Structural Determinants in AUF1 Required for High Affinity Binding to A + U-rich Elements*

AUF1 is an RNA-binding protein that contains two nonidentical RNA recognition motifs (RRMs). AUF1 binds to A + U-rich elements (AREs) with high affinity. The binding of AUF1 to AREs is believed to serve as a signal to an mRNA-processing pathway that degrades mRNAs encoding many cytokines, oncoproteins, and G protein-coupled receptors. Because the ARE binding activity of AUF1 appears central to the regulation of many important genes, we analyzed the domains of the protein that are important for this activity. Examination of the RNA binding affinity of various AUF1 mutants suggests that both RRMs may be required for binding to the human c-fos ARE. However, the two RRMs together are not sufficient. Highest affinity binding of AUF1 to an ARE requires an alanine-rich region of the N terminus and a short glutamine-rich region in the C terminus. In addition, the N terminus is required for dimerization of AUF1. However, AUF1 binds an ARE as a hexameric protein. Thus, protein-protein interactions are important for high affinity ARE binding activity of AUF1.

RNA processing is an important component of regulated gene expression in eukaryotic cells. Together, the rates of transcription, pre-mRNA splicing, mRNA transport, translation, and mRNA degradation determine the steady-state amount of mRNA, and hence protein, that will be present in a cell at a given time. Each of these processes of RNA metabolism involves RNA-binding proteins that exhibit specific protein-RNA interactions (reviewed in Ref. 1). Thus, defining how such proteins interact with their RNA substrates is integral to understanding the complex control of RNA processing.
A variety of conserved protein motifs mediate specific protein-RNA interactions (2). Perhaps the most common and well characterized of these motifs is the RNA recognition motif (RRM), 1 also called a consensus sequence RNA-binding domain. This motif consists of 80 -90 amino acids containing two conserved sequences: a highly conserved octamer motif, RNP-1, and a less conserved hexamer motif, RNP-2 (2, 3). The general structure of an RRM consists of a ␤␣␤-␤␣␤ folding pattern in a four-stranded ␤-sheet with the two ␣-helices packed against one face of the sheet. The RNP-1 and RNP-2 motifs lie on the two central strands at the center of the ␤-sheet. These motifs probably provide general RNA binding activity. The mechanisms by which RRMs provide sequence-specific or structurespecific RNA recognition are unknown. However, recognition of specific targets is thought to be provided by unique amino acids located in intradomain loops and tails. Nonetheless, this seemingly simple view is complicated by two observations. (i) Multiple RRMs within some proteins are required for high affinity RNA binding (see Ref. 3 and references therein); and (ii) complex communication can occur between amino acids in the intradomain loops and tails (4). Thus, the efforts of many laboratories are being directed toward dissecting the molecular mechanisms of specific RNA recognition by RNA-binding proteins.
We previously molecularly cloned the RNA-binding protein AUF1. AUF1 binds with high affinity to a variety of AREs present in the 3Ј-UTRs of mRNAs encoding many cytokines, oncoproteins, and G protein-coupled receptors (5)(6)(7)(8). Binding of AUF1 to an ARE is thought to act as a signal to target the mRNA for rapid degradation. Because the ARE binding activity of AUF1 appears central to the regulation of many important genes, we analyzed the domains of the protein that are important for this activity. Binding assays with various mutants suggest that both RRMs may be required for ARE binding, but they are not sufficient. Amino acids flanking the RRMs seem to contribute to ARE binding, but their contribution is not simply to maintain or allow proper folding of the protein. Instead, the flanking amino acids appear to allow protein-protein interactions that may be crucial for high affinity ARE binding activity. In particular the alanine-rich N terminus is important for homodimerization of AUF1. However, AUF1 binds an ARE as a hexameric protein. Thus, additional protein-protein interactions are involved in the high affinity ARE binding activity of AUF1.

EXPERIMENTAL PROCEDURES
All enzymes and plasmid vectors were obtained from Promega Corp. (Madison, WI) unless otherwise noted. All plasmid constructs were confirmed by both restriction analyses and either dideoxy sequencing with Sequenase (version 2.0, U.S. Biochemical Corp.) or by cycle sequencing with Taq Polymerase (Perkin-Elmer).
