A Novel Function of Molecular Chaperone HSP70

The oncogenic transcription factor FOXM1 is overexpressed in the majority of human cancers, and it is a potential target for anticancer therapy. We identified proteasome inhibitors as the first type of drugs that target FOXM1 in cancer cells. Here we found that HSP90 inhibitor PF-4942847 and heat shock also suppress FOXM1. The common effector, which was induced after treatment with proteasome and HSP90 inhibitors or heat shock, was the molecular chaperone HSP70. We show that HSP70 binds to FOXM1 following proteotoxic stress and that HSP70 inhibits FOXM1 DNA-binding ability. Inhibition of FOXM1 transcriptional autoregulation by HSP70 leads to the suppression of FOXM1 protein expression. In addition, HSP70 suppression elevates FOXM1 expression, and simultaneous inhibition of FOXM1 and HSP70 increases the sensitivity of human cancer cells to anticancer drug-induced apoptosis. Overall, we determined the unique and novel mechanism of FOXM1 suppression by proteasome inhibitors.

The oncogenic transcription factor FOXM1 is overexpressed in the majority of human cancers, and it is a potential target for anticancer therapy. We identified proteasome inhibitors as the first type of drugs that target FOXM1 in cancer cells. Here we found that HSP90 inhibitor PF-4942847 and heat shock also suppress FOXM1. The common effector, which was induced after treatment with proteasome and HSP90 inhibitors or heat shock, was the molecular chaperone HSP70. We show that HSP70 binds to FOXM1 following proteotoxic stress and that HSP70 inhibits FOXM1 DNA-binding ability. Inhibition of FOXM1 transcriptional autoregulation by HSP70 leads to the suppression of FOXM1 protein expression. In addition, HSP70 suppression elevates FOXM1 expression, and simultaneous inhibition of FOXM1 and HSP70 increases the sensitivity of human cancer cells to anticancer drug-induced apoptosis. Overall, we determined the unique and novel mechanism of FOXM1 suppression by proteasome inhibitors.
Forkhead Box M1 (FOXM1) is an oncogenic transcription factor that is also a well recognized cell cycle regulator (1). FOXM1 expression is eliminated from quiescent or differentiated cells, but it is highly expressed in proliferating and malignant cells (1). It is also well documented that FOXM1 is overexpressed in broad types of human cancers and it contributes to several aspects of cancer development (2,3). Consequently, it is not surprising that FOXM1 has evolved as a potential target of anticancer therapy (4).
Our group identified the thiazole antibiotics Siomycin A (5) and thiostrepton (6) as the first chemical inhibitors of FOXM1. Investigation of their mechanism of action revealed that they are proteasome inhibitors (7). Bona fide proteasome inhibitors, including MG132 and bortezomib, have also been found to hinder the transcriptional activity and expression of FOXM1 (7), suggesting that FOXM1 is one of the general targets of these drugs. Proteasome inhibitors stabilize the expression of cellular proteins. Therefore, suppression of FOXM1 is counterintuitive and has not been well understood. As an explanation for this effect, we put forward the hypothesis that a negative regulator of FOXM1 (NRFM) 2 is stabilized by proteasome inhibitors (8), which directly or indirectly inhibits FOXM1 transcriptional activity (9). Here we determined that the NRFM is HSP70. Because a positive autoregulation loop is required for FOXM1 expression (9), HSP70 (NRFM) also represses FOXM1 mRNA and protein expression after treatment with proteasome inhibitors.
