Influence of Nonameric AU-rich Tristetraprolin-binding Sites on mRNA Deadenylation and Turnover*

Tristetraprolin (TTP), a member of the tandem CCCH zinc finger protein family, promotes deadenylation of tumor necrosis factor-α and granulocyte-macrophage colony-stimulating factor mRNAs after binding to the AU-rich elements (ARE) in their 3′-untranslated regions. The high affinity TTP-ARE binding occurs between the tandem zinc finger domain and the preferred nonamer UUAUUUAUU. By mutating a well defined core sequence of 24 bases from the tumor necrosis factor-α ARE, we compared the influence of four possible nonameric TTP-binding sites in the wild-type ARE with that of a single binding site in the mutated probe on the binding of TTP to the RNA and the subsequent deadenylation of the poly(A) tail. By inserting this 24-base ARE into an otherwise stable transcript, we also attempted to determine the extent of the instability conferred by the presence of one or two TTP-binding sites. These sites were created or modified by mutating the As in the UUAUUUAUU nonamer or by changing the central U in the nonamer, in both cases to C residues. The results suggest that even a single nonamer TTP-binding site can confer at least partial sensitivity to the TTP-mediated mRNA turnover on an otherwise stable mRNA, but that two binding sites make the transcript much more unstable. Even though the central U of the nonamer binding site was predicted by structural studies possibly to permit base substitution, mutation of this U to C greatly inhibited the binding of TTP to the ARE, thus reducing the ability of the TTP to promote deadenylation and instability of the mRNA.

Tristetraprolin (TTP) 2 is a member of a small family of three tandem CCCH zinc finger proteins in man that can bind to AU-rich elements in single-stranded mRNA and promote the successive deadenylation and degradation of those RNA species (for review see Ref. 1). Experiments in TTP knock-out mice and cells derived from them demonstrated that tumor necrosis factor-␣ (TNF) and granulocyte-macrophage colonystimulating factor (GM-CSF) transcripts were physiological targets for TTP-mediated destruction (2,3). Both transcripts contain the so-called class 2 AU-rich elements (ARE) (4), and TTP was shown to bind to these elements through its tandem zinc finger (TZF) domain (2,5). This bind-ing is thought to be the first step in the process of subsequent transcript deadenylation and destruction.
More recently, details of the minimal ARE sequence required for TTP binding have emerged. For example, Worthington et al. (6) demonstrated through random selection procedures that the optimum minimum binding site was the nonamer UUAUUUAUU. By using a synthetic 73-amino acid peptide comprising the TTP TZF domain, we showed that the same nonamer was necessary for high affinity binding to the synthetic peptide, and that a 24-base RNA oligonucleotide derived from the TNF mRNA ARE could bind 2 mol of peptide/mol of RNA at two tandem nonamer binding sites (7). In the same paper, we showed by HSQC NMR analyses that the same nonamer represented the minimum ARE-binding site oligonucleotide required for a characteristic conformational shift of the TTP peptide, and that even minor shortening of the oligonucleotide resulted in changes in the peptide-RNA complex conformation. Similarly, by using a solution binding assay with the same synthetic 73-amino acid TTP TZF peptide, Brewer et al. (8) found that shortening the UUAUUUAUU nonamer to a UAUUUAU heptamer increased the K d from 3.0 to 19 nM. On the other hand, lengthening the nonamer by adding two Us to each end did not affect the binding affinity.
Very recently, Hudson et al. (9) solved an NMR structure for the TZF domain of the TTP family member TIS11D (ZFP36L2), which is 73% identical to the TTP TZF domain over the 64 amino acids of the core TZF domain, in complex with the same UUAUUUAUU nonamer. They found that this nonamer bound to the TZF domain peptide in the absence of RNA secondary structure, suggesting that the RNA primary sequence is the critical determinant of effective binding. They also noted that the first U residue in the nonamer was unstructured in the complex, perhaps allowing for substitutions at that site in naturally occurring AREs. We used the coordinates of this complex to model the human TTP TZF domain in complex with the same oligonucleotide, and we found that the amino acid residues in contact with the RNA were identical in the two proteins, increasing the likelihood that this modeled TTP TZF domain-RNA complex is valid (10).
An important question that arises is: What is the relative contribution of each ARE nonameric TTP-binding site to TTP-mediated RNA turnover in naturally occurring mRNAs? In one naturally occurring target, the TNF mRNA, there are often five partly overlapping nonamers in the mammalian transcripts, with some variation among mammalian species (10). There are fewer nonamers in the GM-CSF mRNA, and in the most common mammalian pattern, three nonamers overlap significantly (10). Perhaps coincidentally, GM-CSF mRNA in bone marrowderived stromal cells has a half-life of 99 min (3), whereas TNF in bone marrow-derived macrophages from normal mice has a half-life of 39 min (2).
The main objective of the present study was to determine whether an individual TTP-binding site conferred instability on an otherwise stable transcript, and whether two tandem binding sites conferred even greater instability. We used a well defined core sequence of 24 bases from the mouse TNF ARE, which in its unmutated state is able to bind two tandem molecules of TTP TZF peptide per molecule of RNA (7). This core sequence was then mutated at one or both of the As in the nonamer, or the central Us, to permit 2, 1, or 0 mol/mol of peptide binding, and the effects of these mutations were evaluated in cell cotransfection assays with full-length TTP and in cell-free deadenylation assays. The results suggest that even a single nonamer TTP-binding site can confer some instability on an otherwise stable mRNA, but that two makes the transcript much more unstable. In addition, they indicate the importance of the central U residues in binding of TTP to RNA. This binding and the ability of the TTP to degrade RNA were greatly decreased when these U residues were changed to Cs, even though the central U residue (U5), like the terminal U9, is thought to form only a single hydrogen bond with an amino acid in the protein (9), and thus might be assumed to be replaceable with minimal loss of activity.

Plasmid Constructs
Expression plasmids CMV⅐hTTP⅐tag and its tandem zinc finger (TZF) domain alone (CMV⅐hTTP-(97-173)⅐tag) were made as described (11). The CMV⅐HuR⅐tag, which expresses the widely expressed human ELAV-like HuR protein (12), was constructed as follows. A cDNA coding for the open reading frame of HuR was made using reverse transcription-PCR from HeLa cell total RNA. The 5Ј primer for the PCR amplification was 5Ј-ACGTggtaccACAATGTCT-AATGGTTATGAAGACC-3Ј, and the 3Ј primer was 5Ј-AAGctcgagT-TAAGCGTAATCCGGGACGTCGTATGGGTATTTGTGGGACTT-GTTGG-3Ј, where the lowercase letters indicate the restriction sites for Asp718 and XhoI, respectively. The underlined letters represent the initiator methionine (5Ј primer) and the stop codon (3Ј primer). The italic letters in the 3Ј primer encode the hemagglutinin (HA) epitope. The resulting PCR product was digested with the restriction enzymes and ligated into the Asp718 and XhoI sites of the expression vector CMV⅐BHG3Ј/BSϩ (5). The sequence of the HuR insert corresponds to bp 116 -1196, from Ϫ3 bp before the start codon to the last amino acid, of GenBank TM accession number U38175. CMV⅐HuB⅐FLAG, which expresses the human ELAV-like neuronal protein 1(Hel-N1, HuB), was made by PCR using a plasmid kindly provided by Dr. Jack Keene, Duke University (13), as a template. The 5Ј primer for the PCR amplification was 5Ј-ACGTggtaccATGGAAACACAACTGTCTAATGG-3Ј, and the 3Ј primer was 5Ј-AAGctcgagTTACTTGTCATCGTCGTCCTTGT-AATCGGCTTTGTGCGTTTTGTTTGTC-3Ј, where the lowercase letters indicate the restriction sites for Asp718 and XhoI, respectively. The underlined letters represent the initiator methionine codon (5Ј primer) and the stop codon (3Ј primer). The italic letters in the 3Ј primer encode the FLAG epitope. The resulting PCR product was digested with the restriction enzymes and ligated into the Asp718 and XhoI sites of the expression vector CMV⅐BHG3Ј/BSϩ. The insert sequence of HuB corresponds to bp 585-1661 (from the initiation codon to the last amino acid) of GenBank TM accession number U12431. The correct sequences of both the HuR⅐tag and HuB⅐FLAG DNA inserts were confirmed by dRhodamine terminator cycle sequencing (PerkinElmer Life Sciences).
