Hsp70's RNA-binding and mRNA-stabilizing activities are independent of its protein chaperone functions

Hsp70 is a protein chaperone that prevents protein aggregation and aids protein folding by binding to hydrophobic peptide domains through a reversible mechanism directed by an ATPase cycle. However, Hsp70 also binds U-rich RNA including some AU-rich elements (AREs) that regulate the decay kinetics of select mRNAs and has recently been shown to bind and stabilize some ARE-containing transcripts in cells. Previous studies indicated that both the ATP- and peptide-binding domains of Hsp70 contributed to the stability of Hsp70–RNA complexes and that ATP might inhibit RNA recruitment. This suggested the possibility that RNA binding by Hsp70 might mimic features of its peptide-directed chaperone activities. Here, using purified, cofactor-free preparations of recombinant human Hsp70 and quantitative biochemical approaches, we found that high-affinity RNA binding requires at least 30 nucleotides of RNA sequence but is independent of Hsp70's nucleotide-bound status, ATPase activity, or peptide-binding roles. Furthermore, although both the ATP- and peptide-binding domains of Hsp70 could form complexes with an ARE sequence from VEGFA mRNA in vitro, only the peptide-binding domain could recover cellular VEGFA mRNA in ribonucleoprotein immunoprecipitations. Finally, Hsp70-directed stabilization of VEGFA mRNA in cells was mediated exclusively by the protein's peptide-binding domain. Together, these findings indicate that the RNA-binding and mRNA-stabilizing functions of Hsp70 are independent of its protein chaperone cycle but also provide potential mechanical explanations for several well-established and recently discovered cytoprotective and RNA-based Hsp70 functions.

Tight regulation of gene expression ensures that terminal gene products (protein or RNA) are maintained within levels appropriate for cell function. A critical determinant governing the rate of protein synthesis is the concentration of its encoding mRNA, which depends on both its synthetic and turnover rates. mRNA synthesis is a cumulative function of transcription, pre-mRNA processing, and nucleocytoplasmic transport, with each subject to regulatory control. Cytoplasmic mRNA turnover is also tightly regulated. mRNA decay rates can span 2 orders of magnitude and control gene expression in two ways. First, unstable mRNAs approach new steady states more quickly than stable mRNAs after changes in the synthetic rate (1,2), decreasing the response time between a transcriptional stimulus and phenotypic output. Second, mRNA decay rates can vary in response to diverse stimuli, modulating protein production without altering transcription (3,4). Both constitutive and inducible mRNA turnover are mediated by cis-acting sequences within individual mRNAs.
Among the best-known sequences that regulate mRNA decay in mammalian cells are the AU-rich elements (AREs) 3 located within the 3Ј-untranslated regions (UTRs) of many mRNAs that encode critical regulators of cell growth, differentiation, signaling, and survival (3,5,6). Approximately 8% of all human mRNAs contain AREs (7), which can modulate rates of mRNA decay and translation by recruiting cellular factors collectively known as ARE-binding proteins (ARE-BPs; for review, see Refs. 4 and 8) and possibly select microRNAs (9,10). Whereas Ͼ20 distinct ARE-BPs have been identified (11,12), functional significance has been defined for only a few. For example, tristetraprolin and KH-type splicing regulatory protein (KSRP) generally accelerate degradation of mRNA substrates, whereas HuR is well known for stabilizing targeted transcripts (5,6,(13)(14)(15). AUF1 generally accelerates mRNA decay but can stabilize some targets or enhance their translational efficiency (2, 16 -18), yet other ARE-BPs like TIA-1 and TIAR inhibit translation but do not appear to regulate mRNA decay kinetics directly (19,20).
A more recently identified ARE-BP is the major inducible heat shock protein Hsp70, best known for its role as a protein chaperone that binds exposed hydrophobic polypeptide domains to inhibit aggregation and promote productive protein folding (for review, see Refs. 21 and 22). Binding and release of peptide ligands are controlled through an ATPase cycle and can be regulated by associated co-chaperone proteins (23). In 1999, Hsp70 was also identified as a component of a multisubunit, trans-acting complex recruited to AREs by AUF1 (24). However, subsequent work showed that Hsp70 could bind directly to AREs and similar U-rich RNA substrates with lowto mid-nanomolar dissociation constants despite lacking any canonical RNA-binding domains (25,26). Further biochemical studies suggested that both the N-terminal ATP-binding and C-terminal peptide-binding domains contributed to the stability of Hsp70 ribonucleoprotein (RNP) complexes (27,28). Finally, Hsp70 stabilized select ARE-containing mRNA targets, including vascular endothelial growth factor A (VEGFA) and cyclooxygenase-2 mRNAs, in cultured cell models (28).
The involvement of both the ATP-and peptide-binding domains of Hsp70 in RNA substrate recognition suggested the possibility that interdomain allostery might direct the RNAtargeted functions of Hsp70, similar to such relationships established for peptide ligands (29). This model was also supported by observations that ATP inhibited assembly of Hsp70 -RNA complexes detected after UV cross-linking (25), mirroring the weakened peptide-binding activity of ATPversus ADPbound Hsp70 (21). Conversely, a key difference between Hsp70 functions on protein versus RNA ligands was also noted in a prior study. The compound 2-phenylethynesulfonamide binds Hsp70 and disrupts interactions with co-chaperones and peptide substrates (30). In cisplatin-treated HeLa cells, 2-phenylethynesulfonamide inhibited Hsp70-binding to the tumor suppressor p53 but had no effect on Hsp70's ability to bind RNA in vitro or stabilize an mRNA target in cells (28). Together, these studies typify the current confusion regarding the mechanism(s) responsible for RNA-directed functions of Hsp70 versus the allosteric paradigm defined by its protein chaperone activity.
In this study we have addressed this problem by interrogating the role(s) of the Hsp70 chaperone cycle in its RNA-binding and mRNA-stabilizing functions from multiple perspectives. Biochemical approaches were used to identify the RNA substrate length requirements for Hsp70-binding and to test relationships between its RNA-binding activity and its well-defined nucleotide-binding, ATPase, and peptide-binding functions. Finally, using VEGFA mRNA as a model, we identify the Hsp70 peptide-binding domain as sufficient to bind and stabilize an mRNA target in cells.