Purification of His 6 -AUF1 Proteins, Radiolabeling of RNA Substrates, and Determination of Apparent K d Values-Wild-type and mutant His 6 -AUF1 fusion proteins were expressed and purified as described by Pende et al. (7). The concentration of each His 6 -AUF1 fusion protein was estimated by comparison of the intensities of the full-length polypeptide and dilutions of BSA in Coomassie-stained SDS-polyacrylamide gels. RNA containing the c-fos ARE was synthesized by transcription of BglII-digested plasmid p␣19R␤ϩARE (5), using T3 RNA polymerase and [␣-32 P]UTP (800 Ci/mmol). Apparent K d values were determined by electrophoretic mobility shift assays and PhosphorImager analyses as described by DeMaria and Brewer (8). Free probe concentration was plotted versus His 6 -AUF1 concentration, and apparent K d values were determined as the protein concentration at which 50% of the RNA was bound (9).
Circular Dichroism Measurements-Purified wild-type His 6 -AUF1-(1-286) and His 6 -AUF1-(69 -229) protein samples were diluted into a buffer containing 10 mM Tris-HCl (pH 7.0), 5 mM magnesium acetate, and 100 mM potassium acetate to a final protein concentration of 5 M. CD spectra were recorded from 190 -230 nm at room temperature on a Jasco 720 spectropolarimeter using a cuvette with a 0.05-cm path length. The ellipticity values (mdeg) were converted to mean residue ellipticity (deg⅐cm 2 ⅐dmol Ϫ1 ) according to the following formula (10): [] ϭ /10CL, where [] represents mean residue ellipticity; is ellipticity (mdeg), C is residue molar concentration (i.e. molar concentration of His 6 -AUF1-(1-286) ϫ 326 residues/mol or molar concentration of His 6 -AUF1-(69 -229) ϫ 201 residues/mol), and L is path length of the cell in cm. For denaturation studies, the protein was incubated in buffer containing various concentrations of GdnHCl (0 -7.6 M). For the samples and blanks, the ellipticity (mdeg) at 222 nm was measured, where the signal was collected for a total of 300 points to produce an averaged value for each sample and blank. The appropriate blank value was then subtracted from the sample value at each GdnHCl concentration.
A plot of CD (in mdeg) versus GdnHCl concentration for each protein was constructed. The standard free energy of denaturation (⌬G d 0 ) was derived from each curve according to the method of Sparks et al. (11) based upon the denaturant binding model described by Pace (12). ⌬G d 0 was determined for each protein using two different protein preparations to obtain an average Ϯ S.D.
Gel Filtration Analyses-For all analyses, a 1 ϫ 19-cm column of Sephacryl S-300 (Pharmacia) was equilibrated in buffer A (10 mM Tris-HCl (pH 7.5), 5.5 mM magnesium acetate, 100 mM potassium acetate). Purified recombinant fusion protein at a concentration of 6 M in 300 l of buffer A (final volume) was loaded onto the column, and 60 300-l fractions were collected. The protein elution profile was determined spectrophotometrically at 280 nm using a UA-5 absorbance detector (Isco).
Elution profiles of these proteins were determined spectrophotometrically as above. The void volume was determined using blue dextran.
For examination of the RNA-bound form of the wild-type His 6 -AUF1-(1-286) fusion protein, an RNA-binding reaction was performed using either 32 P-labeled c-fos ARE or rabbit ␤-globin 3Ј-UTR (as a negative control). The reaction was treated with ultraviolet light to cross-link the protein to the RNA, and then it was treated with RNase A as described by DeMaria and Brewer (8). This was loaded onto a calibrated Sephacryl S-300 column in which the column buffer contained 10 mg/ml Torula RNA (Sigma) to reduce nonspecific retention (13). Fractions were collected as described above, and protein-RNA complexes were detected by Cerenkov counting.
