|Home | About | Journals | Submit | Contact Us | Français|
The budding yeast Saccharomyces cerevisiae responds to depletion of iron in the environment by activating Aft1p, the major iron-dependent transcription factor, and by transcribing systems involved in the uptake of iron. Here, we have studied the transcriptional response to iron deprivation and have identified new Aft1p target genes. We find that other metabolic pathways are regulated by iron: biotin uptake and biosynthesis, nitrogen assimilation, and purine biosynthesis. Two enzymes active in these pathways, biotin synthase and glutamate synthase, require an iron-sulfur cluster for activity. Iron deprivation activates transcription of the biotin importer and simultaneously represses transcription of the entire biotin biosynthetic pathway. Multiple genes involved in nitrogen assimilation and amino acid metabolism are induced by iron deprivation, whereas glutamate synthase, a key enzyme in nitrogen assimilation, is repressed. A CGG palindrome within the promoter of glutamate synthase confers iron-regulated expression, suggesting control by a transcription factor of the binuclear zinc cluster family. We provide evidence that yeast subjected to iron deprivation undergo a transcriptional remodeling, resulting in a shift from iron-dependent to parallel, but iron-independent, metabolic pathways.
Iron is an essential nutrient for virtually every organism on earth, because iron participates as a cofactor in numerous essential enzymatic reactions involving the transfer of electrons. Remarkably, organisms can thrive in environments in which the bioavailable iron is extremely scarce, and they can survive tremendous changes in environmental iron conditions, in part by altering patterns of transcription. The budding yeast Saccharomyces cerevisiae responds to depletion of iron in the environment by increasing the expression of systems involved in the uptake of iron. Other metabolic alterations that might occur in response to iron deprivation and the controls for any such alterations are not known. These systems of iron uptake are controlled primarily by Aft1p, the major iron-dependent transcription factor in yeast (Yamaguchi-Iwai et al., 1995 ). Aft1p is constitutively expressed in the cytosol of growing cells, and iron depletion triggers a relocalization of Aft1p to the nucleus, where it binds DNA and activates transcription (Yamaguchi-Iwai et al., 2002 ). A related transcription factor, Aft2p, regulates a subset of the Aft1p target genes, but the role of Aft2p in iron homeostasis is less clear (Blaiseau et al., 2001 ; Rutherford et al., 2001 , 2003 ).
Aft1p activates the transcription of a set of 17 genes involved in the reductive and nonreductive uptake of iron salts and iron-siderophore chelates into the cell. This set includes three cell wall mannoproteins, Fit1p, Fit2p, and Fit3p, which are involved in the retention of iron in the cell wall and enhance the uptake of siderophore-bound iron (Protchenko et al., 2001 ). The reductive system of uptake begins with the reduction of ferric iron salts and chelates to the ferrous form by members of the FRE family of plasma membrane metalloreductases Fre1p, Fre2p, Fre3p, and Fre4p (Dancis et al., 1990 ; Georgatsou and Alexandraki, 1994 ; Yun et al., 2001 ). Two additional FRE family members, Fre5p and Fre6p, are regulated by Aft1p but are not yet functionally characterized, and a seventh family member, Fre7p, is regulated by copper (Martins et al., 1998 ). Reduced iron is then taken up by the high-affinity ferrous transport complex composed of the multicopper oxidase, Fet3p (Askwith et al., 1994 ), and the iron permease, Ftr1p (Stearman et al., 1996 ). Intracellular copper loading of Fet3p occurs through the activities of the copper chaperone Atx1p (Lin et al., 1997 ) and the microsomal copper transporter Ccc2p (Yuan et al., 1995 ). A second, nonreductive system of iron uptake exhibits specificity for a variety of siderophore-iron chelates and is composed of four transporters encoded by ARN1, ARN2/TAF1, ARN3/SIT1, and ARN4/ENB1 (Heymann et al., 1999 , 2000a ,b ; Lesuisse et al., 1998 ; Yun et al., 2000a ,b ).
