Introduction

Transcriptional repression of regulated structural genes in eukaryotes often depends on pleiotropic corepressor complexes which do not bind directly to cis-acting upstream elements but instead are recruited by specific DNA-binding proteins to target promoters. A well-known corepressor conserved from yeast to mammalian systems is Sin3 (reviewed by Grzenda et al. 2009), initially described as a negative regulator of HO expression in S. cerevisiae (Sternberg et al. 1987; Wang et al. 1990) but also required for repression of glucoamylase genes (Yoshimoto et al. 1992), meiotic genes (Strich et al. 1989), HMR/HML loci (Vannier et al. 1996) and phospholipid biosynthetic genes (Hudak et al. 1994). In addition to Sin3, eukaryotic corepressor complexes Ssn6-Tup1 (not found in higher eukaryotes; Malavé and Dent 2006), N-CoR/SMRT (specific for ligand-free nuclear receptors; Jones and Shi 2003), NuRD (in mammalian systems; McDonel et al. 2009), Groucho (in Drosophila and higher metazoa; Gasperowicz and Otto 2005) and Topless (in plants; Liu and Karmarkar 2008) have been identified for which recruitment of various histone deacetylases was demonstrated.

We and others have previously shown that Sin3 is important for repression of genes involved in phospholipid biosynthesis (Slekar and Henry 1995; Wagner et al. 2001). When phospholipid precursors inositol and choline (IC) are limiting (derepressing conditions), a heterodimer of transcription factors Ino2 and Ino4 binds to the regulatory motif ICRE (inositol/choline-responsive element = UASINO; Schüller et al. 1992; Ambroziak and Henry 1994; Schwank et al. 1995) upstream of phospholipid biosynthetic genes such as INO1, CHO1, FAS1 and FAS2 and triggers transcriptional activation (reviewed by Chen et al. 2007). Under these conditions, cellular concentration of phosphatidic acid (PA) as the central metabolite of phospholipid biosynthesis is high, leading to retention of the pathway-specific transcriptional repressor Opi1 at the endoplasmic reticulum (Loewen et al. 2004). In contrast, at a high concentration of IC (repressing conditions), level of PA drops, leading to release of Opi1 repressor which subsequently enters the nucleus and counteracts gene activation mediated by Ino2. Repression is mediated by direct contact of Opi1 domain AID (activator interaction domain) and Ino2 domain RID (repressor interaction domain; Heyken et al. 2005). Via a distinct domain (OSID, Opi1–Sin3 interaction domain; Wagner et al. 2001), Opi1 also binds to corepressor Sin3, allowing recruitment of histone deacetylase Rpd3 (Kadosh and Struhl 1997) which subsequently may modify local chromatin and prevent access of the transcriptional machinery.

Sin3 works as a platform for multiple protein–protein interactions and contains four paired amphipathic helix motifs (PAH1-4) which are not functionally redundant but instead bind to individual partner proteins (Wang et al. 1990; Wang and Stillman 1993). Opi1 specifically contacts PAH1 of Sin3 (Wagner et al. 2001) while PAH2 is the target of repressor/activator Ume6 (Washburn and Esposito 2001), being involved in general repression via URS1 motifs or activation of early meiotic genes (Steber and Esposito 1995). Sin3 is part of at least two protein complexes, designated Rpd3L and Rpd3S, varying with respect of auxiliary proteins (Carrozza et al. 2005b). Specific repressors of other regulatory pathways such as Mig1 (glucose repression), α2 (repression of a-specific genes in α mating type) and Rox1 (repression of hypoxic genes in the presence of oxygen) utilize corepressor complex Ssn6 (=Cyc8) + Tup1 (=Flk1) to execute gene repression (Keleher et al. 1992; reviewed by Malavé and Dent 2006). Ssn6 contains 10 TPR motifs (tetratricopeptide repeat, comprising 34 amino acids; Blatch and Lässle 1999) at its N-terminus which are indispensable for repressor function and mediate protein–protein interactions. A similar function has been shown for 7 WD-40/β-transducin repeats (Smith et al. 1999) present at the C-terminus of Tup1. Importantly, Ssn6 and Tup1 recruit histone deacetylases Rpd3, Hos1 and Hos2 (Davie et al. 2003) while Tup1 also contacts histones H3 and H4 (Edmondson et al. 1996). Thus, a common mechanism of gene repression mediated by Sin3 and Ssn6/Tup1 corepressor complexes is apparent.