Velocity Sedimentation Centrifugation Analyses-Five to twenty percent sucrose gradients in buffer A contained 1 g/ml pepstatin A, 1 g/ml leupeptin, and 0.1 mM phenylmethylsulfonyl fluoride. Purified recombinant fusion protein at a concentration of 6 M in 200 l of buffer A was layered onto a preformed gradient and centrifuged at 39,000 rpm in a SW41 rotor (Beckman) for 20 -30 h at 4°C. Fifty-six 200-l fractions were collected, and the protein elution profile was determined spectrophotometrically at 280 nm using a UA-5 absorbance detector. In addition, protein in selected fractions was precipitated with 10% trichloroacetic acid, 20 g/ml lysozyme (as carrier) by incubation on ice for 30 min and centrifugation at 4°C for 30 min. Each precipitate was resuspended in 10 mM Tris (pH 7.5) and fractionated by SDS-polyacrylamide gel electrophoresis, and the resulting gels were either silverstained or stained with Coomassie Blue. Velocity sedimentation centrifugation was also performed with the protein standards ␤-amylase (8.9 S), BSA (4.3 S), and soybean trypsin inhibitor (2.3 S). Elution profiles of these proteins were determined spectrophotometrically and by gel staining. For each His 6 -AUF1 protein, the sedimentation coefficient, s, was obtained from the following ratio with respect to each standard.
Distance traveled from meniscus by His 6 -AUF1 Distance traveled from meniscus by standard ϭ s of His 6 -AUF1 s of standard (Eq. 1) For each protein, the sedimentation coefficient was determined as the average of the values obtained from comparison to each of the standards in two separate experiments. For examination of the RNA-bound form of the wild-type His 6 -AUF1-(1-286) fusion protein, an RNA-binding reaction was performed using either 32 P-labeled c-fos ARE or rabbit ␤-globin 3Ј-UTR (as a negative control). The reaction was treated with ultraviolet light to cross-link the protein to the RNA and then treated with RNase A as described by DeMaria and Brewer (8). This was run on a sucrose gradient, and fractions were collected as described above. Protein-RNA complexes were detected by Cerenkov counting.
Molecular Mass Determinations of RNA-bound and -unbound His 6 -AUF1 Proteins-The monomer molecular mass of each protein was determined from the amino acid composition. The native molecular mass of each protein was calculated from the equation (14), where is the viscosity (g/cm⅐s), R S is the Stokes radius (cm), N is Avogadro's number, and s is the sedimentation coefficient. The partial specific volumes, v , of His 6 -AUF1-(1-286), His 6 -AUF1-(1-194), His 6 -AUF1-(92-286), and His 6 -AUF1- (29 -194) were calculated as 0.7217 ml/g, 0.7217 ml/g, 0.7301 ml/g, and 0.7270 ml/g, respectively, from amino acid composition by the method of Laue et al. (15). The solvent density, , was measured as 1.0059 g/ml. The frictional ratio, ƒ/ƒ 0 , of each protein was calculated from the equation (14),

RESULTS
Identification of the Domains of AUF1 Critical for Its High Affinity ARE Binding Activity-Using quantitative electrophoretic mobility shift assays, we showed previously that purified, recombinant His 6 -p37 AUF1 fusion protein binds the human c-fos ARE with an apparent K d of 7.8 Ϯ 0.4 nM. (The wild-type fusion protein will hereafter be referred to as His 6 -AUF1-(1-286).) As shown in Fig. 1, the predicted AUF1 polypeptide contains two tandem, nonidentical RRMs. Adja-cent to RRM1 is an alanine-rich, N-terminal region of 68 amino acids. C-terminal to RRM2 is an 8-amino acid region containing six glutamine residues. This is followed by a short C-terminal region. To examine the involvement of the RRMs and other regions of AUF1 in ARE binding, we constructed several N-and C-terminal truncation mutants of His 6 -AUF1 that were purified by Ni 2ϩ -chelate chromatography. Each purified recombinant fusion protein was tested for binding to the wild-type, 32 P-labeled c-fos ARE by gel mobility shift assays. The dissociation constant, or apparent K d , was calculated for each protein by determining the protein concentration at which 50% of the RNA probe was bound (8).
