Thermodynamics and Kinetics of Hsp70 Association with A (cid:1) U-rich mRNA-destabilizing Sequences*

Rapid mRNA degradation directed by A (cid:1) U-rich elements (AREs) is mediated by the interaction of specific RNA-binding proteins to these sequences. The protein chaperone Hsp70 has been identified in a cellular complex containing the ARE-binding protein AUF1 and has also been detected in direct contact with A (cid:1) U-rich RNA substrates, indicating that Hsp70 may be involved in the regulation of ARE-directed mRNA turnover. By using gel mobility shift and fluorescence anisotropy assays, we have determined that Hsp70 directly and specifically associates with U-rich RNA substrates in solution. With the ARE from tumor necrosis factor (cid:2) (TNF (cid:2) ) mRNA, Hsp70 forms a dynamic complex consistent with a 1:1 association of protein:RNA but demonstrates cooperative binding behavior on polyuridylate substrates. Un-like AUF1, the RNA binding activity of Hsp70 is not regulated by ion-dependent folding of the TNF (cid:2) ARE, suggesting that AUF1 and Hsp70 recognize distinct binding determinants on this RNA substrate. Binding of Hsp70 to the TNF (cid:2) ARE is driven entirely by enthalpy at physiological temperatures, indicating that burial of hydrophobic surfaces is likely the principal mechanism stabilizing the Hsp70 (cid:1) RNA complex. Potential roles for the interaction of Hsp70 with ARE-containing mRNAs in the regulation of mRNA turnover and/or translational efficiency are discussed. The

The rate of cytoplasmic mRNA decay constitutes an essential component of regulated gene expression, influencing both the timing and level of expression for many gene products (reviewed in Refs. 1 and 2). In mammals, A ϩ U-rich elements (AREs) 1 direct the rapid turnover of many labile mRNAs, including several encoding oncoproteins, cytokines, inflammatory mediators, and G protein-coupled receptors (reviewed in Ref. 3). AREs comprise a diverse family of cis-acting sequences localized to the 3Ј-untranslated regions (3Ј-UTRs) of these transcripts and promote cytoplasmic degradation of mRNA sub-strates characterized by 3Ј 3 5Ј-exonucleolytic shortening of the poly(A) tail, followed by rapid digestion of the mRNA body (3)(4)(5). Rapid mRNA decay mediated by AREs is essential for maintaining low constitutive expression levels of many gene products. For example, deletion of the ARE from tumor necrosis factor ␣ (TNF␣) mRNA results in TNF␣ overexpression in a transgenic mouse model, concomitant with the development of a systemic inflammatory syndrome in these animals (6). Removing the ARE from the 3Ј-UTR of the c-fos proto-oncogene stabilizes the c-fos mRNA (7) and significantly increases its transforming potential in cultured cells (8,9).
During heat shock, mRNAs containing AREs are stabilized (10). One feature of the cellular heat shock response is an increase in the abundance of several members of the heat shock family of proteins, including the highly conserved protein chaperone Hsp70 (11,12). Several observations indicate that Hsp70 may participate in the regulation of ARE-directed mRNA turnover. First, Hsp70 is present in a cytoplasmic complex containing the ARE-binding factor AUF1 (10). Association of AUF1 with an ARE promotes rapid mRNA decay (13)(14)(15)(16), involving dynamic AUF1 oligomerization on the ARE (17)(18)(19) and recruitment or assembly of a multisubunit, trans-acting complex (10,20). Decreases in the ARE binding activity (21) or abundance (22,23) of AUF1 are associated with stabilization of mRNAs containing AREs, whereas induction of cellular AUF1 levels correlates with accelerated turnover of these transcripts (24,25). Second, the association of Hsp70 with AUF1-containing complexes increases concomitantly with the induction of cellular Hsp70 levels during heat shock and coincident with stabilization of ARE-containing mRNAs (10). Finally, Henics et al. (26) demonstrated that cross-links could be generated between Hsp70 and A ϩ U-rich RNA substrates in an RNA sequence-dependent manner, indicating that Hsp70 may contact AREs directly.