Previously, we have shown that expression of ICRE-dependent genes in an opi1Δ mutant is completely insensitive to high concentrations of IC (Wagner et al. 2001). In a sin3Δ mutant, regulation by IC was clearly alleviated but not entirely abolished. We thus hypothesized that a second mechanism of gene repression acting in parallel to Sin3 may be effective for ICRE-dependent genes. In this work, we show that this is indeed the case, with Ssn6/Tup1 also mediating Opi1-triggered repression. A single domain within Opi1 is able to contact both Sin3 and Ssn6. Interestingly, at least one of both corepressor complexes must be functional to guarantee cellular viability.

Materials and methods

Yeast and E. coli strains, media and growth conditions

Strains of the yeast Saccharomyces cerevisiae used in this work are listed in Table 1. Strain PJ69-4A (James et al. 1996) was used for two-hybrid assays. To test for functional complementation by variants of OPI1, opi1 deletion strain CWY1 (derived from wild-type SIRP3; Schwank et al. 1995) was used. Integrative transformation of CWY1 with expression plasmids containing OPI1 variants (see below) led to strains YDC1-YDC12. Strains with single mutations sin3Δ::kanMX and ssn6Δ::LEU2 were obtained by introduction of the respective null mutant alleles (Wagner et al. 2001) into regulatory wild-type strains isogenic to SIRP3. Yeast extracts used for protein–protein interaction assays were prepared from strain C13-ABY.S86, lacking four vacuolar proteinases.

Table 1 Strains of Saccharomyces cerevisiae used in this work
Full size table

Synthetic complete (SCD) media used for selective growth of transformants and conditions of IC repression or derepression have been described (Schüller et al. 1992). To select against URA3, synthetic complete medium supplemented with 1 mg/ml 5-fluoroorotic acid (5-FOA) was used.

Plasmid constructions

The coding region of OPI1 was mutagenized at selected positions (see below) and gene variants were fused with the natural OPI1 control region (−437/−1). Expression constructs were subsequently inserted as EcoRI/HindIII fragments into integrating plasmid YIplac128 (Gietz and Sugino 1988) to give pDC18–24, pDC40–43, pCM3–4 and pCM17–21 (LEU2 OPI1 var). Deletion variant opi1(Δ2-107) was newly constructed by PCR amplification, fused with its natural promoter and similarly inserted into YIplac128. Plasmids obtained were linearized at a single ClaI site and targeted to the leu2 locus of recipient strain CWY1.

To perform two-hybrid assays, residues 1–106 of Opi1 variants were fused with the DNA-binding domain of Gal4. Thus, specific PCR primers were used to amplify the corresponding region of OPI1, introducing restriction sites EcoRI and BamHI. After activation of these sites, fragments were inserted into plasmid pGBD-C1 (James et al. 1996) to give various Gal4DBD–Opi1(1–106) fusions (pDC29–35, pDC44–47). Plasmid pJW1 [Gal4AD-Sin3(1–300)] has been previously described (Wagner et al. 2001).

To test for synthetic lethality of mutations sin3 and ssn6, single copy plasmids pKH20 (ARS CEN URA3 MET25-SIN3; contains the 4.4 kb SIN3 gene as a BamHI/XhoI fragment) and pYJ37 (ARS CEN URA3 MET25-SSN6; contains SSN6 as a 2.9 kb BamHI/EcoRI fragment) were constructed by insertion of PCR-amplified fragments into expression plasmid p416-MET25 (Mumberg et al. 1994).