These Aft1p target genes involved in iron uptake have been identified both through traditional yeast genetic methods and through cDNA microarrays representing the entire yeast genome. Microarrays designed to identify genes that were differentially expressed after deletion of YFH1 (Foury and Talibi, 2001 ), a gene involved in mitochondrial iron-sulfur cluster metabolism, and arrays designed to identify genes differentially regulated by cobalt stress (Stadler and Schweyen, 2002 ) identified a number of Aft1p target genes. Some genes that were differentially expressed in these arrays were also noted to have Aft1p consensus binding sites in their promoter regions, but these genes were not definitively shown to be Aft1p target genes. Arrays designed to identify genes expressed in the presence of a constitutively active AFT1 or AFT2 allele also identified many Aft1p target genes (Rutherford et al., 2003 ) but did not distinguish between direct and indirect effects of Aft1p expression. Here, we describe the transcriptional response to iron deprivation in yeast and identify both the set of genes directly regulated by Aft1p and genes that are regulated in response to changes in iron availability independently of Aft1p. We report that Aft1p also directs the transcription of genes involved in intracellular iron utilization and homeostasis and that the biotin, glutamate, and purine biosynthetic pathways are transcriptionally regulated in response to changes in iron availability. Promoter mapping data indicate that a transcription factor of the binuclear zinc cluster family may activate transcription under conditions of iron sufficiency.
Strain DBY7286 (MATa ura3 GAL) and the congenic aft1 mutants were used for cDNA microarrays, Northern blots, and enzymatic assays (Yun et al., 2000a ). To generate the vht1Δ strain, the strain RG24695 (MATa/α his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 MET15/met15Δ0 LYS2/lys2Δ0 ura3Δ0/ura3Δ0 VHT1/vht1Δ:KAN; Research Genetics, Huntsville, AL) was sporulated and the spore clones allowed to germinate on YPD medium supplemented with biotin 100 ng/ml. A tetrad containing two clones that failed to grow on SC medium but grew on medium containing biotin and geneticin (vht1Δ) and two clones that grew on SC medium but failed to grow in the presence of geneticin (VHT1+) was used in biotin/KAPA growth experiments. To construct the aft1Δ strain in the constitutively active Fet3p/Ftr1p background, the plasmid YIpDCE1/FET3-HA/FTR1-MYC was linearized and integrated at the ADE2 locus of the strain CPY101 (MATa ura3-52 lys2-801(amber) ade2-101(ochre) trp1-Δ63 his3-Δ200 leu2-Δ1, aft1::TRP1) (Philpott et al., 1998 ) and the congenic parent strain YPH499 as described previously (Stearman et al., 1998 ).
Plasmids 1–6, containing the GLT1 promoter regions including the GLT1 translation start codon fused to lacZ, were constructed by polymerase chain reaction (PCR) by using primers containing an XhoI site in the forward primer and a BamHI site in the reverse primer. PCR products were cloned into pLG699-Z (Guarente and Ptashne, 1981 ) that had been linearized with XhoI and BamHI. Plasmids 7–9 and 11–14 were constructed from PCR fragments containing XhoI sites at both ends, which were cloned into the XhoI site of plasmid 5. Plasmid 10 was constructed by cloning the PCR fragment into plasmid 6. All constructs were confirmed by sequencing. YPD medium, SD medium, and defined-iron medium were prepared as described previously (Philpott et al., 1998 ; Sherman, 1991 ). Defined-iron medium contains 1 mM ferrozine, a ferrous iron chelator that, in the presence of ferrous iron salts, produces a medium containing a constant and reproducible amount of free ferrous iron.
Strains were grown to mid-log phase (A600 of 0.4–0.6) in either SD medium containing complete supplemental mixture (0.8 g/l) or defined iron medium (also contains complete supplemental mixture) containing 20 μM (iron-poor), 100 μM (iron-sufficient), or 500 μM (iron-enriched) ferrous ammonium sulfate. Total RNA was isolated using TRIzol (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. Probe preparation, microarray production, hybridization, and data acquisition were performed as described previously (DeRisi et al., 1997 ). Probes generated from the AFT1-1up strain, cells grown in iron-poor medium, and cells grown in iron-enriched medium were labeled with Cy5 (red), and probes generated from the wild-type strain, the aft1Δ strain, and cells grown in iron-sufficient medium were labeled with Cy3 (green). Data were analyzed using ScanAlyze software (available at http://rana.lbl.gov/EisenSoftware.htm). Each array was performed twice using probes generated from independent cultures grown on different days. The AFT1-1up array was performed once in comparison with the wild-type strain and once in comparison with the aft1Δ strain with similar results, because Aft1p-dependent transcription is repressed in SD medium (Yun et al., 2000a ). The complete data set for the arrays is available at http://genome-www.stanford.edu/microarray.
