Hsp70 Is a Novel Posttranscriptional Regulator of Gene Expression That Binds and Stabilizes Selected mRNAs Containing AU-Rich Elements

Abstract

The AU-rich elements (AREs) encoded within many mRNA 3′ untranslated regions (3′UTRs) are targets for factors that control transcript longevity and translational efficiency. Hsp70, best known as a protein chaperone with well-defined peptide-refolding properties, is known to interact with ARE-like RNA substrates in vitro. Here, we show that cofactor-free preparations of Hsp70 form direct, high-affinity complexes with ARE substrates based on specific recognition of U-rich sequences by both the ATP- and peptide-binding domains. Suppressing Hsp70 in HeLa cells destabilized an ARE reporter mRNA, indicating a novel ARE-directed mRNA-stabilizing role for this protein. Hsp70 also bound and stabilized endogenous ARE-containing mRNAs encoding vascular endothelial growth factor (VEGF) and Cox-2, which involved a mechanism that was unaffected by an inhibitor of its protein chaperone function. Hsp70 recognition and stabilization of VEGF mRNA was mediated by an ARE-like sequence in the proximal 3′UTR. Finally, stabilization of VEGF mRNA coincided with the accumulation of Hsp70 protein in HL60 promyelocytic leukemia cells recovering from acute thermal stress. We propose that the binding and stabilization of selected ARE-containing mRNAs may contribute to the cytoprotective effects of Hsp70 following cellular stress but may also provide a novel mechanism linking constitutively elevated Hsp70 expression to the development of aggressive neoplastic phenotypes.

INTRODUCTION

Cells use a variety of mechanisms to regulate gene expression. The production of mature mRNAs is controlled at the levels of transcription and pre-mRNA processing prior to their export to the cytoplasm. However, the cytoplasmic steady-state level of an mRNA and, hence, its potential to program protein synthesis, is also regulated by its turnover kinetics. Modulation of mRNA decay rates allows cells to control the production of specific proteins independent of transcriptional events but also regulates the speed with which new mRNA steady states are approached following changes in the transcription rate (1–3).

The decay rates of mRNAs are regulated through cis-acting sequence determinants resident within each transcript. The best-characterized examples of these are the AU-rich elements (AREs) that are frequently contained within the 3′ untranslated regions (3′UTRs) of mRNAs encoding cytokines, oncoproteins, and inflammatory mediators (4). Although originally defined by overlapping repeats of AUUUA, AREs from different mRNAs are now known to exhibit a wide range of sequence characteristics, although they are universally enriched in uridylate residues (5). Computational predictions estimate that up to 10% of all cellular transcripts contain AREs (6), although only a small fraction have been experimentally validated. The regulatory consequences of AREs are mediated by association with ARE-binding proteins (ARE-BPs), a diverse collection of cellular factors that vary in RNA substrate selectivity and function. For example, tristetraprolin accelerates the decay of ARE-containing transcripts, while HuR functions antagonistically as an mRNA-stabilizing factor (7–9). Other ARE-BPs, including TIA-1 and TIAR, function as translational suppressors (10). An emerging theme is that ARE-BPs may exert differential effects on mRNA metabolism in an RNA context-dependent manner. For example, AUF1 accelerates the decay of many mRNA targets (11, 12) but can compete with TIAR to regulate the translational efficiency of the MYC transcript (13). ARE-BPs play a significant role in some disease processes, particularly cancers and inflammatory conditions, by controlling the expression of many mRNAs that encode critical regulatory proteins, including p53 and tumor necrosis factor alpha (TNF-α) (14, 15).

Another protein that can directly bind AREs in vitro is the stress-inducible 70-kDa heat shock protein, Hsp70 (also known as Hsp72 and HspA1A) (16, 17). A link between Hsp70 and ARE-mediated mRNA decay was suggested following observations that heat shock can stabilize ARE-containing transcripts and that Hsp70 copurifies with AUF1 (18). However, Hsp70 does not possess any known RNA-binding motif, and the functional significance of its association with ARE substrates has not been determined. In contrast, the role of Hsp70 as a protein chaperone, where it binds hydrophobic peptide domains to prevent aggregation and facilitate protein folding, has been very well documented (19, 20). Peptide binding and release are directed allosterically by the ATPase cycle of Hsp70 and can be regulated by ancillary cochaperone factors.

In thermally stressed cells, Hsp70 enhances survival by preventing protein misfolding and aggregation (21–23). However, while Hsp70 is transiently present in cells following heat shock or other stresses, it is constitutively increased in many tumors and its expression correlates with chemoresistance and poor prognosis (24–26). While the precise mechanisms linking Hsp70 expression to tumor development are largely unknown, recent work indicates that elevated Hsp70 levels can obstruct signaling that drives apoptosis (27, 28) and senescence (27–30). Consistent with these findings, suppression or inhibition of Hsp70 restricts tumor cell growth in a variety of cultured-cell and murine tumor models (31–33).

In this study, we demonstrate that the nucleic acid binding specificity of Hsp70 is highly selective for U-rich RNA sequences, including AREs, and identify Hsp70 protein domains contributing to RNA binding. Next, we show that cellular Hsp70 stabilizes ARE-containing transcripts. Since many mRNAs that encode factors regulating cell growth and division, apoptosis, angiogenesis, and metastasis contain AREs, this mRNA-stabilizing role may thus represent a novel mechanism for the protumorigenic properties of constitutive Hsp70 overexpression. Supporting this hypothesis, we tested the ability of Hsp70 to bind and regulate the decay kinetics of two ARE-containing mRNAs that encode potent tumor-promoting proteins, vascular endothelial growth factor (VEGF) and cyclooxygenase-2 (Cox-2), tested whether these activities required the protein chaperone function of Hsp70, and localized the Hsp70-targeted domain within the VEGF mRNA 3′UTR.

MATERIALS AND METHODS

Synthetic RNA and DNA substrates.

RNA and DNA oligonucleotides were purchased from Dharmacon Research or Integrated DNA Technologies. The RNA substrate ARE[38] (5′-GUGAUUAUUUAUUAUUUAUUUAUUAUUUAUUUAUUUAG-3′) is based on the core ARE sequence from TNF-α mRNA. A DNA version (ARE[38] DNA) substitutes thymidine for uracil, in addition to 2′-deoxyribose at each position (5′-GTGATTATTTATTATTTATTTATTATTTATTTATTTAG-3′). The RNA substrate U[32] incorporates 32 contiguous uridylate residues (5′-GAUCU₃₂A-3′), while Rβ encodes a fragment of the coding sequence from rabbit β-globin (βG) mRNA (5′-UGGCCAAUGCCCUGGCUCACAAAUACCACUG-3′). RNA and DNA substrates containing a 5′ fluorescein tag are prefixed accordingly (5′-Fl). Extended polynucleotides poly(C), poly(A), and poly(I)·poly(C) were from GE Healthcare.

Antibodies.

Mouse monoclonal anti-Hsp70 antibody clones C92F3A-5 and N15F2-5, which do not cross-react with Hsc70, were purchased from Abcam and Enzo Life Sciences, respectively. Rabbit polyclonal anti-Hsc70 antibody, which does not cross-react with Hsp70, was from Cayman Chemical Company. Rabbit polyclonal anti-AUF1 antibody was from Millipore, while anti-glyceraldehyde-3-phosphate dehydrogenase (anti-GAPDH) and anti-FLAG antibodies coupled to horseradish peroxidase were from Sigma, as were all peroxidase-conjugated secondary antibodies.

Generation of recombinant Hsp70 proteins.