To make the mutant TNF␣-(1309 -1332) (A) 50 probes, a series of double-stranded oligonucleotides that encoded the corresponding RNA sequences listed in Fig. 1A was made to substitute for the wild-type TNF␣-(1309 -1332) sequence in the EcoRV and XbaI sites of pTNF␣-(1309 -1332) (A) 50 /SKϪ. The templates for probes were PCR-amplified (primers M13 forward and T50Xba) from the plasmids and were sequenced to verify the 3Ј ends, as described (14). The template for probe ARE was prepared by linearizing plasmid pTNF␣-(1309 -1332) with XbaI, and for probe V by linearizing vector SKϪ.
The RNA probes were transcribed in the presence of [␣-32 P]UTP (800 Ci/mmol). Linearized plasmids or PCR amplification products were used as templates, and the Promega riboprobe in vitro transcription systems protocol was employed. The resulting probes were separated from the free nucleotides using G-50 columns for RNA gel shift assays; for cell-free deadenylation assays, the probes were purified from urea-TBE gels.
Transfection of HEK 293 Cells and Preparation of Cell Extracts-To each 100-mm plate of 293 cells was added 5 g of vector DNA (BSϩ) alone, or 0.2 g of CMV⅐hTTP⅐tag, or 2 g of the TZF domain expression plasmid CMV⅐hTTP-(97-173)⅐tag (11), or 0.5 g of CMV⅐HuB⅐FLAG, or 0.5 g CMV⅐HuR⅐tag. Vector DNA (BSϩ) was added to each plate to make the total amount of co-transfected DNA 5 g per plate by using the calcium-phosphate precipitation method. Twenty four h after the removal of the transfection mixture, the cell monolayers were rinsed with ice-cold Ca 2ϩ -and Mg 2ϩ -free phosphatebuffered saline and then scraped into phosphate-buffered saline. After centrifugation at 600 ϫ g for 3 min at 4°C, the cell pellet was gently rinsed in ice-cold diethyl pyrocarbonate-treated water containing 8 g/ml leupeptin, 0.2 mM phenylmethylsulfonyl fluoride, and 1 g/ml pepstatin A. The cells were lysed in a hypotonic buffer containing 10 mM HEPES (pH 7.6), 5 mM KCl, 5% (v/v) glycerol, 0.25% (v/v) Nonidet P-40, 1 g/ml pepstatin A, 0.1 mM phenylmethylsulfonyl fluoride, and 8 g/ml leupeptin. After a 15-min centrifugation for 15,000 ϫ g at 4°C, KCl and glycerol were added to the supernatant to achieve final concentrations of 40 mM and 15% (v/v), respectively. The cell extracts were stored at Ϫ70°C. The expression of proteins from these constructs was determined by Western blotting with an antibody directed at the HA epitope tag (1:2,000, Santa Cruz Biotechnology) or the FLAG epitope tag (1:10,000, Sigma), essentially as described (11).
In Vitro Deadenylation Assay-Extracts prepared from 293 cells as described above (10 g of protein) were incubated with 5 ϫ 10 4 cpm of RNA probe as described (14). Aliquots of reaction products were analyzed on 6% acrylamide gels containing 7 M urea. The resulting gels were exposed to x-ray films and were analyzed by PhosphorImager to quantify the amount of intact probe remaining.

Co-transfection Assays
Plasmids-A series of plasmids was constructed with each containing the following components: a CMV promoter, the protein coding sequence of the mouse MARCKS-like protein (MLP), used because both the protein and mRNA are relatively stable in the 293 cell cotransfection system; an HA epitope tag attached in-frame to the MLP protein sequence; a 24-base RNA insertion corresponding to the core ARE of human and mouse TNF; and the 3Ј-UTR and polyadenylation signal sequence from the bovine growth hormone mRNA 3Ј-UTR (see Fig. 1B). This was the parent plasmid for subsequent mutants, and served as the "target" for TTP-mediated mRNA turnover in the cell transfection assays. The various mutant forms of this plasmid and their construction are described in detail below, but they basically represented a group of substitution mutants of the wild-type plasmid described above at the 24-base ARE region. They include a series of mutant ARE-containing plasmids based on the mutants of the TNF ARE described in Blackshear et al. (7), as well as two novel mutations of the middle U residue in some of the UUU triplets. A schematic representation of these target plasmids is shown in Fig. 1B.
For the parent MLP expression plasmid CMV⅐MLP-HA, the protein coding region of mouse MLP (GenBank TM accession number NM_010807) was fused at its carboxyl terminus in-frame with the HA epitope coding sequence and inserted into CMV⅐BGH3Ј/BSϩ (5). To create the fusion insert, plasmid pF52.ab, consisting of the mouse Mlp cDNA (16), was used as a PCR template. The 5Ј primer for PCR amplification was 5Ј-AATTggatccCATCATGGGCAGCCAG-3Ј, and the 3Ј primer was 5Ј-TATGctcgagcTTAAGCGTAATCCGGGACGTCGTA-TGGGTACTCATTCTGCTCAGCACTGG-3Ј. The lowercase letters in the primers indicate the restriction sites for Asp718 and XhoI, respectively; the sequences representing the initiator methionine and the stop codon are underlined. The uppercase letters contain the sequences bp 182-198 and bp 767-786 from GenBank TM accession number NM_010807, respectively. The italic sequence encodes the HA epitope. The resulting PCR product was an ϳ0.6-kb cDNA that was digested with Asp718 and XhoI.
To make CMV⅐MLP-HA (ARE), which contained the mouse TNF core ARE (bp 1309 -1332 of GenBank TM accession number X02611, identical to the core ARE of human TNF, see bp 1341-1364 of Gen-Bank TM accession number NM_000594.2), a double-stranded oligonucleotide TTATTTATTTATTATTTATTTATT, including the XhoI and XbaI sequences at its 5Ј or 3Ј ends, respectively, was inserted between the XhoI and XbaI sites 3Ј to the stop codon of CMV⅐MLP-HA. This is designated as plasmid AAAAAA, with each of the six As representing the consecutive As in the native ARE sequence (Fig. 1). The first set of mutant plasmids, representing derivatives of CMV⅐MLP-HA in which single or multiple As in the 24-base ARE (Fig. 1B) were mutated to Cs, was made in a similar fashion. In each case the base substitution is underlined. The last two mutant plasmids contained mutations to the same ARE in which some of the As and the middle U of one or more UUU motifs was replaced by C (Fig. 1, A and B) so that the ARE contained one or two UUAUCUAUU nonamers. Correct insertion of the inserts in the vector, and the correct sequences of the intended mutations, were confirmed by dRhodamine terminator cycle sequencing.