RNA isolation and Northern blot analysis were performed as described previously (Philpott et al., 1998 ) by using 32P-labeled probes that correspond to the entire open reading frames of the indicated genes. For growth assays, strains were precultured in either SC medium or defined-iron medium containing 5 μM ferrous ammonium sulfate and then inoculated into the same medium containing 7-keto, 8-amino pelargonic acid (KAPA; gift of Claude Alban, Aventis, Strasbourg, France) at 0, 1, 5, or 10 ng/ml or biotin at 100 ng/ml and grown at 30°C with shaking. Samples were withdrawn periodically and the A600 determined.
Glutamate dehydrogenase activity was measured spectrophotometrically at 23°C by the oxidation of the reduced coenzyme NADPH as described previously (Doherty, 1970 ). Glutamine synthetase activity was measured at 30°C by the transferase reaction as described previously (Mitchell and Magasanik, 1983 ). Glutamate synthase activity was measured at 23°C by the oxidation of NADH as described previously (Roon et al., 1974 ). β-Galactosidase activity was measured by the production of o-nitrophenol at 28°C as described previously (Adams et al., 1997 ).
To define the set of genes that was directly regulated by Aft1p, we used cDNA microarrays to identify genes that exhibited increased expression in iron-poor medium, decreased expression in iron-enriched medium, or increased expression in a strain expressing AFT1-1up, a constitutively active allele of AFT1 (Yamaguchi-Iwai et al., 1995 ). Expression of AFT1-1up results in high levels of transcription of Aft1p target genes, regardless of the amount of iron in the medium. For each of the identified genes, we examined the upstream DNA sequences within 700 base pairs of the open reading frame for the Aft1p consensus binding site, (T/C)(G/A)CACCC (Yamaguchi-Iwai et al., 1996 ). Genes that exhibited this pattern of iron and Aft1p regulation and also contained an Aft1p consensus site were further examined by Northern blot analysis and were confirmed as Aft1p target genes if they exhibited increased expression in both the AFT1-1up strain and in iron-poor medium. By identifying only those genes that were activated in both of these conditions, we were able to exclude genes that were indirectly activated by iron deprivation in the wild-type strain or by iron accumulation in the AFT1-1up strain. Examination of the entire yeast genome for Aft1p consensus sites by using multiple EM for motif elicitation revealed that many genes that were not regulated by iron or Aft1p contained this minimal binding site in the upstream region (T.F. and C.C.P., unpublished observations). This finding indicated that the presence of an Aft1p site could not be used as the sole criterion to identify the Aft1p regulon. The set of genes regulated by Aft1p, including target genes identified in previous studies, is presented in Figure 1. A clustered analysis of all of the genes exhibiting iron- and Aft1p-dependent regulation is presented in Supplementary Figure 1.
In addition to the 17 genes involved in the uptake of iron at the plasma membrane and FTH1, a gene involved in vacuolar iron transport (Urbanowski and Piper, 1999 ), we identified TIS11, a gene of unknown function; HMX1, which has homology to heme oxygenases; and three additional transporters, SMF3, COT1, and VHT1. TIS11 and HMX1 each exhibited multiple Aft1p consensus binding sites in the upstream sequences of the DNA, whereas SMF3, COT1, and VHT1 each exhibited a single consensus binding site for Aft1p in the 5′ region of the gene (Figure 2A). The transcription profile for each of these genes was similar, with higher mRNA levels present in cells grown in iron-poor medium and in the AFT1-1up strain, suggesting that each of these genes was a direct target of Aft1p (Figure 2B). Transcript levels for SMF3, COT1, and VHT1 were low, but readily detectable, in cells grown in iron-sufficient medium and in an AFT1-deleted strain, indicating that these genes were also under the control of other transcription factors and may have a role outside the response to iron deprivation. SMF3 mRNA levels increased only slightly in the AFT1–1up strain, indicating that much of the iron-dependent regulation occurred through another transcription factor, such as Aft2p (Portnoy et al., 2002 ; Rutherford et al., 2003 ).
