Heat Shock-Independent Induction of Multidrug Resistance by Heat Shock Factor 1

Abstract

The screening of two different retroviral cDNA expression libraries to select genes that confer constitutive doxorubicin resistance has in both cases resulted in the isolation of the heat shock factor 1 (HSF1) transcription factor. We show that HSF1 induces a multidrug resistance phenotype that occurs in the absence of heat shock or cellular stress and is mediated at least in part through the constitutive activation of the multidrug resistance gene 1 (MDR-1). This drug resistance phenotype does not correlate with an increased expression of heat shock-responsive genes (heat shock protein genes, or HSPs). In addition, HSF1 mutants lacking HSP gene activation are also capable of conferring multidrug resistance, and only hypophosphorylated HSF1 complexes accumulate in transduced cells. Our results indicate that HSF1 can activate MDR-1 expression in a stress-independent manner that differs from the canonical heat shock-activated mechanism involved in HSP induction. We further provide evidence that the induction of MDR-1 expression occurs at a posttranscriptional level, revealing a novel undocumented role for hypophosphorylated HSF1 in posttranscriptional gene regulation.

Heat shock factor 1 (HSF1) is the transcription factor responsible for the transcriptional response of vertebrate cells to different stresses including heat shock, ischemia, and aging, all of which have the accumulation of misfolded proteins as a central feature (32, 35). This response includes the induction of a set of highly conserved proteins, heat shock proteins (HSPs), that can act as cytoplasmic chaperones to buffer misfolded proteins and aid their refolding. Two distinct steps have been characterized in HSP induction by mammalian HSF1 in response to heat shock: acquisition of HSF1 DNA binding activity, which results from the conversion of HSF1 from an Hsp90-containing monomeric HSF1 complex to a homotrimeric form (2, 4, 28, 38, 41, 51), and acquisition of HSF1 transcriptional activity, which is correlated with its hyperphosphorylation at various serine/threonine residues (4, 13, 18, 20, 24, 38, 43, 48). Treatment of cells with certain chemicals such as indomethacin (13, 21, 24, 30) or ectopic overexpression of HSF1 (33, 38, 52) activates HSF1 trimerization and DNA binding activity without HSF1 hyperphosphorylation and HSP induction, indicating that these two steps can be uncoupled. In addition, HSF1 activation by heat shock is transitory as it is suppressed by attenuation mechanisms that return active HSF1 to its inactive form (1, 17, 48).

Although HSF1 is generally thought to be involved only in the response to heat shock or other cellular stresses, there is some indirect evidence from studies on animal development that suggests the existence of other important functions of HSF1 that are distinct from HSP induction and do not depend on heat shock or other major cellular stresses. For instance, under normal growth conditions, Drosophila HSF is required for oogenesis and early development, and these functions do not appear to be mediated through the induction of HSPs (22). An HSF1 null mutation in mice results in prenatal lethality, growth retardation, and female infertility, despite unaltered basal HSP expression (49). This suggests that in conditions under which there is no HSF1-dependent induction of HSPs, HSF1 can nonetheless play an important role in development, although the underlying mechanism is still not understood.

The acquisition of multidrug resistance poses a major obstacle to the success of cancer chemotherapy. One important mechanism by which cancer cells resist treatment with anthracyclins or vinca-alkaloids is the overexpression of the MDR-1 gene and its product P-glycoprotein (P-gp), an energy-dependent drug efflux pump. There have been a few reports demonstrating that endogenous P-gp expression could be transiently induced by heat shock (10, 34), suggesting that stress-activated HSF1 could be a regulator of the MDR-1 gene. This is supported by the finding that high-level ectopic overexpression of a constitutively active HSF1 mutant (c-HSF1, which lacks its regulatory domain) that can induce HSP expression in a heat shock-independent manner also induces MDR-1 expression in HeLa human cervical carcinoma cells (46). Although these observations establish the functionality of the heat shock element (HSE) present in the MDR-1 promoter, the biological importance of this mechanism is uncertain. Indeed, HSF1 activation of MDR-1 should be transient in nature due to the attenuation mechanisms that follow HSF1 activation by cellular stress (1, 17, 48). Moreover, in many cell lines (10), including HepG2 and all tested human cervical carcinoma cell lines (10, 29), heat shock was found to be ineffective in the activation of endogenous MDR-1. Thus, although activation of the classical heat shock response can in some cases induce the expression of MDR-1, this pathway is unlikely to play a significant role in the induction of multidrug resistance.

In the present work, we have screened retroviral cDNA expression libraries in human cultured U2-OS osteosarcoma cells to isolate genes that can confer resistance to long-term treatment by the anthracyclin doxorubicin. Two cDNAs that can confer constitutive doxorubicin resistance have been independently recovered from two different retroviral cDNA expression libraries, and both were shown to encode HSF1. Molecular analysis supports a heat shock (or cellular stress)-independent activity of hypophosphorylated HSF1 in conferring drug resistance, which, in the absence of HSP induction, correlates with the constitutive up-regulation of endogenous MDR-1 gene expression. In addition, analyses of the transcriptional activity as well as at the RNA level indicate that the induction of MDR-1 expression occurs at a posttranscriptional level, providing evidence for a new role of HSF1.

MATERIALS AND METHODS

Cell culture, infections, clonogenic assays, and transfections.

U2-OS cells were grown in 4.5 g/liter glucose containing Dulbecco's modified Eagle's medium, and HepG2 cells were grown in Dulbecco's modified Eagle's medium-F12 plus 1 mM aminoguanidine, each supplemented with 10% fetal calf serum and antibiotics, in 5% CO₂ at 37°C. Selections were performed in growth medium supplemented with 400 μg/ml and 1,500 μg/ml G418 (Life technologies) for neo expression and 100 to 150 U/ml hygromycin for hygro expression (Calbiochem). The U2-OS* cell population was obtained by cotransfection of plasmid pBabe-Bleo-EcoR (encoding the receptor for murine type C ecotropic retrovirus) and pBabe-puro into native U2-OS cells with puromycin selection at 1 μg/ml. Production of infectious recombinant retroviruses was performed by transient transfection of Phenix-eco3 or Bing packaging cells with the plasmids carrying the retroviral expression vector and harvesting of cell supernatants 2 days posttransfection. Cells were infected with viral supernatant in the presence of polybrene (8 μg/ml). For clonogenic assays, cells were seeded at 3 × 10⁵ per 6-cm-diameter culture plate the day before treatment, treated with doxorubicin (for 3 to 4 h, depending on the particular experiment) as indicated, passaged 3 days later at very low density, and further cultured for 12 days without changing the growth medium. At this time, cells were fixed and stained with a solution of crystal violet in methanol-acetic acid, and the numbers of clones per plate were counted. Results are given as the calculated total number of cell clones that would have resulted from the plating of all the treated cells or as the calculated cloning efficiency.

