Introduction

A question of central importance for studies of eukaryotic gene expression is to understand how an extra- or intracellular signal is ultimately executed to increase or diminish transcriptional initiation; in other words: How are the regulators regulated? (Falvey and Schibler 1991). The differential expression of yeast structural genes involved in phospholipid biosynthesis by the availability of metabolic precursor molecules is an attractive model system for investigating the transduction of metabolic signals (reviewed by Chen et al. 2007). Genes such as INO1 (biosynthesis of inositol; Lopes and Henry 1991), CHO1 (choline; Bailis et al. 1992), FAS1 and FAS2 (fatty acids; Schüller et al. 1992a) are coordinately activated by ICRE promoter sequences (inositol/choline-responsive element = UASINO; consensus sequence: WYTTCAYRTG; a comprehensive compilation of motifs is given in Hoppen et al. 2005). ICRE motifs are bound by a heterodimer of positive regulators Ino2 and Ino4 (Schüller et al. 1992b; Ambroziak and Henry 1994; Schwank et al. 1995), both containing a basic helix–loop–helix (bHLH) structural motif (Hoshizaki et al. 1990; Nikoloff et al. 1992) which are necessary and sufficient for dimer formation and specific interaction with the ICRE (Dietz et al. 2003). Recent studies have shown that expression of several genes probably unrelated to phospholipid metabolism is also affected by Ino2 and Ino4 (Hoppen et al. 2005; Jesch et al. 2005; Chen and Lopes 2007). Importantly, Ino2 (but not Ino4) contains two separate transcriptional activation domains TAD1 and TAD2 (Schwank et al. 1995) as well as a repressor interaction domain (RID) which is required for functional deactivation of Ino2 by the repressor Opi1 (Wagner et al. 2001; Heyken et al. 2005). Opi1 is necessary for repression of ICRE-dependent transcription when inositol and choline (IC) are present in excess (Greenberg et al. 1982; Wagner et al. 1999). Under these conditions, Opi1 is localized in the nucleus (Loewen et al. 2003) and prevents Ino2 from activation of target genes by recruitment of the pleiotropic corepressor Sin3 (Wagner et al. 2001). Consequently, Ino2 variants defective for interaction with Opi1 mimic the opi1 mutant phenotype and allow constitutive transcription of ICRE-dependent genes (Heyken et al. 2005). Similarly, depletion of IC (derepressing conditions) causes accumulation of phosphatidic acid (PA) which binds to Opi1, leading to its retention at the endoplasmic reticulum or nuclear membrane (Loewen et al. 2003, 2004). Although Opi1 contains a basic leucine zipper motif (Hoshizaki et al. 1990), no evidence for DNA-binding by Opi1 has been obtained yet. In the absence of Opi1, ICRE-bound Ino2 can contact coactivator complexes (Snf1 with its histone kinase function, Lo et al. 2005; SAGA, TFIIB, Dietz et al. 2003) and activates transcription of target genes.

Although previous work has shown that the balance of Opi1 and Ino2 is critical for activation and repression (Hosaka et al. 1994; Schwank et al. 1997; Wagner et al. 1999), the mechanism of Ino2 regulation remained unknown. Eukaryotic activators may be controlled by several single or combined mechanisms: (1) biosynthetic control of the activator itself (e. g. Gal4, Gcn4; Griggs and Johnston 1991; Hinnebusch 1984); (2) control of intracellular localization (e. g. Pho4, Swi5; O’Neill et al. 1996; Moll et al. 1991); (3) regulation of DNA binding of the activator (e. g. Ace1, MyoD; Fürst and Hamer 1989; Benezra et al. 1990); (4) regulation of transcriptional activation (e. g. Gal4, Cat8; Ma and Ptashne 1987; Rahner et al. 1999). We thus wished to uncouple various mechanisms affecting transcriptional regulators Ino2, Ino4 and Opi1 and to separately evaluate their contribution to differential expression of ICRE-containing structural genes. It turned out that autoregulation of INO2, INO4 and OPI1 (mediated by ICRE upstream motifs; Schüller et al. 1992b; Ashburner and Lopes 1995; Wagner et al. 1999) contributes to the regulation of target genes but is not essential. We show that dimerization of Ino2 and Ino4 is regulated by IC, requiring a functional RID (but not an intact DNA-binding domain) within Ino2. We conclude that the integrity of the Ino2–Ino4 heterodimer must be considered as an important trigger of IC-mediated transcriptional regulation.