The newly described Aft1p target genes that have been characterized, along with FTH1, function in the mobilization and reutilization of intracellular iron and in the sequestration of other metals. HMX1 encodes a protein with heme degradation activity that is important in heme iron utilization in yeast (Protchenko and Philpott, 2003 ). TIS11 encodes a protein of the CCCH Zn-finger family, which is represented in mammals, flies, and worms, as well as yeast (Ma and Herschman, 1995 ; Thompson et al., 1996 ). Cot1p is a transporter located in the vacuolar membrane implicated in cobalt resistance (Conklin et al., 1992 ) and in the accumulation of zinc ions in the vacuole (MacDiarmid et al., 2000 ). The role of Cot1p in the response to iron deprivation is not clear, but sequestration of other divalent metals ions in the vacuole may be important for cell survival when iron is scarce (Li and Kaplan, 1998 ). Smf3p is 27% identical at the amino acid level to DMT1, a mammalian iron transporter, and it is also expressed in vacuolar membranes and may be involved in the mobilization of vacuolar stores of iron (Portnoy et al., 2000 ). VHT1 encodes the essential H+-biotin symporter and was a surprising addition to the Aft1p regulon (Stolz et al., 1999 ).
S. cerevisiae is auxotrophic for biotin, an essential nutrient also known as vitamin H, and strains bearing deletions of VHT1 are not viable unless supplemented with certain biotin precursors or high concentrations of biotin (Stolz et al., 1999 ). In higher plants, most fungi, and bacteria, biotin is synthesized from pimelic acid in five steps (Figure 3A). Although the genome of budding yeast does not encode the enzymes for the first two steps of synthesis, yeast can synthesize biotin from the precursors KAPA and 7,8-diamino pelargonic acid (DAPA) through the activities encoded by BIO3, BIO4, and BIO2, respectively. BIO5 encodes a high-affinity transporter for KAPA and DAPA. BIO2 encodes biotin synthase, the rate-limiting step in biotin synthesis, and this protein requires a 4Fe-4S cluster for activity (Marquet et al., 2001 ). The Aft1p-dependent increase in VHT1 transcription during growth in iron-poor medium suggested that yeast increased the uptake of biotin under conditions of iron deprivation. We questioned whether this increase in biotin uptake occurred in response to a decrease in biotin synthesis, and therefore examined whether yeast could synthesize biotin under conditions of iron deprivation.
We constructed a VHT1-deletion strain and measured the capacity of the biotin precursor KAPA to support growth of this strain and the congenic VHT1+ strain in iron-rich medium and iron-poor medium (Figure 3, B–E). The vht1Δ strain cannot take up biotin at the low concentrations present in standard media and does not grow in the absence of supplemental biotin precursors or high concentrations of biotin. As expected, in iron-rich medium, the VHT1+ strain grew well with or without KAPA supplementation (Figure 3B), whereas the vht1Δ strain grew well only in the presence of KAPA supplementation, although only very small amounts of KAPA were required (Figure 3C). When the VHT1+ strain was introduced to iron-poor medium, the cells again grew in the presence or absence of KAPA (Figure 3D). In contrast, when the vht1Δ strain was introduced to iron-poor medium, KAPA supplementation failed to support growth, and the cells grew only in the presence of high amounts of biotin (Figure 3E). These data indicated that, under conditions of iron deprivation, yeast did not synthesize sufficient quantities of biotin from the precursor KAPA to support growth.
