Characterization of zfs1 as an mRNA-binding and -destabilizing Protein in Schizosaccharomyces pombe*

Tristetraprolin is a vertebrate CCCH tandem zinc finger protein that can bind to and destabilize certain mRNAs containing AU-rich element binding sites. zfs1 is the single gene in the fission yeast, Schizosaccharomyces pombe, that encodes a protein containing the critical features of the tristetraprolin zinc finger domain. zfs1 has been linked to pheromone signal transduction control and to the coordination of mitosis, but no biological function has been ascribed to the zfs1 protein. Through a functional genomics approach we compared transcript levels in wild-type and zfs1-deficient S. pombe strains; those elevated in the zfs1-deficient strain were examined for the presence of potential tristetraprolin-like binding sites. One such potential target transcript was encoded by arz1, a gene encoding a protein of unknown function that contains armadillo repeats. arz1 mRNA decay was inhibited in the zfs1-deficient strain when it was expressed under the control of a thiamine-repressible promoter. Mutations within one AU-rich element present in the arz1 3′-untranslated region protected this transcript from zfs1-promoted decay, whereas mutating another potential binding site had no effect. Binding assays confirmed a direct interaction between zfs1 and arz1 mRNA-based probes; this interaction was eliminated when key residues were mutated in either zfs1 zinc finger. zfs1 and its targets in S. pombe represent a useful model system for studies of zinc finger protein/AU-rich element interactions that result in mRNA decay.

Tristetraprolin (TTP) 2 is a well characterized CCCH tandem zinc finger (TZF) domain RNA-binding protein that can regulate the decay of many mRNAs (1)(2)(3)(4)(5)(6). The presence of a conserved TZF domain of this type defines a small family of proteins in man that consists of TTP, ZFP36L1 (7), and ZFP36L2 (8). Interactions between the TZF domain and type II AU-rich elements (AREs) found in the 3Ј-untranslated regions (UTRs) of target mRNAs have been the focus of multiple investigations (9 -14). These have included the use of a synthetic TTP TZF domain peptide to examine specificity of ARE target variants (9,11,12), the evaluation of target mRNA decay as it is influenced by mutated TZF domains expressed in co-transfected cells (14), cell-free deadenylation of ARE-containing mRNAs (10,15), and the use of engineered stable transcripts to assess ARE specificity, as measured through mRNA decay (10). The solution structure of the TZF domain of ZFP36L2 (TIS11D) bound to the UUAUUUAUU nonamer has been solved (13), providing insight into the interaction between the TZF domain and a target ARE. A core UAUUUAU heptamer unit appears to be necessary for optimal binding affinity of the TZF domain for an ARE (10,14).
Examination of the genome of the fission yeast Schizosaccharomyces pombe revealed only one member of the TTP family, encoded by the zfs1 gene, which has a TZF domain at its C terminus (16). zfs1 is necessary for coordinating septum formation with exit from mitosis (17). zfs1-deficient cells exhibit a reduced rate of cell division, with decreased cell size, and do not respond properly to pheromone (16,17). Pheromone insensitivity can be overcome by overexpression of proteins that control the upstream portion of the pheromone signaling cascade (16). To date no specific biochemical function of the zfs1 protein has been demonstrated, but intact zinc fingers are necessary for the biological roles of zfs1 described previously (16). No evidence of DNA or RNA binding has been presented for zfs1.
In contrast to S. pombe, Saccharomyces cerevisiae expresses two members of the TTP family, CTH1 and CTH2 (18,19). Unlike the effects observed with disruption of zfs1 in S. pombe, disruption of either CTH1 or CTH2 had no apparent effect on growth (18). Overexpression of CTH1 or mouse TTP, but not CTH2, in S. cerevisiae reduced growth rates by 50% (18), and increased zfs1 expression in S. pombe resulted in a comparable level of growth reduction after 24 h. CTH2 is important for regulating mRNA stability in response to iron depletion (19). Recent data indicate that zfs1 is not a functional homologue of CTH2 with respect to iron sensitivity (20), and S. pombe and S. cerevisiae differ substantially in their mechanisms of iron metabolism (19 -21).
