The basic helix–loop–helix (bHLH) proteins form a large family of transcriptional regulators. This superfamily is found in all eukaryotes, from the yeast Saccharomyces cerevisiae to mammals (Amati and Land, 1994; Atchley and Fitch, 1997; Bailey et al., 2003; Grove et al., 2009; Heim et al., 2003; Jones, 2004; Ledent et al., 2002; Ledent and Vervoort, 2001; Massari and Murre, 2000; Robinson and Lopes, 2000a; Toledo-Ortiz et al., 2003). These proteins play critical roles in diverse biological processes, including neurogenesis, myogenesis and haematopoiesis (Berkes and Tapscott, 2005; Porcher et al., 1996; Ross et al., 2003). A hallmark of bHLH proteins is the ability to form homo- or heterodimers via two amphipathic α-helices. Dimerization is a prerequisite for DNA-binding via two juxtaposed basic-charged helical domains (Ferre-D'Amare et al., 1993; Ma et al., 1994; Shimizu et al., 1997). The ability to form multiple dimers has long been recognized as a mechanism to coordinate different transcription programmes (Hooker and Hurlin, 2006; Luscher, 2001; Nair and Burley, 2006; Rottmann and Luscher, 2006). However, it is clear that several other mechanisms play important roles in bHLH-mediated regulation of gene expression (Chen and Lopes, 2007). S. cerevisiae has nine bHLH proteins, most of which, have been extensively studied with respect to regulation of single metabolic pathways. This study examines the extent of bHLH-mediated cross-pathway regulation using the CIT2 bHLH target gene as a reporter.
The S. cerevisiae bHLH proteins regulate several important metabolic pathways, including phospholipid biosynthesis, phosphate utilization and glycolysis (Robinson and Lopes, 2000a). The first identified yeast bHLH protein, Pho4p, forms a homodimer required for the induction of phosphate utilization genes when phosphate levels are low (Berben et al., 1990; Johnston and Carlson, 1992; Oshima, 1997). The Ino2p : Ino4p heterodimer induces phospholipid biosynthetic gene expression in response to inositol deprivation (Carman and Han, 2008; Carman and Henry, 2007; Chen et al., 2007). The Sgc1p homodimer activates expression of glycolytic genes (e.g. ENO1) (Nishi et al., 1995; Sato et al., 1999) and is also required for Ty1-mediated gene expression (Lohning and Ciriacy, 1994). Cbf1p binds to several MET gene promoters to regulate their transcription (Ferreiro et al., 2004; Kent et al., 2004; Kuras et al., 1997; Kuras and Thomas, 1995). Cbf1p also functions in chromosome segregation (Cai and Davis, 1990; Foreman and Davis, 1993) and DNA repair (Ferreiro et al., 2004). The two least-studied yeast bHLH proteins are Hms1p, which functions in pseudohyphal growth (Lorenz and Heitman, 1998), and Ygr290wp. Ygr290wp lacks a basic-charged DNA-binding region. While it is classified as a dubious ORF, a recent study revealed that ENO1 expression is altered in an Ygr290w deletion mutant (Chen and Lopes, 2007). The Rtg1p–Rtg3p heterodimer functions to activate nuclear genes (e.g. CIT2) in response to mitochondrial dysfunction (Jia et al., 1997; Liao and Butow, 1993). This response is referred to as retrograde regulation (Liu and Butow, 2006).