Cell Transfection Assays in 293 Cells-These were performed exactly as described (5). The cells were co-transfected with the vector alone plasmid (BSϩ), the wild-type human TTP-expressing plasmid CMV⅐hTTP⅐tag, the human HuB expression plasmid CMV⅐HuB⅐FLAG, and the various target MLP plasmids. The amount of transfected DNA was adjusted so that each plate was transfected with the same amount of DNA (5 g); this was made up of 10 ng of CMV⅐hTTP⅐tag or 30 ng of CMV⅐HuB⅐FLAG and 0.5 g (or otherwise indicated) of MLP plasmids, and the total of 5 g was made up by BSϩ. The cells were harvested, and total cellular RNA was prepared using RNeasy (Qiagen). For Northern blotting, the probes consisted of either a mouse MLP cDNA probe consisting of bp 281-1429 of GenBank TM accession number NM_010807 or a mouse TTP cDNA probe (5), with or without the HuB DNA probe. Northern blotting was performed as described (5).
In some experiments, the total cellular RNA was used for transcript level determination by real time PCR, so that large numbers of experiments could be evaluated conveniently. For this purpose, RNA isolated from transfected 293 cells was treated with DNase to remove plasmid DNA. This was done by treating 1.2 g of RNA with 5 l of RQ1 DNase (1 unit/l; Promega) in 3.35 l of buffer (final concentration, 33.3 mM Tris-HCl (pH 8.3), 50 mM KCl, 2 mM MgCl 2 ); the total reaction volume was 25 l. The RNA and DNase mixture was incubated for 15 min at room temperature. After addition of 2.5 l of 25 mM EDTA to terminate the reaction, the mixture was heated at 65°C for 10 min to inactivate the DNase and then immediately placed on ice for 5 min and centrifuged, and the DNase-treated RNA was converted to cDNA using a high capacity cDNA archive kit (Applied Biosystems). 40 l of cDNA was made by incubating 20 l of DNase-treated RNA with 20 l of reverse transcription mix according to the manufacturer's instructions (per 40 l of reaction: 4 l of 10ϫ buffer, 1.6 l of 25ϫ dNTPs, 4 l of 10ϫ random primers, and 0.5 l of reverse transcriptase, 50 units/l). The reaction was incubated at 25°C for 10 min and then at 37°C for 2 h.
Real Time PCR Analysis-Levels of the bovine growth hormone mRNA sequences expressed in the transfected cells were analyzed by quantitative real time PCR. Primers and the 6-carboxyfluoresein-labeled minor groove binder probe were designed using Primer Express software and synthesis (Applied Biosystems). The primer sequences were as follows: forward 5Ј-GTTGCCAGCCATCTGTTGTTT-3Ј; reverse 5Ј-GACAGTGGGAGTGGCACCTT-3Ј; probe 5Ј-CCTC-CCCCGTGCCTTCCTTGA-3Ј; these corresponded to bp 716 -736, 767-786, and 740 -760, respectively, of the bovine growth hormone mRNA sequence (GenBank TM accession number NM_180996.1).
The Taqman reactions were as follows: 3 l of cDNA, 12.5 l of Taqman Universal PCR Master mix (Applied Biosystems), 1.25 l of probe/primer mix (5 M probe, 18 M each primer), and 8.25 l of H 2 O. The reactions were performed in 96-well plates in an Applied Biosystems Prism 7700 real time PCR instrument. The thermal cycle conditions were as follows: 2 min at 50°C and 10 min at 95°C, followed by 40 cycles at 95°C for 15 s each and 60°C for 60 s. Each sample was analyzed in duplicate. A positive control cDNA was analyzed on each plate in order to minimize plate-to-plate variations. A negative control (293 cells transfected with a TNF plasmid alone) cDNA was also assayed on each plate in order to remove any contribution from nonspecific 293 cell background.
A modified version of the ⌬⌬Ct method of relative quantification (17) was used to determine the fold change in expression in ARE and TTPtransfected cells versus ARE and empty vector control-transfected cells (BS). Briefly, this was done by normalizing the resulting threshold cycle (Ct) values of the samples to the Ct values obtained with the negative control (293 cells transfected with TNF plasmid) to account for 293 cell background (⌬⌬Ct ϭ Ct sample Ϫ Ct 293 control ). The background-normalized Ct values were then calibrated to the Ct values from empty vector control (BS) and ARE-transfected cells that were treated at the same time as the TTP-and ARE-transfected cells; ⌬⌬Ct ϭ ⌬Ct ARE sample Ϫ ⌬Ct BS sample . The fold change in expression was then obtained (2 Ϫ(⌬⌬Ct) ).
Statistical Analysis-The mean fold changes in expression in the six groups of ARE and TTP co-transfected cells (normalized to BS control transfected cells) were analyzed with SAS Enterprise Guide 2.0 software. Differences in real time PCR mRNA levels among all six groups of transfected cells were determined by analysis of variance followed by Tukey-Kramer post hoc analysis, using ␣ level ϭ 0.05. In some experiments, the expression of the MLP protein was determined by Western blotting with an antibody directed at the HA epitope tag. OCTOBER 7, 2005 • VOLUME 280 • NUMBER 40

TTP or TZF Domain Peptide Binding to Mutant ARE Probes-We
showed previously by gel electrophoretic mobility shift assays that both full-length TTP and its TZF domain peptide expressed in 293 cells could bind to the ARE sequence of the TNF␣-(1309 -1332) (A) 50 probe (ARE-A 50 ) (7,14). In the present study, a series of mutant probes was made in which the flanking As of the AUUUA pentamers were mutated to Cs (Fig. 1A); the number of nonamer(s) UUAUUUAUU found in these FIGURE 1. A, ARE portion of wild-type and mutant RNA probes. Plasmids for transcribing the wild-type or the mutant ARE-A 50 probes were made as described under "Experimental Procedures." The names for each ARE-A 50 probe and the RNA sequences comprising the ARE are listed. The underlined RNA nonamers represent the optimal TTP-binding motifs. The octamer in probe ACAACA-A 50 is indicated by the dashed line. In the last column are shown the number of moles of TTP73 peptide bound to the probes, as determined previously (7). N.A., not analyzed. B, schematic representation of TTP target plasmid constructs. Constructs were made as described under "Experimental Procedures." MLP-HA contains the protein coding region of mouse MLP (represented by the solid bar) fused at its carboxyl terminus in-frame with the hemagglutinin (represented by the empty bar) epitope coding sequence. This was driven by the CMV promoter; the protein coding sequence was followed by a 24-base portion of the mouse TNF ARE, and then the bovine growth hormone 3Ј-untranslated region (BGH3Ј). Below the diagram are the names of the constructs (left column) and the DNA inserts (right column) encoding the wild-type (WT) or the mutant AREs. Cs are underlined that replaced A or U in the wild-type sequence.
probes is listed in Fig. 1A. A pair of mutant probes with central U mutations was also made (Fig. 1A); and the number(s) of UUAUCUAUU nonamers in them is also listed (Fig. 1A). Note that for the last two central U probes there are no intact original nonamers remaining in either probe (Fig. 1A). These probes were digested with RNase T1 to release the 24-base ARE from the vector sequence and the poly(A) tail, and were then used in gel shift analysis with extracts prepared from 293 cells transfected with TTP or its TZF domain expression constructs. Extracts from 293 cells transfected with constructs expressing members of the Hu family ARE-binding proteins, HuB and HuR (human members of the ELAV family; for review, see Ref. 18), were also used in the assays.