In a microarray study designed to identify transcripts elevated in a zfs1-deficient strain, we identified a previously uncharacterized gene that we have named arz1 (for armadillorepeat containing zfs1 target 1). We hypothesized that, like TTP and its mammalian family members, zfs1 might function as an ARE-binding, mRNA-destabilizing protein. We show that * This work was supported by the intramural program of NIEHS, National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  zfs1 can bind specifically to AREs within the arz1 mRNA 3Ј-UTR in a zinc finger-dependent fashion, and stimulate the rate of arz1 mRNA decay in intact cells. These data suggest that this system holds promise for more detailed biochemical investigations of the mechanisms of action of this protein family, as well as the physiological function of zfs1 in this organism.

EXPERIMENTAL PROCEDURES
Yeast Strains-A S. pombe zfs1-deficient strain was used in a direct comparison with a representative wild-type strain, which was originally used as the background cell type for zfs1 gene targeting (17). These strains were gifts from Dr. Viesturs Simanis (École Polytechnique Fédérale De Lausanne, Lausanne, Switzerland). Strains were grown at 30°C in Edinburgh minimal medium (EMM) supplemented with 50 mg/liter of adenine, histidine, leucine, lysine, and uracil each (EMMϩ5S) or with SP minus Leu (EMMϩ4S-Leu) when transformed (media and supplements were purchased from Q-biogene, Morgan, CA).
Steady-state Comparison of S. pombe Strains by Microarray-S. pombe wild-type and zfs1-deficient strains were grown to 1.0 A 600 in EMMϩ5S at 30°C for steady-state analysis. Total cellular RNA was prepared from four replicate cultures. For these and all other cultures in this study, cells from 1.5 ml of culture were centrifuged at 16,100 ϫ g for 1 min at room temperature, and total RNA was harvested from the cell pellet after storage at Ϫ80°C using the MasterPure yeast RNA purification kit (Epicenter, Madison, WI). Gene expression analysis was conducted using Yeast Genome 2.0 GeneChip arrays (Affymetrix, Santa Clara, CA) containing both S. cerevisiae and S. pombe probe sets. RNA (1 g) was amplified using the Affymetrix One-Cycle cDNA Synthesis protocol. For each array, 5 g of amplified biotin-cRNAs was fragmented and hybridized to the array for 16 h at 45°C in a rotating hybridization oven using the Affymetrix Eukaryotic Target Hybridization Controls and protocol. Slides were stained with streptavidin/phycoerythrin using a double-antibody staining procedure and washed utilizing the Mini_euk2v3 protocol of the Affymetrix Fluidics Station FS450 for antibody amplification. Arrays were scanned with an Affymetrix Scanner 3000, and data were obtained using the GeneChip operating software. The resulting files (.dat, .cel, and .chp) were imported into the Rosetta Resolver system, and the S. cerevisiae probe sets were masked out prior to data preprocessing and error modeling (22). The data are accessible at the NCBI Gene Expression Omnibus (GEO), as described under "Results," and selected transcripts elevated in the zfs1deficient strain that contain potential binding sites are presented in Table 1.