We considered two explanations for the failure of yeast to synthesize biotin under conditions of iron deprivation: 1) the yeast failed to incorporate iron-sulfur clusters into Bio2p, and thereby produced inactive enzyme; or 2) the enzymes involved in the biotin biosynthetic pathway were poorly expressed. By Northern blot analysis, mRNA levels for each of the genes in the biotin synthetic pathway were reduced in cells grown in iron-poor medium when compared with cells grown in iron-enriched medium, with transcripts of BIO3 and BIO4 being very low in iron-poor and iron-sufficient medium (Figure 3F). These results suggested that transcription of the genes involved in biotin synthesis was repressed when iron was scarce and that biotin uptake and synthesis were reciprocally regulated at the transcriptional level by iron. BIO2, BIO3, and BIO4 may also be transcriptionally repressed when extracellular biotin levels rise (Stolz, personal communication). This could explain the reduced mRNA levels of these genes in the AFT1-1up strain, where increased expression of VHT1 would lead to increased biotin uptake. The small amount of KAPA-dependent growth observed in the iron-deficient cultures (Figure 3D) seems to be consistent with the small amount of BIO2, BIO3, and BIO4 transcription indicated by Northern blot analysis (Figure 3F).
Further examination of the cDNA microarrays suggested that additional metabolic pathways might be regulated by iron. Twelve genes involved in the uptake and metabolism of amino acids and alternative nitrogen sources were more highly expressed in cells grown in iron-poor medium than in cells grown in iron-sufficient medium in at least one array (Figure 4A). Northern blot analysis confirmed that RNA levels for MEP2, an ammonium transporter; CAR1, an arginase; AGP1, an asparagine and glutamine permease; and DUR3, a urea permease were elevated in cells grown in iron-poor medium compared with cells grown in either iron-sufficient or iron-enriched media (Figure 4B). None of these genes was more highly expressed in the AFT1-1up strain by cDNA microarray or by Northern blot analysis, and, with the exception of CAR1, none contained an Aft1p consensus site, indicating that Aft1p did not directly regulate these genes. Further examination of the pathways of nitrogen assimilation suggested that the synthesis of glutamate and glutamine could account for this pattern of activation in iron-poor medium.
S. cerevisiae grows well when ammonium is provided as the sole source of nitrogen. To utilize ammonium, however, cells must first incorporate it into either glutamate or glutamine, because ~88% of cellular nitrogen is derived from the amino group of glutamate and 12% is derived from the amide group of glutamine (Magasanik, 1992 ). The glutamate dehydrogenases encoded by GDH1 and GDH3 catalyze the NADPH-dependent formation of glutamate from ammonia and α-ketoglutarate (Figure 5A, reaction 1). In reactions 2 and 3, glutamine synthetase, encoded by GLN1, catalyzes the formation of glutamine from ammonia and glutamate, and glutamate synthase, encoded by GLT1, catalyzes the formation of glutamate from glutamine and α-ketoglutarate. The sum of reactions 2 and 3 is equivalent to reaction 1, except that ATP is consumed and NADH substitutes for NADPH; therefore, reactions 2 and 3 represent an alternative pathway for the synthesis of glutamate. Glutamate synthases require a 4Fe-4S cluster for activity (Curti et al., 1996 ), and we questioned whether the activity of these enzymes, especially Glt1p, might be affected by the availability of iron. Therefore, we grew cells in media containing varying amounts of iron and measured the activity of glutamate dehydrogenase, glutamine synthetase, and glutamate synthase in crude lysates (Figure 5B). Glutamate dehydrogenase and glutamine synthetase activities exhibited little change after growth in varying amounts of iron. In contrast, glutamate synthase activity was very low in lysates of cells grown in medium containing very low amounts of iron, but increased 20-fold as the concentration of iron in the medium increased.
The paucity of Glt1p activity after growth in iron-poor medium may reflect either a shortage of iron-sulfur clusters or a decrease in transcription. By Northern analysis, GLT1 mRNA levels decreased as the concentration of iron in the medium decreased (Figure 6A). Furthermore, all of the loss of Glt1p activity in cells grown in decreasing amounts of iron could be explained by transcriptional repression of GLT1 under these growth conditions. These changes in GLT1 mRNA levels were not observed in the microarrays, possibly due to inefficient probe synthesis from an exceptionally long (>6 kbp) transcript. In contrast, GDH1, GDH3, and GLN1 mRNA levels changed in a complex manner when cells were grown in varying amounts of iron (Figure 6B). In cells grown in iron-sufficient or iron-enriched medium, transcription from GDH3 produced two species of mRNA. Cells grown in lower amounts of iron exhibited a loss of the smaller mRNA and the appearance of a third transcript that was more abundant and smaller than either of the first two transcripts. This third transcript was also present in cells expressing the AFT1-1up allele. The specificity of each of the detected transcripts for GDH3 was confirmed by Northern analysis of a GDH3-deletion strain (our unpublished data). GDH1 and GLN1 exhibited similar patterns of transcription, with mRNA levels increasing slightly in cells grown in iron-poor (20 μM) medium, but not at the lowest iron concentrations. GLT1, GDH1, GDH3, and GLN1 are extensively regulated at the transcriptional level through multiple nitrogen-responsive transcription factors (Magasanik, 1992 ) but were not known to be regulated transcriptionally by iron.