Transfections of cells were carried out with Lipofectamine 2000 reagent (Life Technologies) according to the manufacturer's instructions. For transient transfection assays with pMDR(−1202), 2 × 10⁵ cells were transiently cotransfected with 50 ng of plasmid pMDR1(−1202) and 50 ng of pRL-Tk (a Renilla luciferase-encoding reporter gene from Promega) and lysed 2 days later to measure luciferase activity of pMDR1(−1202) normalized to that of pRL-Tk with a dual-luciferase assay kit (Promega).

DNA constructs. (i) Construction of retroviral cDNA expression libraries.

Poly(A)⁺ mRNA was isolated from a 129/Sv-derived embryonal stem cell line, clone D3 (39), and from F9 embryonal carcinoma cells using a Dynabeads mRNA purification kit (Dynal). cDNAs were synthesized with EcoRI (5′) and XhoI (3′) linkers using a cDNA synthesis kit (Stratagene) and inserted by ligation into the pFbneo retroviral expression vector (Stratagene) opened by EcoRI and XhoI.

(ii) Construction of hHSF1 expressions vectors.

pFbneo-dn-hHSF1 and pFB-Neo-c-hHSF1 were constructed by insertion of an XhoI (in the 5′ untranslated region of human HSF1 [hHSF1])-XhoI (in the linker) HSF1 cDNA fragment excised from the corresponding pcDNA3 expression vector (a generous gift of Jinhui Wang, Chinese Academy of Sciences, Shanghai, China) into the SalI site of pFB-Neo. pFB-Neo-hHSF1 (wild type) and hHSF1-3F were constructed by insertion of an XhoI (in the 5′ untranslated region of hHSF1)-EcoRI (in the linker, downstream of a hemagglutinin flag) HSF1 cDNA fragment excised from the corresponding pBabe-puro expression vector (a generous gift of R. E. Kingston, Massachusetts General Hospital) between the SalI and EcoRI sites of pFB-Neo. pBabe-hygro-hHSF1-Δ360 and pBabe-hygro-hHSF1-Δ210 were constructed by insertion of a BglII-SalI cDNA fragment (synthesized by PCR amplification performed with Pfu DNA polymerase) (Stratagene) using pBabe-puro-hHSF1-Flag (44) as a template and the following oligonucleotides as primers: GAAGATCTCGAGATGGATCTGCCCGTGGGCCCC for the 5′ primer, TTAGTCGACTCACCCGGGACTCGCCTCCTCTAC for the Δ360 3′ primer, and TTAGTCGACTCAGGGGATCTTTCTCTTCACCCCCAG for the Δ210 3′ primer; the fragment was inserted into the pBabe-hygro expression vector opened by BamHI and SalI. pBabe-hygro-hHSF1 was constructed by insertion of a XhoI (in the 5′ UTR of hHSF1)-HindIII (just downstream simian virus 40 promoter) fragment excised from pBabe-puro-hHSF1 into pBabe-hygro vector opened by SalI (in the linker) and HindIII (downstream simian virus 40 promoter). All constructions were verified by sequencing.

(iii) Construction of pFB-neo-MDR-1.

pFB-Neo-MDR-1 was constructed by inserting a BamHI (in both the 5′ and 3′ the linker) fragment encompassing the complete human MDR-1 cDNA, excised from plasmid pSF-MDR (a kind gift of A. Laurand, Bergonié Institute, Bordeaux, France), into the BamHI site of pFB-neo.

(iv) Construction of pMDR1(−1202)mut.

Mutagenesis of HSE in pMDR1(−1202) was performed using a QuickChange XL Site-Directed Mutagenesis kit (Stratagene) and the oligonucleotides GCCAGAGCATGCCTCCTGGAAATTCAACC and TCCAGGAGGCATCCTCTGGCTTCCGTTGCAC.

Nucleic acids analyses.

Total RNA was extracted by direct lysis of the cells in the classical guanidium-thiocyanate lysis buffer, followed by purification of the RNA by centrifugation through a CsCl cushion. For nuclear and cytoplasmic fractionation, about 4 × 10⁷ cells were washed three times with cold phosphate-buffered saline (PBS) containing 2 mM (each) EDTA and EGTA, harvested by scraping, and lysed in 500 μl of lysis buffer containing 0.14 M NaCl, 10 mM Tris-HCl, pH 8.5, 1.5 mM MgCl2, 0.0625% NP-40, 10 mM dithiothreitol, and 200 U/ml RNasin for 10 min at 4°C. Following centrifugation at 500 × g for 5 min, the nuclear pellet was suspended in 300 μl of lysis buffer and centrifuged once again. The two supernatant fractions were pooled as the cytoplasmic fraction. RNA from the nuclear and cytoplasmic fractions was then extracted by the classical guanidium-thiocyanate method with centrifugation through a CsCl gradient. Northern blot analyses were performed with Hybond-N membranes (Amersham) according to the recommendations of the manufacturer after RNA electrophoresis through a 1% agarose gel containing 2.2 M formaldehyde. For Northern blot analysis of very-high-molecular-weight RNA, it was important to avoid both shearing stresses at all steps and the use of ethidium bromide before nucleic acid transfer to the membrane. α-³²P-labeled DNA probes were synthesized with a multiprime labeling kit (Radprime; Bio-Rad). For MDR-1 RNA analysis, we used either a purified BamHI (in the 5′ linker)-EcoRI (1,176 nt downstream from the initiation codon) fragment encompassing the 5′ part of the MDR-1 cDNA that was previously shown not to cross-hybridize with the MDR-2 gene (9) or, when noted, the complete hMDR-1 open reading frame. The probe encompassing a unique sequence of the MDR-1 gene intron 26 was synthesized by PCR using the oligonucleotides GGAGGATCCTCACAGTAAATATGCATAGAAG and CCTGCTCGAGCGCCTAATACTTCTGAGATGTATC. Run-on analyses were performed essentially as previously described (45). Briefly, dot blots were performed with purified cDNA fragments corresponding to the genes of interest and with λ phage genomic DNA. After hybridization in Church buffer at 65°C, washes were carried out at 65°C in 0.5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) plus 0.1% sodium dodecyl sulfate (SDS), at 55°C in 0.5× SSC plus 0.5% SDS plus 0.2 μg/ml proteinase K, and then at room temperature in 1× SSC plus RNase A at 5 μg/ml for 30 min to ascertain specificity of hybridization. The resulting signal was quantified with a Storm860 PhosphorImager using ImageQuant software.