For the A to C mutant probes, we previously determined the apparent mole of synthetic TTP73 peptide that bound per mol of the respective ARE probe (Fig. 1A, right-hand column) (7). These results were confirmed in the present study by gel shift analysis, using a TZF domain peptide (77 amino acids) that was expressed in 293 cells, with the gel shifts being performed with cytosolic extracts from those cells ( Fig. 2A). For the U to C mutant probes, whose sequences and abbreviations are shown in Fig. 1A, binding to both full-length TTP and its TZF domain peptide is shown in Fig. 2B. These gel shift data were used to calculate the apparent mol/mol of the TZF peptide bound to these RNA probes (Fig. 1A, right-most column), with the presence of the lower band indicating 1 mol per mol of binding and the upper band indicating 2 mol per mol.
The wild-type sequence is labeled AAAAAA to indicate the six A residues in the 24-base sequence; although it contains four possible nonamer binding sites, it was found previously to bind only 2 mol of TTP73 per mol in two nonamer UUAUUUAUU sequences at each end of the 24-base ARE sequence (7). A similar result was obtained in the present experiment ( Fig. 2A, lanes 1-6), in which lane 1 shows the migration position of the free probe in the gel shift assay; lane 2 shows the effect of 293 cell extract proteins after transfection with vector alone; lane 3 shows the migration position of full-length TTP bound to probe, and lane 4 shows the formation of two complexes with the TTP-(97-173) peptide (TZF domain peptide) expressed in the 293 cell extracts. In Fig. 2A, the lower band in lane 4 is thought to represent a complex formed of 1 mol of peptide/mol of probe (presumably comprised of peptide binding to each of the two potential binding sites), and the upper band is thought to represent a complex consisting of 2 mol of peptide per mol of RNA probe, in which both binding sites are occupied (7). Binding of full-length TTP to the wild-type probe also resulted in the formation of two major complexes ( Fig. 2A, lane 3). However, at this time we cannot conclude that the upper band represents a complex consisting of 2 mol of protein per mol of probe, and the exact relationship between the two complexes remains to be determined. In Fig. 2A, lanes 5 and 6 show the migration of RNA-HuB or RNA-HuR complexes, using extracts from 293 cells transfected with those expression constructs, respectively. The RNA-HuR complex migrated to the same position as complex II, which formed between an endogenous 293 cell protein and the probe. Complex II was as abundant as that shown in Fig.  2A, lane 6, when 50 g of vector alone-transfected cell extract was used (not shown), suggesting that complex II possibly contained endogenous 293 cell HuR protein. In contrast, in the presence of large amounts of 293 cell protein, the intensity of complex I remained low with this probe (not shown).
The mutant probe AACCAA, which also contained two normal nonamer binding sites at each end, could also bind 2 mol per mol of synthetic TTP73 peptide (Fig. 1A) and of the TZF domain peptide expressed in 293 cells (Fig. 2A, lane 9). With this probe, full-length TTP formed almost identical RNA-TTP complexes as those seen with probe AAAAAA (Fig. 2A, compare lane 8 to lane 3). Thus, despite the presence of four potential nonamer binding sites in the AAAAAA probe (Fig. 1A), binding appeared to occur only to two of them, presumably the two sites remaining in AACCAA. The complexes formed with either HuB or HuR and this mutant probe were similar to those formed with the wild-type probe ( Fig. 2A, lanes 10 and 11).
Three of the other A to C mutant ARE probes shown in Fig. 2A (AAAACC, AAACCA, and AACCCA) could bind only 1 mol of synthetic TTP73 peptide per mol of oligonucleotide (7). This result was confirmed in the present studies with the TZF domain peptide expressed in 293 cells, with only a single complex being formed in each case ( Fig. 2A, lanes 16, 21, and 28). Thus, in these three cases, the single site binding data from the transfected-expressed peptide agreed with those obtained with the TTP73 synthetic peptide (7). However, there was also a less abundant, slower migrating complex formed with each of these probes and the full-length TTP protein, at approximately the same position as the higher band seen with the wild-type probe (see below). Although there were two possible overlapping binding UUAUUUAUU nonamers in both probes AAAACC and AAACCA (Fig. 1A), there was only one apparent RNA-probe complex formed with the TZF domain peptide in each case (7) (Fig. 2A). The migration patterns of the single HuB-or HuR-RNA complexes with the three mutant probes were very similar to those seen with the wild-type probe ( Fig. 2A, compare lanes  17 and 18, 22 and 23, and 29 and 30 to lanes 5 and 6).
Mutant probe ACAACA lacked the three internal U residues of the essential nonamer unit but contained an octamer UUAUUAUU (Fig.  1A). This change to two internal U residues was found to decrease the affinity of the TTP73 peptide for the mutant probe by about 6-fold when compared with the nonamer (8). Under the present gel shift conditions, the expressed TZF peptide did not appear to form a complex with this probe (Fig. 2A, lane 33). However, a slower migrating, low intensity protein-RNA complex was formed between this probe and full-length TTP that was similar to those seen with the above-mentioned three mutant probes ( Fig. 2A, lanes 15, 20, 27, and 32), even though the major, lower RNA-TTP complex did not form with this probe, because the intensity of the band in that location was no different from that seen with the vector alone-transfected cell extract ( Fig. 2A, compare lane 32  to 31, complex I). The binding of HuB and HuR to this mutant probe was again similar to their binding to the wild-type probes ( Fig. 2A, lanes 34  and 35).
As predicted from the data with the TTP73 synthetic peptide (7), there was essentially no binding of the expressed TZF domain peptide to the last two mutant probes (ACCACA and CCCCCC) ( Fig. 2A, lanes 40  and 45). Again, full-length TTP appeared to form a smaller amount of the slower migrating complex with the first of these two probes (Fig. 2A,  lane 38), and exhibited essentially no binding to the last probe (lane 44) under these conditions. The amount of protein-RNA complex formed between HuB and these probes was very similar to that seen with the wild-type probe ( Fig. 2A, lanes 41 and 46), but the HuR-RNA complex formation was somewhat decreased with probe ACCACA (lane 42), and even more decreased with probe CCCCCC (lane 47).
As listed in Fig. 1A, two additional probes were made that contained one or two UUAUCUAUU nonamers in the 24-base ARE. Similar gel shift experiments using these mutant probes are shown in Fig. 2B. As in Fig. 2A, the wild-type AAAAAA probe shown in lanes 1-6 bound both full-length TTP (lane 3) and the expressed TZF domain peptide (lane 4), in each case with formation of the characteristic two bands. When the UUAUUUAUU motifs in the ARE sequence were mutated to UUAU-CUAUU, there was essentially no RNA-TZF peptide complex formation with either probe (AUCUA)2, containing two UUAUCUAUU motifs, or probe (AUCUA)1, containing one motif (Fig. 2B, lanes 9 and  16). As seen with the other mutant probes, a low intensity upper band formed between full-length TTP and both mutant probes (Fig. 2B, lanes  8 and 15); however, essentially no detectable lower complex formed with full-length TTP (Fig. 2B, compare lane 8 to 7 and 15 to 14). The preserved ability of full-length TTP to bind to these mutant probes ((AUCUA)2 and (AUCUA)1, as well as ACAACA), to which the expressed TZF peptide was not able to bind, was somewhat surprising, because there were no intact nonamers in any of these mutants (Fig.  1A). The binding of HuB to these probes was fairly similar to that of the wild-type probe (Fig. 2B, lanes 10 and 17), but the binding of HuR was obviously decreased (lanes 11 and 18).