Because Northern analysis indicated that transcription of GLT1 was repressed by iron deprivation and by expression of the AFT1-1up allele, we considered whether Aft1p might also act as a transcriptional repressor. We therefore tested whether the iron-dependent regulation of GLT1 occurred in a strain deleted for AFT1. aft1Δ strains normally do not grow in media containing low amounts of iron due to a failure to express the high-affinity transport complex Fet3p/Ftr1p. To bypass this block on high-affinity iron uptake, we deleted AFT1 in a strain in which Fet3p and Ftr1p were expressed constitutively and measured the transcription of GLT1 in cells grown in varying amounts of iron (Figure 6C). No difference in the iron-dependent pattern of GLT1 expression was observed between the aft1Δ strain and the congenic parent strain, indicating that Aft1p was not required for the iron-dependent transcription of GLT1. Subsequent rehybridization with an ACT1 probe indicated equivalent RNA loading (our unpublished data).
Inspection of the cDNA microarrays comparing cells grown in iron-enriched medium to cells grown in iron-sufficient medium revealed that most of the genes involved in the biosynthesis of purines were induced when the concentration of iron was high (Figure 7A). These results were confirmed by Northern analysis of GCV1 and ADE17, which indicated that both genes were more highly expressed in cells grown in iron-enriched medium (Figure 7B). De novo synthesis of purines requires a succession of 10 enzymatic steps carried out by products of the ADE genes. Glycine and one-carbon units produced by the glycine cleavage system and one-carbon metabolism through folate are also required for purine biosynthesis (Denis and Daignan-Fornier, 1998 ). These systems are coregulated by adenine through the transcription factors Bas1p and Bas2p (Daignan-Fornier and Fink, 1992 ). Adenine represses the transcription of the genes of purine biosynthesis and we examined whether iron affected this process. By Northern analysis, growth in media containing high amounts of adenine fully repressed ADE17 transcription, even when iron concentrations were high (Figure 7C). Conversely, growth in media without adenine fully derepressed ADE17 transcription, even when iron concentrations were low. Deletion of BAS1 also resulted in a near complete loss of ADE17 mRNA, with a concomitant loss of iron induction. These results suggested that the effect of iron on the expression of the purine biosynthetic pathway was indirect and might be mediated through the adenine-responsive Bas1p/Bas2p transcription factors.
Although bacterial and mammalian homologues of Ade4p require an iron-sulfur cluster for activity, in yeast, none of the enzymes involved directly in purine biosynthesis is known to require an iron-sulfur cluster (Mantsala and Zalkin, 1984 ). However, Gcv2p of the glycine cleavage system is a lipoylprotein, and lipoic acid synthase, encoded by LIP5, is an iron-sulfur cluster protein (Ollagnier-deChoudens and Fontecave, 1999 ). Microarray data indicate that LIP5 mRNA levels were repressed 2.2-fold by growth in iron-poor medium, raising the possibility that the availability of lipoic acid may influence the expression of the purine biosynthetic pathway. Alternatively, the iron-dependent regulation may reflect the requirement of glutamine as a cosubstrate for purine biosynthesis, because approximately one-half of the glutamine produced by the cell is consumed through the synthesis of purine nucleotides (Magasanik, 1992 ). The available pools of glutamine may increase when growth in iron-poor medium leads to decreases in Glt1p activity, and increased flux through the purine biosynthetic pathway may be offset by lower levels of purine gene expression.