Western blot analyses.

For Western blot analyses, cells (10⁶ to 10⁷) were lysed by a 20-min incubation on ice in 0.6 ml of lysis buffer (1% NP-40, 10 mM Tris-HCl, pH 7.8, 150 mM NaCl, 1 mM EDTA, 2.5 mM EGTA, 10 mM NaF, 10 mM sodium orthovanadate) supplemented with 1% protease inhibitor cocktail (containing 80 μM aprotinin, 4 mM bestatin, 2.2 mM leupeptin, 1.5 mM pepstatin A, 1.4 mM E-64, 100 mM AEBSF [4-(2-aminoethyl)benzenesulfonyl fluoride]; Sigma), followed by vigorous vortexing. The supernatant recovered after a 10-min centrifugation at 13,000 rpm and 4°C was aliquoted for protein quantitation or mixed with 4× Laemmli buffer and boiled before Western blotting. Protein electrophoresis was carried by SDS-8% polyacrylamide gel electrophoresis, using Full-range Rainbow (Amersham) or Prestained Protein (Fermentas) size markers for gel calibration. Proteins were blotted electrophoretically onto a 0.45-μm-pore-size polyvinylidene difluoride membrane (Hybond P; Amersham). Immunodetection was performed with an alkaline phosphatase detection kit (ECF detection kit; Amersham) using commercially available antibodies against FKBP59 (rabbit polyclonal anti-FKBP59/HSP56; Calbiochem), MDR (rabbit polyclonal H-241; Santa Cruz), HSF1 (rabbit polyclonal SPA-901; Stressgen Biotechnologies), GRP75 (rabbit polyclonal H-155; Santa Cruz), HSP70 (mouse monoclonal W27; Santa Cruz), HSP60 (rabbit polyclonal H-300; Santa Cruz), actin (mouse monoclonal C-2; Santa Cruz), and α-tubulin (mouse monoclonal B-7; Santa Cruz). Signal was quantified with a Storm860 Fluorescence Imager using ImageQuant software.

Flow cytometry assays.

Cells were detached from culture dishes by brief trypsin-EDTA treatment, suspended in ice-cold 1× PBS buffer, pelleted by centrifugation at 4°C, and suspended in ice-cold 1× PBS buffer. Fluorescence was measured by flow cytometry with excitation at 488 nm, and fluorescence emission was measured using a 575-nm DF26 band-pass filter as previously described (16). Preliminary experiments demonstrated a linear relationship between doxorubicin-induced cell fluorescence and the concentration of doxorubicin (up to 720 ng/ml) in growth culture medium for a 3-h loading period. All cell fluorescence data were corrected for cell autofluorescence.

RESULTS

Isolation of HSF1 as a doxorubicin resistance-conferring gene.

We performed a genetic screen for cDNA able to confer constitutive doxorubicin resistance to U2-OS* cells (derived from the U2-OS human osteosarcoma cell line by transfection of an expression vector encoding the ecotropic murine retrovirus receptor). As summarized in Fig. 1A (see Materials and Methods for details), U2-OS* cells were infected with retroviral cDNA expression libraries and subjected to two cycles of selection in the presence of doxorubicin. Two independent cDNA libraries were used that had been derived, respectively, from murine embryonic stem and F9 teratocarcinoma cells. After the second selection step (doxorubicin at 75 ng/ml for 2 weeks), only a few doxorubicin-resistant cell clones could be isolated. Southern blot analysis showed that each of the doxorubicin-resistant cell clones contained a single transduced cDNA. Restriction maps were similar for all resistant cell clones. PCR amplification and sequencing of several cDNAs showed that they encoded murine HSF1 (mHSF1). mHSF1 cDNAs derived from F9 and embryonic stem cells contained the entire coding region but differed in the length of their 5′ and 3′ untranslated regions (data not shown).

FIG. 1.

Characterization of mHSF1 as an inducer of doxorubicin resistance. (A) Setup of the genetic screen to isolate cDNAs conferring doxorubicin resistance. (B) mHSF1-transduced cells are resistant to induction of a senescence-like state by doxorubicin treatment. Cells transduced with empty (control) or mHSF1-expressing retroviral vector were treated for 3 days with doxorubicin as indicated and further cultured for 7 days according to a 3T3 protocol before fixation and staining for senescence-associated acidic β-galactosidase, as previously described (14). (C) Analysis of mHSF1-induced doxorubicin resistance by clonogenic assays. Cells transduced with empty (control) or mHSF1-expressing retroviral vectors were treated with the indicated doxorubicin dose for 3 h, and their cloning efficiency was measured as described in Materials and Methods. Error bars indicate standard deviations measured with two independent cell populations. Similar results were obtained in four other independent experiments.

