearing all permutations of gene disruptions were constructed through a systematic series of crosses. Each of the seven OSH deletions was marked by one of four prototrophic markers or by the kanMX4 gene (see below).

Sequence analysis:
OSBP homologues were identified using BLASTP homology searches (ALTSCHUL et al. 1990 ) of the complete yeast genome sequence (GOFFEAU et al. 1996 ). Sequence alignments were performed using the CLUSTALW alignment program. KYTE and DOOLITTLE 1982 plots were used to determine protein hydrophilicity. Secondary structure predictions were determined by the method of CHOU and FASMAN 1978 . Coiled-coil domain probability plots were generated as described by LUPAS 1996 , and transmembrane domains were analyzed using the TMpred program, both offered at www.ch.embnet.org/. Percentage similarity and identity between protein sequences was determined using the BESTFIT program of the GCG sequence analysis package.

The yeast genome encodes seven OSBP homologues (see RESULTS) including the gene referred to here as OSH1. Previously, OSH1 had been reported as two different and separate genes, SWH1 (SCHMALIX and BANDLOW 1994 ) and OSH1 (JIANG et al. 1994 ). OSH1 has also been given the open reading frame designation YAR042w. However, the protein sequences of SWH1 and OSH1 overlap. As determined by the size of epitope-tagged versions of Osh1p detected by immunoblot (our unpublished results), the protein size was consistent with the larger protein sequence reported in EMBL/GenBank Data Library accession no. X74552.

Cloning and plasmid constructions:
Restriction enzymes used for cloning were obtained from New England Biolabs (Beverly, MA), and cloning techniques were performed as described (SAMBROOK et al. 1989 ). To construct plasmids with which to disrupt OSH genes, the genes OSH2, OSH3, OSH6, and OSH7 were amplified by the polymerase chain reaction (PCR) using DNA from yeast strain SEY6210 as template. Primers were designed to generate fragments with restriction sites at the ends to facilitate cloning. All disruptions and chromosomal integrations were verified by DNA blot analysis. The osh1::URA3, osh4/kes1::HIS3, and osh5/hes1::LEU2 strains HAB835, HAB821, and HAB826, respectively (JIANG et al. 1994 ), were gifts of Howard Bussey (McGill University).

To construct the osh7::HIS3 disruption plasmid (pJR2281), the primer combinations used were: 5'-GCTAGGATCCAGTTCTCATAGCTCAATTAACG-3' and 5'-CAGTGGATCCTTCAGGGATGTGCTTG-3'. The 1.2-kb fragment was cloned into the BamHI site of pBluescript KS(+)(pBS-KS+; Stratagene, La Jolla, CA). From this plasmid, a 510-bp BamHI-EcoRI fragment and a 160-bp BamHI-XbaI fragment were subcloned into the EcoRI-XbaI sites of pRS403 (SIKORSKI and HIETER 1989 ) to generate pJR2281. To integrate and disrupt OSH7, pJR2281 was digested with BamHI and the entire plasmid was transformed into SEY6210. In the osh7 allele, 484 bp of the OSH7 coding region were removed and replaced with the HIS3 gene.

The oligonucleotide primers used to generate the osh6::LEU2 plasmid (pJR2282) were 5'-GCTAGGATCCTGCTGGGTTCTGCTTTTCGT-3' and 5'-CAGTGGATCCGCGTGTAGCGACATTTTAC-3'. The 1.6-kb amplified fragment was cloned into the BamHI site of pBS-KS+. From this plasmid, a 220-bp BamHI-XhoI fragment and a 785-bp BamHI-XbaI fragment were subcloned into the XhoI-XbaI sites of pRS405 (SIKORSKI and HIETER 1989 ), generating the plasmid pJR2282. To integrate the disruption construct, pJR2282 was digested with BamHI and the entire plasmid was transformed into SEY6210. The deletion of OSH6 removed 86 bp of upstream sequence and the first 547 bp of coding sequence.

The 2.8-kb OSH3 fragment was amplified and cloned into the SpeI sites of pBS-KS+. OSH3 was amplified using the following two primers: 5'-GCTAACTAGTCCAGTGTAGATGACCATGC-3' and 5'-CAGTACTAGTAACTCTTCGGTCCAGTTATG-3'. The osh3::LYS2 integration construct was produced by inserting the 360-bp SpeI-EcoRI and 240-bp HindIII-ClaI OSH3 fragments from the pBS-KS+ plasmid, together with the 4.8-kb EcoRI-HindIII LYS2 fragment from YIp600 (BARNES and THORNER 1986 ), into the SpeI-ClaI sites of pBS-KS+. The generated plasmid, pJR2283, was digested with SpeI and ClaI to integrate the 5.4-kb nonvector fragment and disrupt OSH3. In the osh3 allele constructed, 1.65 kb of sequence was deleted and replaced with the LYS2 gene.

The 3.8-kb OSH2 DNA was amplified and cloned into the EcoRI-SalI sites of pBS-KS+. The oligonucleotide pair used for PCR was 5'-CTCGAATTCATGTCTAGGGAAGACTTGTCC-3' and 5'-ACGCGTCGACCGTGTTAAAAAATGTCACCACAATC-3'. From pJR987, a PvuII-SphI fragment containing URA3 was subcloned into the SnaBI-SphI sites of OSH2 in pBS-KS+. This disruption plasmid, pJR2287, was digested with EcoRI and SalI to integrate and delete OSH2. In the deletion of OSH2, 1.45 kb of OSH2 coding sequence was removed and replaced with the URA3 gene.

In some osh1 and osh2 strains, the prototrophic marker was converted from URA3 to kanMX4. In these strains, the URA3 gene was replaced with kanMX4 using the disruption construct pJR2284. The construction of pJR2284 was made by first inserting a BglII-SacI kanMX4 fragment from pFA6a-kanMX4 (WACH et al. 1994 ) into the BamHI-SacI sites of pBS-KS+. Then, into AccI-ApaI sites of this plasmid, a 565-bp NotI-ApaI fragment and a 525-bp NotI-AccI URA3 fragment were inserted representing the 5'- and 3'-ends of the gene, respectively. To replace URA3 with kanMX4, pJR2284 was digested with NotI and the entire linearized plasmid was transformed into yeast. Transformants resistant to Geneticin (G148) and 5-fluoroorotic acid (BOEKE et al. 1984 ) were selected.