Materials and methods

Yeast strains, media and growth conditions

All strains of S. cerevisiae used in this work (compiled in Table 1) are isogenic to strain JS91.15–23 (Schwank et al. 1995). Variants of INO2, INO4 and OPI1 were introduced into strains JKY5 and JKY53 by integrative transformation. Synthetic complete (SC) media used for selective growth of transformants and conditions of gene regulation by IC have been described (Schüller et al. 1992a).

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

Plasmid constructions

Plasmids constructed and used for this study are listed in Table 2. Promoter variants of INO2, INO4 and OPI1 were inserted into integrative plasmids pRS403 (HIS3; Sikorski and Hieter 1989), YIplac204 (TRP1; Gietz and Sugino 1988) and YIplac128 (LEU2), respectively. MET25 promoter fusions were obtained by insertion of reading frame cassettes amplified by PCR into p413-MET25 (Mumberg et al. 1994) and subsequent transfer of the expression cassette into integrative vectors. To modify promoter sequences (CANNTG core sequences within ICRE motifs upstream of INO2, INO4 and OPI1 were converted into unrelated cleavage sites of restriction enzymes) or specific residues in the coding regions (ino2 E245A, ino4 E54A and ino4 L94A), the QuikChange site-directed mutagenesis kit from Strategene was used. The INO2 L118A mutation has been described previously (Heyken et al. 2005).

Table 2 Plasmids used in this work
Full size table

The artificial (lexAOp)8-GAL1 upstream region containing eight copies of the lexA operator was amplified from strain L40 (Hollenberg et al. 1995) and fused with lacZ of plasmid YEp356R (Myers et al. 1986). Reporter plasmid pJK119 was obtained by transfer of the (lexAOp)8–GAL1–lacZ fusion to integrative vector YIp352 (Hill et al. 1986). To obtain fusions of the DNA-binding domain of lexA repressor (residues 1–87) with INO2 and INO4, plasmid pSH2-1 (Hanes and Brent 1989) was used as a template for PCR amplifications. Similarly, GFPUV cassettes fused with regulatory genes were amplified from plasmid pBAD–GFPUV (Clontech). Epitope-tagged regulators were constructed by insertion of reading frame cassettes for OPI1, INO2 and INO4 into HA fusion plasmid p426-MET25HA (Mumberg et al. 1994) to give pCW116 (MET25–HA3–OPI1), pMD2 (MET25–HA3–INO2) and pJK51 (MET25–HA3–lexA DBD–INO4), respectively.

Fluorescence microscopy of GFP fusions

Fluorescence microscopy followed the procedure described by Taheri et al. (2000). Yeast strains containing integrated GFPUV fusions (Crameri et al. 1996) with INO2, INO4 or OPI1 were grown to log phase in selective medium (SCD-His) with 1 mM methionine under conditions of IC repression or derepression. Cells from 1 ml of the cultures were concentrated by centrifugation and immediately viewed on a Zeiss Axiovert microscope by either differential interference contrast microscopy (DIC, Nomarski optics) or fluorescence microscopy using a GFP filter set (AHF Analysentechnik AG; Tübingen, Germany). Images were taken using a Xillix Microimager digital camera and the Improvision Openlab software (Improvision; Coventry, UK).

Miscellaneous procedures

Transformation of S. cerevisiae strains, immunoblot analysis, PCR amplification and β-galactosidase assays were performed as previously described (Schwank et al. 1995; Wagner et al. 2001). Enzyme assays were repeated three times, using five independent transformants each. Standard deviations observed with integrated reporter genes ICRE–CYC1–lacZ and (lexAOp)8-GAL1–lacZ were 20% of the mean value or below.