nmt Transformants-The coding region of SPCC1494.03 (arz1) plus 1 kb of downstream sequence predicted to include the 3Ј-UTR was PCR-amplified using primers F1 and R1 (all PCR primers are listed in Table 2), and the product was cloned into a TOPO TA vector system based on the nmt expression system (nmt41) (Invitrogen) (23). The nmt41 vector is known a Indicates the status of the characterization of these genes listed in the S. pombe gene database. "Experimental" indicates some level of experimental characterization, "Homology" indicates a description based on homology to known proteins of other species, and "Hypothetical" indicates a predicted protein that has not been assigned a specific description by homology. b Reflects increased transcript levels in the zfs1-deficient strain relative to wild type. This comparison is based on the means of four separate experiments in each case. to express constructs at ϳ15% of the level of the nmt1 vector (16,23,24). Site-directed mutagenesis was carried out using the QuikChange system (Stratagene), and primers listed in Table 3, to produce the following nmt/arz1 mutant constructs: MuA, which has two A to G mutations; MuB, which has two A to G mutations for each of two nonamers (four total); B1 and B2, which have two A to C mutations at each of the AREb nonamers separately, and CCC, which has three mutations of TAA to CCC. Mutations were confirmed in purified nmt/arz1 plasmids by dideoxy-chain terminator-based sequencing (25). Both S. pombe strains described above were transformed with the unmutated nmt/arz1 plasmid to produce the W1 and Z1 reference transformant strains, and the mutated nmt/arz1 plasmids were transformed into the wild-type S. pombe to produce MuA, MuB, B1, B2, and CCC transformant strains for direct comparison with W1 and Z1 reference strains following previously described methods (26,27). Positive transformants were screened for retained plasmids by PCR using a 5Ј nmt vector specific primer, F2, and a 3Ј arz1 internal primer, R2 (Table 2). S. pombe Iron Depletion-Wild-type S. pombe was grown in EMMϩ5S from ϳ0.1 A 600 to log phase at 30°C (reached ϳ1.0 A 600 at 20 h) in the presence (treatment) or absence (control) of 750 M ferrozine (Sigma), an iron chelator, following an approach used for S. cerevisiae (19). 5 g of total cellular RNA was used for Northern analysis (see below) of zfs1 transcript levels.
Northern Analyses-arz1-, zfs1-, and actin-specific probes were generated by PCR using gene-specific primers (Table 2), and 50 ng of each probe fragment was used for random-primed labeling with a High Prime kit (Roche Applied Science) to incorporate [␣-32 P]dCTP (PerkinElmer Life Sciences). The actin control was used to examine loading consistency, in combination with gel staining using acridine orange, and for quantitative correction during PhosphorImager densitometric evaluation (ImageQuaNT 5.1, Molecular Dynamics, GE Healthcare, Piscataway, NJ).
S. pombe wild-type and zfs1-deficient strains were grown to 1.0 A 600 in EMMϩ5S at 30°C for steady-state analysis. A total of 5 g of total cellular RNA was used in each gel lane for all Northern analyses, following standard methods (28). W1 and Z1 transformant strains were grown to 1.0 A 600 in 30 ml of EMMϩ4S-Leu. A time "0" sample (1.5 ml of culture) was first collected, and the cell suspension was centrifuged at 16,100 ϫ g for 1 min at room temperature, after which the pellet was frozen in liquid N 2 . Thiamine (10 M final concentration) was then added to the cultures. These were then sampled such that pellets were frozen at 10, 20, and 40 min after thiamine addition. A second screen focused on 4-min time points (4,8,12,16, and 20 min) was carried out in an effort to determine the half-life of the arz1 mRNA in the W1 and Z1 transformant strains. Cell pellets of transformants were used for total cellular RNA isolation and Northern analysis. Endogenous mRNA control samples were evaluated in direct comparison with each set of transformant strain comparisons and used for correction during quantification of nmt/arz1-expressed mRNA.
Total cellular RNA from W1, MuA, MuB, B1, B2, or CCC mutant nmt/arz1 constructs was compared directly to RNA from Z1 by Northern analysis. Direct comparisons of the four samples were performed following the same protocol used for the initial W1 and Z1 initial time-course comparisons. Northern blots were washed extensively with two final washes at 60°C in 1ϫ SSC, 0.1% SDS before exposure to film or phosphorimaging screens. Blots were exposed to phosphorimaging screens for 2-3 h, and these were scanned using the Typhoon system (Molecular Dynamics, GE Healthcare) and analyzed using ImageQuaNT and GraphPad (San Diego, CA) Prism, version 4. All graphs show paired comparisons using Students t test (for direct comparison of individual means) or one-way analysis of variance (used for comparisons of multiple means), and error bars indicate standard deviations.