To generate an allele of OSH2 under regulated control of the MET3 promoter, the plasmid pJR2285 was constructed. In pJR2285, the EcoRI-PvuII fragment from pJR2286, containing the OSH2 open reading frame, replaced the EcoRI-MseI RAS2 fragment from pJR1786. In addition to the P^MET3-OSH2 promoter fusion, pJR2285 encoded TRP1 and, after linearizing with a BstXI partial digest, the construct could be integrated at sequences adjacent to trp1901. Potential transformants were selected on plates containing solid synthetic medium without tryptophan. In the absence of methionine, the TRP1-integrated P^MET3-OSH2 construct suppressed the lethality of a strain lacking all OSH genes. In medium supplemented with 100 mg/liter methionine, the P^MET3-OSH2 construct could not rescue an osh1-7 strain, indicating that the promoter was sufficiently repressed by methionine to deny cells the essential function of the OSH genes.

Reporter gene analysis:
The construction of the 96 plasmids containing the green fluorescent protein (GFP) fused in frame to 96 different open reading frames, and fluorescence detection and analysis is described by DIMSTER-DENK et al. 1999 . The 96 plasmids were transformed into a kar1-1 strain (MS2339) and at least 10 independent transformants were pooled and then inoculated onto solid medium made with 0.5% agarose. These plates were made with solid synthetic complete medium (SC-Ura) and were supplemented with 50 mg/liter tryptophan. The arrayed transformants were grown at 30°.

The 96 plasmids were transferred to each of the seven deletion strains and their controls by exceptional cytoduction (DUTCHER 1981 ) producing the CBX strain series (see Table 1). The plasmids were efficiently introduced into all recipient strains from an original set of kar1-1 transformants. Recipient strains were grown on plates as a lawn and replica-printed onto plates with the kar1-1 donor strains. Mating occurred on solid YPD at 30° for 6–10 hr after which plasmid transfer was selected on SC-Ura containing 3 mg/liter cycloheximide. Multiple papillae (>20) from each recipient were resuspended in sterile water and reapplied onto SC-Ura with cycloheximide.

Gene expression was normalized to the corresponding wild-type strain with matching prototrophies. A set of Osh⁺ control strains was used in which each was paired for analysis with the appropriate osh mutant strains bearing identical prototrophic markers to avoid marker effects on gene expression. In general, differences between profiles of the various prototrophic wild-type strains were nominal, indicating that marker differences contributed insignificantly to any profile similarities between OSH mutants.

For each plasmid in each strain, GFP fluorescence was averaged from at least 16 measurements. Pairwise profile comparisons between normalized expression ratios from each OSH deletion mutant were quantified by correlation coefficients, r. The significance of r was determined by the test statistic t (where t = r[]^0.5 and n is the number of data pairs for which there was a twofold or greater effect for at least one OSH mutant strain). Using Student's t-distribution, the probability of t at n - 2 d.f. was determined at a 99% confidence level. Color representations of gene expression were produced using the Dot Display program (TOD FLAK, personal communication).

Sterol lipid analysis:
Sterol lipids were saponified and extracted using a modification of a published method (HAMPTON and RINE 1994 ). For duplicate analysis of the same culture, 200 ml of exponentially growing yeast (0.6 to 1.0 OD₆₀₀) were split into two equal volumes and harvested by centrifugation. The cells were washed once with an equal volume of distilled water. Pellets were resuspended in 2.5 ml 0.1 M HCl and placed in a boiling water bath for 20 min. After centrifugation, cells were washed twice with 5 ml distilled water and then the cell pellets were resuspended in 0.5 ml 67% methanol. Glass beads were added to the mixture, and cells were lysed by vortexing twice for 3 min each. To the glass bead slurry, 2.5 ml methanol and 1.25 ml 60% KOH were added, and the suspension was heated at 70° for 90 min. Free sterols were isolated with four 2.5-ml hexane extractions after which 0.5 g anhydrous Na₂SO₄ (Sigma) was added to the pooled extracts.

Within the extracts, the identity of sterol lipids present was determined by tandem gas chromatography-mass spectroscopy (GC-MS) and the amount of each sterol was determined by quantitative GC. As an internal standard, 50 µl of a 1.00 mg/ml solution of cholesterol (Sigma) in ether was added to each hexane extract. The solvent was then evaporated at 40° under a stream of N₂, and the residue was dissolved in 100 µl CH₂Cl₂. The quantitative measurement of underivatized sterol was performed using GC, with flame ionization detection, under these conditions: injector temperature 280°, splitless injection (1.5 min); column 5% phenyl-95% methylpolysiloxane WCOT capillary, 0.25 mm ID x 30 m; temperature program 180° (1.5 min isothermal), to 240° at 20°/min, to 300° at 3°/min, 10-min hold at 300°; carrier gas He at 0.84 kg/cm²; detector temperature 300°. Individual compounds were quantified from peak area ratios compared to the internal standard peak, with the assumption of equal 1.0 response factors for all sterols. To ensure accuracy, duplicate injections of each sample were performed. Using GC-MS, individual underivatized sterols were identified by retention time and/or MS comparison with literature data (BARD et al. 1977 ; NES et al. 1989 ; NBS EPA/NIH Mass Spectral Data Base) or, in the case of ergosterol, with an authentic sample. GC conditions were the same as those used for quantitative analysis; MS detection was by electron impact at 70 eV.