Plasmid pBAD/HisC-Hsp70, which contains the complete coding sequence of human Hsp70 encoded by the HSPA1A gene downstream from a His₆ tag motif and was described previously (17), was used to transform Escherichia coli Rosetta 2 cells (Novagen). Cultures were grown in a 37°C shaking incubator in SOB medium (20 g tryptone, 10 g yeast extract, 0.5 g NaCl, and 0.5 g KCl per liter, pH 7.5) containing MgCl₂ (10 mM), chloramphenicol (34 μg/ml), and ampicillin (50 μg/ml) to an optical density at 600 nm (OD₆₀₀) of 0.6. Production of His₆-Hsp70 was then induced by the addition of arabinose (0.02%), and cultures were transferred to a 30°C shaker for 4 h. His₆-Hsp70 was then purified under denaturing conditions using a modification of a protocol by Jindal et al. (34). Briefly, cells were pelleted and then resuspended in column wash buffer (50 mM sodium phosphate, 500 mM NaCl, 20 mM imidazole, pH 8.0) containing 8 M urea and disrupted by sonication. Urea stocks were made fresh daily from ultrapure-grade crystals (American Bioanalytical) and deionized by gently mixing with Amberlite MB-150 mixed-bed exchanger (1%, wt/vol; Sigma) at room temperature for 1 h to remove cyanate breakdown products that can carbamylate proteins (35). After cellular debris was pelleted (20,000 × g for 30 min), His₆-Hsp70 was purified by Ni²⁺ affinity chromatography over a HiTrap chelating affinity column (GE Healthcare) equilibrated in column wash buffer containing 8 M urea. The column was washed in the same buffer to remove unbound material, and then bound His₆-Hsp70 was refolded on the column by applying a reverse urea gradient (8 M to 0.2 M) in column wash buffer over 3 h. Purified His₆-Hsp70 was then eluted in imidazole elution buffer (20 mM sodium phosphate, 500 mM NaCl, 580 mM imidazole, pH 6.3) containing 0.2 M urea, concentrated in an Amicon ultra concentrator (30-kDa molecular-mass cutoff), and quantified using Coomassie blue-stained SDS-PAGE against a titration of bovine serum albumin. Protein was stored at −80°C in the elution buffer. The purity of full-length His₆-Hsp70 was judged to be >90% by Coomassie blue-stained SDS-PAGE (Fig. 1A). Dynamic light scattering analyses (Zetasizer Nano series) verified that recombinant His₆-Hsp70 samples did not contain high-molecular-mass complexes or aggregates. Typical protein preparations displayed particles 8 to 10 nm in diameter, which constituted >98% of the total sample volume (Fig. 1B). For RNA-binding assays, His₆-Hsp70 stocks were diluted in 10 mM Tris-HCl (pH 8) containing acetylated bovine serum albumin (0.1 mg/ml) immediately before use.

Fig 1.

Specific and direct binding of Hsp70 to U-rich RNA substrates. (A) Coomassie blue-stained SDS-PAGE of recombinant His₆-Hsp70 generated via Ni²⁺-affinity chromatography under denaturing conditions followed by on-column refolding. Molecular mass markers (in kDa) are indicated at left. (B) A dynamic light scattering volume trace showing purified His₆-Hsp70 as a monodisperse species in solution with an average diameter of 9.7 nm. (C) EMSAs performed using 5′-³²P-labeled ARE[38] (left) or Rβ (right) RNA substrates with titrations of His₆-Hsp70. Lanes marked “NP” contain no protein. Arrowheads indicate bands corresponding to complexes formed between RNA substrates and His₆-Hsp70. (D) Fluorescence anisotropy assays of His₆-Hsp70 binding to Fl-ARE[38] RNA and DNA substrates. RNA binding was resolved using a single-site model to determine dissociation constants given in Table 1. (E) His₆-Hsp70 (30 nM) was incubated with Fl-ARE[38] RNA (0.5 nM) for 30 min at 25°C prior to adding various concentrations of unlabeled competitor RNAs. Incubation continued for a further 30 min before measuring reaction mixture anisotropy. ARE[38], U[32], and Rβ probes were added at the indicated molar ratios to the Fl-ARE[38] substrate. Extended poly(A), poly(C), and poly(I·C) substrates were added based on molar excess of 38-nt binding sites. Points represent the means ± SD of three separate trials.

To generate His₆-Hsp70 deletion mutants, relevant cDNA fragments were amplified from pBAD/HisC-Hsp70 by PCR and subcloned into pBAD/HisC to generate in-frame His₆ fusion cassettes. Production and characterization of recombinant His₆-Hsp70 deletion mutant proteins were performed as described above for the full-length protein.

Protein-nucleic acid binding assays.

Qualitative assessments of His₆-Hsp70 binding to nucleic acid substrates were performed using electrophoretic mobility shift assays (EMSAs) essentially as described previously (17). RNA substrates ARE[38] and Rβ were 5′-³²P-radiolabeled using [γ-³²P]ATP (PerkinElmer Life Sciences) and T4 polynucleotide kinase (Promega) to specific activities of 3 × 10³ to 5 × 10³ cpm/fmol. To generate substrates derived from VEGF mRNA, cDNA fragments tagged with 5′-T7 promoter sequences were first amplified from IMAGE clone 6006890 (American Type Culture Collection), which contains nucleotides 574 to 3422 of the VEGFA variant 6 transcript (GenBank nucleotide sequence accession number NM_001171628.1), by PCR using primers listed in Table S1 in the supplemental material. ³²P-labeled RNA substrates were generated from these cDNA fragments by in vitro transcription using the Ambion MAXIscript kit and [α-³²P]UTP to specific activities of 1× 10⁴ to 1.5 × 10⁴ cpm/fmol. VEGF mRNA fragment substrates were heated to 70°C for 5 min and then immediately placed on ice prior to use in EMSA reaction mixtures in order to limit RNA secondary structure.

Binding of His₆-Hsp70 proteins to fluorescein-labeled nucleic acid substrates was quantitatively measured using a fluorescence anisotropy-based assay. Briefly, RNA or DNA oligonucleotides containing 5′-Fl groups (0.5 nM) were incubated with various concentrations of His₆-Hsp70 in 10 mM Tris-HCl (pH 8) containing 50 mM KCl, 0.1 μg/μl acetylated bovine serum albumin (BSA), 2 mM dithiothreitol (DTT), and 0.5 mM EDTA for 30 min at 25°C. Heparin (1 μg/μl) was added to all reaction mixtures as a nonspecific polyanionic competitor. The fluorescence anisotropy and total fluorescence intensity values of the binding reaction mixtures were measured using a Beacon 2000 fluorescence polarization system (Panvera) equipped with fluorescein excitation (490 nm) and emission (535 nm) filters. The total fluorescence emission did not vary significantly as a function of protein concentration in any of the experiments described in this study (data not shown), permitting the affinity of single-site protein-nucleic acid binding events to be calculated using equation 1 (36), as follows:

$$A_t=\frac{A_R+A_{PR}K[P]}{1+K[P]}$$ (1)

Here, A_t is the total measured anisotropy of the reaction mixture, A_R is the intrinsic anisotropy of the free nucleic acid substrate, A_PR is the anisotropy of the protein-nucleic acid complex, [P] is the protein concentration, and K is the equilibrium association constant (K_d = 1/K, where K_d is dissociation constant). Nonlinear regression of A_t versus [P] data sets was performed using PRISM version 3.03 software (GraphPad).

Analyses of folded-protein stability by chemical denaturation.

The stability of protein folding was assessed for selected recombinant proteins by equilibrium denaturation in guanidine hydrochloride (GdnHCl), essentially as we have previously described (37). Briefly, protein samples (1 to 5 μM) were incubated for 1 h at room temperature in 10 mM Tris-HCl (pH 8) containing 50 mM KCl, 2 mM DTT, and 0.5 mM EDTA in the presence of various concentrations of GdnHCl. The extent of protein unfolding was then assessed by measuring protein fluorescence (excitation wavelength [λ_ex] = 270 nm and emission wavelength [λ_em] = 290 to 400 nm; 10-nm bandwidth) using a Cary Eclipse spectrofluorometer. Although His₆-Hsp70 contains three Trp residues, preliminary experiments using a λ_ex of 295 nm demonstrated that Trp emission was not significantly altered during protein unfolding (data not shown). As such, the excitation wavelength was shifted to 270 nm to include excitation of tyrosine residues, 16 of which are dispersed throughout the protein.

Protein denaturation was modeled as a two-state, GdnHCl-dependent transition between native and unfolded conformations which exhibited fluorescence emission intensities F_native and F_unfolded, respectively. Thermodynamic parameters describing the unfolding transition were then estimated from the change in protein fluorescence measured at 325 nm (F₃₂₅) as a function of GdnHCl concentration using equations 2 and 3 adapted from the linear extrapolation method (38), as modified by Manyusa and Whitford (39):

$$F_{325}=F_{native}-\left[\frac{(F_{native}-F_{unfolded})\mathrm{.}{e}^{-\mathrm{\Delta }{G}_u/RT}}{1+e^{-\mathrm{\Delta }{G}_u/RT}}\right]$$ (2)

$${\mathrm{\Delta }G}_u={\mathrm{\Delta }G}_{uw}-m_{eq}\left[GdnHCl\right]$$ (3)

ΔG_u is the free energy of protein denaturation at each GdnHCl concentration, ΔG_uw is the extrapolated free energy of unfolding in the absence of denaturant, m_eq represents the sensitivity of ΔG_u to GdnHCl concentration, R is the gas constant (1.986 × 10⁻³ kcal · mol⁻¹ · K⁻¹), and T is the absolute temperature. All parameters were resolved from plots of F₃₂₅ versus [GdnHCl] values by nonlinear regression using PRISM version 3.03 software (GraphPad).

