The Schizosaccharomyces pombe Corepressor Tup11 Interacts with the Iron-responsive Transcription Factor Fep1*

The Schizosaccharomyces pombe fep1+ gene encodes a GATA transcription factor that represses the expression of iron transport genes in response to elevated iron concentrations. This transcriptional response is altered only in strains harboring a combined deletion of both tup11+ and tup12+ genes. This suggests that Tup11 is capable of negatively regulating iron transport gene expression in the absence of Tup12 and vice versa. The tup11+- and tup12+-encoded proteins resemble the Saccharomyces cerevisiae Tup1 corepressor. Using yeast two-hybrid analysis we show that Tup11 and Fep1 physically interact with each other. The C-terminal region from amino acids 242 to 564 of Fep1 is required for interaction with Tup11. Within this region, a minimal domain encompassing amino acids 405-541 was sufficient for Tup11-Fep1 association. Deletion mapping analysis revealed that the WD40-repeat sequence motifs of Tup11 are necessary for its interaction with Fep1. Analysis of Tup11 mutants with single amino acid substitutions in the WD40 repeats suggested that the Fep1 transcription factor interacts with a putative flat upper surface on the predicted β-propeller structure of this motif. Further analysis by in vivo coimmunoprecipitation showed that Tup11 and Fep1 are physically associated. In vitro pull-down experiments further verified a direct interaction between the Fep1 C terminus and the Tup11 C-terminal WD40 repeat domain. Taken together, these results describe the first example of a physical interaction between a corepressor and an iron-sensing factor controlling the expression of iron uptake genes.

Iron is required for a number of biological functions in most life forms, from microbes to mammals (1,2). Because its chemistry allows a reversible one-electron oxidation-reduction reaction, iron plays a crucial role in electron-transfer reactions (3). On the other hand, the redox active nature of iron can make it toxic due to its ability to unleash highly reactive species in the presence of oxygen that can damage biological cellular compo-nents (4). Because of its essential but toxic nature, highly regulated homeostatic controls have evolved across different species in order to maintain intracellular concentration of iron at levels needed for essential biochemical processes while preventing its accumulation to toxic levels (5).
Although iron is an abundant transition metal on earth, under aerobic conditions, it is usually present as insoluble ferric hydroxides (6). Because of its low bioavailability, organisms have developed a variety of mechanisms for iron acquisition including secretion and utilization of siderophores, ferrireductase activity at the cell surface, and heme-iron introduction inside the cell (7)(8)(9)(10). In the model organism Schizosaccharomyces pombe, two pathways for iron uptake have been identified (11)(12)(13). The first one consists of a siderophore-iron transport system. Although S. pombe does not produce siderophores, the fission yeast can utilize siderophorebound iron complexes produced by other microorganisms using three transmembrane proteins, Str1, Str2, and potentially Str3 (11). Of these three siderophore transporters, Str1 exhibits specificity for ferrichrome-iron, while Str2 is specific for ferroxiamine B-iron and to a lesser extent ferrichrome-iron. Although Str3 may participate in the mobilization of iron bound to siderophores, its substrate specificity has not been determined (11). The second pathway for iron assimilation in S. pombe involves three components, Frp1, Fio1, and Fip1 (12,13). The Frp1 protein reduces Fe 3ϩ to Fe 2ϩ at the cell surface, rendering iron accessible to iron-binding extracellular ligands found in the iron-transport proteins in the plasma membrane (13,14). Once reduced, iron is transported by two high affinity iron uptake proteins encoded by the fio1 ϩ and fip1 ϩ genes (12). The Fio1 protein is a multicopper ferroxidase that converts Fe 2ϩ to Fe 3ϩ , which is then transported by Fip1, a transmembrane iron permease (12). The frp1 ϩ , fio1 ϩ , and fip1 ϩ genes are regulated by the cellular need for iron. When cells are grown under iron-deficient conditions, frp1 ϩ , fio1 ϩ , and fip1 ϩ mRNA levels are induced. In contrast, iron-replete conditions repress the expression of these genes (12,13,15). As expected, str1 ϩ , str2 ϩ , and str3 ϩ gene expression is up-regulated under conditions of iron starvation and down-regulated under conditions of iron repletion (11). It was recently shown that the transcription factor Fep1 1 mediates the iron-dependent repression of str1 ϩ , str2 ϩ , str3 ϩ , frp1 ϩ , fio1 ϩ , and fip1 ϩ transcription (11,16). The N-terminal region of Fep1 is highly similar to the N-terminal regions of the Urbs1, SREA, and SRE proteins that have been shown to negatively regulate, in an iron-dependent manner the siderophore biosynthesis pathways of Ustilago maydis, Aspergillus nidulans, and Neurospora crassa, respectively (17)(18)(19)(20)(21)(22). This region of Fep1 harbors two consensus GATA-type zinc finger motifs that has been shown to be required for DNA binding to the consensus sequence 5Ј-(A/T)GATAA-3Ј (16). Thus, Fep1 is related to a family of GATA-type transcription factors (23).