Sterol lipid content was calculated either as a function of culture optical density (OD) or normalized to protein mass. For the samples in which sterol content was normalized to OD₆₀₀, optical density and cell size were shown to be equivalent for all the strains tested by plating dilutions of equal OD₆₀₀ of cells on solid rich medium and counting the colonies formed. Yeast cells depleted of Osh proteins, however, were significantly larger than wild-type cells. Therefore, comparisons between sterol content of wild-type and Oshp depleted cells were calculated relative to cytosolic protein concentration. For protein determination, 0.5 ml of culture was pelleted, resuspended in 0.25 ml water, glass beads were added, and cells were then lysed by vortexing for 3 min. Insoluble debris was discarded after centrifugation in a microcentrifuge. Protein concentration was determined by Bradford assay using bovine serum albumin (Sigma) as standard. The data for each analysis represented an average of at least four measurements.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The OSBP superfamily of genes:
To define and identify OSBP homologues, the protein sequence of the first-identified OSBP, rabbit OXYB, was used in sequence database searches. BLASTP searches identified 39 nonredundant protein homologues from a diverse set of eukaryotes including plants, metazoans, and fungi. Within these proteins, similarity was highest in a small domain of 150–200 amino acids. The derived consensus sequence for this "OSBP domain" is shown in Fig 1A. Within the OSBP domain, sequence identity was concentrated within three smaller subdomains separated by a region of variable size and sequence unique to each protein (Fig 1A). We defined OSBP homologues by virtue of their similarity to all three subdomains of the OSBP consensus sequence.

View larger version (104K): In this window In a new window Download PPT slide   Figure 1. Sequence alignments of the 7 yeast OSBP homologues compared to other related proteins. (A) Alignment of amino acid sequences from all yeast OSBP homologues (Osh1p–7p), OXYB, and the consensus sequence compiled from 39 independent OSBP homologues. The 39 OSBP homologues included the following: (A. thalius) At2g31020, At2g31030, At4g08180, At4g12460, At4g22540, At4g25850, At4g25860, F3F19.19, F3L24.17; (Candida albicans) OBPa, OBPalpha; (C. elegans) C32F10.1, F14H8.1, Y47D3A.17, ZK1086.1; (D. melanogaster) Cg1513, Cg3860, D.m.OSBP; (H. sapiens) BAA91118.1, BAA91496.1, DJ430N08.1, KIAA0704, KIAA0772, H.s.OXYB; (Mus musculus) M.m.OSBP; (N. crassa) osbP; (Oryclolagus cuniculus) O.c.OXYB; (S. cerevisiae) see RESULTS; (S. pombe) SPAC23H4.01c, SPBC1271.12, SPBC2F12.05c, SPBC354.07c, SPCC23B6.01c. Each residue in the consensus represented the amino acid found in at least 90% of OSBP homologues at that position. Solid boxes indicated identity and shaded boxes indicated similarity between a majority of aligned amino acids. Residues identified by an asterisk (*) signified invariant amino acids found in all 39 OSBP homologues. The bars (1–3) indicated OSBP subdomains within which the greatest sequence identity between OSBP homologues was located. The three subdomains were separated by regions less conserved in sequence and in length. (B) Alignment of the ankryin repeat domains of Osh2p and Osh1p with human Ankyrin-2. The N-terminal regions of both Osh2p and Osh1p contained two motifs with similarity to ankyrin repeats. The motifs from Osh2p and Osh1p with the highest similarity to the canonical ankyrin repeat are indicated by the bar numbered 1. The second, less homologous repeats of Osh2p and Osh1p are indicated with bar numbered 2. (C) The pleckstrin homology (PH) domains of Osh3p, Osh2p, Osh1p, and OXYB, aligned with mouse pleckstrin-2.

The yeast OSBP homologues are encoded by seven "OSH" genes:
The S. cerevisiae genome encoded seven OSBP homologues. These genes corresponded to yeast open reading frames YHR001w, YKR003w, YHR073w, YDL019c, YAR042w, YPL145c, and YOR237w, respectively designated OSH1–OSH7.

All yeast OSHs encoded proteins with small domains that shared high overall similarity to the OSBP consensus (Fig 1A). Within the OSBP subdomains some residues were invariant in all OSBP homologues. The yeast Osh proteins differed widely in size. The largest proteins, Osh3p, Osh2p, and Osh1p, contained PH domains (Fig 1C) amino-terminal to the OSBP domains (SCHMALIX and BANDLOW 1994 ; LEVINE and MUNRO 1998 ). PH domains regulate protein targeting to membranes and thereby serve as membrane adapters (HEMMINGS 1997 ; LEMMON et al. 1997 ). In addition to PH domains, ankyrin repeats were found within the N-terminal sequences of Osh2p (SCHMALIX and BANDLOW 1994 ) and Osh1p (Fig 1B). Ankyrin repeats mediate protein-protein interactions and are found in many proteins including cytoskeletal proteins and some transcription factors (reviewed by SEDGWICK and SMERDON 1999 ). Thus the structure of Osh2p and Osh1p is suggestive of being able to bind both a phosphoinositide lipid through their PH domain and a protein partner through their ankyrin repeats.

If the OSBP domain is responsible for binding oxysterols, it would appear to be a unique sterol-binding motif. By paired BLASTP sequence comparisons, no similarity was found between the OSBP domain and oxysterol-binding steroid nuclear hormone receptors (RUSSELL 1999 ). In addition, the sterol-binding motif common to SCAP, NPC1, Patched, and HMG-CoA reductase, a motif consisting of five membrane-spanning helices (LANGE and STECK 1998 ), was not present in any of the OSBP homologues.

On the basis of overall sequence homology the yeast Osh proteins were divided into four subfamily groups: (1) Osh1p and Osh2p, (2) Osh3p, (3) Osh4p and Osh5p, and (4) Osh6p and Osh7p. Over the region of homology, members of each subfamily were at least 55% identical; between subfamilies, identity was <30% (Fig 2). Osh1p, Osh2p, and Osh3p shared greatest homology to OXYB. Like OXYB, these yeast Osh proteins share regions of similarity, such as the PH domain, that lie outside the OSBP consensus domain.

View larger version (21K): In this window In a new window Download PPT slide   Figure 2. The identities and similarities between OXYB and the yeast Osh proteins. Percentage identity and similarity (number in parentheses) between the canonical O.c. OXYB and its yeast homologues as determined by BESTFIT sequence analysis. The yeast Osh proteins were grouped into four sequence subfamilies on the basis of protein identities exceeding 55% (indicated by highlighted numbers); Osh3p defined its own subfamily and therefore was not >55% identical to another Oshp. Although all Osh proteins shared significant similarity to OXYB over the OSBP domain, the closest identity with OXYB was limited to two subfamilies (corresponding to Osh1p/2p and Osh3p, respectively).