Hsp70 bound with the highest affinity to ARE-containing RNA substrates >30 nucleotides in length
To assess interrelationships between the RNA-binding and protein chaperone activities of Hsp70, it was first necessary to select a model high-affinity RNA substrate. As a relative newcomer to the ARE-BP family, far less is known about the RNAbinding properties of Hsp70 than better-characterized ARE-BPs. Previous competition assays using recombinant Hsp70 demonstrated a preference for U-rich RNA targets (28) but did not resolve the RNA substrate length required to form stable Hsp70 RNPs. This parameter can vary widely among different ARE-BPs and strongly influences their substrate preferences. For example, both tristetraprolin and HuR form stable com-plexes with short (9 -13 nt) AU-rich motifs, although HuR cooperatively assembles into multimers on extended (Ն18 nt) target sites (31,32). By contrast, AUF1 proteins require Ͼ30 nucleotides of RNA sequence to form high-affinity RNPs (33).
To resolve the RNA substrate length-dependence for Hsp70, binding assays were performed using purified recombinant His 6 -tagged Hsp70 and a panel of RNA ligands based on the ARE from tumor necrosis factor ␣ mRNA, varying in length from 20 to 44 nucleotides (Table 1). Electrophoretic mobility shift assays (EMSAs) with both 38-and 24-nt ARE substrates supported a single major Hsp70 -RNA complex in each case (Fig. 1A), consistent with earlier reports (26,28). Binding to the 24-nt ARE substrate appeared much weaker based on the elevated protein concentration required to generate detectable RNPs. Consistent with previous findings, Hsp70 binding to the ARE substrates included sequence-specific contributions, as RNPs were not detected on an RNA substrate lacking AU-rich sequence (R␤), even at 2 M protein. Quantitative binding assays using fluorescence anisotropy corroborated the ARE length-dependence of Hsp70 binding, with high-affinity complexes (K d Ͻ15 nM) formed on all ARE substrates of 30 nucleotides or greater (Fig. 1B, Table 2). Shortening the ARE ligand from 30 to 24 nucleotides decreased the affinity for Hsp70 by a factor of 30, with further reductions in RNA substrate length (Fl-ARE [20]) yielding even weaker RNP complexes.
Hsp70 proteins are composed of an N-terminal ATP-binding domain that is separated from a nearby peptide-binding domain by a short but highly conserved interdomain linker (34). The protein's short C-terminal domain ends with a sequence (EEVD) responsible for recruiting a number of cochaperone proteins (21). Previously we demonstrated that these C-terminal sequences did not contribute to the AREbinding activity of Hsp70; however, both the ATP-and peptidebinding domains were required for optimal RNP assembly (28). For our experiments, the ATP-binding domain was defined as amino acid residues 1-385, which terminate immediately before the interdomain linker, whereas the peptide-binding domain spans amino acids 386 -613 ( Fig. 2A). These domain assignments were based on their similarity to Hsp70 protein fragments successfully crystallized in complexes with cognate ligands (35,36). Furthermore, chemical denaturation experiments showed that N-terminal His 6 -tagged versions of each domain were stably folded in solution (28).
Previously, we demonstrated that the N-terminal His 6 tag did not affect the affinity of Hsp70wt for ARE substrates (26). However, given the relatively low affinity of the isolated ATP-