zfs1 Protein Production and RNA Gel Shift Assays-The zfs1 coding region was PCR-amplified using primers F6 and R6 (Table 2), containing SspI and SpeI restriction sites, respectively, for cloning into restriction enzyme-digested expression vector. A PCR-derived restriction digest fragment was cloned into an expression vector LacZ promoter upstream of the maltose-binding protein (MBP) coding sequence (cytoplasmic) along with a tetracycline resistance cassette (AstraZeneca, Wilmington, DE). The expression construct was transformed into Escherichia coli BL21(DE3) containing a TF16 (trigger factor) expression construct and a chloramphenicol resistance cassette. A clone bearing both expression constructs was grown in selective medium at 37°C to 0.4 A 600 . Expression of TF16 was induced by the addition of 1 ml of 0.5 g/ml L-arabinose for 1 h, followed by induction and expression of the MBP/zfs1 construct with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside for 20 h at 18°C (shaking at 200 rpm). Cultures were chilled on ice, and cells were pelleted by centrifugation at 5,000 ϫ g for 20 min at 4°C and then lysed in 40 ml of cold Buffer A (20 mM Tris-HCl, pH 7.4, 200 mM NaCl, 0.1% Triton X-100) by ultrasonication. The cellular debris was pelleted by centrifugation at 10,000 ϫ g for 20 min at 4°C. The supernatant was then applied to 5 g of amylose resin (New England Biolabs, Ipswich, MA) that had been washed with Buffer A in preparation for batch method purification in a 50-ml polypropylene tube, following the manufacturer's protocol. Approximately 30 mg of protein was obtained (5-6 mg/ml) from the amylose resin. Protein concentrations were assessed using the BCA protein assay (Pierce) and were confirmed by SDS-PAGE, using albumin as a standard, and purity was estimated at Ͼ90% by SDS-PAGE Coomassie Blue staining. Site-directed mutants of the MBP/zfs1 construct were made (29), using primer pairs Fzfs1C and Rzfs1C, and Fzfs1H and Fzfs1H (Table 3), that would change Cys-370 to Gly (C370G) and His-351 to Ile (H351I), respectively, using the amino acid numbering system in NCBI accession number CAB75997.1, which we predicted would eliminate RNA binding based on previous studies in mammalian family members (3) and homology modeling presented in Fig. 5. Recombinant mutant proteins were expressed and purified using the same strategy that was used for the wild-type MBP/zfs1 protein and for MBP alone. The C370G and H351I amino acid mutant proteins yielded ϳ24 mg and 18 mg of total protein, respectively.
Single-stranded RNA oligonucleotides were obtained from Invitrogen that represented the arz1 AREa and AREb sequences, as well as binding site mutants of each of these sequences (Table 4). A human tumor necrosis factor-␣ AREbased oligonucleotide of comparable length was also tested in gel shift assays (Table 4). Each oligonucleotide was end-labeled with [ 32 P]pCp at the 3Ј-end using T4 RNA ligase (New England Biolabs, Ipswich, MA). Approximately 100 fmol of each probe (1 ϫ 10 4 cpm) was used per reaction in binding experiments.
Approximately 4.25 g of each recombinant MBP/zfs1 protein (50 pmol) was used in RNA electrophoretic mobility shift assays in comparison with 4.25 g of MBP alone (ϳ100 pmol). In each case, the designated amount of probe was preincubated with binding buffer (10 mM HEPES (pH 7.6), 40 mM KCl, 5% (v/v) glycerol, 50 g of heparin, 3 mM MgCl 2 , 1.2 g of yeast tRNA) for 20 min. An aliquot of each protein was then combined with each probe separately, along with 2 units of Ribonuclease Inhibitor (Invitrogen), and the mixture was incubated at room temperature for 20 min before gel loading. The reactions were separated on 8% polyacrylamide gels in 0.4ϫ TBE (40 mM Tris-HCl (pH 8.3), 40 mM boric acid, and 8 mM EDTA) running buffer for 1.5 h at 250 V. Gels were then dried at 80°C for 30 min on absorbent paper and exposed to film.
Cladogram Construction-TZF domains from select proteins were aligned based on blastp (31, 32) results, and a representative cladogram was constructed using the neighbor-joining method and the p-distance substitution model with MEGA 3.1 (33). All TZF domains surveyed in this analysis were the same length, with the exception of the Cryptococcus neoformans TZF, which is three amino acids longer in the normally 18-amino acid spacer between zinc fingers and thus served as a natural out-group for comparison.