We first sought to verify the ability of mHSF1 to confer drug resistance. Three different mHSF1 cDNAs amplified by PCR were cloned into the pFbneo retroviral expression vector, and the corresponding viral stocks were used to infect native U2-OS* cells. Cell populations infected with empty or mHSF1-expressing pFbneo retroviral vectors were selected with G418 and then tested for doxorubicin resistance. In a first set of experiments, cells were treated for 3 days with 20 ng/ml doxorubicin and further cultured for an additional week according to a 3T3 protocol. Under these conditions, control cells infected with an empty vector nearly completely stopped growing, acquired a senescence-like morphology, and expressed senescence-associated acidic β-galactosidase as previously observed for tumor cells exposed to anticancer agents (7, 25) (Fig. 1B). In contrast, all mHSF1-transduced cell populations resumed a normal growth rate after doxorubicin treatment so that, after a week, the doxorubicin-treated cells could not be distinguished from untreated controls (Fig. 1B). To quantify doxorubicin resistance conferred by mHSF1, we measured the cloning efficiency of cells following a short exposure (3 to 4 h, depending on the particular experiment) to various doses of doxorubicin. As illustrated in Fig. 1C, the cloning efficiency of control cells measured 3 days after a 3-h doxorubicin exposure exhibits a sharp decrease at doxorubicin doses around 40 to 80 ng/ml. By contrast, the cloning efficiency of mHSF1-transduced cells was much less affected, so that at 80 ng/ml there was more than a 200-fold difference in cloning efficiency between control and mHSF1-transduced cells.

HSF1-induced drug resistance and HSP induction are uncoupled.

The constitutive induction of doxorubicin resistance triggered by mHSF1 in the absence of any intentional heat shock or cellular stress led us to investigate whether it could be uncoupled from the induction of HSPs. The C-terminal HSF1 transcriptional activation domain has been shown to be required for the induction of the heat shock genes. We first focused our analysis on two previously reported human HSF1 mutants. The dn-hHSF1 mutant (see Fig. 2A) harbors a C-terminal deletion in the transcriptional activation domain and displays constitutive DNA binding activity to HSEs of HSP promoters but is unable to transactivate, even under stress conditions (42, 47, 52). The hHSF1-3F mutant, in which phenylalanines 418, 492, and 500 are mutated, was previously shown to be dramatically impaired in its ability to induce HSP70 mRNA upon heat shock as a result of a defect in transcript elongation (6, 11). As control for HSP70 induction, we used a human HSF1 mutant harboring an internal deletion encompassing the regulatory domain (c-HSF1) that renders it constitutively active for transactivation of HSP genes, even in the absence of heat shock. Clonogenic assays showed that dn-hHSF1 and hHSF1-3F mutants were as active as wild-type HSF1 (wt-HSF1) (of murine or human origin) to confer doxorubicin resistance (Fig. 2B, C, and D). In the same assay, the c-hHSF1 mutant (constitutively active for HSP induction) (Fig. 2F) tended to be less effective than the wild-type mHSF1 (Fig. 2C and data not shown). Nonetheless, some sequences in the C-terminal part of hHSF1 are probably required, since mutants with larger C-terminal deletions that included all of the transcriptional activation domain involved in HSP induction (Fig. 2A, mutant hHSF1Δ360) and/or the regulatory domain (Fig. 2A, mutant hHSF1 Δ210) were unable to confer doxorubicin resistance (Fig. 2E).

FIG. 2.

HSF1-induced drug resistance is independent of the heat shock response. (A) Schematic representation of wild-type (wt) hHSF1 and of the corresponding mutants. LZ1 to LZ4, hydrophobic repeats (LZ, leucine zipper); CTR, C-terminal region (see references 38 and 52 and references therein). The red boxes represent the two transcriptional activation subdomains and the blue one indicates the regulatory domain (RD), as defined by Newton et al. (37). Numbering refers to the amino acids at the border of the domain. (B to E) Clonogenic assays to measure doxorubicin resistance of cell populations infected with the indicated HSF1 derivatives. Assays were conducted as described in Materials and Methods. (F) Western blot analysis of U2-OS* cell populations (U) transduced with the indicated HSF1 derivative. dox80^r indicates that the cell populations were selected for doxorubicin resistance by a 3-h treatment at 80 ng/ml doxorubicin and subsequent drug-free culture for 2 weeks. For heat shock, cells were lysed after incubation at 42°C for 1 h or following an additional culture period at 37°C for 3 h. U2-OS* cells infected with an MDR-1-expressing retroviral vector and selected for doxorubicin as described above (U-MDR dox80^r) were a positive control for P-gp expression (see Results). The indicated cell extracts were analyzed with antibody against HSF1, HSP70, and α-tubulin. The asterisk identifies a nonspecific band at about 73 kDa revealed by antibodies against HSF1, as suggested by its invariable intensity upon heat shock or infection with various HSF1-expressing retroviral vector.

To further confirm this uncoupling between the doxorubicin resistance and HSP expression induction, we performed Western blot analysis. We observed that ectopic expression of mHSF1, hHSF1, and dn-hHSF1 resulted in all cases in less than a twofold increase of endogenous HSP70, HSP75, HSP60, and HSP59 levels (Fig. 2F and data not shown). In cells overexpressing the hHSF1-3F mutant, the level of all tested HSPs was inferior or equal to those detected in control cells. Similar observations were made with HSF1-transduced cells that had been selected for doxorubicin resistance (Fig. 2F, dox80^r). In contrast, the constitutively active c-hHSF1 mutant induced a twofold increase in endogenous HSP70 expression, and selection for doxorubicin resistance resulted in an additional two- to threefold increase in expression of both c-hHSF1 and HSP70 (Fig. 2F). Since the transcriptional activity of HSF1 has been correlated with its hyperphosphorylation status, we analyzed phosphorylation of HSF1 by taking advantage of its slower mobility on SDS-polyacrylamide gels associated with its hyperphosphorylation (4). As a positive control for HSF1 hyperphosphorylation, we analyzed U2-OS* heat shocked for 1 h at 43°C. In control U2-OS* cells, HSF1 displayed a somewhat diffuse migration (around 80 kDa), indicating a hypophosphorylated state (Fig. 2F, U-control cells). Heat shock treatment for 1 h altered the migration properties of the endogenous hHSF1, which migrated as a sharp band with an apparent 10-kDa increase in molecular weight (Fig. 2F). Three hours post-heat shock, the original hHSF1 migration profile was partially restored, whereas HSP70 expression was induced more than threefold. In contrast, ectopic expression of mHSF1 did not result in any detectable HSF1 hyperphosphorylation (Fig. 2F, U-mHSF1 cells). In cells overexpressing HSF1 derivatives marked by a hemagglutinin epitope at their C terminus (hHSF1-wt and hHSF1-3F), the pattern of hHSF1 migration was similar to that observed for mHSF1 except for additional bands around 120 kDa (data not shown). In cells overexpressing the c-hHSF1 and dn-hHSF1 mutants, altered and complex patterns of migration specific for the corresponding mutant could be observed (Fig. 2F).