RNA recognition and mRNA stabilization by Hsp70
and peptide-binding domains for RNA targets (28) coupled with previous findings where His 6 tags altered the ligand binding and/or enzymatic activities of other recombinant proteins (37,38), it was important to verify that these weak interactions were not the result of (or limited by) the purification tag. Accordingly, we used enterokinase (EK) digestion to excise N-terminal His 6 tags from His 6 -Hsp70(1-385) and His 6 -Hsp70(386 -613). SDS-PAGE analyses verified digest completion (Fig. 2B), whereas fluorescence anisotropy-based binding assays were used to compare the affinities of His 6 -tagged and EK-cleaved proteins for the Fl-ARE [38] RNA substrate (Fig.  2C). These experiments demonstrated that removing the His 6 tag had no significant effect on the affinities of either the ATPbinding domain (1-385; K d app ϭ 78 Ϯ 10 nM (n ϭ 3) for His 6tagged versus 80 Ϯ 7 nM (n ϭ 3) for EK-cleaved) or peptidebinding domain (386 -613; K d app ϭ 65 Ϯ 8 nM (n ϭ 4) for His 6 -tagged versus 77 Ϯ 7 nM (n ϭ 3) for EK-cleaved) for the RNA ligand.

RNA binding by Hsp70 was independent of its ATPase cycle
An intrinsic component of Hsp70's chaperone function is mechanistic coupling between the nucleotide occupancy of the ATP-binding domain and distinct protein conformations. When ATP is bound, the Hsp70 molecule can loosely associate with client peptide. However, upon ATP hydrolysis, the molecule closes around the peptide substrate, and fast exchange is prevented (for review, see Ref. 23). This allosteric linkage of nucleotide binding, ATPase activity, and peptide binding by Hsp70 are a paradigm reminiscent of the effects that nucleotide-binding and hydrolysis exert on RNA association by RNA helicases (39). If the interaction between Hsp70 and RNA substrates is also regulated by the protein chaperone cycle, it would be expected that the nucleotide-bound status of Hsp70 would influence its RNA-binding affinity. To test this possibility, we first confirmed the nucleotide-binding activity and selectivity of recombinant His 6 -Hsp70, which is purified under denaturing conditions to remove associated protein and nucleotide cofactors before on-column refolding (28). Fluorescence anisotropy-based binding assays demonstrated that His 6 -Hsp70 bound the fluorescein-conjugated ATP analog N 6 -(6-amino) hexyl-ATP-6-carboxyfluorescein (Fl-N 6 -ATP) with an apparent K d ϭ 1.29 Ϯ 0.21 M (n ϭ 3) (Fig. 3A). Hsp70 -Fl-N 6 -ATP  [38] and Fl-ARE [24] RNA substrates. Isotherms were resolved using single-site binding models to determine dissociation constants listed in Table 2. Ͼ 500 2 a Apparent equilibrium dissociation constants were resolved using fluorescence anisotropy-based assays as described in Fig. 1 and are listed as the mean Ϯ S.D. from n independent experiments.

RNA recognition and mRNA stabilization by Hsp70
complex formation was effectively competed by inclusion of unlabeled ATP, ADP, and to a slightly lesser extent by the nonhydrolysable ATP analog AMP-PNP (Fig. 3B). By contrast, CTP could not compete with Fl-N 6 -ATP for Hsp70 binding, consistent with specific recruitment of adenosine-based nucleotides by this protein.
Next, we tested whether excess ATP, ADP, or AMP-PNP could modulate the RNA-binding activity of Hsp70 by measuring the effects of each nucleotide on His 6 -Hsp70 affinity for an ARE ligand. The relatively compact Fl-ARE [34] substrate was selected for these experiments in light of an inherent restriction to Hsp70 RNP formation. Hsp70 proteins coordinate bound nucleotides using Mg 2ϩ (40), however, longer ARE sequences favor folded states in the presence of multivalent cations (41) that can inhibit Hsp70 binding (42). For example, mFold (43) returns a predicted ⌬G fold for the ARE [38] substrate of Ϫ2.44 kcal/mol at 25°C, close to the value of Ϫ1.52 kcal/mol measured using thermal denaturation assays (42). By contrast, mFold calculates ⌬G Ϸ 0 kcal/mol for the ARE [34] substrate at 25°C. This prediction that the ARE [34] RNA ligand would be less susceptible to Mg 2ϩ -stabilized folding was supported by anisotropy isotherms, where His 6 -Hsp70 bound Fl-ARE (34) with an apparent K d of 11.7 Ϯ 2.7 nM in the presence of 2 mM Mg 2ϩ (Fig. 3C, Table 3), indistinguishable from the affinity measured in reactions lacking Mg 2ϩ (Table 2). Including ATP, ADP, or AMP-PNP in binding reactions did not significantly affect the affinity of His 6 -Hsp70 for the Fl-ARE [34] ligand ( Table 3), suggesting that the protein contacts involved in RNA recognition and binding are not significantly altered as Hsp70 moves through the conformational transitions associated with its peptide refolding activity.
Observations that both the ATP-and peptide-binding domains of Hsp70 can independently associate with RNA ligands ( Fig. 2 and Ref. 28) also raised the possibility that nucleotide-dependent inhibition of RNA recruitment by the ATPbinding domain might be compensated by interactions elsewhere in the protein. To test whether RNA contacts specifically within the ATP-binding domain were influenced by its nucle-  Table 3. D, anisotropy isotherm of His 6 -Hsp70(1-385) binding to Fl-N 6 -ATP as described in A. E, nucleotide competition assays measuring Fl-N 6 -ATP displacement from His 6 -Hsp70(1-385) as described in B. F, anisotropy isotherms of His 6 -Hsp70(1-385) binding to the Fl-ARE [34] RNA ligand (0.5 nM) in buffer containing 2 mM MgCl 2 in the absence (black) or presence (red) of 2 mM ATP. G, ATPase activity of His 6 -Hsp70 (2 M) measured as described under "Experimental Procedures" after complex formation with the indicated RNA or peptide (NLLRLTG) ligands (5 M). Bars represent the mean Ϯ S.D. of ATPase activities measured across four independent reactions.