Secondary structure predictions indicated that all yeast Osh proteins are likely to be soluble proteins. Like mammalian OXYB, the yeast Osh proteins lack any predictable membrane-spanning domains. Most hydrophobic spans were too short to traverse a membrane bilayer or were not predicted to be -helical, and no N-terminal secretory signal sequences were found (Fig 3). Since mammalian OSBP (RIDGWAY et al. 1992 ) and yeast Osh4/Kes1p (FANG et al. 1996 ) have been detected on intracellular membranes, OSBPs are likely to be peripheral rather than integral membrane proteins. Also, like OXYB, potential coiled-coil regions were identified in most of the yeast homologues (Fig 3). On the basis of these secondary structure predictions, OXYB and the yeast Osh proteins were not expected to be integral membrane proteins, but some might bind other membrane proteins through coiled-coil domain interactions. Membrane association may also be conferred by a combination of interactions with membrane proteins, through ankyrin repeats and coiled-coil domains and through lipid/PH domain interactions.

View larger version (44K): In this window In a new window Download PPT slide   Figure 3. Predicted secondary structure of the yeast Osh proteins. Secondary structure motifs for each yeast Oshp, as well as OXYB, were predicted from their amino acid sequence by three methods. For each protein indicated, the top graph plots the probability of coil-coil domain formation vs. amino acid residue number. The second illustration defines blocks of potential -helical regions. The bottom graph plots hydrophilicity vs. residue number. A likely membrane-spanning domain would constitute a contiguous stretch of 19–20 residues predicted to form an -helix, with a hydrophilicity score of <-1.6 over the entire length (KYTE and DOOLITTLE 1982 ). By these criteria, none of the OSBPs was likely to be an integral membrane protein. The bottom figure depicts important sequence motifs and their relative positions within each protein. Pleckstrin homology motifs (PH) are indicated by green boxes, ankyrin repeats (ANK) are indicated by shaded boxes, and the OSBP domains (OSBP) are indicated by solid boxes.

Disruption of the yeast OSH genes:
To determine whether any of the seven OSH genes was necessary for growth, strains in which each OSH was deleted and substituted with a prototrophic marker were constructed (see MATERIALS AND METHODS). Haploid cells lacking any single OSH gene grew normally regardless of growth medium or temperature. Therefore OSHs either were involved in nonessential processes or performed one or more essential but overlapping functions.

Some of the previously characterized OSH deletions had marginal changes in cellular ergosterol concentrations (JIANG et al. 1994 ; FANG et al. 1996 ; DAUM et al. 1999 ). To identify and quantify sterol lipids, saponified lipid extracts from each deletion strain were analyzed by tandem gas chromatography-mass spectroscopy. Relative to extracts from wild type, extracts from most of the deletion mutants had nearly the same level of ergosterol and sterol precursors (Fig 4). Compared to wild type, however, osh5 and osh6 strains contained a statistically significant elevation in steady-state ergosterol levels.

View larger version (31K): In this window In a new window Download PPT slide   Figure 4. Sterol lipid concentrations of individual OSH deletion mutants. From wild type and OSH deletion strains, free membrane sterol lipids were extracted and quantified by GC-MS (see MATERIALS AND METHODS). Steady-state amounts of three representative lipids are shown: squalene, lanosterol, and ergosterol. The levels of the other detected sterol lipids were equivalent between wild-type and mutant extracts (our unpublished observations). , ergosterol; , lanosterol; , zymosterol.

Gene expression profiles of the osh strains:
To determine whether there were any phenotypic differences between each OSH deletion mutant, we compared their expression profiles utilizing a collection of promoter-fusion reporter plasmids representing 96 yeast genes (DIMSTER-DENK et al. 1999 ). In each plasmid, a specific gene was fused at the position of the fourth amino acid to the coding region of green fluorescent protein and yeast colonies carrying each individual reporter were independently cultured and assayed for GFP fluorescence. The reporters represented genes encoding all known mevalonate pathway proteins, proteins involved in lipid metabolism, and proteins that respond to other well-established cellular responses (e.g., heat-shock, pheromone induction, and DNA repair; Table 2).

In the analysis of the OSH deletion mutants, expression profiles provided a sensitive measure of phenotype, a "fingerprint" of changes in expression in response to the loss of a particular gene. If each of the OSHs performed exactly the same cellular function, then the overall expression profiles for each deletion mutant would be identical to each other and to wild type. If the OSHs performed different cellular functions, then the profiles for each mutant would be distinct. The expression profiles demonstrated clear differences between deletion mutants and between deletion mutants and wild type (Fig 5). Of the 96 reporter plasmids, 39 were induced or repressed at least twofold in one or more OSH deletion mutants (Fig 5). In most cases, the profiles of each osh were unique, indicating that the deletion of most OSH genes had distinct consequences. The one exception involved osh5 and osh6, whose expression profiles correlated. By this analysis, OSH5/HES1 and OSH6 appear to share some functional relatedness, as suggested by the similar sterol lipid composition of osh5 and osh6 strains, described above.

View larger version (32K): In this window In a new window Download PPT slide   Figure 5. Expression profile analysis of individual OSH deletion mutants. The responses of specific genes to the deletion of individual OSHs are represented by color. The magnitude of changes, relative to wild-type controls, is color coded according to the graduation shown. Of the 96 genes listed in Table 2, only the 39 that exhibited a twofold or greater change in expression in at least one mutant profile are displayed. The degrees of correlation between the 21 pairwise comparisons of OSH mutant profiles shown are as follows (see MATERIALS AND METHODS): r(osh6&osh7) = 0.090 and t = 0.55; r(osh3 &osh6) = 0.24 and t = 1.52; r(osh2&osh3) = -0.15 and t = -0.91; r(osh1&osh2) = -0.0057 and t = -0.034; r(osh1 &osh4) = -0.11 and t = -0.67; r(osh4&osh5) = 0.17 and t = 1.1; r(osh3&osh7) = -0.073 and t = -0.45; r(osh2&osh6) = 0.018 and t = 0.11; r(osh1&osh3) = 0.32 and t = 2.0; r(osh2&osh4) = 0.15 and t = 0.93; r(osh1&osh5) = 0.048 and t = 0.29; r(osh2&osh7) = 0.21 and t = 1.3; r(osh1&osh6) = -0.053 and t = -0.32; r(osh3&osh4) = 0.11 and t = 0.65; r(osh2&sh5) = 0.072 and t = 0.44; r(osh1&osh7) = -0.24 and t = -1.5; r(osh4&osh6) = 0.11 and t = 0.66; r(osh3&osh5) = 0.30 and t = 1.9; r(osh4&osh7) = 0.10 and t = 0.61; r(osh5&osh6) = 0.91 and t = 13.7; r(osh5&osh7) = 0.082 and t = 0.50. Only the profiles of osh5 and osh6 strains showed statistically significant correlation (at a 99% confidence level).