Results

Influence of autoregulation of INO2, INO4 and OPI1 on expression of target genes

Regulatory genes INO2, INO4 and OPI1 each contain a single ICRE motif in their respective upstream regions and therefore control their own expression (Schüller et al. 1992b; Ashburner and Lopes 1995; Schwank et al. 1997; Wagner et al. 1999). Thus, the ratio of positive factors Ino2 + Ino4 and repressor Opi1 may vary under repressing and derepressing conditions, respectively. However, it remained unclear whether autoregulation of regulatory genes subsequently affects the expression pattern of ICRE-dependent target genes. We therefore wished to systematically compare the regulation of an ICRE-activated reporter gene in the presence of INO2, INO4 and OPI1 controlled by upstream regions devoid of their natural ICREs.

For each regulatory gene, three integrating plasmids were constructed: (1) a plasmid containing the entire wild-type control region (autoregulated); (2) a plasmid of identical length but mutated at the ICRE core element upstream of the respective gene (autoregulation eliminated); (3) a plasmid containing the coding region downstream of the heterologous MET25 promoter (cf. Fig. 1a). All regulatory genes controlled by promoter variants were confirmed as functional (not shown). We also constructed the triple mutant JKY5 (Δopi1 Δino2 Δino4, integrated ICRE–CYC1–lacZ reporter gene) and re-introduced regulatory genes OPI1, INO2 and INO4 driven by different control regions.

Fig. 1
figure 1

Variation of INO2, INO4 and OPI1 promoters and influence on the regulation of an ICRE-dependent reporter gene. a Comparison of promoter variants upstream of regulatory genes INO2, INO4 and OPI1. b INO2, INO4 and OPI1 promoter variants on plasmids containing selection markers HIS3, TRP1 and LEU2 were integrated at the respective loci of strain JKY5 (contains chromosomal null mutations Δino2::kanMX Δino4::kanMX Δopi1::kanMX and an integrated ICRE–CYC1–lacZ reporter gene). Transformants were cultivated under selective conditions in the presence of 1 mM methionine. Specific β-galactosidase activities are given in nmoles oNPG hydrolyzed per min per mg of protein (U/mg). R conditions of inositol/choline repression (200 μM inositol + 2 mM choline); D conditions of inositol/choline derepression (5 μM inositol + 5 μM choline); D/R ratio of enzyme activities under derepressing and repressing conditions, respectively; SD standard deviation of the mean value

Full size image

The resulting strains were cultivated under conditions of IC repression and derepression, respectively, and assayed for expression of the ICRE-dependent reporter gene. As shown in Fig. 1b, loss of autoregulation partially weakened regulation by IC, especially when INO2 variants with a mutated ICRE were used. Nevertheless, even when autoregulation was completely eliminated (ΔICRE triple mutant), target genes were clearly regulated by IC (derepression factor D/R = 6.1 compared with D/R = 14.0 when wild-type promoters controlled INO2, INO4 and OPI1). We conclude that autoregulation of regulators affects ICRE-dependent gene expression as a mechanism of signal amplification but is not responsible for triggering IC repression. To rule out other contributions of the natural control regions, we also investigated ICRE-mediated gene regulation with INO2, INO4 and OPI1 each driven by the MET25 promotor. This promoter is of moderate strength and can be further weakened by addition of methionine (Mumberg et al. 1994), allowing an expression level which is typical of regulatory genes. As is also apparant from Fig. 1b, expression and regulation of the ICRE-dependent reporter gene is almost identical when regulatory genes are activated by the MET25 promoter or by ICRE-mutated natural promoters (D/R = 6.5 and 6.1).

