Phosphorylation of p40AUF1 Regulates Binding to A + U-rich mRNA-destabilizing Elements and Protein-induced Changes in Ribonucleoprotein Structure*

Messenger RNA turnover directed by A + U-rich elements (AREs) involves selected ARE-binding proteins. Whereas several signaling systems may modulate ARE-directed mRNA decay and/or post-translationally modify specific trans-acting factors, it is unclear how these mechanisms are linked. In THP-1 monocytic leukemia cells, phorbol ester-induced stabilization of some mRNAs containing AREs was accompanied by dephosphorylation of Ser83 and Ser87 of polysome-associated p40AUF1. Here, we report that phosphorylation of p40AUF1 influences its ARE-binding affinity as well as the RNA conformational dynamics and global structure of the p40AUF1-ARE ribonucleoprotein complex. Most notably, association of unphosphorylated p40AUF1 induces a condensed RNA conformation upon ARE substrates. By contrast, phosphorylation of p40AUF1 at Ser83 and Ser87 inhibits this RNA structural transition. These data indicate that selective AUF1 phosphorylation may regulate ARE-directed mRNA turnover by remodeling local RNA structures, thus potentially altering the presentation of RNA and/or protein determinants involved in subsequent trans-factor recruitment.

The steady-state level of any mRNA population is a collective function of its synthetic and degradation rates. Eukaryotic mRNAs decay across a broad kinetic spectrum, discriminated largely by the presence of specific cis-acting stability determinants within each transcript (reviewed in Refs. 1 and 2). For many labile mammalian transcripts, rapid cytoplasmic mRNA turnover is directed by A ϩ U-rich elements (AREs) 1 contained within their 3Ј-untranslated regions (reviewed in Ref. 3).
AREs constitute a diverse population of mRNA sequences that may interact with a wide variety of cellular protein factors. Binding of some factors, including AUF1 and tristetraprolin, is associated with acceleration of mRNA decay (4 -7), whereas factors like HuR protect ARE-containing transcripts from degradation (8,9). Furthermore, many additional ARE-binding factors exist for which specific roles in mRNA metabolism have not been defined (reviewed in Ref. 10). These myriad options for trans-factor occupancy on ARE targets present opportunities for multifactoral regulation of mRNA decay kinetics through these elements, including discrimination of mRNA targets based on ARE sequence composition, and differential transfactor activation or inhibition by specific signal transduction pathways. The latter possibility is further supported by recent findings that some ARE-binding proteins may be post-translationally modified in response to diverse stimuli. For example, phosphorylation of tristetraprolin by components of the p38 mitogen-activated protein kinase pathway may alter its ARE binding activity (11,12) and/or subcellular distribution (13). For HuR, phosphorylation events involving the AMP-dependent protein kinase are implicated in nuclear retention of the protein (14), whereas lipopolysaccharide treatment induces methylation of HuR (15). Finally, AUF1 may be modified by both ubiquitination (16) and phosphorylation (17,39). However, whereas some potential regulatory events have been linked to these modifications, it remains unclear how they modulate the functions of ARE-binding proteins at the biochemical level, which conceivably may include alterations in the ability of each protein to interact with RNA substrates or other cellular components.
In the monocytic leukemia cell line, THP-1, interleukin-1␤ and tumor necrosis factor ␣ (TNF␣) mRNAs are stabilized following treatment with phorbol esters. Stabilization of these mRNAs was accompanied by changes in the activity of cytoplasmic ARE-binding complexes containing AUF1 and loss of phosphate from Ser 83 and Ser 87 of the predominant polysomeassociated AUF1 isoform, p40 AUF1 (39). Current evidence indicates that AUF1 selectively binds and oligomerizes on ARE substrates (18,19) and has the potential to remodel local RNA structure (20). In addition, cytoplasmic AUF1 is present in a multisubunit complex (17) containing other factors involved in the regulation of mRNA decay and translation, including the translation initiation factor eIF4G, poly(A)-binding protein, the heat shock proteins Hsp70 and Hsc70 (21), and lactate dehydrogenase (22). Together, these findings suggest that AUF1 oligomers may function by recruiting additional trans-acting factors to ARE-containing mRNAs. It follows, therefore, that changes in AUF1 phosphorylation may influence mRNA decay kinetics by a number of mechanisms, including alterations in ARE-binding affinity, oligomerization potential, ribonucleoprotein (RNP) conformation, and interaction with other cytoplasmic proteins. Unlike the examples involving tristetraprolin and HuR (13,14), phosphorylation of p40 AUF1 does not appear to influence its nucleocytoplasmic distribution, since (i) both phosphorylated and nonphosphorylated p40 AUF1 proteins were recovered from THP-1 cell polysomes, and (ii) no significant changes in the levels of nuclear or cytoplasmic AUF1 proteins were detected following phorbol ester treatment of these cells (39).
In this study, we have biochemically examined the influence of p40 AUF1 phosphorylation at Ser 83 and Ser 87 on its interaction with the ARE from TNF␣ mRNA. First, we show that recombinant His 6 -p40 AUF1 may be specifically, quantitatively, and independently phosphorylated in vitro at Ser 83 by glycogen synthase kinase 3␤ (GSK3␤) and at Ser 87 by protein kinase A (PKA). Second, we employed gel mobility shift assays (GMSAs) and measurements of fluorescence anisotropy to show that phosphorylated His 6 -p40 AUF1 retains specific ARE-binding activity but that the affinity of the p40 AUF1 /ARE interaction and the flexibility of the protein-bound RNA substrate varies among the different phosphorylated forms of the protein. Finally, using fluorescence resonance energy transfer (FRET), we demonstrate that the global conformation of the ARE substrate is compacted following binding of unphosphorylated, Ser 83phosphorylated, or Ser 87 -phosphorylated His 6 -p40 AUF1 . By contrast, His 6 -p40 AUF1 phosphorylated at both Ser 83 and Ser 87 does not significantly condense the conformation of associated ARE substrates. Together, these data constitute the first biochemical evidence that post-translational modification of p40 AUF1 regulates its ARE-targeting role and provide a potential mechanism for inhibition of ARE-directed mRNA turnover concomitant with loss of phosphate from polysome-associated p40 AUF1 .