The deletion of OSH4/KES1 appeared to affect expression of the 96 genes to the greatest degree, whereas OSH2 affected gene expression the least (Fig 5). Several of the genes examined were induced in some osh strains but repressed in others (e.g., ERG8, SOD1). Only a few genes (COQ1, CPS1, GSC2, SUC2, YDR516C) were either uniformly repressed or uniformly induced in most OSH mutants, and none of these genes function directly in sterol lipid biosynthesis. Only one of these genes, COQ1, was involved in isoprenoid biosynthesis (coenzyme Q biosynthesis). If the OSH deletions have a common effect on sterol homeostasis, it was not revealed by changes in expression of mevalonate pathway genes.

Bypass suppression of sec14-1 temperature sensitivity by OSH deletion:
SEC14 encodes a phospholipid transfer protein capable of binding both phosphatidylcholine and phosphatidylinositol (BANKAITIS et al. 1990 ). The essential function of SEC14, required for secretory export from the Golgi complex, can be bypassed by the deletion of OSH4/KES1 (FANG et al. 1996 ). To determine if other OSH deletions could restore viability to a SEC14 mutant, each of the seven osh strains and their wild-type parent were crossed to a strain (CTY1-1A) bearing the temperature-sensitive allele, sec14-1. Diploids were sporulated and tetrads were dissected onto solid rich medium (>15 tetrads analyzed) and incubated at 23°. To test for temperature sensitivity, the dissection plates were replica-printed onto solid rich medium and incubated at 37°. Consistent with previous findings (FANG et al. 1996 ), all osh4 sec14-1 spores grew at the restrictive temperature. All spores carrying sec14-1 and any of the other OSH deletions were still temperature sensitive, indicating that the bypass suppression of sec14 was specific to the osh4 allele. Moreover, when osh4 sec14-1 strains were transformed with high-copy plasmids containing any of the OSHs, only the transformant strain with the OSH4/KES1 plasmid was temperature sensitive. Thus, none of the other OSH genes on high-copy-number plasmids could restore OSH4/KES1 function in the context of the osh4 suppression of sec14-1 lethality.

Sterol-related phenotypes of single and multiple osh mutants:
Inspired by previous studies (JIANG et al. 1994 ) and our finding that some OSH deletions affected ergosterol levels, we explored whether all single osh mutants manifested sterol-related defects. On rich medium, many sterol-related mutants exhibit a defect in tryptophan transport when grown at low temperatures (GABER et al. 1989 ). Some osh strains were reported to grow poorly due to a defect in tryptophan uptake (JIANG et al. 1994 ). In our experiments, however, such growth defects were observed only in mutants with multiple OSH deletion alleles (Table 3). We also compared the growth of individual OSH mutants (and multiple deletions) to the growth of wild type in the presence of lovastatin, nystatin, or high concentrations of NaCl. All results are catalogued in Table 3 and examples are shown in Fig 6. Lovastatin inhibits HMG-CoA reductase, the rate-limiting step in isoprenoid and sterol lipid biosynthesis (ALBERTS et al. 1980 ) and confers growth sensitivity to strains defective for sterol biosynthesis and its regulation. Nystatin is a polyene antifungal drug that binds directly to ergosterol in the cell membrane (WOODS 1971 ; WALKER-CAPRIOGLIO et al. 1989 ). Strains resistant to nystatin have reduced levels of ergosterol exposed on the cell surface. Osmotic stress and potential defects in small ion or metabolite transport were examined on 1.2 M NaCl plates. Serial dilutions of each single OSH deletion mutant were spotted onto various rich medium plates, and their growth relative to the parental control was recorded (Table 3). Compared to wild type, the osh1 strain was lovastatin sensitive and somewhat salt sensitive and the osh2 mutant was nystatin resistant as was the osh4 strain. Deletion of just OSH3, OSH5/HES1, or OSH7 had minimal effects on growth under these conditions. Since three of the deletion mutants had distinguishable phenotypes, these results reaffirmed that the OSHs were functionally distinct.

View larger version (75K): In this window In a new window Download PPT slide   Figure 6. Sterol-related OSH mutant phenotypes. Tenfold serial dilutions of wild type and various OSH deletion mutants were spotted onto rich medium and onto rich medium containing (A) 150 µg/ml lovastatin, (B) 20 units/ml nystatin, (C) 1.2 M NaCl, and (D) 1.2 M NaCl, 0.7 M KCl, 1.0 M sorbitol, and 0.15 M LiCl. All plates were supplemented with excess tryptophan (see MATERIALS AND METHODS) and incubated at 30°. The strains were incubated for the times described in Table 3 and then photographed.