Interestingly, a mutant fission yeast strain with deletions in both tup11 ϩ and tup12 ϩ genes exhibited a fio1 ϩ gene expression that was highly derepressed and unresponsive to repression by iron (16). Elimination of either Tup11 or Tup12 alone was not sufficient to annihilate the iron-mediated repression of fio1 ϩ (16). These observations suggest that, tup11 ϩ and tup12 ϩ , known to encode transcriptional corepressors, are functionally redundant in down-regulating the expression of the iron transport genes. Recent work has demonstrated that Tup11 assembles with itself or the Tup12 protein (24). The S. pombe Tup11 and Tup12 proteins exhibit 39.9 and 43.5% identity to Saccharomyces cerevisiae Tup1, respectively (25). Based on the extended homology of Tup11 and Tup12 to Tup1, putative functional domains have been designated for the Tup1 homologues in S. pombe (26). The N-terminal 70 and 87 amino acids of the S. pombe Tup11 and Tup12 proteins, respectively, bear homology to a similar region in Tup1 that is known to interact with Ssn6. Like Tup1, both Tup11 and Tup12 contain a homologous region located in the middle part of their Nterminal halves that is required for interaction with histones H3 and H4. Consistent with this observation, it has been shown that Tup11 associates with histones H3 and H4 of S. cerevisiae in vitro (26). Within their C termini, Tup11 and Tup12 contain seven copies of a repeated amino acid motif, named WD40 repeat (27), found in Tup1 and other Tup1-like corepressors as well as in other proteins that are unlinked to transcription like the ␤-subunit of the trimeric G-proteins (28). Based on the x-ray crystal structures of three proteins with seven WD40 repeats, including the Tup1 corepressor, each repeat folds into four antiparallel ␤ strands, forming a blade structure (29 -31). Each WD40 repeat unit is interconnected to the other by a loop to form a seven-bladed ␤-propeller, structure that is highly symmetrical and assumes an overall donut shape (29,31). Although biochemical data has shown that the sides of the ␤-propeller are capable of establishing protein-protein contacts, the flat upper surface has been shown to bind to some proteins (30,32).
Previous studies have shown that S. cerevisiae Tup1 and Ssn6 proteins physically interact in vivo to form a corepressor complex (33). Furthermore, it has been demonstrated that Tup1 forms a tetramer with one Ssn6 molecule (34). Although the Ssn6-Tup1 corepressor complex is incapable of binding to DNA, it is attracted to the regulatory regions of different genes by interacting with transcription factors that function in specific metabolic pathways (28). A number of potential mechanisms by which the Ssn6-Tup1 complex can repress gene expression have been described (28). First, upon location to the target promoter, the corepressor complex may inhibit the activities of the basal transcription factors that activate gene expression through the RNA polymerase II holoenzyme (35). Second, the Ssn6-Tup1 may promote the formation of a repressive chromatin structure (36,37). Third, the formation of a complex between Tup1 and the mRNA 5Ј-triphosphatase Cet1 may abrogate mRNA stability and translation through the association of Tup1 with the mRNA capping enzyme (38).
Based on our previous findings that the S. pombe Tup11 protein has the potential to regulate the expression of the fio1 ϩ iron transport gene (16), we sought to determine its ability to interact with the iron-sensing DNA-binding repressor Fep1. Using two-hybrid analysis, we demonstrate that Tup11 interacts with the C-terminal region of Fep1, but not with the two adjacent zinc fingers in the N-terminal region. The WD40repeat domain in Tup11 is necessary for its interaction with Fep1. In addition, we showed that the amino acid residues Tyr 362 and Leu 542 , located on the same surface of the predicted ␤-propeller structure of the Tup11 WD40 repeats, are critical for the Tup11-Fep1 interaction. When coexpressed in fission yeast, the Tup11 and Fep1 proteins were detected in a heteroprotein complex by coimmunoprecipitation experiments. Furthermore, in vitro protein binding analysis using fusion proteins that were expressed in and purified from Escherichia coli revealed that the Tup11 and Fep1 C-terminal regions directly interact with each other. Taken together, our findings indicate that Tup11 and Fep1 are components of a heteromeric complex that is required for transcriptional down-regulation of genes that are critical for iron acquisition in fission yeast.
Two-hybrid Analysis-To study the association between Tup11 and Fep1, the complete or truncated versions of the tup11 ϩ open reading frame were inserted downstream of and in-frame to the E. coli lexA gene as bait. The prey plasmid, pVP16 (45,47), contains the VP16 acidic activation domain followed by a polylinker in which different fragments of the fep1 ϩ gene were introduced. Each L40 transformant strain harboring the indicated bait and prey plasmids was tested for the interaction of the two fusion proteins using previously described standard procedures (48). For quantitative measurements, ␤-galactosidase activity was determined using early logarithmic phase cultures (OD 600 nm of ϳ0.5) of yeast cells transformed with the indicated plasmids. 4-ml samples were withdrawn from the cultures, and the cells were harvested, washed with sterile water, and resuspended in 700 l of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , pH 7.0, 10 mM KCl, 1 mM MgSO 4 , 50 mM ␤-mercaptoethanol). The cells were permeabilized by adding 50 l of chloroform and 50 l of 0.1% SDS. After the cell suspension was vortex-mixed for 10 s, 200 l of 4 mg/ml o-nitrophenyl-␤-D-galactopyranoside was added to each sample. Following a 10-min incubation at 30°C, 350 l of 1 M Na 2 CO 3 was added to stop the reactions. After clarification by centrifugation at 4°C, the A 420 nm was measured within the linear response range and expressed in standard units (49). Values shown are the average of triplicate assays of three independent transformants. In addition to liquid ␤-galactosidase assays, a riboprobe derived from the plasmid pKSlacZ (50) was used to monitor the steadystate levels of lacZ mRNAs from the integrated (lexAop) 8 -lacZ reporter construct in the L40 strain. Total RNA was extracted by the hot phenol method as described previously (51). RNase protection analyses were carried out as described previously (52) using the plasmid pKSACT1 (50) to probe ACT1 mRNA as an internal control. For protein expression analysis of the LexA-Tup11 and VP16-Fep1 fusion protein derivatives, and PGK, the following antisera were used for immunodetection: monoclonal anti-LexA antibody 2-12; monoclonal anti-VP16 antibody 1-21 (both from Santa Cruz Biotechnology, Santa Cruz, CA); and monoclonal anti-PGK antibody (Molecular Probes, Eugene, OR).