To identify the sequences that conferred iron-dependent activation of transcription on GLT1 (the FeAS), 700 base pairs of the DNA sequence upstream of the open reading frame were fused to the lacZ reporter gene and introduced into the wild-type strain. We then measured β-galactosidase activity in cells grown in varying concentrations of iron, and found a 5.5-fold increase in activity as the iron concentration was increased from lowest to highest (Figure 8, plasmid 1). To further define the FeAS, a series of 5′ and 3′ deletions were made in the upstream sequences, fused to lacZ, and tested. The sequences from –700 to –550 and from –450 to –260 were deleted without loss of iron-dependent activation, but deletion of the sequences from –550 to –450 resulted in a complete loss of activation (Figure 8, plasmids 2–9). Previous investigators had identified two putative TATA boxes in the GLT1 upstream DNA and shown that the 5′-most TATA box controls the preferred transcription start site (Valenzuela et al., 1998 ). Deletion of the region containing the 5′-most TATA box also resulted in a loss of iron-dependent activation (Figure 8, plasmid 10). Further deletions of the sequences between –550 and –450 indicated that a 44-base pair region from –511 to –468 contained the FeAS (Figure 8, plasmids 11–14). The GLT1 FeAS was not tested as a fusion to the CYC1 “minimal” promoter, because this minimal promoter exhibited iron-dependent activity in the absence of GLT1 sequences (J.H.T., unpublished observations).
The 44-base pair region that contained the FeAS was noted to contain a CGGN15CCG palindrome, and previous investigators have questioned whether this site might have a role in nitrogen-dependent regulation of GLT1 (Valenzuela et al., 1998 ). A CGG palindrome separated by a variable number of nucleotides constitutes a recognition site for a family of fungal transcription factors known as the binuclear Zn-cluster family (Todd and Andrianopoulos, 1997 ). We tested whether the CGG palindrome had a role in the iron-dependent regulation of GLT1 by mutating the half-sites individually and in combination, then measuring their activity as lacZ fusions (Table 1). Mutation of either half-site resulted in a significant loss of iron-dependent β-galactosidase activity, and mutation of both half-sites resulted in a near total loss of activity, indicating a role of the CGG palindromic sequences in the iron-dependent regulation of GLT1.
A set of genes under the control of Aft1p is actively transcribed during growth in an iron-poor environment, and the products of these genes function in the acquisition of iron, mobilization of intracellular iron stores, sequestration of other metals, and uptake of biotin. The processes of biotin uptake and biosynthesis were reciprocally regulated by iron, with uptake being activated when iron was scarce and biosynthesis being activated when iron was abundant. The effect of these changes in transcription was to transfer the essential process of biotin accumulation to an iron independent system when iron was scarce, thereby allowing the cell to conserve iron. A similar phenomenon was observed in the regulation of glutamate synthesis. When iron was scarce, GLT1 transcription was low, and glutamate biosynthesis was accomplished by the iron-independent activities of GDH1 and GDH3. When iron was abundant, GLT1 transcription was high, and glutamate was synthesized from glutamine and α-ketoglutarate. Iron depletion activated genes involved in the uptake of alternative nitrogen sources and the metabolism of amino acids, whereas iron enrichment activated genes involved in purine synthesis. These changes may have occurred in response to changes in glutamate synthase activity. The transcriptional remodeling described here has the effect of shutting off nonessential metabolic pathways that consume iron during periods of iron deprivation. This suggests that the cell does not allow all iron-dependent pathways to become less active during iron deprivation, but rather the cell undergoes a transcriptional remodeling to allow the limited remaining iron to be selectively used in essential iron-requiring processes. These essential processes, such as the synthesis of deoxyribonucleotides, ergosterol, and long chain fatty acids, were not found to be regulated at the mRNA level by the alterations in iron described here.
In E. coli, iron deprivation results in the down-regulation of three iron-sulfur cluster-containing enzymes of the tricarboxylic acid cycle, succinate dehydrogenase (sdhCDAB), aconitase (acnA), and fumarase (fumA), as well as the Fe-dependent superoxide dismutase (sodB), although the homologous enzymes encoded by acnB and fumB are not regulated by iron. This regulation occurs through the activities of the small RNA RyhB, which may destabilize the RNAs transcribed from these loci (Masse and Gottesman, 2002 ). The microarrays that are reported in this study indicated that the iron-sulfur cluster enzymes of the tricarboxylic acid cycle (Sdh2p, Aco1p, and Fum1p) or of other amino acid biosynthetic pathways (Leu1p and Lys4) or Rli1p were not regulated by iron at the transcriptional level, neither were the enzymes involved in heme biosynthesis. However, these experiments were performed using a relatively mild degree of iron deprivation, and additional iron-dependent pathways may be down-regulated by more severe iron deprivation.