His 6 -AUF1-(92-286) and His 6 -AUF1-(1-194) lack a portion of either RRM1 or RRM2, respectively, and these mutant proteins bound the c-fos ARE with respective 100-and 70-fold lower affinities than wild-type His 6 -AUF1 (Fig. 1). These results suggest that both RRMs are important for high affinity ARE binding. In support of this suggestion, His 6 -AUF1-(1-257), which retains both RRMs but lacks the C-terminal 29 amino acids, bound the c-fos ARE with an apparent K d of 5.3 nM, similar to the 7.8 nM apparent K d for wild-type His 6 -AUF1 ( Fig. 1 Fig. 4), in which most of the amino acids comprising the RRMs were deleted. Comparison of this mutant with the wildtype protein using UV-cross-linking analysis with 32 P-labeled c-fos ARE revealed undetectable binding by His 6 -AUF1-(1-29/ 195-286) (data not shown). Together, these experiments indicate that the RRMs of AUF1 are necessary for ARE binding, but they are not sufficient.
Since the C-terminal 29 amino acids of AUF1 are dispensable for high affinity ARE binding, the importance of both the N-terminal and glutamine-rich regions for ARE binding was examined. His 6 -AUF1-(69 -257) bound the c-fos ARE with 180fold lower affinity than the wild-type protein (Fig. 1). His 6 -AUF1-(69 -257) and His 6 -AUF1-(1-257) differ only by the presence of the N-terminal region in the latter; therefore, the Nterminal region of AUF1 is required for high affinity ARE binding activity. His 6 -AUF1-(1-229) was then created to assess the importance of amino acids C-terminal to RRM2 for ARE binding. His 6 -AUF1-(1-229) bound the c-fos ARE with an affinity of 207 nM, 27-fold lower than the affinity displayed by the wild-type protein (Fig. 1). Thus, amino acids C-terminal to RRM2 are important for high affinity ARE binding. To determine if the glutamine-rich region contributes to high affinity binding, this 8-amino acid region (amino acids 240 -247) was deleted from the full-length protein. This mutant, His 6 -AUF1-(1-239/248 -286), bound the c-fos ARE with an apparent K d of 36 nM, a 4.5-fold lower affinity than that of the wild-type protein (Fig. 1). This result suggests that the glutamine residues located in this domain do contribute to the ARE binding activity of AUF1. However, based upon the low binding affinity of His 6 -AUF1-(1-229), amino acids in the C terminus flanking the glutamine residues also contribute to the RNA binding activity.
As a control for the possibility that one or more of the mutant proteins with low affinity binding had lost RNA binding activity during purification, ARE binding activity was analyzed in aliquots of the same bacterial extracts used for purification of the wild-type and mutant polypeptides. By UV-cross-linking analysis using these bacterial lysates and 32 P-labeled c-fos ARE (8), ARE binding was either undetectable or occurred with at least 100-fold lower affinity than wild-type His 6 -AUF1 for the mutant proteins that bound the c-fos ARE with low affinity in the mobility shift assays shown in Fig. 1 (data not shown). Therefore, putative loss of RNA binding activity during purification of the mutant proteins is unlikely. The results can be summarized as follows. (i) While the RRMs appear important for high affinity ARE binding activity, both RRM1 and RRM2 together are not sufficient for ARE binding. (ii) The N-terminal region of AUF1 is required for high affinity ARE binding activity. (iii) Amino acids C-terminal to RRM2, particularly the glutamines and amino acids flanking the glutamines, are also important for high affinity ARE binding. However, the C-terminal 29 amino acids appear not to contribute to RNA binding, since their truncation has no effect on ARE binding affinity.