Protein Coimmunoprecipitation Experiments-To determine whether Fep1 interacts with Tup11 in fission yeast cells, plasmid pNTAP-fep1 ϩ codons2-564 was constructed as follows. The fragment containing the fep1 ϩ gene (codons 2-564) was isolated by PCR using Pfu Turbo polymerase (Stratagene, La Jolla, CA), purified, and cloned into the BamHI and NotI sites of pREP1-NTAP (53). Plasmid pNTAP-fep1 ϩ codons319 -564 was generated by using a similar approach, except that the DNA fragment harboring the fep1 ϩ gene corresponded to codons 319 through 564. Plasmid pGST-tup11 ϩ containing the fulllength S. pombe tup11 ϩ gene fused downstream of and in-frame to the GST coding region was created by inserting a BamHI-SacI fragment encompassing the entire coding sequence of the tup11 ϩ gene into the corresponding sites of pAAUGST (54). The coding region of the tup11 ϩ gene that contains either the mutation Tyr 362 3 Cys or Leu 542 3 Ser was also swapped for an identical DNA region into the pGST-tup11 ϩ plasmid, creating pGST-tup11Tyr 362 3 Cys and pGST-tup11Leu 542 3 Ser. For co-immunoprecipitation experiments, JY741 cells were cotransformed with pGST-tup11 ϩ or mutant derivatives or pAAUGST, and pNTAP-fep1 ϩ codons2-564, or pNTAP-fep1 ϩ codons319 -564. Cells were grown to OD 600 nm of 1.0 in the absence of thiamine after 18 -20 h in culture. Cells were broken with glass beads in buffer G (30 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10 mM EDTA, pH 8.0, 5 mM dithiothreitol, 10 mM ␤-mercaptoethanol, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride) in the presence of a complete protease inhibitor mixture (P8340, from Sigma) followed by a centrifugation at 3500 rpm at 4°C for 5 min. Protein extracts (ϳ3 mg of total protein) were incubated for 16 h with 50 l of glutathione-Sepharose 4B beads (Amersham Biosciences, Uppsala, Sweden). Beads were washed five times with 1.5 ml of buffer G, with the beads completely transferred to a fresh microtube before the last wash. The immunoprecipitates were resuspended in 50 l of SDS loading buffer, incubated for 5 min at 95°C, and resolved by electrophoresis on a 9% SDS-polyacrylamide gel. Proteins were electrophoretically transferred to nitrocellulose membranes. Blots were incubated with primary antibody for 2 h at room temperature. For protein analysis of GST-Tup11, NTAP-Fep1, and PCNA, the following primary antisera were used: polyclonal anti-GST antibody Z-5 (Santa Cruz Biotechnology); polyclonal anti-mouse IgG antibody (ICN Biomedicals, Aurora, OH); and monoclonal anti-PCNA antibody PC10 (Sigma). After immunoblots were washed in TBS (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% bovine serum albumin), membranes were incubated with appropriate horseradish peroxidase-conjugated secondary antibodies (Amersham Biosciences) for 30 min, washed, and developed with enhanced chemiluminescent detection reagents.
Expression and Purification of the MBP-Fep1 and GST-Tup11 Fusion Proteins in E. coli-The plasmid pMAL-fep1 ϩ (codons 319 -564) was constructed by cloning a 738-bp BamHI-PstI DNA fragment containing the C-terminal 319 -564 codons of Fep1 into the same sites of pMAL-c2X (New England BioLabs, Beverly, MA). The expression and purification of MBP-319 Fep1 564 fusion protein from E. coli were carried out essentially as described previously (16). Plasmids pGEX-tup11N (codons 1-298) and pGEX-tup11C (codons 297-614) were described previously (38). These fusion plasmids were transformed into E. coli TB1 cells. Fresh transformants of TB1 cells were grown to early growth phase (OD 600 nm of 0.5) in Luria-Bertani (LB) medium supplemented with 0.2% glucose and 100 g/ml ampicillin. At this early logarithmic growth phase, the cells were induced with 0.2 mM isopropyl-␤-D-thiogalactopyranoside for 2 h at 25°C. Purification of the GST-Tup11 fusion proteins from E. coli cells was performed using glutathione-Sepharose 4B beads following the manufacturer's protocol, with minor modifications. Cells were broken with glass beads in buffer G as described above. Protein extracts (ϳ1.4 mg of total protein) were incubated for 30 min at 4°C with 50 l of glutathione-Sepharose 4B using siliconized microtubes (Fisher). Beads were washed extensively with buffer G as described above. The different GST fusion proteins bound to the beads were eluted with 2-bed volumes of 5 mM glutathione in 30 mM Tris-HCl (pH 8.0), separated by SDS-PAGE, and visualized by Coomassie Brilliant Blue stain, or used for Western blot analysis. Purified fusion proteins produced in E. coli were used for GST pull-down experiments. The MBP and MBP-Fep1 fusion proteins (ϳ3.2 g) were incubated with equivalent amount of GST-Tup11N, GST-Tup11C, or GST proteins bound to glutathione-Sepharose 4B beads. After 16 h at 4°C, the resin was sedimented, and the supernatant removed. The resin was washed five times with 1.5 ml of buffer G. The bound protein complexes were resuspended in Laemmli buffer and heated at 95°C for 5 min prior to SDS-polyacrylamide gel electrophoresis and immunoblot analysis.