Heme is an enzyme cofactor that also acts in the transcriptional regulation of heme-dependent genes. Heme, acting through the HAP family of transcription factors, can induce the expression of a number of genes involved in respiration and aerobic growth, and several of these gene products require heme as a cofactor (Zhang and Hach, 1999 ). The role of heme in the regulation of transcription may be mainly as an indicator of oxygen levels, as heme synthesis is an oxygen-dependent process and these pathways are induced during aerobic metabolism (Kwast et al., 1998 ). However, heme-dependent transcription may also be affected by the availability of iron. Respiration and oxidative phosphorylation require the expression of iron-rich respiratory complexes and yeast exhibit an increased requirement for iron when forced to rely on respiration for energy production. Recent work indicates that transcription of the respiratory cytochrome Cyc1p is repressed during iron deprivation, and this transcriptional repression is due in part to the degradation of regulatory pools of heme by the Aft1p target gene, HMX1 (Protchenko and Philpott, 2003 ). Furthermore, transcription of HEM15 is also repressed by iron deprivation (Lesuisse et al., 2003 ). Thus, iron deprivation may indirectly repress the expression of heme-dependent metabolic pathways.
Metabolic pathways are frequently regulated at the transcriptional level by both substrates that are consumed in the pathway and by the products of the pathway. For example, galactose can activate the expression of a number of genes involved in galactose utilization through the Gal4p transcription factor (Johnston and Carlson, 1992 ), whereas glutamine can repress the transcription of glutamine synthetase through Gln3p (Magasanik, 1992 ) and adenine can repress the transcription of genes in the purine biosynthetic pathway through Bas1p and Bas2p (Daignan-Fornier and Fink, 1992 ). More complex modes of regulation of metabolic pathways also operate in yeast. The transcriptional control of tricarboxylic acid cycle genes shifts from the HAP transcription factors to the heterodimeric Rtg1/Rtg2 transcription factor when mitochondrial function is impaired (Liu and Butow, 1999 ). Here, we have described a mode of regulation in which the availability of a cofactor, iron, can activate, directly or indirectly, the expression of pathways that are dependent on iron-containing prosthetic groups.
The iron-dependent transcriptional activation of GLT1 required sequences that consisted of a CGG inverted repeat (a palindrome) that was separated by 15 nucleotides. This sequence is similar to the DNA binding sites that have been characterized for transcription factors of the binuclear zinc cluster family. Members of this fungal-specific family of transcription factors contain a characteristic CysX2CysX6CysX5–16CysX2CysX6–8Cys motif in the DNA binding domain that coordinates two zinc atoms (Todd and Andrianopoulos, 1997 ). These factors bind as homodimers to inverted, direct, or everted repeats of trinucleotide sequences, especially CGG, that are separated by a variable but defined number of nucleotides. Members of this family include Gal4p and Hap1p, and inspection of the genome of S. cerevisiae has revealed a total of 54 potential members of this gene family, many of which remain uncharacterized (Akache et al., 2001 ). Although the upstream regions of the BIO genes do not contain CGG palindromes identical to that of GLT1, they may be regulated by the same or similar transcription factors, as some diversity in the sequence of the terminal repeats occurs.
We thank Jerry Kaplan and Alan Hinnebusch for generously providing plasmids, Claude Alban for generously sharing reagents, Juergen Stolz for sharing data before publication, and Jerry Kaplan for critically reading the manuscript.
Article published online ahead of print. Mol. Biol. Cell 10.1091/mbc.E03–09–0642. Article and publication date are available at www.molbiolcell.org/cgi/doi/10.1091/mbc.E03-09-0642.
Online version of this article contains supplementary material for some figures. Online version available at www.molbiolcell.org.