At least five different properties of yeast bHLH proteins contribute to their ability to coordinate gene expression (Chen and Lopes, 2007). First, yeast bHLH proteins are able to form homodimers as well as heterodimers with several different partners. For example, Ino4p functions as a hub by interacting with the other eight bHLH proteins (Robinson et al., 2000; Lepeak and Lopes, unpublished results). Second, most yeast bHLH proteins bind with limited specificity to the well-characterized E-box, 5′-CANNTG-3′ (Robinson and Lopes, 2000a), although the Rtg1p–Rtg3p dimer binds an alternative conserved sequence (R-box: 5′-GTCAC-3′) (Jia et al., 1997). It is well established that binding to the E-box requires interactions between conserved amino residues in the basic DNA-binding domain and invariant nucleotides of the E-box (Blackwell and Weintraub, 1990; Burley et al., 1993; Cave et al., 2000; Ferre-D'Amare et al., 1993,1994; Fisher and Goding, 1992; Hakoshima et al., 1993; Shimizu et al., 1997). Other residues in the basic domain interact with E-box central variant nucleotides or nucleotides flanking the core sequence to dictate binding specificity (Cave et al., 2000; Fisher and Goding, 1992; Hakoshima et al., 1993; Shimizu et al., 1997). For example, while Pho4p and Cbf1p homodimers both bind a 5′-CACGTG-3′ sequence, a flanking ‘T’ nucleotide inhibits Pho4p binding (Fisher and Goding, 1992). Third, some bHLH proteins autoregulate their expression. The Ino2p–Ino4p heterodimer is required for derepression of INO2 reporters during inositol deprivation, whereas Ino4p is required for expression of INO4 reporters (Ashburner and Lopes, 1995a; Robinson and Lopes, 2000b; Schüller et al., 1992; Schwank et al., 1997). Fourth, some HLH proteins lack the basic region and therefore function as inhibitors by preventing DNA-binding. The mammalian Id protein is an example of this subclass (Jen et al., 1992; Lin et al., 2000; Ross et al., 2003). While an HLH protein has not been documented in yeast, Ygr290w could be a yeast representative of this subclass. Lastly, some bHLH proteins are also regulated by cytoplasmic-nuclear shuffling. Rtg3p : Rtg1p and Pho4p are examples of yeast bHLH proteins that translocate to the nucleus under inducing conditions (Kaffman et al., 1998a, 1998b; Komeili and O'Shea, 1999; O'Neill et al., 1996; Sekito et al., 2000).
This study focused on the CIT2 gene, which is a well-studied target of the retrograde response. The CIT2 gene encodes citrate synthase, which functions in the TCA and glyoxylate cycles (Kim et al., 1986). The retrograde signalling cascade adjusts carbohydrate and nitrogen metabolism in response to a change in the functional state of the mitochondria (Butow and Avadhani, 2004; Liu and Butow, 2006). The Rtg1p–Rtg3p bHLH dimer binds directly to two R-boxes (5′-GTCAC-3′) in the CIT2 promoter (Jia et al., 1997). The Rtg1p and Rtg3p activities are regulated by phosphorylation and nuclear translocation (Sekito et al., 2000). In wild-type mitochondrial DNA-competent (ρ+) cells, the hyperphosphorylated Rtg3p is retained in the cytoplasm by Rtg1p. However, in mitochondrial DNA-deficient (ρ0) strains, Rtg3p is dephosphorylated and translocates to the nucleus, where it dimerizes with Rtg3p. Rtg3p contains two activation domains that activate genes such as CIT2 (Chelstowska et al., 1999; Rothermel et al., 1997). Superimposed on this, transcription of the RTG3 gene is also upregulated two-fold in a ρ0 strain (Jia et al., 1997).
While the Rtg1p–Rtg3p regulation of CIT2 is well documented, there is also evidence of cross-regulation with other bHLH regulated pathways. Rtg1p and Rtg3p have both been shown to interact with Ino4p (Robinson and Lopes, 2000a). Expression profiling identified two PHO genes as targets of retrograde regulation (Epstein et al., 2001). Moreover, ENO1 (a target of the Sgc1p homodimer) is also regulated by Rtg1p and Rtg3p (Chen and Lopes, 2007). Yeast genome-wide ChIP–chip studies do not identify these regulatory circuits because they do not use growth conditions that affect bHLH protein activities (Lee et al., 2002; Ren et al., 2000). Here, we found that CIT2 expression was affected by additional bHLH proteins in both ρ+ and ρ0 strains, including Ino2p, Ino4p, Pho4p, Hms1p and Sgc1p. We observed that CIT2 expression in ρ0 strains is also induced by the presence of inositol. We also found that, in addition to CIT2, RTG3-cat expression is regulated by Ino2p and Ino4p.
Materials and methods
Strains, media and growth conditions
Plasmid-containing Escherichia coli DH5α cells (Invitrogen, Carlsbad, CA, USA) were grown in LB—Amp medium (10% w/v Bacto-tryptone, 5% w/v yeast extract, 10% w/v NaCl, 50 µg/ml ampicillin) at 37 °C.
The S. cerevisiae strains used in this study are listed in Table 1. All strains listed exist as ρ+ and ρ0 derivatives of BY4742, which is based on yeast strain S288C (Baudin et al., 1993; Brachmann et al., 1998) and isogenic strains containing ino2Δ, ino4Δ, pho4Δ, cbf1Δ, sgc1Δ, rtg1Δ, rtg3Δ, hms1Δ and ygr290wΔ (Giaever et al., 2002; Winzeler et al., 1999). The ρ0 derivatives were obtained by several passages of the ρ+ strains through YEPD medium (2% yeast extract, 1% peptone, 2% glucose) containing 15–20 mg/ml ethidium bromide (Goldring et al., 1970), then checked for the presence of mitochondrial DNA by staining with 4′,6′-diamino-2-phenylindole (DAPI).