Influence of phospholipid precursors on intracellular localization of Ino2, Ino4 and Opi1

Several transcriptional regulators are controlled at the level of nuclear import/export, thereby allowing or preventing access to their DNA target sites. For Opi1, anchoring at the endoplasmic reticulum in the absence of IC has been previously shown (Loewen et al. 2003). For a systematic comparison of intracellular localization of Ino2, Ino4 and Opi1 under conditions of IC repression and derepression, we constructed bifunctional GFP fusion genes driven by the MET25 promoter which were subsequently integrated into the genome of the respective null mutants. To obtain fluorescent signals with weakly expressed fusion genes, the GFPUV variant of the green fluorescent protein was used (Crameri et al. 1996). The resulting strains MDY50 (GFP-INO2), MDY49 (INO4-GFP) and MDY54 (OPI1-GFP) were able to grow in the absence of inositol (not shown) and showed regulated expression of an ICRE-dependent reporter gene with a weakened MET25 promoter (cf. Fig. 2a). Thus, fluorescence of GFP should reliably correspond with the localization of the respective regulator under both conditions of inositol supply.

Fig. 2
figure 2

a GFP fusion proteins with full-length Ino2, Ino4 and Opi1, respectively, allow regulated expression of an ICRE-dependent reporter gene. GFP fusion genes on integrating HIS3 plasmids (pMD122: INO4-GFPUV, pMD155: GFPUV-INO2, pMD160: OPI1-GFPUV) were expressed under control of the MET25 promoter. Plasmids were transformed into derivatives of strain SIRP3 (integrated ICRE–CYC1–lacZ reporter gene), deleted for the chromosomal copy of the regulatory gene fused to GFP. To avoid overproduction of regulatory proteins, transformants were grown in the presence of 1 mM methionine under repressing (200 μM inositol + 2 mM choline) and derepressing conditions (5 μM inositol + 5 μM choline), respectively. Specific β-galactosidase activities are given in nmoles oNPG hydrolyzed per min per mg of protein (U/mg). M25Pr, MET25 promoter; SD standard deviation of the mean value. b Subcellular localization of regulatory proteins Ino2, Ino4 and Opi1 under conditions of inositol/choline repression and derepression. Representative cells of the strains indicated expressing fusion proteins GFP-Ino2, Ino4-GFP or Opi1-GFP were viewed by differential interference contrast microscopy (DIC, Nomarski optics) or by fluorescence microscopy (GFP)

Full size image

As shown in Fig. 2b, nuclear fluorescence could be observed with the GFP–Ino2 fusion protein under repressing and derepressing conditions (strain MDY50). Interestingly, a uniform fluorescence signal in the cytoplasm was observed in the absence of INO4 (MDY53). Similar to Ino2, Ino4-GFP constitutively localized to the nucleus (MDY49), even in the absence of INO2 (not shown). These results indicate that Ino4 is required for nuclear import of Ino2 while, in contrast, Ino2 does not affect intracellular localization of Ino4. Opi1-GFP fluorescence is clearly concentrated in the nucleus under repressing conditions while the signal is much more distributed under derepressing conditions. This agrees with previous results, arguing for a retention of the Opi1 repressor at the endoplasmic reticulum or the nuclear membrane when maximal ICRE-dependent gene expression is required (Loewen et al. 2003).

Influence of phospholipid precursors on transcriptional activation by Ino2

Even when activator proteins constitutively localize to the nucleus, binding to their DNA target sites or transcriptional activation may be regulated. We thus fused INO2 to the DNA-binding domain of bacterial lexA repressor and assayed Ino2-dependent gene activation of a lexAOp-containing reporter gene. These tests were performed with three strains (wild-type, Δino4, Δopi1), each cultivated under repressing and derepressing conditions, respectively. Nuclear import of the resulting lexADBD–Ino2 fusion protein in the absence of Ino4 should be guaranteed by an intrinsic nuclear localization sequence within lexA (Rhee et al. 2000).

With constitutive DNA-binding mediated by lexA, Ino2-dependent activation was not influenced by IC supply (cf. Fig. 3). Although TAD2 of Ino2 partially overlaps with the Opi1 repressor interaction domain (RID; Heyken et al. 2005), no increase of transcriptional activation was observed with an isogenic opi1 null mutant. Similarly, absence of Ino4 did not alter Ino2-dependent activation. We conclude that activation mediated by Ino2 occurs constitutively, at least in the context of the full-length protein as analyzed here.