To identify possible genetic interactions among the OSHs, crosses were performed to create all combinations of the seven deletion alleles. Of the 127 possible mutant strains (n = 2⁷ - 1), 122 were viable on tryptophan-supplemented rich medium, 3 were viable only on synthetic medium, and 2 were inviable regardless of growth medium. Viable strains were systematically examined on solid medium for sterol lipid and membrane defects (Fig 6 and Table 3). When compared to wild type, and taking into account growth defects observed on rich medium (containing excess tryptophan), the most common defect noted was lovastatin sensitivity. When incubated with lovastatin at 23°, 30°, or 37°, the growth of 45 mutant combination strains was inhibited 100-fold or more (e.g., Fig 6A). At the temperatures tested, 40 combination strains exhibited a 100-fold or greater resistance to nystatin (e.g., Fig 6B). Although a few strains were resistant to NaCl (Fig 6C), a larger number (19) exhibited a 100-fold or greater sensitivity to NaCl (e.g., Fig 6D). Since many of the strains shared the same nutritional prototrophies, it was unlikely that marker effects accounted for any of the observed phenotypes. For example, osh2 osh3 osh4 osh7 and osh1 osh3 osh4 osh7 shared the same prototrophies but only the latter strain was lovastatin sensitive (Fig 6A). Only a few of the mutants were nystatin sensitive or lovastatin resistant and the effects were relatively small.

Specific effects of deleting various sets of OSHs were also evident (Table 3). Under most conditions, mutant combinations that included osh4 were the most severely affected. For instance, the genotypes of almost all cold-sensitive strains, most salt- or lovastatin-sensitive, and most nystatin-resistant deletion combinations included osh4. Under all conditions tested, none of the 11 deletion combinations that grew comparably to wild type included osh4. In contrast, 10 of these 11 strains included osh3. These results suggested that of the OSH family members, deletion of OSH4/KES1 had the greatest impact on yeast cells and deletion of OSH3 the least. The genotypes of all salt-resistant strain combinations included osh6 and/or osh7 and never included osh4. NaCl-sensitive strain combinations were not necessarily sensitive to other salts and many were not osmosensitive. For instance, the strain osh1 osh2 osh3 osh4 osh6 was sensitive to 1.2 M NaCl, 0.7 M KCl, 1.0 M sorbitol, and 0.15 M LiCl (Fig 6D). Despite only a minor difference in genotype, the strain osh1 osh3 osh4 osh5 osh6 was sensitive only to 1.2 M NaCl. In general, NaCl-sensitive strain combinations that included osh2 were more likely than others to also be sensitive to 0.7 M KCl. Thus, some deletion combinations caused pleiotropic membrane defects, but others exhibited selective ion sensitivities presumably by disrupting specific ion transport processes. Compared to wild type, mutant strains with larger multiples of OSH deletions grew poorly and exhibited both germination defects and extensive flocculation (our unpublished observations). The defects of some osh combinations were rescued by additional deletions. For example, the strain osh1 osh2 osh3 was temperature sensitive, lovastatin sensitive, and nystatin resistant but under the same conditions the strain osh1 osh2 osh3 osh5 grew as well as wild type. Also, the deletion of OSH1 caused lovastatin sensitivity (at 37°) but in combination with the deletion of OSH2, OSH6, or OSH7, there was no defect. Taken together, these results suggest multiple roles for OSH gene family involving sterol lipids and the cell membrane.

Antagonistic interactions between OSH genes:
In the absence of all the other OSHs, strains containing only one of OSH2, OSH3, OSH4/KES1, OSH6, or OSH7 were viable on both rich and synthetic medium (Table 3). In contrast, the strain containing only the OSH5/HES1 gene was viable on synthetic medium but inviable on rich medium. Spores predicted to have the genotype osh1-4 (Osh5⁺) osh6-7 did not germinate on tryptophan-supplemented solid rich medium. They did germinate on synthetic medium, but were inviable when streaked onto solid rich medium (Fig 7). Strains containing both OSH1 and OSH5/HES1, in the absence of the other OSH genes, were viable on all media tested (Fig 7). As shown below, without other OSHs, OSH1 was itself insufficient to maintain viability on rich or on synthetic medium. The OSH1 gene seemed to augment the functions of OSH5/HES1 such that together the genes could impart growth on both types of media when the other OSH genes were disrupted.

View larger version (54K): In this window In a new window Download PPT slide   Figure 7. Media sensitivities of the OSH deletion combinations. Specific OSH deletion strains could be propagated on YM synthetic medium but were unable to grow on YPD rich medium or on solid medium containing a mixture of both YM and YPD. Wild-type (Osh+) (SEY6210), osh1-5 (Osh6+) osh7 (JRY6324), osh1-4 (Osh5+ and Osh6+) osh7 (JRY6308), osh1-2(Osh3+)4-7 (JRY6323), osh1-2 (Osh3+) osh4 (Osh5+) osh6-7 (JRY6304), osh1-4 (Osh5+) osh6-7 (JRY6319), and (Osh1+) osh2-4 (Osh5+) osh6-7 (JRY6299) were streaked onto YM synthetic, YPD/YM mixed, and YPD rich media. The strains grew at 30° for 3 days before they were photographed.

Remarkably, some genetic interactions between the OSHs were antagonistic. Most dramatically, the lethality of some specific deletion combinations on rich medium was suppressed by deletion of an additional gene. Spores containing OSH5/HES1 and OSH6 or OSH3 and OSH5/HES1 as the only OSH genes germinated only on synthetic medium and did not grow when streaked onto solid rich medium (Fig 7). However, when OSH5/HES1 was disrupted in these strains, growth (albeit poor) was restored on both rich and synthetic media (Fig 7) and spores containing only OSH3 or OSH6 germinated on rich medium. Thus OSH5/HES1 was functionally antagonistic to OSH3 and OSH6 in cells grown on rich medium.

Media mixing experiments were also conducted to better determine the medium component limiting the growth of these "media-sensitive" strains. OSH deletion strains were streaked onto a solid mixture of rich and synthetic media and onto each individual constituent medium. To ensure that tryptophan was not limiting, excess tryptophan was added to all media (see MATERIALS AND METHODS). All strains unable to grow on rich medium, but viable on synthetic medium, were also unable to grow on the mixed medium (Fig 7). Since both the mixed and synthetic media shared the same pH (5.5), and these strains grew only on synthetic medium, growth was not restricted by pH differences. These results indicated that rich medium contained an inhibitor to the growth of these strains.