Homology Modeling-A model of the TZF domain segment of zfs1 (amino acids 327-390, NCBI accession number CAB75997.1) was obtained using Swiss-Model (34 -36) by submitting the sequence through the First Approach mode. The resultant PDB file was examined, and the known structure of the ZFP36L2 TZF domain complexed with the ARE (13) was fit onto the zfs1 model using DeepView v3.7 (34), resulting in a combined file. A layer of this file was saved that included the zfs1 TZF domain and the ARE nonamer. Distances between core atoms present in specific functional groups were measured using DeepView v3.7, and POV Ray v3.6 was used for renderings.

RESULTS
Transcripts Elevated in zfs1-deficient S. pombe-mRNAs obtained from S. pombe wild-type and zfs1-deficient strains were compared directly using the Yeast 2.0 array (Affymetrix, Santa Clara, CA). Results were deposited in the NCBI GEO (37, 38) (www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO series accession number GSE8122. 123 transcripts were elevated in the zfs1-deficient strain at mean levels Ն1.5fold greater than in the wild-type strain; in all cases, the mean levels were significantly different as determined by Rosetta Resolver (22); p values are presented in Table 1. These transcripts were then examined for AREs of the TTP binding type present in the hypothetical 3Ј-UTR comprising 1 kb downstream of the stop codon. From the 123 significantly elevated transcripts, 12 contained two or more UAUUUAU sequences in this region (Table  1). From this analysis we chose to examine in more detail the most elevated candidate transcript containing likely binding sites, encoded at SPCC1494.03 in the S. pombe genome.

JOURNAL OF BIOLOGICAL CHEMISTRY 2589
Elevation of arz1 mRNA Levels in the Absence of zfs1-Northern analysis of total RNA isolated in parallel from wild-type and zfs1-deficient strains (four separate cultures of each) revealed that the mRNA encoded by the SPCC1494.03 locus was elevated by 3.6-fold in the zfs1-deficient strain (Fig. 1, A and B), comparable to the microarray results (4.3 fold) ( Table 1). We have named the protein encoded by SPCC1494.03 arz1 (signifying armadillo-repeat containing zfs1 target 1), based on the presence of armadillo-like (ARM) repeats in the arz1-predicted protein sequence (NCBI accession number CAA19301.1), and on apparent arz1 mRNA targeting by zfs1 (see below). arz1 has two different types of ARM repeat. The first repeats are found in three areas that encompass amino acids 7-177, 205-317, and 401-486, and the second is found at amino acids 346 -428 (SPCC1494.03). The ARM repeats contribute to the high percentage of leucine (12%) in the arz1 protein sequence and result in limited amino acid sequence identity to ARM repeat-containing proteins such as human ␤-catenin (39,40). Blast (31,32) searches showed that the entire length of the deduced arz1 sequence aligns with several vertebrate Rap1-like sequences (at 20% identity) and revealed a stretch of 151 amino acids in arz1 with 25% identity to several ARM repeat-containing proteins in Drosophila, including Vimar (41). The function of S. pombe arz1 has not been explored, and it does not appear to have an S. cerevisiae orthologue.
Effect of zfs1 Deficiency on arz1 Transcript Decay-We measured arz1 mRNA stability in the presence or absence of zfs1, using the nmt (no message with thiamine) repressible expression system (23). Wild-type and zfs1-deficient strains transformed with the nmt/arz1 expression construct (named W1 and Z1, respectively) were examined for relative arz1 mRNA abundance after thiamine addition (repression), which was determined by normalizing arz1 mRNA levels to actin and calculating the percentage of arz1 mRNA present at each time point after repression. The first screen focused on evaluating arz1 mRNA levels at 10 or 20 min relative to 0 min (0 min equals 100% relative intensity). Ten minutes after the addition of thiamine, the relative abundance of the arz1 transcript in the ⌮1 strain was significantly greater than in the W1 strain (Fig. 1, C  and D). By 20 min most of the arz1 transcript was degraded in both strains (Fig. 1, C and E). Data from five independent replicates of this experiment, in which mRNA levels from W1 and Z1 transformants at 10 min were compared directly, are summarized in Fig. 1D. The half-life of the arz1 mRNA in the W1 and Z1 strains was then evaluated by sampling earlier time points (at 4 and 8 min) continuing on to 20 min in 4-min intervals (Fig. 1, E and F). Data from four independent replicates of this experiment, in which mRNA levels from W1 and Z1 transformants were compared directly, are summarized in Fig. 1F. The arz1 mRNA half-life was extended from ϳ6.3 min in the W1 strain to 14.3 min in the Z1 strain (Fig. 1F).