Circular Dichroism Analyses-For many RRM-containing proteins, either one RRM or a combination of the RRMs is usually sufficient for high affinity RNA binding (see "Discussion"). Thus, the low (2200 nM) ARE binding affinity of the His 6 -AUF1-(69 -229) mutant (containing RRM1 plus RRM2 alone) is unusual among RRM-containing proteins. To examine the possibility that the low ARE binding affinity of this mutant might be simply due to its misfolding, the thermodynamic stabilities of the wild-type protein and the AUF1-(69 -229) mutant were measured. Both proteins have CD spectra similar to other RRM-containing proteins (3,16,17), suggesting that the RRMs are folded in a structure similar to that observed for this class of proteins (Fig. 2, upper panels). A combination of chemical denaturation with various concentrations of GdnHCl and ellipticity measurements at 222 nm was used to reveal changes in ␣-helical content as a function of GdnHCl concentration (Fig. 2, bottom panels). Application of the denaturant binding model of Pace (12) with these denaturation curves revealed that the thermodynamic stabilities (⌬G d 0 of unfolding) of the wild-type protein and the AUF1-(69 -229) mutant are similar (1.941 Ϯ 0.008 kcal/mol versus 1.9 Ϯ 0.1 kcal/mol, respectively). Therefore, it is unlikely that the low affinity of the AUF-(69 -229) mutant for the ARE is due to misfolding of the polypeptide. We conclude from the ARE-binding studies that amino acids flanking the RRMs are necessary for the high affinity ARE binding activity of AUF1. However, the analyses of the thermodynamic stabilities suggest that they do not simply serve to permit proper folding of the RRMs. Nonetheless, the amino acids flanking the RRMs must play some role in the RNA binding function of AUF1. Since some RNA-binding proteins self-associate (e.g. see Ref. 18), it is possible that protein domains flanking the RRMs of AUF1 are involved in AUF1 self-association. To address this question, the hydrodynamic properties of wild-type and mutant His 6 -AUF1 proteins were examined to determine their native molecular mass in solution.
Determination of Molecular Masses of His 6 -AUF1 Proteins-Gel filtration analysis of His 6 -AUF1-(1-286) was used to determine the Stokes radius of the native protein. Four protein standards were used to calibrate the Sephacryl S-300 column (Fig. 3, upper panel). His 6 -AUF1-(1-286) at a concentration of 6 M was loaded onto the column, and fractions were collected. The amount of His 6 -AUF1-(1-286) present in each fraction was determined spectrophotometrically, and relative amounts were plotted versus fraction number (Fig. 3, upper panel). Analysis of the gel filtration data by the method and equations of Ackers (19) yielded a Stokes radius of 3.6 nm (Table I). Gel filtration using 500 nM His 6 -AUF1-(1-286) produced a protein elution profile identical to that shown in Fig. 3 (data not shown), demonstrating that the protein did not form detectable aggregates at the higher (6 M) concentration. In addition, both purified cellular AUF1 and recombinant AUF1 (without a His 6 tag) synthesized by translation in vitro displayed the same elution profile as that in Fig. 3 (data not shown). This control FIG. 1. Analysis of ARE binding activity of mutant His 6 -AUF1 proteins. AUF1 contains two nonidentical RNA recognition motifs (labeled as RRM1 or RRM2, respectively), each of which contains two conserved RNP boxes depicted as solid black bars. AUF1 also contains an 8-amino acid glutamine-rich region C-terminal to RRM2 (labeled Q). Binding of each protein to the c-fos ARE was analyzed by electrophoretic mobility shift assays in which increasing concentrations of purified fusion protein were incubated with 100 pM 32 P-labeled RNA. Apparent K d values were demonstrates that the gel filtration profile shown in Fig. 3 is not unique to the bacterially expressed protein.
Gel filtration analyses were also performed for His 6 -AUF1-(1-194) and His 6 -AUF1-(92-286), since these mutant proteins do not bind an ARE with high affinity and together span the length of the wild-type protein. For each, 6 M protein was loaded onto, and eluted from, the column, and the amount of protein present in each fraction was determined spectrophotometrically (Fig. 3, upper panel). Analysis of these data yielded Stokes radii of 3.5 and 2.6 nm for His 6 -AUF1-(1-194) and His 6 -AUF1-(92-286), respectively ( Table I). The larger Stokes radii of His 6 -AUF1-(1-286) and His 6 -AUF1-  suggested that either each of these proteins contains more than one polypeptide or that each is a highly anisotropic monomer. To distinguish between these two possibilities, velocity sedimentation analyses were performed to determine sedimentation coefficients, which were then used in combination with the Stokes radii to calculate molecular mass values that are not dependent upon the shape of the molecule (14,20).