The protein chaperone activity of Hsp70 has been extensively investigated, whereby Hsp70 transiently associates with localized hydrophobic domains of polypeptides to promote correct protein folding and inhibit aggregation (reviewed in Refs. 27 and 28). Interaction of Hsp70 with peptide substrates is essential for cellular protection against apoptosis (29,30), protein transport across membranes (31), and targeted protein degradation (32) and contributes to the function of selected regulatory proteins in transcription (33,34) and translation (35). Recently, the identification of a specific binding site for sulfogalactolipids within the ATP-binding domain of Hsp70 (36) indicated that this protein also performs functions beyond intracellular peptide folding and trafficking.
To date, no direct role has been demonstrated for Hsp70 in mRNA catabolism. However, the identification of Hsp70 in cytoplasmic complexes containing AUF1, coupled with evidence of a direct interaction between Hsp70 and A ϩ U-rich mRNA-destabilizing sequences, prompts the possibility that Hsp70 may function as a trans-acting regulator of AREdirected mRNA turnover or as an ancillary factor for AUF1. Assignment of a definitive role for Hsp70 in this system will require a thorough understanding of the biochemical interactions between Hsp70, the RNA substrate, and other components of the AUF1-containing complex. Toward this end, we have characterized the thermodynamics of Hsp70 interactions with model RNA substrates, using assays of fluorescence anisotropy to evaluate binding parameters in solution under true equilibrium or approach to steady-state conditions. In addition, this study examined the binding specificity and dynamics of Hsp70 interactions with RNA substrates and provided insight into the physical forces driving these binding reactions.

EXPERIMENTAL PROCEDURES
RNA Substrates-RNA substrates encoding the core ARE from TNF␣ mRNA (TNF␣ ARE), a polyuridylate sequence (U 32 ), or a fragment of the rabbit ␤-globin coding region (R␤) were synthesized by Dharmacon Research (Boulder, CO). The nucleotide sequence of each RNA substrate is listed in Table I. Replicates of each RNA oligonucleotide were synthesized containing a fluorescein (Fl) group conjugated at the 5Ј-end and are designated by the prefix "Fl-". All RNA substrates were 2Ј-Odeprotected and quantified by A 260 as described previously (18,19). For gel mobility shift assays, 5Ј-OH RNA substrates were radiolabeled using T4 polynucleotide kinase and [␥-32 P]ATP to specific activities of 3-5 ϫ 10 3 cpm/fmol as described (18).
Preparation of Recombinant Hsp70 -Standard recombinant DNA procedures were employed for all plasmid DNA constructions (37). The fidelity of each plasmid construct was verified by restriction mapping and automated DNA sequencing. The coding sequence of human hsp70 cDNA was excised from the plasmid pH2.3 (American Type Culture Collection, Manassas, VA) by digestion with BamHI ϩ HindIII and was subcloned into the corresponding restriction sites in p(T3/T7)␣19 (Life Technologies, Inc.) to generate the plasmid p␣19-hsp70. A fragment containing the 5Ј-390 base pairs of the coding sequence but lacking the initiator AUG was then amplified from pH2.3 by polymerase chain reaction using Pfu DNA polymerase (Stratagene, La Jolla, CA) and primers 5Ј-CGTCGGATCCTGCCAAAGCCGCGGCAGTC-3Ј and 5Ј-GGCGATCTCCTTCATCTTGG-3Ј. This fragment was then digested with NcoI ϩ BamHI and subcloned into similarly digested p␣19-hsp70 to generate the plasmid p␣19-hsp70(ϪATG) and served to position a BamHI restriction site immediately upstream of codon 2 of the hsp70 cDNA. This modified hsp70 cDNA was then excised by digestion with BamHI ϩ HindIII and was subcloned into pBAD/HisC (Invitrogen, Carlsbad, CA) digested with BglII ϩ HindIII to generate the plasmid pBAD/HisC-hsp70.