HSF1 activates MDR-1 expression and induces a corresponding multidrug resistance phenotype.

An important mechanism of resistance to anthracyclins is the overexpression of the MDR-1 gene and its product P-gp, an energy-dependent drug efflux pump. Northern blot analysis of cell populations infected with retroviral vectors expressing wt-HSF1, dn-HSF1, or HSF1-3F derivatives (in the absence of any doxorubicin exposure) showed a two- to threefold increase of endogenous MDR-1 RNA levels (Fig. 3A). However, subcloning experiments demonstrated that the transduced HSF1 RNA could not be detected by Northern blot analysis in about 50% of the infected cells (data not shown), suggesting that only about half of the cells were effectively transduced. Thus, a four- to sixfold increase in the MDR-1 mRNA level could be expected in cells expressing the transduced HSF1. Indeed, when uninfected or HSF1-nonexpressing cells were eliminated by a short exposure to 80 ng/ml doxorubicin and subsequent drug-free culture for 2 weeks, the levels of transduced HSF1 and endogenous MDR-1 mRNAs were doubled, as demonstrated by Northern blot analysis (Fig. 3A). Interestingly, we observed that the level of HSP27 mRNA was also increased about twofold in U2-OS* cells transduced with wild-type HSF1 in the absence of heat shock. However, HSP27 mRNA was not increased upon hHSF1-3F mutant ectopic expression (Fig. 3A), suggesting distinct mechanisms for HSP27 and MDR-1 induction by HSF1 overexpression. Western blot analysis confirms that the increase in the MDR-1 mRNA level in all cases correlates with a corresponding increase of P-gp (Fig. 3B and data not shown). It is noteworthy that we were unable to induce P-gp expression upon heat shock in U2-OS* cells (Fig. 3B), further suggesting that P-gp induction by HSF1 in U2-OS* cells is independent from the heat shock response.

FIG. 3.

HSF1 activates MDR-1 gene expression. (A) Northern blot analysis of MDR-1 and HSP27 (left panel) and HSF-1 (right panel) expression in U2-OS* cell populations transduced with the indicated HSF1 derivative. MDR1 indicates the endogenous MDR-1 mRNA, HSP27 indicates the endogenous HSP27 mRNA, and gRNA and sRNA indicate the genomic and spliced mRNAs (respectively) transcribed from the retroviral vectors. Endogenous, endogenous hHSF1 mRNA; r, 28S and 18S rRNA. Northern blots were rehybridized with a GAPDH probe to ascertain equal RNA loading in each lane. Numbers below the panels are a measure of the relative amounts (as normalized to GAPDH mRNA levels) of MDR-1-, HSP27-, and HSF1-hybridizing spliced retroviral mRNAs. (B) Western blot analysis of P-gp expression in cell populations transduced with the indicated retroviral expression vector. Data shown correspond to the same experiment as described in the legend to Fig. 2F. dox80^r indicates the cells that have been selected for doxorubicin resistance (see the legend to Fig. 2F).

To test the relevance of this increase in P-gp expression to HSF1-induced drug resistance, we first estimated doxorubicin accumulation by flow cytometry in control and HSF1-transduced cells treated with doxorubicin. This assay takes advantage of the strong doxorubicin fluorescence that allows its detection at doses relevant to that used in the clonogenic assays (16). As shown in Fig. 4A, treatment of the cells for 3 h with doxorubicin followed by a 3.5-h culture in drug-free medium demonstrated that fluorescence was on average twofold lower in HSF1-transduced cells than in control cells. Moreover, U2-OS* cells infected with an MDR-1-expressing retroviral vector and further selected for doxorubicin (Fig. 4A, U-MDR dox80^r) displayed the same decrease of drug accumulation as U2-OS* mHSF1-transduced cells selected for doxorubicin resistance (Fig. 4A, U-mHSF1 dox80^r). To confirm that the differences in drug accumulation are due to P-gp activity, we performed the same assay in the presence of verapamil, an inhibitor of P-gp activity. When 22 μM verapamil was added during the loading and efflux phases, no difference between control and mHSF1 or MDR-1 cells was observed. It is worth noting that in this case, the average fluorescence was higher than in the control cells treated with doxorubicin alone, suggesting that control cells also harbor a significant basal P-gp activity. To correlate these changes in drug accumulation to doxorubicin resistance, we analyzed the effect of the same nontoxic dose of verapamil on long-term resistance to doxorubicin. As illustrated in Fig. 4B, verapamil abrogated doxorubicin resistance of mHSF1-infected cells as measured by clonogenic assays.

FIG. 4.

P-gp overexpression is involved in HSF1-induced multidrug resistance. (A) Analysis of cellular accumulation of doxorubicin by flow cytometry. The indicated cell populations were loaded with doxorubicin by a 3.5-h culture in the presence of doxorubicin (80 ng/ml) at 37°C, washed twice, and further incubated in drug-free growth medium for an additional 3.5 h before flow cytometry analysis. For analysis of verapamil effects, verapamil was added at a concentration of 22 μM during the entire course of the experiment. Fluorescence is given as percentage of U2-OS* (U-control) cell fluorescence just after the 3-h loading period. One of three experiments with similar results is shown. (B) Reversal of doxorubicin resistance by verapamil. A clonogenic assay was conducted as described in Materials and Methods except for the presence of 22 μM verapamil during doxorubicin treatment and the 3-day recovery period before plating at a low cell density. In the presence of verapamil, cloning efficiency (given as the number of cell clones for an initial population of 3 × 10⁵ cells; see Material and Methods) was below 10² at doses of doxorubicin superior or equal to 40 ng/ml for both control and HSF1-transduced cells. Error bars indicate standard deviations for two independent cell populations. Another experiment with two other independent cell populations gave similar results. (C) Clonogenic assay to measure vincristin resistance. Cells were treated for 3 days in the presence of the indicated dose of vincristin and then passaged at various dilutions to measure cloning efficiency. Error bars are standard deviations for two independent cell populations. (D) Reversal of doxorubicin resistance upon down-regulation of MDR-1 gene expression. Cell clones isolated following the stable transfection of U2-OS* mHSF1-transduced (U-mHSF1) cell population with a vector expressing a small hairpin small interfering RNA directed against MDR-1 (sh28) or with a control vector (sh28 mismatch) were analyzed by Northern blotting with the indicated probes (upper panel) or for doxorubicin resistance as described in the legend to Fig. 2D.