Construction of βG reporter and Hsp70 expression vectors.

Plasmid pTRERβ-wt, containing the rabbit βG gene downstream from a tetracycline-responsive promoter, and pTRERβ-ARE[38], in which the ARE[38] sequence (described above) was subcloned downstream from the βG stop codon, were described previously (40). Additional βG reporter plasmids used in this study were constructed by amplifying specific regions of the VEGF 3′UTR from IMAGE clone 6006890 (described above) by PCR, using primers listed in Table S1 in the supplemental material, and then subcloning these fragments into the unique BglII restriction site in the proximal βG 3′UTR.

To generate Hsp70 expression vectors for mammalian cells, the complete open reading frame of human Hsp70 was amplified from pBAD/HisC-Hsp70 (described above) by PCR and subcloned downstream from three copies of the FLAG epitope in p3xFLAG-CMV10 (Sigma) to generate pCMV-FLAG-Hsp70wt. To construct plasmid pCMV-FLAG-Hsp70sm, expressing a small interfering RNA (siRNA)-resistant Hsp70 mRNA, a DNA duplex was synthesized that spanned a unique 925-bp BglII fragment of Hsp70 cDNA but contained codon-neutral mutations in the region targeted by the Hsp70 siRNA (Genscript). This fragment was then substituted for the corresponding BglII fragment in the wild-type Hsp70 vector. The fidelity of all plasmid constructs was verified by restriction mapping and automated DNA sequencing.

Cell culture and transfection.

HeLa cells (American Type Culture Collection) were maintained at 37°C and 5% CO₂ in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum. The HeLa/Tet-Off cell line (Clontech) was grown under the same conditions but in medium containing 100 μg/ml G418. HL60 cells (generous gift from Fey Rassool) were grown in RPMI 1640 supplemented with 10% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. Plasmids and siRNA duplexes were transfected into cells using the DharmaFECT Duo reagent (Dharmacon), which is optimized for simultaneous transfection of plasmid and silencing RNA sequences. The targeting (antisense) strand of the Hsp70 siRNA was 5′-UUUCUCUUGAACUCCUCCAUU-3′ (Dharmacon). The negative-control duplex had the targeting sequence 5′-ACGAAAUUGGUGGCGUAGGdTdT-3′ (Bioneer). The final concentration of siRNA in transfection mixtures was 50 nM. Cells were exposed to the transfection mixture for 48 h prior to sample harvest or initiation of time courses.

mRNA decay assays.

The decay kinetics of βG reporter mRNAs were analyzed using doxycycline (Dox) time course assays essentially as described previously (41). Plasmids expressing βG reporter transcripts and enhanced green fluorescence protein (EGFP) were cotransfected into HeLa/Tet-Off cells. After 48 h, transcription of the βG cassette was selectively inhibited by adding fresh growth medium containing Dox (2 μg/ml). DNA-free total RNA was purified at various time points thereafter using the NucleoSpin RNA II kit (Macherey-Nagel). From each sample, βG reporter and EGFP mRNAs were quantified by multiplex quantitative reverse transcription (qRT)-PCR using the qScript One-Step qRT-PCR kit (Quanta Biosciences) with primers and probes listed in Table S2 in the supplemental material. The βG reporter mRNA level was normalized to the level of EGFP mRNA in each amplification reaction mixture and then plotted as the percentage of βG reporter mRNA remaining at each time point based on the mean ± standard deviation (SD) of four independent qRT-PCRs from each RNA sample. First-order decay constants (k) were solved by nonlinear regression to a single exponential decay function using PRISM version 3.03 software, yielding mRNA half-lives (t_1/2) of ln2/k.

Decay of endogenously expressed mRNAs was evaluated using actinomycin D (ActD; Calbiochem) time course assays, where global cellular transcription was repressed by adding ActD (5 μg/ml) directly to the growth medium. Total RNA was purified using TRIzol reagent (Invitrogen) according to the manufacturer's instructions at various time points thereafter. Time courses were limited to 4 h to avoid complicating mRNA decay kinetics by the initiation of an ActD-based apoptosis program (42). RNA samples were analyzed for VEGF and Cox-2 mRNAs by multiplex qRT-PCR using the qScript One-Step qRT-PCR kit with the primers and probes listed in Table S2 in the supplemental material, with each data point taken as the mean ± SD of four qRT-PCRs. Transcript levels were normalized to the level of GAPDH. First-order decay constants (k) and associated mRNA half-lives were calculated by nonlinear regression as described above. Where indicated below, steady-state mRNA levels were compared using total RNA harvested from cultures lacking ActD.

Protein coimmunoprecipitation assays.

Complex formation between FLAG-Hsp70 and p53 was assessed by coimmunoprecipitation. HeLa cells expressing FLAG-Hsp70 were washed 3 times with 1× phosphate-buffered saline (PBS) and then scraped into ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, containing 100 mM KCl, 2 mM EDTA, 1% Igepal CA-630 [octylphenoxypolyethoxyethanol], 1 mM phenylmethylsulfonyl fluoride, and 10 μg/ml leupeptin). Cells were lysed by being passed through a 23-gauge needle 3 to 5 times, incubated at 4°C for 30 min on a tumbling mixer, and then centrifuged at 11,000 × g for 10 min at 4°C. The supernatant was recovered, and the protein concentration measured using the RC DC protein assay kit (Bio-Rad) according to the manufacturer's protocol. Complexes containing p53 were recovered by incubating this lysate (1 mg) with anti-p53 antibodies (3 μg; Abcam) on a tumbling mixer overnight at 4°C before adding it to 50 μl of packed protein A/G resin (Pierce) equilibrated in lysis buffer. After incubation on a tumbling mixer for 2 h at room temperature, beads were washed twice with lysis buffer. Copurifying FLAG-Hsp70 was detected by Western blotting.

Ribonucleoprotein immunoprecipitation.

Ribonucleoprotein (RNP) complexes containing specified protein components were purified from crude HeLa cell lysates by immunoprecipitation (RNP-IP) essentially as described previously (43) with the following modifications. The antibodies used included control mouse IgG (BD Pharmingen), as well as anti-Hsp70 and anti-AUF1 antibodies (described above). The relative levels of specific mRNAs copurifying with immunoprecipitated RNPs were quantified by RT-qPCR using the iScript one-step RT-PCR SYBR green kit (Bio-Rad) with the primer sets listed in Table S2 in the supplemental material. The relative levels of VEGF and Cox-2 mRNAs were calculated from threshold cycle (C_T) numbers after normalization to endogenous GAPDH mRNA abundance using the 2^ΔΔCT method. Each data point was taken as the mean ± SD from quadruplicate qRT-PCRs for each RNA sample.

Biotin-RNA pulldown assays.

Cytoplasmic extracts containing FLAG-Hsp70 were generated by first transfecting HeLa cells with plasmid pCMV-FLAG-Hsp70wt using the Attractene reagent (Qiagen). Forty-eight hours after transfection, cells were washed twice in PBS before being scraped into buffer L (10 mM Tris [pH 7.5] containing 100 mM KCl, 2.5 mM MgCl₂, 2 mM DTT, 1% Igepal CA-630, 1 μg/ml leupeptin, 1 μg/ml pepstatin A, and 0.1 mM phenylmethylsulfonyl fluoride) and then disrupted using a Dounce homogenizer and centrifuged at 1,000 × g for 10 min to pellet nuclei. The protein concentration of the supernatant was quantified using the Bio-Rad protein assay kit. Selected fragments of VEGF cDNA were amplified by PCR downstream from T7 promoter sequences from IMAGE clone 6006890 using primers listed in Table S1 in the supplemental material. These cDNA fragments were used to program in vitro transcription reactions using the Ambion MAXIScript kit incorporating UTP and biotin-16-UTP (Roche) at a 9:1 ratio. RNA products were quantified by measurement at OD₂₆₀.