When the number of Cs substituted for flanking As or center Us in the AUUUA motif was increased to three or more, there was increased binding of an endogenous 293 cellular protein to the probes, shown in Fig. 2, A and B, as complex I ( Fig. 2A, lanes 26, 39, and 43; Fig. 2B, lanes  7 and 14). The lower TTP-RNA complex co-migrated with this endogenous protein complex I ( Fig. 2A, lanes 39 and 44), possibly masking the appearance of the lower TTP-RNA complex with probes ACCACA-A 50 or CCCCCC-A 50 . However, significant binding of TTP to either of these probes seems unlikely, because there was no observable binding of the TZF domain peptide to these probes, and TTP did not induce deadenylation of these mutant probes, nor did it destabilize the hybrid mRNAs containing these mutated AREs in cell transfection studies (see below).
Another observation from these gel shift assays is that the formation of the TTP-RNA complex that migrated to the position of complex I required the presence of the perfect nonameric sequence UUAUU-UAUU in the RNA probe as seen in probes AAAAAA, AACCAA, AA-AACC, AAACCA, and AACCCA ( Fig. 2A, lanes 3, 8, 15, 20, and 27). This sequence was not found in probes ACAACA, ACCACA, CCC-CCC, (AUCUA)2, and (AUCUA)1, and with these probes there was no formation of this faster migrating TTP-RNA complex ( Fig. 2A, lanes 32,  38, and 44; Fig. 2B, lanes 8 and 15).
It is interesting to note that, although the binding of HuB did not seem to be affected by the mutated nucleotides in any of the mutant probes tested here, the binding of HuR seemed diminished somewhat with probes that did not contain any UUAUUU or UUUAUU motifs (Fig. 1A), such as in probes CCCCCC or (AUCUA)2 ( Fig. 2A, lane 47;  Fig. 2B, lane 11), or probes containing only one of the pentamers UUAUUU or UUUAUU, as in probe (AUCUA)1 (Fig. 2B, lane 18). Fig.  2C shows Western blots of 293 cytosolic extracts from cells transfected with expression constructs of these ARE-binding proteins to document their expression.
To assess whether the upper band formed with TTP-expressing extracts and mutant probes that did not contain a perfect UUAUUUAUU motif and did not bind the expressed TZF peptide (probes ACAACA, ACCACA, (AUCUA)2, and (AUCUA)1) actually contained TTP, we performed "supershift" assays with the anti-HA antibody (Fig. 3). Both of the upper RNA-protein complexes formed with expressed full-length TTP, and probes AAAAAA or AACCAA (Fig. 3A, lanes 3 and 6) migrated slower and appeared in the gel as supershifted bands (lanes 4 and 7), indicating that protein component of the double band complexes formed with either probe was indeed TTP. The amount of RNA-TTP complex shifted to a slower migrating position (Fig. 3A, SS) was dependent upon the amount of antibody present (data not shown). No supershifted band was formed when the anti-HA antibody was incubated with 50 g of extract protein (a 10-fold greater amount of protein than used for the usual assay) from vector alonetransfected 293 cells incubated with any probes (not shown).
As in Fig. 2A, TTP formed two complexes with probes AAAACC, AAACCA, or AACCCA, one migrating as an upper band of low intensity and a slightly more intense lower band (Fig. 3A, lanes 10,  15, and 18). When the anti-HA antibody was included in the incubation, the upper RNA-TTP complex formed with any of these probes was supershifted (Fig. 3A, lanes 11, 16, and 19; indicated as SS). However, although the lower complexes formed with the AAAAAA and AACCAA probes were apparently supershifted in this assay (Fig. 3A, compare lanes 3 and 4 and lanes 6 and 7), the lower complexes formed with the other probes did not supershift (compare lanes 11 to 10, 16 to 15, and 19 to 18, indicated as I), even when the amount of antibody was increased in the assays (not shown). The identity of this faster migrating complex formed with these mutant probes in the presence of full-length TTP is unclear. As shown in both Fig. 2 and Fig. 3, this complex co-migrated with that formed between the probes and an endogenous 293 cell protein; it is conceivable that the presence of TTP could increase the binding of the endogenous 293 cell protein to these probes. Another possibility, suggested by the presence of lower M r proteolytic fragments of TTP observed in the transfected cell extracts blotted with the antibody to the carboxyl-terminal HA epitope tag, is that an untagged aminoterminal fragment of TTP was present that contained the TZF domain and caused the gel shift. The presence of such a fragment has been noted in similar cell extracts probed with an antibody to the intact TTP protein rather than to the HA epitope tag. However, such a possibility seems highly unlikely, because both of the TTP-RNA complexes formed with probes AAAAAA or AACCAA were supershifted in the presence of the anti-HA antibody (Fig. 3A, lanes 4 and For each probe, a lane was loaded with a probe sample that was incubated under the same conditions but without the added cell extract (PЈ). The gels were run at 250 V for 115 min to allow the supershifted complexes to run into the gels. FP, free probe. OCTOBER 7, 2005 • VOLUME 280 • NUMBER 40 7). Further experiments will be necessary to establish the identity of the protein component of the lower band with certainty.

Tristetraprolin-binding Sites and mRNA Turnover
Probes ACAACA and ACCACA did not contain a UUAUUUAUU motif, although ACAACA contained an octamer UUAUUAUU. Nevertheless, there was still detectable upper complex formation between TTP and both of these probes (Fig. 3A, lanes 22 and 27). Even though the binding was weak, as shown by the considerable amount of free probe remaining in each case (Fig. 3A, lanes 22 and 27), this complex nonetheless supershifted when the anti-HA antibody was included in the assay (lanes 23 and 28). The intensity of complex I formed with probe CCCCCC was very similar when extracts from 293 cells either transfected with vector alone or with full-length TTP were used (Fig.  3A, lanes 29 and 30), and this complex did not supershift when the anti-HA antibody was included in the incubation mixture (lane 31).
In probes (AUCUA)2 and (AUCUA)1, the ARE had been changed to contain two or one mutant nonamer UUAUCUAUU (Fig. 1A), with no intact wild-type nonamers. As in Fig. 2B, these formed low intensity RNA-TTP complexes that migrated to the same position as the upper band when wild-type probe AAAAAA was used (Fig. 3B, lanes 6 and 10,  compared with lane 3). The upper complex formed with these probes also supershifted with the anti-HA antibody (Fig. 3A, lanes 7 and 11, indicated as SS) indicating that the upper band contained TTP, whereas the intensity of the lower band was unchanged.