Heat shock proteins (HSPs) are stress proteins that are mainly expressed in response to different cellular and environmental insults. HSPs are divided by their molecular weight into the subfamilies HSP110, HSP90, HSP70, HSP60, and small HSPs. Besides their major function of aiding protein folding as molecular chaperones, HSPs also play important roles in cell signaling, protein trafficking, and blocking cell death (10 -13). Elevated levels of HSPs are linked not only to cytoprotection but to neurodegenerative disorders and cancer (10,12). Among the several members of the HSP70 family, HSC70 is constitutively expressed under non-stressful conditions as well, whereas the expression of stress-inducible HSP70 is mainly detectable following different types of stress (11,12). Overexpression of anti-apoptotic HSP70 is often seen in human cancers. Therefore, targeting HSP70 has emerged as a tempting approach to improve anticancer therapy (12). However, our current findings suggest that caution should be employed with the incorporation of HSP70 inhibitors into chemotherapeutic drug combinations for the treatment of cancer because they lead to the induction of oncogenic transcription factor FOXM1.
In this study, we identified HSP70 as the negative regulator of FOXM1 after proteotoxic stress. We demonstrate that treatment with proteasome and HSP90 inhibitors or heat shock results in decreased FOXM1 and increased HSP70 protein expression. We also show that following proteotoxic stress FOXM1 and HSP70 bind to each other. We concluded that HSP70 negatively regulates the transcriptional activity of FOXM1 by interfering with the DNA-binding ability of FOXM1. Because of FOXM1 autoregulation, HSP70 also inhibits FOXM1 expression. In addition, we demonstrate that the combination of FOXM1 and HSP70 inhibitors increases the sensitivity of human cancer cells to cell death. Our results reveal a novel regulatory function of HSP70 after proteotoxic stress. osteosarcoma and osteosarcoma-derived C3 (14) and C3-luc cells (5) were grown in DMEM (Cellgro). The MDA-MB-231 (ATCC) breast and RPMI 8266 multiple myeloma cancer cell lines were grown in RPMI medium (Cellgro). The media were supplemented with 10% fetal bovine serum (Atlanta Biologicals) and 1% penicillin-streptomycin (Gibco). All cells were maintained at 37°C in 5% CO 2 . Bortezomib (Velcade, Millennium Pharmaceuticals), thiostrepton (Sigma), PF-4942847 (Pfizer), 9-aminoacridine (9AA, Fluka) and quercetin (Sigma) were dissolved in dimethyl sulfoxide (Fisher Scientific), and doxycycline (LKT Laboratories) was dissolved in PBS.
Plasmids, siRNAs, and Transfection-The FOXM1 expression plasmid was a gift from the late Dr. Robert H. Costa, and the HSP70 expression plasmid was purchased from Addgene (plasmid 15215, Ref. 16). Control (universal negative control 1) siRNAs and siRNAs specific to HSP70 (17)(18)(19) and FOXM1 (GGACCACUUUCCCUACUUUUU) were synthesized by Sigma. Transient transfections were carried out with Lipofectamine 2000 (Invitrogen) according to the recommendation of the manufacturer. Cells were processed as described in the figure legends after transfection.
IP-Cells were harvested, lysed in IP buffer supplemented with protease inhibitors, and protein concentration was determined by Bio-Rad protein assay reagent. Between 200 and 400 g of lysates in a 500-l volume were precleared by incubation with 1 g of rabbit IgG and 30 l of protein A-agarose (Santa Cruz Biotechnology) on a rotary mixer for 30 min at 4°C. The samples were centrifuged for 1 min at 1000 rpm, and the clarified lysates were transferred to fresh microcentrifuge tubes. 2 g of FOXM1 antibody and 30 l of protein A-agarose were added, and the samples were incubated on a rotary mixer overnight at 4°C. The agarose beads were washed four times with IP buffer containing protease inhibitors and centrifuged for 1 min at 1000 rpm at 4°C each time. The beads were then resuspended in 100 l of 2ϫ Laemmli sample buffer (Bio-Rad), boiled, and centrifuged, and then the supernatant was separated on a 10% SDS-polyacrylamide gel and immunoblotted with FOXM1-and HSP70-specific antibodies.
Flow Cytometry: Propidium Iodide Staining-Cells were treated as indicated and harvested by trypsinization. Then cells were washed in PBS and fixed in ice-cold 95% ethanol. Following fixation, cells were stained with propidium iodide (50 g/ml) (Invitrogen) in PBS/RNase A/Triton X-100 for 30 min at room temperature and analyzed by flow cytometry.