Yeast cultures were grown at 30 °C in a complete synthetic medium lacking inositol, choline, KH2PO4 and uracil (for reporter plasmid selection) (Chen and Lopes, 2007; Kelly and Greenberg, 1990; Toh-E et al., 1973). Where indicated, 75 µM inositol (I+) and/or 1 mM choline (C+) were added. Pi low medium contained 0.22 mM KH2PO4 and 20 mM KCl, and Pi high medium contained 11 mM KH2PO4. In the case of the ino2Δ and ino4Δ mutants, I−C− medium contained 10 µM inositol, which is required for the growth of these inositol auxotrophs but still yields partially derepressed expression (Ashburner and Lopes, 1995a; Eiznhamer et al., 2001; Robinson and Lopes, 2000b). To screen for RTG3–cat integrants at the GAL4 locus, strains were grown on complete synthetic-Gal medium (contained inositol, choline, Pi high, 2% galactose instead of glucose, but lacked uracil).
Yeast cells were grown to early log phase and lysed using zymolyase (Zymo Research, Orange, CA, USA). The spheroplasts were fixed in 70% ethanol at room temperature for 30 min, and stained with 1 µg/ml DAPI for 5 min. The cells were viewed with a Zeiss Fluorescence Axioscope (Carl Zeiss, NY) and a × 100 oil-immersion objective. Images were captured with a RETIGA 2000 R camera (Carl Zeiss, NY, USA).
Plasmid YEp357R–CIT2 contained 1000 bp of the sequences upstream of the CIT2 ORF and the first codon fused in-frame to the lacZ reporter gene in YEp357R, a multicopy episomal plasmid with a URA3-selectable marker (Table 2) (Myers et al., 1986). The 1000 bp upstream of the CIT2 gene contains two previously characterized R-boxes required for retrograde regulation (Jia et al., 1997). This fusion plasmid was constructed by first amplifying 1000 bp of the CIT2 promoter from S. cerevisiae genomic DNA (Invitrogen, Carlsbad, CA, USA), using primers CIT2 F and CIT2 R (Table 3). The 1000 bp PCR product was initially cloned into pGEM®-T EASY (Promega, Madison, WI; Table 2) and confirmed by sequencing. The 1000 bp CIT2 promoter fragment was excised from pGEM-T–CIT2 and inserted into YEp357R using HindIII and EcoRI to create YEp357R–CIT2 (Table 2).
pBM2015 containing 500 bp of the RTG3 promoter fused to the cat reporter
Table 3. Oligonucleotides used in this study
Plasmids that complement the rtg1Δ and rtg3Δ mutant alleles were constructed by cloning each ORF and promoter sequences into pRS315 (Table 2). Plasmids pRS315–RTG1 and pRS315–RTG3 were constructed by excising 1.03 kb SalI–BamHI and 1.9 kb SalI–BamHI fragments from previously constructed YCp50–RTG1 and YCp50–RTG3 (Koepke and Lopes, unpublished) and inserting them into the same sites in pRS315, respectively. Plasmids pRS315–INO2 and pRS315–INO4 were constructed previously (Table 2) (Chen and Lopes, 2007).
A fusion of 500 bp of the RTG3 promoter and the cat reporter gene was created by PCR amplification, using the RTG3-catF′ and RTG3-catR′ primers (Table 3). The 500 bp PCR product, flanked by BamHI and BglII sites, was inserted into pGEM-T–EASY to create pGEM-T–RTG3 (Table 2). The RTG3 promoter insert was excised from pGEM-T–RTG3 by digestion with BamHI and BglII and inserted into the BamHI site of pBM2015 (Griggs and Johnston, 1993) to create pBM–RTG3 (Table 2). Yeast strains containing the RTG3 promoter–cat fusion integrated at the GAL4 locus were created by selecting for Ura+ transformants, as previously described (Ashburner and Lopes, 1995a). Typically, integrants were screened for by slow growth on complete synthetic-Gal medium, resulting from integration at the GAL4 site. However, this screen did not work for the ρ0 derivative strains, so the integrants were confirmed by PCR using a primer in the middle of the RTG3 promoter and another primer in the genome region outside the GAL4 locus (GAL4 M and GAL4 F, respectively)(Table 3).