RNA recognition and mRNA stabilization by Hsp70
otide-bound status, we repeated the nucleotide-and RNAbinding experiments described above but using the isolated ATP-binding domain alone. His 6 -Hsp70(1-385) formed a complex with Fl-N 6 -ATP with K d app ϭ 1.15 Ϯ 0.12 M (n ϭ 2) (Fig. 3D), very close to the affinity observed for the full-length protein, and was similarly displaced by competition with unlabeled ATP, ADP, and AMP-PNP (Fig. 3E). Furthermore, the affinity of His 6 -Hsp70(1-385) binding to the Fl-ARE [34] RNA substrate was not significantly affected by inclusion of any tested adenosine nucleotide (Fig. 3F, Table 3), indicating that RNA contacts within the Hsp70 ATP-binding domain are not impacted by associated nucleotide ligands.
Finally, the ATPase activity of Hsp70 was measured to determine whether it might be influenced by RNA binding, analogous to the RNA-dependent ATPase activity of RNA helicases (44,45). ATPase assays were performed under subsaturating nucleotide concentrations (5 M ATP) to maximize resolution of any ligand-dependent effects. Previous studies resolved K m ϭ 10 M and V max ϭ 0.095 min Ϫ1 for the ATPase activity of purified human Hsp70 (46), and initial reaction velocities measured for His 6 -Hsp70 in our assays were consistent with these parameters (Fig. 3G). However, preincubation with saturating concentrations of RNA ligands had no significant effect on ATPase activity, further confirming that it is independent of Hsp70's RNA-binding role. Additionally, it was noted that Hsp70 ATPase activity was also unaffected by the addition of a peptide ligand. This differs from the chaperone cycle paradigm typified by Escherichia coli DnaK, where ATPase function is potently enhanced after peptide substrate-binding (47) but is consistent with previous findings with purified human Hsp70 (48). These findings suggest some distinctions in catalytic mechanisms between the bacterial and mammalian Hsp70 homologs, at least in the absence of co-chaperone proteins.

RNA binding by Hsp70 was independent of its peptide-binding activity
Observations that the peptide-binding domain contributed to RNP formation by Hsp70 in vitro (Refs. 27 and 28 and Fig. 2) also raised the possibility of competition or allostery between peptide-and RNA-binding roles. This possibility was tested using the peptide ligand NLLRLTG, an Hsp70 substrate identified by phage display (49) and previously shown to bind with low-to mid-micromolar affinity, similar to other aliphatic peptides (50). Anisotropy-based binding assays using an Fl-tagged peptide demonstrated slow association kinetics for His 6 -Hsp70 at 25°C (Fig. 4A). This necessitated extended incubation times (1 h) for equilibrium binding measurements, which revealed an apparent K d ϭ 25 Ϯ 5 M (n ϭ 3) (Fig. 4B). However, including near-saturating concentrations of peptide ligand in RNA-binding reactions had no detectable effect on the affinity of His 6 -  Table 3 Effect of nucleotides on the ARE-binding affinity of select Hsp70 proteins Apparent equilibrium dissociation constants for the indicated proteins binding to the Fl-ARE͓34͔ RNA substrate were resolved using fluorescence anisotropy-based assays as described in Fig. 3 and are given as the mean Ϯ S.D. from n independent experiments.