E. coli expression plasmids pCW86 [GST–Opi1(1–106)] and pCW136 (tetp/o HA3-OPI1) have been described (Wagner et al. 2001). Bacterial expression of SSN6 length variants representing various TPR motifs was obtained by PCR amplification and subsequent insertion of BamHI/EcoRI fragments into plasmid pGEX-2TK (GE Healthcare) to give GST–Ssn6var plasmids (pJuS5, 11, 13, 14, 15, 16 and 20, pYJ95, 99 and 107).

For yeast expression of Ssn6 length variants, plasmid p426-MET25HA (Mumberg et al. 1994) was used. DNA representing residues 1–398 (comprising TPR motifs) and 399–966 was amplified by PCR (1.2 kb BamHI/EcoRI fragment and 1.7 kb BamHI/EcoRI fragment, respectively) and inserted into the vector cleaved correspondingly to give plasmids pYJ65 [MET25-HA3-SSN6(1–398)] and pYJ64 [MET25-HA3-SSN6(399–966)]. The same vector was used for epitope-tagging of Opi1 missense variants L56A, V59A and V67A. The coding region of OPI1 was cut with BamHI/HindIII from pDC18 (V59A), pDC20 (L56A) and pDC21 (V67A) and inserted into p426-MET25HA to give pJuS7, pYJ108 and pYJ109, respectively. Expression plasmids pCM35-pCM40 containing epitope-tagged Opi1 variants mutated within the Ino2 interaction domain (L333A, L340A, V343A, V350A, L354A and V361A) were constructed correspondingly.

Site-directed mutagenesis

To alter selected residues in the coding region of OPI1, the QuikChange site-directed mutagenesis kit of Stratagene was used. To obtain mutations within OSID and AID, plasmid pWTH99 containing the OPI1 coding region was used (Heyken et al. 2005). To replace selected residues against alanine, we used pairs of mutagenic primers introducing a GCG codon instead of the natural codon, flanked by 15–19 nucleotides on both the sides. DNA sequencing was used to confirm the presence of the desired mutant alleles of opi1 (L53A, L56A, D57A, R58A, V59A, I63A, V67A, F70A, Y71A, L81A, L88A, L333A, L340A, L343A, M347A, V350A, L354A and V361A) and the absence of any other change in the plasmids obtained (pDC11–17, pDC36–39, pCM1–2, pCM12–16).

In vitro-interaction assays (GST pull-down)

To perform interaction assays by affinity chromatography, heterologous expression of GST- and HA-tagged proteins was performed in E. coli strain BL21 (Stratagene/Agilent). Gene expression in transformants containing GST fusions regulated by the tac promoter was induced with 1 mM IPTG. To produce HA-Opi1 in E. coli, tet promoter-dependent expression of pCW136 was activated by 0.2 mg/l anhydrotetracycline. For maximal derepression of MET25-dependent HA fusion gene variants, yeast transformants were cultivated in a selective medium in the absence of methionine.

Following bacterial biosynthesis, GST fusion proteins containing length variants of either Opi1 or Ssn6 were immobilized on glutathione (GSH) Sepharose (GE Healthcare). The procedure of affinity chromatography and subsequent washing steps at intermediary stringency has been described (Wagner et al. 2001). GST fusions and bound HA-tagged proteins were finally eluted with free GSH (10 mM) and subsequently separated by SDS/PAGE. Following transfer to filters, HA fusion proteins were incubated with anti-HA-peroxidase conjugate and detected using POD chemiluminescent substrate (both supplied by Roche Biochemicals).

Miscellaneous procedures

Transformation of S. cerevisiae strains, immunoblot analysis, PCR amplification and β-galactosidase assays was performed as previously described (Schwank et al. 1995; Wagner et al. 2001).