The sedimentation coefficients for His 6 (Table I). These values, along with the Stokes radii and partial specific volumes for each protein were used to calculate native molecular masses and frictional ratios (see "Experimental Procedures").
As presented in Table I, the molecular masses calculated for M of wild-type or mutant His 6 -AUF1 in a 300-l total volume was loaded onto a Sephacryl S-300 column, and 60 300-l fractions were collected ("Experimental Procedures"). The amount of His 6 -AUF1 protein in each fraction was assayed spectrophotometrically, and relative protein abundance was plotted versus fraction number. The Stokes radius of each protein standard used for column calibration is shown with an arrow, designating the peak fraction in which each standard eluted. Bottom panel, sedimentation velocity analyses of His 6 -AUF1 proteins. Sedimentation velocity analysis was performed twice; shown is the elution profile for one experiment. For each protein, the sedimentation coefficient was determined as the average of the values obtained from comparison with each of the standards in two separate experiments. For each protein, 6 M wild-type or mutant His 6 -AUF1 in a 200-l total volume was layered onto a 5-20% sucrose gradient and centrifuged at 39,000 rpm, and 56 200-l fractions were collected (see "Experimental Procedures"). The amount of His 6 -AUF1 protein in each fraction was assayed both spectrophotometrically and by gel staining of precipitated fractions. Relative protein abundance was plotted versus fraction number. The sedimentation coefficients of protein standards are each shown with an arrow, designating the peak fraction in which each standard eluted.  Fig. 4) were used to examine the importance of the N-terminal alanine-rich region for dimerization. As calculated from the experimentally determined values shown in Table I, the molecular mass of His 6 -AUF1- (29 -194) is consistent with that of a monomer. Thus, amino acids 1-28 appear required for dimerization. While this analysis does not allow one to predict the actual shape of the molecule, these proteins are clearly not spherical structures.
Since the native form of the AUF1 protein in solution is a dimer, the form of AUF1 protein bound to its RNA target sequence was next determined. A binding reaction was performed with the wild-type fusion protein and 32 P-labeled c-fos ARE. The protein concentration used was 20 nM, which is in the range of the apparent K d for binding to the c-myc and c-fos AREs (8). The protein was cross-linked to RNA using ultraviolet light to prevent dissociation of the protein from RNA during sample processing (22). Unbound RNA was digested with RNase A. Reactions were fractionated by gel filtration or by sucrose gradient centrifugation. RNA-bound His 6 -AUF1-(1-286) protein was detected by Cerenkov counting due to label transfer. The Stokes radius of the 32 P-labeled His 6 -AUF1-(1-286) protein is 6.4 nm, and its sedimentation coefficient is 8.6 S (Table II). From these values, the calculated molecular mass of the 32 P-labeled His 6 -AUF1-(1-286) protein is 227,700 Da. Since its monomer mass is 35,922 Da, the mass of the labeled protein is consistent with a hexameric structure for AUF1 when bound to the c-fos ARE. Three controls were performed. (i) A binding reaction with His 6 -AUF1 protein and 32 P-labeled rabbit ␤-globin 3Ј-UTR was analyzed by gel filtration and sucrose gradient centrifugation. No 32 P-labeled protein was recovered from the gel filtration column or the sucrose gradient, since AUF1 does not bind this RNA substrate. (ii) Gel filtration and sucrose gradient analyses of a binding reaction with 32 Plabeled c-fos ARE in the absence of His 6 -AUF1-(1-286) protein yielded only background counts in the fractions expected of a protein complex with a Stokes radius of 6.4 nm and a sedimentation coefficient of 8.6 S. (iii) A UV-cross-linked binding reac-

FIG. 4. Summary of apparent K d values for His 6 -AUF1 proteins.