ARE Dependence of zfs1-mediated arz1 mRNA Decay-Three UUAUUUAUU nonamers are present in the predicted arz1 3Ј-UTR, including one at 85 bases 3Ј of the stop codon (AREa), and two others at 215 and 232 bases 3Ј of the stop codon (named AREb, collectively) ( Fig. 2A). These ideal TTP binding sites were mutated to UUGUUUGUU in each ARE ( Fig. 2A). We compared transcript levels at 10 min after thiamine addition between the following strains: W1 (described above), MuA (wild-type S. pombe expressing arz1 mRNA with mutations A87G and A91G) or MuB (wild-type S. pombe expressing arz1 mRNA with four mutations A217G, A221G, A224G, and A238G), and Z1 (described above) ( Fig. 2A). For each comparison, mean arz1 mRNA levels at 10 min are shown as normalized to actin control mRNA levels and as a calculated percentage of the respective 0-min mRNA level, to indicate relative arz1 mRNA abundance at 10 min. Mean arz1 mRNA levels observed in the MuA mutant were identical to those observed for W1 (Fig. 2B), indicating that mutations at AREa did not confer a protective effect on the arz1 mRNA. The MuB mutant arz1 mRNA levels were comparable to those observed for ⌮1 and significantly greater than those observed for W1 (Fig.  2C), indicating that the mutations at AREb protected against zfs1-dependent decay. This strongly supports a role for AREb, but not AREa, in zfs1-mediated arz1 transcript destabilization.
Mutations in each of the core UUAUUUAUU sequences within AREb separately (B1 and B2 mutants (Fig. 2, D and E) resulted in only partial protection from zfs1-mediated decay (Fig. 2D). The B1 mutant, expressing arz1 mRNA containing both the A217C and A221C mutations, appeared to confer greater protection from decay than did the MuB2 mutant (Fig.  2, D and E). A one-way analysis of variance of mean arz1 mRNA levels from W1, B1, B2, and Z1 strains indicated significant variance across multiple means (p Ͻ 0.0001). Using the Bonferroni multiple comparison test (Fig. 2E), we showed that B1 and B2 mutant arz1 mRNA levels were different than those seen in both the W1 and Z1 transformant strains (p Ͻ 0.001, with the exception of W1 versus B2, which was p Ͻ 0.01, and B1 versus B2 was p Ͻ 0.05).
Two separate stem-loop structures were predicted in the AREb sequence (Fig. 2F): one (initial ⌬G ϭ 2.3 kcal/mol) that could juxtapose the two nonamers into an overlapping binding site, and another within the B2 nonamer (initial ⌬G ϭ 2.0 kcal/ mol). To examine the role of these possible structures in zfs1mediated mRNA decay, we mutated three residues (U229 -A231) predicted to participate in both structures (Fig. 2, F and   G), resulting in a change from 5Ј-UAA-3Ј to 5Ј-CCC-3Ј in the expressed mRNA (Fig. 2F), thus preventing the formation of both of the predicted stem-loop structures. However, mutating these residues had no effect on zfs1-dependent arz1 transcript decay (Fig. 2, H  and I).
zfs1 Binding to ARE Probes-The ability of recombinant zfs1 to bind directly to ARE-based RNA probes was assessed using RNA gel shift assays. MBP/zfs1, two separate zinc finger mutant forms of MBP/zfs1 (C370G and H351I), and MBP alone were expressed in and purified from E. coli. Fig. 3A shows that the recombinant proteins were quite pure based on SDS-PAGE.