Results

Mutational analysis of Opi1 domain interacting with Sin3

We have previously shown that the N-terminus of Opi1 (residues 1–106) is able to interact with PAH1 of Sin3 in vivo and in vitro (Wagner et al. 2001). Within this domain (designated OSID), an amphipathic pattern of hydrophobic amino acids could be identified (residues 53–88). To investigate the possible importance of these residues for regulated expression of Opi1 target genes, we performed a site-directed mutagenesis at the selected positions, leading to replacement of hydrophobic amino acids to alanine (L53, L56, V59, I63, V67, F70, L81 and L88). As a control, the importance of two charged side chains (D57 and R58) and a hydrophilic amino acid (Y71) was also investigated.

To assay for interaction of Opi1 wild-type and missense variants with Sin3 in vivo, we fused Opi1 residues 1–106 with the DNA-binding domain of Gal4. The resulting plasmids were co-transformed into strain PJ69-4A with a plasmid encoding Sin3 residues 1–300 (comprising PAH1) fused behind Gal4 activation domain. In the case of Opi1-Sin3 interaction, the resulting hybrid constructs should be able to activate the Gal4-dependent reporter gene GAL7-lacZ of strain PJ69-4A. As is shown in Table 2, Opi1 variants L56A, V59A and V67A were defective for wild-type activation of the reporter gene.

Table 2 Mutational analysis of Opi1–Sin3 interaction using two-hybrid analyses
Full size table

We next investigated the consequences of defective interaction of Opi1 and Sin3 for repression of an Opi1-dependent reporter gene. Full-length OPI1 wild-type and missense mutants were transferred into an integrating yeast vector (YIplac128). The resulting plasmids (LEU2 OPI1 variant) were then transformed into strain CWY1 (leu2 opi1Δ::HIS3), containing an integrated ICRE-CYC1-lacZ reporter gene. In agreement with results of the two-hydrid analysis, Opi1 variants L56A, V59A and V67A were unable to restore IC-dependent repression of the reporter gene (Table 3). Opi1 variants which were functional for interaction with Sin3 could also complement the opi1 null mutation of recipient strain CWY1.

Table 3 Influence of opi1 missense mutations within OSID on ICRE-mediated gene regulation
Full size table

In a previous work, we characterized Opi1 length variants for functional complementation of an opi1 null mutation (Wagner et al. 1999) and described that the opi1(Δ2-107) allele lacking OSID could partially restore gene repression. Since this result contradicts the findings with Opi1 missense variants reported above, we re-constructed the deletion allele and repeated the assay for reporter gene expression with transformants obtained. As is also apparent from Table 3, our assays show that loss of repressor function was observed following integration of the new opi1(Δ2-107) allele. The reason for the discrepancy of previous data and results reported in this work remains unclear. Nevertheless, our new findings show that Opi1 lacking residues 2–107 do no longer support repression. This result agrees with the phenotypes of OSID missense mutants.

