Regulation of A + U-rich Element-directed mRNA Turnover Involving Reversible Phosphorylation of AUF1*

Proteins binding A + U-rich elements (AREs) contribute to the rapid cytoplasmic turnover of mRNAs containing these sequences. However, this process is a regulated event and may be accelerated or inhibited by myriad signal transduction systems. For example, monocyte adherence at sites of inflammation or tissue injury is associated with inhibition of ARE-directed mRNA decay, which contributes to rapid increases in cytokine and inflammatory mediator production. Here, we show that acute exposure of THP-1 monocytic leukemia cells to the phorbol ester 12-O-tetradecanoylphorbol-13-acetate mimics several features of monocyte adherence, including rapid induction and stabilization of ARE-containing mRNAs encoding interleukin-1β and tumor necrosis factor α. Additionally, TPA treatment alters the activity of cytoplasmic complexes that bind AREs, including complexes containing the ARE-specific, mRNA-destabilizing factor, AUF1. Analyses of AUF1 from control and TPA-treated cells indicated that post-translational modifications of the major cytoplasmic isoform, p40AUF1, are altered concomitant with changes in RNA binding activity and stabilization of ARE-containing mRNAs. In particular, p40AUF1 recovered from polysomes was phosphorylated on Ser83 and Ser87 in untreated cells but lost these modifications following TPA treatment. We propose that selected signal transduction pathways may regulate ARE-directed mRNA turnover by reversible phosphorylation of polysome-associated p40AUF1.

encoding regulatory proteins like cytokines, inflammatory mediators, and oncoproteins are constitutively unstable. This ensures that the steady-state levels of these mRNAs, and hence their potential for translation, remain low but also that new steady-state levels are approached quickly following changes in the rate of mRNA synthesis (reviewed in Ref. 1). In mammals, a common feature of many unstable mRNAs is the presence of an A ϩ U-rich element (ARE) 1 within the 3Ј-untranslated region (3Ј-UTR). These elements range from 40 to 150 nucleotides in length and exhibit significant variability in sequence composition, but they usually include one or more AUUUA motifs within a U-rich context (2). In general, mRNA turnover mediated by AREs consists of rapid 3Ј 3 5Ј shortening of the poly(A) tail, followed by decay of the mRNA body (3,4).
The regulation of mRNA decay kinetics by AREs involves their association with any of a number of cellular ARE-binding factors (reviewed in Ref. 5). One such factor, AUF1 (also referred to as heterogeneous nuclear ribonucleoprotein D), is expressed as a family of four protein isoforms resulting from alternative splicing of a common pre-mRNA (6). The larger isoforms, designated by their apparent molecular weights as p42 AUF1 and p45 AUF1 , are largely nuclear (7), probably due to the presence of a binding determinant for components of the nuclear scaffold (8). By contrast, p37 AUF1 and p40 AUF1 lack this sequence determinant and, as such, may be found in both nuclear and cytoplasmic compartments. AUF1 binding to an ARE is linked to acceleration of mRNA decay, based on extensive studies correlating mRNA turnover rates with AUF1 abundance (9 -13) or ARE-binding activity (14,15). Association of p37 AUF1 with an ARE induces the formation of protein oligomers (16) and local conformational changes in the RNA substrate (17), which in turn may recruit additional factors to generate a multisubunit, trans-acting complex on the mRNA (7,18). Ultimately, the mRNA is degraded by catabolic activities, which may include specific nucleases (19 -21) or the proteasome (18). Besides AUF1, other RNA-binding factors have also been implicated in the regulation of mRNA decay rates through AREs. For example, association of tristetraprolin (TTP) with some ARE-containing transcripts enhances their decay (22,23). By contrast, mammalian factors related to the Drosophila Elav (embryonic lethal abnormal vision) protein, including the ubiquitously expressed HuR and the neuronspecific Hel-N1, are thought to inhibit ARE-directed mRNA turnover (24 -26). Thus, associated trans-acting proteins may modulate the decay kinetics of ARE-containing mRNAs positively or negatively. Additionally, a growing number of AREbinding proteins have been described to which no specific functions have been ascribed (27)(28)(29)(30)(31).
Modulation of ARE-directed mRNA turnover has been observed in response to a plethora of stimuli. For example, the rapid decay of interleukin-3 mRNA in mast cells is inhibited by Ca 2ϩ influx (32). Similarly, mRNAs containing AREs are stabilized during heat shock (18). Specific intracellular signaling pathways have also been identified, which contribute to the regulation of ARE-directed mRNA turnover. In particular, components of the p38 mitogen-activated protein kinase (p38 MAPK ) and c-Jun N-terminal kinase pathways are required for stabilization of ARE-containing mRNAs associated with the inflammatory response (33)(34)(35)(36)(37) and tumor cell metastasis (38). Both tyrosine kinase and p38 MAPK activities are required for the stabilization of interleukin-1␤ (IL-1␤) and GRO mRNAs induced by monocyte adherence (15). By contrast, ARE-dependent stabilization of cyclooxygenase-2 mRNA by G␣ q -coupled receptor signaling in smooth muscle is mediated by the p42/p44 MAP kinases and is independent of p38 MAPK activity (39).