Preparation of Recombinant His 6 -p40 AUF1 -Plasmid pBAD/HisB-p40 AUF1 was constructed by inserting the complete coding sequence of human p40 AUF1 cDNA into the polylinker of pBAD/HisB (Invitrogen) using standard subcloning techniques (23). The identity of the insert and continuity of the open reading frame were verified by automated DNA sequencing. Recombinant p40 AUF1 was prepared as an N-terminal His 6 -fusion protein by arabinose induction of E. coli TOP10 cells transformed with pBAD/HisB-p40 AUF1 . His 6 -p40 AUF1 was then purified by Ni 2ϩ -affinity chromatography and quantified by Coomassie Bluestained SDS-PAGE against a titration of bovine serum albumin as described previously (19,24). Where indicated, the N-terminal fragment of His 6 -p40 AUF1 containing the His 6 tag was excised using the Enterokinase Cleavage Capture Kit (Novagen, Madison, WI).
In Vitro Phosphorylation of His 6 -p40 AUF1 -32 P label transfer assays consisted of His 6 -p40 AUF1 (50 pmol) in 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 1 mM EGTA, 2 mM dithiothreitol, 0.1% Nonidet P-40, and 200 M ATP (50-l final volume). [␥-32 P]ATP was added to yield a final specific activity of 4000 cpm/pmol ATP. After 5 min at 30°C, reactions were initiated by adding 2500 units of either the recombinant catalytic subunit of murine PKA (Calbiochem) or recombinant rabbit GSK3␤ (New England Biolabs, Beverly, MA) and then returned to 30°C. At selected time points, aliquots were removed and spotted onto P81 filters (Whatman, Clifton, NJ), which were then immediately washed three times for 2 min each in freshly made 75 mM H 3 PO 4 and air-dried. Filter-bound 32 P was quantified by liquid scintillation counting. Nonspecific retention of 32 P was determined by performing replicate assays in the absence of kinase. Substrate-linked 32 P was then calculated as the difference in filter-retained 32 P between reactions containing and lacking kinase. Additional reactions lacking the His 6 -p40 AUF1 substrate indicated no significant retention of 32 P via phosphorylation of the kinases themselves (data not shown). Phosphorylation reactions requiring both PKA and GSK3␤ were performed in tandem, with the PKA reaction allowed to proceed for 60 min prior to the addition of GSK3␤ and incubation for a further 60 min. For samples analyzed by mass spectrometry, [␥-32 P]ATP was omitted from reactions, and products were desalted and concentrated using ZipTipC 4 columns (Millipore, Bedford, MA) according to the manufacturer's instructions.
Large scale preparation of phosphorylated His 6 -p40 AUF1 for biochemical analyses was performed similarly, but using 5 nmol of His 6 -p40 AUF1 in a total volume of 300 l. PKA and/or GSK3␤ (50,000 units) were added as necessary, with samples incubated as above. Following phosphorylation reactions, samples were diluted 10-fold in Ni 2ϩ -affinity binding buffer (50 mM sodium phosphate (pH 8.0), 500 mM NaCl, 20 mM imidazole, 5% polyethylene glycol 6000). Mock-or kinase-modified His 6 -p40 AUF1 was then purified by Ni 2ϩ -affinity chromatography and quantified by Coomassie Blue-stained SDS-PAGE as described (19,24), except that the loaded Ni 2ϩ column was given an additional wash with 6 column volumes of Triton washing buffer (50 mM sodium phosphate (pH 8.0), 500 mM NaCl, 20 mM imidazole, 1% Triton X-100) prior to His 6 -p40 AUF1 elution to ensure complete removal of the kinases. Following purification, a sample from each preparation was analyzed by MALDI-TOF to verify quantitative phosphate transfer.
Mass Spectrometry-In-gel tryptic digestion, immobilized metal ion affinity chromatography, alkaline phosphatase reactions, carboxypeptidase Y digests, and detection of proteins and peptide fragments by MALDI-TOF mass spectrometry were all performed exactly as described previously (39). The apparent molecular weight of His 6 -p40 AUF1 was calculated from the predicted amino acid sequence (GenBank TM accession number NM_002138 and pBAD/His vector system literature) (Invitrogen) using the AAStats program of the Biology Workbench version 3.2 (San Diego Supercomputer Center; available on the World Wide Web at www.workbench.sdsc.edu).
RNA-Protein Binding Assays-GMSAs using unmodified or phosphorylated His 6 -p40 AUF1 and 32 P-labeled RNA oligonucleotide substrates were performed essentially as described (18,24), except that magnesium ions were not included in binding buffers, and heparin (5 g/l) and yeast tRNA (0.2 g/l) were included to compete for nonspecific RNA binding activities. Also, flanking regions of RNA substrates were not excised by nucleases prior to gel fractionation.