Recombinant His 6 -Hsp70 was purified under non-denaturing conditions by Ni 2ϩ -affinity chromatography from lysates of Escherichia coli TOP10 cells transformed with pBAD/HisC-hsp70 following arabinose induction essentially as described (19). Desalted protein preparations were quantified by comparison of Coomassie Blue-stained SDS-polyacrylamide gels containing His 6 -Hsp70 and a titration of bovine serum albumin as described (38). Where indicated, the N-terminal fragment of His 6 -Hsp70 containing the His 6 tag was removed using the Enterokinase Cleavage Capture Kit (Novagen, Madison, WI). Immunodetection of recombinant Hsp70 was performed by Western blotting using goat anti-human Hsp70 (Santa Cruz Biotechnology, Santa Cruz, CA).
RNA Binding Assays-The RNA binding activity of His 6 -Hsp70 was qualitatively assessed using gel mobility shift assays as described (18). For quantitative analyses, RNA binding activity was monitored by changes in the fluorescence anisotropy of fluorescent RNA substrates across a titration of protein concentrations essentially as described (18,19). Fluorescence anisotropy was measured using the Beacon 2000 Variable Temperature Fluorescence Polarization System (Panvera, Madison, WI). Equilibrium binding experiments were performed with a range of recombinant His 6 -Hsp70 concentrations and 0.15 nM fluorescent RNA in a final volume of 100 l containing 10 mM Tris⅐HCl (pH 8.0), 100 mM KCl, 2 mM dithiothreitol, and 0.1 g/l acetylated bovine serum albumin. Heparin (0.2 g/l) was required to inhibit nonspecific association of recombinant proteins with RNA substrates (19). MgCl 2 was used as a source of magnesium ions where necessary. Reactions lacking Mg 2ϩ contained 0.5 mM EDTA. Samples were read as blank prior to addition of fluorescent RNA substrates. After probe addition, samples were incubated for 90 s before anisotropy was measured. Preliminary on-rate experiments demonstrated that anisotropic equilibrium was reached within 15-30 s under these conditions (data not shown). Each data point represents the mean of 10 anisotropy measurements for each binding reaction.
Anisotropy of Fl-conjugated RNA substrates was measured using ex ϭ 490 and em ϭ 535 nm. Association of His 6 -Hsp70 with Fl-TNF␣ ARE and Fl-U 32 did not significantly alter the fluorescence quantum yields of these RNA substrates (data not shown). Accordingly, the total measured anisotropy (A t ) of each binding reaction was interpreted as a function of the intrinsic anisotropy (A i ) and fractional concentration (f i ) of each fluorescent species as described in Equation 1 (39 -41).
The association of His 6 -Hsp70 with these RNA substrates was considered in terms of a general binding scheme xP ϩ R º P x R, in which a single RNA molecule may associate with x protein molecules.
where K represents the equilibrium constant; x is the Hill coefficient, and A R and A PR represent the intrinsic anisotropy values of the free and protein-associated Fl-labeled RNA substrates, respectively. Although this algorithm cannot discriminate individual binding steps in oligomerization mechanisms (18), it will detect significant deviation from a single-site binding model (i.e. x ϭ 1). A R was measured as the anisotropy of each RNA substrate in the absence of Hsp70 (n Ն 3), and all other constants were solved by nonlinear least squares regression using PRISM version 2.0 (GraphPad, San Diego, CA). For off-rate experiments, binding reactions containing His 6 -Hsp70 and Fl-labeled RNA substrates were assembled as described above, with anisotropy measured at equilibrium (t ϭ 0). Following addition of a 5000-fold excess of unlabeled RNA, anisotropy was measured in intervals of 15 s, with five measurements taken for each time point.