Currently, the molecular mechanisms linking activation or inhibition of these signaling pathways to modulation of mRNA decay rates are largely unknown. One possibility is that these systems function by altering the abundance or RNA-binding properties of ARE-binding factors responsible for initiating the decay process. To test this hypothesis, we have examined the stabilization of ARE-containing mRNAs encoding the cytokine IL-1␤ and the inflammatory mediator tumor necrosis factor ␣ (TNF␣), using a cultured cell model that mimics key features of monocyte adherence. In particular, we show that acute exposure of THP-1 monocytic leukemia cells to the phorbol ester 12-O-tetradecanoylphorbol-13-acetate (TPA) rapidly induced cell adhesion, accompanied by dramatic increases in levels of IL-1␤ and TNF␣ mRNAs. Accumulation of these transcripts included significant mRNA stabilization, similar to that observed following adhesion of primary monocytes (15). Using this model system, we show that the distribution of cytoplasmic ARE binding activities containing AUF1 is altered following TPA treatment, coincident with changes in post-translational modifications of cytoplasmic p40 AUF1 . Finally, we have identified two sites of phosphorylation on p40 AUF1 purified from THP-1 polysomes, which are dephosphorylated following TPA treatment, concomitant with inhibition of ARE-directed mRNA turnover. From these data, we propose that reversible phosphorylation of AUF1 may constitute a critical mechanism for regulating mRNA turnover rates through AREs.
Cell Culture-THP-1 cells were maintained in RPMI 1640 supple-mented with 10% fetal calf serum and 1 mM L-glutamine in the absence of antibiotics. Defined fetal bovine serum was lot-selected for minimal endotoxin content by HyClone (Logan, UT) and then incubated at 56°C for 30 min to inactivate residual endotoxins prior to use. mRNA Quantitation by RNase Protection Assay (RPA)-A plasmid template for synthesis of a riboprobe complimentary to a portion of the IL-1␤ 3Ј-UTR was generously provided by Dr. Mary Vermeulen. A cDNA fragment encoding a portion of the TNF␣ 3Ј-UTR was amplified by PCR from plasmid pE4 (ATCC, Manassas, VA), and a human ␤-actin 3Ј-UTR fragment was amplified by reverse transcription-PCR from THP-1 cell total RNA. The resulting cDNA fragments were then subcloned into pGEM7Zf(ϩ) to generate plasmid templates for preparation of antisense riboprobes. 32 P-Labeled riboprobes used for RPAs were generated by in vitro run-off transcription as described previously (40). Riboprobes complementary to IL-1␤ and TNF␣ mRNAs were synthesized to specific activities of 1-2 ϫ 10 4 cpm/fmol, whereas ␤-actin antisense riboprobes were synthesized to 200 cpm/fmol. Cellular RNA samples were purified first as the total RNA fraction, extracted using TRIzol reagent according to the manufacturer's instructions, and then enriched for mRNA using the PolyATtract mRNA purification system. Specific cellular mRNAs were quantified using RPAs (8 g of poly(A) ϩ RNA/lane) with RNases P 1 and T 1 as described previously (41). Protected RNA fragments were fractionated by denaturing gel electrophoresis and were visualized and quantified using a PhosphorImager (Amersham Biosciences).
mRNA Decay Assays-Decay rates of IL-1␤ and TNF␣ mRNAs were measured by an actD time course assay. Briefly, transcription was inhibited in control or TPA-treated THP-1 cells by the addition of actD (5 g/ml) to the culture medium, and poly(A) ϩ RNA samples were harvested at selected time points thereafter. The relative abundances of IL-1␤ and TNF␣ mRNAs were determined at each point by RPA and were normalized using ␤-actin mRNA levels. First-order decay constants (k) were solved by nonlinear regression of the percentage of IL-1␤ or TNF␣ mRNA remaining versus time of actD treatment using PRISM version 2.0 (GraphPad, San Diego, CA). Errors about regression solutions (S.E.) were calculated by the software using n Ϫ 2 degrees of freedom, with replicate experiments yielding similar results. Comparisons of mRNA decay constants for control versus TPA-treated cells were performed using the unpaired t test, with differences exhibiting p Ͻ 0.05 considered significant.
Fractionation of THP-1 Cells and Gel Mobility Shift Assays-Nuclear and cytoplasmic fractions of control or TPA-treated (10 nM, 1 h) THP-1 cells were prepared by resuspension (control samples) or scraping (TPAtreated samples) of PBS-washed cells in lysis buffer (10 mM Tris-HCl (pH 7.5), 150 mM KCl, 2.5 mM EDTA, and 1% IGEPAL-CA630) containing mixtures of protease inhibitors (1 g/ml each of leupeptin and pepstatin A, 0.1 mM phenylmethylsulfonyl fluoride) and kinase/phosphatase inhibitors (50 mM sodium fluoride, 5 mM disodium pyrophosphate, 1 mM sodium orthovanadate). Lysis was performed with 7-10 strokes of a loosely fitting Dounce homogenizer and verified by phasecontrast microscopy. Nuclei were pelleted by centrifugation at 1000 ϫ g for 10 min and resuspended directly in SDS-PAGE loading buffer for Western analyses. Protein concentrations were determined for cytoplasmic extracts (42).