A diagram of each protein is shown with its corresponding K d value. For the proteins whose affinities are labeled ND (not detected), no binding was detected by UV-cross-linking analyses using 32 P-labeled c-fos ARE as probe. a The 32 P-labeled c-fos ARE was incubated with wild-type His 6 -AUF1 protein, cross-linked using ultraviolet light, and then incubated with RNase A. The Stokes radius and sedimentation coefficient were determined from gel filtration and velocity sedimentation centrifugation, respectively, as described under "Experimental Procedures." The sedimentation coefficient was determined as the average of the values obtained in two separate experiments. The native molecular mass value, M, was determined from the R s and the s value, and the frictional ratio was calculated from M and R s values as described under "Experimental Procedures." The physical properties of the wild-type His 6 -AUF1 protein in the absence of RNA are shown for comparison. tion containing radiolabeled c-fos ARE and His 6 -AUF1-  was also analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography. The radiolabeled His 6 -AUF1-(1-286) polypeptide exhibited the same mobility as the protein alone (i.e. in the absence of RNA) (data not shown; e.g. see Fig. 1 in Ref. 8); no higher molecular weight forms were observed. This suggests that the large mass of the 32 P-labeled protein is not due to RNA bridging. Thus, we conclude that the hydrodynamic properties of the RNA-bound form of AUF1 are consistent with it being a hexamer. The frictional coefficients of His 6 -AUF1-(1-286) and the ARE-bound form of the protein were also compared (Table II). A frictional ratio of 1.3 for the unbound form corresponds to either a prolate or an oblate ellipsoid with an axial ratio of 6 (see Ref. 20). By contrast, a frictional ratio of 1.6 for the AREbound form of His 6 -AUF1 corresponds to either a prolate or an oblate ellipsoid with an axial ratio of 13. Again, while this analysis does not allow us to predict the actual shape of the molecule, the His 6 -AUF1 oligomer bound to RNA is clearly not spherical in structure. DISCUSSION We have shown previously that AUF1 binds AREs with high affinity and that the magnitude of this binding affinity is comparable with the affinities exhibited by several other RNAbinding proteins that recognize specific sequences or structures (8). In this study, our objective was to elucidate the structural features of AUF1 that mediate its ARE binding activity. Our results (summarized in Fig. 4) suggest that while both RRMs may be required for high affinity ARE binding, they are not sufficient. Additionally, our results suggest that amino acids adjacent to the RRMs participate in AUF1 binding to its cognate RNA target.
The prevailing paradigm for RNA-binding proteins that contain more than one RRM is that RRMs are essentially modular. The number of RRMs typically found within this class of RNAbinding proteins ranges between two and four (23). RRMs can be tandemly arranged or separated by intervening polypeptide sequence. In some RNA-binding proteins with two or more RRMs, a subset of one or more RRM-containing regions is generally sufficient for RNA binding (24 -29). Some other RNAbinding proteins do require all of their RRMs for RNA binding activity (30,31), but as a general rule, other regions of the protein are dispensable (31)(32)(33).
Our experiments suggest that AUF1 does not follow this paradigm of RRM modularity. In this respect, AUF1 exhibits a greater similarity to RNA-binding proteins that contain only a single RRM, which often do require amino acids adjacent to the RRM for high affinity target binding (2,23,34). For instance, the Drosophila Tra2 protein contains one RRM that is necessary but not sufficient for RNA binding (34). Like AUF1, Tra2 requires at least part of the region C-terminal to the RRM for RNA binding activity. Deleting a 90-amino acid region N-terminal to the Tra2 RRM has no effect on RNA binding. By contrast, AUF1 requires regions N-terminal to RRM1 and Cterminal to RRM2 for high affinity ARE binding activity.
Analyses of the hydrodynamic properties of various mutant proteins in the absence of target RNA indicate that the alaninerich, N-terminal 28 amino acids of AUF1 are necessary for dimerization, while the glutamine-rich region and the C-terminal domain appear not to be important for dimerization. Truncation of the N-terminal 28 amino acids from the full-length protein abolishes dimerization (Table I) and lowers its ARE binding affinity almost 10-fold to 62 nM (data not shown; Fig.  4). This result suggests that self-association may be crucial for high affinity ARE binding activity.