Fluorescence anisotropy was employed for all quantitative assessments of RNA-protein binding equilibria. Binding reactions containing the fluorescent RNA substrate Fl-TNF␣ ARE and varying concentrations of unmodified or phosphorylated His 6 -p40 AUF1 were assembled in a final volume of 100 l containing 10 mM Tris-HCl (pH 8.0), 50 mM KCl, 2 mM dithiothreitol, 0.5 mM EDTA, 0.1 g/l acetylated bovine serum albumin, and 1 g/l heparin. Following incubation at 25°C for 1 min, anisotropy was measured using a Beacon 2000 variable temperature fluorescence polarization system (Panvera, Madison, WI) equipped with fluorescein excitation (490-nm) and emission (535-nm) filters. Preliminary on-rate experiments verified that anisotropic equilibrium was attained within 10 -20 s at this temperature for all binding equilibria described herein (data not shown). For equilibrium binding experiments, the polarimeter was operated in static mode, with each sample read as blank prior to the addition of fluorescent RNA substrates to correct for intrinsic fluorescence from other reaction components. Each anisotropy data point represents the mean of 10 measurements. Replicate experiments using enterokinase-digested His 6 -p40 AUF1 (see Fig.  1B) yielded results similar to the undigested protein (data not shown), indicating that the His 6 and flanking N-terminal sequences do not significantly influence the ARE binding activity of His 6 -p40 AUF1 .
The total measured anisotropy (A t ) of a mixture of fluorescent species exhibiting similar fluorescence quantum yields may be interpreted based on the intrinsic anisotropy and fractional concentration of each fluorescing species, given by A i and f i , respectively, using Equation 1 (25)(26)(27).
For all equilibria described in this work, total fluorescence intensity did not vary significantly as a result of protein binding (data not shown), thus validating interpretation of anisotropy by Equation 1. A sequential dimer binding mechanism yielding a tetrameric protein-RNA complex is defined by two equilibrium constants, K 1 and K 2 , and presents three potential fluorescent species, R, P 2 R, and P 4 R (see Fig.  4A). Under conditions of limiting RNA (i.e. [protein] free Ϸ [protein] total ), measured anisotropy is thus related to the concentration of protein dimer (P 2 ) by Equation 2 (18).
Anisotropy data sets were also analyzed using a binary binding model described by Equation 3, where a single protein dimer interacting with an RNA substrate yields a single equilibrium constant (K).
Application of binding algorithms to A t versus [P 2 ] data sets was performed by nonlinear least squares regression using PRISM version 2.0 (GraphPad, San Diego, CA). The appropriateness of all mathematical models was monitored by the coefficient of determination (R 2 ) and analysis of residual plot nonrandomness to detect any bias for data subsets (PRISM version 2.0). Where indicated, pairwise comparisons of sum-of-squares deviations between mathematical models were performed using the F test, whereas pairwise comparisons of binding or anisotropy parameters between experiments used the unpaired t test (PRISM version 2.0). In both cases, differences exhibiting p Ͻ 0.05 were considered significant. Protein off-rate experiments to measure the dynamics of RNA-protein complexes were performed as described previously (18,28).
Measurement of RNA Folding by FRET-Interpretation of RNA conformation in the P 2 R and P 4 R complexes by FRET required comparison of the protein dependence of FRET efficiency (E FRET ) to the fractional concentration of each fluorescent species. Since higher concentrations of RNA (2 nM) were required for FRET experiments, the approximation that [P 2 ] free Ϸ [P 2 ] total , used to derive Equation 2, becomes invalid. Accordingly, the fractional concentrations of free RNA, P 2 R, and P 4 R were determined by solution of an equation system relying exclusively on K 1 , K 2 , [RNA] total , and [P 2 ] total . Details of this system are described in detail elsewhere (20). Solution sets for selected values of [P 2 ] total with constant [RNA] total were generated using Mathematica version 4.1 (Wolfram Research, Champaign, IL).
Protein-dependent changes in the scalar distance between the 5Ј and 3Ј termini of the Cy-TNF␣ ARE-Fl RNA substrate were evaluated by measuring E FRET between the fluorescent moieties as a function of protein concentration. Binding reactions were assembled as described for fluorescence anisotropy analyses (above), except that RNA substrates were included at higher concentration (2 nM) and were labeled with either (i) the FRET donor-acceptor pair (Cy-TNF␣ ARE-Fl) or (ii) the FRET donor alone (TNF␣ ARE-Fl). E FRET was calculated from the decrease in fluorescence emission of the FRET donor ( ex ϭ 490 nm, em ϭ 518 nm, 5-nm bandwidth) in the presence of the acceptor using Equation 4 (29,30).
F DA and F D represent the blank-corrected fluorescence of paired binding reactions containing the Cy-TNF␣ ARE-Fl and TNF␣ ARE-Fl RNA substrates, respectively. All fluorescence readings were taken using a Cary Eclipse fluorescence spectrophotomer (Varian Instruments, Walnut Creek, CA), using the Peltier multicell holder and temperature controller accessories. Where indicated, absolute scalar distances between fluorophores were calculated using Equation 5, where r represents the scalar interfluorophore distance and R 0 is the Förster distance, at which the fluorescent donor-acceptor pair yields 50% FRET efficiency. R 0 for the Cy3 and Fl linked to single-stranded DNA has been calculated as 55.7 Å (31).

Recombinant p40 AUF1 Is Quantitatively Phosphorylated at Ser 87 by Protein Kinase A and at Ser 83 by Glycogen Synthase
Kinase 3␤ in Vitro-In unstimulated THP-1 cells, polysomeassociated p40 AUF1 is phosphorylated on Ser 83 and Ser 87 (39). These modifications are lost following phorbol ester treatment of this cell line, concomitant with stabilization of some AREcontaining mRNAs and alterations in cytoplasmic ARE-binding activities containing AUF1. Three observations suggest that the effects of AUF1 phosphorylation on cytoplasmic mRNA turnover are likely to be manifested through p40 AUF1 . First, Ser 83 and Ser 87 are encoded by exon 2 of the AUF1 gene (32) and, due to alternative pre-mRNA splicing, are specific for the p45 and p40 isoforms of AUF1. Second, p45 AUF1 is almost exclusively nuclear in THP-1 cells, whereas p40 AUF1 is detected predominantly in the cytoplasm (39). Finally, p40 AUF1 is the major polysome-associated AUF1 isoform in this cell line. Accordingly, the next objective was to determine how phosphorylation of p40 AUF1 at Ser 83 and Ser 87 might alter its biological function. Given that these residues are located immediately upstream of the p40 AUF1 RNA-binding domain (Fig. 1A), alterations in ARE-binding activity were postulated to be a possible consequence of these phosphorylation events.