Protein
Nucleotide

RNA recognition and mRNA stabilization by Hsp70
Hsp70 for the Fl-ARE [38] substrate (Fig. 4C), indicating that Hsp70 binding to RNA is independent of associated peptide cargoes. Parallel experiments were performed using the isolated peptide-binding domain to test the possibility that compensatory contacts in full-length Hsp70 might mask the effects of peptide ligands on RNA interactions within the peptide-binding moiety. Complexes between the Fl-NLLRLTG substrate and His 6 -Hsp70(386 -613) formed more slowly than those involving the wild-type protein, although equilibrium was still approached within 1 h (Fig. 4D). The peptide ligand also showed slightly better affinity for His 6 -Hsp70(386 -613) than the wild-type protein with an apparent K d ϭ 16 Ϯ 3 M (n ϭ 2) (Fig. 4E). However, similar to observations with the wild-type protein (Fig. 4C), 50 M unlabeled NLLRLTG ligand had no detectable effect on the affinity of His 6 -Hsp70(386 -613) for the Fl-ARE [38] RNA substrate (Fig. 4F), confirming that peptide ligands do not impair RNP assembly by the peptide-binding domain of Hsp70.

The peptide-binding domain of Hsp70 was sufficient to bind and stabilize an mRNA target in cells
Although the ATP-and peptide-binding domains of Hsp70 can independently interact with RNA (Fig. 2), both are required to form optimal RNP complexes with the Fl-ARE [38] substrate (28), prompting the hypothesis that RNA-binding might be coupled to the Hsp70 chaperone cycle. However, because the experiments described above indicate no allosteric linkage between these processes, it follows that the ability of Hsp70 to bind and stabilize mRNA substrates in cells might not require both major protein domains. To test this hypothesis, we first analyzed the interaction between select His 6 -Hsp70 truncation mutants ( Fig. 2A) and the ARE sequence within the 3Ј-UTR of VEGFA mRNA (Fig. 5A) shown to mediate Hsp70directed stabilization of this transcript (28). EMSAs demonstrated that both the ATP-binding (1-385) and peptide-binding (386 -613) domains independently formed complexes with the VEGFA ARE fragment in vitro (Fig. 5B), but not with an RNA substrate spanning the VEGFA-coding sequence (Fig.  5C). However, RNPs formed on the ARE ligand at lower protein concentrations for full-length His 6 -Hsp70wt or a mutant spanning both the ATP-and peptide-binding domains (1-613), suggesting increased complex stability when both major Hsp70 domains were present, consistent with previous results observed using the ARE [38] substrate (28).
To assess the roles of Hsp70 subdomains in cellular mRNA binding, FLAG-tagged Hsp70 truncation mutants were individually expressed from siRNA-resistant cDNAs in HeLa cells where endogenous Hsp70 had been suppressed using siRNA. Transfection conditions were optimized such that ectopic FLAG-Hsp70wt was expressed at levels comparable with endogenous Hsp70 in control cells (Fig. 6A, left). Because some Hsp70 deletion mutant proteins were not recognized by the anti-Hsp70 antibody, similar levels of expression from FLAGtagged shRNA-resistant Hsp70 deletion cassettes versus ectopically expressed wild-type protein were verified by Western blot using anti-FLAG antibodies (Fig. 6A, right). In FLAG-targeted ribonucleoprotein immunoprecipitation (RIP) experiments, endogenous VEGFA mRNA was robustly enriched in immunoprecipitated pellets from cells expressing FLAG-Hsp70wt versus cells lacking the transgene (Fig. 6B), consistent with previ-

RNA recognition and mRNA stabilization by Hsp70
ous data showing this transcript binding to endogenous Hsp70 (28). VEGFA mRNA was similarly recovered in complexes with the Hsp70(1-613) mutant, consisting of both the ATP-and peptide-binding domains but lacking the C-terminal sequences responsible for recruiting select co-chaperone proteins (21). However, when testing individual Hsp70 subdomains, significant enrichment of VEGFA mRNA was noted in RIP assays targeting the FLAG-tagged Hsp70 peptide-binding domain (386 -613) but not the ATP-binding moiety (1-385). The data suggest that, unlike in vitro reactions where both the Hsp70 ATP-and peptide-binding domains can form stable RNPs with ARE-containing RNA substrates (Fig. 5 and Ref. 28), only the peptide-binding domain appears to retain this activity in cells, at least for the VEGFA transcript.
In a final series of experiments, the functional consequences of specific Hsp70 domains on VEGFA mRNA decay kinetics were assessed in HeLa cells using actinomycin D (actD) time course assays. The half-life of endogenous VEGFA mRNA decreased by 50% when Hsp70 was suppressed in HeLa cells using siRNA (Fig. 6C and Table 4), consistent with an mRNA-stabilizing role for this protein (28). Functional specificity for Hsp70 was confirmed by ectopic expression of siRNA-resistant FLAG-Hsp70wt at levels comparable with the endogenous protein, which increased the VEGFA mRNA halflife to a value indistinguishable from that measured in siControl-transfected cells. Expression of the Hsp70(1-613) mutant and the peptide-binding domain alone (386 -613) similarly slowed the rate of VEGFA mRNA turnover. However, this transcript was not stabilized in cells expressing the Hsp70 ATPase domain alone (1-385). Together, these data indicate that the Hsp70 peptide-binding domain alone is sufficient to bind and stabilize an mRNA target in cells.  Table 4.