Fig. 3
figure 3

Influence of phospholipid precursors on transcriptional activation. Isogenic strains containing an integrated lexAop–GAL1–lacZ reporter gene were transformed with plasmid pJS519, containing a lexA DBD–INO2 fusion gene (INO2 full length). Specific β-galactosidase activities are given in nmoles oNPG hydrolyzed per min per mg of protein (U/mg). R conditions of inositol/choline repression (200 μM inositol + 2 mM choline); D conditions of inositol/choline derepression (5 μM inositol + 5 μM choline); D/R ratio of enzyme activities under derepressing and repressing conditions, respectively. bHLH basic helix–loop–helix; DBD DNA-binding domain; SD standard deviation of the mean value; TAD transcriptional activation domain

Full size image

Influence of phospholipid precursors on dimerization of Ino2 and Ino4

For basic helix–loop–helix proteins, dimerization is a prerequisite for DNA binding (Littlewood and Evan 1998). Since Ino2 must be considered as a constitutive activator, we wished to investigate whether dimerization of Ino2 and Ino4 is affected by the availability of phospholipid precursors. We used a simplified two-hybrid system with a lexAOp-dependent reporter gene and a lexA–Ino4 fusion protein. Transcriptional activation of the reporter gene is entirely mediated by full-length Ino2 which is recruited to promoter-bound lexA-Ino4 via its HLH domain.

To rule out a biosynthetic variation of regulators under our experimental conditions, MET25–HA fusions of INO2, lexA–INO4 and OPI1 were transformed into the corresponding single mutants. After cultivation of transformants under repressing and derepressing conditions, protein extracts were analyzed by immunoblot analysis. As is apparent from Fig. 4, biosynthesis of regulatory proteins Ino2, lexA–Ino4 and Opi1 was indeed unaffected by the variation of phospholipid precursors.

Fig. 4
figure 4

Constitutive biosynthesis of Ino2, lexA–Ino4 and Opi1 regulators expressed under MET25 promoter control. Plasmids pMD2 (MET25–HA3–INO2), pJK51 (MET25–HA3–lexA DBD–INO4) and pCW116 (MET25–HA3–OPI1) were transformed into mutants SS92.3-1 (Δino2), JS91.15-9 (Δino4) and MBY1 (Δopi1), respectively. Yeast transformants were grown in selective media under conditions of inositol/choline repression (R 200 μM inositol + 2 mM choline) and derepression (D 5 μM inositol + 5 μM choline), respectively. Fifty micrograms of total yeast protein was separated in each lane and immunoblot analyses were performed using anti-HA monoclonal antibody. Two antigenic signals are usually obtained for Ino2 (Ambroziak and Henry 1994)

Full size image

For a direct comparison of fusion constructs, we used the authentic ICRE-dependent reporter gene (requiring DNA-binding by Ino2 and Ino4) as well as the heterologous lexAOp-dependent reporter gene (requiring formation of the Ino2/Ino4 heterodimer). To avoid deregulation as a result of unphysiological levels of Ino2 and Opi1 (Schwank et al. 1997; Wagner et al. 1999), all regulatory genes were controlled by the MET25 promoter, weakened by supplementation of media with 1 mM methionine.