Effects of ARE Mutations on TTP-stimulated Transcript Deadenylation in a Cell-free Assay-We then tested the same extracts from 293 cells transfected with either vector alone (BSϩ) or TTP in a cell-free deadenylation assay. In this experiment, each target RNA consisted of 58 bases transcribed from the SKϪ vector, 24 bases of the same wildtype or mutant ARE sequences used in the gel shift reactions, and 50 bases of a poly(A) tail (the ARE sequence of each probe is listed in Fig.  1A). Each RNA target was incubated with 293 cell extracts for 60 min at 37°C in the presence or absence of 20 mM EDTA. At this concentration of EDTA, the RNA probes were stable in extracts prepared from cells transfected with vector alone or with other constructs that expressed a variety of different proteins, including TTP (not shown). In the absence of EDTA, the disappearance of the RNA target and the accumulation of the deadenylated RNA were monitored by electrophoresis and autoradiography, as described previously (14). The amount of target RNA remaining was quantified by PhosphorImager analysis, and the probe radioactivity in the absence of EDTA was compared with that measured in the presence of EDTA for each individual RNA probe. The migration positions of the ARE probe that did not have a poly(A) tail (Fig. 4, lanes
The effect of TTP on the wild-type AAAAAA-A 50 probe in this assay is shown in Fig. 4A, lanes 4 -6. When extracts from 293 cells transfected with vector alone (Ϫ) were incubated with the AAAAAA-A 50 probe for 1 h in the absence of EDTA, there was a slight decrease in the amount of full-length probe and no detectable increase in the accumulation of the deadenylated species (Fig. 4A, compare lane 5 to lane 4). In four individual experiments, this decrease averaged 22.6 Ϯ 1.9% (S.D.) when compared with the identical incubation in the presence of EDTA. However, when identical amounts of 293 cell protein from cells transfected with TTP (ϩ) were used, there was an obvious decrease in the amount of the remaining full-length probe, as well as the accumulation of the deadenylated RNA, indicated by the arrow (Fig. 4A, lane 6). This decrease averaged 62.1 Ϯ 1.6% (n ϭ 4 experiments) compared with the sample in the presence of EDTA (TABLE ONE ). When the decreased amount of full-length probe in extracts from cells transfected with TTP was compared with that seen in extracts from cells transfected with vector alone (both in the absence of EDTA), the average decrease was 2.74-fold greater in the presence of TTP (TABLE ONE). Virtually identical results were observed when the mutant ARE probe was used that contained two tandem nonamer TTP73-binding sites (AACCAA-A 50 ) (Fig. 4A, lanes 7-9; see also TABLE ONE).
In contrast, when mutant AREs were used that contained a single binding site for TTP73, TTP caused less disappearance of the full-length probe and less accumulation of the deadenylated RNA when compared with the probes containing two TTP73-binding sites (Fig. 4A, lanes  10 -12 for AAAACC-A 50 ; lanes 13-15 for AAACCA-A 50 ; lanes 16 -18  for AACCCA-A 50 ). When the two TTP73 nonbinding mutants were

Effect of ARE mutations on TTP-stimulated transcript deadenylation in a cell-free assay
Shown in the 3rd column are the averages from four independent deadenylation experiments, as described in the legend to Fig. 4. The volume of the remaining intact probe in each sample was estimated by PhosphorImager analysis, and the percentage by which the amount of each intact probe decreased was calculated by comparing the volume of the remaining probe in the absence of EDTA to that in the presence of EDTA (EDTA ϭ 100%, Fig. 4). Shown in the 4th column are the ratios representing the fold change by comparing the percentage of probe decrease in extracts from TTP (or other expression constructs)-transfected cells to that in extracts from cells transfected with vector BSϩ alone.
used, ACCACA-A 50 (Fig. 4A, lanes 19 -21) and CCCCCC-A 50 ( lanes  22-24), there was no apparent effect of TTP to decrease the probe levels or increase the accumulation of the deadenylated probe (see also A different series of deadenylation assays was performed using mutant probes that contained either two nonamers or one nonamer in which the center U in the AUUUA motif was substituted by C or a probe in which there was one UUAUUAUU octamer (Fig. 1A). In addition to extracts from 293 cells transfected with vector alone or TTP, extracts from 293 cells transfected with the CMV⅐hTTP-(97-173)⅐tag, expressing the 77-amino acid TZF peptide, and constructs expressing the Hu family proteins HuB or HuR, were used. The results with these expressed proteins using the wild-type probe AAAAAA-A 50 are shown in Fig. 4B, lanes 4 -9. The ability of TTP to increase AAAAAA-A 50 probe degradation, and cause the accumulation of the deadenylated RNA, was similar to the series shown in Fig. 4A (Fig. 4B, lanes 4 -6; TABLE ONE). As we have shown previously (14), the presence of the TZF domain peptide did not induce more probe degradation when compared with vector alone (Fig. 4B, compare lane 7 to 5), with the fold change in the presence of this peptide averaging 1.14-fold of control (TABLE ONE). When compared with extracts from cells transfected with vector alone, there was no change in probe stability in the presence of HuB or HuR (Fig. 4B, lanes 8 and 9 compared with lane 5), with no decreases in probe observed compared with control (TABLE ONE). This result is in line with a previous report that HuR had little effect on the rate of deadenylation of mRNAs containing AREs (19).
The three mutant probes in this series did not bind the expressed TZF peptide, but their binding to HuB or HuR (probe ACAACA) was similar to that seen with the wild-type probe. The gel shift assays had demonstrated weak binding to full-length TTP. In the presence of full-length TTP, their degradation was slightly increased when compared with that seen with the vector alone-transfected cell extract (Fig. 4B, lanes 12, 18,  and 24 compared with lanes 11, 17 and 23; TABLE ONE). The expressed TZF peptide and the Hu proteins did not stimulate the degradation of these probes (Fig. 4B, lanes 13-15, 19 -21, and 25-27 ; TABLE ONE). Fig.  1B, the TTP target constructs for co-transfection in 293 cells consisted of a CMV promoter, the protein coding region of mouse MLP, an HA epitope tag in-frame with the carboxyl-terminal amino acid of MLP, a 24-bp sequence encoding the wild-type or a mutant ARE, and the bovine growth hormone 3Ј-UTR and polyadenylation signal. The ARE consisted of the same 24 bases of core TNF sequence whose binding to the TTP73 synthetic TZF domain of human TTP was described (7), and which was present in the probe used in the gel shift and deadenylation assays.

Effect of ARE Mutations on MLP-HA mRNA Stability in the Presence and Absence of TTP in Co-transfection Experiments-As shown in
These constructs were co-transfected into 293 cells in the presence or absence of TTP, and the effect of TTP on deadenylation of the target mRNA was assessed by Northern blotting and real time PCR. A control used the same construct in which the 24-base ARE sequence was omitted altogether. As shown in Fig. 5, co-transfection of TTP with the MLP construct containing the AAAAAA probe sequence (wild-type ARE) resulted in a marked decrease in MLP mRNA accumulation in 293 cells (Fig. 5, compare lane 2 to lane 1, top panel). The MLP artificial construct was used in preference to the previously used TNF, GM-CSF, and interleukin-3 constructs because there was no detectable accumulation of a deadenylated mRNA fragment, which can complicate PhosphorImager analysis of the mRNA levels. PhosphorImager analysis of this experiment indicated a 66% decrease in steady-state MLP mRNA levels in the presence of TTP when the mRNA contained two TTP73 nonamer binding sites. There was a commensurate decrease in expressed MLP protein (Fig. 5, bottom panel, lanes 1 and 2) in the presence of TTP, as assessed by immunoblotting with an antibody to the HA epitope. The expression level of TTP mRNA is shown in the middle panel of Fig. 5.
Co-transfection and expression of TTP caused a similar decrease (68%) in steady-state MLP mRNA levels when the AACCAA mutant was present, which also contains two TTP73-binding sites (Fig. 5, top  panel, lanes 3 and 4). This also resulted in marked decreases in the expression of MLP protein (Fig. 5, bottom panel), whereas the expression of TTP mRNA is shown in lane 4 of the middle panel. These data suggested that the presence of two nonamer TTP-binding sites in the ARE were capable of mediating similar decreases in mRNA stability, whether or not the intervening sequence contained two Cs instead of two As.