Tup11
Associates with the S. pombe Fep1 Iron-sensing Transcription Factor-We have previously shown by genetic analysis that Fep1 requires the presence of the cofactor Tup11 or Tup12 to repress the expression of the iron transport genes under conditions of iron repletion (16). Based on this observation, we examined the possibility that Tup11 physically interacts with Fep1. We performed two-hybrid analyses using the full-length tup11 ϩ gene fused to the LexA coding region as bait and different coding regions of the fep1 ϩ gene fused to the coding region of the VP16 acidic activation domain as prey. Interestingly, coexpression of the full-length Fep1 fused to VP16 with the LexA-Tup11 fusion protein, produced high levels of ␤-galactosidase activity (ϳ448 Miller units) (Fig. 1), indicating a strong interaction between these proteins. To gain insight into the specific region of Fep1 that interacts with Tup11, we first tested an N-terminal segment of Fep1 (residues 2-281) that contains its DNA-binding domain (16). This region, which bears two Cys 2 /Cys 2 -type zinc finger motifs failed to interact with the LexA-Tup11 fusion protein. In contrast, the C-terminal portion of Fep1 (residues 242-564) interacted with the LexA-Tup11 fusion protein. Although the induction of lacZ was lower by ϳ30% compared with the full-length Fep1, a clear transactivation of the reporter gene expression was observed (Fig. 1). Further removal of the amino-terminal residues 242-318 of Fep1 had little effect (ϳ8% decrease) on the activity of the reporter gene. Based on this observation, we used the chimeric protein that contains the N-terminal 78 amino acids of VP16 and the C-terminal 245 amino acids of Fep1 (residues 319 -564) in subsequent analyses. To ensure that the fusion proteins were expressed in the transformed cells, immunoblot analyses of protein extracts were performed using anti-LexA and anti-VP16 antibodies (Fig. 1C). Although all of the VP16-Fep1 fusion proteins used in this study were consistently detected by immunoblotting, we were unable to detect the VP16 polypeptide alone, perhaps, owing to its low predicted molecular weight of ϳ8 kDa. As shown in Fig. 1D, the association between the LexA-Tup11 and VP16-Fep1 fusion proteins in bakers' yeast cells using a two-hybrid system was not modulated by cellular iron status. Indeed, the LexA-Tup11 and VP16-Fep1 fusion proteins that specifically associate with each other gave a constitutive steady-state level of lacZ mRNA as assayed by RNase protection experiments under both ironstarvation and iron-replete conditions.
To further delineate the region of Fep1 that is necessary for physical interaction with Tup11, we constructed six chimeric VP16-Fep1 molecules, denoted VP16-242 Fep1 564 , VP16-319 Fep1 564 , VP16-360 Fep1 564 , VP16-390 Fep1 564 , VP16-405 Fep1 564 , and VP16-432 Fep1 564 , all of which contain the intact C terminus of Fep1 with progressive truncations from amino acid Asp 242 . As shown above (Fig. 1), when co-expressed with LexA-Tup11, the VP16-242 Fep1 564 and VP16-319 Fep1 564 fusion proteins produced high levels of ␤-galactosidase activity. When the N-terminal region of 319 Fep1 564 was truncated to position Val 360 , the ␤-galactosidase activity decreased by ϳ42%. Further deletions to positions Thr 390 and Ser 405 diminished the ␤-galactosidase activity by ϳ61 and ϳ80%, respec-tively, compared with VP16-319 Fep1 564 , while deletion to position Ala 432 completely abolished the activity of the reporter gene (Fig. 2). We then tested if the amino acids located at the C terminus of Fep1 were involved in the interaction with Tup11. We cloned a DNA fragment of the fep1 ϩ gene corresponding to amino acids 360 -541 into pVP16 and tested for interaction with LexA-Tup11. Interestingly, the chimeric VP16-360 Fep1 541 molecule gave ϳ24% increase in the activity of the reporter gene compared with the VP16-319 Fep1 564 fusion protein (Fig.  2). However, when the C terminus of the VP16-360 Fep1 541 fusion protein was further truncated by removing either the last 50 or 84 amino acids (VP16-360 Fep1 491 or VP16-360 Fep1 457 ), the ␤-galactosidase activity was extinguished. Based on the results of our analysis of reporter gene expression, we predicted that the C-terminal region of Fep1 from residues 405 to 541 would be sufficient for interaction between Tup11 and Fep1. Indeed, as shown in Fig. 2A, VP16-405 Fep1 541 conferred expression on the (lexA-operator) 8  tein. To gain additional insight into the region that is responsible for interaction with Tup11, two other chimeric proteins were generated using a portion of the Fep1 protein between residues 242 and 457 (VP16-242 Fep1 457 ), and a second one between residues 319 and 457 (VP16-319 Fep1 457 ). When the ␤-galactosidase activity of VP16-242 Fep1 457 and VP16-319 Fep1 457 coexpressed with LexA-Tup11 was assayed, only very weak levels of activity relative to the VP16-319 Fep1 564 fusion protein were detected. As shown in Fig. 2B, all the fusion proteins tested for two-hybrid interactions were expressed as confirmed by immunoblot analyses. Taken together, we conclude that the C-terminal segment of Fep1 from amino acids 242-564 is required for interaction with Tup11 in an ironindependent manner by two-hybrid assay. Furthermore, within this region, a domain corresponding to amino acids 405-541 constitutes a minimal module, that is sufficient for interaction between Tup11 and Fep1.