Biotin-RNA pulldown reactions were performed using a modification of the procedure of Wang et al. (44). Protein extract (200 μg) was mixed with 20 pmol biotin-RNA in 1.5 ml buffer L containing 80 U RNasin (Promega) and 0.2 mg/ml yeast tRNA (Sigma) and incubated for 30 min at room temperature. Streptavidin-agarose beads (Sigma) were then added, and the samples incubated for a further 30 min at room temperature on a tumbling mixer. Following incubation, beads were pelleted by centrifugation, washed four times with 1× PBS, and then suspended in 2× SDS sample buffer and boiled for 5 min to dissociate protein-RNA complexes. FLAG-Hsp70 copurifying with biotin-RNA substrates was detected by Western blotting.

Bioinformatics analyses.

Nucleotide sequences for mRNAs encoding human Hsp70 family members were retrieved from the National Center for Biotechnology Information website (www.ncbi.nlm.nih.gov) using the Entrez utility. Applications at the San Diego Supercomputer Center Biology Workbench (http://workbench.sdsc.edu) were used to translate each sequence (Sixframe), as well as for pairwise (Align) and global (Clustal W) sequence alignments.

Statistics.

Comparisons of mRNA levels and decay kinetics were done using the unpaired t test. Differences yielding a P value of <0.05 were considered significant.

RESULTS

Hsp70 binds specifically and with high affinity to U-rich RNA substrates.

Quantitative studies of Hsp70 binding in vitro are complicated by many factors, including its extreme conservation among nearly all extant species. As such, preparations of the recombinant human protein from any host cell model can include associated nucleotide cofactors (ATP or ADP), as well as host cochaperone proteins and peptide or nucleic acid ligands. Since previously published measurements of human Hsp70 binding to RNA probes were performed using native preparations of the protein from E. coli, they may have been influenced by the presence of copurifying factors (17). To address this problem, we adapted a strategy to purify His₆-tagged Hsp70 under denaturing conditions and then refolded the protein on-column prior to elution. Using this method, soluble monodisperse human His₆-Hsp70 was isolated from E. coli with high purity (Fig. 1A and B), permitting assessment of its RNA-binding properties under conditions in which minimal copurifying elements were present.

EMSAs detected two closely migrating complexes containing His₆-Hsp70 and a canonical ARE sequence derived from TNF-α mRNA (ARE[38]) at protein concentrations as low as 2 nM (Fig. 1C, left), with free RNA largely depleted at 200 nM protein. In contrast, His₆-Hsp70 did not bind a non-ARE substrate of similar length derived from rabbit β-globin mRNA (Rβ) (Fig. 1C, right), confirming that Hsp70 can selectively form stable complexes with an ARE-containing RNA target. Findings from EMSA experiments were complemented by fluorescence anisotropy-based assays, permitting quantitative assessment of binding under true equilibrium conditions. Using this technique, His₆-Hsp70 binding to the Fl-ARE[38] RNA substrate resolved a K_d of 12 ± 2 nM (mean ± SD) (Fig. 1D, closed circles), significantly lower than the K_d previously reported for His₆-Hsp70 purified under native conditions (17) and approaching the low nM range observed for other ARE-BPs, including tristetraprolin and AUF1 (45, 46). Robust ARE binding by His₆-Hsp70 purified under denaturing conditions also demonstrates that nucleotide or protein cofactors are not required for this activity. Although EMSAs detected two closely migrating His₆-Hsp70–ARE complexes, resolution of the anisotropy isotherm to a single-site binding model suggests that Hsp70 may bind this RNA target through multiple conformationally distinct yet thermodynamically equivalent binding modes.

In addition to measuring binding affinity directly, fluorescence anisotropy-based assays were also used to evaluate the nucleic acid binding specificity of His₆-Hsp70. First, His₆-Hsp70 showed minimal binding to the Fl-ARE[38] DNA substrate (Fig. 1D, open circles), indicating that the nucleic acid binding activity of this protein is probably limited to RNA targets. Second, the protein did not bind the Fl-Rβ substrate at concentrations up to 500 nM (data not shown), consistent with the EMSA results (Fig. 1C). Finally, competition assays demonstrated that in addition to Rβ RNA, homopolymeric single-stranded poly(C) and poly(A) RNAs were also unable to compete for His₆-Hsp70 binding to the Fl-ARE[38] substrate, as was poly(I·C) duplex RNA (Fig. 1E). In contrast, unlabeled ARE[38] RNA efficiently competed for Fl-ARE[38] RNA binding to His₆-Hsp70, as did the poly(U)-containing substrate U[32], suggesting that Hsp70 may also interact effectively with noncanonical ARE sequences, which lack AUUUA motifs but remain rich in uridylate residues. These relaxed sequence requirements for RNA binding are similar to those observed for AUF1 and HuR (47, 48) and suggest that the number of potential Hsp70 binding targets among the cellular mRNA population may be very large.

RNA binding by Hsp70 is mediated by sequence-specific contacts to multiple protein domains.

Hsp70 proteins do not contain canonical RNA-binding modules, such as RNA recognition motifs, K-homology domains, or zinc fingers. Rather, they consist of an N-terminal ATP-binding domain followed by a peptide-binding domain and a short C-terminal extension ending with an EEVD sequence that is targeted by selected cochaperone proteins (20). To identify Hsp70 domains responsible for its RNA-binding activity, we generated and purified a series of His₆-Hsp70 deletion mutant proteins (Fig. 2A and B). By EMSA, both the ATP-binding (amino acids 1 to 385) and peptide-binding (amino acids 386 to 613) domains of Hsp70 [mutant proteins His₆-Hsp70(1-385) and His₆-Hsp70(386-613)] formed single complexes with the ARE[38] RNA (Fig. 2C). However, RNPs formed between the ARE[38] RNA and these domains appeared weaker than those formed with full-length Hsp70 (Fig. 1C), since significant depletion of free RNA was not observed, even at 500 nM protein. The dissociation constants derived from fluorescence anisotropy experiments indicated that the Fl-ARE[38] substrate bound to the ATP- and peptide-binding domains of Hsp70 individually 5- to 6-fold more weakly than to the full-length protein (Fig. 2D and Table 1). In contrast, the ARE-binding activity of His₆-Hsp70 was not significantly affected by deletion of the EEVD sequence either with (amino acids 1 to 637) or without (amino acids 1 to 613) the C-terminal domain. These data indicated that both the ATP- and peptide-binding domains of Hsp70 are required for optimal ARE-binding activity but that sequences C terminal to these modules do not contribute to RNP complex stability.

Fig 2.

RNA-binding activity of Hsp70 domain mutants. (A) Organization of wild-type (wt) Hsp70 protein is shown at top. Domains included in Hsp70 protein mutants used in this study are indicated below. (B) Coomassie blue-stained SDS-PAGE of purified His₆-Hsp70 and mutants, with molecular mass markers (in kDa) shown at left. (C) EMSAs of Hsp70 ATP-binding (left) or peptide-binding (right) domains with ³²P-labeled ARE[38] or Rβ RNA substrates. (D) Fluorescence anisotropy assays measuring the affinity of His₆-Hsp70 mutant proteins for the Fl-ARE[38] RNA substrate. All isotherms were resolved by single-site binding models, yielding dissociation constants listed in Table 1. (E) RNA competition assays for Fl-ARE[38] binding to His₆-Hsp70(1-385) and His₆-Hsp70(386-613) performed as described in the legend for Fig. 1E, except that higher protein concentrations were used [His6-Hsp70(1-385), 100 nM; His₆-Hsp70(386-613), 60 nM] because of the weaker affinity of these proteins for the ARE substrate. Points represent the means ± SD of three separate trials.

Table 1.

Affinities of Hsp70 and selected deletion mutants for the Fl-ARE[38] RNA substrate

Protein	Kd (nM)a	ΔG (kcal · mol−1)b	n
His6-Hsp70wt	13 ± 2	−10.7	5
His6-Hsp70(1-637)	18 ± 1	−10.6	3
His6-Hsp70(1-613)	16 ± 5	−10.6	3
His6-Hsp70(1-385)	96 ± 12	−9.6	3
His6-Hsp70(386-613)	76 ± 8	−9.7	3

^aEquilibrium dissociation constants were resolved using fluorescence anisotropy-based assays as described in Fig. 1 and 2 legends and are given as the means ± SD of n independent experiments.

^bFree energy of protein binding to the Fl-ARE[38] substrate was calculated using ΔG = −RTlnK.

