Purification and Characterization of β-Adrenergic Receptor mRNA-binding Proteins*

β-Adrenergic receptors (β-ARs), like other G-protein-coupled receptors, can undergo post-transciptional regulation at the level of mRNA stability. In particular, the human β1- and β2-ARs and the hamster β2-AR mRNA undergo β-agonist-mediated destabilization. By UV cross-linking, we have previously described an ∼M r 36,000 mRNA-binding protein, βARB, that binds to A/C+U-rich nucleotide regions within 3′-untranslated regions. Further, we have demonstrated previously that βARB is immunologically distinct from AUF1/heterogeneous nuclear ribonucleoprotein (hnRNP) D, another mRNA-binding protein associated with destabilization of A+U-rich mRNAs (Pende, A., Tremmel, K. D., DeMaria, C. T., Blaxall, B. C., Minobe, W., Sherman, J. A., Bisognano, J., Bristow, M. R., Brewer, G., and Port, J. D. (1996) J. Biol. Chem. 271, 8493–8501). In this report, we describe the peptide composition of βARB. Mass spectrometric analysis of an ∼M r36,000 band isolated from ribosomal salt wash proteins revealed the presence of two mRNA-binding proteins, hnRNP A1, and the elav-like protein, HuR, both of which are known to bind to A+U-rich nucleotide regions. By immunoprecipitation, HuR appears to be the biologically dominant RNA binding component of βARB. Although hnRNP A1 and HuR can both be immunoprecipitated from ribosomal salt wash proteins, the composition of βARB (HuR alone versus HuR and hnRNP A1) appears to be dependent on the mRNA probe used. The exact role of HuR and hnRNP A1 in the regulation of β-AR mRNA stability remains to be determined.

For many protooncogenes, lymphokines, and cytokines, as well as for mRNAs encoding G-protein-coupled receptors that undergo regulated mRNA stability, a common feature is the existence of AϩU-rich elements (AREs) within their 3Ј-untranslated regions (3Ј-UTRs). Although the exact sequence of an ARE can be variable (20), it is generally defined by the presence of AUUUA pentamers or UUAUUUAUU nonamers (21)(22)(23)(24)(25)(26)(27), often within a generally U-rich context. An example within the family of ␤-ARs is a 20-nucleotide AϩU-rich region within the hamster ␤ 2 -AR 3Ј-UTR (16). In contrast, non-canonical nucleotide sequences such as that found in c-Myc can also be highly destabilizing to an mRNA (23). Similarly, a noncanonical (48) 160-nucleotide CϩU-rich sequence with the human ␤ 1 -AR 3Ј-UTR has been described (29). 2 Each element, within the context of its full-length 3Ј-UTR, has been demonstrated to be necessary for ␤-agonist-mediated mRNA destabilization as well as being capable of acting in cis to destabilize a normally stable ␤-globin mRNA in chimeric (␤-globin coding region/heterologous 3Ј-UTR) constructs (3,18).
In previous studies using hamster smooth muscle DDT1-MF2 cells, we and others have described an ARE binding activity called ␤ARB, for ␤-AR mRNA-binding protein (16, 17, 29 -31). Several observations originally led to the hypothesis that ␤ARB is an mRNA destabilizing protein. First, in DDT1-MF2 cells, the relative abundance of ␤ARB is reciprocally related to the abundance and stability of the hamster ␤ 2 -AR mRNA (6,29,30,32). Further, treatment with ␤-AR agonist or direct activators of protein kinase A concurrently decreases the half-life of the hamster ␤ 2 -AR mRNA and increases the abundance of ␤ARB protein (30). Conversely, treatment with either insulin or dexamethasone, agents that up-regulate ␤ 2 -AR mRNA, decreases the abundance of ␤ARB (30). More recently, ␤ARB has been shown to bind to a number of other highly regulated AϩU-rich mRNAs (17), 2 including other G-proteincoupled receptor mRNAs that undergo agonist-mediated destabilization, including the human ␤ 1 -AR (18,29) 2 and the thrombin receptor mRNA (31).