The WD40 Motif Repeats of Tup11 Are Necessary for Its Interaction with the C-terminal Region of Fep1-Our studies on Fep1 prompted us to determine the region on Tup11 that is required for interaction with Fep1. Truncations were created from the C-terminal end of the LexA-1 Tup11 614 fusion protein (Fig. 3A). The first three constructs in which the last 522, 313, and 258 amino acids of Tup11 were removed showed no ␤-ga-lactosidase activity when coexpressed with VP16-319 Fep1 564 (Fig. 3B). These results suggest that the predicted Ssn6-binding region (amino acids 1-69) and another region that is potentially required for interaction with histones H3 and H4 (amino acids 69 -301) were not needed to establish contacts with the Fep1 C-terminal region. Furthermore, when the first WD40 sequence motif (amino acids 301-356) was added to the C terminus of LexA-1 Tup11 301 , no activity was observed. LexA-69 Tup11 301 and LexA-69 Tup11 356 also exhibited no ␤-galactosidase activity. Moreover, the presence of only one WD40 motif (LexA-301 Tup11 356 ) resulted in the loss of (lexA-operator) 8 -lacZ expression. Thus, these results suggest that the region encompassing amino acids 301-614 of Tup11, which contains seven WD40-repeat domains, is required for interaction with the Fep1 C-terminal region. As shown in Fig. 3C, all of the LexA-Tup11 fusion derivatives were expressed as verified by immunoblot analysis. The results from the two-hybrid analyses were further confirmed by assessing the activation or lack of lacZ reporter gene expression as a function of the iron status of the cells by RNase protection assay (Fig. 3D).
Given the results from the analysis of the truncations generated from the C-terminal end of Tup11, we next tested constructs encoding various amino-terminal deletions in Tup11 for their ability to interact with VP16-319 Fep1 564 (Fig. 4). The

FIG. 2. Minimal C-terminal region of Fep1 required for interaction with LexA-Tup11 fusion protein by two-hybrid assay. A,
LexA-Tup11 fusion protein was co-expressed with a series of VP16-Fep1 truncations in the L40 strain. Positive interactions between the proteins were detected by liquid ␤-galactosidase assays. B, the LexA-Tup11 fusion protein and truncated versions of VP16-Fep1 fusion proteins were detected by immunoblotting using either anti-LexA or anti-VP16 antibody. As a control, total extract preparations were probed with anti-PGK antibody.
N-terminal deletions LexA-134 Tup11 614 , LexA-192 Tup11 614 , LexA-245 Tup11 614 , LexA-267 Tup11 614 , and LexA-284 Tup11 614 showed high levels of ␤-galactosidase activity when coexpressed with VP16-319 Fep1 564 . Although expression of the (lexA-operator) 8 -lacZ reporter by LexA-245 Tup11 614 , and LexA-267 Tup11 614 were lower by ϳ51 and ϳ34%, respectively, compared with the full-length LexA-1 Tup11 614 , the reporter gene was still highly expressed (Fig. 4D). All deletion derivatives were detected by immunoblotting (Fig. 4C). Furthermore, cells carrying this series of pLexA-Tup11 truncations with pVP16-319 Fep1 564 were assayed for iron-regulated expression of lacZ mRNA (Fig. 4D). The results showed that the interactions between these proteins in an S. cerevisiae two-hybrid system were independent of the iron status of the cells. We therefore conclude, on the basis of these data, that the WD40 repeats found in the C terminus of Tup11 specifically interact with the C-terminal region of Fep1 that includes amino acids 319 -564.
Conserved Residues in S. pombe Tup11 WD40 Repeats 2 and 6 Are Essential for the Association between Tup11 and the C-terminal Region of Fep1-The primary amino acid sequence of Tup11 shows that it has seven WD40 repeat motifs (amino acids 298 -614) (26). As shown in other proteins containing seven WD40 repeats (27), these domains can fold into a sevenbladed ␤-propeller structure where each blade is a rigid skeleton formed by four antiparallel internal ␤-strands. According to this model, the overall structure (or ␤-propeller) constitutes a platform with a relatively flat top and bottom. On the top, conserved amino acids are exposed to provide contact points with other proteins (55). To determine if specific residues within this region play a role in the interaction between Tup11 and Fep1, two amino acids of the Tup11 protein that are highly conserved in the C-terminal WD40 domains of S. cerevisiae Tup1 (31,32), Caenorhabditis elegans UNC-37 (56), Drosophila Groucho, and human TLE1 and TLE2 proteins (29) were individually altered by site-directed mutagenesis. These residues correspond to Tyr 362 and Leu 542 in Tup11. Mutation of Tyr 362 to cysteine and Leu 542 to serine were chosen based on previous genetic analyses that identified point mutations in proteins containing WD repeats that disrupt protein-protein interactions (29,32,57,58). When Tyr 362 was changed to cysteine, the mutant protein failed to interact with Fep1 (Fig. 5). Similarly, when the Leu 542 3 Ser mutation was tested for interaction with pVP16-319 Fep1 564 by two-hybrid analysis, no ␤-galactosidase activity was detected (Fig. 5B). Consistent, with the loss of ␤-galactosidase activity observed in cells harboring the mutant proteins (Fig. 5C), lacZ mRNA levels were undetectable by RNase protection analysis (Fig. 5D). The predicted structural model of the WD repeats in Tup11 obtained using Swiss-Model (59 -61) showed that both Tyr 362 and Leu 542 are extending out on the top surface of WD repeats 2 and 6 of Tup11, respectively (Fig. 6). Although only two point mutations of Tup11 were analyzed, our results suggest that Fep1 interacts with the flat upper surface of the predicted Tup11 ␤-propeller. The involvement of other residues located on the other side of the putative Tup11 propeller or in the other blades in mediating the Tup11-Fep1 contacts must await a fine mapping dissection of the amino acids that form the propeller structure.