The localization of ARE-binding activity to two distinct domains of Hsp70 prompted two additional questions. First, is the RNA-binding activity contributed by each domain sequence specific? The non-ARE Rβ substrate showed no detectable binding to either the ATP- or peptide-binding domains of Hsp70 in EMSAs (Fig. 2C) or anisotropy-based binding assays (data not shown), and furthermore, only the U[32] and ARE[38] substrates effectively displaced binding of either Hsp70 domain to the Fl-ARE[38] RNA in competition assays (Fig. 2E). Second, does the weakened ARE-binding activity observed for individual Hsp70 domains result from impaired protein folding? This was tested by measuring the changes in protein fluorescence as selected His₆-Hsp70 proteins were unfolded across titrations of GdnHCl (Fig. 3), permitting the calculation of thermodynamic parameters describing the folding stability of each protein (Table 2). While the free energy of denaturation for the ATP-binding domain (amino acids 1 to 385) was slightly less sensitive to GdnHCl (m_eq) than for either the peptide-binding domain (amino acids 386 to 613) or full-length His₆-Hsp70 protein, no significant differences were observed among the extrapolated free energies of protein unfolding in the absence of denaturant (ΔG_uw) for any protein tested. Together, these data demonstrate that both the ATP- and peptide-binding domains are stably folded structures that individually contribute relatively weak RNA-binding activity with preferences for U-rich RNA substrates.

Fig 3.

Chemical denaturation of His₆-Hsp70 and selected deletion mutant proteins. (A) Blank-corrected fluorescence emission spectra (λ_ex = 270 nm) of His₆-Hsp70wt in the absence or presence of 1 M, 2 M, or 3 M GdnHCl. The arrow highlights the decrease in emission near 325 nm associated with increasing denaturant concentration. (B to D) Relative changes in fluorescence emission at 325 nm from His₆-Hsp70wt (B), His₆-Hsp70(1-385) (C), and His₆-Hsp70(386-613) (D) as a function of GdnHCl concentration. Thermodynamic parameters describing the folded stability of each protein were resolved by nonlinear regression using equations 2 and 3 as described in Materials and Methods and are listed in Table 2.

Table 2.

Thermodynamic parameters describing GdnHCl-induced unfolding of His₆-Hsp70 and selected deletion mutant proteins^a

Protein	ΔGuw (kcal · mol−1)	meq (kcal · mol−1 · M−1)
His6-Hsp70wt	2.86 ± 0.33	1.85 ± 0.29
His6-Hsp70(1-385)	2.58 ± 0.19	1.48 ± 0.17
His6-Hsp70(386-613)	2.52 ± 0.15	1.90 ± 0.13

^aThe free energy of protein unfolding in the absence of denaturant (ΔG_uw) and the sensitivity of protein-folding free energy to GdnHCl (m_eq) were resolved from F₃₂₅ versus [GdnHCl] data sets (Fig. 3) using equations 2 and 3 and are expressed as the means ± SD of three independent experiments.

Hsp70 stabilizes an ARE-containing reporter mRNA in cells.

Given that a major function of AREs is to control mRNA turnover rates, we assessed the effect of Hsp70 on the decay kinetics of a reporter transcript containing the high-affinity ARE[38] binding site in HeLa/Tet-Off cells (Fig. 4A). Hsp70 is robustly expressed in these cells, consistent with the constitutive overexpression of Hsp70 observed in many clinical tumor samples and cultured cancer cell lines (24, 30, 33). An siRNA targeting the coding sequence of HSPA1A mRNA potently suppressed Hsp70 protein expression in HeLa/Tet-Off cells without silencing the expression of the closely related family member Hsc70 (Fig. 4B). In Dox time course assays, βG mRNA lacking an ARE was extremely stable (Fig. 4C, left), decaying with a half-life of >10 h (n = 3) in cells transfected with either siHsp70 or a control siRNA (siControl). Insertion of the ARE[38] sequence into the 3′UTR of βG mRNA (βG-ARE[38]) accelerated mRNA decay by approximately 10-fold in siControl-transfected cells (Fig. 4C, right), yielding a half-life of 1.10 ± 0.14 h (n = 4), consistent with previous measurements of βG-ARE[38] mRNA turnover in cells lacking siRNA (40). However, in cells cotransfected with siHsp70, the βG-ARE[38] mRNA decay rate was enhanced by a further 2-fold, yielding a half-life of 0.54 ± 0.09 h (n = 5, P = 0.0001 versus results for siControl).

Fig 4.

Hsp70 stabilizes an ARE-containing reporter mRNA. (A) Schematic of the rabbit βG reporter mRNA. Introns are indicated by thin lines and exons with thick boxes. The ARE[38] sequence was inserted downstream from the translation termination codon to generate βG-ARE[38] mRNA. (B) Western blots of Hsp70 and Hsc70 in whole-cell lysates from untransfected HeLa/Tet-off cells (ut) or cells transfected with siHsp70 or a control siRNA (siControl) (left). Panels at right show rescue of Hsp70 expression in siHsp70-transfected cells by cotransfecting plasmid pCMV-FLAG-Hsp70sm. Ectopically expressed FLAG-Hsp70 was detected using both anti-FLAG and anti-Hsp70 antibodies, with GAPDH as a loading control in all cases. (C) Representative plots of βG-wt (left) and βG-ARE[38] mRNA decay resolved from Dox time courses performed as described in Materials and Methods. HeLa/Tet-Off cells were cotransfected with reporter mRNAs and either siControl or siHsp70 as indicated. For Hsp70 rescue experiments, cells were cotransfected with siHsp70 and plasmid pCMV-FLAG-Hsp70sm. Points indicate the means ± SD of four separate qRT-PCRs for each RNA sample. Nonlinear regression to single exponential decay functions yielded first-order mRNA decay constants and associated half-lives. Average decay constants measured across replicate independent experiments are given in the text.

The observation that suppression of Hsp70 expression accelerates the decay of βG-ARE[38] mRNA suggests that Hsp70 represents a novel ARE-binding, mRNA-stabilizing activity. However, since Hsp70 encompasses a family of closely related proteins, it is likely that mRNAs encoding multiple homologues may be targeted by this siRNA. For example, HSPA1B mRNA contains a perfectly conserved binding site for siHsp70 (see Fig. S1 in the supplemental material) but also encodes a protein identical to HSPA1A. The proteins encoded by the HSPA1L and HSPA2 genes share 89% and 84% identity, respectively, with HSPA1A, and these mRNAs also contain sequences with significant complementarity to the siHsp70-targeted site. Conversely, HSPA6 and Hsc70 mRNAs, which encode proteins sharing 82% and 86% identity with HSPA1A, respectively, do not include a likely siHsp70 target site. Consistent with divergence in this region, Hsc70 levels were not suppressed in siHsp70-transfected HeLa/Tet-Off cells (Fig. 4B, left). To confirm that ARE-directed mRNA stabilization could be mediated by a specific Hsp70 variant, we expressed an siRNA-resistant HSPA1A cDNA (see Fig. S1 in the supplemental material) downstream from the FLAG epitope, which restored Hsp70 expression when cotransfected with siHsp70 in HeLa/Tet-Off cells (Fig. 4B, right). In this Hsp70-rescued cell background, βG-ARE[38] decayed with a half-life of 1.14 ± 0.13 h (n = 3) (Fig. 4C, right), similar to its half-life in siControl-transfected cells but significantly stabilized relative to its half-life in cells transfected with siHsp70 alone (P = 0.0002). These data demonstrate that a reporter mRNA containing a high-affinity Hsp70 binding site is destabilized in cells lacking Hsp70 but can be restabilized by expression of the HSPA1A-encoded member of the Hsp70 protein family.

Hsp70 binds and stabilizes endogenous ARE-containing transcripts in cells.

For Hsp70 to be a physiologically relevant mRNA-stabilizing factor, it must interact with and inhibit the decay of endogenous substrate transcripts. To this end, we tested the ability of Hsp70 to bind and regulate the expression of VEGF and Cox-2 mRNAs in HeLa cells. Both transcripts encode proteins that are frequently overexpressed in aggressive neoplasms but also exhibit divergent sequence features in their 3′UTRs. Cox-2 mRNA (GenBank accession number NM_000963) includes a canonical ARE immediately downstream from the translation termination codon, while the VEGF 3′UTR (GenBank accession number NM_001025370) does not contain overlapping AUUUA-like ARE motifs but, instead, has three heterogeneous AU-rich domains dispersed along its length. Both VEGF and Cox-2 mRNAs were significantly enriched in RNP-IPs programmed with anti-Hsp70 antibodies in comparison to the results using nonspecific IgG controls (Fig. 5A), demonstrating that endogenous Hsp70 can associate with these transcripts. Both mRNAs were also enriched in RNP-IPs using anti-AUF1 antibodies, a positive-control result consistent with previous findings that these transcripts are each targeted by AUF1 (49, 50). In contrast, the non-ARE-containing mRNA encoding ribosomal protein l13a (rpl13a) was not enriched in RNP-IPs containing either anti-Hsp70 or anti-AUF1 antibodies.