In examining the hamster ␤ 2 -AR in greater detail, Tholanikunnel et al. (16) have demonstrated that a 20-nucleotide AϩU-rich region within the context of its 3Ј-UTR is necessary for ␤ARB binding. Point mutations (U3 G) within the AUUUUA hexameric sequence effectively disrupt ␤ARB binding as well as completely abrogate ␤-agonist-mediated destabilization of the hamster ␤ 2 -AR mRNA (16). Thus, there is strong correlative evidence between the binding of ␤ARB and the ability of that mRNA to undergo ␤-agonist-mediated destabilization. These findings also support an alternative hypothesis, that disruption of the AϩU-rich region of the hamster ␤ 2 -AR 20-mer likely disrupts the binding of other ARE-binding proteins such as AUF1/hnRNP D (discussed below). In fact, based upon our own work (18) 2 and that of others (18,33,34), 2 disruption of the hamster ␤ 2 -AR ARE would almost certainly significantly diminish the affinity of AUF1 for the mRNA and presumably, its ability to destabilize the target mRNA (35).
Recently, a number of trans-acting factors presumed to either stabilize or destabilize mRNA have been described. In many cases, these proteins display high affinity for AϩU-rich or CϩU-rich elements within 3Ј-UTRs. Based upon an association between the presence of an AϩU-rich mRNA binding activity and the destabilization of target mRNAs e.g. c-Myc, Brewer et al. (36) purified and cloned a protein called "AUF1" for AϩU-rich binding factor 1, known now to be identical to hnRNP D (37)(38)(39)(40)(41). AUF1 is similar in class to hnRNPs A1 and A2/B1 in that it contains two RNA-binding domains (RBDs) and an area rich in glycine and arginine residues (an RGG box) (39,40). In general, hnRNP proteins, such as hnRNP A1, have been associated with the pre-mRNA, small nuclear ribonucleoprotein complex where they facilitate the formation of basepaired double strands (42). Thus, hnRNPs have been implicated in modulating mRNA splicing. However, more recent evidence that hnRNPs (as well as other classes of mRNA-binding proteins) can shuttle from the nucleus to the cytoplasm (43)(44)(45) has led to the speculation that these proteins may play an additional role in affecting the turnover of AϩU-rich mRNAs (46,47).
Further investigation of AUF1 indicates that there is a good correlation between the relative affinity of the p37 splice form of AUF1 for specific AϩU-rich mRNAs and their intrinsic stability (34). These data fit well with the recent findings of Loflin et al. (35), indicating that the p37 and p42 isoforms of AUF1 display a greater ability to destabilize an mRNA than the lower affinity isoforms (p40 and p45).
Previously, we have demonstrated that both endogenous and purified recombinant AUF1 binds with high affinity to the A/CϩU-rich regions of the human ␤ 1 -AR, human ␤ 2 -AR, and hamster ␤ 2 -AR mRNAs (29). Further, we have demonstrated that, in cells treated with ␤-agonist or in tissues obtained from failing human heart, a condition associated with increased circulating catecholamines, AUF1 mRNA and protein are upregulated (29). By these criteria, the mRNA binding activity, ␤ARB, and p37AUF1 are highly similar. However, by immunological methods, we have demonstrated that ␤ARB and AUF1 are distinct proteins (29). We have speculated, therefore, that ␤ARB may either be another hnRNP know to bind AϩUrich regions (46,47), or a member of the elav family of mRNA-binding proteins, known to bind to A/CϩU-rich regions and to affect mRNA stability (49 -59).
In this manuscript, we describe for the first time the peptide composition of ␤ARB, as determined by both direct chemical sequencing (mass spectrometry) and by immunological methods. Additionally, by nondenaturing gel-shift analysis, we have determined the affinity of ␤ARB protein components for different ␤-AR mRNA substrates. Finally, we demonstrate that the composition of ␤ARB is dependent on the RNA target used for its identification.
Preparation of Cytosolic (S100) Fraction-Cells were washed twice with phosphate-buffered saline (PBS: 8 mM Na 2 HPO 4 , 1.4 mM KH 2 PO 4 , 2.6 mM KCl, 136 mM NaCl) and removed from the cell culture plate with PBS, 1 mM EDTA. The cells were then pelleted gently, resuspended in PBS, transferred to a sterile ultracentrifuge tube, and again pelleted gently. Excess buffer was removed, and 5 l each of the protease inhibitors (10 mg/ml) aprotinin and leupeptin were added to the cells (ϳ final (12.5 g/ml) for each protease inhibitor). The cells were then spun at 100,000 ϫ g using a Beckman Ti50 rotor and thick-walled, sterile, polycarbonate tubes, for 90 min at 4°C. The resulting supernatants were then transferred to Eppendorf tubes and placed on ice. Protein concentration was determined using the method of Lowry (60).