When one of the mutant AREs containing a single intact nonamer TTP-binding site (AACCCA) was tested in the same experiment, there was very little decrease in steady-state MLP mRNA levels (Fig. 5, lanes 5  and 6). PhosphorImager analysis of this experiment revealed that the MLP mRNA in the presence of TTP was decreased by only 12%, as compared with the presence of co-transfected vector alone. Similarly, there was little detectable decrease in MLP protein in the presence of TTP (Fig. 5, lower panel, lanes 5 and 6). TTP mRNA levels were similar in all three groups (Fig. 5, middle panel). These data suggest that the presence of a single TTP73-binding site contributed only slightly to the increased turnover of the ARE-containing MLP mRNA, whereas the presence of two TTP73-binding sites had a marked effect to decrease the MLP mRNA levels. These findings correlated well with the results of the cell-free deadenylation assays, in which there was an increased degradation of the intact probe in the presence of TTP, as compared with extracts from cells transfected with vector alone. In those experiments, TTP caused a decrease in probe of 2.74-fold compared with vector alone for the wild-type probe AAAAAA-A 50 , and 2.77-fold for AACCAA-A 50 , whereas with probe AACCCA-A 50 the decrease with TTP was only 1.38-fold (TABLE ONE).
Similar experiments were performed with the other ARE mutants (Fig. 6A). In every case, the expression of TTP mRNA is shown in the lower panel of Fig. 6A. As expected, the co-transfection of TTP with the MLP construct lacking an ARE altogether had no effect on MLP mRNA levels (Fig. 6A, lanes 1 and 2). The presence of either the wild-type ARE 24-base insert (AAAAAA) or the AACCAA mutant resulted in 71.7 and 61.5% decreases in MLP mRNA levels, respectively, as shown in Fig. 6A, lanes 3-6 of the upper panel. These data are similar to those shown in Fig. 5. In contrast, two constructs containing mutant AREs with single TTP73-binding sites were only slightly affected by the co-expression of TTP. In the case of the AAACCA construct, TTP decreased the steadystate level of MLP mRNA by 32.4%, as determined by PhosphorImager analysis (Fig. 6A, upper panel, lanes 7 and 8); in the case of the AACCCA construct, the TTP-induced decrease in MLP mRNA levels was only 24% (Fig. 6A, upper panel, lanes 9 and 10). When no TTP73-binding sites were present in the "ARE," as in the cases of the ACCACA and the CCCCCC mutant AREs, there was essentially no effect of the co-transfected TTP to promote the turnover of the respective MLP transcripts (7 and 6% decreases, respectively; Fig. 6A, upper panel, lanes 11-14). These results also correlated well with the results in the deadenylation assay (TABLE ONE).
In order to determine the average effect of having zero, one, or two TTP73-binding sites in the ARE with susceptibility to TTP-induced degradation, we developed a real time PCR assay for MLP transcripts; this involved developing a procedure for removing plasmid DNA, as well as normalization for the presence of 18 S RNA. This type of assay was conducted on several separate transfection experiments, as shown in Fig. 6B. In Fig. 6B, the 1st bar in the histogram, labeled with a minus sign, indicates that the presence of co-transfected TTP caused an average increase of about 50% in the levels of an MLP mRNA containing no ARE when compared with co-transfected vector alone (n ϭ two experiments). Similarly, when the mutant ARE containing no TTP73-binding sites was present (ACCACA), TTP caused an increase in mean MLP mRNA levels of about 35% compared with vector alone (Fig. 6B, last bar on the histogram). When two binding sites were present, as in the case of the AAAAAA and AACCAA AREs (Fig. 6B, 2nd and 3rd bars on the histogram), TTP co-expression caused ϳ5-fold decreases in mean MLP mRNA levels in each case, when compared with the construct lacking an ARE, and ϳ5-fold decreases when compared with the ACCACA mutant; these changes were highly statistically different, as determined by one-way analysis of variance followed by Tukey-Kramer correction (p Ͻ 0.01).
The mean values for the constructs containing a single TTP73-binding site were decreased by an average of 55% in the case of the AAACCA construct, and 38% in the case of the AACCCA construct, when compared with the mean values determined with the construct containing no ARE (in Fig. 6B compare the 4th and 5th bars in the with the 1st bar). They were decreased on average by 47 and 29% when compared with the values obtained from the construct containing the mutant ARE (Fig. 6B, compare the 4th and 5th bars to the last bar). Although these decreases FIGURE 6. Effect of ARE mutations on MLP-HA mRNA stability in the presence of TTP. Expression constructs CMV⅐MLP-HA (0.5 g/plate), with or without the wild-type ARE, or with mutant AREs in which various flanking-As were substituted by Cs as indicated, were co-transfected into 293 cells with vector alone (lanes 1, 3, 5, 7, 9, 11, or 13) or with CMV⅐hTTP⅐tag (ϩ hTTP, lanes 2, 4, 6, 8, 10, 12, or 14), as described in the legend to Fig. 5. Each gel lane was loaded with 10 g of total cellular RNA. A, electrophoresis and Northern hybridization were performed as described in the legend to Fig. 5. B, statistical analysis of real time PCR using RNA samples described in A from two (1st bar) and five (2nd to 6th bars) co-transfection experiments.
in steady-state mRNA levels were substantial compared with the expression of the MLP mRNA containing either no ARE (Fig. 6B, 1st bar) or the mutant ARE lacking any TTP-binding sites (last bar), none of these decreases was statistically significant.
These experiments demonstrated that the presence of two TTP73 binding nonamers in the MLP transcript conferred significant TTP susceptibility on the MLP transcript, resulting in 4 -5-fold decreases in steady-state mRNA levels. Although the situation with the single binding site constructs was less obvious, they were clearly less sensitive to TTP-mediated mRNA destruction than the transcripts containing two TTP73-binding sites.
MLP expression constructs containing the center U to C (UUAUC-UAUU) mutations in the ARE insert, and a flanking A mutant that contained an octamer UUAUUAUU in the ARE insert, were also tested in similar co-transfection experiments, followed by Northern blotting and PhosphorImager analysis. ARE mutant probes from this group did not bind the expressed TZF peptide but showed weak binding to fulllength TTP (Figs. 2 and 3), and their binding to HuB was not different from that of the wild-type ARE probe. A Northern blot from a typical co-transfection experiment (n ϭ 4) is shown in Fig. 7. The co-transfection of the TTP expression construct with the MLP construct lacking an ARE caused a slight increase in steady-state MLP mRNA levels to 107.8 Ϯ 19.4% (mean Ϯ S.D.) of control cells that were co-transfected with vector alone (Fig. 7, lanes 1 and 2). The resulting steady-state MLP mRNA level was similarly unchanged when this construct was co-transfected with the HuB expression construct (119.1 Ϯ 8.7%, lane 3). The presence of the 24-base wild-type ARE insert, AAAAAA, resulted in an average 77.2 Ϯ 3.0% TTP-induced decrease in MLP mRNA levels (Fig. 7,  lanes 4 and 5). The co-expression of HuB did not decrease the steadystate MLP mRNA levels of this construct, with mRNA levels 122.8 Ϯ 14.7% of control (Fig. 7, lane 6). For the two center U to C mutant ARE inserts (AUCUA)2 and (AUCUA)1, the presence of TTP induced a decrease in steady-state MLP mRNA levels by 30.2 Ϯ 6.2 and 31.7 Ϯ 15.5%, respectively (Fig. 7, lanes 7 and 8, 10, and 11). The presence of HuB did not decrease the MLP mRNA levels expressed from either construct (116.2 Ϯ 27.0 and 141.7 Ϯ 62.2% of control, respectively; Fig.  7, lanes 9 and 12, respectively). Although the mutant ARE insert ACAACA contained no apparent nonamer binding sites for TTP, there was still a modest decrease in steady-state MLP mRNA levels in the presence of TTP (by 30.8 Ϯ 13.5%; Fig. 7, lanes 13 and 14), whereas HuB expression did not affect the steady-state MLP mRNA levels (105.4 Ϯ 13.1%; Fig. 7, lane 15). The expression of TTP and HuB mRNAs in this experiment is shown in the lower panel of Fig. 7.