As expected, activation of either reporter gene failed in the absence of INO2 or INO4 while regulated activation was found with authentic INO2, INO4 and OPI1. Although concentration of these regulators was no longer influenced by phospholipid precursors, the ICRE-dependent reporter gene was still regulated 6,5-fold (Fig. 5a). An almost identical regulatory pattern was observed with a lexA–Ino4 fusion (6,1-fold derepression). Importantly, a considerable degree of derepression was also found with the lexAOp-driven reporter gene (4,0-fold increase; Fig. 5b), indicating that regulation could be transferred to a heterologous cis-acting element. Next, we replaced functional genes INO2 and INO4 by variants with a detrimental mutation in the basic region of the DNA-binding bHLH-domain (mutant alleles ino2E245A and ino4E54A; R. Rohde et al. unpublished). Indeed, activation of the ICRE-dependent reporter gene was completely abolished (Fig. 5a) while the lexAOp-dependent reporter gene showed an even stronger expression compared to wild-type INO regulators. This increase can be explained by the failure of mutant Ino2(E245A) and lexA-Ino4(E54A) to bind to chromosomal ICRE target sites, allowing a higher degree of occupation at the lexAOp. In contrast, wild-type lexA–Ino4 is a bi-functional protein which can bind to both ICRE (together with Ino2) and lexAOp. Importantly, activation with b-domain mutants of Ino2 and Ino4 was still fully regulated, indicating that DNA-binding of Ino2/Ino4 is not required for the IC-response. As a control, we also used mutant allele ino4L94A, encoding an Ino4 variant which is no longer able to interact with Ino2 (unpublished). Indeed, regulated activation was abolished with both reporter genes. Finally, we investigated the influence of a repression-resistant Ino2 variant which is no longer bound by the Opi1 repressor due to a replacement at a critical position of the RID (INO2 mutant allele L118A; Heyken et al. 2005). In the presence of this Ino2 variant, repression of either reporter gene by phospholipid precursors was no longer detectable, mimicing the effect of an opi1 null mutation. While permanent recruitment of lexA–Ino2 to a lexAOp-dependent promoter lead to constitutive gene activation (one-hybrid situation, Fig. 3, see above), regulated activation was observed with promoter-bound lexA–Ino4 and its dimerization partner Ino2 (simplified two-hybrid situation, Fig. 5b). From these results we conclude that dimerization of Ino2 and Ino4 is regulated by phospholipid precursors via Opi1-interaction with Ino2. Moreover, regulated dimerization is not influenced by the ability of Ino2 and Ino4 to bind DNA.

Fig. 5
figure 5

Regulation of ICRE- and lexAOp-dependent reporter genes by phospholipid precursors. Regulatory genes INO2, INO4 and OPI1 were controlled by the MET25 promoter. Variants of MET25-INO2, MET25-(lexA DBD)-INO4 and MET25-OPI1 were integrated at loci his3, trp1 and leu2, respectively. Transformants were uniformly cultivated under selective conditions in the presence of 1 mM methionine to achieve an adequate level of regulators. Interactions among promoter elements and regulatory proteins are depicted for each type of experiment. a Use of an integrated ICRE–CYC1–lacZ reporter gene (strain JKY5). b Use of an integrated lexAop–GAL1–lacZ reporter gene (strain JKY53). Specific β-galactosidase activities are given in nmoles oNPG hydrolyzed per min per mg of protein (U/mg). R conditions of inositol/choline repression (200 μM inositol + 2 mM choline); D conditions of inositol/choline derepression (5 μM inositol + 5 μM choline); D/R ratio of enzyme activities under derepressing and repressing conditions, respectively. bHLH basic helix–loop–helix; DBD DNA-binding domain; SD standard deviation of the mean value; TAD transcriptional activation domain

Full size image

Discussion

The analysis of signalling systems of transcriptional control in eukaryotes has shown that the observed regulatory phenotype often results from several partial mechanisms which may act additively, cooperatively and/or redundantly (e. g. glucose repression and galactose induction of GAL structural genes; Johnston et al. 1994; Bhat and Murthy 2001). In this work, we wished to systematically analyse whether biosynthetic variation, intracellular localization, transcriptional activation and DNA binding of regulators Ino2, Ino4 and Opi1 contribute to IC-dependent expression of target genes.