Preparation of Polysomes and Ribosomal Salt Wash (RSW) Preparations-Cytosolic S130 and polysome fractions, and RSW preparations were produced as described previously (29,30,61). Briefly, cells were washed with cold, serum-free medium, resuspended in 3.5 ml of Polysome Buffer A (10 mM Tris-HCl (pH 7.6), 1 mM potassium acetate, 1.5 mM magnesium acetate, 2 mM dithiothreitol). Cells were homogenized and centrifuged to pellet the nuclei. The supernatant was removed and layered gently over a 1.5-ml sucrose (30% w/v) cushion and centrifuged at 130,000 ϫ g in a Beckman SW60Ti rotor. The S130 supernatant was removed from above the S130/sucrose interface, aliquoted, and stored at Ϫ80°C. To recover the polysome fraction, the sucrose cushion was aspirated off and the tubes allowed to drain while kept at 4°C. To prepare an RSW, polysomes were resuspended in Polysome Buffer A and 4 M KCl was added dropwise to a final concentration of 0.3 M. After stirring, KCl-treated polysomes were re-pelleted through Polysome Buffer B (Polysome Buffer A with 30% w/v sucrose) as described above. The supernatant (RSW), generally containing ϳ80% of polysome-associated 3Ј 3 5Ј exonuclease activity, was aliquoted and stored at Ϫ80°C.
In Vitro Transcription-The cDNAs for the AϩU-rich hamster ␤ 2adernergic receptor 20-mer (16) and human ␤ 1 -160-mer (29) (Fig. 1) were each contained in pBSIIKS plasmid vectors. Plasmids were linearized immediately 3Ј to the target DNA insert, and in vitro transcription was performed using T3 DNA-directed RNA polymerase to produce 5Ј-capped, uniformly labeled mRNAs based on the technique of Melton et al. (62). mRNAs were transcribed in the presence of RNasin with nucleotide concentrations and buffer conditions as detailed by the manufacturer (Promega). Co-transcriptional capping was achieved by using the cap analogue m 7 (5Ј)Gppp(5Ј)G (New England Biolabs) at a concentration that was 10-fold in excess to the concentration of GTP. Radiolabeled probes were prepared using [␣-32 P]UTP (800 Ci/mmol, NEN Life Science Products). Biotinylation of the probes was achieved using biotin-14-CTP (Life Technologies, Inc.). Following transcription, RNase-free DNase (Promega) was added to the mixture to remove the template DNA and the probe was precipitated with 2.5 volumes of 0.5 M ammonium acetate in ethanol. Radiolabeled probes were resuspended in RNase-free water. Biotinylated probes were resuspended in Column Buffer (75 mM KCl, 10 mM Tris-HCl (pH 7.5), 1.5 mM MgOAc, 1 mM KOAc, 2 mM dithiothreitol) with a 1:1000 dilution of each phenylmethylsulfonyl fluoride (10 mg/ml), pepstatin (5 mg/ml), leupeptin (10 mg/ ml), and RNasin (40 units/l). All probes were maintained at Ϫ80°C and used within 24 h of synthesis.
UV Cross-linking-UV cross-linking was performed as described previously (29,30). Briefly, an aliquot of radiolabeled RNA (1-4 ϫ 10 6 cpm) was added to a mixture containing 20 l of RSW (ϳ2 ϫ 10 6 cell eq/l) from DDT1-MF2 cells, 5 g of yeast tRNA, 4 mM dithiothreitol, 5 g of heparin, and 65 units of RNasin in a total volume of 50 l. After incubation for 10 min at 22°C, samples were placed in an ice slurry and exposed to short wave (254 nm) UV radiation for 3 min in a Stratagene (La Jolla, CA) model 1800 UV Stratalinker. The cross-linked RNA was digested with RNase A (0.5 mg/ml) and RNase T1 (10 units/ml) at 37°C for 30 min. Samples were solubilized in 50 l of Laemmli loading buffer for 10 min at 70°C, and proteins were resolved by SDS-PAGE. Gels were stained with Coomassie R-Blue (Sigma), followed by destaining and drying, and subjected to autoradiography for 1-5 days.