In order to test this hypothesis, it was first necessary to generate purified, selectively phosphorylated forms of p40 AUF1 . Using the PhosphoBase version 2.0 data base (33), PKA was predicted to specifically phosphorylate p40 AUF1 at Ser 87 (Fig.  1A). Furthermore, Ser 83 presented a consensus phosphorylation site for GSK3, provided that Ser 87 was previously phosphorylated. Recombinant, N-terminal His 6 -tagged p40 AUF1 was expressed in Escherichia coli TOP10 cells and purified by Ni 2ϩ -affinity chromatography to Ͼ95% purity as described under "Experimental Procedures" (Fig. 1B). In label transfer assays with [␥-32 P]ATP, both GSK3␤ and PKA readily and independently phosphorylated the recombinant His 6 -p40 AUF1 substrate, approaching a stoichiometry of 1 mol of phosphate/ mol of protein in each case (Fig. 1C). For both phosphorylation reactions, SDS-PAGE analysis indicated a single principal 32 Plabeled product (Fig. 1D).
To further confirm His 6 -p40 AUF1 as the phosphorylated substrate, kinase reactions were repeated using unlabeled ATP and then analyzed by MALDI-TOF. A mock phosphorylation reaction (i.e. containing no enzyme) yielded a major product with an M r of 37,803 (Fig. 1E), close to 37,793, the calculated M r of unmodified His 6 -p40 AUF1 . By contrast, products identified in PKA-or GSK3␤-programmed reactions gave M r values indicating increases of 80 and 79 Da, respectively, consistent with the addition of a single phosphate to His 6 -p40 AUF1 by each kinase. Since unphosphorylated His 6 -p40 AUF1 was not detected in either experiment, phosphorylation by each kinase was deemed to be quantitative. Finally, a reaction programmed sequentially with PKA and then GSK3␤ yielded a principal product with M r of 37,963. The 160-Da difference relative to the unmodified protein indicates the addition of two phosphates to His 6 -p40 AUF1 in these reactions.
Having demonstrated that His 6 -p40 AUF1 may be quantitatively phosphorylated by PKA and GSK3␤, it was subsequently necessary to confirm the sites of modification on the substrate protein. To this end, PKA-, GSK3␤-, and PKA ϩ GSK3␤-phosphorylated His 6 -p40 AUF1 were digested with trypsin. Released phosphopeptide fragments were then purified by immobilized metal ion affinity chromatography across a Ga 3ϩ -charged matrix and analyzed by MALDI-TOF ( Fig. 2A). PKA-phosphorylated His 6 -p40 AUF1 yielded a single Ga 3ϩ -binding fragment (m/z 1051), consistent with a single phosphate linked to peptide HSEAATAQR, spanning residues 86 -94 of p40 AUF1 . Phosphorylation with GSK3␤ yielded a distinct fragment (m/z 1539), characteristic of monophosphorylated peptide NEEDEGHSNS-SPR, spanning p40 AUF1 residues 73-85. From PKA ϩ GSK3␤-phosphorylated His 6 -p40 AUF1 , Ga 3ϩ -binding fragments corresponding to both the PKA-and GSK3␤-modified tryptic peptides were identified. For each Ga 3ϩ -binding peptide, the presence of a single phosphorylated residue was further confirmed by alkaline phosphatase digestion, which produced an ϳ80-Da decrease in fragment mass (Fig. 2B). Finally, the specific phosphorylation sites within each modified fragment were determined by limited carboxypeptidase Y digest (Fig. 2C). For the PKA-phosphorylated peptide (m/z 1051), phosphoserine was exclusively retained within a His-Ser P -Glu tripeptide fragment (m/z 444), thus confirming Ser 87 as the sole residue modified by PKA phosphorylation. In the GSK3␤-phosphorylated peptide (m/z 1539), phosphoserine was released within a Ser P -Pro-Arg tripeptide (leaving m/z 1113), permitting assignment of the GSK3␤-modified residue to Ser 83 .
Together, these experiments confirm that recombinant His 6 - Residues that are phosphorylated in polysome-associated p40 AUF1 purified from unstimulated THP-1 cells are indicated, along with consensus phosphorylation target sites for PKA and GSK3. B, purified His 6 -p40 AUF1 was fractionated by SDS-PAGE prior to (undigested) or following (enterokinase (EK)-digested) excision of the N-terminal His 6 tag and additional amino acids by digestion with enterokinase. Proteins were detected by staining with Coomassie Brilliant Blue R-250. C, in vitro phosphorylation of His 6 -p40 AUF1 by PKA and GSK3␤ was measured by label transfer assay as described under "Experimental Procedures" and plotted as the molar ratio of phosphate/protein as a function of incubation time. D, products from His 6 -p40 AUF1 phosphorylation reactions containing the kinases indicated were fractionated by SDS-PAGE. 32 P-labeled protein products were detected by autoradiography. The migration position of His 6 -p40 AUF1 is indicated by the arrowhead (right). E, His 6 -p40 AUF1 phosphorylation reactions programmed without (mock) or with the indicated kinases were analyzed by MALDI-TOF mass spectrometry. M r values for the predominant peptides are indicated. p40 AUF1 may be specifically and quantitatively phosphorylated in vitro by PKA at Ser 87 and by GSK3␤ at Ser 83 . In contrast with the data base prediction and a previous report (34), we find that phosphorylation of His 6 -p40 AUF1 at Ser 83 with GSK3␤ does not require prior phosphate linkage at Ser 87 and have confirmed this finding by label transfer assay, SDS-PAGE, and MALDI-TOF analyses. Finally, since the in vitro phosphorylation sites concur with those identified on polysome-associated His 6 -p40 AUF1 purified from THP-1 cells, selective phosphorylation of His 6 -p40 AUF1 with PKA and/or GSK3␤ will be a versatile system for biochemically dissecting the influence of p40 AUF1 phosphorylation on its interaction with its RNA substrates or other macromolecules.