RESULTS
Specific Association of Hsp70 with U-rich RNA Substrates-To biochemically characterize the interaction of Hsp70 with RNA substrates, it was first necessary to generate large quantities of the protein. To this end, Hsp70 was expressed in E. coli TOP10 cells as an N-terminal His 6 fusion protein and was purified by Ni 2ϩ -affinity chromatography to Ͼ95% purity as described under "Experimental Procedures." By SDS-polyacrylamide gel electrophoresis, the His 6 -Hsp70 fusion polypeptide resolved to a single band migrating at ϳ70 kDa (Fig. 1A), which was immunoreactive with anti-Hsp70 antibodies (Fig. 1B).
Gel mobility shift assays were performed to evaluate the ability of recombinant His 6 -Hsp70 to associate with model RNA substrates. By using a 38-nucleotide RNA substrate containing the core ARE from the 3Ј-UTR of TNF␣ mRNA, increasing concentrations of His 6 -Hsp70 generated a single RNA⅐protein complex (Fig. 2, left panel). This is in marked contrast to the Hsp70 Association with A ϩ U-rich Elements association of AUF1, which forms a tetrameric protein structure by sequential binding of two protein dimers to this RNA substrate (18,19). A complex of similar mobility was also observed with the U 32 substrate (Fig. 2, center panel), although the increased breadth of this band relative to the TNF␣ ARE⅐Hsp70 complex suggests that additional RNA-protein binding events may occur on the polyuridylate sequence. Hsp70 did not associate with a fragment of the ␤-globin coding region (Fig. 2, right panel), indicating that the RNA binding activity of this protein shows some specificity for U-rich RNA substrates.
To quantitatively assess the binding activity of His 6 -Hsp70 for these RNA substrates, fluorescence anisotropy assays were performed employing Fl-conjugated versions of each substrate. Regression of total measured anisotropy (A t ) versus the concentration of His 6 -Hsp70 protein [P] for each data set allowed the association binding constant (K) and the Hill coefficient (x) to be calculated using Equation 2. Changes in the fluorescence anisotropy of the Fl-TNF␣ ARE substrate due to association of His 6 -Hsp70 were well described by this binding algorithm (Fig.  3A, top panel), with residuals randomly distributed about the regression solution (Fig. 3A, bottom panel). The Hill coefficient for this interaction resolved to x ϭ 1.1 Ϯ 0.1 (n ϭ 3), indicating that the association of Hsp70 with the ARE substrate did not significantly deviate from a single-site binding model. Whereas solution of x ϭ 1 may also be indicative of multiple, noncooperative protein binding sites on this RNA substrate, the identification of a single RNA⅐protein complex by gel mobility shift assay (Fig. 2, left panel) indicates that the association of Hsp70 with the TNF␣ ARE likely represents a 1:1 interaction of protein:RNA. The association binding constant for this interaction resolved to K ϭ 3.6 Ϯ 0.4 ϫ 10 7 M Ϫ1 , corresponding to a dissociation binding constant (K d ϭ 1/K) of 28 nM. To determine whether the His 6 tag might influence the RNA binding activity of His 6 -Hsp70, similar experiments were performed using either enterokinase- (Fig. 1A) or mock-digested His 6 -Hsp70 proteins. For both protein samples, calculated values of K and x did not differ significantly from those describing the association of undigested His 6 -Hsp70 with the Fl-TNF␣ ARE substrate (data not shown), indicating that the presence of the N-terminal His 6 tag does not affect the ARE binding activity of Hsp70.
To the Fl-U 32 substrate, His 6 -Hsp70 demonstrated cooperative binding activity (Fig. 3B), with x ϭ 1.68 Ϯ 0.09 (n ϭ 2). Solution of x Ͼ 1 indicates that multiple protein binding events may occur on this RNA substrate, and that initial protein binding events facilitate association of subsequent protein molecules. Furthermore, association of His 6 -Hsp70 with Fl-U 32 indicates that U-rich RNA sequences represent high affinity targets for Hsp70 binding. Addition of His 6 -Hsp70 to binding reactions containing the Fl-R␤ substrate did not significantly increase the fluorescence anisotropy of the RNA (Fig. 3C), indicating that Hsp70 does not associate with this substrate. This is consistent with the lack of R␤ binding activity, detected by the gel mobility shift experiments (Fig. 2, right panel), and indicates that Hsp70 does not bind indiscriminately to RNA targets.