Discussion
High-affinity complexes between Hsp70 and AU-rich RNA ligands required ϳ30 nucleotides of RNA sequence, as affinity dropped precipitously in binding reactions with shorter RNA substrates ( Table 2). This RNA length requirement is similar to that previously described for AUF1 (33) but much longer than that required to recruit individual molecules of HuR (32) or tristetraprolin (31). Interactions across such long stretches of RNA may contribute to the extreme sensitivity of Hsp70 and AUF1 binding to local RNA folding (42), as occluding any subset of RNA-binding determinants would be expected to prevent stable RNP formation. Similarly, an extended RNA footprint may contribute to Hsp70's mRNA-stabilizing function by precluding access of competing mRNA-destabilizing factors for large segments of individual ARE targets, which can exceed 100 nucleotides in some mRNAs (11,51). An analogous mechanism is exploited by HuR, which forms cooperative, multimeric protein complexes on extended ARE substrates (32).
Several findings from this work indicate that the RNA-binding and mRNA-stabilizing activities of Hsp70 are independent of its well-characterized protein chaperone functions, where substrate occupancy at the ATP-and peptide-binding domains are intercommunicated via small scale allosteric changes (52)(53)(54)(55). In the classical model exemplified by DnaK, the hydrolysis of ATP to ADP converts the peptide-binding domain from an open to a closed conformation (56). As such, affinity for peptide ligands is greatest when Hsp70 is in its ADP-bound form (29). By contrast, the affinity of Hsp70 for the Fl-ARE [34] RNA ligand was unaffected by nucleotide co-factors (Table 3), suggesting (i) that RNA contacts with the Hsp70 ATPase domain do not involve the nucleotide binding pocket, and (ii) that ARE binding is not sensitive to the allosteric molecular changes associated with the chaperone cycle. These data directly contradict a previous study by Henics et al. (25), who reported that Hsp70 binding to an ARE target was inhibited in reactions supplemented with ATP. However, we suggest that a technical detail might be responsible for the ATP-dependent decrease in UV cross-linking efficiency that formed the basis for their conclusion. At 260 nm, ATP has an extinction coefficient (⑀ 260 ) of 1.54 ϫ 10 4 M Ϫ1 ⅐cm Ϫ1 . Henics et al. (25) noted that UV crosslinking between Hsp70 and a model ARE was severely inhibited at ATP concentrations as low as 1 mM. Assuming that a 10 -20-l cross-linking reaction in a 1.5-ml sample tube pres-ents a fluid depth of Ϸ1 mm, by Beer's law this sample would yield an absorbance value of 1.54 at 260 nm, which converted to transmittance (A ϭ log(1/T)) indicates that only 2.9% of the applied UV radiation actually penetrates to the bottom of the reaction tube. We submit, therefore, that the ATP-dependent inhibition of Hsp70:ARE cross-linking reported in the previously published work was likely the result of an inner filter effect that prevented the vast majority of incident photons from reaching and covalently coupling target RNP complexes. By contrast, our fluorescence anisotropy measurements are taken using excitation and emission wavelengths in the 490 -530-nm range, where ATP exhibits no significant absorbance.
A second apparent divergence between the RNA-binding mechanism of Hsp70 and its canonical chaperone functions is the inability of RNA substrates to modulate protein ATPase activity. Traditionally, most biochemical features of Hsp70 chaperone functions have been defined using the E. coli DnaK model, where the hydrolysis of ATP is intrinsically very slow but dramatically stimulated by association of peptide ligands or co-chaperone proteins (47,53,57). Our observations that neither an RNA nor a peptide substrate could stimulate ATP hydrolysis by human Hsp70 (Fig. 3G) suggests that the DnaK paradigm does not accurately reflect this allosteric relationship for the human protein. This point was supported by an independent group who also observed no enhancement of human Hsp70 ATPase activity by peptide ligands (48) but does not exclude the possibility that these activities might be coupled by the inclusion of select co-chaperone proteins and/or specific post-translational modifications (58 -62). Independence of Hsp70's RNA-binding roles from its chaperone functions may also account for the apparent lack of competition between RNA and peptide ligands (Fig. 4), which likely reflects distinct molecular determinants for each of these interacting partners.
Although both the ATP-and peptide-binding domains individually form moderately stable RNPs with RNA substrates in vitro (Figs. 2 and 5 and Ref. 28) despite lacking any known RNAbinding motif, only Hsp70 mutants containing the peptidebinding domain formed stable and functional complexes with VEGFA mRNA in cells (Fig. 6). In fact, the peptide-binding domain alone (386 -613) was sufficient to bind and stabilize this target transcript. Because this truncation mutant lacked binding sites for the major Hsp40/DnaJ family co-chaperones, located within the ATP-binding domain and the C-terminal tail (58,59), it is likely that these co-chaperones are not required for Hsp70's RNA-binding and -stabilizing functions. Curiously, the ATP-binding domain alone was unable to stabilize VEGFA mRNA in cells despite binding ARE targets in vitro at affinities comparable with the peptide-binding domain (Figs. 2 and 5 and Ref. 28). Although levels of ectopically expressed Hsp70 truncation mutants were similar in transfection experiments (Fig.  6A), we cannot yet exclude the possibilities that the ATP-binding domain is inappropriately localized in cells when expressed independently or that its RNA-binding potential might be influenced by other cellular binding partners.
Cumulatively, the relatively large length of RNA required for stable association of Hsp70 together with the lack of any detectable linkage between its RNA-binding activity and chaperone cycle are consistent with a model whereby Hsp70 stabilizes