Statistical Analysis-Statistical analysis was performed using one-way analysis of variance followed by Tukey's multiple comparison post test or unpaired t test. p values of Ͻ0.05 were considered to be statistically significant.

FOXM1 Protein Expression Is Suppressed, whereas HSP70
Protein Expression Is Elevated following Proteotoxic Stress-We have found previously that FOXM1 is down-regulated (7), whereas it is known that HSP70 is up-regulated after treatment with proteasome inhibitors (21). To confirm this observation, multiple myeloma (RPMI 8266) and breast cancer cells (MDA-MB-231) were treated with the proteasome inhibitors bortezomib or thiostrepton (Fig. 1, A-C). We found that following treatment with the proteasome inhibitors FOXM1 suppression correlated with HSP70 up-regulation. Because inhibition of HSP90 leads to the up-regulation of HSP70 (12), we wanted to determine whether treatment with HSP90 inhibitors also represses FOXM1. MIA PaCa-2 pancreatic and MDA-MB-231 breast cancer cells were treated with the HSP90 inhibitor PF-4942847, and FOXM1 was detected by immunoblotting (Fig. 1, D and E). We found that a decrease in FOXM1 expression was indeed linked to an increase in HSP70 expression following treatment with the HSP90 inhibitor PF-4942847. In agreement with previous studies, we also confirmed that the HSP90 inhibitor PF-4942847 is not a proteasome inhibitor (data not shown). Therefore, it induces HSP70 (Fig. 1, D and E) by a mechanism different from proteasome inhibitors. Because both proteasome inhibitors and the HSP90 inhibitor PF-4942847 up-regulated HSP70 and suppressed FOXM1, it suggests that after proteotoxic stress the down-regulation of FOXM1 by HSP70 is a universal mechanism. Additionally, when breast (MDA-MB-231) cancer cells were subjected to heat shock, we also observed that the induction of HSP70 correlated with the suppression of FOXM1 (Fig. 1F). These data suggest that HSP70 might negatively regulate FOXM1 during proteotoxic stress.
Inhibition of HSP70 Alleviates FOXM1 Suppression during Proteotoxic Stress-To evaluate the role of HSP70 in the negative regulation of FOXM1 during proteotoxic stress, we inhibited HSP70 using drugs and RNAi. Prostate (DU145), breast (MDA-MB-231), and osteosarcoma (U2OS) cells were treated with thiostrepton or PF-4942847 in the presence and absence of the HSP70 inhibitors 9AA or quercetin, which are commonly used to inhibit HSP70 (10 -12, 24). Western blotting showed that inhibition of HSP70 led to the restoration of FOXM1 protein level to basal level following treatment with the proteasome inhibitor thiostrepton or the HSP90 inhibitor PF-4942847 (Fig.  2, A-D). Because both 9AA and quercetin are not specific inhibitors of HSP70, we also utilized RNAi against HSP70. Similarly to the pharmacological inhibitors of HSP70, RNAi-mediated knockdown of HSP70 in pancreatic and breast cancer cells restored the protein expression of FOXM1 after thiostrepton or PF-4942847 treatment (Fig. 2, E-G). These data suggest that HSP70 suppresses FOXM1 during proteotoxic stress.

Inhibition of HSP70 in Combination with Suppression of FOXM1 Promotes Apoptosis of Human Cancer Cells-HSP70
can block both the extrinsic and intrinsic apoptotic pathways at different levels. Consequently, inhibition of HSP70 is believed to counteract its anti-apoptotic functions and make cells more sensitive to cell death (10 -12). However, according to our current findings, HSP70 inhibition leads to the up-regulation of FOXM1 (Fig. 2). FOXM1 is also known to confer resistance to apoptosis induced by various drugs, including proteasome inhibitors, Herceptin, paclitaxel, cisplatin, and epirubicin (25)(26)(27)(28). Consequently, FOXM1 suppression increased the sensitivity of cancer cells to different cell death-inducing stimuli (3,4).