Association of AUF1 to RNA substrates containing the TNF␣ ARE sequence is significantly inhibited by Mg 2ϩ or other multivalent cations (19), because of the cation-dependent stabilization of a spatially condensed, higher order RNA structure within this substrate (42). Binding of His 6 -Hsp70 to this substrate, however, was independent of ion-induced changes in the structure of the ARE (Table II), indicating that some differences exist between the mechanisms of U-rich RNA recognition by AUF1 versus Hsp70. By contrast, the Fl-U 32 substrate does not exhibit significant structural changes in the presence of Mg 2ϩ (42). Like AUF1 (19), the association of His 6 -Hsp70 with Fl-U 32 was unaffected by the presence of Mg 2ϩ (Table II). Because Hsp70 binds with high affinity to the TNF␣ ARE (Fig.  2, left panel; Fig. 3A) and U 32 RNA substrates (Fig. 2, center  panel; Fig. 3B) but not R␤ (Fig. 2, right panel; Fig. 3C), we conclude that Hsp70 functions as a sequence-specific RNAbinding protein and that U-rich RNA sequences are sufficient for protein binding. In addition, the observation that Hsp70 binds with high affinity to U-rich RNA substrates lacking any Hsp70 Association with A ϩ U-rich Elements significant higher order structure (Table II: Fl-TNF␣ ARE, ϪMg 2ϩ ; Fl-U 32 , ϮMg 2ϩ ) indicates that Hsp70 likely makes direct contact with the uracil bases of these substrates.
Kinetics of Hsp70 Binding to the TNF␣ ARE-To evaluate the dynamics of the interaction between Hsp70 and an ARE, off-rate analyses were performed to measure the stability of Hsp70⅐Fl-TNF␣ ARE complexes in solution. Following incubation of 20 nM His 6 -Hsp70 with 0.15 nM Fl-TNF␣ ARE, dissociation of the fluorescent RNA substrate was monitored by the time-dependent decrease in anisotropy observed following addition of a 5000-fold excess of unlabeled TNF␣ ARE or R␤ RNA. The unlabeled R␤ substrate was unable to compete for His 6 -Hsp70 binding to Fl-TNF␣ ARE (Fig. 4, open circles), whereas the Hsp70⅐Fl-TNF␣ ARE complex was rapidly dissociated following addition of the unlabeled TNF␣ ARE RNA (Fig. 4, solid circles). The rate of complex dissociation was well described by single phase, first-order decay (Fig. 4, solid line). Simultaneous regression of triplicate independent experiments yielded a dissociation rate constant (k Ϫ1 ) of 6 Ϯ 1 ϫ 10 Ϫ2 s Ϫ1 (mean Ϯ 95% confidence interval), corresponding to a complex half-life of ϳ12 s in solution. By using the equilibrium association constant K ϭ 3.6 Ϯ 0.4 ϫ 10 7 M Ϫ1 for binding of His 6 -Hsp70 to Fl-TNF␣ ARE (Table II) and assuming a 1:1 binding stoichiometry of protein:RNA, the corresponding second-order association rate constant (k 1 ) may thus be estimated from K ϭ (k 1 /k Ϫ1 ) as 2 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 . The rapid kinetics of the Hsp70-ARE interaction indicates that this binding event is very dynamic in nature, similar to the ARE binding and RNA-dependent protein oligomerization activities of AUF1 (18). Regulatory implications of highly dynamic equilibria involving these AREbinding factors and their cognate RNA sequences are addressed under "Discussion." Thermodynamics of Hsp70 Binding to the TNF␣ ARE-The stability of the Hsp70⅐TNF␣ ARE interaction is reflected in the free energy of binding (⌬G 0 ), readily extractable from the equilibrium association constant (K) using Equation 3, where R is the gas constant (1.987 ϫ 10 Ϫ3 kcal⅐mol Ϫ1 ⅐K Ϫ1 ), and T is