The yeast OSH genes shared at least one common essential function:
Although each OSH gene was dispensable for viability, together the OSHs defined an essential gene family. In crosses with each OSH deletion marked with a prototrophic marker, spores lacking all seven genes could not be isolated, regardless of growth medium. To evaluate independently the effect of deleting all OSHs, an integrated P^MET3-OSH2 construct was used to suppress osh1-7 strain inviability. When the OSH2 gene was expressed, the osh1-7 P^MET3-OSH2 strain (JRY6326) was viable. In the presence of added methionine, however, expression of OSH2 was repressed and the strain failed to grow (Fig 8). If osh1-7 P^MET3-OSH2 cells were grown for 24 hr in methionine-containing medium and micromanipulated onto solid medium lacking methionine, 75% (36 out of 48) recovered the growth arrest and formed colonies (96% of wild-type cells formed colonies under the same conditions). Thus, the inviability of most osh1-7 cells could be reversed if OSH2 expression was reactivated after growth arrest. These results confirmed the essential requirement of yeast for the OSHs.

View larger version (100K): In this window In a new window Download PPT slide   Figure 8. Lethal OSH deletion allele combinations. Two of the OSH deletion combinations were inviable regardless of growth medium. Strains including wild type (Osh+) (SEY6210), PMET3-OSH2 osh1-7 (JRY6326), and PMET3-OSH2 (Osh1+) osh2-osh7 (JRY6321) were streaked onto synthetic solid medium without (-Met) or with (+Met) added methionine. Plates were photographed after incubation for 4 days at 30°.

Although only two individual deletions of OSH genes even modestly affected sterol lipid levels, we examined whether depletion of all Osh proteins from yeast would have a greater effect. To deplete yeast of Osh proteins, methionine was added to exponentially growing osh1-7 P^MET3-OSH2 cells. Following growth arrest (corresponding to about four culture doublings), lipids were extracted, saponified, and quantified. Analysis of the extracted sterol lipids by GC-MS indicated a severe perturbation of normal sterol levels. For example, ergosterol concentrations increased 3.5-fold, and 22-dihydroergosterol levels increased 13-fold relative to wild type and, by varying degrees, there were steady-state increases in the levels of many other sterols (Table 4). Some sterol lipids remained largely unaffected by Oshp depletion (e.g., episterol). These results were also consistent with observations by microscopy using the fluorescent sterol-binding polyene, filipin. Fixed cells depleted of Osh proteins and treated with filipin appeared to have significantly greater filipin/sterol fluorescence (our unpublished observations). These results established that an important function of all OSHs, perhaps their essential function, is the maintenance of cellular sterol lipid composition.

Of all seven OSH genes, only OSH1 was incapable of maintaining viability in the absence of the other genes, regardless of growth medium. In crosses with marked OSH deletions, spores in which OSH1 was the only OSH gene left intact could not be isolated. An (Osh1⁺) osh2-7 P^MET3-OSH2 strain could be propagated in the absence of methionine but not in its presence (Fig 8). However, multicopy plasmids containing any of the OSH genes, including OSH1 itself, were able to support growth of the osh1-7 P^MET3-OSH2 strain on a medium repressing OSH2 expression (our unpublished observations). Thus, each of the seven OSH genes, including OSH1 if overexpressed, had the capacity to maintain the essential function(s) common to all yeast Osh proteins.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

This study provided a comprehensive evaluation of the gene family defined by its homology to the mammalian OXYB oxysterol-binding protein. Although this family is present in all eukaryotes examined, the in vivo role of these proteins is unclear. Homozygous OXYB knockout mice fail to develop beyond the first zygotic divisions (J. GOLDSTEIN, personal communication) indicating that the protein carries out an essential function.

Each OSH performed a specific and unique function:
In yeast, null alleles of any single OSH gene had no discernible effect on growth on standard media. However, using broad-based assays guided by the biochemical clue that the family may have a role in sterol metabolism, clear phenotypes were found for the OSH genes. The phenotypes of osh deletion mutants allowed some functions of the yeast Osh proteins to be deduced. First, three single deletion mutations caused cells to be resistant to nystatin, a polyene antibiotic whose toxicity to yeast is proportional to the amount of ergosterol in the cell membrane. Although osh2 and osh4 mutants were resistant to nystatin, they contained wild-type levels of ergosterol. This result suggested that less ergosterol was exposed at the cell membrane in these mutants and, since total ergosterol levels were the same as wild type, some resided at other locations shielded within the cell. Thus, OSH2 and OSH4/KES1 may facilitate the transfer of ergosterol to the cell membrane. In contrast, osh1 strains were sensitive to lovastatin, an inhibitor of an early step in sterol biosynthesis, yet had wild-type levels of sterol lipids. Therefore, osh1 strains had no defect in sterol biosynthesis per se, but the lovastatin sensitivity indicated a defect in the postsynthetic regulation of sterol lipid function. Indeed osh5 and osh6 mutants had elevated sterol levels. Thus, at some level OSH1, OSH5/HES1, and OSH6 were required for the proper regulation of sterol biosynthesis.

As a second and independent evaluation of the relationships among the OSH genes, we compared the expression profiles of 96 genes selected to include genes involved in all aspects of sterol biosynthesis and a range of other processes. This analysis was not based on any assumption about Osh proteins being regulators of transcription. Rather, transcription of the selected genes was simply used as a broad-based molecular phenotype. This analysis revealed clear and distinct differences between all OSH mutants and wild type. With one exception, the pattern of each mutant expression profile was distinctly different from the other mutants, indicating that each OSH had a specific role. The most similar expression profiles were those of osh5 and osh6, which were the only two mutations that individually affected sterol levels.

The use of expression profiles and subtle phenotypes allowed us to dismiss a conventional interpretation of mutants that have little or no phenotype. Specifically, when individual members of a gene family are disrupted and have no readily discernible phenotypic consequences, functional redundancy is often offered as an explanation. As shown here, by expanding the range of phenotypes examined, there was no difficulty establishing that every family member executed a unique function in the cell. Thus, by definition, these genes were not functionally redundant. This analysis did not exclude the possibility that each member of the gene family carried out a common function and, as discussed below, that was exactly the case for the OSH genes.