When equal amounts of MBP/ zfs1, C370G and MBP alone were allowed to bind the tumor necrosis factor-␣ probe, the probe was completely shifted by the recombinant MBP/zfs1 protein (Fig. 3B), whereas neither the MBP/zfs1 zinc finger mutant, C370G, nor the MBP protein alone caused analogous shifting of the probes (Fig. 3B). A small fraction of probe was shifted into the gel by both the C370G mutant and the MBP protein alone, indicating that some zinc finger independent or nonspecific binding to these probes can occur in these conditions. Several probes based on arz1 AREa and AREb were examined by gel shift in the presence of MBP/zfs1 recombinant proteins and MBP alone. A probe based on the AREa sequence was partially shifted by 50 pmol of MBP/zfs1 (2 M) (Fig. 3C), but not by either zinc finger mutant form or by MBP alone (Fig. 3C). A mutant probe with alterations in the predicted binding site at AREa (Table 4) was not shifted by MBP/zfs1 (Fig. 3C). An AREb-based probe was completely shifted by 50 pmol of MBP/ zfs1 (2 M) (Fig. 3D), but not by equal amounts of either zinc finger mutant or by MBP alone (Fig. 3D). A mutant probe with alterations in both predicted binding sites at AREb (Table 4) was not shifted by MBP/zfs1 (Fig. 3D). AREb mutant probes, based on the mutations examined in Fig. 2 (D and E) using the nmt/arz1 expression system, were also tested ( Table 4). Fig. 3E shows a complete shift of the AREb probe by MBP/zfs1, with partial shifts observed for the B1 and B2 mutant probes.
A serial dilution of the MBP/zfs1 protein starting from 2 M was applied to the AREb probe to examine the affinity of this interaction. Half-maximal binding appeared to occur at 125-250 nM, which was thus in excess over the 4 nM of probe used in the assays. It is possible that either the assay conditions or the recombinant production system yielded a fraction of protein that would not bind probe.
zfs1 mRNA and Iron Deprivation-Following methods described previously for the study of S. cerevisiae CTH2 (19), the S. pombe wild-type strain was grown in iron-deficient conditions in parallel with normal conditions. Growth rates were similar under the two different conditions. Previously, zfs1 mRNA was not observed to be affected by iron addition or depletion (20). Our results confirmed those findings (20) (data not shown). This result contrasts sharply with the dramatic increase of CTH2 mRNA observed in S. cerevisiae (19) after iron depletion.
Comparison of the zfs1 TZF Domain to Other Eukaryotic TZF Domains-Analysis of selected eukaryotic sequences revealed that the S. pombe zfs1 TZF domain exhibited more amino acid sequence identity to TZF domain sequences from metazoans than to those from other yeasts (Fig. 4), whereas the two S. cerevisiae TZF domain sequences (from CTH1 and CTH2) (18) cluster with TZF domain sequences from two Candida species, C. albicans and C. glabrata (Fig. 4B).
Conservation of Key TZF Domain Motif Characteristics in zfs1-A model of the zfs1 TZF domain made using Swiss-Model (36) revealed that conserved residues between TTP and zfs1 (Fig. 4) are found in the general ARE interaction areas (Fig.  5A). Stacking interactions are predicted between highly conserved aromatic side chains and RNA bases (Fig. 5A). Main polypeptide chain interactions with RNA bases are predicted to be involved in the TZF domain-ARE interaction (Fig. 5B), as observed for ZFP36L2 (13). The polypeptide chain of the N-terminal zinc finger at zfs1 amino acids Glu-330 and Pro-331 may hydrogen bond with the adenosine residue (A7), an interaction that is in close proximity to the zinc binding core. The side chain of a conserved Arg residue in the C-terminal finger (Fig.  5A, Arg-383) extends into the N-terminal zinc-finger (Fig. 5, A  and B). This interaction may be important to the overall configuration of the TZF domain (Fig. 5B).

DISCUSSION
The most important findings of this study are that zfs1 is an ARE-specific mRNA-binding protein that exhibits binding site selectivity and can promote mRNA decay, and that the arz1 transcript is the first target identified for zfs1. A function has not been determined for the arz1 protein, but the regulation of arz1 transcript levels by zfs1 may eventually provide insights into the role of zfs1 in cell division and mating. The system we describe for assessing zfs1-mediated mRNA decay holds promise for the evaluation of the mechanisms of action of all the TTP family proteins. S. pombe expresses only a single TTP family member, whereas humans express three TTP family members. zfs1 can be studied at native expression levels in vivo, whereas most studies of TTP and its relatives have relied on transfection and overexpression studies in cultured cells.