A single potential E-box and three R-boxes in the CIT2 promoter (positioned at − 809, − 804, − 346 and − 312, respectively) were mutagenized using the QuikChange® XL-Site-directed Mutagenesis Kit, the pGEM-T–CIT2 plasmid described above, and the primer pairs named for each of the four positions (Table 3). The elements at − 809, − 804, − 346 and − 312 were mutated to SmaI, SphI, StuI and XhoI sites, respectively. Note that the R-boxes at − 346 and − 312 had previously been shown to be active and bind the Rtg1p–Rtg3p heterodimer (Jia et al., 1997). The mutant CIT2 promoters were cloned into YEp357R as described above (Table 2).
Reporter enzyme assays
To assay β-galactosidase activity, yeast strains were grown in 5 ml of appropriate medium to mid-log phase (OD600 = 0.6–0.8) and pelleted by centrifugation at 5000 × g for 10 min. The cell pellet was suspended in 200 µl β-gal assay buffer (20% glycerol, 0.1 M Tris–HCl, pH 8.0, 1 mM DTT, and 1 mM PMSF), transferred to a sterile 1.5 ml microcentrifuge tube and stored at − 80 °C overnight. The cells were thawed on ice and 100 µl glass beads (0.45 mm diameter) was added. The cells were lysed by vortexing six times for 15 s. Cellular debris was removed by centrifugation at 14 000 rpm for 15 min at 4 °C. The supernatant was transferred to another 1.5 ml microcentrifuge tube. To assay β-galactosidase activity, reactions were set up with 20 µl yeast extract and 80 µl β-gal assay buffer and incubated for 5 min at 28 °C. The reaction was initiated by the addition of 40 µl ONPG (4 mg/ml). The absorbance of the reaction was measured at OD420nm at 12 s intervals for a total of 30 min. Protein concentration was determined by the BioRad Protein Assay (Bio-Rad, Rockville Center, NY). Both the β-galactosidase activity reactions and the protein concentration reactions were quantified using the SOFTmax® PRO Software and a VERSAmax Tunable Microplate Reader (Molecular Devices, Sunnyvale, CA, USA). Units of β-galactosidase activity = (A420 nm/min/mg total protein) × 1000.
Chloramphenicol acetyltransferase (Cat) assays were performed as described previously (Ashburner and Lopes, 1995a). Units of Cat activity were defined as percentage conversion of [14C]-chloramphenicol/total amount of protein (mg)/h.
Real-time quantitative PCR
Transcript abundance was quantified by real-time quantitative PCR (Q-PCR) from wild-type strain BY4742 grown in appropriate medium. Total RNA was isolated from yeast using glass-bead disruption and hot phenol extraction (Elion and Warner, 1984) and purified using an RNeasy RNA Extraction Kit (Qiagen, Valencia, CA, USA). First-strand cDNA was synthesized using Superscript II reverse transcriptase (Invitrogen) as previously described (Jani and Lopes, 2008). The resulting cDNA was analysed using real-time Q-PCR, using a Mx3000P Q-PCR thermocycler and MxPro Q-PCR software (Stratagene, La Jolla, CA, USA). Primers spanning the ORFs of CIT2 and ACT1 (denoted by the ORF designation in Table 3) were designed using Primer 3 software (http://frodo.wi.mit.edu/primer3/input.htm). Amplification reactions were carried out with Brilliant SYBR Green Master Mix (Stratagene), as previously described (Jani and Lopes, 2008).
CIT2 gene expression is induced by inositol and regulated by INO2, INO4, PHO4, HMS1 and SGC1 in addition to RTG1 and RTG3
A recent study demonstrated that expression of the ENO1 gene, which was known to be regulated by the Sgc1p bHLH protein, is also regulated by several other bHLH proteins and growth conditions that affect their activity (Chen and Lopes, 2007). The increase in CIT2 gene expression in response to dysfunctional mitochondria has been a model for understanding retrograde regulation via Rtg1p and Rtg3p (Liu and Butow, 2006). We sought to determine whether other bHLH proteins and growth conditions that affect their function also regulated CIT2 expression. We tested the presence (ρ+) and absence (ρ0) of mitochondrial DNA, presence and absence of inositol and choline (I+C+ and I−C−, respectively) and high (Pi high) and low (Pi low) phosphate. In order to carry out these studies, it was necessary to first create ρ0 versions of the strains to be tested. This was done by growing ρ+ strains in the presence of ethidium bromide (Goldring et al., 1970) (Table 1). The absence of mitochondrial DNA was assessed by DAPI staining (Figure 1).
Both ρ+ and ρ0 strains were transformed with a CIT2–lacZ reporter (Table 2) and β-galactosidase activity was assayed after growth in the four conditions described above. As expected, CIT2–lacZ expression was induced 5- to 10-fold in the ρ0 wild-type strain under the four growth conditions tested (cf. Figure 2A to 2B; note difference in y axes scales) (Jia et al., 1997). Also as expected, the RTG1 and RTG3 genes were required for the expression of CIT2–lacZ under all four conditions in both ρ0 and ρ+ strains (Figure 2) (Jia et al., 1997; Liao and Butow, 1993).