These observations led us to test whether HSF1-transduced cells also displayed a multidrug resistance phenotype, as would be expected from an increase in functional P-gp drug export activity. We observed that mHSF1-transduced cells displayed a drastic increase in resistance to vincristin, a vinca-alkaloid drug with cell toxicity mechanisms unrelated to those of doxorubicin but which is also a P-gp substrate (Fig. 4C). Similar results were obtained for cells infected with the hHSF1-3F mutant (data not shown). By contrast, we observed no cross-resistance to cisplatin or UV irradiation as expected for an MDR-1-induced multidrug resistance phenotype (two independent experiments in each case) (data not shown). Finally, to further substantiate a role of P-gp in the observed multidrug resistance phenotype, we tested whether down-regulation of endogenous MDR-1 gene expression in mHSF1-transduced cells could reverse drug resistance. Two independent mHSF1-transduced cell populations were stably transfected with a vector expressing a small hairpin (sh) interfering RNA against MDR-1 (sh-28 vector) or with a corresponding control mismatch vector (sh-28 mismatch vector) (50), and 16 clones were isolated following hygromycin selection. As illustrated in Fig. 4D, most (7 out of 8) of the clones transfected with the sh-28 vector displayed a reduction of MDR-1 RNA to a level comparable with that of parental U2-OS* cells as analyzed by Northern or Western blotting. Upon doxorubicin exposure at 80 ng/ml for 4 h, it appears that the cloning efficiency of these clones was on average 20.1-fold lower than that of the control clones transfected with the sh-28 mismatch vector (Fig. 4D and data not shown). Overall, these results suggest that MDR-1 activation is an important mechanism of HSF1-induced multidrug resistance, even if it might not be the sole mechanism.

HSF1 activates MDR-1 expression also in HepG2 cells.

To extend these observations to another cell line, we chose the HepG2 human hepatocarcinoma cell line in which MDR-1 expression is inducible by arsenite treatment (29). Cells were infected with an empty or dn-hHSF1-expressing pFbneo retroviral vector using an amphotropic retroviral packaging cell line, and two independently infected cell populations were selected with G418. Northern blot analysis showed that MDR-1 mRNA was induced twofold using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA levels as a control (Fig. 5A). Western blot analysis confirmed that endogenous P-gp was induced two- to threefold (2.8 ± 0.4; n = 2) compared to α-tubulin levels in cells ectopically expressing dn-hHSF1 (Fig. 5B), which was also comparable to what we observed in U2-OS* cells. The relative level of endogenous HSP70 was 1.3 ± 0.3 in dn-hHSF1 cells, which was also in the range of that found in U2-OS* cells. Interestingly, in the HepG2 cell line, heat shock or treatment by doxorubicin, tetradecanoyl phorbol acetate, or H₂O₂ were all reported to be incapable of inducing endogenous MDR-1 (29), supporting a stress-independent MDR-1 activation by HSF1.

FIG. 5.

dn-hHSF1-induced MDR-1 gene expression in HepG2 (Hep) cells. (A) Northern blot analysis of MDR-1 expression in cell populations infected with empty (control) or dn-hHSF1 retroviral expression vectors. The arrowhead indicates the endogenous MDR-1 mRNA; r indicates the 28S and 18S rRNAs. Northern blots were rehybridized with a GAPDH probe to ascertain equal RNA loading in each lane. Numbers below the panels are a measure of the relative amount (as normalized to GAPDH mRNA levels) of the main MDR-1 mRNA transcript. (B) Western blot analysis of protein extracts isolated from the same cell populations with antibodies against P-gp, HSF1, HSP70, and α-tubulin.

Heat shock-independent induction of MDR-1 expression by HSF1 occurs at a posttranscriptional level.

We found that HSF1 overexpression in U2-OS* cells strongly enhanced HSF1 DNA binding activity as previously reported in other cell types (see Fig. S1 in the supplemental material). This led us to test whether the activity of a previously characterized MDR-1-luciferase reporter gene [plasmid pMDR1(−1202) (23, 46) containing the MDR-1 promoter sequence from −1202 to + 118] was regulated by HSF1 in the absence of cellular stress or heat shock. Transient transfection experiments failed to demonstrate any enhanced expression of the pMDR1(−1202) luciferase reporter gene in HSF1-transduced cells, even in cells selected for doxorubicin resistance and in which endogenous MDR-1 transcript is induced about sixfold (Fig. 6A). Similar results were obtained upon stable transfection of pMDR1(−1202) into control and HSF1-transduced cells. In fact, we observed a significant decrease of luciferase activity in HSF1-transduced cell populations that were stably transfected with pMDR1(−1202) and subsequently selected for doxorubicin resistance (Fig. 6B). Additionally, we found that disruption by mutagenesis of the HSE previously implicated in the regulation of MDR-1 promoter by HSF1 (29, 46) has no significant effect on the luciferase activity of the reporter gene as assayed upon transient or stable transfections [Fig. 6B, pMDR1(−1202)mut]. To ascertain the functionality of the luciferase reporter genes, we verified that luciferase activity could be induced upon treatment of the cell populations with trichostatin as previously reported (Fig. 6B, U-control/TSA) (23, 46). This failure to detect any regulation of MDR-1 promoter activity with the MDR-1-luciferase reporter gene led us to examine whether transcriptional activity of the endogenous MDR-1 gene was increased in HSF1-transduced cells. Run-on analyses revealed no increase of endogenous MDR-1 gene transcription in HSF1-transduced cells or in HSF1-transduced cells that were further selected for doxorubicin resistance and in which endogenous MDR-1 mRNA level is increased about sixfold (Fig. 6C and D). As an internal control, we readily detected about a sevenfold increase of HSF1 sequence transcription in HSF1-transduced cells compared to control cells (Fig. 6C), which fits well the about eightfold-increased level of HSF1-hybridizing transcripts measured in the doxorubicin resistant HSF1-transduced cells by Northern blot analysis (Fig. 3A, right panel).