Cytoplasmic ARE-binding activities were identified in THP-1 cytoplasmic extracts using gel mobility shift assays as described previously (43). Oligoribonucleotide RNA substrates were 5Ј-end-labeled using [␥-32 P]ATP and T4 polynucleotide kinase to specific activities of 3-5 ϫ 10 3 cpm/fmol as described (16). Identical reactions were assembled for antibody supershift assays except that preimmune serum or antiserum was added to a maximum of 10% total reaction volume prior to incubation.
Analyses of Polysome-associated p40 AUF1 and Peptide Fragments by Mass Spectrometry-A ribosomal salt wash was prepared by first isolating polysomes from control or TPA-treated THP-1 cells by hypotonic lysis and centrifugation through 30% sucrose and then releasing polysome-associated proteins using 0.5 M potassium acetate as described previously (41). From this material, AUF1 was purified by tandem heparin and poly(U) chromatography as described (43).
Tryptic digests were performed in SDS-PAGE gel slices using the Montage In-Gel Digest 96 kit according to the manufacturer's instructions. Briefly, the SDS-PAGE band corresponding to p40 AUF1 was excised, destained, and dehydrated and then digested with trypsin overnight at 30°C. Liberated peptide fragments were lyophilized in a SpeedVac. Tryptic phosphopeptide fragments were enriched using an immobilized metal ion affinity chromatography (IMAC)-based strategy (44). Briefly, ZipTip MC columns were charged with 60 mM GaCl 3 , washed with 10% ACN containing 1% acetic acid, and equilibrated with 10% ACN containing 0.1% acetic acid. Lyophilized peptide fragments were resuspended in 0.1% acetic acid and bound to the columns through 10 application cycles. Phosphopeptides were eluted with freshly prepared 0.3 M NH 4 OH. Where indicated, phosphopeptide fragments selected by Ga 3ϩ -IMAC were dephosphorylated using on-target dephosphorylation reactions (45). Here, phosphopeptides spotted on the MALDI target plate (described below) were dissolved in 0.5 l of 50 mM NH 4 HCO 3 (pH 8.9) containing calf intestinal alkaline phosphatase (0.01 units). Samples were then incubated for 2 h at 37°C in a high humidity chamber. The dephosphorylation reaction was stopped by the addition of 0.5 l of ACN. Specific sites of phosphorylation were determined by limited carboxypeptidase Y digest (46) of Ga 3ϩ -IMAC-selected phosphopeptide fragments. Briefly, phosphopeptides were dried and resuspended in 50 mM sodium citrate (pH 6.0). Carboxypeptidase Y was added (enzyme/peptide ratio of 2:1 by mass), and the mixture was incubated at 37°C for 8 h. The reaction was stopped by the addition of trifluoroacetic acid to 0.5%.
MALDI-TOF mass analyses of all samples were performed on a Voyager-DE PRO work station equipped with a PerkinElmer Biosystems DE-PRO mass spectrometer (PerSeptive Biosystems, Framingham, MA). Phosphopeptides isolated by Ga 3ϩ -IMAC and products from carboxypeptidase Y digests were purified and concentrated with Zip-TipC 18 columns according to the manufacturer's instructions and then mixed (1:4, v/v) with a saturated ␣-cyano-4-hydroxycinnamic acid solution in ethanol/water/formic acid (45:45:10). Polysomal p40 AUF1 purified from THP-1 cells was concentrated using ZipTipC 4 columns and then mixed (1:4 v/v) with saturated sinapinic acid in 0.1% trifluoroacetic acid, 50% ACN. Samples from each mixture (2 l) were spotted onto MALDI target plates and air-dried. The samples were analyzed by the mass spectrometer in the reflector positive or linear negative ion delayed extraction mode. Reflector positive mode was preferred for detection of phosphopeptide fragments due to its improved mass resolution (47), although this mode also contributed to increased background signals. Mass calibration was performed using angiotensin II as an external standard. The apparent molecular mass of unmodified p40 AUF1 was calculated from the predicted amino acid sequence (GenBank TM accession number NM_002138) 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).

IL-1␤ and TNF␣ mRNAs Are Rapidly Induced and Stabilized
in THP-1 Cells following Acute TPA Treatment-Circulating monocytes adhere at sites of infection or tissue injury. This event rapidly triggers the expression of a battery of cytokines and inflammatory mediators, involving both transcriptional and post-transcriptional mechanisms (15,48). By contrast, THP-1 monocytic leukemia cells grow constitutively in suspension but may become adherent following treatment with phorbol esters. Stimulation of THP-1 cells with phorbol esters like TPA is a popular model of monocytic differentiation to macrophage-like cells, based on the manifestation of adherence, loss of proliferation, phagocytosis, and enhanced production of proinflammatory cytokines in response to lipopolysaccharide (49 -52).