Synthetic lethality of sin3 and ssn6 null mutants

The phenotype of missense mutants opi1L56A, opi1V59A and opi1V67A mapping to OSID shows a complete loss of function for the respective Opi1 variants. It appears reasonable to assume that a sin3 null mutant should phenocopy this defect. However, this assumption is not entirely supported by the previous finding of severely reduced but, nevertheless, residual gene regulation observed with a sin3 null mutant (Wagner et al. 2001). We thus hypothesized that Opi1 OSID may recruit not only Sin3 but a second mediator of repression as well. Possible candidates are Ssn6 and Tup1 which together affect a large number of target genes repressed by various stimuli (Malavé and Dent 2006). Indeed, previous testing of ICRE-dependent gene regulation in ssn6 and tup1 null mutants also showed an increased gene expression under repressing conditions (Wagner et al. 2001). To further confirm the importance of both corepressor complexes Sin3 and Ssn6/Tup1 for repression of ICRE-dependent genes, we wished to construct a double mutant sin3 ssn6. However, we were unable to obtain this double mutant by introduction of a ssn6Δ::LEU2 disruption cassette into a sin3 deletion strain. Similarly, no meiotic products of the desired genotype were obtained after crossing single mutants sin3 and ssn6 and subsequent sporulation. Thus, mutations sin3 and ssn6 may be synthetically lethal. To confirm this, we constructed plasmids containing markers ARS CEN URA3 SIN3 (pKH20) and ARS CEN URA3 SSN6 (pYJ37) which were subsequently transformed into strain JS05.4 (ura3/ura3 sin3/SIN3 ssn6/SSN6). Following sporulation, we selected for haploid progeny with sin3 ssn6 double mutations, carrying plasmid copies of either SIN3 or SSN6. Similarly, ssn6 and sin3 single mutants were transformed with SSN6 or SIN3 plasmids. While transformants ssn6 + SSN6 and sin3 + SIN3 were still viable after selecting against URA3 plasmids using 5-fluoroorotic acid (5-FOA), no growth was observed after plasmid curing of the double mutant (Fig. 1). Thus, at least one corepressor complex Sin3 or Ssn6 + Tup1 must be functional to support viability of the yeast cell. We conclude that mutations sin3 and ssn6 are indeed synthetically lethal. The same result was obtained for mutations sin3 and tup1 (not shown).

Fig. 1
figure 1

Synthetic lethality of mutations sin3 and ssn6. Serial dilutions of regulatory wild-type strain JS91.15-23 and strains with single or double mutations ssn6 and sin3 were transferred to YEPD, SCD-Ura and SCD + 5-FOA media. The following transformants were analyzed: JS91.15-23 + p416-MET25 (wild-type + empty vector), JS05.5-2B + pYJ37 (ssn6Δ + SSN6 plasmid), JS05.1-3 + pKH20 (sin3Δ + SIN3 plasmid), JS05.4-3C + pYJ37 (ssn6Δ sin3Δ + SSN6 plasmid) and JS05.7-4A + pKH20 (ssn6Δ sin3Δ + SIN3 plasmid). Relevant genotypes and times of incubation (d, days) are indicated

Full size image

OSID interacts with TPR motifs of Ssn6

To confirm our assumption of Ssn6 being important for Opi1-mediated gene repression, we assayed for in vitro-interaction of GST–Opi1 (1–106) with HA-tagged Ssn6. It has been previously shown that TPR motifs of Ssn6 are responsible for interaction with various repressor proteins (Tzamarias and Struhl 1995; Smith et al. 1995). We thus separately expressed epitope-tagged Ssn6 residues 1–398 (comprising its TPR motifs) and 399–966 in yeast. Although both fragments of Ssn6 could be stably detected in protein extracts (Fig. 2a, lanes 1, 2), only the TPR-containing fragment was able to interact with GST–Opi1 (Fig. 2a, lane 4). No evidence for direct binding of HA-tagged Tup1 to GST–Opi1 could be obtained (not shown).

Fig. 2
figure 2

In vitro interaction of Opi1 and Ssn6. a Input control and assay for Opi1-binding of Ssn6 fragments 1–398 and 399–966 (plasmids pYJ65 and pYJ64). GST–Opi1(1–106) was synthesized in E. coli (BL21 transformed with plasmid pCW86), immobilized on GSH Sepharose and incubated with yeast total protein extract. b Length variants of Ssn6 comprising various combinations of TPR motifs were fused with GST and incubated with bacterially synthesized HA-Opi1 (plasmid pCW136). c Comparative analysis of interaction of GST–Ssn6(79–175) comprising TPR motifs 2–3 (pYJ95) with HA-tagged Opi1 wild-type and missense variants (full-length) expressed in S. cerevisiae (plasmids pCW116, pYJ108, pJuS7 and pYJ109). Extracts containing 75 μg of total protein were analyzed for the input control. To achieve comparable amounts of HA-Opi1 variants for the interaction assay, total protein was adjusted accordingly