the absolute temperature of the binding reaction. However, to understand the forces driving the interaction between Hsp70 and an ARE substrate, the contributions of enthalpy and entropy to this binding event were determined by measurement of association binding constants between His 6 -Hsp70 and Fl-TNF␣ ARE at various temperatures. Under conditions where the change in enthalpy (⌬H 0 ) is independent of temperature, a van't Hoff plot of ln(K) versus 1/T yields a linear relationship (43,44). Based on the solutions of K for the interaction of His 6 -Hsp70 with Fl-TNF␣ ARE as temperature was varied (Fig. 5A, solid circles), it was clear that ln(K) exhibited a nonlinear relationship with respect to 1/T, indicating that ⌬H 0 was not constant. Accordingly, the contributions of enthalpy and entropy at different temperatures were resolved by calculation of the change in heat capacity of the system (⌬C P,obs 0 ) and the characteristic temperatures at which enthalpy and entropy contribute no energy to the system, given by T H and T S , respectively, as defined in Equation 4 (44).  Whereas the total free energy of Hsp70⅐TNF␣ ARE complex formation changed only slightly as temperature was increased (Fig. 5B, open circles), the enthalpic (Fig. 5B, solid line) and entropic (Fig. 5B, dashed line) contributions to free energy varied strongly with temperature. For example, between 20 (293 K) and 24°C (297 K), binding of Hsp70 to the TNF␣ ARE is associated with favorable changes in both enthalpy (⌬H 0 Ͻ 0) and entropy (⌬S 0 Ͼ 0). However, above 24°C, Hsp70⅐ARE complex formation is driven entirely by large negative changes in ⌬H 0 , which compensate for unfavorable changes in entropy above this temperature. At 37°C, these functions resolve to ⌬H 0 ϭ Ϫ39 kcal⅐mol Ϫ1 and ⌬S 0 ϭ Ϫ99 entropy units (1 entropy unit ϭ 1 cal⅐mol Ϫ1 ⅐K Ϫ1 ). For site-specific, noncovalent interactions between nucleic acids and proteins, negative changes in molar heat capacity are largely attributable to significant burial of hydrophobic surfaces in the protein and/or nucleic acid components upon binding (44). However, hydrophobic interactions also generally contribute favorable changes in entropy, due to the expulsion of ordered water molecules from the binding surfaces (45). As such, the unfavorable changes in ⌬S 0 upon Hsp70⅐TNF␣ ARE complex formation may reflect the associa-  The equilibrium constant (K) and Hill coefficient (x) were calculated using Equation 2 and are listed as the mean Ϯ spread for n ϭ 2 or mean Ϯ nϪ1 for n ϭ 4 independent experiments, each consisting of a minimum of 15 data points.
Hsp70 Association with A ϩ U-rich Elements tion of ions, ordering of additional water molecules by hydrogen bonding, or structural changes in the protein and/or RNA that restrict their flexibility (46). Based on the large negative changes in ⌬H 0 upon Hsp70⅐ARE complex formation at physiological temperatures, we conclude that hydrophobic interactions, likely including burial of uracil bases on the RNA substrate, provide the major energetic contribution to the association of Hsp70 with the TNF␣ ARE. However, the significant entropic penalty associated with Hsp70 binding to the ARE substrate at these temperatures indicates that structural changes in one or both binding partners occur concomitantly with the binding event, which either restrict molecular motion or significantly augment ionic and/or hydration shells enveloping the complex.