Whereas phorbol ester treatment induces THP-1 cells to adopt many macrophage-like characteristics over a period of days, adherence per se may be observed within a mere 15-min exposure to TPA (10 nM, data not shown). To determine whether this acute TPA-induced adherence of THP-1 cells was accompanied by increased expression of cytokine and inflam-matory mediator mRNAs, similar to that observed in adherent monocytes, the levels of IL-1␤ and TNF␣ mRNAs were measured across a time course of TPA treatment (Fig. 1). Increases in IL-1␤ mRNA levels were observed within 30 -60 min and reached a maximum induction of nearly 50-fold within 8 h (Fig.  1, A and C). TNF␣ mRNA levels were also rapidly increased, attaining 30-fold induction within 4 h (Fig. 1, B and C). In both cases, these inductions were transient, with preinduction mRNA levels reached (TNF␣) or approached (IL-1␤) after 24 h of TPA treatment.
During monocyte adherence, mRNA stabilization contributes to the induction of mRNAs encoding cytokines and other inflammatory mediators (15). To establish whether similar post-transcriptional mechanisms contributed to induction of these mRNAs in the acute TPA-treated THP-1 model, the decay kinetics of IL-1␤ and TNF␣ mRNAs were monitored prior to and following TPA-induced cell adherence by actD time course assay (Fig. 2). In all cases, mRNA decay kinetics were well approximated by first-order decay functions (Fig. 2C). Both IL-1␤ and TNF␣ mRNAs are relatively unstable in unstimulated THP-1 cells, decaying with half-lives of 29 and 8 min, respectively. However, after 1 h of TPA treatment, both mRNAs were stabilized 6 -7-fold (Fig. 2C). In addition, stabilization of each mRNA was sustained as mRNA levels accumulated (Table I). Taken together, the similarities in mRNA accumulation kinetics, the transient nature of mRNA induction, and the rapid and prolonged stabilization of each mRNA following TPA treatment of THP-1 cells indicate that expression of IL-1␤ and TNF␣ mRNAs are coordinately regulated in this system, involving both transcriptional and post-transcriptional mechanisms. As such, these data suggested that acute TPA treatment of THP-1 cells mimics some features of the adhesiondependent induction of cytokine and inflammatory mediator mRNA levels in primary monocytes.
The Activity of Cytoplasmic ARE-binding Complexes Including AUF1 Is Modified Concomitant with Inhibition of AREdirected mRNA Turnover-Several observations indicated that stabilization of IL-1␤ and TNF␣ mRNAs in TPA-treated THP-1 cells is probably due to regulation of ARE-directed mRNA turnover. First, these mRNAs are coordinately stabilized following TPA treatment of THP-1 cells (above). Second, both IL-1␤ and TNF␣ mRNAs contain AREs in their 3Ј-UTRs (Fig.  3A), which constitute critical determinants of mRNA turnover for many labile transcripts (2,3). Third, additional ARE-containing transcripts, encoding the GRO family of chemokines, are also stabilized coordinately with IL-1␤ mRNA during adherence of primary monocytes (15). Finally, activation of the p38 MAPK pathway stabilizes several labile transcripts by inhibiting the mRNA-destabilizing effects of their AREs (33,34). This is further demonstrated by transgenic mouse models where the ARE from TNF␣ mRNA has been deleted. Here, the synthetic rate of TNF␣ is dramatically increased but is no longer responsive to p38 MAPK activity (53).
Since the ability of an ARE to initiate mRNA decay in cis is linked to its association with trans-acting factors (3,5), gel mobility shift assays were employed to determine whether changes in cytoplasmic ARE-binding activities were exhibited concomitant with TPA-induced mRNA stabilization. In cytoplasmic extracts from untreated THP-1 cells, two distinct activities were detected that bound the ARE from TNF␣ mRNA (Fig. 3B, lane 2). These activities were specific for the ARE, since they did not associate with an unrelated RNA substrate spanning a portion of the ␤-globin coding sequence (Fig. 3B,  lane 10). Using cytoplasmic extracts from TPA-treated cells, two ARE-specific binding activities were also detected (Fig. 3B,  lane 6). Whereas RNA-protein complexes formed using extracts from control or TPA-treated cells displayed similar migration by gel mobility shift assay, their distribution was significantly different (Fig. 3, B (cf. lane 6 versus lane 2) and C). In particular, the faster mobility complex (complex I) accounts for only 20% of the total ARE substrate bound using extracts from TPA-treated cells, with the remaining 80% contained within the slower mobility complex (complex II). By contrast, the ARE substrate is roughly evenly distributed between these complexes using extracts from control cells.