Our findings provide evidence of binding site selectivity in a native mRNA targeted by a TTP family member. Specifically, non-binding mutations within the arz1 5Ј-most binding site present in AREa had no effect on zfs1-dependent decay, despite the fact that its sequence corresponds to that of an optimal TTP family member binding site (11,13). In contrast, non-binding mutations within the two 3Ј-most arz1 binding sites (AREb) essentially eliminated zfs1-dependent decay. Further evidence of binding site selectivity was observed when mutations within each of the binding sites present in AREb were introduced, each resulting in partial elimination of zfs1-dependent decay. Based on what is known from the NMR structure of the ZFP36L2 TZF domain bound to the optimal ARE nonamer, we can postulate that several interactions occur between amines in the adenine bases and both main chain and side chain functional groups in the zfs1 TZF domain. These interactions probably form the basis for binding affinity that results in biological function (i.e. mRNA decay).
Binding site selectivity may be explained by the propensity for tandem nonamers to contribute in an interactive fashion to the targeting of this region by zfs1 in a way that facilitates completely shifted the tumor necrosis factor-␣-based probe into the gel, whereas neither the C370G amino acid mutant of MBP/zfs1, nor MBP alone, caused the same gel shift. C, a partial shift is observed for an arz1 AREa-based probe with MBP/zfs1, but not with either zinc finger mutant MBP/zfs1 protein or MBP alone. Mutations in the binding site of the AREa probe eliminated the partial shift observed with MBP/zfs1. D, an AREb-based probe is completely shifted into the gel by MBP/zfs1, but not with either zinc finger mutant MBP/ zfs1 protein or MBP alone. Mutations in the two binding sites of the AREb probe eliminated the shift observed with MBP/zfs1. E, mutations in either AREb binding site (B1 or B2) resulted in a partial shift. 50 pmol of each protein was assayed. F, a serial dilution of 50 pmol of MBP/zfs1 was used in the gel shift assay with the AREb probe, resulting in the range of concentrations shown. Half-maximal binding of the AREb probe (100 fmol total, 4 nM final concentration) was observed with ϳ3.125-6.25 pmol of MBP/zfs1 (125-250 nM final concentration). For each gel (B-F) the position of the free probe at the bottom of each gel is indicated. addition or depletion (20). Moreover, zfs1 deletion had no effect on transcripts encoding proteins involved in iron metabolism or on their regulation by iron availability (20). The other TZF protein in S. cerevisiae, CTH1, has not yet been characterized functionally, but resembles zfs1 no more than CTH2, with both proteins exhibiting 52% identity to zfs1 in the TZF domain. Interestingly, the TZF domain of zfs1 resembles that of human TTP (59% identity) more closely than do the S. cerevisiae CTH1 or CTH2 TZF domains (52 and 50%, respectively), making it a better root to metazoan TZF domain sequences than either of the S. cerevisiae proteins in phylogenetic trees. These factors emphasize the potential that S. pombe zfs1 presents as a model for the study of TTP family proteins.
The S. pombe nmt expression system provides a useful alternative to mammalian transient transfection studies for the examination of TZF domain-ARE interactions. The limitations of mammalian transfection systems include difficulties in controlling the timing and extent of the co-expression and the artificially high levels of expression of the TTP family members. The nmt expression system appears to be well suited to such studies because of the rapidly repressible nature of the promoter, permitting direct assessment of mRNA decay, in the context of the single TTP family member, zfs1, expressed at endogenous levels. Furthermore, the genetic tractability of yeasts permits the development of important strains, such as the zfs1-deficient strain, which may facilitate the evaluation of the importance of cofactors and modifying proteins in the regulation of accelerated target transcript decay.
The function of zfs1 in S. pombe to regulate the pheromone response and cell division should provide new insights into the relevance of mRNA decay in the control of these processes. Future investigations will focus on the biochemical role of zfs1 targets in these physiological events.