We were interested whether concurrent inhibition of FOXM1 and HSP70 make cancer cells more prone to cell death. Transient FOXM1 knockdown cells were treated with thiostrepton in the presence or absence of the HSP70 inhibitor 9AA. Apoptosis was assessed by immunoblotting for cleaved caspase-3 (Fig. 3, A-C) and by flow cytometry after propidium iodide staining (Fig. 3D). In agreement with earlier studies (10 -12), we also observed that HSP70 inhibition sensitized cancer cells to cell death, in this case to thiostrepton-mediated apoptosis. But cancer cells with FOXM1 knockdown in the presence of the HSP70 inhibitor 9AA were more sensitive to cell death induced by thiostrepton (Fig. 3). These data suggest that simultaneous suppression of FOXM1 and HSP70 could be greatly beneficial for cancer treatment to circumvent any drug-related resistance.
FOXM1 and HSP70 Interact with Each Other following Proteotoxic Stress-Because treatment with both thiostrepton and heat shock resulted in a low level of endogenous FOXM1 protein (Fig. 1), we took advantage of the tetracycline/doxycyclineinducible system to obtain high levels of exogenous FOXM1 (5), which is not affected by either stressor (data not shown). U2OS osteosarcoma-derived, tetracycline/doxycycline-inducible C3 cells (14) were treated with doxycycline and thiostrepton and immunoprecipitated with a FOXM1-specific antibody. Analysis of the TCA-acetone-purified immunoprecipitates by mass spectrometry suggested that FOXM1 interacts with HSP70 (data not shown). To further confirm the binding between FOXM1 and HSP70, the U2OS-C3 (14) and U2OS-C3-luc cell lines (5) were treated with doxycycline and the proteasome inhibitor thiostrepton (Fig. 4A) or heat shock (Fig. 4B), respectively. Immunoprecipitations were performed with a FOXM1-specific antibody followed by immunoblotting for FOXM1 and HSP70. HSP70 coimmunoprecipitated with FOXM1 in both thiostrepton-and heat shock-treated cells (Fig. 4, A and B).
HSP70 Inhibits DNA-binding of FOXM1 to Its Own Regulatory Elements-In preliminary experiments, HSP70 inhibited the transactivation of a specific reporter by exogenous FOXM1 (data not shown). Because FOXM1 is involved in a positive autoregulatory loop in which it induces its own transcription, inhibition of FOXM1 transcriptional activity also leads to the down-regulation of its mRNA and protein levels (5,9). Next we investigated the mechanism by which HSP70 inhibits the transcriptional activity of FOXM1. ChIP experiments were performed to determine whether HSP70 interferes with the binding of FOXM1 to its own regulatory elements and to the Cdc25B target gene promoter. A promoter region outside of the FOXM1 binding site of the direct transcriptional target Aurora B kinase was used as the negative control. The C3 cell line with stable expression of the doxycycline-inducible FOXM1-GFP fusion protein (14) was transiently transfected with the control enhanced GFP and the HSP70 expression plasmids. Forty-eight hours after transfection, FOXM1 expression was increased by doxycycline addition for 24 h, after which cells were processed for ChIP. These quantitative ChIP assays revealed that HSP70 reduces up to 2-fold the ability of FOXM1 to bind to its regula-tory elements (Fig. 4D), whereas the level of exogenous FOXM1 was not affected (Fig. 4C). These data suggest that HSP70 interacts with FOXM1 and inhibits its DNA-binding activity.