Sensitivity of Hsp70-ARE Interaction to Ionic Strength-As an anionic polymer, RNA substrates present abundant opportunities for electrostatic interactions between their phosphodiester backbones and positively charged amino acid side chains on binding proteins (47,48). To determine whether ionic interactions contribute to the binding of Hsp70 to an ARE substrate, the affinity of His 6 -Hsp70 for Fl-TNF␣ ARE was measured across a range of KCl concentrations. Association binding constants (K) were determined at each concentration of KCl and then plotted as logK versus Ϫlog[K ϩ ]. The slope of the data set (ϪѨlogK/Ѩlog[K ϩ ]) is influenced by the number of electrostatic interactions contributing to the stability of the RNA⅐protein complex (49). In cases where ionic interactions play minor roles in RNA⅐protein complex stability, such as the association of ribosomal protein L11 (49), the trp RNA-binding attenuation protein (50), and RNase A (48) with RNA substrates, the value of ϪѨlogK/Ѩlog[ion ϩ ] is typically between 0 and 2. Some other RNA⅐protein complexes involve larger numbers of ionic interactions, displaying values of ϪѨlogK/Ѩlog[ion ϩ ] Ͼ3. Examples of these include the interaction of U1A with the U1 hairpin II RNA substrate (51) and the binding of ribosomal protein S8 to its rRNA recognition site (49). For the interaction of His 6 -Hsp70 with Fl-TNF␣ ARE, ϪѨlogK/Ѩlog[K ϩ ] resolved to ϳ1.4 (Fig. 6) across a range of 100 -300 mM KCl. Binding constants were difficult to confidently resolve for reactions containing Ͼ300 mM KCl, because of the high concentrations of protein required to approach saturated binding. Based on the sensitivity of the Hsp70⅐TNF␣ ARE binding equilibrium to monovalent ion concentration, we conclude that some electrostatic interactions are involved in the association of Hsp70 with the ARE substrate, although they make relatively minor contributions to the overall stability of the Hsp70⅐ARE complex.

DISCUSSION
The results presented here demonstrate that Hsp70 can form high affinity complexes with RNA sequences involved in regulating rapid cytoplasmic mRNA turnover. On the TNF␣ ARE, a single protein⅐RNA complex was formed (Fig. 2, left panel), which was well described by a bimolecular association function (Fig. 3A). Binding to the ARE was highly dynamic (Fig. 4, solid circles), which may contribute to the efficient recognition of specific ARE targets in complex RNA populations. Hsp70 bound cooperatively to a polyuridylate substrate (Fig. 3B), indicating that U-rich RNA sequences were sufficient for recognition by this protein and suggesting that larger Hsp70 complexes may be assembled on some target RNAs. However, the observation that Hsp70 did not associate with the R␤ substrate (Fig. 2, right panel; Fig. 3C) indicates that the protein does not bind RNA indiscriminately but rather shows some selectivity for U-rich sequences.
Proteins of the Hsp70 family are composed of an N-terminal ATP/ADP-binding domain, responsible for the ATPase activity of the protein, an internal peptide-binding domain, and a Cterminal domain of unknown function (27). Currently, the region of Hsp70 that directly contacts RNA substrates is unknown. The thermodynamic analyses described in this work, however, permit some interesting comparisons between the peptide and RNA binding activities of this protein. The favorable change in enthalpy upon protein binding at physiological temperatures (Fig. 5B, solid line) likely reflects burial of the hydrophobic RNA bases (43,44), analogous to the mechanism of peptide binding by Hsp70. The peptide-binding domain of the bacterial Hsp70 homologue, DnaK, contains two sheets of four antiparallel ␤-strands each, which are stacked to form a ␤-sandwich (52). This region of the Hsp70 molecule is highly conserved, with 60% sequence identity between bacterial and mammalian homologues. Hydrophobic residues from both ␤-sheets contribute to binding of hydrophobic amino acid side chains, and substrates are maintained in an elongated conformation (52). These features evoke some appealing parallels with sequence-specific RNA-binding mediated by RNA-recognition motifs (RRMs), in which a four-stranded anti-parallel ␤-sheet centrally positions two highly conserved sequences rich in hydrophobic amino acid residues (53). Association of RNA with this domain includes stacking interactions between these hydrophobic side chains and the RNA bases (51,54). Furthermore, extended RNA substrate conformations are promoted by RRMs in some cases, like the RNA-binding domain of the splicing regulator sex-lethal (55) or the two N-terminal RRMs from poly(A)-binding protein (56).