In order to identify specific ARE-binding proteins within these complexes, antibody supershift assays were performed. Using antiserum specific for AUF1, supershifted complexes were detected in ARE-binding reactions containing cytoplasmic extracts from both control and TPA-treated cells. The addition of anti-AUF1 antiserum to reactions containing control extracts effectively shifted ϳ90% of complex I and 70% of complex II, indicating the presence of AUF1 in both ARE-binding activities (Fig. 3, B (lane 4) and C). In comparable experiments with extracts from TPA-treated cells, complex I was also effectively shifted by anti-AUF1 antiserum (Ͼ90%); however, the abundance of complex II was only diminished by ϳ25% (Fig. 3B,  lane 8). Furthermore, increasing the concentration of anti-AUF1 antiserum in the reaction did not significantly increase the fraction of complex II shifted (data not shown). These data indicate that a portion of complex II derived from the cytoplasm of TPA-treated THP-1 cells is not immunoreactive with anti-AUF1 antibodies. This could be the result of conformational differences in complex II between extracts from control or TPAtreated cells, such that AUF1 is largely inaccessible to antibod-ies in the ribonucleoprotein complex formed using cytoplasm from TPA-treated cells. Alternatively, complex II may represent a heterogeneous population of ARE binding activities, of which only a small fraction contain AUF1. To evaluate this possibility, additional supershift assays were performed using antibodies specific for the ARE-binding proteins HuR, TTP, and Hsp70; however, no supershifted or antibody-neutralized complexes were detected (data not shown). Whereas this may indicate that these proteins do not associate with the TNF␣ ARE in this system, it is also possible that the epitopes targeted by these antibodies are occluded as a result of interactions between target proteins and RNA or other cytoplasmic components. In any case, differences in the distribution and anti-AUF1 reactivity of cytoplasmic ARE-binding complexes from control versus TPA-treated cells suggest that the abundance and/or RNA-binding activity of cytoplasmic AUF1 is altered in THP-1 cells following TPA treatment.
Post-translational Modifications of p40 AUF1 Are Altered by TPA Treatment-Conceivably, the abundance of cytoplasmic AUF1 could be modified by regulation of either its expression or subcellular distribution. To test these possibilities, Western blots were used to determine whether the level or nucleocytoplasmic distribution of any AUF1 isoform was significantly affected by TPA treatment (Fig. 4A). From THP-1 cells, p37 AUF1 and p40 AUF1 were recovered almost exclusively in cytoplasmic extracts, whereas p42 AUF1 and p45 AUF1 were nuclear. Acute exposure to TPA had no detectable effect on either the abundance or subcellular distribution of any isoform of AUF1. Since the overall amount of cytoplasmic AUF1 available for RNA binding did not change following TPA treatment, it follows that some feature(s) of the ARE-binding activity of AUF1 was probably affected, possibly involving post-translational modifications of cytoplasmic AUF1 or modulation of some essential co-factor. The latter hypothesis is less likely, since recombinant AUF1 proteins are capable of binding U-rich RNA substrates with high affinity in the absence of biological co-factors (17,54). However, since TPA is a potent activator of signaling pathways involving protein kinase C (55), and AUF1 exists as a phosphoprotein in K562 erythroleukemia cells (7), it was plausible that TPA might influence AUF1 post-translationally. To test this possibility, two-dimensional Western blots were used to determine (i) whether cytoplasmic AUF1 was post-translationally modified in THP-1 cells and (ii) whether such modifications of AUF1 were changed following TPA treatment. In cytoplasmic extracts from untreated THP-1 cells, sev-eral forms of p40 AUF1 were detected, which differed in their isoelectric points (Fig. 4B, left), demonstrating that post-translationally modified variants of p40 AUF1 were present. By contrast, fewer variants of p40 AUF1 were detected in cytoplasmic extracts from TPA-treated cells. In particular, the most acidic form of cytoplasmic p40 AUF1 in control cells was undetectable in TPA-treated cells (Fig. 4B, right), indicating that post-translational modifications of cytoplasmic p40 AUF1 are altered by TPA treatment of THP-1 cells. In both control and TPA-treated cells, the concentration of cytoplasmic p37 AUF1 was too low to permit accurate assessment of post-translational modifications by this method.