RNA recognition and mRNA stabilization by Hsp70
mRNAs by competing with ARE-targeted mRNA-destabilizing factors for cognate binding sites. Although contributions of this activity to Hsp70's cytoprotective effects remain unknown, selective stabilization of Hsp70-associated mRNAs could explain, in part, an early report showing that the ATP-binding domain is dispensable for Hsp70-mediated thermotolerance in a rat fibroblast model (63). However, the ability of Hsp70 to bind RNA and peptide targets independently presents appealing mechanical hypotheses for additional well-established and newly identified functions of this protein. For example, separate RNA-and peptide-binding interfaces could enable roles in cotranslational protein folding (64) by tethering Hsp70 to the protein synthesis machinery, analogous to that recently described for the yeast ribosome-associated Hsp70 protein SSB (65), or even by recruitment via the 3Ј-UTRs of translating mRNAs (e.g. Fig. 5). Reciprocally, the peptide-binding functions of Hsp70 may contribute to their engagement in stress granule cores, recently demonstrated in both yeast and mammalian cells (66). Although Hsp70 is required for efficient disassembly of these complexes after stress (67), it is conceivable that its RNA-binding function might also contribute to the selective RNA triage roles of these structures. We envision that future efforts to comprehensively identify the RNA interactome of Hsp70 will provide essential data to direct downstream efforts aimed at resolving the functional significance of Hsp70 binding to RNA.

Materials
RNA oligonucleotide substrates were purchased from GE Dharmacon or Integrated DNA Technologies. Sequences are listed in Table 1. Probes prefixed "Fl-" include a 5Ј-fluorescein (Fl) tag. Free nucleotides (ATP, ADP, AMP-PNP) were from Sigma. The Fl-tagged ATP analogue Fl-N 6 -ATP was purchased from Jena Bioscience, whereas peptide substrates NLLRLTG and an N-terminal Fl-tagged variant were from Genscript. Hsp70 siRNA was from GE Dharmacon with targeting (antisense) strand 5Ј-UUUCUCUUGAACUCCUCCAUU-3Ј, whereas siControl was from Bioneer with targeting strand 5Ј-ACGAAAUUGGUGGCGUAGGdTdT-3Ј.

Recombinant proteins
N-terminal His 6 -tagged Hsp70 (encoded by the human HSPA1A gene) and truncation mutant proteins were expressed from pBAD/His vectors (Invitrogen) in E. coli Rosetta cells (Novagen), purified under denaturing conditions using Ni 2ϩaffinity chromatography, refolded on-column using a reverse urea gradient (8 M to 0.2 M), concentrated, and quantified as described previously (28). Where indicated, N-terminal His 6 tags were removed using the Enterokinase Cleavage Capture Kit (Novagen) following the manufacturer's instructions.
Quantitative assessment of equilibrium binding between His 6 -Hsp70 proteins and Fl-labeled RNA, nucleotide, or peptide ligands was performed using fluorescence anisotropybased assays with a Beacon 2000 fluorescence polarization system as described previously (28). When RNA-protein binding was assayed in the presence of nucleotide, EDTA was removed from the reaction mixture and replaced with MgCl 2 (2 mM final concentration) and the desired nucleotide (2 mM final). Binding reactions containing RNA or nucleotide substrates were incubated for 30 min at 25°C before measurement of fluorescence anisotropy, as preliminary on-rate experiments indicated that equilibrium was attained during this period (data not shown). However, the slower association kinetics of peptide ligands under these conditions (Fig. 4, A and D) necessitated 1-h incubations for any reactions that included peptide components. Because total fluorescence emission from Fl-tagged ligands did not significantly vary as a function of protein concentration, apparent equilibrium association constants (K ϭ 1/K d app ) were resolved from plots of total measured anisotropy (A t ) versus protein concentration ([P]) using Equation 1, where A L is the intrinsic anisotropy of the free ligand (RNA, nucleotide, or peptide at concentrations Ͻ ϽK d app ) and A PL is the anisotropy of the protein-ligand complex (69).
PRISM v3.03 software (GraphPad) was used for all nonlinear regression analyses.