Recombinant p40 AUF1 Associates with an ARE Substrate by Sequential Dimer Binding-To determine whether phosphorylation of His 6 -p40 AUF1 influenced its RNA binding activity, mock-, PKA (Ser 87 )-, GSK3␤ (Ser 83 )-, and PKA ϩ GSK3␤ (Ser 83 ϩ Ser 87 )-phosphorylated His 6 -p40 AUF1 proteins were first purified by Ni 2ϩ -affinity chromatography. Binding to RNA substrates was then qualitatively analyzed by GMSA. In each case, His 6 -p40 AUF1 bound to a substrate containing the core ARE from TNF␣ mRNA (TNF␣ ARE) (Fig. 3B, lanes 2-5). Substrate specificity was also preserved, since binding was not observed to a fragment of the ␤-globin mRNA coding sequence (R␤) (Fig. 3B, lanes 7-10). In the ARE-binding reactions, an RNA-protein complex of intermediate mobility was detected (Fig. 3B, complex a), characteristic of the sequential dimer binding mechanism employed by His 6 -p37 AUF1 to associate with RNA substrates (18,19). However, significant differences in ARE-binding activity were noted for the slowest mobility complexes. With the mock-phosphorylated protein, most bound ARE substrate migrated as a single complex (Fig. 3B, complex   FIG. 2. Identification of His 6 -p40 AUF1 residues phosphorylated in vitro by PKA and GSK3␤. A, His 6 -p40 AUF1 was phosphorylated in vitro with the indicated kinases and then fragmented by in-gel tryptic digest. Phosphopeptide fragments were selected by Ga 3ϩ -immobilized metal ion affinity chromatography and analyzed by MALDI-TOF. M r values corresponding to the principal Ga 3ϩ -binding peptide fragments are indicated. B, Ga 3ϩ -binding tryptic peptide fragments of each His 6 -p40 AUF1 preparation were partially dephosphorylated with alkaline phosphatase and then analyzed by MALDI-TOF. The ϳ80-Da, phosphatase-dependent shifts in fragment mass characteristic of loss of a single phosphate group from each peptide are indicated by the arrows. C, Ga 3ϩ -binding tryptic peptide fragments of His 6 -p40 AUF1 phosphorylated with PKA plus GSK3␤ were partially digested with carboxypeptidase Y. Products were then analyzed by MALDI-TOF, with the M r values of specific peptide digestion products indicated. Amino acid residues released from each peptide fragment were identified based on the resulting shift in M r and are denoted by the schematics at the top. c). However, ARE substrate bound to the phosphorylated forms of His 6 -p40 AUF1 was largely distributed across a population of two complexes (Fig. 3B, complexes b and c). Whereas the ARE substrate binds His 6 -p40 AUF1 regardless of phosphorylation status, the variations in complex mobility observed by GMSA suggest that the conformation and/or flexibility of the resulting His 6 -p40 AUF1 -ARE complexes may be influenced by phosphorylation of the protein components.
To address these possibilities, complex formation between an ARE substrate and mock-or kinase-modified forms of His 6 -p40 AUF1 was quantitatively assessed under solution equilibrium conditions by fluorescence anisotropy. Protein binding to a 5Ј-Fl-conjugated RNA substrate increases the anisotropy of the fluorophore, due to an increase in the rotational correlation time of the RNA-protein complex relative to the RNA alone and a decrease in the flexibility of single-stranded RNA molecules following protein binding (18,25,35). Accordingly, as the concentration of His 6 -p40 AUF1 increases under conditions of limiting fluorescent RNA substrate, total measured anisotropy (A t ) of the RNA substrate increases (Fig. 4B). The mathematical relationship between A t and protein concentration is dependent on the binding model describing their association. In the case of His 6 -p37 AUF1 binding to the 5Ј-Fl-labeled TNF␣ ARE (Fl-TNF␣ ARE), the protein dependence of anisotropy is well described by a highly dynamic, sequential dimer binding model (Fig. 4A), resolving to a tetrameric AUF1-RNA complex (18,19). The equations relating A t and protein concentration for this model are described under "Experimental Procedures" and permit solution of both equilibrium constants (K 1 and K 2 ) as well as intrinsic anisotropy values for the free, protein dimer-bound, and protein tetramer-bound RNAs (A R , A P 2 R , and A P 4 R , respectively).