Immunodetection of HuR and hnRNP A1-RSW from DDT1-MF2 cells, and radiolabeled mRNAs, were prepared as described above. Western blotting and immunoprecipitation of UV cross-linked protein were performed using a modification of previously described methods (29,36). Briefly, following UV cross-linking of RSW proteins to either the hamster ␤ 2 -AR 20-mer or the human ␤ 1 -AR 160-mer, the mixture was pre-cleared with mouse non-immune serum followed by the addition of 40 l of protein G-Sepharose beads (Amersham Pharmacia Biotech) pre-equilibrated in RIPA Buffer (50 mM Tris (pH 8.0), 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS). Proteins were immunoprecipitated with either anti-HuR antibody (provided by Dr. Henry Furneaux) or anti-hnRNP A1 (9H10, provided by Dr. Gideon Dreyfuss) or with non-immune serum. The lysate, antibody, protein G-Sepharose bead complexes were washed three times in RIPA buffer followed by RIPA buffer with 0.5 M NaCl. The pellet was resuspended in Laemmli buffer, boiled for 5 min, and the proteins resolved by SDS-PAGE (10%). Gels were dried, and radiolabeled proteins were visualized by autoradiography. RNA Affinity Chromatography-Dynabeads ® M-280 streptavidin (Dynal) were washed twice with 0.1 M NaOH, 0.5 M NaCl and twice with 0.1 M NaCl to remove any RNases present on the beads. Beads were equilibrated in Column Buffer (see "In Vitro Transcription") with 1:1000 dilutions of phenylmethylsulfonyl fluoride, pepstatin, leupeptin, and RNasin. In vitro transcribed, biotinylated (50%) human ␤ 1 -AR 160-mer RNA was added to the Dynabead column, and the reaction tube was mixed while flipping at 22°C for 10 min. The RNA was "fixed" to the column with 2 ϫ 1 M NaCl washes in Column Buffer. The column was equilibrated in three washes of Column Buffer and RSW (ϳ3 ϫ 10 8 cell eq) was added to the column and the tube was mixed while flipping at 4°C for 30 min. Dynabeads were again washed three times: once with Column Buffer, and twice with Column Buffer containing 0.15 M KCl, while rotating at 4°C for 15 min each wash. Protein was eluted from the column with 0.3, 0.5, 1.0, and 2.0 M KCl washes in Column Buffer while rotating at 4°C for 15 min. All fractions extracted from the column were stored at 0°C. Fractions were dialyzed into Column Buffer A with 150 mM KCl for 1 h, and UV cross-link was performed on the dialyzed fractions using the radiolabeled hamster ␤2-AR 20-mer to detect the presence of ␤ARB protein (as described above).
Protein Isolation and Mass Spectrometry Analysis-Following resolution of RSW proteins by SDS-PAGE (10% acrylamide, Tris glycine buffer), proteins were stained with Colloidal Coomassie G-250 Stain as described by the manufacturer (Novex). A single band with an M r nearly identical to the autoradiographic signal obtained by UV crosslinking RSW proteins to both the human ␤ 1 -AR 160-mer and the hamster ␤ 2 -AR 20-mer was removed from the gel. The gel fragment was washed twice with 50% HPLC-grade acetonitrile in water and frozen at Ϫ80°C for subsequent analysis. Peptide composition analysis was performed by Dr. William Lane at the Harvard Microchemistry Facility, Harvard University. The specific method utilized was microcapillary reverse-phase HPLC nano-electrospray tandem mass spectrometry using a Finnigan LQC quadrupole ion-trap mass spectrometer.
Production of Recombinant GST-HuR and GST-hnRNP A1 Proteins-Recombinant GST-HuR was produced and purified as described previously (56). Briefly, a cDNA encoding amino acids 2-326 of the HuR coding region was ligated into pGEX2T. An overnight culture of Escherichia coli transformed with pGEX-HuR was diluted 1:50 in LB medium, and upon reaching an A 600 of 0.4 O.D. units, was induced with isopropyl-1-thio-␤-D-galactopyranoside for 4 h. Cells were pelleted, resuspended in buffer A (50 mM Tris, pH 8.0, 200 mM NaCl, 1 mM EDTA), and lysed with lysozyme (0.2 mg/ml), and 1% Triton. The lysate was centrifuged for 30 min at 12 000 ϫ g, and the resulting supernatant loaded onto a glutathione-agarose column. Following column washing with buffer B (50 mM Tris, pH 8.0, 200 mM NaCl, 1 mM EDTA, 1% Triton), GST-HuR was eluted with 50 mM Tris, pH 8.0, 5 mM glutathione. Protein concentration was determined by comparison with known quantities of BSA on a Coomassie-stained SDS-PAGE, as well as by the method of Bradford (Bio-Rad). Appropriate protein purification was confirmed by Western blot using a monoclonal antibody (generously provided by Dr. Henry Furneaux).