RNA recognition and mRNA stabilization by Hsp70
ATPase assays ATP hydrolysis by His 6 -Hsp70 was monitored by measuring 32 P i release from [␥-32 P]ATP. His 6 -Hsp70 (2 M) with or without peptide or RNA ligands (5 M each) was added to ATPase reaction buffer (50 mM TrisHCl (pH 8.0) containing 100 mM KCl and 5 mM MgCl 2 ) and incubated at 37°C for 1 h to allow formation of Hsp70 -ligand complexes. ATPase reactions were then initiated by adding ATP (5 M final) containing [␥-32 P]ATP (10 4 cpm/pmol ATP). After 20 min at 37°C, reactions were terminated by transfer to an equal volume of ice-cold 100 mM EDTA (pH 8.0). Liberated 32 P i was then measured using the EasyRad Phosphate Assay kit (Cytoskeleton, Inc.). Pilot time-course experiments indicated that this incubation period was well within the linear range of reaction progress (data not shown). His 6 -Hsp70-dependent phosphate release was calculated by subtraction of 32 P i counts from parallel reactions lacking protein and expressed as pmol of ATP hydrolyzed/min/pmol of His 6 -Hsp70.
Decay of endogenous VEGFA mRNA was measured using actD time-course assays. After inhibiting global transcription by adding actD (5 g/ml; Calbiochem) directly to the growth media, total RNA was purified using TRIzol reagent (Invitrogen) according to the manufacturer's instructions at select time points. Time courses were limited to 4 h to minimize the complicating effects of actD-induced apoptosis signaling (70). RNA samples were analyzed for VEGFA and GAPDH mRNAs by multiplex RT-qPCR using the qScript One-Step RT-qPCR kit (Bio-Rad) programmed with the following primer/probe sets: VEGFA sense 5Ј-GCACCCATGGCAGAAGG-3Ј, antisense 5Ј-CTCGATTGGATGGCAGTAGCT-3Ј, and probe Fl-CTGATAGACATCCATGAACTTCACCACTTCGT-Black Hole Quencher-1; GAPDH sense 5vЈ-GAGAGTCAGCCG-CATCTTC-3Ј, antisense 5Ј-ACTCCGACCTTCACCTTCC-3Ј, and probe Texas Red-CGCCAGCCGAGCCACATCGC-Black Hole Quencher-2. VEGFA mRNA levels were normalized to GAPDH, averaged across four RT-qPCR reactions, and then plotted as the %VEGFA mRNA remaining as a function of actD treatment time. First-order decay constants (k) and associated mRNA half-lives (t1 ⁄ 2 ϭ ln2/k) were calculated by nonlinear regression as described (71).

Ribonucleoprotein immunoprecipitation
HeLa cells were transiently co-transfected with Hsp70 siRNA and expression vectors encoding shRNA-resistant FLAG-tagged wild-type Hsp70 or the indicated truncation mutants as described above. RNP complexes containing specified FLAG-tagged proteins were then purified from crude lysates by immunoprecipitation with anti-FLAG antibodies essentially as described (72). Relative levels of VEGFA and GAPDH mRNAs co-purifying with immunoprecipitated RNPs were quantified by multiplex RT-qPCR as described above. Recovered VEGFA mRNA levels were normalized to GAPDH mRNA and averaged across four RT-qPCR reactions, then expressed relative to VEGFA mRNA immunoprecipitated from cell extracts lacking FLAG-Hsp70 proteins.

Statistics
Comparisons of equilibrium-binding constants, ATPase rates, mRNA enrichment in RIP reactions, and mRNA halflives were performed using the unpaired t test based on the mean Ϯ S.D. of each parameter measured across at least three independent replicate experiments. Differences yielding p Ͻ 0.05 were considered significant.
Author contributions-A. K. and G. M. W. designed the study and wrote the paper. A. K., A. E. M., and G. M. W. prepared the recombinant proteins. A. K., A. E. M., Z. Y., and B. T. collected and analyzed fluorescence anisotropy data. A. K. and G. M. W. performed the ATPase assays. A. K. and E. J. F. W. performed the EMSAs and analyzed the RNA-binding and mRNA-stabilizing activities of Hsp70 mutants in mammalian cells. All authors reviewed the results and approved the final version of the manuscript.