The fluorescence anisotropy assays yielded three principal observations indicating that His 6 -p40 AUF1 and His 6 -p37 AUF1 bind AREs by similar mechanisms. First, binding of mockphosphorylated His 6 -p40 AUF1 to the Fl-TNF␣ ARE substrate was well described by sequential dimer binding (Equation 2; Fig. 4B, solid line), further confirmed by the random distribution of residuals about the regression solution (Fig. 4B, bottom). Second, the sequential dimer binding model was significantly preferred (p Ͻ 0.0001 by F test) over an algorithm describing a single protein/RNA interaction (Equation 3; Fig. 4B, dotted line), demonstrating that the simpler binding model is clearly inappropriate in this case. Finally, interactions between His 6 -  4. Characterization of His 6 -p40 AUF1 binding the TNF␣ ARE by fluorescence anisotropy. A, the sequential dimer binding model of AUF1-ARE complex formation. By this model, an AUF1 protein dimer (P 2 ) interacts with an ARE substrate (R) to generate the P 2 R complex, described by the equilibrium association constant K 1 . This complex, in turn, may associate with a subsequent protein dimer to yield the P 4 R complex, described by K 2 . B, binding reactions containing the fluorescent RNA substrate Fl-TNF␣ ARE were assembled across a titration of mock-phosphorylated His 6 -p40 AUF1 as described under "Experimental Procedures." Fluorescence anisotropy was measured for each reaction and plotted as a function of protein concentration. The anisotropy data set was resolved by nonlinear regression using Equation 2 (solid line), with a residual plot prepared by subtracting the regression-derived anisotropy solution (A calc ) from the experimentally observed values (A obs ) at each tested protein concentration (lower panel). An additional regression solution is indicated for a single-site binding model, solved using Equation 3 (dotted line). p40 AUF1 and the Fl-TNF␣ ARE substrate were highly dynamic, with off-rate experiments yielding complex dissociative halftimes of Ͻ15 s (data not shown), similar to those measured with His 6 -p37 AUF1 -ARE complexes (18). Together, the similarities in the ARE-binding activities of His 6 -p40 AUF1 and His 6 -p37 AUF1 support a common RNA-binding mechanism for these proteins. Furthermore, they validate the use of the fluorescence anisotropy system to analyze His 6 -p40 AUF1 /ARE solution binding equilibria, thus permitting quantitative assessments of ARE binding affinity, protein oligomerization potential, and RNA dynamics in His 6 -p40 AUF1 -ARE complexes.
Phosphorylation of Recombinant p40 AUF1 Alters ARE-binding Affinity and RNA Dynamics-ARE binding of all phosphorylated forms of His 6 -p40 AUF1 conformed to the sequential dimer binding model, based on random residual distribution and poor representation by the binary binding algorithm (p Ͻ 0.0001 by F test; data not shown). The intrinsic anisotropy and association equilibrium constants for each protein binding to the Fl-TNF␣ ARE substrate are listed in Table I. Comparing these parameters between the various phosphorylated forms of His 6 -p40 AUF1 revealed significant distinctions in their AREbinding characteristics. For Ser 87 -phosphorylated His 6 -p40 AUF1 , an ϳ2-fold increase in affinity at the second binding step (K 2 ) was observed relative to mock-phosphorylated protein. This was accompanied by a significant decrease in the intrinsic anisotropy of both the dimer-bound (A P2R ) and tetramer-bound (A P4R ) ARE substrates. These decreases in intrinsic anisotropy indicate that the mobility of the TNF␣ ARE 5Ј-end is enhanced when complexed with His 6 -p40 AUF1 phosphorylated on Ser 87 relative to complexes formed with the unphosphorylated protein. Furthermore, the highly dynamic nature of the His 6 -p40 AUF1 /ARE equilibrium (described above) raises the possibility that changes in A P2R and K 2 are linked, since enhanced RNA flexibility in the Ser 87 -phosphorylated P 2 R complex may kinetically improve opportunities for the second dimer-binding event.
Based on the equilibrium binding studies, phosphorylation at Ser 83 also influenced the ARE-binding activity of His 6 -p40 AUF1 , although in a manner quite different from Ser 87 phosphorylation. First, phosphorylation of His 6 -p40 AUF1 at Ser 83 did not detectably alter RNA flexibility in His 6 -p40 AUF1 -ARE complexes relative to the mock-phosphorylated protein (based on A P2R and A P4R ). Second, binding of the initial protein dimer to the ARE was inhibited by ϳ40% (K 1 ) when phosphorylated at Ser 83 . Third, comparison of ARE-binding parameters between the doubly phosphorylated protein and each singly phosphorylated species indicates that phosphorylation at Ser 83 dominates the effects of Ser 87 phosphorylation. This assertion is supported by (i) restriction of RNA mobility in the P 2 R (A P2R ) and P 4 R (A P4R ) complexes by phosphorylation at Ser 83 plus Ser 87 , relative to Ser 87 alone, (ii) diminution of P 2 R affinity for a second His 6 -p40 AUF1 dimer (K 2 ) when phosphorylated at Ser 83 plus Ser 87 relative to Ser 87 , and (iii) maintenance of Ser 83 -mediated inhibition of the initial protein dimer-binding step (K 1 ), regardless of Ser 87 phosphorylation. Taken together, these data indicate that phosphorylation of His 6 -p40 AUF1 induces several changes in its interactions with RNA substrates, including influences on the binding affinity of the initial protein dimer (Ser 83 ), RNA flexibility in His 6 -p40 AUF1 -ARE RNP complexes (Ser 87 ), and enhancement of subsequent protein dimer recruitment (Ser 87 ). In addition, the phosphospecific effects on p40 AUF1 activity are not additive, since phosphorylation at Ser 83 abrogates the influence of modification at Ser 87 .
p40 AUF1 -dependent Changes in ARE Conformation Are Inhibited When both Ser 83 and Ser 87 Are Phosphorylated-In addition to forming protein oligomers, the interaction of His 6 -p37 AUF1 with an RNA substrate induces structural condensation of the bound RNA, detectable as a protein concentrationdependent decrease in the distance between the 5Ј and 3Ј termini of the RNA substrate (20). This distance is calculated from the FRET efficiency (E FRET ) between 3Ј-Fl donor and 5Ј-Cy3 acceptor moieties of appropriately labeled RNA substrates (Fig. 5A). To determine whether His 6 -p40 AUF1 conferred similar conformational restraints following RNA binding, fluorescence emission of a 5Ј-Cy3-plus 3Ј-Fl-labeled TNF␣ ARE substrate was measured across a range of His 6 -p40 AUF1 concentrations (Fig. 5B). As protein concentration increased, fluorescence emission from the Fl moiety ( max ϭ 518 nm) significantly decreased. Loss of emission from Fl was not due to nonquantum quenching, since no significant change in fluorescence was observed for a TNF␣ ARE substrate labeled only with 3Ј-Fl (data not shown). The protein-dependent loss of Fl emission from the Cy3-plus Fl-labeled ARE substrate was thus interpreted as FRET, with the decrease in Fl emission indicating shortening of the distance between the RNA termini upon His 6 -p40 AUF1 binding.