Full size image

We next asked whether a combination of specific TPR motifs or a minimal number of them is required for Opi1 OSID to bind. For reasons of protein stability, Ssn6 TPR length variants were fused with GST and subsequently incubated with HA-Opi1. As is shown in Fig. 2b, Ssn6 fragments containing non-overlapping TPR motifs were able to bind Opi1 with comparable efficiency. Similarly, the number of motifs contained in the fusion constructs did not influence binding efficiency. Finally, we arbitrarily selected individual TPR motifs 1, 5 and 8 for fusion with GST and were able to show that each motif was sufficient for interaction with Opi1.

As shown above, Opi1 variants L56A, V59A and V67A led to IC-insensitive expression of the ICRE-dependent reporter gene. We thus compared binding of Opi1 wild-type and missense variants to TPR motifs 2–3 of Ssn6. As is shown in Fig. 2c, Opi1 full-length mutant proteins could be expressed in S. cerevisiae but were unable to interact with bacterially synthesized GST–Ssn6(TPR2–3), explaining the loss of repression observed with Opi1 mutants L56A, V59A and V67A.

Mutational analysis of Opi1–Ino2 interaction

We have previously shown that carboxy-terminal residues of Opi1, designated AID (amino acids 321–380), are necessary and sufficient for binding to Ino2 (Heyken et al. 2005). A closer inspection of AID revealed the existence of a periodic pattern of hydrophobic amino acids, reminiscent of a leucine zipper (L-N6-L-N6-M-N6-L-N6-V). It should be emphasized that this motif is distinct from the genuine leucine zipper located at the core of Opi1 (residues 139–160). Similar to our analysis of OSID, we replaced all hydrophobic residues within AID compatible with an amphipathic α-helix (L333, L340, V343, M347, V350, L354 and V361) by alanine and integrated the OPI1 variants obtained into the chromosomal DNA of an opi1 null mutant. Stability of HA-tagged variants was confirmed by immunoblot analysis (cf. Fig. 3). As is apparent from Table 4, all Opi1 variants investigated were unable to trigger repression of the reporter gene in the presence of IC (exception: missense mutation M347A allowed a substantial repression but did not completely restore wild-type regulation). These findings support the hypothesis of a leucine-zipper-related structural motif as the critical core of Opi1 repressor required for deactivating Ino2.

Fig. 3
figure 3

Control of stable expression of Opi1 variants containing missense mutations within the Ino2 activator interaction domain (AID). Plasmids pCW116 (wild-type) and pCM35–pCM40 were transformed into strain C13-ABY.S86. For Western blot analysis with anti-HA antibodies, extracts containing 75 μg of total protein were used

Full size image
Table 4 Influence of opi1 missense mutations within AID on ICRE-mediated gene regulation
Full size table

Discussion

Previous analyses have shown that pathway-specific repressor proteins require interaction with corepressors to finally fulfil their function. In yeast, corepressors Sin3 and Ssn6/Tup1 can be recruited to pleiotropically downregulate a large number of unrelated structural genes (Green and Johnson 2004; Grzenda et al. 2009; Malavé and Dent 2006). Genetic evidence for a role in repression and demonstration of physical interaction with Sin3 has been shown for repressors Ume6 (Kadosh and Struhl 1997), Ash1 (Carrozza et al. 2005a), Whi5, Stb1 (Takahata et al. 2009), Fkh1, Fkh2 (Ho et al. 2002) while Ssn6/Tup1 directly mediate repression by α2, Crt1, Mig1, Rox1 (Malavé and Dent 2006), Nrg1 (Berkey et al. 2004), Sfl1 (Conlan and Tzamarias 2001), Sko1 (Proft and Struhl 2002), Rgt1 (Polish et al. 2005), Hmo1 (Krogan et al. 2006) and Cup9 (Xia et al. 2008). Both corepressors may also interact with activators Pho4, Hac1 (recruiting Sin3; Graumann et al. 2004; Schröder et al. 2004), Hap1 and Aft1 (recruiting Ssn6/Tup1; Zhang and Guarente 1994; Fragiadakis et al. 2004). Consequently, a positive influence on transcription for yeast corepressor complexes Sin3 and Ssn6/Tup1 has been described as well (Vidal et al. 1991; Proft and Struhl 2002).