Recombinant GST-hnRNP A1 protein was produced as for GST-HuR, above, and similar to the method described previously by Hamilton et al. (47). The plasmid for GST-hnRNP A1 was generously provided by Dr. William Rigby.
Determination of Protein Affinity for Substrate mRNAs-The affinity of purified, recombinant GST-HuR and GST-hnRNP A1 proteins for substrate mRNAs was determined as described previously by DeMaria and Brewer (34). Briefly, 1 fmol of 32 P-labeled, in vitro transcribed RNA was incubated with increasing concentrations of purified recombinant protein. Complexes were resolved by electrophoresis through non-dena-turing 5-6% polyacrylamide gels and visualized by autoradiography. Bands representing the remaining "free" RNA were used for densitometric quantification. The signal for remaining "free probe" was plotted versus protein concentration. Apparent K d values were defined as the protein concentration at which ϳ50% of the "free" target RNA was shifted above base line. Numerical values were determined using the nonlinear regression modeling software program, GraphPad Prizm (GraphPad Software, San Diego, CA). In cases where there was an abrupt shift of labeled RNA instead of a gradual decrease of free RNA and increase in bound RNA (likely due to cooperative binding), the apparent K d was estimated as the protein concentration at which this immediate shift occurred (n ϭ 3-4 for each RNA).

RESULTS
Previously, we have described an ϳM r 36,000 mRNA binding activity, ␤ARB, that is up-regulated by ␤-AR stimulation and that displays preferential affinity for AϩU-rich regions, (30). We and others have hypothesized that ␤ARB may be involved in ␤-agonist-mediated destabilization of ␤-AR mRNAs (3, 16, 17, 29 -31), as well as potentially, other AϩU-rich mRNAs. Other than our previous results indicating that ␤ARB and AUF1 are not immunologically the same protein, the exact identity of ␤ARB has remained unknown (29).
Further characterization of ␤ARB was pursued by two independent means: subcellular fractionation and electrophoresis, and RNA affinity chromatography. In each case, to enrich for ␤ARB binding activity, hamster DDT1-MF2 cells were treated with 10 Ϫ5 M (Ϫ)-isoproterenol for 48 h, which we have shown previously to up-regulate cytosolic and polysome-associated ␤ARB (30). Cells were homogenized and RSW preparations were produced as described previously (29,61).
As a first attempt at purification of ␤ARB protein, RNA affinity chromatography was performed using two different biotinylated substrate RNAs, the hamster ␤ 2 -AR 20-mer (16), which in our hands was unsuccessful (data not shown), and the human ␤ 1 -AR 160-mer (29), which was successful. Nucleotide sequences for each of these RNAs are depicted in Fig. 1. To produce target RNA, the ␤ 1 -AR 160-mer was in vitro transcribed in the presence of biotinylated CTP. The RNA was incubated with RSW followed by passage through an agarosebead streptavidin column (Dynal Corp.) and the column was washed, followed by elution with increasing concentrations of KCl. To detect the presence of ␤ARB protein (bioassay), a portion of each fraction was incubated with radiolabeled, in vitro transcribed hamster ␤ 2 -AR 20-mer followed by UV crosslinking and resolution by SDS-PAGE. As seen in Fig. 2, a distinct single band at an ϳM r of 36,000 was present in low salt (0.15 M KCl) washes 2 and 3, and in the first 0.3 M KCl elution fraction.
FIG. 1. ␤-AR RNA sequences used to detect ␤ARB protein binding activity. In vitro transcribed RNA sequences encoding CϩU-and AϩU-rich regions within the 3Ј-UTRs of the human ␤ 1 -AR (160-mer) and the hamster ␤ 2 -AR (20-mer) were used purify and/or to detect ␤ARB protein binding.
Using the affinity chromatography approach, the amount of protein available for peptide analysis was very limited. Therefore, in parallel experiments, we used the more direct approach of isolating ␤ARB protein by one-dimensional SDS-PAGE. Silver staining of SDS-PAGE gels of highly purified RSW preparations revealed a distinct, isolatable band at an M r essentially identical to the autoradiographic signal for ␤ARB obtained by UV cross-linking of the human ␤ 1 -AR 160-mer or hamster ␤ 2 -AR 20-mer to RSW proteins, Fig. 3. Using 3 ϫ 10 8 cell eq, RSW protein was again isolated by SDS-PAGE followed by protein detection with Novex Colloidal Coomassie Stain (data not shown). The gel fragment containing the single stained band was excised, processed in acetonitrile, and shipped frozen to Dr. William Lane at the Harvard Microchemistry Facility. Mass spectrometric analysis of the tryptic digest fragments contained sequences indicative of two distinct proteins: a short splice form of hnRNP A1, and the elav-related protein, HuR. The exact peptide fragments and their relative location within the full-length HuR and hnRNP A1 proteins are shown in Fig.  4. An extensive search of the mass spectrometry data base of unknown peaks failed to identify any other known proteins.