FIG. 6.

HSF1 activates MDR-1 gene expression at a posttranscriptional level. (A) pMDR1(−1202) luciferase reporter gene activity is identical in control and HSF1-transduced cells as measured in transient transfection assays. Transfections and luciferase assays were performed as described in Materials and Methods. Error bars indicate standard deviations of the measurements with two independent cell populations in each case. U-mHsf1, U2-OS* mHSF1-transduced cells; U-mHsf1 dox^80r, U2-OS* mHSF1-transduced cells resistant to doxorubicin (see the legend to Fig. 2F). (B) The activity levels of pMDR1(−1202) and pMDR1(−1202)mut luciferase reporter genes are identical in both control and HSF1-transduced cells as measured by stable transfection assays. The indicated luciferase reporter was stably transfected into the control or mHSF1-transduced cell populations upon cotransfection with plasmid pBabe-hygro, followed by selection for hygromycin resistance. U-mHSF1 dox160^r cell populations were derived from the above hygromycin-resistant U2-OS* mHSF1-transduced cell populations upon treatment with doxorubicin at 160 ng/ml for 4 h and additional cell culture for 2 weeks. U-control/TSA are the above hygromycin-resistant control cell populations treated with trichostatin at 100 ng/ml for 1 day before luciferase activity measurement. Relative luciferase activity was normalized to protein content in the cell extract. Error bars represent standard deviations for four independent cell populations in each case. Similar results were obtained by comparing control and hHSF1-3F-transduced cells. (C) Nuclear run-on analysis of the endogenous MDR-1 gene in control and the indicated HSF-1-transduced cell populations. Run-on transcripts were synthesized and hybridized to dot blots of lambda phage DNA as a negative control or cDNA derived from HSF1, GAPDH, and MDR-1 genes. Similar results were obtained by comparing control, hHSF1-3F-transduced cells and hHSF1-3F-transduced doxorubicin-resistant cell populations (two independent cell populations were analyzed in each case). (D) Northern blot analysis of control and mHSF1-transduced doxorubicin-resistant cells with the indicated probes. Actinomycin D at 10 μg/ml was added as indicated to the cell culture medium 30 min before cell lysis. Numbers below the panels are a measure of the relative amount (as normalized to GAPDH mRNA levels) of the main transcript hybridizing with the intronic (position indicated by arrow a in the left panel) or with the MDR-1 (position indicated by the arrowhead in the right panel) probe. r indicates the positions of 28S and 18S rRNAs.

Overall, these results suggested that HSF1 overexpression induces an increase in the MDR-1 mRNA level in the absence of any effect on MDR-1 promoter activity. We surmised that if this was true, we should be able to detect MDR-1 mRNA precursors at similar levels in HSF1-transduced and control cell populations. We performed Northern blot analysis of control and HSF1-transduced doxorubicin-resistant cell populations using as a probe a nonrepeated sequence of MDR-1 gene intron 26 (the MDR-1 gene contains 28 introns [8]; see Materials and Methods). A very-high-molecular-weight transcript could be detected using this probe in both control and HSF1-transduced doxorubicin-resistant cell populations (Fig. 6D, left panel, arrow a). The level of this transcript measured in different HSF1-transduced cell populations further selected for doxorubicin resistance was found to be 1.2-fold ± 0.2-fold (n = 3) that found in control U2-OS cell populations (normalized to GAPDH mRNA levels). A similar result was found when HSF1-transduced cell populations not selected for doxorubicin resistance were analyzed (see Fig. S3 in the supplemental material). In contrast, the level of MDR-1 mRNA was on average sixfold higher in the HSF1-transduced cell populations compared to control (Fig. 6D, right panel). Two additional observations support the view that this high-molecular-weight transcript is an intron-containing MDR-1 gene transcript. First, this transcript has a short half-life, as would be expected from a high-molecular-weight mRNA precursor undergoing splicing. Indeed, a 30-min treatment of the cells by actinomycin-D (to inhibit polymerase II transcription) markedly reduced its level in both control and HSF1-transduced cells (Fig. 6D, left panel), whereas the mature MDR-1 mRNA was only slightly affected. Additionally, rehybridization of the blot with a probe encompassing the 5′ part of the MDR-1 cDNA or a unique sequence of MDR-1 gene intron 4 allowed us to detect a signal at the position of the intron 26 hybridizing species, suggesting that this latter encompasses the major part of the MDR-1 gene (Fig. 4D, right panel, and data not shown; see Fig. S2 in the supplemental material).

HSF1 targets nuclear processing of MDR-1 transcripts.

To further define the level at which HSF1 regulates MDR-1 expression, we isolated nuclear and cytoplasmic RNA fractions from control and hHSF1-3F-transduced doxorubicin-resistant cells and analyzed 10 μg of RNA from each fraction by Northern blotting. Staining of the membrane with ethidium bromide showed that the nuclear RNA fraction was considerably enriched with 28S rRNA precursors (Fig. 7A). Conversely, the nuclear fraction was shown to contain a very low amount of mature 18S rRNA (Fig. 7A) or GAPDH mRNA (Fig. 7B) compared to the cytoplasmic fraction. The ratio between the levels of nuclear and cytoplasmic GAPDH mRNA was 0.19 ± 0.05 (n = 4), suggesting that, under these experimental conditions, contamination of the nuclear RNA by cytoplasmic RNA was less than 20% of the level of signal measured in the cytoplasmic fraction. By contrast, MDR-1 mRNA (Fig. 7B, arrowhead) was relatively more abundant in the nuclear fraction than in the cytoplasmic one, suggesting that most of the MDR-1 mRNA detected in the nuclear fraction was truly nuclear. Indeed, the presence in the nucleus of mature mRNA has been documented for many genes (see reference 3 and references therein). Interestingly, the level of MDR-1 mRNA observed in the nuclear RNA fraction was higher in HSF1-transduced cells than in control cells (Fig. 7B), suggesting that induction of MDR-1 mRNA by HSF1 is a nuclear event. Another polyadenylated transcript hybridizing with an MDR-1 probe but of higher molecular weight than the MDR-1 mRNA was also more abundant in the HSF1-transduced cells than in control cells (Fig. 7B, arrow b) (this RNA species was hidden by the 28S RNA in total RNA). In addition, MDR-1 mRNA stability was not found to be increased upon HSF1 overexpression (see Fig. S2 in the supplemental material). Overall, these findings suggest that some nuclear events affecting the stability and/or the splicing of MDR-1 transcripts are modulated by HSF1 overexpression.