Alanine-rich regions of proteins are known to mediate pro-tein-protein interactions. For example, the alanine-rich domain of the transcriptional repressor Dr1 binds the TATA-binding protein (35), and the Oct-1 protein contains a C-terminal, alanine-rich domain that acts as a transcriptional repressor (36). However, to our knowledge, the structure(s) of these alaninerich regions and how they mediate protein-protein interactions are unknown. Domains N-or C-terminal to RRMs are known to be involved in self-association of RNA-binding proteins (27,34,37). In the case of Tra2, it was proposed that the region C-terminal to the RRM may play a role in formation of both homodimers and heterodimers and also cooperate in RNA binding (34). While the C-terminal 29 amino acids of AUF1 are dispensable for high affinity ARE binding, removal of all the amino acids C-terminal to RRM2 lowers its ARE binding activity 26-fold (Fig. 4). Amino acids C-terminal to RRM2 may thus be important for RNA interactions rather than protein dimer formation, since the entire C-terminal domain is dispensable for dimerization. The opposite is true for the C terminus of the Xenopus poly(A)-binding protein. For this protein, the large domain C-terminal to the four RRMs is not involved in RNA binding but is required for homodimerization (27).
While AUF1 is not unique in its ability to self-associate, it does appear to have unique structure-function relationships that govern self-association. Among RNA binding proteins containing multiple RRM domains, AUF1 is the only example, to our knowledge, in which abolishing dimerization significantly lowers RNA target affinity. Tra-2, which contains a single RRM, does require amino acids C-terminal to its RRM for RNA binding activity, but these residues also participate in maintaining the structure of the protein. By contrast, a mutant AUF1 containing only RRM1 and RRM2 appears, by hydrodynamic criteria, to be as structurally stable as the full-length protein. For most RNA-binding proteins that contain two or more RRMs, the polypeptide domains that participate in protein-protein interactions are not required for high affinity binding to RNA target sequences.
The unique structure-function relationships of AUF1 elucidated by these experiments are consistent with phylogenetic analysis of AUF1 cDNA and genomic sequences. When AUF1 RRMs are compared with other RRMs from a variety of species, they form a unique class of RRM-containing proteins unrelated to other proteins, such as small nuclear ribonucleoproteins or heterogenous nuclear ribonucleoproteins (hnRNPs), involved in post-transcriptional RNA processing. The RRMs of AUF1 show significantly greater homology to the RRMs of a Xenopus RNA-binding protein, Nrp-1, than to human hnRNP A, B, or C RRMs (5), and when the RRMs in AUF1 are compared with these hnRNP proteins, there is very little homology outside the more highly conserved RNP-1 and -2 motifs.
Although AUF1 apparently forms dimers in solution in the absence of target RNA, our experiments suggest that AUF1 binds the c-fos ARE as a hexamer, based upon its hydrodynamic properties. While our experiments do not yet allow us to establish the precise spatial orientations of the RRMs within the hexamer in relation to the ARE, they do allow us to draw some general conclusions. For example, the hexamer has a large Stokes radius (6.4 nm), which might permit contact of the protein to an RNA fragment of 120 nucleotides (14,38). Based upon the size range of known AREs (approximately 9 -150 nucleotides; reviewed in Ref. 39), AUF1 might be able to contact most AREs from one end of the RNA sequence to the other. Additionally, while we do not yet know whether each polypeptide in the complex contacts RNA, it is possible that such a large protein complex provides a packaging function. For example, the hnRNP C proteins, which form tetramers in solu-tion, also bind RNA as tetramers. These package 700-nucleotide increments of heterogenous nuclear RNA into triangular complexes (22). The hnRNP (A1) 3 /B2 tetramers and (A2) 3 /B1 tetramers also package heterogenous nuclear RNA in a fashion similar to the hnRNP C proteins (40). Alternatively, since AUF1 forms complexes with a number of intracellular proteins (5,41), AUF1 bound to an ARE as a hexameric protein complex might provide a large surface for interacting, effector proteins to bind. This function, however, need not be exclusive of a putative packaging role for AUF1.
In summary, our experiments indicate that the RRMs in AUF1 are not modular in function and that the ability of AUF1 to dimerize via its alanine-rich N-terminal domain is important for high affinity ARE binding. Furthermore, they suggest that AUF1 binds to its RNA target as a hexameric complex. Future experiments will examine how these protein-protein interactions work in concert with the RRMs to permit ARE recognition by AUF1.