As stated above, Coomassie-stained gels revealed single discrete band at the same relative M r as the autoradiographic signal indicative of ␤ARB binding activity (ϳM r 36,200 and 36,400, respectively). The calculated molecular weights for the identified splice form of hnRNP A1 and for HuR are 34,152 and 36,069 daltons, respectively. Additionally, the two proteins have almost identical pI values (9.27 and 9.23, respectively); thus, it is not overly surprising that, in isolating the single visible protein band by one-dimensional SDS-PAGE, both two proteins were recovered. However, it is entirely possible that only HuR or hnRNP A1 was the only protein actually visible.
To verify by an independent means the presence of these two proteins in RSW, Western blots were performed. As seen in Fig.  5, both proteins are readily detectable in RSW preparations of (Ϫ)-isoproterenol-treated hamster DDT1-MF2 cells.
To begin to address the issue of which protein may be more biochemically and physiologically relevant, immunoprecipitations of HuR and hnRNP A1 proteins UV cross-linked to radiolabeled ␤-AR mRNAs were performed. As seen in Fig. 6, an overnight exposure (short exposure) of the autoradiogram revealed immunoprecipitated HuR protein cross-linked to the human ␤ 1 -AR 160-mer probe. However, after a 1-week exposure (long exposure), the autoradiogram revealed (minimally) immunoprecipitated hnRNP A1 (Fig. 6). Given that approxi-mately the same signal intensity is seen for HuR and hnRNP A1 proteins upon Western blotting of equivalent amounts of RSW protein (Fig. 5), it appears that, at least for this mRNA substrate (human ␤ 1 -AR 160-mer), HuR is the dominant mRNA-binding protein component of ␤ARB.
It should be recalled that ␤ARB binding activity was first described for the 3Ј-UTR of the hamster ␤ 2 -AR mRNA (30). Further, the 20-mer AϩU-rich region, within the context of the full-length hamster ␤ 2 -AR 3Ј-UTR, has been shown to be necessary for ␤ARB binding and for ␤-agonist-mediated mRNA destabilization (16). Therefore, additional immunoprecipitation reactions were performed using both the hamster ␤ 2 -AR 20-mer and the full-length hamster ␤ 2 -AR 3Ј-UTR as probes. As seen in Fig. 7, immunoprecipitated HuR protein is evident  (29). To detect ␤ARB protein by UV cross-linking, RSW protein (4 ϫ 10 7 cell eq) was incubated with radiolabeled RNA encoding the hamster ␤ 2 -AR 3Ј-UTR ARE (20-mer) followed by SDS-PAGE as described previously (29). For silver staining, RSW (2 ϫ 10 7 cell eq) was electrophoresed side-by-side with the UV cross-linked products. The gel was cut in half, and the UV cross-linking side used for autoradiography (2-h exposure); in parallel, non-UV-cross-linked RSW protein was subjected to silver staining. A single protein band (within the box) at an apparent at ϳM r 36,000 co-migrates precisely with the single band on the autoradiogram. for both radiolabeled mRNA probes. However, immunoprecipitated hnRNP A1 protein is only evident when the probe used was the full-length hamster ␤ 2 -AR 3Ј-UTR. From these experiments, we conclude that the protein component(s) of the mRNA binding activity formerly described as ␤ARB depends on the mRNA substrate used. However, regardless of substrate probe, the dominant protein binding to RNA appears to be HuR.