Tup11 and Fep1 Proteins Found in a Complex in S. pombe-Given the fact that in a two-hybrid assay LexA-Tup11 interacts with VP16-Fep1, we examined the physical association between Tup11 and Fep1 in vivo. The GST protein was fused to the N terminus of the Tup11 protein or its mutant derivatives.
The TAP tag was added in-frame to both the N terminus of the full-length Fep1 and to the C-terminal residues 319 -564 of Fep1. We coexpressed GST-Tup11 with TAP-Fep1 in S. pombe and performed a GST pull-down experiment using glutathione-Sepharose beads that selectively binds GST-tagged proteins. Analysis of the proteins bound to the beads by immunoblotting with purified rabbit anti-mouse IgG showed that the TAP-Fep1 was present in the bound fraction (Fig. 7). TAP-Fep1 was not observed in the bound fraction of control cells expressing only TAP-Fep1 with GST alone (Fig. 7). When a reciprocal coimmunoprecipitation with IgG-Sepharose was performed and probed with the anti-GST Z-5 antibody, the full-length GST-Tup11 fusion protein was also specifically precipitated with TAP-Fep1. 2 To ascertain the specificity of the pull-down experiments, the total cell lysates and immunoprecipitates or bound fractions were analyzed by immunoblotting using an antibody directed against PCNA, a soluble protein like the GST-Tup11 and TAP-Fep1 fusion proteins. As shown in Fig. 7, PCNA is present in the total cell extracts but not in the immunoprecipi-2 S. Znaidi, B. Pelletier, and S. Labbé, unpublished data. tates. This suggests that the GST-Tup11 and TAP-Fep1 fusion proteins specifically interact with each other to form a stable heteroprotein complex that can be coimmunoprecipitated from whole cell extracts. To assess GST or GST-Tup11 steady state levels, immunoblot analysis of the protein preparations and coimmunoprecipitation reactions was conducted using anti-GST antibody (Fig. 7). We then tested if the mutations of Tyr 362 to cysteine and Leu 542 to serine affected the ability of Tup11 to interact with Fep1 in S. pombe. We conducted GST pull-down experiments using protein lysates that were prepared from cells coexpressing TAP-319 Fep1 564 with either GST-Tup11Tyr 362 3 Cys or GST-Tup11Leu 542 3 Ser. As shown in Fig. 8, the interaction between Tup11 and Fep1 in vivo was extinguished when the Tyr 362 or Leu 542 residue within the C-terminal WD40-repeat domain of Tup11 was mutated. Taken together, these data reveal that S. pombe Tup11 and Fep1 are physically associated in vivo, and that at least the Tup11 Tyr 362 and Leu 542 residues are essential for this association.
Association of the Tup11 and Fep1 C-terminal Regions in Vitro-The use of two-hybrid analyses in S. cerevisiae or GST pull-down or coimmunoprecipitation experiments on S. pombe cell extracts cannot rule out the possibility that the interaction between Tup11 and Fep1 was mediated by an intermediate yeast protein. To determine if these proteins directly interact with each other, we examined the ability of purified, bacterially expressed forms of these proteins to physically associate in an in vitro system. The C-terminal region of Fep1 from residues 319 to 564 was fused to the maltose-binding protein. The fusion protein was purified using two rounds of one-step affinity chromatography based on the affinity of MBP for maltose (62). The first 298 amino acids of Tup11 and the C-terminal 317 amino acids of Tup11 were separately expressed as bacterial fusion proteins with GST. The purified MBP protein or MBP-319 Fep1 564 fusion protein was incubated with GST-1 Tup11N 298 , GST-297 Tup11C 614 , or GST that had been immobilized on glutathione-Sepharose resin. Following washing, bound GST fusion proteins, and any associated proteins, were eluted with glutathione and analyzed by immunoblotting using anti-MBP antibody. As shown in Fig. 9, MBP-319 Fep1 564 , but not MBP alone, was bound to the immobilized GST-297 Tup11C 614 . No interaction was detected when MBP-319 Fep1 564 was incubated with GST-1 Tup11N 298 or GST alone. A blot probed with anti-GST Z-5 antibody confirmed that all reactions contained the GST protein, GST-1 Tup11N 298 , or GST-297 Tup11C 614 fusion protein. Interestingly, we note that the GST-297 Tup11C 614 fusion protein exhibits a slightly faster electrophoretic mobility to that observed for the GST-1 Tup11N 298 , despite a higher predicted molecular mass. Perhaps, the nature of the WD40repeat domain affects its mobility. Consistent with the results obtained by two-hybrid analysis, GST pull-down, and coimmu- The sequence of the WD40 regions in Tup1 that were used to model the structure of the Tup11 WD40 repeats contains internal deletions (indicated by a triangle below) because these regions were not structurally interpretable (31). One deletion corresponding to residues 514 -528 was made in noprecipitation experiments using S.pombe cell extracts, the C-terminal WD40 repeat domain of Tup11 is required for interaction with the C terminus of the Fep1 protein. These data clearly indicate that Tup11 and Fep1 directly interact with each other. No other protein is required for this association. DISCUSSION In S. pombe, the reductive iron transport genes frp1 ϩ , fio1 ϩ , and fip1 ϩ are transcriptionally repressed by iron and induced by iron deprivation (12,13,15). Similarly, str1 ϩ , str2 ϩ , and str3 ϩ , whose products are involved in uptake of siderophorebound iron are regulated by iron at the transcriptional level in the same direction (11). Iron-mediated