We and others have previously shown that INO2, INO4 and OPI1 genes each contain a functional ICRE upstream sequence, arguing for autoregulation of regulators (Schüller et al. 1992b; Ashburner and Lopes 1995; Wagner et al. 1999). It should be mentioned that the ICRE variants mediating IC-regulated expresson of INO2 (AATTCACATGT), INO4 (TATTCACATGT) and OPI1 (TCTTCATATGC) deviate from the optimal UAS sequence (WYTTCACATGS; Schüller et al. 1995). Nevertheless, Ino2 + Ino4 could efficiently bind to these motifs in vitro (Schwank et al. 1997; Wagner et al. 1999). It should be mentioned that autoregulation of INO4 has been questioned by others, suggesting a posttranscriptional mechanism of IC control of INO4 expression, instead (Robinson and Lopes 2000). However, using the MET25 promoter, we found constant levels of HA-tagged Ino4 (not shown) and lexA-Ino4 (this work) under repressing and derepressing conditions. We conclude that IC-mediated regulation of INO4 expression (about threefold) is a result of its specific upstream region and the ICRE therein. In this work, we replaced regulatory genes of phospholipid biosynthesis by variants mutated at their ICRE promoter motifs. The resulting variants were able to functionally complement their respective null mutations and led to regulation of an ICRE-dependent reporter gene by IC which was weakened but still effective (14-fold derepression vs. sixfold derepression observed in the presence of three promoter mutants). Loss of INO2 autoregulation (8–10-fold with its wild-type control region) showed the most significant contribution. Almost identical results were obtained with regulatory genes driven by methionine-repressed MET25 fusions. We conclude that target genes are differentially expressed upon variation of IC supply even with Ino2, Ino4 and Opi1 regulators at a constitutive steady-state level.

The functional compartmentation of a eukaryotic cell provides a regulatory means of transcriptional control so that many activators and repressors show signal-dependent regulation by nuclear import/export. A related mechanism has been shown for Opi1 which interacts with the Scs2 protein of the endoplasmic reticulum (ER) and/or nuclear membrane in the presence of phosphatidic acid (PA) via its FFAT structural motif (two phenylalanine residues in an acidic tract; Loewen et al. 2003, 2004). Our results with a bifunctional OPI1–GFP fusion are in agreement with these findings. While the signal observed with repressed cells concentrated in the nucleus, fluorescence under derepressing conditions was no longer exclusively nuclear. A cluster of basic amino acids (residues 109–138) extending the leucine zipper domain of Opi1 may be responsible for PA binding as well as for nuclear import. While Opi1 shows signal-dependent shuttling between ER and nucleus, Ino4-GFP was localized in the nucleus constitutively. Ino4 contains an extremely basic region (residues 29–40) which may function as a nuclear localization sequence (NLS). In contrast, the GFP–Ino2 fusion protein was nuclear only when INO4 was functional. A basic cluster of amino acids as a conventional NLS is absent from the Ino2 sequence. We thus assume that Ino4 heterodimerizes with Ino2 in the cytoplasm and may subsequently function as a mediator of nuclear import for the resulting complex. In conclusion, our data argue for a nuclear translocation of Ino2 which depends on Ino4 but is not affected by phospholipid precursors.

We have previously shown that the Opi1 repressor interaction domain within Ino2 partially overlaps transcriptional activation domain TAD2 (Heyken et al. 2005). This appears reminiscent of Gal80 binding to the Gal4 TAD prior to galactose induction (Ma and Ptashne 1987) and Dig1 inhibiting activation by Ste12 in the absence of a pheromone signal (Olson et al. 2000). However, transcriptional activation by a lexA–Ino2 fusion was not significantly influenced by IC availability or by deletion of OPI1 and INO4, respectively (cf. Fig. 3). Nevertheless, it should be mentioned that Gardenour et al. (2004) isolated INO2 missense mutants mapping to TAD1 (D20H, F21L) which led to partially increased target gene activation (threefold) under repressing conditions.

We finally investigated whether interaction between Ino2 and Ino4 is influenced by IC regulation. To do this, we fused Ino4 to the DNA-binding domain of bacterial lexA repressor and used a lexAOp-dependent reporter gene to monitor gene expression. Since Ino4 lacks a transcriptional activation domain by itself, activation of the reporter gene should occur only if Ino2 forms a heterodimer with the promoter-recruited lexA–Ino4. No biosynthetic variation of Ino2, lexA–Ino4 and Opi1 under repressing and derepressing conditions was found with our experimental set-up (MET25-driven regulatory genes). In a similar one-hybrid experiment, transcriptional activation by lexA–Ino2 turned out as constitutive (see above). Our experimental approach further requires that DNA-binding of lexA fusion proteins is not affected by IC regulation. Although it is known that fusion domains may affect DNA-binding by lexA (Golemis and Brent 1992), previous results from our group showed that DNA-binding at least by lexA–Ino4 was not regulated by IC (Schwank et al. 1995).