The two-dimensional Western blots indicated that posttranslational modifications of cytoplasmic p40 AUF1 were altered following TPA treatment. However, since ARE-binding complexes immunoreactive with anti-AUF1 antibodies were detected in the cytoplasm of both control and TPA-treated THP-1 cells (Fig. 3B), such modifications are unlikely to function by completely abrogating the ARE-binding activity of p40 AUF1 . Given that the focus of this study was to identify mechanisms contributing to the regulation of ARE-directed mRNA turnover in the THP-1 model system, the next question was thus to determine how RNA-bound (i.e. polysome-associated) AUF1 was modified before and after TPA treatment of this cell line. The loss of an acidic p40 AUF1 variant without a significant change in the apparent molecular weight of the protein is consistent with a net loss of phosphate from p40 AUF1 following TPA treatment. Accordingly, post-translational modifications of polysome-associated p40 AUF1 were identified using  MALDI-TOF mass spectrometry. MALDI-TOF analyses of polysome-associated p40 AUF1 purified from control versus TPAtreated THP-1 cells gave relative molecular weights (M r ) of 32,976 and 32,814, respectively (Fig. 5). The difference of 162 Da between these values is consistent with the loss of two phosphate groups from polysome-associated p40 AUF1 following TPA treatment. Furthermore, the determined M r of p40 AUF1 purified from TPA-treated cells (32,814) was not significantly different from 32818, the predicted M r of a completely unmodified p40 AUF1 protein. The unphosphorylated protein (M r 32,814) was not detected in polysome-associated AUF1 purified from control cells (Fig. 5A), nor was the modified protein (M r 32976) identified in analyses of p40 AUF1 purified from polysomes of TPA-treated cells (Fig. 5B). Whereas detection of additional cytoplasmic p40 AUF1 variants by two-dimensional Western blot (Fig. 4B) indicates that other p40 AUF1 modifications may occur in THP-1 cells, their absence in the polysomeassociated AUF1 population suggests that alternatively modified p40 AUF1 proteins do not significantly associate with mRNA. This finding, taken together with the two-dimensional Western analyses and the MALDI-TOF data, indicates that polysome-associated p40 AUF1 (i) is post-translationally modified in untreated THP-1 cells, most likely involving two phosphorylation events, (ii) loses these modifications following TPA treatment of THP-1 cells, and (iii) apparently lacks other posttranslational modifications in this cell line.
Polysome-associated p40 AUF1 Is Phosphorylated on Ser 83 and Ser 87 in Untreated THP-1 Cells-Since the preceding experiments strongly suggested the presence of two phosphate groups on polysome-associated p40 AUF1 in untreated THP-1 cells and their loss following TPA treatment, the next objectives were to verify the identity of the p40 AUF1 post-translational modifications as phosphates and to map their location(s) within the protein. To this end, polysome-associated p40 AUF1 purified from untreated THP-1 cells was digested with trypsin. Phosphorylated peptide fragments were then selected by IMAC across a Ga 3ϩ -charged resin. MALDI-TOF analyses of Ga 3ϩselected tryptic fragments indicated the presence of two phosphopeptides (Fig. 6A). The larger fragment (m/z 1539) was consistent with a single phosphate linked to peptide NEE-DEGHSNSSPR, spanning amino acid residues 73-85 of p40 AUF1 . The smaller fragment (m/z 1051) was consistent with monophosphorylated HSEAATAQR, spanning residues 86 -94. Single phosphate groups were conclusively indicated on each peptide by treatment with alkaline phosphatase, which caused a decrease of ϳ80 Da in the mass of each fragment (Fig. 6B). Since multiple potential sites of phosphorylation are present within each of these peptide fragments, specific modification sites were identified by limited carboxypeptidase Y digest (Fig.  6C). In the larger phosphopeptide fragment (m/z 1539), phosphoserine was exclusively released as a Ser P -Pro dipeptide (m/z 1381-1110), indicating that Ser 83 was the site of phosphorylation in this fragment. Following digestion of the smaller fragment (m/z 1051) with carboxypeptidase Y, phosphoserine was retained in a His-Ser P -Glu-Ala tetrapeptide (m/z 516), permitting assignment of a single phosphate to Ser 87 in this fragment. Based on these observations, we conclude that polysome-associated p40 AUF1 is phosphorylated on Ser 83 and Ser 87 in un-   6 -8). For a separate gel, comparable reactions were assembled using a 32 P-labeled R␤ RNA substrate (lanes 10 and 11). To identify ARE-binding components that contained AUF1, additional binding reactions were treated with anti-AUF1 antiserum (␣-AUF1; lanes 4 and 8) and compared with binding reactions containing preimmune serum (p. i.; lanes 3 and 7). The location of the ␣-AUF1specific supershift is indicated. Since the supershifted complex(es) did treated THP-1 cells and that these phosphate groups are lost following TPA treatment. Furthermore, observations that TPA treatment of THP-1 cells concomitantly effects changes in phosphorylation of polysome-associated p40 AUF1 (Figs. 4B and 5), alterations in the distribution of cytoplasmic ARE-binding activities containing AUF1 (Fig. 3B), and stabilization of mRNAs containing AREs (Fig. 2), suggests the presence of a mechanistic link between AUF1 phosphorylation and the regulation of ARE-directed mRNA turnover.

DISCUSSION
Mechanisms regulating ARE-directed mRNA turnover rates are of substantial interest, since their manipulation presents an attractive target for restricting the expression of gene products contributing to the development of cancer and inflammatory diseases (3). Whereas recent studies have identified some signaling pathways contributing to these regulatory events, few details are available regarding either their molecular targets or the biochemical basis of their function in modulating mRNA turnover rates.