To address directly the biological issue of mRNA-binding protein affinity, nondenaturing gel-shift experiments were performed using purified, recombinant GST-HuR and GST-hnRNP A1 proteins and in vitro transcribed, radiolabeled ␤-AR mRNAs. By nondenaturing gel-shift analysis, HuR binds with high affinity to both the human ␤ 1 -AR 160-mer and to the full-length hamster ␤ 2 -AR 3Ј-UTR (ϳ20 nM) (Fig. 8A), and with relatively high affinity to the hamster ␤ 2 -AR 20-mer (ϳ50 nM). Similarly, GST-hnRNP A1 binds to the human ␤ 1 -AR 160-mer and the full-length hamster ␤ 2 -AR mRNAs with high affinity (ϳ20 nM) (Fig. 8B). In contrast, hnRNP A1 does not bind the hamster ␤ 2 -AR 20-mer with detectable affinity (data not shown). This result is consistent with our immunoprecipitation and affinity purification data. As an appropriate control, we have demonstrated previously that GST alone does not bind to the c-Myc ARE. 2 In non-denaturing gel shifts of ␤-AR mRNAs with both HuR and hnRNP A1 proteins, increasing protein appears to produce a progressively greater degree of shift. One potential explana-tion for this observation is that HuR and hnRNP A1 may undergo multimerization, as has been demonstrated for AUF1 (33,77). DISCUSSION By direct peptide sequence analysis, by immunological techniques (Western blotting and UV cross-linking/immunoprecipitation), and by functional analysis (mRNA binding activity), we have identified the protein components of ␤ARB as HuR and hnRNP A1. The data suggest that HuR is the dominant biologically active constituent of ␤ARB, at least in terms of the relative abilities of the two proteins to bind to the mRNA substrates examined. In this context, the relative signal strength for HuR in UV cross-linking/immunoprecipitation assays is severalfold greater for HuR than that for hnRNP A1. Although this is not a direct measurement of protein abundance, given that the relative signal strength of hnRNP A1 and HuR are not detectably different by Western blotting of equivalent amounts of RSW proteins (Fig. 5), the greater signal strength of immunoprecipitated HuR may be a function of higher affinity mRNA binding. Additionally, only HuR binding, not hnRNP A1 binding, is detectable by UV cross-link/immunoprecipitation to the isolated hamster ␤ 2 -AR 20-mer ARE. However, differential efficiencies of immunoprecipitation can by no means be ruled out. As described by Hamilton et al. (46,47), hnRNP A1 is inefficiently immunoprecipitated with antibodies directed against C-terminal epitopes when preceded by UV cross-linking. Therefore, it is possible that our data indicating the dominance of HuR binding instead reflect inefficiencies in immunological detection of hnRNP A1 protein. This argument can be extended by our finding that both HuR and hnRNP A1 bind to the human ␤ 1 -AR 160-mer with similar affinities (ϳ20 nM). However, this argument is inconsistent with the finding that only HuR and not hnRNP A1 bind to the isolated AϩU-rich hamster ␤ 2 -AR 20-mer, data that seem to indicate again that either HuR is the dominant binding protein or that HuR is considerably less constrained in its binding requirements than is hnRNP A1. In other experiments, we have found similar differences in the relative constraints for high-affinity binding between HuR and AUF1 with HuR binding being less dependent on exact sequence or on the composition of flanking sequence. 2 The above data present an interesting quandary in that HuR, or at least its overexpression, has been generally associated with stabilization, not destabilization, of ARE-containing mRNAs (51,52,59,63). Similarly, although the cytosolic function of hnRNP A1 remains obscure, the assumption has been that it, too, is associated with the stabilization of polyadenylated, cytoplasmic mRNAs (47). Although this has not been tested directly, overexpression of a closely related protein, hnRNP A0, when compared simultaneously to HuR, has been found not to stabilize ARE-containing mRNAs (51). These observations, although far from definitive, are in distinct contrast to the hypothesis that ␤ARB is associated with mRNA destabilization (16, 17, 29 -31).
There are a number of alternative explanations for these findings, including the possibility that HuR and/or hnRNP A1, much like other mRNA-binding proteins, may be multifunctional (35, 39, 47, 64 -69). As an example, although AUF1 has clearly been associated with mRNA destabilization, its role is potentially far more complex. Kiledjian et al. (39) have recently identified AUF1/hnRNP D as a component of the multi-protein ␣-globin mRNA stabilization complex. Further, Dempsey et al. (68,69) have found that hnRNP D (AUF1) and nucleolin constitute the component parts of the LR1 heterodimer, a B-cellspecific DNA binding complex that acts as a transcriptional activator. Thus, AUF1 may subserve different roles, i.e. mRNA FIG. 4. ␤ARB peptide sequences within the context of hnRNP A1 and HuR. RSW protein from isoproterenol-stimulated hamster DDT1-MF2 cells were separated by SDS-PAGE, and the protein band co-migrating with the autoradiographic signal for ␤ARB was sent to the Harvard Microchemistry facility for peptide sequencing. Tryptic digest of the protein reveals mass spectrometric derived spectra for peptide fragments (bold) with exact identity for two RNA-binding proteins known to interact with AϩU-rich mRNAs.  (30). RSW preparations (4 ϫ 10 6 cell eq) were used to immunologically confirm the presence of hnRNP A1 and HuR proteins. destabilization or stabilization (and/or DNA binding), depending on the specific context, including: cell type, intracellular location, specific substrate mRNA, specific mRNA binding site, and state of phosphorylation.