Fig 5.

Binding and stabilization of endogenous ARE-containing mRNAs by Hsp70. (A) RNP-IP reaction mixtures from HeLa cell lysates were programmed with nonspecific IgG, anti-Hsp70, or anti-AUF1 antibodies. Levels of VEGF, Cox-2, and rpl13a mRNA recovered were quantified by qRT-PCR and normalized to the level of GAPDH mRNA (means ± SD of four reactions; *, P < 0.01 versus results with control IgG). Independent replicate experiments yielded similar results. (B) Western blots of Hsp70 in whole-cell lysates from HeLa cells transfected with siHsp70 versus results for transfection with control siRNA. (C) Steady-state VEGF and Cox-2 mRNA levels measured by qRT-PCR in HeLa cells transfected with either siControl or siHsp70. Bars show the means ± SD of three independent samples normalized to GAPDH mRNA (*P < 0.01 versus results for transfection with siControl). (D) ActD time course assays measuring VEGF and Cox-2 mRNA decay in HeLa cells transfected with siControl or siHsp70. Sample errors and regression analyses are as described in the Fig. 4 legend. Average mRNA half-lives from replicate independent experiments are given in the text.

To determine whether Hsp70 binding to VEGF and Cox-2 mRNAs influences their expression, steady-state levels of each transcript were measured in HeLa cells transfected with control or Hsp70-targeted siRNA. qRT-PCR analyses showed that VEGF mRNA levels were decreased by 52% in siHsp70-transfected cells, while Cox-2 mRNA levels were 33% lower when Hsp70 was depleted (Fig. 5C). Next, ActD time course assays assessed whether diminution of VEGF and Cox-2 mRNAs in Hsp70-depleted cells was a result of accelerated mRNA decay (Fig. 5D). In siControl-transfected cells, VEGF mRNA decayed with a half-life of 4.01 ± 0.35 h (n = 4). In contrast, the decay rate of this transcript was 2-fold faster in cells where Hsp70 expression was suppressed (t_1/2 = 2.05 ± 0.32 h, n = 5; P < 0.0001 versus results for siControl). Similarly, the rate of Cox-2 mRNA turnover was significantly faster in cells transfected with siHsp70 (t_1/2 = 2.83 ± 0.29 h, n = 4) than with control siRNA (t_1/2 = 4.42 ± 0.42 h, n = 4; P = 0.0008 versus results for siHsp70). Together, these data show that Hsp70 can bind VEGF and Cox-2 mRNAs in HeLa cells and that this interaction increases their steady-state levels by stabilizing each transcript.

Control of mRNA decay by Hsp70 is independent of its protein chaperone function.

Hsp70 is best known as a protein chaperone, an activity that is coupled to nucleotide hydrolysis and association with selected cochaperone proteins (19, 20). To determine whether posttranscriptional gene regulation by Hsp70 was coupled to its protein chaperone cycle, we measured the effect of the Hsp70 chaperone inhibitor 2-phenylethynesulfonamide (PES) on its RNA-binding and mRNA-stabilizing activities. PES selectively binds Hsp70 and disrupts interactions with cochaperones and peptide substrates (51). Using coimmunoprecipitation experiments, we observed that PES strongly suppressed FLAG-Hsp70 binding to p53 in cisplatin-treated HeLa cells (Fig. 6A), similar to the effect of this compound in other cell models (51). However, in fluorescence anisotropy-based binding assays (Fig. 6B), His₆-Hsp70wt bound the Fl-ARE[38] substrate with a K_d of 10.2 ± 1.8 nM (n = 3) in the presence of PES, an affinity indistinguishable from the results of reactions performed in the presence of vehicle (DMSO) alone (K_d = 9.5 ± 2.6 nM, n = 3) (data not shown). Furthermore, the addition of PES did not affect the mRNA-stabilizing activity of endogenous Hsp70 in HeLa cells. The decay kinetics of VEGF (t_1/2 = 3.77 ± 0.21 h, n = 4) and Cox-2 (t_1/2 = 4.37 ± 0.35 h, n = 4) mRNAs (Fig. 6C) in PES-treated cells were indistinguishable from the mRNA turnover rates measured in cells containing functional Hsp70 (Fig. 5D). Consistent with this observation, PES treatment did not significantly alter the steady-state levels of either VEGF or Cox-2 mRNAs (Fig. 6D). These data suggest that the RNA-binding and mRNA-stabilizing activities of Hsp70 are independent of its protein chaperone function, since neither was affected by treatment with the chaperone inhibitor.

Fig 6.

Effect of the Hsp70 chaperone inhibitor PES on RNA-binding and mRNA-stabilizing activities. (A) HeLa cells transfected with siHsp70 and plasmid pCMV-FLAG-Hsp70sm were treated with or without PES (20 μM) for 1 h prior to addition of cisplatin (50 μM) and then incubated at 37°C for an additional 8 h. Top panels show expression of FLAG-Hsp70 and p53 with GAPDH as a loading control. Bottom panel shows FLAG-Hsp70 recovered in immunoprecipitation reactions programmed with anti-p53 antibodies. (B) Fl-ARE[38] RNA substrate binding to His₆-Hsp70 measured by fluorescence anisotropy after preincubating the protein in PES (20 μM) for 30 min. Data are fit to a single-site binding model. (C) ActD time course assays measuring VEGF and Cox-2 mRNA decay in HeLa cells following treatment with PES (20 μM, 2 h). Average mRNA half-lives from replicate independent experiments are given in the text. (D) Steady-state levels of VEGF and Cox-2 mRNAs in HeLa cells treated with or without PES (20 μM, 2 h) measured by qRT-PCR. Bars represent the means ± SD of four independent samples normalized to GAPDH mRNA.

Hsp70 stabilizes VEGF mRNA by binding to a proximal AU-rich 3′UTR domain.

Given the limited information available on the RNA substrate selectivity of Hsp70, it is conceivable that this protein may interact with any U-rich domains within the cellular mRNA population. To evaluate whether Hsp70 can discriminate between subsets of U-rich RNA targets, we surveyed an extended RNA substrate for Hsp70-binding sites using the VEGF mRNA 3′UTR as a model. The VEGF transcript does not contain a canonical ARE with overlapping AUUUA motifs but does encode several AU-rich tracts that could potentially be targeted by Hsp70 (Fig. 7A). One AU-rich tract near the 3′ end of VEGF mRNA has already been identified as a target for the mRNA-stabilizing factor HuR (52, 53). We constructed 10 biotin-tagged RNA probes of approximately 300 nucleotides each that cumulatively spanned the coding and 3′UTR sequences of VEGFA variant 6 mRNA in an overlapping fashion (Fig. 7A). Four of these probes (fragments 4, 8, 9, and 10) included significant AU-rich subdomains and, as such, were considered likely Hsp70 targets. However, in biotin-RNA pulldown assays with cytoplasmic extracts from HeLa cells expressing FLAG-Hsp70, only fragment 4 consistently displayed robust binding to this protein (Fig. 7B). Although weak Hsp70 binding to some other RNA probes was also observed, these data demonstrate that Hsp70 will not bind indiscriminately to all U- or AU-rich sequences but, rather, must have a more stringent set of RNA substrate requirements.

Fig 7.

Hsp70 binding determinants on VEGF mRNA. (A) Schematic of VEGFA variant 6 mRNA. AU-rich domains common to the 3′UTRs of all known VEGFA mRNA variants are indicated above (open boxes). Sequences encoded by biotin-tagged RNA substrates are shown below (solid boxes). (B) FLAG-Hsp70 proteins recovered from HeLa cytoplasmic lysates by biotin-RNA pulldown assays detected by Western blotting. (C) EMSAs performed using indicated ³²P-labeled VEGF RNA substrates and titrations of recombinant His₆-Hsp70. Open arrowheads demarcate the positions of unbound probes, while solid arrowheads indicate bands corresponding to His₆-Hsp70 RNP complexes.