repression of the abovementioned genes requires a wild-type DNA-interacting Fep1 transcription factor (11,16). To further understand the mechanism of Fep1 repression, we investigated the possibility that Fep1 interacts with other transcription factors to repress ironresponsive genes. Because deletion mutations in the transcriptional corepressors tup11 ϩ and tup12 ϩ were phenotypically similar to mutations in fep1 ϩ , resulting in robust transcription of the iron transport genes even in the presence of high iron concentrations (16), we examined the possibility that these proteins are physically associated. Here, we demonstrate that Fep1 interacts specifically and directly with the Tup11 protein. This represents the first example of a corepressor and an ironsensing factor that physically associate to regulate the expression of iron assimilation genes. Consistent with these results, it has been demonstrated that iron-dependent gene regulation in Candida albicans requires a corepressor protein orthologous to S. cerevisiae Tup1 (63,64). Although no Fep1 homolog has been reported in C. albicans, putative negative-acting GATA factorencoding genes exist according to the C. albicans Genome Project Database. In S. cerevisiae, the Tup1 corepressor is part of a complex that is attracted by repressor proteins to the promoter regions of diverse genes that function in specific pathways (28). For example, the ␣2 DNA-binding factor brings the Tup1 corepressor complex to two distinct sets of genes to control mating type silencing (65). Recently, the Tup1 corepressor complex of S. cerevisiae has been implicated in the down-regulation of siderophore-bound iron transport (66), however, the mechanism by which the Tup1 corepressor complex is delivered to the siderophore-iron transporter promoters is unclear. In S. pombe, besides the iron transporter genes, only the fbp1 ϩ and cta3 ϩ genes encoding fructose 1,6-biphosphatase and a putative Ptype ATPase transporter, respectively, have been identified as targets for Tup11-Tup12-mediated repression. While the Mig1 homolog in S. pombe, designated Scr1, has been implicated as the DNA-binding partner of the Tup11-12 corepressor complex, the DNA-interacting protein that recruits the corepressor complex to the cta3 ϩ promoter has not been determined (24,25). Although we have detected that the S. pombe Tup12 also interacts with Fep1 by two-hybrid analysis and coimmunoprecipitation experiments, we have not mapped the location of the minimal region in Tup12 that is necessary for Tup12-Fep1 interaction. 2 Deletion mapping studies of VP16-2 Fep1 564 fusion protein have shown that the C-terminal amino acids from residues 242 Tup11, allows the propeller of Tup11 (B and C, shown in green) to overlay remarkably well with that of Tup1. Predicted regions of Tup11 that differ from Tup1 are shown in red (B and C). Using the DeepView program (59), ribbon representations (B and C) of the three-dimensional structure of Tup11 were obtained. B, view from the top surface of the ␤-propeller. The blades are numbered according to the Tup1 propeller structure. The positions of the mutations that affect the Tup11-Fep1 interaction are indicated with arrows. C, view from the side highlighting the fact that Tyr 362 and Leu 542 residues are exposed and located on the same surface of the ␤-propeller. D, space-filled representation of the C-terminal WD40 domain of Tup11 emphasizing the accessibility of the Tyr 362 and Leu 542 residues (shown in green) for interaction with Fep1. FIG. 7. Tup11 and Fep1 physically interact in S. pombe. Strain JY741 was transformed with TAP-tagged full-length Fep1 or TAP-tagged 319 Fep1 564 (retaining amino acids 319 -564) containing plasmids expressing either GST-tagged Tup11 or GST alone. Extracts (Total) were subjected to GST pull-down using glutathione-Sepharose. The proteins bound to the beads were heated, loaded, and separated on 9% SDS-polyacrylamide gels, blotted to nitrocellulose, and visualized with the anti-mouse IgG antibody. TAP-2 Fep1 564 and TAP-319 Fep1 564 are indicated with arrowheads. As specific controls, aliquots of whole cell extracts and bound fractions were probed with anti-PCNA antibody and anti-GST antibody (shown with arrowheads). Extracts (Total) loaded on gels represent ϳ2% of the whole extracts used in the binding reactions. WB, Western blot.
to 564 are required for physical interaction of Fep1 with the corepressor Tup11. Within this region of Fep1 lies a domain from residues 405 to 541 that constitutes the minimal module for this interaction. Interestingly, this minimal module contains several leucine-proline dipeptide repeats. One of these repeats, 414 Leu-Pro-Pro-Ile-Leu-Pro 419 , is highly conserved in other iron responsive transcriptional repressors including Urbs1 from U. maydis (22), Srea from A. nidulans (18), and Srep from P. chrysogenum (67). Furthermore, in S. cerevisiae, this dipeptide repeat is also found in the Mig1 and Rox1 sequence-specific DNA binding transcription factors necessary for glucose and oxygen repression, respectively (68,69). In fact, it has been suggested that this di-leucine-proline motif may play a role in protein-protein interactions with the Tup1 corepressor complex (68). The contribution of these residues or other residues in the C-terminal region encompassing amino acids 405-541 of Fep1 to the interaction between Fep1 and Tup11 must await a comprehensive dissection of that minimal module.