Other data presented in this work, however, indicate significant differences between the peptide versus RNA binding activities of Hsp70. First, based on the crystal structure of the Hsp70 peptide-binding domain, the peptide-binding pocket is flanked by acidic amino acid residues, suggesting that cationic substrates would be preferred at this site (52). Peptide substrates flanked by acidic amino acid residues show significantly poorer binding affinity than those flanked by basic residues (57). By contrast, RNA substrates are inherently anionic in solution, and electrostatic interactions play only a minor role in the stability of the Hsp70⅐TNF␣ ARE complex (Fig. 6). Second, the binding affinities of RNA substrates for Hsp70 (K d ϭ 28 nM for TNF␣ ARE) are comparable or better than those calculated for binding of peptide substrates to Hsp70 (K d Ն 60 nM) (57), despite the lack of positive charges on the RNA molecules. Taken together, these observations admit the possibility that a site on Hsp70 distinct from the peptide-binding domain may be responsible for contacting RNA substrates. Experiments aimed at delimiting the region(s) of Hsp70 essential for RNA binding activity and the involvement of the ATPase cycle of this protein will be necessary to further understand the structural events contributing to RNA recognition by this protein.
Although the biological significance of the Hsp70⅐ARE interaction remains unknown, the RNA-binding properties of Hsp70, together with observations of Hsp70 induction (11,12) concomitant with stabilization of ARE-containing transcripts during heat shock (10), support several functional possibilities. First, Hsp70 may directly compete with other ARE-binding proteins for these RNA substrates, thus influencing the ability of these factors to induce or inhibit the mRNA decay process. For example, association of some ARE-binding proteins, such as AUF1 (14,16,21,23) and tristetraprolin (58,59), promote rapid turnover of ARE-containing mRNAs. Others, including HuR, are associated with inhibition of the ARE-directed mRNA decay pathway (60 -62). Considering that the cellular levels of Hsp70 are enhanced during heat shock (11,12), and that the association of both Hsp70 (Fig. 3) and AUF1 (18) is highly dynamic in solution, it is possible that the competitive equilibrium between these proteins for ARE substrates is rapidly shifted in favor of Hsp70 under heat shock conditions and contributes to stabilization of these transcripts at elevated temperatures (10). Alternatively, competition between Hsp70 and ARE-binding regulators of mRNA translation, such as TIAR (63) and TIA-1 (64), may serve to modulate the efficiency of translation during heat shock, because most mRNAs are translationally silent under these conditions (reviewed in Ref. 11). A second possibility is that association of Hsp70 may influence the accessibility of RNA substrates for other AREbinding proteins by remodeling local RNA structures, analogous to its protein chaperone activity. For example, ion-stabilized folding of the TNF␣ ARE inhibits association of AUF1 with this RNA substrate (19,42). However, binding of Hsp70 to the ARE is independent of the folded state of the RNA (Table  II). Within the context of cellular mRNAs, AREs may extend up to 150 nucleotides in length (3). Accordingly, structural changes in an ARE resulting from Hsp70 binding at one site within the element might serve to influence the kinetics or thermodynamics of protein binding at proximal sites. Based on co-immunoprecipitation experiments, a physical relationship has been detected between AUF1 and Hsp70 (10), raising the possibility that these proteins may co-associate with a subset of AREs. Finally, association of Hsp70 with ARE-containing transcripts may function to re-localize cytoplasmic mRNAs to stress granules during heat shock. Evidence supporting this hypothesis is given by the identification of Hsp70 in the stress granules (65), and the recruitment of mRNAs to these structures during the heat shock response in higher eukaryotes (66) concomitant with shuttling of specific ARE-binding proteins (64). Considered together, some or all of these mechanisms may participate in the stabilization and translational silencing of specific mRNAs, which are characteristic of the cellular heat shock response (reviewed in Ref. 11).