The OSH family members each performed a common essential function:
The phenotypes of all 127 possible combinations of OSH mutants revealed some simple conclusions and a wealth of phenotypic complexity. To emphasize the most salient result, the lack of all seven OSH genes caused growth arrest. In cells without OSH genes, growth could be restored by any one OSH gene on a multicopy plasmid. Thus, the seven yeast OSH genes, which together shared only a small region of sequence identity, shared at least one essential overlapping function. At face value, it seems incongruent that the knockout of a single OSBP gene in mouse was lethal, but the deletion of all OSH genes was required to kill a yeast cell. A simple explanation might be that the mouse ovum does not carry a maternal store of OSBPs and OXYB is the only OSBP expressed during the early stages of development.

The regulated expression of a single OSH gene in a strain lacking all seven chromosomal copies of OSH genes allowed us to grow these cells and determine what happens when the last remaining OSH gene is shut off. The most striking result was a 3.5-fold increase in the total level of ergosterol in the cell, a dramatic enhancement over the modest increases observed in the osh5 and osh6 single mutants. This induction is astonishing in comparison to wild-type levels of ergosterol, which are normally present at 3.3-fold molar excess over all plasma membrane phospholipids (ZINSER et al. 1991 ). How the yeast cell accommodates the elevation of ergosterol levels to 3.5-fold above normal is a challenge to conventional models of membrane organization. Either the cell membrane adapts to the increased level of ergosterol, perhaps through compensatory changes in other lipid concentrations, or the excess ergosterol accumulates within the cell.

Because the high level of ergosterol overproduction was observed only when the entire OSH family was deleted, each single gene could prevent the massive overproduction of ergosterol and hence each OSH had a common regulatory role. It was unclear, however, whether this ergosterol regulatory role was a direct or indirect part of the common essential function shared by all OSHs.

Lessons from combinations of OSH genes:
The phenotypes of all mutant combinations are described in Table 3, but certain phenotypes warrant particular attention. First, although the primary structures of the seven OSHs fall into four subfamilies, there was little evidence that phenotypes were apt to be more similar among mutants representing the same sequence subfamily. For example, OSH4/KES1 and OSH5/HES1 define a subfamily, yet the deletion of OSH4/KES1, but not OSH5/HES1, had appreciable impacts on growth, salt and lovastatin sensitivity, and nystatin resistance. Second, although most OSHs shared a common essential function, deletion of particular OSH genes often had completely different consequences in viable deletion strains. For example, in contrast to OSH4/KES1 deletion mutants, most of which were NaCl sensitive, all NaCl-resistant strains lacked OSH6 or OSH7.

The NaCl sensitivity of many OSH deletion strains indicated that the yeast Osh proteins affected the cell membrane in several different ways. Some of these strains were sensitive to other salts and conditions that affect cellular osmolarity. Some of the other strains were sensitive to NaCl and only to other specific salts, indicating ion-specific sensitivities. The results suggest that certain deletion combinations generally affect membrane permeability, while others specifically affect ion transport.

Even more surprising was evidence that some OSHs work at cross purposes. This conclusion was based in part on an unanticipated dependence of certain mutant combinations on a particular medium to grow. Specifically, certain mutant combinations grew on synthetic medium but not on rich medium, a phenomenon we currently do not understand. However, clearly rich medium contained an inhibitory substance. Some pairs of OSH genes (OSH5/HES1 and OSH6; or OSH3 and OSH5/HES1) supported growth on minimal medium but not on rich medium, whereas OSH3 or OSH6 alone supported growth on both media. Thus, some individual OSH genes are better than two. Similarly, based upon other phenotypes, cells containing only OSH4/KES1, OSH6, and OSH7 grew better than cells containing only OSH4/KES1, OSH5/HES1, OSH6, and OSH7. In both of these cases, the presence of OSH5/HES1 seemed to antagonize the function of the other OSHs.

Although unusual, the ability of smaller subsets of a gene family to be better for a cell than larger subsets has been seen before in the case of kinesins. Kinesins are microtubule motors that have a characteristic polarity and can move toward only one end of a microtubule. Plus-end movement requires one class of motors and minus-end movement requires a different class of motors (reviewed in HILDEBRANDT and HOYT 2000 ). Cells lacking too many kinesins of a particular class grow poorly or die, but viability can be restored either by adding back the missing motors or by removing a subset of the motors of the other class (SAUNDERS and HOYT 1992 ; HOYT et al. 1993 ). Apparently, a balance of motor types is more critical for cells than the presence or absence of any particular kinesin. If this principle applies to OSH genes, in some contexts the OSH5/HES1 gene would appear to carry out a process in opposition to OSH3 or OSH6. The direction of lipid transport could underlie this phenomenon.

The structure of OSBPs:
The function of the defining motif of the OSBP protein family, the 150-amino-acid tripartite consensus, remains unknown. Although the OXYB protein binds oxygenated sterols directly, it is not clear whether binding is mediated by this conserved motif or by an adjacent region. OSBP family members lack any apparent transmembrane domains, but may function at the surface of membranes. Indeed the Osh1p, Osh4p, and the mammalian OXYB protein associate with vesicle and Golgi membranes (RIDGWAY et al. 1992 ; FANG et al. 1996 ; LEVINE and MUNRO 1998 ). Although the intracellular localization of Osh2p and Osh3p is unknown, they have a PH domain like Osh1p. Moreover, the PH domain of Osh1p binds phosphatidylinositol lipids (LEVINE and MUNRO 1998 ) suggesting that Osh2p and Osh3p may also bind membranes and specifically interact with phosphatidylinositol. Ankyrin repeats, motifs that mediate protein-protein interactions (SEDGWICK and SMERDON 1999 ), found in some of the Osh proteins, and the coiled-coil domains found in all of them, certainly invite the notion that they are a part of larger protein complexes.

Prokaryotes lack sterols and their genomes lack genes encoding OSBPs. OSBP homologues have been found in all eukaryotic genomes examined, including organisms such as Drosophila melanogaster and Caenorhabditis elegans, which have lost the ability to synthesize sterols. C. elegans contains four OSBP homologues and Drosophila contains three (C. ELEGANS SEQUENCING CONSORTIUM 1998; ADAMS et al. 2000 ). In both these cases, the total number of OSBPs is fewer than found in genomes of eukaryotes that can synthesize sterols (