In this study, we have used acute TPA treatment of THP-1 cells as a model system to investigate mechanisms regulating ARE-directed mRNA decay. In particular, we queried how TPA-activated signaling systems might be transduced directly through the ARE-binding proteins responsible for catabolic targeting of these mRNAs. Treatment of THP-1 cells with TPA induced rapid stabilization of ARE-containing transcripts, concomitant with alterations in the activity of cytoplasmic AREbinding complexes containing AUF1, and loss of phosphate from Ser 83 and Ser 87 of the major cytoplasmic AUF1 isoform, p40 AUF1 . These serine residues are both encoded within exon 2 of the human AUF1 gene (Fig. 6D), a domain that is conserved among the AUF1 genes sequenced to date. In rodents, Ser 83 is retained, and the conservative Ser 3 Thr substitution at position 87 maintains the potential for phosphorylation at this site. Alternative pre-mRNA splicing of exon 2 sequences distinguishes p37 AUF1 from p40 AUF1 (6), the two cytoplasmic AUF1 isoforms expressed in THP-1 cells (Fig. 4A). Accordingly, p37 AUF1 cannot be phosphorylated at comparable positions, indicating that signal transduction pathways regulating cytoplasmic AUF1 function through these phosphorylation events are specific for p40 AUF1 .
Although polysome-associated p40 AUF1 was modified exclu- FIG. 4. Post-translational modifications of cytoplasmic p40 AUF1 are altered following TPA treatment. A, one-dimensional Western blots were constructed using equal cellular equivalents (2 ϫ 10 7 ) of cytoplasmic (cyto) or nuclear (nuc) extracts prepared from untreated (Control) or TPA-treated (10 nM, 1 h; TPA) THP-1 cells and then probed using anti-AUF1 antiserum. AUF1 isoforms were identified by their apparent molecular weights (right). B, two-dimensional Western blots were constructed using cytoplasmic extracts (50 g of protein) prepared from untreated (left) or TPA-adhered (right) THP-1 cells and then probed for AUF1. The most acidic form of p40 AUF1 detected in untreated cells is denoted by the arrow. No signal was detected at the corresponding position in extracts from TPA-treated cells (ellipse). sively by phosphate in THP-1 cells, other data maintain the possibility for additional post-translational modifications. First, the two-dimensional Western analyses clearly displayed at least three p40 AUF1 species in the cytoplasm of untreated THP-1 cells (Fig. 4B, left), yet only a single polysome-associated p40 AUF1 variant was detected by mass spectrometry (Fig. 5A). However, since the p40 AUF1 analyzed by mass spectrometry was purified based on (i) its polysomal localization in THP-1 cells and (ii) its ability to bind a poly(U) column, p40 AUF1 variants with poor or highly dynamic RNA-binding properties would not be retained for analysis. Second, in HeLa cells, a portion of the AUF1 population is ubiquitinated (18), yet no evidence for such modification was detected in THP-1 cytoplasm by two-dimensional Western blot or mass spectrometry. In the THP-1 system, however, it is conceivable that ubiquitinated AUF1 might not be detected if it accumulated in nuclei or the perinuclear space and was thus excluded from the soluble or polysomal cytoplasmic fractions or if ubiquitinated forms of FIG. 6. Ser 83 and Ser 87 are phosphorylated in polysome-associated p40 AUF1 from untreated THP-1 cells. A, tryptic fragments of polysome-associated p40 AUF1 purified from untreated THP-1 cells were selected for binding to Ga 3ϩ using IMAC and then analyzed by MALDI-TOF. M r values corresponding to the principal Ga 3ϩ -binding peptide fragments are indicated. Enhanced background signal was due to data collection in reflector positive mode, selected for its improved resolution of fragment mass. B, Ga 3ϩ -binding tryptic peptide fragments of p40 AUF1 were partially dephosphorylated using alkaline phosphatase as described under "Experimental Procedures" and then analyzed by MALDI-TOF.
The arrows indicate ϳ80-Da, phosphatase-dependent shifts in fragment mass, consistent with the loss of a single phosphate group from each peptide. C, limited carboxypeptidase Y digests of Ga 3ϩ -binding tryptic peptide fragments of p40 AUF1 were analyzed by MALDI-TOF. M r values correspond to peptide digestion products. Amino acid residues excised from each peptide were identified based on the resulting shift in M r and are denoted by the schematics at the top. D, localization of p40 AUF1 phosphorylation sites relative to its domain structure. Amino acid residues encoded by exon 2 (shaded) distinguish p40 AUF1 from p37 AUF1 and are located immediately N-terminal of the first RNA-binding domain (RRM1). Both sites of phosphorylation are contained within this alternatively expressed region of the protein. Alignment of homologous peptide sequences from mouse (GenBank TM accession number I49070) and rat (GenBank TM accession number BAB03467) indicate that the potential for phosphorylation at these sites is conserved (boxed), with Ser 83 absolutely conserved, and a Thr residue substituted for Ser 87 in rodent p40 AUF1 . AUF1 do not accumulate to significant levels in THP-1 cells due to rapid degradation by the proteasome. In any case, the potential exists for additional post-translational modifications of AUF1, which will be addressed in later studies.