FIG. 7.

Induction of MDR-1 mRNA by HSF1 is a nuclear event. Nuclear and cytoplasmic RNA isolated from control (U-control) and hHSF1-3F-transduced doxorubicin-resistant (U-hHSF1-3F dox^80r) cells were analyzed by Northern blotting. (A) Staining of the blot membrane with ethidium bromide (EtB). (B) Hybridization of the blot with an MDR-1 probe encompassing the complete MDR-1 open reading frame (upper panels) and rehybridization with a GAPDH probe (lower panel). The right panel is a Northern blot analysis of the polyadenylated RNAs extracted from the corresponding fraction with a Dynabeads kit (Dynal). Arrowhead marks the MDR-1 mRNA. Arrow b indicates the position of a higher-molecular-weight MDR-1-hybridizing mRNA species.

DISCUSSION

In the present work, we set up a functional genetic screen in mammalian U2-OS osteosarcoma cells to select genes that confer constitutive doxorubicin resistance. Unexpectedly, HSF1 was isolated as a strong inducer of doxorubicin resistance in two independent genetic screens with different cDNA expression libraries. We showed that HSF1 overexpression induces a multidrug resistance phenotype independently of any exogenous cellular stress (or heat shock). This phenotype was associated with an enhanced expression of the MDR-1 gene, both at the mRNA and protein level, and silencing of MDR-1 by RNA interference partially abrogated the resistance. This observation was extended to HepG2 cells.

The activity of HSF1 was unexpected since it is generally considered that HSF1 is transcriptionally inactive in the absence of heat shock or cellular stress. This study indicates, however, that the multidrug resistance and MDR-1 induction can be uncoupled from the heat shock response. Indeed, in U2-OS and HepG2 cells, no induction of MDR-1 was observed upon heat shock (reference 29 and the present study). Furthermore, ectopic HSF1 overexpression did not lead to a significant induction of heat shock-responsive genes, in agreement with the absence of HSF1 hyperphosphorylation (Fig. 2F, mHSF1). Additionally, two different HSF1 derivatives, with mutations or a deletion in the transcriptional activation domain that compromise their ability to activate HSP gene transcription in response to heat shock, were found to be as effective as wild-type HSF1 in the induction of MDR-1.

These observations suggested that the induction of MDR-1 expression does not result from the previously characterized transcriptional activity of HSF1. This was further supported by a set of experiments looking at the transcriptional activity of the MDR-1 promoter. First, HSF1 overexpression has no effect on the expression of an MDR-1-luciferase reporter gene, pMDR1(−1202) (23), as assayed in transient and stable transfections in control and HSF1-transduced U2-OS* cell populations in which the endogenous MDR-1 transcript is induced about sixfold. Second, disruption by mutagenesis of the HSE, previously implicated in the regulation of MDR-1 promoter by HSF1 (29, 46), has no significant effect on the luciferase activity of a pMDR1(−1202) stably transfected into U2-OS cells that also carry a transduced HSF1. Third, run-on analyses failed to detect an increase in MDR-1 transcription in HSF1-transduced cells. These results led us to investigate whether HSF1 could regulate MDR-1 gene expression at a posttranscriptional level. We first attempted to detect MDR-1 mRNA precursors to explore whether or not their abundance was affected by HSF1 overexpression. At least one such high-molecular-weight RNA transcript was identified using as a probe a unique sequence derived from intron 26, the abundance of which was unchanged. Northern blot analysis of the nuclear RNA fraction further showed that the increase of MDR-1 mRNA is a nuclear event. These data suggest that HSF1 modulates the MDR-1 mRNA maturation pathway and could act either on splicing or on the stability of some MDR-1 mRNA precursor. This is in agreement with our finding that the stability of MDR-1 mRNA is unchanged in HSF1-transduced cells (see Fig. S2 in the supplemental material). Although a role for HSF1 in RNA processing has not been fully documented, it is noteworthy that stress-activated HSF1 has already been suspected of modulating RNA splicing due to its incorporation into nuclear stress bodies (5, 12). The present study, however, suggests that hypophosphorylated HSF1 complexes can modulate gene expression through posttranscriptional mechanisms, as the main effect of HSF1 overexpression is a greatly enhanced formation of hypophosphorylated HSF1 complexes with DNA binding capabilities (references 33, 38, and 52 and the present study).

The biological relevance of the heat shock-independent activity of HSF1 remains to be evaluated, as there is little information in the literature regarding the status of HSF1 during normal development or pathological events. However, as stated in the introduction, there is indirect evidence for the involvement of non-heat-shock-related HSF1 activity during development (22, 49). In addition, it has been reported that some nonsteroidal anti-inflammatory drugs induce MDR-1 expression in human Molt-4 lymphoma (15), HepG2 hepatocarcinoma (31), and LNCaP prostate cancer (40) cells. Since these nonsteroidal anti-inflammatory drugs can induce the DNA binding activity of HSF1 but not HSP expression (13, 21, 24, 30), we propose that the induction of MDR-1 by these drugs results from a posttranscriptional mechanism similar to what we observed in U2-OS cells. Accordingly, we demonstrated a heat shock-independent constitutive induction of MDR-1 upon ectopic dn-hHSF1 expression in HepG2 cells. This strongly suggests the occurrence of heat shock-independent biological activity of HSF1 in several human cell types involved, in particular, in the activation of MDR-1 expression. The possible involvement of HSF1 in the constitutive activation of MDR-1 (and perhaps other genes) in some human cancers may warrant clinical investigation. In this regard, it is worth noting that a correlation between HSE DNA binding activity and the basal level of P-gp expression was previously established in mouse mammary lymphoma and leukemia cell lines (26, 27). Furthermore, the demonstration of a frequent up-regulation of HSF1 in malignant prostate adenocarcinoma cells without a corresponding increase of HSP70 (19) is consistent with our results and may point to HSF1 as participating in tumor progression via heat shock-independent mechanisms.