Importantly, lexA–Ino4 + Ino2 led to a fourfold regulation of the lexAOp-dependent reporter gene. With a reporter gene containing the ICRE as the authentic binding site of Ino2 + Ino4 and the same set of regulators, a sixfold derepression was observed (cf. Fig. 5a, b). Gene regulation by IC of the lexAOp-containing promoter was completely maintained with missense mutants of Ino2 and Ino4, each defective for DNA-binding due to alteration of an essential glutamate residue of both basic regions (E245 and E54). Thus, differential activation of the heterologous reporter gene does not require binding of Ino2 + Ino4 to the homologous cis-element, arguing for their regulated dimerization as a critical means of target gene control. Expectedly, disruption of the Ino2–Ino4 interaction by mutation of an essential residue within helix II (L94) completely abolished activation by either promoter motif. In contrast, deletion of OPI1 or mutation of L118 necessary for Opi1 binding to the Ino2 RID similarly caused constitutive gene expression with both reporter genes.

Since dimerization is a prerequisite for DNA-binding by bHLH proteins, occupancy of the ICRE would be also affected by regulated formation of the Ino2 + Ino4 heterodimer in a native promoter situation. As a conclusion from our findings we hypothesize that interaction of Opi1 with Ino2 RID may trigger a conformational change within Ino2 which interferes with Ino4 binding, finally leading to less occupation of genomic ICRE target sites and gene repression. In a previous study, Brickner and Walter (2004) investigated the influence of the unfolded protein response (UPR) on INO1 transcription. Using chromatin immunoprecipitation, these authors describe constitutive binding of Ino2 + Ino4 to the INO1 promoter. Instead, intranuclear relocation of the INO1 locus from the nucleoplasm to the nuclear periphery is postulated as a critical step of gene derepression. However, to our surprise, we find that the original data of Brickner and Walter clearly show a twofold increase of Ino2 binding to the INO1 promoter under derepressing conditions which is in partial agreement with our results. It should be also emphasized that both hypotheses assuming regulated heterodimerization of Ino2 and Ino4 (this work) and subsequent alteration of nuclear localization of INO1 (Brickner and Walter 2004) are compatible with each other. A model which summarizes regulatory transitions based on findings from this and previous work is shown in Fig. 6. In this model, we distinguish between an early phase of repression in which a chain of interactions (ICRE–Ino2–Opi1–Sin3–HDACs) leads to a specific local alteration of chromatin structure and a late phase (steady state conditions) with Opi1 weakening formation of the Ino2–Ino4 heterodimer.

Fig. 6
figure 6

Hypothesis on regulatory transitions upon establishment of IC repression. a Under conditions of IC limitation, accumulation of phosphatidic acid (PA) leads to retention of Opi1 at the ER and/or nuclear membrane via Scs2 (Loewen et al. 2003, 2004), allowing Ino2 + Ino4 to fully activate ICRE-driven target genes. b Opi1 is released from its membrane anchoring upon increase of IC concentration and subsequent consumption of PA. Interaction with the RID of Ino2 specifically targets Opi1 to ICRE-containing genes (Wagner et al. 2001; Heyken et al. 2005). Since Opi1 simultaneously contacts the pleiotropic corepressor Sin3, local modification of chromatin via histone deacetylases (HDACs) initiates the repressed state (Kadosh and Struhl 1997). c Opi1 also weakens interaction of Ino2 and Ino4 (this work), leading to release of the heterodimer from its ICRE target sites in late repression and under steady state conditions. Autoregulation of regulators as a putative means of signal amplification is not considered here. DBD DNA-binding domain; NLS nuclear localization sequence; RID repressor interaction domain; TAD transcriptional activation domain

Full size image