A hallmark of the fungal GATA factors Fep1 (16), Urbs1 (70), Srea (18), Sre (71), and Srep (67) is that, like the vertebrate GATA transcription factors, two Cys 2 -Cys 2 -type zinc finger motifs are found in their DNA-binding domain. In Fep1, these motifs are within the N-terminal portion of the protein (residues 12-220). In the mammalian GATA-1 transcription factor, the C-terminal zinc finger has been shown to be necessary and sufficient for DNA binding (72,73). Although the N-terminal zinc finger has been shown to increase stability and influence specificity of the binding of GATA-1-DNA complexes, this zinc finger has also been demonstrated to mediate specific proteinprotein interactions (73)(74)(75)(76)(77). Using two-hybrid analyses and in vitro protein binding experiments, we showed that Tup11 interacts with the C-terminal portion of Fep1, which does not contain the two adjacent zinc finger motifs. This suggests that FIG. 8. In vivo association between Tup11 and Fep1 requires the conserved Tyr 362 and Leu 542 residues within the WD40 repeat domain of Tup11. Total cell extracts (T) were prepared from JY741 cells harboring TAP-319 Fep1 564 fusion protein and the indicated GST-tagged-1 Tup11 614 derivatives or GST alone. Lysates were incubated with a glutathione affinity resin and washed, and the bound fraction (B) was analyzed by immunoblot assay using anti-mouse IgG antibody. A portion of the total cell extracts (ϳ2%) was included to verify the presence of the immunoblotted proteins prior to chromatography. Furthermore, aliquots of all samples (extracts and bound fractions) were analyzed by immunoblotting using anti-GST antibody. WB, Western blot. the N-terminal zinc finger is not required for association between Fep1-Tup11. Further studies will be needed to assess the nature of the function of the zinc fingers in Fep1 with respect to its DNA binding activity and ability to mediate interactions with other proteins.
In this study, we have identified and demonstrated a role for the C-terminal WD40 repeat domain of the S. pombe Tup11 protein. This domain consists of seven WD40 repeats. The model predicts that these repeats clasp together into a circular, seven-bladed propeller with a relatively flat top and bottom. Deletion mapping analyses revealed that the WD40 repeat sequence motifs of Tup11 are requisite for the Fep1-Tup11 interaction. The first Tup11 WD40 repeat failed to interact with Fep1 when fused to the LexA domain by two-hybrid assay, suggesting that amino acids from more than one WD repeat are required for binding Fep1. Analysis of the WD40 repeat structure using computer algorithms predicted that the Tup11 Tyr 362 and Leu 542 residues in the WD40 repeats 2 and 6, respectively, were exposed on the upper surface of the ␤-propeller. We tested the roles of these residues in Tup11-Fep1 interaction by replacing Tyr 362 with cysteine or Leu 542 with serine and found that these mutations abolished the ability of Tup11 to interact with Fep1. Although we have not ascertained the contributions of each of the seven WD40 repeats that comprise the putative ␤-propeller, taken together, these data suggest that proteins with WD40 repeats may use different blades of the propeller to provide binding sites for different partners. Indeed, the WD40-repeat protein RACK1 interacts with the PDE4D5 protein through its WD40 repeats 5-7, while the S. cerevisiae Tup1 can associate with ␣2 using only WD40 repeat 1 (30,78).
It is currently unknown how sequence-specific DNA-interacting proteins are capable of recruiting the Tup corepressor complex. Recent studies have suggested a mechanism whereby relief from repression can be achieved (79,80). In S. cerevisiae, under non-osmotic stress conditions, the Tup1 corepressor complex and its interacting partner protein Sko1 are required to negatively regulate expression of a subset of osmotic stressresponsive genes (81). It was shown that the Hog1 kinase can overcome the repressive action of the Tup1 corepressor complex by phosphorylating the DNA-binding transcription factor Sko1 (80,81). Under osmotic stress, the phosphorylation of Sko1 by Hog1 converts the Sko1-Tup1 corepressor complex into an activator that recruits the SAGA histone acetylase and SWI/ SNF nucleosome remodeling complexes to derepress osmoticinducible promoters (80). Interestingly, S. cerevisiae Hog1 and S. pombe Spc1 (also named StyI) share a high percentage of sequence identity (82%). 2 Both are homologs of the mammalian stress-activated protein kinases (82). The Hog1 kinase is activated in response to osmotic stress, whereas activation of Spc1 is mediated by a wide range of stimuli including oxidative stress, nutrient limitation, and heavy metal toxicity (83). We tested the possibility that spc1 ϩ plays a role in the regulation of the genes that are known to be involved in iron uptake in S. pombe. Using cells bearing a disruption of the sty1 ϩ gene, we have been unable to demonstrate a loss of expression or lack of iron regulation of mRNA levels in each of these genes. 3 Nevertheless, this observation leaves open the possibility that an iron starvation-activated cellular kinase may be involved in modulating iron-responsive gene expression of the fission yeast iron transport genes through the Fep1-Tup11 corepressor complex.