The results also show that most of the bHLH knock-out strains generally yielded elevated CIT2–lacZ expression in the ρ+ strains, with the pho4Δ mutant having no effect (Figure 2A). However, in the ρ0 strains, CIT2–lacZ expression yielded several interesting effects. First, CIT2–lacZ expression in the wild-type strain was increased in I+C+ compared with I−C− conditions, regardless of the phosphate concentration (Figure 2B). This induction was not evident in the ρ+ wild-type strain (Figure 2A). This inositol/choline-mediated induction was the reverse of that which occurs with the phospholipid biosynthetic genes via Ino2p and Ino4p (Jesch et al., 2005; Santiago and Mamoun, 2003). However, the ino2Δ and ino4Δ deletion mutants did not display inositol/choline-mediated induction but, interestingly, did reveal phosphate induction (Figure 2B). Another interesting result was that the pho4Δ mutant did not display the inositol induction or phosphate regulation (Figure 2B). Lastly, the hms1Δ mutant yielded significantly increased expression in I+C+ conditions, while the sgc1Δ mutant yielded elevated expression under all four growth conditions (Figure 2B). For the remainder of this study, we decided to focus on the inositol/choline-mediated induction because this is such a well-established system.
To confirm that the inositol induction of CIT2–lacZ expression accurately reflects transcription regulation, CIT2 transcript levels were quantified by real-time Q-PCR. As expected, the data show that the CIT2 transcript levels in a ρ0 wild-type strain was induced relative to a ρ+ wild-type strain and undetectable in rtg1Δ and rtg3Δ strains (Figure 2C). As seen with the CIT2–lacZ reporter, CIT2 expression was induced in the presence of inositol and choline regardless of the phosphate concentration (Figure 2C).
A complementation test was used to confirm that the phenotypes were indeed due to the deleted bHLH genes (Figure 3). The ρ+ versions of the rtg1Δ and rtg3Δ and the ρ0 versions of the ino2Δ and ino4Δ strains containing the CIT2–lacZ reporter gene were transformed with centromeric pRS315 plasmids containing the RTG1, RTG3, INO2 and INO4 genes under the control of their own promoters (Table 2). In almost every case, the clones restored CIT2–lacZ expression levels back to the normal levels (Figure 3). Plasmid pRS315–RTG1 complemented the rtg1Δ, but not the rtg3Δ, phenotype (Figure 3A). Likewise, pRS315–INO2 and pRS315–INO4 complemented the CIT2–lacZ expression phenotypes of the ino2Δ and ino4Δ mutants (cf. Figure 3B to 2B) as well as the inositol auxotrophy (data not show). The pRS315–RTG3 plasmid partially restored expression in the rtg3Δ mutant strain but not the rtg1Δ mutant strain (Figure 3A). It is unclear why this plasmid did not yield complete complementation; however, it may reflect that RTG3 expression from the pRS315 plasmid is not at normal levels.
It has been determined that, while inositol represses phospholipid biosynthetic gene expression and choline enhances this regulation, inositol and choline can affect gene expression separately (Jesch et al., 2005; Santiago and Mamoun, 2003). To determine whether the elevated expression of CIT2 were affected by inositol and choline separately or together, β-galactosidase activity was assayed from wild-type ρ+ and ρ0 cells grown in various combinations of inositol and choline with low and high phosphate. The data show that CIT2–lacZ expression was induced in ρ0 strains and that the inositol/choline induction was also observed when inositol was supplied alone. Choline by itself had no effect (Figure 4).
Ino2p, Ino4p and Pho4p regulate CIT2–lacZ expression through R-boxes
Promoter mutants were used to define the role of potential bHLH binding sites in the CIT2 promoter. While most yeast promoters require regulatory sequences within 500 bp the ORF, some promoters do utilize more distal regulatory elements (Kristiansson et al., 2009). Consequently, we inspected 1000 bp upstream of the CIT2 ORF for potential E-boxes (5′-CANNTG-3′) and R-boxes (5′-GGTCAC-3′) and found a single E-box at − 809 next to a previously uncharacterized R-box at − 804, in addition to the previously studied R-boxes at − 346 and − 312 (Figure 5A) (Jia et al., 1997). These four sites were individually mutated to restriction sites in the CIT2–lacZ plasmid (see Materials and methods) and assayed after growth in a ρ0 strain.