Whereas cytoplasmic AUF1 proteins are involved in the regulation of ARE-directed mRNA turnover, nuclear roles for AUF1 have also been described, including the regulation of transcription (56 -58). Sequences encoded by exon 2 are important for transactivation through p40 AUF1 , although this activity may also be regulated through selected signal transduction systems (56,59). Recently, the potential for phosphorylation of p40 AUF1 at Ser 83 and Ser 87 was independently described based on in vitro kinase assays, which indicated that Ser 87 was phosphorylated by protein kinase A, whereas Ser 83 could be phosphorylated by glycogen synthase kinase-3␤, although this was dependent on prior phosphorylation of Ser 87 (60). Expression of Ser 3 Ala and Ser 3 Asp mutants of p40 AUF1 suggested that phosphorylation of Ser 87 enhances the transactivation potential of the protein but that subsequent phosphorylation of Ser 83 inhibits this function. Whereas our study confirms that Ser 83 and Ser 87 are targets for phosphorylation on p40 AUF1 , we furthermore provide the first evidence that these residues are phosphorylated in cells and link the loss of these phosphate groups with alterations in the activity of cytoplasmic AREbinding complexes and concomitant inhibition of ARE-directed mRNA turnover.
ARE-directed mRNA decay was significantly slowed by TPA treatment of THP-1 cells concomitant with the loss of phosphate from p40 AUF1 . However, it is important to note that ARE-containing transcripts were still degraded relatively quickly. For example, TNF␣ mRNA was not stabilized beyond a half-life of about 1 h (Table I), although no phosphorylated p40 AUF1 was detected in the polysomal population at this point (Fig. 5B). This observation argues that multiple, partially redundant ARE-directed mRNA decay mechanisms may operate concurrently in this system, some of which are sensitive to TPA, whereas others remain active. For example, the other cytoplasmic AUF1 isoform, p37 AUF1 , shows high affinity for ARE substrates (54) yet does not contain residues homologous to Ser 83 and Ser 87 of p40 AUF1 . The presence of distinct transacting factors is also supported by the population of cytoplasmic ARE-binding activities in TPA-treated THP-1 cells, some of which were not recognized by anti-AUF1 antiserum (Fig. 3B). If these ARE-binding activities represent additional protein factors (like TTP, HuR, etc.), this supports a model for global regulation of ARE-directed mRNA decay mediated by the interplay of diverse factors operating in a combinatorial or mutually exclusive manner, thus greatly enhancing the sophistication of cellular control over the decay of individual mRNAs. For the TNF␣ mRNA, recent evidence also indicates the presence of an additional destabilizing element downstream of the ARE (61), which probably presents distinct protein binding preferences. Together, such integrated control mechanisms could be finely tuned by the activity of many different signaling systems yet define selectivity for specific mRNAs based on relative trans-factor binding affinity. Further support for this model is given by emerging examples of post-translational mechanisms influencing the activity or availability of individual ARE-binding proteins for interaction with cytoplasmic mRNA. For example, phosphorylation of TTP by p38 MAPK correlates with rapid ARE-directed mRNA decay, possibly by enhancing the RNA-binding activity of TTP (62). However, an independent study indicates that TTP phosphorylation may be mediated by the p38 MAPK -activated protein kinase 2 rather than p38 MAPK itself (63). In any case, interactions between 14-3-3 proteins and TTP are linked to the phosphorylation status of TTP (64), which may influence its nucleo-cytoplasmic distribution. Recent work also suggests that p38 MAPK -activated protein kinase 2 phosphorylation of heterogeneous nuclear ribonucleoprotein A0 may influence its interaction with cytokine mRNAs (65). The activity of the AMP-dependent protein kinase contributes to nuclear retention of HuR, thus restricting the ability of HuR to stabilize ARE-containing transcripts in the cytoplasm (66). In addition, HuR may be methylated in macrophages in response to lipopolysaccharide treatment (67). At present, the biochemical consequences of these post-translational modifications remain poorly defined. However, these examples illustrate how the activity or availability of selected ARE-binding factors may be modulated in response to diverse stimuli, thus altering the cytoplasmic population of ARE-binding proteins available to interact with, and hence direct the catabolic fate of, their mRNA substrates.
Accordingly, a major question concerns the biochemical significance of phosphorylation at Ser 83 and Ser 87 of p40 AUF1 . Since these phosphorylation events alter neither the abundance nor the nucleo-cytoplasmic distribution of this protein (Fig. 3A), we are interested in how these post-translational modifications influence the interaction of p40 AUF1 with its RNA substrates or other cellular factors. The localization of these phosphorylation sites immediately adjacent to the RNA-binding domains of p40 AUF1 (Fig. 6D) prompts the hypothesis that they regulate some feature(s) of its ARE binding activity, possibly by altering local charge distribution or by inducing conformational changes in the protein. In an accompanying article (68), we describe how phosphorylation at Ser 83 and Ser 87 modulates both (i) the RNA-binding thermodynamics of p40 AUF1 and (ii) the influence of this binding event on the conformation of ribonucleoproteins containing this protein.