To correlate changes in RNA structure with the formation of protein-ARE complexes, it was first necessary to construct fractional concentration plots for each His 6 -p40 AUF1 -ARE complex under all phosphorylation conditions (Fig. 5C), using the solutions for K 1 and K 2 in each case (Table I). Then E FRET was measured for the Cy3-plus Fl-labeled ARE substrate across titrations of each phosphorylated form of His 6 -p40 AUF1 (Fig.  5D). For mock-phosphorylated His 6 -p40 AUF1 , E FRET increased concomitant with the decrease in free RNA concentration, resolving to ϳ0.60 -0.65 at protein concentrations of Ͼ50 nM dimer. E FRET did not appreciably increase further, although P 4 R becomes the dominant complex Ͼ100 nM dimer. As such, it is likely that structural condensation of the ARE substrate results from the initial dimer-binding event and is not significantly altered by association of the second dimer. This correlation between initial dimer binding and RNA remodeling is virtually identical to that previously described for His 6 -p37 AUF1 (20). His 6 -p40 AUF1 phosphorylated at Ser 87 or Ser 83 induces similar constraints on the conformation of the bound ARE substrate, but with the principal distinction that in-  creases in E FRET are shifted to higher protein concentrations.
In the case of the Ser 87 -phosphorylated protein, this may be coupled to the increased flexibility of the ARE in each of the RNP complexes (described above). By contrast, Ser 83 -phosphorylated His 6 -p40 AUF1 requires higher protein concentrations to achieve complex formation, due to the 40% decrease in ARE binding affinity relative to the mock-phosphorylated protein ( Table I). The most notable distinction, however, was observed with His 6 -p40 AUF1 -ARE complexes containing the Ser 83 -plus Ser 87 -phosphorylated protein, for which little change in E FRET was detected as a function of protein concentration. These data indicate that, unlike mock-phosphorylated or singly phosphorylated His 6 -p40 AUF1 , the doubly phosphorylated protein does not induce significant structural condensation of associated RNA substrates. Rather, RNA substrates bound by the doubly phosphorylated His 6 -p40 AUF1 are retained in a relatively elongated conformation. In this manner, selective phosphorylation of p40 AUF1 at Ser 83 and Ser 87 may dramatically alter not only the thermodynamics of formation but also the overall structure of RNPs resulting from its association with its RNA substrates.  Table I, with the total RNA concentration set to 2 nM. D, E FRET of the Cy-TNF␣ ARE-Fl RNA substrate was monitored in solution as a function of His 6 -p40 AUF1 concentration for each phosphovariant of the protein as described under "Experimental Procedures." Replicate experiments yielded similar results.

DISCUSSION
Cellular signaling pathways involving phosphorylation cascades are common themes in the regulation of gene expression. For many years, phosphorylation of specific transcription factors has been closely associated with their ability to induce or inhibit transcription of selected target genes (reviewed in Ref. 36). By comparison, current understanding of such mechanisms regulating mRNA turnover is far less developed. For some ARE-containing transcripts, selective stabilization or destabilization has been observed in response to the activity of specific signaling systems. Also, post-translational modifications of some ARE-binding proteins have been detected (described in the Introduction). The next step, then, is to discern how modification of specific ARE-binding proteins influences their biochemical activities, thus permitting these regulatory mechanisms to be considered and scrutinized within the broader context of mRNA catabolism. To this end, we have queried the biochemical significance of p40 AUF1 phosphorylation on Ser 83 and Ser 87 , prompted by observations that polysome-associated p40 AUF1 is phosphorylated at these residues in unstimulated THP-1 cells but loses these modifications following phorbol ester treatment, concomitant with stabilization of some ARE-containing mRNAs (39).
Previously, GMSA-based analyses indicated that His 6 -p40 AUF1 bound an RNA substrate containing the ARE from c-fos mRNA at ϳ30-fold lower affinity than did His 6 -p37 AUF1 (32). In this study, however, unmodified His 6 -p40 AUF1 bound the TNF␣ ARE with affinity comparable with His 6 -p37 AUF1 (19), measured by protein-dependent changes in fluorescence anisotropy. Some of this difference may result from the use of GMSAs in earlier studies, since this method may be prone to underestimation of equilibrium binding for highly dynamic interactions (18). Alternatively, His 6 -p40 AUF1 and His 6 -p37 AUF1 may exert distinct binding preferences among different ARE substrates. Also, improvements in bacterial expression of recombinant AUF1 proteins (pBAD/His vector system; Invitrogen) and streamlined purification procedures (19) may be contributing to a greater proportion of active His 6 -p40 AUF1 protein than was previously possible.