The results supported previous findings showing that R-boxes at − 346 and − 312 are both required for expression of the CIT2–lacZ gene in a ρ0 strain, with the − 312 element playing the more prominent role (Figure 5B) (Jia et al., 1997). Conversely, mutating the − 809 and − 804 elements had little effect on overall levels of expression (Figure 5B). However, expression from − 809 and − 804 mutant promoters yielded the same pattern of expression observed with the wild-type promoter with ino2Δ and ino4Δ mutant strains (cf. Figures 5B and 2B). That is, the inositol-mediated induction was replaced by phosphate-mediated regulation. This suggests that Ino2p and Ino4p are required for inositol-mediated induction through the distal sites. Because of the close proximity between these distal sites, it was not possible to conclude which sites were functional. The results further suggest that phosphate regulation is exerted through the − 346 and − 312 elements, since their deletion eliminates the response while deletion of the two distal elements does not affect the response. Expression from all of the promoter constructs was completely eliminated in rtg1Δ and rtg3Δ mutant strains (data not shown).
RTG3 expression is also regulated by bHLH proteins
Regulation of INO2 bHLH gene expression has been shown to be important for expression levels of the phospholipid biosynthetic genes (Ashburner and Lopes, 1995b; Schwank et al., 1997). Thus, one possible explanation for the inositol or phosphate regulation of CIT2 is that it is mediated indirectly through regulation of RTG1 and/or RTG3 expression. We focused on RTG3 because Rtg3p levels has been shown to be limiting relative to Rtg1p (Rothermel et al., 1997) and RTG3 transcription is regulated in response to mitochondrial DNA deficiency (Jia et al., 1997), while Rtg1p levels are not affected (Rothermel et al., 1995). To test this, we used an RTG3–cat reporter stably integrated in single copy at the GAL4 locus (Ashburner and Lopes, 1995a; Griggs and Johnston, 1991, 1993). Both ρ0 and ρ+ versions of wild-type, rtg1Δ, rtg3Δ, ino2Δ, ino4Δ and pho4Δ strains were grown as described above and assayed for Cat activity.
The results show RTG3–cat expression is induced approximately two-fold in a wild-type strain in response to mitochondrial DNA-deficiency (cf. Figures 6A and 6B) (Jia et al., 1997). This is consistent with a previous report showing two-fold induction of RTG3 mRNA levels (Jia et al., 1997). RTG3–cat expression was also elevated approximately two-fold in all of the ρ+ bHLH knock-out strains (Figure 6A). The data also show phosphate, but not inositol, induction in every strain tested (Figure 6). This suggests that the phosphate induction of CIT2 expression might be due to indirect regulation of RTG3. However, it raises the question of why the phosphate induction of CIT2 is only observed in the ρ0ino2Δ and ino4Δ strains (Figure 2B). One possible explanation is suggested by the observed two-fold decrease in RTG3–cat expression ρ0ino2Δ and ino4Δ strains (Figure 6B). That is, Rtg3p levels may become limiting in the ρ0ino2Δ and ino4Δ strains, allowing the phosphate regulation of RTG3 to have an effect on CIT2 expression. Interestingly, deletion of the PHO4 gene did not eliminate phosphate regulation of RTG3–cat (Figure 6). Lastly, deletion of RTG3 and RTG1 did not significantly affect RTG3–cat expression, suggesting that RTG3 does not autoregulate its expression, as has been shown for the INO2 gene (Ashburner and Lopes, 1995a, 1995b; Eiznhamer et al., 2001; Gardenour et al., 2004; Heyken et al., 2005; Miller and Lopes, 2001; Schwank et al., 1997) and other bHLH-encoding genes (Shetty, He, Chen and Lopes, unpublished data).