EMSAs were then used to determine whether Hsp70 binds directly to the VEGF mRNA fragment 4 substrate. Fragment 1, which appeared to retain a small amount of FLAG-Hsp70 in biotin-RNA pulldown reactions (Fig. 7B), displayed no binding to recombinant His₆-Hsp70. In contrast, EMSA reaction mixtures containing His₆-Hsp70 and VEGF mRNA fragment 4 resolved two distinct RNP complexes (Fig. 7C). The first Hsp70-shifted complex dominated in reaction mixtures containing as little as 5 nM His₆-Hsp70, while a second RNP complex appeared at higher protein concentrations. Consistent with results from biotin-RNA pulldown experiments, no RNP complexes were observed in reaction mixtures containing His₆-Hsp70 and VEGF mRNA fragment 9, despite the AU-rich domain present in this RNA substrate. The fragment 9 probe migrated as two bands even in the absence of added protein, probably as a result of distinct conformational subpopulations.

The Hsp70-induced changes in mobility of the VEGF mRNA fragment 4 substrate were very modest, probably because of the large size of this RNA ligand (300 nucleotides [nt]) relative to the size of the protein. To determine whether the embedded AU-rich domain of fragment 4 was responsible for binding Hsp70, as well as improving the resolution of these RNP complexes by EMSA, a 148-nt RNA substrate spanning this AU-rich subdomain, termed fragment 4A, was constructed (see Fig. S2 in the supplemental material). Similar to fragment 4, His₆-Hsp70 binding to fragment 4A yielded two distinct RNP complexes in a concentration-dependent manner (Fig. 7C), indicating that Hsp70 can interact with VEGF mRNA by directly associating with this proximal AU-rich sequence.

The functional significance of selective Hsp70 binding to the VEGF mRNA fragment 4 AU-rich domain was tested by measuring the decay kinetics of βG reporter mRNAs containing various segments of the VEGF 3′UTR in HeLa/Tet-Off cells (Fig. 8). In cells transfected with control siRNA, the complete VEGF 3′UTR dramatically accelerated turnover of the βG reporter mRNA, indicating that binding sites for mRNA-destabilizing factors reside within this sequence (Table 3). However, cotransfection of siHsp70 decreased the half-life of the βG-VEGF 3′UTR transcript by an additional factor of two, consistent with the effect of suppressing Hsp70 expression on the decay of endogenous VEGF mRNA (Fig. 5). A reporter transcript containing only the Hsp70-binding domain of VEGF mRNA within the βG 3′UTR (βG-fgt 4) was similarly destabilized in cells lacking Hsp70. Conversely, degradation of reporter mRNAs containing only VEGF 3′UTR sequences downstream from the Hsp70 binding site in fragment 4 (βG-3′UTR Δ4) or the distal AU-rich domain (βG-fgt 9) was not significantly affected by Hsp70, although their short half-lives also indicate susceptibility to mRNA-destabilizing activities. Together, these data demonstrate that Hsp70 directly and specifically binds to the proximal AU-rich sequence within the VEGF 3′UTR and that this association is responsible for stabilizing mRNAs containing this sequence.

Fig 8.

Hsp70-dependent stabilization of βG-VEGF reporter mRNAs. (A) Schematic of VEGF 3′UTR segments cloned into the βG reporter vector. Fragments 4 and 9 are identical to those shown in Fig. 7. (B) Western blots of Hsp70 in lysates from HeLa/Tet-Off cells transfected with siHsp70 versus results for transfection with control siRNA. (C) Dox time course assays resolving the decay kinetics of βG reporter mRNAs in HeLa/Tet-Off cells cotransfected with siControl or siHsp70. Error bars and regression solutions are as described in the Fig. 4 legend. Average mRNA half-lives from replicate independent experiments are listed in Table 3.

Table 3.

Regulated decay of reporter mRNAs by Hsp70

mRNA	siRNA	t1/2 (h)a	n	P vs result for siControlb
βG-wt	siControl	>10	3
	siHsp70	>10	3	NS
βG-VEGF 3′UTR	siControl	1.41 ± 0.14	4
	siHsp70	0.74 ± 0.11	4	0.0003
βG-3′UTR Δ4	siControl	1.13 ± 0.27	4
	siHsp70	0.93 ± 0.27	4	NS
βG-fgt 4	siControl	2.83 ± 0.14	4
	siHsp70	1.42 ± 0.37	3	0.0008
βG-fgt 9	siControl	2.20 ± 0.34	4
	siHsp70	2.24 ± 0.41	4	NS

^aTurnover kinetics of βG reporter mRNAs were measured in HeLa/Tet-Off cells using Dox time course assays as described in Materials and Methods and Fig. 8 legend. Listed mRNA half-life values represent the means ± SD from n independent time course experiments.

^bNS, no significant difference between samples (P > 0.05).

VEGF mRNA is stabilized concomitant with Hsp70 accumulation in HL60 cells recovering from heat shock.

The results of the experiments described above show that Hsp70 can bind and stabilize selected mRNA substrates containing AU-rich sequences, which may enhance the levels of some mRNAs that encode protumorigenic factors when Hsp70 is constitutively expressed. However, this model would also predict that stabilization of substrate mRNAs should accompany the normal transient increases in cellular Hsp70 observed following thermal stress. To test this prediction, we first examined the temporal control of Hsp70 protein accumulation following heat shock in the promyelocytic leukemia cell line HL60. Unlike HeLa cells (Fig. 5B), Hsp70 protein levels were very low in untreated HL60 cells and were not significantly affected by incubation at 42°C for 1 h (Fig. 9A). If heat-shocked cultures were allowed to recover at 37°C, Hsp70 protein levels remained minimal for several hours but were then dramatically increased starting 8 h after cessation of the thermal stress. A similar temporal pattern of Hsp70 protein accumulation was reported following heat shock in lymphoblastoid cell models (54). Next, ActD time course assays were used to measure the decay kinetics of VEGF mRNA at selected time points throughout this heat shock and recovery regimen (Fig. 9B). In untreated HL60 cells, VEGF mRNA decayed with a half-life of 1.62 ± 0.38 h (n = 4), remarkably similar to the rapid turnover rate of this transcript in HeLa cells following the suppression of endogenous Hsp70 (above). Following heat shock, VEGF mRNA decay was not significantly altered after 1 h of recovery at 37°C (t_1/2 = 1.68 ± 0.48 h, n = 3), consistent with the absence of appreciable amounts of Hsp70 protein. However, following 8 h of recovery, VEGF mRNA was stabilized by approximately 2-fold (t_1/2 = 3.65 ± 0.18 h, n = 4; P < 0.0001 versus results for untreated cells and P = 0.0006 versus results for heat shock followed by 1 h of recovery). These data show that VEGF mRNA is stabilized concurrently with the accumulation of Hsp70 in HL60 cells recovering from heat shock and may reflect a more general role for this protein in posttranscriptional gene regulation following cell stress.

Fig 9.

(A) Western blot analysis of Hsp70 expression in HL60 cells prior to and at various times following application of thermal stress (42°C, 1 h) with or without recovery at 37°C. GAPDH protein levels were used as loading controls. (B) ActD time courses measuring VEGF mRNA decay kinetics in untreated HL60 cells or heat-shocked (hs) cells following recovery at 37°C for 1 or 8 h. Sample errors and regression analyses are as described in the Fig. 4 legend. Decay parameters averaged from replicate independent experiments are given in the text.

DISCUSSION

Although binding between Hsp70 and ARE-like RNA ligands was first reported over a decade ago (16, 17), neither the functional significance of these interactions nor evidence of their occurrence in a cellular environment had been described. Furthermore, little was known regarding the nucleic acid binding specificity of Hsp70 in solution. Here, we have shown that Hsp70 tightly binds a model ARE substrate in the absence of ancillary nucleotide cofactors (ADP/ATP) or protein cochaperones. High-affinity binding is specific for RNA versus DNA ligands (Fig. 1D) and is directed by interactions with U-rich sequences (Fig. 1E) that involve sequence-specific RNA contacts from both the ATP- and peptide-binding domains (Fig. 2). However, contributions to RNP stability from these protein moieties are not additive. While each individually binds the Fl-ARE[38] ligand with ΔG ≈ −9.6 kcal/mol, adding the second RNA-binding determinant only contributes ΔΔG ≈ −1 kcal/mol to binding energy (Table 1), suggesting that protein and/or RNA conformational restrictions may preclude optimal RNA contacts with both domains when tandemly arranged.