In this work, we have shown that a single domain of yeast repressor Opi1 (OSID) is able to interact with both corepressor complexes Sin3 and Ssn6/Tup1. Mutational analysis of OSID led to the identification of Opi1 variants L56A, V59A and V67A which could be stably expressed in S. cerevisiae but were defective for interaction with Sin3 and entirely insensitive to gene repression by IC. In contrast, a sin3 deletion mutant was strongly, but not completely deregulated (Wagner et al. 2001). We thus concluded that OSID may be involved in a second pathway of repression acting in parallel to Sin3. Indeed, we could show that OSID not only binds to PAH1 of Sin3 but also interacts with TPR motifs of corepressor Ssn6. No evidence for interaction of Opi1 with Tup1 could be obtained. The functional importance of Opi1–Ssn6 interaction was further supported by our observation that Opi1 variants L56A, V59A and V67A were defective for in vitro interaction with Ssn6. This finding explains why target gene repression in the presence of these variants was completely abolished. Interaction of two corepressor complexes with a single specific repressor may be the exception but not a general rule, at least in yeast. Besides Opi1, recruitment of Sin3 and Ssn6/Tup1 has been described only for Cti6 involved in regulation of metal acquisition by yeast (Papamichos-Chronakis et al. 2002; Puig et al. 2004). While Ssn6 binds to the C-terminus of Cti6, no mapping of its Sin3 binding site has been performed. Thus, it is unclear whether a single domain of Cti6 mediates interaction with both corepressors.

To further confirm the contribution of both corepressors, we wished to assay ICRE-dependent gene expression in a sin3 ssn6 double mutant. However, we were unsuccessful to construct such a mutant and were able to show synthetic lethality of both mutations. The same is true for sin3 and tup1 mutations. These results are not entirely surprising, considering the substantially reduced growth rate of each single mutant. Our finding is in agreement with synthetic lethality of ssn6 and rpd3 described in a study on genetic interactions among factors of histone modification (Lin et al. 2008).

Mapping of the Opi1 binding domain within Ssn6 using GST pull down assays clearly allowed us to demonstrate OSID-TPR interaction which agrees with results previously obtained for repressors such as α2 and Mig1. However, two-hybrid fusion constructs containing Gal4DBD–Opi11–106 and Gal4TAD–Ssn6TPR1–10 could not activate a GAL7-lacZ reporter gene (data not shown). Presumably, the presence of highly potent TPR repression domains which are autonomously able to recruit histone deacetylases (Watson et al. 2000) counteracts gene activation as the working principle of the two-hybrid approach. Mapping studies in vitro using combinations of several TPR motifs and even single TPR motifs revealed a surprising degree of functional redundancy among them. Although a strong conservation at the sequence level is apparently absent, single TPR motifs 1, 5 and 8 chosen arbitrarily could interact with OSID of Opi1. These findings agree with the results previously described for interaction of α2 repressor and TPR6 of Ssn6 (Smith and Johnson 2000), based on identical methods. On the other hand, a combination of defined TPR motifs required for repression of specific genes has been also described (Tzamarias and Struhl 1995). For α2-Ssn6 interaction, the tripeptide motif SRI reminiscent of the peroxisomal targeting sequence PTS1 has been described as important (Smith and Johnson 2000). OSID does not contain a PTS1 consensus motif but the sequence SNV (residues 65–67) is present. Although V67 is indispensable for Opi1–Ssn6 interaction and target gene repression, it is questionable whether SNV may be considered as PTS1-like because asparagine is biochemically too distant from genuine basic amino acids.