Biochemical analyses of interactions between the TNF␣ ARE substrate and unphosphorylated versus Ser 83 -plus Ser 87 -phosphorylated His 6 -p40 AUF1 revealed several mechanistic features. In both cases, His 6 -p40 AUF1 was competent to interact with an ARE substrate in vitro, based on GMSAs and fluorescence-based assays. Given the similarities in complex migration by GMSA, and the confident resolution of anisotropy data by the sequential dimer binding model in both cases, we concluded that the ability of His 6 -p40 AUF1 to form oligomeric His 6 -p40 AUF1 -ARE complexes was not abrogated by changes in its phosphorylation status. That both Ser 83 -plus Ser 87 -phosphorylated and unphosphorylated p40 AUF1 proteins are competent for RNA binding is further supported by their recovery from polysomal fractions of untreated and phorbol ester-stimulated THP-1 cells, respectively (39). However, two significant differences were observed in the ARE binding activities of the Ser 83plus Ser 87 -phosphorylated versus unphosphorylated His 6 -p40 AUF1 (Fig. 6). First, the unphosphorylated protein exhibited a small but significant decrease in affinity for the ARE substrate relative to Ser 83 -plus Ser 87 -phosphorylated His 6 -p40 AUF1 , although this was only evident for interactions between the initial protein dimer and the ARE substrate (K 1 ). More dramatic, however, was that the ARE substrate is retained in an extended conformation when complexed with the Ser 83 plus Ser 87 -phosphorylated protein but is folded into a more compact structure by unmodified His 6 -p40 AUF1 . Extrapolating to the phosphorylation status of p40 AUF1 in THP-1 cells, these data suggest that the ability of p40 AUF1 to promote rapid mRNA decay may be coupled to its ability to maintain associated RNA substrates in elongated conformations, based on (i) stabilization of some ARE-containing mRNAs in THP-1 cells concomitant with loss of phosphate from Ser 83 and Ser 87 of polysome-associated p40 AUF1 following phorbol ester stimulation (39) and (ii) retention of elongated RNA substrate conformation in vitro solely when both Ser 83 and Ser 87 are phosphorylated (this work). Whereas phosphorylation of His 6 -p40 AUF1 separately at Ser 83 or Ser 87 also yielded alterations in RNA binding affinity (lowered K 1 for Ser 83 -phosphorylated protein; elevated K 2 for Ser 87 ) and RNA dynamics in the protein-bound state (increased RNA flexibility when Ser 87 phosphorylated), in neither case was adoption of the condensed RNA conformation significantly prevented in the RNP complex (Fig. 5). At present, physiological conditions yielding significant proportions of p40 AUF1 phosphorylated exclusively at Ser 83 or Ser 87 have not been described. However, enhancement of K 2 by phosphorylation at Ser 87 and its subsequent inhibition by Ser 83 phosphorylation reflects the influences of similar modifications on the FIG. 6. Phosphorylation-induced changes in His 6 -p40 AUF1 -ARE RNP conformation and thermodynamics. A schematic illustrating differences in the binding of Ser 83 -plus Ser 87 -phosphorylated versus unphosphorylated His 6 -p40 AUF1 proteins to the TNF␣ ARE substrate. Estimates of scalar distance between the termini of RNA substrates were calculated using Equation 5, based on E FRET ranges of 0.37-0.41 for the free RNA (R), 0.40 -0.44 for RNA substrate bound by Ser 83 -plus Ser 87 -phosphorylated His 6 -p40 AUF1 , and 0.63-0.68 for the unphosphorylated His 6 -p40 AUF1 -bound RNA. E FRET values corresponding to the protein-folded ARE substrate were estimated from the average E FRET for reactions containing Ն250 nM His 6 -p40 AUF1 dimer, since the tetramer-bound ARE substrate is the dominant RNP species in these cases, and free RNA concentrations are negligible.
DNA-binding activity of the cyclic AMP response element binding protein. Phosphorylation of this transcription factor by PKA at Ser 119 induces a 2-3-fold improvement in binding affinity for its cognate DNA substrate in vitro, but this enhancement is prevented by subsequent phosphorylation at a proximal Ser residue by GSK3 (37).
The next question is to identify the mechanisms whereby phosphorylation of p40 AUF1 alters its RNA binding characteristics. Whereas additional studies will be necessary to define the contributions of different contact points to the overall stability and conformation of the AUF1-ARE complex, details accrued thus far suggest some putative foci of interest. One possibility is that changes in protein conformation are induced by phosphorylation at these residues. Another, and perhaps more likely scenario is that the introduction of localized negative charges in a region contiguous with the upstream RNA recognition motif of p40 AUF1 (Fig. 1A) may directly influence interactions between this region and RNA substrates. By this mechanism, phosphorylation may induce localized repulsion from the RNA phosphodiester backbone. This could account for the enhanced ARE mobility observed in complexes with His 6 -p40 AUF1 phosphorylated at Ser 87 . In this case, however, compensation must be made for the enthalpic penalty arising from the unfavorable ionic interactions, since no loss of binding affinity was observed. Conceivably, this may involve favorable changes in entropy resulting from enhanced RNA flexibility and/or introduction of constructive ionic interactions at different sites. Localized electrostatic repulsion may also account for the decrease in binding affinity observed following phosphorylation at Ser 83 (K 1 ). Finally, the presence of phosphate groups conjugated to both Ser 83 and Ser 87 may interfere with protein-RNA contacts essential for adoption of the condensed RNA structure, provided that protein determinants in this region contribute to local remodeling of RNA substrates.
To place the influence of p40 AUF1 phosphorylation within a broader perspective, we may speculate on means by which differences in AUF1-ARE complex conformation or dynamics contribute to alterations in mRNA decay rates. Previous data indicate that AUF1 initiates ARE-directed mRNA turnover through the targeted assembly of a multisubunit, trans-acting complex, involving AUF1 oligomerization and maximization of complex surface area (reviewed in Ref. 38). It follows, therefore, that phosphospecific alterations in the architecture of the AUF1-ARE RNP complex may expose or obscure specific RNA and/or protein determinants involved in subsequent factor recruitment. Furthermore, the rapid dynamics of AUF1/ARE interactions would ensure that changes in p40 AUF1 phosphorylation status were quickly reflected in the cytoplasmic population of AUF1-ARE RNPs. Taken together, the influence of p40 AUF1 phosphorylation on the conformation of AUF1-ARE RNPs, coupled with the multiplicity of ancillary factors that may subsequently interact with these complexes, provides a plethora of downstream binding events that may be sensitive to differential phosphorylation of p40 AUF1 . Finally, the observation that both phosphorylated and nonphosphorylated forms of p40 AUF1 interact with polysomes in cells (39) raises the possibility that these modifications may serve as a switching mechanism for p40 AUF1 , converting it from an "mRNA-destabilizing" to an "mRNA-stabilizing" factor.