Discussion
The oncogenic transcription factor FOXM1 is one of the most overexpressed genes in cancer (29), and it contributes to several aspects of oncogenesis (3), including metastasis in pancreatic (30) and liver cancer (31). Therapeutic targeting of FOXM1 is discussed widely in the literature, and a few years ago we identified proteasome inhibitors as the first chemical inhibitors of FOXM1 (5, 7). However, the mechanisms of FOXM1 suppression by proteasome inhibitors were not well understood. We previously proposed a hypothetical model for the suppression of FOXM1 by proteasome inhibitors. Proteasome inhibitors stabilize a negative regulator of FOXM1 (NRFM), which, in turn, inhibits the activity of FOXM1 as a transcription factor (8,32). In this study, we identified the molecular chaperone HSP70 as the NRFM after proteotoxic stress (Fig. 4E). We determined that increased HSP70 protein expression correlates with decreased FOXM1 protein expression after exposure to proteotoxic stress (Fig. 1) and that inhibition of HSP70 up-regulates FOXM1 expression (Fig. 2). We demonstrated that HSP70 binds to FOXM1 (Fig. 4, A and B) and it inhibits FOXM1 DNA-binding to its promoter elements (Fig. 4D). As we have shown before, FOXM1 is involved in a positive feedback loop and activates its own transcription (9), so inhibition of FOXM1 transcriptional activity by HSP70 also leads to a decrease in its expression after proteotoxic stress (Figs. 1 and 4E). Recently, Cheng et al. (33) confirmed the importance of FOXM1 autoregulation for cancer development (34). They showed that the tumor suppressor SPDEF competed with FOXM1 for the FOXM1 binding site responsible for the autoregulation of FOXM1. Furthermore, SPDEF inhibited FOXM1 expression and consequently cancer development in a mouse model of prostate cancer (33). These data suggest that the autoregulation of FOXM1 might be required for cancer development and could be a promising target for anticancer treatment.
Suppression of FOXM1 (22,23) and accumulation of ubiquitin conjugates (35) are two of the accepted hallmarks of proteasome inhibition. Here we confirm that HSP90 inhibitors are not proteasome inhibitors because they do not induce the accumulation of ubiquitin conjugates (data not shown). However, they target FOXM1 (Fig. 1, D and E), suggesting that they belong to a novel class of FOXM1 inhibitors. Both proteasome and HSP90 inhibitors up-regulate HSP70 and suppress FOXM1, supporting the idea that HSP70 is the universal NRFM that inhibits FOXM1 after proteotoxic stress.
Under normal conditions, HSP70 accumulates after stress to protect cells from adverse effects, therefore promoting cell survival. However, high levels of HSP70 are also often found in different types of cancer, providing protection against cell death and resistance to chemotherapeutic drugs. Therefore, HSP70 appeared as one of the important targets to improve the outcomes of anticancer therapy, and the search for inhibitors of HSP70 has been on the rise (12). However, this work calls for caution in respect to the prospective administration of HSP70 inhibitors to cancer patients in the future. We showed that inhibition of HSP70 leads to the up-regulation of the oncogenic transcription factor FOXM1 (Fig. 2) and that FOXM1 knockdown cancer cells in the presence of the HSP70 inhibitor 9AA became more sensitive to apoptosis (Fig. 3). Consequently, these findings strongly support the combination of novel types of FOXM1 and HSP70 inhibitors to avoid detrimental effects and to achieve a more effective therapeutic strategy for the successful treatment of cancer.
Several parallels exist between FOXM1 and HSP70. Both are overexpressed and considered to be oncogenic in different types of cancer, and their elevated expression is associated with a poor prognosis for cancer patients. Increased expression of both makes cancer cells resistant to cell death and chemotherapeutic drugs, and both have emerged as attractive targets of anticancer therapy (3,4,12,36). Altogether, according to the existing literature, both FOXM1 and HSP70 are oncogenes. Also, the authors of a recent review on HSP70 argued that HSP70 might contribute to cancer development by regulating multiple signaling pathways, such as HIF1-␣ and NF-B (36). However, according to our findings, by inhibiting the FOXM1 oncogene, HSP70 can also act as a tumor suppressor. Taken together, our current data suggest that HSP70 belongs to the group of proteins with "antagonistic duality" (37), including c-myc and p21, which depending on the circumstances may act as oncogenes or tumor suppressors.