Binding of OSID to PAH and TPR motifs suggest that both interaction modules should be able to contact related targets. Solution structures have been determined for PAH1 and PAH2 of mammalian Sin3, both of which generate an amphipathic four-helix-bundle with an up-and-down topology (Brubaker et al. 2000; also designated wedged helical bundle, Spronk et al. 2000). PAH motifs form a hydrophobic pocket which provides a suitable binding surface for a helical partner protein. According to their structural analysis and modelling of NRSF-mSin3B interaction, Sahu et al. (2008) proposed a consensus sequence for binding to PAH1 (ΦXΦΦSXΦS; Φ, bulky hydrophobic residue; S, residue with short side-chain; X, any residue except proline). Within OSID of Opi1, two sequence motifs correspond to this consensus (residues 53–60: LNILDRVS and residues 68–75: VTFYDEIN). Interestingly, the mutant phenotype found for Opi1 variants L56A and V59A nicely fits to this proposition (V67 maps outside the consensus). Mammalian NRSF (neural-restrictive silencer factor) contains identical residues at corresponding positions which are involved in hydrophobic interactions with α-helical segments of PAH1 (L49 and V52 of NRSF with F93 and F58 of mSin3B, respectively). The second motif within OSID may be less important since variants F70A and Y71A showed unimpaired interaction with Sin3.

The 34 amino acids of a TPR motif form a pair of antiparallel α-helices with limited sequence conservation (reviewed by D’Andrea and Regan 2003). Using structural data from TPR motifs of protein phosphatase 5, Gounalaki et al. (2000) presented a model of Ssn6-Tup1 interaction, emphasizing hydrophobic interactions. We thus conclude that interaction modules PAH and TPR both employing helical segments with hydrophobic-apolar surfaces should exhibit sufficient structural flexibility to allow binding of the same partner domain (OSID of Opi1; summarized in Fig. 4).

Fig. 4
figure 4

Summary of regulatory interactions effective under conditions of inositol/choline repression. AID activator interaction domain; bZIP basic leucine zipper; DBD DNA-binding domain; FFAT phenylalanines in acidic tract; HDAC histone deacetylase; HID HDAC interaction domain; ICRE inositol/choline-responsive element; OSID Sin3/Ssn6-Opi1 interaction domain; P1–P4 paired amphipathic helices; RID repressor interaction domain; TAD transcriptional activation domain; TPR tetratricopeptide repeat; WD40 tryptophan/aspartate-containing repeat. Protein–protein interactions are depicted by arrows

Full size image

We finally performed a mutational analysis of the Opi1 AID, required to silence Ino2. We have previously characterized mutant alleles altered at the position of charged amino acids (double mutants K337 E338 and R372 E373; Heyken et al. 2005) which are strongly conserved among Opi1 proteins from various yeasts. Although these Opi1 variants showed 3–4-fold increase of ICRE-dependent gene expression under repressing conditions, a significant degree of regulation remained. In this work, we describe the regulatory phenotype of variants altered at the position of a sequence motif with similarity to a leucine zipper (residues 333–361: L-N6-L-N2-V-N3-M-N2-V-N3-L-N6-V). Among seven mutant proteins each containing a single alanine mutation at the positions indicated six turned out as completely deregulated. This result is in agreement with the phenotype of an opi1 allele encoding a V343Q mutation (rum1-25; Kaadige and Lopes 2006), isolated by a strategy selecting for loss of Opi1 function. We thus conclude that the hydrophobic-amphipathic motif is indeed the functional core of Opi1 AID which binds to the repressor interaction domain (RID) of Ino2, counteracting target gene activation. Interactions among regulators specific for phospholipid biosynthesis and pleiotropic factors as characterized in this and previous work are summarized in Fig. 4.