Previous studies on bHLH function have largely been focused on DNA-binding specificity and dimer selection (Blackwell and Weintraub, 1990; Burley et al., 1993; Cave et al., 2000; Ferre-D'Amare et al., 1993, 1994; Fisher and Goding, 1992; Hakoshima et al., 1993; Robinson et al., 2000; Robinson and Lopes, 2000a; Shimizu et al., 1997). However, it is clear that auto- and cross-regulation of bHLH gene expression, interorganellar movement of bHLH proteins and sequestration of bHLH proteins from binding DNA also play important roles (Ashburner and Lopes, 1995a; Chen and Lopes, 2007; Jen et al., 1992; Kaffman et al., 1998a, 1998b; Komeili and O'Shea, 1999; Lin et al., 2000; O'Neill et al., 1996; Robinson and Lopes, 2000b; Ross et al., 2003; Schüller et al., 1992; Schwank et al., 1997; Sekito et al., 2000). S. cerevisiae is an ideal system to evaluate the contribution of each of these mechanisms, since it contains a relatively small number of bHLH proteins. This study used the CIT2 gene as a reporter of bHLH function. The results faithfully recapitulated the published results with this gene. First, CIT2–lacZ and CIT2 mRNA levels were induced in ρ0 strains and this induction required the Rtg1p–Rtg3p bHLH dimer and two R-boxes (−346 and − 312) in the CIT2 promoter (Figures 2, 5)(Chelstowska and Butow, 1995; Jia et al., 1997). The fact that the degree of induction differed slightly from previous reports might be due to several factors. It has been shown that retrograde regulation is strain-sensitive (Dilova and Powers, 2006). Moreover, the current study utilized a variety of growth conditions (low/high phosphate, with/without inositol), which affect bHLH protein function and are different from previous media (typically, complete synthetic media). Nevertheless, the general pattern of regulation reported here is in excellent agreement with published studies.
The results here show that multiple bHLH proteins known to regulate a variety of biological processes also regulate CIT2. While bHLH proteins did play a role in ρ+ strains, the more dramatic effects were observed in ρ0 strains (Figure 2). Interestingly, both HMS1 and SGC1 were found to have repressive effects in I+C+ medium and all four media, respectively (Figure 2B). Relatively little is known about Hms1p beyond its role in regulating pseudohyphal differentiation (Lorenz and Heitman, 1998). A more recent study revealed that overexpression of HMS1 induced pheromone-responsive genes and some metabolic genes, both of which are consistent with a role in filamentation (Chua et al., 2006). The published study focused on genes induced by overexpression of HMS1 and therefore, as expected, did not identify the CIT2 gene, which is repressed by HMS1. Another study revealed that HMS1 is required for full expression of the ENO1 gene, which encodes a glycolytic enzyme (Chen and Lopes, 2007). Coincidental with this, Sgc1p is the primary regulator of ENO1 expression (Nishi et al., 1995; Sato et al., 1999). Consequently, the current analysis of CIT2 expression identified a novel target for these two relatively understudied bHLH proteins and shows that CIT2 expression is coordinated with ENO1 expression.
One of the more dramatic observations was that, in ρ0 strains, CIT2–lacZ expression was derepressed in response to inositol (Figure 2). The inositol response in yeast is known to be mediated by the Ino2p–Ino4p bHLH dimer, which regulates the phospholipid biosynthetic genes (Carman and Henry, 1999; Chen et al., 2007; Greenberg and Lopes, 1996; Henry and Patton-Vogt, 1998; Jesch et al., 2005; Robinson and Lopes, 2000a; Santiago and Mamoun, 2003). Therefore, the current study suggests that CIT2 expression is coordinated with phospholipid biosynthesis. It is worth noting that the role of INO2 and INO4 in regulating CIT2 was not revealed by two expression-profiling studies that identified inositol-regulated or Ino2p : Ino4p-regulated genes (Jesch et al., 2005; Santiago and Mamoun, 2003) and a genome-wide study of transcription factor binding sites (ChIP–chip) (Harbison et al., 2004). This is not entirely surprising, since these genome-wide studies did not use ρ0 strains, which is where the effect of Ino2p and Ino4p is most prominent (Figure 2B). The expression-profiling studies each identified a significant number of inositol/choline-induced genes (sets of 35 and 11 in the two studies) (Jesch et al., 2005; Santiago and Mamoun, 2003), but the sets did not overlap significantly. The lack of consistency has been attributed to differences in experimental design and media composition (Jesch et al., 2005). Regardless, only three genes were found to be induced in the presence of inositol alone (HO and two unknown ORFs) (Jesch et al., 2005). Thus, it is not surprising that these published studies did not identify CIT2 as a target.
The results suggest that Ino2p and Ino4p regulate CIT2 via two different mechanisms (Figure 7). They mediate the inositol response via the distal E-box (−809) or R-box (−804) (Figures 2B, 5). Because of the proximity of the two binding elements, it was not possible to definitively state whether both elements are required for this regulation. INO2 and INO4 were also required for full expression of the RTG3–cat gene in a ρ0 strain (Figure 6B). It has been previously shown that inositol-mediated regulation of INO2 expression requires both INO2 and INO4 (Ashburner and Lopes, 1995a, 1995b; Eiznhamer et al., 2001; Gardenour et al., 2004; Heyken et al., 2005; Jackson and Lopes, 1996; Miller and Lopes, 2001; Schwank et al., 1997). Similarly, RTG3–cat expression was induced in a ρ0 strain which supports previously published results (Figure 6) (Jia et al., 1997). However, this induction did not appear to require the RTG3 gene (Figure 6B). RTG3–cat expression in the ρ0 wild-type strain differed from CIT2 expression, in that it appeared to be induced by inositol and phosphate (Figures 2B, 6B). The inositol induction of RTG3–cat expression was subtle but, predictably, was eliminated in the ino2Δ and ino4Δ strains (Figure 6B).