Although we have demonstrated that ␤ARB is composed of HuR and hnRNP A1, it is possible, however unlikely, that ␤ARB contains other as yet unidentified ARE-binding proteins capable of affecting ␤-agonist-mediated mRNA destabilization. Thus, one potential explanation is that the HuR and hnRNP A1 components of ␤ARB are actually stabilizing but under conditions of ␤-agonist stimulation, another protein, either an unidentified component of ␤ARB or another protein such as AUF1/ hnRNP D, is necessary for mRNA destabilization. This situation is complicated by previous data that support the hypothesis that constitutive mRNA turnover and agonist-mediated destabilization may be separable processes, utilizing different decay pathways, and affected by different trans-acting factors. In this context, we have published previously that  6. UV cross-linking and immunoprecipitation of hnRNP A1 and HuR proteins. RSW proteins (ϳ4 ϫ 10 6 cell eq) were incubated with an in vitro transcribed RNA probe encoding the human ␤ 1 -AR 160-mer. Following UV cross-linking, reactions were immunoprecipitated with antiserum to hnRNP A1 or HuR or with non-immune (N.I.) serum. Shown are short (24 h, left panel) and long exposure (1 week, right panel) autoradiograms. Immunoprecipitated HuR protein is evident after a short exposure, whereas hnRNP A1 protein is evident only following a long exposure.
In light of the above, we believe one of three hypotheses will ultimately be substantiated. First, HuR (and/or hnRNP A1) could indeed be destabilizing ␤-AR mRNA upon agonist stimulation. Overexpression studies of HuR, to date, have not examined stability of any G-protein-coupled receptor mRNAs. Thus, it is at least theoretically possible that HuR may differentially modulate target mRNA stability dependent upon the specific mRNA substrate. Furthermore, localization and phosphorylation state (among other factors) are highly likely to ultimately influence the effect of HuR (or any other mRNA-binding protein) on target mRNA stability, factors that may or may not be masked in overexpression studies.
Second, it is possible that HuR (and/or hnRNP A1) may not play any direct role in modulating ␤-AR mRNA stability. There are numerous mRNA-binding proteins, and although HuR (and/or hnRNP A1) are apparently up-regulated following ␤-agonist stimulation (30), this increased expression may be unrelated to ␤-AR mRNA destabilization. For example, ␤-AR stimulation has recently been demonstrated to be "mitogenic," in that it can stimulate various tyrosine kinase signaling pathways, including p38 mitogen-activated protein kinase (70 -74). Indeed, p38 mitogen-activated protein kinase has recently been demonstrated to affect changes in mRNA stability assumed to be caused by AUF1 or other ARE-binding proteins (67,75,76).
Third, HuR (and/or hnRNP A1) may normally play a protective role in modulating ␤-AR mRNA stability. Upon ␤-agonist stimulation, although HuR (and/or hnRNP A1) are significantly up-regulated as detectable by UV cross-linking or immunoblot, an mRNA destabilizing protein (i.e. AUF1) may gain an advantage in binding to ␤-AR mRNAs. This advantage may be gained through increased affinity (higher than that of HuR and/or hnRNP A1) for ␤-AR mRNAs, perhaps by phosphorylation (28), or co-localization/shuttling with ␤-AR mRNAs. To this end, we have preliminary data suggesting a co-localization of AUF1 with hamster ␤ 2 -AR mRNA in polysomes following agonist stimulation. 3 In summary, the mRNA binding activity, ␤ARB, is composed of (at least) two mRNA-binding proteins with a high affinity for A/CϩU-rich elements, HuR and hnRNP A1. The evidence to date suggests that these two proteins may generally function to stabilize, not destabilize, ARE-containing mRNAs. However, it is also clear that the full functional potential and mechanisms by which these proteins affect mRNA stability are not fully understood. What remains to be determined is the role, if any, of either of these proteins, as well as the role of other proteins (i.e. AUF1/hnRNP D), in modulating the stability of G-proteincoupled receptor mRNAs.