DISCUSSION
It has become clear from several lines of evidence that the minimum TTP-binding site, and presumably that for its other family members, is well represented by the single-stranded RNA nonamer UUAUUUAUU. This binding site was identified independently by random selection techniques (6) and by the successive truncation of the larger TNF ARE (7); in the latter case, the nonamer was optimal for changing the NMR conformation of a 73-amino acid synthetic peptide comprising the TZF domain of human TTP, whereas even slightly shorter RNA oligonucleotides produced more degenerate conformations. This conclusion was further supported by the data of Brewer et al. (8), who showed that shortening the nonamer to a heptamer by removing the two outer Us decreased the binding affinity for the TTP73 peptide by ϳ6-fold. These data were recently extended by the determination of the three-dimensional structure of the analogous TZF from the TTP relative ZFP36L2 (TIS11D) in complex with the same nonamer (9). Theoretical modeling suggested strongly that the RNA contact amino acids were identical when comparing the TTP TZF domain to that of ZFP36L2, supporting the idea that the structure of the TTP TZF domain complex with the nonamer is virtually identical to that of the ZFP36L2 TZF domain (10). The study of Hudson et al. (9) also supported the results of our previous study in which we showed that mutations of any of the CCCH residues, or any of the aromatic residues located within the Cx5C and Cx3C subdomains of the zinc fingers, resulted in the loss of TTP binding to the ARE (20). These zinc-binding CCCH residues, and the aromatic residues involved in base-stacking interactions with RNA (9), are identical between TTP and TIS11D.
The primary goals of the present study were to determine whether the presence of a single nonameric TTP-binding site in the 3Ј-UTR of an otherwise stable mRNA would confer TTP-dependent instability on that mRNA, and whether the presence of a second binding site would further increase the TTP susceptibility of the mRNA. This information is important for our ongoing attempts to elucidate the mechanism of action of TTP, as well as to inform bioinformatics and experimental approaches to determining bona fide physiological mRNA targets for TTP and its related proteins.
We attempted to address these goals by using several experimental approaches, including analysis of RNA binding by gel shift assays, analysis of TTP-stimulated target mRNA deadenylation in a cell-free assay, and an assessment of the ability of the TTP to promote the deadenylation and instability of mRNA targets in a 293 cell co-transfection assay. For the last experiments, we devised a target molecule based on the FIGURE 7. Effect of ARE mutations on MLP-HA mRNA stability in the presence of TTP. Expression constructs CMV⅐MLP-HA (0.5 g/plate), with or without the wild-type ARE, or with mutant AREs in which various center Us were substituted by Cs, or a with a mutant ARE with two flanking As mutated as indicated, were co-transfected into 293 cells with vector alone (lanes 1, 4, 7, 10, or 13) or with CMV⅐hTTP⅐tag (ϩ hTTP, lanes 2, 5, 8, 11, or 14), or with CMV⅐HuB⅐FLAG (lanes 3, 6, 9, 12, or 15). Total cellular RNA was harvested as described under "Experimental Procedures." Electrophoresis and Northern hybridization were performed as described in the legend to Fig. 5. An identical filter to that shown in the upper panel was blotted and probed at the same time with 32 P-labeled mouse TTP cDNA and the HuB cDNA. The expressed HuB mRNA and TTP mRNA are indicated. stable mRNA encoding the mouse MLP (16,21); one of the advantages of using this RNA target over more conventional TNF-based RNA targets was that Northern analysis of the resulting samples was easier to interpret because of the apparent lack of accumulation of the deadenylated RNA body after exposure to TTP.
The gel shift data demonstrated that the expression of a 77-amino acid TZF domain in 293 cells resulted in identical gel shift data to those obtained with a synthetic 73-amino acid TZF peptide in an otherwise protein-free assay (7). In both cases, the presence of one or two TZF domain nonamer binding sites was reflected in the gel shift pattern, using A to C substituted probes. However, the pattern of full-length TTP binding to these probes was somewhat different from that seen with both TZF peptides, and at this time we are unable to conclude that the full-length protein binds the wild-type and AACCAA probes with 2 mol/mol occupancy. Further studies will be necessary before we can conclude that the two sites in the wild-type probe can be occupied simultaneously by 2 mol/mol of intact TTP, as suggested by the peptide binding studies.
In general, the results of the functional assays showed that the presence of even one intact nonamer binding site in an otherwise stable mRNA conferred a modest degree of TTP "sensitivity" to that mRNA. In the co-transfection studies, this was apparent when the samples were analyzed by either Northern blotting or by real time PCR. On average, the presence of a single TTP-binding site decreased the amount of MLP mRNA accumulation in the presence of TTP to about 12-50% of a control that consisted of a plasmid expressing an RNA with either no ARE or a nonbinding mutant ARE. Similar modest but definite TTP effects were seen in the cell-free deadenylation assays, using RNA targets containing AREs with single TTP-binding sites. These results suggest that the presence of a single nonamer is sufficient to confer modest TTP susceptibility on that mRNA. Whether this modest experimental effect will translate into identification of true, physiological binding targets for TTP and its family members that contain only a single binding site remains to be determined.
The modest response of the mRNA containing a single TTP-binding site to co-expressed TTP was in contrast to the response observed when two binding sites were present. In the co-transfection experiments, TTP decreased the accumulation of MLP mRNAs containing two TTP-binding sites by ϳ5-fold, compared with an MLP mRNA containing no ARE or a nonbinding mutant ARE. Similar data were generated when RNA probes based on the same sequences were used in the TTP-sensitive, cell-free deadenylation assays. These data demonstrate that the presence of two binding sites clearly confers increased TTP susceptibility on an mRNA, compared with the presence of one binding site. This information should be taken into consideration when searching for or evaluating potential RNA targets for TTP or its family members. Whether additional binding sites makes an mRNA still more sensitive to TTPinduced deadenylation and turnover is not clear from our studies, but seems likely. Our data support the earlier findings of Zubiaga et al. (15), who found that two copies of the UUAUUUAUU nonamer destabilized ␤-globin mRNA more effectively than a single copy of the nonamer in NIH-3T3 cells, which express all three TTP family members It is interesting to compare these results to the naturally occurring AREs present in the TNF and GM-CSF mRNAs, still the only demonstrated physiological targets for TTP (as determined by the rather strict criterion of mRNA stabilization in cells derived from TTP knock-out mice). Both of these AREs contain multiple copies of the optimum binding nonamer, although there are species differences in how the nonamers are arranged (10). Although the current study cannot answer the question of why the half-life of GM-CSF mRNA was more than twice as long as the half-life of TNF mRNA in cells derived from the TTP knockout mice, it should be possible, using the assays described here, to determine the stoichiometry of TTP peptide binding to the naturally occurring TNF and GM-CSF AREs from various animal species and, by extension, the corresponding TTP sensitivity of those AREs in the intact mRNAs. Obviously, many other factors influence whether these interactions will occur in vivo, including the presence of TTP (or related protein) in the same cells as the mRNA target, the presence of adequate concentrations of TTP, the occurrence in the same part of the cell at the critical times, and others.