The phosphate induction of RTG3–cat expression, however, was evident in every strain tested including a pho4Δ strain (Figure 6B). In the PHO system, Pho4p mediates induction of target genes in response to phosphate starvation (Johnston and Carlson, 1992). Thus, RTG3–cat expression is unusual in two respects, phosphate induction and Pho4p independence. There is evidence in the literature of Pho4p-mediated phosphate repression. A promoter shared by two stationary phase genes (SNZ1 and SNO1) and two proline catabolic genes (PUT1 and PUT2) are all repressed by Pho4p (Nishizawa et al., 2008; Popova Iu et al., 2000). Pho4p was found to bind to the shared SNZ1/SNO1 promoter (Nishizawa et al., 2008). However, to our knowledge RTG3 provides the first evidence of phosphate regulation that does not require the Pho4p transcription factor. It is not unreasonable to consider a model for Pho4p-independent phosphate regulation, since Ino2p/Ino4p-independent inositol regulation has been reported (Graves and Henry, 2000).
An obvious important question to address is why CIT2 expression is regulated by such a complex network that includes a response to inositol and phosphate. One possible explanation is that phospholipid synthesis is regulated by inositol and phospholipids are a source of acetyl-coA, which is a substrate for citrate synthase (Epstein et al., 2001). Arguably, the regulation by phosphate was only observed under very specific conditions in our experiments (absence of INO2 or INO4 and ρ0), which may not happen in nature. However, it is noteworthy that expression profiling of mitochondrial dysfunction does identify the PHO89 gene (Na+-dependent phosphate transporter) as a target, suggesting that mitochondrial function is coordinated with phosphate uptake (Epstein et al., 2001; Traven et al., 2001). Another possible explanation is that this is simply a mechanism to balance phospholipid synthesis (regulated by inositol) with energy production, since citrate synthase is the rate-limiting step in the TCA cycle.
The results presented here support the following model (Figure 7). Retrograde regulation of CIT2 expression is predominantly dictated by the Rtg1p/Rtg3p heterodimer through the two proximal R-boxes (−312 and − 346). The Ino2p/Ino4p heterodimer is required for inositol-mediated induction through the distal E-box/R-box (−809 and − 804) but inositol induction is manifested only in mitochondrial DNA-deficient cells. It is not known whether they bind directly to the CIT2 promoter, or function by sequestering the binding of another factor, or regulating the expression of a repressor protein that binds the CIT2 promoter. The observation that CIT2 expression is induced by phosphate when INO2 or INO4 are deleted may be explained by indirect regulation via RTG3. That is, in the presence of Ino2p and Ino4p, Rtg3p levels are sufficiently high that the effect of phosphate on RTG3 expression is not manifested on CIT2 expression. Conversely, in ino2Δ and ino4Δ strains, RTG3 expression is reduced below a threshold that allows for phosphate regulation of CIT2 to be manifested. However, this would not explain why deletion of the distal E-box/R-box (−809 and − 804) eliminates inositol induction and reveals phosphate induction. Therefore, the simpler model is that phosphate regulation requires that Pho4p directly induce CIT2, but only when inositol induction is eliminated by deleting Ino2p/Ino4p or the cognate CIT2 promoter element. The other results that should be addressed is why the inositol regulation, and the effect of other bHLH proteins, are most dramatic in the ρ0 strains. The simplest explanation is that regulation of CIT2 expression is dominated by Rts1p and Rtg3p, and that the other bHLH proteins function by modulating the regulation by Rts1p/Rtg3p. Since there is more Rts1p/Rtg3p complex bound in ρ0 strains, the effect of the other bHLH proteins is magnified. The auxiliary effect of other proteins may be because they function through a distant site (e.g. Ino2p, Ino4p and inositol reponse function through − 809 and/or − 804 elements) or regulate by dimerization the Rts1p/Rtg3p levels available for binding the promoter.
The authors thank Ying He, Ameet Shetty, Naved Munir and Tina Paul for helpful discussions and editing of the manuscript. We dedicate this study to the memory of Dr Ronald Butow, who led the field of retrograde regulation. This work was supported by a National Science Foundation grant (MCB-0718608) to J.M.L.