The results of RNP-IP experiments indicate that Hsp70 can associate with endogenous ARE-containing mRNAs encoding VEGF and Cox-2 (Fig. 5A). Biotin-RNA pulldown experiments identified the binding site within VEGF mRNA as a distinct AU-rich sequence in the proximal 3′UTR (Fig. 7B), while EMSAs using recombinant protein indicated that Hsp70 interacted directly with this RNA substrate. A general preference for U-rich RNA ligands is a common theme among many ARE-BPs. Like Hsp70, the proteins AUF1, HuR, and TIAR show little discrimination between bona fide ARE and generic poly(U) sequences (48, 55, 56). In contrast, stable association of TTP-related proteins requires direct contact with the interspersed adenylate residues typical of many AREs (45, 57). Although all ARE-BPs show preferences for U-rich sequences to some extent, recent surveys of cellular ARE-BP binding sites using RNP-IP or cross-linking and immunoprecipitation-based approaches reveal distinct but frequently overlapping mRNA subpopulations associating with these factors (58–60). Complicating this theme further is the frequently modular nature of mRNA 3′UTRs, which may contain multiple but distinct ARE-like sequences. The VEGF mRNA 3′UTR has three extended AU-rich tracts distributed along its length but, interestingly, only interacted with Hsp70 through the most upstream of these domains (Fig. 7), while the 3′-most AU-rich sequence is a functional target of HuR (52). These observations indicate that although similar in overall nucleotide composition, the VEGF mRNA ARE-like domains can efficiently discriminate between these binding proteins, likely involving embedded sequence motifs or local RNA structural determinants.

While demonstrating that Hsp70 binds directly to U-rich RNA substrates in vitro and can associate with VEGF and Cox-2 mRNAs in cells, this study is, to our knowledge, also the first to show that Hsp70 can stabilize ARE-containing mRNA targets in mammalian cells. A reporter mRNA containing a model ARE sequence in its 3′UTR was destabilized in HeLa cells when Hsp70 expression was suppressed but was restabilized when Hsp70 levels were ectopically restored (Fig. 4C). Similarly, transfection with Hsp70-targeted siRNA destabilized reporter transcripts containing an Hsp70-binding fragment of VEGF mRNA (Fig. 8, VEGF 3′UTR and fgt 4) but had no effect on βG-VEGF chimeric reporters that lacked the Hsp70-targeted 3′UTR sequence (3′UTR Δ4, fgt 9). Finally, suppression of Hsp70 destabilized endogenous VEGF and Cox-2 mRNAs, resulting in significant decreases in the steady-state levels of these transcripts in HeLa cells (Fig. 5). Recent findings from other groups also suggest an mRNA-stabilizing role for Hsp70 in the regulation of ENPP1 and SMAR1 mRNAs (61, 62). Although mRNA decay kinetics were not explicitly resolved in either case, Hsp70 binding to SMAR1 mRNA involved binding to a 5′UTR sequence, raising the possibility that mRNA stabilization is mediated by Hsp70 association with a range of mRNA sequence or positional contexts.

The role of Hsp70 in posttranscriptional gene regulation can be considered analogous to that of HuR in several ways. First, both target AREs or similar sequences. Second, both function to stabilize mRNA substrates. Third, both Hsp70 and HuR are inducible activities in most healthy cells. For Hsp70, induction is normally based on increased expression during or following cell stress. In contrast, HuR is generally constitutively expressed and principally localized to the nucleus but may, in response to a variety of cellular stresses, translocate to the cytoplasm, where it can bind and stabilize target mRNAs (reviewed in reference 63). The data presented in this study, however, also suggest some differences between the mRNA-stabilizing functions of Hsp70 and HuR. The observation that each protein recognizes a different AU-rich subdomain in VEGF mRNA (described above) indicates that they target similar but separate molecular determinants on RNA ligands. This divergence in RNA recognition suggests that acute stresses which induce Hsp70 expression or relocalize HuR protein probably regulate distinct but overlapping mRNA subpopulations through each factor. However, Hsp70 and HuR binding to VEGF mRNA at two disparate sites also raises the possibility that mRNA stabilization by these trans factors is additive for some mRNA substrates.

Transient accumulation of Hsp70 and other inducible heat shock proteins is a hallmark of the thermal stress response in many cell types and confers survival advantages by preventing the unfolding and aggregation of cellular proteins (22). However, the data presented in this study indicate that another role of Hsp70 in the stress response may be to stabilize a subset of cellular mRNA targets. The decay of VEGF mRNA in heat-shocked HL60 cells is consistent with this model, where mRNA turnover was rapid both in untreated cells and in those recovering at 37°C for 1 h after heat shock (Fig. 9). Stabilization of VEGF mRNA was only observed after prolonged (8 h) recovery at 37°C, concomitant with the accumulation of Hsp70 protein. Since many factors that control cell proliferation, survival, angiogenesis, and metastasis are encoded by ARE-containing mRNAs (64), the constitutive overexpression of Hsp70 observed in many advanced cancers may thus subvert this otherwise transient stress-induced mRNA-stabilizing mechanism. If this is true, then Hsp70 would join a growing number of ARE-BPs whose regulated expression is essential for preventing the exacerbation of tumorigenic phenotypes by aberrant stabilization of ARE-containing mRNAs. For example, elevated levels of the mRNA-stabilizing protein HuR are associated with higher tumor grades and poor patient outcomes in breast cancer (65, 66). Conversely, the expression of the mRNA-destabilizing protein tristetraprolin is suppressed in many tumor types (43), which increases the production of a variety of angiogenic and metastatic factors, including VEGF, urokinase plasminogen activator, and matrix metalloproteinase-1 (67, 68). Comprehensive identification of the cellular mRNA subpopulation targeted by Hsp70 will determine how constitutive enrichment of Hsp70 protein levels may also promote tumor progression or survival by stabilization of selected mRNA substrates.

Beyond highlighting the need to define the range and diversity of mRNA targets for Hsp70, the findings reported in this study prompt other intriguing questions. First, what is the mechanism by which Hsp70 binding stabilizes mRNA substrates? The trivial model may be by dynamic competition with mRNA-destabilizing factors for selected ARE target sites, particularly since ARE binding and stabilization of VEGF and Cox-2 mRNAs by Hsp70 was not affected by a chemical inhibitor of its protein chaperone activity (Fig. 6). However, following heat shock, it is also possible that this requires a more complex series of interactions involving cellular components directing mRNA triage and release from stress granules (reviewed in reference 69). Second, since Hsp70 is a member of a closely conserved family of proteins, could different members exhibit similar or divergent functions in posttranscriptional control of gene expression? Evidence for the former hypothesis was given by findings that Hsc70 binding to an ARE sequence in Bim mRNA was responsible for stabilizing that transcript in a pro-B cell model deprived of interleukin-3 (70). Interestingly, Hsp70 was unable to bind this mRNA target, suggesting that these proteins might recognize different transcript subpopulations. Hsp70 family members were also observed to induce different protumorigenic cellular phenotypes. While both HSPA1A (Hsp70) and HSPA2 (also known as Hsp70-2) enhanced the proliferation of both HeLa and MCF-7 breast cancer cells, HSPA1A contributed to cell cycle transit through the G₂/M checkpoint, while HSPA2 prevented arrest in G₁ (30). Also, substantially different changes in gene expression patterns were noted when cells were transfected with siRNAs targeting HSPA1A versus those targeting HSPA2. A final example is given by the yeast Hsp70 homologue Ssa1p, which targets an ARE-like sequence in MFA2 mRNA but accelerates decay rather than stabilizing the transcript (71).

In conclusion, we have demonstrated that Hsp70 can form direct, high-affinity RNP complexes with ARE substrates in vitro and associates with ARE-containing mRNAs in cells. The functional consequence of Hsp70 recruitment to these sequences is to stabilize the mRNA substrate. Given that (i) constitutive expression of Hsp70 is associated with increased tumor aggressiveness and (ii) the stability of many mRNAs that encode factors directing protumorigenic phenotypes is regulated by AREs, we propose that sustained elevation of Hsp70 may contribute to tumor progression by posttranscriptionally enhancing the expression of a subset of ARE-containing transcripts. Conceivably, this gene regulatory function could complement the well-characterized prosurvival effects of Hsp70 protein chaperone activities.